The lectures series of the33rd Professor Harry MesselInternational Science School3-16 July 2005

The Science Foundation for Physicswithin The University of Sydney

“ I n t h e P u r s u i t o f E x c e l l e n c e ”

EDITORS

Dr Chris Stewart

Executive Officer,

The Science Foundation for Physics,

The University of Sydney, Australia

Associate Professor Robert Hewitt

Director

The Science Foundation for Physics

The University of Sydney, Australia

Editorial assistance from Alison Thorn and Alex Viglienzone

A course of lectures given at the

33nd Professor Harry Messel International Science School for High School Students,

organised by The Science Foundation for Physics within

The University of Sydney, at the University of Sydney

3-16 July 2005

The Science Foundation for Physics

The University of Sydney NSW 2006

Australia

http://www.physics.usyd.edu.au/foundation

© Copyright Science Foundation for Physics

June 2005

All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any

means, electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from

the Science Foundation for Physics, The University of Sydney.

Designed and produced by Peter Thorn Design, Sydney

Waves of the Future

ISBN: 1 86487 725 1

The Science Foundation for Physics gratefully acknowledges the Telstra Foundation's Community Development Fund

for their generous support in the production of this book.

3

ContentsRadio telemetry in the study of wildlife 09Dr Frank Seebacher

Catch, Move and Twist with Optical Tweezers: Biophotonics at work 23Professor Halina Rubinsztein-Dunlop

The Treatment of Cancer Using Ionising Radiation 39Dr Clive Baldock

The psychophysics of real and virtual auditory spaces 51Associate Professor Simon Carlile

Telecommunications: the here and now 67Professor Martijn de Sterke

Telecommunications: looking to the future 79Professor Martijn de Sterke

The Science of the Aerosols we breathe 91Professor Lidia Morawska

Creating and overcoming invisibility: scrutably personalised ubiquitous computing 107Associate Professor Judy Kay

Seeing in the Nanoworld 119Professor David Cockayne

Building in the Nanoworld 133Professor David Cockayne

Understanding Brain Dynamics 147Professor Peter Robinson

Wind, Waves and Beaches 157Professor Andrew D Short

The ever changing life of galaxies 173Dr Raffaella Morganti

Monsters lurking in the centre of galaxies 191Dr Raffaella Morganti

Quantum Mechanics: The Wild Heart of the Universe 207Dr Joseph Hope

Einstein and the Quantum Spooks 221Professor Huw Price

The Messel Endowment 235

4

MESSEL ENDOWMENT DONORS

Extra-Galactic Donors: A$1 000 000 and overThe Department of Education, Science and TrainingMulpha Australia Limited

Galactic Donors: A$100 000 to A$999 999The Science Foundation for PhysicsNell & Hermon Slade TrustHermon Slade Foundation

Stellar Donors: A$10 000 to A$99 999Mr Terrey P. ArcusMr Michael MesselJames N. Kirby FoundationOneSteel LimitedCochlear LimitedAustralian Business LimitedCecil & Ida Green FoundationMacquarie Charitable Foundation LimitedMr Robert ArnottMr John A L Hooke, CBEUSA FoundationMr Jim O’ConnorWestpac Banking Corporation

I n 1999 the Science Foundation for Physics within the University of Sydney established the Messel Endowment to honour Professor Harry Messel and to fund the International Science School in perpetuity.

The Messel Endowment is managed to preserve the real value of all its donations. The surplus interest isused to support the Professor Harry Messel International Science School (ISS). If income exceeds therequirements of the ISS, the Foundation will use the funds to support other initiatives within the School ofPhysics. Such activities will be named to honour Professor Harry Messel.

The Science Foundation for Physics sincerely thanks all supporters of the Messel Endowment. Allcontributions to the Endowment are important to its success and the Foundation acknowledges thefollowing for their generosity. For more information on the Messel Endowment please contact The ScienceFoundation for Physics on +61 2 9351 3622, email scifound@physics.usyd.edu.au or visitwww.physics.usyd.edu.au/foundation/ <http://www.physics.usyd.edu.au/foundation/>

Emeritus Professor Richard Collins and Mrs Marilyn Collins

Emeritus Professor Maxwell G Brennan, AO,and Mrs Ionie M Brennan

Associate Professor Robert G Hewitt and Mrs Helen Hewitt

Planetary Donors: A$1 000 to A$9 999Mrr Reginald J Lamble, AOMr Anthony M JohnstonIBM Australia LimitedMr Trevor E DanosMr David B HerrmanDr Jenny A NichollsMr Basil Sellers AM through Sellers Pty LtdDr Emery Severin & Mrs Sharman SeverinDr Brian J O’BrienMr Nicholas Manettas through Nick’s Seafood

Bar & GrillMr Raymond Walton & Mrs Margit WaltonMs Valma G StewardMrs Kathy ManettasMrs Georgina Donaldson

5

Lahili Pty LimitedEmeritus Professor John DavisMr Steven K EckowitzMs Danielle M LandyMr Thomas M F Yim on behalf of Alex YimDr Bruce McAdam & Mrs Janice McAdamDr Joseph A BeunenMr John A VipondProfessor Lawrence E CramMr Peter ManettasMr Graham H HallMr Christopher C VonwillerKenneth Coles FoundationDr Stephen D SegalDr Robin B FitzsimonsDr Robert L EveryDr David MalinMr David C Davidson

through David C Davidson Pty LimitedSouthcorp LimitedASA ITF Foundation for the Advancement of Astronomy

Asteroidal Donors: up to $999Mrs Iona S DoughertyProfessor Roger V ShortMs Yvonne PitsikasMrs Irene P GibsonMr Wen W MaMr John H ValderMr Enrico PiccioliMr Arun AbeyDr P E RolanBarker CollegeDr Robert H MastermanThe Australian Association of Phi Beta KappaDr George F BrandSir Walter BodmerMs Margaret A DesgrandMs Julie K EllinasMs Belinda H AllenMrs Chrissie AthisMr Julian J DrydenMr John W L RawsonMr John PatersonMr George AthisMr Gavin M ThomsonMr Arthur J BuchanMr Allan F RainbirdMrs Helen BellMr Thomas M F Yim on behalf of Jerome YimEmeritus Professor Louis C Birch

Dr Jennifer J TurnerMr Tim M SmythDr David R V WoodMrs June PapadopoulosMr Frank PapadopoulosToni R Kesby Pty LtdProfessor David R FraserMs Jennifer H F WanlessMrs R LambertMrs Mary MooreMr Spiro J PandelakisMr Robert R B MurphyMr Peter C PerryMr Ian G DennisMr Ian A DysonMr Harry J PembleMr Geoffrey D PopleMr Alan K Milston, OAMDr Xian ZhouDr Kevin C AllmanDr David Z RobinsonDr Christopher J E PhillipsAssociate Professor Donald D MillarDr David G BlairMs Tomoko KikuchiMr Steven KambourisMr Jeff CloseFr Mervyn J F ZiesingDr Hugh S MurdochDr Claire E CupittMs Anne WoodsMs Elana BontMr George Papadopoulos

6

SupportersThe Science Foundation for Physics warmly thanks the supporters ofthe 2005 ISS: Waves of the Future

The Messel EndowmentIBM Asia PacificNSW Department of Education and TrainingCollege of Sciences and Technology, The University of SydneyFaculty of Science, The University of SydneyScientific Services Pty LtdTelstra Foundation Community Development FundRalph’s CaféThe Kirby Foundation

Australian students were selected with the support of the Science Teachers Associations in Victoria,Tasmania, South Australia, the Northern Territory, the ACT and Western Australia, and the NSWDepartment of Education and Training.

The following institutions assisted in the selection and travel of the overseas students:

The Affiliated Middle School of Beijing University, ChinaMONBUSHO, JapanMinistry of Education MalaysiaThe Royal Society of New ZealandMinistry of Education, SingaporeMinistry of Education, ThailandThe Association for Science Education, UKThe Royal Institution of Great BritainNESTA (The National Endowment for Science, Technology and the Arts)Department of Energy, USA

Webcasting is made possible with generous gifts from:

Emeritus Professor Harry MesselIBM Asia PacificNSW Department of Education and TrainingAssociate Professor Bob HewittDr Jenny Nicholls

7

Preface

THE SCIENCE Foundation for Physics within the University of Sydney is delighted to present the 33rdProfessor Harry Messel International Science School for High School Students, named after the manwho had the vision to initiate them in 1958. The United Nations has designated 2005 as the

International Year of Physics since it is the centenary of the year in which Albert Einstein published hisgroundbreaking papers in three areas of physics which are the foundation for modern science. Thereforethe 2005 Science School, titled “Waves of the Future”, will give scholars the opportunity to encounter andexperience research in some of the wide range of scientific fields that have developed rapidly as a resultof Einstein’s contributions.

The primary aim of all of the Science Schools is to acknowledge the excellence of the scholars, who havebeen selected on the basis of their academic abilities. The presence of gifted young people from manycountries will allow the scholars to experience the values of many cultures and to learn new ways ofdoing things. The Science Foundation stands for the Pursuit of Excellence, and is always pleased to havean opportunity to acknowledge this spirit in young people.

The International Science School can only be held because of the generous financial contributions of theSupporters and the Donors to the Messel Endowment, and because of the time and energy given by theLecturers. Like the Science Foundation itself, the Supporters, Donors and Lecturers are committed topromoting science education at the very highest level of excellence. On behalf of the Foundation, Iexpress our grateful thanks to all three classes of benefactors.

Robert G HewittSydney, June 2005

DR FRANK SEEBACHERSchool of BiologicalSciences, University ofSydney. After completinga Ph.D. at the Universityof Queensland on thethermal biology of theAustralian freshwatercrocodile, Dr Seebacherworked as a post-doctoral researcher atJames Cook Universityand The University ofQueensland. He has

been at the University of Sydneysince 2001. His research isconcerned with the response ofanimals to changing environments,with a resolution spanning from geneexpression to behaviour in the wild.Frank is a member of the IUCNCrocodile Specialist Group.

8

What got you interested inscience in the first place?

I was a little field biologist as akid, partly for the adventure ofexploring and partly because of myinterest in the natural environment. Igrew up with a great interest inanimals and plants, and as a kidspent most of my free time roamingaround collecting lizards, frogs, andinsects, breeding tadpoles andplanting a garden. This interest turnedinto inquiry to find out how animalsand natural systems work, and Idecided to study science at university.

What’s the best thing about beinga researcher in your field?Biology is a diverse field and theresearch can be really varied. Forexample, my research coverseverything from gene expression, tofieldwork studying animal behaviour.It is this multidisciplinary approachthat reveals how natural systemswork, and that makes being aBiologist very compelling.

Who inspires you – either inscience or in other areas of your life?I get great inspiration from books,and one of my favourites is the 16thcentury thinker Michel de Montaigne.

If you could go back andspecialise in a different field,what would it be and why?I always wanted to be a mathematicianbecause of the logical structure ofmathematics, and its ubiquitousimportance in explaining the world.

What’s the ‘next big thing’ inscience, in your opinion? What’scoming up in the next decade or so?The ongoing advances in molecularbiology will change the way humansinterpret their environment, andmolecular biology will also impact onour personal lives by its secondarytechnological applications.

Introduction T H E W O R L D I S becoming an increasingly dire place for wildlife. Increasesin human population around the globe and the increasing demand fornatural resources means that there is less and less room for animalsand plants to live in their natural environment. Australia has a uniqueassemblage of wild animals and plants that have evolved in isolationfrom other continents for many millions of years. Like most countries inthe world, there is an urgent need for Australia to manage its wildlife toensure that the biodiversity of the continent is preserved into the future.A prerequisite for effective wildlife management is an understanding ofhow animals work in their natural environment. Stunning technologicaladvances in physics over the past 30 years have made radio telemetryan essential tool for wildlife research allowing the monitoring ofundisturbed animals in the wild.

Radio telemetry in the studyof wildlifeDr Frank Seebacher

9The cane toad Bufo marinus

10

WAV

ES O

F TH

E FU

TURE

I n this chapter I will give a broad outline of howliving organisms interact with their environment,and of the importance of environmental change

on the functioning of organisms. This backgroundwill help us understand why it is important tostudy animals in their natural environment. I willthen introduce radio telemetry and its uses inwildlife research. The second half of the chapterwill review three case studies in which telemetryhas been used to address important wildlifemanagement and scientific questions.

The Earth is not a Stable Place:Evolution is Driven by EnvironmentalChangeLiving systems are made up of thousands ofbiochemical reactions that must work together forthe whole organisms to function. Even if a singleone of those reactions is disrupted, thefunctioning of the whole organism may becompromised. There are many examples frommedicine where a single enzyme is lacking sothat a reaction cannot proceed to produce therequired product. In humans, such disruptions areoften treated as medical conditions, but variationsin biochemical function are found in all organismsand they are not necessarily bad – in fact,variation is essential for organisms to be able torespond to changes in their environment.

In nature, the ultimate goal of animals is toreproduce so that a particular set of genes ispassed on to the next generation. The capacity ofan organism to achieve reproductive success isreferred to as its ‘fitness’. Fitness is not onlydetermined by reproduction in its immediate sense,but also by the general well-being or performanceof an individual in its environment. For example, ifan individual lizard is a very slow runner, it may notbe able to escape from a bird that intends to eat itas well as a slightly faster lizard may. If the formerlizard gets eaten it cannot reproduce any more sothat its fitness is compromised compared to thelatter lizard because of its slow running speed. Torefer back to the example above, a disruption in abiochemical pathway may similarly decrease thefitness of an animal so that, in the worst case, itmay not be able to reproduce and its heritablematerial will disappear from the gene pool of thatspecies or population.

The rate at which chemical reactions proceedfrom a set of reagents to a product is dependenton many physical factors such as pH, pressure,and particularly temperature. Temperature is ameasure of heat energy, and the Laws ofThermodynamics state that the more energy iscontained within a chemical system, the morereadily the reactions within that system will occur.Considering that living organisms may be definedas a complex system of biochemical reactions, itbecomes obvious that the functioning, or fitness,of living beings is dependent on temperature andon other factors of the environment.

Most of the daily life of animals is preoccupiedwith interactions with the environment. Thinkabout your own life: many of your decisionsevery day depend on the climate where you live.For example, your choice of what to wear willdepend on the weather outside. Mammals are ofcourse a special case because we must maintainour internal temperature more or less constant,and that is achieved by producing metabolicheat. In other words, we breakdown food – someof this is used as building blocks and nutrients,but most of it is dissipated as heat for us toregulate our body temperature. As mentionedabove, mammals, and birds as well, are a specialcase: most animals do not regulate their internaltemperature by producing metabolic heat, butchoose suitable environments in which theirbodies will adopt a temperature that is within arange which will permit the individual to maintainits fitness. Of course, all living organisms willproduce some metabolic heat, but in the case ofectotherms – that is, most invertebrates, fish,frogs, lizards and snakes – metabolic rates aretoo low to produce sufficient heat forthermoregulation. However, regardless ofwhether an animal is ectothermic or endothermic(those that produce metabolic heat forthermoregulation) the environment, andparticularly climate, will dictate much of itsbehaviour and physiology.

The climate on the Earth is in constant flux.Climate fluctuations occur at different temporalscales from millions of years to days, and at allscales these are important for the fitness ofanimals. For example, climate during theMesozoic, the time when dinosaurs dominated

11

RAD

IOTE

LEM

ETRY

INTH

EST

UDY

OF

WIL

DLI

FE

terrestrial ecosystems from 220 to 65 millionyears ago, was much warmer than today.

[FIGURE 1 HERE]

The Cretaceous period (120 to 65 million yearsago) in particular was characterised by very warmconditions, and even at very high arctic latitudesthe climate was more similar to present dayMelbourne than to the permafrost of today’sarctic regions. Similarly, the Antarctic continent, atpresent covered by permanent ice up to 3 kmthick, was once very equable with extensivevegetation cover and diverse terrestrial wildlife.

Given the instability of the earth’s climate, andthe dependence of organisms’ fitness onenvironmental parameters, it becomes clear thatthe environment represents a major selectionpressure for organismal evolution. A commonperception is that animals have evolved to dobest in the environment they currently live in. Byand large this is true, but how do we know whatthe response of organisms will be as the climatechanges at sudden short time scales, and alsogradually over thousands and millions of years?It is crucial for our understanding of ‘life’ to knowthe mechanisms that are at work to help animalsmaintain their fitness in changing environments.These mechanisms are ultimately molecular,based on the expression of genes withinindividuals, and the recombination of genomesbetween generations. These genetic changesmay manifest themselves at a physiological level, for example in the functioning of musclesor the cardiovascular system, or at a behavioural level.

Before it is possible to assess the molecularmechanisms that underlie the evolutionaryresponse to changing environments, we mustknow the ecology of species and how theyinteract with their environment. Goodecological knowledge is not only at the basis ofa scientific understanding, but it is also crucialfor management. The natural fluctuations inclimate and the environment are exacerbatedby human induced changes. Human activitymay change environments rapidly and severely,so that the pressures on organisms may be fargreater than those originating form naturallyoccurring fluctuations alone. For example,weekly activity of humans perceptibly changesthe climate in the USA so that there are short-term weekday and weekend cycles. Humaninduced climate change in combination withpollution and habitat destruction means thatthere are immense pressures on wildlife.Sophisticated and effective wildlifemanagement will be the challenge for thiscentury, and at its core lies a goodunderstanding of the movement, habitat use,thermal requirements, and physiologicalresponses of focal species under naturalconditions.

TelemetryThe challenge for wildlife scientists andmanagers is to collect data from animals thatare undisturbed in their natural environment.Radio telemetry is a technique that allowsmonitoring of animals from a distance, therebypermitting data collection from animals that areunaware of the researcher. Telemetryencompasses a transmitter that sends signalsto a remote receiver. The most basicapplication of telemetry is monitoring thelocality of a transmitter. Transmitters emitregular signals, and they broadcast at a knownfrequency, so that a transmitter attached to ananimal will not only tell where the animal is,but it will also be uniquely identified by thetransmitter frequency. The signal is received byan investigator with a specialised telemetryreceiver up to several kilometers away, and theuse of a unidirectional antenna (one thatreceives the signal over a narrow range ofincident angles) will allow the tracking of the

Figure 1: Mean global climate has fluctuated considerablyduring the earth’s history.

12

WAV

ES O

F TH

E FU

TURE

transmitter. In recent years, biological applicationsof telemetry have become increasinglysophisticated, and it is possible now to trackanimals with satellites, and to measure bodytemperature, heart rate and blood flow. Aexpansion of radio telemetry are data loggers thatare deployed on animals to measure similarvariables - but instead of emitting a signal,dataloggers store the information and must beretrieved so that the information can bedownloaded onto a computer.

Wildlife telemetry was first developed in the1950s in studies on birds. The particularAustralian perspective, and one that is certainlyrelevant for this Science School, is the work ofProfessor Harry Messel starting in the 1960s. In1968 Harry was commissioned by one of theleading space electronics industries to design asmall, long-range, and long-lasting transmitterthat could be used under extreme environmentalconditions. The purpose of this transmitter wasnot so much wildlife tracking as tracking humans,in rescue operations for example. Nonetheless,wildlife provided a good test system. The firstprototype was tested on polar bears in the Arctic– and failed. After several years, a secondattempt was made, this time using crocodiles inNorthern Australia as the guinea pigs. This starteda major research program on the development ofwildlife tracking devices, and also on crocodilemanagement and conservation.

In the early 1970s when the telemetry work wasstarted, crocodiles were nearly extinct in Australiaas a result of hunting for their skins. Australia hastwo species of crocodile: the endemic freshwatercrocodile, Crocodylus johnstoni, and the estuarinecrocodile, Crocodylus porosus. Freshwatercrocodiles are the smaller of the two species andlive primarily in inland billabongs and rivers, butthey may also enter saltwater.

Crocodylus porosus is the largest living crocodilespecies with reputable records of maximumlength of seven metres! Populations of bothspecies, and particularly of C. johnstoni, haverecovered very well and they are again found ingood number in the Northern Territory, WesternAustralia and Queensland. Increasing populationsof estuarine crocodiles or ‘salties’ also increasesadverse encounters between people andcrocodiles, particulalry around populated areassuch as Cairns in North Queensland. Effectivemanagement of populations is therefore essentialfor the wellbeing of humans and to conservecrocodile populations into the future. Researchand management of crocodiles continues at amuch more professional level than in the huntingdays before 1970, and two examples of currentwork on crocodiles are given below.

Biotelemetry, encompassing both radio-telemetryas well as data loggers (also known as archivaltelemetry), is of course not restricted to crocodiles

Figure 2: Freshwater crocodile, Crocodylus johnstoni.

13

RAD

IOTE

LEM

ETRY

INTH

EST

UDY

OF

WIL

DLI

FE

but it is widely used in wildlife research, as well asin laboratory monitoring. The examples below aretaken from the recent scientific literature and willhopefully provide a good introduction to thepotential uses and importance of telemetry inscience and management.

Telemetry in Wildlife Studies(a) Tracking cane toadsCane toads (Bufo marinus) are a large terrestrialamphibian, originally from South and CentralAmerica. They were introduced into many Pacificareas as a control agent for sugar cane pests.Cane toads were introduced into Australia in the1930s as a ‘natural’ pest controlmeasure for the cane beetle thatinfested the extensive sugarcane crops in North Queensland.In the event, the toads ateanything but the cane beetle,and they have spread throughoutQueensland, into New SouthWales, and cane toads have justarrived at the World Heritagelisted Kakadu National Park inthe Northern Territory. As aresult, there is a strong desire in the general andscientific communities to control populations of B.marinus in Australia.

What makes matters much worse is that canetoads are potentially lethal for native predatorsowing to their venom glands situated just behindthe head. Although the direct impact of canetoads on native species is not well documented,there are considerable management effortsaimed at controlling toad infestations. As a firststep for any effective management, it is essentialto understand the ecology of toads, particularlytheir movements and the environmentalconditions that favour their dispersal.

It is not clear whether toads in northern Australiaare nomadic, with restricted movements during dryperiods but long and unpredictable migrationsduring wet seasons. In their native habitat in SouthAmerica, toads move only short distances eachnight and remain within limited activity areas evenduring the wet season. A number of toad speciesshow quite an impressive ability to navigate and

return repeatedly to the same shelter, breeding orfeeding sites. It may be advantageous for toads toreturn to familiar sites, which they know to providefood shelter or mates, rather than incurring theunknown and potentially more expensive cost oflocating alternative sites. For managementpurposes, the difference between random long-distance movements and restricted site-specificactivity is crucial.

Regardless of their pattern of movement, terrestrialamphibians are always faced with the threat ofdehydration. This is particularly true for toadsbecause, unlike many native species of frogs (thereare no native Australian toads) that secrete a waxy

coating on their skin to reducewater loss, toads have no skinresistance to evaporative waterloss – in other words, water canevaporate through their skin thesame as from an open bowl ofwater or a puddle.

Radio-telemetry is a perfect toolfor collecting data that willresolve whether toads moverandomly or whether they

display homing behaviour within a restrictedactivity range. In combination with environmentalmeasurements, such as soil moisture or airtemperature, tracking data will also reveal thebest conditions for movement. In a study Iconducted on Orpheus Island with Ross Alfordfrom James Cook University in Townsville, we

Figure 3: Orpheus Island in North Queensland where thestudy on movement of cane toads (insert) was conducted.

“... cane toads have justarrived at the WorldHeritage listed KakaduNational Park in theNorthern Territory...”

14

WAV

ES O

F TH

E FU

TURE

surgically implanted radio transmitters into canetoads to determine whether the toads movedrandomly in their environment, and whether theirmovements were limited by environmentalfactors.

We found that movement of B. marinus dependedon the moisture level of the ground. Toads nevermoved when the ground was dry (less than 0.1ml of water per gram of dry soil) which, ofcourse, is not surprising given the high water lossrates of terrestrial amphibians. Interestingly, rateof movement increased as the soil becamewetter, but there was also a maximum (i.e. anasymptote) that probably indicates the physicallimit to how far toads could move per night(Figure 4).

To resolve the question of whether toads movedrandomly, as opposed to staying within a well-defined home range, we compared the movementdata collected by radio tracking animals in thefield to a random walk. A random walk in physicsis a process that is defined as a sequence ofdiscrete steps in random directions. Randomwalks vary greatly depending on the dimension ofthe steps. Using distances and angles randomlydetermined within the range of measured valuesfrom toads, random movement patterns can beconstructed, and these can be compared toactual movement over a given period of time(Figure 5).

As it turned out, the toads moved randomly, andalthough they did frequently return to the same

daytime shelter site after moving at night, theydid not return more often than would be expectedfrom completely random movement.

(b) Satellite tracking of crocodilesTracking toads was a relatively easy thing to do.Toads are abundant and the animals are easilycaught by hand, and their small size means thatthe distances traveled are not too great fortracking on foot. The situation becomes moredifficult when dealing with large crocodiles: theanimals are potentially dangerous to deal withand catching even a two to three metre longcrocodile requires considerable logistic effort toset traps and handle the animals. To makematters worse, crocodiles travel in water, andthey travel long distances in a relatively shorttime period.

One of the shortcomings of radio telemetry is thatthe signal emitted by the transmitter is attenuatedby freshwater reducing the range over whichsignals may be received; for example, signalsfrom the transmitters inside the cane toads couldbe received from 1-2 km away, but if animalswere in freshwater, this range would be reducedto about 300-500 m. Radio signals do not travelat all in salt water so that when radiotrackingmarine animals, the investigator must rely on theanimal being at the water surface with theantenna protruding. An alternative to radiotransmitters are sonar transmitters. Sonar worksby sound waves that do travel in water, andwhich are emitted by the transmitter. In fact,some marine animals, such as whales and

Figure 5: Patterns of movement by cane toads. Measuredmovement data obtained with radio telemetry (red line)were not different from movement patterns generated by arandom walk model (blue line).

Figure 4: Movement of cane toads. Toads did not move atall when the soil was dry, and there was a maximum rateof movement.

15

RAD

IOTE

LEM

ETRY

INTH

EST

UDY

OF

WIL

DLI

FE

dolphins, use sonar to communicate and to locatepredators and prey. Sonar tracking is usuallymore involved than radiotracking on land,because several listening devices, orhydrophones, must be employed to locate thetransmitter under water, and it is not really usefulfor animals that move long distances.

Clearly, tracking estuarine crocodiles is quite achallenge. Nonetheless, obtaining movement datafrom C. porosus is absolutely essential formanagement, and it is of great scientific interestwith respect to species radiation and interactionsbetween populations of crocodiles living atdifferent parts of the world. From a managementperspective, crocodiles that could potentiallybecome a threat to humans are often removed tomore remote areas. It is not unusual for thoseanimals to return to their original residency. Thesebehaviour patterns raise the question of how farand how often crocodiles move naturally, andhow much site fidelity different sized animals andanimals of different sexes display.

Knowing movement patterns also becomesimportant in conservation of endangeredcrocodiles. Although C. porosus appear to be safein Australia, they are endangered or threatenedover most of their natural range, which includesIndia and Southeast Asia, the Pacific Islands,Indonesia, Papua New Guinea, and Australia. Is itpossible that animals from Australia supplementdepleted populations in New Guinea andIndonesia?

Mark Read from the Queensland EnvironmentalProtection Agency in Cairns recently initiated aprogram of tracking C. porosus via satellite.Satellite tracking works similarly to ‘ordinary’radiotracking, and most wildlife tracking uses theArgos Satellite System. Several satellites thatcarry the necessary reception equipment orbit theearth at any time. The satellites receive the‘message’ from the transmitter on the animal andrelay the data to the ground in real time or asstored information. At any point in time, aparticular satellite can receive signals fromtransmitters on earth over a ground area of 5000km2, and owing to the earth’s rotation thereception area moves in swathes around theglobe. This means that if there are two

operational satellites, signals from eachtransmitter on earth will be received for about tenminutes on each of 20 to 30 occasions per day.The satellites send data to ground stations thatforward them by e-mail to the researcher.

To date, Mark has deployed satellite transmitterson 15 animals at different locations on Cape YorkPeninsula in North Queensland, Australia. Thetransmitters are sutured to an area of protrudingbony scutes on the crocodile’s neck called thenuchal shield.

The data from the satellite trackers provide afascinating snapshot of crocodile life that wouldnot have been possible without the developmentof this technology. The first example of a trackingrecord (Figure 7) comes from a small (2.5 metre)male crocodile that was released 50 to 60kilometres further south from its capture site.

Soon after release, within 2.5 weeks, the animalmoved back north to its original capture site, withthe occasional foray back south. Clearly,movement of 50 kilometres or so is no problemeven for relatively small crocodiles, and theOctober records (blue triangles) in particular showthat the animal moved between initial capture andrelease sites within the same month. Nonetheless,most of the time, the crocodile moved within thefairly restricted area of its original river system.Crocodiles are territorial animals and althoughthey will travel about occasionally, they remain intheir territories for many years and will defendthem against other, intruding animals, particularly

Figure 6: Estuarine crocodile, Crocodylus porosus, with asatellite transmitter attached to its nuchal shield.

16

WAV

ES O

F TH

E FU

TURE

Figure 8: Satellite tracking data of a 3.9 metre saltwatercrocodile (example 2) on Cape York Peninsula, Queensland.

Figure 7: Satellite tracking data (example 1) from a smalladult estuarine crocodile that was released 60 km from itscapture site.

during the breeding season between October/November and February. Most crocodile attackson humans are not stimulated by the crocodile’sintention to feed but by territorial defense.

The second example (Figure 8) stems from a 3.9metre animal caught in the Nesbit River on CapeYork Peninsula. The map is on a finer scale thanin the previous example, and the tracking dataclearly emphasise that adult crocodiles staywithin well-defined territories most of the time,but that they do travel occasionally.

The crocodile traveled quite a long way upstreaminto the freshwater reaches of the river on oneoccasion. Although they are named estuarine orsaltwater crocodiles, C. porosus often enterfreshwater, and they build their nests only infreshwater. The other striking feature about thistracking record is the animal’s movement out tosea. Saltwater crocodiles – this time as theirname suggests – routinely swim out into theocean and are well able to navigate their wayback to their home river. In fact, crocodiles aresighted quite frequently on tourist beaches inCairns and Darwin.

17

RAD

IOTE

LEM

ETRY

INTH

EST

UDY

OF

WIL

DLI

FE

On a different note, a recent report by the BBC(September 2004) in the United Kingdomdescribes quite a different sort of tracking usingsatellites. In three pilot schemes initiated late in2004, satellite tracking technology is used totrack convicted burglars, robbers and car thieves.For the first time in Europe, offenders will befitted with transmitters as part of a communitysentence, or as a condition of their release fromprison, so that they can be tracked by satellites24 hours a day. Transmitters are monitored by acontrol station that records the location of theperson to within a few metres, and if the offender strays into a prohibited area, the policeare alerted.

(c) Physiological monitoringIn addition to tracking the movement of animals,technology has been developed to measurephysiological responses of animals by radiotelemetry. A common application is measuringbody temperatures of animals behaving naturally inthe wild. In temperature telemetry, a temperature-sensing device, or thermistor, is incorporated intothe transmitter. The thermistor causes thetransmitter to emit signals at a rate that dependson the temperature of the transmitter. Bycalibrating the interval between two signals againsttemperature before deploying the transmitter, it ispossible to obtain the body temperature of ananimal after surgically implanting a transmitter intothe animal’s body cavity (Figure 9).

In addition to body temperature, anotherimportant physiological parameter is heart rate.Heart rate is routinely monitored by veterinariansto assess the wellbeing of their patients. Itprovides insight not only into their cardiovascularphysiology, but also into an animal’s metabolicstatus. Each time the heart beats, a very weakelectric current is generated. Specialised heartrate transmitters can detect that current, and theelectric pulse generated by the heart triggers thetransmitter to emit a signal. Hence, heart rate canbe detected from a distance by the signalsreceived from a heart rate transmitter. I haveused this technology frequently to study thecardiovascular response to different thermalenvironments by free-ranging reptiles. In a recentstudy at Lakefield National Park on Cape YorkPeninsula with colleagues from the University ofQueensland and the Queensland EnvironmentalProtection Agency, we were interested inmeasuring physiological responses – bodytemperature regulation and heart rate – inrelation to diving in wild freshwater crocodiles.

Many lineages of terrestrial vertebrates, forexample crocodiles, seals, and otters, havesecondarily recolonised aquatic environments (thatis, they have evolved from ancestors that lived onland, and they have since adapted back to thewater again). They have presumably gainedselective advantages by adopting a semi-aquaticlifestyle. The pronounced physical differencesbetween air and water impose different challengeson anatomical and physiological characteristics ofsemi-aquatic animals. For example, many semi-aquatic vertebrates have independently evolvedanatomical features that assist movement in water(webbed feet, fin-like appendages, dorso-ventrallyflattened tails, etc), although behavioural patterns,such as foraging behaviour and avoidance ofpredation often encompass movement betweenwater and land. Behavioural responses aresupported by physiological functions, particularlymetabolic and cardiovascular, that must respondto the unique characteristics of both aquatic andterrestrial environments.

Crocodiles and alligators evolved from ancestralterrestrial animals in the Late Triassic andsecondarily became aquatic. Although moderncrocodiles are proficient in terrestrial locomotion,

Figure 9: Heart rate transmitters were attached tocrocodiles in the field (A), and body temperaturetransmitters were surgically implanted into the animals (B).The interval between two signals from a temperaturetransmitter depends on the temperature of the transmitter– a typical relationship is shown in (C).

18

WAV

ES O

F TH

E FU

TURE

Crocodiles perform most behaviours, such asfeeding and mating, under water and they mainlydo so in the early morning. Thermal sensitivity oflocomotor performance originates primarily fromthe temperature-induced constraints on muscle

much of their ecology is geared towards anaquatic lifestyle. Crocodiles possess a number ofcardiovascular characteristics that make themwell disposed for aquatic behaviour and diving. Inlaboratory trials, crocodiles can slow their heartrate dramatically to only 4 or 5 beats per minutein response to submergence in water. Thisslowing of heart rate, or bradycardia, is thought tobe an adaptation to diving because it results indecreased blood flow to the tissues and,therefore, less use of oxygen. The problem withlaboratory experiments is that the animals areoften stressed, and stress may elicit similarresponse of the cardiovascular system. This iswhere telemetry provides the solution, enablingmeasurements of heart rate in undisturbedcrocodiles.

Additionally, ectotherms such as crocodiles mustalso reconcile essential terrestrialthermoregulatory behaviour such as basking inthe sun on land with reproduction, socialinteractions, and feeding that are almost entirelyrestricted to water. In many reptilesthermoregulation is closely tied to thecardiovascular system. Numerous laboratorystudies have shown that heart rates duringheating are significantly higher than duringcooling. Animals gain an advantage from thisheart rate pattern (known as heart ratehysteresis) by being able to control rates ofheating and cooling, thereby increasing the timespent within a preferred thermal range during theday. Given the reliance of crocodilians on waterfor thermoregulation, we were particularlyinterested in determining the interaction betweendiving ecology and physiologicalthermoregulation.

A further challenge, beyond successfullydeploying body temperature and heart ratetransmitters, was to measure diving behaviour ofwild crocodiles. To achieve this, we used arecently developed device called a time-depthrecorder. These are archival devices; that is, theystore data in their internal memory for laterdownloading to a computer, rather than sendingsignals to a receiver. The recorders sensechanges in pressure with time, which can berelated to water depth. By attaching a time-depthrecorder to an animal it is possible to infer the

water depth the animal has been at, andtherefore obtain a record of natural divingbehaviour (Figure 10).

Interestingly, we found that animals were mostactive when their body temperature was low andbefore they basked in the sun. It was alwaysbelieved that reptiles and other ectotherms mustwarm up in the sun before they can becomeactive. Our data clearly showed that this is notthe case (Figure 11).

Figure 11: Diving activity of freshwater crocodiles inrelation to body temperature. Activity, expressed as numberof vertical movements per hour (blue area), was greatestbefore the animals warmed up in the late morning.

Figure 10: A typical example of a daily dive profile from afreshwater crocodile.

19

RAD

IOTE

LEM

ETRY

INTH

EST

UDY

OF

WIL

DLI

FE

performance and metabolic potential. It may bethat metabolic and muscular demands ofcrocodiles during their normal behaviour neverreach their full potential and that the reducedpotential at the lower body temperature does notpose a physiological constraint on activity.

Additionally, metabolic processes may berelatively temperature insensitive so that activitymay not be significantly curtailed as a result oflower body temperature in the morning. Warmingup later in the day, while not necessary foractivity, may be important for other physiologicalfunctions such as digestion of food, and recoveryfrom exercise.

Regardless of the asynchrony in the timing ofactivity and peak body temperatures, the animalsdid regulate their body temperature by emergingto bask, and by displaying the typicalcardiovascular changes in response to heat.Faster heart rates during heating than duringcooling significantly increases the efficacy ofthermoregulation by conferring control overheating and cooling rates. Interestingly, thedecrease of heart rates during diving wouldaugment the temperature induced decrease inheart rate in animals that dive after basking,although the two mechanisms operateindependently from each other (Figure 12).

Waves of the Future

New physiological applications of radio telemetryare constantly being developed. A radiotransmitter that measures blood flow is beingtested at the moment. Applications of suchtechnology are not restricted to wildlife studies.Monitoring of cardiovascular parameters andbody temperature are routine procedures inmedical and biological research laboratories, andincreasing numbers of telemetry systems arebeing developed to cater for that demand.

The multidisciplinary approach of applyinginnovative research on the physics of signaltransmission and environmental sensing to thestudy of natural systems has clearly led to majoradvances in biological research in recent years.The continuing development of new telemetrydevices will ensure the progress of wildliferesearch and management into the future.

References and Further Reading

Seebacher, F. and Alford, R. A., 1999, Movement and

microhabitat use of a terrestrial amphibian (Bufo marinus)

on a tropical island: seasonal variation and environmental

correlates, Journal of Herpetology, 33, pp 208-214.

Seebacher, F., Franklin, C. E. and Read, M., 2005, Diving

behaviour of a reptile (Crocodylus johnstoni) in the wild:

interactions with heart rate and body temperature,

Physiological and Biochemical Zoology, in press.

Cooke, S. J., Hinch, S. G., Wikelski, M., Andrews, R.D., Kuchel,

L. J., Wolcott, T. G. and Butler, P. J., 2004, Biotelemetry: a

mechanistic approach to ecology, Trends in Ecology and

Evolution 19, pp 334-343.

Grigg, G. C., Seebacher, F. and Franklin, C. E. (eds.), 2001,

Crocodilian Biology and Evolution, (Surrey Beatty, Chipping

Norton), ISBN 0 949324 89 2.

Figure 12: Heart rate and body temperature in relation todiving behaviour. The red shading indicates times when thecrocodile basked in the sun. Both heart rate and bodytemperature increased during basking. When the crocodileentered the water (blue shading), heart rate decreasedimmediately, thereby slowing cooling of the body.

20

WAV

ES O

F TH

E FU

TURE

Initially the Schools were annualevents, and the first four Schools,held between 1958 and 1961, werefor teachers. In 1962 ProfessorHarry Messel, the founder of theISS, changed the focus to honourexcellence in senior high schoolstudents and to encourage them toconsider careers in science.

A Truly International ScienceSchool One student from New Zealandattended the very first ScienceSchool in 1962, and overseasstudents have been a feature of theISS ever since. In 1967, tenstudents traveled from the USA toattend the School; the followingyear they were joined by five fromthe United Kingdom and five fromJapan. South-East Asia joined theISS in 1985 when studentsattended from Singapore, Malaysia,Thailand and the Philippines –sadly, that was the only time thePhilippines has participated. Chinahas sent five students to every ISSsince 1999, except for 2003 whenthe SARS epidemic restricted travelin the region and they reluctantlywithdrew.

The ISS2005 has studentsattending from nine countries intotal: Singapore, Malaysia, Thailand,Japan, China, the USA, the UK,New Zealand and, of course, everystate and territory of Australia.

The Great Lecturers One of the features of theInternational Schools is the lectureseries. Past ISS lecturers includeJames Watson, who won a NobelPrize for discovering the structure ofDNA, and Jerome Friedman, also aNobel laureate his for work onparticle physics. Sir Hermann Bondi(physicist and astronomer atCambridge University), MargaretBurbidge (astronomer with theHubble Space Telescope), Carl Sagan(famous astronomer and sciencebroadcaster) and Lord May(President of the Royal Society in theUK) have all given talks at the ISS.

And of course, who could forget thebrilliant science demonstrations ofProfessor Julius ‘Why is it so?’Sumner Miller, which were such apopular feature of the ISS that theyspawned a television show! Thesedays, Dr Karl Kruszelnicki – theFoundation’s Julius Sumner MillerFellow – entertains and enthusesthe ISS Scholars with his famousGreat Moments in Science.

History of the ISST H E P R O F E S S O R H A R R Y Messel International ScienceSchool has a long and distinguished history. The 140students attending Waves of the Future are the 33rd group togather at the University of Sydney for the Science School – inall, almost 4000 have attended a Science School since theybegan in 1958.

Between 1960 and 1979 the ISSlectures were shown on television – in fact, many people recallwaking up early on Sundays tomake sure they didn’t miss thetelecast! One member of theSchool of Physics here at theUniversity of Sydney is adamantthat the lectures shown on TV werea key part of her decision tobecome an astronomer.

Today, the ISS is no longer afeature of the television schedule – but we have moved on toembrace new technology. In 2003part of the lecture series wasbroadcast on the internet as a trialrun, and in 2005 the entire serieswill be webcast. Which means theISS has once again moved out ofthe lecture halls and out intopeople’s homes.

21

HIS

TORY

OF

THE

ISS

Science Schools for High School TeachersYear Teachers Theme1958 123 Selected Lectures in Modern Physics and the Astronomer’s Universe1959 123 Lecture notes on an introductory course in modern physics and nuclear

power and radioisotopes1960 123 From Nucleus to Universe1961 123 Space and the Atom

TOTAL 492

International Science Schools For High School StudentsYear Boys Girls Total Theme

1962 108 45 153 A Journey through Space and the Atom1963 104 51 155 The universe of Time and Space1964 106 53 159 Light and Life in the Universe1965 114 42 156 Time (and Relativity)1966 104 52 156 Atoms to Andromeda1967 101 57 158 Apollo and the Universe1968 109 20 129 Man in Inner and Outer Space1969 118 21 139 Nuclear Energy Today and Tomorrow1970 99 33 132 Pioneering in Outer space1971 87 35 122 Molecules to Man1972 95 28 123 Brain Mechanisms and the Control of Behaviour1973 93 29 122 Focus on the Stars1974 90 33 123 Solar Energy1975 76 43 119 Our Earth1977 54 50 104 Australian Animals and their Environment1979 63 52 115 Energy for Survival1981 50 65 115 Biological Manipulation of Life1983 67 51 118 Science Update 19831985 71 59 130 The Study of Populations1987 70 56 126 Highlights in Science1989 69 58 127 Today’s Science Tomorrow’s Technology1991 61 70 131 Living with the Environment1993 60 72 132 Carbon: Element of Energy and Life1995 55 80 135 Breakthrough! Creativity and Progress in Science1997 72 65 137 Light1999 73 66 139 Millennium Science2001 70 71 141 Impact Science2003 54 85 139 From Zero to Infinity

TOTALS 2293 1442 37352005 ?? ?? 140 Waves of the Future

PROFESSOR HALINARUBINSZTEIN-DUNLOPis Head of Physics andDirector of the Centre forBiophotonics and LaserScience at the Universityof Queensland. She is aprogram manager of oneof the scientificprograms of the Centreof Excellence inQuantum Computertechnology. ProfessorRubinsztein-Dunlop is

also a Research Director of theFaculty of Engineering, PhysicalSciences and Architecture at theUniversity of Queensland. Aftercompleting her PhD at the Universityof Gothenberg and ChalmersUniversity of Technology in SwedenHalina worked on the development oflaser based methods for ultra-sensitive trace element analysis andestablished a strong research groupin this area. She moved to theUniversity of Queensland in 1989.Halina’s research interests are inlaser physics, lasermicromanipulation, atom optics,quantum computing, linear andnonlinear high resolution laserspectroscopy, and nano-optics. Sheis one of the originators of laserenhanced ionisation spectroscopy,and is known for her work in lasermicromanipulation and atom optics.Professor Rubinsztein-Dunlop hasmore than 130 international journalpublications, 3 book chapters, and alarge number of internationalconference contributions.

22

Catch, Move and Twist withOptical Tweezers: Biophotonicsat workProfessor Halina Rubinsztein-Dunlop

IntroductionT H E P A S T C E N T U R Y H A S brought about an unprecedented number oftechnological breakthroughs, one of which is photonics. Photonics usesphotons instead of electrons to transmit, process and store information,providing a great gain in capacity and speed on information technology.This all-encompassing light-based optical technology is predicted tobecome the dominant technology for this new millennium. The inventionof lasers, which represent a concentrated source of monochromatic andhighly directional light, has had a tremendous impact on photonics.Since the demonstration of the first laser in 1960 and its very firstapplication in the correction of the retinal detachment, lasers haveilluminated all aspects of our lives, from barcode scanners in thesupermarkets and home entertainment, through high capacityinformation storage, to fibre optics communications, thus opening upnumerous opportunities for photonics.

23

24

WAV

ES O

F TH

E FU

TURE

Anew extension of photonics is biophotonics,the science of generating and harnessinglight to image, detect and manipulate

biological materials. Biophotonics is used inbiology to probe for molecular mechanisms,function and structure. It is used in medicine tostudy tissue and blood at the macro (large-scale)and micro (very small scale) organism level todetect, diagnose and treat diseases in a way that are non-invasive to the body.

Nature has used biophotonics as a basic principleof life from the beginning. Harnessing photons toachieve photosynthesis and the conversion ofphotons through a series of complex steps tocreate vision are the best examples ofbiophotonics at work. Biophotonics offers greathope for the early detection of diseases and thedevelopment of novel techniques for light-guidedand light-activated therapies. Lasers have alreadymade a great impact on general, neural andplastic surgeries. Laser technology allows theadministration of a burst of ultrashort pulses thatcan be used for improved imaging and for tissueengineering. Furthermore, biophotonics may beused to produce retinal implants for restoringvision. New and exciting applications ofbiophotonics are emerging very quickly and manylaboratories around the world are involved inthese rapidly expanding field.

Imagine focusing a laser beam specifically on to anorganelle, a structure within a living cell. Considerfurther that the beam can actually grasp that tinyentity and hold it in place. Nowimagine that while this microbeamacts as tweezers, a second beamserves as scalpel or scissors toconduct a delicate surgery on theorganelle. Is this possible, or just afigment of our imagination? Inother words, we are asking: canwe catch, move or rotatemicroscopic objects withouttouching them? Is it possible touse light to do all that?

Radiation pressureIt is well known and easily verifiable that electromagnetic waves carry energy. Perhaps you

also know that such waves can transport linearmomentum. If light is absorbed or reflected by anobject, momentum is transferred to this object.This transfer of momentum from light to anobject creates a force on the object. It is like acollision between two objects, only this collisionis slightly unusual in that one of the objects isactually light! That is, it is possible to exert apressure, called radiation pressure, on an objectby shining light on it.

Does it mean that we have to worry aboutopening the door on a bright day and beingknocked over by the light pouring in? Do we feela recoil force when we turn on a flashlight?Obviously not. The recoil force involved is toosmall in relation to the forces of our dailyexperience for us to feel it. So how strong are theforces exerted by light? How can we describethese forces?

How strong are the forces exerted by light?Assume that a parallel beam of light falls on anobject for some time t, and is entirely absorbedby the object. Maxwell showed that, if an amountof energy E is absorbed during this time interval,the magnitude of the momentum change, p, ofthe object due to the absorption will be given byp = E/c (where c is the speed of light). If theradiation is totally reflected back along itsoriginal path, the magnitude of the momentumchange of the object is twice the amount weestimated for the absorption case – so in thiscase the momentum change will be p = 2E/c.

This is similar to the situationwhere you throw a ball at anobject, like a milk bottle. Whenyou bounce a perfectly elastictennis ball off the bottle, itreceives twice as muchmomentum as when it isstruck by a perfectly inelasticball (a lump of putty, forexample, which would stickand not bounce off) of the

same mass and velocity. If the incidentradiation is partially absorbed and partiallyreflected, the momentum change of the objectis somewhere between the two of the aboveestimated values.

“... can we catch, move orrotate microscopicobjects without touchingthem? Is it possible touse light to do all that?”

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

From Newton’s second law we know that achange in momentum is related to a force give byF = p/t. To find expressions for the force exertedby radiation pressure in terms of the lightintensity, I, we have to consider radiation fallingon a flat area A that is perpendicular to the pathof the radiation. In a time interval t, the energyintercepted by the area A is E = IAt. If the energyis completely absorbed then we can see that theforce on an area A is given by the intensity Itimes the area A divided by the proportionalityconstant, which is c, the speed of light – that is,F = IA/c. If instead the radiation is totallyreflected, the force will be equal to twice theforce in the absorbing case.

Knowing this, we can estimate, for example, themagnitude of the force exerted on our body whenwe open the door on a bright day. For simplicitywe can assume that the area exposed to the sunis 40 cm wide by 1.65 m tall, and that averagepower per unit area delivered by the Sun isaround 0.2 W/m2. Knowing that the speed of lightis 3 x 108 m/s, the force exerted on our body willthen be 4.4 x 10-10 N. We could compare this withthe force that is needed to move an objectaround: if we want to lift a person weighing 60 kgthe force required to overcome gravity would be588 N. The difference between these two forcesis twelve orders of magnitude, so it would beimpossible to use the light force to do this task!

If we instead ask how much force would beneeded to move a microscopic object, such as abiological cell, a rough estimate shows that thesituation would be very different. If we assumethat we can approximate the cell by a cube ofvolume 2 x 10-6 m3, then it will have a mass ofabout 8 x 10-15 kg. So the force on it due togravity will be 7.84 x 10-14 N. If we go back to thequestion of what sort of intensity of light would beneeded to move around such an object, we willfind that we need an intensity of about 11.7W/m2, and the momentum transfer to producesuch a force would be 7.84 x 10-14 kg.m/s.Average light intensity from sun light at theEarth’s surface is about 0.2W/m2, which suggeststhat it is not even enough to move a microscopicobject such as a biological cell around. Thequestion is whether it is possible to create lightwith sufficient intensity, and therefore sufficient

momentum, for moving microscopic objectsaround in some way?

Optical micromanipulationThe concept that light can exert forces is not anew one. In the seventeenth century Kepplerproposed that light pressure is what causescomet tails to always point away from the Sun.In later times Newton’s corpuscular theory of lightwas the inspiration for early experiments to try tomeasure the radiation pressure of light, but all ofthese early attempts failed as the measuredforces could always be explained by moremundane causes such as convection. Theradiometer invented by Sir William Crookes in1873 was thought for some time to demonstrateradiation pressure, but in fact its action is due tothe forces of molecular bombardment onsurfaces heated by light rather than light directly.

[FIGURE 2 HERE]

The first experimental measurement of radiationpressure was made between 1901 and 1903 by

Figure 1. Tail of a comet points away for the Sun.

Figure 2. The arrangement of torsion balance technique asproposed by Nichols and Hull for a measurement ofradiation pressure.

25

26

WAV

ES O

F TH

E FU

TURE

Nichols and Hull1 and by Lebedev2, thirty yearsafter Maxwell’s theoretical predication of thiseffect. In order to measure the radiation pressureforce they used a torsion balance technique. Amirror (a perfect reflector) and a black disk (aperfect absorber) are connected by a horizontalrod suspended from a fine fibre. Normal-incidence light striking the black disk iscompletely absorbed, so all of the momentum istransferred to the disk. Normal-incidence lightstriking the mirror is totally reflected, and hencethe momentum transferred to the mirror is twiceas great as that transferred to the disk. Theradiation pressure is determined by measuringthe angle through which the horizontal connectingrod rotates. The measured radiation pressuresare very small (about 5x10-6 N/m2). How can weincrease the radiation pressure force that isavailable to us? Maybe one way of doing it is tofocus the light beam.

The invention of the laser in 1960 enabled anunprecedented development within the opticalresearch and applications. With researchers nowbeing able to focus very strong beams ofcoherent light on to objects, the effects ofradiation pressure were no longer limited to afeeble force observable only under high vacuumconditions. Early demonstration of the radiationpressure by Ashkin3 involved optical levitation of asmall particle by means of vertically directed laserbeam. In this famous experiment, which appearsin many undergraduate physics textbooks today, atransparent glass sphere of about 20 microns indiameter is lifted in air about 1 cm above a glassplate. The scattered laser light makes thelevitated spheres visible to the naked eye, astriking display of laser light force.

We can do a simple calculation of the magnitudeof the radiation pressure force. Assume we havea 10 mW Helium-Neon (He-Ne) laser beam,focused to a spot one wavelength (λ) across,onto an object of the same diameter with densityρ = 1 g/cm3. If the particle were 10% reflectiveand acted as a flat mirror, then the momentumchange of the laser light each second would be2x10-3/(3x108) kg.m/s. By conservation ofmomentum, this results in a force on the particleof approximately 6.7x10-12 N. Estimating the massof the particle by ρπr3/6 gives an acceleration on

the particle of 5x104 m/s2, more than 1000 timesthe acceleration due to gravity!

Such calculations led to many experiments in theearly 1970s that used laser beams in variousconfigurations to move microscopic particlesaround. The particles used in these experimentswere transparent, as the heating effects on anabsorbing particle were thought likely to obscureradiation pressure effects similar to the way thatheating and convection foiled earlier attempts tomeasure radiation pressure. Ashkin’s paper from1972 entitled ‘The Pressure of Light’ (seereference 3) discusses some of theseexperiments in detail. One of the experimentsdescribed there considers small transparentplastic spheres dispersed in water, placed in aglass cell under a microscope and illuminatedfrom below by a focused laser beam. In hisobservation Ashkin noted that not only was therea force propelling transparent particles in thedirection of beam propagation, but spheres nearthe edge of the beam experienced a force pullingthem into the centre of the beam.

[FIGURE 3 HERE]

This effect can be explained by considering a pairof rays of a Gaussian-shaped laser beam a and bsituated symmetrically with respect to the centreof a sphere having refractive index higher thanthat of the medium. (A Gaussian beam is onewhere the cross-section intensity profile is aGaussian, or ‘normal function’, shape – bright inthe centre, becoming less intense radiallyoutwards.) For a sphere situated off the axis of a

Figure 3. Refraction of a pair of rays by a high refractiveindex sphere. A pair of rays from a Gaussian beam arerefracted as the pass through a sphere with refractiveindex higher than the surrounding medium. Since ray a isstronger than ray b, the force on the sphere due torefraction of ray a, Fa, is larger than Fb, and the sphere ispulled into the centre of the beam propelled in the directionof beam propagation by radiation pressure.

27

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

Gaussian beam, ray a is stronger than ray b. Ontransmission through the sphere, rays a and bare bent, giving rise to forces Fa and Fb in theopposite directions to the momentum change ofthe rays. Since ray a is stronger than ray b, theforce Fa is greater than force Fb. The resultantforce F = Fa + Fb has components both in thedirection of the beam propagation and towardsthe centre of the beam, so one expects a nettransverse force pulling the high index sphereinto the centre of the beam where the intensity ismaximum (see figure 3).

That means that a particle that is mostlytransparent and has refractive index greater thanthe surrounding medium experiences a gradientforce, which is greater than the absorption force,when irradiated by a tightly focused beam oflight, and is thus trapped in the region of highestintensity. Under favourable circumstances this willeven mean that such particles will move backalong the beam axis to a beam waist, giving threedimensional trapping.

A highly absorbing particle, however, willexperience a much greater absorption force thangradient force and will be ‘pushed’ away from theregion of high intensity. These particles cannottherefore be trapped three-dimensionally usingGaussian laser beams, the most common profileof laser light. They can, however, be trappedusing a ring of light, a so called ‘doughnut beam’.Absorbing particles are pushed away from theintense region of the beam, either away from thebeam or, preferably, into the central dark regionwhere they will be trapped.

With the ordinary Gaussian beam, in 1978Ashkin4 showed theoretically that a gradient forcecould be produced in the direction opposite tobeam propagation, and that small dielectricparticles could be three-dimensionally trappedusing a single laser beam. In 1986 Ashkin et al.5

reported the observation of the first single beamgradient force optical trap, also known as optical tweezers.

The distinguishing feature of this trap was that itwas the first all-optical single-beam trap. Agradient force, proportional to and in the directionof the beam intensity gradient, is produced by

strongly focusing the beam. This gradient forcehas both radial and axial components, and theaxial component can be made strong enough toovercome the gravitational and scattering forceson a small dielectric particle, and thus confining itto the most intense region, the beam waist. Thistrap does not rely on the balance of thescattering force (radiation pressure) and gravity.

Another, simple way of treating forces acting on amicroscopic particle exposed to a highly focusedlight is by analysing the rays a and b of lightincident on the particle as shown in figure 4.

[FIGURE 4 HERE]

When the light ray entering a particle withrefractive index greater than that of thesurrounding medium is bent toward the normal,the change in momentum of the light results in aforce on the particle. If we follow the rays a andb, as shown in the figure 4, we can see theforces produced by change of the momentum. Ifthe centre of the particle is below the focal spotof the beam, the particle will be moved upwardsto the most intense part of the beam. If theparticle is above the centre of the beam it will bepushed down, and if it is situated to the side ofthe waist of the beam it will be pulled sidewaysinto the centre. The total effect will give a three-dimensional trap.

The essential elements of the single-beamgradient optical trap, or optical tweezers, is a highnumerical aperture lens, necessary to bring thetrapping beam to the tightest possible focus. This

Figure 4. Ray optics model of optical trapping oftransparent spheres. Restoring force on a sphere when it isdisplaced form the focus.

28

WAV

ES O

F TH

E FU

TURE

tight focussing is responsible for the axialgradient force and a microscope objective istypically used for this purpose. The typical set upfor optical tweezers is shown below, where thesame objective is used for both producing thetrap and imaging.

In figure 6 we show a photograph of a simpleoptical tweezers set-up. It consists of a highnumerical aperture objective, a piezo-driven stagewith microscope slide and cover slip containingparticles dispersed in water, the laser source andsome optics to shape the beam entering theobjective. The same objective is used here forobservation of the trapped particles.

Because it consists only of a single laser beam,the optical tweezers trap is extremely simple to

use – once a particle is trapped at the beamwaist it can be manipulated relative to itssurroundings by either changing the angle atwhich the beam enters the back pupil of theobjective (that is, by moving the laser beam) or bymoving the cell containing the sample. Both ofthese are relatively uncomplicated, making opticaltweezers the first truly practicablemicromanipulation tool.

Following the discovery in 1987 that livebiological specimens could be optically trappedfor quite long periods of time (from minutes tohours) and still remain viable, the single-beamgradient trap has found many applications inbiologically related fields, and its popularity as abiophysics and biology tool is steadily increasing.In response to the demand of researchers, thereare several commercially produced opticaltweezers instruments on the market.

Early application of optical micromanipulationmostly made use of the advantage that livespecimens can be manipulated in a closedenvironment in a very controlled way, leading tothe use of optical tweezers in applications whichpreviously employed micropippetes. With theadditional tool of a cutting beam, which can beachieved with the same arrangement as tweezersbut with the laser light delivered in short pulsesand the wavelength chosen so that it is absorbedby the specimen, optical tweezers have foundtheir way to wide variety of research.

Optical tweezers have been used to trap dielectricspheres, viruses, bacteria, living cells, organelles,small metal particles, and even strands of DNA.Applications include confinement andorganization (for example, for cell sorting),tracking of movement (of bacteria, for example),application and measurement of small forces,and altering of larger structures (such as cellmembranes). Other uses for optical traps havebeen the study of molecular motors and thephysical properties of DNA. In both areas, abiological specimen is biochemically attached to amicron-sized glass or polystyrene bead that isthen trapped. By attaching a single molecularmotor (such as kinesin, myosin or RNApolymerase) to such a bead, researchers havebeen able to probe molecular motor properties.

image

Laser light

Microscope objective N.A.~1.3

Figure 6. Photograph of the microscope constructed foroptical tweezers. The set-up is versatile with respect toadding and removing optics and photodetectors.

Figure 5. The essential elements of a single beam opticaltweezers trap. The same objective is used for bothproducing the trap and imaging.

29

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

The studies have started to answer questionssuch as: Does the motor take individual steps?What is the step size? How much force can themotor produce? Similarly, by attaching the beadsto the ends of single pieces of DNA, experimentshave measured the elasticity of the DNA, as wellas the forces under which DNA breaks orundergoes a phase transition.

When two trapping beams are introduced (adouble-beam optical tweezers), the microscopestudies can be performed on objects that have tobe stretched, aligned or turned. Using thistechnique, Chu et al.6 studied recoil behaviourand viscoelastic properties of DNA molecules. Themolecules were stretched out and fixed with thetweezers and then viewed with scanningtunnelling or atomic force microscopes.

When optical tweezers are combined with a lasermicro-beam (also called an optical scalpel)controlled cell fusion can be carried out7. Thebasic idea of the use of a laser scalpel forintracellular microsurgery on biological objects isthat the laser cutting beam is strongly focusedonto the object from a large numerical aperturemicroscope objective (like the beam of the opticaltweezers). This ensures that the laser scalpel hasvery short effective depth of field, implying thatthere will be enough power density for thedesired effect only at very limited depth in theobject (1-2 µm). Above and below this depth, thelight intensity will cause no harm to the tissue.This ensures that the micro-beam surgery will becarried out in the interior of an unperforated cellwith no damage to cell walls or membranes.

[FIGURE 7 HERE]

Another use of optical tweezers and scissors isfor cell fusion. In this case the optical trap iscombined with a pulsed UV laser micro-beam.The two selected cells are brought into closecontact by the optical tweezers. Once inside thetrap, the two cells can be fused by applyingseveral pulses of the UV laser micro-beam. Withthis technique a selective fusion of two cells isdone without critical chemical or electricaltreatment. Laser induced cell fusion shouldprovide an increased selectivity and efficiency ingenerating viable hybrid cells in the future8.

An important application of combined usage ofoptical tweezers and UV-laser micro-beam ismanipulation of gametes and early embryos.Using these techniques the fertilization processescan be studied in more detail leading toincreased efficiency of in-vitro fertilization. Thecombination of a UV-laser micro-beam and anoptical tweezers was first suggested by Ng et al.9.Subsequently Schütze et al.10 successfully drilleda hole into the zone pellucid and inserted a singlesperm through the laser drilled hole into thepervitelline space using these combinedtechniques.

Figure 9 below shows another example of the useof a micro-beam. Here sperm motility is stoppedwith a few laser pulses placed close to thewaving tail (a reversible process). Cutting thesperm tail is possible by focusing the laser ontothe tail and performing a single laser shot (anirreversible process).

We can conclude that a combined system ofoptical tweezers, laser micro-beam and laser

(a) (b) (c) (d)

Figure 8. Trapping and cutting beam for the cell fusion. Twohuman lymphocyte cells (No. 1 and 2) were broughttogether by means of the optical tweezers (a). Laser pulsesof the cutting beam (dark spot in b denoted by the arrow)perforated the outer cell membrane and both cells fusedtogether (c, d). 40 seconds later the cells start to fuse (c).160s after the wall perforation (d).

Figure 7. Laser microbeams is used as scissors to cutchromosomes. In the visible range (400-700 nm), thesebeams can be coupled with imaging methods and opticaltweezers. The arrows indicate where the cutting beam wasfocused on mitotic chromosomes in a living cell.

30

WAV

ES O

F TH

E FU

TURE

induced fluorescence detection converts theordinary light microscope from being a passiveanalytical tool into a preparative and manipulativeinstrument that allows micro-manipulation ofbiological objects without any mechanical contact.

Optically driven micromachinesIt isn’t uncommon for scientists to be inspired byscience fiction, and good science fiction has astrong basis in real science. Modern sciencefiction is well populated with ‘nanobots’ and‘microbots’, tiny robots that are injected into thebloodstream where they perform tasks rangingfrom repairing damage and curing diseases tocontrolling human thoughts and actions. Theminiature robot concept may seem far-fetchedtoday, yet technological advances in the area ofmicromachine research are bringing thepossibility of devices like these much closer toreality than fiction.

Micromachines have potential advantages overmacromachines in their mobility, informationtransfer rates and energy efficiency. The best-known micromachines are themicroelectromechanical systems (MEMS) thatincorporate mechanics together with electronicson a miniature scale. More along the lines of the

fictional nanobots, microrobots have beenfabricated with ‘elbows’ and ‘fingers’ that arecapable of manipulating micrometer-sizedobjects11. Production of working micromachineshas motivated research into microdeviceproduction methods, surface engineering of thesubstrates and driving mechanisms for themachines. Microscopic electromagnetic motorsand piezoelectric and electrostatic actuators havebeen incorporated into MEMS. More novel drivingmechanisms that avoid contact with the macroworld suggested for micromachines include adielectric fluid motor based on convection, anopto-micro-engine based on the same principlesas a Crookes’ radiometer, and the use of opticaltorque from strong sources of laser light.

Microdevices are getting smaller and are actuallybecoming nanodevices, and laser light has playeda role in many of the advances that are makingthis possible. Together with other moderntechnologies, lasers have been used in almost allaspects of micro and nanodevice research,including their fabrication, construction, and as asource of torque to provide a driving mechanismfor the devices. Laser light has been used topolymerize resins to produce structures withnanometer-sized features12 and to bring togetherparts of a two-element moving microsystem13. Inboth these and other experiments light was alsoused to drive the rotation of the elements.

Light as a source of torqueThere are two basic ways that light can be used todrive the rotation of an object. In the first case, thetorque originates from the light itself carryingangular momentum, which can then be transferredto the object by processes such as transmission,reflection and absorption. Types of light carryingangular momentum include elliptically polarizedbeams and beams with helical phase structure. Inthe second category, the torque originates from theshape of the object. Radiation pressure can act onasymmetries in the object’s shape in a similar wayto wind on the blades of a windmill: light deflectedby the particle exerts torque to drive rotation.

Angular momentum due to polarizationScientists have known for a long time that certaintypes of light carry angular momentum, and so in

Figure 9. Trapping of sperm for in vitro fertilization.Stopping sperm motility is done with a few laser shotsplaced close to the waving tail (reversible process). Cuttingthe sperm tail is possible by focusing the laser onto the tailand performing a single laser shot (irreversible process).

31

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

principle that light could be used to exert torqueon an object. In practice though, the effects ofoptical angular momentum are hard to observe, asthey represent very small quantities. For example,the angular momentum flux carried by a circularlypolarized 10 mW He-Ne laser beam is of the orderof 10-18 Nm, which is millions of times smaller thanthe torque driving the balance wheel of amechanical wrist watch. Each photon of circularlypolarized light carries +h of angular momentum(where h =h/2π, and h is Planck’s constant).

In 1936 Richard Beth carried out the firstmeasurement of torque produced by light14, inwhich he used a series of waveplates suspendedon a torsion fiber (see figure 10a). When linearlypolarized light is passed through a birefringentmaterial of the correct thickness (λ/4), itbecomes circularly polarized. If that circularlypolarized light is passed through a linear polariser(λ/2 plate), the handedness of the polarization isreversed and, in the process, angular momentum

is transferred to the plate. In Beth’s experimentcircularly polarized light was passed twicethrough a wave plate suspended on a torsionfibre, and the resulting period of oscillationmeasured. In his experiments he was able toconfirm that the sign and magnitude of the effectagreed with theory – the light did indeed possessand transfer angular momentum.

