The Secret Lives of Stars - Physics and Astronomyphysics.bgsu.edu/~layden/TMP/A321/Articles/SciAm/secret_stars_2004.pdfTHE SECRET LIVES OF STARS 3Letter from the Editor YOUTH 4The

WWW.SCIAM.COM

COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

http://www.sciam.com/index.cfm?ref=digitalpdf

T H E S E C R E T L I V E S O F S T A R S

3 Letter from the Editor

Y O U T H

4 The First Stars in the UniverseBY RICHARD B. LARSON AND VOLKER BROMM

Exceptionally massive and bright, the earliest stars changed the course of cosmic history.

12 Fountains of Youth:Early Days in the Life of a Star

BY THOMAS P. RAY

To make a star, gas and dust must fallinward. So why do astronomers see stuff

streaming outward?

18 Companions to Young Stars

BY ALAN P. BOSS

The surprising finding that even the youngest starscommonly exist in sets of two or three has revised

thinking about the birth of star systems.

26 The Discovery of Brown DwarfsBY GIBOR BASRI

Less massive than stars but more massive than planets,brown dwarfs were long assumed to be rare. New sky surveys, however, show that the objects may be as common as stars.

SCIENTIFIC AMERICAN Volume 14, Number 4

The Secret

Lives of Stars co

nten

ts20

04

42


2 T H E S E C R E T L I V E S O F S T A R S

M I D L I F E

34 The Stellar DynamoBY ELIZABETH NESME-RIBES, SALLIE L. BALIUNAS AND DMITRY SOKOLOFF

Sunspot cycles—on other stars—are helping astronomersstudy the sun’s variations and the ways they might affect Earth.

42 The Fury of Solar StormsBY JAMES L. BURCH

Shock waves from the sun can endanger Earth’s satellites and astronauts.

50 When Stars CollideBY MICHAEL SHARA

When two stars smash into each other, it can be a verypretty sight. Once considered impossible, theseoccurrences have turned out to be common in certaingalactic neighborhoods.

58 X-ray BinariesBY EDWARD P. J. VAN DEN HEUVEL AND JAN VAN PARADIJS

When ultradense neutron stars feed on their more sedatecompanions, the binary systems produce outpourings of x-rays and drastically alter the evolution of both stars.

68 MagnetarsBY CHRYSSA KOUVELIOTOU, ROBERT C. DUNCAN ANDCHRISTOPHER THOMPSON

Magnetized so intensely, some stars alter the very natureof the quantum vacuum.

O L D A G E

76 Supersoft X-ray Stars andSupernovaeBY PETER KAHABKA, EDWARD P. J. VAN DEN HEUVEL AND SAUL A. RAPPAPORT

Supersoft sources—which spew unusually low-energy x-rays—are now thought to be white dwarf stars thatsiphon matter from their stellar companions and then, inmany cases, explode.

84 Binary Neutron StarsBY TSVI PIRAN

The inevitable collapse of these paired stellar remnantsgenerates runaway heating that, for a few weeks, emitsmore light than an entire galaxy.

92 The Brightest Explosionsin the UniverseBY NEIL GEHRELS, LUIGI PIRO AND PETER J. T. LEONARD

Every time a gamma-ray burst goes off, a black hole is born.

Scientific American Special (ISSN 1048-0943), Volume 14, Number 4, 2004, published by ScientificAmerican, Inc., 415 Madison Avenue, New York, NY 10017-1111. Copyright © 2004 by ScientificAmerican, Inc. All rights reserved. No part of this issue may be reproduced by any mechanical,photographic or electronic process, or in the form of a phonographic recording, nor may it be storedin a retrieval system, transmitted or otherwise copied for public or private use without writtenpermission of the publisher. Canadian BN No. 127387652RT; QST No. Q1015332537. To purchaseadditional quantities: U.S., $10.95 each; elsewhere, $13.95 each. Send payment toScientific American, Dept. STARS, 415 Madison Avenue, New York, NY 10017-1111.Inquiries: fax 212-355-0408 or telephone 212-451-8890. Printed in U.S.A.

Cover illustration by Don Dixon

9234


E D I T O R I N C H I E F : John Rennie E X E C U T I V E E D I T O R : Mariette DiChristina I S S U E E D I T O R : Mark Fischetti I S S U E C O N S U L T A N T : George Musser

A R T D I R E C T O R : Edward Bell I S S U E D E S I G N E R : Patti NemotoP H O T O G R A P H Y E D I T O R S : Bridget Gerety Small, Emily Harrison, Smitha AlampurP R O D U C T I O N E D I T O R : Richard Hunt

C O P Y D I R E C T O R : Maria-Christina Keller C O P Y C H I E F : Molly K. Frances C O P Y A N D R E S E A R C H : Daniel C. Schlenoff, Michael Battaglia, David Labrador, Sara Beardsley

E D I T O R I A L A D M I N I S T R A T O R : Jacob Lasky S E N I O R S E C R E T A R Y : Maya Harty

A S S O C I A T E P U B L I S H E R , P R O D U C T I O N : William Sherman M A N U F A C T U R I N G M A N A G E R : Janet Cermak A D V E R T I S I N G P R O D U C T I O N M A N A G E R : Carl Cherebin P R E P R E S S A N D Q U A L I T Y M A N A G E R : Silvia Di Placido P R I N T P R O D U C T I O N M A N A G E R : Georgina Franco P R O D U C T I O N M A N A G E R : Christina Hippeli C U S T O M P U B L I S H I N G M A N A G E R : Madelyn Keyes-Milch

A S S O C I A T E P U B L I S H E R / V I C E P R E S I D E N T , C I R C U L A T I O N :Lorraine Leib Terlecki C I R C U L A T I O N D I R E C T O R : Katherine Corvino F U L F I L L M E N T A N D D I S T R I B U T I O N M A N A G E R : Rosa Davis

V I C E P R E S I D E N T A N D P U B L I S H E R : Bruce Brandfon W E S T E R N S A L E S M A N A G E R : Debra SilverS A L E S D E V E L O P M E N T M A N A G E R : David Tirpack W E S T E R N S A L E S D E V E L O P M E N T M A N A G E R : Valerie BantnerS A L E S R E P R E S E N T A T I V E S : Stephen Dudley, Hunter Millington, Stan Schmidt

A S S O C I A T E P U B L I S H E R , S T R A T E G I C P L A N N I N G : Laura Salant P R O M O T I O N M A N A G E R : Diane Schube R E S E A R C H M A N A G E R : Aida Dadurian P R O M O T I O N D E S I G N M A N A G E R : Nancy Mongelli G E N E R A L M A N A G E R : Michael Florek B U S I N E S S M A N A G E R : Marie Maher M A N A G E R , A D V E R T I S I N G A C C O U N T I N G A N DC O O R D I N A T I O N : Constance Holmes

D I R E C T O R , S P E C I A L P R O J E C T S : Barth David Schwartz

M A N A G I N G D I R E C T O R , O N L I N E : Mina C. Lux O P E R A T I O N S M A N A G E R , O N L I N E : Vincent MaS A L E S R E P R E S E N T A T I V E , O N L I N E : Gary Bronson

D I R E C T O R , A N C I L L A R Y P R O D U C T S : Diane McGarvey P E R M I S S I O N S M A N A G E R : Linda Hertz M A N A G E R O F C U S T O M P U B L I S H I N G : Jeremy A. Abbate

C H A I R M A N E M E R I T U S : John J. Hanley C H A I R M A N : John Sargent P R E S I D E N T A N D C H I E F E X E C U T I V E O F F I C E R : Gretchen G. Teichgraeber V I C E P R E S I D E N T A N D M A N A G I N G D I R E C T O R ,I N T E R N A T I O N A L : Dean Sanderson V I C E P R E S I D E N T : Frances Newburg

The Secret Lives of Stars is published by the staff of Scientific American, with project management by:

w w w . s c i a m . c o m T H E S E C R E T L I V E S O F S T A R S 3

Established 1845

®

FOR PURE THEATRICS andspectacle, Hollywood ce-lebrities have nothing onthe denizens of the heav-ens. Stars are born, live anddie in fiery and fascinatingways—ways that we haveonly recently been able tostudy in greater detail, likeso many swarming papa-razzi, using the long-rangelenses created by improvedtechniques and new, sharp-er observatories.

We’ve observed someeye-popping behavior fromthese brazen orbs. As withmany glamorous beings,stars often pair off, but thetwo may blow apart—liter-ally—one star ripping theshine from its former part-

ner in a highly public display. It seems that only about half the stars live as cou-ples—no better than the average state of human relationships. Or consider thewondrous strange doings of the magnetar, a name itself worthy of a characterin the blockbuster X-Men films. These intensely magnetized stars emit hugebursts of magnetic energy that can alter the very nature of the quantum vacu-um. Like rebels without a cause, they are furiously active for only a short time:after 10,000 years they wink out. We’ve also been able to see how, after a longera of flamboyant x-ray emission, x-ray binaries settle down to become someof the most steady, unchanging entities in the cosmos. Not all of these celestialobjects make it as luminaries, however. We may feel some sympathy for browndwarfs, failed stars that glow so dully nobody could even find one until 1995.

In this special edition from Scientific American, we invite you to forget abouteveryday life to spend some time with the stars. In the pages that follow, you’llfind the latest gossip on the glitterati, written by the astronomer shutterbugsthemselves. Although the stars have revealed more than ever, they’ve beencareful not to tell all, lest we grow bored with their antics. As authors ChryssaKouveliotou, Robert C. Duncan and Christopher Thompson so aptly put it in“Magnetars”: “What other phenomena, so rare and fleeting that we have notrecognized them, lurk out there?” We can hardly wait to find out.

MAR

K A

. G

ARLI

CK

ENCOUNTER with a tempestuous starlet tears apart our sun.

letter from the editor

Exposé on the Stars

Mariette DiChristinaExecutive Editor

Scientific [email protected]



mailto:[email protected]

4 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e D e c e m b e r 2 0 0 1 i s s u e

UNIVERSESTARSIN THE

FIRST

BY RICHARD B. LARSONAND VOLKER BROMM

ILLUSTRATIONS BY DON DIXON

Exceptionally massive and bright, the earliest starschanged the course of cosmic history

We live in a universe that is full of bright objects. On a clear night one can see thousands ofstars with the naked eye. These stars occupy mere-ly a small nearby part of the Milky Way galaxy; tele-scopes reveal a much vaster realm that shineswith the light from billions of galaxies. According toour current understanding of cosmology, howev-er, the universe was featureless and dark for a longstretch of its early history. The first stars did notappear until perhaps 100 million years after thebig bang, and nearly a billion years passed beforegalaxies proliferated across the cosmos. Astron-omers have long wondered: How did this dramat-ic transition from darkness to light come about?

THE


EARLIEST COSMIC STRUCTURE most likely took the form of a network of filaments. The first protogalaxies, small-scale systems about 30 to 100 light-yearsacross, coalesced at the nodes of this network. Inside the protogalaxies, the denser regions of gas collapsed to form the first stars (inset).

EARLIEST COSMIC STRUCTURE most likely took the form of a network of filaments. The first protogalaxies, small-scale systems about 30 to 100 light-yearsacross, coalesced at the nodes of this network. Inside the protogalaxies, the denser regions of gas collapsed to form the first stars (inset).


After decades of study, researchershave recently made great strides towardanswering this question. Using sophisti-cated computer simulation techniques,cosmologists have devised models thatshow how the density fluctuations leftover from the big bang could haveevolved into the first stars. In addition,observations of distant quasars have al-lowed scientists to probe back in timeand catch a glimpse of the final days ofthe “cosmic dark ages.”

The new models indicate that the first

stars were most likely quite massive andluminous and that their formation wasan epochal event that fundamentallychanged the universe and its subsequentevolution. These stars altered the dynam-ics of the cosmos by heating and ionizingthe surrounding gases. The earliest starsalso produced and dispersed the firstheavy elements, paving the way for theeventual formation of solar systems likeour own. And the collapse of some of thefirst stars may have seeded the growth ofsupermassive black holes that formed inthe hearts of galaxies and became thespectacular power sources of quasars. Inshort, the earliest stars made possible theemergence of the universe that we see to-day—everything from galaxies and qua-sars to planets and people.

The Dark AgesTHE STUDY of the early universe ishampered by a lack of direct observa-tions. Astronomers have been able to ex-amine much of the universe’s history bytraining their telescopes on distant galax-ies and quasars that emitted their lightbillions of years ago. The age of each ob-ject can be determined by the redshift ofits light, which shows how much the uni-verse has expanded since the light wasproduced. The oldest galaxies andquasars that have been observed so far

date from about a billion years after thebig bang (assuming a present age for theuniverse of 13.7 billion years). Re-searchers will need better telescopes tosee more distant objects dating from stillearlier times.

Cosmologists, however, can makedeductions about the early universebased on the cosmic microwave back-ground radiation, which was emittedabout 400,000 years after the big bang.The uniformity of this radiation indicatesthat matter was distributed very smooth-ly at that time. Because there were nolarge luminous objects to disturb the pri-mordial soup, it must have remainedsmooth and featureless for millions ofyears afterward. As the cosmos expand-ed, the background radiation redshifted

to longer wavelengths and the universegrew increasingly cold and dark. As-tronomers have no observations of thisdark era. But by a billion years after thebig bang, some bright galaxies andquasars had already appeared, so the firststars must have formed sometime before.When did these first luminous objectsarise, and how might they have formed?

Many astrophysicists, including Mar-tin Rees of the University of Cambridgeand Abraham Loeb of Harvard Universi-ty, have made important contributions to-

ward solving these problems. The recentstudies begin with the standard cosmo-logical models that describe the evolutionof the universe following the big bang. Al-though the early universe was remarkablysmooth, the background radiation showsevidence of small-scale density fluctua-tions—clumps in the primordial soup.These clumps would gradually evolveinto gravitationally bound structures.Smaller systems would form first and thenmerge into larger agglomerations. Thedenser regions would take the form of anetwork of filaments, and the first star-forming systems—small protogalaxies—

would coalesce at the nodes of this net-work. In a similar way, the protogal-axies would then merge to form galaxies,and the galaxies would congregate intogalaxy clusters. The process is ongoing:although galaxy formation is now most-ly complete, galaxies are still assemblinginto clusters, which are in turn aggregat-ing into a vast filamentary network thatstretches across the universe.

According to the models, the firstsmall systems capable of forming starsshould have appeared between 100 mil-lion and 250 million years after the bigbang. These protogalaxies would havebeen 100,000 to one million times moremassive than the sun and would havemeasured 30 to 100 light-years across.These properties are similar to those of

6 S C I E N T I F I C A M E R I C A N T H E S E C R E T L I V E S O F S T A R S

■ Computer simulations show that the first stars should have appeared between100 million and 250 million years after the big bang. They formed in small protogalaxies that evolved from density fluctuations in the early universe.

■ Because the protogalaxies contained virtually no elements besides hydrogenand helium, the physics of star formation favored the creation of bodies thatwere many times more massive and luminous than the sun.

■ Radiation from the earliest stars ionized the surrounding hydrogen gas. Somestars exploded as supernovae, dispersing heavy elements throughout theuniverse. The most massive stars collapsed into black holes. As protogalaxiesmerged to form galaxies, the black holes possibly became concentrated in the galactic centers.

Overview/The First Stars

It seems safe to conclude that the FIRST STARS IN THE UNIVERSE WERE TYPICALLY MANY TIMES

more massive and luminous than the sun.


the molecular gas clouds in which starsare currently forming in the Milky Way,but the first protogalaxies would havediffered in fundamental ways. For one,they would have consisted mostly ofdark matter, the putative elementary par-ticles that are believed to make up 90 per-cent of the universe’s mass. In present-day large galaxies, dark matter is segre-gated from ordinary matter: over time,ordinary matter concentrates in thegalaxy’s inner region, whereas the darkmatter remains scattered throughout anenormous outer halo. But in the proto-galaxies, the ordinary matter would stillhave been mixed with the dark matter.

The second important difference isthat the protogalaxies would have con-tained no significant amounts of any ele-ments besides hydrogen and helium. Thebig bang produced hydrogen and helium,but most of the heavier elements are cre-ated only by the thermonuclear fusion re-actions in stars, so they would not havebeen present before the first stars hadformed. Astronomers use the term “met-als” for all these heavier elements. Theyoung metal-rich stars in the Milky Wayare called Population I stars, and the oldmetal-poor stars are called Population II

stars. The stars with no metals at all—thevery first generation—are sometimescalled Population III stars.

In the absence of metals, the physicsof the first star-forming systems wouldhave been much simpler than that of pres-ent-day molecular gas clouds. Further-more, the cosmological models can pro-vide, in principle, a complete descriptionof the initial conditions that preceded thefirst generation of stars. In contrast, thestars that arise from molecular gas cloudsare born in complex environments thathave been altered by the effects of previ-ous star formation. Several researchgroups have used computer simulations toportray the formation of the earliest stars.

A team consisting of Tom Abel, GregBryan and Michael L. Norman (now atPennsylvania State University, ColumbiaUniversity and the University of Califor-nia at San Diego, respectively) has madethe most realistic simulations. In collab-oration with Paolo Coppi of Yale Uni-versity, we have done simulations basedon simpler assumptions but intended toexplore a wider range of possibilities.Toru Tsuribe, now at Osaka Universityin Japan, has made similar calculationsusing more powerful computers. The

work of Fumitaka Nakamura and Ma-sayuki Umemura (now at Niigata andTsukuba universities in Japan, respec-tively) has yielded instructive results. Allthese studies have produced similar de-scriptions of how the earliest stars mighthave been born.

Let There Be Light!THE SIMULATIONS show that the pri-mordial gas clouds would typically format the nodes of a small-scale filamentarynetwork and then begin to contract be-cause of their gravity. Compression wouldheat the gas to temperatures above 1,000kelvins. Some hydrogen atoms would pairup in the dense, hot gas, creating traceamounts of molecular hydrogen. The hy-drogen molecules would then start tocool the densest parts of the gas by emit-ting infrared radiation after they collidedwith hydrogen atoms. The temperaturein the densest parts would drop to 200 to300 kelvins, reducing the gas pressure inthese regions, allowing them to contractinto gravitationally bound clumps.

This cooling plays an essential role inallowing the ordinary matter in the pri-mordial system to separate from the darkmatter. The cooling hydrogen would set-

w w w . s c i a m . c o m S C I E N T I F I C A M E R I C A N 7

FROM THE DARK AGES . . .After the emission of the cosmic microwave background radiation (about 400,000 years afterthe big bang), the universe grew increasingly cold and dark. But cosmic structure graduallyevolved from the density fluctuations left over from the big bang.

COSMIC TIMELINE

1 million years100 million years

1 billion years12 to 14 billion years

BIGBANG

Emission of cosmic background

radiation Dark ages

First stars

Protogalaxymergers

Modern galaxies

First supernovaeand

black holes. . . TO THE RENAISSANCEThe appearance of the first stars and protogalaxies(perhaps as early as 100 million years after the big bang) set off a chain of events that transformed the universe.




PRIMEVAL TURMOILThe process that led to the creation of the first stars wasvery different from present-day star formation. But theviolent deaths of some of these stars paved the way for theemergence of the universe that we see today.

2 The cooling of the hydrogen allowed the ordinary matter to contract, whereas

the dark matter remained dispersed. The hydrogen settled into a disk at the centerof the protogalaxy.

THE BIRTH AND DEATH OF THE FIRST STARS

3 The denser regions of gas contractedinto star-forming clumps, each hundreds

of times as massive as the sun. Some of theclumps of gas collapsed to form verymassive, luminous stars.

4Ultraviolet radiation from the starsionized the surrounding neutral hydrogen

gas. As more and more stars formed, thebubbles of ionized gas merged and theintergalactic gas became ionized.