A modern version of this experiment (figure 10b)uses laser light, optical tweezers and microscopicwaveplates to observe the same torque, and isthe basic idea behind the use of circularlypolarized light to drive the rotation of microscopicelements15. A laser beam is focussed to a verysmall spot, providing an extremely intense lightsource, and is passed through sphericalbirefringent calcium carbonate crystal. The calcite

λ/4

λ/4

λ/2

Aluminiumlayer

fast

fast

fast

slow

slow

slow

Torsionfibre

λ/4

λ/4

λ/2

Aluminiumlayer

fast

fast

fast

slow

slow

slow

Torsionfibre

(a)

Figure 11. 11(a) The electric field vector of circularlypolarized light. Each cycle the electric field vector rotates2π radians. The rotation of the electric field vector at heoptical frequency is associated with the spin angularmomentum of a circularly polarized photon.11(b) The phase fronts of a helical beam of charge l = 3,and the intensity pattern when interfered with a planewave. The intensity pattern when interfered with a planewave is the configuration used in the propeller beamexperiments.

Figure 10. Schematic diagramfor measurement of torqueproduced by polarized light asin the original Bethexperiment, (a), and asperformed when usingspherical birefringent crystalsin optical tweezers withpolarized laser beams, (b).

(a)

(b)

32

WAV

ES O

F TH

E FU

TURE

particles can act as wave-plates. On passagethrough a crystal, different components of theincident light will undergo different phase shiftsand the electric field vector will rotate (figure11a). This will introduce a change in the angularmomentum carried by the light, and there will bea corresponding torque on the material.Depending on the polarization of the incidentbeam, the particles either become aligned withthe plane of polarization (and thus can be rotatedthrough specified angles) or spin with constantrotation frequency. The light transmitted throughthe crystal will become elliptically polarised.Effectively, the measurement of the change inpolarisation state of light on transmission thoughthe birefringent crystal will allow determination ofthe rotation rate of the crystal and also an opticalmeasurement of torque.

Polarized light can also exert torque on an objectif the material is absorbing – in that case if thelight carries both linear and angular momentum,both will be absorbed by the material. In thiscase, not only does the rotating object feel forcein the direction of light propagation, it is alsoheated as it absorbs energy from the field. Thismeans that high rotation speeds cannot beachieved, since the increase in laser powerrequired to increase the spinning rate will burnthe rotating element. So the phenomenon is quiteinteresting scientifically but due to these extraeffects is less useful as a possible drivingmechanism for rotating microscopic objects.

Helical light and propeller beamsThe source of the polarization torque is therotation in time of the electric field vector of thelight field. Light can also exert torque if thewavefront associated with the field is rotating intime. Here the wave has a helical phasestructure (figure 11b) while its electric field maybe rotating or not. We call the angularmomentum from polarization spin angularmomentum, and the angular momentum due tothe helical structure orbital angular momentum.Just as passing light through a wave plate canproduce circularly polarized light, helical light canbe produced when light is passed through aphase plate. Combinations of the two are ofcourse possible if the light is prepared usingboth wave and phase plates.

Depending on the type of phase plate used, lightwith many intertwined helices can be created.Orbital angular momentum can be transferred toan object by absorption or reflection, and canalso be used to spin tiny particles trapped usingoptical tweezers 16. When light absorption is thecause of the angular momentum transfer, bothspin and orbital angular momentum can betransferred to the material at the same time. It isthen possible to change the rotation rates of thespinning objects by a simple rotation of the waveplates, or by reversal of the helicity of the wave.

These helical wavefronts can be interfered with aplane wave so that a spiral interference patternwith three arms is produced within an opticaltweezers trap (see figure 11b). The arms of thepattern rotate when the path length of theinterferometer is changed, and so particles thatare trapped in the bright regions of the arms arerotated too17. These ‘propeller beams’ offer atechnique that does not depend on intrinsicproperties of the particles (e.g. birefringence) andalso avoid absorption of the field – the methodrequires that the particles be transparent andhave a higher refractive index than theirsurrounding medium.

Optical tweezers and light-drivenmachinesMicroscopic particles trapped in the tightly focusedbeam of an optical tweezers trap often tend torotate, either due to their own shape or throughinteraction with light carrying angular momentum.Both these effects have been used to drive therotation of microscopic machine elements.

Recently, a method to build microscopic light-driven rotors was reported18 where opticaltweezers were used in both the production andmanipulation of the rotors. Microscopic particlesof arbitrary shape were produced by a two-photon polymerization method. In this experimentthe researchers used a resin that, whenpolymerized, results in a glass-like material withrefractive index n=1.56, which is ideal forhandling using optical tweezers. The light sourcefor the polymerization process was at the focusintense enough to initiate two-photon excitation.To build the structures, the beam focus is moved

33

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

along a preprogrammed trajectory and a three-dimensional shape is built up from the line alongwhich the resin is hardened. Arbitrary shapescould be constructed using this method withfeatures of about 0.5 µm in size. The researchersexperimented with different shapes includinghelices, sprinklers and propellers, and found thatthe most efficient shape for rotation was thesprinkler shape, with an added central linear axisto improve stability in the optical tweezers trap.Using 20 mW in the optical trap producedrotation rates of several Hertz.

[FIGURE 12 HERE]

Using this method, a complex micro machineconsisting of two engaged cogwheels rotated by alight drive rotor was constructed, where the rotorwas held and driven by optical tweezers and thecogwheels rotated on axes fixed to the glasssurface (see figure 12). The same techniques couldbe used to produce much more complicatedarrangements, offering a promising method forconstructing micron sized light driven machines.

Micromachine elements have also been driven bylight by transferring the angular momentum fromphotons to a microscopic particle, which wasthen used to drive rotation of a microfabricatedelement19. Optical tweezers again played animportant role in the experiment. Two fullysteerable traps were used to hold and manipulatethe light driven rotor and the microfabricatedelement. The rotor consisted of a microscopicfragment of birefringent calcite, which could beinduced to spin at hundreds of Hertz whentrapped in circularly polarized light. The

microfabricated elements were cog-like shapes10µm in diameter and 0.5µm thick, made of SiO2

using a photolithographic double-liftoff technique.

Once trapped, the rotor was moved next to atrapped cog (see figure 13). Then, when the lightwas made circularly polarized through rotation ofa λ/4 plate, the calcite spun, inducing the nearbycog to rotate also. In this experiment the torque istransferred from the rotor to the machine elementvia the fluid between them. For both particles, theoptical tweezers acts as an axle for them torotate about.

This proof of principle experiment has at leastone obvious drawback: energy is being lost to thefluid that could be used to drive the element. Ifthe machine elements could be made from abirefringent material, then light could be used todrive the elements directly, avoiding the need forthe calcite. Simple birefringent structures of asimilar size to these cogs have already beenproduced20 that show the same behaviour inpolarized light as calcite, so production of morecomplicated shapes may soon follow.

The advantages of light driven microscopic rotorsand machine elements are obvious: the non

spinning CaCO3

crystal

MicrofabricatedSiO2 “cog”

spinning CaCO3

crystal

MicrofabricatedSiO2 “cog”

Figure 13. Optically driven and assembled micromachine.Dual fully steerable optical tweezers are used. The first trapis used to trap a birefringent crystal, the other one is usedto trap the “cog”. The light is circularly polarized and thecrystal rotates. The other trap brins the “cog” close to thespinning crystal. The rotation of the cog is induced. Thetorque here is transferred from the rotor to the machineelement via the fluid surroundings. The dual opticaltweezers provide axles for the rotor and the machineelement.

Figure12. Light-driven micromachinery produced anddriven by light. The arrows point to the threeinterconnected machine cogs.

34

WAV

ES O

F TH

E FU

TURE

contact nature of the driving mechanism meansthat these micromachines can be operated in anymicroscopic systems that are accessible by laserlight. Micro-rotors may find application asinstruments for measuring properties of biologicalsystems such as torsional elasticity of biologicalpolymers or microscopic viscosity. Spinningbirefringent particles have already been used toturn biological specimens around to view themfrom different angles21 and for studies of cellmembranes.

By monitoring the change in circular polarization oflight passing through an object, the reaction torqueon the object can be found. This idea wasexploited as the basis for an optically drivenviscometer, an instrument that can measureviscosity of liquids. The viscosity of a liquid can bedetermined by measuring the torque required torotate a sphere immersed in the liquid at aconstant angular velocity – this concept has beenimplemented at the micrometer scale using asystem based on optical tweezers. Birefringentspheres of synthetically grown material (see figure14) were trapped three dimensionally in acircularly polarized optical tweezers trap and thefrequency of rotation as well as the polarizationchange of light passing through the particle weremeasured to determine the viscosity of thesurrounding fluid. This method, which is based ondirect optical measurement of the rotation of aspherical probe particle in a stationary position,

allows highly localized measurements to be made,since the probe particle does not move in thesurrounding medium. At the same time the use ofspherical probe particles greatly simplifies thetheoretical analysis of the fluid flow. Also, since theprobe particle is rotationally driven by the opticaltrap, the rotation rate can be readily controlled.

Using this technique the viscosity of water andsome other liquids in extremely small volumes ofsample have been measured, demonstratingfurther possible applicability of this method.

Recently we have combined the techniques ofoptical scalpel and rotating tweezers to measurethe properties of liquid inside a biological cell. Inthis experiment a biological cell was exposed tovery sharply focused pulsed laser with the pulsesof the order of femtosecond (that is, 10-12 of asecond) in length for a short period of time. Theselaser pulses created a hole in the membrane ofthe cell. Subsequently optical tweezers were usedto insert a birefringent spherical crystal into thecell, and then rotating tweezers were used tomeasure the viscosity of the liquid inside the cell.Figure 15 shows the cell with the crystal inside it.

With further advances in the fabrication materialsand methods, the manufacture of microscopicfluid pumps for extremely localized delivery ofchemicals will become possible. Optical tweezersand optical scissors have been joined by the

Figure 14. (a) Optical microscope image of a typical vaterite crystal used for viscosity measurements. (b) Scanning electronmicroscope image of a vaterite crystal.

35

CAT

CH

,MO

VEAN

DTW

IST

WIT

HO

PTIC

ALTW

EEZE

RS:

BIO

PHO

TON

ICS

ATW

ORK

optical spanner and the propeller beam, providingtools to trap, manipulate, cut, align, turn over androtate a wide range of micro- and nano-objects. Itseems certain that these laser tools will continueto make their mark in the world of microengineering.

References

1 E. Nichols and G. Hall, Phys. Rev. 13:307 (1901).

2 P. Lebedev. Ann. Phys. 6:433 (1901).

3 A. Ashkin, Scientific American 226,(2): 62 (1972).

4 A. Ashkin, Phys. Rev. Lett. 19: 283 (1978).

5 A. Ashkin, JM. Dziedzic, JE Bjorkholm and S. Chu, Opt. Lett.

11(5 : 288 (1986).

6 TT Perkins, De. Smith and S. Chu, Science 253, 861

(1991).

7 K. Schütze and A. Clement_Sengewald, Nature 368, 667

(1994).

8 WH. Wright, GH. Sonek, Y. Tadir and MW. Berns, IEEE

Journal of Quantum Electronics 26, 2149 (1990).

9 Y. Ng et al., J. Assist. Reprod. Genet. 9, 191 (1992)

10 K. Schutze, A. Clement-Sengewald and A. Ashkin, Fert.

Steril. 61, 783 (1994)

11 E.W.H. Jager, O. Inganäs and I. Lundström, Science 288,

2335 (2000)

12 P. Galajda and P. Ormos, Appl. Phys. Lett. 78, 249 (2001)

13 M.E.J. Friese et al., Appl. Phys. Lett. 78, 1 (2001)

14 R.A. Beth, Phys. Rev. 50, 115 (1936)

15 M. Friese, T.A.Nieminen, N.R. Heckenberg and H.

Rubinsztein-Dunlop, Nature 394, 348 (1998)

16 H. He, M.E.J. Friese, N.R. Heckenberg and H. Rubunsztein-

Dunlop, Phys. Rev. Lett. 75, 826 (1995)

17 L. Paterson et al, Science 292, 912 (2001)

18 P. Galajda and P. Ormos. Appl. Phys. Lett. 78, 1 (2001).

19 MEJ. Friese, TA. Nieminen, NR. Heckenberg and H.

Rubinsztein-Dunlop, Nature 394, 348 (1998).

20 E. Higurashi, R. Sawadaand T. Ito, J. Michromech Microeng

11, 140 (2001)

21 S. Bayoudh et al., J Microsc 203, 214 (2001)

Figure15. Biological cell with a birefringent spherical crystal inside it.

36

WAV

ES O

F TH

E FU

TURE

prestigious Federal PolytechnicSchool (or Academy) in Zurich,Switzerland. He was 16, two yearsyounger than his fellow applicants.He did outstandingly well in physicsand mathematics, but failed thenon-science subjects, doingespecially badly in French – so hewas not accepted. So in that sameyear, he continued his studies atthe Canton school in Aargau (alsocalled Aarau). He studied well, andthis time, he passed the entryexams into the Federal PolytechnicSchool.

So the next year, he finally startedstudying at the Federal Polytechnicin Zurich (even though he was nowone year younger than most of hisfellow students). Also in the year1896, even though he was only 16 years old, he wrote a brilliantessay that led directly to his laterwork in relativity.

So he definitely did not fail his highschool, and definitely was not apoor student.

So how did the myth that he failedhigh school start?

First, Einstein did not win the 1921Nobel Prize in Physics for his workon Relativity. Let’s back up a little.Back in 1905, Einstein had thebiggest year of his life. He wrote,with the help of his wife, Mileva,five ground-breaking papers that,according to the EncyclopaediaBritannica “forever changed Man’sview of the Universe”. Any scientistwould have been proud to writeeven one of these magnificentpapers – but Albert published fiveof them in one year!

One paper, of course, dealt withRelativity – what happens toobjects as they move relative toother objects. Another paper provedthat atoms and molecules had toexist, based on the fact that youcould see tiny particles jiggingaround when you looked at a dropof water through a microscope. Athird paper looked at a strangeproperty of light – the PhotoelectricEffect. Plants and solar cells do thePhotoelectric Effect, when they turnlight into electricity. His paperexplained the Photoelectric Effect.Relativity may have captured thepublic’s consciousness, but it was

the unglamorous PhotoelectricEffect that got him the Nobel Prize.Well, that’s one myth out of the way.

Second, Einstein definitely did notfail at high school. Einstein wasborn on 14 March in Ulm, inGermany, in 1879. The next year,his family moved to Munich. At theage of 7, he started school inMunich. At the age of 9, he enteredthe Luitpold-Gymnasium. By the ageof 12 he was studying calculus.Now this was very advanced,because the students wouldnormally study calculus when theywere 15 years old. He was verygood at the sciences. But, becausethe 19th-century German educationsystem was very harsh andregimented, he didn’t really develophis non-mathematical skills (such ashistory, languages, music andgeography). In fact, it was hismother, not his school, whoencouraged him to study the violin– and he did quite well at that as well.

In 1895, he sat the entranceexaminations to get into the

A T T H E E N D O F the 20th Century, Time magazine votedAlbert Einstein to be the Man of The Century. Albert was thedude who came up with all that really weird Relativity stuff –and he was your genuine certified Mega Brain. After all, weare told that he even won the Nobel Prize for his work inRelativity. On the other hand, generations of school kids haveconsoled themselves over their poor school marks with thebelief that Einstein failed at school. Some motivationalspeakers also make this claim – but this claim is as wrongas the claim about the Nobel Prize.

Einstein Failed SchoolBy Dr Karl Kruszelnicki

37

EIN

STEI

NFA

ILED

SCH

OO

L

Easy. In 1896, which was Einstein’slast year at the school in Aargau,the school’s system of marking was reversed.

A grading of “6”, which hadpreviously been the lowest mark,was now the highest mark. And so,a grading of “1”, which had beenthe highest mark, was now thelowest mark.

And so, anybody looking upEinstein’s grades would see that hehad scored lots of grades around“1” – which under the new markingscheme, meant a “fail”.

And that means that schoolkidscan’t use that mythconception as acrutch any more – they’ll just haveto work harder... FROM Dr Karl’s book Mythconceptions

(Harper Collins Publishers)

DR CLIVE BALDOCKBefore attendinguniversity, Clive wasemployed as a medicalphysics technician in theNuclear MedicineDepartment of St. Paul’sHospital, London and theInstitute of Urology,University of London. Hesubsequently obtainedhis BSc (Hons) in Physicsfrom the University ofSussex, Brighton in

1987. He then worked as a BasicGrade Medical Physicist in theDepartment of Clinical Physics andBioengineering, Guy’s Hospital,London and the United Medical andDental Schools (UMDS), University of London.

In September 1993, Clive moved tothe Medical Physics and NuclearMedicine Departments, Royal SussexCounty Hospital, Brighton Health CareNHS Trust, as part of the teamproviding the scientific servicesupporting the Nuclear MedicineDepartment and Magnetic ResonanceImaging (MRI) scanner. From January1997, he was a Lecturer, andsubsequently Senior Lecturer, inMedical Physics in the Centre forMedical, Health and EnvironmentalPhysics, School of Physical Sciences,Queensland University of Technology(QUT) in Brisbane, Australia.

38

In June 2003 Clive was appointed asSenior Lecturer and Director of theInstitute of Medical Physics within theSchool of Physics at the University ofSydney. The Institute acts as anumbrella organization for medicalphysics activities including research,postgraduate supervision andundergraduate teaching within theSchool of Physics. The Institute hasstrong links to University of SydneyHospitals as well other hospitals inNew South Wales.

Clive’s current research interestsinclude radiotherapy gel dosimetry,radionuclide dosimetry, motioncorrection in medical imaging andradiotherpy, electronic portal imaging,kilovoltage dosimetry Monte Carlocalculation and applications ofSPECT/CT and PET/CT inradiotherapy.

The Treatment of Cancer UsingIonising RadiationDr Clive Baldock

Radiation and what it does to cancerC A N C E R I S C H A R A C T E R I Z E D by a prolific and uninhibited replication ofcells, which can interfere with the function of normal cells and organs,thereby endangering the life of the patient. The aim of radiationtreatment is to deliver a sufficient radiation dose to sterilize cancer cellswhile limiting accidental damage to adjacent healthy tissue.

39

40

WAV

ES O

F TH

E FU

TURE

But what is ‘dose’, and what does it mean tosterilize a cell? Radiation dose is the energyabsorbed by cells from an incident radiation

beam, per unit mass of tissue, and is a measureof molecular damage inflicted on the cells. It ismeasured in joules per kilogram (Jkg-1). Manydifferent types of radiation have been investigatedfor cancer-cell killing efficacy with varying results:these include beams of photons, electrons,neutrons, protons, pions and even heavy atomslike carbon-12. The majority of modern clinicsuse photon and/or electron beams to treatcancer, and these beams disrupt normal cellularfunction principally by breaking chemical bondsthrough ionizing interactions. These ionizationscan corrupt key molecules required for cellularreplication and metabolism. A cell is sterilizedwhen sufficient molecular damage has beeninflicted that the cell can no longer replicate. Notethat while cancer cell-death may be preferable,cell sterilization is often sufficient to preserve thelife of the patient.

Radiation damage to DNA molecules in the cellnucleus is the primary mechanism for cellsterilization. DNA contains coded instructions forthe entire functionality of the cell, and when a cellreplicates the DNA divides and an identical copyis made for the offspring cell. Radiation damageto DNA can prevent cell replication and the limitthe viability of the offspring cell.

Radiation biologists have developed the linear-quadratic (LQ) mathematical model to describeradiation sterilization of cancer cells. The modelassumes that a cell is sterilized when bothstrands of a DNA double helix molecule arebroken; this can occur either by a single particlethat breaks both strands at the same time, or bytwo particles that each break one of the strandswith a short time interval between breakages. Thesignificant difference between this ‘single’ or‘double’ DNA hit is that normal cellular repairmechanisms can repair some or all of the doublehit damage (the first broken strand may berepaired before the second strand is broken) butthese mechanisms are unable to repair thesubstantial damage when both strands arebroken at the same time. According to the LQmodel, the surviving fraction of cells Sfx after adose of radiation D is given by

Sfx = exp(–αD – βD2)

where α is the coefficient of non-repairabledamage and β is the coefficient of repairabledamage. A plot of log Sfx against dose is called acell-survival curve – an example is shown infigure 1.

Empirical tests determine that DNA repairmechanisms of tumour cells are significantly lesseffective than those of normal healthy cells (thatis, tumour cells exhibit low values of β). Thiseffect is exploited by ‘fractionating’ the radiationtreatment into a succession of small doses,typically one dose-fraction a day for six to sevenweeks. After each fraction the normal healthycells are able to repair some of the damagecaused by the radiation, whereas the tumour cellsare unable to repair this damage, leading tocompounded decimation with each new fraction.The fractionation is optimized when maximumcancer cell sterilization is achieved with minimaldamage to normal tissues. Determining theoptimal trade-offs between dose per fraction,time interval between fractions and totaltreatment dose is the subject of a great deal ofcurrent research.

How do we generate radiation in ahospital setting?Photon beams are used for both the localizationand treatment of human tumours. Localization isthe process of defining the physical extent and

Figure 1: Cell survival curve illustrating the percentage ofsurviving cells after doses of radiation.

41

THE

TREA

TMEN

TO

FC

ANC

ER U

SIN

GIO

NIS

ING

RAD

IATI

ON

location of the tumour and is performed using lowenergy ‘diagnostic’ x-rays, typically 0.03 MeVmean energy, which give relatively high contrastbetween soft-tissue and bone. Low energydiagnostic beams are ineffective for therapy ofmost cancers, however, as they have poorpenetration (especially through bone) and asignificant fraction of radiation scatters out of thetreatment region. Radiotherapy therefore useshigh energy photon and electron beams, typicallywith a mean energy of 2 to 10 MeV. The problemof generating radiation therefore falls into twodivisions: the production of low and high energyx-rays. Interestingly, although the two divisionsdiffer dramatically in technical implementation,they both employ the same physical mechanismof radiation generation: the bremsstrahlunginteraction.

Bremsstrahlung, or braking radiation, is thephenomenon whereby a fast moving electron issuddenly decelerated by interaction with a heavy,positively charged target nucleus (a coulombicinteraction). During deceleration the electronradiates away some of its energy in the form of aphoton. In general, faster moving electronsradiate a greater fraction of their energy asbremsstrahlung photons and deposit less heat tothe heavy target atoms. In a low-energy x-raygenerator (Figure 2) electrons emitted from aheated filament via thermionic emission areaccelerated across an evacuated tube to strike atungsten target where the bremsstrahlunginteraction occurs. Notice how the radiationemerges almost at right angles to the direction ofthe incident electrons.

The Linear Accelerator: at the heart ofradiation treatment!The low energy x-ray tube of figure 2 is onlyrequired to accelerate electrons to a few tenths ofan MeV, which can be achieved with this simpledesign. Therapy machines are required toaccelerate electrons up to 25 MeV, and thisrequires very high electromagnetic (EM) fieldsand a substantial number of high technologycomponents.

Figure 3: The linear accelerator. (a) Treatment-room view of a travelling wave accelerator; (b) Schematic diagram of a standingwave accelerator.

Figure 2: Low-energy diagnostic and superficial therapy x-ray tube.

42

WAV

ES O

F TH

E FU

TURE

The Linear Accelerator is a technological wonderat the heart of the radiotherapy department, andgenerates both the high-energy photon andelectron beams used in cancer therapy. It isusually located in a thick concrete undergroundbunker to minimize exposure to hospital staff andthe general public. A treatment-room view and aschematic design are shown in Figure 3. Echoesof the basic x-ray circuit are seen in the electrongun (corresponding to the filament), the mainaccelerator waveguide (where electrons areaccelerated, necessarily longer than in an x-raytube) and the high-Z (high atomic number)tungsten target. Intense EM fields generated by acomplex resonant microwave cavity combinationcalled a klystron accelerate electrons almost tothe speed of light along the mainwaveguide of the accelerator. Avariety of beam modifiers areplaced in the beam to adapt it forclinical use. The most importantmodifiers are the flattening filter,which achieves a flat, uniformbeam, and the multi-leafcollimator, which shapes thecross section of the beam tomatch the projection of thetumour.

The klystronA cross-sectional drawing of an elementary two-cavity klystron is shown in Figure 4. The klystron isable to massively amplify low-power inputmicrowaves using two coupled resonant microwavecavities. Thermionic emission from a heatedcathode introduces electrons into the first cavity,where they are ‘bunched’ together by the low-power input microwaves. The first cavity also exerts‘velocity modulation’ on the electron bunches,

which become progressively more distinct as theytravel along the drift tube. The second resonant‘catcher’ microwave cavity has intense electricfields induced on it by the electron bunches, whichare consequently decelerated. Energy is thustransferred from the kinetic energy of the electronbunches to EM microwave power, which then exitsthe klystron and is transported to the mainaccelerator waveguide.

Treatment with photon beamsPhotons are regarded as indirectly-ionizingradiation – that is to say, photon interactions intissue generate fast moving electrons thatpropagate through tissue, directly damaging cells

via multiple ionizations along theelectron’s track. Theseinteractions are predominantlywith orbital electrons of atomsand molecules in the tissue.Photons have greater penetrabilityinto tissue than electrons (seeFigure 6) and therefore createsignificant electron fluence atdepth, without giving excessivedose to the intervening healthytissue. Although photonspenetrate tissue quite well, theystill deposit a significant dose

superficially (that is, at the surface). To maintain thesurface dose below acceptable levels severalphoton beams are generally used in crossfiretechniques. All the beams in the treatment plancontribute dose to the tumour, at the point ofcrossfire, but surface tissue generally only receivesdose from one or two beams.

Physical interactionsThere are three main physical interactionmechanisms by which a therapeutic energyphoton can interact with human tissue: thephotoelectric effect, Compton scattering and pairproduction. The relative probabilities of eachinteraction are simple functions of the energy ofthe incident photon and the atomic mass numberof the target atom. More complicated relationsconnect interaction probability to parameters likethe scattering angle and energy distributionamongst particles. Over the therapeutic range ofenergies considered here, Compton scattering isFigure 4: The klystron.

The Linear Accelerator isa technological wonder ...and generates both thehigh-energy photon andelectron beams used incancer therapy.

43

THE

TREA

TMEN

TO

FC

ANC

ER U

SIN

GIO

NIS

ING

RAD

IATI

ON

generally the dominant mechanism, althoughphotoelectric absorption becomes important atlower energies (< 100 keV) and pair productionbecomes important at high energies (> 5 MeV).

A schematic diagram of the Compton scatteringinteraction is shown in Figure 5. An incidentphoton scatters off an outer orbital electron withreduced energy; the electron leaves the atomcarrying off the energy difference between theincident and scattered photons. In thephotoelectric effect, the incident photon interactswith an inner bound orbital electron. The incidentphoton is completely absorbed in the interaction(there is no scattered photon) and the electronleaves the atom carrying off the differenceenergy, accounting for the binding energy of theelectron to the atom.

In pair production, the incoming photon is onceagain completely absorbed, this time in aninteraction with the field of the nucleus, resultingin the production of an electron-positron pair. Theincident photon therefore has to have energygreater than the rest masses of the electron andpositron. The positron causes ionization similar toan electron until it is brought to rest, when itannihilates with an electron producing acharacteristic back-to-back 511 keV photon pair.

A typical depth-dose curve for a clinical photonbeam is illustrated in figure 6.

Clinical examples of photon beamtreatmentBreast cancerBreast cancer is the most common malignantdisease of women in the western world – thelifetime probability of a woman developing breastcancer is estimated to be about 11%. Breastcancer is a particularly dangerous diseasebecause of its tendency to break throughbasement membranes of small tissue-ducts,leading to metastatic spread through thelymphatic system. Typical treatment will involvelumpectomy of the gross tumour followed bypost-operative irradiation of the entire breasttissue. The entire breast is treated because of themicroscopic invasive potential of breast cancer tothe surrounding tissue to uniform dose.

A typical photon beam treatment arrangementand dose distribution are illustrated in Figure 7.Two opposed tangential 6 MV radiation beamsdeliver a near-uniform dose to the breast andlumpectomy cavity. Dose uniformity is enhancedby placing metal wedge filters in the beam, whichreduce the intensity of radiation progressively

Figure 7: (a) Transaxial view through a CT scanillustrating a typical radiation treatment forbreast cancer. Two opposed wedged tangential6 MV radiation beams (only one shown) delivera near-uniform dose to the entire breast.Isodoses illustrate dose uniformity across thebreast. (b) The radiation fields are shaped tominimize dose to lung and heart tissue.

Figure 6: Photon depth-dose curves.

Figure 5: Schematic diagram of inelastic Comptonscattering. An outer orbital ‘Compton’ electron is ejectedfrom the atom by a high-energy incident photon, whichloses energy in the interaction and is scattered.

44

WAV

ES O

F TH

E FU

TURE

towards the thick end of the wedge. The wedgecompensates for the ‘missing tissue’ towards theapex of the breast. Although both beams arewedged, only one is shown in Figure 7. A typicalradiation treatment course is a dose of 180 cGyper day, five days a week, for a total dose of 45Gy (the Gray, or Gy, is the SI unit of radiationdose, equal to 1 joule of energy deposited in 1 kgof tissue or other material). A boost dose tomicroscopic tumour residue around the surgicalscar is sometimes given with an electron beam.Recent advances in breast radiotherapy (seeKestin et al 2000) use intensity-modulatedradiation therapy (IMRT), which leads to improveddose homogeneity and better cosmetic outcome.

Prostate cancerProstate cancer is the most common malignantdisease of men in the US and Europe, with alifetime probability of development of about 12%.Radiotherapy has been found to be as effectiveas radical prostatectomy for tumours limited tothe prostate, and has considerably less toxicity.The prostate is located close to the rectum andbladder, and sophisticated treatment planningtechniques are employed to shape the dose

distribution to the shape of the prostate and toavoid excessive damage to these critical organs.

An example of a four-field prostate treatment isshown in Figure 8. The treatment plan consists oftwo pairs of beams: an anterior-posterior pair anda right-left lateral pair. The beams crossfire onthe prostate to give a uniform therapeutic dose.Often higher energy photon beams (> 18 MeV)are used to take advantage of their greaterpenetrability to the prostate, which can lie atdepths of 18 cm or more. In a typical fractionatedcourse of radiotherapy the patient will receive adaily dose of 180 cGy, five days a week, for eightweeks. Areas of current research interest includeincorporating organ motion into treatmentplanning (see Yan and Lockman 2001, Yan et al2000), and online image-guided therapy, whereanatomical information is obtained at the time oftreatment (see Jaffray and Siewerdsen 2000).

Treatment with electron beamsElectron beams are obtained from the linearaccelerator by simply moving the tungsten targetout of the path of the accelerated electrons

Figure 8: Typical four-field radiation treatment for cancer of the prostate. Isodose lines show a uniform dose delivered to thewhole prostate organ. The dose-limiting healthy structures are the rectum, bladder and femoral heads.

45

THE

TREA

TMEN

TO

FC

ANC

ER U

SIN

GIO

NIS

ING

RAD

IATI

ON

travelling along the waveguide of the accelerator.The electrons exit the head of the treatmentmachine and are incident on the patient afterpassing through a scattering foil which widensthe beam and increases uniformity. Electronbeams are useful for treating superficial lesionswithin 6 cm of the skin surface. The depth-dosecurve illustrated in Figure 6 shows that, unlikephotons, the dose decreases rapidly with depthafter a maximum value. Both the depth of themaximum and the gradient of dose fall-off varywith the incident energy of the electrons.

Electrons lose energy in tissue through coulombicinteractions with atomic electrons and atomicnuclei (bremsstrahlung). In low-Z materials (e.g.tissue) energy loss is predominantly by inelasticionizing events with atomic electrons. The meanrate of energy loss of electrons in tissue is about2 MeVcm-1. This means that a 10 MeV beam willhave a maximum range of 5 cm. The dosedistribution from electron beams can be difficultto predict, especially for small fields, due toelectron scattering and build-up effects. Electrons‘backscatter’ at an interface to a high-Z material(e.g. metal dental implant or hip prosthesis),causing a significant increase in dose upstreamfrom the interface. In low-density regions like thelungs, electrons can travel three times furtherthan in normal tissue. In such cases it may benecessary to modify the angle of incidence of thetreatment beam or the treatment dose tominimize the dose to healthy tissue.

A clinical example of the application of anelectron beam would be to boost the dose to thesurgical scar after lumpectomy of the breast. Thesurgical scar may be contaminated with sub-clinical tumour deposits during the surgicalprocedure. Electrons are also often used to boostthe dose to superficial lymph nodes after apredetermined photon dose has been delivered.This ‘mixed-beam’ approach achieves dose atdepth with the photon component, with a boostedsuperficial dose from the electron component.

What’s it like to have radiationtreatment?We have seen that a tremendous amount ofactivity occurs at the atomic and subatomic level

when a patient is irradiated with a photon or anelectron beam. The patient, however, does notfeel anything at all at the time of treatment. Allthe millions of interactions that take place intissue are completely undetectable to the humannervous system at the time of treatment. Patientsonly feel the effects of their radiation treatmentwhen significant numbers of targeted cells havebeen sterilized and removed by the immunesystem. This time interval depends on manyissues, like cell-cycle time and the structure andsensitivity of the irradiated tissue.

Patients undergoing palliative treatment, whenthe aim is to improve quality of life rather than tocure the patient, often receive pain relief within afew days. Negative symptoms from radiationtreatment normally occur after 2-3 weeks if at all.Side effects that may occur are red and tenderpatches of skin, mild nausea and vomiting forabdominal treatments, diarrhoea and rectaldiscomfort for pelvic irradiation.

The medical physicist and the radiation oncologytreatment teamThe patient interacts with a highly trained team ofspecialists who form the radiation oncology team.The primary roles in the team are the physician whodetermines the nature of the treatment, the physicistwho advises on technical aspects, the dosimetristwho formulates the computer treatment plan, thetherapist who delivers the treatment to the patient,and the nurse who is the primary care-giver.

The medical physicist makes critically importantcontributions at many stages of the treatmentprocess. The physicist is responsible for correctfunctioning of all aspects of radiation equipment,and for the purchasing, clinical acceptance andcommissioning of all new equipment. Physicistsare responsible for analysing treatment efficacyand for developing and implementingimprovements, whether from technologicaladvances or new possibilities in treatmenttechnique. An example is gel dosimetry, a newtechnique for obtaining high resolution 3D imagesof complex dose distributions.

Measuring three-dimensional absorbed dose inradiotherapy using gel dosimetryAs long ago as the 1950s, radiation-induced

46

WAV

ES O

F TH

E FU

TURE

colour change in dyes was used to investigatedoses in radiation sensitive gels (Day 1950,Andrews 1957). Subsequently, nuclear magneticresonance (NMR) relaxation properties ofirradiated gels infused with conventional Frickedosimetry solutions were measured (Fricke 1927,Gore 1984). In Fricke gels, Fe2+ ions in ferroussulphate solutions are usually dispersedthroughout a gel matrix. Fe2+ ions are convertedto Fe3+ ions with a corresponding change inparamagnetic properties that may be quantifiedusing NMR relaxation measurements (Gore 1984)or optical measurement techniques (Appleyby 1991).

Due to predominantly diffusion-related limitations(Baldock 2001), alternative polymer geldosimeters were subsequently suggested(Maryanski 1993, 1994). In these polymer gels,now commonly known as BANG-type (Maryanski1994) or PAG-type (Baldock 1998), monomersare usually dispersed in an aqueous gel matrix.The monomers undergo a polymerisation reactionas a function of absorbed dose resulting in a 3Dpolymer gel matrix. The radiation-inducedformation of polymer influences NMR relaxationproperties and results in other physical changesthat may be used to quantify absorbed radiationdose with the potential for true 3D dosimetry (DeDeene 1998, 2000).

As the polymerization is inhibited by oxygen(Maryanski 1994, Baldock 1998) in BANG-type orPAG-type polymer gel dosimeters, all free oxygenhas to be removed from the gels. For many yearsthis was achieved by bubbling nitrogen throughthe gel solutions and by filling the phantoms in aglove box that is perfused with nitrogen. Analternative polymer gel dosimeter formulation,known as MAGIC gel (Fong 2001) was proposedin which oxygen is bound in a metallo-organiccomplex thus removing the problem of oxygeninhibition and enabling polymer gels to bemanufactured on the bench-top of the laboratorywith the potential for true 3D dosimetry(Gustavsson 2003).

As well as MRI (Vergote 2004), other quantitativetechniques for measuring dose distributionsinclude optical (Oldham 2003) and x-raycomputer tomography (Audet 2002), vibrational

spectroscopy (Rintoul 2003) and ultrasound(Mather 2003). Numerous clinical applications ofthese radiologically tissue equivalent (Keall 1999)gel dosimeters have been reported in thescientific literature and the international DOSGELconference series on radiotherapy gel dosimetry(DOSGEL 1999, 2001, 2004).

In conclusionThe role of the medical physicists is often a veryinteresting and challenging one, given the rapiddevelopment of computing and technicalhardware and software. The physicist also plays acritical role as the department trouble-shooter,available at immediate notice to solve problemsas they arise at treatment time and during thetreatment planning process. Such a role oftendemands quick thinking in a stressful situation.The role of the medical physicist demands highprofessional standards; at the same time it canbe an exciting and rewarding profession, whichmakes a real difference to the quality of life ofpatients treated in the Radiotherapy department.

References

Andrews HL, Murphy RE and LeBrun EJ 1957. Gel dosimeter

for depth dose measurements. Review of Scientific

Instruments 28 329-332.

Appleby A and Leghrouz A 1991. Imaging of radiation dose by

visible color development in ferrous-agarose-xylenol orange

gels. Med. Phys. 18 309-312.

Audet C, Hilts M, Jirasek A and Duzenli C 2002. CT gel

dosimetry technique: Comparison of a planned and

measured 3D stereotactic dose volume. J. Appl. Clin. Med.

Phys. 3 110-118.

Baldock C, Burford RP, Billingham NC, Wagner GS, Patval S,

Badawi RD and Keevil SF 1998. Experimental procedure

for the manufacture of polyacrylamide gel (PAG) for

magnetic resonance imaging (MRI) radiation dosimetry.

Phys. Med. Biol. 43 695-702.

Baldock C, Harris PJ, Piercy AR and Healy B 2001.

Experimental determination of the diffusion coefficient in

two-dimensions in ferrous sulphate gels using the finite

element method. Australasian Physical & Engineering

Sciences in Medicine 24 19-30.

47

THE

TREA

TMEN

TO

FC

ANC

ER U

SIN

GIO

NIS

ING

RAD

IATI

ON

Day MJ and Stein G 1950. Chemical effects of ionizing

radiation in some gels. Nature 166 146-147

De Deene Y, De Wagter C, Van Duyse B, Derycke S, De Neve

W and Achten E 1998. Three-dimensional dosimetry using

polymer gel and magnetic resonance imaging applied to

the verification of conformal radiation therapy in head-and-

neck cancer. Radiother. Oncol. 48 283-291.

De Deene Y, De Wagter C, Van Duyse B, Derycke S,

Mersseman B, De Gersem W, Voet T, Achten E and De

Neve W 2000. Validation of MR-based polymer gel

dosimetry as a preclinical three-dimensional verification

tool in conformal radiotherapy. Magnetic Resonance in

Medicine 43 116-125.

DOSGEL 1999. Proceedings of the 1st International Workshop

on Radiation Therapy Gel Dosimetry. (Canadian

Organization of Medical Physicists, Edmonton)

Eds. LJ Schreiner and C Audet.

DOSGEL 2001. Proceedings of the 2nd International

Conference on Radiotherapy Gel Dosimetry. (Queensland

University of Technology, Brisbane, Australia) Eds. C

Baldock and Y De Deene.

DOSGEL 2004 Proceedings of the 3rd International

Conference on Radiation Therapy Gel Dosimetry eds De

Deene, Y and Baldock, C (Ghent University, Ghent, Belgium)

Fong P M, Keil D C, Does M D and Gore J C 2001. Polymer

gels for magnetic resonance imaging of radiation dose

distributions at normal room atmosphere. Phys. Med. Biol.

46 3105-3113.

Fricke H and Morse S 1927. The chemical action of roentgen

rays on dilute ferrous sulfate solutions as a measure of

radiation dose. Am. J. Roentgenol. Radium Therapy Nucl.

Med. 18 430-432.

Gore JC, Kang YS and Schulz RJ 1984. Measurement of

radiation dose distributions by nuclear magnetic resonance

(NMR) imaging. Phys. Med. Biol. 29 1189-1197.

Gustavsson H, Karlsson A, Back SAJ, Olsson LE, Haraldsson P,

Engstrom P and Nystrom H, 2003. MAGIC-type polymer gel

for three-dimensional dosimetry: Intensity-modulated

radiation therapy verification. Med. Phys. 30 1264-1271.

Jaffray D A and Siewerdsen J H 2000. Cone-beam computed

tomography with a flat-panel imager: initial performance

characterization Med. Phys. 27 1311-23

Keall PJ and Baldock C 1999. A theoretical study of the

radiological properties and water equivalence of Fricke and

polymer gels used for radiation dosimetry. Australasian

Physical & Engineering Sciences in Medicine 22 85-91.

Kestin L L, Sharpe M B, Frazier R C, Vicini F A, Yan D, Matter

R C, Martinez A A and Wong J W 2000. Intensity

modulation to improve dose uniformity with tangential

breast radiotherapy: initial clinical experience Int. J. Radiat.

Oncol.Biol. Phys. 48 1559-68

Maryanski MJ, Gore JC, Kennan RP and Schulz RJ 1993.

NMR relaxation enhancement in gels polymerized and

cross-linked by ionizing radiation: a new approach to 3D

dosimetry by MRI. Magn. Reson. Imaging 11 253-258.

Maryanski MJ, Schulz RJ, Ibbott GS, Gatenby JC, Xie J, Horton

D and Gore JC 1994. Magnetic resonance imaging of

radiation dose distributions using a polymer-gel dosimeter.

Phys. Med. Biol. 39 1437-1455.

Mather ML and Baldock C 2003. Ultrasound tomography

imaging of radiation dose distributions in polymer gel

dosimeters. Med. Phys. 30 2140-2148.

Oldham M, Siewerdsen JH, Kumar S, Wong J and Jaffray DA

2003. Optical-CT gel-dosimetry I: Basic investigations.

Med. Phys. 30 623-634.

Oldham M, Siewerdsen J H, Shetty A and Jaffray D A 2001.

High resolution gel-dosimetry by optical-CT or MR

scanning Med. Phys. 28 1436-45

Rintoul L, Lepage M and Baldock C 2003. Radiation Dose

Distribution in Polymer Gels by Raman Spectroscopy. Appl.

Spectrosc. 57 51-57.

Vergote K, De DeeneY, Duthoy W, De Gersem W, De Neve W,

Achten E and De Wagter C 2004. Validation and application

of polymer gel dosimetry for the dose verification of an

intensity-modulated arc therapy (IMAT) treatment. Phys.

Med. Biol. 49 287-305.

Yan D and Lockman D 2001. Organ/patient geometric

variation in external beam radiotherapy and its effects Med.

Phys. 28 593-602

Yan D, Lockman D, Brabbins D, Tyburski L and Martinez A

2000. An off-line strategy for constructing a patient-specific

planning target volume in adaptive treatment process for

prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 48 289-302

For example, 2004 was theInternational Year of Rice – andgiven the number of people whodepend on that grain for their dailysustenance and survival, it’s a worthyrecipient of its own year. 2003 wasthe Year of Fresh Water, for similarreasons. Ecotourism got its year in2002; that was also theInternational Year of Mountains, andthe International Year of CulturalHeritage – 2002 was a big yearindeed. And levering the symbolismof a new millennium, the UnitedNations declared 2001 to be theInternational Year of Mobilizationagainst Racism, RacialDiscrimination, Xenophobia andRelated Intolerance (or IYoMRRDXRIfor short).

So what has the UN decided for 2005?This is the year of Microcredit(www.yearofmicrocredit.org)focussing on the practice ofdelivering small loans to needyindividuals or groups.

It’s also the International Year ofSport and Physical Education(www.un.org/sport2005), aiming topromote the value of sport to healthand culture worldwide.

But most importantly (for this book,at any rate) 2005 is theInternational Year of Physics!

Why physics? What makes thisyear more physics-y than anyother? This year we’re celebratingone hundred years since a youngSwiss patent clerk wrote a series ofresearch papers that changed theway we view the world around us.His insights into the machinery ofthe physical universe changed howwe understand light, and matter,and energy, and even space andtime – all in this one incredible yearof 1905.

A century later scientists aroundthe globe got together andconvinced the UN to declare 2005the International Year of Physics(IYOP) in honour of that great man– Albert Einstein. All over the world,groups of scientists, educational

2005: The International Year of ...E A C H Y E A R T H E United Nations declares that this isofficially the International Year of Something Important. Withthe UN’s backing, an announcement like that draws theworld’s attention to a cause, helps raise awareness andgalvanises support and action.

institutions, museums and otherpublic organizations are runningevents to showcase the excitementand achievements of the physicalsciences, to get the collective mindworking and to honour the legacyof Einstein. To see what events arehappening near you, visitwww.wyp2005.org.

Throughout this book, tuckedbetween the chapters, you will findshort articles about Einstein’sdiscoveries of a hundred years ago.We hope you enjoy them, and thatyou make the most of this, theInternational Year of Physics.

48

WAV

ES O

F TH

E FU

TURE

2005

:TH

EIN

TERN

ATIO

NAL

YEAR

OF

...

Official United Nations International Years

2006: International Year of Deserts and Desertification2005: International Year of Physics

International Year for Sport and Physical EducationInternational Year of Microcredit

2004: International Year to Commemorate the Struggle against Slavery and its AbolitionInternational Year of Rice

2003: International Year of Freshwater2002: International Year of Ecotourism

International Year of MountainsUnited Nations Year for Cultural Heritage

2001: International Year of Mobilization against Racism, Racial Discrimination, Xenophobia and Related Intolerance

United Nations Year of Dialogue among CivilizationsInternational Year of Volunteers

2000: International Year of ThanksgivingInternational Year for the Culture of Peace

1999: International Year of Older Persons1998: International Year of the Ocean1996: International Year for the Eradication of Poverty1995: World Year of Peoples’ Commemoration of Victims of the Second World War

United Nations Year for Tolerance1994: International Year of Sport and the Olympic Ideal

International Year of the Family1993: International Year of the World’s Indigenous People1992: International Space Year1990: International Literacy Year1987: International Year of Shelter for the Homeless1986: International Year of Peace1985: Year of the United Nations1985: International Youth Year1983: World Communications Year1982: International Year of Mobilization for Sanctions against South Africa1981: International Year for Disabled Persons1979: International Year of the Child1978/1979: International Anti-Apartheid Year1975: International Women’s Year1974: World Population Year1971: International Year for Action to Combat Racism and Racial Discrimination1970: International Education Year1968: International Year for Human Rights1967: International Tourist Year1965: International Cooperation Year1961: International Health and Medical Research Year1959/1960: World Refugee Year 49

SIMON CARLILE did hisundergraduate andgraduate training at TheUniversity of Sydney. HisPhD thesis workexamined thebioacoustic andphysiological basis ofthe representation ofauditory space in themammalian auditorynervous system. Hespent five years inOxford as a postdoctoral

fellow and then a Beit MemorialFellow and Junior Research Fellow ofGreen College Oxford. During thistime he worked in themultidisciplinary sensoryneuroscience group led by ColinBlakemore. In 1993 he moved backto The University of Sydney as aLecturer in Neuroscience in theDepartment of Physiology and hasestablished the AuditoryNeuroscience Laboratory. TheLaboratory has a broad focus withcurrent work ranging from thebioacoustics of outer ear, thepsychophysics of real and virtualauditory space as well as theneurophysiological mechanisms thatresult in neural representations ofauditory space. He also has interests

in the applications of informationtechnology to medical

education and has lecturedand tutored in History

and Philosophy ofScience.

The psychophysics of real andvirtual auditory spacesSimon Carlile

IntroductionI N T H I S C H A P T E R we will be considering how we perceive the spacearound us using our sense of hearing. The environments in which wenormally live are full of many different sounds, often occurring at thesame time. Some of these sounds are of interest to us, but many arenot. Take for example a noisy club or party where you are concentratingon what your friend is saying (the foreground sound of interest) andtrying to ignore the concurrent background music and conversations(the background noise). This issue has been called the ‘cocktail partyproblem’ by E. C. Cherry in the 1950s (Cherry 1953, see Figure 1) – hisideas about how the nervous systems solves this problem have beenvery influential in guiding our development of communications devicessuch as telephones as well as more advanced virtual reality displays.The way in which our auditory system sorts out the different informationfrom each of these concurrent sources using only two ears is amagnificent feat of information processing. As we shall see, our abilityto determine where the different sources are located in space plays animportant role in this process.

51

52

WAV

ES O

F TH

E FU

TURE

Before proceeding, it is probably worthspending a little time exploring the meaningof the title of this chapter, as the terms are

not ones used in everyday conversation. Firstly,‘psychophysics’ is the study of human perceptionthat is concerned with establishing quantitativerelations between physicalstimulation and perceptualevents – in other words,attaching sounds we hear toevents we see or feel.Psychophysics has a quite longhistory as a science and arose inthe early part of the twentiethcentury. In the context of thischapter, we will be examiningthe performance of humansubjects undergoing differenttests of hearing. The results ofsuch experiments allowinferences about what therelevant physical stimuli are and how they areprocessed by the auditory system.

Secondly, the term ‘auditory space’ refers to theacoustic environment that surrounds the listenerat any particular time. This will not only include

multiple different sound sources but also thepassive acoustics of the environment such asreverberation. Our perception of auditory spacecan be described along a number of differentdimensions. Three of the principal dimensions aredirection, distance and spaciousness. While wecan objectively characterise direction anddistance using the standard three dimensions(also known as a Euclidian description), this doesnot necessarily mean that our perception ofauditory space maps onto this coordinate system.

For instance, spatial coordinates could bedescribed using spherical coordinates thatindicate the direction to a sound source and itsdistance from the listener. However, in ourqualitative description of a sound’s location ineveryday life we generally talk about thehorizontal direction and distance, and the heightabove or below the audio-visual horizon. Thisdescription implies that we use a more cylindricalview of space when talking about sound. Inaddition to the perception of location (directionand distance), the extent or ‘spaciousness’ of thespace inhabited by the listener is also a powerfulelement of perception.

Thirdly, we need also to consider what is meantby ‘real’ and ‘virtual’ spaces. A real space is thesort of space where you find yourself now. The

sounds are coming fromindividual sources that occupydifferent locations in the worldaround you. They may be movingor static, and the passiveacoustic environment (theobjects and spaces surroundingyou) may also be shaping andmodifying the sounds that reachyour ears. By contrast, in avirtual auditory space the soundsusually come from a limitedrange of sources – such as apair of headphones. However,the sounds are processed in a

way that leads you to perceive different sourcesat different locations in particular acousticenvironments. This is very different to the normalexperience of listening to sounds overheadphones.

Figure 1: The cocktail party problem – how does thisperson manage to hear the conversation next to himthrough all the background noise?

“In general, properlymatching the display and the sensory system requires anunderstanding of theperceptual limits of theauditory system.”

53

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

For instance, when you play music or speechover headphones, the sound image is heardinside or very close to the head, rather than froma location away from the head, as is the casewhen you are listening to the same or speechover stereo speakers in your lounge room (seeFigure 2). We will see later just how we canprocess the sound to generate the illusion ofsounds outside the head in ‘virtual auditoryspace’. In this context the perception ofspaciousness plays an important role in thegeneration of the sense of ‘presence’ enjoyed bythe listener (see Durlach 1991) – that is, thefeeling of actually being in a virtual environmentcan be generated using auditory and visualdisplays.

One of the interests of some of the researcherswhose work I will describe in this chapter is howto implement virtual auditory displays that matchthe capacity of the human listener. From atheoretical point of view, the fidelity of anyreproduction is a key issue in the generation ofthe representation of an object. In a practicalsense, the fidelity of reproduction is oftendetermined by the use to which the reproductionis to be put.

For instance, where the objective is to produce arepresentation that is as close as possible to aperfect copy of the original object, then, from aperceptual point of view, the limiting factor is thefidelity of the sensory system encoding therepresentation. On the one hand, when the fidelityof reproduction is low the quality of the resulting

perception can be degraded – in the case ofgenerating virtual auditory space, listeners wouldnot get a compelling or accurate illusion of anexternal world when listening over headphones.On the other hand, where fidelity of therepresentation is higher than that of the sensorysystem, then the effort of producing such a highfidelity reproduction is wasted. In the context ofauditory virtual displays, significant computingpower is needed to generate these displays so‘over-engineering’ the fidelity places a seriousoverhead on performance.

In general, properly matching the display and thesensory system requires an understanding of theperceptual limits of the auditory system. One wayin which this can be determined is by making

Figure 2: (a) Sound from stereo speakers seems to comefrom somewhere outside your head, typically between thespeakers; (b) in contrast, sound from a pair of headphonesseems to originate from somewhere inside or very close toyour head.

(a)

(b)

54

WAV

ES O

F TH

E FU

TURE

making objective measures of the user’s psycho-physical performance on tasks in real spacecompared with virtual space – an approach thatwe will examine in more detail below.

In the next sections we will consider the physicalacoustic properties of a sound that the auditorysystem can use as cues to the location andnature of the sound. We will also consider howwell we can locate sounds in space and the rolethat the differences in the locations of the soundsources plays in sorting out and understandingspeech in acoustically complex or ‘cluttered’environments. We will then look at how thesounds can be processed before they arepresented over headphones so that they give riseto the illusion of virtual auditory space. In the finalsection we will look at a few of the moreinteresting applications of virtual auditory space.

Physical cues to the perception ofauditory spaceThe acoustic cues used by the auditory systemin generating our perceptions of space are basedon the interactions of the sound with the twoears, the head and torso as well as with thereflecting surfaces in the immediateenvironment. A very powerful set of cues to a

sound’s location is called the binaural cues.These arise as a consequence of the ears beingseparated by the acoustically dense head andthe fact that each ear is located at slightlydifferent locations in space.

What this means is that each ear simultaneouslysamples the sound field under slightly differentconditions (Figure 3). For a sound located off themidline, there is a difference in the length of thepath from the sound source to each ear, whichresults in a difference in the arrival time of thesound at each ear. This difference is called theinteraural time difference (or ITD) and isdependent on the horizontal location of thesource of the sound with respect to the head. Fora sound located directly ahead there is nodifference in the path lengths from the source toeach ear. By contrast, when the sound source islocated opposite one ear then the interauraldifference is at a maximum. As sound travels ataround 330 m/s, this difference is generally verysmall – the maximum difference is less than 1 ms.

However the auditory system is able to extractthis information from the sound encoded at eachear and compute the horizontal position of thesound. In addition, as the head is relatively largewith respect to the wavelengths of the sound,

Figure 3: The interaction of the head and ears with a soundfield. For a sound located away from the mid-line plane(dashed line) there is a path length difference between thesound source and each ear which gives rise to a differencein the time of arrival of the sound at each ear. Likewise, thehead is an effective obstacle for the sound field andshadows the ear farthest from the source producing adifference in the level at each ear.

55

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

lower frequency range where the wavelengths arelong compared to the dimensions of the outer ear.

A critical feature of these spectral cues is thatwhen considered across the whole range offrequency sensitivity of the listener, the filtering isunique for each possible location in space (Figure5). Of course, a limitation of this cue is that for itto be unambiguous there needs to be a widerange of frequencies in the source sound andthat the original spectrum of the source is alsoknown by the auditory system.

Fortunately, this is the case for sharp transientsounds, which contain a very wide range offrequencies and are generally spectrally flat –that is, there is a reasonable, even distribution ofenergy across the wide frequency range of thesound. Given that there is probably significantevolutionary pressure to locate the transientsounds of approach of a predator (say, the snapof a breaking twig) then encoding and processingthis type of cue is likely to have played animportant role in the survival of the species. Onthe other hand, sounds that have a more limitedspectral range are not going to provide much of amonaural or spectral cue and indeed will not thenhelp in resolving the ambiguities in the binauralcues. As such it is no surprise that spectrallylimited sounds are very hard to localise and, infact, many animals produce warning calls with avery narrow spectrum to sound an alert whileavoiding detection!

A range of physiological studies havedemonstrated that neural representations ofauditory space in the mammalian midbrain aredependent on the integration of these binauraland monaural cues (King and Carlile 1994). Thisconvergence and integration also appears to leadto a form of spatial channel processing that mayunderlie the perception of auditory space (Carlile,Hyams et al. 2001) although a more completediscussion of this very interesting processing bythe nervous system is beyond the scope of this chapter.

Cues for DirectionCoding of the direction of the source of a soundarises principally as a result of the interaction of

Figure 4: The outer earhas a very convolutedshape that helps us tolocate sounds. (Thanksto Natalie for posing forthe photograph.)

the ear furthest from the source will be acoustically shadowed giving rise to an interaurallevel difference in the level of the sound in each ear.

While these binaural cues to sound location arepowerful, it is also well known that these cuesare ambiguous. That is, because of the symmetryof the placement of the ears on the side of thehead, any particular interaural interval will specifythe surface of a cone centred on the interauralaxis – the so called ‘cones of confusion’. Forinstance, a sound in front of the listener willgenerate zero differences in the binaural cues butthen so will a source located directly above ordirectly behind the listener. Likewise, a sound at45° to the left of the midline will generate thesame interaural intervals as the same sourcelocated behind the listener at 135° to the left ofthe midline. As we don’t usually confuse thelocation of a sound in front with one behind, theauditory system must have some other way ofresolving the ambiguities in the binaural cues.

This third set of cues, in addition to the twobinaural cues, is produced by the asymmetricaland highly convoluted shape of the outer ear(Figure 4). This complex physical structure givesrise to a location-dependent filtering of the soundsand produces the so called spectral or monauralcues to location. Reflections from the shoulder andtorso may also contribute to the filtering of the

56

WAV

ES O

F TH

E FU

TURE

the sound with the auditory periphery. Bycontrast, the coding of the distance of the soundsource is dependent on the detection of theinteractions of the sound with objects in thelistening environment. Four acoustic cues todistance have been identified for stationarysources (Mershon and King 1975). Under idealconditions the intensity of a sound decreases withdistance according to the inverse square – thisleads to a 6 dB loss with a doubling of distance.However, in practice this relationship isdependent on the reflective characteristics of theenvironment.

A second and similar cue for distance resultsfrom the transmission characteristics of the air –that is, high frequencies (> 4 kHz) are absorbed,leading to a reduction of around 1.6 dB perdoubling of distance (Zahorik 2002). Notably for

both of these cues, there is a confounding ofsource characteristics (intensity and spectrum)with distance. That is, the overall level of thesound at the ears or the level of the highfrequency content will only provide a reliable cueif the level at the source is known. Consequently,these cues may only be reliable for particularclasses of familiar sounds. Interestingly, we aregenerally very good a judging the distance ofsomeone speaking – probably because if wesubstantially decrease (whisper) or increase(scream) the level of the voice the spectrum ofthe sound also changes.

A third cue to distance is the ratio of the direct toreverberant energy – that is, the proportion of theenergy reaching the listener directly decreaseswith the distance of the source from the listener(i.e. is subject to the inverse square law of

Figure 5: The variation in filter functions of the outer ear are shown as a function of the location of the source on theaudiovisual horizon where 0° is directly ahead and 90° is on the interaural axis opposite the ear. The gain of the filter functionsis indicated by the contour colour. The filter function for each location is unique. Note in particular that the filter functions forthe frontal portion of space (0 - 90°) are very different for the back portion of space.

57

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

distance) while the level of reverberation in aroom is determined principally by thecharacteristics of the room. This is a particularlypowerful cue for distance but obviously isdependent on the reverberant characteristics ofthe listening environment.

For sound sources close to the head (say within a1 metre radius), recent work has confirmed theobservations in the 1920s (Hartley and Fry 1921)that substantial variation in the interaural leveldifferences can occur with variation in thedistance of the source from the head (see forinstance Shinn-Cunningham 2000). As the sourcegets closer to the head, the wave front becomesmore spherical and will bend around the headand interact at the far ear – this is very differentfrom the case when the wave front is effectivelyparallel, as occurs for a source some distancefrom the head. In contrast to the effects on level,the distance effect on interaural time differenceappears to be much less salient (Brungart andRabinowitz 1999). This is what would bepredicted acoustically, given that the path lengthsfrom the source to each ear would not varysubstantially as a function of the distance fromthe centre of the head.

Cues for SpaciousnessThe final cues considered here are those that resultin our perception of spaciousness (see for exampleBlauert and Lindermann 1986, Potter, Raatgever etal. 1995). This has been characterized by (i)‘apparent source width’, which is related to earlylateral reflections and the level of low frequencysound, and (ii) ‘listener envelopment’, which isrelated to the reverberant sound field, particularlyfor sounds arriving more than 80 ms after thedirect sound (see Okano, Beranek et al. 1998).Interestingly, it is these sorts of reflection patternsthat architects have sought in designing the bestopera and concert halls in the world, as thequalities of apparent width and envelopmentcontribute significantly to the appreciation ofmusical performance in these spaces.

In addition to the range of purely acoustic cuesdiscussed above it is also important to note that arange of other factors can play a role in theperception of space. Such so-called top-down or

cognitive factors can depend on expectations thatthe listener may have about the nature of thesounds or the listening environment. Indeed,recent listening experience in a particular spacecan also subsequently affect the perception ofsound in that space. A most straightforwardexample is the well-known ventriloquist effectwhere the source location of a talker is capturedby the image of the talker (as in say the cinemaor TV).

The fidelity of the perception ofauditory spaceIn this section we will consider how well we areable to determine where a sound is located inspace. There are a large number ofpsychophysical studies of the resolution andaccuracy of the auditory system relating toauditory space (for reviews see Middlebrooks andGreen 1991; Carlile 1996). The resolution of theauditory system has been examined bymeasuring how well the auditory system canresolve differences in the locations of asequentially presented sound source (Mills 1958).

This is referred to as a minimum audible angle(MAA) detection task and provides informationabout the just-noticeable differences in the cuesto a sound’s location in space. MAA studies havedemonstrated that the resolution of the auditorysystem is highly dependent on the range offrequency in the stimulus and the absolute spatiallocation about which the change in location isbeing determined (see Grantham 1995). Subjectsdemonstrate the smallest MAA for sounds with abroad range of frequencies (1-2° for soundlocations directly in front of the listener) withsignificant increases in the MAA for narrow bandstimuli and for locations well off to the side.

More recent work has also examined the ability ofsubjects to discriminate concurrent sounds asoriginating from different locations (Best, Schaiket al. 2004). This study demonstrates that theability to separate out two concurrent broadbandstimuli is dependent on the magnitude of thedifferences in the interaural cues generated bythe two stimuli rather than differences in thespectral cues generated by the filtering of theouter ear.

58

WAV

ES O

F TH

E FU

TURE

The accuracy in determining the absolute locationof a sound source has been assessed by allowingsubjects to indicate by pointing or calling outcoordinates of the perceived location of a soundsource (e.g. Makous and Middlebrooks 1990;Carlile, Leong et al. 1997). Absolute localisationaccuracy has generally been determined underanechoic conditions – that is, in special environ-ments where there are little to no echoes orreverberation that are known to degrade this ability.

In general, results from such studies indicate thatthere are two broad classes of localisation errors:(i) large so called ‘front-back’ confusions or ‘coneof confusion’ errors where the perceived locationis in quadrant different from the actual soundsource, and perceived to come from a source atan angle reflected about the interaural aural axis;and (ii) local errors where the sound location isperceived to be located in the vicinity of the actualtarget. The average localisation errors for 19subjects are illustrated on spherical plots in Figure6 (Carlile, Leong et al. 1997). These plots illustrateboth the systematic errors in the localisation(indicated by the difference between the actuallocation and the average of the localisationestimates) and a dispersion of the responsesabout the average perceived location (indicatedby the ellipses surrounding each average).

In general, localisation accuracy decreases forlocations from the anterior to posterior midlineand as the elevation of the target deviates fromthe audio-visual horizon (see also Makous andMiddlebrooks 1990). When stimuli contain abroad range of frequencies only about 3% to 6%

of the trials result in ‘front-back’ errors. Of note,localisation performance is strongly related to thecharacteristics of the stimulus; in particular, theperformance levels noted above are for shortduration (150 ms) broadband stimuli presentedunder anechoic conditions. Where sounds are ofa longer duration then performance can improveas the auditory system has the opportunity totake multiple samples of the sound source andhence the cues to the location are more reliable.However, with narrowband stimuli or lesscontrolled listening conditions localisationperformance can degrade substantially.

The perception of distance has been studiedmuch less exhaustively than the perception of therelative direction of a sound source. In general,listeners tend to underestimate the sourcedistance for far distances and overestimate fornear distances, with a cross over point around1.5 metres. This is called the specific distancetendency (Mershon and King 1975). Recent workhas indicated that the distance cues thatdominate in a particular distance judgment taskvary according to the reliability of the set ofdistance cues. Depending on the spectral contentof the sound source and the nature of theacoustic environment, different cues will be moreor less reliable indicators of distance (Zahorik2002). It appears that the auditory systemdynamically weights its reliance on different cuesdepending on how well the different cuescontribute to the overall assessment of thedistance of the sound source.

In addition to the static cues to a sound’s location,there are also a range of dynamic cues to bothstatic sound sources and moving sound sources.Dynamic cues to the location of static sources areproduced by moving the head to allow moresamples of the sound field. This provides binauralinformation that can be used to resolve theambiguities in the binaural cues and can result insignificant reduction in the cone-of-confusion orfront-back error (Wallach 1940; Lambert 1974).

However, some other findings indicate that headmovement (self-induced or otherwise) only leadsto a reduction in local errors when the sound isrelatively narrow band or the spectral cues tolocation are degraded in some way (see Pollak

Figure 6: Pooled localisation responses from 19 subjectsshown for front and back hemispheres of space. The actualtarget locations are shown by the small ‘*’ and the mean ofthe pooled localisation responses for each location isshown by the small filled circle. The ellipse surroundingeach mean response indicates the standard deviation.

59

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

and Rose 1967, Fisher and Freedman 1968, andsee also Wightman and Kistler 1999). Theavailability of such cues of course requires thatthe sound of interest is of sufficient duration toallow multiple sampling.

The second class of dynamic cues are thoseproduced by moving stimuli. The minimumaudible movement angle (MAMA) has beendefined as the smallest angle a sound must travelbefore its direction is correctly discriminated. Theability to detect the movement of a sound sourceis dependent on the same sorts of cues used inlocating a stationary stimulus, with the advantagethat multiple sampling is possible with a relativelylong-duration stimuli. In addition there may alsobe a Doppler shift of the frequencies of the sounddependent on the movement trajectory of thesource with respect to the listener.

It has been consistently found that MAMAs arelarger than the static MAA (generally 2 to 4degrees of arc) using both noise bursts and tonesemitted from a moving speaker (Harris andSergeant 1971). It has also been reported thatincreased source velocity results in a larger MAMA(Perrott and Musicant 1977; Grantham 1986).This apparent loss in spatial acuity with rapidlymoving sources might indicate a minimumintegration time required to perform these tasks.However, there is much debate about the durationof that integration time. Recent work exploitingvirtual auditory space stimulation techniques (seebelow) has also indicated that human listeners arerelatively insensitive to changes in the velocity ofmoving sound sources and that absolute sensitivityis velocity dependent (Carlile and Best 2002).

The generation of virtual auditoryspaceAs discussed above, an understanding of thefidelity of the perceptual encoding of auditoryspace can be used to guide the development ofauditory displays that exploit the spatial nature ofthe auditory perception. Quite simply, to createthe illusion of external auditory space usingheadphones, the principal processing goal is toreproduce at the ear drums the pattern of soundwaves that would have occurred had the soundsources actually been in the free field. All thingsbeing equal, the auditory system should then beable to extract the necessary acoustic cues toidentify the sources and their relative locationsalong with the characteristics of the auditoryenvironment.

The top panel in Figure 7 illustrates the situationwhen normally listening to, for example, a clicksound over headphones. However, imagine thatwe have also inserted small microphones close tothe ear drums of the listener so we can recordthe sound waves at the ear drums – the outputsof these microphones are shown on the right andleft of each panel. By varying the binaural cues ofinteraural time difference or the interaural leveldifference (that is, the timing or amplitude of theclicks presented to each headphone) the soundimage heard by the listener is perceived to movecloser to the ear where the click arrives firstand/or is loudest.

Despite the fact that this manipulation is varyingthe cues to sound location, the listener still hearsthe sound inside the head. The middle panelillustrates the situation for a sounds presentedfrom a single loudspeaker located away from thelistener. Again we can see the differences in theinteraural level and time cues, but in this case theclick sound is filtered by the outer ears and thesound source is heard outside the head – so ifwe simply take the recordings of sounds at theear made in the middle panel, and play themback over headphones, then, following anycorrection for distortions by the headphones, thesound waves at the eardrums should be thesame as in the middle panel. Doing this, we findthat the listener also hears the sound outside thehead and at the correct location in space. This isthe basis of Virtual Auditory Space (VAS).

Figure 7: The differences between free field and headphonelistening are illustrated in the top two panels. When thefiltering functions of the outer ear are properly accountedfor with headphone listening, the sound image is heardoutside the head in virtual auditory space (bottom panel).

60

WAV

ES O

F TH

E FU

TURE

Of course, if a virtual environment is to be usefulwe want to be able to present any sound at anylocation around the listener and not simply playback what has been previously recorded as in ourdescription above. Lets begin by describing forthe simplest scenario: a single source placed at aparticular location with respect to the head of thelistener. As we have seen above, the sound froma particular source will be transformed or filteredby the outer ears by the time it reaches theeardrum. Acoustically, this is be characterised asthe head related transfer function (HRTF). Asmentioned above, there is a unique HRTF foreach ear and for every direction in spacesurrounding the listener (Carlile and Pralong 1994).

The HRTFs are measured in the laboratory byplacing small recording microphones in anearplug secured flush with the outer end of theear canal for each ear (Møller, Sorensen et al.1995, see also Pralong and Carlile 1994). Themeasurements are performed inside an anechoicchamber with the subject placed at the centre –

see Figure 8. A speaker, mounted on the robotarm, delivers the measurement stimuliautomatically from up to 400 locations evenlydistributed on an imaginary sphere surroundingthe subject. The resulting HRTFs are storeddigitally and can then be used to filter any soundstimuli before presentation to the left and rightears using high quality headphones. In manycases the transfer function of the headphones tothe microphones in the ears are also measuredso that any frequency distortion due to theheadphones can also be accounted for.

The fidelity of such an auditory display isdependent on how well the HRTFs used togenerate the VAS match the actual HRTFs of thelistener. However, the HRTFs are highlyindividualised because the filtering is dependenton the precise shape of the outer ear, andeverybody’s ears are slightly different (Jin, Carlileet al. 2000). Therefore, if we want to generatehigh fidelity VAS, the specific HRTFs of thelistener need to be accounted for.

Figure 8: The anechoic chamber at the University of Sydney’s Auditory Neuroscience Laboratory. The walls are speciallydesigned to reflect as little sound as possible. The chamber is equipped with a robot arm carrying a small speaker that can be placed at almost any location on the surface of an imaginary sphere surrounding a test subject located in the middle of the chamber.

61

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

One way of doing this is to record the HRTFs foreach user of an auditory display and use these togenerate the listener’s own personalised VAS. Ofcourse this is not practical for more general usesof this technology because the facilities andexpertise required to record the HRTFs are notwidespread. On the other hand, methods arebeing developed for predicting the HRTFs ofindividual listeners by making physicalmeasurements of the shapes of the outer earsand their placement on the head (e.g. Jin, Carlileet al. 2000).

Alternatively, a listener could select from arange of HRTF libraries and select those thatbest match their own by using some form ofperformance task – indeed, this is the approachthat is often taken in assessing the fidelity of theVAS rendered by a particular display. In theseassessments, the ability of a subject todetermine exactly where a sound is coming fromis compared for sounds presented in real spacewith sounds presented in virtual space. In moststudies where the recordings have beencarefully controlled, performance is nearlyidentical under both conditions (Wightman andKistler 1989; Bronkhorst 1995; Carlile 1996;Carlile, Leong et al. 1996; Langendijk andBronkhorst 2000).

Examples of the uses of virtualauditory space technologiesAs the popularity of so-called virtual environmentshas blossomed, the uses for 3D audio, as VAS isalso sometimes referred to, have becomeincreasingly more widespread (see Shinn-Cunningham and Kulkarni 1996; Shinn-Cunningham,Lehnert et al. 1997). In addition to the obviousgames and entertainment applications of socalled ‘virtual realities’, some of the earlyexperimental applications of 3D audio have beenin navigational and collision avoidance systemsby the air force (see for example Begault andPittman 1994).

For instance, in a high performance aircraftcockpit, there are numerous visual displays thatneed to be monitored almost simultaneously. Inhigh-G manoeuvres, vision can often becompromised along with the pilot’s gravitational

sense of up and down. Obviously, in low altitudesituations this can produce high risks for thepilot and the aircraft. One experimentalapplication of 3D audio has been to map theaircraft’s horizon indicator into an audio ‘icon’presented over headphones to the pilot. Whenthe plane is flying level and upright then the pilothears the ‘icon’ above the head of the pilot, butif the plane has banked heavily to say the leftthen the ‘icon’ will appear to be to the right ofthe pilot – that is, the sound presented in VASindicates which way is up for the pilot.

This is an example of a man-machine interfacewhich has mapped data from the gravitationaldomain (usually signalled by the organs ofbalance) into the auditory domain and thatprovides a very natural means for the pilot ofdetermining which was is up. More widespreadapplications of this technology awaits theimplementation of binaural sound systems intothe cockpits of high performance aircraft –something that is a target of the next generationof aircraft being developed in North America andEurope.

More recently, there has been growing interest inthe use of 3D audio to enhance speechdiscrimination in multi-talker communicationsystems to support teleconferencing andcommand and control activities. For instance, inair traffic control the ground control personneland the pilots often have to monitor multipleconversations over their headphones. Anyone whohas participated in a multi-person teleconferenceover the phone will know that it is very difficult tofollow individual conversations or to interactnaturally when people talk over each other in asituation where all of the talkers appear to becoming from one place (i.e. the telephonespeaker).

However, if the individual talkers are spatialisedso that they appear at different locations aroundthe listener then the ability to distinguish theindividual conversations is greatly enhanced.Under normal listening conditions, we naturallyuse the positions of different talkers to assist usin focussing our attention on what is being saidand to ignore other talkers or to shift ourattention from talker to talker. In fact such uses of

62

WAV

ES O

F TH

E FU

TURE

VAS technologies in multi-talker environmentshelp simulate normal communicationenvironments to allow us to naturally solve the‘cocktail party’ problem discussed at the start ofthis chapter.

References

Begault, D. R. and M. T. Pittman (1994). 3-D audio versus

head down TCAS displays. San Jose, San Jose State

University.

Best, V., A. v. Schaik, et al. (2004). ‘Separation of concurrent

broadband sounds sources by human listeners.’ J. Acoust.

Soc. Am. 115: 324-337.

Blauert, J. and W. Lindermann (1986). ‘Auditory spaciousness:

Some further psychoacoustic analyses.’ J. Acoust. Soc.

Am. 80(2): 533-542.

Bronkhorst, A. W. (1995). ‘Localization of real and virtual

sound sources.’ J. Acoust. Soc. Am. 98(5(Pt 1): 25422553.

Brungart, D. S. and W. M. Rabinowitz (1999). ‘Auditory

localisation of nearby sources. Head related transfer

functions.’ J. Acoust. Soc. Am. 106: 1465-1479.

Carlile, S. (1996). The physical and psychophysical basis of

sound localization. Virtual auditory space: Generation and

applications. S. Carlile. Austin, Landes: Ch 2.

Carlile, S., Ed. (1996). Virtual auditory space: Generation and

applications. Austin, Landes.

Carlile, S. and V. Best (2002). ‘Discrimination of sound source

velocity by human listeners.’ J. Acoust. Soc. Am. 111(2):

1026-1035.

Carlile, S., S. Hyams, et al. (2001). ‘Systematic distortions of

auditory space perception following prolonged exposure to

broadband noise.’ J. Acoust. Soc. Am. 110: 416-425.

Carlile, S., P. Leong, et al. (1997). ‘The nature and distribution

of errors in the localization of sounds by humans.’ Hearing

Res. 114: 179-196.

Carlile, S., P. H. W. Leong, et al. (1996). High fidelity virtual

auditory space: an operational definition. SimTecT,

Melbourne, Australia, DSTO.

Carlile, S. and D. Pralong (1994). ‘The location-dependent

nature of perceptually salient features of the human head-

related transfer function.’ J. Acoust. Soc. Am. 95(6): 3445-

3459.

Cherry, E. C. (1953). ‘Some experiments on the recognition of

speech with one and two ears.’ J. Acoust. Soc. Am. 25:

975-979.

Durlach, N. (1991). ‘Auditory localization in teleoperator and

virtual environment systems: ideas, issues and problems.’

Perception 20: 543-554.

Fisher, H. G. and S. J. Freedman (1968). ‘The role of the pinna

in auditory localization.’ Journal of Auditory Research. 8:

15-26.

Grantham, D. W. (1986). ‘Detection and discrimination of

simulated motion of auditory targets in the horizotal plane.’

J. Acoust. Soc. Am. 79(6): 1939-1949.

Grantham, D. W. (1995). Spatial hearing and related

phenomena. Handbook of Perception and Cognition. B. C.

J. Moore. San Diego, Academic.

Harris, J. D. and R. L. Sergeant (1971). ‘Monaural/binaural

minimum audible angles for a moving sound source.’ J.

Spch. Hear. Res. 14: 618-629.

Hartley, R. V. L. and T. C. Fry (1921). ‘The binaural location of

pure tones.’ Physics Rev. 18(6): 431-42.

Jin, C., S. Carlile, et al. (2000). Enabling individualised virtual

auditory space using morphological measurements. Int

Symp on Multimedia Info Processing, Sydney, Australia.

King, A. J. and S. Carlile (1994). Neural coding for auditory

space. The Cognitive Neurosciences. M. S. Gazzaniga.

Boston, MIT Press: 279-293.

Lambert, R. M. (1974). ‘Dynamic theory of sound-source

localization.’ J. Acoust. Soc. Am. 56(1): 165-171.

Langendijk, E. H. A. and A. W. Bronkhorst (2000). ‘Fidelity of

three-dimensional-sound reproduction using a virtual

auditory display.’ J. Acoust. Soc. Am. 107(1): 528-537.

Makous, J. and J. C. Middlebrooks (1990). ‘Two-dimensional

sound localization by human listeners.’ J. Acoust. Soc. Am.

87(5): 2188-2200.

63

THE

PSYC

HO

PHYS

ICS

OF

REAL

AND

VIRT

UAL

AUD

ITO

RYSP

ACES

Mershon, D. H. and L. E. King (1975). ‘Intensity and

reverberation as factors in the auditory perception of

egocentric distance.’ Percept Psychophys 18(6): 409-415.

Middlebrooks, J. C. and D. M. Green (1991). ‘Sound

localization by human listeners.’ Annu. Rev. Psychol. 42:

135-159.

Mills, A. W. (1958). ‘On the minimum audible angle.’ J.

Acoust. Soc. Am. 30(4): 237-246.

Møller, H., M. F. Sorensen, et al. (1995). ‘Head-related transfer

functions of human subjects.’ J. Audio Eng. Soc. 43(5):

300-321.

Okano, T., L. L. Beranek, et al. (1998). ‘Relations among

interaural cross-correlation coefficient (IACC(E), lateral

fraction (LFE), and apparent source width (ASW) in concert

halls.’ J. Acoust. Soc. of Am 104(1): 255-265.

Perrott, D. R. and A. D. Musicant (1977). ‘Minimum audible

movement angle: Binaural localization moving sound

sources.’ J. Acoust. Soc. Am. 62(6): 1463-1466.

Pollak, I. and M. Rose (1967). ‘Effect of head movement on

the localization of sounds in the equatorial plane.’

Perception and Psychophysics 2(12): 591-596.

Potter, J. M., J. Raatgever, et al. (1995). ‘Measures for

Spaciousness in Room Acoustics Based on a Binaural

Strategy.’ Acta Acustica 3(5): 429-443.

Pralong, D. and S. Carlile (1994). ‘Measuring the human

head-related transfer functions: A novel method for the

construction and calibration of a miniature ‘in-ear’

recording system.’ J. Acoust. Soc. Am. 95(6): 3435-3444.

Shinn-Cunningham, B. and A. Kulkarni (1996). Recent

developments in virtual auditory space. Virtual auditory

space: Generation and applications. S. Carlile. Austin,

Landes: Ch 6.

Shinn-Cunningham, B. G. (2000). Distance cues for virtual

auditory space. IEEE 2000 International Symposium on

Multimedia Information Processing, Sydney, Australia.

Shinn-Cunningham, B. G., H. Lehnert, et al. (1997). Auditory

displays. Spatial and Biaural Hearing. R. Gilkey and T.

Anderson. New Jersey, Lawrence Erlbaum Associates,

Publishers: 611 - 664.

Wallach, H. (1940). ‘The role of head movements and

vestibular and visual cues in sound localization.’ Journal of

Experimental Psychology 27(4): 339-368.

Wightman, F. L. and D. J. Kistler (1989). ‘Headphone

simulation of free field listening. II: Psychophysical

validation.’ J. Acoust. Soc. Am. 85(2): 868-878.

Wightman, F. L. and D. J. Kistler (1999). ‘Resolution of front-

back ambiguity in spatial hearing by listener and source

movement.’ J. Acoust. Soc. Am. 105(5): 2841-53.

Zahorik, P. (2002). ‘Assessing auditory distance perception

using virtual acoustics.’ J. Acoust. Soc. Am. 111(4): 1832-

1846.

64

WAV

ES O

F TH

E FU

TURE

conscious part of our brain, withthe remaining 90% thesubconscious part. (In reality, thereis no such neat division.)

In the 1980s, Yorkshire TV in theUK showed a documentary called,Is Your Brain Really Necessary? Itdescribed the work of the lateBritish neurologist, Professor JohnLorber. He was a kids’ doctor, andhe saw many cases ofhydrocephalus. Hydrocephalushappens to about one of two kidsout of every 1,000 kids that areborn alive. There is a constantcirculation of cerebro-spinal fluidaround the brain and spinal cord. Iftoo much fluid is produced, or ifthere is a blockage to its outflowfrom the brain, then it can build upinside the skull. This excess fluidusually makes the skull growbigger, but sometimes it just makesthe brain meat get thinner as themeat gets squashed up against thebony skull.

Professor Lorber discussed manycases where young people had notmuch brain, but normal intelligence.

This myth has been going fornearly a century, and it keeps re-emerging. Over the last decade,some motivational speakers haveshamelessly recycled this myth,and they claim that if you take theirexpensive course, you will suddenlybe able to use all of yourbrainpower.

One of the earliest popularmentions of this myth is in DaleCarnegie’s 1936 book, How To WinFriends and Influence People. Hewanted to back up his claim that ifyou worked your brain just a littleharder, you could improve your lifeenormously. Without anyneurological proof whatsoever, heboldly claimed that most peopleused only 15% of their brains. Hisbook sold very well indeed, so thathelped push the myth.

Dale Carnegie might have got hismisinformation by wronglyinterpreting the experiments of acertain Karl Lashley, back in the1920s. He was trying to find outjust where in the brain this strangething called “memory” is stored. He

trained rats to run through mazes,and then measured how well theydid as he removed more and moreof the cortex of their brains. Hefound that memory is not stored inone single place, but existsthroughout the entire cortex, andprobably a few other places aswell. In fact, his results showed thatremoval of any of the cortex causedmemory problems. Karl Lashley’sfairly-straightforward result wassomehow changed to read that ratsdid fine until they had only 10% oftheir brains left. First, he neverclaimed that. Second, he neverremoved as much as 90% of thebrain.

This myth has been constantlyreinvented every decade or so. Soone version might have thatcertified mega-brain, AlbertEinstein, saying (guess what?) that“we use only 10% of our brain”.But I have also heard the versionwhere some anonymous scientist(who is never named) supposedlydiscovered that we use only 10% ofour brain. Another version is that10% of the mass of the brain is the

Use Your BrainBy Dr Karl Kruszelnicki

T H E H U M A N B R A I N is one of the most complicateddevices that you could think of. It’s a very expensive organ –from a metabolic point of view. It takes a lot of energy to runthe brain. Even though it weighs only about 2% of our bodyweight, it uses about 20% of our blood supply and 20% ofour energy – and it generates about 20% of our heat. Thereare many myths about this mysterious organ. One persistentmyth is that we really use only 10% of our brain – and that ifwe could use the remaining 90% we could each win a NobelPrize or a gold medal at the Olympics, or even unleash oursupposed psychic powers.

65

USE

YOUR

BRAI

N

In one extraordinary case, a youngman had only one millimeterthickness of gray matter in hisbrain, instead of the average 45mm. Even so, he had an IQ of 126(the average is 100) and had gainedan honours degree in mathematics!

But this does not prove that most ofyour brain is useless. Instead, itshows that in some cases, thebrain can recover from, orcompensate for, quite majorinjuries.

The myth that we use only 10% ofour brain is finally being proveduntrue, because over the last fewdecades, we have invented newtechnologies (such as PositronEmission Tomography andFunctional Magnetic ResonanceImaging) that can show themetabolism of the brain. In any onesingle activity (talking, reading,walking, laughing, eating, looking,hearing, etc) we use only a few percent of our brain – but over a 24-hour day, all the brain will light upon the scan.

In fact, if you did use all of yourbrain at the same time, you wouldprobably have a Grand Malepileptic fit. And finally, have youever heard a doctor say, “Luckily,he had a stroke in that 90% of thebrain we never use, so I think he’llbe alright.”?

FROM Dr Karl’s book Mythconceptions

(Harper Collins Publishers)

66

PROFESSOR MARTIJNDE STERKE studiedEngineering and AppliedPhysics at DelftUniversity in theNetherlands beforereceiving his doctoratein optics from theUniversity of Rochesterin the USA in 1988. Heheld a position as aresearch fellow at theUniversity of Toronto,Canada until 1990,

before moving to Sydney in 1991.

Martijn is currently a Reader with theSchool of Physics’s Centre forUltrahigh-bandwidth Devices forOptical Systems – CUDOS – doingresearch into the next generation ofcommunication systems that will uselight, not electricity, to carry andprocess information. He is atheoretical physicist, whoseapproach to his research ischaracterized by actively seekingcollaborations with experimentalists.He has authored papers in the fieldsof optics and photonics, solid-statephysics, and acoustics, and thesepapers have appeared both in thephysics and in the engineeringliterature.

What got you interested inscience in the first place?

I did not see suddenly see thelight, so to speak. I did physics in highschool and thought it was really neat.I also saw a number of TV programson particle physics for TV in which thediscovery of the J/psi particle wasdescribed – that also got me interested.

What were you like as a kid? Wereyou curious, pulling apart stuff tosee how it worked?I never pulled things apart – whichmight explain why I evolved into atheorist.

What’s the best thing about beinga researcher in your field?This is a good time to be aresearcher in optics. Many of theproblems are very fundamental, yetthe applications are real and not toofar in the future.

Who inspires you – either in scienceor in other areas of your life?Early in my career I had the greatprivilege to work with one of thegreats in the field. Someone whomade time for everyone in his (verylarge) group on a weekly basis,worked in half a dozen differentareas, and is one of nicest guys inthe universe.

If you could go back andspecialise in a different field, whatwould it be and why?During high school biochemistry wascoming up strongly, but it did nothave the rigorous basis of physics forexample. In another life I might havedone biochemistry.

What’s the ‘next big thing’ in science,in your opinion? What’s coming upin the next decade or so?Essentially unlimited bandwidth forall, artificial life forms, and mobilephones that dispense coffee in themorning.Fig 1. Representation of the links that are powering the internet.

Telecommunications: the here and nowMartijn de Sterke

P E O P L E C O M M U N I C A T E V I A many different means: speaking to each otherin person, over the phone, mobile or fixed line, or increasingly throughthe world-wide web, sending each other letters or email messages,pictures and movies – Figure 1 shows a schematic of the main internetconnections in the world. Here I will discuss aspects of communicationsscience and engineering, and some of the issues involved in creatingthe high-speed communications that allowed the development of theinternet. I also want you to get some appreciation of the relevantnumbers related to quantities of information that come up in this area.In this lecture I will predominantly look to the present: how are thingsdone at present and why. In the next chapter I will discuss thechallenges coming up if we want to achieve even larger data transferrates than now, and some of the ideas that have appeared on thehorizon to deal with some of these issues.

67

68

WAV

ES O

F TH

E FU

TURE

As I mentioned, there are many means ofcommunications; let us first use long-distance telephone as an example. Any

long-distance phone call makes use of an opticalfibre. First I want to discuss why fibres are usedand why they have displaced other means ofcommunication. To understand the uniqueadvantages of optical fibres we need to understandsome communication theory, an area thatstraddles mathematics and electrical engineering.Communication theory was developed after WorldWar II by, among others, Claude Shannon in theUS and Norbert Wiener in Europe. Here we willfollow some of the ideas developed by Shannon(his picture is shown in Figure 2). I will start withsome of the elementary ideas and then fast-forward to the result that we require.

[FIGURE 2 HERE]

Defining InformationSince communication is about taking informationfrom one place to another, we first need a gooddefinition of information. The definition we adoptis perhaps initially somewhat counterintuitive: itessentially measures the amount of uncertaintythat can be resolved. Let us initially take digitalcommunications, where we communicate viasignals that can only have discrete values, as anexample. Suppose the variable that we measurecan take one of N values, each with a probabilitypn, such that, of course, ∑ pn =1. As a simpleexample, if the variable is a dice then N=6, and if

the dice is unbiased then pn=1/6 for each n = 1,..., 6. In contrast, if the dice is biased then not allpn are the same – but the probabilities must stilladd up to unity. The question is now: if you aregiven just one outcome and nothing else, what isthe expected information that you have beengiven?

Let us first see if we can find some generalproperties that the expected information shouldsatisfy. First, negative information makes nosense, and so the expected information must benon-negative.1 Now, when N=1 (which would belike tossing a coin with heads on both sides) thenno information at all is given, because the answerwas known in advance (since the coin wouldalways come down heads!).

Further, if N=6 and pn=1/6 for all n (an unbiaseddice) then the expected information per outcomeis larger than when N=2, corresponding tothrowing an unbiased coin – larger, becausethere were more possible outcomes for the roll ofthe dice than the toss of the coin, so the actualoutcome of the dice roll carries, in a sense, moreinformation than a toss of a coin.

Let us now consider not N=1, but a situationclose to it: we take N=2, but p1=1-ρ and p2=ρ.When ρ Y0, or when ρ Y1, one of the outcomesis certain and the other impossible – it becomesthe same as the N=1 case above. For thisparticular case of N=2, then, the expectedinformation is zero when ρ is zero or one, and itpresumably peaks somewhere in between.

Arguments of this type can be continued for awhile, but let us not further beat around the bush– the expected information H per outcome isgiven by

H = -∑ pn log pn. (1)

Note that this function satisfies the requirementsdiscussed earlier: since all pn are between 0 and1, H cannot be negative.2 The logarithm ensuresthat knowing two outcomes gives twice theexpected information as that of a single knownoutcome, which can be checked for example bycomparing H for the tossing of an unbiased coinor dice once and twice.

Fig 2. Claude Shannon.

69

TELE

CO

MM

UNIC

ATIO

NS:

THER

EH

ERE

AND

NO

W

In communication theory we’re talking abouttransferring messages of some kind, andmessages come as a string of symbols. In thiscase, the equivalent of seeing the outcome of aroll of the dice is seeing what symbols appear inour message – so in communications we calloutcomes ‘symbols’.

The only thing left to discuss now is the base ofthe logarithm: the obvious ones are base 10 logsor natural logarithms with base e. However, incommunication theory one uses base 2logarithms, and the unit of information is then abit. Let us consider an unbiased coin: in that caseN=2 and p1=p2=1/2. Since log2(1/2) = -1 wefind that H=1 bit per symbol. While Equation (1)gives the expected information per symbol,multiplication by the number of symbols persecond gives, of course, the expected informationper second.

Information or Entropy?Some of you may think that Equation (1) looksfamiliar – indeed, apart from a factor k=1.38x10-23 J/K (the Boltzmann constant, whichplays a central role in thermodynamics) and afactor 1/ln 2 (the difference between base 2 logsand natural logs), essentially the same expressionis used in physics for the definition of entropy. Infact, the expected information is often referred toas the entropy, a fundamental function inthermodynamics. In this definition the N can referto the number of particles (atoms) in the systemunder study, and the pn to the probability for thesystem to be in a particular state. For an atomthis could refer to the energy levels that theelectrons have access to. This relation betweenthe expected information and the entropy is nocoincidence, as the entropy can be thought of asmeasuring the degree of disorder of a physicalsystem – a large degree of disorder means that alot of information is required to describe thesystem. Indeed, many of the more advancedconcepts in communication theory are couched interms of the entropy.

It is now easy to make the link between theentropy and the way we can send informationthrough an information channel, the details ofwhich do not matter at this point. Suppose the

input into an information channel is indicated byZ, which signifies all possible inputs zi each withprobability pi. For a binary signal, the zi can takethe values ‘0’ and ‘1’, each of which is likely tohave a probability of 50% (though we could easilygeneralise to cases where z is not 50%). Theoutput of the communication link is similarlyindicated by W, with the symbols now indicatedby wj.

Message sent, message receivedWe now address the following question, whichgoes to the heart of communications: if we knowwhat message is received, what does that sayabout the message that was sent? That is, giventhat we know the output W, what can we sayabout the input Z? To start answering thisquestion, we introduce the idea of ConditionalEntropy, which is defined as

H(Z / W) = -∑ij p(zi,wj) log(p(zi / wj)). (2)

Here p(a/b) is the conditional probabilityp(a/b)=p(a,b)/p(b).3 The Conditional Entropy canbe interpreted to be the uncertainty in ourknowledge of Z if W is known. As an extremecase, consider when Z and W are independent,so the output of the channel bears no relation tothe input (which would not be a very usefulcommunications channel, it’s true). ThenH(Z/W)=H(Z,W)/H(W)=H(Z)H(W)/H(W)=H(Z), andknowing W has not cleared up anything about Z.Here the first equality is the definition of H(Z/W)and the second uses the definition ofindependent events from Footnote 3.

The other extreme case occurs when W givesperfect information about Z – in that case there isno uncertainty left when W is known andH(Z/W)=0, since every term in Equation (2) isproportional to log(p(z/w)) = log(1) = 0.

Having defined the conditional entropy allows usfinally to introduce the Mutual Information, whichcorresponds to the information regarding theinput W that is obtained by knowing the output Z.The mutual information is defined to be

I(W;Z) = H(Z) – H(Z/W). (3)

70

WAV

ES O

F TH

E FU

TURE

This is not surprising if you think about it: if H(Z)is the uncertainty that would be cleared up if Zwere known, and H(Z/W) is the uncertainty that isleft regarding Z if W is known, then the differencebetween these two is the information that Wprovides about Z. Considering the two extremecases discussed in the previous paragraph again:in a very bad channel where the output does notdepend on the input, H(Z/W)=H(Z) and thusI(W;Z)=0 – no information whatsoever isprovided. In a perfect channel, though, H(Z/W)=0,and so I(W;Z)=H(Z), and so the MutualInformation equals the Expected Information. Inpractice, of course, the properties of a channelare somewhere between these extremes.

Capacity of the Information ChannelNow that we know a little about communicationtheory, let us concentrate on a channel that takesinformation from one place to another. We like toknow the Capacity of this channel, the maximumamount of information that can possibly besqueezed through this channel given its physicallimitations. This does not necessarily mean thatinformation is actually lost. Using clever codingtechniques, for example error correcting code, itis possible to correct the errors that were madein transmitting the information.4 However, codingtheory is well outside the scope of this lectureand we will not discuss it further; here we areinterested in the physical characteristics of theinformation. Using the concepts introduced earlierthe Capacity of a channel is defined as

C = max | P(Z) I (z, w); (4)

Here max|P(Z) means the maximum value that canbe reached by adjusting the P(Z). This means thatthe number of input symbols and their probabil-ities are chosen such that the mutual informationof the input and the output is maximized.

As an example, consider a system with binaryinformation encoded in 0s and 1s as discussed inFootnote 3. During the travelling through thechannel the signal is subject to noise and otherdeleterious effects. As a consequence, there is asmall chance p of an error, so that a 0 is receivedas 1 or vice versa. Since the error probability issymmetric between 0s and 1s, the mutual

information between input and output ismaximized when p(0)=p(1)=1/2.

Using a bit of thought you can find from Equation(3) that, in this case, C=1+(1-p)log(1-p)+plog(p)and so C=1 bit when p=0, but drops to C=0.92bit when p=1%. When p=0.5 – that is, wheneach bit has a 50% chance of being wrong –then C=0 and all information is lost in thechannel. (When the symmetry between 0s and 1sis dropped, that is when the noise in the 0s and1s are different, then the capacity is not achievedfor p(0)=p(1)=1/2.)

Capacity and BandwidthWith the formal definition of the Capacity of achannel we are now in a position to discuss oneof the most celebrated theorems incommunication theory, derived by Shannon. Awell known form of this theorem does not applyto the digital signals that we have beenconsidering up to now, but with analogue signals,but this does not matter much. The theoremrefers to an analogue signal of power S that isaffected by noise. The noise is assumed to bewhite noise, which means that the value of thenoise can change arbitrarily quickly in time, andto have average power N.

At each particular time, the noise is assumed tohave a Gaussian distribution, which indicates theprobability that the noise has some particularvalue at a given time.5 These assumptions areidealizations, but this can be turned into apositive: the situation that is analysed is, in somesense, the best possible one, and the situation toaim for in practice. Under these conditions, thecapacity was proven by Shannon to have thevalue

C = B log ( 1 + S/N ), (5)

where B is the bandwidth. Equation (5) has unitsof inverse seconds, and thus refers to thecapacity per unit time. It is important toappreciate the conceptual difference betweenEquations (4) and (5): the former is simply adefinition of the Capacity, whereas the secondgives the value of the actual capacity in acommunication system under the conditions

71

TELE

CO

MM

UNIC

ATIO

NS:

THER

EH

ERE

AND

NO

W

specifically discussed above. It is clear that, asexpected, when the noise power increases, thecapacity goes down. However, since the signal-to-noise ratio S/N enters in logarithmic form, inpractice it does not vary by very much – I returnto this below.

There are now two things to discuss: what ismeant by the bandwidth, and what exactly doesEquation (5) mean? Since most of us have atleast a vague notion of bandwidth, let us startwith the latter. The meaning of Equation (5) is thatif a data stream with H<C is launched throughthe channel then, even though noise is present, acoding scheme exists by which the error betweeninput and output can be made arbitrarily small. Incontrast, if H>C then no such code exists andtransmission errors inevitably occur.

The theorem is obviously very powerful but itshares the infuriating characteristics of all‘existence theorems’: proving that something canbe done does not necessarily mean that youknow how to do it. In fact, codes that can achievethe theoretical limit have not yet been devised!

Let us now return to the bandwidth B thatappears in Shannon’s result: it basically tells ushow fast the signal can change with time. If thebandwidth is low, this means that the signalvaries only slowly with time and not muchinformation can be transferred. Suppose, forexample, that we have some signal that carriesinformation, say that in Figure 3. It is carried byan information channel at some rate. Nowcompress the time axis by a factor of 2 and do

the same thing. Then the same information iscarried through the channel in half the time; thusthe information is being transferred at a ratetwice as large as before. In the language that wehave developed here this means that the entropyH that can be transferred per unit time hasdoubled too.

This procedure would imply that arbitrarily highrates can be achieved, but this is of courseimpossible – the bandwidth of the channel limitsthe rate at which the signal can change its value,and thus limits the rate at which information canbe transferred. This is why the bandwidth appearsin Equation (5). The bandwidth B is defined to bethe range of frequencies that can be sent throughthe channel. One way of understanding theoccurrence of the bandwidth in Shannon’s result(5) is to imagine a scheme in which theinformation is transferred using short pulses. Thepresence of a pulse would indicate a 1, while theabsence would indicate a 0. Clearly, the shortereach pulse, the more tightly the pulses can bepacked, and the higher the information transfer.By definition, a short pulse needs to have steepsides, and a steep side can only be generated byhigh frequencies.

Let us use Equation (5) to see how we canoptimize the capacity of a channel. You might firstwant to imagine why you would want to increasethe capacity in the first place. Take as an examplethe internet, and the time it takes to downloadvarious files. Movies are massive files, and yetyou do not want to wait more than a few minutesto download one of them. Future applications ofthe internet will require even larger bandwidths.In fact, because new telecommunicationstechnology always has unforseen applications,the historical record of predicting demand forcapacity is poor and has always been much lower than the actual outcome, as illustrated inFigure 4.

Returning now to the question as to how thecapacity may be increased, note that this can bedone by making the signal-to-noise ratio S/N aslarge as possible. However, this is not aparticularly good way to go about this, since theratio appears in the argument of a logarithm, afunction that varies very slowly as a function of

Fig 3. Example of a signal with Gaussian noise as afunction of time.

72

WAV

ES O

F TH

E FU

TURE

the argument. For example, to make the capacityten times larger, the signal to noise needs toincrease by roughly a factor 210=1024. Instead, itis more efficient to try and increase the capacityvia the bandwidth, as it enters into Shannon’sresult linearly: a ten-fold increase in the Capacitycan be achieved by a tenfold increase in W. Thus,the way to increase the capacity is to increasethe bandwidth.6

Optimising BandwidthThe question now thus is: how can we optimizethe bandwidth of a communications channel? Letus think about AM radio waves: these have awavelength of about λ≈1 km and thus afrequency f=c/λ≈300 kHz, where c is the speedof light in vacuum. Clearly, the maximumbandwidth that can be achieved usingmicrowaves is a few times 105 Hz, and thecapacity is thus roughly similar. To make thebandwidth any larger you would need waves withhigher frequencies, which would no longer be AMradio waves. This is a general result: themaximum bandwidth that can be achieved isroughly equal to the highest frequency that canbe used to carry the information.

The result from the previous paragraph nowclearly shows how to optimize the capacity: wewant to use waves with the highest possiblefrequency. Of course the communication channelneeds to be transparent to these waves, and wemust be able, somehow, to encode informationon (or ‘modulate’) this wave. So if we want highfrequencies, let’s consider X-rays: they havewavelengths λ≈1 nm, and so f=3x1017 Hz. X-rays

go through all kinds of materials, so the channelwould be transparent.

The problem is that we simply cannot manipulateX-rays particularly well to encode them with anydecent amount of information. Other high-frequency waves are in the Ultraviolet part of thespectrum: but now the problem is that UVradiation is easily scattered or absorbed, so thechannel isn’t transparent anymore, and, as withX-rays, we do not know how to manipulate UVradiation well enough. The class of waves withthe next lower frequency range is visible light. Wecan easily generate these, we understand themwell, and some materials are very transparent.We also know how to modulate light and so wemay consider this option in some more detail.

Though light propagates through air quite well(after all, we can see each other!), over longdistances the light tends to diffract, which meansthat it spreads while it propagates. Lightpropagation through air also needs a clear line ofsight at all times. Both of these problems can beovercome when light propagates through anoptical fibre. In an optical fibre the light isconfined in the fibre core and cannot escapefrom it, and it therefore does not diffract.

This can briefly be understood using a schematicof an optical fibre, shown in Figure 5(a). It showsa small core with a diameter of 8.3 µm,surrounded by a much larger cladding with adiameter of 125 µm. The cladding is made ofpure silica and has a refractive index around1.45. The cladding has a refractive index that isslightly higher since it is doped with a smallamount of germanium. Germanium is chemicallysimilar to silicon, since they appear in the samecolumn in the periodic table; the addition ofgermanium increases the refractive index since agermanium atom has more electrons than asilicon atom. In the standard SMF28 optical fibrethis difference in refractive index amounts to only0.36%, i.e., the refractive index differencebetween 1.45 and 1.455.

Now remember Snell’s law:

n1 sin θ 1 = n2 sin θ 2 (6)

Fig 4. Evolution of installed bandwidth with time, comparedto various predictions. The latter have always been severeunderestimates.

73

TELE

CO

MM

UNIC

ATIO

NS:

THER

EH

ERE

AND

NO

W

It applies when light, travelling through a mediumwith refractive index n1, is incident at angle (1 ona medium with refractive index n2 (see Figure5(b)). Angle θ2 then gives the direction in whichthe light propagates through medium 2. Supposenow that light is in the core, which has a slightlyhigher refractive index than the cladding. Then ifangle θ1 is sufficiently large the left-hand side ofEquation (6) can be larger than n2 – and so wecannot find an angle θ2 that satisfies theequation. As a consequence, the light is totallyinternally reflected off the interface between thecore and the cladding, and must stay inside thecore. This can only work if the light is initially inthe high-index medium and is the reason why thefibre has a slightly higher refractive index thanthe cladding.

The light remains confined to the fibre core, evenwhen the fibre is bent (by a modest amount, afew centimetres or so), and so there is no needfor a clear line of sight. Figure 6 shows the lossin a number of optical fibres as a function ofwavelength. Note that it reaches its lowest valuearound λ=1.55 µm – not in the visible part of thespectrum, but the near-infrared.

Here the losses are 0.2 dB/km, where dB standsfor decibels – to understand what this means youneed to know that 10 dB corresponds to a factor10, that 20 dB corresponds to a factor 100, etc.In general, the value in dB of a factor F is10log10(F). Since 100.3≈2, that means 3 dBcorresponds to a factor of 2. Coming back to theloss of 0.2 dB/km, this corresponds to 3 dB per15 km – and so the (infrared) light loses half ofits strength after travelling an astonishing 15 km.Once the losses get too large the radiation cannotbe used, so the bandwidth of infrared light in anoptical fibre corresponds to roughly 200 nm,which is approximately 25 THz.7 AssumingS/N=100, corresponding to 20 dB, Equation (5)then gives a capacity of 166 Tb/s.

As per our previous discussion, this bandwidth isorders of magnitude higher than can be achievedwith, for example, AM radio waves, since thesehave much lower frequencies. All optical fibresmade of silica have their minimum loss aroundλ=1.55 µm, and so all long distance opticalcommunication uses light in this part of thespectrum.

Bandwidth in PerspectiveLet us put the number obtained in the previousparagraph in some sort of perspective. Have alook at Figure 7, showing what physical form theinformation in the world takes. Note that thisgraph has a logarithmic vertical scale and thatthe units are Petabytes (1015 bytes=8x1015 bits).Here the length of one character of information is1 byte = 8 bits.

The Library of the US Congress contains a totalamount of information of a few Petabits. Whichsounds like a lot of information – until you realisethat an optical fibre can transfer all this

Fig 6. Propagation losses in a number of different optical fibres.

Fig 5. (a) Schematic of the cross section of an optical fibre.(b) Illustration of Snell’s law and total internal reflection.Here n1 is taken to be larger than n2.

74

WAV

ES O

F TH

E FU

TURE

information in 10-15 seconds or so. The totalmemory of all people has been estimated to be1000 Pb, which, with 6 billion people on thisplanet, gives a modest 100-200 Mb each.Though this is more than a floppy disc, it roughlycorresponds to the amount of information thatcan be stored on a flash memory device, and isdwarfed by any type of memory disc – a soberinginsight for members of the genus homo sapiens!

Note from the table that the total digital storagecapacity available, mostly disc space, is anotherfew orders of magnitude larger. Let us alsoconsider the total amount of information that isavailable in the entire universe. Lloyd hasestimated this to be (t/tp)3/2 bits, where t is theage of the universe (about 1010 years), andtp=√(Gh/c5) – in this equation G=6.67x10-11

Nm2/kg2 is the gravitational constant that entersNewton’s theory of gravity, and h=1.05x10-34 Jsis Planck’s constant, the constant that governsquantum processes. The time tp is the Plancktime, which enters into research to understandnature at its most basic level. Substituting thenumbers we find the resulting storage capacityof the universe is something like 1091 bits.8 Thisputs the results in Figure 7 somewhat inperspective.

Bandwidth: theory vs. realityMuch of the above (in this context particularlyEquation (5), but also the estimated number ofbits in the universe) was based on existencetheorems. This means something can be done inprinciple, but does not say how it is to beachieved. So the natural question to ask is: how

well does it work in practice? The theoretical limitgiven by Shannon’s theorem has not yet beenreached, but the capacity of an optical fibre isnonetheless incredibly high.

Below I discuss some of the ways in whichinformation is sent through optical fibres. Thekey word to know here is that of WavelengthDivision Multiplexing, or WDM for short. Beforeexplaining this let us use an analogy. Earlier wediscussed AM radio and found that the capacityis quite low because of the low frequencies thatare involved.

However, what we did not discuss is that the totalAM band carries roughly 10 stations, each ofwhich has a different centre frequency. Eachradio station needs a certain bandwidth of tens ofkHz, and so the frequencies of the differentstations in a particular geographical area must beat least this far apart. Different radio stations canbe picked up by changing the centre frequencythat the radio receives. Wavelength divisionmultiplexing works similarly.

We saw in the discussion of Figure 6 that thetotal bandwidth available in a fibre is about 25THz. In practice this bandwidth is chopped up intonarrow bands, the analogue of a radio station,with a width of typically 50 or 100 GHz. Each ofthese bands is, confusingly, referred to as a“channel”. Each channel corresponds to a slightlydifferent colour than the other bands, and ismeant to transfer information independently ofthe others. A schematic of the setup is shown inFigure 8. A set of different lasers, each operatingat slightly different frequencies, encodes theinformation for each of the channels. These arethen combined in a multiplexer and sent throughan optical fibre. To account for the losses, eventhough these are very small, a number ofamplifiers need to be included in the fibre link,but the details do not matter here.

At the far end of the fibre link the differentchannels are separated again and fall on differentreceivers. In this case the data rate of each WDMchannel is about 100 Gb/s, and there are roughly100 channels, so the total aggregate data rate is10 Tb/s – not that close to the theoreticalmaximum of Shannon.

Fig 7. Estimates of amounts of information available storedin various ways.

75

TELE

CO

MM

UNIC

ATIO

NS:

THER

EH

ERE

AND

NO

W

But let us finish with some perspective on thisdata rate of 10 Tb/s from the previous paragraph.A total capacity of 10 Tb/s corresponds to tens ofCDs per second. What can be transferred in onehour? The entire library of the US congress; thememory of about 10 millionpeople; the total genomic codeof 1 million people. However, itcorresponds to only about 10-17

of the information needed toteleport a human.

In the next chapter we will see anumber of effects that limit thedata rates described here andthat limit us increasing the datarates even more, which wouldallow applications such astelemedicine and otherapplications that link to remotelocations. My personal favouriteapplication, though somewhatremote, is for people to toucheach other at a distance –wouldn’t it be great ifgrandparents could hold theirnew grandchild at a distance and have it feel as ifthey had the child in their arms? In lecture twowe will also consider possible solutions to theseproblems.

I am grateful to Dr Richart Slusher from BellLabs, Lucent Technologies, for the use if some ofhis figures. I am also grateful to my colleagues atCUDOS, particularly Ben Eggleton for discussionsregarding some of the issues raised here.

Notes

1 Let us here not consider certain reality shows on TV which

may violate this assumption.

2 Remember that, for a logarithm of any base, log(x)>0 for

x>1 and log(x)<0 for 0<x<1. In this definition we have to

be careful since log(0) diverges. However it does not

diverge very quickly and so 0log(0)=0, which can easily be

seen using a limiting procedure.

3 Recall that P(a,b) is the probability that both a and b occur,

whereas P(a/b) is the probability that a occurs, given that

we know that b occurs. When a and b are independent of

each other, then P(a,b)=P(a)P(b) and thus P(a/b)=P(a): the

fact that b occurs does not affect a. In the other extreme

case event a always occurs in conjunction with b and thus

P(a,b)=P(b), so that P(a/b)=1. An example of two

correlated events is the observation of a Christmas tree,

and the date being 25 December. In contrast the events

that the moon is full and the date being 25 December are

uncorrelated. An often used example in a communications

context is a binary channel in which the

probability of the bit arriving incorrectly

is P(0/1)=P(1/0)=p and the probability

that they arrive correctly is thus

P(0/0)=P(1/1)=1-p.

4 A simple type of error correcting code

is to send every symbol 3 times, say. If

one of the three symbols arrives

incorrectly than this can easily be

detected. This method assumes that the

probability of receiving the wrong signal

is small, so that it is very unlikely that

two of the three symbols are received

incorrectly, since then the wrong

conclusion would be drawn. In practice

the error correcting codes are much

more sophisticated, but outside the

scope of the discussion here.

5 This means that the probability that the

noise has a particular value s is

proportional to exp(-s2/σ2) where σ is a

characteristic width that here is related to S.

6 Amongst all this discussion about high Capacity and

bandwidth, we may want to pause for a minute and think

about channel with very low Capacity. We may want to

discuss this during the lecture, and perhaps a question to

think about is the following: what is the lowest capacity

channel that you can think off?

7 Here ‘T’ is short for Tera, 1012

8 S. Lloyd, “Computational capacity of the universe,” Phys.

Rev. Lett. 88, 237901 (2002). Note that there is a typo in

this paper: the 3/4 exponent should be 3/2.

Fig 8. Schematic of a WDM communications system.

What can be transferredin one hour? The entirelibrary of the UScongress; the memory ofabout 10 million people;the total genomic codeof 1 million people.However, it correspondsto only about 10-17 of theinformation needed toteleport a human.

76

WAV

ES O

F TH

E FU

TURE

events into dynamic leading playersin their own right. His were trulygroundbreaking ideas that pavedthe way for the era of modernphysics.

General RelativityThe ideas of 1905 were just thepreamble to what was arguablyEinstein’s greatest achievement: histheory of General Relativity. In thismost beautiful of works he showedhow space and time, and energyand matter, act and react toproduce the effect we know asgravity. The theory’s consequences– warped space, stretched time,bent light beams and black holes –seem very strange and non-intuitive, yet today the theory is avital part of science and technology.For example, Global PositioningSatellite (GPS) systems rely onGeneral Relativity for theirincredible accuracy.

The questions left open by theGeneral Theory of Relativity are stilltantalizing scientists today, as theysearch for evidence of gravitywaves, attempt to figure out if theUniverse is open or closed, and tryto find ways to reconcilefundamental difficulties with mixingGeneral Relativity with the QuantumTheory.

If there is such a thing as asuperstar scientist, Albert Einsteinmust be it. His shock of wild hair,his great nose and bushymoustache, his craggy features –Einstein’s image gazes wisely outfrom posters, t-shirts, coffee mugs... even cuddly toys and actionfigurines!

What is it about him that capturesthe public’s imagination this way?Perhaps it’s that he seems so sageand yet so gentle, he’s theintellectual giant but also the kindlygrandfather-figure. Perhaps it’s theway he tried – naively some wouldsay – to influence politics the wayhe influenced global science.Perhaps it’s that he’s the archetypalabsent-minded professor, toopreoccupied with the secrets of theuniverse to be concerned withhaircuts or matching socks.

An inauspicious beginning Albert Einstein was born in 1879 atUlm in Württemberg, Germany, theson of middle-class Jewishparents. He was a quiet, curiouschild; his parents were concernedthat he didn’t start speaking untilhe was three years old. At schoolhe didn’t exactly stand out amongsthis peers, and left at the age offifteen. He worked in his parents’

business, spent some timetravelling and eventually enrolled inthe Federal Institute of Technology inZurich. He graduated in 1900 with afairly ordinary academic record.

Afterwards, he did what manygraduates do: he looked for work,taking various odd jobs beforelanding a position as a patentexaminer with the Swiss patentoffice. And there he worked, forseveral years ...

The Miraculous YearPrior to 1905, nothing in Einstein’sacademic career foreshadowed theevents to come. In just one year, hewrote a series of five papers thatchanged our view of the universeforever and cemented hisreputation as one of the finestminds of his or any othergeneration.

In 1905, Einsein wrote aboutBrownian Motion and demonstratedconclusively that atoms really doexist; he took the notion ofquantised energy, previously anabstract mathematical trick, andshowed how it explained perfectlythe interactions of light and matter;and he merged space and time,transforming them from abackground arena for physical

I WANT TO know how God created this world. I am notinterested in this or that phenomenon, in the spectrum of this or that element. I want to know His thoughts; the rest are details.

Albert Einstein: Scientist or Superstar?

77

ALBE

RTEI

NST

EIN

:SC

IEN

TIST

OR

SUPE

RSTA

R?

Einstein the Political ScientistEinstein’s political reputation attimes matched his scientificreputation. When the First WorldWar broke out in August 1914,Einstein was vilified as a traitor bysome in Germany for his pacifisticideas and his open criticism of thegovernment’s ambitions. Though heremained in Germany throughoutthe war, life as a Jewish scientistbecame increasingly difficult and,eventually, untenable. In 1933 herenounced his German citizenshipand moved to America. For manyyears the Nazi government bannedeven teaching Einstein’s ideas,along with those of other Jewishscientists.

He was approached to becomePresident of Israel (though hedeclined the offer), and he wrote tothe American President, F.D.Roosevelt, warning him aboutGermany’s plans to build nuclearweapons. He encouraged the USAto develop on their own nuclearresearch programme – his earlierbelief in total pacifism wastempered by his belief that Germanythe aggressor would not hesitate touse their own atomic weapons.

In his later life, Einstein was astrong supporter of the United

Nations and the need for nucleardisarmament. Throughout his lifehe wrote many essays on politicalphilosophy, alongside his physicsresearch.

Einstein died in 1955 at the ageof 76.

In January 2000, TIME magazinenominated Albert Einstein as thePerson of the Century. It’s hard tothink of another individual who hasgained so much attention, who isso well recognised across bordersand generations ... and who, in theend, is so deserving of the honour.

One thing I have learned in along life: that all our science,measured against reality, isprimitive and childlike – and yetit is the most precious thing we have.

78

Telecommunications: looking tothe futureMartijn de Sterke

79

I N T H E P R E V I O U S chapter I reviewed how the capacity of a channel isdefined, and Shannon’s expression for the capacity for a channel withGaussian noise. We saw that the main contribution to the capacitycomes from the bandwidth of the channel, but that the capacity alsodepends more weakly on the signal to noise ratio. I also showed howmodern telecommunications systems make use of wavelength divisionmultiplexing (WDM), in which the total available bandwidth is choppedup in narrow channels, each of which, using separate laser sources anddetectors, carries a signal independent of the other channels. Forreasons that we will discuss shortly, the properties of the fibre causeShannon’s result not apply to this case, and we thus require ageneralization – we’ll tackle this in the first part of this chapter. In thesecond part we discuss how to deal with these effects.

80

WAV

ES O

F TH

E FU

TURE

Linear and nonlinearIn the previous chapter we made a number oftacit assumptions about the optical fibres that weare dealing with; one of the most important ofthese is the assumption of linearity. Tounderstand what this means we need tounderstand something about the physics of howlight interacts with matter. The model peopleoften use to describe this interaction is theLorentz model, developed by the famous Dutchphysicist Hendrik Lorentz who lived from 1853-1928.

Consider the atom as in Figure 1(a). It has anucleus in the centre, and an electron, with someprobability distribution, around it. We now applyan electric field as indicated in the figure, and asa consequence the nucleus and the electron aresubject to a force. However since the nucleus andthe electron are oppositely charged, the forcespoint in opposite directions. Let us now considerthe situation in which the force is small.1 Then,the effect of the electric field is that the nucleusand ‘centre of charge’ of the electron no longercoincide (see Figure 1(b)). Provided the force issmall, the distance d between these two isproportional to the applied field and so d ∝ E.

In this model, the light is considered to be a wavewith an associated electric field (the effect of themagnetic field is weak and is neglected here);thus, for the purposes of our discussion, the lightis modelled as an electric field with a frequencyof roughly 1015 Hz. Under this condition oflinearity (d ∝ E), the properties of a medium thatconsists of a collection of such atoms can be

described by its refractive index and itsabsorption coefficient (though, as we shall see,these quantities generally depend on frequency).Note that if the proportionality constant betweend and E is small, then the refractive index is closeto unity; in contrast, if it is easy to move thepositive and negative charges with respect toeach other, then the refractive index is large. Inorder to prepare us for the nonlinear effects wediscuss below let me point out that one of thehallmarks of linear systems is that no newfrequencies are generated: if the light has afrequency f or a series of frequencies f1, f2, ...,then light with other frequencies is not generatedby the medium.

We now consider applied fields that are stronger,but not so strong that the atom is destroyed (seeFootnote 1). In this case the linear result nolonger applies and can be generalized to d ∝ E +α1 E2 + α2E3, where α1 and α2 are smallconstants; the terms that are nonlinear in theelectric field are then thus small corrections tothe linear relation between d and E. Notice thatthe linear relation d ∝ E has the property thatwhen E flips sign then so does d; it is thus left-right symmetric. The cubic term in the nonlineargeneralization has the same property, but thequadratic term does not; its presence causeselectric fields that point to the left and to right tolead to different values of d. Here we areinterested in materials that are left-rightsymmetric and we therefore drop the quadraticterm; the first correction to linearity is then cubic.

The question we now wish to answer is thefollowing: what is the effect of the additional cubicterm on the optical properties of the material? Theanswer is that, unlike the linear responsediscussed earlier, the nonlinear response doeslead to the generation of new frequencies. Thesenew frequencies can be found as follows: considera light field with frequencies f 1, f 2 and f 3; thenthe new frequencies are those of2

(cos(2πf1t ) + cos(2πf2t ) + cos(2πf3t ))3 (1)

In order to find these frequencies you first needto expand to third power, leading to 10 differentterms, all of the form cos(2πfit ) cos(2πfjt )cos(2πfkt ), where i,j,k can be any of 1,2,3. The

Figure 1: (a) Schematic of an atom, consisting of anelectron, indicated by its probability distribution, around thenucleus. (b) In an applied electric field the electron and thenucleus are subject to forces that point in oppositedirections so that the centres of the positive and negativecharges no longer coincide.

81

TELE

CO

MM

UNIC

ATIO

NS:

LOO

KIN

GTO

THE

FUTU

RE

next step is to use the trigonometric identities3,which show that expression (1) can be written asa sum of terms of the type cos(2π(fi±fj±fk)t).

Frequency MixingLet us look at some of the consequences of thisby first taking i = j = k =1, say. Then we findcos(6πfjt ), i.e., the medium generates a wavewith frequency 3f1 (third-harmonic generation, orTHG). More generally, the nonlinear mediumgenerates the new frequencies |fi ± fj ± fk |. Thisgeneral process is referred to as frequencymixing; since in total there are four frequenciesinvolved, in the optical context it is referred to asfour-wave mixing (FWM). Though we identifiedthe new frequencies that are generated by themedium, we have not discussed the amplitudesof each of these. This is a more difficult problemthat we do not consider here.4

The example of THG is actually not a very goodone. For reasons that are beyond the scope ofour discussion here, THG is very difficult toaccomplish in practice. In practice the threeincoming frequencies f ÿ k are quite similar, andtwo of them may be identical, and the newlygenerated frequency is similar again. Animportant example here is that of two incidentfrequencies f1 and f2, which generate the newfrequencies (2f1 – f2) and (2f2 – f1), leading tofour equally spaced frequencies. Thesefrequencies can be considered to follow fromenergy conservation: the first process can beconsidered be of the type

(a) (f1 + f1) Y (f2 + (2f1 – f2)),

and the second

(b) (f2 + f2) Y (f1 + (2f2 – f1)).

Since each photon has an energy E = hf, where his Plank’s constant, in each of these processesenergy is conserved.

Suppose now that f1 is much stronger than f2 –in terms of electric fields this means that the fieldat f1 is much stronger than that at f2; in terms ofphotons it means that there are many morephotons at f1 than at f2. What can be said about

the probabilities of the processes (a) and (b)? Onemight expect that the probability of one of theprocesses occurring depends on the density ofphotons that act as ‘input’ into the FWM process.Therefore, since there are more photons at f1than at f2, we expect process (a) to be more likelythan process (b) – and indeed this is so. In fact,the probability of the process is proportional tothe square of the intensity, or, using photonsagain, proportional to the square of the number of photons.

Frequency Mixing and WDMNow we know about frequency mixing, what is itsrelation to the WDM that we discussed earlier?The key thing here is to recall that the channelsare equally spaced. Therefore, in a generalizationthe processes (a) and (b) from the previousparagraph, any of the newly generatedfrequencies also corresponds to one of thechannels. Thus, the effect of FWM is expected tobe particularly detrimental in WDM systems. Thedetails of this have been considered by a numberof people, starting with Mitra and Stark. Here,however, we quote a result derived by Stark,Mitra, and Sengupta5: they show that for a WDMsystem, in the presence of FWM, Shannon’sresult from Lecture 1 is generalized to

C = B log(1 + (1-β2)S ). (2)N + β2 S

Note that this reduces to Shannon’s result whenβ = 0. Parameter β corresponds to the rate atwhich energy is transferred between differentchannels; note that in the numerator of thesecond term in the argument of the logarithm,β2S corresponds to signal power taken out of achannel by FWM, whereas in the denominator, itcorresponds to this energy being deposited in theother channels and adding to the noise. Thegeneral expression for β is quite complicated, butin one regime it reduces to β = ψSLe. Here ψ isa coefficient that measures the degree of non-linearity of the fibre and is proportional to α2, andLe is an effective length that corrects the actuallength of the fibre for losses.

82

WAV

ES O

F TH

E FU

TURE

DispersionBelow I describe ways in which one can deal withFWM – as we will see this is rather subtle.Therefore, before doing so, I will quickly discussdispersion, another way in which the signal in afibre deteriorates, and how to fix it. The contrastbetween these two is quite revealing. Tounderstand dispersion we first need to rememberthat the information in a fibre is encoded in shortoptical pulses. The precise number depends onthe type of system, but a rough pulse length is100 picoseconds (ps). Now recall from theprevious lecture that in optical fibres the light isencoded in short light pulses, and that these light pulses consist of a number of differentfrequencies.

Let us now return to the Lorentz model that weconsidered earlier in this lecture: the atomconsists of a nucleus that is bound to electronsthat surround it. Earlier we argued that for smallfields the atom behaves linearly: d ∝ F, where dis the distance between the electrons and F is theforce (which, in turn, is proportional to the appliedfield). Now it so happens that systems in whichthe deviation from equilibrium is proportional tothe applied force are well known: such systemsgive rise to simple harmonic motion (SHM). Agood example of a system that exhibits SHM is amass hanging off a spring (a ‘mass-spring’system), and so let us look at this example for aminute. If we give the mass a small bump then itstarts to oscillate backwards and forwards.Because this oscillation is SHM, it is known that(1) the position of the mass varies sinusoidallywith position at a characteristic frequency (fa,say), and that the frequency depends on thespring’s strength and the mass, but that (2) thefrequency is independent of the amplitude. If, onthe other hand, we impose a sinusoidal motion onthe mass, for example by a small motor, then themass oscillates at this frequency (recall thatlinear systems such as this one cannot generateits own frequencies).

However, the system does have a characteristicfrequency, fa. The effect of this is that if theimposed frequency is close to the characteristicfrequency, then the mass oscillates with largeamplitude. In contrast, for an imposed frequencythat differs significantly from the characteristic

frequency, the amplitude of the mass is small (itis as if the mass really does not want to oscillateat this frequency).

Let us now translate these properties of mass-spring systems to the properties of atoms whenirradiated by light. Just as for a mass-springsystem, the light, with frequency f, imposes itsfrequency onto the atom; in addition, if f is closeto fa then the amplitude of the atomic motion islarge. In contrast, when these two frequenciesdiffer significantly, then the amplitude is small.

We also saw that the linear effect leads to arefractive index of the medium. Now we combinewhat we have learned: since (a) the applied fieldforces a sinusoidal oscillation onto the atom, and(b) this oscillation leads to the refractive indexwhen considering a collection of atoms, then (c) ifthe amplitude of the oscillation depends onfrequency, then so must the refractive index –this is referred to as dispersion.

Let us now combine this with our earlier findingthat a light pulse consists of differentfrequencies. You may remember that the velocityof light in a medium is v=c/n, where c is thevelocity of light in a vacuum. So it follows that ifdifferent frequencies have different refractiveindices in a medium, they must travel at differentvelocities. The consequence of this is thatdispersion has the effect of broadening a pulse,since, upon propagation, the different frequenciesstart to march out of step. Once the pulses startto broaden so that adjacent pulses begin tooverlap, the information would seem to be lost.

Overcoming FWM and DispersionSo we have seen two different physical effectsthat affect pulses when they propagate throughan optical fibre: dispersion, a linear effect thatcauses light pulses to broaden, and FWM, anonlinear effect that causes the quality of anoptical pulse to decrease more generally.6 Ifuntreated, both lead to the loss of information –this may be OK when you talk to friend over thephone, but is unacceptable when your bankstatement information, for example, is carried bythe fibre. Researchers have therefore studiedcountermeasures, and I will discuss some of

83

TELE

CO

MM

UNIC

ATIO

NS:

LOO

KIN

GTO

THE

FUTU

RE

these. We will see that compensation ofdispersion is straightforward, at least in principle,but that the compensation of FWM is muchharder.

Let us first, however, look at the simplestmethods for overcoming dispersion and FWM. Weknow roughly the strength of the dispersion andthe FWM, and we thus know over what distancethey destroy the pulses. Suppose we now includein our fibre a detector at some position beforethis happens. The detector essentially turns thelight into a current; we can then use any methodfrom electronics to fix up the information. We nowlet the current associated with the fixed-up signaldrive a laser (as in Figure 8 of the previouschapter), and a clean pulse once againpropagates through the fibre. This ploy maysound too good to be true, but at some level it isnot: until a few years ago fibre systems workedlike this!

There is a problem though: the optical pulses lastabout 100 ps, and though may sound short, itcorresponds to about 100,000 periods of thelight wave, and so for the light it could beconsidered to be quite long! In contrast, forelectronics, which as we saw in the previouschapter operates at much lower frequencies, 100ps is very short, and it is at these time scales thatelectronics starts to run out of steam. Thoughelectronics as fast enough for the presentgeneration of optical networks, it will not besufficiently to deal with future systems. Thereforeanother approach is needed.7

Our task is to negate the effects of dispersionwithout the use of electronics – we require all-optical solutions. Fortunately, this is easy, inprinciple. In a standard fibre the dispersion issuch that the blue frequencies (shortwavelengths) travel faster than the red ones (longwavelengths). What is needed to undo the effectof dispersion is to let the pulses propagatethrough a fibre with the opposite dispersion: thered frequencies that lag behind the bluefrequencies in standard fibre would be able tocatch up again. The effect is that the pulsesnarrow down again and that the information canbe read out. Though this may sound easy, inpractice it is tricky since it is difficult to negate

the fibre dispersion accurately over the entiretransparency bandwidth shown in Figure 6 of theprevious chapter.8 Note that, as required, thismethod can be applied to very short pulses.

As mentioned earlier, dealing with FWM is muchmore difficult, essentially because it is a nonlinearprocess. The first step in dealing with FWM is tohave some idea how badly the pulses havedeteriorated. This issue of monitoring is one ofthe most important since otherwise anycorrection is a stab in the dark. The aim of themonitoring process can be described as findingout to what degree the energy in the fibre isconcentrated in pulses. This is illustrated inFigure 2, which shows (a) a high-quality signalwith well-defined pulses, and (b) a badlydeteriorated signal with the same average energy.To be able to distinguish between these two(without using electronics!) requires a processthat depends nonlinearly on the intensity.

We earlier encountered such a process thatdepends nonlinearly on intensity: it is FWM!Recall that the process (a) f1 + f1 Y f2 + (2f1 –f2) depends quadratically on the intensity of f1.The monitoring process that we require istherefore straightforward, at least in principle. Westart with the signal frequency fs and add anotherfrequency fp, and we then monitor the intensity oflight at the frequency (2fs – fp), the intensity ofwhich depends quadratically on the intensity of fs.In turn, this tells us whether the pulses in thesignal are narrow (so that they can bedistinguished) or broad (so that distinguishingthem is difficult).

Such an experiment was done by students andcolleagues in CUDOS, here in the School ofPhysics at the University of Sydney9. The crucial

Figure 2: (a) Low-noise pulse train, and (b) a pulse trainwith noise, after propagation through an optical fibre.

84

WAV

ES O

F TH

E FU

TURE

result is shown in Figure 3, showing the intensityof frequency (2fs – fp) versus a control parameterthat determines how narrow the signal pulsesare. At the zero setting the pulses attain theirminimum value, as in Figure 2(a), and theintensity of the frequency (2fs – fp) peaks. Atother settings, the pulses are much wider, as inFigure 2(b), and the intensity at frequency (2fs –fp) is significantly lower.

Regenerating the signalLet us now take this a step further and discusshow the signal can be regenerated, rather thanmonitored. It is well known that this can beachieved through the use of a suitable transferfunction. The transfer function of a device is the

power of the output signal versus the power ofthe input signal. Suppose that we have a devicewith a transfer function as shown in Figure 4(a),which has the property that all input signalsbelow the switching power Ps leads to no output,whereas all input signal above Ps lead to theoutput Pt. Consider now an input consisting of aseries of pulses, possibly of poor quality. Then inthe output all input values below Ps disappear,while the values above Ps get the value Pt.Provided that Ps is sensibly chosen, this canregenerate the input signal, as shownschematically in Figure 4(b).

A practical problem with this approach is that nooptical systems exist that have a transfer functionas in Figure 4(a), and one therefore settles forsystems that have a transfer function as in Figure5(a). Clearly, when the width of the transitionregion vanishes, it turns into the ideal transferfunction in Figure 4(a). A transfer function as inFigure 5(a) can in fact be achieved in FWM, aswas first pointed by Ciaramella and Trillo10. Wesaw earlier that in FWM the transfer function ofthe input signal at frequency fs, to the output at(2fs – fp) is quadratic – this takes care of the low-input part of the transfer function in Figure 5(a).

Fig. 5: Schematic of an approximation to the ideal transferfunction in Figure 4(a). (b) Actual transfer function that is ofthe form in (a). Powers are expressed in dBm, which is adB scale in which 1 mW corresponds to 0 dB.

Fig. 4: (a) Ideal transfer function, leading to no outputunless the input power exceeds Ps, in which case theoutput power is Pt. (b) Operation of the transfer function in(a) to the noisy signal from Figure 2(a).

Figure 3: Intensity at frequency (2fs - fp) (solid line) versus acontrol parameter that determines the widths of the pulses.With the control parameter set to zero, the pulses arenarrow, as in Figure 2(a), and the frequency (2fs - fp) isintense. At other settings, the pulses are wider, as in Figure2(b), and the frequency (2fs - fp) is much weaker. Thedotted line gives the pulse width. The data is given by thedots, with the curve following from theory.

85

TELE

CO

MM

UNIC

ATIO

NS:

LOO

KIN

GTO

THE

FUTU

RE

To get the tailing-off at high-power inputs, recallwhat the FWM actually does: two photons atfrequency fP disappear, generating photons atfrequency fs and at (2fP - fs). Clearly this processmust slow down once a significant fraction of thephotons at fP have disappeared. This is indeed so,and it leads to the desired saturation effect athigh input powers.

Let us now turn to some practical implementationof this idea, which is shown in Figure 5(b)10,which shows a measured transfer function in aFWM experiment. The measurement is similar tothat described earlier, but not completely thesame – however, that does not really matter here.It may not look very much like Figure 5(a), butthat is because both axes are logarithmic (thistends to hide some of the differences in thecurves). Nonetheless, at high input intensities theoutput clearly saturates, whereas at the lowestinput powers, the output increases only slowly.Let us now take this transfer function and apply itto a noisy pulse as in Figure 6(a), which leads tothe output in Figure 6(b). Clearly it has less noiseboth at low and at high intensities, as desired.

[FIGURE 6 HERE]

Eliminating FWMOf course the best way to deal with FWM is toeliminate it completely, and the last topic I wantto discuss deals with just that. We saw earlierthat FWM is a nonlinear process, caused by thenonlinear response of the medium. It isinteresting to point out that the strength of thenonlinear effects in silica glass is much smallerthan in just about any other material; thus, tryingto reduce FWM by using a material other thansilica glass is doomed. The best way to do so, infact, is to eliminate the glass altogether and letthe light propagate in air.