1 The first star-forming systems—smallprotogalaxies—consisted mostly of the

elementary particles known as dark matter(shown in red). Ordinary matter—mainlyhydrogen gas (blue)—was initially mixed with the dark matter.

6Gravitational attraction pulled theprotogalaxies toward one another.

The collisions most likely triggered starformation, just as galactic mergers do now.

7 Black holes possibly merged to form asupermassive hole at the protogalaxy’s

center. Gas swirling into this hole might havegenerated quasarlike radiation.

5 A few million years later, at the end oftheir brief lives, some of the first stars

exploded as supernovae. The most massivestars collapsed into black holes.

Black hole

Supernova

Ultravioletradiation


tle into a flattened rotating configurationthat was clumpy and filamentary and pos-sibly shaped like a disk. But because thedark-matter particles would not emit ra-diation or lose energy, they would remainscattered in the primordial cloud. Thus,the star-forming system would come toresemble a miniature galaxy, with a diskof ordinary matter and a halo of darkmatter. Inside the disk, the densest clumpsof gas would continue to contract, andeventually some of them would undergoa runaway collapse and become stars.

The first star-forming clumps weremuch warmer than the molecular gasclouds in which most stars currentlyform. Dust grains and molecules con-taining heavy elements cool the present-day clouds much more efficiently to tem-peratures of only about 10 kelvins. Theminimum mass that a clump of gas musthave to collapse under its gravity is calledthe Jeans mass, which is proportional tothe square of the gas temperature and in-versely proportional to the square root ofthe gas pressure. The first star-formingsystems would have had pressures similarto those of present-day molecular clouds.But because the temperatures of the firstcollapsing gas clumps were almost 30times higher, their Jeans mass would havebeen almost 1,000 times larger.

In molecular clouds in the nearbypart of the Milky Way, the Jeans mass isroughly equal to the mass of the sun, andthe masses of the prestellar clumps areabout the same. If we scale up, we can es-timate that the masses of the first star-forming clumps would have been 500 to1,000 solar masses. The computer sim-ulations mentioned above showed theformation of clumps with masses of sev-eral hundred solar masses or more.

Our group’s calculations suggest thatthe predicted masses of the first star-forming clumps are not very sensitive tothe assumed cosmological conditions.The predicted masses depend primarilyon the physics of the hydrogen moleculeand only secondarily on the cosmologi-cal model or simulation technique. Onereason is that molecular hydrogen can-not cool the gas below 200 kelvins, mak-ing this a lower limit to the temperatureof the first star-forming clumps. Anoth-

er is that the cooling from molecular hy-drogen becomes inefficient at the higherdensities encountered when the clumpsbegin to collapse. At these densities thehydrogen molecules collide with otheratoms before they have time to emit aninfrared photon; this raises the gas tem-perature and slows the contraction untilthe clumps have built up to at least a fewhundred solar masses.

Did the first collapsing clumps formstars with similarly large masses, or didthey fragment and form many smallerstars? The research groups have pushedtheir calculations to the point at whichthe clumps are well on their way to form-ing stars, and none of the simulations hasyet revealed any tendency for the clumpsto fragment. This agrees with our under-standing of present-day star formation;the fragmentation of clumps is typicallylimited to the formation of binary sys-tems (two stars orbiting around eachother). Fragmentation seems even lesslikely to occur in the primordial clumps,because the inefficiency of molecular hy-drogen cooling would keep the Jeansmass high. The simulations, however,have not yet determined the final out-come of collapse with certainty, and theformation of binary systems cannot beruled out.

Precise estimates of just how massivethe first stars might have been are diffi-cult because of feedback effects. In gen-eral, a star forms from the “inside out,”by accreting gas from the surroundingclump onto a central protostellar core.But when does this accretion processshut off? As the star grows in mass, itproduces intense radiation and matteroutflows that may blow away some ofthe gas in the collapsing clump. Yet theseeffects depend strongly on the presenceof heavy elements, and therefore theyshould be less important for the earlier

stars. In collaboration with Loeb of Har-vard, one of us (Bromm) has recentlyused numerical simulations to study theaccretion onto a primordial protostar.The calculations show that a PopulationIII star grows to roughly 50 solar mass-es within the first 10,000 years after theinitial core forms. Although we could notfollow the accretion further because ofnumerical limitations, it is likely that thestar continues to grow, perhaps to 100 to200 solar masses. It seems safe to con-clude that the first stars were typicallymany times more massive and luminousthan the sun.

The Cosmic RenaissanceWHAT EFFECTS did these first starshave on the rest of the universe? An im-portant property of stars with no metalsis that they have higher surface temper-atures than stars with compositions likethat of the sun. The production of nu-clear energy at the center of a star is lessefficient without metals, and the starwould have to be hotter and more com-pact to produce enough energy to coun-teract gravity. Because of the more com-pact structure, the surface layers of thestar would also be hotter. In collabora-tion with Loeb and Rolf-Peter Kudritzkiof the University of Hawaii Institute forAstronomy, Bromm devised theoreticalmodels of such stars with masses between100 and 1,000 solar masses. The modelsshowed that the stars had surface tem-peratures of 100,000 kelvins—about 17times higher than the sun’s surface tem-perature. Thus, the first starlight in theuniverse would have been mainly ultra-violet radiation from very hot stars, andit would have begun to heat and ionizethe neutral hydrogen and helium gasaround these stars soon after they formed.

We call this event the cosmic renais-sance. Although astronomers cannot yet


RICHARD B. LARSON and VOLKER BROMM have worked together to understand the processesthat ended the “cosmic dark ages” and brought about the birth of the first stars. Larson, a pro-fessor of astronomy at Yale University, joined the faculty there in 1968 after receiving his Ph.D.from the California Institute of Technology. His research interests include the theory of star for-mation as well as the evolution of galaxies. Bromm earned his Ph.D. at Yale in 2000 and is nowan assistant professor of astronomy at the University of Texas at Austin, where he focuseson the emergence of cosmic structure. The authors acknowledge the many contributions ofPaolo Coppi, professor of astronomy at Yale, to their joint work on the formation of the first stars.

THE

AU

THO

RS



estimate how much of the gas in the uni-verse condensed into the first stars, evenas little as one part in 100,000 could havebeen enough for these stars to ionizemuch of the remaining gas. Once the firststars started shining, a growing bubble ofionized gas would have formed aroundeach one. As more and more stars formedover hundreds of millions of years, thebubbles of ionized gas would havemerged, and the intergalactic gas wouldhave become completely ionized.

Scientists from the California Insti-tute of Technology and the Sloan Digi-tal Sky Survey have found evidence forthe final stages of this ionization process.They observed strong absorption of ul-traviolet light in the spectra of quasarsthat date from about 900 million yearsafter the big bang. The results suggestthat the last patches of neutral hydrogengas were being ionized at that time. Adifferent probe has recently providedclues to the earliest stages of reioniza-tion, already occurring only 200 million

years after the big bang. In an importantbreakthrough, NASA’s Wilkinson Mi-crowave Anisotropy Probe (WMAP) hasmeasured the fundamental properties ofthe universe with high precision. Theseinclude the age of the universe—precise-ly 13.7 billion years—the proportions ofdark and luminous matter, and dark en-ergy in the cosmos. The biggest surprise:scrutinizing the subtle patterns that wereimprinted into the photons of the cosmicmicrowave background, WMAP has in-dicated that ultraviolet radiation fromthe first stars ionized atomic hydrogenand helium, providing an abundance offree electrons early in cosmic history. Mi-crowave background photons were po-larized as they interacted with these elec-trons. An early generation of massivePopulation III stars seems to be requiredto account for the surprising strength ofthe polarization patterns.

Helium requires more energy to ion-ize than hydrogen does, but if the firststars were as massive as predicted, they

would have ionized helium at the sametime. On the other hand, if the first starswere not quite so massive, the heliummust have been ionized later by energeticradiation from sources such as quasars.Future observations of distant objectsmay help determine when the universe’shelium was ionized.

If the first stars were indeed very mas-sive, they would also have had relativelyshort lifetimes—only a few million years.Some of the stars would have explodedas supernovae, expelling the metals theyproduced. Stars that are between 100and 250 times as massive as the sun arepredicted to blow up completely in ener-getic explosions, and some of the firststars most likely had masses in this range.Because metals are much more effectivethan hydrogen in cooling star-formingclouds and allowing them to collapseinto stars, the production and dispersalof even a small amount could have hada major effect on star formation.

Working in collaboration with An-


Computer simulations have given scientists some indication of the possible masses, sizes and other characteristicsof the earliest stars. The lists below compare the best estimates for the first stars with those for the sun.

SUNMASS: 1.989 × 1030 kilogramsRADIUS: 696,000 kilometersLUMINOSITY: 3.85 × 1023 kilowattsSURFACE TEMPERATURE: 5,780 kelvinsLIFETIME: 10 billion years

FIRST STARSMASS: 100 to 1,000 solar massesRADIUS: 4 to 14 solar radiiLUMINOSITY: 1 million to 30 million solar unitsSURFACE TEMPERATURE: 100,000 to 110,000 kelvinsLIFETIME: 3 million years

COMPARING CHARACTERISTICS

STAR STATS


drea Ferrara of the Astrophysical Obser-vatory of Arcetri in Italy, we have foundthat when the abundance of metals instar-forming clouds rises above onethousandth of the metal abundance inthe sun, the metals rapidly cool the gas tothe temperature of the cosmic back-ground radiation. (This temperature de-clines as the universe expands, falling to19 kelvins one billion years after the bigbang and to 2.7 kelvins today.) This ef-ficient cooling allows the formation ofstars with smaller masses and may alsoconsiderably boost the rate at which starsare born. It is possible that the pace ofstar formation did not accelerate until af-

ter the first metals had been produced. Inthis case, the second-generation starsmight have been the ones primarily re-sponsible for lighting up the universe andbringing about the cosmic renaissance.

At the start of this active period of starbirth, the cosmic background temperaturewould have been higher than in present-day molecular clouds (10 kelvins). Untilthe temperature dropped to that level—which happened about two billion yearsafter the big bang—the process of starformation may still have favored massivestars. As a result, many such stars mayhave formed during the early stages ofgalaxy building by successive mergers ofprotogalaxies. A similar phenomenonmay occur in the modern universe whentwo galaxies collide and trigger a star-burst—a sudden increase in the rate ofstar formation—producing relativelylarge numbers of massive stars.

Puzzling EvidenceTHIS HYPOTHESIS about early starformation might help explain some puz-zling features of the present universe. Oneunsolved problem is that galaxies containfewer metal-poor stars than would be ex-pected if metals were produced at a rateproportional to the star formation rate.This discrepancy might be resolved if ear-ly star formation had produced relatively

more massive stars; on dying, these starswould have dispersed large amounts ofmetals, which would have then been in-corporated into most of the low-massstars that we now see.

Another puzzling feature is the highmetal abundance of the hot x-ray-emit-ting intergalactic gas in clusters of galax-ies. This observation could be accountedfor most easily if there had been an earlyperiod of rapid formation of massivestars and a correspondingly high super-nova rate that chemically enriched the in-tergalactic gas. This case also dovetailswith the recent evidence suggesting thatmost of the ordinary matter and metals

in the universe lies in the diffuse inter-galactic medium rather than in galaxies.To produce such a distribution of matter,galaxy formation must have been a spec-tacular process, involving intense burstsof massive star formation and barragesof supernovae that expelled most of thegas and metals out of the galaxies.

Stars that are more than 250 timesmore massive than the sun do not ex-plode at the end of their lives; insteadthey collapse into massive black holes.Several of the simulations mentionedabove predict that some of the first starswould have had masses this great. Be-cause the first stars formed in the densestparts of the universe, any black holes re-sulting from their collapse would havebecome incorporated, via successivemergers, into systems of larger and larg-er size. It is possible that some of theseblack holes became concentrated in theinner part of large galaxies and seeded thegrowth of the supermassive black holes

that are now found in galactic nuclei.Furthermore, astronomers believe

that the energy source for quasars is thegas whirling into the black holes at thecenters of large galaxies. If smaller blackholes had formed at the centers of someof the first protogalaxies, the accretion ofmatter into the holes might have gener-ated “mini quasars.” Because these ob-jects could have appeared soon after thefirst stars, they might have provided anadditional source of light and ionizing ra-diation at early times.

Thus, a coherent picture of the uni-verse’s early history is emerging, althoughcertain parts remain speculative. The for-

mation of the first stars and protogalax-ies began a process of cosmic evolution.Much evidence suggests that the periodof most intense star formation, galaxybuilding and quasar activity occurred afew billion years after the big bang andthat all these phenomena have continuedat declining rates as the universe hasaged. Most of the cosmic structure build-ing has now shifted to larger scales asgalaxies assemble into clusters.

In the coming years, researchers hopeto learn more about the early stages of thestory, when structures started developingon the smallest scales. Because the firststars were most likely very massive andbright, instruments such as the JamesWebb Space Telescope—the planned suc-cessor to the Hubble Space Telescope—

might detect some of these ancient bod-ies. Then astronomers may be able to ob-serve directly how a dark, featurelessuniverse formed the brilliant panoply ofobjects that now give us light and life.


The formation of the first stars and protogalaxiesBEGAN A PROCESS OF COSMIC EVOLUTION.

Before the Beginning: Our Universe and Others. Martin J. Rees. Perseus Books, 1998.

The First Sources of Light. Volker Bromm in Publications of the Astronomical Society of the Pacific,Vol. 116, pages 103–114; February 2004. Available at www.arxiv.org/abs/astro-ph/0311292

The First Stars. Volker Bromm and Richard B. Larson in Annual Reviews of Astronomy and Astrophysics,Vol. 42, pages 79–118; September 2004. Available at www.arxiv.org/abs/astro-ph/0311019

Graphics from computer simulations of the formation of the first luminous objects can be found athttp://cfa-www.harvard.edu/~vbromm/

M O R E T O E X P L O R E



FountainsFountains


Go out on a winter’s night in the North-ern Hemisphere and look due southaround midnight. You will see the con-stellation of Orion the Hunter, probablythe best-known group of stars after theBig Dipper. Just below Orion’s Belt,which is clearly marked by three promi-nent stars in a line, is the Sword of Ori-on, and in the center of the sword is afaint fuzzy patch. This region, the OrionNebula, is a giant stellar nursery em-bracing thousands of newborn stars.

Orion is a convenient place to studythe birth of stars because it is relative-ly close by—a mere 1,500 light-yearsaway—and has a good mix of low- andhigh-mass stars. It also contains a vastquantity of gas and dust in the form ofa so-called molecular cloud. Such cloudsare known to provide the raw material

for new stars. What is now happening inOrion probably replicates what tookplace in our part of the galaxy five bil-lion years ago, when the sun and itsplanets first came into being.

Understanding how stars and plan-ets form is one of astronomy’s quintes-sential subjects yet, until recently, one ofthe most poorly understood. Twentyyears ago astronomers knew more aboutthe first three minutes of the universethan they did about the first three billiondays of our solar system. Only in thepast decade have they started to get an-swers. Infant stars, it turns out, look likescaled-down versions of the heart of aquasar, with powerful jets of materialflung outward by sweeping magneticfields. These stellar fountains of youthnot only make for spectacular pictures

but also help to resolve paradoxes thathave long dogged astronomers.

The Journeywork of the StarsTHE THEORY OF HOW STARS andplanets form has a venerable history.Just over 200 years ago French mathe-matician Pierre-Simon Laplace put for-ward the idea that the solar system wascreated from a spinning cloud of gas. Heproposed that gravity pulled most of thegas to the center, thereby creating thesun. At the same time, some of the ma-terial, because of its spin, could not beabsorbed by the young sun and insteadsettled into a disk. Eventually these dregsbecame the planets. According to mod-ern numerical simulations of the pro-cess, once the spinning cloud starts tocollapse, it proceeds quickly to the for-mation of one or more stars, a proto-planetary disk, and a leftover envelopeof gas (individual atoms and molecules)and dust (very large clumps of atoms).

Laplace’s model was not universallyaccepted. The uncertainty was mainlyobservational: testing the model waswell beyond the astronomical capabili-ties of, say, 30 years ago, for two rea-sons. First, the leftover cloud of gas anddust blocks our view of the very regionthat must be studied. Second, proto-planetary disks subtend minute angleson the sky: if the distance between thesun and Pluto (six billion kilometers) is

w w w . s c i a m . c o m U p d a t e d f r o m t h e A u g u s t 2 0 0 0 i s s u e 13

C.

RO

BE

RT

O’D

ELL

, M

ARK

MC

CAU

GH

RE

AN A

ND

JO

HN

BAL

LY H

ub

ble

Sp

ace

Tel

esco

pe

AND

NAS

A (d

isks

);

GAR

YB

ER

NST

EIN

Un

iver

sity

of

Pen

nsy

lva

nia

, M

EG

AN C

. N

OVI

CK

I U

niv

ersi

ty o

f M

ich

iga

n A

ND

LU

CE

NT

TEC

HN

OLO

GIE

S (n

ebu

la)

STELLAR BIRTHING GROUND in the Orion Nebula (opposite page) has given rise to hundreds of newstars. Surrounding it is an invisible but immense molecular cloud—a million suns’ worth of gas anddust in a volume 300 light-years across. Young stars in Orion are swaddled in disks of material aboutthe size of our solar system (above); around some, planets may even now be forming.

To make a star, gas and dust must fall inward. So why do astronomers see stuff streaming outward?

of YouthEarly Days in the Life of a Star

By Thomas P. Ray



representative of the scale of the disks,conventional ground-based telescopescan resolve them to a distance of only200 light-years, less than halfway to thenearest star-forming region. Simplybuilding bigger telescopes without spe-cial instrumentation does not help, be-cause the blurring of detail occurs in theatmosphere.

Theoretical problems also stymiedastronomers. Sunlike stars at the youth-ful age of 100,000 years rotate onceevery few days and are four or five timesbigger than the mature sun. As such starscontract, they should spin faster, just likeice skaters pulling in their arms. Yet thesun has evidently slowed down, current-ly taking a month to rotate once. Some-thing must have drained away its angu-lar momentum. But what?

Another puzzle is how molecularclouds survive so long. Gravity is tryingto force them to collapse, and withoutsupport they should implode withinabout a million years. But clouds seemto have endured for a few tens of mil-lions of years. What holds them up?Thermal pressure is woefully inadequatebecause the clouds are far too cold, just10 or 20 kelvins. Turbulence might dothe trick, but what would generate it? Ingiant molecular clouds such as Orion,winds and shock waves produced byembedded massive stars would stir thingsup, but many smaller, sedate clouds haveno massive stars.

The first observational obstacle yield-

ed in the late 1970s, when astronomersbegan to observe star-forming regions atwavelengths that penetrate the dustshroud. Studying regions such as theOrion molecular cloud at millimeterwavelengths—a previously unexploredpart of the spectrum—astronomers iden-tified dense, cold clumps typically mea-suring a light-year across. Such clumps,known as molecular cores, contain asmuch as a few suns’ worth of gas andquickly became identified with Laplace’sspinning clouds.

As is often the case in astronomy,new mysteries immediately emerged. Al-though a few of the molecular cores seemto be in the process of collapsing, most ofthem seemed stabilized by means that arenot entirely understood. What triggerstheir eventual collapse is equally uncer-tain, but it may involve some outsidepush from, for example, a nearby super-nova explosion. Or turbulence may sim-ply die away, letting gravity take over.The biggest conundrum concerns the di-rection in which material is moving. Ac-cording to Laplace’s hypothesis, starsarise from gravitational accretion, so as-tronomers expected to see signs of gasplummeting toward the cores.

To their astonishment, they discov-ered that gas, in the form of molecules, isactually moving outward. Usually twogiant lobes of molecular gas were foundlying on either side of a young star. Theselobes, typically a few light-years in length,have masses similar to or even largerthan that of the young star itself, andthey move apart at speeds of tens ofkilometers per second.