Now we saw in the previous chapter that theessence of a fibre is that the light propagates inthe core, and this cannot be done using just airalone – we need refractive index jumps betweensome outer layer and the core. The solution tothis in the context of optical fibres was providedby Russell in the 1990s11. His solution does noteliminate the glass altogether, but at least lets thelight travel predominantly through air. Figure 7,shows a cross-section of a photonic crystal fibre(PCF), and is thus the equivalent of Figure 5(a) ofthe previous chapter. The hole in the centre ofthese figures is the fibre core, and the region withthe smaller holes in it acts as the cladding.

Notice that, according to our argument in the lastchapter, such fibres could never guide light; totalinternal reflection only works if the light travelsform a high-refractive index to a low-refractiveindex material. In a PCF the core has the lowest

Fig. 7: (a) Electron micrograph of a PCF. (b) Close up aroundthe core region of the fibre in (a).

Fig 6: (a) Noisy optical pulse, and (b), regenerated opticalpulse, using transfer functions as in Figs 5, with less noiseat both high and low power levels.

86

WAV

ES O

F TH

E FU

TURE

refractive index, and total internal reflection thusdoes not occur. Though this argument is perfectlycorrect, PCFs still guide light – to understandhow it works recall that for a fibre to work itneeds some form of ‘mirror’ that prevents thelight from escaping. In a conventional fibre themirror comes from total internal reflection; inPCFs the mirror comes from Bragg reflection,which does not require the core to have thehighest refractive index.

Bragg reflection is a ubiquitous phenomenon thatrelies on periodicity (as in the PCF cladding inFigure 7). It is most easily explained in a one-dimensional periodic structure, as in Figure 8,which shows the refractive index versus positionwith period Λ. Let us assume that the refractiveindex jump ∆n is much smaller than the averagerefractive index . This is not essential, but itmakes the argument slightly simpler. Suppose wehave a wave coming from the left, as indicated bythe arrow. Now each time the light encounters ajump in refractive index, some of the light isreflected. In the case we are considering here thelight reflected at each jump is minute, but thatdoes not matter.

Now consider the light that is reflected off twojumps that are one period apart, and let’s workout under what condition these two contributionsinterfere constructively. To answer this, note thatthese two contributions are the same in everyway, except for one difference: one of themtravels further, back-and-forth, by one period.Using the assumption that ∆n is small, we findthat the two waves have an optical path lengthdiffering by 2n̄Λ. For these two waves to interfereconstructively, they must be in phase, whichmeans that the optical path lengths must differ by

n

the wavelength λ (or, actually, a multiple thereof,but we’ll ignore this). We thus require

λ = 2n̄Λ. (3)

But if the light reflected off two adjacent periodsinterferes constructively, then the light reflectedoff all periods does so. A periodic structure madeto the Bragg condition, equation (3), will thereforereflect light very strongly. Though our argumentapplies to one-dimensional structures, it isperfectly general: a wave propagating throughany type of medium that is periodic undergoesBragg reflection – the condition for this to occurdiffers from (3) in general, but is essentially of thesame form.

Let us now return to PCFs, with their periodiccladding (Figure 7). It is now clear how it canconfine the light to the core. The cladding has tobe designed such that Bragg reflection occurs atthe wavelengths of interest. In fact, this issomewhat trickier than the one-dimensionaltreatment above might suggest – in a PCF yourequire Bragg reflection not just in one direction,but in all directions: if Bragg reflection occurs inall direction but one, then the light can still seepout of the core, and confinement is lost. In someof the most recent work in this area, the claddingconsists almost entirely of air, with some very thinglass membranes making up the fibre (as inFigure 7(b)). In these fibres around 99% of thelight propagates in air, rather than in glass, whichjust about completely eliminates FWM and othernonlinear optical effects. For this and otherreasons, many groups around the world are doingresearch on the properties and fabrication of PCF.

ConclusionThis completes my overview of some of theissues that are coming up in telecommunications,and some of the ways to deal with these. Inclosing I should give a few general comments.The first is that I obviously only have been able toscratch the surface and that many challengesexist that need to be solved (and they are beingsolved!). Following from this, the general area oftelecommunications is not ‘finished’, and muchmore research needs to be done. The overarchingcomment is the issues that we are dealing with

Fig. 8: Schematic of Bragg reflection in a one-dimensionalperiodic medium. The symbols are discussed in the text.

87

TELE

CO

MM

UNIC

ATIO

NS:

LOO

KIN

GTO

THE

FUTU

RE

here belong to general physics and mathematics– telecommunications is brought to you by thebasic sciences.

I am grateful to my colleagues Justin Blows andBen Eggleton for discussions regarding some ofthe issues discussed here, and I thank Trina Ngfor providing me with Figure 3.

Notes

1 This means that the applied electric field should be small

compared to the internal field of the atom. The latter can

be estimated roughly from Coulomb’s law for the hydrogen

atom: E = 1/(4πε0)(q/a2). Here ε0 is the permittivity of

vacuum, q is the elementary charge, and a is the Bohr

radius, the average distance between the nucleus and the

electron in hydrogen. An applied field of this strength would

rip the atom apart. Though this is interesting, we are not

interested in such fields here.

2 In Equation (2) the factors 2πare included in the argument

to make sure that the time T=1/f corresponds to a period

of the wave.

3 The particular trigonometric identity that we require, a

generalization of those that you may have seen before, is

cos(a+b+c) = (cos(a+b+c) + cos(a+b–c) + cos(a–b+c) +

cos(a–b–c))/44 Note that the discussion here is not the entire story. Apart

from the frequency mixing we get other effects which can

be gleaned by taking i = j = k = 1 and the signs such that

fi ± fj ± fk = f1. This is interesting: the nonlinear effect

leads, indirectly, to one of the input frequencies. This

nonlinear contribution has the effect of making the

refractive index depend on the intensity of the light.

5 The references are: P.P. Mitra and J.B. Stark, ‘Nonlinear limits

to the information capacity of optical fibre communications’,

Nature 411, 1027-1030 (2001); J.B. Stark, P. Mitra, and A.

Sengupta, ‘Information capacity of nonlinear wavelength

division multiplexing fiber optic transmission line’, Optical

Fiber Technology 7, 275-288 (2001).

6 Many other things of course also happen when light travels

through a fibre. For example, even though the losses are

small (see Figure 6 of the previous chapter), they do not

vanish. However, as we saw in the last chapter, an

amplifier can be used to boost the power.

7 In addition to this, electronics processing starts to become

very cumbersome – the required equipment becomes very

large, expensive and requires a lot of power.

8 In other strategies the dispersion is compensated in each

WDM channel separately. But in this case the dispersion is

required to be tunable since the fibre dispersion can vary

with time. For example, even though the fibre is buried

underground, varying temperatures during the day lead to

small variations in the dispersion.

9 T.T. Ng, J.L. Blows, J.T. Mok, P. Hu, J.A. Bolger, P. Hambley,

and B.J. Eggleton, ‘Simultaneous residual chromatic

dispersion monitoring and frequency conversion with gain

using a parametric amplifier’, Optics Express 11, 3122-

3127 (2003). In fact, in this work the broadening of the

pulses is due to dispersion, rather than to FWM, however,

this does not matter for us.

10 Some relevant papers include E. Ciaramella and S. Trillo,

‘All-Optical Signal Reshaping via Four-Wave Mixing in

Optical Fibers’, Photonic Technology Letters 12, 849-851

(2000), and S. Radic, C.J. McKinstrie, R.M. Jopson, J.C.

Centanni, and A.R. Chraplyvy, ‘All-optical regeneration in

one- and two-pump parametric amplifiers using highly

nonlinear optical fiber’, Photonic Technology Letters 15,

957-959 (2003).

11 Such a fibre was first demonstrated in R.F. Cregan, B.J.

Mangan, J.C. Knight, T.A. Birks, P.St.J. Russell, P.J. Roberts

and D.C. Allan, ‘Single-mode photonic band gap guidance

of light in air’, Science 285, 1537-1539 (1999).

88

WAV

ES O

F TH

E FU

TURE

Chromosomes 45 and 46 provide aspare set of plans for each other.The fertilized egg uses these plansas a blueprint as it grows itself intoa baby human over 9 months.

If one X-chromosome has adamaged section, the growingembryo in the uterus can use anundamaged section from the otherX-chromosome as it builds itself.

But boys can’t do this. After all,while their chromosome 45 is a bigboofy X-chromosome, theirchromosome 46 is a minuscule Y-chromosome. The Y-chromosomedoesn’t have a matching partner. Ifthe Y-chromosome is damaged, itcan’t read the X-chromosome nextdoor to find an undamaged section.

So the Y-chromosome looks like adisaster waiting to happen. Andindeed, boys suffer quite a fewdiseases (such as red-greencolour-blindness) that girls don’t get.

Just about every cell in your bodyhas the entire DNA needed to makeanother you. Our DNA is lots ofthings at once – it’s a blueprint formaking a human being, it’s ahistory book of our ancestry, it’s agiant medical book, and it’s awhole lot more.

When a cell gets ready to split intotwo, the DNA also prepares to splititself into two. The 2 to 3-metrelength of the DNA bunches itself upinto 46 little packages. We callthem “chromosomes” because asearly as 1848, the geneticistswould colour them with variousdyes, to show up various features – and “chromosome” means“coloured body”.

If you look at these 46chromosomes with a microscope,you’ll see that 45 of them are thesame in boys and in girls, but thatthe 46th chromosome is different.It’s quite large in girls where itcarries about 1,500 genes, but in

boys it’s tiny, and carries only 78genes. So guys, next time you gothinking that you are the Lord ofCreation, just remember that yourpulsating masculinity is madepossible by just 78 genes. In fact,you can think of women as theluxury model with all the options,while guys are the cheap economymodel without the air-con, powersteering and cruise control.

Women have an advantage withtheir chromosomes 45 and 46being big bold matching X-chromosomes.

You can understand this advantageif you think about constructing abuilding from the architect’s plans.Now suppose that the plans havebeen made harder-to-read bysmudges, burger stains and spiltcoffee. It would be very hard tocomplete your building with thesedamaged plans. It would be veryhandy to have a spare undamagedset of plans.

Y-Chromosome: Waste- or Wonder-Land?By Dr Karl Kruszelnicki

IF YOU’VE FOLLOWED the news over the last few years, you’llknow that we humans have just about mapped our own DNA.Now the big difference between Boy- and Girl-DNA is thatone of the chromosomes is different – boys have a Y-chromosome, while girls don’t. For a while the scientiststhought that the tiny Y-chromosome was relentlessly headedfor oblivion, and that it would be gone in a few million years.But the latest research shows that for all its faults, the Y-chromosome (and its nasty byproduct, boys) is probablyhere to stay.

89

Y-CH

ROM

OSOM

E:W

ASTE

- OR

WON

DER-

LAND

?

But the news gets worse for thepathetic little Y-chromosome.

It mutates very rapidly, about 1,000times faster than any otherchromosome. It seems that it wasvirtually identical to the X-chromosome about 300 millionyears ago. In fact, there are a fewshort sections on the very ends ofthe Y-chromosome that are identicalto the very ends of the X-chromosome – but over 95% ofthe Y-chromosome is very differentfrom the X-chromosome. It hasmutated massively. Over the last300 million years, the once-proud Y-chromosome has shrunk fromabout 1,500 genes to its currentmiserly 78 genes. And if it keepsshrinking and mutating at thepresent rate, it’ll be totally use-lessin just a few million years.

Yup, for a while there, the scientistsreally thought that the Y-chromosome was a regularwasteland.

But in June 2003, Nature publishedsome reassuring research by Dr.Page from the MassachusettsInstitute of Technology, whichturned the Y-chromosome fromWasteLand to WonderLand. He andhis team worked out, after a lot ofvery hard work, that the Y-chromosome has a special trick,which means it doesn’t need amatching partner to cover for anydamaged sections (or bits).

They discovered that large sectionsof the Y-chromosome are“palindromes”. You mightremember from your high-schoolEnglish days that a “palindrome” isa word, or phrase, that reads thesame in each direction. The word“radar” is a palindrome, as is theword “level”.

It is very exciting that sections ofthe Y-chromosome arepalindromes. It means that if onesection of the Y-chromosome isdamaged, sometimes it can find anundamaged version of the samesection somewhere else in the Y-chromosome. It can ignore thedamaged section, and use theundamaged section somewhere upat the other end of the Y-chromosome.

But while this is good news, it isalso a little disturbing. Yes, the Y-chromosome can fix itself up – but only by having sex with itself... FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

Illustration courtesy Adam Yazxhi

LIDIA MORAWSKA is aProfessor at the Schoolof Physical and ChemicalSciences, QueenslandUniversity of Technology(QUT) in Brisbane,Australia, and theDirector of theInternational Laboratoryfor Air Quality and Healthat QUT (ILAQH, which isa WHO CollaboratingCentre of the WorldHealth Organization on

Research and Training in the field ofGlobal Burden of Disease due to AirPollution). She conducts fundamentaland applied research in the interdis-ciplinary field of air quality and itsimpact on human health and theenvironment, with a specific focus onscience of airborne particulate mat-ter. Professor Morawska is a physi-cist and received her doctorate at theJagiellonian University, Krakow,Poland for research on radon and itsprogeny. Prior to joining QUT shespent several years in Canada con-ducting research first at McMasterUniversity in Hamilton as a Fellow ofthe International Atomic EnergyAgency, and later at the University ofToronto. Dr Morawska is an author ofover hundred fifty journal papers,book chapters and conferencepapers. She has also been involvedat the executive level with a numberof relevant national and internationalprofessional bodies and has beenacting as an adviser to the WorldHealth Organization. She is theimmediate Past President of theInternational Society of the Indoor AirQuality and Climate.

90

The Science of the Aerosolswe breatheProfessor Lidia Morawska

IntroductionW A I T I N G F O R A C H A N G E of traffic lights at a busy intersection in any largecity we can smell vehicle exhaust; standing on a hill above the city wecan often see brown haze blanketing the urban site at our feet;travelling on a plane we pass through pollution plumes from an urbanmetropolis, industrial chimneys or forest fires. In situations like these wemay contemplate whether air pollution is a new phenomena of theindustrialised world, whether nations are gaining control over theproblem of the pollution – or whether in fact it is getting worse.

91

92

WAV

ES O

F TH

E FU

TURE

A irborne pollution was reported long ago asa nuisance problem in ancient Romancities, and has continued through the

development of modern civilization ever since. Yetthe impact of pollution, nor the necessity tocombat it, was not realised until recent times; it is only in the last fifty years that it has becameobvious that pollution introduced to the air byhumans has a significant negative impact onhuman health and the environment. Since then,the subject has became the focus of anincreasing number of scientific studies conductedworldwide, and with new knowledge emerging ithas also became evident how complex, and howscientifically and technically challenging thisinterdisciplinary field is.

Over the last decade, one specific aspect ofairborne pollution has been increasinglypreoccupying scientists as well as politicians: thestudy of airborne particles, particulate matter, andmore specifically very small particles. Thephysics, chemistry and biology of these particles– which can vary in size by up to a hundredthousand times between the smallest and thelargest and are involved in multitudes of reactionsand interactions – are not only fascinating topicsfor scientific studies, but ones with importantpractical implications. This chapter aims tointroduce the scientific world of airborne particlesand explain how such particles affect humanheath and the environment.

The Complex World of AtmosphericAerosol ParticlesIn scientific terminology, airborne particles arecalled aerosols – these are defined as anassembly of liquid or solid particles suspended ina gaseous medium long enough to enableobservation or measurement. Airborne particlesrange in diameter from about 0.001 µm, to about100 µm, with the lower limit describing the sizeof large molecules and the upper limit the size ofparticles that deposit very quickly due togravitational force.

Particles can be characterized by their physicalproperties, such as size, shape or sizedistribution, but also by their chemicalcomposition or biological nature. For example

viruses, bacteria or pollens, when in the air,belong to the world of airborne particles. Thescientific complexity of the world of atmosphericaerosol particles is depicted by the many differentterms that are used to describe or identify them.Some of the terms identify particles by theirsizes, others by the processes that led to theirgeneration and some by a particle’s ability toenter human respiratory tract.

Considering even the basic parameter of particlesize, there are various classifications andterminologies used to define particle size rangesfor the purpose of scientific discussion orpractical applications. The division mostcommonly used is between fine and coarseparticles; however, the boundary between thesetwo ranges is somewhat arbitrary and has beendefined differently by various authors in relationto different aerosols and applications. The divisionline often used in aerosol science and technologyis somewhere between 1 and 2 µm. This is therange of the natural division between smallerparticles, which are generated mainly fromcombustion and other process leading to gas toparticle conversion, and larger particles generatedfrom mechanical processes. Obviously any suchdivision is somewhat arbitrary, as nature itselfdoes not provide a perfect division.

The terminology that has been used in thewording of the ambient air quality standards, andalso in characterization of indoor and outdoorparticle mass concentrations includes PM2.5 andPM10 fractions. PM2.5 (or fine particles) is themass concentration of particles with aerodynamicdiameters smaller than 2.5 µm, while PM10 refersto the mass concentration of particles withaerodynamic diameters smaller than 10 µm(more precisely, the definitions specify the inletcut-offs for which 50% efficiency is obtained forthese sizes). Total suspended particulate (TSP) isthe mass concentration of all particles suspendedin the air. Other size classifications of particlesmay be into submicrometer and supermicrometerparticles, which are particles smaller than andlarger than 1 µm, respectively, and ultrafineparticles, which are smaller than 0.1 µm

93

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

Where do Aerosols come from?While the above classification of particlesconsiders only their sizes, usually a particle’s sizeis a consequence of the process that led to itsgeneration. Particles in the submicrometer range(smaller than 1 µm) are generated mainly bycombustion processes such as burning timber,smoking a cigarette or combusting fossil fuel invehicle engines, and by processes which result ingas or vapour molecules combining to form anew particle (gas-to-particle conversion,nucleation processes or photochemicalprocesses). Particles in this size range typicallycontain a mixture of components including soot,acid condensates, sulfates and nitrates, as wellas trace metals and other toxins. Larger particles,in the supermicrometer range, result mainly frommechanical processes such as cutting, grinding,breaking and wear of material and dustresuspension, and contain largely earth crustalelements and compounds.

From the above it can be concluded that some ofthe particles in the air originate from direct airemissions, while others are formed in theatmosphere by the chemical reactions of gases,particularly sulphur dioxide, nitrogen oxides,ammonia and volatile organic compounds. Theformer are called primary, and the latter aresecondary particles.

Particle sources are also indicated in thecommonly used descriptive terms for airborneparticles, and in particular:

■ Dust – solid particles formed by crushing orother mechanical breakage of a parentmaterial, larger than about 0.5 µm (see belowfor definition of house dust)

■ Fog – liquid particle aerosol formed bycondensation or atomization

■ Fume – particles that are usually the result ofvapour condensation in low-temperaturecombustion processes, with subsequentagglomeration, usually smaller than 0.05 µm

■ Smog – an aerosol consisting of solid andliquid particles created, at least in part, by theaction of sunlight on vapour triggeringphotochemical reactions and generatingsecondary pollutants

■ Smoke – a solid or liquid aerosol, the result ofincomplete combustion or condensation ofsupersaturated vapour; usually considered theprecursor of smog

The above classifications are an indication of themany different processes leading to generation ofparticles with different physical and chemicalproperties. It is interesting to note that while inscientific language these terms clearly identify themechanisms by which the particles weregenerated, in common everyday language theyare used loosely to identify the presence ofparticle matter in the air without, however,attaching to them their scientific meaning.

Physical Properties of AerosolsThe most important physical properties of aerosolparticles include: number and number sizedistribution, mass and mass size distribution,surface area, shape, and electrical charge. To alarge extent these are the physical properties ofparticles that determine particle behaviour in theair and, ultimately, removal from the atmosphericsystems. The efficiency of various forces actingon particles and processes to which they aresubjected in the air depends strongly on theparticles’ physical properties, of which size is oneof the most important.

Health and environmental effects of particles arestrongly linked to particle size, as it is the sizethat is a predictor of the region in the lung wherethe particles would deposit, or the outdoor andindoor locations to which the particles canpenetrate or be transported. Also sampling ofparticles and choice of an appropriateinstrumentation and methodology is primarilybased on a particle’s physical properties. Next we’lldiscuss aerosol shape and size in more detail.

Aerosol Shape and Equivalent DiameterParticles vary significantly in shape, with some ofthe shapes fairly simple and regular; however themajority display a varying degree of irregularity orcomplexity. In general, shape relates to aparticle’s formation process or its origin. Particlesresulting from coagulation and agglomeration ofsmaller solid particles are usually highly irregular.Examples of these are particles resulting from

94

WAV

ES O

F TH

E FU

TURE

combustion process such as vehicle emissions(which are agglomerates of carbonaceousspherical particles), or tobacco smoke. Similarly,dust particles and particles resulting frommechanical processes typically have irregularshapes. In contrast, liquid aerosol particles areusually spherical, while simple fibres are rodshaped. Biological particles are complex in shapeand differ significantly between various types. InFigure 1, microscopic images of different types ofparticles are presented.

Particles of complex or irregular shape can becharacterized by many parameters, but forpractical applications these are usually reducedto one or two parameters that can be measured.Most commonly these are particle diameter, or forfibres, their length and width. Diameter is acharacteristic of spherical objects; however, asexplained above only a small fraction of airborneparticles are spherical. Therefore a way ofrepresenting particles of irregular shape has beenintroduced by means of a particle’s equivalentdiameter.

This process is schematically presented in Figure2. Particle equivalent diameter is the diameter ofa sphere having the same value of a physicalproperty as the irregularly or complex shapedparticle being measured. Equivalent diameterrelates to particle behaviour (such as inertia,electrical or magnetic mobility, light scattering,radioactivity or Brownian motion) or to particleproperties (such as chemical or elementalconcentration, cross-sectional area or volume tosurface ratio). Therefore the particle diameterdetermined experimentally depends on the choiceof particle properties or behaviour measured.

Application of different methods for measurementof particle diameter usually results in somewhatdifferent values of the diameter obtained.Therefore understanding of which particlediameter is actually determined in a particularstudy is important for interpretation of the resultsand for comparison between different studies.The most commonly used equivalent diameter isthe aerodynamic diameter, the diameter of a unit-density sphere having the same gravitationalsettling velocity as the particle being measured.

Aerosol Size and Size DistributionAs explained above, particles in ambient air aremixtures generated by a large number of sourcesincluding motor vehicles, power plants,windblown dust, photochemical processes,cigarette smoking, nearby quarry operation, and

Figure 1: Particles collected in residential houses inBrisbane, Australia, and examined with an energy-dispersive X-ray analyser attached to a transmissionelectron microscope. (a) The three types of particles in thispicture are: a square particle (NaCl crystal), many bigfibrous particles (dominant elements: Mg, Cl, S, O), andmany fine fibrous particles (dominant elements: Ca, S). (b)Particles of the same type with the dominant elements: Mg,Cl, S, K, Na. (c) Small NaCl particles (crystals) and a longparticle with dominant elements: C, O, Ca, Mg, S, P(possibly a fragment of an insect or a plant). (c) Dry drop.The dominant elements of the big particles are Cl, K, Naand of the small ones Mg, K, Cl, S, O, Na, Si. (Images fromMorawska and Salthammer, 2003.)

Figure 2: Representation of particles of irregular shape bymeans of particle equivalent particle diameter.

95

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

so on – and particle size is dependant on particleformation processes. For example the diameter ofparticles produced by various indoor andcombustion sources is very small and typicallymay lie in the ranges (according to source): forgas burners, gas ovens, and toasters in the orderof 0.01 to 0.02 µm; for frying and grilling in therange from 0.05 to 0.2 µm; natural gas, propane,and candle flames generate particles between0.01 and 0.1 µm; for cigarette smoking in therange from 0.01 to 0.1 µm; and incense burningabout 0.1 µm.

Particles generated by outdoor combustionsources are also generally small. A significantproportion of diesel emission particles havediameters smaller than 0.1 µm. Gasoline particlesare mostly carbonaceous sphericalsubmicrometer agglomerates ranging from 10 to80 nm. Particles from CNG emissions are smallerthan those from diesel or even petrol emissionsand range from 0.01-0.7 µm, with the majoritybeing between 0.020 and 0.060 µm. Themajority of particles emitted from biomassburning, which includes controlled burning anduncontrolled fires (as well as burning for thepurpose of heating and cooking) are ultra fine,with only a small fraction in the larger size rangeand with most of the mass present in particlesless than 2.5 (m in aerodynamic diameter.

In contrast, particles generated by mechanicalprocesses are larger. For example in indoorenvironments, walking, moving one’s arms, andeven sitting in front of a computer produces

particles in the range of 5 to 10 µm. Outdoors,processes such as wear on vehicle tires, roaddust resuspension or coal processing result inairborne particles typically in the size range ofseveral micrometers. Figure 3 presents the largevariation in the sizes of airborne particles.

Almost all sources generate particles with somedistribution of particle sizes (so calledpolydisperse aerosol) rather than particles of asingle size (monodisperse aerosol). A veryinteresting phenomenon related to thisdistribution of sizes is that particles generated bymost sources have a log-normal size distribution– this means that the curve of particleconcentration versus particle size is ‘normal’ (bellshaped) when the particles are plotted on alogarithmic scale. When a single pollution sourceis investigated and when it operates under steadyconditions (for example steady parameters of thecombustion process), the size distribution ofparticles obtained is likely to have one distinctivepeak, and sometimes additional, usually muchsmaller peaks. These peaks are called modes ofthe distribution.

Different emission sources are characterised bydifferent size distributions, and while thesedistributions are not unique to these particlesources alone, the information on the sizedistribution can help to identify their contributionto particle concentrations in ambient air, and alsoserve as a source signature.

Figure 4 presents examples of size distributionsof particles generated by different combustionprocesses. Figure 5a presents an example of atypical, urban, air particle number size distributionmeasured in Brisbane, Australia.

Particle distributions can be presented either interms of number or mass distributions. In termsof number, the vast majority of airborne particlesare in the ultrafine range. For example, in urbanoutdoor air where motor vehicle emissions are adominant pollution source, over 80% ofparticulate matter in terms of number is in theultrafine size range. The total mass of theultrafine particles is, however, often insignificantin comparison with the mass of a small numberof large particles, with which most of the mass ofFigure 3: Large variation in the sizes of airborne particles.

96

WAV

ES O

F TH

E FU

TURE

Figure 5: Typical measured urban air particle number size distribution (a), and mass distribution (b) calculated from the number distribution.

Figure 4: Examples of particle size distribution spectra of environmental tobacco smoking (ETS), coal, petrol and, diesel.

97

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

airborne particles is associated. Particle surfacearea in turn is largest for particles somewhatabove the ultrafine size range. This relationshipbetween particle number and volume ispresented in Figure 5a and b – the relationshipwas derived using measured particle number sizedistribution (Figure 5a) and calculating particlevolume distribution assuming their sphericity(Figure 5b). Particle mass can be calculated fromthe volume when particle density is known or canbe assumed. It can be seen from Figures 5a andb that the peak in the number distributionspectrum appears in the area where there isalmost no volume in the volume distributionspectrum, and vice versa, the peak in the volumedistribution spectrum is where the particlenumber is very low.

The form of presentation of particle sizedistribution used in Figures 4 and 5 is quitecommon, but it is somewhat simplistic, as it doesnot properly reflect the logarithmic nature of thedistribution. The proper and most common way ofpresenting particle distributions is by plotting (inlogarithmic scale) particle number, surface areaand volume respectively, per logarithmic intervalof size.

Chemical and Biological Properties ofAerosolsThe chemical composition of particles is multi-factorial and depends, as discussed already, onparticle sources as well as post-formationprocesses. For example, some types of particleslike asbestos and glass fibres consist of inorganicmaterials, while other types like cellulose fibresare purely organic. In many cases, the behaviour

of organic and inorganic compounds associatedwith the particle ‘body’ (adsorption/desorption,water solubility, extractability) is of interest. Themost important chemical properties of particlesinclude:

■ Elemental composition■ Inorganic ions■ Carbonaceous compounds (organic and

elemental carbon)

In general, interest in the elemental compositionderives from the potential health effects of heavyelements like lead, arsenic, mercury andcadmium, and the possibility of using theelements as source tracers. Water-soluble ionssuch as potassium, sodium, calcium, phosphates,sulfates, ammonium and nitrates associatethemselves with liquid water in indoorenvironments and can also be used for sourceapportionment.

Carbonaceous compounds are composed oforganic and elemental carbon. The former cancontain a wide range of compounds such aspolycyclic aromatic hydrocarbons, pesticides,phthalates, flame retardants and carboxylic acids,some of which are tracers for certain sources,while the latter is sometimes termed ‘soot’, ‘blackcarbon’ and ‘graphitic carbon’.

Another important aspect of airborne particles istheir biological properties. Individual bacteria orpollens suspended in the air are examples ofbiological particles. But bacteria or pollen sporescan be attached to other, non-viable particles,such as smoke or dust, which become carriers ofbiological particles.

Outdoor sources

Surfaces of living and dead plants (fungal spores,

bacteria), soil

Natural and anthropogenic waters such as sewage lagoons or

cooling towers (bacteria)

Aerosolation of water

Building exhaust and sanitary vents

Indoor sources

Occupational environment where organic materials

are handled

Agriculture (processing of products)

Microbial growth in buildings (heating, ventilation and air

conditioning systems, building structure)

Ornamental fountains, showers

Humans

Pets

Hospital procedures

Indoor plants

Table 1: Sources of biological particles

98

WAV

ES O

F TH

E FU

TURE

There are different classes of biological particles,which include many different species, varying notonly in terms of their biological composition butalso size. In general, viruses range from 0.02 to0.3 µm, bacteria from 0.5 to 10 µm, fungi from0.5 to 30 µm, pollen 10 to 100 µm and housedust mites are about 10 µm. Biological particlescan originate from a number of indoor or outdoorsources that can be broadly classified aspresented in Table 1.

Aerosol DynamicsFollowing formation or resuspension, airborneparticles undergo a range of physical andchemical processes that change their chemicalcomposition, physical characteristics andconcentration in the air. The most important ofthese processes include: coagulation, whichresults from Brownian motion and collision ofparticles, mainly of similar sizes; deposition ofsmaller particles on the surface of biggerparticles; changes to particle size due to changesin moisture content (the latter could behygroscopic growth or shrinking by evaporation);sedimentation and deposition on surfaces. Thecomplexity of assessing the role of individualprocesses or quantifying their rates relates to thefact that all of them take place simultaneously,affect differently particles of different size rangesand are dependent on a large number of factorsand characteristics of indoor environments.

Some emission products, for example those thatare combustion related, are highly dynamicmixtures of hot gases and particles undergoingrapid changes; others, like mechanical dust, areless so. Particles measured away from theemission site, or particles generated indoors andmeasured some time after emission, will havedifferent characteristics to those measuredimmediately after formation.

The residence time of emission products in theair depends on the nature of the processes theyare involved in, and varies in outdoor air fromseconds or minutes to days or weeks. Largerparticles (of a micrometer size range inaerodynamic diameter and more) are removedfrom the atmosphere mainly through gravitationalsettling (with particles above 100 µm settling

almost immediately after becoming airborne),while smaller particles are removed byprecipitation or diffusional deposition.

Aerosol Detection MethodsMeasurement is a means of detecting or, in otherwords, seeing the airborne particles and learningabout their composition. Depending on whichmeasuring technique is used, the particles areseen somewhat differently. For example, asexplained above, an irregular particle can bedescribed by a number of diameters, calledequivalent diameters, which relate to the physicalmethod used for approximating the irregularshape as a regular, spherical object. There aremany different measurement and samplingtechniques for airborne particles andunderstanding the principles of operation of thevarious techniques is of importance not only tothose who design and conduct experimentalstudies, but also to anyone involved withanalysing and interpreting the data generatedthrough the experiments.

In general, the experimental techniques formeasurements of particle characteristics arecomplicated and expensive. Such measurementsshould cover a broad range of particle sizes fromnanometers up to at least several micrometers.However, due to the different physical propertiesof particles in this size range, various methodsand different instruments have to be applied forcomprehensive measurements of particlecharacteristics. Investigations of the particlenumber size spectrum yield information about therelationship between particle number anddiameter. This information is usually lost incommonly applied measurement methodstargeted at the determination of the TSP (TotalSuspended Particles), PM10 or even PM2.5

fractions, especially for submicrometer particles.As explained above, the larger particles contributestrongly to the total mass of the air particles, butoften the number of such particles is severalorders of magnitude lower than the number ofsmall particles.

Measurements of particles can be conducted byactive or passive sampling followed by applicationof appropriate analytical techniques for analyses

99

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

of the material collected through sampling.Passive sampling means that the particlesdeposit on the sampling medium throughgravitation, diffusion or other natural processes.By contrast, active sampling means application ofpumps or other devices for drawing a certainvolume of air through the sampling medium.Passive sampling is less commonly usednowadays for particles, as its outcomes areassociated with large uncertainties.

Measurements of airborne particles can also beperformed using in-situ methods where thesample is temporarily captured by the instrument,which measures some particle characteristics inreal time. In this case there are no furtheranalytical techniques required. Manycharacteristics of particles can be measured inreal time, including particle mass or numberconcentration, particle number size distributionand surface area. Some of these methods providea direct measure of the property investigated;others measure a parameter, which is thentranslated into the property of interest. Forexample, optical counters directly count particlescrossing the sensing area of the instrument, whilemicrobalances measure changes in the oscillatingfrequency of a crystal on which particles aresampled, and translate the change of thefrequency into the mass collected. Not all particleproperties can be measured in real time, and ingeneral, while there are readily available methodsfor real time measurements of many physicalproperties of the particles, there are only very fewsuch methods for measurements of chemical orbiological properties.

However, there are new methods being developedand new instruments becoming available, so it islikely that in the future it will be possible tomeasure more particle properties in real time. Forexample, the time-of-flight spectrometer enablesmeasurements of the size and chemicalcomposition of individual aerosol particles in nearreal time, while an ultraviolet aerodynamicparticle sizer (UV-APS) measures the fluorescencecharacteristics of individual particles in an aerosolsample, which makes possible real timeidentification of biological aerosol particles, asdistinct from inanimate particles. As an example,Figure 6 presents a diagram of the size

distribution of viable and non-viable particlesmeasured in a piggery. It took only two minutes toconduct this measurement and obtain this graph,which is significantly faster than using classicalmicrobiological techniques.

[FIGURE 6 HERE]

Aerosols in Outdoor AirParticle concentration levels in environmentswhich are not influenced by human activities(clean environments) are usually of the order of afew hundreds particles per cm3. In urbanenvironments, background particle numberconcentrations range from a few thousands toabout twenty thousand particles per cm3.Background concentrations refer to theconcentrations measured at monitoring stations,which are not influenced by a nearby emissionsource. Near roads and in tunnels, vehiculartraffic constitutes the most significant urbansource of particle numbers. Here particle numberconcentrations can be more than ten timeshigher than in other urban environments and canreach or exceed levels of 105 particles per cm3.This is in contrast to PM10 and PM2.5 massconcentrations, which at such roads have beenshown to be no more than 25 to 30% abovebackground levels (calculated as the differencebetween the maximum at the road and thebackground levels).

When considering emissions of particles into theair, the formation of new particles in that air, theresulting airborne particle concentration, and

Figure 6: Size distribution of viable and non-viable particlesmeasured in a piggery.

100

WAV

ES O

F TH

E FU

TURE

ultimately the impact that these particles have onhuman health and the environment, differentspatial scales have to be examined. In particular,it is important to distinguish between local airpollution in magnitude and impact, and global airpollution and its effects. Global air pollutionimpacts are related to the total emissions on acontinental or worldwide scale and include thegreenhouse effect, effects on climate and ozone depletion.

The effects of local air pollution are localenvironmental problems (for example visibilityreduction), but most important are the impacts onhuman health. Figure 7 presents a situation ofdense brown smog over a central European citylocated in a valley surrounded by mountainswhere, in addition to vehicle emissions, coal andwood burning for heating purposes during winterresult in significant local air pollution. The globalscale of pollution resulting from combustionprocesses (vehicles, domestic heating andcooking, burning of fossil fuels for industrial andenergy generation processes) can be appreciatedwhen flying over continents and seeing largeareas under similar brown smog, extending to theareas beyond urban metropolises.

While it is evident there are many significantsources of ambient airborne particles, vehicleemissions in particular will be discussed here inmore detail to illustrate the differences between

local and global impacts of emissions. In mosturban environments, motor vehicle emissions arethe main anthropogenic source of air pollutionand specifically particles, significantly contributingto the deterioration of urban air quality. Sincepopulation density is higher in urban than in ruralareas, the number of people exposed world wideto elevated levels of vehicle emissions isenormous. The impact of vehicle emissions onambient urban concentrations of particles andgaseous pollutants could be considered in termsof two spatial scales:

■ Large scale, which is the total urban airshed, and which means contribution ofvehicles to the background urbanconcentration levels of pollutants, and

■ Small scale, which is the close proximity toa road, where the concentrations areelevated above the urban backgroundlevels

Estimates of the total vehicle emission levels,which include the vehicle emission inventory andthe total contribution of vehicles to the urbanbackground concentration of pollutants, arederived using transport and traffic models withthe relevant vehicles’ emission factors. The smallspatial scale of vehicle emission impacts isconcerned with the areas adjacent to the roadscarrying considerable amounts of traffic, or roadintersections and other traffic congestion areas.In order to assess the impact of the emissionsfrom a nearby road on air quality in theneighbouring buildings, the following aspectsneed to be taken into account:

■ Total vehicle flow on the road and thespeciation of the flow into individual vehicleclasses in terms of vehicle size and the fuelon which they operate

■ Emission factors of individual classes ofvehicles

■ Variation of vehicle flow with time of theday

It has been shown that while spatial distributionof pollutants in the urban environment appears tobe highly homogenous based on the data from airquality monitoring stations, closer analyses ofpollutants’ dispersion and transport reveals thatfor gaseous and particle number concentrationsthere is a high level of heterogeneity displayed,

Figure 7: Dense brown smog over a central European citycaused by vehicle emissions, and coal and wood burningfor heating purposes during winter.

101

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

with significantly elevated concentrations in theimmediate vicinity of roads. Particle numberconcentration, like the concentration of gaseouspollutants and other surrogates for very smallparticles, decreases significantly with the distancefrom a road.

Decay in particle concentration was approximatedby exponential curves in a number of studies andit was shown that the impact of a road on particlenumber concentration, while significant in theimmediate vicinity of the road, is notdistinguishable past about 300 metres from theroad. It was shown that dispersion is the mainfactor responsible for the decrease in particleconcentration with distance from the road.

A practical implication from these findings is thatthe exposure to gaseous pollutants and numberconcentration of particles emitted by vehicletraffic on roads is significantly increased withinthe distance of the first 100 to 200 meters fromthe road, compared to the urban averageexposure levels, and reduces to the urbanbackground level at distances larger than about300 to 400 metres from the road. On this basis,it is reasonable to assume that people living andworking in close proximity to an urban arterialroad will likely be exposed to levels of ultrafineand submicrometer particles beyond what couldbe considered ‘normal’ ambient levels.

This is a finding that needs to be taken intoconsideration in future urban land and transportplanning. The situation is somewhat different inrelation to particle mass, which is not so stronglyelevated close to roads compared to thebackground level. The reason for this is thatnewer vehicle technologies result in loweremissions of particle mass and in addition, thereis little dust generated from modern, sealedroads.

Aerosol Inhalation and Deposition in theLungThe main mechanism for intake of airborneparticles by the human body is inhalation ofparticles and deposition in the respiratory tract.Factors influencing the deposition of inhaledparticles can be classified into three main groups:

1) The physico-chemistry of aerosols2) The anatomy of the respiratory tract 3) The airflow patterns in the lung airways

In relation to the physico-chemistry of aerosols,the forces acting on a particle and its physicaland chemical properties such as size or sizedistribution, density, shape, hygroscopic orhydrophobic character and chemical reactions ofthe particle will affect its deposition. With respectto the anatomy of the respiratory tract, importantparameters are the diameters, the lengths, andthe branching angles of airway segments, whichdetermine deposition. Physiological factorsinclude airflow and breathing patterns that furtherinfluence particle deposition.

Particle penetration and deposition in therespiratory tract is, to a large extent, governed byparticle size. Large-size particles deposit mainlyin the upper part of the respiratory tract due toimpaction, interception, gravitationalsedimentation, as well as turbulent dispersion.Very small particles that can follow the gasstreamlines, such as those generated throughcombustion, can penetrate to deeper parts of therespiratory tract processes, and have a highprobability of deposition there due to their highdiffusivities.

The understanding of the mechanisms of particledeposition in the human respiratory tract and theability to quantify the deposition in individual partsof the respiratory tract is of principal importancefor assessment of human intake of particles dueto inhalation, which can then be used for riskassessment. Over the last three decades or so, alarge number of studies have been conducted toinvestigate particle deposition in the humanrespiratory tract, with a somewhat larger numberfocused on theoretical modelling than on theexperimental determination of the deposition.Figure 8 presents the results of recentexperimental studies conducted at theInternational Laboratory of Air Quality and Health,Queensland University of Technology, Brisbane,showing the dependence of deposition of dieselparticles in the human respiratory tract onparticle size, and also individual differencesbetween people.

102

WAV

ES O

F TH

E FU

TURE

Health Effects due to Exposure to Aerosols Inhalation of airborne particles has been shownto be detrimental to human heath due to theirtoxic or carcinogenic properties, but alsobecause they act as irritantscausing discomfort and affectinggeneral human well being. Alarge number of epidemiologicalstudies conducted in cities indifferent parts of the world havelinked daily mortality statisticswith increased particleconcentrations measuredoutdoors (epidemiology is thescience concerned with thestudy of disease in a generalpopulation, and determination ofthe incidence or rate ofoccurrence and distribution of aparticular disease by age, sex,or occupation, which may provide informationabout the cause of the disease). An increase of 1to 8% in deaths per 50 (g µm-3 increases inoutdoor air particle mass concentrations hasbeen a common conclusion from these studies.

There is still, however, only limited informationavailable in terms of quantitative links between

exposure to airborne particulate matter,particularly of small sizes and at lowerconcentrations, and the health effects theycause. Both fine and ultrafine particles appear toaffect health outcomes such as mortality, and

respiratory and cardiovascularmorbidity, and appear to do soindependently of each other.However, the database atpresent is too limited (both innumbers of studies andnumbers of subjects) andgeographically restricted toallow clear conclusions on themode of action or generalizationto other settings. All of thestudies demonstrate that theprimary determinant of theeffect of ultrafine particles istheir number and their surfacearea and not the weight of

particles present. This means that the traditionaluse of particle mass weight measures isinappropriate in evaluation of the likely biologicaleffects of ultrafine particulates.

The mechanisms by which the particles inducethe range of health effects are still hypothesised,but not fully proven. The size of airborne particles

Figure 8: Size dependent deposition of diesel smoke particles in the lung of fourteen volunteers for the study.

A large number ofepidemiological studiesconducted in cities indifferent parts of theworld have linked dailymortality statistics withincreased particleconcentrations ...

103

THE

SCIE

NC

E O

F TH

E AE

ROSO

LS W

E BR

EATH

E

determines in which parts of the respiratory tractthe particles are deposited: small airborneparticles less than one micrometer in diameter(submicrometer particles) have a high probabilityof deposition deeper in the respiratory tract andare likely to trigger or exacerbate respiratorydisease. Small particles also have higher burdensof toxins which, when absorbed in the body, canresult in health consequences other thanrespiratory health effects.

Effects on Climate ChangeGlobal warming has been linked to theintroduction into the air of substances that resultin trapping heat in the atmosphere, the so-calledgreenhouse effect. Different substances havedifferent effects on global warming, expressed interms of global warming potential. For example,over a 100-year time frame, nitrous oxide is 310times more effective than carbon dioxide attrapping heat in the atmosphere.

There is growing recognition that the very smallparticles also contribute to the greenhouseeffect, but the degree of contribution has not yetbeen established. There are many challenges inquantification of particle contribution to thegreenhouse effect. For example, calculation ofthe potential of atmospheric aerosols resultingfrom motor vehicle emissions for causing earth’sclimatic change through radiative forcing cannotbe done in the same way as for greenhousegases. This forcing is comparable, but ofopposite sign to, the radiative forcing due togreenhouses gases (IPCC, 1995). However,unlike greenhouse gases, the atmosphericaerosol is not uniformly distributed about the globe.

In summary, as stressed in the report of theScientific Committee on Problems of theEnvironment 2000:

Measurements of tropospheric andstratospheric aerosol particles, the vastmajority of those of anthropogenic origincoming from combustion sources, is thus ofcritical importance for developing anunderstanding of climatic effects of particlesand establishing effective ways for minimisingthe effects.

Future Directions in Aerosol ResearchIn summary, despite the significant increase innational and international efforts towardsextending our understanding of various aspectsrelated to airborne particles, there is still onlylimited understanding of the production rate,airborne concentrations and composition indifferent environments, the theory of particledynamics, the fate of very small particles in theair, on the exposure-response relationship and onthe mechanisms by which particles induce arange of health effects.

One important reason for the current difficultiesrelates to the scientific complexity ofinvestigations on airborne particulate matter at alllevels of approaches, including: instrumentation,measurement and modelling, model validation,data interpretation, exposure assessment and riskquantification. All these aspects requireconsiderably more scientific knowledge andunderstanding, and will constitute the futuredirections for research in this broad,interdisciplinary field.

References and Further Reading

Baron, P.A and Willeke, K., 2001, Aerosol Measurement:

Principles, Techniques, and Applications, (Wiley: New York,

USA), ISBN 0471356360 (cloth).

Morawska, L and Salthammer, T., 2003, Indoor Environment:

Airborne Particles and Settled Dust, (Wiley-Vch: Weinheim,

Germany), ISBN 3-527-30525-4.

IPCC, 1995, Summary for Policymakers: The Science of

Climate Change – IPCC Working Group I,

http://www.ipcc.ch/pub/sarsum1.htm.

Web-based resources

Aerosol Related Home Pages: http://www.aaar.org/hplinks.htm

104

WAV

ES O

F TH

E FU

TURE

the wave’s frequency was, as longas it was big enough – in otherwords, the photoelectric currentshould depend on amplitude of thelight (that is, how bright the lightis), but not on the frequency (thatis, the light’s colour).

This all seems straightforward ...except Nature had other ideas. Ifyou do the experiment, it is easy toshow that, first, the number of

But no – Albert Einstein’s firstscientific paper in March 1905, Aheuristic point of view concerningthe production and transformationof light (Annalen der Physik, 17pages 132-148), earned him theNobel Prize for Physics in 1921.

At the heart of this paper lie thevery nature of light and matter, andthe interactions between them.Prior to 1905, the debate aboutwhether light was a wave or astream of particles seemed to havebeen decided. Isaac Newton, twocenturies earlier, had proposed thatlight was really comprised of tinyparticles, like pellets from an airgun. However, physicists such asHooke, Huygens, Euler, Poisson,Fresnel and, later, Faraday andMaxwell showed that light hadvarious properties that were betterexplained if light is considered awave. In particular, when lightpasses a fine edge, or through athin slit, it forms a complex patternof light and dark bands called aninterference pattern. Interference isa property easily demonstrated withother kinds of waves, like waterwaves or sound waves – and solight, it seemed, was also,conclusively, a kind of wave too,because that explained many of theexperimental results.

But there were experiments thatcouldn’t be explained this way. Inparticular, there was thephotoelectric effect. The idea isthis: shine light onto a polishedmetal surface and you can causeelectrons to jump off the surface.Set things up in the right way andyou can use this to produce anelectric current – this is theprinciple behind some solar cellsand light-detectors.

Albert and the Electrons – The Photoelectric Effect

THE MORE YOU think about it, the less likely it seems: apatent clerk, an unknown in science, submits his firstacademic paper to the Annalen der Physik (Annals ofPhysics). For a first effort from an unheard-of scientist, youmight expect something simple, something naïve, somethingmundane and forgettable.

Einstein’s Miraculous Year /Part 1

Now, light waves could knockelectrons off a metal surface –certainly ocean waves are capableof bashing things around. So whatwould you expect if light waveswere responsible for thephotoelectric effect? You’d expectthat the larger the wave, the moreelectrons would be knocked off thesurface. It shouldn’t matter what

electrons knocked off does dependon the frequency – the higher thefrequency of the incident light, themore electrons you get off themetal surface. But this is only trueabove a certain cut-off frequency;below this cut-off, no electrons areejected at all. And here’s the weirdthing – this is true no matter howbig the wave is.

105

EIN

STEI

N'S

MAR

VELL

OUS

YEA

R,PA

RT 1

It’s a bit like a huge ocean wavesmashing into a sand castle, yetwhen the water washes away, thecastle is intact.

Einstein thought about this anddecided there was a better way tolook at this problem – you have tostop thinking of light as a wave,and think of it as a particle instead.If light consists of little bundles, orphotons, of energy, he said, and if

that energy increases in proportionto the frequency of the light, then itall makes sense. An electron willonly ever absorb one photon at atime – if that photon has enoughenergy (that is, if the light is highenough frequency), it will knock theelectron off its atom in the metalsurface. But if the frequency is toolow, the photons each have small

amounts of energy, and so theelectron will never have enoughenergy to break free – no matterhow many photons there are!

Einstein used the photon idea toexplain the photoelectric effect. Solight is made of particles, right? Notso fast, there – remember, otherproperties of light, like interference,are explained by the wave model,not the particle model. So it seems

light isn’t a wave, or a particle ...somehow, strangely, it’s both.This first paper of Einstein’s in1905 paved the way fordevelopments in the QuantumTheory, which has become one ofmodern science’s most spectacularachievements. The Nobel Academythought the effort worthy of theNobel Prize for Physics in 1921.

JUDY KAY is anAssociate Professor atthe School of InformationTechnologies at theUniversity of Sydney.She is a principal in the Smart InternetTechnology ResearchGroup, which conductsboth fundamental andapplied research in user-adapted systems.The core of her work isto represent and manage

personalisation ensuring the usermaintains control while also beingable to scrutinise and control thewhole process: the user candetermine what is modelled aboutthem, how this is managed and howit is used.

She applies this in ubiquitous,pervasive computing as well asintelligent teaching systems.

106

What got you interested in ITin the first place?

Near the end of high school, Ibegan hearing about computers anddecided that they were going to animportant part of our future world,and that I wanted to be part of that.

What were you like as a kid? Wereyou curious, pulling apart stuff tosee how it worked? Were youalways interested in science/IT, ordid this come on later?I delighted in learning about mostthings. I recall a few bad experiencesin taking something apart and beingunable to reassemble so steeredaway from that sort of thing.

What’s the best thing about beinga researcher in your field?A chance to take a small part ininventing the future, or influencing it.

Who inspires you - either in IT orin other areas of your life?People who come up with elegant,simple and powerful ideas.

If you could go back andspecialise in a different field, whatwould it be and why?Ah! If I’m allowed to replay my lifeonce, well why not twice or more? Inthat case, I would have loved moretime for music, art, history, ... I wouldwork my way through all of themsystematically.

What’s the ‘next big thing’ in IT, inyour opinion? What’s coming up inthe next decade or so?Lots of computation throughout ourlives, in every endeavour, makingreal changes to the way we do andsee things.

Creating and overcoming invisibility:scrutably personalised ubiquitouscomputingAssociate Professor Judy Kay

IntroductionM O O R E ’ S L A W D E S C R I B E S the observed exponential growth in computingpower, with the capacity of new chips doubling every 18 to 24 monthsfor more than forty years. This applies to processing power, mainmemory and discs: computers have become dramatically more powerfuland cheaper, and this trend appears likely to continue. This has meantthat we now have many special purpose computers embedded in carsand home appliances.

107

Figure 1: SharPic coffee table interfaces

108

WAV

ES O

F TH

E FU

TURE

This trend will continue, with increasingnumbers of invisible computers in everyvenue from the home to workplace, working

quietly and unobtrusively. We are just beginningto explore the meaning of invisible computing,also described by Weiser, the much quoted fatherof this emerging area, as ubiquitous or calmtechnology.

One part of the vision for invisible computing wasdescribed by Nicholas Negroponte, whoenvisaged butler-like agents within ourenvironment:

... computer surrogates that possess a body ofknowledge both about something (a process, afield of interest, a way of doing) and about youin relation to that something (your taste, yourinclinations, your acquaintances). Namely, thecomputer should have dual expertise, like acook, gardener, and chauffeur using their skillsto fit your tastes and needs in food, planting,and driving. (Negroponte 1995, p151).

Such computers should look after your needswithout bothering you unduly, just as a goodbutler should. Negroponte also described apersonal newspaper:

Imagine a future in which your interface agentcan read every newswire and newspaper andcatch every TV and radio broadcast on theplanet, and then construct a personalisedsummary. This kind of newspaper is printed inan edition of one. (Negroponte 1995, p153)

The envisioned invisible computing will providethat personal news report in very flexible modes;it should be delivered as audio if you are drivingand need to watch where you are going, or it maybe in almost conventional print that appears onyour coffee table. Of course, this all relies on acomputer learning what you like, want and knowas well as how you want information presented.

Another part of the vision is described by theComputer Research Association in their 2002Grand Research Challenges in Computer Scienceand Engineering over the next twenty years. Theyidentified the goal of ‘A Teacher for everyLearner’. This strives to achieve the famous

2-sigma improvement in learning that Bloomobserved: the result that a learner who is at the50th percentile in typical classroom settings canachieve at the 98th percentile with expertpersonal tuition. To move closer to this, we needbetter, personalised learning support fromcomputers; they should be within our normalenvironment, so that when and where we have alearning need, we can consult our personalisedteacher.

There is a huge amount of computer scienceresearch needed to achieve these visions andmuch of the technical progress needs to beintegrated with human, social and legal demands.This chapter explores two sides of invisibility:some of the current research exploring how tocreate invisibility and, in anticipation of achievingthat, explorations of ways to overcomeinvisibility’s problems.

Creating InvisibilityOne invisibility goal is for computationalresources to fit so well into our environment thatthey feel like a natural part of it: we will not eventhink of them as computers. This contrasts with,and complements, current personal computerswith their screen, keyboards and mice. Althoughthese are becoming increasingly common inworkplaces and homes, they are quite differentfrom the computers in cars, washing machines,ovens and the like. We are barely aware ofcomputers integrated into appliances. Theysimply make those appliances increasinglyconvenient and useful (at least when everythingworks properly). The challenge of this invisibilitygoal is to build systems so that they are useful,work well and do not require any effort to learn anything.

A somewhat different, and conflicting, invisibilitygoal is that the computers in our environment areinvisible in the sense that we can do the thingswe want to without needing to think about thecomputer. An expert user of a personal computerdoes this when they type an essay, being able tofocus on the task of writing, barely thinking aboutthe computer at all – they can be in the flow ofwriting the essay and, in a sense, the computer isalmost invisible. One problem is that the expert

109

CRE

ATIN

GAN

DO

VERC

OM

ING

INVI

SIBI

LITY

:SCR

UTAB

LY P

ERSO

NALI

SED

UBIQ

UITO

US C

OMPU

TING

only gets to this level after making a real effort tolearn and practice using those computing tools.The pay-off for this work is that the power usercan achieve so much more.

The coffee table interfaceThis is a prototype for a future coffee table wherepeople interact with digital images naturallywithout conventional computer interfaces. Figure1 shows two people using this coffee table’sSharePic application to collaboratively work withphotographs. At the top left of the table you cansee a triangular red area beneath the images.This is the personal space of the user at the rightside of the table. He has just dragged a pictureinto that personal space: he simply placed hisfinger near the middle of the image and then slidit along toward the red personal space. Once animage is in his personal space, only he can dragit out.

At the right middle of the table, the swirlingimage, the black hole, swallows any picture thatcomes near it. Near the black hole is a small bluerectangle, ‘the frame’. It is like any other imageon the table: you can enlarge or shrink it byputting a finger at a corner and moving outwardsor inward. You can rotate it by putting a finger ina corner and sliding around an axis through thecentre of the image. (And you can do both atonce, in a rosize gesture.) However, the frame isalso special in that if you put it over any otherimages and allow your finger to dwell in themiddle, it creates a new picture of whatever isunder it. So, for example, if it is over a corner ofanother image, you get a cropped part of thatimage. If it is over a set of three images, theframe creates a composite of those images.

SharePic works by projecting onto the specialtable, which can detect where the user has placedtheir finger or fingers. It can distinguish whichperson is touching the screen. SharePic was partof a research project to support remembranceactivities. An important motivation for the work isthe potentially important role of reminiscenceactivities, based on the huge amounts of digitalmemorabilia that people will own. The researchaims to help people build personal histories andshare memories with others.

Another part of the same project is the prototypedigital scrapbook being used by the person in themiddle back of Figure 2 – this uses the Anotopen and paper systems (www.anoto.com). Theresearch explores natural pen and paper-basedinterfaces for a scrapbook. It allows people tocreate a physical scrapbook with photographsand annotations written with them, rather like anyother scrapbook. However, it also enables theuser to record audio that is associated with thepages of the book and, on revisiting thescrapbook, the user can replay it. The systemalso synchronises the physical scrapbook with anonline version. So, for example, I could create ascrapbook with pictures and associated text andaudio. Then it appears on the web, available forrelatives overseas.

The Anoto pen works with conventional paperthat has a special pattern on it. The pen has acamera that enables it to capture what the userhas written. It uses Bluetooth wirelesscommunication to send this to the processingpart of the intelligent environment. So, forexample, if the user draws the audio recordingshape, they can then record audio for that pointin the scrapbook and later, by ticking thatlocation, they can replay the audio. Of course,this requires a microphone and sound systemwithin the intelligent environment. A similar

Figure 2: The scrapbook pen interface is used by theperson at the middle back, while other people useSharePic. In the background is a large 3-D wall, which isused for fly through visualisations and other activities,like games.

110

WAV

ES O

F TH

E FU

TURE

process can be used to associate video with partsof the scrapbook. Figure 3 shows an example ofa notebook page with audio and video.

Magic Mirror and Keep-in-Touch (KiT)The KiT project supports families to Keep inTouch. The research is multi-disciplinary andaims to help distributed members of cross-generational families to easily maintain contact. Itprovides message-based asynchronouscommunication between the members of afamily. The prototype normally looks like a mirror.For strangers, it blends into the environment, as anormal mirror in a typical house, as in Figure 4.However, if a family member approaches themirror, the screen behind the half-silvered mirrorbecomes active.

Figure 5 shows one KiT prototype, with adistributed family – most of the family lives inSydney, one child is in England and agrandparent lives in Queensland. Otherprototypes use different displays that look likeframed pictures, and KiT also operates onconventional computers.

[FIGURE 4 HERE]

Family members can interact with KiT by usinghand gestures to play messages and make newones. For example, the gesture of moving yourhand down selects the grandmother’s image andthen another gesture starts the recording of amessage. The gestures are intended to be easyto learn and to customise (and they work withoutanyone putting finger prints on the mirror). KiTdelivers messages to a similar appliance in thegrandmother’s Queensland home. She can tellwhen there is message as a red dot appears onthe screen for each message she has not played.

There are many other possibilities for usefulambient displays, like the orb shown in Figure 6.This glows in different colours and light patternsto present digital information, like changes in thestock market. For example, the family that usesthe KiT also has an orb in their Sydney home; thisglows green when their child in England is online,with the orb signal enabled – that way, they knowwhen to try to get in touch. Other examplesinclude art works and even subtle light dots.

Figure 3: Example of a page from the scrapbook.

Figure 4: Magic mirror in mirror state.

Figure 5: Magic mirror display when a family memberactivates it.

111

CRE

ATIN

GAN

DO

VERC

OM

ING

INVI

SIBI

LITY

:SCR

UTAB

LY P

ERSO

NALI

SED

UBIQ

UITO

US C

OMPU

TING

The examples above have focused on ways tointeract with machines that might be integratedinto homes. This is the user’s view of researchsystems that will help create the invisible future.There is considerable technological challengeunder the hood, with research needed onnetworks and operating systems. One of the realchallenges is to make appliances that are ‘zero-configuration’ devices (or Zeroconf – seewww.zeroconf.org), meaning that when you buyyour magic mirror, you can install it at your homeas easily as current consumer electrical devices,like toasters or TVs.

Overcoming Invisibility – building cloaksof visibilityIn parallel with work toward invisible computing,there is exploration of the expected problems thatwill come with it. For example, you might come toa new place and wonder what services areavailable there. Or, you might be surprised bysomething that the intelligent environment does,and wonder why that happened. You might alsoobserve that the same environment treats yourbrother differently, and you might wonder whythat happens and what you can do about it. Wenow describe two examples of research towardsthese challenges of invisibility.

Augmented Reality PhoneSuppose you are in a room that has a hiddendoor. Suppose that you are allowed to open thatdoor by saying the magic words. What happens if

you have forgotten the magic words? Theaugmented reality phone (AR phone) is aprototype that tackles this problem. At the sametime, it might be useful for much more mundaneinvisibility problems – for example, if you are in aroom with a printer and it seems not to beworking, the AR phone could deliver additionalinformation to help you overcome the problem.

It needs special so-called fiducial markers like thelarge symbol at the right of Figure 7 – Figure 8shows what happens when the phone recognisesone of these. The AR-phone has a camera, likemany current phones; the user points this at thefiducial marker. The phone sends this image to asystem, which processes the image andrecognises the marker. It then looks into itscollection of information and images associatedwith this marker, and displays the scene with theadditional retrieved information. In the case ofFigure 7, this additional information is a picture ofa teapot.

[FIGURE 7 HERE]

[FIGURE 8 HERE]

Figure 6: Ambient display of information with the glowingorb.

Figure 7: Augmented reality phone and the fiducial marker it recognises, with the image associated with this marker(the teapot).

112

WAV

ES O

F TH

E FU

TURE

With a device like the AR phone, it is easy topersonalise the information to be delivered. So, forexample, if you are not supposed to know about ahidden door in the wall in front of the camera, theinformation presented may be a description of thepainting on the wall that is there as well. On theother hand, if you are allowed to use the door, thephone might give you a clue to the magic words – for example, if the magic word is teapot, theimage in Figure 7 should help you remember thewords to open the door.

Scrutably personalised museum tourAt the core of any of personalised systems is auser model in which the computer holdsinformation about the user. Figure 9 shows how auser model is related to theworld, the person it models andthe programmer(s) who build thesystems. The top of the figureshows the real world; in the caseof our emerging invisiblecomputing environment, thisworld has many sensors that cancollect information about aperson. Some of the informationfrom these can flow straight intothe user model. Most goes alongthe path shown at the right,being interpreted by programs,which use it to draw conclusionsabout the user. As the figureindicates, the programmer’s ownunderstandings, goals andprejudices are inherently builtinto their code. The figure showsthat the user also interacts withthe world and may provide direct

input to the user model: for example, the usermay answer questions about their preferences.An important thing to notice is that the usermodel is an artificial model, intended toeffectively capture a tidy, useful but dramaticallysimplified view of the real world, which isexceedingly messy.

A personalised system is driven by the user modelthat holds the system’s beliefs about the user.There are many reasons for enabling people toscrutinise personalisation and the user model itself.Essentially, this is a matter of ensuring that whenwe build future personalised ubiquitous computingenvironments, we want to ensure that people arestill in control. This is a fine balancing act: on theone hand, the whole invisibility enterprise aims to

ensure that people need notbother with technical details andthe technology will simply do theright thing, as generally happenswith the many existing applianceswith embedded computers. Onthe other hand, it is immenselyirritating when systems are toosmart and make mistakes: it iseven worse if the user cannotwork out just what is happening,and why it is happening, and howto get back in control. In the caseof user model in the intelligentenvironment, there are alsolegislative requirements related tothe use of personal information.

Figure 10 shows an example ofwhat a user can see when theydecide to scrutinise thepersonalisation provided by a

Figure 8: Architecture of the AR-phone. Figure 9: User models, users, programmers and the world.

“... on the one hand, thewhole invisibilityenterprise aims toensure that people neednot bother with technicaldetails and thetechnology will simply dothe right thing ... On theother hand, it isimmensely irritatingwhen systems are toosmart and makemistakes ...”

113

CRE

ATIN

GAN

DO

VERC

OM

ING

INVI

SIBI

LITY

:SCR

UTAB

LY P

ERSO

NALI

SED

UBIQ

UITO

US C

OMPU

TING

museum guide for the Nicholson Museum at theUniversity of Sydney, Australia’s first museum. Thisdisplay enables this user to see information thatwas not presented to them in the normaloperation of the system: this has a yellowbackground. The information with a greybackground was presented because of this user’smodel. Other parts, with no background, arepresented to everyone.

To the right of the display, there is a descriptionof the parts of the user model – if the user placestheir mouse over one of personalised parts of thescreen, they see the ‘mouse-over’, as in thisfigure, explaining why the material was omitted.Of course, in the normal mode, the information issimply presented as in any personalised webpage – and, then it looks rather like the manypersonalised sites on the internet. These usecookies to connect each visit to a web site fromthe same computer. The web site can build upuser models or profiles: people are generallyunaware of what is personalised, and oftencannot easily find out.

Your own avatar guide to a personalisedmuseumThe PEACH project is exploring ways to personalisemuseum tours. Figure 11 shows one part of PEACHthat is working towards helping help people figureout where they look in a rich environment. Thefemale figure at the left is projected onto the wallof the museum and moves around, leading theuser’s eye towards relevant parts of theenvironment. In this case, she is standing next to alarge display that presents information. The figureat the right is an artist: he presents information

Figure 10: Scrutably personalised information delivery.

Figure 11: Personal tour guide avatar at left and personalinformation delivery avatar at right.

114

WAV

ES O

F TH

E FU

TURE

only to those visitors who want an artistic emphasisin their tour.

An interesting part of the PEACH work is toexplore ways to deal with groups of people –people do tend to visit museums in groups and,for that matter, they spend much of their time athome, work and elsewhere in groups as well.Figure 12 shows three people at a large screendisplay with audio information delivered by theartist figure at the right of the display. PEACH usesthe large display for common information; this isfor all members of this group. In addition, eachperson has a personal device, with personalisedinformation. In this case, an invisibility problemoccurs since people may not know when andwhere to look. The PEACH project uses animatedavatars to help with this too.

Scrutinising large user modelsSuppose you wanted to know about your usermodel, as used in a complex system. You need todeal with the problem of seeing a large amountof information. Scrutable Inference Visualisation

(SIV) helps with a visualisation like that in Figure13. This user model could well be derived fromone of the many movie-recommender web sitesthat build up models of what a person likes ordoes not like, based on collections of informationabout other people’s preferences. Such sites arean example of another group of inscrutable, butpersonalised, systems that are widely available.

If a recommender were enhanced with a SIVdisplay, like that in Figure 13, you could easily seethat most of the titles displayed are green,indicating the user model represents them as‘liked’ by this user. The red titles are ones modelledas ‘not liked’, and Great Catherine stands out inred. The movie that is currently selected isLawrence of Arabia – all the titles that are in largerprint are more similar to this film than those shownin smaller print. The visualisation was designed togive a quick overview of a user model so the usercan then select parts of interest to scrutinise inmore detail.

Figure 12: Personalised information delivery that exploitslarge displays for common information and individualdisplays for personalised extras.

115

CRE

ATIN

GAN

DO

VERC

OM

ING

INVI

SIBI

LITY

:SCR

UTAB

LY P

ERSO

NALI

SED

UBIQ

UITO

US C

OMPU

TING

ConclusionThis paper has presented a user’s view of aselection of research projects in the area ofpersonalised ubiquitous computing. Some involvehardware that is still novel but is likely to be aprecursor of appliances of the future. In many ofthe projects, much of the research explores arange of deep technical issues as well as thepragmatics of how these systems will fit intopeople’s lives. In this brief overview, it has notbeen possible to describe these; nor has therebeen the space to analyse the impact this willhave on every aspects of home, work and leisurelife. There are extraordinary possibilities forinvisible computers, as well as the conventionalvisible ones, to support activities in all professionsand other aspects of our lives.

AcknowledgementsSeveral of the projects described in this paperwere funded by the Smart Internet TechnologyCooperative Research Centre.

References

Assad M, D J Carmichael, D Cutting, A Hudson. (2003).

Accessible Augmented Reality in the Intelligent

Environment. Proceedings of OZCHI, pp 268-287.

Bloom, B. S. (1984). The 2-sigma problem: The search for

methods of group instruction as effective as one-to-one

tutoring. Educational Researcher, 13(6), 4-16.

Negroponte, N. (1995). Being Digital, Hodder and Stoughton.

Donald Norman, The invisible computer, why good products

can fail, the personal computer is so complex, and

information appliances are the solution, Cambridge, Mass.

MIT Press 1998.

Weiser, M. (1991) The Computer for the Twenty-First Century,

Scientific American, 94-10.

Web sites to explore

www.ambientdevices.com/cat/index.html

Examples of current products that are ambient devices

www.media.mit.edu

MIT Media Lab

oxygen.lcs.mit.edu

MIT Project Oxygen: pervasive human-centred computing

home.cc.gatech.edu/ubicomp

Georgia Tech Ubiquitous Computing Research Group

www.ubiq.com/hypertext/weiser/UbiHome.html

A summary of foundation work on ubiquitous computing with

links to major papers by Mark Weiser.

www.doc.gov/ecommerce/eudir.htm

peach.itc.it/consortium.html

PEACH Personal Experience with Active Cultural Heritage

Consortium

Figure 13. Visualisation of alarge user model for aperson’s movie preferences

116

WAV

ES O

F TH

E FU

TURE

But it was a bit of a surprise to findthat this theory of “excitable media”could deal with something ascomplicated as human behaviour –but then again, we are talkingabout a Sports Event.

You can think of a person as beingan excitable unit – and that’s notso unusual. This simply means thata person can be prompted intoaction by some sort of externalstimulus. In general, the closer andmore powerful the stimulus is, themore the human being, as anexcitable unit, would respond. So toexplain how a Mexican Wave canget started, and keep rolling, eachperson needs only three internalrules that they obey, one after theother. First, they wait in their restingstate, ready to be excited by theright stimulus. Second, whenstimulated, they go through theactive phase where they stand upand wave their arms. Third, they goto the refractory phase, which is

This so-called “Mexican Wave” firstbecame famous during the 1986Soccer World Cup in Mexico. Infact, that’s how the Mexican Wavegot its name, because it got its firstworld-wide exposure at this event – and soon enough, there was aswell of interest in far-away Europe.

Scientific work was done by TámasVicsek and his colleagues from theEötvös University in Budapest inHungary, and the University ofTechnology at Dresden in Germany.They video taped, and thenanalysed, 14 separate MexicanWaves in football stadia eachholding more than 50,000 people.They noticed that the wavegenerally went in a clockwisedirection – that it spread from oneperson to the next person on theirleft. The Mexican Wave usuallymoved at around 12 metres (or 20 seats) per second, and wasabout 9 metres wide (about 15seats wide).

Now mathematicians have mademathematical models of much ofthe world around us, and some ofthem did mathematical models ofwhat are called “excitable media”.

One example of “excitable media”is the dry trees and dry litter in aforest. This particular theory of“excitable media” describes how aforest fire starts, and then spreads.

Another example of “excitablemedia” is the set of muscles thatmake up your heart tissue. Whenthey’re given a little jolt ofelectricity, the muscles in yourheart will contract. It turns out thatthere are a quite a few electricalabnormalities of the heart thatinvolve problems with theexcitability of the heart tissue –either it’s too excitable, or it’s notexcitable enough. In fact, the theoryof “excitable media” can alsopredict that some people will haveextra heart beats.

Mexican WaveBy Dr Karl Kruszelnicki

IF YOU’VE EVER been to, or watched a major sporting event,you’ll probably have seen the famous Mexican Wave. Thiswave sweeps around the audience in a stadium or sportsarena, as firstly one group of people leap to their feet withtheir arms up and then sit down and then the group ofpeople next to them does the same thing, and so on. If you’reon the other side of the stadium looking across, you can seethis beautifully rhythmic and synchronised movement rollingthrough the audience. On a good night, you can see multiplewaves winding their way around the terraces. Somescientists have studied this strange phenomenon, and notonly do they now understand it, they can probably dosomething useful with this knowledge.

117

MEX

ICAN

WAV

Ewhat you and I would call “sittingdown again”.

You now might say that thisresearch is incredibly useless. Butit is important with regard tocrowd control. When you have100,000 people at a sports event,simply moving them in and out ofthe stadium can be dangerous ifit’s not done wisely. Evensomething as simple as putting ahand rail along the centre of acorridor can speed up themovement of people in or out of astadium. So knowing this theory ofexcitable media could help if a

So if you do see, or are involved in,a Mexican Wave, this research alsotells you that you’re not gettingyour money’s worth – because thegame is boring, and the crowd areentertaining themselves...

small group of agitators tried toget a large crowd over-excited.

The scientists did find a goodrelationship between their theory,and what they actually saw in reallife. Basically, you need a minimumcritical mass of two or three dozenpeople to get the wave going. Andeven that’s not enough – you needa lull happening in the sportsevent. After all, if the football gameor athletics event is incrediblyriveting, the audience is not goingto pay attention to the person nextto them suddenly jumping up.

FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

Illustration courtesy Adam Yazxhi

David, what got youinterested in science in

the first place?I was always interested in ‘howthings worked’, and theencouragement of teachers at school was important in fosteringthis interest.

What’s the best thing about beinga researcher in your field?You get paid to pursue your hobby.

Who inspires you – either inscience or in other areas of your life?When I was younger, I was inspiredby older eminent scientists whoknew so much more than I did. NowI’m older, I’m inspired by youngstudents who know so much morethan I do.

If you could go back andspecialise in a different field,what would it be?Today it would be aboriginalmedicine; yesterday it wasinternational law; tomorrow it mightbe legal aid; and then there’s howthe memory works, and carpentryand school teaching and ... Whythose fields? Because of theenthusiasm that people I come intouch with have for these topics (mychildren, in three of the cases).

What’s the ‘next big thing’ inscience, in your opinion? What’scoming up in the next decade or so?The next big thing in science needsto be a significant development inrenewable energy sources.

PROFESSOR DAVIDCOCKAYNE FRShas been a Professor inthe Department ofMaterial at the Universityof Oxford since 2000.Before that he was aProfessor at theUniversity of Sydney.He studiedundergraduate physicsat the University ofMelbourne, and hecarried our research at

the University of Oxford for his D Phil.In 1999 he was elected to the RoyalSociety for his pioneeering work inelectron microscopy.

118

Seeing in the NanoworldProfessor David Cockayne FRS

IntroductionA D V E N T U R O U S Y O U N G people in the 18th and 19th centuries travelled theworld, exploring unknown countries, and seeing wonderful and strangesights. When they first landed in Sydney where we are today, theydiscovered not only a new country, but also an amazing number ofanimals never before seen by Europeans. When they wrote of theirdiscoveries, many of their countrymen did not believe them. Indeed,when the skin of a platypus was sent back to England, scientiststhought it was a fraud.

119

120

WAV

ES O

F TH

E FU

TURE

Nowadays every corner of the earth hasbeen ‘discovered’, and we might think thatthere’s nowhere left to explore. But there

is – scientists are the explorers of the 21stcentury and the nanoworld is waiting to beexplored. And just as Cook and Magellandepended upon specially designed ships for theiradventures, so scientists rely upon advancedinstrumentation and techniques to explore thenanoworld. And just as the early explorers sawstrange and unfamiliar animals, plants and landformations, so the scientist-explorer is finding awealth of unfamiliar objects – from bucky balls tonanotubes, and carbon nanotrees tonanomatches. In this lecture, we follow thedevelopment of the most important tool forexploring the nanoworld – the electronmicroscope – and we discuss the closerelationship between diffraction and imaging.

The relationship between diffractionand imagingTo understand why we can’t see the nanoworldwith our naked eyes, or even with a lightmicroscope, we must consider the relationshipbetween diffraction and imaging. We see diffractionin various forms all the time – in the colours ofbutterfly wings and of opals, and in some aspectsof a rainbow. So diffraction is not unfamiliar to us.And we can clearly see the relationship betweendiffraction and imaging if we carry out a simpleexperiment of firing a laser through a grid of wires.On a distant wall, the pattern we see is an array ofspots; a hexagon if the grid of wires is hexagonal,and a line of spots if the grid is a set of parallelwires. This array of spots (a line or a hexagon)carries information about the wires, displayed in aparticular way – the diffraction pattern, as we callthe pattern on the wall. If we now place a lensbetween the grid of wires and the wall, the sameinformation can be displayed in a different way –an image. This image can be in focus or out offocus, depending upon exactly where we place thelens. But the important point to realise is that it isthe same information in the scattered light whichresults in these patterns, the images and thediffraction pattern. As we shall see, sometimes it ismore useful to study the information as an image,and sometimes it is more useful to study it as adiffraction pattern.

This close relationship between diffraction andimaging has been understood since the work ofAbbe in the late 19th century. He showed that iflight of wavelength λ is scattered by an object,then the information carried by that scatteredlight can tell us about the details of that object,but only down to a level of detail d, where

d = 0.5λ [1]

This means that light of wavelength 500 nm, say,will carry no information about any feature of anobject that is less than 250 nm in size. At the endof the 19th century, this was of great concern toscientists (and indeed to at least one poet (Belloc1912)), because it meant that much of themicroscopic world would never be visible. Itwasn’t that scientists of those days didn’t havegood light microscopes – they had perfectmicroscopes – the problem was that theinformation simply wasn’t there in the scatteredlight.

This question of how much information can becarried by a wave is easily understood if you thinkabout the following experiment. You are standingon the sea shore, looking at the waves coming insmoothly from far out at sea. Suddenly the wavesstart to arrive with froth in them. Seeing thisdisturbance, you realise that there must be someobject over the horizon that is causing thisdisturbance. Could it be a small rowing boat? No,because a small rowing boat would just bob upand down on the wave, without causing any froth.It must be something bigger – perhaps a whaleor a ship.

So it is only if the object disturbs the wave thatyou will know it is there – and you can imaginethat something smaller than about half thedistance between the wave tops (for example, arowing boat or a small fish) won’t disturb thewave. Of course, whether or not you candifferentiate between the froth from a whale andthe froth from a ship is a question of howintelligent you are – it isn’t a problem that thewaves can solve for you! And it’s the same withlight. If the object doesn’t disturb the light, theimage or the diffraction pattern won’t showanything about it.

121

SEEI

NG

INTH

EN

ANO

WO

RLD

The situation improved with the discovery of X-rays (Roentgen, 1895), since they can have awavelength of 1 nm or less. Equation 1 showsthat the scattered X-rays would then carryinformation at the level of interatomic spacings,about 0.3 nm. As a result, X-ray diffraction (thestudy of the scattered X-ray intensity as afunction of the scattering angle) has been apowerful tool for investigating the structure ofmaterials down to the atomic level for over half acentury. It is a major tool in chemistry, materialsscience mineralogy and molecular biology.

But the discovery of X-rays did not lead to seeingobjects in the nanoworld because, although thescattered X-rays carried information at thenanoscale, no lens was available to form animage.

Things might have remained there – resolution inimaging at 0.25 micron with light, and at theatomic level with X-ray diffraction. But then deBroglie (1924) showed that electrons could beconsidered as waves, with their wavelengthgiven by

λ = h/(mv ) [2]

(h is the Planck constant, and m and v the massand velocity of the electron respectively). Forelectrons travelling at 75% the speed of light, thisgives a wavelength shorter than that of X-rays.The experimental proof of this hypothesis resultedfrom experiments by Davisson (with Germer andKunsmann) in 1927 (and by Thomson at thesame time, but with a different kind ofexperiment), which demonstrated that periodicobjects (crystals) give rise to diffraction patternsfor which the distribution of diffraction spots isgiven by the Bragg equation

2d sin(ϑ /2) = nλ [3]

where d is the repeat pattern spacing and n is aninteger. (This same equation can be used todescribe the distribution of light spots caused bythe scattering of the laser by the grid of wires weconsidered earlier.) If we know the light or X-rayor electron wavelength λ, and we measure theangles θ, then we can determine the spacing d.In addition, the intensities of the diffraction spots

can be used to determine the shape of therepeating object (the grid wires in our case). So ifwe use X-rays or electrons, the diffraction patterncan be used to study the arrangements of atomsin crystals.

But now we should reflect on the next step. Aswe have seen, X-ray diffraction existed, but (untilvery recently) there were no X-ray lenses and sono images with X-rays – you could not seeobjects much smaller than the wavelength oflight. There are no lenses for water waves. Nonefor gamma rays. It was not inevitable that therewould be lenses for electrons – not at all. But itwas known that electrons can be deflected by amagnetic field, and in 1926, Busch showed thatthe path of electrons through a magnetic field issimilar to that of light through a glass lens, i.e. amagnetic field has a lens action.

Very quickly electron lenses were made, and puttogether to construct an electron microscope. Forthis, Ruska received the Nobel Prize (in 1986,some 50 years after his invention!). Thearrangement of the lenses is identical to that fora light microscope, with magnetic fields replacingthe glass lenses, and fast electrons replacinglight. How good were these electron lenses?Absolutely terrible – about as good a lens forelectrons as the bottom of a Coca Cola bottlewould be for light. In fact electron lenses aresuch terrible lenses that they hardly deserve tobe called lenses at all. But the reason they areused is because if they were perfect, we wouldbe able to see detail about half the wavelength ofthe electron – about 0.002 nm, or 1/100th of thesize of an atom. So with terrible lenses, perhapswe can see an atom. And if we didn’t haveelectron lenses – and as I’ve said, there’s noinevitability that they would have existed – thenwe would not have seen the structure of cells, ortheir organelles, we would not have seen buckyballs or carbon nanotubes, and certainly notatoms.

So at this stage, scientists had electron diffraction,needing no lenses, and electron lenses (eventhough they were terrible) to form images atresolutions previously unachievable. And so theyhad the possibility of both imaging and diffractionwith electrons. Just like excited explorers given

122

WAV

ES O

F TH

E FU

TURE

new kinds of ships, what new explorations couldthey carry out with these new microscopes? Wecan certainly see objects below the limit of thelight microscope such as viruses and molecules,small catalyst particles and the components ofmodern computers and DVDs. And with the bestmicroscopes, we can even see atoms.

Diffraction with electronsOne of the first proofs that electrons had a wavenature (as given by Equation 1) was in a series ofexperiments by Davisson and his coworkers, whoscattered electrons off a nickel target, and lookedat the diffraction pattern. The pattern (Figure 1)was very uninteresting! But one night theapparatus exploded, and the entire experimentwas destroyed. They put it back together again,and heated the sample to a high temperaturebecause it had itself become destroyed. Andwhen they repeated the experiment, the resultswere completely different – the diffraction patternnow showed maxima of intensity at angles givenby the Bragg equation (Equation 3). And becausethe Bragg equation describes scattering bywaves, this provided proof of the wave nature ofelectrons. This kind of luck is called serendipity,and is an essential attribute for successfulscientists! (The word serendipity was made up bythe writer Horace Walpole in 1754, from a fairytale in which three princes from a countrySerendip – an old name for Sri Lanka – madeaccidental discoveries about things that they werenot looking for.)

One of the most important advantages of electrondiffraction over X-ray or neutron diffraction is thatit can be carried out in an electron microscopefrom a region of sample chosen from within theimage. That is, we can compare the diffractionpattern from small regions of the same sample,and so use selected area diffraction as a meansfor characterising differences in structure within asample. Because X-rays and neutrons have nolenses (although X-ray lenses are now underdevelopment), X-ray and neutron diffractionpatterns are necessarily average patterns fromthe entire sample.

Different ways of seeingTransmission Electron Microscopy Since most of you will know the construction ofthe light microscope, the best way to explain theconstruction of a transmission electronmicroscope (TEM) is by analogy with thetransmission light microscope (LM). If we turn thelight microscope upside down (see Figure 2), then

i. At the top we have a source of light – aheated wire. This wire emits both light andheat, and perhaps a few electrons. If the wireis made of tungsten (W) it will give off heat andlight, but also a lot of electrons (because of itslow work function). So we use a W wire in aTEM as the source of electrons. (Increasingly,other brighter sources, such as lanthanumhexaboride crystals and field emission sources,are being used.)

ii. In the LM, the light travels at the speed oflight (!) towards the sample. In the electronmicroscope, the electrons that come off the Wwire are ‘thermal’ electrons, and are nottravelling very fast. We want them to travelfast, because we want them to have a shortwavelength so that they will carry informationabout the small details of objects (see Equation1). And since the wavelength of an electron isrelated to its velocity (Equation 2), weaccelerate it across a large voltage difference(30,000 to 1 million volts). The electrons, likethe light in the LM, then travel rapidly towardsthe sample.

iii. In the LM, a glass lens (the condenser lens)

Figure 1: The scattering curves from nickel as reported byDavisson and Germer before and after their apparatusexploded.

123

SEEI

NG

INTH

EN

ANO

WO

RLD

concentrates the light and directs it onto thesample. As we have seen, glass does not actas a lens for electrons, but a magnetic fielddoes. So for electrons we use a magnetic fieldas the condenser lens, to concentrate theelectrons onto the sample.

iv. The sample scatters the light (electrons),which is then collected by the objective glass(magnetic) lens to form an image (always areal image in the case of the TEM). In the caseof the LM, this image is a variation in lightintensity across the field of the image. In thecase of the TEM, it is a variation in electronintensity across the field of the image.

v. This image is then treated as an object, andimaged once more into a final image by aseries of lenses (the eyepiece or projectorlenses) onto a photographic film in the case ofa LM. Similarly, for a TEM, the image is formedon photographic film or onto the screen of acamera. For the LM, the final projector lens canbe our eye, and the final screen our retina. Ifwe use our eye and retina for the final lens andscreen for the TEM, they will be destroyed!

vi. Even though the electrons are travelling veryfast, they are not massive, and so they arestopped by about a centimetre of air.Consequently we must evacuate the electronmicroscope. Inserting the sample requirespassing it through an air lock, to maintain thevacuum.

vii. Because the electrons are so easilystopped, the sample for an electronmicroscope must be very thin, typically lessthan 1 micron in thickness.

The lens action in the electron microscope comesfrom the magnetic field of the lenses, which isformed by passing electric current through a wire(windings) which are wrapped around thecylindrical metal lens. Consequently the strengthof the electron lens (its ability to magnify) can bevaried by adjusting the current through thewindings. This has obvious advantages over alight microscope, where each lens has a fixedmagnifying power. Just as in a light microscope,the image is focussed by adjusting the objective

lens (moving it in the case of the lightmicroscope, and changing the current throughthe windings in the case of the TEM).

Scanning electron microscopy (SEM)There is a second way of forming images thatdoes not depend upon having a physical imaginglens after the sample. If we use a lens toconcentrate the electrons into a very small spot(say 3 nm in size) on the surface of a solidsample, they can be scattered backwards fromthe sample surface and collected on a detector (aback scatter detector). Their number is used tocontrol the brightness of a dot on a cathode raytube (CRT). The spot is then scanned across thesample surface in a raster, and the dot on theCRT is scanned in synchronism.

If there are variations in the sample surface (e.g.different elements at different positions, differenttopology) then the number of back-scatteredelectrons will vary with position, and so theintensity on the CRT will vary. In this way abackscattered (BS) image is formed. At the sametime, some of the electrons penetrate thesurface, and electrons (secondary electrons, SE)are ejected from the sample. A different detectorcan count these secondary electrons, as afunction of the spot position, and so form a SEimage on the CRT.

The SEM has a much greater depth of field (thatis, the depth over which the object remains infocus) than the TEM or the LM, and consequentlythe SEM is very useful for investigating the

Figure 2: A comparison of a modern TEM and a transmittedlight microscope.

124

WAV

ES O

F TH

E FU

TURE

surface of objects which have a complicatedtopography, as in Figure 3. Because the electronsin a transmission EM (TEM) pass through thesample, the TEM is used to look at the internalstructure of the sample, while the scanning EM(SEM) looks at the surface (or the near-surface).The resolution of the SEM is limited by the size ofthe incident electron spot, and how much itspreads out inside the sample when it generatesthe detected signal.

Seeing in 3DThe electron microscope techniques we havediscussed so far give us two-dimensional images;but since many objects are three-dimensional, wewould like to be able to see in 3D. How do wenormally see in 3D? By using two eyes, with eacheye viewing the object from a slightly differentdirection (interestingly, this means that beyond afew metres, you can’t see in 3D!). Each eyereceives a different image from the one object,

and our brain puts these together into a 3Dimage. So the question is, how can we obtain twoimages in the electron microscope?

It is not difficult – we take two pictures with thesample in slightly different orientations, and thenshow one image to our left eye and the other toour right eye, simultaneously. We can view themwith a special viewer (a stereo-viewer), or we cancross our eyes, or we can print one image ingreen and the other in red, and use filteringglasses. We can go further than this: we canpresent the two images to a computer, and let itreconstruct the 3D object for us. Then on thecomputer screen we can rotate it, view it from achosen direction, and analyse it in full detail.

To improve resolution, we can collect not two buttens, or tens of thousands, of different views – asthough we had tens of thousands of eyes – andso build up a detailed 3D structure of the object,at a resolution approaching 1 nm. An example of

Figure 3: A scanning electron microscope image of a radiolarian.

125

SEEI

NG

INTH

EN

ANO

WO

RLD

such a procedure for a magnetite nanoparticlefrom a meteorite from Mars is shown in Figure 4,from various viewing directions. This technique isknown as electron tomography. The collection ofthe many images can be automated, andautomated electron tomography is now animmensely powerful tool for revealing thestructure of objects, and especially in biology forinvestigating the structure of viruses andmacromolecules.

Atomic resolutionSince the development of the first electronmicroscope, the goal has been to see atoms. Theimportance of being able to see atoms tounderstand the physical world cannot beoveremphasised. According to Feynman, if hewere to have a choice of leaving only one briefstatement to future inhabitants of the world if acataclysmic event were to occur, it would be thatall things are made of atoms. And the fact thatwe can see atoms, and how they interact tomake crystals and molecules, makes the electronmicroscope perhaps the most important scientifictool to have been developed in the 20th century.

To achieve this, the major problem of lensaberrations, referred to earlier, had to beovercome. And unlike glass lenses for light, therewas no possibility that a perfect electron lens

could be made by spending a lot of money and alot of time. Round magnetic lenses necessarilyhave aberrations. Initially aberrations werereduced by improving the lens design to minimisetheir effect, but as improvements occurred,further improvements proved more and moredifficult to achieve.

Recently, two approaches have been developedwhich completely overcome the majoraberrations. The first approach was to accept thatevery image is distorted by the aberrations of thelens, and to realise that the form of this distortionchanges if the image is taken slightly out offocus. So, by taking a series of images each witha different focus – a through focal series – thetotal set of images can be used to reconstructwhat the image would have looked like if the lenshad no aberrations.

An example is seen in Figure 5 for a boundarybetween two crystalline grains of gold. Any singleimage is unclear (see the first two in Figure 5)because the lens aberrations have distorted it,

Figure 4: A nanocrystal of magnetite reconstructed from aseries of tomographic electron microscope images. Thesample is from a meteorite from Mars.

Figure 5: Two high resolution TEM images taken atdifferent settings of focus, and the image reconstructedfrom a set of such images. The images are of a boundaryin Au.

126

WAV

ES O

F TH

E FU

TURE

but the final image, which is reconstructed fromthe full set of images, allows us to see theboundary clearly (the last image in the Figure).

The second approach is similar to what we dowhen we have poor eyesight: we wear correctorlenses (glasses). In a similar way, correctors havebeen developed for electron lenses tocompensate for the aberrations. In themicroscope shown in Figure 2, these correctorsare located about half way down the column.

Using these techniques, it is now relativelystraightforward to form images showing detail atthe atomic level. So we can see crystallinestructures and their defects, we can seeindividual atoms on surfaces, and even watchthem move, and we can investigate atomicinteractions at the atomic scale.

Atomic resolution can also be achieved usingscanning electron microscopy. As we have seen,SEM involves scanning a small probe of electronsacross the surface of a sample and recording thescattered signal as a function of probe position. Ifthe sample is very thick, the electron beam isscattered within the sample, and resolution islost. But if the sample is very thin (say 10 nm)then it passes through the sample scattering inall directions.

Electrons that have passed close to the nucleusare scattered to high angles, and can becollected on an annular detector. Their intensity isproportional to Z2 (Z is the atomic number) and sothe scanning image shows heavy elements (largeZ) brightly. We call this technique scanningtransmission electron microscopy (STEM). In thisway we can investigate the arrangements ofheavy elements within a matrix of lighterelements, and the example of Figure 6 shows Luatoms that have segregated to the surface ofsilicon nitride crystals in a ceramic used forautomotive bearings.

What kind of atoms?Seeing atoms is one thing – but can we tell whatatoms they are? We can answer this question byreferring to the Bohr model of the atom, in whichthe atomic electrons circle around the nucleus inwell-defined orbitals, with each orbital having awell-defined energy. Different atoms havedifferent electronic shell structures and differentorbital energies. If we fire a fast electron at anatom, we can knock out one of the inner orbitalelectrons (Figure 7) and make the atom unstable.It can regain its stability by having an electronfrom a higher orbital ‘fall’ into the vacant orbital;in doing so, the electron loses energy equal tothe energy difference between the two shells –for example, 8 keV for Cu and 6 keV for Fe.This lost energy can express itself as an X-ray ofthe same energy. If this X-ray is collected by adetector that can measure the energy, then wehave a means of knowing whether the electronhas hit a Cu atom or a Au atom. Figure 8 showsa spectrum of X-ray energies collected by a solidstate X-ray detector from a metal alloy. We see alarge number of X-rays corresponding to anumber of different elements. This spectrum canbe converted to a compositional analysis if weknow about the physics of X-ray generation(which we do!).

Figure 7: A fast electron (1) ejects a core electron (2) fromthe inner shell of an atom (left) and loses energy. Then anelectron from another shell drops into the inner shell, andemits an X-ray (right).

Figure 6: The interface between asilicon nitride crystal (bottom) and aglass (top) in a silicon nitrideceramic. The bright atoms at theinterface are Lu, used to engineerthe strength of the interface.

In an electron microscope, we can focus all theelectrons into a very small region of the sample,and so collect an X-ray spectrum from just thatarea. By scanning the electron spot across thesurface, and recording the number of X-rays of aparticular energy (the X-ray spectrum) while thescan proceeds, we can generate an elementalmap of the sample. Figure 9 shows an example

127

SEEI

NG

INTH

EN

ANO

WO

RLD

of the Nb and Cu elemental distributions in amultilayered material, where the layers are 1.5nm in thickness.

Of course, the incident electron in Figure 7 willlose energy when it knocks the orbiting electronfrom the atom, and the amount of energy it loseswill be equal to the energy required to knock outthe orbiting electron. This energy will dependupon the atomic species. By placing aspectrometer after the sample, we can displaythe number of electrons that have lost differentamounts of energy (the electron energy lossspectrum or EELS) as in Figure 10. Just as wecan use particular X-ray energies to givecompositional maps, so we can use the energyloss signal in a similar way.

It is clear from the X-ray spectrum and EELS ofFigures 8 and 10 that the more there is of aparticular element, the larger will be the signal atthe energy corresponding to that element. Therelative signal strengths can be used to give therelative amounts of each element, if we knowhow easy it is to generate signals of each energy(the cross section). The ability to determine thecross sections for experiment or theory is nowvery sophisticated, so that quantitative X-ray andEELS microanalysis is widely used formicrostructural elemental characterisation.

Lorentz MicroscopyMagnetic materials are among some of the mostimportant for technological applications e.g. forrecording data or for acting as fast switches.These materials are generally composed of localregions (domains) in which the direction ofmagnetisation is well defined and is different ineach domain. The information is stored by thesedomains by their direction of magnetisation, andthe aim of the materials scientist is to developmaterials in which this information is stable, andcan easily be written and read.

Figure 8: An X-ray spectrum from a multi-element alloy.

Figure 9: An elemental map of Cu (red)/Nb (green) mutilayers (1.5 nm period).

128

WAV

ES O

F TH

E FU

TURE

In order to visualise these domains, we can usethe fact that electrons are deflected by magneticfields. An electron beam passing through adomain will be deflected by the domain in adirection that depends upon the direction of thedomain’s magnetic field. Adjacent domains, withdifferent directions of magnetic polarisation,deflect the electrons in different directions. If theimage is formed from only those electrons thatare scattered into a particular direction, thendomains which scatter in that direction willappear in the image, and domains which don’t,won’t. Figure 11 shows an example for smallmagnetic domains in a FePd alloy, wherealternating domains appear bright and dark,depending upon whether or not the electrons arescattered into the direction for which theelectrons are allowed to reach the image.

ConclusionThe realisation that electrons have a wave nature,and the invention of electron lenses, opened upthe possibility of seeing objects smaller thancould ever be seen with light microscopes. Thelimitation of the resolution that can be achieved isset, in the first instance, by the aberrations of theelectron lenses.

Over the past 60 years, these aberrations havegradually been reduced, but very recently,methods for eliminating the effects of the majoraberrations have been developed. It is nowpossible to see individual atoms, and how theycome together to form structures. At the sametime, the X-ray and energy loss signals from

scattered electrons give signals that are sensitiveto the atomic number, Z, and because of this, wecan determine the local composition, almost tothe atomic level.

In the next chapter, we will see how thiswonderful instrument can be used to explore thenanoworld.

Reference

H Belloc H 1912 “The Microbe” from More Beasts For Worse

Children. Duckworth 1912

Website

For a clear description of the works of Davisson and of de

Broglie and of Ruska and other Nobel Laureates, see

http://nobelprize.org/physics/laureates/

Figure 11: Lorentz microscopy (Foucault) magnetic domainimage of an FePd alloy.

Figure 10: An energy loss spectrum from BN.

129

SEEI

NG

INTH

EN

ANO

WO

RLD

130

WAV

ES O

F TH

E FU

TURE

suspended particles, their Brownianmotion, could be explained if youimagine what’s going on at an evensmaller scale, down at the level ofwater molecules.

Einstein argued that the watermolecules would be in constantthermal motion, the random kineticenergy associated withtemperature, all jiggling about andcolliding with the much largerparticles floating in the water. Andthose many collisions, averagedover time, could explain themeandering paths of the larger

But back in 1905, the case for theatom wasn’t quite so strong – infact, you could say that the wholefield of science focused on theproperties of matter was in a bit ofa bother. Boltzmann had astatistical theory ofthermodynamics (the study oftemperature and the flow of energywithin physical systems) that statedthat irreversible changes, like themelting of an ice cube, could beexplained statistically by thereversible motions of tiny, unseen,hypothetical atoms. But there weremany other physicists who doubtedthe idea (including the great MaxPlanck, at first). So the atomichypothesis was very much up forgrabs in 1905.

In another, very different corner ofscientific inquiry – more down thebiological end of things – there wasa seemingly unrelated phenomenonsearching for an explanation. In1828 a botanist named RobertBrown had been observing pollengrains suspended in water. Duringhis observations, Brown noticedthat the tiny particles of dust andpollen in the water seemed to be in

constant, random, zig-zaggingmotion. No obvious reason for thiswas seen at the time, and so thephenomenon was dubbed Brownianmotion and it remained a curiosityfor over 75 years.

In May 1905, Einstein published hissecond paper of that remarkableyear: On the movement of smallparticles suspended in stationaryliquids required by the molecular-kinetic theory of heat (Annalen derPhysik 17, pages 549-560). In thispaper, Einstein showed that themysterious haphazard wanderings of

Einstein’s Miraculous Year /Part 2Albert and the Atoms – Brownian MotionTHE CONCEPT OF the atom is so commonplace today, it’shard to imagine not believing in these tiny bundles of matter.Yet believe we do, and the belief is founded on a somedegree of faith, because unless you use something like apowerful electron microscope, you can’t actually see atomsat all. Instead, from our earliest lessons in science at school,we’re assured that atoms exist in all their different elementalvarieties, and we believe in the evidence built over years of scientific research that overwhelmingly supports their existence.

131

EIN

STEI

N'S

MIR

ACUL

OUS

YEAR

,PAR

T 2

suspended particles. His paperbegan:

... according to the molecular-kinetic theory of heat, bodies ofa microscopically visible sizesuspended in liquids must, as aresult of thermal molecularmotions, perform motions ofsuch magnitudes that they canbe easily observed with amicroscope. It is possible thatthe motions to be discussed hereare identical with so-calledBrownian molecular motion;however, the data available tome on the latter are so imprecisethat I could not form a judgmenton the question.

But more than offering amechanism, Einstein calculated themagnitude of the effect – he showedwith rigorous statistics that themolecular thermal motion gave justthe right amount of kick to result inthe observed Brownian motion.Einstein’s great insight came fromthe way he tackled the problem: heignored the actual randomwanderings of the suspendedparticles, which are hard to defineand even harder to observe

accurately, and instead concentratedon the average displacement of theparticle from a starting position overtime. This quantity is much easier toobserve through a microscope thantrying to track the incessant zig-zagging itself.

Einstein’s theoretical work, andsubsequent experimentalmeasurements of Brownian motionthat confirmed the theory, providedsome of the first real evidence forthe existence of atoms andmolecules. As a bonus, because themathematical treatment relatedBrownian motion to physicalparameters and constants,Einstein’s theory provided a newway to measure Avogadro’snumber, too!

So Albert Einstein’s second paper of1905 gave a real boost to thestatistical treatment of thermo-dynamics and, through the lens of amicroscope, showed the world someof the first real evidence for theexistence of atoms and molecules –their microscopic thermal jigglingmanifest as the macroscopicmeanderings of Brownian motion.

132Opposite: (Fig 2) Five different ways ofarranging a repeating pattern (five 2-DBravais lattices).

Building in the NanoworldProfessor David Cockayne FRS

IntroductionN A N O T E C H N O L O G Y I S about building on the nanometre scale. And just asit is important to be able to see what you are building if you are abricklayer (would you hire a blind bricklayer?), so it is important to beable to see if we want to build in the nanoworld. And one of the mainreasons that the nanoworld is of such great research interest nowadaysis because of the many new instruments we have for seeing in thenanoworld, such as atom probes and atomic force microscopes and theelectron microscopes that we discussed in the earlier chapter.

133

134

WAV

ES O

F TH

E FU

TURE

But nature has been building at this level foran aeon, and so the best approach is tolearn from what can be found there. At the

simplest level, this involves understanding howatoms pack together, both in the crystalline andin the amorphous state. From there we canconsider how systems assemble at the atomicand molecular level. And, because the surfaces ofobjects take on an even more important role atthe nanoscale than they do in larger systems, weconsider the structure of interfaces and how tobuild them to control properties. From these kindsof observations, we can use the knowledgegained to build artificial structures, such asquantum dot arrays, nanowires and quantumcomputers.

Atomic packingHow is the material world constructed? We willstart with a simple system and work towards themore complex. Figure 1 is an electron micrographof gold, in which we see the atoms tightly packedtogether in a regular array. The atoms are likepeople – they like to snuggle up to each other.That’s because atoms interact through the forcesthat they exert on each other. In the simplestterms, these are (i) an attractive force due to theelectrostatic attraction of the negative chargedistribution, and (ii) a repulsive force between thenuclei. The opposing action of these two forcesresults in the atoms having a preferred distanceapart at which these forces balance, and thisdistance depends upon the atoms involved. Forgold, the preferred distance is 0.29 nm. Cu andPt and Al pack in exactly the same way as Au asin Figure 1, but because these atoms havedifferent atomic structures (different atomicnumbers Z), they have different interatomic forcesand so they have different preferred distances totheir nearest neighbour – 0.29 nm for Cu, 0.27nm for Pt, and 0.28 nm for Al.

Because the atoms have a preferred distance ofseparation, when a large number come togetherthey arrange themselves in a repeating pattern,as we see in Figure 1. This raises the question ofhow many possible repeating patterns there are.

Let us first consider this question in twodimensions (we will consider three dimensions

later). Next time you are visiting a shop that sellswallpaper, look to see how many different designsthere are. I don’t mean whether the paper haspictures of roses or horses or cars, but how thepattern repeats itself. It will be very surprising ifyou find any more than the five different patternrepeats shown in Figure 2 (shown at the start ofthis chapter). Each of these patterns is composedof a set of objects (the motif), repeated on aregular grid of points called the lattice. You canhave an infinite number of different motifs –horses, flowers, houses, squiggles, and in red,blue, green – so that there are an infinite numberof different wallpapers (which is why it takes solong to choose), but I doubt that you will find anyother lattice than one of the five shown. We writethis arrangement mathematically as (motif*lattice)where * is a mathematical operation calledconvolution. You can see that the pattern ofFigure 1 is (oblong) * (one atom). We can refer tothe repeating unit as the unit cell.

Atoms pack in a regular array not only in two, butalso in three dimensions; and in three dimensionsthere are not 5 but 14 ways of packing identicalobjects together in a repeating pattern – theseare called the 14 Bravais lattices. There is aninfinite number of 3D motifs, so that the possiblenumber of different crystals = (motif )*(Bravaislattice) is infinite.

In 3D, gold has the ‘cubic F’ Bravais lattice with amotif of one atom (Figure 3), as do Cu and Pt andAg, although the sizes of their unit cells areslightly different because the atoms have different

Fig 1: High resolution image of crystalline gold showing theatomic structure.

135

BUIL

DIN

GIN

THE

NAN

OW

ORL

D

interatomic forces. And since they all have thissame structure, it is not too difficult to replace anAu atom by a Pt atom without causing too muchstress for the system. (There will be some stress,because the Pt atom will not be at its optimumdistance from the Au atoms, since it is forced tohave the Au-Au rather than the Au-Pt distance.)So Pt and Cu are commonly used in Au forjewellery (wedding rings, for example), since Au istoo soft on its own.

So crystalline materials are very common innature – the sand on the sea shore is crystallineSiO2, common salt (NaCl) is crystalline, as arealmost all the metals you know, and as are mostsemiconductors in your cell phones and watches.And all of them have one of the 14 Bravaislattices as the building block, repeated over andover to form a crystal. If a crystal grows slowly,starting at one point and gradually expanding involume, then very large single crystals can begrown (for instance, precious diamonds grownover a long time in the earth’s crust). But usuallystructures grow from many sites at once, formingsmall crystalline grains that meet at grainboundaries. Consequently a large block of thematerial might be composed of many smallcrystal grains (an example is shown in Figure 4).

Nowadays, because we can see atoms, we caninvestigate the structure of these materials byforming images with a variety of microscopes,including scanning tunnelling microscopes,electron microscopes and atom probes. Butbecause they are crystals, we can alsoinvestigate their structure by scattering waves(such as neutrons, electrons, X-rays) from themand forming diffraction patterns, as we discussedin the previous chapter. Because the crystalshave a repeating pattern, the diffraction patternwill be an array of spots, as we saw in the firstchapter. Electron or X-ray or neutron diffractionstudies involve the analysis of these diffractionpatterns to determine both the motif and thelattice of the structure.

The diffraction pattern we obtain depends uponthe direction of the incident ray relative to theorientation of the crystal, and so patternscollected with different incident beam directionscarry different information about the crystal.Consequently we can use a set of differentdiffraction patterns, collected for different incidentray directions, to increase the information and soobtain a more detailed analysis of the crystalstructure. Figure 5 shows the resultingconvergent beam electron diffraction pattern, inwhich each diffraction spot is converted into adisc, equivalent points in each disk correspondingto a different direction of the incident electronbeam.

Fig 4: Crystalline grainsin silicon showing theatomic structure.

Fig 3: The cubic F Bravais lattice.

Fig 5: Convergent beam electron diffraction pattern.

136

WAV

ES O

F TH

E FU

TURE

Crystal deformationAtoms arrange themselves into a crystal becauseit is the lowest energy configuration they canassume. (Indeed we can try to predict theirstructure by studying which model system hasthe lowest energy.) Consequently to change thecrystal from this state requires us to do somework – we must stretch or even break theinteratomic bonds. To move the crystal from oneenergy state to another requires work toovercome an energy barrier. Consequently theperfect array of atoms in a crystal makes itdifficult to deform (do you think a leaf or your skinare crystalline?).

So we might wonder how it is that we can, forexample, bend or stretch a piece of metal. Sincethe structure of a metal is repeated layers ofatoms like those seen in Figure 1, it might bethought that the deformation could occur bysliding some of the layers of atoms over oneanother, like sliding the cards in a pack over eachother. But this would require that all the bondsbetween the atoms on the two layers break at thesame time, and an estimate of the force to dothat shows that it is far greater than what isrequired in practice.

Figure 6 shows what actually happens: the bondsbreak along a line (into the page) in the crystal,with all the deformation concentrated along thatline. All the atoms on the left hand side haveslipped sideways by one repeat unit, and, as thedislocation line moves to the right, the regionwhich has slipped gradually increases, untileventually the deformed region reaches the righthand side, and the entire set of atoms above theslip plane has moved to the right relative to theatoms below the slip plane. We call this localisedregion of deformation the dislocation line, anddeformation proceeds by the dislocation linemoving across the crystal from one side to theother. Since each dislocation causes a movementof only about one interatomic distance (about 0.2nm), a deformation of 1 mm involves the motionof about 106 dislocations.

We can see these dislocations by using acombination of diffraction and imaging in theelectron microscope. As we saw in the firstchapter, if we have a periodic object (we looked

at the two cases of a copper grid and a crystal),then scattering occurs in well-defined directions(given by the Bragg equation). If we shineelectrons onto a crystal which contains adislocation, then the amount of scattering thatoccurs into a particular direction will be differentfor the region where there is a dislocation,compared to the region where there isn’t adislocation (Figure 7), because of the atomdisorganisation near the dislocation line.

Fig 7: The formation of an image of a dislocation by thescattering of electrons.

Fig 6: The motion of a dislocation through a crystal.

If we then prevent all electrons from reaching theimage except for those which are scattered intothis particular direction, the dislocation willappear as a dark line, as in Figure 8. We can

137

BUIL

DIN

GIN

THE

NAN

OW

ORL

D

deliberately hinder the movement of thedislocations by putting ‘boulders’ in their way, sothat it then would become more difficult todeform the crystal. This is the principle ofprecipitation hardening, in which the ‘boulders’are nano-sized precipitates. On the other hand, ifthere are sufficient dislocations travelling indifferent directions through the crystal, they canbecome entangled and restrict each other’smovement, and the result is called workhardening. Understanding the role thatdislocations play in controlling the strength andproperties of materials has been one of the majoradvances in materials science over the past 50 years.

[FIGURE 8 HERE]

AllotropesThe most well-known crystal structure of carbonis diamond, which can be described as (cubicF)*(C-C). That is, it has the same Bravais latticeas Au (cubic F), but the motif is a pair of C atomsinstead of one Au atom. Carbon crystallises inseveral other forms – one is graphite, in the formof sheets of C atoms arranged in a hexagonalarray, with these sheets held together by weak(Van der Waals) forces. If we take a sheet ofgraphite, it can be rolled into a tube and joined,with no discontinuities. These carbon nanotubeswere first discovered by Iijima in the electronmicroscope in 1991. They are very strong, and socan be used as strengthening fibres withincomposites; and they can be used as containersinside which thin crystal wires can be grown, withproperties which are different to the bulk, orwithin which individual molecules can be packed.If a piece with 60 carbon atoms is cut from asingle graphite sheet, it can be bent into afootball shaped molecule, C60, known as a buckyball (after the architect Buckminster Fuller whohad previously built large structures with thisshape). Indeed the bucky ball is small enough topack inside a nanotube (see Figure 9), and anatom is small enough to place inside a buckyball. So individual atoms can be held insidebucky balls, which in turn are packed within ananotube. Scientists are investigating whetherthis structure might be the basis for the nextgeneration of (quantum) computers.

Fig 9: Models of a carbonnanotube and a bucky ball, andelectron microscope image ofbucky balls inside a singlewalled carbon nanotube.

Fig 8: An electron microscope image of dislocations in silicon.

138

WAV

ES O

F TH

E FU

TURE

Quantum dots Quantum dots are formed by depositing a layer ofone material (e.g. Ge) onto the surface of anothermaterial with a very similar structure (e.g. Si).These two materials have the same Bravaislattice and the same motif. Their only differenceis that they have different atoms (Si or Ge), andthis difference results in them having differentunit cell sizes.

When Ge is deposited as a thin film onto a Sicrystal surface, this size difference can beaccommodated by the film of Ge beingconstrained into adopting the unit cell size of Si.We call this accommodation epitaxial growth. Asthe thickness of the Ge film is increased,eventually the strain in the film becomes toogreat to support, and the surface film collapses.One way it can collapse is for the Ge to form intosmall islands or dots, an example of which isseen in Figure 10 – in this case, an InAs dot on GaAs.

When crystals have such a small volume (say100x100x100 atoms), they can have unusualproperties not found in the bulk. For example, ifwe have a small crystal of Ge into which we injectan electron, the electron can have certain well-defined energies that depend upon the size andcomposition of the crystal. These energies arecalled quantised energies (from the Latin wordquantus, ‘how much’), and the object is called aquantum dot. Because these quantised energiesremind us of the quantised energies of electronsin the shells of an atom, quantum dots aresometimes referred to as giant atoms. And just asatoms emit X-rays of a given wavelength whenthey become unstable, as we saw in the first

chapter, so to when the electron in the quantumdot makes a transition between the quantisedenergy levels, it emits light of a well-definedwavelength. In this way intense, efficient sourcesof light can be produced for use not only as laserpointers and fluorescent dies, but as replacementsfor brake lights on cars and traffic lights aroundcities of the world. The structure of these quantumdots – their shape and composition profile –determines their quantised states, and so it isimportant not only to be able to investigate whatstructures are obtained for a given method ofgrowth, but also to be able to predict whatstructure will be obtained from a particular set ofexperimental conditions.

Modelling and simulationThis predictive aspect of materials science is oneof the most important tools for progress indeveloping new materials, because it allows us todesign materials with specific properties withouthaving to first make samples of them. To do thisrequires a close interaction between modellingand experimentation. For example, in the case ofquantum dots, a model atomic structure can beset up in the computer, with the computerprogramme describing the forces acting betweenthe atoms, either using quantum mechanics or asempirical equations. The total energy of thestructure can be calculated, and then an iterativeprocess of moving atoms one by one can becarried out, at each step recalculating the energyof the structure to determine whether or not themove has taken the system to a lower energy. Onthe assumption that the physical structure will bestable when the energy is a minimum, thepredicted stable structure can be obtained.

Fig 11: A model of the distribution of Ge in a Ge-Siquantum dot.

Fig 10: An InAs quantum dot on a GaAs surface.

139

BUIL

DIN

GIN

THE

NAN

OW

ORL

D

Figure 11 shows the result of such a procedurefor a quantum dot composed of Ge and Si, inwhich the Ge and Si atoms were initially randomlydistributed. The modelling has resulted in aredistribution of Ge and Si atoms, with Ge (bluecolour) concentrating near the surface. This resultis in agreement with the experimental resultsobtained from energy-loss and X-ray mapping(discussed in the previous chapter), and theagreement gives us confidence in the modelling.These calculations are extremely time consuming,even for the most powerful computers, becauseof the large number of atoms involved inmodelling a realistic structure.

A second important role of computers is as anaid in the interpretation of images and diffractionpatterns obtained from the electron microscope.Although we have not mentioned it to this point, amajor difference between electron scattering andneutron and X-ray scattering is that electrons arescattered orders of magnitude more strongly thanX-rays or neutrons (we say that atoms have alarger cross section for electron scattering). Thishas the advantage for electrons that, whenstudying small objects, the scattered signal isgreater for electrons; but it has the disadvantagethat the electrons, in passing the sample, willusually be scattered many times. This multiplescattering often makes it difficult to interpretimages and diffraction patterns straightforwardly.Fortunately we have a thorough understanding ofhow this multiple scattering occurs, and so wecan describe it mathematically and writecomputer programmes to simulate it. In this way,model structures can be set up in the computer,and simulated images and diffraction patternscan be obtained for comparison withexperimental images. Then a choice betweenalternative model structures can be made bycomparing the simulated image or diffractionpattern from each of them with the experimental

image or diffraction pattern. Figure 12 shows anexample of a comparison of experimental imageof a dislocation with images calculated fordifferent models of the dislocation.

Amorphous materialsAs we have seen, atoms like to come together ascrystals. However there are many situationswhere the conditions for forming crystals do notexist, or where the crystalline structure isdestroyed. For example, in DVDs, a laser pulse isused to blast the surface of a crystal, creating ananovolume of amorphous material, which thenserves as the information bit. Other examples arethin amorphous films used as hard coatings onsurfaces, for example on high speed drills.

It might be thought that, with microscopes thatcan resolve atoms, we could study thesematerials atom by atom. But there are twoshortcomings to this approach. Firstly, because ofthe disorganised arrangement of the atoms, anatom-by-atom description is of little use – astatistical description is needed. And secondly, itis very difficult to interpret the images of thesematerials because the individual images of all theatoms superimpose on each other and obscureany detail.

So let us consider what we might do withdiffraction. We start by considering scattering ofelectrons by an individual atom. Many of theelectrons are scattered in the forward direction,

Fig 13: The electron diffraction pattern of amorphous carbon.Fig 12: Experimental (left) and simulated (right) images of adislocation in brass.

140

WAV

ES O

F TH

E FU

TURE

while some are scattered through large angles asin Figure 13. If we have a large number of atomsscattering independently, the diffraction patternlooks very much the same, but stronger becauseof the larger number of atoms.But if we look closely at the diffraction patternfrom this array of atoms, we see (Figure 14) thatthere are oscillations which we might not expectfrom the diffraction from an array of randomlydistributed atoms. These oscillations are telling usthat the carbon atoms are not arranged randomly,but have some degree of organisation, arisingfrom the interatomic forces that gave rise to thecrystals discussed earlier. A mathematicalanalysis of the diffraction pattern (a Fouriertransformation) allows us to display this atomicarrangement as a ‘distribution function’ (Figure15). We see that even in this amorphous material,the carbon atoms like to be 0.17 nm apart and0.25 nm apart, but none like to be 0.20nm apart.Why is this? We can understand it if the atomsare in a crystal – specifically, diamond. For indiamond the atoms are arranged in a very regulararray, as we have discussed earlier. And indiamond, the nearest neighbours are at 0.17 nm,

and the second-nearest neighbours are at 0.25 nm. But why are these same distancesobserved in the amorphous state?

The reason is that in diamond, each carbon atomis at the centre of a tetrahedron of four othercarbon atoms, and this arrangement isexceedingly strong. And this tetrahedron of atomssurvives, with some distortion, into theamorphous state of carbon. Adjacent connectedtetrahedra can rotate about their common bond(2-5 in Figure 16), and this random rotation,when repeated for all bonds, results in theamorphous structure. During this rotation, the firstnearest distance (1-2 in Figure 16), and thesecond nearest distance (1-5) of the tetrahedronare retained, and appear in the distributionfunction of the amorphous state, but the thirdnearest neighbour distance (1-8) is not retainedand does not appear.

The amorphous state is of increasingtechnological importance as it plays a central role

Fig 16: The tetrahedral arrangement of atoms in diamond,showing the rotation around bonds that occurs inamorphous diamond.

Fig 15: The radial distribution function of amorphouscarbon; the vertical lines mark the nearest neighbourdistances in crystalline diamond.

Fig 14: The intensity profile of Figure 13, showingoscillations due to structure in the amorphous carbon.

141

BUIL

DIN

GIN

THE

NAN

OW

ORL

D

in materials such as optical fibres, DVDs, solarcollectors and ceramics, and experiments of thekind described allow us to study these materialsin detail.

An interesting case results from throwing buckyballs into a heap and scattering electrons fromthem. The distribution function that results fromthe diffraction pattern (Figure 17) has numerousmaxima on the length scale of 0.2 to 1 nm,which arise from the various distances betweenatoms on the surface of the bucky ball. Using thisdata about the interatomic distances within theball, we can build a model structure to agree withthe data. But in Figure 17 there is also a verylarge, broad maximum at 1.1nm which is largerthan the size of a single bucky ball; so it can’tarise from interatomic distances within a ball.This feature shows us that the balls themselveslike to arrange themselves into one of the Bravaislattices – a cubic lattice – with a repeat distanceof 1.1nm. So again we see nanostructures self-organising into periodic structures.

The meeting of the crystalline andamorphous nanoworldsAs we can understand from Figure 4, interfacesbetween crystals occur frequently in technologicalmaterials, and we consider the case of siliconnitride as an example. Silicon nitride ceramics areused in many important applications includingbrake linings and bearings in cars, because oftheir strength and toughness. One of their mostimportant features is that cracks do not easilypass through the material. The way this isachieved can be understood from Figure 18.

The ceramic is made up of a high density of smallneedle-like silicon nitride crystals, separated by athin layer of amorphous glass. By a method thatwe shall discuss, the interface between the glassand the crystal is made weaker than both theglass and the crystal. So, if a crack forms at theedge of the glass, it will follow a path such as thatshown by the yellow line. The longer the path thebetter, because, to advance, the crack must breakthe interatomic bonds between the glass and thecrystal, and each broken bond absorbs energy.Eventually, if the crack doesn’t have enoughenergy, it comes to a halt. So the aim is toproduce long crystals with a relatively weak

Fig 19: The atomic structure of the interface of silicon nitridegrain and an intergranular film; the bright dots are Lu atomswhich act as a “zipper” to guide the path of a crack.

Fig 18: The path of a crack in a silicon nitride ceramic.

Fig 17: The experimental radial distribution function ofbucky balls.

142

WAV

ES O

F TH

E FU

TURE

interface with the glass. Rare earth elements suchas lanthanum or lutetium are introduced for thepurpose, and the high-resolution image of Figure19 shows how they act. They ‘paint’ the surface ofthe silicon nitride grain, firstly stopping it growingsideways, so that it forms a long crystal, and thenacting as a ‘zipper’, guiding any crack along thecrystal surface.

Building with atoms The ability to manipulate the arrangement of atomsusing a variety of tools is becoming increasinglysophisticated. One of the most amazingdemonstrations is the use of a scanning tunnellingmicroscope to manipulate single atoms to formgeometric shapes such as quantum corrals andletters of the alphabet. While such manipulationsare fascinating, they are likely tobe technologically useful only ifthey can be replicated on a largescale. Obvious targets are thegrowth of bone and skin, so thatrepair can be carried out followingdamage to the human body. Butothers include the self-assemblyof nanoparticles to form arrays forquantum computing or formemory and recording devices.

An example is the growth ofarrays of quantum dots. We havealready discussed the growth of quantum dotsand their ability to produce quantised lightemissions. If a large number of dots can bearranged to pack tightly, so that their density ishigh, then bright light sources could result. Oneway to pre-determine where the quantum dotsmight grow on a surface is to bury dislocations(which we discussed earlier) just below thesurface of a crystal. Each dislocation results in aslight disturbance to the positions of the atoms atthe crystal surface, and this disturbance causesthe quantum dot to form above the dislocationrather than elsewhere on the surface. By having aregular array of dislocations beneath the surface,a regular array of quantum dots can be formed.Frances Ross and her colleagues at IBM havesuccessfully demonstrated this approach to theself assembly of quantum dots, and an exampleof their work is seen in Figure 20.

The assembly of individualatoms into small halide crystalshas recently been achieved byinserting them into carbonnanotubes, which act as tiny testtubes. An example is shown inFigure 21, where a wire of KI,one unit cell wide, has beenassembled within the tube. KI isnormally cubic in the bulk, butfrom the figure we see that theunit cell has different sizes in thedirections along and across thewire. This is an example of how

the structure and properties of materials arechanged when they form at the nano scale.

Nature itself has developed many methods forself-assembly, and we can often learn techniquesfrom the natural world. One example of nano

Fig 21: A KI crystal grown within a carbon nanotube.

Fig 20: Ge quantum dots on a Si substrate. The quantumdots have grown in rows above dislocations which are inthe Si.

“The assembly ofindividual atoms intosmall halide crystals hasrecently been achieved byinserting them intocarbon nanotubes, whichact as tiny test tubes ...”

143

BUIL

DIN

GIN

THE

NAN

OW

ORL

D

objects serving a purpose in nature is the smallmagnetite crystals found in certain bacterium.These small magnetic crystals form a chain as abackbone to the bacterium, as seen in Figure 22.These crystals are referred to as magnetosomes.They are usually arranged in a linear chain insideeach bacterium and they result in a permanentmagnetic moment to the cell, which results in ithaving the ability to align itself and move parallelto the local geomagnetic field lines. It is thoughtthat this benefits the bacteria by making themable to more efficiently locate the optimumoxygen concentrations in the water columns ofthe sea where they are found.

ConclusionThere are many fascinating objects in thenanoworld which occur naturally, and which canbe seen only with powerful microscopes. This is areal part of the physical world, so far relativelyunknown to us because we have only recentlydeveloped the tools needed to explore in thisworld. Many of the objects we find, such ascarbon nanotrees and nano-onions observed tooccur naturally under certain carbon growthconditions, have no obvious use to date. Butwhether or not a discovery proves immediatelyuseful, the important point to remember is that,for the nanoworld, we are in the age ofexploration and you can be the explorers.

Fig 22: Magnetic particles arranged as a backbone within asingle cell of Magnetospirillum magnetotacticum.

144

WAV

ES O

F TH

E FU

TURE

neutrons, the thorium-232 wouldturn into uranium-233. So hebought thousands of gas mantlesand turned them into thorium ashwith a very hot gas flame. How didhe purify the thorium? Simple – hebought a few thousand dollarsworth of lithium batteries, cut themopen, and did some simplechemistry to concentrate thethorium. But alas, the effort waswasted. His neutron gun didn’thave enough grunt to turn thorium-232 into uranium-233.

Time for Plan B. Radium deliversheaps of ∝ -particles, and he hadbeen told if you blast these ∝ -particles onto beryllium, you getenormous numbers of neutrons.But how could he get someradium? Well, until the late 1960s,the glow-in-the-dark faces ofclocks, and car and airplanedashboard instruments glowedbecause they were painted withradium. So he started the slowprocess of haunting junk andantique shops, surreptitiouslychipping off the glowing radium.But one day, he got lucky when hisGeiger Counter went off its brain.He bought the clock for $20, andinside, found a complete vial ofradium paint conveniently leftbehind.

At the age of ten he was given abook called The Golden Book ofChemistry Experiments. Somethingclicked, and by the age of 12 hehad mastered his father’sUniversity-level chemistry books,and by 14, he had madenitroglycerine. His father thoughtthat David needed a stabilisinginfluence, so he advised him to tryfor the goal of Eagle Scout – whichneeds a total of 21 Merit Badges.Some Merit Badges (likecitizenship, first aid and personalmanagement) are compulsory,while some (from AmericanBusiness to Woodchuck) arechosen by the scout. David chosethe Merit Badge in Atomic Energy.To get this badge, you have toknow about nuclear fission, knowwho the important people in thehistory of Atomic Energy were, andmake a few models of some atoms,and other stuff. David built a modelof a nuclear reactor with some tincans, drinking straws and rubberbands, and earned his Merit Badgein Atomic Energy on May 10, 1991(when he was 14 years and 7 months old). Then he decided toaim higher.

Atoms have a core of positively-charged protons, and neutralneutrons. Some of the biggeratoms (like uranium, for example)have unstable cores. If a neutron

hits this core, it splits into twosmaller atoms and a neutron or twoand also gives off a huge amountof energy.

So David started off by making aneutron gun. He pretended to be aPhysics Lecturer, and got lots ofprofessional help from industrialcompanies, the American NuclearSociety and the Nuclear RegulatoryCommission. He found out that hecould get radioactive Americium-241 from household smokedetectors – so he bought 100broken ones at a dollar each. Thefriendly customer-servicesrepresentative told him exactlywhere the Americium-241 was, andhow to remove it from its inert goldmatrix. He then put his tiny pile ofAmericium-241 inside a hollowlead block, and drilled a small holein it. As Americium decays, it givesoff a-particles. When a-particles hitaluminium, the aluminium gives offneutrons. So he put a strip ofaluminium in front of the hole in thelead block where the a-particlescame out, and bingo – he had aneutron gun.

He had found out that the clothmantles in gas lanterns are coveredwith thorium-232 (because thoriumis very resistant to hightemperatures). He also knew that ifyou hit thorium-232 with enough

Radioactive Boy ScoutBy Dr Karl Kruszelnicki

MOST KIDS HAVE some kind of hobby – a sport, collectingstamps or computer games. But David Hahn, who lived inCommerce Township in Michigan, about 40 km out of Detroit,had a scientific hobby – chemistry. And so he tried to build anuclear reactor.

145

RAD

IOAC

TIVE

BOY

SCO

UT

FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

Illustration courtesy Adam Yazxhi

And David? Well, while he was awhiz at science, he never wasmuch good with maths and English.So today, he’s a juniorsailor/deckhand on the aircraftcarrier, USS Enterprise, which has 8 nuclear reactors.

And if George Bush ever needs tocall in the Heavy Artillery, maybe heshould forget the SEALs and theSAS, and call in the Boy Scouts todo their Bob-A-Job...

So he rigged up a more powerfulneutron gun with a hollow lead blockwith a hole, his precious radiuminside, and some beryllium to get hitby the ∝ -particles and give offneutrons. What did he use for atarget? Some uranium ore he gotfrom a friendly supplier. But failureagain. The neutrons were moving toofast (about 27 million kilometres perhour) and just zipped through theuranium. So he slowed them downto about 8,000 km/h by runningthem through tritium (which hepainstakingly scraped off modernglow-in-the-dark gun and bowsights) – and the uranium ore gotmore radioactive.

By this time, David Hahn was 17,and he decided to stop foolingaround. He mixed his radium withhis americium and aluminium,wrapped it in aluminium foil, andthen wrapped the whole mess inhis thorium and uranium – ofcourse, all held together with gaffertape. Finally he had success – thebizarre ball got more radioactiveevery day. Perhaps too muchsuccess – he could pick up theradioactivity 5 houses away. Hepanicked, and began to dismantlehis creation.

At 2.40 am on the 31st of August,1994, the local police were calledbecause a young man was doingsomething suspicious near a car.David told the police to be carefulof his toolbox, because it wasradioactive. Soon some men inventilated white moon suits werechopping up his radioactive shedwith chainsaws, and stuffing theparts into thirty nine 200-litresealed drums which they tookaway to a nuclear waste repository.The clean-up cost about $120,000– but it did protect the 40,000nearby inhabitants from harm.

PROFESSOR PETER ROBINSONreceived his PhD intheoretical physics fromthe University of Sydneyin 1987, then held apostdoc at the Universityof Colorado at Boulderuntil 1990. He thenreturned to Australia,joining the permanentstaff of the School ofPhysics at the Universityof Sydney in 1994, and

obtaining a chair in 2000. He iscurrently an Australian ResearchCouncil Federation Fellow working ontopics including brain dynamics,space plasma physics, wave theoryand self-organized criticality.

146

Understanding Brain DynamicsProfessor Peter Robinson

IntroductionO N E O F T H E K E Y questions of neuroscience is how to relate brain activityto what that activity is accomplishing. In other words, when brain cells(neurons) are active, what are they doing? Before one can attempt toanswer this question, one must be able to measure activity, and torelate activity to the stimuli that cause it.

147

Figure 1: The brain sectioned along a vertical plane running approximately from ear to ear. The gray matter is seen as a thinlayer on the outside, immediately surrounding the white matter. The location of the thalamus is also shown.

148

WAV

ES O

F TH

E FU

TURE

The work described here attempts to relatestimuli to brain activity, and brain activity tomeasurements of that activity, in ways that

enable measurements to pin down details of theactivity and underlying stimuli. These insights willhelp in understanding brain function (informationprocessing, etc.) and malfunction (e.g., disease).

In this chapter, I will first briefly describe some ofthe ways in which brain activity can bemeasured. Then I will discuss how theconnections between stimuli, activity, andmeasurements can be modelled theoretically.Some of the key predictions, and theirrelationships to experiment, will be reviewed,followed by a short discussion of commercialapplications currently under way.

Measuring Brain Activity One measure of brain activity is electrical signalsof neurons. When neurons become active, smallcurrents flow briefly between their interior andexterior. These currents are associated withelectric potential (voltage) changes, which takethe form of spikes on a voltage-time plot. Whenaveraged over many neurons these voltages canbe detected by electrodes placed at the surfaceof the brain or on the scalp.

The first detection of brain electrical activity wasreported in 1875 by a British physiologist,Richard Caton, who used a galvanometer tomeasure electrical activity in the brains of rabbitsand monkeys. Little more was done until themid-1920s when Berger showed that similaractivity was present in humans, and depended onmental activity. Since then, recordings of electricalactivity from the scalp – electroencephalograms,or EEGs – have been widely used to measurebrain activity and its changes with state ofarousal (e.g. awake, asleep), brain function(information processing, attention, type of sensoryinput involved), and disease (e.g. Alzheimer’s,epilepsy, attention deficit disorder).

It has long been recognized that the electricalsignals seen at the scalp are ultimately producedby brain activity at the neural level. What hasbeen lacking, however, is a quantitativeconnection between these two levels. In the

absence of a proper theory that enablesmeasurements to be predicted (i.e. calculated)starting from the cellular level, all that can befound is a (very large!) set of associationsbetween experimental situations and the resultingmeasurements. The primary aim of the Universityof Sydney group is to enable detailed predictionsto be made and compared with EEG experiments.

Another type of measurement of brain activity,one that has become available in the last 20years, is provided by functional magneticresonance imaging, or fMRI. This techniquemeasures tiny differences in the magneticproperties of oxygen atoms, between those withinthe oxygen-carrying blood chemical haemoglobin,and those that are detached from haemoglobin.Because of the slight differences in theirenvironment, oxygen atoms behave slightlydifferently in these two situations, therebyenabling the amounts of haemoglobin with andwithout oxygen to be measured in various partsof the brain. The amount of deoxygenatedhaemoglobin is highest in areas that are active,where oxygen is being used, thereby enablingbrain activity to be mapped.

Again, in the case of fMRI, our aim was to makethe connection between the stimuli, activity, andmeasurements of activity sufficiently precise thatthey can be used to obtain new insights into brain function.

Modeling The BrainThe largest part of the human brain is thecerebral cortex, seen in Figure 1. It is composedof a thin folded layer (2 to 3 mm thick) of ‘greymatter’, which is involved in high-levelinformation processing. This is underlaid by‘white matter’, which comprises mainly bundlesof neurons that interconnect areas of greymatter. Because it is closest to the skull, corticalgrey matter gives rise to most of the observableelectroencephalographic signals seen at the scalp.

Another key brain structure is the thalamus (seeFigure 1), which relays all sensory informationexcept smell to the cortex. It also receives

150

WAV

ES O

F TH

E FU

TURE

Overall, we obtain a set of equations that can besolved to predict brain properties in terms of amodest number of physiologically measurableparameters, such as ranges and signal speeds ofaxons, numbers, speeds, and strengths ofsynapses, and time delays in travelling betweenneural populations.

Results Once equations for all the above features arewritten down, standard analytic andcomputational techniques can be used to solvethem and extract predictions for the types ofbehaviour that the brain should exhibit whenstimulated. Some of these are discussed in thefollowing subsections.

Steady Brain ActivityThe first question we ask is whether the brain canshow stable, steady-state activity on average if itreceives a constant average level of externalstimulation. Solution of our equations shows thatthere are three possible steady states, two ofwhich are stable. Of these two, one correspondsto all the neurons firing (spiking) flat-out, whichrelates to some kind of epileptic seizure. The otherhas average firing rates of a few spikes persecond per neuron, which turns out to be in goodagreement with what is actually observed whenelectrodes are inserted into living brains.