Jetting from the CribTHE MOLECULAR LOBES bear astrange resemblance to the vastly largerlobes of hot plasma seen near active gal-axies such as quasars. Astronomers hadknown for years that jets produce theselobes. Squirting outward at velocitiesclose to the speed of light, jets from ac-tive galaxies can stretch for millions oflight-years. Might a miniature version ofthese jets also drive the molecular lobesin star-forming regions?

This idea harked back to a discoveryin the early 1950s by astronomersGeorge H. Herbig and Guillermo Haro.Herbig, then working at Lick Observa-tory in northern California, and Haro, atTonantzintla Observatory in Mexico, in-dependently found some faint fuzzy


FROM DUST TO A STAR

A star begins to coalesce when a disturbance, suchas a nearby supernova explosion, causes a cloudof gas and dust to collapse.

Gas and dust clump at the center, surrounded byan envelope of material and a swirling disk.Magnetic forces direct jets along the axis.

Material continues to rain onto the disk.Roughly a tenth of it streams out in an unevenflow, shoving aside ambient gas.

THOMAS P. RAY is astronomy professor at the Dublin Institute for Advanced Studies, hav-ing also worked at the University of Sussex in England and the Max Planck Institute for As-tronomy in Heidelberg, Germany. Ray has been the principal or co-investigator on nu-merous Hubble observations of jets from young stars and is helping design the JamesWebb Space Telescope, Hubble’s successor. His other interests include quasars, comets,archaeoastronomy (the study of sites such as Stonehenge), sailing and Guinness.TH

E A

UTH

OR


patches in Orion. Now known as Her-big-Haro objects, these small cloudswere initially thought to be specific sitesof star formation. (Some popular as-tronomy books still repeat this error.) In1975, however, Richard D. Schwartz,then at the University of California atSanta Cruz, realized that the spectrum ofa Herbig-Haro object closely resemblesthat of the material left over from a su-pernova. From the Doppler shifting ofthe spectral lines, he found that Herbig-Haro objects are moving at speeds up toa few hundred kilometers per second.

That is slower than the motion of atypical supernova remnant, but Schwartzreckoned that the principles are thesame—namely, that the Herbig-Haro ob-jects are heated gas flowing away from astar. The heat, as in supernova remnants,comes from the motion of the gas itself;shock waves convert some of the kineticenergy of motion into thermal energyand then into radiation. Schwartz’s ideagained further support when astronomerslooked at photographs of Herbig-Haroobjects taken a number of years apart.They were indeed moving. By extrapo-lating backward in time, astronomers de-duced their source. Invariably it was a staronly a few hundred thousand years old.

Verification of this connection camewith another technological revolution:the charge-coupled device (CCD), thelight-sensitive chip found in camcordersand digital cameras. For astronomers,CCDs offer greater sensitivity and con-

trast than the traditional photographicplates. In 1983 Reinhard Mundt andJosef Fried of the Max Planck Institutefor Astronomy in Heidelberg, Germany,made the first CCD observations of stel-lar jets. Subsequent work by Mundt, BoReipurth of the European Southern Ob-servatory in Santiago, Chile, and others(including me) showed that jets fromyoung stars stretch for several light-years. These jets are closely related toHerbig-Haro objects. In fact, some suchobjects turned out to be nothing morethan the brightest parts of jets. Otherswere discovered to be bow shockscaused by jets as they plow their way su-personically through ambient gas, likethe shock wave that surrounds a bulletzinging through the air. The jets typical-ly have a temperature of about10,000 kelvins and contain 100atoms per cubic centimeter—

denser than their surroundingsbut still thinner by a factor of10,000 than the best vacuum inlabs on Earth. Near the star thejets are narrow, opening withan angle of a few degrees, butfarther from the star they fanout, reaching a diameter widerthan Pluto’s orbit.

Out of the WayH O W A R E T H E J E T S andHerbig-Haro objects, whichare mostly made up of atomsand ions, related to the molecu-

lar flows? The leading explanation isthat a molecular lobe consists of ambi-ent gas that got in the way of the jet andwas accelerated.

None of these observations reach theheart of the matter, however: the diskaround the nascent star. Astronomershad long been gathering circumstantialevidence for disks. In the early 1980s theInfrared Astronomical Satellite discov-ered that many new stars had excess in-frared radiation over and above whatshould be produced by the star alone.Warm dust in a disk seemed the mostlikely source. Around the same time, mil-limeter-wave telescopes began to mea-sure the mass of gas and dust aroundthese stars, typically finding 0.01 to 0.1solar mass—just the right amount of ma-terial needed to form planetary systems.In the mid-1980s Edward B. Churchwellof the University of Wisconsin–Madisonand his colleagues observed the OrionNebula at radio wavelengths. Theyfound sources comparable in size to ourown solar system and suggested that thesources were clouds of hot gas that hadevaporated from a disk.

Sighting the disks themselves, how-ever, ran up against the second observa-tional obstacle: their comparativelysmall size. For that, astronomers had toawait the clarity afforded by the HubbleSpace Telescope and by ground-basedinstruments equipped with adaptive op-tics. In 1993 C. Robert O’Dell of RiceUniversity and his collaborators ob-


ALFR

ED

T.

KAM

AJIA

N (

illu

stra

tion

s);

CH

RIS

TOP

HE

R J

. B

UR

RO

WS

Spa

ce T

eles

cop

e Sc

ien

ce I

nst

itu

te A

ND

NAS

A (p

hot

ogra

ph

)

TIME-LAPSE IMAGES of a hatchling star, Herbig-Haro 30,taken a year apart show pockets of gas moving away fromthe center. These jets are clearly perpendicular to the darkdisk that hides the star.

Disk material agglomerates into planetesimals.The envelope and the jets dissipate. By thispoint, one million years have passed.

The high pressure and temperature at the center of the star trigger nuclear fusion. The planetesimalshave assembled into planets.

0.01 light-year



served Orion with Hubble and finallysaw the disks that Laplace had predict-ed [see illustrations on page 13]. Theirmaterial, where it was buffeted by the in-tense radiation and winds from nearbymassive stars, was seen to be evaporat-ing. O’Dell christened these disks “pro-plyds” for protoplanetary disks. Thename may actually be a misnomer be-

cause some disks will evaporate withina million years, probably before planetscan form. But similar disks in milder en-vironments should indeed survive longenough to give birth to planets.

With the discovery of all the basiccomponents of a modern version ofLaplace’s theory—spinning clouds, out-flows, disks—astronomers could begin to

study the relationships among them. Mycolleagues and I, along with anothergroup led by Christopher J. Burrows,then at the Space Telescope Science In-stitute, turned the Hubble telescope onHerbig-Haro 30, which consists of a pairof oppositely directed jets. To our sur-prise, the images revealed two smallcusp-shaped nebulae where the source ofthe jets should have been. Cutting acrossthe nebulae was a dark band. It soon be-came clear that we were looking at a diskperpendicular to the jets. As seen fromour edge-on view, the disk obscures thecentral star. The nebulae are dust cloudsilluminated by starlight. Jets stream out-ward, culminating in the Herbig-Haroobjects. The jigsaw puzzle of star forma-tion was coming together.

In active galaxies, disks are crucial tothe formation of jets. But how does thisprocess work for an embryonic star? Anintriguing coincidence has provided acrucial clue. All the jets and flows locat-ed near Herbig-Haro 30, with one oddexception, have roughly the same orien-tation. In fact, they are aligned with themagnetic field of the parent cloud. Thisseems to support ingenious suggestions—

made by Ralph E. Pudritz and Colin A.Norman, both then at the University ofCambridge, and by Frank H. Shu of theUniversity of California at Berkeley—forhow magnetic fields could drive an out-flow from a young star.

Astronomy abounds with examplesof magnetic fields guiding ionized gas.For example, auroras are caused bycharged particles that stream downEarth’s magnetic field lines and hit the up-per atmosphere. In the same way, ionizedparticles from a circumstellar disk couldattach themselves to the field lines of ei-ther the disk or the star. Because the diskis spinning, the particles would experi-ence a centrifugal force and would thusbe flung out along the field lines. Morematter would flow in to replace whatwas lost, and so the process would con-tinue. Although most of the matter wouldend up being accreted by the star, some10 percent might be ejected. In comput-er simulations the process may proceed infits and starts, which would account forthe knotty structures seen in many jets.


Observing the star-forming region NGC 2264 at millimeter wavelengths, astronomers see twolobes of molecular gas moving at tens of kilometers per second. Red indicates the fastestvelocities, violet the slowest.

Complex jet patterns, as evident in Herbig-Haro 47, can arise because of variations in theoutflow rate and the gravitational effect of companion stars.

In the core of the active galaxy Messier 87, the driving force is thought to be a black hole abillion times more massive than the sun.

Mysterious though their detailed mechanisms may be, jets always involve the samebasic physical process: a balance of power between gravity, magnetic fields andangular momentum. Gravity tries to pull matter toward the center of mass, butbecause of centrifugal forces, the best it can do is gather material into a swirlingdisk. Narrow streams of gas shoot out along magnetic field lines, the direction inwhich matter can most easily move. The escaping matter carries away angularmomentum, thereby allowing less footloose matter to settle inward.

JET ACTION

On a much smaller scale, a newborn star whips up and sprays out a current of gas known asHerbig-Haro 34. The jet may push ambient molecular gas outward.

T H O M A S P . R A Y

W I L L I A M B . S P A R K S S p a c e T e l e s c o p e S c i e n c e I n s t i t u t e

M I C H E L F I C H U n i v e r s i t y o f W a t e r l o o A N D G E R A L D M O R I A R T Y - S C H I E V E N J o i n t A s t r o n o m y C e n t e r , H a w a i i

J O N A . M O R S E S p a c e T e l e s c o p e S c i e n c e I n s t i t u t e A N D N A S A

1 light-year

0.01 light-year

1,000 light-years

0.01 light-year


Nebulous No MoreTHE REALIZATION that jets are inte-gral to star formation may solve severalof the theoretical puzzles. As particlestravel outward, they carry angular mo-mentum away from their source—whichwould partially explain why maturestars such as the sun rotate so slowly.Jets may also churn up the surroundingcloud, supplying the necessary turbulentsupport to slow down its collapse.

At the same time, many questions re-main. For example, only about 50 per-cent of evolved young stars are found tohave disks. The others may have haddisks in the past, but these disks couldhave already coalesced into planets. Ob-servers, however, have been unable toconfirm this. Another problem in starformation is the distribution of stellarmasses. Why is the ratio of high- to low-mass stars pretty much the same irre-spective of location in the galaxy? Thisratio seems to be a fundamental proper-ty of the way molecular clouds frag-ment, but for unknown reasons. Re-searchers know little about the early lifeof high-mass stars—partly because theyare rarer, partly because they evolvefaster and are difficult to catch in the actof forming. We do know, however, thatsome high-mass young stars are sur-rounded by disks and produce jets.

With these caveats, astronomers cannow sketch out nature’s recipe for stars.They form in interstellar clouds thatconsist largely of the ashes of earlier gen-erations of stars. The dust was manu-factured in the cool winds and outer at-mospheres of stars as they approachedthe ends of their lives. The clouds arealso laced with heavy elements such asiron and oxygen that were forged deepin the nuclear furnaces of bygone stars.

Magnetic fields or turbulent motionshold up the clouds, but eventually theycollapse under their own weight, per-haps because the magnetic fields leakaway, the turbulence dissipates or a su-pernova goes off nearby. As the materi-al falls in, the clouds fragment intocloudlets, each of which settles into aprimitive star system. In massive molec-ular cores, such as those that gave rise tothe cluster in the Orion Nebula, thesesystems are spaced every few light-weeks(as opposed to light-years) apart. Moststars in the galaxy, including the sun,probably formed in such clusters.

Jets carry away angular momentumand allow the accretion to continue. Oursun must once have had narrow jets thatstretched for several light-years. Whatturned them off is not certain. The storeof infalling material may simply have runout. Some of it may have been drivenaway by the outflows; if so, the jets mayhave served to limit the sun’s final mass.Around the same time, large dust grainswere beginning to stick together to formplanetesimals, the building blocks of theplanets. The planetesimals swept up anyremaining gas, further choking off thejets. The outflows from the sun and itsstellar contemporaries blew away theleftover gas and dust that threaded thespace between them. This weakened thegravitational glue that bound them to-gether, and over a few million years thestars dispersed. Today the nearest star tothe sun is about four light-years away.

Two centuries after Laplace put for-ward his nebular hypothesis, the piecesare beginning to fall into place. Studiesof young stars suggest not only thatplanet formation is going on today butthat planets are very common through-out our own and other galaxies.


ALFR

ED

T.

KAM

AJIA

N

MAGNETIC, PERIPATETIC

The generation of jets may begin whenmaterial—a mixture of ions, atoms,molecules and dust—rains onto thecircumstellar disk along magnetic field lines.

As the disk contracts under gravity, the lines(which are frozen into the material) arepulled in, taking on an hourglass shape.

When the field lines are bent to an angle of30 degrees from the perpendicular,centrifugal force overcomes gravity andflings material outward along the lines.

The inertia of the swirling material twists thefield lines into a helix, which helps tochannel the outward-flowing material in avertical direction.

Young Stars and Their Surroundings. C. Robert O’Dell and Steven V. W. Beckwith in Science, Vol. 276,No. 5317, pages 1355–1359; May 30, 1997.

Jets: A Star Formation Perspective. Thomas P. Ray in Astrophysical Jets, Open Problems. Edited bySilvano Massaglia and Gianluigi Bodo. Gordon and Breach Scientific Publishers, 1998.

The Origin of Stars and Planetary Systems. Edited by Charles J. Lada and Nikolaos D. Kylafis.Kluwer Academic Publishers, 1999.

Protostars and Planets IV. Edited by Vince Mannings, Alan P. Boss and Sara S. Russell. University ofArizona Press, 2000.

For links to World Wide Web sites, visit astro.caltech.edu/~lah/starformation.html




18 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e O c t o b e r 1 9 9 5 i s s u e

DAV

ID M

ALIN

RO

E/An

glo

-Au

stra

lia

n O

bse

rva

tory

CompanionsYoung Stars

The surprising finding that even the youngest stars commonly exist in sets of two or three has revised thinkingabout the birth of star systems By Alan P. Boss

to

A minor revolution in astronomy occurred on April 6, 1992. It did not

take place at a mountaintop observatory but happened at an unlikely

location—the Callaway Gardens Inn on Georgia’s Pine Mountain (el-

evation: 820 feet). Astronomers had gathered there for an international

meeting on the normally slow-paced research topic of double stars, a

field where discoveries often require decades to allow for many of these

systems to complete their orbits. While azaleas flowered outside in the

spring rain, astronomers inside presented results pointing to the star-

tling conclusion that young stars frequently have stellar companions.

This realization was the product of painstaking observations by many

different people using a host of clever techniques and new devices. That

morning in Georgia, the separate works of these numerous researchers

appeared magically to dovetail.

RHO OPHIUCHUSmolecular cloud harborscolorful reflectionnebulae and numerousstars in the process offormation. Becausethese stellar nurserieslie relatively close toEarth, observations ofthem can provideimportant insights intothe birth of doublestars.



The finding that binary systems are atleast as common for young stars as forolder ones might seem reasonable enough,but to astronomers it came as a shock.Most notions of double star formationhad predicted that stellar companionsare produced or captured well after a starhas formed; hence, the youngest starswould be expected to exist singly in space.Such theories no longer bear weight.There remains, however, at least one ideafor the formation of double stars thatholds up to the recent observations. Itmay be the sole explanation for why bi-nary star systems are so abundant in theuniverse.

The sun, which is a mature star, hasno known stellar companions, eventhough most stars of its age are found ingroups of two or more. In 1984 RichardA. Muller of Lawrence Berkeley Na-tional Laboratory and his colleagues hy-pothesized that the sun is not truly a sin-gle star but that it has a distant com-panion orbiting it with a period of about30 million years. He reasoned that grav-itational forces from this unseen neigh-bor could disturb material circling in theoutermost reaches of the solar system,sending a shower of comets toward theinner planets every time the star neared.Muller suggested that this effect mightexplain periodic mass extinctions:comets generated by the sun’s compan-ion would hit the earth every 30 millionyears or so and—as with the demise ofthe dinosaurs—would have wiped outmuch of life on our planet. Because itsapproach would have sparked suchwidespread destruction, Muller calledthe unseen star “Nemesis.”

Most scientists have not accepted

Muller’s interesting idea. For one, theclosest known stars (the Alpha Centauritriple star system, at a distance of 4.2light-years) are much too far away to bebound to the sun by gravity. In fact,there is no astronomical evidence thatthe sun is anything other than a singlestar whose largest companion (Jupiter)is one thousandth the mass of the sun it-self. But living on a planet in orbitaround a solitary sun gives us a distort-ed view of the cosmos; we tend to thinkthat single stars are the norm and thatdouble stars must be somewhat odd.For stars like the sun, this turns out tobe far from true.

Doubles, Anyone?IN 1990 Antoine Duquennoy andMichel Mayor of the Geneva Obser-vatory completed an exhaustive, de-cade-long survey of nearby binary stars.

They considered every star in the sun’s“G-dwarf” class within 72 light-years,a sample containing 164 primary starsthat are thought to be representative ofthe disk of our galaxy. Duquennoy andMayor found that only about one thirdof these systems could be consideredtrue single stars; two thirds had com-panions more massive than one hun-dredth the mass of the sun, or about 10Jupiters.

Binary star systems have widely vari-able characteristics. Stars of some dou-ble G-dwarf systems may be nearlytouching one another; others can be asfar apart as a third of a light-year. Thosein contact may circle each other in lessthan a day, whereas the most widely sep-


CAR

EY

BAL

LAR

D (

top

); W

OLF

GAN

G B

RAN

DN

ER

Un

iver

sity

of

Wü

rzb

urg

(b

otto

m)

YOUNG BINARIES are at least as ubiquitous as mature double stars. For all orbital periods yetdetermined, young doubles found in star-forming regions (blue) are even more common than solar-type binaries that have been surveyed in the sun’s neighborhood (red).

Ratio

of B

inar

ies

to T

otal

Sta

r Sys

tem

s

Period (days)

0.2

0.1

01 100 10,000 1,000,000 100,000,000

Youngstars

Maturesolar-typestars

■ In our galaxy, two thirds of mature stars have one or more companions. But astronomers have been shocked to find that binary systems are just as common for young stars.

■ The collapse of protostellar gas clouds into binary systems occurs quickly—over a few hundred thousand years during a stellar lifetime lasting several billion years.

■ Whether binary, triple or quadruple systems eventually form depends on the initial shape of the cloud and the precise amount of thermal and rotationalenergy in it. Planets may readily form around these systems, too.

Overview/Twin Suns

GLASS-I proved to be a young double star when imaged by an infrared camera at a wavelength of 0.9 micron.


arated double stars may take tens of mil-lions of years to complete a single orbit.Duquennoy and Mayor showed thattriple and quadruple G-dwarf stars areconsiderably rarer than double stars.They counted 62 distinct doubles, sev-en triples and two quadruple groupings.They further determined that each of thetriple and quadruple sets had a hierar-chical structure, composed of a relative-ly close double orbited by either a moredistant single star (forming a triple sys-tem) or another close double star (form-ing a quadruple system). The separationbetween distant pairs needs to be at leastfive times the gap of the close doubles forthe group to survive for long. Arrange-ments having smaller separations arenamed Trapezium systems, after a youngquadruple system in the Orion nebula.These arrangements are orbitally unsta-ble—they will eventually fly apart. Forinstance, if the three stars of a triple sys-tem come close enough together, theywill tend to eject the star of lowest mass,leaving behind a stable pair.

Double stars thus seem to be the rulerather than the exception. This conclu-sion does not, however, mean that plan-ets must be rare. A planet could travelaround a double star system providedthat it circles either near one of the twostars or far away from both of them.Imagine living on such a world orbitingat a safe distance from a tightly boundbinary, where the two stars complete anorbit every few days. The daytime skywould contain a pair of suns separatedby a small distance. Sunrises and sunsetswould be fascinating to watch as firstone and then the other glowing orbcrossed the horizon. Other strange ce-lestial configurations might also occur.If, for example, the planet orbited in thesame plane as did two stars of equalmass, the two suns periodically wouldappear to merge as they eclipsed eachother, briefly halving the amount ofcombined sunlight reaching the planet.

Stellar NurseriesTHE SUN FORMED about 4.6 billionyears ago and has about five billion yearsremaining of its so-called main-sequencelifetime. After it reaches the end of its


ALFR

ED

T.

KAM

AJIA

N

ALAN P. BOSS began modeling the formation of stellar and planetary systems as a phys-ics graduate student at the University of California, Santa Barbara, where he received a doc-torate in 1979. After two years at the NASA Ames Research Center, he joined the departmentof terrestrial magnetism at the Carnegie Institution of Washington (where, despite its name,terrestrial magnetism has not been studied for decades). Boss chairs the International Astronomical Union committee that maintains the organization’s list of extrasolar plan-ets (available at www.ciw.edu/boss/IAU/div3/wgesp/planets.html).

THE

AU

THO

R

PLANETS in double or triple star systems would be excluded from special regions (blue) within whichthey could not orbit stably. Inside this zone, a planet would eventually be tossed out by gravitationalinteractions. For a double system (top), planets could reside either near each of the stars or far fromthem both. For a triple system (bottom), planets could orbit close to either of the paired stars, in a more extended region around the single member or far from all three.



main sequence, it will expand to becomea red giant that will engulf the innerplanets. This configuration will be some-what akin to one that occurred early inthe sun’s history, when it extended farbeyond its present radius. At that time,before it had contracted to its currentsize, the sun was similar to the T Tauri

class of stars that can be seen in those re-gions of our galaxy where stars are nowforming. During its T Tauri stage, thesun’s radius was about four timesgreater than its present measurement ofsome 700,000 kilometers. And even ear-lier still, the protosun must have ex-tended out to about 1.5 billion kilome-

ters, or 10 times the distance betweenthe earth and the sun (that span, 150million kilometers, is known as an as-tronomical unit, or AU).

Present-day T Tauri stars offer as-tronomers an opportunity to learn whatthe sun was like early in its evolution.The nearest T Tauri stars are in two lo-cations, known as the the Taurus mo-lecular cloud and the Rho Ophiuchusmolecular cloud, both about 460 light-years from Earth. The fact that young


ALAN

P.

BO

SS

a b

c d

BINARY STARS can form from the collapse of a molecular cloud. Computer simulation of a slightlyprolate, magnetically supported cloud (a) shows that the cloud becomes increasingly prolate oncemagnetic support is lost and the collapse begins (b), leading to an intermediate bar-shaped entity(c) that fragments into a binary protostar (d).


stars are always embedded in such dustyconcentrations of gas gives us convinc-ing testimony on the nature of their ori-gin—stars are born from the contractionand collapse of the dense cores of mo-lecular hydrogen clouds.

Because young stars are typically en-shrouded by dust, astronomers usuallyhave difficulty viewing them in visiblelight, no matter how powerful the tele-scope. But these sites can be detectedreadily using infrared wavelengths thatare characteristic of the emission fromheated dust grains surrounding thenearby star. Progress in understandingthe formation of stars has thus been de-pendent to a large extent on the devel-opment of detectors capable of sensinginfrared radiation. At the 1992 meetingin Georgia, the first results were pre-sented from several different infraredsurveys specifically designed to detectcompanions to the T Tauri stars in Tau-rus and Ophiuchus.

Subsequently, Andrea M. Ghez ofthe University of California at Los An-geles and her colleagues Gerry F. Neuge-bauer and Keith Matthews, both at theCalifornia Institute of Technology, useda new indium antimony array camera onthe five-meter Hale telescope to photo-graph the regions around known T Tau-ri stars at the near-infrared wavelengthof 2.2 microns. (Visible light has a wave-length between about 0.4 and 0.7 mi-cron.) Using a so-called speckle imagingtechnique to minimize the noise intro-duced by fluctuations in Earth’s atmo-sphere above the telescope, Ghez andher colleagues found that almost half of the 70 T Tauri stars in their sampleshowed stellar companions. For the lim-ited range of separations considered,about 10 to 400 AU, this study indicat-ed that for the youngest systems, bina-ries are twice as common as for main-sequence stars.

Christoph Leinert of the Max PlanckInstitute for Astronomy in Heidelbergalso conducted a near-infrared speckleimaging survey. He found that 43 of the106 T Tauri stars examined had nearbycompanions, again implying that bina-ries were much more common in thesestars than in G-dwarf stars like our sun.

Hans Zinnecker, now at PotsdamAstrophysics Institute, Wolfgang Brand-ner, now at European Southern Obser-vatory, and Bo Reipurth, now at the Uni-versity of Hawaii Institute for Astrono-my, used a high-resolution digital camerawith the European New TechnologyTelescope to image 160 T Tauri stars atan infrared wavelength of one micron.After analyzing the data, they uncovered28 companions lying from 100 to 1,500AU from the T Tauri stars, about a thirdmore than circle around older, solar-typestars in that distance range.

Michal J. Simon of the Stony Brook

University, N.Y., along with Wen PingChen of the National Central Universi-ty in Taiwan and their colleagues, re-ported a novel way to find young doublestars. When the moon passes over, or oc-cults, a distant star system, careful mon-itoring of the light received can revealthe presence of two or more sources, asfirst one and then another star slips be-hind the sharp edge of the lunar face. Si-mon and Chen’s measurements detectedcompanions much closer to T Tauristars than was possible with infraredimaging. Their work again showed thata large fraction of the entities are bina-ries. Robert D. Mathieu of the Univer-sity of Wisconsin–Madison employed amore traditional means for detectingclose double stars, the same method asthat used by Duquennoy and Mayor.Mathieu used spectroscopic measure-ments of the periodic Doppler shift toshow that some T Tauri stars have com-panions located within 1 AU. Oncemore, closely spaced binaries provedmore common for young T Tauri sys-tems than for solar-type stars.

Search for a TheoryHOW DID ALL THESE stellar com-panions come to be? Why did they formso abundantly and so early in their evo-lution? The wealth of observations ofyoung stars requires that binary starsmust form well before even their pre-main-sequence (T Tauri) phase. More-over, the finding that binaries are socommon demands that the mechanismgenerating them—whatever it is—mustbe very efficient.

In principle, a double star systemcould arise from two stars that pass closeenough together so that one forces theother into a stable orbit. The celestialmechanics of such an event, however, re-quires the intervention of a third objectto remove the excess energy of motion


CAR

EY

BAL

LAR

D DENSITY in star-forming clouds was thought tofollow a power law that concentrates masstightly, but better data now suggest it matchesa bell-shaped Gaussian curve that allowsbinaries to form.

Gaussian

Power law

Dens

ity

Distance from Center of Cloud

EARTHLINGS HAVE A DISTORTED VIEW OF the cosmos. Double stars are not odd;

THEY ARE ALMOST AS COMMON AS SINGLES.



between the two stars and ultimatelyleave them trapped in a gravitationallybound system. Nevertheless, such three-body encounters are too rare to accountfor very many binary stars. Cathy J.Clarke and James E. Pringle of the Uni-versity of Cambridge studied a morelikely way that companion stars mighthave paired up. They investigated thegravitational coupling that occurs be-tween two young stars that still had flat-tened disks of dust and gas surroundingthem. That geometry would be far morecommon than three-body encountersand could, in theory, remove enough en-ergy from the stars’ motions. But in theiranalysis they found that such interac-tions are much more likely to end up rip-ping apart the circumstellar disks thanto result in one star’s neatly orbitingwith the other. So this embellishment

seems to help little in explaining the ex-tensive existence of binary star systems.

Failure of the capture mechanismhas forced most astronomers to thinkabout processes that might form bina-ry stars more directly. Consideration ofthis notion goes back over a century. In1883 Lord Kelvin proposed that doublestars result from “rotational fission.”Based on studies of the stability of bod-ies in rapid rotation, Kelvin suggestedthat as a star contracted, it would spinfaster and faster until it broke up into abinary star.

Astronomers now know that pre-main-sequence stars contract consider-ably as they approach the hydrogen-burning main sequence, but T Tauristars do not rotate fast enough to be-come unstable. Furthermore, Kelvin’sfissioning would act too late to explain

the frequency of binariesamong young stars. RichardH. Durisen of Indiana Univer-sity and his colleagues showedthat fission fails on theoreticalgrounds as well—a reasonablecalculation of this instabilityshows that the ejected matterwould end up as trailing spiralarms of gas rather than as aseparate cohesive star.

In contrast to the century-old fission theory, the more re-cent idea for creating binarystars is called fragmentation.This concept supposes that bi-nary stars are born during aphase when dense molecularclouds collapse under theirown gravity and become pro-tostars. The obscuring gas anddust then clear away, and anewly formed binary star (ofthe T Tauri class) emerges. Incontrast to older theories ofthe birth of binary systems,fragmentation fully agreeswith the latest observations ofyoung stars.

The protostellar collapsethat enables fragmentation oc-curs relatively suddenly in thescale of a several-billion-yearstellar lifetime; the event takes

place in a few hundred thousand years.This violent transformation of a diffusecloud into a compact star thus offers aspecial opportunity for a single object tobreak into several distinct members. As-trophysicists have identified two mecha-nisms that might operate on the transi-tion. Very cold clouds can fragment di-rectly into binary systems, whereaswarmer clouds with substantial rotationcan first settle into thin disks and thenlater break up as they gain more mass orbecome progressively flattened.

Cloudy IdeasA KEY OBJECTION to the fragmenta-tion theory involved the distribution ofmatter in protostellar clouds. It was pre-viously thought that this material wasdistributed according to a so-calledpower law. That is, there would be an


ALAN

P.

BO

SS

COLLAPSE of an initially oblate-shaped, magnetically supported cloud leads to fragmentation into a multipleprotostar system. Here four protostars have formed with separations on the order of a few AU. This system ishighly unstable and will probably decay into a close binary system and two ejected single stars.


extremely high concentration of mater-ial near the center of the cloud and arapid decrease in density with increasingdistance. This objection was apparentlyremoved, however, by high-resolutionradio observations made in 1994 usingsubmillimeter wavelengths. Derek Ward-Thompson, now at the University ofCardiff in Wales, and his colleagues de-termined the distribution of material in-side several precollapse clouds. Theyfound that the density follows a classicGaussian (bell-shaped) distribution

rather than a power law. Hence, matterwould be less tightly concentrated to-ward a central point when the star sys-tem began to form.

Elizabeth A. Myhill, then at the Uni-versity of California at Los Angeles, andI had shown separately that the highdensity at the center of a cloud that fol-lows a power law makes it almost im-possible for a second or third star to co-alesce. It proves much easier for frag-mentation to occur with an initialGaussian distribution.

Astrophysicists can predict whethermultiple fragments will ultimately formby solving the set of equations that gov-ern the flow of gas, dust and radiationin a protostellar cloud. The calculationsare sufficiently complex to require accu-rate software and a powerful computerfor their solution.

I began modeling the collapse ofdense clouds with Gaussian density pro-files in 1986 and found that fragmenta-tion could readily occur provided cer-tain conditions were met. As long as aGaussian cloud has sufficient rotationto give the binary system the angularmomentum it requires and the precol-lapse material is cold enough (less than10 kelvins) to make its thermal energyless than about half its gravitational en-ergy, the cloud will fragment during its

gravitational contraction. The condi-tions appear to be nothing out of the or-dinary for the clouds found in stellarnurseries.

Whether a binary, triple or quadru-ple system eventually forms depends onmany details, including the three-di-mensional shape of the original cloud,how lumpy it is, and the precise amountof thermal and rotational energy avail-able. In general, prolate, or football-shaped, clouds tend to form bars thatfragment into binary systems, whereas

more oblate, or pancake-shaped, cloudsflatten to disks that later fragment intoseveral members.

The collapse is thought to occur intwo separate steps. The first phase gen-erates protostars with a radius on theorder of 1 AU or so. These bodies thenundergo a second collapse to form thefinal protostars that have stellar dimen-sions. Fragmentation is believed to bepossible only during the first collapsephase and appears to be capable of gen-erating most of the entire range of sep-arations observed in young binary stars.The very closest systems appear to bethe result of the orbital decay of multi-ple protostar systems.

Dwarfs and Giants WHAT ABOUT FINDING companionsof even lower mass? Duquennoy and

Mayor produced evidence that as manyas 10 percent of solar-type stars arebound to brown dwarfs—that is, theyhave stellar companions with massesranging from 0.01 to 0.08 times themass of the sun. Brown dwarfs are toosmall to ignite hydrogen the way the sundoes, but they could still be massiveenough to burn deuterium soon afterformation. After that, their radiationwould cease, and they would becomecool and extremely difficult to detect.Brown dwarf stars have been found to

be commonplace in the solar neighbor-hood, though seldom as companions tosunlike stars. The best place to find abrown dwarf is in orbit around anotherbrown dwarf.

The original publication of this arti-cle occurred at a propitious time. Octo-ber 1995 began the era of the discoveryof planets around binary as well as sun-like stars: Mayor and his colleague Didi-er Queloz announced they had found ahalf–Jupiter mass planet orbiting aroundthe solar twin 51 Pegasi.

In the nine years since then, well over100 extrasolar planets have been dis-covered, all of them gas giants similar toJupiter. As the effort continues to ex-pand, the race is on to find the firstEarth-like planet outside our solar sys-tem, and it could indeed be foundaround a binary system.


Formation of Binary Stars. Alan P. Boss in The Realm of Interacting Binary Stars. Edited by J. Sahade, G. E. McCluskey, Jr., and Y. Kondo. Kluwer Academic Publishers, 1993.

Stellar Multiple Systems: Constraints on the Mechanism of Origin. P. Bodenheimer, T. Ruzmaikinaand R. D. Mathieu in Protostars & Planets, Vol. 3. Edited by E. H. Levy and J. I. Lunine. University ofArizona Press, 1993.

A Lunar Occultation and Direct Imaging Survey of Multiplicity in the Ophiuchus and Taurus Star-Forming Regions. M. Simon, A. M. Ghez, C. Leinert, L. Cassar, W. P. Chen, R. R. Howell, R. F. Jameson, K. Matthews, G. Neugebauer and A. Richichi in Astrophysical Journal, Vol. 443, No. 2,Part 1, pages 625–637; April 20, 1995.

Pre-Main-Sequence Binary Stars. R. D. Mathieu in Annual Review of Astronomy and Astrophysics,Vol. 32, pages 465–530; 1995.


FOOTBALL-SHAPED GAS CLOUDS TEND TO FORM binary systems. Pancake-shaped clouds

FRAGMENT INTO SEVERAL STARS.




CR

ED

IT


THE DISCOVERY OF

w w w . s c i a m . c o m U p d a t e d f r o m t h e A p r i l 2 0 0 0 i s s u e 27

PAT

RAW

LIN

GS

(op

pos

ite

pa

ge)

; T.

NAK

AJIM

A C

ali

forn

ia I

nst

itu

te o

f Te

chn

olog

y A

ND

S.

DU

RR

ANC

E J

ohn

s H

opki

ns

Un

iver

sity

(lef

t);

S. K

ULK

ARN

I C

alt

ech

, D

. G

OLI

MO

WSK

I Jo

hn

s H

opki

ns

Un

iver

sity

AN

DN

ASA

(rig

ht)

Abrown dwarf is a failed star. A star shines becauseof the thermonuclear reactions in its core, whichrelease enormous amounts of energy by fusing hy-drogen into helium. For the fusion reactions to besustained, though, the temperature in the star’score must reach at least three million kelvins. And

because core temperature rises with gravitational pressure, thestar must have a minimum mass: about 75 times the mass of theplanet Jupiter, or about 7 percent of the mass of our sun. Abrown dwarf just misses that mark—it is heavier than a gas-gi-ant planet but not quite massive enough to be a star.

For decades, brown dwarfs were the “missing link” of ce-lestial bodies: thought to exist but never observed. In 1963 Uni-versity of Virginia astronomer Shiv Kumar theorized that thesame process of gravitational contraction that creates stars fromvast clouds of gas and dust would also frequently producesmaller objects. These hypothesized bodies were called blackstars or infrared stars before the name “brown dwarf” was sug-gested in 1975 by astrophysicist Jill C. Tarter, now director ofresearch at the SETI Institute in Mountain View, Calif. Thename is a bit misleading; a brown dwarf actually appears red,not brown. But the name “red dwarf” was already taken. (It isused to describe stars with less than half the sun’s mass.)

In the mid-1980s astronomers began an intensive search forbrown dwarfs, but their early efforts were unsuccessful. It wasnot until 1995 that they found the first indisputable evidence oftheir existence. That discovery opened the floodgates; sincethen, researchers have detected hundreds of the objects. Now

observers and theorists are tackling a host of intriguing ques-tions: How many brown dwarfs are there? What is their rangeof masses? Is there a continuum of objects all the way down tothe mass of Jupiter? And did they all originate in the same way?

The search for brown dwarfs was long and difficult becausethey are so faint. All astrophysical objects—including stars,planets and brown dwarfs—emit light during their formationbecause of the energy released by gravitational contraction. Ina star, the glow caused by contraction is eventually supplantedby the thermonuclear radiation from hydrogen fusion; once itbegins, the star’s size and luminosity stay constant, in most cas-es for billions of years. A brown dwarf, however, cannot sus-tain hydrogen fusion, and its light steadily fades as it shrinks[see box on page 31]. The light from brown dwarfs is primari-ly in the near-infrared part of the spectrum. Because browndwarfs are faint from the start and dim with time, some scien-tists speculated that they were an important constituent of“dark matter,” the mysterious invisible mass that greatly out-weighs the luminous mass in the universe.

Astronomers assumed that a good place to look for very faintobjects would be close to known stars. More than half the starsin our galaxy are in binary pairs—two stars orbiting their com-mon center of gravity—and researchers suspected that some starsthat seemed to be alone might actually have a brown dwarf as acompanion. One advantage of such a search is that astronomersdo not have to survey large sections of sky for brown dwarfs—

they can focus their telescopes on small areas near known stars.The strategy looked good early on. A likely candidate ap-

BROWN DWARF GLIESE 229B gives off a red glow in this artist’s conception (oppositepage). The object is believed to be slightly smaller than Jupiter but about 10 timeshotter and 30 to 40 times more massive. It was discovered in 1995 as a companion tothe red dwarf star Gl 229A (shown in background). Astronomers detected the browndwarf in images from the Palomar Observatory’s 1.5-meter telescope (near right) andfrom the Hubble Space Telescope ( far right) that show the object as a faint spot next tothe red dwarf. Gl 229B is actually more than six billion kilometers from its companionstar—farther than Pluto is from our sun.

BROWN DWARFSLess massive than stars but more massive than planets, brown dwarfs werelong assumed to be rare. New sky surveys, however, show that the objectsmay be as common as stars By Gibor Basri



peared in 1988, when Eric Becklin andBenjamin Zuckerman of the University ofCalifornia at Los Angeles reported thediscovery of GD 165B, a faint red com-panion to a white dwarf. White dwarfsare unrelated to brown dwarfs: they arethe corpses of moderately massive starsand are smaller, hotter and much heav-ier than brown dwarfs. GD 165B may in-deed be a brown dwarf, but astronomershave been unable to say for certain be-cause the object’s inferred mass is close tothe 75-Jupiter-mass boundary betweenlow-mass stars and brown dwarfs.