EEG SpectraSmall departures from steady states can bestudied by assuming that they are ‘linear’ – thatis, twice the stimulus produces twice the changein the brain. If this is done, a large number ofproperties of brain activity can be calculated interms of the ‘linear transfer function’, which is ameasure of the ratio of activity change tostimulus.

Because the inputs to the brain are extremelycomplex functions of space and time, weapproximate them as having an equal mix of allpossible spatial and temporal scales, at least to afirst approximation. If this is done, we can use thetransfer function to calculate the spatial andtemporal scales of the brain’s response to them.In particular, we can calculate the amount of eachfrequency of response is present – the ‘powerspectrum.’ Figure 4 shows the EEG powerspectrum obtained from a resting adult subject,overlaid with a predicted curve from our model.We see prominent peaks near 10 and 20 Hz,which are the so-called ‘alpha’ and ‘beta’rhythms, first discovered by Berger in the 1920s.At high frequencies, there is a rapid fall-off inpower, while power climbs steeply at lowfrequencies, before levelling off below 1 Hz.

Figure 4: Example power spectrum (solid) from a typicaladult subject in the relaxed, eyes-closed state. The dottedcurve is a fit of our theoretical predictions to these data.

Figure 3: Schematic of connections (indicated by arrows)between the cortex and thalamus, showing the relay andreticular nuclei of the thalamus separately. External stimulienter at the bottom.

151

UND

ERST

AND

ING

BRA

IN D

YNAM

ICS

The good match between model predictions anddata seen in Figure 4 supports the validity of themodel and means that we can try to interpret theabove features in terms of the physics in themodel. In particular, we note the following:(i) The alpha and beta peaks are due to

resonances in the corticothalamic closed loops,mentioned above. These are similar to violinstrings, in having only certain frequencies atwhich they can oscillate.

(ii) The fast high-frequency fall-off is due to therelatively slow rate at which dendrites andsynapses can respond to incoming spikes –which means higher frequencies are blocked.

(iii) The rapid rise at low frequencies is a sign thatthe brain is close to instability, where it wouldcross into a seizure of some sort. Thefrequency at which the spectrum levels off isa measure of the degree of stability; the lowerthis frequency, the less stable the brain. If ourbrains were too stable, we would not be ableto be flexible in our responses to complexinputs – but they mustn’t become unstable.Hence, they operate close to the edge ofstability.

Evoked Responses to StimuliThe transfer function also enables us to calculatethe response of the brain to a short stimulus,such as a flash of light or a sudden tone –- theso-called ‘evoked response.’ Figure 5 shows a

comparison between experimentally measuredand theoretically predicted evoked responses toan auditory tone. Again, there is goodagreement, and the parameters for whichagreement are found turn out to be very similarto those found for EEG spectra. This is notsurprising, since both phenomena are producedby the same brains. What is perhaps moresurprising is that these two types of phenomenahave been studied almost independently for manyyears, in the absence of a theory that links themtogether.

States of ArousalIt has long been known that the EEG shows majordifferences between sleep, wake, and otherstates of arousal. Figure 6 compares some of ourmodel predictions for EEG time variations withcorresponding experimental results. Again, goodagreement is found, based on the transferfunction.

Figure 6: Left column: Model time series representative of(a) alert, eyes-open, (b) relaxed, eyes-closed, (c) normalsleep, and (d) deep sleep. Right Column: Correspondingtime series from human subjects (Nunez, 1995). Note thatthe vertical scales differ between the various frames.

Figure 5: Experimental (solid) and theoretical (dotted)evoked response potentials in response to an auditorystimulus.

Epileptic SeizuresIf we allow disturbances to become large, thebehaviour of the model is no longer linear –doubling a stimulus could result in increase in theresponse by more (or less) than a factor of two,and may even send the brain off into anoscillatory or chaotic state. We have found thatthe stability of the brain can be plottedapproximately in a three-dimensional space, as

152

WAV

ES O

F TH

E FU

TURE

shown in Figure 7. The x, y and z axes measurethe strengths of cortical feedback on itself,corticothalamic feedback, and feedback withinthe thalamus, respectively.

All stable states of the brain lie within the tent-shaped region in Figure 7, with several specificexamples shown. If the boundaries of the tent arecrossed, the brain either zoomsoff to a sustained high-firing ratestate (some kind of extremeepileptic seizure) or it enters anoscillatory mode. Brain oscillationsare predicted to occur atfrequencies between roughly 3and 10 Hz, and can correspond toseizures if their amplitudesbecome large enough. Figure 8shows an example of a roughly 3Hz oscillation obtained oncrossing the lower left boundaryin Figure 7. The time course ofthis oscillation is very similar tothat seen in ‘petit mal’ seizures,which affect about 1% of childrenat some point. During such aseizure the subject losesconsciousness, but does not collapse or undergoconvulsions. On return of consciousness, after 20to 30 seconds, they may even continue what theywere doing beforehand. We have also found

similarly realistic results for 10 Hz ‘grand mal’seizures that do involve convulsions.

New Measures of Brain ParametersIf we make a prediction from our theory andcompare it with experiment, we usually have toadjust the parameters of the theory slightly to getthe best possible match to data from eachindividual subject. Used in the reverse direction,this provides a new way to measure thoseparameters for that subject. We have used thisapproach to estimate a number of parameters,such as axonal ranges and speeds, and synapticproperties, for the population at large and forindividuals. These estimates have proved to be inagreement with independent ones obtained byphysiologists and anatomists, thereby further

strengthening the basis of our theory.

One powerful potentialapplication of this approach is touse it to measure parameterchanges between states ofarousal, or between differentclinical conditions. EEGs areknown to be very different indifferent arousal and diseasestates, and our method promisesto enable these differences to beinterpreted in terms of changesin the underlying physiology andanatomy. Initial work shows thatthis approach has great promisefor opening up a new, non-invasive window on these subtle

brain changes. Corrective changes induced bymedications can also be monitored objectively tooptimise drug dose and get the best outcomeswhile avoiding overdosing.

Figure 8: Sample time series from the model in a regimecorresponding to a petit mal seizure. This stronglyresembles experimental time series from such seizures.

Figure 7: Brain stability zone. Stable states lie inside thetent-shaped surface. Approximate locations are shown ofalert eyes-open (EO), relaxed eyes-closed (EC), normalsleep (NS), and deep sleep (DS) states, with each statelocated at the top of its bar. The quantities x, y, and z arediscussed in the text.

EEGs are known to bevery different indifferent arousal anddisease states, and ourmethod promises toenable these differencesto be interpreted interms of changes in theunderlying physiology andanatomy.

153

UND

ERST

AND

ING

BRA

IN D

YNAM

ICS

Functional Magnetic Resonance ImagingSo far, I have mentioned only EEG-basedmeasurements. Recently, we have extended thetheory to calculate the change in blood oxygenlevel caused by neural activity. This enables us toestimate the response measured in functionalmagnetic resonance imaging (fMRI) experimentsthat are sensitive to blood oxygenation levels. Weare currently comparing the predictions in detailwith experimental measurements.

Commercial ApplicationsThe ability to measure brain parameters using ourmethods is already being commercialised via aspinoff company, the Brain Resource Company,which was set up in 2001. A database of brainfunction and structure measures on over 4000normal and clinical subjects has been amassed,against which individual subjects’ measurementscan be standardized. This enables deviations fromnormality to be rapidly detected and quantified,and provides data that serve to test and calibratetheoretical efforts such as our own.

Already, a number of significant features of thedata have been predicted and confirmed viamodelling efforts. For example, it has long beenknown that the frequency of the alpha rhythmincreases in childhood, peaks at an age of around20 years, and then declines slightly thereafter. Inour model, this is predicted to be connected witha speeding up, then gradual slowing of conductionvelocities in axons, particularly those linking cortexand thalamus. The physics of the situationimmediately implies that there should be a parallelspeeding up, then slowing down, of evokedresponses, and this proves to be borne out by thedata. Interestingly, this effect is only discernable inenormous databases, such as the one operatedby the BRC, because individual variations obscureit if only a few subjects are examined.

ConclusionRecent physiologically-based investigations ofbrain dynamics have resulted in a model thatincorporates the main relevant features ofcorticothalamic physiology and anatomy inrelatively few parameters, which can all beindependently estimated by physiologists and

anatomists. The predictions of the model providea successful, unified description of a wide rangeof phenomena relating to brain activity, includingthe existence of steady states, EEG spectra,evoked responses, time dependence of EEGs,seizures, arousal changes, fMRI, and otherphenomena.

Fitting the model’s predictions to data provides anon-invasive probe of the physiology andanatomy, yielding parameter values that areconsistent with independent measures, and whichwill enable new objective measures of differencesin brain state due to arousal and disease, forexample.

This work represents an extension of biologicalphysics into a new field where quantitativemethods have been lacking, despite over 130years of effort. A host of questions can beaddressed with the methods now available, and itis possible to adopt a unified approach topreviously unconnected sub-areas ofneuroscience relating to brain activity.

References and Further Reading

Kandel, ER, Schwartz, JH, and Jessell, TM, Principles of

Neural Science, 3rd Ed, Appleton and Lange, Norwalk,

Connecticut, 1991, ISBN 0-8385-8034-3.

Nunez, PL, Neocortical Dynamics and EEG Rhythms, Oxford

Univ. Press, Oxford, 1995, ISBN 0-19-505728-7.

Robinson, PA, Rennie CJ, Rowe, DL, and O’Connor, SC,

Estimation of Multiscale Neurophysiologic Parameters by

Electroencephalographic Means, Hum. Brain Mapp. 23,

53-72, 2004.

154

WAV

ES O

F TH

E FU

TURE

This led to another significant resultin the lead-up to 1905: the resultof the Michelson-Moreleyexperiments. If light travelledthrough the aether, and the aetherformed a background referenceframe, then the earth must also betravelling through the aether as itorbits the Sun each year. Michelsonand Moreley reasoned that thiswould mean light travels atdifferent speeds relative to anobserver on earth, depending onwhether the light was going in thedirection of the Earth’s passagethrough the aether, or at rightangles to it.

Electrodynamics? Moving bodies?What about the famous cosmicspeed limit, c, the velocity of light in avacuum? What about bendable spaceand warpable time? What about themind-bending twin paradox?

It makes sense once youunderstand the context in whichEinstein produced this remarkablepiece of theoretical physics. Twophysical results set the stage forthe revolution in 1905: one fromexperiment, the other theoretical.Both had to do with the nature of light.

The first grew out of Maxwell’s workon the laws of electromagnetism,which were still fresh, and theirastounding success in unifying awide range of diverse phenomenawas still being celebrated acrossthe scientific community. But oneconsequence of the electro-magnetic equations caused someconcern: they predicted electro-magnetic waves with a preciselydefined velocity in the vacuum. Thetrouble was, the theory didn’tspecify what you were supposed tomeasure the velocity relative to.

With every other kind of wave, thespeed is relative to the mediumthrough which the wave travels:soundwaves through air, ocean

waves through water. If the mediumis in motion relative to an observer,this will affect the perceived speedof the wave – think of the circularripples from a rock thrown in ariver, travelling downstream withthe current.

But electromagnetic theory predictsthe speed of light is c, about300,000 km/s – but there’s noexplicit medium, no referenceframe. It was clear, then, that sincelight must travel through something,therefore this something – calledthe aether – must form abackground, an absolute referenceframe against which all othermotion could be compared.

Special Relativity – Einstein’s = magnificent c2reationTHIS IS THE one everyone knows about: Einstein’s theory ofrelativity. So it might come as a surprise that the actual paper,published in June 1905, was titled On the electrodynamics ofmoving bodies (Annalen der Physik 17, pages 891-921).

Einstein’s Miraculous Year /Part 3

155

EIN

STEI

N'S

MIR

ACUL

OUS

YEAR

,PAR

T 3

So they performed experiments todetect the effect of our motionthrough the aether – and despitesome really clever experiments,they found nothing: their resultsshowed no significant change inthe speed of light when measuredin different directions. Which cameas a bit of a surprise at the timeand led many people to propose allsorts of reasons for the null result.

Lorentz and Poincare in particularpushed Newtonian physics as faras it could go with theirmathematical insights. Using theusual ideas of Newtonian dynamics,Lorentz came up with the equationsof time dilation and lengthcontraction, the classic equationsusually attributed to specialrelativity, years before Einstein everwrote them down. Poincare showedthat the ideas helped to explainwhy the effects of our motionthrough the aether weren’t seen bythe Michelson Moreley experiments– in effect, the idea was thatNewton was still correct, but Naturewas being sneaky in ‘hiding’ theeffects of the aether from us.

So if Lorentz and Poincare werethere first, what’s the big dealabout Einstein’s work in 1905? Thedifference is in the underlyingassumptions, and what those implyabout the physical universe. Lorentz

and Poincare were working from aclassical Newtonian point of view,and saw the constancy of thespeed of light as an observed factthat needed to be explained.Einstein’s great leap of insight wasto assume the speed of light isconstant in all frames of referencefrom the start, as a principle – andthen to see what happens as a result.

The insights in Einstein’s first paperon relativity are contained withintwo very simple postulates: thelaws of physics are the same in allinertial reference frames, and thespeed of light (through a vacuum)has the same value, c, in anyreference frame, no matter how thelight source is moving relative tothe observer.From these two ideas, the restflows: time slowing down andspeeding up, space shrinking andstretching – in fact, the classicalnotions of space and time as anarena for physical events went outthe window: space and timebecame players in the events ofEinstein’s universe. Space couldbecome time, time could becomespace, and physics could never bethe same.

PROFESSOR ANDREW D SHORT,School of Geosciences,University of Sydney is aSydneysider whodeveloped an attractionfor beaches and surf atan early age. He foundas an undergraduate atSydney in the 1960s youcould actually studythem, and did anHonours thesis onMcMasters Beach. He

then took his surfboard and headedto Hawaii when he completed hisMasters at the University of Hawaii,followed by a PhD in MarineSciences at Louisiana StateUniversity where he was sent tostudy the usually frozen beaches ofthe north slope of Alaska. He hassince conducted beach research inthe USA, Brazil, Europe, New Zealandand around the entire Australiancoast. After a postdoctoral position atMacquarie University he has been atthe University of Sydney since 1977.His research and publications relateto the evolution, nature anddynamics of coastal systems,particularly beaches, as well as itsapplication to managing our coast.

156

What got you interested in science

in the first place?I was good at geography in highschool and decided to study it atUniversity, which once I foundout I could study the coast, leadto coastal geomorphology andmarine science.

What were you like as a kid?Were you curious, pullingapart stuff to see how itworked? Were you alwaysinterested in science, or didthis come on later?I was always interested in thesurrounding environment – as achild I loved going for longwalks, exploring the bush andharbour foreshore. When I was ateenager I hitchhiked everypossible day to the beach to gosurfing. Once I was 17 I boughta car and started exploringAustralia and its coast, which Iam still doing.

What’s the best thing aboutbeing a researcher in yourfield?I get paid to go to the beach.

Who inspires you – either inscience or in other areas ofyour life?People who lead and do things,they don’t wait to be told.

What’s the ‘next big thing’ inscience, in your opinion?What’s coming up in the nextdecade or so?A better understanding of ourmarine environment – we mightfinally get to know as muchabout our oceans and seabed aswe do about Mars and theMoon.

Wind, Waves and BeachesProfessor Andrew D Short

IntroductionH A V E Y O U E V E R sat on sandy beach and looked at the waves, perhapsbreaking well out to sea before finally expending their energy as theyreach the shore, collapsing and running up the beach as a thin layer ofswash? That final thrust of energy and water, some of which runs backdown the beach, while some soaks into the sand, may have had abeginning many thousands of kilometres away and many days ago.

157

Figure 1: As a wind streamline flows over thesurface of a wave it pushes (stresses) harder theexposed windward face, while the pressure andstress is less on the leeward side of the crest.This produces shear stress on the face, as well aspressure differences between the windward andleeward sides. Both the shear stress and pressuregradient combine to increase wave height.

158

WAV

ES O

F TH

E FU

TURE

The energy that propels waves across theoceans and is finally expended at the shoreis derived from the atmosphere, from wind

blowing across the ocean surface. The energytransferred to the ocean surface as waves canthen transport this energy across vast oceandistances with surprisingly little loss of energy.Once these streamlined, energy-efficient wavesreach the coast, they rapidly transform as theyinteract with the shallow seabed. The energyreleased in shoaling and breaking sets off awhole new chain of events that ultimately shapesthe coastline and shore. Within the surf zone,ocean waves and secondary longwaves, edge waves, standingwaves and shear waves, andassociated currents, travel in alldirections and, in doing so,arrange every grain of sand into apredictable suite of bars, troughs,rip channels, sloping beaches,berms, cusps, megacusps ... Thebeach, and in particular the surf,may look a little confusing anddaunting, but it is behaving in avery predictable way governed bythe laws of physics.

Sandy beaches make up 50% of Australia’s30,000 km long coastline. In fact, there are10,685 of them and each and every one hasbeen classified into one of 15 different beachtypes: six are wave-dominated beaches thatoccupy much of the high wave energy southernhalf of Australia; three are tide-modified and fourare tide-dominated beaches, both of which aremost prevalent across northern Australia wheretides are high and waves lower; and the final twoare fronted by intertidal rocks or fringing coralreefs. The type of beach produced is a function ofthe incoming waves, the tide range and the typeof sand that forms the beach, acting within thelocal coastal environment. It is the waves,however, that supply the bulk of the energy to ourbeach systems.

In this chapter we begin where waves begin,where the wind blows over the ocean surface,transferring energy to form waves – or, morecorrectly, sea. The sea then is transformed intoswell and sets off across the ocean surface. We

will then look at what happens to waves as theyreach the shallow waters of the coast and finallybreak on a sand beach, and then what happensto form the typical wave and tide-dominatedsandy beaches that surround Australia and muchof the world’s coastline.

Wind and wavesWaves are disturbances of a fluid mediumthrough which energy is moved.

The lower atmosphere contains four major beltsof pressure which encircle the earth: the polar

high pressure, the subpolar lowpressure, the subtropical highsand, around the equator, theequatorial lows. Air iscontinuously flowing from theareas of high pressure towardsand into the areas of lowpressure and in doing sogenerates the worlds greatwind systems: the low velocitypolar easterlies, the highvelocity westerly of the roaring40s, and the moderate velocitytrade winds which flow over the

huge expanse of the subtropics. As 71% of theearth’s surface is ocean, most of these winds areblowing across large expanses of the Pacific,Atlantic, Indian, Southern and Arctic oceans, aswell as the many smaller seas. As the windsinteract with the ocean surface they generatewaves.

Wind-wave interactionWind energy is transferred to the ocean surfaceby both pressure fluctuations and tangentialstress exerted by the streamlines (Figure 1).Pressure fluctuations are expected to dominatebecause in nature wind is always turbulent at thesurface, which permits pressure differences toinitiate and enhance wave motion. Once seasbegin to grow they increase surface roughness;that is, the undulations in the surface manifest aswave crests and troughs. The increasedroughness in turn increases the shear stressbetween the wind and ocean surfaces, whichbuilds larger waves. It also increases the pressuredifference through the sheltering effect, as theincreasingly high waves shelter the backing

The beach, and inparticular the surf, maylook a little confusing anddaunting, but it isbehaving in a verypredictable way governedby the laws of physics.

159

WIN

D,W

AVES

AND

BEAC

HES

trough, which generates additional pressuredifferences. Since the sheltering effect isdependent on wave height there will be aconstantly changing balance between the two aswaves grow to a fully aroused sea.

There are, however, two factors that limit the heightwaves can reach, or to which they can be aroused.First is gravity, which, as wind is building thewaves, attempts to restore the water to its originalhorizontal state. As crests of waves are pulleddown by gravity the momentum continues belowthe flat-water surface, developing a trough, and sogravity waves are developed – another name forocean waves. Second is wave steepness, the ratiobetween the wave height H and length L. When H/Lis about 1/7 the wave will form white caps andbreak, thereby limiting its height.

Fully arisen seasAll winds flowing over a water surface willgenerate waves, however the size of the wavesthe wind can build is dependent on a number offactors. First is the wind velocity – the higher thevelocity the higher the waves, and as wind energyis transferring to waves at an exponential rate,very high waves require very strong winds.Second is the duration of the wind, the period of

time the wind blows – the longer the duration thebigger the waves. Third is the wind directionwhich will determine the direction of wave travel.Fourth is the fetch, or length of sea or ocean overwhich the wind can blow unimpeded – the longerthe fetch the higher the waves. And fifth is waterdepth, as shallow water (less than half the length)will mean wave interaction with the seabed andcause the wave to break.

A fully arisen sea is therefore the maximum waveconditions that can be produced by a given windvelocity, duration and fetch blowing over a deepocean. For light winds the waves will be very low,while only very strong winds, blowing in aconstant direction, for a low period (days) over along fetch of deep water can generate the biggerwaves. Figure 2 illustrates the spectral energytransferred to a fully arisen sea for a range ofwind velocities.

Deepwater wavesWhile wind is blowing over the ocean surface itgenerates waves called a sea. By nature, seaconsists of waves that are relatively high butshort in length, and therefore steep. They areprone to breaking and travel in a broad range of

Figure 2: The amount of wave energy generated by windsfrom 20 to 40 knots. Wave energy increases exponentiallywith wind velocity. (From Neuman and Pierson, 1966.)

160

WAV

ES O

F TH

E FU

TURE

directions either side of the wind direction. Oncethe wind stops blowing, or the seas leave thearea of wind action, they rapidly transform intowhat we call swell – swell waves are lower,longer and therefore flatter, and tend to travel in amore uniform direction. Once formed into swell,waves can theoretically travel around the globe,though in reality they always run into something,such as an island or the coast. When both seaand swell are travelling in wave depths (ho) whereho/L > 0.5 they are called deepwater waves,meaning the water is too deep for them tointeract with the seabed (Figure 3). In reality mostwaves have lengths less than 350 m, while theoceans average over 4 km deep, meaning in thedeep ocean all waves are deepwater waves.

Deepwater waves have a number of predictablecharacteristics, which are illustrated in Figure 3.First, while the waveform is moving in thedirection of wave travel, the water in the waveundergoes an orbital motion – that is, the watermoves around and around, as the wave crestsand troughs pass overhead. This is why when youare in a boat at sea the wave passes underneath,while the boat just bobs up and down. The orbitalmotions extend down into the ocean, butdecrease exponentially in size and orbit withdepth, so at depths greater that half thewavelength (L/2) no orbital motion remains.

Wavelength, as illustrated in Figure 3, is thedistance between successive wave crests ortroughs. It is also related to the wave period T –the time between successive crests – such that

L = g/2πT2 tanh (2πho/L) (1)

where g is the gravitational constant and tanhrefers to the hyperbolic cotangent.

The speed at which the wave travels, C, is called thewave phase velocity and is given by the ratio L/T

C =L/T = g/2πT tanh (2πh/L) (2)

Both equations can be approximated indeepwater when h becomes large, causingtanh(h) ≈ 1, which enables equations 1 and 2 toreduce to

Lo = gT2/2π (3)

and

Co = gT/2π (4)

Since g and πare constant, Lo = 1.56 T2 and Co

= 1.56 T, meaning L increases at the square ofthe period, and C is directly related to T. It alsomeans that waves with a longer T and L travelfaster than shorter waves. Typical deepwaterwaves arriving around the southern Australiancoast have T between 10 and 14 s, and thereforelengths between 156 and 306 metres and speedsbetween 15.6 and 21.8 m/s, or 56 and 78 km/hr.Figure 3: Orbital motion beneath a deepwater wave. So

long as the depth is greater then L/2 the wave will notinteract with the seabed. (From Sager, 1998.) Figure 4: Seas (to the left of the

crossover) are relatively high andshort in period and length. As theytransform to swell (right ofcrossover) they become lowerand longer in period and length,and as such can theoreticallytravel for thousands of kilometres.(From Davies, 1980.)

161

WIN

D,W

AVES

AND

BEAC

HES

Figure 4 illustrates the transformation of the higherbut shorter and slower seas into lower but longerand faster swell, which can then travel greatdistances with minimal additional loss of energy.

Wave climatesBecause waves are dependent on the global windregimes they can be classified according to thepart of the ocean, or wind regime, that producesthem. Figure 5 illustrates the global distribution ofwaves that occur 10%, 50% and 90% of thetime. Note how the world’s biggest waves arelocated year round in the Southern Oceanbetween 40ºS and 60ºS, where they aregenerated by the subpolar lows andaccompanying ‘roaring 40s’ and ‘raging 50s’, the

strong westerly winds that encircle the southernhemisphere at those latitudes. They blow yearround across long ocean fetches to generate theworlds biggest wave factory. Because of theCoriolis effect (which is due to the rotation of theEarth) these westerly waves are deflected to theleft towards the equator, and so can travel fromsouth of Australia to Hawaii and California.Likewise in the northern hemisphere similarwaves are generated, but only in the northernwinter, as the low pressure regions reside overthe landmasses during summer.

The waves are termed storm wave environments,and the swell emanating from their stormy wavefactories are called west coast swell, as they tendto arrive on the western side of the oceans’basins, as well as southern Australia. The swellthat has to bend and travel up the eastern side ofthe continents, like the east coast of Australia, iscalled east coast swell. The mighty thoughmoderate velocity trade winds, which dominatethe world’s wind systems, generate only low tomoderate waves in the subtropics. The tropics arethe site of the ‘doldrums’, calm balmy conditionswhich produce few waves; consequently thetropics only receive larger waves when they aregenerated outside the tropics by the systemsmentioned above. Likewise in polar regions thepolar easterlies are not only low in velocity formuch of the year, they also blow over a frozensea, so only low waves are generated in summer.

The world’s deepwater wave climates aretherefore very closely related to latitude and theprevailing wind systems, as well as location andorientation of the coastline. Those facing towardsthe west coast swell receive these energeticwaves even though they may be locatedthousands of kilometres distant. The deepwaterwave height is, however, only half the story – tohave beaches those waves have to get to theshore, and in crossing the shallow waters of thecontinental shelf a range of factors act to hindertheir arrival.

Shallow water wavesAll waves eventually run into shallow water and inthe process transform into shallow waves.Shallow water occurs when water depth h < L/2,

Figure 5: Global distribution of wave heights occurring10%, 50% and 90% of the time. The biggest waves aregenerated year round in the southern oceans. (From Short,1999, based on Young and Holland, 1996.)

162

WAV

ES O

F TH

E FU

TURE

since at this depth the wave orbits begin toincreasingly interact with the seabed, and in theprocess undergo a range of predictabletransformations based on the depth andconfiguration of the seabed. In shallow watertanh(h) becomes very small, enabling thefollowing shallow water approximations toequations 1 and 2:

Ls = T √gh (5)Cs = √gh (6)

Both shallow water wavelength Ls and speed Cs

are dependent on depth, and both decrease as

depth decreases – in other words, waves shortenand slow down as they approach the shore.

This leads to a range of shallow water impacts.First, waves shoal, meaning L shortens, which inturn concentrates the energy over a shorterlength, thereby increasing wave height. Wavesalso attenuate, meaning they transfer some oftheir energy to the sea bed, for example by

Figure 6: Waves shoaling, refracting and breaking alongsections of the Tasmanian coast.

Figure 7: Wave breaking - a) surging, b) plunging and c)spilling waves.

163

WIN

D,W

AVES

AND

BEAC

HES

moving sand, which will lead to a loss of waveenergy and a decrease in wave height. Overvariable seabed, the waves move faster in deeperwater and slower in shallower water, causing thewave crests to bend and change direction, aprocess known as wave refraction (Figure 6).Finally, if the water becomes too shallow thetrough slows faster than the backing crest, whichattempts to overtake the trough and in theprocess breaks.

Breaking wavesWave tend to break when h = 0.78Hb, where Hb

is the breaker wave height. However the type ofbreaker will depend on the slope of the seabed orsand bar over which the wave is breaking. Whenthe slope is gentle waves gradually peak and thetip of the crest spills forwards producing spillingbreakers. Moderate slopes result in a plungingbreaker where the crest thrusts forward and curlsover the face of the wave, producing what surferscall a tubing wave. On steeper slopes the frontface and crest of the wave remain smooth andthe wave slides directly up the beach withoutbreaking as a surging breaker (Figure 7).

TsunamiThe tsunami on 26 December 2004 focusedworldwide attention to this type of wave and itspotentially deadly impact. While winds havenothing to do with tsunamis, a brief description iswarranted for those interested in their science.

Tsunamis are waves generated by a suddenimpulse or impact in the sea. The source of theimpulse is usually an earthquake that displacesthe sea floor up or down. In the case of Aceh, a1000 km long by 200 km wide area of seafloorwas suddenly lifted by ten metres. Tsunamis arealso generated by large undersea and coastallandslides (usually triggered by an earthquake),volcanic explosions (Krakatoa near Javagenerated a massive tsunami in 1883) and, morerarely, meteorite-comet impacts, which cangenerate mega-tsunamis. The single impulsegenerates several waves as the ocean surfacegradually returns to normal, in much the sameway as the waves generated by a rock throwninto a pond.

In the deep ocean tsunamis travel very fast (up to800 km/hr), are spaced approximately 200 kmapart and have a period of about 15 minutes.Tsunamis have a crest and a trough like allwaves. The trough usually arrives first, resulting ina draw-down in the water and sea level, whichunfortunately can attract people onto the exposedseafloor. However within five to ten minutes thefirst crest will arrive. While in the deep oceantsunamis may only be a few decimetres high,because they are so long and contain so muchwater they build in height as they approach theshallow coastline and slow down to 40-50 km/hr.

Once it hits the coast the wave not only increasesin height, up to ten metres and more in Sumatraand Sri Lanka, but because of its great length itjust keeps on coming for several minutes, raisingthe water level and flowing inland until it reachesground higher than the tsunami. It then retreatsfor several minutes before the second and usuallylargest wave arrives. This is followed by a seriesof increasingly smaller waves.

Tsunamis occur relatively frequently, particularlyin the Pacific Basin, and major waves the size ofthe Aceh tsunami occur on average every onehundred years. The devastation and lost of lifecaused by the Aceh tsunami was a result of alarge (8.9 Richter scale) undersea earthquakeand the associated uplift of the seafloor occurringclose to a densely populated, low-lying coast. Theearthquake generated a large tsunami, whicharrived at the Sumatra coast within 10 to 15minutes as a ten metre high wall of water thatrapidly flooded the densely populated low-lyingcoastline. The tsunami wreaked havoc furtherafield in Thailand, Sri Lanka, India and east Africa,with smaller remnants circulating right around theglobe within 24 hours. An early warning systemcould have warned many of the locations in timefor evacuation of the local population.

Waves and beachesOnce waves break they undergo a rapid almostinstantaneous transformation, from a progressivewave composed of potential energy that mayhave travelled thousands of kilometres, to atranslatory wave full of kinetic energy that can dowork in the surf zone. A translatory wave is one

164

WAV

ES O

F TH

E FU

TURE

where the water moves with the waveform, andits energy is translated shoreward as a brokenwave, also called a wave bore or white water. Thesurf zone is the area between the wave breakingand the shoreline – it’s the most energetic anddynamic part of the Earth’s surface.

The energy released as the wave breaks is in theform of turbulence, sound (the roar of the surf)and even heat. The turbulence moving towardsthe shore stirs sand into suspension and carries itshoreward with the wave bore. The wave boredecreases in height shoreward, eventuallycollapsing into swash as it reaches the shoreline.

Breaking waves, wave bores and swash, togetherwith unbroken and reformed waves, all contributeto a shoreward momentum in the surf zone. Asthis energy moves shoreward it is transferred intoother forms of surf zone currents, namelylongshore, rip feeder and rip currents, as well aslong waves and associated currents, all of whichcan move shoreward, alongshore or seaward.Eventually, all this water has to return seaward; itdoes so through three related mechanisms: itmay simply reflect off the beach face and travelseaward as a reflected wave; it may flowsideways and then turn and pulse seaward as arelatively strong, narrow rip current; or it maycontribute to the growth of standing (long) wavesagainst the shoreline, which also pulses waterseaward via bed return flow, which is waterflowing seaward underneath the breaking wavesand wave bores.

So while waves deliver potential energy to theshore, it is the transformation of the wave energyfrom potential to kinetic energy and a range ofsecondary wave and currents processes that dothe work in the surf zone to shape the beaches ofthe world. The most important secondary wavescan be grouped under the general term of longwaves, with subdivision into standing waves andedge waves. Long waves are by definition long inperiod, usually greater than 20 seconds and up toseveral minutes. They are also known asinfragravity waves – that is, beyond or longerthan ocean gravity waves.

Long waves form in the surf zone due to acombination of factors. First, there are long

waves in the deep ocean related to the groups orsets of higher and lower waves, called wavegroups. These groups cause the sea level to beslightly depressed under the high waves, andelevated under the lower waves, forming a longwave with the same period as the wave group(which is typically several minutes). These longwaves enter the surf zone unbroken. As wavestravel shoreward, break and transform theytransfer energy and water from the shorter gravitywaves to these longer waveforms. This transferreaches a maximum at the shore where the longwaves reach their maximum size and amplitude.As a rule of thumb the height of the long wave atthe shore is one-third to one-half of the height ofthe breaker waves.

Long waves can also be manifest as standingand/or edge waves. Standing waves are formedwhen an incoming long wave interacts with anoutgoing (reflected) long wave in the surf zone, soas to form a wave whose crest simply moves upand down, but does not progress – it just standsin place. The water under the crest, however,must move sidewards and return to the crest witheach oscillation, thereby generating horizontalcurrents associated with the standing waves.What we have, therefore, is a standing wave,periodically moving up (set-up) and down (set-down), extending across the surf zone and a fewwavelengths out to sea.

Standing waves are important because, on higherenergy beaches with bars and surf, theydetermine the location of the outer bar and thespacing of multibars. The bar crests form underthe antinode or crest of the wave, while thedeeper troughs of the bar form under the node ortrough of the wave. The horizontal currentsmoving between the standing crest and trough,as mentioned above, are responsible for movingsediment towards the antinodes and forming thebar crests.

Standing waves increase in spacing exponentiallyoffshore, so when there are two or more barstheir spacing likewise increases exponentially(Figure 8). The wavelength of a standing wave is

Ls = g/2πTs2 (2n+1) tanβ (7)

165

WIN

D,W

AVES

AND

BEAC

HES

where Ts is the standing wave period, n standingwave mode and β the slope of the surf zone. Wecan therefore predict the standing wave lengthand bar spacing: the longer the period and lowerthe beach slope, the greater the wave length andtherefore bar spacing.

Another form of long wave motion is edge waves.Edge waves are standing waves that are trappedin the surf zone and also propagate along shoreas a series of standing crests and troughs. Theytherefore stand and oscillate both perpendicularand parallel to the beach. Like standing waves,edge waves are smallest and shortest at theshoreline and increase exponentially in sizeoffshore. At the shoreline they generate a cellularcirculation in the upper swash zone, which canlead to the formation of beach cusps. On oceancoasts cusps are usually spaced every 20 to 40metres, which matches the edge wave spacing(Figure 9a). In the surf zone the edge waves havewavelengths ranging from 100 to 500 metresand are responsible for the cellular circulationthat leads to the formation of rip currents spacedat similar distances, and the associatedintervening crescentic bars (Figure 9b).

Therefore, while it is the shorter period gravitywaves (sea and swell) that drive sand shorewardand form the general beach profile, it is the

Figure 8: Standing waves form in the surf zone inassociation with breaking waves, with the horizontalcurrents under the waves (arrows) leading to the formationof bars crests under the antinodes.

Figure 9: a) Well developed beach cusps on Pearl Beach,NSW; b) Crescentic bars and regular rip currents alongForster Beach, NSW.

166

WAV

ES O

F TH

E FU

TURE

longer period standing and edge waves that areresponsible for rearranging this sand into shoreparallel bars, and segregating the swash to formbeach cusps and the surf zone currents to formrip currents and crescentic bars.

Sand and beachesWaves are only half the equation when it comesto beaches. A beach is a wave-depositedaccumulation of sediment lying between wavebase and the swash limit. They are usuallycomposed of sand but can form from cobblesand boulders. Wave base is the depth at whichwaves can pick up and move sediment

shoreward, usually at a depth equal to a quarterof the ocean wave length – between depths of 20to 30 metres along the New South Wales coast.The swash limit is as far as the swash reachesup the beach, usually an elevation of about 3metres. Therefore to have a beach you musthave sand and waves – and as mentionedpreviously this combination forms half theAustralian coastline.

Just as the size of waves will vary around thecoast, depending on the deepwater wave climateand shoaling processes, so too the sediment thatcomposes the beaches ranges from fine sand tocoarse sand, cobbles and boulders. The impact of

Figure 10: Distribution of wave-dominated (WD), tide-modified (TM), tide-dominated (TD) and beaches fronted byrock or reef flats (RF) beaches around Australia. (FromShort, 2005.)

167

WIN

D,W

AVES

AND

BEAC

HES

sediment size is to change the gradient or slopeof the beach, from a low 1º gradient on a finesand beach to a few degrees on medium sand, to10º or more with coarse sand, and up to 15º oncobbles and boulder beaches. All this alsocontrols the slope and width of the surf zone,with the widest surf zones composed of finesand, while coarse sand, cobble and bouldersform a slope so steep that no bar or surf zonecan form.

It is therefore the variation in waves and sedimentthat produces the seemingly wide range ofbeaches present along the coast, ranging fromthe steep, narrow, protected beaches to thebroad, low gradient beaches with wide surfzones, large rips and massive breakers. Yet everybeach follows a predictable pattern of response,largely governed by its sediment size andprevailing wave height and length. The nextsection discusses the types of beaches that canbe produced by waves and sand aroundAustralia, followed by the impact of increasingtide range, as is typical of northern Australia.

Australian beachesAround southern Australia the coast is exposed tothe persistent high swell originating in theSouthern Ocean. This generates a west coastswell that dominates from North West Cape in thewest around to Tasmania in the east, and an eastcoast swell running up the southeast coast toFraser Island. This energetic wave climatecombined with a low (less than two metre) tiderange results in an overwhelming occurrence ofwave-dominated beach systems in the south.This contrasts with northern Australia wherelower subtropical and tropical wave climates andtides up to ten metres produce tide-modified andtide-dominated beach systems (Figure 10).

Wave-dominated beaches Wave-dominated beaches by definition have lowtide ranges, as a result of which the breaker zone,surf zone and shoreline are relatively stationary.This enables all the wave and long-wave processesto imprint themselves upon the surf zone andbeach morphology – bars, rip currents and beachcusps all form in a relatively stationary surf zonearea. In addition, once rips are formed they usually

generate a reworking of the shoreline to formmegacusps, regular undulations in the shorelinethe same spacing as the rip currents.

There are three types of wave-dominatedbeaches. Under low waves (less than one metre)and particularly if the sand is coarse, the beacheswill be reflective. Reflective beaches have no barsor surf zone; the waves arrive at the base of thebeach, surge up a relatively steep beach face andreflect back out to sea – hence the name‘reflective’. Beach cusps are usually present, asillustrated in Figure 9a. They represent the low-energy end of the beach spectrum.

When waves are between 1 and 2.5 metres theytend to produce intermediate beaches, which arecharacterised by bars, breaking waves, surf andcellular rip circulation. The rip and bar spacing iscontrolled by the edge waves and, in Australia,range from a small 100 metres in the north to upto 500 metres on high-energy southern Australiabeaches (Figure 11a). This is the most common

Figure 11: a) Rip dominated beach with alternating barsand rip channels, Victoria; b) Dissipative beach with shoreparallel bars and trough, South Australia.

168

WAV

ES O

F TH

E FU

TURE

beach type around southern Australia and resultin the formation of 13,500 beach rips around the coast.

Waves greater than 2.5 metres, when breakingover fine sand, produce the highest energy beachtype called a dissipative beach, because the highwaves dissipate their energy over a wide, lowgradient multibar surf zone (Figure 11b). Largestanding waves and vertically-segregated bedreturn flow dominates the surf zone circulation.

Tide-modified beachesTide-modified beaches occur when the tide rangeis between three and twelve times the breakerwave height. As shown in Figure 10, they occur inthe higher tide range areas of northern Australiaas well as in the South Australian gulfs and partsof Tasmania. The major difference between wave-and tide-modified beaches is the periodicoscillation of the tide, which shifts the shoreline,surf zone and breakpoint backwards andforwards across the intertidal zone, over adistance of hundreds of metres. As a result, whileall the breaking, standing and edge waveprocesses can occur on tide-modified beaches,because they aren’t stationary they have difficultyimprinting themselves upon the shoreline and

Figure 12: a) Tide-modifiedbeach with low tide riphighlighted by dye,Queensland; b) tide-dominatedbeach with 500 m wide verylow gradient intertidal zone,north Western Australia.

(a)

(b)

169

WIN

D,W

AVES

AND

BEAC

HES

intertidal-surf zone. Consequently, while cuspsare commonly found on the high tide part of thebeach, bars and rip channels are usually absentfrom the intertidal-surf zone (Figure 12a). There isinsufficient time to form them because of thecontinual shift in the processes, and any incipientform is reworked and smeared by the shiftingshoreline and breaker zone.

Tide-modified beaches do however have a steepreflective beach at high tide typically containingcusps, and a wide low gradient intertidal zone;only at low tide do some form rips and bars asthe tide slows and turns.

Tide-dominated beachesTide-dominated beaches are the most commonbeach type across northern Australia (Figure 9).They require areas of very low waves and hightide, such that the tide range is between 12-50times the wave height. These conditions occur ona third of the Australia’s beaches. These beachesare typified by a small sandy high tide beach,fronted by a very wide, low gradient sandyintertidal zone, which may have subdued sandridges. They may be flat and featureless (Figure12b), have the imprint of tidal currents and,finally, they may be formed of mud. Beyond, theshoreline grades into true tidal flats.

SummaryMost energy is transferred in some form of wavemotion, whether it be electromagnetic wavesemanating from the Sun to warm our planet, orthe sound waves you hear when listening tosomeone talking. Wind energy is derived fromdifferences in atmospheric pressure, which aredue to the differential heating of our planet – andso is ultimately derived from the Sun’s rays. Thewind is able to transfer some of this energy intothe surface of the ocean in the form of wavesand ocean currents, both of which can continueto travel long after the wind has stopped blowing.This energy is therefore stored in the oceansurface. The wave energy is able to rapidlyarrange itself into energy-efficient swell waves,and can potentially travel great distances to farshores and coasts. The potential energy carriedby the waveform, is converted into kinetic energyas the waves shoals and break across the shore.

This transformation both releases and transformsthe wave energy into a range of longer-periodsecondary waves and currents, which areprimarily responsible for shaping the finer detailof the surf zone and shoreline. The interaction ofwaves, sand and tides are responsible for fifteendifferent types of beaches, and within this systemof categorization we can classify all 10,685Australian beach systems.

References

Papers and books

Davies, J L, 1980, Geographical Variation in Coastal

Development. 2nd Ed (Longman, London).

Neuman, G and Pierson, W J, Jr, 1966, Principles of Physical

Oceanography (Prentice Hall, Englewood Cliffs, NJ).

Sager, D A, 1998, Introduction to Ocean Sciences (Wadsworth

Publishing Co., Belmont, CA.).

Short, A D (ed), 1999, Beach and Shoreface Morphodynamics

(John Wiley and Sons, Chichester).

Short, A D, 2005, Australia beach systems – nature and

distribution. Journal of Coastal Research.

Young, I R and Holland, G J, 1996, Atlas of the Oceans, Wind

and Wave Climate (Elsevier, UK).

Websites

NSW wave buoys

http://www.mhl.nsw.gov.au/www/real_quick.htmlx

Beach video images

http://www.wrl.unsw.edu.au/coastalimaging/public/tweed/

index.html

170

WAV

ES O

F TH

E FU

TURE

brain in various birds andanimals. One day, he happenedto be stuck for 5 hours inChicago Airport. He noticed thatwhen a particular couple kissed,they each tilted their head tothe right – so that eachperson’s nose was to their rightof the other person’s nose. Hesuddenly realized that he hadjust seen a left-right preference,and that public places were agood place to collect data.

Now being a good scientist (asopposed to a perve in a brownraincoat) he immediately set upsome criteria. There had to belip-to-lip contact, the faces hadto be aimed at each other, therehad to be an obvious directionof head-tilting, and finally, theycouldn’t carry any luggagebecause this could influence thedirection in which they tiltedtheir head. Over the next 2.5years, he collected data on 124scientifically-valid “kissingpairs” at airports, parks,beaches and railway stations.The results were clear-cut –two-thirds of people tilt theirheads to the right. This, ofcourse, means that one-third ofus kiss the wrong way.

We humans have a long historyof kissing. Early Christianskissed each other wheneverthey met. The bride and groomkiss after the marriageceremony. Eskimos andPolynesians kiss by rubbingnoses together, while theinhabitants of SoutheasternIndia do their version of the kissby each pressing the noseagainst the cheek, with theactive person inhaling deeply.There’s deep (or French) kissingwhere one person’s tonguegoes a’roaming in the otherperson’s mouth – and we’ll stopright there. And as a completecontrast, kissing was not veryvisible in the old days in Asia,because the bow was the all-purpose greeting, and peoplewould kiss only in private – soof course, you wouldn’t see it.

There are a few theories onhow this habit came to be. Onetheory claims that it all began inancient times with motherschewing up food to pass itdirectly into the mouth of theirbaby. A second theory reckonsthat kissing gets you closeenough to smell the mood, foodand recent adventures of

whomever you are kissing, soyou can work out how to handlethem. And a third theory saysthat we used to believe thatyour soul lived in your breath,and that kissing would marrythe breaths together, and fuseyour souls for all eternity. Thislast theory is really cute, butlike all the other theories, istotally unprovable.

On the other hand, somepeoples have thought it waswrong to kiss. The state ofIndiana in the USA has a lawmaking it illegal for a man witha moustache to “habitually kisshuman beings”. In Hartford inthe state of Connecticut, it isstill illegal for a man to kiss hiswife on a Sunday. And in 16thcentury Naples, in Italy, kissingwas an offence that carried thedeath penalty.

Even so, most of us today willstill kiss – so how can there bea “wrong way to kiss”? Well,that’s the opinion of OnurGunturkun from the RuhrUniversity in Bochum, Germany.His speciality is studyingdifferences between the left andright sides of the body and

Right Way to KissBy Dr Karl Kruszelnicki

THERE ARE MANY different types of kisses – the soft flutterykisses between a child and parent, the chaste closed-mouthkiss given to a grandparent or aunt, and of course, the wildrollicking hungry open-mouthed kisses between two peoplewho love each other in a very special way. In an averagelifetime, we spend about two weeks in kissing. So how come,if we do it so often, one third of us kiss the wrong way?

171

RIG

HT

WAY

TOKI

SS

What’s going on? Firstly, about 90% of us areright-handed. About two-thirdsof us (about 60-70%) prefer touse our right foot, right eye orright ear. So on average, wehumans tend to use our right side.

Secondly, we humans tend toturn our heads to the right(rather than the left) for our last weeks in the uterus, andour first six months after being born.

So Dr. Gunturkun reckons thatwe humans start off with apreference to look to the right,and pay more attention toevents on our right. He alsoreckons that this preferencecarries right on into our fertile

years, so that when we kiss, wewill tilt our heads to the right.

Maybe for once this is too muchscience, and we would bebetter off listening to the wordsof the 18th-century Scottishpoet, Robert Burns:

“Honeyed Seal of soft affections,Tenderest pledge of future bliss,Dearest tie of young connections,Love’s first snowdrop, virgin kiss”,

and go right ahead and kiss inwhatever way seems natural...

FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

Illustration courtesy Adam Yazxhi

DR RAFFAELLAMORGANTI has workedfor the last few years atthe NetherlandsFoundation for Researchin Astronomy (ASTRON)in Dwingeloo, as amember of theWesterbork RadioObservatory group. Shewas born in Italy andobtained her PhD inAstronomy at theUniversity of Bologna.

Afterwards she enjoyed several yearsin Munich at the European SouthernObservatory (ESO) and at theAustralia Telescope National Facilityin Sydney. Before moving to TheNetherlands, she worked as staffastronomer at the institute of RadioAstronomy in Bologna. Raffaella’scurrent research interests include thestudy of Active Galactic Nuclei (radiogalaxies in particular), as well as thestudy of the interstellar medium innormal and active galaxies.

172

What got you interested inscience in the first place?

Perhaps the enthusiasm of myfather for astronomy (and science ingeneral). He passed this on to me. Istill remember when a total solareclipse was visible in Italy, when Iwas only a few years old, and howhe tried very hard to explain to mehow unique this phenomenon wasand make sure it would remain in mymemory!

What is the best thing about beinga researcher in your field? I think doing scientific research isgreat, regardless of the field! Ofastronomy, I like the fact that thecommunity is so international, oneworks with people and instrumentsthat are spread all over the world(and beyond!). Some of thetelescopes are located in veryremote and fascinating places!

If you could go back andspecialize in a different field, whatwould it be?When I had to decide which courseto take at the University, I wasactually uncertain between geology,biology and astronomy. In my nextlife, I would probably like to trygeology!

What ‘s the ‘next big thing’ inscience? It is going to be a great time to be anastronomer. The astronomicalcommunity is busy planning andbuilding the new generation oftelescopes. They will represent sucha major step forward that will makepossible to almost look at thebeginning of the Universe, somethingthat now we can only predict throughnumerical simulations.

The ever changing life ofgalaxiesRaffaella Morganti

ONE NIGHT I WAS, as usual, observing the sky with my telescope. Inoticed that a sign was hanging from a galaxy a hundred million light-years away. On it was written: I SAW YOU. I made a quick calculation:the galaxy’s light had taken a hundred million years to reach me, andsince they saw up there what was taking place here a hundred millionyears later, the moment when they had seen me must date back twohundred million years. Even before I checked my diary to see what I hadbeen doing that day, I was seized by a ghastly presentiment [...]

Italo Calvino, Cosmicomics

173

Figure 1 – The MilkyWay and the galacticcentre as seen fromLa Silla ESOobservatory in Chile.

Figure 2 (opposite,background image)The Hubble DeepField (HDF) was takenwith the HubbleTelescope by staringat one tiny piece ofsky for ten days. Theimage reveals morethan 1500 galaxies.

174

WAV

ES O

F TH

E FU

TURE

Timescales and distances in the Universe areamazingly large. Not only are the distancesbetween objects enormous, the objects

themselves are also huge. Most of the astronom-ical objects appear immutable in the sky but thisis only because the timescales of changes areusually much too long to be directly measurableby humans. There are, however, other ways tofind out how these objects change during theirlife. In this chapter we will see how astronomersthink some of the largest astronomical objects, thegalaxies, are evolving as a result of the spectacularencounters – that can happen once or more intheir lifetime – with other neighbouring galaxies.

Galaxies are among the largest objects in the sky.We live inside one of them, the Milky Way, andthe great majority of what we see in the sky withthe naked eye are stars that belong to our galaxy.The Milky Way is quite an ordinary spiral galaxythat, nevertheless, contains about 100 billionstars. The bright band that crosses the sky –from which the name Milky Way (see Figure 1) isderived – comes from the myriad of stars thatform its disk. To travel from its centre to thelocation of the Sun would take almost 30thousand million years ... that is, if one couldtravel at the speed of light.

The Milky Way is just one of the many galaxies inthe Universe. It is quite amazing to realize that ithas only been only 80 since years we discoveredthat outside our Milky Way many similar worldsexist (Figure 2). The discussion about theexistence of ‘island universes’ similar to the Milky

Way actually went on for centuries, but the finalanswer only came when astronomers were able tomeasure with reasonable accuracy the distancesto these systems, thanks to Edwin Hubble in1925. Owing to the very deep images (see forexample Figure 2) that can be taken by the verypowerful telescopes available at present, we canobserve galaxies so far away that they must haveformed very soon after the Universe started.

Galaxies are big! The number of stars that form agalaxy can go from 108, for a ‘dwarf’ galaxy, up to1012 (a one followed by twelve zeroes!) for a largeone. A galaxy can be so large that, for the starsat the periphery, it takes more than 100 millionyears to make a full orbit.

Given the huge distances, the only way to studygalaxies is through the information carried to usby the light that they emit. The term ‘light’ inastronomy has a very broad meaning, as it doesnot represent only the ‘visible light’ to which oureyes are sensitive (and that is only a tiny bit ofthe electromagnetic spectrum), but also theemission with wavelengths ranging from radio toX-rays and gamma rays (see Figure 3).

Not all wavelengths emitted by astronomicalobjects can reach the earth (see Figure 4)because the atmosphere absorbs some of them.Thus, the telescopes capable of observing thesewavelengths have to be located outside theatmosphere using satellites or building them athigh elevation. Compared to only 50 years ago,astronomers now have instruments – both

Figure 3 - Light consists of electromagnetic waves that can have very different wavelengths. The figure shows what thedifferent portions of the electromagnetic spectrum are called. The visible light we see with our eyes has wavelengths rangingfrom about 400 nm (1 nanometer is one billionth of a meter) at the blue end to about 700 nm at the red end. Light withsomewhat longer wavelengths than red light is called infrared. Radio waves are the longest-wavelength of light – they can bea meter long! On the other end of the spectrum (with wavelength shorter than the blue light) is the ultraviolet and, with evenshorter wavelengths, the X-ray and the gamma rays. Different bands of the electromagnetic spectrum can be observed usingdifferent instruments.

Figure 5 - The Milky Way seen in different wavebands fromradio to X-ray and observed using different instruments.The different morphologies in the different wavebands areclearly visible.

175

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

ground based and in orbit around the earth – toobserve the sky in all these different wavelengthbands. This has opened completely newpossibilities to understand astronomical objects(like galaxies). In particular, it has allowed the

discovery that these objects can look verydifferent when observed in different bands of theelectromagnetic spectrum! Figure 5 shows theexample of our Milky Way, but there are caseseven more spectacular that we will see in thisand the following chapter.

These differences are because the emission in thedifferent wavebands is generated by differentphysical mechanisms, or by the same mechanismbut under different conditions. For example, theemission from electrons moving at relativisticspeed in a magnetic field can be responsible forpart of the emission at radio wavelengths. Gas atdifferent temperatures emits in differentwavebands: the higher the temperature, the shorterthe wavelength emitted. Stars are responsible formost of the emission at the optical wavelengths.

How to make a galaxyOne of the most outstanding problems inextragalactic astronomy, and one of the bigpuzzles that astronomers are trying to solve, ishow galaxies form. Galaxies are believed toemerge from some small perturbations in the

Figure 4 - Diagram showing the approximate depths towhich different wavelengths of light can penetrate Earth’satmosphere. A large part of the electromagnetic spectrum– except for visible light, a small portion of the infrared,and radio – can be observed only from very high altitudesor from space.

176

WAV

ES O

F TH

E FU

TURE

(otherwise very smooth) Universe soon after theBig Bang. Computer simulations show that theseperturbations can grow to produce filamentarystructures, and from the higher density regions inthese filaments, galaxies form. The distribution ofthe regions of galaxy formation (if we let thesimulations continue on!) would look remarkablysimilar to the distribution of galaxies (and galaxyclusters) that we observe in the present day Universe.

Two major scenarios can describe how galaxiesactually formed out of these condensations (seeFigure 6). The first, considered by a largercommunity of astronomers as the most likely, isthe so-called hierarchical scenario. In this picture,there is no single epoch of galaxy assembly;rather, galaxies form and evolve continuously. Thebuilding up of galaxies is done by mergingsmaller clumps that, over time, will form muchmore massive structures. The second scenario is

Figure 7 - Examples of spiraland elliptical galaxies. Onthe left is NGC 2997, a spiralgalaxy that contains a lot ofgas and young stars, whileon the right is the ellipticalgalaxy M87 (also known asNGC 4486, or Virgo A). Thelatter contains relatively littlegas but has a powerfulactive nucleus in the centreassociated with a super-massive black hole (seemore about this galaxy inthe next chapter).

Figure 6 - A schematic view of the two competing scenarios of galaxy formation.

177

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

the monolithic scenario, with an early and rapidcollapse of matter when the Universe was stillyoung, followed by passive evolution of thegalaxies thereafter. It is not easy to distinguishbetween these two scenarios and there are prosand cons in both cases.

Galaxies are mainly classified in three groups:elliptical galaxies that seemingly have little or nostructure and little or no star formation or gas;spiral galaxies that are gas-rich and that haveongoing star formation; and a third group of so-called ‘irregular’ galaxies. Examples of a spiraland an elliptical galaxy are given in Figure 7.

If both elliptical and spiral galaxies have formed atthe early stage of the Universe, as predicted bythe monolithic scenario, how can we explain sucha major difference in their appearance, or manyother of their characteristics like, for example, thetype of stars – young vs old – that dominatesthese systems? A possibility is that in the twosystems the speed at which stars form isdifferent. If the initial star formation happened veryfast and the gas was used up in an initial burst,the galaxy will have a round shape and very littlegas left – and we would classify it as ellipticalgalaxy. Otherwise, if the star formation is slower,the gas has time to settle and form a rotatingdisk, and therefore the result is a spiral galaxy.

However, as mentioned above, the scenarioconsidered more likely by the astronomers is thehierarchical one, in which small condensationsform first and then, through interaction andmerging, they form larger structures like ellipticalgalaxies. One of the reasons why this isconsidered to be the way galaxies form is thatobservations are increasingly showing thatgalaxies are indeed still forming and evolving inthe local Universe under the effect of interactionand/or merging.

What does this mean? Galaxies seldom live in acompletely isolated environment. Very often, theylive in clusters or groups, regions where thedistances between galaxies is, on average, only afew times the size of a typical galaxy. Because oftheir motion inside these structures, galaxies arelikely to interact with each other. This means thatthey feel each other’s gravity up to the point that

a larger galaxy may completely tear apart asmaller one. In other extreme cases, the galaxiesget so close that they cannot avoid merging andforming one object. Although less frequent, theseevents also occur to galaxies that are located inless dense environments (so called field galaxies).

Figure 8 - Three impressive examples of interactingsystems caught by the Hubble Telescope. (Top) The so-called Tadpole Galaxy with the impressive 280 thousandlight-year tail likely formed by a small intruder galaxy thatcrossed in front of galaxy and was slung around by theirgravitational attraction. (Middle) A rare and spectacularhead-on collision between two galaxies – the CartwheelGalaxy. The striking ring-like feature is a direct result of asmaller intruder galaxy that passed through the core of thehost galaxy. The collision sent a ripple of energy into space,blowing gas and dust in front of it. The Cartwheel Galaxywas probably just a normal spiral galaxy before thecollision. (Bottom) Two colliding galaxies nicknamed ‘TheMice’ because of their long tails. These galaxies willeventually merge in one single galaxy.

178

WAV

ES O

F TH

E FU

TURE

Figure 8 shows a few examples caught by theHubble Space Telescope; in these cases, thedistortion is clearly visible from the distribution ofthe stars. In less extreme cases, the stars do notshow distortions and one has to look at thedistribution of the gas to find some more subtleeffects. Spectacular cases of galaxies distortedby the effect of a collision or by interaction havebeen known for a long time, but they werepreviously regarded as kind of ‘weird’ cases.Deeper and deeper observations are showing thatthese are probably just the ‘tip of the iceberg’,and interactions and merging – albeit in a lessspectacular way – could affect the life of aconsiderable fraction of the galaxies in the sky.

Evolving galaxiesIt then follows from the above that even hugesystems like galaxies can have a ‘private life’,evolving and changing exactly like a humanbeing. The times involved in these changes areenormous compared to, for example, the humanlifespan or even timescales of human civilization.However, from their morphology and physicalcharacteristics it is possible to derive informationabout their evolutionary stage and, in particular,whether their quiet life has been disturbed by aclose encounter.

As shown in Figure 8, interaction with a nearbycompanion can produce tail-like structures. Thetails are the result of stars and gas pulled out ofthe objects during the gravitational interaction.Before looking in more detail at which kind ofinteractions can affect the life of a galaxy, it isimportant to understand that these tails and thepresence of interaction are not always so evidentfrom the optical images, because they trace onlythe distribution of stars. A much more powerfulway is to observe the distribution of atomichydrogen. The most common element in theUniverse, atomic hydrogen is believed to beassociated with the very origin of the galaxies.

Atomic hydrogen (or ‘neutral’ hydrogen) emits aspectral line with a wavelength of 21 cm, and istherefore observable in the radio band using radiotelescopes (see Figure 9). In galaxies, the atomichydrogen usually extends out to much larger radiithan the stellar component (see Figure 10) and

because of this, it is less gravitationally bound tothe galaxy. Thus, atomic hydrogen is much moresensitive to disturbances produced by the passageof another galaxy and this is shown by thepresence of tails, plumes and, in general, by thestrange distribution of the gas. Figure 11 showsthe example of NGC 4631 where the effects ofinteraction are only visible in the neutral hydrogenwavelengths; observing in the optical band, thegalaxies appear more or less undisturbed.

Neutral hydrogen has another interestingproperty. Because it can extend to much largerradii compared to the main stellar distribution, itcan also provide a better measure of how much

Figura 9 - Observations of the neutral hydrogen 21-cm lineare done using radio telescopes. In order to reach therequired resolution (in other words, to be able to producesharp enough images that can reveal as many as possibledetails in the structure of the radio sources) many radiotelescopes are working as interferometers. This means thatthey are formed by a large number of single antennas andthe final image is made by combining the signal from allthese antennas. The Figure shows five of the six dishes ofthe Australia Telescope Compact Array located about 500km from Sydney.

Figure 10 - The almost face-on spiral galaxy NGC 6946. Onthe left, a true-colour optical image (based on images fromthe Digital Sky Survey) is shown while on the right a deepimage of the neutral hydrogen obtained with theWesterbork Synthesis Radio Telescope reveals how muchextended the gas is compared with the stars.

179

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

mass is in the galaxy. Essentially, the faster theneutral gas rotates (in regions as far as possiblefrom the centre), the more mass is present insidethat radius. This can be compared with the massof the visible matter (stars or gas) to estimatewhether the mass of the galaxy is all due to this(visible) matter, or whether there is more hidingsomewhere. The rotation velocity of the neutralhydrogen is often observed to stay almostconstant outside the region where the stars areobserved. But if the matter in the galaxy weredue predominantly to the presence of stars, thevelocity of the gas would have been expected todecrease with distance from the edge of theoptical body of the galaxy. The fact that this doesnot happen means that there is some othermatter – the so called dark matter, as it has notyet been observed directly, only indirectly – thatkeeps the velocity of the gas from decreasing.The presence of dark matter around galaxies is acrucial ingredient in the theoretical models ofgalaxy formation.

Coming back to the evolution of a galaxy, thereare different ways in which a galaxy can change:■ Slow and quiet accretion of gas due to

gravitational attraction can happen if a smallgalaxy passes near a large one. The smallcompanion may or may not be completely

destroyed. In any case the large galaxy can geta new input of gas that can be used, forexample, to form new stars.

■ If the interaction is much closer, the largegalaxy swallows a small satellite that hits orpasses very close to it (a so-called ‘minormerger’). This process will not only bring newgas but may also have a major impact on thekinematics (and perhaps morphology) of themain galaxy.

■ A ‘major merger’ happens when two galaxies,of about equal mass, get close enough tocollide and merge. This gives rise to a ‘new’galaxy that can be quite different from theprogenitors!

■ Finally, we should not forget that the normalcycle (from life to death) of stars affects theinterstellar medium of the galaxy. For example,the supernovae phase is crucial in ‘polluting’the medium. This happens in the life of everygalaxy. In addition to this, some galaxies mayexperience a phase of intense star formation(the ‘starburst phase’).

The nearest interacting system: ourMilky WayIdeally, one would like to study the theory ofgalaxy formation and evolution by observing thedistant Universe. However, this does not alwaysgive all the physical details needed by theastronomers. Thus, the objects in the nearbyUniverse become important targets; the extremeexample is the study of stars and gas in our ownMilky Way.

A giant stellar stream surrounds the Milky WayGalaxy, originated by a companion dwarf galaxycalled the Sagittarius dwarf. Furthermore, a

Figure 11 - An example of galaxy interaction detected byobserving the neutral hydrogen: NGC 4631 (upper galaxy)and NGC 4656 (lower galaxy). On the left, an optical imageof the field where no obvious sign of interaction is seen, onthe right the same field but with superimposed contoursrepresenting the emission from the neutral hydrogenobtained with the Westerbork Synthesis Radio Telescope(regions with more contours correspond to a strongeremission). The large tails of gas between the interactinggalaxies are clearly visible.

180

WAV

ES O

F TH

E FU

TURE

narrow tail of neutral hydrogen, called theMagellanic Stream, has been known for manyyears. This stream trails the (Small and Large)Magellanic Clouds – the two closest companionsof the Milky Way – in an orbit around the Milky Way.

More recent studies with the Parkes radiotelescope in Australia have shown that thisstream of neutral hydrogen is much moreextended than previously thought, and it extendsin the opposite direction to what astronomersexpected (see Figure 12). Astronomers considerthis a further confirmation of tidal interactionsbetween the Magellanic Clouds and the MilkyWay. This means that the Magellanic Clouds willbe slowly torn apart and destroyed, and their gasand stars will become part of our Galaxy.

A stellar stream has been also found around theAndromeda Galaxy (or M31), the nearest largegalaxy to the Milky Way. The stream found there,coming from a small companion of this galaxy,supplies Andromeda with gas and stars. This

supports the idea that the epoch of galaxyformation is continuing even now, although at amuch slower rate. These streams are perhaps ageneric feature of almost all galaxies.

While it is very difficult to see the stellar streamsin the visible spectrum in galaxies outside theregion of our Local Group, it is much easier todetect them in neutral hydrogen – as we shallsee shortly.

Stealing from your neighbourAs our telescopes become increasingly sensitivethanks to improved technology, we discover thatmany galaxies have very likely experienced (atleast once in their lifetime) some kind ofinteraction. Some examples have been alreadyshown in Figure 8 and a few more will bedescribed below – but in fact, there is an endlesslist of cases. The web links at this end of thislecture will guide you to some of the sites whereyou can see more examples of them.

Figure 13 - An optical image of the galaxy NGC 3359 (greyscale) with superimposed contours representing theintensity of the emission from the neutral hydrogen (densercontours represent regions of more intense emission). Thedata were obtained with the Westerbork Synthesis RadioTelescope. The faint tail-like emission connecting the biggalaxy NGC 3359 with the faint companion can be seen.

Figure 12 - Image of thedistribution of the neutralhydrogen in theMagellanic Stream.Darker regions representzones where the signal ismore intense. The twodark clouds correspondto the Magellanic Cloudsconnected by a bridge ofneutral hydrogen. The tailgoing up from theseClouds is the MagellanicStream. The Milky Way isjust outside the figure atthe bottom.

181

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

As mentioned above, galactic interactions can bequite gentle. A large galaxy can gravitationally pullgas out of a smaller companion. Figure 13 showsa case where a faint connection of neutralhydrogen is visible between a large galaxy, NGC3359, and its distorted companion (almostinvisible in the optical image). The neutralhydrogen transferred from this small companionis less than 3% of the atomic hydrogen mass ofthe large galaxy. This kind of interaction is verymuch reminiscent of that seen in the case of theMilky Way and M31.

This interaction is not producing any majorchange in the morphology and characteristics ofthe large galaxy. The effect of this slow accretion(that in some cases can go on for billions ofyears) is mainly to stimulate new star formation inthe large galaxy, both through the new gassupplied by the unlucky companion and from thecompression of the gas produced by theinteraction itself. This is one of the ways tocontinuously supply gas to a galaxy and keep it growing.

Swallow a fly The small companion can, however, be much lessfortunate. Depending on how the encounterhappens (the so-called initial conditions, like therelative direction of the encounter, the relativerotation of the two objects, etc.) the effects forthe main galaxy can be very different. The middleimage in Figure 8 shows one of the nicer examplesof head-on collision, with a small galaxy hitting alarger one – the famous ‘Cartwheel’ galaxy. Thecollision has produced a shock wave that is nowpropagating at a speed of 300,000 km/h throughthe surrounding gas. The compression of thisshock wave in the gas is making the attractivering-like feature around the galaxy.