Another advantage of looking forbrown dwarfs as companions to stars isthe brown dwarf itself doesn’t necessar-ily have to be observed. Researchers candetect them with the same method usedto find extrasolar planets: by observingtheir periodic effects on the motions ofthe stars they are circling. Astronomersdetermine the variations in the stars’ ve-locities by measuring the Doppler shiftsin the stars’ spectral lines. It is actuallyeasier to detect brown dwarfs than plan-ets by this technique because of theirgreater mass.

Nevertheless, famed planet hunterGeoffrey W. Marcy of San FranciscoState University and the University ofCalifornia at Berkeley found no browndwarfs in a survey of 70 low-mass starsconducted in the late 1980s. In the mid-

1990s Marcy discovered half a dozen ex-trasolar gas-giant planets in a survey of107 stars similar to our sun but still sawno clear-cut evidence of brown dwarfs.The failure of these efforts gave rise to theterm “brown dwarf desert” because theobjects appeared to be much less com-mon than giant planets or stars.

Only one of the early Doppler-shiftsearches detected a brown dwarf candi-date. In a 1988 survey of 1,000 stars,David W. Latham of the Harvard-Smith-sonian Center for Astrophysics found astellar companion at least 11 times asmassive as Jupiter. The Doppler-shiftmethod, though, provides only a lowerlimit on a companion’s mass, so Lath-am’s object could be a very low mass starinstead of a brown dwarf. This issue willremain unresolved until scientists can de-termine stellar positions more precisely.

Meanwhile other astronomers pur-sued a different strategy that took advan-tage of the fact that brown dwarfs arebrightest when they are young. The bestplace to look for young objects is in starclusters. The stars in a cluster all form atthe same time but have very different life-times. The most massive stars shine foronly a few million years before runningout of hydrogen fuel and leaving themain-sequence phase of their lifetimes,whereas low-mass stars can keep shiningfor billions, even trillions, of years. The

standard method for estimating the age ofa cluster amounts to finding its most mas-sive main-sequence star. The age of thecluster is roughly the lifetime of that star.

Once researchers locate a young clus-ter and determine its age, they need onlylook for the faintest, reddest (and there-fore coolest) objects in the cluster to iden-tify the brown dwarf candidates. Theoryprovides the expected surface tempera-ture and luminosity of objects of variousmasses for a given age, so by measuringthese properties astronomers can esti-mate each candidate’s mass. Severalteams began the search, imaging the ar-eas of sky containing young clusters andpicking out faint red objects.

The research teams made a series ofannouncements of brown dwarf candi-dates in young clusters, including thestar-forming region in the Taurus con-stellation and the bright cluster called thePleiades (better known as the Seven Sis-ters). Unfortunately, closer scrutinyshowed that none of the candidates wasreally a brown dwarf. Some turned outto be red giant stars located thousands oflight-years behind the cluster; becausethese background stars are so distant,they appear faint even though they arequite luminous. Others were low-massstars behind or in front of the cluster.Some of the “discoveries” made it intothe press, but the later retractions were


T. N

AKAJ

IMA

Ca

ltec

h A

ND

S.

DU

RR

ANC

EJo

hn

s H

opki

ns

Un

iver

sity

(le

ft);

SP

ACE

TE

LESC

OP

E S

CIE

NC

E I

NST

ITU

TE (

top

, ce

nte

r);

JOH

N S

TAU

FFE

R H

arv

ard

-Sm

ith

son

ian

Cen

ter

for

Astr

oph

ysi

cs (

bot

tom

, ce

nte

r);

EU

RO

PE

AN S

OU

THE

RN

OB

SER

VATO

RY

(rig

ht)

OBSERVING FAINT OBJECTS such as brown dwarfs requires specialstrategies. One approach is to focus telescopes on areas near known starsand to look for companions; astronomers used this method to find Gl 229B(above left). Another strategy is to concentrate on young star clusters,because brown dwarfs are brightest when they are young. Scientistssearched the 120-million-year-old Pleiades cluster (above center) to findthe brown dwarf PPl 15 (center inset) as well as many others. Last,

astronomers can find “field” brown dwarfs by imaging large sections of skywith instruments that are sensitive to faint, red sources. The discovery ofthe first field brown dwarf, Kelu-1 (above right), was announced in 1997.

KELU-1

PPL 15

GL 229B

FINDING BROWN DWARFS: SEARCH METHODS


not given much play. This led to furtherskepticism among astronomers towardall brown dwarf announcements and re-inforced the widespread view that theobjects were rare.

Looking for LithiumIN 1992 RAFAEL REBOLO, EduardoL. Martín and Antonio Magazzú of theAstrophysics Institute in Spain’s CanaryIslands proposed a clever new method tohelp distinguish low-mass stars frombrown dwarfs. Called the lithium test, itexploits the fact that below a mass ofabout 60 Jupiter masses, a brown dwarfnever achieves the conditions necessaryto sustain lithium fusion in its core. Thisnuclear reaction occurs at a slightly low-er temperature than hydrogen fusiondoes; as a result, stars quickly consumewhatever lithium they originally had.Even the lowest-mass star burns all itslithium in about 100 million years,whereas all but the most massive browndwarfs retain their lithium forever. Thus,the continued presence of lithium is asign that the object has a substellar mass.

The spectral lines produced by lithi-um are fairly strong in cool red objects.The Canary Islands group looked for

these lines in all the coolest objects in thesky that are also bright enough to pro-vide a spectrum of the needed quality.None showed evidence of lithium. In1993 another team—consisting of my-self, Marcy and James R. Graham ofBerkeley—began to apply the lithium testto fainter objects using the newly built10-meter Keck telescope on Mauna Keain Hawaii. We, too, met with failure atfirst, but our luck changed when we fo-cused on the Pleiades cluster.

A group of British astronomers hadjust conducted one of the broadest, deep-est surveys of the cluster. They found sev-eral objects that by all rights should havehad substellar masses. They showed thatthese objects shared the proper motion ofthe cluster across the sky and thus had tobe members of the cluster rather thanbackground stars. We went right to thefaintest one, an object called HHJ 3, ex-pecting to find lithium. It was not present.But Harvard-Smithsonian Center as-tronomer John Stauffer supplied us withanother target. He, too, had been survey-ing the Pleiades for low-mass objects andhad detected an even fainter candidate,dubbed PPl 15 (the 15th good candidatein the Palomar Pleiades survey). At last,

we were successful: for the first time wedetected lithium in an object for which itspresence implied a substellar mass. We re-ported the discovery at the June 1995meeting of the American AstronomicalSociety. Our results indicated that thecluster was about 120 million years old,giving PPl 15 an inferred mass at the up-per end of the brown dwarf range.

In one of the interesting convergencesthat seem to occur regularly in science,other research teams also reported strongevidence of brown dwarfs in 1995. TheCanary Islands group had also been con-ducting a deep survey of the Pleiadescluster and had detected two objects evenfainter than PPl 15: Teide 1 and Calar 3,both named after Spanish observatories.Each had an inferred mass just below 60Jupiter-masses. By the end of the year Ihad teamed up with the Canary Islandsgroup, and we confirmed the expectedpresence of lithium in both objects. Theastronomical community retained someskepticism about these objects for thefirst few months—after all, they stilllooked like stars—until further discover-ies made it clear that now the browndwarfs were for real.

At the same time, a very different


BR

YAN

CH

RIS

TIE

(le

ft);

LAU

RIE

GR

ACE

(ri

gh

t)

3.0

2.5

2.0

1.5

1.0

0.5

0.0667 668 669 670

Lithiumline

Brown dwarf PPl 15

Low-mass star HHJ 3

Wavelength (nanometers)

Rela

tive

Inte

nsity

671 672 673

ANALYZING THE SPECTRA of faint objects can reveal whether they arestars or brown dwarfs. All stars destroy the lithium in their cores; inthis reaction, a proton collides with the isotope lithium 7, which thensplits into two helium atoms (left). In contrast, all but the mostmassive brown dwarfs cannot achieve the core temperature needed

for lithium destruction, so they retain the element forever. Thespectrum of HHJ 3 (right, yellow line), a low-mass star in the Pleiadescluster, shows no sign of lithium. The spectrum of brown dwarf PPl 15(red line), however, has a strong absorption line indicating thepresence of the light metallic element.

Proton

Lithium 7Helium 4

Impact

CONFIRMING THE DISCOVERIES: THE LITHIUM TEST

LITHIUM DESTRUCTION



search bore spectacular fruit. A groupfrom the California Institute of Technol-ogy and Johns Hopkins University hadbeen looking for brown dwarf compan-ions of nearby low-mass stars. The as-tronomers had equipped the Palomar 1.5-meter telescope with an instrument thatblocked most of the light of the primarystar, allowing a faint nearby companionto be more easily seen. In 1993 they ob-served several brown dwarf candidates.They took second images a year later. Be-cause the targets are relatively close to oursolar system, their movements throughthe galaxy are perceptible against thebackground stars. If a candidate is truly acompanion, it will share this motion. Oneof the companions confirmed was 1,000times fainter than its primary, the low-mass star Gliese 229A. Because the pri-mary was already known to be faint, thecompanion’s luminosity had to be wellbelow that of the faintest possible star.The group kept quiet until it obtained aninfrared spectrum of the object.

At a meeting of the CambridgeWorkshop on Cool Stars, Stellar Systemsand the Sun in October 1995, the Cal-tech/Johns Hopkins group announced

the discovery of Gl 229B, the browndwarf companion to Gl 229A. It wasclearly substellar by virtue of its faint-ness, and the clincher was the detectionof methane in its spectrum. Methane iscommon in the atmospheres of the giantplanets, but all stars are too hot to allowit to form. Its strong presence in Gl 229Bguaranteed that this object could not bea star. At the same meeting the CanaryIslands group reported the observationof several new brown dwarf candidatesin the Pleiades cluster, suggesting thatthese objects might be fairly numerous.In addition, a group led by Michel May-or of the Geneva Observatory in Switz-erland announced the discovery of thefirst extrasolar planet, a gas giant circlingthe star 51 Pegasi. In one morning, thefrustrating search for substellar objectscame to a dramatic conclusion.

Most astronomers view Gl 229B as

the first indisputable brown dwarf dis-covered because it is a million timesfainter than the sun and has a surfacetemperature under 1,000 kelvins—far be-low the minimum temperature that eventhe faintest star would generate (around1,800 kelvins). It has reached this statebecause it is a few billion years old. It isprobably 30 to 40 times more massivethan Jupiter. In contrast, PPl 15, Teide 1and Calar 3 in the Pleiades are more mas-sive (from 50 to 70 Jupiter masses) andalso much hotter (with surface tempera-tures between 2,600 and 2,800 kelvins),primarily because they are much younger.

Once the methods for detectingbrown dwarfs had been proved, the dis-coveries came at an increasing pace. Sev-eral groups returned to the Pleiades. TheCanary Islands group, now includingMaria Rosa Zapatero Osorio of the Lab-oratory for Space Astrophysics and The-oretical Physics, near Madrid, discovereda Pleiades brown dwarf only 35 timesmore massive than Jupiter—the lightestbrown dwarf found in the cluster. Moreimportant, the Canary Islands group con-ducted the first useful assessment of thenumber of brown dwarfs in the Pleiadesby counting the most likely candidates ina small surveyed area and then extrapo-lating the tally for the entire cluster. Theresults indicated comparable numbers ofstars and brown dwarfs in the Pleiades. Iftrue in general, this would mean that ourgalaxy alone contains about 100 billionbrown dwarfs. But it also means thatbrown dwarfs are not the dominant con-stituent of the universe’s mass, becausethey are much lighter than stars. Thehope that they would provide an answerto the dark matter mystery has faded.

Other researchers focused on howthe brown dwarfs are distributed bymass. What is the lowest mass a browndwarf can attain? Is there a continuum ofobjects down to the planetary range—be-low 13 Jupiter masses—or is there a gap


LAU

RIE

GR

ACE

GIBOR BASRI is a professor in the astronomy department at the University of California,Berkeley. He received his Ph.D. in astrophysics from the University of Colorado at Boulderin 1979. Basri has studied solar-type stars, low-mass stars and star formation using a va-riety of telescopes, including the Lick and Keck observatories, as well as spaceborne in-struments, including the Hubble Space Telescope. He is keen on promoting an interest inscience among groups currently underrepresented in the field.TH

E A

UTH

OR

LUMINOSITY HISTORY of low-mass stars (yellow lines), brown dwarfs (red lines) and planets (black line)shows that only stars are massive enough to achieve a stable luminosity. The light from brown dwarfs andplanets fades as they age. Data from brown dwarfs (black crosses) indicate how old and heavy they are.

PPl 15

Mass of Object(Jupiter masses)

Teide 1

10–2

1

10–4

10–6

10–8

106 107 108

Age of Object (years)

Frac

tion

of S

un’s

Lum

inos

ity

109 1010

200

125

90

80

70

50

30

15

10

Kelu-1

Gl 229B


between the lightest brown dwarf andthe heaviest planet because they areformed by different mechanisms?

The best place to answer these ques-tions is in newly forming star clusters,where even very low mass brown dwarfsare still bright enough to see. Surveys ofvarious nearby star-forming regions haveturned up a number of objects that seemcool and dim enough to lie near, or evenbelow, the fusion limit, as predicted bymodels. The models are not very reliablefor such objects, however, so there hasbeen some skepticism, and some objects

have turned out to not actually be in thestar-forming regions. Recently, though,work by me, Subhanjoy Mohanty of theHarvard-Smithsonian Center and ourcollaborators has taken a more directtack. We do not rely on evolutionarymodels but deduce the surface gravity ofconfirmed members by studying theirspectral lines (which broaden as gravityand pressure increase). We confirm verylow masses in some cases. Thus, it ap-pears that brown dwarfs are produced inall possible masses between planets andstars [see box on next page].

Continuing searches for browndwarfs around solar-type stars have con-firmed the initial impression that they arefairly rare in this situation. Brown dwarfsappear to be more common, though, ascompanions to lower-mass stars. In 1998Rebolo and his co-workers discoveredone orbiting the young star G196-3. De-spite its youth, this brown dwarf is al-ready quite cool, which means it must belight, perhaps only 20 Jupiter masses.

The first binary system involving twobrown dwarfs was identified by Martínand me. We determined that the Pleiades


BR

YAN

CH

RIS

TIE

1 MILLION YEARSBrown Dwarf and Accretion Disk

10 MILLION YEARSBrown Dwarf and Planet

Radius of planet's orbit: 2 million–500 million kmRadius of brown dwarf: 300,000 kmTemperature: 2,900 K

100 MILLION YEARSGravitational Contraction

Radius of brown dwarf: 100,000 kmTemperature: 2,500 K

1 BILLION YEARSRadiative Cooling

Radius of brown dwarf: 65,000 kmTemperature: 1,200 K

10 BILLION YEARSFading to Oblivion

Radius of brown dwarf: 60,000 kmTemperature: 550 K

Radius of disk: 1 billion–5 billion kmRadius of brown dwarf: 350,000 kmTemperature: 2,900 K

100,000 YEARSInterstellar Molecular CloudRadius: 100 billion kilometersTemperature: 10 kelvins

The early lives of brown dwarfs and stars follow thesame pattern. Both are believed to originate from thegravitational collapse of interstellar clouds of gas anddust. These clouds are composed primarily ofhydrogen and helium, but they also initially containsmall amounts of deuterium and lithium that areremnants of the nuclear reactions that took place a few minutes after the big bang.

As young stars and brown dwarfs contract, theircores grow hotter and denser, and the deuteriumnuclei fuse into helium 3 nuclei. (Deuterium fusioncan occur in brown dwarfs because it requires a lowertemperature—and hence a lower mass—thanhydrogen fusion.) The outpouring of energy fromthese reactions temporarily halts the gravitationalcontraction and causes the objects to brighten. Butafter a few million years the deuterium runs out, andthe contraction resumes. Lithium fusion occurs next in starsand in brown dwarfs more than 60 times as massive as Jupiter.

During the contraction of a brown dwarf, thermal pressurerises in its core and opposes the gravitational forces. All theelectrons are freed from their nuclei by the heat. Because notwo electrons can occupy the same quantum state, when thecore is very dense the low-energy states are filled, and manyelectrons are forced to occupy very high energy states. Thisgenerates a form of pressure that is insensitive to temperature.Objects supported in this manner are called degenerate. Oneconsequence of this process is that all brown dwarfs are roughlythe size of Jupiter—the heavier brown dwarfs are simply denserthan the lighter ones.

In stars the cores do not become degenerate. Instead hydrogenfusion provides the pressure that supports the star against its owngravity. Once fusion begins in earnest, the star stops contractingand achieves a steady size, luminosity and temperature. In high-mass brown dwarfs, hydrogen fusion begins but then sputters out.As degeneracy pressure slows the collapse of brown dwarfs, theirluminosity from gravitational contraction declines. Although verylow mass stars can shine for trillions of years, brown dwarfs fade

steadily toward oblivion. This makes them increasingly difficult tofind as they age. In the very distant future, when all stars haveburned out, brown dwarfs will be the primary repository ofhydrogen in the universe. —G.B.

BROWN DWARF IS BORN from the contraction of a vast cloud of gas and dust.After a million years the object is a glowing ball of gas, possibly surroundedby an accretion disk from which an orbiting planet could later arise. (So farno planets have been detected around brown dwarfs; their existence andpossible orbits are strictly hypothetical.) Over time the brown dwarf shrinksand cools. The radii and surface temperatures shown here are for an objectof 40 Jupiter masses.

The Life Cycle of Brown Dwarfs

Accretion disk and planetaryorbit not drawn to scale.




Fullyconvective

Fullyconvective

Thermo-nuclear

reactions

Metallichydrogen

Jupiter

Browndwarf

Reddwarfstar

Lithium

No lithium

Molecular hydrogenand helium

Name

Type of object

Mass (Jupiter masses)

Radius (kilometers)

Temperature (kelvins)

Age (years)

Hydrogen fusion

Deuterium fusion

Jupiter

Gas-giant planet

1

71,500

100

4.5 billion

No

No

Gliese 229B

Brown dwarf

30–40

65,000

1,000

2–4 billion

No

Yes

Teide 1

Brown dwarf

55

150,000

2,600

120 million

No

Yes

Gliese 229A

Red dwarf star

300

250,000

3,400

2–4 billion

Yes

Yes

Sun

Yellow dwarf star

1,000

696,000

5,800

4.5 billion

Yes

Yes

Planets versus Brown DwarfsIs there a fundamental difference between the

largest planets and the smallest brown dwarfs? Theclassical view is that planets form in a different way thanbrown dwarfs or stars do. Gas-giant planets are thought tobuild up from planetesimals—small rocky or icy bodies—amid adisk of gas and dust surrounding a star. Within a few millionyears these solid cores attract huge envelopes of gas. Thismodel is based on our own solar system and predicts that allplanets should be found in circular orbits around stars and thatgas-giant planets should travel in relatively distant orbits.

These expectations have been shattered by the discoveryof the first extrasolar giant planets. Most of these bodies havebeen found in close orbits, and most travel in eccentric ovals

rather than in circles. Some theorists have even predictedthe existence of lone planets, thrown out of their

stellar systems by orbital interactions with

sibling planets. This makes it very hard for observers todistinguish planets from brown dwarfs on the basis of how orwhere they formed or what their current location and motion is.We can find brown dwarfs by themselves or as orbitalcompanions to stars or even other brown dwarfs. The samemay be true for giant planets.

An alternative approach is gaining adherents: todistinguish between planets and brown dwarfs based onwhether the object has ever managed to produce any nuclearfusion reactions. In this view, the dividing line is set at about13 Jupiter masses. Above that mass, deuterium fusion occursin the object. The fact that brown dwarfs seem to be lesscommon than planets—at least as companions to moremassive stars—suggests that the two types of objects mayform by different mechanisms. A mass-based distinction,however, is much easier to observe. —G.B.

CONTINUUM OF OBJECTS from planets to stars (below) shows that older brown dwarfs,such as Gliese 229B, are fairly similar to gas-giant planets in size and surfacetemperature. Younger brown dwarfs, such as Teide 1, more closely resemble low-massstars, such as Gliese 229A. Brown dwarfs and low-mass stars are fully convective,meaning that they mix their contents (left). Thermonuclear reactions in the stars’ coresdestroy all their lithium, so its presence is a sign that the object may be a brown dwarf.



BR

YAN

CH

RIS

TIE

brown dwarf PPl 15 is really a close pairof brown dwarfs, with an orbital peri-od of six days. Together with Germanastronomer Wolfgang Brandner, we alsoresolved the first nearby binary pair ofvery cool objects with the Hubble SpaceTelescope. Such systems should providethe first real dynamical masses of browndwarfs within a few years.

Several subsequent studies haveshown that the fraction of very cool ob-jects that turn out to be double in spacetelescope images is about 20 percent (thefraction of stellar binaries at any separa-tion is about 50 percent). These observa-tions suggest that the brown dwarf desertis only a lack of brown dwarfs as com-panions to more massive stars. Whenlooking near low-mass objects (eitherstars or brown dwarfs), the likelihood offinding a brown dwarf companion ismuch greater. This variance probably re-sults from the process that gives birth tobinary systems, which is still poorly un-derstood. Apparently this process is lesslikely to produce a system in which theprimary object is more than about 10times the mass of the secondary. Remark-ably, the brown dwarfs are never foundwith separations more than about the sizeof our solar system, even though that isonly the median separation for stars.

Brown Dwarfs EverywhereASTRONOMERS FOUND still morebrown dwarfs using another search tech-nique: looking for them at random loca-tions in the sky. These “field” browndwarfs are easily lost among the myriadstars of our galaxy. To locate such objectsefficiently, one must image large sectionsof sky with great sensitivity to faint redsources. The first field brown dwarf wasannounced by Maria Teresa Ruiz of theUniversity of Chile in 1997. She dubbedit “Kelu-1” from a South American Indi-an word for “red” and noted that it showslithium. At about the same time, theDeep Near-Infrared Survey (DENIS)—

a European project that scans the south-ern hemisphere of the sky—found threesimilar objects. Researchers quickly con-firmed that one contains lithium.

The Two Micron All Sky Survey(2MASS), managed by the University ofMassachusetts at Amherst, has detectedeven more field brown dwarfs. The team,including J. Davy Kirkpatrick of NASA’sIPAC center in Pasadena, Calif., hasfound hundreds of new extremely coolobjects and confirmed lithium in over 50.Most of these objects have surface tem-peratures greater than 1,500 kelvins andso must be younger than about a billionyears. They are relatively bright and eas-ier to observe than older objects.

The hunt for older field brown dwarfswas frustrated until the summer of 1999,when the Sloan Digital Sky Survey (whichuses optical detectors) turned up twobrown dwarfs containing methane intheir atmospheres. The presence of meth-ane indicates a surface temperature below1,300 kelvins and hence an age greaterthan one billion to two billion years. Atthe same time, the 2MASS group report-ed the observation of four similar objects.Most of the brown dwarfs in our galaxyshould be methane-bearing, because themajority formed long ago and shouldhave now cooled to that state. Thus, thesediscoveries are just the tip of the iceberg.Adam Burgasser of Caltech and othershave now been able to collect a largeenough sample of older brown dwarfs togive us a preliminary idea of how theylook as their atmospheres cool to near-planetary temperatures.

Further study of very cool objects hasyielded clues to the composition and evo-lution of brown dwarf atmospheres. Their

optical spectra lack the molecules of tita-nium oxide and vanadium oxide thatdominate the spectra of low-mass stars.These molecules do not appear, becausetheir constituent heavy elements condenseinto hard-to-melt dust grains. The prima-ry optical spectral lines are from moleculesof hydrides, instead of oxides, and fromneutral alkali metals. The dust grains forminto clouds, whose height appears to dropas the objects cool. There is some evidencethe clouds may not always cover thewhole object, so one might be able to study“weather” on brown dwarfs. The dustclouds sink below the visible surface in themethane objects, whose optical spectrathen are dominated by sodium and potas-sium. These spectral lines are broadenedeven more than in high-pressure streetlamps, making the color of the objects ma-genta (not brown!).

The initial discovery phase for browndwarfs is now almost over. Astronomershave good methods for detecting themand many targets for detailed study. In-deed, the 2MASS and DENIS teamshave found that the number of fieldbrown dwarfs in the surveyed areas issimilar to the number of low-mass starsin those areas. Brown dwarfs seem to benearly as common as stars.

Over the next few years, scientists willget a better handle on the basic factsabout brown dwarfs: their numbers,masses and distribution in our galaxy.Researchers will also try to determinehow they form as binary or solo objectsand what processes take place as theiratmospheres cool. It is remarkable thatthese nearby and common objects, asabundant as stars, have only now begunto reveal their secrets.

Brown Dwarfs: A Possible Missing Link between Stars and Planets. S. R. Kulkarni in Science, Vol. 276, pages 1350–1354; May 30, 1997.

Brown Dwarfs and Extrasolar Planets. Edited by R. Rebolo, E. L. Martín and M. R. Zapatero Osorio.Astronomical Society of the Pacific Conference Series, Vol. 134, 1998.

More on brown dwarfs is available at astron.berkeley.edu/~basri/bdwarfs/


We now have an idea of how brown dwarfs look AS THEIR ATMOSPHERES COOL TO

almost planetary temperatures.




TheStellar

Dynamo

In 1801, musing on the vagaries of English weather, astronomer William Herschelobserved that the price of wheat correlated with the disappearance of sunspots. Butthe pattern soon vanished, joining what scientists at large took to be the mytholo-gy connecting earthly events with solar ones. That the sun’s brightness might pos-sibly vary, and thereby affect Earth’s weather, remained speculative.

Thus, in the mid-1980s, when three solar satellites—Solar Maximum Mis-sion, Nimbus 7 and Earth Radiation Budget—reported that the sun’s radiancewas declining, astronomers assumed that all three instruments were failing. Butthe readings then perked up in unison, an occurrence that could not be attributedto chance. The sun was cooling off and heating up; furthermore, the variationwas connected with the number of spots on its face.

In recent years one of us (Baliunas) has observed that other stars undergo rhyth-mic changes much like those of our sun. Such studies are helping refine our un-derstanding of the “dynamo” that drives the sun and other stars. Moreover, theyhave revealed a strong link between “star spots” and luminosity, confirming thepatterns discovered in our sun. And yet astrophysicists, including the three of us,are still debating the significance of the sun’s cycles and the extent to which theymight influence Earth’s climate.

MAGNETIC FIELDS on thesun are rendered visible in

this x-ray photograph by the curving contours of solar flares.

The lines of magnetic fields erupt from the sun’s surface and heat the

gases of the surrounding corona to up to 25 million degrees C, causing them to glow.

Flares are more frequent during sunspot maxima.

Sunspot cycles—on other stars—are helpingastronomers study the sun’s variations and theways they might affect Earth

By Elizabeth Nesme-Ribes, Sallie L. Baliunas and Dmitry Sokoloff

U p d a t e d f r o m t h e A u g u s t 1 9 9 6 i s s u e 35COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

SunspotsTHE EARLIEST KNOWN sunspot rec-ords are Chinese documents that goback 2,000 years, preserving observa-tions made by the naked eye. From 1609to 1611 Johannes Fabricius, ThomasHarriot, Christoph Scheiner and Galileo

Galilei, among others, began telescopicstudies of sunspots. These records, asGerman astronomer Samuel HeinrichSchwabe announced in 1843, displayeda prominent periodicity of roughly 10years in the number of observed sunspotgroups. By the 20th century George

Ellery Hale of the Mount Wilson Ob-servatory in California found those darksurface irregularities to be the seat of in-tense magnetic fields, with strengths ofseveral thousand gauss. (Earth’s magnet-ic field is, on the average, half a gauss.)

Sunspots appear dark because they


1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880

Suns

pot N

umbe

r

Maunder minimum

ELEVEN-YEAR CYCLES of sunspot activity were interrupted between 1645 and 1715 by a period of quiescence. This dearth of sunspots, called the Maunder minimum,coincided with unusually cool temperatures across northern Europe, indicating thatsolar fluctuations influence Earth’s climate. The regular pulsing of the sun’s activity(right) was observed over one cycle at the Paris Observatory. These photographs weretaken in violet light emitted by ionized calcium; the technique that produced them isnow used to study the magnetic activity of other stars.

NAT

ION

AL S

OLA

R O

BSE

RVA

TOR

Y/SA

CR

AME

NTO

PE

AK;

NAT

ION

AL S

OLA

R O

BSE

RVA

TOR

IES/

KIT

T P

EAK

(in

set)

SUNSPOTS are relatively cool regions formed where magnetic fields emerge from the sun, thereby suppressing the upwelling of hot gases from the interior. Elsewhere on the surface, tightly coiled cells ofcyclonically flowing gases show up as granules. Near a sunspot the

magnetic fields organize the gaseous flow into lines resembling iron filings near a bar magnet. The magnetogram (inset) shows field linesemerging at one sunspot (yellow) and reentering at another (blue); such sunspot pairs are common.

1979 1982 1986

MAXIMUM MINIMUM


are 2,000 degrees Celsius cooler than thesurrounding surface of the sun; theywould glow orange-red if seen againstthe night sky. The spots form whenstrong magnetic fields suppress the flowof the surrounding gases, preventingthem from carrying internal heat to thesurface. Next to the sunspots are oftenseen bright areas called plages (after theFrench word for “beach”). The magnet-ic field lines tend to emerge from the sur-face at one spot to reenter the sun at an-other, linking the spots into pairs that re-semble the two poles of a bar magnetthat is oriented roughly east-west.

At the start of each 11-year cycle,sunspots first appear at around 40 de-grees latitude in both hemispheres; theyform closer to the equator as the cycleprogresses. At sunspot minimum, patch-es of intense magnetism, called active re-gions, are seen near the equator. Asidefrom the sunspots, astronomers have ob-served that the geographic poles of thesun have weak overall magnetic fields ofa few gauss. This large-scale field has a“dipole” configuration, resembling thefield of a bar magnet. The leading sun-spot in a pair—the one that first comesinto view as the sun rotates from west toeast—has the same polarity as the pole ofits hemisphere; the trailing sunspot has

the opposite polarity. Moreover, as Haleand Seth B. Nicholson had discovered by1925, the polarity patterns reverse every11 years, so that the total magnetic cycletakes 22 years to complete. But the sun’sbehavior has not always been so regular.In 1667, when the Paris Observatorywas founded, astronomers there begansystematic observations of the sun, log-ging more than 8,000 days of observa-tion over the next 70 years. These rec-ords showed very little sunspot activity.This important finding did not raisemuch interest until the sunspot cycle wasdiscovered, prompting Rudolf Wolf ofZürich Observatory to scrutinize therecords. Although he rediscovered thesunspot lull, Wolf’s finding was criticizedon the grounds that he did not use all theavailable documents.

During the late 1880s, first Gustav F. W. Spörer and then E. Walter Maun-der reported that the 17th-century solaranomaly coincided with a cold spell inEurope. That astonishing observationlay neglected for almost a century, withmany astronomers assuming that theirpredecessors had not been competentenough to count sunspots. It was only in1976 that John A. Eddy of the High Al-titude Observatory in Boulder, Colo., re-opened the debate by examining theParis archives and establishing the va-lidity of what came to be known as theMaunder minimum.

Minze Stuiver, while at Yale Univer-sity, Hans Suess of the University of Cal-ifornia at San Diego and others had dis-covered that the amount of carbon 14 intree rings increased during the dearth ofsunspots. This radioactive element is cre-ated when galactic cosmic rays trans-mute nitrogen in the upper atmosphere.Their findings suggested that when themagnetic fields in the solar wind—theblast of particles and energy that flowsfrom the sun—are strong, they shieldEarth from cosmic rays, so that less car-

bon 14 forms; the presence of excess car-bon 14 indicated a low level of magneticactivity on the sun during the Maunderphase. Eddy thus reinforced the connec-tion between the paucity of sunspots anda lull in solar activity.

Aside from the rarity of sunspotsduring the Maunder minimum, the Parisarchives brought to light another oddi-ty: from 1661 to 1705, the few sunspotsthat astronomers sighted were usually inthe southern hemisphere. They were alsotraveling much more slowly across thesun’s face than present-day sunspots do.Only at the beginning of the 18th cen-tury did the sun assume its modern ap-pearance, having an abundance ofsunspots rather evenly distributed be-tween the two hemispheres.

The Solar DynamoTHE MAGNETIC ACTIVITY of thesun is believed to reside in its convectivezone, the outer 200,000 kilometers wherechurning hot gases bring up energy fromthe interior. The fluid forms furiouswhorls of widely different sizes: the bestknown is an array of convective cells orgranules, each 1,000 kilometers acrossat the surface but lasting only a few min-utes. There are also “supergranules” thatare 30,000 to 50,000 kilometers acrossand even larger flows. Rotation gives riseto Coriolis forces that make the whorlsflow counterclockwise in the northernhemisphere (if one is looking down atthe surface) and clockwise in the south-ern hemisphere; these directions are calledcyclonic.

Whether similar cyclones exist un-derneath the surface is not known. Deepwithin, the convective zone gives way tothe radiative zone, where the energy istransported by radiation. The core of thesun, where hydrogen fuses into heliumto fuel all the sun’s activity, seems to ro-tate rigidly and slowly compared withthe surface.


1920 1940 1960 1980 2000

Suns

pot N

umbe

r

NO

AA;

PE

TER

SAM

EK

Sli

m F

ilm

s

PAR

IS O

BSE

RVA

TOR

Y

1988 1990

SUNSPOTS ARE 2,000 DEGREES C COOLER than the surrounding surface; they would

GLOW ORANGE-RED AGAINST A NIGHT SKY.

MAXIMUM



The first description of how the sun’sgases conspire to create a magnetic fieldwas proposed in 1955 by Eugene N.Parker of the University of Chicago. Be-cause of the high temperature, the atomsof hydrogen and helium lose their elec-trons, thereby giving rise to an electri-cally charged substance, or plasma. Asthe charged particles move, they gener-ate magnetic fields. Recall that the linesdescribing magnetic fields form contin-uous loops, having no beginning orend—their density (how closely togeth-er the lines are packed) indicates the in-tensity of the magnetic field, whereastheir orientation reveals the direction.Because plasma conducts electricity veryefficiently, it tends to trap the field lines:if the lines were to move through theplasma, they would generate a large, andenergetically expensive, electric current.

Thus, the magnetic fields are carried

along with the plasma and end up get-ting twisted. The entwined ropes wraptogether fields of opposite polarity, whichtend to cancel each other. But the sun’srotation generates organizational forcesthat periodically sort out the tangles andcreate an overall magnetic field. This au-tomatic engine, which generates mag-netism from the flow of electricity, is thesolar dynamo.

The dynamo has two essential ingre-dients: the convective cyclones and thesun’s nonuniform rotation. During themid-1800s, Richard C. Carrington, anEnglish amateur astronomer, found thatthe sunspots near the equator rotate fast-er, by 2 percent, than those at midlati-tudes. Because the spots are floating withthe plasma, the finding indicates that thesun’s surface rotates at varying speeds.The rotation period is roughly 25 days atthe equator, 28 days at a latitude of 45degrees and still longer at higher latitudes.This differential rotation should extendall the way through the convective zone.

Now suppose that the initial shape ofthe sun’s field is that of a dipole orientedroughly north-south. The field lines getpulled forward at the equator by thefaster rotation and are deformed in theeast-west direction. Ultimately, they lieparallel to the equator and float to the

surface, erupting as pairs of sunspots.But Coriolis forces tend to align the

cyclones and thereby the sunspots, whichare constrained to follow the plasma’sgyrations. The cyclones arrange the sun-spots so that, for example, a trailing sun-spot in the northern hemisphere lies ata slightly higher latitude than a leadingone. As the equatorial field lines arestretched, they eventually unwind anddrift outward. The trailing sunspotreaches the pole first, effectively revers-ing the magnetic field there. (Recall thatthe trailing spot has a polarity oppositethat of the nearest pole.) Those field linesthat initially extended far beyond thesun reconnect into loops and are blownaway by the solar wind. In this manner,the overall magnetic field flips, and thecycle begins again.

There is, however, a caveat. This sim-ple picture seems to be at odds with re-sults from helioseismology, the science ofsunquakes. The model requires the sunto rotate faster at the interior; in contrast,results from the Global Oscillation Net-work Group, an international collabora-tion of observatories, show that the ro-tation velocity near the equator decreas-es downward. Such experiments areproviding accurate information on inter-nal motions of the sun and thereby help-


PE

TER

SAM

EK

Sli

m F

ilm

s

NorthPole

NorthPole

NorthPole

SOLAR DYNAMO generates the sun’s magnetic field and also causes it to change orientation every 11 years. Suppose that the initial magnetic field (a) resembles that of a bar magnet with its northpole (+) near the sun’s geographic north pole. The magnetic field lines are carried along with theelectrically charged gases. The faster flow at the equator therefore distorts the field lines (b) untilthey wrap tightly (c) around the sun. But the field lines then resist the stretching and unwind,

QUADRUPOLE DIPOLE

a b c


ing Mausumi Dikpati and Peter Gilmanof the High Altitude Observatory andothers to refine dynamo theory.

But what happened during the Maun-der minimum? To explain this lull, twoof us (Nesme-Ribes and Sokoloff) notedthat apart from a dipole pattern, themagnetic field must also have a smallquadrupole component, resembling thefield of two bar magnets placed side byside. If the quadrupole oscillates at a

slightly different rate than the dipole, thesunspots in one hemisphere are producedslightly earlier than those in the otherhemisphere—precisely what we observenow. Furthermore, over the past fourcenturies, a few solar cycles showed dif-ferent numbers of sunspots in the north-ern and southern hemispheres. This pat-tern seems to repeat every century or so,exactly what one would expect if the di-pole “beats” with a weak quadrupole.

But suppose that the quadrupolefield is as strong as the dipole. The equa-torial field lines that result from stretch-ing this combination will then cancel outin one hemisphere yet remain in the oth-er. And the few spots that do appear willall be in one hemisphere, just as 17th-century astronomers noted during theMaunder minimum.

We can encapsulate the relation be-tween the dipole and quadrupole fieldsin a “dynamo number” D. It is the prod-uct of the helicity, or spiraling motion,of the plasma and the local rate ofchange of rotation. When D is verysmall, the magnetic field tends to dieout; as D increases, the quadrupole fieldshows up, with the dipole following. Be-yond a critical value, both componentsof the field are steady. But as D increas-es further, the dynamo becomes period-ic, increasing and decreasing; this is theregime in which the sun now lies. Aweak quadrupole field, beating in phasewith the dipole, leads to short and in-tense cycles; a stronger quadrupole field,if slightly out of phase with the dipolefield, lengthens and weakens the sunspotcycle. Far beyond the critical dynamonumber, chaos results.

Dynamic StarsAS WE NOW KNOW, the sun’s bright-ness increases with the magnetic activity


PE

TER

SAM

EK

Sli

m F

ilm

s

NorthPole

NorthPole

ROTATION of the sun’s surface is faster at the equator and slower near thepoles. This differential rotation (as measured by means of sunquakes by the Global Oscillation Network Group) extends through the outer layers. Thesun’s core, in which fusion generates the energy that ultimately powers thedynamo, most likely rotates at a constant angular velocity, like a rigid body.

moving up toward the surface and erupting as sunspot pairs (d). The sunspots drift toward the poles,with the trailing sunspot reaching first; as a result, the overall field flips (e). In addition to the dipolefield above, the sun probably also has a “quadrupole” field (opposite page, red) whose “beating” withthe dipole field was responsible for the Maunder minimum.