Finally, the small companion can end upcompletely captured and swallowed by the largegalaxy. This must have happened in one of themost famous galaxies in the sky, Centaurus A.The main body of this elliptical galaxy is dividedin two by a spectacular lane of dust. At theoptical wavelengths, the dust appears as darkpatches where the light from the stars has beenabsorbed (see Figure 14a). Structures like dust

lanes are thought to be the left over from a small,gas-rich companion that was swallowed by thebig galaxy. This galaxy, only about 10 millionlight-years away from us, is the nearest largegalaxy of this type (despite the large dust lane, itcan be considered in many respects an elliptical

Figure 14 - (a) Optical image of the galaxy Centaurus A(also known as NGC 5128) obtained with the 8-m VLT ESOtelescopes in Paranal (Chile); (b) an infrared image of thesame galaxy obtained by the Spitzer Space Telescope; (c)optical image of Centaurus A (taken from the Digital SkySurvey) with superimposed contours representing theemission from the neutral hydrogen obtained with the VeryLarge Array (more contours indicate stronger emission).Most of the neutral hydrogen follows the dust lane butsome ‘left-overs’ can be seen in a kind of half-ringstructure.

(a)

(b)

(c)

182

WAV

ES O

F TH

E FU

TURE

galaxy) – it also hosts at its core an active blackhole emitting copious amounts of radio waves!

It represents, therefore, a great laboratory formany phenomena. We will meet this galaxy againin the next lecture. Figure 14b shows howdifferent the galaxy looks when observed in theinfrared. At these wavelengths, the dustreradiates the light absorbed from the stars.Astronomers believe that the peculiar geometricalshape of the dust emission is the result of thetwisting and warping of the infalling spiral galaxyas it was captured by the large elliptical.

Another reason why this galaxy is considered theresult of a merger with a small companion is thatelliptical galaxies are usually found to be verypoor or lacking in gas. However, when observedin the 21-cm line, Centaurus A shows emissionfrom the neutral hydrogen (see Figure 14c)indicating that a relatively large amount of thisgas is present in this galaxy. This can onlyhappen if the gas has been recently ‘donated’ tothis galaxy by a merger with a gas-rich object.Figure 14c shows that the neutral hydrogen ismainly concentrated along the dust lane, and thisis consistent with the idea that this structure (aswell as the neutral hydrogen) has an externalorigin. However, some gas is also found at evenlarger distances from the centre (60,000 light-years or more) and is considered to be ‘left overs’from the merger.

Both in the case of the Cartwheel and in the caseof Centaurus A, the merger has had a clear effectalso on the large galaxy – that is, the galaxy thathas swallowed the fly! In particular, after themerger the stars and gas in the galaxy can displaysome very peculiar kinematics. Astronomers havefound amazing cases of stars rotating in onedirection while gas rotates in the opposite direction!In other cases, stars and gas rotate aroundcompletely different (often perpendicular) axes. It isvery difficult to imagine how such complexkinematics can be possible, and the only wayappears to be if the gas has been acquired later(once the stars in the galaxy were already wellformed) through an encounter. In this case, thekinematics of the gas depends on the geometry ofthe encounter, and therefore can be completelydifferent from the kinematics of the stars.

Violent CollisionsThe most impressive galaxy mergers are thosebetween two galaxies of similar size (known asmajor mergers). The system known as theAntennae represents one of the most amazingcases; Figure 15 shows the extremely complexstructure in detail. The system consists of thecollision between two galaxies, and the strikingfeatures of the picture are the two huge‘antennae’ that represent gas and stars pulledout of the galaxies during the encounter.Numerical simulations show that these structuresare indeed the result of such an encounter if thegalaxies passed by each other in the same senseas the rotation of each disk of stars (so-calledprograde encounter). In this case, the outer ringsof stars and gas are ripped off from each of thegalaxies, resulting in the formation of thespectacular tails.

Figure 15 - The two colliding galaxies known as Antennae(NGC 4038 on the top and NGC 4039 at the bottom). Thefigure on the left shows the neutral hydrogen in blue (fromobservations with the NRAO Very Large Array) super-imposed on an optical image from the CTIO 0.9m in greenand white. This image clearly shows the long tails createdby the interaction. The image on the right shows the centreof the two interacting galaxies as observed by HST.

There are many cases known of major mergers andby studying their characteristics it has been possibleto construct a sort of evolutionary sequence.For example, the Antennae still show the twoseparate bodies of the merging galaxies andprominent tails of stars and gas. Large quantitiesof neutral hydrogen (many times bigger than our

183

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

Figure 17 - The optical image of the elliptical galaxy NGC 5266 (yellow) superimposed with the emission from the neutralhydrogen (red) as observed with the Australia Telescope Compact Array. More than 10 billion solar masses of neutral hydrogenare observed around the galaxy, believed to be the result of a merger between two large and gas-rich spiral galaxies. Notehow much more extended is the emission of the neutral hydrogen compared to the stellar light.

Figure 16 - Acomposite image ofthe merger remnantNGC 7252 showingthe optical light(green), star formingregions (yellow andpink) and the coldatomic hydrogengas (blue) observedwith the Very LargeArray. This system isthe result of twospiral galaxies,which collided andmerged into a singleobject.

184

WAV

ES O

F TH

E FU

TURE

Galaxy) are observed from the centre of themerger to the end of the tails. This merger likelystarted a few hundred million years ago.

Figure 16 shows instead the case of NGC 7552,also the result of a major merger, but one that ismuch more evolved. The two progenitor galaxiesare already so well amalgamated that now theyare just one object. This is extremely interestingin our quest to understand how galaxies formbecause the resulting object now resembles anelliptical galaxy (although the long tails still clearlyindicate its merging origin). The merger in NGC7252 probably started about a billion years ago.Interestingly, a large amount of neutral hydrogenis still present but it is mainly concentrated alongthe tails. By now, the atomic hydrogen in thecentral region has been converted into stars; themerger has gone through an intense phase ofstar formation as a result of the violentcompression of the gas during the initial phase ofthe merger.

If we are patient enough to wait few more billionyears (!) the merger will finally produce a realelliptical galaxy as shown in Figure 17. By now,the long tails have faded away and the atomichydrogen (once that the starburst phase is over)has slowly moved back to the galaxy under theeffect of gravity. The resulting elliptical galaxy will

still be rich in atomic hydrogen and thischaracteristic allows galaxies that have formedthrough major mergers to be recognised. Indeed,NGC 5266 (the galaxy shown in Figure 17) hasmore than 10 billion solar masses of atomichydrogen, one of the largest amounts found inan elliptical galaxy.

Astronomers have put this scenario togetherbased on observations. However, computersimulations also show us that, under certain initialconditions (admittedly some of them quiteuncertain), the characteristics of the observedmergers can be nicely reproduced. In Figure 18we can see one example of how a numericalsimulation reproduces different steps of themerger. The similarities with the images of realobjects in Figures 16, 17, 18 are remarkable.

A final important remark: it is interesting to noticethat the gas brought by the merger will be usednot only to form stars but it will also provide fuelto the central regions of the galaxy and, inparticular, to the central black hole. In this way,as we will see in the next chapter, the black holecan become active and produce enormousamounts of energy at different wavelengths andin different forms!

New gas, new stars!One of the effects of accreting new gas either ina quiet way or through a violent merger, is tostimulate new star formation. In fact, mergers notonly bring new gas but also induce compressionof the gas that then can trigger star formation –in the high density, condensed gas, gravity canstart to dominate, giving rise to the subsequentproduction of stars. This process, if taken to theextreme as in the case of the so-called starburstgalaxies, can have a major impact on thestructure, morphology and evolution of the galaxy.

There are many spectacular cases of galaxies goingthrough a starburst phase. Figure 19 shows theimpressive case of NGC 3079. Some of theobserved starburst galaxies have a star formationrate exceeding 100 stars per year, more than 100times the rate in our galaxy. It is clear that thisextreme rate has a major influence on thegaseous medium in the galaxy, and on the

Figure 18 - A computer simulation showing the evolutionwith time of a merger between two gas rich spiral galaxies.The time interval (time increasing from left to right)between frames is roughly 100 Million years. Thesimulation shows clearly the creation of large tidal tails(similar to those observed in the Antennae) that fade awayas time goes by. At the end only the central object (mergerof the two progenitors) will remain.

185

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

galaxy’s structure. At such high rates of starformation, the number of supernovae that willexplode (when the stars reach the end of their life)is also about 100 times the rate in our galaxy.

Each supernova that explodes will produce ashock wave through the gas medium around it,which creates a kind of bubble. If many

supernovae explode at about the same time, theoverall effect is the creation of a super bubble,a bubble of hot gas (heated by the shock waves)so large that it will travel through the galaxy until itbreaks free in the intergalactic medium. It thenbecomes more of a wind of hot gas that can pushaside everything in its path. The consequence forthe galaxy can be enormous. Figure 19 showshow spectacular these winds and their associatedgas outflow can be.

Large amounts of dust, the first by-product ofstellar formation, characterise star formingregions. The important characteristic of dust isthat it can obscure these regions when they areobserved at optical wavelengths: dust absorbsoptical radiation and re-emits it at longerwavelengths, in particular in the infrared.Observations in this band of the electromagneticspectrum can only be done either from outsidethe atmosphere or from telescopes located athigh altitude.

Figure 19 - The spectacular starburst galaxy NGC 3079. AChandra’s X-ray image (blue) has been combined with anoptical image from the Hubble Space Telescope (red andgreen). The filaments consist of warm (about ten thousanddegrees Celsius) and hot (ten million degrees Celsius)formed by a super wind produced by a burst of supernovaactivity or from the super massive black hole.

Figure 20 - The distribution of morphology of galaxiesextracted from the Hubble Deep Field (see Figure 2) butgrouped in order of increasing redshift, corresponding toincreasing distance from us.

186

WAV

ES O

F TH

E FU

TURE

Only in recent years have technical improvementsmade possible the study of starburst galaxies,which are otherwise mostly obscured at theoptical wavelengths. These capabilities areparticularly crucial for the study of thisphenomenon in the distant Universe.

Galaxies in the distant UniverseThe objects considered so far are part of the localUniverse – this means that what we observecorresponds to how they look today. Thedistances from Earth to the galaxies we haveseen so far are not more than a billion light years,and therefore the light we see now left thesegalaxies a billion years ago. This is less than atenth of the estimated age of the Universe, whichis about 13 billion years.

On one hand, it is very important for astronomersto study these objects because of the detailedinformation that can be obtained and thencompared with theoretical models. On the otherhand, if we want to know how galaxies form andevolved, the distant universe is the place wherethe main ‘action’ is happening. Interactionhappens more often in dense environments, andso interactions between galaxies were probablyextremely frequent (and therefore extremelyimportant in shaping the galaxies) in the early

Universe. In this phase, the Universe was muchmore dense than the one of today, because it was much smaller. It is, therefore, not surprisingthat a lot of effort from the astronomicalcommunity is now concentrated on the study ofdistant objects. Because of their distance, theseobjects are small and faint and making theseobservations, even with state-of-the-artinstrumentation, is extremely challenging.

One could write an entire lecture (even a wholeseries!) on the recent studies of distant galaxies,but let’s just mention here two of mostinteresting recent results. Extremely remoteobjects have been observed that are so far awaythat it corresponds to when the Universe wasonly about one-tenth of its current age. Themorphology of these distant galaxies appears, asexpected, much more distorted than in thegalaxies that populate the local Universe (seeFigure 20). Almost all the observed distantgalaxies show some kind of distortions and theseparation in spiral and elliptical galaxiesbecomes much more complicated, if almostimpossible. This indicates that interaction andmerging do seem to play a major role in thisphase of the Universe. However, some of theobjects observed seem to have at least thecentral bright part, looking very much like anelliptical galaxy. This would suggest that elliptical

187

THE

EVER

CH

ANG

ING

LIFE

OF

GAL

AXIE

S

galaxies can form quite early in the life of theUniverse – but to make a nice, proper spiralgalaxy could take much longer.

Another remarkable discovery of the last fewyears is the existence of a population of distantgalaxies that appear to be forming stars at veryhigh rate. These galaxies are among the mostluminous objects in the Universe and are poweredby a star burst (and probably also by a nuclearblack hole), with star formation rates up tohundred times higher than the typical starburstgalaxy observed in the present-day Universe (soup to 10,000 times the star formation rate of ourgalaxy). These galaxies were discovered usingsub-millimetre observations (wavelengthsbetween radio and infrared). Now called ‘sub-mmgalaxies’, these galaxies are mostly invisible (orvery faint) in the optical band while they are verybright in sub-mm – this is thought to be theresult of the large amount of dust originatingfrom the rapid star formation going on in these objects.

The few of these galaxies that have been studiedin detail almost exclusively exhibit disturbed andunusual morphologies, in line with the idea thatthe large rate of star formation is the result ofmergers, as expected in the hierarchical structureformation scenario.

In conclusion, many open questions still remainfor astronomers over the formation and evolutionof galaxies. Nevertheless, they will continue onthis endeavour using both the detailed studies ofthe nearby spectacular cases and the study ofthe mysterious objects at the edge of theUniverse. Improvements in technology that willproduce the next generation of telescopes will becrucial in making this difficult task possible.

Some interesting Web links:

Hubble Telescope Picture Gallery and Outreach:

http://hubblesite.org/gallery and

http://www.stsci.edu/outreach

ESO outreach page: http://eso.org/outreach

Astronomical photographs from David Malin images:

http://www.davidmalin.com/

NRAO image gallery: http://www.nrao.edu/imagegallery/

A living HI Rogues Gallery:

http://www.nrao.edu/astrores/HIrogues/RoguesLiving.shtml

Numerical simulation of interacting galaxies in the home page

of Joshua Barnes

http://www.ifa.hawaii.edu/faculty/barnes/barnes.html

Web links to see how radio telescopes work:

http://www.astron.nl/p/astronomy2,

http://outreach.atnf.csiro.au, http://www.nrao.edu/students

188

WAV

ES O

F TH

E FU

TURE

But when you add up the mass ofthese two atoms and the extraneutrons, the total is less than theoriginal Uranium atom’s mass.Where did the extra mass go?

When the Uranium splits, theneutrons that escape have a lot ofenergy – this energy is the“missing mass”. When you use thefamous E = mc2 equation, andfactor in the energy carried by theneutrons, you find everything adds up.

This famous equation didn’t appearin Einstein’s first paper on SpecialRelativity, which was described inPart 3 of this series. In Septemberof 1905, he published yet anotherpaper, titled Does the Inertia of aBody depend on its EnergyContent? (Annerlen der Physik 19,page 639). In this paper, Einsteinconsiders a body emitting aquantify of energy in the form ofradiation; he concludes that:

If a body releases the energy [E]in the form of radiation, its massdecreases by [E/c2]. Sinceobviously here it is inessentialthat the energy withdrawn fromthe body happens to turn intoenergy of radiation rather thaninto some other kind of energy,we are led to the more generalconclusion: The mass of a bodyis a measure of its energycontent; if the energy changesby [E], the mass changes in thesame sense by [E/c2].

If the theory [this derivation of m= E/c2] agrees with the facts,then radiation transmits inertiabetween emitting and absorbingbodies.

In this quote from the paper, thesquare brackets indicate where themodern symbols for mass, energyand the speed of light are usedinstead of Einstein’s originalsymbols.

Moving around this derivation, we ofcourse wind up at E = mc2. Notonly does mass seem to beequivalent to a form of energy, butenergy ‘transmits inertia’ betweenbodies. We’re used to thinking ofmass as something fixed and solid,so Einstein’s conclusion that themass of an object depends on howmuch energy it has came as a bitof a revelation. In particular, if anobject is moving, it has kineticenergy – and so an object’s massincreases the faster it moves!

But this isn’t just about the energyan object has through motion.Other kinds of energy –gravitational potential energy,electric potential energy and so on– change the mass of an object aswell. One particularly interestingapplication of this equation occursin nuclear physics, when an atomof Uranium is split, for example.The Uranium atom decays into twolighter atoms – Barium and Krypton– and several stray neutrons.

The Final Paper of 1905: that Famous EquationASK ANYONE ON the street, What equation do you associatewith Albert Einstein?, and most will answer, E = mc 2. Theymay not know what the equation means, but they know it’sclassic Einstein.

Einstein’s Miraculous Year /Part 4

The quantity of energy you can getin this way is apparent when youconsider the energy output ofnuclear power stations and, evenmore frighteningly, nuclearweapons.

189

EIN

STEI

N'S

MIR

ACUL

OUS

YEA

R,PA

RT 4

190

Monsters lurking in thecentre of galaxiesRaffaella Morganti

T H E U N I V E R S E I S F U L L with strange objects, but Active Galactic Nuclei (or AGN) are among the most spectacular of them all. When we look atgalaxies in the sky using an optical telescope, most of them may appearlike relatively normal elliptical or spiral galaxies. However, some of themhide a monster in their very centre. In these galaxies, the nucleus alone– that means a region not more than a few light-years in size – can beup to 10,000 times more luminous than the rest of the entire galaxy!

191

Figure 1 - Four examples of Active Nuclei: (below left) the Seyfert galaxy NGC 1068; (below centre) the quasar 3C273observed by Hubble Space Telescope and two radio galaxies Centaurus A (opposite) and 3C31 (below right). The image of3C31 is the superposition of a radio image (red) and an optical (blue). The image of NGC 1068 is a composite X-ray (blue andgreen) and optical (red) image. The X-ray shows gas blowing away in a high-speed wind from the vicinity of the centralsupermassive black hole. A composite image (opposite) is shown for Centaurus A: Chandra X-ray image (blue); radio 21 cmimage and continuum (pink and green); optical (yellow).

192

WAV

ES O

F TH

E FU

TURE

F or a while this incredible amount of energyreleased by these nuclei puzzledastronomers. The energy from stars (even

considering a large number of them squeezed inthis tiny space) is not enough to explain it. Amachine capable of doing this amazing job has tobe exceedingly massive and the only candidate isa super-massive black hole. Thus, we candescribe an AGN as a ‘very compact region at thecentre of a galaxy emitting a large amount ofenergy that cannot be produced only by stellaremission’. The galaxy (either elliptical or spiral)hosting this active nucleus is then called anactive galaxy.

The phenomenon of AGN can manifest itself inmany different ways and it is impossible tosummarize all in one chapter of this book. Figure1 shows some examples of different types ofAGN: a so-called Seyfert galaxy, a quasi-stellarobject (quasar) and two galaxies particularlystrong at radio wavelengths (radio galaxies).Some of the effects of the active nucleus in theseobjects are shown in the pictures:■ the strong X-ray emission coming from the

active nucleus of the Seyfert galaxy, that wouldotherwise be a normal spiral galaxy;

■ the very bright nucleus of a quasar – whichmakes this object look like an ordinary star, butit is in fact several billion light-years away. Inthis object nothing stops us from lookingdirectly at the powerful active nucleus;

■ the striking difference between the emissiondetected at radio and X-ray wavelengths,compared to the morphology of the opticalemission in the Centaurus A galaxy (an objectthat we have met already in the previouschapter); and

■ the different morphology, again, but also thestrikingly different size of the radio emissioncompared to the optical emission in the radiogalaxy 3C31.

AGN emit not only in the optical but also in everyother waveband. In fact, Active Galactic Nucleiemit an enormous amount of energy over almostthe entire electromagnetic spectrum. Theradiation observed at different wavelengthscomes from different regions inside the activenucleus, and from different mechanisms. To makeeverything even more complicated, there are AGN

that emit a lot of their energy in the radio band,while others emit more in the X-ray band. Thereason for these differences is still not completelyunderstood; it could be related to some subtledifferences in the structure of the various AGN orin the way they have formed. This makes, ofcourse, the job of classifying and understandingAGN very complicated. Despite these diversities,astronomers think that all AGN have somecharacteristics in common and therefore they canbe grouped under only few categories. We shallsee why shortly.

How many galaxies host an AGN? The presenceof a central black hole in galaxies could berelatively common. Even our galaxy, the MilkyWay, hosts one, although it isn’t so very big.However, active galaxies are only a small fractionof the total number of galaxies in the Universe. Ofthe relatively nearby galaxies, only at most a fewpercent of them host an AGN.

Thus, the presence of a very massive andcompact object, the black hole, in the centre of agalaxy is not enough to turn a quiet nucleus in anAGN. The black hole must be active – into otherwords it should be supplied with fuel that allowsit to produce the huge amount of energy that weobserve. Although they are relatively rare, thestudy of this phenomenon is extremely important:the phenomenon is so extreme that we cannotavoid being curious and making an effort tounderstand the physics behind it. Moreover, it canhave a major influence in the life of the galaxy.The AGN phase is likely to be very destructive forthe galaxy – the release of so much energy canhave a major impact on, for example, theinterstellar medium and therefore even affect thegalaxy’s subsequent evolution. So the study of theAGN phase is necessary for understanding thelife of a galaxy.

It is clear that AGN are exotic and extremelyinteresting objects. Despite being one of the ‘hottopics’ in astronomy for many years, AGN stillkeep astronomers very busy trying to answer themany open questions about this complexphenomenon. Because the subject is so vast, thischapter can only cover a very limited part. Inparticular, we will see what astronomers think thestructure of an AGN looks like, how the energy

193

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

may be produced and released, and someexamples of these incredible objects, in particularthe radio galaxies.

Just a little bit of historyIt is interesting to see how AGN were discovered.In fact, this is a relatively recent discovery thathappened when the astronomers finally were ableto look at the sky at wavelengths different fromthe optical.

Radio astronomers were the first to realize thatsomething very strange was happening in somegalaxies. Starting in 1946 with very basicinstruments, Australian radio astronomers wereable to detect at least two strong radio sourcesthat were perhaps associated with extragalacticobjects. This seemed unbelievable, as it meant anincredible amount of energy was being released.The first radio telescopes, because of their smallsize, could only produce very fuzzy images, so itwas very difficult to identify with certainty the‘optical counterpart’ of the strong radio signal –in other words, to figure out which optical object(star or galaxy) was producing the radio signal.

It took some effort and creativity (even using theocean as a mirror!) before the radio imagesbecame sharp enough to be able to solve thedilemma. When finally the distances of theseobjects from Earth were estimated (at that timethought to be up to 100 million light years), it wasclear that the amount of energy that they wereemitting was far beyond any human imagination!The fact that such a distant radio source could bedetected made it clear that radio astronomy couldextend the boundary of the observed Universe farbeyond what was accessible to opticalastronomers.

Since then, a zoo of AGN has been discovered –we now know that AGN indeed come in differenttypes and flavours, and can live in the centre ofdifferent types of galaxies. As it turned out, AGNthat are also strong radio sources are only afraction of the active nuclei, but they still form animportant and fascinating group.

How does an AGN work?If stars alone are not sufficient to produce theenergy released by an AGN, another mechanismthat is more efficient needs to be found.Astronomers believe that the radiation producedby an AGN comes from converting matter intoenergy. This is made possible by the presence ofa very massive black hole. Such a massive bodyworks like a cosmic sink: all the stars and gasthat happen to be close enough are ‘sucked in’by the black hole’s gravity field. This is the onlymechanism that can give the huge amount ofenergy we observe. The emitted radiation isproduced because gravity converts the potentialenergy of the infalling matter into kinetic energy.Collisions between infalling particles convert thekinetic energy into thermal energy, and photonscarry this energy away.

The matter falls toward the black hole in anorganized way: angular momentum causes matterfrom the stars and gas to circle around the blackhole in a so-called accretion disk (see Figure 2). Bydefinition, a black hole is an object with such a highdensity (we are talking about 100 million solarmasses compressed in a volume about the size ofour solar system!) that its gravitational field isincredibly strong – at a certain point, the infallingmatter passes through the black hole’s event horizonand, at that point, the gravity becomes so strong thatnot even light can escape. For this reason, a blackhole cannot be observed directly. Even the regionaround it is so small that today’s telescopes cannotsee its structure.

Figure 2 - An artistic illustration of the centralregions of an AGN.

194

WAV

ES O

F TH

E FU

TURE

So how do we know that black holes really exist?The existence of a massive object in the centre ofgalaxies has been derived mainly throughobservations of the kinematics of the gas in theinner regions of the galaxy – at least, thoseregions that we can observe. The gas in theseregions has been seen to rotate: on one side ofthe orbit the gas comes toward us, on the otherside it moves away from us. From the amplitudeof the rotation velocity, the amount of mass withinthat region can be derived. Similar measurementswere done using maser emission from watermolecules (see Figure 3 – maser stands for‘microwave amplification by stimulated emission ofradiation’, a mechanism similar to the laser beam,but with microwaves instead of visible light).

The super-massive black holes in the centre ofpowerful AGN are estimated to have a massbetween 10 million and few billion times themass of our Sun.

Radio JetsSo far, we have seen that material goes into theblack hole, and this infalling material is convertedin energy. What happens then to the energyproduced? A variety of mechanisms can explain

the energy radiated across the electromagneticspectrum by these nuclear regions. For example,because of their high temperature, the accretionregions are mainly observed at the X-raywavelengths. On the other hand, at radiowavelengths, the energy is ejected through verycollimated (narrow) jets of radio waves. How doesthis happen?

The radio emissions from AGN are due tosynchrotron radiation, which occurs whenrelativistic electrons (electrons accelerated tovelocity close to the speed of light) are present ina magnetic field. The structure of the magneticfield (see Figure 4) probably channels theelectrons into the collimated structure that wesee, called jets.

To the ‘first order’ (which is a scientist’s way ofsaying that in reality it can be much morecomplicated than this) the direction of thesecollimated structures is along the rotation axis ofthe accretion disk. This is illustrated in Figure 5 inthe case of the radio galaxy NGC 4261.

The active nucleus also emits ultraviolet (UV)radiation, which can profoundly influence the gasaround the AGN by ionising it. This means it canprovide the energy to extract electrons from theatoms and give rise to emission lines that areobserved mainly at optical wavelengths. Thestudy of these emission lines has played, sincethe discovery of AGN, a crucial role in ourunderstanding of these objects. Two examples of

Figure 4 - An artistic concept of the formation of a radio jet.

Figure 3 - The detection of fast rotating gas in the nuclearregions (0.2 parsec in size corresponding to only about 0.6light year) of the galaxy NGC 4258 indicating the presenceof a black hole of 40 million solar masses.

195

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

optical spectra observed in AGN are shown inFigure 6. The obvious difference between the twois that in one case the lines are very broad. Thesebroad lines have been interpreted as coming fromgas that is located only at most few light yearsaway from the active nucleus, and it is thereforemoving quickly under the effect of the gravity ofthe black hole.

Galactic DonutsAt this point, it is also important to note thatastronomers think that outside the accretion disk,a thicker, donut-shaped structure surrounds theAGN (called the torus). This is illustrated, in anartistic way, in Figure 2. A thick structure like thiscan obscure the very nuclear regions around theblack hole. This is a crucial element because itmeans that, depending how this structure isoriented, one can or cannot detect the emissioncoming from the very inner part of the AGN (forexample the broad emission lines). Astronomersthink that orientation effects can explain some ofthe differences observed between AGN.

The exact size of the torus is not really known asits inner edge is supposed to be quite close tothe black hole, while the outer edge could extenda thousand light years or more from the centre.Indeed the high resolution provided by the opticalobservations of the Hubble Space Telescope(which, because it orbits the earth in space, doesnot suffer from the effect of the atmosphere inblurring the images) has shown the existence ofthese structures in many galaxies (see Figure 5).

If we can detect the broad lines, we can be surethat we are looking directly at the nucleus of theobject. When we do not observe such broad linesit indicates that part of the very central region isperhaps obscured from view because of thetorus. Optical and UV radiation can be emitted, inprinciple, in all directions, but the presence ofstructures around the black hole, like the torus,can force this emission to be confined to a cone-like structure. Evidence of this comes from thedetection of just such a cone structure in theSeyfert galaxy NGC 1068 (see Figure 7). We willsee later how all this influences our knowledgeand classification of AGN.

Figure 6 -Two examples of optical spectra of AGN. In bothemission lines from many different elements are detected.However, in the spectrum at top some of the emission linesappear very broad while on the spectrum above all theemission lines are quite narrow. The units in the horizontalaxis are Ångstroms (1 Å = 10-8 cm).

Figure 5 - Panel showing (left) a ground based image withsuperimposed a radio image and (right) an image from theHubble Space Telescope of the radio galaxy NGC 4261.

196

WAV

ES O

F TH

E FU

TURE

One can already imagine that the injection of allthis radiation and particles from the AGN into thesurrounding medium of the host galaxy can havea major impact on this medium. Fast gaseousoutflows reaching speeds of many thousands ofkilometres per second have been observed inmany AGN. Astronomers are beginning to realizethe presence of an active nucleus can have amajor impact in the life of the galaxy: forexample, it may blow away gas from the centralregions and therefore prevent the formation ofstars, or it may even create a self-regulatingmechanism that after a while stops the feeding ofthe AGN.

In the zoo of AGNGalaxies emitting strong radio signals were thefirst AGN discovered – however, after that anumber of other types of AGN were found. Thediscovery of quasi-stellar radio source (or quasarsfor short) was particularly surprising as theseobjects appear to be associated with star-likeobjects – but in fact they represent very distantextragalactic objects. One of the first quasars forwhich the distance was estimated turned out to

have luminosity of the order of 1039 watts,corresponding to 1012 times our Sun’s energyoutput – or a hundred times more powerful thanour entire galaxy.

Seyfert galaxies (from the name of theastronomer that classified them) are nearby AGN,mainly hosted by spiral galaxies. They wereoriginally identified as unusual because of thevery bright point-like nucleus. Theircharacteristics are, therefore, similar to those ofquasars although the AGN in Seyfert galaxieshave a much lower power. As with quasars,Seyferts also show very strong emission lines.

It is impossible in the short space of this chapterto give a full overview of all the different AGNtypes, so we will focus mainly on one kind: thegalaxies that are strong radio emitters, knownsimply as radio galaxies. It is important toremember, though, that the fact that they are soimpressive at the radio wavelengths does notnecessarily mean that the majority of the energythat they emit ends up in the radio band. Theradio photons are not very energetic, as theenergy carried out by photons is inverselyproportional to their wavelength (the longer thewavelength, the lower the energy). Thus, while‘radio-loud’ objects can look extremely impressivewhen observed with radio telescopes, the bulk oftheir energy is not at radio wavelengths.

The spectacular radio galaxiesThe first thing one notes when looking at a radiogalaxy with a radio telescope is the strikinglydifferent morphology of the radio emissioncompared to the optical (stellar) morphology.

Figure 8 shows the superposition of the radioemission of Fornax A (red) and surroundingoptical field (blue-white). The radio source isdominated by two large lobes well outside theoptical galaxy, the large elliptical galaxy visible inthe centre of the image. Figure 9 shows a muchmore extreme radio source and illustrates howgreat the difference can be between the size ofthe radio emission and that of the optical galaxy.

This difference clearly points to two completelyseparate mechanisms responsible for the origin

Figure 7 - Cone-like region of emission from gaseousclouds ionised by the intense radiation from the nucleus ofthe Seyfert galaxy NGC 1068. The nucleus is located nearthe base of the cone. This region is only few hundred lightyears in size, therefore many times smaller than the regionshown in Fig.1 for the same galaxy. This image has beenobtained by using a “narrow-band” filter that lets throughonly the light from the emission lines.

197

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

for these emissions. Electrons and protonsspiralling around magnetic field lines at nearly thespeed of light (synchrotron emission) produce theradio waves from these galaxies.

Figure 10 shows one of the most spectacular andwell-known radio galaxies, Cygnus A, as observedwith radio telescopes at different frequencies and,more importantly, with increasing resolution, inorder to explore the regions closer to the nucleus.

The host (optical) galaxy would be only about atenth of the size of the full radio emission, shownin the top image in Figure 10. The structure ofthe radio emission is very interesting and hasbeen crucial in giving clues on the structure ofthe AGN that we have described above. Thetypical radio structure shows a nucleus

Figure 11 - The radio galaxy Cygnus A with the variouscomponents the radio emission indicated.

Figure 10 - The radio source Cygnus A. This radio source isproduced in a galaxy some 600 million light-years away.The images show the radio structure seen at differentfrequencies and different resolution. Red (blue) coloursrepresent stronger (weaker) intensity. The two remainingimages have been obtained with high resolutionobservations (with the Very Long Baseline Interferometry).These images allow investigating in more detail the innerpart of the jet, the region closer to the nucleus. The imageat the bottom can resolve details of only 0.1pc in size(about 0.3 light years).

Figure 9 - Panel showing (left) the image of a radio galaxytaken at the wavelength of 20 cm with the Very LargeArray; (middle) the radio image (red) superimposed onto anoptical image (blue, from the Digital Sky Survey) and (right)the optical image marked with the galaxy responsible forthe huge radio emission. This last image is to illustrate theamazing difference in size between the optical galaxy andthe radio emission.

Figure 8 - The superposition of the radio emission ofFornax A (red) and surrounding optical field (blue-white)The radio source consists of two large and complex radiolobes. At the centre of the optical field is the ellipticalgalaxy from where the extended radio emission originates.

198

WAV

ES O

F TH

E FU

TURE

(coincident with the region of the black hole,accretion disk and torus) and two very narrowjets emerging from it. Two large lobes are seenwhere the jets end. The presence of the jetsclearly indicates that the energy (at least theradio plasma) is emitted, not in every direction,but in collimated structures. Typically, these jetsare also perpendicular to the nuclear dusty disksfound by HST as seen in Figure 5.

The radio jets connect the nucleus to the lobes,so although the radio lobes are far away they areindeed fed by the active nucleus. Jets are thoughtto be the channels along which the acceleratedelectrons travel to very large distances from thenucleus. If the jet is powerful and fast enough,the strong interaction with the medium around itwill produce a ‘shocked’ and bright region –which shows up as ‘hot spots’. Figure 11illustrates all these different components in thecase of Cygnus A.

In other cases, the jet will just fade away andproduce diffuse regions (lobes) that are formedby ‘old’ electrons, not so energetic anymorebecause they have been decelerated. Thismorphology can be seen in the radio galaxy M87(also known as Virgo A) illustrated in Figure 12.This radio source is hosted by an elliptical galaxythat we met in the previous Chapter. The radioemission here is much more diffuse and no hotspots are seen at the edge of the lobes,indicating that the jets do not impact stronglywith the outside medium. Note that here we canalso see the presence of only one jet, which isstraight at first but then bends dramatically.These characteristics are often observed and canbe partly due to the effect of the medium around

Figure 13 - Example of compact source observed with theVery Long Baseline Interferometer. The size of this source isonly of the order of 100 pc (300 light years) and its age isestimated to be only about 1000 years.

Figure 12 - The panel shows view of M87 (Virgo A) atdifferent spatial scales. The bottom image (obtained usingthe Very Long Baseline Array) shows the very centralregions (red represents the regions of brighter emission,blue of fainter emission). The white bar indicates distancesof about 0.03 light years or about only 2 times the distancefrom the Earth to the Sun! The top left image is obtainedusing the Very Large Array and shows the radio jet andlobes at much lower spatial resolution. The distortionobserved in the radio emission on the large scale isamazing and it is likely due to the effect of the mediumaround the radio source. The top right image shows theoptical jet observed by the Hubble Space Telescope.

199

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

the radio source, or to projection effects due tothe orientation of the jets with respect to the observer.

The radio emission can reach amazingly largedistances from the nucleus, many times the sizeof the stellar body of the galaxy. In some cases,it can reach a few million light-years from the nucleus.

From the characteristics of the radio emission inthese regions, astronomers can estimate how oldthey are and therefore the age of the radiosource. Typical ages are a few tens of millions tohundreds of millions of years – if you comparethese with the time-scales of merging galaxies,you can see that the radio source seems to livefor a much shorter time.

Not all the galaxies are so extended – baby radiosources also do exist! These are tiny radiosources, with a size of only few tens of light-years but with morphology already essentiallyidentical to that of the grown-up sources; theycan be considered miniature radio galaxies. Anexample is shown in Figure 13: without the linearscale, it would be impossible to distinguish thisfrom a grown-up radio galaxy.

The study of these sources allows investigationsof how the radio plasma expands through themedium in the initial phase of the life of a radiosource. In some cases, it looks like it is quitedifficult for the baby radio source to make its waythrough the dense gas that surrounds it. Many ofthem have quite distorted morphologies indicatingthat they are forced to interact with a densemedium around them. The gas that surroundsthese sources is likely to be the same materialthat feeds the black hole and keeps the nucleus active.

Narrow and powerful jets Through the technique of interferometrydeveloped some years ago, radio astronomy hasreached the ability to observe objects with veryhigh spatial resolution (that is, being able todistinguish two very close objects from far away).Today’s radio telescopes can resolve the mostinner structures of the AGN jets.

Very high resolution observations becamepossible when radio astronomy developed theVery Long Baseline Interferometry technique.This consists of performing radio observationswith telescopes that are very far apart (thousandof kilometres) – the signal from the radiosources detected by the different antennas isrecorded separately, and later on combined toproduce the image. This system essentiallycreates a radio telescope as large as the Earth –actually, even bigger if you use antennas in orbitaround the Earth, as was done recently with aJapanese telescope! Consequently, the imagescan reach resolutions that would allow you tosee a football on the moon (if there was one, andif it could emit at radio wavelengths). Theresolution that astronomers can reach usingradio telescopes is still unsurpassed at other wavelengths.

For the study of AGN, this high resolution iscrucial, as we want to explore really tiny regions.The jet observed in M87 (Figure 12, bottomimage) shows the structure of the jet in theregion very close to the black hole where the jetis forming. The formation of the jet appears tooccur within a few tenths of a light-year of thegalaxy’s core. While the jet seems to start as awide structure, it then quickly becomes verycollimated: this suggests that something very likethe magnetic field mechanism (see Figure 4) isneeded to make this happen.

The ability to observe such small features hasallowed astronomers to track the motion of theplasma in the jet, confirming that indeed the jetsare the channels of supply for the radio lobes. Byobserving the region of radio jets very close tothe nucleus every few years, astronomers havedetected the change in the jets and the shift ofsome features due to their motion at very highspeed, in some cases close to the speed of light.Figure 14 shows high-resolution observations,carried out monthly for 16 months, of the jet inthe radio source called 3C120 – the nucleus isbelieved to be the bright spot on the left. First, itis clear that the jet is not really a continuousstructure, rather it is formed by a series of brightblobs. Every now and then the nucleus emits oneof these blobs that then starts moving away toreach the radio lobes.

200

WAV

ES O

F TH

E FU

TURE

A very exciting discovery (made more than 20years ago) is that jets observed in powerful radiosources can have an apparent velocity higherthan the speed of light, so-called superluminalvelocities. This is clearly not possible, accordingto Albert Einstein’s theory of Relativity, and theexplanation for this phenomenon is that it is dueto a projection effect (that is, it is due to the waywe observe the jet, not the jet itself). The objectsthat show this effect are sources where the radiojet is ejected in a direction very close to our lineof sight, in other words almost straight in thedirection of the observer. If such an ejectionhappens with velocity close to the speed of light(and some jets have been found to have avelocity more than 90% of light speed) thenrelativistic effects make it appear as if the velocityis greater than the speed of light.

One more effect comes from the high speed ofthe jet, combined with its projection or direction.As noted above, the jets can be very asymmetricin their luminosity: one jet is often much brighterthan the other. This characteristic has been againinterpreted to be due to projection effects.Because the jets are very narrow, the way we seethem depends very much on whether they are onthe plane of the sky or they are ejected by thenucleus in a direction close to the observer’s lineof sight. When we observe them close to the lineof sight, and when they travel at very highspeeds, relativistic effects act in a way that theapproaching jet looks much brighter than thereceding jet. Thus, the asymmetry observed hasbeen used as critical parameter to work out thedirection (relative to the observer) in which the jethas been ejected.

Are all AGN different?For many astronomical objects, like stars, it doesnot matter very much from which direction theyare viewed. For the AGN this is not the case. As

Figure 14 - Images of the radio source 3C120 obtainedwith the VLBA at 22 GHz (0.3 milliarsecond resolution), atmonthly intervals. The 3C120 likes to eject componentswith apparent superluminal velocities of about 5 times thespeed of light. The images also show rapid variations(scale of months) in total intensity, that we interpreted asinteraction with the external medium.

201

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

we have seen, most of the energy produced byan AGN is radiated through collimated jets orcones, and therefore it is radiated only in acertain direction. Thus, one can picture AGNs abit like a lighthouse: we can see the light onlywhen it is emitted in (or close to) our direction.For the same reason, the direction in which theenergy from an AGN is emitted becomesextremely important in determining the way AGNlooks when observed from the Earth.

Figure 15 illustrates these effects. When theemission from the AGN is radiated almost in ourdirection, we can see directly into the nucleus.This means that the nucleus is so bright that itovershines almost all the rest of the galaxy – inthis case, what we observe is a quasar object.The broad emission lines that are detected in theoptical spectra of quasars support the idea thatwe are looking directly into the nucleus.

Figure 16 shows what happens if we observe aquasar but we block (with a special instrument)the light coming from the bright nucleus. We cannow see the underlying galaxy, suggesting thatindeed the apparent difference between quasars

Figure 15 - Effect of orientation on the way AGN look. The filled and empty circles represent the region of gas producingemission lines (respectively broad or narrow). In both cases, the astronomers are happy!

and galaxies might be simply because theamazingly bright nucleus is making us blind to allthe rest!

However, if we are not looking straight into thenucleus, then the presence of the torus aroundthe accretion disk and black hole can obscurewhat is happening in the very inner regions. Thus,we do not see the broad lines and many othersigns, but the fact that the nucleus does notovershine the rest allows the astronomers to viewthis more easily than other phenomena.

As result of these orientation effects, astronomersthink that many different types of AGN (forexample quasars and radio galaxies) may in factbe the same kinds of objects only seen from adifferent direction!

Well-fed and starving black holes The question is, are active nuclei located only in asmall fraction of galaxies (that can therefore beconsidered as special, lucky objects) or do theyrepresent a phase in the life of every (or almostevery) galaxy?

202

WAV

ES O

F TH

E FU

TURE

The main consensus now is for the secondhypothesis. As we have seen above, at least theradio emission seems to have a relatively shorttime scale compared to the life of a galaxy. Butthere’s more – there could even be more thanone active period in a galaxies lifespan!

Evidence of this can be seen in the morphologyof some giant radio galaxies. The radio galaxy inFigure 17 has two giant (and old) radio lobesabout 1.5 billion light years in size. The period ofradio activity that created those lobes is over andthe giant lobes are now relics. The image shows,however, that the nucleus has recently becomeactive again – bright new jets are being emittedby the black hole and new lobes are now seenadvancing through the old ones. For somereason, new fuel has reached the back hole andthis galaxy could happily start its nuclear activityall over again.

The idea that almost every galaxy could, at leastonce in its lifetime, go through an AGN phase isalso supported by another fact. Astronomers nowthink that almost every galaxy may actually host amassive black hole in its centre. The presence ofthis massive body can be seen from the

Figure 16 - Images of the nearby quasar 3C273 taken with the Hubble Space telescope. With the higher resolution of the rightimage, and by using a device to block the light from the central AGN, the host galaxy can be seen.

Figure 17 - Giant radio galaxy (called B1545-321, based onits position on the sky) showing indication of a recentrestarted activity. This image has been obtained using theAustralia Telescope Compact Array.

203

MO

NST

ERS

LURK

ING

INTH

EC

ENTR

EO

FG

ALAX

IES

kinematics of the gas in the regions around it.This means that the presence of a central blackhole does not automaticallyensure the production of the hugeamount of energy detected froman AGN. There are, therefore,starving as well as well-fed blackholes and only the latter canproduce the amazing amount ofenergy observed in AGN.

If we calculate how much mattera black hole needs to eat in orderto produce the energy necessaryto shine as a bright AGN, it turns out to besomething like few tens of suns every year. Thisis assuming that the black hole can convert massin energy with a 10% efficiency.

This amount of food for a black hole can beeasily available in a galaxy, but the main problemis actually to make sure that the matter reachesthe very central nuclear regions to be able toenter the gravitational field of the black hole.Interaction between galaxies, as we have seen inthe previous chapter, seem to be an ideal way todo this job by ‘pushing’ some of the materialdown to the central regions. Due to this, thepresence of interaction and the nuclear activityare thought to be closely related.

Thus, from all we have seen in these twochapters, the life of a galaxy it is very

complicated indeed. Throughinteraction and merging, notonly can galaxies change theirstructure during their lifetimes... but the same phenomena canalso help turn their nuclei intofascinating monsters!

Some interesting web sites:

Alan Bridle’s page with impressive pictures of radio galaxies

(and links to other interesting sites):

http://www.cv.nrao.edu/~abridle/image.htm ; more

explanations about radio galaxies in

http://www.cv.nrao.edu/~abridle/dragnparts.htm

Radio images of AGN obtained with Very Long Baseline

Interferometry techniques can be seen also at the

European VLBI Network site:

http://www/evlbi.org/gallery/images.html

Nice pictures of AGN from the X-ray Chandra mission’s Photo

Album site: http://chandra.harvard.edu/photo

NASA site: http://legacy.gsfc.nasa/ with link to astronomy for

students

Nice images of radio, X-ray and optical jets can be found at:

http://hea-www.harvard.edu/XJET/index.cgi

“Astronomers now thinkthat almost every galaxymay actually host amassive black hole in itscentre.”

204

WAV

ES O

F TH

E FU

TURE

muscles of the hand. He gave aseries of simple clinicalquestions about the threenerves that run the hand(median, radial and ulnarnerves) to 20 junior doctors.They could get any markbetween 0 and 10. The averagemark was 3, with no doctorgetting higher than 5.

So Davidson wrote a paper inthe surgical journal called Injury,entitled “Rock, Paper, Scissors”.

Rock, Paper, Scissors is a veryancient game. Back beforescissors were invented therewas a similar game called“Earwig, man, elephant”. Today,Rock, Paper, Scissors is oftenused to decide matters betweentwo people in much the sameway they might toss a coin. Onthe count of three, each playerhas to select a hand position.

Now rock (which is a clenchedfist) will break scissors. Butscissors (which has the indexand second fingers open andoutstretched, and the little andring fingers tucked in) will cutpaper. And paper (the open

What makes the hand veryspecial is the thumb, which isprobably as important as all theother fingers put together. Ifyou activate one set ofmuscles, you form your handinto the “static hook” – whereyou grab a briefcase. Anotherset of muscles gives you the“pinch grip”, where the indexfinger and thumb combine tolet you make precisionmovements, such as passing athread through the eye of aneedle. You make the “powergrip” when you grab the handleof a hammer. And all thedifferent muscles are controlledby nerves.

It all begins in the spine in theneck, where five nerves run outfrom the four cervical vertebrae(C5, C6, C7, and C8) and thefirst thoracic vertebra (T1). Ifeach of the five nerves ran oneof the five fingers of the hand, itwould be easy to remember.But if you damaged one ofthese nerves, you would lose allfunction to one of your fingers.So in an area near yourshoulder called the BrachialPlexus, the five nerves all mix-

and-match and get all crossedover to give just three nervesthat control the hand – themedian nerve, the radial nerveand the ulnar nerve.

As an example of this mixing-and-matching thingie, the ulnarnerve is made up of C8 and T1(and sometimes, from C7) – but no nerves from C5 andC6. This crossing-over givessome protection in case ofinjury – a finger or two mightget weaker, but probably won’tgo totally floppy.

The arrangement of thesenerves and the muscles theycontrol is quite complicated,and memorising them hasworried medical students for thelast century. It’s important forthe junior doctors to knowwhich nerves run whichmuscles of the hand, becausethey are often the first to seepatients with injuries to thearm. Dr. A. W. Davidson, fromthe Department of Trauma andOrthopaedics at the RoyalLondon Hospital, reckons thatjunior medical doctors don’treally know the nerves and

Rock, Paper, ScissorsBy Dr Karl Kruszelnicki

IF YOU EVER study anatomy, they’ll teach you that the hand isthat incredibly important prehensile (or grasping) organ atthe far end of the multi-jointed lever called the “upper limb”– the rest of us call it an “arm”. The human mind lets usthink about the world, but the hand makes our dreams cometrue. It’s very necessary for medical doctors to know whichnerves control which muscles of the hand – and now asimple game has come to help their memory.

the fingers, which gives you therock position. The radial nervewill extend or stretch out fingersfrom the closed position to theopen position – so that givesyou the paper position. And theulnar nerve does two things.Firstly it makes the little fingerand the finger next to it, thering finger, tuck in. Andsecondly, it spreads the indexand middle finger. So overall,your hand looks like a pair ofscissors with the ring and littlefingers clawed up and the indexand middle fingers opening andclosing.

So with a simple game, morejunior doctors will be able tomake a decent fist ofdiagnosing hand injuries.

FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

Illustration courtesy Adam Yazxhi205

ROC

K,PA

PER,

SCIS

SORS

hand held flat with the fingerstogether and outstretched) willcover the rock.

So rock will break scissors,scissors will cut paper, andpaper will cover the rock – andaround and around it goes. Soyou’ll win, lose or make a draw.

Now it turns out that there is aWorld RPS or Rock, Paper,Scissors Society. They give youon their homepage variousbasic rules to the game, as wellas advanced gambits. Theyreckon that you should not play

random moves. You shouldeither use psychology toanticipate your opponent’smoves, or use a certainsequence of moves to influenceyour opponent’s responses.

And getting back to the handitself, you can work out mostinjuries to the nerves thatcontrol the muscles to the handby getting the patient to makeeither a rock, a paper or ascissors.

Now it turns out that themedian nerve clenches all of

DR JOE HOPE wasawarded his PhD fromthe ANU in 1997, andtook up a brief researchposition at the Universityof Queensland beforebeginning a lecturingposition at the AucklandUniversity. He hasworked in many areas ofquantum and atomoptics, and is currentlyresearching methods ofdetecting atoms non-

destructively, controlling the quantumstate of atomic samples, andproducing a high quality atom laser.He has won awards forcommunicating science, includingthe 2003 Dialogica Award from theAcademy of Science. In 2003, hebecame a founding member of theAustralian Research Council Centreof Excellence for Quantum-AtomOptics, in which he leads the atomlaser theory group located at the ANU.

206

What got you interested inscience in the first place?

I’m curious about nearlyeverything, and science covers avery large part of ‘everything’. I thinkmy fascination grew as I discoveredthat the huge number of experiencesin the world were based on such asmall number of fundamentalprinciples.

What’s the best thing about beinga researcher in your field?It’s a very fast-moving field at theconvergence of many traditionalareas of physics. It’s a chance tomix the fundamental with theapplied.

Who inspires you – either in scienceor in other areas of your life?I’m inspired by people who keepgrowing and learning throughouttheir lives – in all areas of their life.Such people have infectiousenthusiasm, and they always makeme want to engage.

If you could go back andspecialise in a different field,what would it be and why?Probably writing. My artisticenergies are packed into the cornersof my life, and I’d relish the chanceto see what I could manage if I gavemyself the time and space toconcentrate on them.

What’s the ‘next big thing’ inscience, in your opinion? In my own area, I’m hoping therecent development of systems inwhich we can test quantum fieldtheories will lead to applied quantumtechnology of a dramatically differentkind than has previously beenpossible – from quantum computersto usable superconductors. On themore applied side, I think there willbe a massive expansion in the areasof complex systems in general andbiotechnology in particular.

Quantum Mechanics: The WildHeart of the UniverseJoseph Hope

IntroductionQ U A N T U M M E C H A N I C S I S one of the two fundamental theories underlyingour current understanding of the universe, the other being Einstein’stheory of relativity. In terms of accurately predicting nature, it is by farthe most successful theory of all time, with some experiments agreeingwith quantum mechanical predictions to a dozen significant digits.Whenever quantum mechanics has made a preposterous claim aboutthe way things behave that goes counter to intuition, physicists havealways found that the prediction is correct and that it is our intuitionthat has been at fault.

207

Figure 1: The Poisson Spot - (a) simulated image formed from calculations using Fresnel’stheory (thanks to Dauger Research’s Fresnel Explorer simulator), (b) experimental image usingred laser light.

208

WAV

ES O

F TH

E FU

TURE

DESPITE SUCH OUTLANDISH successes, noone really believes the theory. In thestandard description of the quantum

theory, there is a little flaw that is easy (and quiteprofitable) to ignore, but that has neverthelesstroubled physicists for the last eighty years.Worse than that, the nature of this flaw is that itmakes quantum mechanics incompatible withanother cherished physical theory: relativity. Atleast one of them must be wrong, but noexperiment to date has managed to demonstratea failure of either quantum mechanics orrelativity. With that in mind, there are very goodreasons for the current form of quantummechanics, and we can definitely show that it issuperior to other (some would say ‘more sane’)theories that have been proposed.

Wave/particle dualityOne of the best-known ideas in quantummechanics is that objects can behave like wavesand particles at the same time. Where did thisidea come from, and how can it possibly be true?To turn that question around, what does it meanto behave like a particle or a wave? The mentalimage of a particle is somethingthat has a definite position andvelocity; it travels in straight linesunless it interacts with a force; itcarries energy, mass, andsometimes charge. A wave canalso carry energy1, but ratherthan having a definite position, itis spread out over the mediumthat is waving. The fact that itdoes not have a single positionmeans that it also tends not tohave a single velocity. Forexample, different parts of a water wave at thebeach often travel in different directions atdifferent speeds.

The struggle to use these ideas to classify aparticular phenomenon as particle-like or wave-like is epitomised in the history of our under-standing of the nature of light. In Newton’s earlyinvestigations into light, he saw how it travelled instraight lines, and decided that it must be madeof particles. The famous experiment in which hediffracted sunlight with a prism to show the

different colours led him to postulate theexistence of different kinds of light particle.

Between 1805 and 1815 Laplace and Biot, fromthe French Societe d’Arcueil in Paris, created anextensive theory of ‘neo-Newtonian’ optics basedon Newton’s ideas. This theory explained all theobserved optical phenomena at the time. As itwas nearing completion in 1817, Laplace andBiot arranged for the French Academy of Scienceto offer a prize for the best work on the theme of‘diffraction’, the apparent bending of light rays atthe boundaries between different media. In 1818Fresnel submitted his thesis for the prize,suggesting that light was a wave with anamplitude and phase at every point. This wavelikepicture of light explained diffraction and refractionwith ease, and without the complexity of the neo-Newtonian theories.

Fresnel’s thesis produced a hot debate when hedefended it in front of the judging panel. Criticsnoted that the wave theory of light had a hardertime explaining the tendency of light to travel instraight lines, and form sharp shadows. Fresnelcountered by pointing out that waves with a very

small wavelength would have acorrespondingly small blurringon the edge of shadows, andtherefore they would seem totravel in straight lines. In thebulk of the shadow, the phase ofthe wave from differentunblocked areas would add uprandomly, and on averagecancel out, making a dark area.

Poisson, a mathematicianworking with Laplace, dealt what

he thought was the killer blow: if Fresnel’s theorywere correct, then his argument workedeverywhere except in the exact centre of theshadow formed by a circular disk or a sphere. Atthe centre of such a shadow, the waves wouldhave to add constructively: there would be abright spot right in the middle of the shadow. Thisabsurd consequence of Fresnel’s theory causedhim to leave in apparent defeat.

Soon afterwards, one of the judges, FrancoisArago, actually performed the experiment – and

One of the best-knownideas in quantummechanics is thatobjects can behave likewaves and particles atthe same time.

209

QUA

NTU

MM

ECH

ANIC

S:TH

EW

ILD

HEA

RTO

FTH

EUN

IVER

SE

there, in the centre of the shadow, he observed abright spot. Fresnel went on to win the prize, andthe spot was ironically named Poisson’s spot, afterthe man who had ‘predicted’ it. The moral fromthis story is that any simple tests should alwaysbe performed rather than relying on intuition. Afterthese events, light came to be known as a wave,because while some waves can move in straightlines like particles, particles have a very difficulttime mimicking the constructive and destructiveinterference of waves.

The two-slit experimentThe phenomenon of interference is simpler whenthere are only two sources of waves. This can beachieved by making the waves pass through twosmall gaps in a blocking material that are verysmall compared to the wavelength. This meansthat the waves on the other side will tend toradiate outwards as though from two separatepoints. If the initial waves have flat wave-fronts,then an equivalent experiment in two dimensionsis to pass the waves through two parallel slits.

This experiment is easy to perform with a laserand shows a pattern of bright and dark lines,corresponding to places where the light from thetwo slits is interfering constructively anddestructively. Covering either of the slits causesthe bright and dark lines to be replaced by adiffuse blob of light. The dark lines, where light

coming from the two slits add together to give nolight, is the feature that most convinces us thatlight is indeed a wave - this ‘destructiveinterference’ is certainly a property of waves.

So what about things other than light? Is a pair ofsunglasses a particle or a wave? It is easiest toanswer this by ripping up the sunglasses intopieces and doing experiments on the bits. Let usassume that we are all comfortable with the ideathat the resulting pieces are electrons, protonsand neutrons. The very language implies thatthese things are particles, but we do not have totake this on faith – we can test it. It turns out tobe quite easy to perform a two-slit experimentwith electrons, as the inside of every televisioncontains an electron gun. If very thin slits aremade2, then we can set up a phosphorus screenthat glows to show where the electrons arrive,and perform the same experiment that we didwith light.

Taken separately, the electron beam comingthrough each slit will produce a diffuse glow on thescreen as they spread out. If the slits are almostnext to each other, the resulting spots can be inalmost identical positions. However, when both slitsare uncovered at once, the electrons form a familiarpattern of bright and dark lines. This is proof thatthe electrons must be a wave. But that is not thewhole story. When the intensity of the electron gun

Left slit only Right slit only Both slits

Figure 2: The two-slit experiment for electrons, when only one electron is in the apparatus at the same time. The left columnshows the pattern produced when the right slit is covered, and the second shows the pattern when the left slit is covered.Notice that these two patterns are broad enough that they overlap strongly. The third column shows the pattern producedwhen the electrons can go through both slits. The first image in each column shows the arrival of the first electron as a smallspot on the screen. The next four images show the screen after 2, 10, 500 and 3000 electrons respectively. While eachelectron arrives randomly as a single spot, when both slits are uncovered, the overall pattern of their arrival shows theinterference pattern of a wave going through both of the slits.

210

WAV

ES O

F TH

E FU

TURE

is turned down, the smooth interference patternresolves into single, tiny dots appearing on thescreen. In other words, we see what appear to beparticles arriving individually. We cannot just changeour minds and decide that electrons are particleswithout explaining the interference pattern. If wewait a long time, then the arrivals of these littleblobs eventually average out to that interferencepattern that was visible at high intensity.

Analysis of this experiment quickly leads to theconclusion that neither the wave nor the particlemodel of electrons is suitable. The electronscannot be travelling as the little blobs that we seeon the screen, because to make the interferencepattern they must be travelling through both slitsat the same time. The pattern we would see ifthey were passing through each slit randomlywould be two overlapping blobs, not a series ofbright and dark lines. It sounds ridiculous for aparticle to travel through both slits at the sametime, but it is a very natural thing for a wave todo, and the wave description describes the shapeof the interference pattern exactly. On the otherhand, a spread-out wave passing through twoslits to make an interference pattern is alsospread out when it reaches the screen, andwould not cause a single, point-like dot.

The famous American physicist Richard Feynmanhad this to say about the two-slit experiment:

‘We choose to examine a phenomenon whichis impossible, absolutely impossible, toexplain in any classical way, and which has init the heart of quantum mechanics. In reality,it contains the only mystery. We cannot make

the mystery go away by explaining how itworks . . . In telling you how it works we willhave told you about the basic peculiarities ofall quantum mechanics.’

The Feynman Lectures, Vol III

We can’t even explain the interference pattern interms of some complicated interactions betweenthe electrons. We can turn the electron gun downso low that only one electron is present in thesystem at a time, and the dots and theinterference pattern occur regardless.

Can electrons really be going through both slits atthe same time? We can modify our originalexperiment to try to detect the electrons as theygo through the slits. We can do this carefully sothat we don’t disturb them too much, and thenwe’ll know which slit they go through, and try tounderstand how this interference pattern can form.The amazing result of this experiment is that theinterference pattern no longer forms. If thedetectors are removed, the interference patternreappears. Not only are the electrons trying toconfuse us, they can tell when we are looking!

The same results can be found, with carefullyconstructed apparatus, for protons, neutrons,entire atoms and even complex molecules.Turning down the intensity of our light source willalso show the appearance of little, localisedpackets of light, called photons. It seems thateverything has this duality between wave andparticle behaviour.

As always, apparent paradoxes like this existbecause of our preconceptions. As with alltheories that challenge our preconceptions, it tooka lot of evidence to change them. It was almosthalf a century of debate before anyone seriouslyargued that quantum mechanics was adescription of reality rather than just a goodcalculational tool.

The axioms of quantum mechanicsLeaving aside the details, we can write down thewhole theory of quantum mechanics very simply:

Figure 3: Richard Feynman, Nobel Prize-winning physicistand infamous bongo player.

211

QUA

NTU

MM

ECH

ANIC

S:TH

EW

ILD

HEA

RTO

FTH

EUN

IVER

SE

1. Every object is described by a wavefunction.2. These wavefunctions evolve according to

Schrödinger’s equation.3. When a measurement is made on the

wavefunction, it collapses to give a definiteresult.

4. Immediately after the measurement, the newwavefunction is consistent with the result ofthe measurement.

Let’s examine three of these points in a littlemore detail. The first point says that every object,whether it is an electron, a beam of light or aniPod, is described by a wavefunction. This meansthat everything acts like a wave, and it is thisaxiom that explains the interference pattern in thetwo-slit experiment. These waves carry energylike all waves, but they can also have mass,electrical charge and other physical properties,depending on the kind of object they aredescribing. The thing that they all have incommon is that they have amplitude and aphase, and can form constructive and destructiveinterference.

The main detail left out of these axioms is theform of Schrödinger’s equation, which alsodepends on the particular object, as well as its

environment. Mathematically it is not verycomplicated, so once the concepts of quantummechanics have been accepted it is not terriblyhard to learn the calculational details.

The third and fourth points are required tounderstand the appearance of little ‘blobs’ in thetwo-slit experiment. After the electronwavefunction has travelled through the slits andhas formed interference fringes, the screenperforms a measurement on the electronposition. The wavefunction then suddenly jumpsto a single position randomly, with a probabilitythat depends on the amplitude of thewavefunction. This explains the appearance of thesmall spots on the screen (the wavefunctioncollapsed) and also explains why they form aninterference pattern over time, as more and moreelectrons are measured (the bright fringes wouldappear where the wavefunction’s amplitide islarge, so there’s a greater chance for the dot toappear in those regions).

One common misconception about this random‘wavefunction collapse’ is that it represents somekind of ignorance about where the electron‘really’ was just before the measurement. This isa natural idea, as that is what probabilitiesusually mean in physics, or in any other field – probabilities normally indicate that you aremissing some information, not that reality itself isuncertain. For example, if you know the secretformula then the computer-generated randomnumbers in a gambling machine are actuallypredictable.

But the case of the electron position is different,as it really does have to go through both slits atthe same time in order for the wave to make aninterference pattern. The randomness in quantummechanics is fundamental randomness. This ideaexplains all the phenomena that we can see, butmany people have felt that there must be adeeper theory that describes these effectsdeterministically:

The theory yields a lot, but it hardly brings us any closer to the secret of the Old One. Inany case I am convinced that He does not throw dice.

Albert Einstein to Max Born,4 December 1926

Figture 4: Erwin Schrödinger, Nobel Prize-winning physicistand creator of the Schrödinger Equation.

212

WAV

ES O

F TH

E FU

TURE

Measurement has a very special effect onwavefunctions in quantum mechanics, and this isthe explanation for the surprising ability ofelectrons to know when they are being observed.When the measurement devices look at theelectrons before they reach the slits, thewavefunction collapses to go through either oneslit or the other – but only one slit – with equalchance for either to occur. This means that theinterference pattern disappears. When themeasurement is not made at the slits, the wavegets a chance to make an interference patternbefore it is measured by the final screen.

Taking a definition of ‘reality’ as ‘things that arethere even when you’re not looking’, and notingthat quantum mechanics says that when you’renot looking, there are no particles, some peoplehave seriously suggested that the act ofmeasurement actually creates reality:

[The] moon is demonstrably not there whennobody looks.

Prof. David Mermin

What is waving?One of the first questions that arise whendeciding that all matter is made of waves is this:what is waving? The answer is obvious when yousee a water wave at the beach, or a standingwave in a guitar string, but what is waving whenthe wave is an electron? The answer is ‘an

electron field’. This is imagined to exist throughoutall space, much like the electromagnetic field.There is also a proton field and a neutron field.These fields sound very abstract, but the reasonthat we believe they exist is fundamentally thesame as the reason we think walls, air andelectricity exist – the world behaves as if they do.The difference is that we need only postulate anelectron field after a careful examination of theworld, whereas walls are a little easier to believe.Electricity would be hard to explain to someoneborn a thousand years ago, but an attempt wouldbe a little like this article: a description of a seriesof experiments and their conclusions.

The electron field has, at each point in space, anamplitude and a phase. All of these amplitudesand phases together define a function that wecall the wavefunction of the electron.Mathematically, we define this function to becomplex-valued, as complex numbers also havean amplitude and a phase.

Some consequences of quantummechanicsQuantum mechanics can be used to explainmany things about our everyday world. It alsodescribes strange behaviours that we do notnormally see in our everyday world, and one ofthe larger mysteries of the theory is how theworld can look so classical in everyday life whenthese strange things are happening all the time at a microscopic level.

The existence of atomsQuantum mechanics explains why atoms, andtherefore everything made of them, can exist. Ifelectrons really were negatively charged particlescircling a positively charged nucleus, thenclassically they would emit electromagneticwaves and spiral inwards towards destruction. Aselectrons are actually waves, it is fairly easy toshow that for waves with a well-defined energy,only certain shapes fit around the nuclei. Thismeans that there are certain stable configurationsfor the electrons in atoms, and that they havecertain very specific energies, including a lowestallowable energy. The energy of electronsattached to atoms can only be changed indiscrete amounts, called quanta. This is where

Figure 5: Visualisation of a field, making up a wavefunction.At each point, the field has an amplitude and phase,represented here by the size of a circle and the angle ofthe interior line respectively. A continuous series of thesepoints make up the wavefunction.

213

QUA

NTU

MM

ECH

ANIC

S:TH

EW

ILD

HEA

RTO

FTH

EUN

IVER

SE

the theory gets its name. Ironically, a ‘quantumjump’, which in ordinary usage means a largeand sudden change, is actually the smallestpossible change that is allowed.

Quantum tunnellingQuantum tunnelling is a phenomenon that wasfirst used to describe radioactivity, and isextremely hard to explain without quantummechanics. Some nuclei contain alpha particlesthat lack the energy to overcome the short-rangeattractive force attracting them to the centre. Thiscan be described as an energy barrier, like a hillthat is too high for them to climb. If they everfound themselves on the other side, however,they would be able to escape the nucleus. If theywere particles, they could only do this if theycould somehow get enough energy to get overthe hill. Quantum mechanically, however, they arewaves, and while most of the wave is locatedinside the nucleus, a small part of the wave is onthe outside. Measuring the position of the alphaparticle will cause the wavefunction to collapse,and there will always be a probability that this willoccur well outside the nucleus. The alpha particlecan therefore ‘tunnel’ through the energy barrierand appear on the outside without ever havingenough energy to be in between.

This seems outrageous enough for a tiny particle,but why don’t we see it all the time? If everycomponent of our bodies is a wave, then we areall made of waves. When we walk towards adoor, some part of us is already on the other side.Plenty of people walk into doors every day, sowhy don’t they occasionally appear on the otherside without opening it? The answer to thisquestion is simply a matter of the numbers. Eachwave comprising our body does indeed extendthroughout the whole universe, but it is exponent-ially less likely to be found somewhere unusual,with a natural length scale of about a nanometre.It is therefore extraordinarily unlikely for one ofthe many electrons in our bodies to be a wholemillimetre out of place, never mind all of them.