480

440

400

3600.4 0.6 0.8 1

30° latitude

Equator

60° latitudeRota

tion

Freq

uenc

y (n

anoh

ertz

)

Distance from Center/Solar Radius

d e

0

0.3

CORE

RADIATIVE ZONE

CONVECTION ZONE

1

DISTANCE/RADIUS

0.7



over a cycle: the bright plages over-whelm the dark sunspots. (Presumably,as the sun brightens and darkens, its to-tal energy is temporarily channeled intodifferent reservoirs—kinetic, magnetic,thermal or potential.) During the past 24years of satellite observations, the sun’stotal energy output has varied roughly0.1 percent between a brighter, magneti-cally active phase and a fainter, quiet one.

Because of the brevity of the satelliterecords, we do not know the variabilityof the sun’s brightness over decades.Richard Willson and his colleagues atColumbia University’s Center for Cli-mate Systems Research recently found aslight, 0.05 percent increase in brightnessat the observed solar minima in 1986and 1996. Finding a longer-term valuefor brightness variability, however, is vi-tal to evaluating the sun’s influence onEarth. One possible way to answer thisquestion is to examine “star spot” cycleson other stars.

It is not easy to map the features onthe surface of stars. But as magneticfields heat the outer layers of a star’s at-mosphere, they radiate the energy in cer-

tain spectral lines. For example, on oursun, the intensity of the two violet emis-sion lines of calcium (having wave-lengths of 396.7 and 393.4 nanometers)closely follows the strength and extentof the magnetic fields. Variations inthese lines thus give us a measure of thechanging surface magnetism of a star.

At Mount Wilson Observatory in1966, Olin C. Wilson began a programof measuring the magnetic activity ofroughly 100 so-called main-sequencestars—those that, like the sun, are burn-ing hydrogen. (When the hydrogen runsout, a star expands into a red giant.)Most of these stars show obvious signsof magnetic activity, by way of varia-tions in their violet calcium emission

lines. The fluctuations vary greatly inamplitude and duration, depending pri-marily on the age and mass of the star.

All these stars have a dynamo num-ber, D, higher than the critical value re-quired for sustaining magnetic fields. Fora young star of one or two billion years,the rotation period is fast, roughly 10 to 15 days. The resulting high value of D means that these young stars have er-ratic fluctuations in magnetic activityover intervals as short as two years andno well-defined cycles. The fluctuationssometimes repeat, however, having pe-riods between two and 20 years or sothat lengthen with age.

But as a star ages, it slows down—

because its angular momentum is carriedoff by the magnetic wind—and D falls.Then a consistent dynamo cycle beginsto appear, with a period of about six toseven years and sometimes even withtwo independent periods. Later on—foran even lower D—one period starts todominate, lengthening with age fromeight to 14 years. In addition, there areoccasional Maunder minima. If rotationwere to slow further, in the very oldeststars, we predict that the magnetic fieldshould be steady. The Wilson samplecontains a few very old stars, but theystill show cycles, indicating that thesteady dynamo would not be reached in10 billion years—soon after which theywill expand into red giants.

To focus on the solar dynamo, Ba-liunas and her collaborators restrictedWilson’s broad sample of stars to thosesimilar to our sun in mass and age. Thatgroup currently comprises 10- to 20-yearrecords of 150 stars, depending on thecriteria defining similarity to the sun.Many of these stars show prominent cy-cles similar in amplitude and period tothose of the sun. About one quarter of


MO

UN

T W

ILSO

N O

BSE

RVA

TOR

Y H

K P

RO

JEC

T; P

ETE

R S

AME

K S

lim

Fil

ms

● ● ● ● ● ● ● ● ● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0.14

0.16

● ● ●

● ● ●●

● ●● ●

●● ● ● ● ● ● ●

●

●● ● ● ● ●

●●

●●

●

●

● ●

●

●0.3

0.35

0.4

1966 1975 1985 1990 1995 2000

● ● ●

●● ●

●●

●● ●

●●

●

●

●●

●

●

●●

● ●●

●●

●●

●

● ● ● ●●

●

●

0.2

0.25

0.3

1970 1980

●

●

Young star, HD 20686, variable cycle

Mature star, HD 4628, 8.6-year cycle

Mature star, HD 14376, flat0.18

Year

INTERANNUAL MAGNETIC VARIABILITY of stars over the years is detected by way of violet calciumemission lines. Here activity from three nearby stars reveals the likely states of our own sun: thevariable magnetic cycles of a young star (top); the steady cycles of a star at an age comparable to our sun’s (middle); and the subsidence of a sunlike star into a Maunder-type minimum phase(bottom). The magnetic, and therefore sunspot, activity of other stars indicates that our sun iscapable of far greater variability than it has shown in the past century.

ELIZABETH NESME-RIBES, SALLIE L. BALIUNAS and DMITRY SOKOLOFF all have been ac-tive in unraveling connections between the sun’s variations and Earth’s climate. Nesme-Ribes, who recently passed away, was an astronomer at the Paris Observatory and the Na-tional Center for Scientific Research in France. Apart from studying the solar dynamo, sheconducted extensive searches into the 17th-century archives on sunspots at her home in-stitution. Baliunas is a scientist at the Harvard-Smithsonian Center for Astrophysics in Cam-bridge, Mass. She observes the variations of sunlike stars at the Mount Wilson Observato-ry in Pasadena, Calif. Sokoloff is professor of mathematics in the department of physics atMoscow State University in Russia.

THE

AU

THO

RS

Inte

nsity

of C

alci

um E

mis

sion

Lin

es


the records show that the stars are in adead calm, suggesting a phase similar toour sun’s Maunder minimum. This find-ing implies that sunlike stars spend aquarter of their lives in a lull—consistentwith radiocarbon results.

We may have captured one star, HD3651, in transition between the cyclicand Maunder minimum phases. HD3651’s cycles have weakened andlengthened dramatically (from 12 to 15years) as its surface activity has rapidlydropped to very low levels. Sunlike starssuch as HD 3651, observed over a fewdecades, offer us “snapshots” of therange of variability that our sun—andwe—might experience over a timescaleof centuries.

The brightness of these sunlike starscan also be compared with their mag-netic activity. In 1984 thorough and pre-

cise photometric observations of someof the Wilson stars began at the Lowelland Sacramento Peak observatories.Since 1992 those of us at the Smithson-ian Astrophysical Observatory and atTennessee State University have used au-tomated telescopes to observe some ofthese stars. Nearly all the older stars, likethe sun, are brightest near the peak ofthe activity cycle. Some stars vary as lit-tle as our sun does—only 0.1 percentover the last 11-year cycle—but othersunlike stars have varied by as much as0.6 percent in a cycle. Thus, the sun’scurrent changes might be a poor indica-tor of the full range of fluctuations ofwhich it is capable.

Over the decades, researchers haveinferred the evolutionary history of asunlike star from the collection of stellarrecords. A young star has a relativelyrapid rotation period of several days andhigh, irregular levels of surface magnet-ism. Changes in brightness of severalpercent accompany the magnetic varia-tions. The young star is, however, dark-

est during the peak of magnetic activity,presumably because the dark spots areso large that they, not the plages, domi-nate. As the sunlike star ages, it rotatesmore slowly, and the magnetic activitydecreases. Maunder minima appear inthese “older” stars; furthermore, radi-ance now peaks at sunspot maximum,with fluctuations of 1 percent or lessover a cycle.

Influencing EarthTHE STAR-SPOT RESULTS point to achange in brightness of at least 0.4 per-cent between the cyclic phase and theMaunder minimum phase. This valuecorresponds to a decrease in the sun’snet energy input of one watt per squaremeter at the top of Earth’s atmosphere.Simulations performed at the Laborato-ry of Dynamic Meteorology in Paris and

elsewhere suggest that such a reduction,occurring over several decades, is capa-ble of cooling Earth’s average tempera-ture by 1 to 2 degrees C—enough to ex-plain the observed cooling during theMaunder minimum.

But greenhouse gases generated byhuman activity may be warming ourplanet, by trapping heat that would oth-erwise radiate into space. This warmingis equivalent to Earth’s receiving radia-tion of two watts per square meter at thesurface. The sun has apparently deliveredto Earth no more or less than 0.5 to 1.0watt per square meter over the past fewcenturies. Therefore, if direct heating isthe only way in which the sun affectsEarth’s climate and is presumed to actthe same as the enhanced greenhouse ef-

fect, the added greenhouse gases shouldalready be dominating the climate, wash-ing out any correlation with the sun’svarying activity.

The sun’s energy reaches Earth as ra-diation and particles and varies overmany frequencies and periods. Yet thelink between climate and solar magnet-ic activity seems rather persistent. Thelength of the sunspot cycle, for example,correlates closely with global tempera-tures over the past 240 years. Minima insolar magnetism, as traced by radiocar-bon dating in tree rings and beryllium 10in ice cores, coincide with roughly 1,500-year intervals of cooler climate, seen in environmental changes going back10,000 years. In addition, the sunspotcycle correlates with stratospheric windpatterns, for reasons not yet well under-stood. All these pieces of evidence induce

some scientists, including us, to arguethat the sun must be influencing Earthby powerful indirect routes as well as theobvious ones.

Variations in the sun’s ultraviolet ra-diation, for example, may be changingthe ozone content of our upper atmo-sphere, as well as its dynamics. Recentsimulations also indicate that winds inthe lower stratosphere can convey vari-ations in solar radiance down to the tro-posphere, where they interact more di-rectly with the weather system. Suchmatters are now the subject of vigorousdebate. Unraveling the ways in which thesun warms Earth provides vital informa-tion concerning the role played by hu-mankind—and the role played by thesun—in the process of climatic change.


The Variable Sun. Peter V. Foukal in Scientific American, Vol. 262, No. 2, pages 34–41; February 1990.The Paradox of the Sun’s Hot Corona. Bhola N. Dwivedi and Kenneth J. H. Phillips in Scientific American, Vol. 284, No. 6, pages 40–47; June 2001.The Maunder Minimum and the Variable Sun-Earth Connection. Willie Wei-Hock Soon and Steven H. Yaskell. World Scientific Publishing, 2003.


UNRAVELING THE INFLUENCES OF THE SUN provides vital information on the role

OUR STAR PLAYS IN CLIMATE CHANGE.



42 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e A p r i l 2 0 0 1 i s s u e

THE FURY ofSHOCK WAVES FROM THE SUN CAN TRIGGER SEVERE TURBULENCE INTHE SPACE AROUND EARTH, ENDANGERING SATELLITES AND ASTRONAUTS IN ORBIT. A NOVEL SPACECRAFT IS SHOWING HOW SPACESTORMS DEVELOP BY JAMES L. BURCH

The tempest began on a date known for its violent events:Bastille Day, the anniversary of the beginning of the FrenchRevolution. On the morning of July 14, 2000, the Space En-vironment Center in Boulder, Colo., detected a warning signfrom the GOES-8 satellite, which monitors x-rays from thesun as well as weather conditions on Earth. At 10:03 Uni-versal Time the center’s forecasters saw a sharp jump in theintensity of x-rays emanating from active region 9077, a sec-tion of the sun’s surface that had been roiling for the pastweek. The data indicated the onset of a solar flare, a brief butpowerful burst of radiation.

The flare, which reached its maximum intensity at 10:24UT, was also sighted by the Solar and Heliospheric Obser-vatory (SOHO), a spacecraft stationed between the sun andEarth, about 1.5 million kilometers from our planet. Half anhour later, as the flare was waning, SOHO observed an evenmore ominous phenomenon: a bright, expanding cloud thatsurrounded the sun like a halo. It was a coronal mass ejec-tion (CME), an eruption in the corona—the sun’s outer at-mosphere—throwing billions of tons of electrically chargedparticles into interplanetary space. The halo signature meantthat the particles were heading directly toward Earth, at anestimated speed of 1,700 kilometers per second.

As the CME plowed into the solar wind—the flow of ion-ized gas continuously streaming from the sun—it created ashock wave that accelerated some charged particles to evenhigher velocities. In less than an hour a deluge of high-ener-gy protons struck SOHO, temporarily blinding its instru-ments. The bombardment also damaged the spacecraft’s so-lar arrays, causing a year’s worth of degradation in 24 hours.But this torrent of particles was only the leading edge of thesquall. The CME-driven shock wave arrived the next day,slamming into Earth’s magnetic field at 14:37 UT. The im-pact marked the start of a severe geomagnetic storm, whosefull fury was unleashed by the arrival, a few hours later, ofthe CME itself. According to the index of geomagnetic ac-tivity used by the Space Environment Center, the Bastille

Day storm was the largest such event in nearly a decade.Most people on the ground were completely unaware of

the celestial fireworks, but researchers were following thetempest closely, collecting data from instruments on Earthand in space. Among the satellites tracking the storm wasthe Imager for Magnetopause-to-Aurora Global Explo-ration (IMAGE), which NASA had launched just fourmonths earlier. IMAGE is the first satellite dedicated to ob-taining global images of the magnetosphere, the region ofspace protected by Earth’s magnetic field. By providing anoverall picture of the activity in the magnetosphere, IMAGEdoes for space what the first weather satellites did for Earth’satmosphere.

In 1996 I had been selected by NASA to lead a team of en-gineers and scientists in developing the IMAGE spacecraftand analyzing the data that it transmits. As the Bastille Daystorm progressed, we received astounding images of ions cir-cling Earth and pictures of the brilliant aurora borealis—thenorthern lights—that occurred when the charged particlesstruck the upper atmosphere. The results will help scientistsanswer long-standing questions about how CMEs and thesolar wind interact with Earth’s magnetosphere. The find-ings may also have practical applications. Space storms candisable satellites, threaten the safety of astronauts and evenknock out power grids on the ground [see box on page 44].Indeed, the Bastille Day storm caused the loss of the Ad-vanced Satellite for Cosmology and Astrophysics, an x-rayobservatory launched in 1993 by the Japanese space researchagency. In hopes of mitigating such effects in the future, sci-entists are keenly interested in improving the accuracy ofspace weather forecasts.

VIOLENT ERUPTION in the sun’s outer atmosphere on November 8,2000, spewed billions of tons of charged particles toward Earth.The event was observed by the Solar and HeliosphericObservatory (SOHO); the spacecraft’s coronagraph uses a disk(dark circle) to block direct light from the sun (white circle) sothat its atmosphere can be seen.


SOLAR STORMS

S C I E N T I F I C A M E R I C A N 43COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

It’s Not the Heat or the HumidityLIKE WEATHER ON EARTH, weather in space is extremelyvariable. Conditions can turn from quiet to stormy in a matterof minutes, and storms can last for hours or days. And just asterrestrial weather changes with the seasons, space weather,too, follows its own cycles. Solar magnetic activity, which caus-es flares and CMEs, rises and falls every 11 years, and there-fore geomagnetic storms follow the same pattern. The BastilleDay storm took place during the solar maximum, the most ac-tive part of the cycle. Space weather also varies, though lessdramatically, according to the sun’s 27-day rotation period, as alternating streams of fast and slow solar wind sweep past our planet.

Space weather, however, arises from physical processes thatare profoundly different from those responsible for terrestrialweather. The medium for terrestrial weather is the dense, elec-trically neutral gas in Earth’s lower atmosphere, whose behav-ior is governed by the laws of fluid dynamics and thermody-namics. The medium for space weather, in contrast, is plasma—

very sparse gases consisting of equal numbers of positively

charged ions and negatively charged electrons. Unlike the atomsand molecules of the atmosphere, these plasma particles are sub-ject to the influence of electric and magnetic fields, which guideand accelerate the particles as they travel through the space sur-rounding Earth.

Terrestrial weather is driven by the sun’s radiation as it heatsEarth’s atmosphere, oceans and landmasses. But in the mag-netosphere, weather results from the interaction betweenEarth’s magnetic field and the solar wind. The solar wind hasits own magnetic field, which travels with the outflowing plas-ma into interplanetary space. As the wind carries this inter-planetary magnetic field (IMF) away from the sun, the field linestypically stretch out so that they are directed radially (pointingtoward or away from the sun). Under certain conditions, how-ever, the IMF’s field lines can tilt out of the equatorial plane ofthe sun, taking on a northward or southward component. Astrong and sustained southward IMF direction is a key factorin triggering geomagnetic storms. The IMF was oriented south-ward for many hours during the Bastille Day storm.

Protons are the dominant constituents of the solar wind, ac-counting for about 80 percent of its total mass. Helium nucleimake up about 18 percent, and trace quantities of heavier ionsare also present. The average density of the solar wind at Earth’sorbit is nine protons per cubic centimeter. The wind’s averagevelocity is 470 kilometers per second, and the average strengthof the IMF is six nanoteslas (about one five-thousandth thestrength of Earth’s magnetic field at the planet’s surface). Theseproperties, along with the orientation of the IMF, are highlyvariable, and it is this variability that ultimately explains thechanging weather in space.

All the bodies in the solar system are immersed in the solarwind, which continues flowing outward until it encounters theionized and neutral gases of interstellar space. The solar winddoes not impinge directly on Earth and its atmosphere, how-ever. The planet is shielded by its magnetic field, which formsa kind of bubble within the stream of charged particles ema-nating from the sun. The shape of this cavity—the magneto-sphere—is determined by the pressure of the solar wind and bythe IMF [see illustration on opposite page]. The wind com-presses Earth’s magnetic field on the dayside of the planet—theside facing the sun—and stretches the field on the nightside toform a long tail resembling that of a comet. This magnetotailextends more than one million kilometers past Earth, well be-yond the orbit of the moon.

Between the solar wind and the magnetosphere is a thinboundary called the magnetopause, where the pressure of thegeomagnetic field balances that of the solar wind. On Earth’sdayside, this boundary is usually located about 64,000 kilome-ters from the planet’s center, although the distance varies withchanges in the solar-wind pressure. When the pressure increas-es, as occurred during the Bastille Day storm, the dayside mag-netopause is pushed closer to Earth, sometimes by as much as26,000 kilometers. C

OU

RTE

SY O

F N

ASA/

ESA

, SO

HO

/LAS

CO

CO

NSO

RTI

UM

(p

rece

din

g p

ag

e)

THE EFFECTS OF

SPACE STORMSDuring geomagnetic storms, charged particles swirl aroundEarth and bombard the upper atmosphere, particularly at the higher latitudes. The gusts of particles can havesevere effects on:■ POWER GRIDS. As electrons cascade toward Earth, they

create a strong current in the upper atmosphere calledthe auroral electrojet. This current causes fluctuations inthe geomagnetic field, which can induce electrical surgesin power lines on the ground. During an intensegeomagnetic storm on March 13, 1989, a surge knockedout the Hydro-Quebec power grid, plunging large parts ofCanada into darkness.

■ SATELLITES. When particles strike a satellite, the craft’ssurface becomes charged. This buildup sometimestriggers sparks that can short-circuit the satellite’selectronics. Also, space storms heat Earth’s atmosphere,causing it to expand. If the atmospheric density at asatellite’s orbit becomes high enough, friction will slowthe craft and drag it downward. This process led to thepremature fall of Skylab in 1979.

■ ASTRONAUTS. A severe storm could expose the InternationalSpace Station to protons that could penetrate a spacesuitor even the station’s walls. To protect its astronauts, NASA

monitors space weather data. If an oncoming stormposes a risk, NASA will postpone or cancel any plannedspace walks and may order the astronauts to seekshelter in a shielded part of the station.