While the chance of walking through a wall isunsurprisingly low, one can imagine conductingexperiments that deliberately amplify the quantum world.

Heisenberg’s uncertainty principleOne of the best-known results in quantummechanics is Heisenberg’s Uncertainty Principle.It has many forms, but the most common is this:The uncertainty in the position x of an objectmultiplied by the uncertainty in the momentum pof that object must be greater than a constant, or∆x.∆p > h

2πIn this equation, h is Planck’s constant. Thisequation means that it is impossible for an objectto have a well-defined position at the same timeas having a well-defined momentum. This limit isoften ascribed to the uncertainty caused by themeasurement process, but it is more correct tothink of it as an intrinsic property of the wave-function. It is true that measuring the position ofa wavefunction requires interaction with themeasurement device and that measuring theposition with increasing accuracy means that theobject must be given an increasing spread inmomentum, but this is closer to an explanation ofthe dramatic effects of making a measurement inquantum mechanics than a statement of theuncertainty principle. The uncertainty principle saysthat not only is it impossible to measure a preciseposition and momentum simultaneously, but that anobject cannot have them simultaneously.

Like all surprises in quantum mechanics, this ismuch easier to understand if you think of thingsas wavefunctions rather than particles. We havealready discussed the fact that waves don’tnormally have a single position but are spread

Figure 6:WernerHeisenberg,Nobel Prize-winningphysicist andsource of theUncertaintyPrinciple.

214

WAV

ES O

F TH

E FU

TURE

out. The same is true for momentum. To look atthe momentum of a wave, we look at thederivative of the wavefunction. A state that doesnot change if we apply the derivative is theexponential function, so a wavefunction that hasa well-defined momentum p is given by

ψ(x) = e (2πp.x)/h

But the amplitude of this wavefunction is constanteverywhere, and the phase rotates linearly as wemove in the direction of travel. This is clearly verydifferent to a wave with an amplitude that isnearly zero everywhere except for a well-definedposition – having a definite momentum isn’tcompatible with being in a definite place.

The reason we do not see this uncertaintybehaviour all the time is that Planck’s constant isa very small number indeed. A car travelling at aspeed known to the nearest nanometre per yearcan have a wavefunction localised to less than abillionth of the size of an atom – which isn’t veryuncertain at all. Heisenberg’s uncertainty principleis vastly more important for very small thingssuch as electrons or atoms. An electron in anatom has a size of about an Angstrom (10-10 m),so it must have an uncertainty in speedcorresponding to about 1000 km/s!

Experiments on ultracold atoms and BECIn the laboratories at the ANU in Canberra, as wellas many others around the world, we cool atomsdown using a combination of techniques including

laser cooling and old-fashioned evaporation. Whenan atom absorbs a photon from a laser beam, italso absorbs the momentum of the photon, and itgets a kick in the direction of the laser. Thisprocess would normally heat a cloud of atoms,making them go faster.

If we make the frequency of the laser just a littleless than the natural absorption frequency of theatoms, then the atoms will be more likely toabsorb a photon when they are moving towardsthe laser beam than they are when moving awayfrom it. So the laser is more likely to slow the

I

I

σ

σ σ

σσ

σ+

+ +

−

− −

Figure 8: The laser cooling and magnetic trapping isperformed in high vacuum in the glass cell shown. Red-detuned lasers are incident from all sides, and currents inthe coils allow the atoms to be trapped magnetically for thefinal evaporative stage. Image courtesy of Nick Robins.

Figure 7: A simple version of Heisenberg’s uncertainty principle. A wave with a single momentum is different to a wave withwell-defined position. The compromise between the uncertainty in each variable is quantified in Heisenberg’s equation.

atoms down than speed them up. Shining red-detuned laser light from all directions cantherefore cool a cloud of gaseous atoms fromnearly a thousand degrees Kelvin to a milliKelvin

215

QUA

NTU

MM

ECH

ANIC

S:TH

EW

ILD

HEA

RTO

FTH

EUN

IVER

SE

in just a few milliseconds. With the lasers thenturned off and the atoms captured in a magnetictrap, they can be cooled even further by allowingthe highest energy ones to escape and waitingfor the gas to reach thermal equilibrium.

This evaporation can cool the atoms to less thana microKelvin, or an average atomic speed ofapproximately a tenth of a millimetre per second.This means that the wavefunction of these atomsis of the order of tens of microns across1, almostas wide as a human hair, and virtually detectableto the naked eye. It stands to reason that a lot ofstrange things occur when atomic wavefunctionsare big enough to be seen by the human eye,and examining some of these effects is a majormotivation for producing them. If you look atthem, the wavefunctions collapse; they forminterference fringes when split up andrecombined; they can tunnel through barrierswithout having enough energy to get over them;and they can reflect off attractive potentials.

These ultracold atoms do something else quiteextraordinary. When the atom cloud gets so coldand dense that the individual wavefunctions ofthe atoms overlap significantly, then the fact thatall fundamental particles are indistinguishable3causes them to behave very strangely. Someatoms, like electrons, are Fermions: that is, they

cannot have two particles in the same quantumstate. This means that they start to repel eachother. Other atoms are bosons (like photons), andthey can happily occupy the same state. In fact, ifwe cool a cloud of bosons so that the individualwavefunctions overlap, then they tend to all gointo the same quantum state. This is known asBose-Einstein condensation (BEC), and it allowsus to enter a world where we can manipulatelarge wavefunctions in the laboratory. This newability should allow us to test some of the strangeclaims in quantum mechanics; build ultra-sensitive measurement devices based on atominterference; and model previously inaccessiblequantum systems like superconductors, blackholes and neutron stars in systems where we cancontrol most of the physical parameters.

Current work on these ultracold atoms isfocussed on exploring the possibilities of makinga useful atom laser. The formation of a Bose-Einstein condensate in an atom trap is directlyanalogous to the stimulated emission thatcauses a laser, as both processes use the sameproperty of bosons (photons in the laser case,and bosonic atoms in the other). Letting acondensate out of a trap therefore produces acoherent beam of atoms with similar propertiesto the coherent beam of light emitted by a laser.Atom lasers are still in their infancy, however,and there is much to learn about controllingthese new atomic sources.

TeleportationEver hear the joke about the office worker whoran out of fax paper and asked a friend to faxthem some more? It is a surprisingly commonmisconception that fax machines are actuallyteleporting paper. What they really do, of course,is make as accurate a copy of a piece of paperas they can by measuring the paper and thensending the information via the phone line. Inprinciple, therefore, we could build a machinethat did the same for a human, building her upfrom vats of chemicals. This is a little more likecopying the human rather than teleporting her,but so long as we destroy the original, no one willbe the wiser.

Figure 9: The momentum spread of a cloud of atoms as wemake it colder. The leftmost picture shows a thermaldistribution. In the next picture, as we make the atoms onlyslightly cooler, we see a new, sharp distribution sticking upin the middle. This is the formation of the Bose-Einsteincondensate. The last picture, slightly cooler again, showsthe pure BEC. This is a single wavefunction, large enoughto see, containing a hundred thousand atoms. Imagecourtesy of Nick Robins.

216

WAV

ES O

F TH

E FU

TURE

People who have recently read half an article onquantum mechanics may notice a little snag,however. A combination of the process ofmeasurement and Heisenberg’s uncertaintyprinciple means that it will never be possible tomeasure the position and momentum of eachwave in the human’s body. After a positionmeasurement, the momentum spread will belarge, and unrelated to the original momentumspread of the wave. After a momentummeasurement, the wave will be spread out,irrespective of where it may have initially beenlocalised. In general terms, after any number ofmeasurements, it is only possible to ever obtainhalf of the information about a given initialwavefunction. This would make teleporting evena single, microscopic object impossible by simplymeasuring it, unless we were happy with losinghalf of the information about it. We are morethan happy to lose information about the precisepattern of paper and ink molecules in a letter tobe faxed, which is why fax machines have neverhad to confront this problem. However,teleporting a quantum state or a brain (whichmay rely on particular quantum states) appears doomed.

Using another deep property of quantum systems,called entanglement, it is possible to teleport aquantum state without learning anything about it.This was first achieved in CalTech, but has sincebeen repeated in two other laboratories, includingone at ANU. Researchers demonstrated thesequantum effects using light instead of atoms.They generated pairs of photons that were quiterandom, but strongly correlated. This meant thatmeasuring any property of the two beams would

give a very noisy signal, but the noise on one sidewould be the same as the noise on the other.

Taking a laser beam and mixing it with one of thenoisy beams allowed measurements to be made,and these could then be used to recreate the firstnoisy beam’s state out of the second, correlatedbeam – in effect, the first’s state has beenteleported to the second. This seems tocontradict our earlier argument, but themeasurements made on the original beam gaveno useful information about it, because it wasdrowned out by the noise on the first correlatedbeam. While this will probably never be used as amethod of transportation, the technology requiredto do quantum teleportation is extremely usefulfor manipulating quantum states.

Quantum computersThe most ambitious reason to want to manipulatequantum states would be to build a quantumcomputer. Any computation can be boiled downto taking a number as an input and calculatingsome arbitrarily complicated function of thatnumber. Physically, this number is represented bysome quantity such as the alignment of varioustiny magnetic domains in a magnetic material –the ones and zeros on a hard-drive. This assumesthat these quantities exist, of course, which is notnecessarily true when you realise that themagnetic material is made of waves, and can beboth ‘one’ and ‘zero’ at the same time. Thissuperposition of states must be carefully avoidedin normal computers or else they get confused.As the components of computers get smaller, andtheir wave-like behaviour become increasingly

Figure 10: Schematic of the process of teleportation.

217

QUA

NTU

MM

ECH

ANIC

S:TH

EW

ILD

HEA

RTO

FTH

EUN

IVER

SE

important, it is hard to keep the machinebehaving classically. A quantum computer tries todo the exact opposite and keep the quantumnature of a machine intact as the machine ismade larger.

A quantum computer is a computer that performsall of its calculations so carefully that if asuperposition of different inputs is put into themachine, then the result is a superposition of allthe possible corresponding results. It is easy toshow that when the input number has a fewthousand digits, a classical computer the size of the known universe could not calculate all ofthose possible combinations in a reasonable time;a quantum computer, though, could do it almost instantly.

Of course, having a superposition of all of thepossible outputs does not obviously help, asmeasuring the output will collapse thewavefunction to one of the individual resultsanyway, wasting the hard work that went intoperforming all of the calculations at once.However, there are some algorithms that canobtain important results from testing suchsuperpositions, the most famous of whichinvolves factorising large numbers. This isextremely interesting to anyone with a secret, oran interest in other people’s secrets, as mostmajor codes base their security on the difficulty ofthat factorisation problem. It may be possible touse a quantum computer to solve many otherproblems that, like code-breaking, were onceconsidered permanently out of reach.

The real question with quantum computers iswhether they can actually be made. In practice,any interaction with the environment messes uplarge superpositions, and that makes it very hard to build a quantum computer of anysignificant size. A microscopic quantumcomputer has already been used to factorise 15 into prime numbers, but the real problemseems to be scaling up the system withoutletting it become buried in the noisy world of macroscopic objects.

The future of quantum mechanicsQuantum mechanics already underpins most ofour technology, including electronics and otherengineering standards such as lasers. These toolsrequire a practical understanding of quantummechanics to build. Usually, though, we tend touse only the more mundane properties of thesedevices. The actual control and manipulation ofwavefunctions themselves is only just becomingfeasible. Although the future is notoriously hard topredict, we might find that our current centurybecomes known as the time for the birth ofquantum engineering.

Before quantum engineering can begin inearnest, it is important to put the finishingtouches on quantum science. The conflict withrelativity and the strange behaviour ofmeasurement are particularly worrying becausesome of the ways people have attempted toresolve these problems have suggested that thereare fundamental physical processes that destroylarge superpositions. It is important to knowwhether this is true. If not, and if we are carefulenough, we just might manage to out-computethe universe and walk through walls after all.

Notes

1 Technically, it is possible to invent waves that carry mass

and charge as well, but we will ignore this issue for the

moment for the sake of simplicity.

2 ‘Very thin’ in this case means much smaller than a human

hair, but much larger than the wires on a computer chip.

3 An extra axiom of many-body quantum mechanics is that

all particles are indistinguishable. For example, if two

people swap all of their electrons they are unchanged, as

electrons are all the same.

218

WAV

ES O

F TH

E FU

TURE

of the minerals that reach the ocean.

The very heavy rains combinedwith enormously steep slopescause huge erosion. The carbondioxide that is dissolved in therain drops forms a weak acid –carbonic acid. This carbonicacid combines with granite andlimestone which come from themassive erosion. Thecombination of carbon dioxideand granite/limestone makesminerals which wash downhilltowards the ocean. Theseminerals are very rich incarbon. So Tibet takes carbondioxide out of the atmosphere,and shoves it not into trees, butinto minerals.

So, according to this theory, theTibetan plateau is really a hugepump that takes carbon dioxideout of the atmosphere, anddeposits it on the ocean floorwhere it stays locked away formillions of year.

The Tibet theory was created byoceanographer Maureen Raymofrom MIT and her colleague BillRuddiman, a paleoclimatologistof the University of Virginia.They claim that chemicalreactions caused by the Tibetanplateau have removed so much

Now the climate of the worldhas been fairly predictable overmost of the last few hundredmillion years. Until recently, itwas warm and wet, like thetropics. Back then, the level ofcarbon dioxide was twice thelevel that it is today. Thedinosaurs, who lived from 200million to 65 million years ago,enjoyed a temperature about 8-11oC warmer, and swam inseas about half a metre higherthan we do today.

But all this changed 50 millionyears ago when India collidedwith Asia at the frighteningspeed of 20 centimetres peryear (roughly 4 times fasterthan your fingernails grow). As aresult of this slow but giganticcollision, the Himalayas andTibet relentlessly and graduallyrose above sea-level as Indiaploughed northwards another2,000 km. India slowed itsnorthward speed to a moresedate 5 cm per year.

During this enormous collision,the Antarctic began to ice upand the world cooled down. Theworld’s temperature kept ondropping. About 2-3 millionyears ago, our human brainbegan to double in size from600 ml to about 1,200 ml.

It could be just a coincidence,but Big Brains do need a lot ofcooling. After all, we humansreally need our big brains. Wecan’t see very clearly, we can’trun very fast, our skin won’teven stand up to a rose bushand our nails are pathetic asclaws. Compared to the otheranimals on the planet, our bigbrain is our only worthwhileasset. But while our brainweighs only 2% of our bodyweight, it takes 20% of ourblood supply, and so 20% ofour waste heat gets dumpedfrom our head.

Now a new theory claims thatthe Tibetan Plateau isresponsible for cooling theworld by taking carbon dioxideout of the air. The TibetanPlateau is a huge area, roughlyhalf the size of Australia, andmostly higher than 5 km abovesea level. Clouds run into thisplateau, and dump their wateras rain. In fact, the Tibetanplateau causes the annual Asianmonsoons. As a result, eighthuge rivers, which include theGanges, Mekong, Indus and theYangtze, drain from the TibetanPlateau and its approaches.These rivers drain a total areaof less than 5% of our Earth’sland area, but they dump 25%

Tibet Cooled The WorldBy Dr Karl Kruszelnicki

TIBET IS A spiritual place. It sits on the roof of the world – the 5 km high Tibetan plateau. Some researchers nowbelieve that this plateau cooled the whole planet, and maybehelped the evolution of the human brain.

219

TIBE

TC

OO

LED

THE

WO

RLD

carbon dioxide from theatmosphere, that thetemperature has dropped – notthe Greenhouse Effect but theTibetan Ice Block Effect.

Now the theory is in its earlydays, and it’s not rock solid, butwe do know that after about 50million years of a steadydownward drop in bothtemperature and carbon dioxidelevels, the Earth’s climateseems to have stabilised into anoscillating series of Ice Ages

and non-Ice Ages. And at theend of that drop, our brainsbegan to evolve larger. Somaybe Tibet not only chilled outthe world, it also gave usswollen heads.

FROM Dr Karl’s book Bum Breath, Botox

and Bubbles (Harper Collins Publishers)

PROFESSOR HUWPRICE was born inOxford, and came toAustralia on an Italianliner when he was 13.He attended the ISS in1969, and went on tostudy Mathematics,Physics and Philosophyat ANU, Oxford andCambridge. He hasworked at the Universityof Sydney since 1989,except for a two-year

break early this century, when hewas Professor of Logic andMetaphysics in the Department ofPhilosophy at the University ofEdinburgh. He is now ARC FederationFellow and Challis Professor ofPhilosophy at Sydney, and heads theCentre for Time in the Department of Philosophy.

Huw has written several booksincluding Facts and the Function ofTruth (Blackwell, 1988), Time’s Arrowand Archimedes’ Point (OxfordUniversity Press, 1996), as well as arange of articles in academic journalssuch as The Journal of Philosophy,Mind, The British Journal for thePhilosophy of Science and Nature. heis a Fellow of the Australian Academyof the Humanities, and a PastPresident of the AustralasianAssociation of Philosophy. He is alsoa consulting editor for the StanfordEncyclopedia of Philosophy, anassociate editor of The AustralasianJournal of Philosophy, and on theeditorial board of The PhilosophicalQuarterly and Logic and Philosophyof Science.

220

Huw, what got youinterested in science in the

first place?I’ve always been very interested inanything that starts with “s”. “sc” iseven better. I was quite good at allmy school subjects, starting withhome economics (making scones),and by adolescence I was into sportsscience (Scalectrix, sculling andscuba diving). So when I lookedaround for career opportunities,science seemed like the go. Physicsis a particularly fruitful areas for“sc”s: we have plenty of schisms, forexample, several of them involvingSchrödinger.

What were you like as a kid? Wereyou curious, pulling apart stuff tosee how it worked?I often tried to pull other kids apart tosee how they worked.

What’s the best thing about beinga researcher in your field?I think I’d have to say the adulation.The adulation and the parking spot.

Who inspires you – either inscience or in other areas of yourlife?I think it’s fair to say that mostscientists are inspired by fairlyethereal things. Science is notsomething you plan, like climbing amountain: it’s a field in which youhave to be patient, and willing tofollow your curiosity wherever itgoes. And in both those things, I feel— like most scientists, probably —that reality TV is the best inspirationone could ask for. The way in whichperfect strangers form strongattachments and even strongerhatreds within minutes of meetingeach other has inspired me, forexample, to investigate themathematical symmetry betweenwaves which converge on a point inspace and those which diverge from

continues page 232

Einstein and the QuantumSpooksHuw Price

IntroductionT H E I N T E R N A T I O N A L Y E A R O F Physics celebrates the centenary ofEinstein’s amazing debut: the three groundbreaking papers he publishedin 1905. The most famous paper introduced the first of the two greatrevolutions in twentieth century physics, the special theory of relativity.Another paper, studying the statistics of Brownian motion, providedcrucial new support for the (then still controversial) hypothesis thatmatter was made of atoms. And the third, proposing a newunderstanding of something called the photoelectric effect, was one ofthe important steps towards the century’s second great revolution, thetheory of quantum mechanics. So Einstein is not only the father of thetheory of relativity. He’s also one of the grandparents of quantum theorythat, after gestating for about a generation, was born into the world inanother remarkable twelve months for physics, between June 1925 and June 1926.

221Albert Einstein

222

WAV

ES O

F TH

E FU

TURE

We know that grandparents tend to bemore indulgent than parents. In the caseof Einstein and his famous theoretical

offspring, however, it was the other way round.Far from being a doting grandparent to quantummechanics, Einstein always disliked it, or at leastthe interpretation of what it meant that becamewidely accepted in physics during his lifetime.This popular view of the meaning of quantummechanics was called the CopenhagenInterpretation, because it was developed andchampioned by the ‘Great Dane’ of twentiethcentury physics, Niels Bohr (1885-1962).

Bohr was another grandfather of quantum theory,and his disagreement with Einstein about themeaning of the theory was very much like a bitterfamily feud. It led to a personal rift between thesetwo former friends, two of the giants of twentiethcentury physics, which persisted until Einstein’sdeath in 1955. Even more like a family feud,perhaps, Einstein’s unhappiness with quantummechanics had a lot to do with tensions betweenquantum mechanics and his own brainchild, thetheory of relativity, although the full extent of thattension didn’t become clear until at least adecade after Einstein’s death.

This chapter is about why Einstein was unhappywith quantum mechanics, and about the amazingsequel to his objections that later came to light.This sequel, unearthed by an Irish physicist calledJohn Bell (1928-1990) in 1965, is still one of themost puzzling things in contemporary physics.Nobody really knows what it means. Worse still, itreveals a deep tension between quantummechanics and Einstein’s own special theory ofrelativity. As the special theory reaches its one-hundredth birthday, in other words, we still don’tknow how to reconcile it with its illustrious eighty-year-old cousin. It is as if these two great theorieshave lived side by side for eighty years, neverproperly speaking the same language.

It is true that for many purposes this conflictdoesn’t matter very much. Working physicistsknow how to deal with one or the other theory, asnecessary. But the tension is still there, and it isone of the deepest mysteries that Einstein’scentury has bestowed on the one we now callours. It is impossible to say whether theresolution of this mystery will one day lead tonew revolutions in physics, in the way thatEinstein himself developed relativity as a solutionto tensions in nineteenth century physics. But Ithink we can be sure that there is somethingimportant we don’t understand about the physicalworld, until we find a better understanding ofthese quantum mysteries.

Fuzzy Pictures versus Fuzzy RealityQuantum mechanics had several parents andgrandparents, but the two with best claim to befathers of the new theory were young Germanphysicist, Werner Heisenberg (1901-1976), andthe Austrian physicist, Erwin Schrödinger (1887-1961). Heisenberg and Schrödinger discoveredwhat turned out to be different but equivalentforms of the new theory in 1925 and 1926,respectively. You’ve probably come across thesenames already. You’ve heard about Heisenberg’sUncertainty Principle, and the idea that quantummechanics shows that properties often don’t havesharp values in the quantum world – that aparticle can’t have both a sharply defined positionand a sharply defined momentum, for example.You may have also heard, at least briefly, aboutSchrödinger’s unlucky cat.

Niels Bohr

223

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

I said earlier that Einstein wasn’t at all a dotinggrandfather to quantum mechanics. AndSchrödinger tended to side with Einstein on theseimportant family matters. Certainly, he wasn’t atall happy with the interpretation that was soonbeing placed on quantum mechanics by peoplesuch as Bohr. His famous cat first turns up in1935, as an objection to the Copenhagen view.This is what Schrödinger says:

One can even set up quite ridiculous cases. Acat is penned up in a steel chamber, along withthe following diabolical device ... In a Geigercounter there is a tiny bit of radioactivesubstance, so small that perhaps in the courseof one hour one of the atoms decays, but also,with equal probability, perhaps none; if ithappens, the counter tube discharges andthrough a relay releases a hammer whichshatters a small flask of hydrocyanic acid.

If one has left this entire system to itself for anhour, one would say that the cat still lives ifmeanwhile no atom has decayed. The firstatomic decay would have poisoned it. The ψfunction for the entire system would expressthis by having in it the living and the dead cat(pardon the expression) mixed or smeared outin equal parts.

It is typical of these cases that anindeterminacy originally restricted to the atomicdomain becomes transformed into macroscopicindeterminacy, which can then be resolved bydirect observation. That prevents us from sonaively accepting as valid a “blurred model” forrepresenting reality. In itself it would notembody anything unclear or contradictory.There is a difference between a shaky or out-of-focus photograph and a snapshot of cloudsand fog banks. (Schrödinger 1935)

Schrödinger’s basic point is if we understandquantum mechanics as saying (in the way thatBohr’s Copenhagen school recommended) thatreality is ‘fuzzy’, like a cloud or a fog-bank, then itis easy to think of cases in which large things,like cats, would have to be fuzzy too. And we’renot talking about the usual kind of felinefuzziness, of course: in the experiment asdescribed, the cat would have to be in someindeterminate state, neither alive nor dead, for

example. If that’s really absurd, as Schrödingerthought it was, then it follows that the fuzzyreality interpretation of quantum mechanics mustbe wrong. The fuzziness of quantum mechanicsmust be in the picture, not in the world.

Schrödinger’s Copenhagen opponents tended tosay that in quantum mechanics, reality stoppedbeing fuzzy when we make a measurement –when we decide to measure either the position orthe momentum of an electron, for example, andthereby make it the case that it has a definitevalue for one or the other. However if that’s right,what about the poor cat? Does it only stop beingneither alive nor dead when we open the box,and make a measurement – when we look to seehow it is fairing, inside the ‘diabolical device’?

The usual answer was that the cat itself isperfectly capable of making a measurement.Bohr and his supporters said that ordinaryclassical physics applied to big or ‘macroscopic’things, like measuring devices, and that surelyincluded cats. However, this answer just raises afurther question. What does ‘big’ mean here?How big does something have to be to count

Erwin Schrödinger

224

WAV

ES O

F TH

E FU

TURE

as a measuring device and to stop the worldbeing fuzzy?

One way to make this problem vivid is to imaginea range of variants on the cat experiment, inwhich we replace the cat with progressivelysimpler ‘detectors’, all the way down tomicroscopic objects, such as an amoeba, or avirus, or a molecule, or an atom. A few of thesepossible variations are shown in Figure 1. Thenotation ψyes + ψno is just a way of writing thepossibility that quantum mechanics describes, inwhich it is not yet a determinate matter whetherthe radioactive atom in the source has decayed(‘yes’) or not (‘no’). The quantum state or wavefunction thus contains two components, onecorresponding to each possibility.

For good measure, I’ve also included a variant inwhich the cat is replaced by something a littlemore complex. This version of the experiment hasa name. It is called the ‘Wigner’s Friend’ thoughtexperiment, after the physicist Eugene Wigner,who suggested replacing Schrödinger’s cat with ahuman observer. Wigner thought that onlyconsciousness could stop the world being fuzzy.

Here’s the problem, in this new form. In which ofthese various experiments does a measurementtake place, to stop the world being fuzzy, before

the box is opened? ‘What counts as ameasurement?’ turns out to be one of thehardest problems to answer in quantummechanics. Today, it is called the QuantumMeasurement Problem. Many people think that itstill doesn’t have a satisfactory solution (althoughthis is a controversial issue). We don’t have timeto explore this debate here. Before we move on Iwant to emphasise two points.

First, I want to stress the main reason why this isa problem: according to the Copenhagen view ofquantum mechanics that Schrödinger wascriticising (which remains popular today), nothingdefinite happens in the quantum world until ameasurement is made. If that’s true, then it is avery important matter what counts as ameasurement. Until we know that, we haven’tunderstood why the world isn’t just ‘fuzz’, all theway up to the level of our experience.

The second thing I want to emphasise is that it isSchrödinger’s famous feline, seventy-years-oldthis year, which first puts her paw on this crucialissue. That’s why she’s so important, and that’swhy, as far as we can tell, she’ll have apermanent place in the mythology of physics,along with Archimedes’ bath, Galileo’s featherand Newton’s apple.

Figure 1 - variants on the Schrödinger cat experiment, with different 'detectors'.

225

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

Einstein and the Completeness ofQuantum MechanicsAs we’ve just seen, Schrödinger favoured theview that quantum mechanics gives us a fuzzypicture of a sharper reality. In other words, hethought that quantum mechanics is anincomplete description – a description that leavesout some of the details. But Schrödinger isn’t themost famous opponent of the view that quantummechanics is a complete description. That honourgoes to Einstein, and his strongest argument is ina famous paper written with two of his Princetoncolleagues, Boris Podolsky and Nathan Rosen,that appeared in the same year as Schrödinger’sCat. (In fact, Schrödinger’s paper was a responseto the Einstein-Podolsky-Rosen paper, in whichSchrödinger was offering further arguments forthe same conclusion.)

The Einstein-Podolsky-Rosen (EPR) paperintroduces a class of experiments that turn out toinvolve some of the strangest consequences ofquantum mechanics. Now known collectively asEPR experiments, the crucial feature of thesecases is that they involve a pair of particles thatinteract and then move apart. Provided theinteraction is set up in the right way, quantummechanics shows that the results ofmeasurements on one particle enable us topredict the results of correspondingmeasurements on the other particle. For examplewe might predict the result of measuring theposition of particle 1 by measuring the positionof particle 2, or predict the result of measuringthe momentum of particle 1 by measuring themomentum of particle 2 (see Figure 2).

This was the feature of these cases that interestedEinstein. In philosophical terms, Einstein was arealist – in other words, he believed that the worldexists independently of minds and observations.He had no time for Bohr’s view that realitydepends on what we humans choose to observe.And he thought that the features of quantummechanics that Bohr and others took as evidenceof deep entanglement between observation andreality were really a result of the fact that thetheory gives only a fuzzy description of reality. Ashe saw it, then, the crucial question is thereforewhether the quantum mechanical description ofreality can be considered to be complete. Does itsay all there is to be said about a physical system,or are there further facts about the physical worldnot captured by quantum mechanics?

The two-particle systems seemed to provide thedecisive argument that Einstein was looking for.He argued like so. First of all, he laid down whathe called a criterion of reality – in other words, aprinciple that tells us when there is somethingreal, ‘out there’ in the world. This criterion saysthat if we can predict with certainty what theresult of some measurement would be then theremust be an element of reality responsible for thatmeasurement. Let’s write this down explicitly.

Criterion of Reality: If we can predict withcertainty the result of a measurement ofsome physical quantity F, then there issomething in reality corresponding to F.

This is how Einstein, Podolsky and Rosen expressthis criterion: ‘If, without in any way disturbing a

Figure 2 - An EPR experiment

226

WAV

ES O

F TH

E FU

TURE

system, we can predict with certainty (i.e., withprobability equal to unity) the value of a physicalquantity, then there exists an element of realitycorresponding to that quantity.’ (Einstein,Podolsky and Rosen, 1935) (You might like tothink about whether you agree with this principle.If not, why not?)

The second important ingredient in the EPRargument is an assumption. It says that so longas the two particles are sufficiently far apart,what we do to one of them doesn’t affect theother one. Another way to put this is to say thatanything we do to one particle only has effectslocally – and that is why this assumption is calledthe assumption of locality.

Assumption of Locality: There is no action ata distance. Or as EPR put it: ‘If at the time ofmeasurement ... two systems no longerinteract, no real change can take place in thesecond system in consequence of anythingthat may be done to the first system.’(Einstein, Podolsky and Rosen, 1935)

At this point it’s important to keep in mindEinstein’s strongest reason for believing in thislocality principle. One of the fundamentalprinciples of his theory of relativity is that nothing– no particle, signal, or causal influence – cantravel faster than the speed of light. So if twoparticles are a long way apart – say, as far apartas the Earth and the Sun – then no change inone particle can affect the other particle for atleast eight minutes (the time it takes light totravel this distance). So from the EPR point ofview, the locality assumption seemed to beguaranteed by the biggest Swiss bank in town,the theory that Einstein himself had thought upwhile working in the patent office in Berne, thirtyyears earlier.

Let’s see how the EPR argument goes, giventhese two principles:

A. If we measure the position of Particle 1, wecan infer with certainty the result of aposition measurement on Particle 2.(Quantum mechanics tells us this.)

B. So, if we measure the position of Particle 1,then there is an element of reality

corresponding to the position of Particle 2.(This follows from A, by the Criterion ofReality).

C. But by the Assumption of Locality, what wedo to Particle 1 doesn’t affect Particle 2. Soit follows from B that there must be anelement of reality corresponding to theposition of Particle 2, regardless of whetherwe measure the position of Particle 1.

D. Similarly, going through the same threesteps for momentum instead of position,that there must be an element of realitycorresponding to the momentum of Particle2, regardless of whether we measure themomentum of Particle 1.

E. So, even if we don’t measure anything onParticle 1, there must be elements of realitycorresponding to both the position and themomentum of Particle 2. Or in other words,there is a sharp reality out there after all,and quantum mechanics is just a fuzzypicture of it.

Thus, Einstein believed that he had given aconclusive argument that quantum mechanicsonly gives us an incomplete description of reality.Accordingly, he thought that there must be extra‘hidden variables’, of which quantum mechanicsdidn’t provide us with any account. He seems tohave thought that quantum mechanics was just astatistical theory, describing the averagebehaviour of large collections of particles, a bitlike the theory of gases provided in statisticalmechanics (where properties such as temperatureand pressure are just averages, not fundamentalproperties of the real constituents of gases).

The response to this argument from Bohr and hisCopenhagen Interpretation followers isn’t easy todescribe. In fact, many people say that theysimply don’t understand it. However, becauseEinstein’s argument is so simple, and obviouslycorrect if you accept the two principles of theCriterion of Reality and the Assumption ofLocality, Bohr could only challenge it by rejectingone of these principles, or both. In fact, he seemsto have been committed to rejecting both: torejecting Locality, on the grounds that until ameasurement is made, the two particles are notgenuinely independent; and to rejecting theCriterion of Reality, on the grounds that there isn’t

227

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

a definite element of reality, until a measurementis actually made.

Einstein was aware that his opponents would tryto evade the argument in this way, but hewouldn’t have any of it, and for a very goodreason: as I noted earlier, his own theory ofspecial relativity provided the strongest argumentin favour of Locality. His attitude is nicelysummed-up in a famous remark in a letter to hisold friend and colleague, Max Born (1882-1970)in 1947. (Max Born was another of the fathers, orgrandfathers, of quantum theory. He was also thematernal grandfather of the Australian singer andactress, Olivia Newton-John - which means thatshe’s a kind of second cousin of quantummechanics!) Einstein writes to Born that he can’taccept quantum theory in its current form,because ‘the theory cannot be reconciled with theidea that physics should represent a reality inspace and time, free from spooky actions at adistance.’ (Letter to Born, 3 March 1947, myemphasis.)

However, the spooks were going to turn out to bemuch harder to eradicate than Einstein hadthought and, ironically, he himself had got halfway to showing why by focusing attention on thekinds of experiments involved in the EPRargument. The Irish physicist John Bell made thenext crucial step in 1965, about ten years afterEinstein’s death. The crucial part of Bell’sargument is remarkably straightforward, as we cansee by examining a case that has nothing to dowith quantum mechanics, at least on the surface.

The Scandinavian Institute ofSynchronised SwimmingImagine it is the year 2021. There’s turmoil in theworld of sport. Many countries are still reelingfrom their dismal performance at the AucklandOlympics the previous year where, for the firsttime in Olympic history, the host nation won everysingle medal! The three Scandinavian countries –Sweden, Norway and Denmark – decide to joinforces, and to concentrate on improving a singlesport. They choose synchronised swimming. Atgreat expense, the three governments establishthe Scandinavian Institute of SynchronisedSwimming (SISS) – demolishing the famous Neils

Bohr Institute (NBI) in Copenhagen, foundedexactly a century earlier, to make way for severalnew swimming pools.

From all over Scandinavia, hopeful pairs ofswimmers arrive in Copenhagen, hoping to bechosen for the elite training squad. Of course,there’s a rigorous selection procedure. Only themost committed and synchronised teams arewanted. Here’s how it works. The two candidatesare separated, and isolated in different interviewrooms. Each of them is then asked just onequestion, chosen from a list of three:

1. Would you like to swim for Sweden?2. Would you like to swim for Norway?3. Would you like to swim for Denmark?

The two questions are chosen at random,independently, in the two rooms, so sometimesthey’re the same (three times out of nine, in fact,on average) and sometimes they’re different. Andneither candidate knows what question the othercandidate is being asked.

Of course, if the two candidates are asked thesame question and give different answers, thenthat’s the end of the matter. They’re notsufficiently synchronised, and they’re both shownpolitely to the door. So consistency is absolutelyvital in this case. It’s better if they both say No tothe same question than if one says Yes and theother says No.

But what if they’re asked different questions? Inthis case, it’s no use if they both say Yes. Thatwould show that they wanted to swim fordifferent countries, and again, they’d be shownthe door. And it’s not much better if they both sayNo. That reduces their chances of getting pickedfor any of the national teams, even if they do getinto the Institute. So they have to try to givedifferent answers in this case: have one say Yesand the other say No.

So far, this is just simple sport psychology. Nowlet’s introduce a little bit of mathematics. Let’sfigure out the maximum possible success rate forthe second part of the strategy – the part thatapplies if they are asked different questions –given that they need to guarantee that theyalways give the same answer when they’re asked

228

WAV

ES O

F TH

E FU

TURE

the same question. Obviously, the only way toguarantee that they give the same answer ifthey’re asked the same question is for them toagree in advance what they’ll say, in response toeach of the three possible questions. We canwrite down their possible policies in this form:YYN, YNY, etc. Thus YYN means that they wouldanswer Yes to Questions 1 and 2, and No toQuestion 3. YNY means that they would answerYes to Questions 1 and 3, and No to Question 2.And so on.

It is easy to see that there are just eight possiblestrategies of this kind: YYY, YYN, YNY, NYY, YNN,NYN, NNY and NNN. Of these eight strategies,two (YYY and NNN) ensure that the candidatesgive the same answer, no matter what twoquestions they are asked. So these two strategiesare bad strategies. Remember, the candidates aretrying to maximise their chances of giving differentanswers when they are asked different questions.That leaves just six possible strategies: YYN, YNY,NYY, YNN, NYN, and NNY. Let’s pick one of these,say NYY, as an example, and think about what itimplies about the chances of a pair of candidateswho chose that strategy giving different answers,when they are asked different questions.

There are six possible ways the two candidatescan be asked different questions, as in thefollowing table; this also shows, for eachcombination of questions, whether the NYYcandidates manage to give different answers.

Candidate A Candidate B Different answers?

Q1 Q2 YesQ1 Q3 YesQ2 Q1 YesQ2 Q3 NoQ3 Q1 YesQ3 Q2 No

So with the strategy NYY, pairs of candidates canexpect a success rate of about 66% in theirattempt to give different answers when they’reasked different questions. Some will be lucky,some won’t, but on average, if the questions arechosen at random, there’ll still be a failure rate ofaround 33%.

It’s not hard to see that the same applies to anyof the other five strategies we just listed: YYN,YNY, YNN, NYN or NNY. In each case, there’ll betwo out of six possible combinations of questionsin the table for which the strategy doesn’t givedifferent answers – so still a failure rate of 33%.

Let’s summarise these conclusions. We’vededuced that there is no strategy for making surethat the two candidates give the same answerwhen they’re asked the same question, whichalso has a success rate higher than about 66%when they are asked different questions (where‘success’ means giving different answers, inthese cases). To put it another way, if they makesure they give the same answers to the samequestions, then they’ll also give the sameanswers to different questions, at least about33% of the time. As we’ve seen, it’s just a matterof arithmetic.

Is there any way to cheat the arithmetic? To dobetter than a 66% success rate, in the differentquestion cases? Notice that it would be easy todo better if the two swimmers couldcommunicate, and tell each other what their ownquestion is – if they were telepathic, for example,and could flash a message such as ‘I’ve got theSweden question!’ to their partner in the otherinterview room. But we’ve assumed that they’regenuinely isolated. In other words, we’veassumed that the question asked in one roomcan’t make any difference to the answer theother swimmer gives in the other room.

Does this assumption sound familiar? It should,because effectively it is the Assumption ofLocality, as used by Einstein, in his argument thatquantum mechanics is incomplete. In otherwords, we should really put our conclusion likethis:

SISS Theorem: If the Assumption of Localityis true for candidates interviewed for theSISS, then the maximum success rate in thedifferent-question cases is 66%.

If the Assumption of Locality is true, in otherwords, then there’s no way for the candidates tocheat the arithmetic.

229

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

Quantum Mechanics to the Rescue?What has this got to do with quantum theory?It’s simple. Somehow, quantum mechanics doesmanage to cheat the arithmetic. In fact, if thefounders of SISS hadn’t demolished the NielsBohr Institute so hastily, they would havediscovered that there’s a way to use quantummechanics to get a higher success rate than thearithmetic seems to allow. And the bit ofquantum mechanics we need is very similar tothe kind of experiments discussed by Einstein,in the EPR argument of 1935. Like thatargument, it involves physical systems in whichwe have two particles produced in somecommon source, which can then be measured in different locations.

In the original EPR experiment, we had a choiceof two measurements on each particle, eitherposition or momentum. But there are similarexperiments in which we can find three possiblemeasurements we can perform on each particle,each of them mutually exclusive with the othertwo. In other words, just as we can measureeither the position or the momentum but not bothin the original experiment (remember, this isHeisenberg’s Uncertainty Principle at work), so wecan measure just one of these three newproperties in the new EPR experiments.

One example is provided by the polarisation ofphotons, or particles of light. We can measure thepolarisation of a photon by putting a polarisinglense in front of it, and detecting whether itpasses through. And we can rotate the lense, andthus measure the polarisation in differentdirections. What we can’t do is measure thepolarisation in more than one direction at thesame time.

Thus versions of the EPR experiment withpolarisation measurements work just as well asEinstein’s original, for the purposes of Einstein’sargument. If we choose our pairs of photons inthe right way, we can find out the polarisation ofone photon, for a particular orientation of thelense, by measuring the polarisation of the otherphoton in the same orientation. Again, there’s aperfect correlation between the two results – ormore exactly, an anti-correlation, in the sensethat if one photon goes through the polariser, the

other one is always blocked, and vice versa. (Wecan turn it into a perfect correlation by rotatingone polariser through 90°.)

Another version of the EPR experiment useselectrons. In this case we measure a property ofthe electrons called ‘spin’, which is related toangular momentum. Like polarisation, it ismeasured in a chosen direction perpendicular tothe direction of travel of the particles. Whateverdirection we choose, electrons always turn out tohave a spin of either +1/2 or -1/2 (don’t worrytoo much about what the numbers mean) in thechosen direction; and if a pair of electrons isproduced in the right way, the total spin must bezero, and so there must be one of each. So aspin measurement on one particle enables us topredict the result of a correspondingmeasurement on the other particle (i.e., a spinmeasurement at the same angle, perpendicular tothe line of flight of the electron), just as in theoriginal EPR case.

In the spin case, the physics gives us a perfectanti-correlation. In other words, if we get a resultof +1/2 on one side we get -1/2 on the other. Butby making the measurement on one side reveal‘minus-spin’ (that is, making the device display+1/2 when it measures -1/2, and vice versa) weeasily turn this into a perfect correlation. Betterstill, we can set things up so that on one side themeasurement device shows YES when it records+1/2 and NO when it records -1/2, and on theother side, the same in reverse.

We then have something with exactly the sameform as the SISS case, if we let the threedifferent questions in the SISS case correspond tomeasuring the spin in three directions spaced at120° with respect to each other, perpendicular tothe line of flight. In effect, there are threedifferent ‘questions’ we can ask each electron,and the measuring device produces either a ‘Yes’or a ‘No’. And if we ask the same question, weget the same answer, on both sides.

The simple arithmetic we used in the SISS caseproves that if the questions are chosen at randomon each particle, then when the two particles areasked different questions they will produce thesame answer at least 33% of the time. As in the

230

WAV

ES O

F TH

E FU

TURE

SISS case, the argument depends on theAssumption of Locality – but as long as thatholds, it is just arithmetic.

This piece of simple arithmetic now has a namein quantum theory. It’s called Bell’s Inequality,after the physicist who first saw its importance. (Itis called an ‘inequality’ because it says that thecorrelation has to be at least 33%.) Why is itimportant? Because, as Bell realised, quantummechanics predicts something different.Depending on how we set up the experiment,quantum mechanics predicts a correlation as lowas 25%, in these cases in which the two particlesare subject to different measurements.

One way to see how surprising this is is to noticethat if SISS hadn’t demolished the Niels BohrInstitute, they could have used a real-life device,based on quantum mechanics, to cheat thearithmetic we derived above. In principle, it couldwork like this. Electrons or photons would beproduced in the right kind of pairs, and directedinto mirrored boxes, where they could be storeduntil needed.

Each candidate would take one box, and a three-setting measurement device. They’d be instructedto base the measurement setting on which of thethree possible questions they were asked, and tobase the answer to the question on the result ofthe measurement on the particle in the box. Withcareful experimental design, they could certainlydo better than the theoretical limit of 33% – inprinciple, according to quantum mechanics, theycould reduce the number of occasions on whichthey gave the same answer to different questionsto around 25%.

In other words, quantum mechanics enables ourswimmers to do something that is mathematicallyimpossible, if the Assumption of Locality is true.So quantum mechanics must imply that theAssumption of Locality is false! That was JohnBell’s great discovery.

Synchronised SpookinessThus Einstein had assumed Locality, and used it,in an ingenious argument based on these two-particle EPR experiments, to argue that quantummechanics is incomplete – that quantum theorymust be a fuzzy picture of a sharper reality. ButBell showed that those same EPR experimentscould be used to show that the predictions ofquantum mechanics were inconsistent with theAssumption of Locality. If quantum mechanics isright, then the Assumption of Locality is wronganyway, and Einstein’s argument for the fuzzyinterpretation collapses.

For this reason John Bell is sometimes called theman who proved Einstein wrong. But it isimportant to be clear what Bell actually provedEinstein to be wrong about. Bell did show thatEinstein must be wrong about the Assumption ofLocality (at least if quantum mechanics is true).But he didn’t show, as people often wronglyassume, that Einstein was wrong about quantummechanics being incomplete. It could still be true,as Einstein thought, that there are extra ‘hiddenvariables’, not described by quantum mechanics.It is just that they couldn’t be local hiddenvariables, satisfying the Assumption of Locality.Somehow, the measurement made on oneparticle would have to affect the hidden variablesof the other. (There are some well-developed

John Bell

231

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

extensions of quantum mechanics of this kind.The best-known was invented by the physicistDavid Bohm (1917-1992), who also invented theversion of the EPR experiment described above,on which Bell’s analysis was based.)

Where do we stand, then, in the light of Bell’sresult? First, Einstein’s best argument for thefuzzy picture view of quantum mechanics hasbeen seriously undermined. Secondly, and moreimportantly, Bell has put his finger on a simpleand basic fact: if quantum mechanics is true,then the Assumption of Locality is false, and viceversa. If quantum mechanics is true, in otherwords, there really is spooky action at a distance.

At the time of Bell’s original work,in the 1960s, experiments to testthe predictions of quantummechanics were not technicallyfeasible. But by the 1970s, variousexperimenters were devising waysto do it, and since then thepredictions of quantum mechanicshave been confirmed many times.(The best-known results are those of a team led bythe French physicist, Alain Aspect.) So very fewpeople doubt that ‘Non-locality’ is here to stay.

As we saw earlier, however, the strongest reasonfor believing in Locality is Einstein’s own specialtheory of relativity. Accordingly, the realimportance of Bell’s results is that they expose avery deep tension between the two mostimportant theories in twentieth century physics.Roughly speaking, quantum mechanics doessomething that simply shouldn’t be possible,according to special relativity. Quantum systemscan be ‘entangled’ in some strange way, evenwhen they are very long distances apart. (Inprinciple, we could arrange an EPR experiment inwhich the two particles had travelled light yearsapart. Almost everybody in physics now believesthat even in this case, the bizarre effects ofquantum entanglement would still apply.)

It is true that there are some subtleties here,which soften the blow a little. It looks as if thestrange non-local correlations that Bell noticed inEPR experiments can’t be used to send faster-than-light messages, for example – there’s noprospect of an instantaneous ‘Bell telephone’, assomeone once put it. But the conflict is there allthe same, and Bell himself thought that it impliedthat Einstein’s own understanding of the meaningof special relativity was wrong – that we had togo back to the ‘pre-revolutionary’ ideas thatphysicists such as Lorentz had developed before Einstein.

Back From the Future?To end this chapter, I want to describe anothercurious idea, sometimes suggested as a way ofresolving the tension between quantummechanics and relativity. I’ll introduce it by goingback to our synchronised swimmers in 2021,

trying to ensure that they givedifferent answers when they’reasked different questions at theSISS admission interviews. Thinkhow easy it would be for theswimmers if they hadprecognition – if they could just‘see’ in advance what questionthey were going to be asked. Ifthey knew this before they left

each other’s company, then it would be easy forthem to collaborate, to make sure they gavedifferent answers. (For example, suppose the two

David Bohm

“...the strongest reasonfor believing in Locality isEinstein’s own specialtheory of relativity.”

232

WAV

ES O

F TH

E FU

TURE

swimmers foresee that they are going to beasked questions 1 and 2 from the list of three.They could then adopt a strategy such as YNY, which gives different answers to these two questions).

This possibility reveals another hiddenassumption in the mathematical argument weused to show that the maximum possible successrate was around 66%. We were assuming,implicitly, that the swimmers didn’t know thequestions in advance – that their strategy had tobe independent of the choice of question in the future.

In the case of quantum particles, this amounts tothe assumption that the properties of the twoparticles in an EPR experiment cannot be affectedby the kind of measurement they’re going toencounter in the future. (If they’re affected by thefuture measurement, then they ‘know about it’, atleast metaphorically speaking; and again, there’sa possible way of cheating the arithmetic –getting a success rate higher than 66%.)

Of course, this assumption seems uncontroversial.How could photons and electrons possibly knowanything about what is going to happen to them

in the future? But it’s worth examining this issuea little more closely. After all, we don’t findanything controversial about the idea thatphotons and electrons know something aboutwhat happened to them in the past; or in otherwords, less metaphorically, that their propertiesdepend on what happened to them in the past.So why not the future, too?

At this point, we get to a fascinating tug ofintuitions. On the one hand, it seems just obviousthat causation only works one-way, from past tofuture – the past can affect the future, but thefuture can’t affect the past. On the other hand,however, the basic laws of physics seem to makeno distinction between the past and the future. Atthe fundamental level, physics is almost entirelytime-symmetric, in the sense that if it allows aprocess to happen then it also allows the reverseprocess to happen (roughly, what we would see ifwe reversed a video of the first process). Sowhere does the past-future bias of causationcome from, if it isn’t in the fundamental physics?

We don’t have time to explore these issues here.They would take us deep into philosophy, as wellas physics. But to finish up with, let’s think aboutwhat it would mean for the conflict between Bell’s

a point, such as when you throw astone into a still pond or spa bath.These symmetries turn out to begreater than many have previouslythought, and from this idea one canget almost a whole theory ofbackwards-in-time causation ...something which might come in veryhandy for contestants on Survivor.

If you could go back andspecialise in a different field,what would it be and why?When I was appointed to a PersonalChair at the University of Sydney, Iactually had the opportunity to namethe chair, and I thought of callingmyself Professor of ManagementConsulting, so that I could make a lotof money on Friday afternoons. So

that would be one option. It ispossible to make a bit of money inscience, but the really big bucks goto the less honest among us. Anotherthing I might have liked to specialisein would have been philosophy. Thatmight have been fun.

What’s the ‘next big thing’ inscience, in your opinion? What’scoming up in the next decade or so?As someone at the very theoreticalend of the spectrum it’s hard for meto know what will be next. Intheoretical science you can never saywith much confidence what the nextbig thing will be, because if youcould you’d already have it. But it’sfun to guess.

continued frompage 220

233

EIN

STEI

NAN

DTH

EQ

UAN

TUM

SP0O

KS

result and relativity, if we allowed ‘backwardcausation’. Suppose the properties of our twoelectrons are affected by the measurements wechoose to perform on them in the future. Thenwhen we choose what measurement to make onParticle 1, we affect its properties, all the wayback to the source. However, at the source,Particle 1 interacts with Particle 2. So byaffecting the properties of Particle 1, it ispossible, at least in theory, that we could affectParticle 2, as well; and Particle 2 could then carrythose effects into the future, to the time at whichits properties are measured, on the other side ofthe experiment.

But this means, the choice of measurement onone side of the EPR experiment could affect theresults on the other side – and all without anyspooky actions at a distance! All the actions areordinary local actions, and the only novelty is thatone of them works backwards in time.

So here’s a possible resolution of the mystery, aresolution which ought to make Einstein happy, atleast in one sense, because it avoids spookyaction at a distance. On this account, the ‘non-local’ effects that look like action at a distanceactually turn out to be the result of a combination

of two component actions, each of which isthoroughly compatible with relativity. (In otherwords, the proposal shows how we could have akind of pseudo-non-locality, that isn’t really intension with relativity. Apparent action at adistance gets resolved into a kind of zig-zageffect, where the ‘zig’ goes backwards in time.)

I stress that at this stage, this is just an intriguingidea. It hasn’t been developed very far, and mostphysicists and philosophers seem to think that thiskind of backward causation is even more spookythan the actions at a distance that Einstein hatedso much. But as I mentioned a moment ago, thetemporal bias of ordinary forward causation is itselfa bit spooky, or at least mysterious, in the light ofthe apparent time-symmetry of fundamentalphysics. So it’s just possible that this strange ideawill turn out to rid physics of two spooks, thoughadmittedly at the cost of some considerabledamage to naive ideas of cause and effect. If so,then it will have turned out that Einstein was rightafter all, in two ways: first, in thinking that there isno genuine action at a distance; and second, inbelieving that quantum mechanics is incomplete(for if this proposal is right, then ordinary quantummechanics leaves out the mechanisms responsiblefor these zig-zag causes).

I’d like to speculate that time-symmetric theories of physics,incorporating backwards causation,will be one big thing in theoreticalphysics. Also probably esoteric theoriesof computation: that field’s been a bitstatic since shortly after Turingfounded the field in the 1930s, but ifquantum computing works thenTuring’s whole theory will have to bereworked. Turing was an interestingcharacter. He is perhaps most famousfor his saying, “It’s amazing whatpeople will believe if they read it onthe web, even though they wouldn’tbelieve it if they read it anywhereelse.” Widely recognised as a genius atan early age, he was driven to suicidein the 1950s by the homophobia of hissociety. I can’t predict whether the

same thing will happen to the newtheoreticians of computing.

The third big thing I predict issomething like a cross between old-skool rap and progressive house,only with a bit of an acid jazzbackbeat. Either that or yet anotherblatant Christmas song, perhaps withRolf Harris.

Q&A with Professor Huw Price (with a little help from JasonGrossman)

234

WAV

ES O

F TH

E FU

TURE

Image creditsAll images are © the original owners, or asstated below.

Chapter 1Figure 1: C. R. Scotese, Paleomap Project,UniversityTexas, Arlington, U.S.A.Figure 4: Seebacher, F. and Alford, R. A.,1999, Movement and microhabitat useaterrestrial amphibian (Bufo marinus) on atropical island: seasonal variation andenvironmental correlates, Journal ofHerpetology, 33, pp 208-214.Figure 6, 7, 8: Mark Read, EPA, QLD.Figures 10, 11, 12: Seebacher, F., Franklin, C.E. and Read, M., 2005, Diving behaviourareptile (Crocodylus johnstoni) in the wild:interactions with heart rate and bodytemperature, Physiological and BiochemicalZoology, 78(1), pages 1-8, 2005, withpermission from Publisher: University ofChicago Press.

Chapter 2Figure 4: adapted from A. Ashkin, ScientificAmerican 226 (2), page 62 (1972).Figure 7: Michael Berns, Irvine, California.Figure 8, 12: P. Ormos, Institute ofBiophysics, Biological Research Centre of thethe Hungarian Academy of Sciences.Figure 9: K. Schütze, PALM Corporation,Germany.Figure 11b: K. Dholokia, School of Physics andAstronomy, St Andrews University, Scotland.

Chapter 3Figures 1, 2, 4, 5, 6: from Biomedical Uses ofRadiation, William R. Hendee, (ed.), Wiley-VCH: Brisbane (1999) ISBN 3-527-29668-9.

Chapter 5Figure 1: Stephen G. Eick, from the book Atlasof Cyberspace, M Dodge and R Kitchin(Addisom-Wesley), 2001 ISBN 0-201-74575-5.Figure 2: Lucent Technologies Inc.Figure 5, 6, 7, 8: Dr. R Slusher, Bell Labs,Lucent Technologies.

Chapter 6Figure 3: T.T. Ng, J.L. Blows, J.T. Mok, P. Hu,J.A. Bolger, P. Hambley, and B.J. Eggleton,Simultaneous residual chromatic dispersionmonitoring and frequency conversion withgain using a parametric amplifier, OpticsExpress 11, 3122-3127 (2003).Figure 5: S. Radic, C. J. McKinstrie, R. M.Jopson, J. C. Centanni, and A. R. Chraplyvy,All-Optical Regeneration in One- and Two-Pump Parametric Amplifiers Using HighlyNonlinear Optical Fiber, Photonic TechnologyLetters 15, 957-959 (2003).Figure 6: S. Radic, C. J. McKinstrie, R. M.Jopson, J. C. Centanni, and A. R. Chraplyvy,All-Optical Regeneration in One- and Two-Pump Parametric Amplifiers Using HighlyNonlinear Optical Fiber,” Photonic TechnologyLetters 15, 957-959 (2003).Figure 7: (a) from R.F. Cregan, B.J. Mangan,J.C. Knight, T.A. Birks, P.St.J. Russell, P.J.Roberts and D.C. Allan, “Single-modephotonic band gap guidancelight in air,Science 285, 1537-1539 (1999). (c)Photonics and Photonic Materials Group,University of Bath.

Chapter 7Figure 1: Morawska, L and Salthammer, T.,2003, Indoor Environment: Airborne Particlesand Settled Dust, (Wiley-Vch: Weinheim,Germany), ISBN 3-527-30525-4.

Chapter 9Figure 1: Davisson C and Germer L H,Physical Review 30(6), (1927) 705Figure 2, 3, 7: Cockayne DJH, PhysicsEducation 40 (2) (2005) 134 Figure 4: P. R. Buseck, et al., Proceedings ofthe National Academy of Science 98 (24)(2001) 13491Figure 5: C. Hetherington and A. Kirkland, Oxford.Figure 6: G. Winkelman and C. Dwyer, Oxford.Figure 8: copyright Materials Evaluation andEngineering, Inc, Plymouth.Figure 9: V. Keast, Sydney.Figure 10: from TEM: A textbook of materialsscience, Plenum Press (ISBN 03064532XX,1996).

Chapter 10Figure 1: Brian P. Gorman, Department ofMaterials Science and Engineering, Universityof North Texas.Figure 5: Tanaka, Terauchi and Kaneyama,JEOL Ltd, Tokyo.Figure 9: J. Brook, J. Sloan, A. Briggs,Department of Materials, Oxford University.Figure 10: K. Jacobi, Max Planck Institute.Figure 11: C. Lang, Department of Materials,Oxford University.Figure 18: G. Winkelman, Department ofMaterials, Oxford University.Figure 19: G. Winkelman, C. Dwyer,Department of Materials, Oxford University.Figure 20: from F Ross, IBM J. Res. Develop.44(4) July 2000.Figure 21: J. Sloan and A. Kirkland,Department of Materials, Oxford UniversityFigure 22: P. Midgely, Department of MaterialsScience and Metallurgy, University of Cambridge.

Chapter 11Figure 1, 2: Kandel, ER, Schwartz, JH, andJessell, TM, PrinciplesNeural Science, 3rd Ed,Appleton and Lange, Norwalk, Connecticut,1991, ISBN 0-8385-8034-3.Figure 6: Nunez, PL, Neocortical Dynamicsand EEG Rhythms, Oxford Univ. Press, Oxford,1995, ISBN 0-19-505728-7.

Chapter 12Figure 1, 3: Sager, D A, Introduction to OceanSciences (Wadsworth: Belmont, CA) 1998.Figure 2: Neuman, G and Pierson, W J, Jr,Principles of Physical Oceanography (PrenticeHall: Englewood Cliffs, NJ) 1966.Figure 4: Davies, J L, Geographical Variationin Coastal Development. 2nd Ed (Longman,London) 1980.Figure 5: Short, A D (ed), Beach andShoreface Morphodynamics (John Wiley andSons: Chichester) 1999. Reprinted from Atlasof the Oceans, Wind and Wave Climate,Young, I R and Holland, G J, page241,Copyright 1996, with permission fromElsevier.Figure 10: Short, A D, Australia beachsystems – nature and distribution. Journal ofCoastal Research, in press, 2005.

Chapter 13Figure 1: Nico Housen, European SouthernObservatory.

Figure 2: R. Williams and the Hubble DeepField Team (STScI, NASA).Figure 3, 4: Bennet et al.Figure 7: Anglo-Australian Observatory,photograph by David Malin.Figure 8: (a) ACS Science & EngineeringTeam, NASA; (b) Kirk Borne (STScI) andNASA; (c) NASA, H. Ford (JHU), G. Hillingworth(UCS/LO), M. Clampin (STScI), G. Hartig(STScI), the ACS Science Team and ESA.Figure 9: R. Boonsma & T. Oosterloo,RUG/ASTRON.Figura 10: CSIRO-ATNF.Figure 11: T.A. Oosterloo and K. Kovac,ASTRON/RUG.Figure 12: Putman et al. 2003 ApJ 586, 170.Figure 13: J.M. van de Hulst, RUG.Figure 14: (a) M. Rejkuba (ESO) et al. (b)J.Keene (SSC/Caltech) et al.(JPL,Caltech,NASA), (c) from Schiminovich D.van Gorkom J.H., van der Hulst J.M., KasowS. 1994, ApJ 423, L101.Figure15: NRAO/AUI and STScI/NASA,investigators Hibbard J.E., van der Hulst J.M.,Barnes J.E., Rich-Whitmode B. & Schweizer F.Figure 16: NRAO/AUI and J. Hibbard & J. vanGorkom.Figure 18: J. Barnes.Figure 19: NASA/CXC/STScI/U.NorthCarolina/G. Cecil.Figure 20: Simon Driver & Alberto Fernandez-Soto, UNSW.

Chapter 14Figure 1: X-ray - NASA/CXC/MIT/UCSB/P.Ogleet al; Optical - NASA/STScI/A.Capetti et al.;John Bahcall (IAS), NASA); Alan Bridle;NASA/CXC/M. Karovska et al.); NRAO/VLAvan Gorkom/Schiminovich et al.; J.Condon etal.; Digital Sky Survey U.K. Schmidt/STScI.Figure 2: CXC/M.Weiss.Figure 3: NRAO/AUI Miyoshi et al. and GeraldCecil and Holland Ford.Figure 4: NASA and Ann Field (SpaceTelescope Science Institute).Figure 5: W. Jaffe (Leiden), H. Ford(JHU/STScI) and NASA).Figure 8: NRAO/AUI, Fomalont et al.Figure 10: R. Perley, C. Carilli & J. Draher,NRAO/AUI; Krichbaum.Figure 12: NRAO/AUI.Figure 13: Polatidis et al.Figure 14: Gomez Jose1-Luis et al.Figure 16: NASA and J. Bachall (AS) - NASA,Martel et al., the ACS Science Team and ESA.Figure 17: L. Saripalli, R. Subrahmanyan,Udaya Shankar (ATCA).

Chapter 15Figure 4: AIP Emilio Segré Visual ArchivesFigure 6: photograph by Francis Simon,courtesy AIP Emilio Segré Visual Archives,Francis Simon Collection.Figure 8: Nick Robins, ANU.Figure 9: Nick Robins, ANU.

Chapter 16Figure 1: (b) Alan Henderson, copyrightMuseum Victoria, AustraliaErwin Schrodinger: photograph by FrancisSimon, courtesy AIP Emilio Segré VisualArchives, Francis Simon Collection.Neils Bohr: Neils Bohr Archive, Copenhagen.

Einstein on a bicycle, page 36: the Archives,California Institute of Technology.

235

THE

MES

SEL

END

OW

MEN

T

THE INTERNATIONAL Science Schools were established byProfessor Harry Messel in 1962 to recognise and rewardtalented senior high school students, and to encourage themto pursue careers in science. Alumni of the InternationalScience Schools can be found in senior positions in all walksof life, with many of them acknowledging that ‘their’ ScienceSchool was responsible for changing their lives, and recallingits two weeks as an exciting developmental experience.

2005. There is information in thebooklet given to you on how toapply. Secondly, you couldapproach someone in yourneighbourhood, for example acompany or philanthropic institutionor even a member of your familywith the capacity to contribute, toseek a donation for the MesselEndowment. A personal approach isalways the best method ofobtaining such support and aperson such as yourself, who hasjust benefited from an InternationalScience School, makes the bestambassador. Finally, we realise thatyou may not be in a position, atpresent, to give to the Endowmentyourself – however, at some stagein the future, if you are in a positionto do so, we ask that you considerit then. A donation form has beenincluded on the next page to assistyou if you choose to proceed withsome of these suggestions. Moredonation forms are available onwww.physics.usyd.edu.au/messel/index.htm.

A contribution of A$25,000 to theMessel Endowment will ensure theparticipation of one student inperpetuity.

Please join us today in our visionfor the young scientists oftomorrow through the MesselEndowment.

Associate Professor Robert G HewittDirector,Science Foundation for PhysicsJune 2005

The Messel Endowment

The Science Schools, renamed the Professor Harry MesselInternational Science Schools in1999 to honour Harry for hisforesight, have continueduninterrupted since then. In orderto ensure the future of theInternational Science Schools, sothat future students may alsobenefit, the Science Foundation forPhysics has established the MesselEndowment. The Foundation aimsto raise $4,500,000 in 2005Australian dollars. At the time ofprinting, the fundraising campaignhas come a long way: in 2005, theEndowment sits at just over half ofthe amount needed. Despite thissuccess, for which we areenormously grateful, theEndowment still has a long way togo, and the Foundation isdetermined to achieve the target.

The Messel Endowment will bemanaged so that the real value ofthe capital is preserved. Bequestswill also be sought to ensure thegrowth of the capital in the years tocome. The primary purpose of theEndowment is to support theInternational Science Schools. Ifincome from the Endowmentexceeds the requirements of theScience Schools the funds may,with the approval of the Science

Foundation, be used to supportother initiatives names to honourProfessor Harry Messel. Examplesof such initiatives include aProfessor Harry Messel VisitingChair and a Professor Harry MesselLectureship.

As acknowledged with gratitude inthe front of the book, many peopleand companies have alreadycontributed to the MesselEndowment, and the 2005Professor Harry MesselInternational Science School hasreceived considerable benefit fromthe amount already raised. Donorsare also honoured on a web page,www.physics.usyd.edu.au/messel/index.htm; they will beacknowledged in all books of thelectures presented at futureInternational Science Schools, andon a permanent display in theSchool of Physics.

We hope you have enjoyed yourtime at the International ScienceSchool. Many people, afterexperiencing the ScienceSchools, wish to give somethingback, but are not sure how to dothis. Here are some suggestions.First, you could consider comingback as staff for the nextInternational Science School in

236

WAV

ES O

F TH

E FU

TURE

THE MESSEL ENDOWMENT DONATION FORMYou can contribute to the Messel Endowment in one of a number of ways. All donations to the Messel Endowment by Australianresidents are fully tax deductible. Donations by cheque must be posted to the address given below. Donations by credit card may befaxed to the number given below.

NAME

COMPANY

ADDRESS

STATE POSTCODE COUNTRY

TELEPHONE (BH) (AH)

PLEASE FILL IN THIS SECTION FOR A ONCE-ONLY DONATION

I enclose a cheque, made out to the University of Sydney for A$Please charge my credit card nominated below for the sum of A$❍ Bankcard ❍ Visa ❍ MastercardCARD NUMBER / / /EXPIRY DATE /CARDHOLDER NAME DATE /CARDHOLDER SIGNATURE

OR

PLEASE FILL IN THIS SECTION TO GIVE A YEARLY DONATION(AN ANNUAL REMINDER WILL BE ISSUED).

I undertake to give A$ once a year for ❍ 2, ❍ 3, ❍ 4 or ❍ 5 years and enclose a cheque, made out to the University of Sydney as a first installment.

OR

Please charge the credit card nominated below for the sum of A$ once a year from / (insert month & year)for ❍ 2, ❍ 3, ❍ 4 or ❍ 5 years.❍ Bankcard ❍ Visa ❍ MastercardCARD NUMBER / / /EXPIRY DATE /CARDHOLDER NAME DATE /CARDHOLDER SIGNATURE

The Science Foundation for Physics Residents of the USA should contactSchool of Physics A28 The SecretaryThe University of Sydney NSW 2006 The University of Sydney USA Foundation AustraliaFacsimile: +612 9351 7726 Telephone: (607) 277 1346

Facsimile: (607) 277 0078Email: semmler@clarityconnect.com

Thank you for your investment in the future of young people through The Messel Endowment.