44 S C I E N T I F I C A M E R I C A N T H E S E C R E T L I V E S O F S T A R SCOPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

Just as the passage of a supersonic jet through the atmo-sphere produces a shock wave, the encounter of the solar windwith the magnetosphere forms a shock wave—known as thebow shock—some 13,000 kilometers upstream (that is, closerto the sun) from the dayside magnetopause. The region of so-lar-wind plasma between the bow shock and the magnetopauseis known as the magnetosheath. Because of its passage throughthe shock, the magnetosheath plasma is slower, hotter and moreturbulent than the plasma farther upstream.

Satellite detectors have indicated that the charged particlessurrounding Earth are a mix of plasma from the magnetosheath(mostly protons) and plasma that flows out of the upper atmo-sphere above the North and South poles (mostly protons andoxygen ions). The proportions of this mix vary according towhether the magnetosphere is in a quiet or a disturbed state.During geomagnetic storms, charged particles bombard Earthat high latitudes. The resulting electric currents heat the upperatmosphere, pumping increased amounts of protons and oxy-gen ions into the magnetosphere. This plasma is stored, togetherwith the solar-wind plasma that has gained entry into the mag-netosphere, in a great reservoir known as the plasma sheet,which extends for tens of thousands of kilometers on Earth’snightside.

At the heart of the study of space weather is a question: Howdo changes in the solar wind affect conditions in the space sur-rounding Earth? In other words, how can the wind overcome

the barrier of the geomagnetic field and drive the motions of theplasma inside the magnetosphere?

Blowing in the Solar WindTHE MOST ACCEPTED ANSWER was proposed in 1961 byBritish physicist James W. Dungey. In this process, called mag-netic reconnection, the field lines of the IMF become tem-porarily interconnected with the geomagnetic field lines on thedayside of the magnetopause [see illustration above]. This tan-gling of the field lines allows large amounts of plasma andmagnetic energy to be transferred from the solar wind to themagnetosphere.

Reconnection is most efficient when the IMF has a compo-nent that is directed southward—that is, opposite to the north-ward direction of Earth’s magnetic field at the dayside of themagnetosphere. Under these circumstances, reconnection takesplace along a wide equatorial belt, opening up nearly the entireouter boundary of the magnetosphere to the solar wind. For oth-er orientations of the IMF, reconnection still happens, but it maybe more localized in the higher latitudes, where the released en-ergy mainly flows around the magnetosphere rather than into it.

The transfer of magnetic energy from the solar wind radi-cally alters the shape of the magnetosphere. When reconnec-tion is initiated on the dayside magnetopause, the intercon-nected IMF and geomagnetic field lines are swept back by thesolar wind over Earth’s poles, pouring energy into the north-


JOE

ZE

FF

DISTURBANCES IN THE MAGNETOSPHERE occur when the interplanetarymagnetic field (IMF) carried by the solar wind turns southward. In a processcalled reconnection, the field lines of the IMF connect with the northwardgeomagnetic field lines at the dayside of the magnetopause (1). Energy

and particles from the solar wind rush into the magnetosphere, enlargingits northern and southern lobes and narrowing the plasma sheet. Then thegeomagnetic field lines themselves reconnect (2), accelerating ions andelectrons toward Earth.

MagnetopauseBow

shock

Lobe

Cusp

Earth

■ Plasma sheet

■ Radiation belts and ring current

■ Plasmasphere

Interplanetarymagnetic

field

Connectedfield lines

1 2Solarwind

Geomagneticfield



ern and southern lobes of the long magnetotail on the night-side. As the lobes swell with the added magnetic energy, theplasma sheet that lies between them begins to thin. The pro-cess continues until the field lines of the northern and southernlobes, which have opposite directions, are pressed together andthemselves reconnect.

This second reconnection releases the solar wind’s magnet-ic field, enabling it to continue its flow through the solar system.At the same time, it allows Earth’s magnetic field lines, whichhave been stretched tailward during the loading of the lobes,to snap back into their normal configuration. The abrupt move-ment of the field lines heats and accelerates the ions and elec-trons in the plasma sheet, injecting them into the inner part ofthe magnetosphere. Some of these particles, traveling along geo-magnetic field lines, dive into the upper atmosphere above

Earth’s poles, stimulating auroral emissions at x-ray, ultravio-let, visible and radio wavelengths as they collide with oxygenatoms and nitrogen molecules. The entire sequence of events—

from dayside reconnection to nightside reconnection to auro-ras—is known as a magnetospheric substorm.

In addition to transferring magnetic energy to the tail lobes,dayside reconnection also intensifies the electric field across themagnetotail. The stronger field, in turn, increases the flow ofions and electrons from the plasma sheet to the inner magne-tosphere. This flow feeds into Earth’s ring current, which is car-ried by charged particles circling the planet above its equator ataltitudes between 6,400 and 38,000 kilometers. During longerperiods of dayside reconnection—which occur when the IMF’sorientation remains consistently southward—the sustained en-hancement of the earthward plasma flow greatly increases the

A SOLAR STORMIN ACTIONFirst warning of the Bastille Daystorm was a solar flare on July 14,2000. Images of the sun fromSOHO’s Extreme Ultraviolet ImagingTelescope (top) show activeregion 9077 (in white box) beforeand during the flare. At about thesame time, SOHO’s coronagraphobserved a coronal mass ejection(CME) that soon deluged thespacecraft with high-speedprotons, temporarily blinding itsinstruments (middle). The shockwave and CME slammed intoEarth’s magnetic field the nextday, triggering auroras observedby the IMAGE spacecraft’sWideband Imaging Camera(bottom) and a sharp drop ingeomagnetic field strength at theplanet’s surface (on oppositepage, middle). In this graph, calledthe disturbance storm time index,zero represents the normalsurface field strength. As thestorm progressed, IMAGE’s HighEnergy Neutral Atom instrumentmonitored the waxing and waningof the ring current around Earth’sequator (on opposite page, top).

July 14, 2000 9:00:00 10:00:00 11:00:00 12:00:00

09:24:10

BEFORE FLARE

BEFORE CME DETECTION HALO CME PROTON DELUGE

DURING FLARE

09:54:05

14:37:27 14:39:29

10:54:07 11:55:12

10:24:10

NAS

A/E

SA,

SOH

O (

sola

r im

ag

es);

NAS

A/IM

AGE

FU

V TE

AM (

au

rora

l im

ag

es);

N

ASA/

IMAG

E H

EN

A TE

AM (

rin

g c

urr

ent

ima

ges

); W

OR

LD D

ATA

CE

NTE

R F

OR

GE

OM

AGN

ETI

SM,

KYO

TO,

JAP

AN (

gra

ph

)

46 S C I E N T I F I C A M E R I C A N T H E S E C R E T L I V E S O F S T A R SCOPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

number and energies of the charged particles in the ring current.An extended period of southward IMF can also lead to a rapidsuccession of substorms, each of which injects more particlestoward Earth. The resulting growth in strength of the ring cur-rent is the classic hallmark of a full-fledged geomagnetic storm.

Here Comes the SunTHE ORIENTATION OF THE IMF turns southward quitefrequently, so magnetospheric substorms are fairly common:on average, they happen a few times every day and last for oneto three hours. But major geomagnetic storms such as theBastille Day event are much rarer. Although they can occur any-time during the 11-year solar cycle, they are most frequent inthe solar maximum period.

Until the early 1990s, it was widely believed that solar flares

triggered geomagnetic storms. Space and solar physicists, how-ever, had been assembling evidence that pointed strongly to an-other culprit, and in 1993 John T. Gosling of Los Alamos Na-tional Laboratory wove the various threads of evidence togetherin an article in the Journal of Geophysical Research that chal-lenged the “solar flare myth.” Gosling set forth a compelling ar-gument for the central role of coronal mass ejections in settingoff large geomagnetic storms. Scientists still do not know whatcauses these violent eruptions in the sun’s corona, but the phe-nomenon most likely involves a reconfiguration of the magnet-ic field lines there. CMEs are often, but not always, associatedwith solar flares.

Not all CMEs cause geomagnetic storms. Most of the erup-tions are not directed at Earth, and of those that are, only aboutone in six is “geoeffective”—strong enough to trigger a storm.


PRE-STORM MAIN PHASE EARLY RECOVERY LATE RECOVERY25

0

–25

–50

–75

–100

–125

–150

–175

–200

–225

–250

–275

–300

–325

a b c d

Stre

ngth

of G

eom

agne

tic F

ield

at E

arth

’s S

urfa

ce (n

anot

esla

s)

14:47:40 14:59:56 15:12:12 15:36:44

July 15 July 16 July 17 July 18

a

b

c

d



The primary factor is the CME’s speed relative to that of the so-lar wind. Only fast CMEs are geoeffective. Why? When fastCMEs plow through the slower solar wind, they produce inter-planetary shock waves, which are responsible for the high-en-ergy particle showers and the severe deformations of Earth’smagnetic field. Even more important, a fast-moving CME com-presses the solar wind ahead of it, thereby increasing the strengthof the magnetic field in the compressed region and in the frontpart of the CME itself. Moreover, this draping of the field aroundthe CME tends to tilt the IMF more along the north-south di-rection, which causes a stronger reconnection when the IMF en-counters Earth’s magnetic field.

A weaker kind of geomagnetic storm occurs during the

declining phase of the solar cycle and near thesolar minimum period. These disturbances,which tend to recur in phase with the sun’s 27-day rotational period, are set off by the inter-action between fast solar winds emanatingfrom holes in the corona and slower winds aris-ing from the sun’s equatorial streamer belt. Al-though CMEs are not the primary cause of re-current geomagnetic storms, they may con-tribute to their intensity.

With the launch of IMAGE in 2000, re-searchers finally had the means to obtain glob-al views of the minute-by-minute progress of alarge geomagnetic storm. The satellite travelsin an elliptical polar orbit, with its altitudevarying from 1,000 to 46,000 kilometers. Thisorbit allows the craft to observe a large part ofthe magnetosphere, including the dayside mag-netopause, the inner reaches of the magnetotailand the polar cusp regions, which are the mainentryways for the particles from the solar wind.

The Perfect Solar StormIMAGE’S INSTRUMENTS are designed to observethe magnetosphere’s plasmas, but they do so indifferent ways. The craft contains three Ener-getic Neutral Atom (ENA) imagers that indi-rectly measure ion flows. When a fast-movingion (such as an oxygen ion) collides with one ofthe neutral hydrogen atoms in the magneto-sphere, it sometimes strips away the hydrogenatom’s lone electron and becomes an energetic

neutral atom. Because this atom no longer carries a charge, itdoes not have to move along the geomagnetic field lines. In-stead it travels in a straight path from where it was created.The ENA imagers record the number and energies of the neu-tral atoms coming from a particular region, and researcherscan deduce from those data the mass, speed, direction anddensity of the ions in that region.

The satellite also carries several instruments that monitoremissions in the ultraviolet part of the spectrum. The ExtremeUltraviolet (EUV) imager measures the density of singly ion-ized helium atoms in the plasmasphere—a doughnut-shapedregion of the inner magnetosphere containing low-energyplasma—by detecting the solar ultraviolet light that they absorband then reradiate. The Far Ultraviolet (FUV) imaging systemhas two instruments for viewing auroras—the WidebandImaging Camera and the Spectrographic Imager—as well asthe Geocorona Photometers for detecting emissions from neu-tral hydrogen atoms. Last, the Radio Plasma Imager sends outpulses of radio waves that bounce off clouds of charged par-ticles. It works like a state trooper’s radar gun: the returningradio signals convey information about the direction, speedand density of the plasma clouds.

During the Bastille Day event in 2000, IMAGE began


NAS

A/IM

AGE

HE

NA

TEAM

(io

n c

lou

d);

E.

MAY

FIE

LD (

spa

cecr

aft

)

THE

AU

THO

R JAMES L. BURCH is vice president of the Space Science and Engi-neering Division of the Southwest Research Institute in San Anto-nio, Tex., and principal investigator for the IMAGE mission. Burchearned his Ph.D. in space science from Rice University in 1968. Hismain research interests are auroral processes, magnetic recon-nection and magnetospheric imaging. He is a Fellow of the Ameri-can Geophysical Union and former chairman of the Committee onSolar and Space Physics of the National Research Council.

High-speed ions

Medium EnergyNeutral Atom Imager

Axial antenna

Radial antennas

Far Ultravioletimaging system

Earth

IMAGE SPACECRAFT is shown above a cloud of high-speed ions circling Earth in thisillustration. Researchers produced the ion image using data from the satellite’sHigh Energy Neutral Atom Imager (on the side opposite the Medium Energy NeutralAtom Imager). IMAGE’s Radio Plasma Imager charts the clouds of charged particlesby sending pulses of radio waves from two 10-meter-long axial antennas and four250-meter-long radial antennas. Although the spacecraft’s body is only 2.25 meterswide, the antennas make IMAGE one of the biggest sensors ever flown.


recording the storm’s effects less thantwo minutes after the CME-drivenshock wave hit Earth’s magnetic field onJuly 15. The Wideband Imaging Cam-era sent back stunning photographs ofthe aurora borealis triggered by thecompression of the field [see bottom il-lustrations on pages 46 and 47]. Amovie created from the images shows asudden dramatic brightening of a ringabove the Arctic region—the auroraloval—with brilliant emissions racinglike brushfires toward the North Pole.The aurora quieted less than an hour af-ter the storm began but flared up againwhen a second shock hit at about 17:00UT. Powerful substorms followed, asenergy stored in the magnetotail wasexplosively released into the upper at-mosphere. Substorms and the attendantauroral displays continued through therest of July 15 and into the morning ofJuly 16.

During the storm’s main phase,which began four hours after its start,the magnetic field strength on Earth’s surface fell precipitous-ly, dropping 300 nanoteslas below its normal value. This phe-nomenon, the defining feature of geomagnetic storms, resultedfrom the rapid growth of the ring current. IMAGE’s EnergeticNeutral Atom imagers produced vivid pictures of this flow ofions and electrons around Earth as it reached its peak on July16 and then began to diminish [see top illustrations on page 47].Once the transfer of energy from the solar wind abates, the flowof plasma into the inner magnetosphere slows, and ions are lostfrom the ring current more rapidly than new ones arrive. As thecurrent weakens, the magnetic field on Earth’s surface re-bounds. The return to pre-storm levels usually takes one to afew days but may require more than a month in the case of ma-jor tempests.

Geomagnetic storms also change the shape of the plasma-sphere. The enhanced flow of plasma from the magnetotail to-ward Earth erodes the plasmasphere by sweeping its chargedparticles toward the dayside magnetopause. When a stormsubsides, the plasmasphere is refilled by ion outflow from theupper atmosphere. Scientists had hypothesized from modelingstudies that the eroded material from the plasmasphere wouldform a long tail extending to the dayside magnetopause andthat from there, it would become lost in the solar wind. Glob-al images of the plasmasphere from IMAGE’s EUV instrumenthave confirmed this decades-old hypothesis [see illustrationabove]. At the same time, the images have revealed structuresin the plasmasphere that raise new questions about its dynamicresponse to magnetospheric disturbances. Perhaps these ques-tions will be answered: NASA has extended IMAGE’s missionthrough October 2009.

On the HorizonALTHOUGH IMAGE has opened anew window on the magnetosphere, ourview of space weather is still imperfect.Unlike terrestrial clouds, the clouds ofplasma seen by IMAGE are complete-ly transparent: nothing is hidden fromsight, but depth perception is lacking.Thus, there will always be the need forsatellites that make local measurementsof the plasmas, as well as the fields andcurrents that govern their motion.

The next step for space weather ob-servation will involve clusters of satellitesacting like hurricane-hunter planes—

they will go where the action is. The Eu-ropean Space Agency began conductingthe first such mission, called Cluster II,in the summer of 2000. (A predecessormission, Cluster I, was destroyed in arocket explosion just after liftoff in1996.) Cluster II consists of four close-ly grouped identical spacecraft designedto probe turbulent plasma phenomenain the magnetosphere and nearby solar

wind. NASA is also planning a cluster mission for launch in2010. The Magnetospheric Multiscale mission will study re-connection, charged particle acceleration, and turbulence at thedayside magnetopause and at specific locations in the magne-totail where substorms are triggered.

The space agencies are considering even more ambitiousmissions involving constellations of satellites: dozens of tinyspacecraft that will monitor large regions of space, just as theglobal weather networks now monitor conditions on Earth.The first constellations would most likely observe the innermagnetosphere and the dayside magnetopause, with each cake-size craft recording the basic characteristics of the plasmas andmagnetic fields.

Earth’s magnetosphere is at once protective and dangerous.Its strong magnetic field shields humans from penetrating ra-diation that would otherwise be lethal. But the field is not strongenough to ward off the most powerful shock waves from the sun.Like the tornado belt or the tropical cyclone zone, the magneto-sphere is a place of sudden storms. And that’s why storm watch-ers such as the IMAGE satellite are so important.


NAS

A/IM

AGE

EU

V TE

AM

PLASMASPHERE appears as a pale blue cloud ofions surrounding Earth (yellow circle) in thispicture obtained by IMAGE during a moderategeomagnetic storm on May 24, 2000. The resultsconfirmed the existence of the hypothesizedplasma tail and revealed an unexpected structurethat was dubbed the “shoulder.”

Plasma tail

Shoulder

From the Sun: Auroras, Magnetic Storms, Solar Flares, Cosmic Rays. Edited by S. T. Suess and B. T. Tsurutani. American Geophysical Union, 1998.

The 23rd Cycle: Learning to Live with a Stormy Star. Sten Odenwald. Columbia University Press, 2001.

More pictures and data from the IMAGE mission are available athttp://image.gsfc.nasa.gov/

General information on space weather can be found atwww.spaceweather.com/




IMPACT CRASHSHOCKS M A S H I N T ORAM COLLIDEVIOLENT BLOWS M A C K J O LT

THIS IS NOT A SIGHT you would ever want to see. If a white dwarf star hit the sun, it would trigger a calamitous series of events—despite the fact thatthe dwarf is barely a hundredth the sun’s diameter. As the dwarf approached,it would suck matter toward it and distort the sun into a pear shape.Thankfully, such a collision is unlikely. But similar events occur regularly in denser parts of the galaxy, such as globular star clusters.

WHEN TWO STARS SMASH INTO EACH OTHER, IT CAN BE A VERY PRETTY SIGHT (AS LONG AS YOU’RE NOT TOO CLOSE BY).

THESE OCCURRENCES WERE ONCE CONSIDERED IMPOSSIBLE, BUT THEY HAVETURNED OUT TO BE COMMON IN CERTAIN GALACTIC NEIGHBORHOODS

BY MI C H A E L S H A R A

W H E N S TA R S



Of all the ways for life on Earth to end, the collisionof the sun and another star might well be the mostdramatic. If the incoming projectile were a whitedwarf—a superdense star that packs the mass ofthe sun into a body a hundredth the size—the res-

idents of Earth would be treated to quite a fireworks show.The white dwarf would penetrate the sun at hypersonicspeed, over 600 kilometers a second, setting up a massiveshock wave that would compress and heat the entire sunabove thermonuclear ignition temperatures.

It would take only an hour for the white dwarf to smashthrough, but the damage would be irreversible. The super-heated sun would release as much fusion energy in that houras it normally does in 100 million years. The buildup of pres-sure would force gas outward at speeds far above escape ve-locity. Within a few hours the sun would have blown itselfapart. Meanwhile the agent of this catastrophe, the whitedwarf, would continue blithely on its way—not that we wouldbe around to care about the injustice of it all.

For much of the 20th century, the notion that stellar colli-

sions might be worth studying seemed ludicrous to as-tronomers. The distances between stars in the neighborhoodof the sun are just too vast for them to bump into one anoth-er. Other calamities will befall the sun (and Earth) in the dis-tant future, but a collision with a nearby star is not likely tobe one of them. In fact, simple calculations carried out earlyin the 20th century by British astrophysicist James Jeans sug-gested that not a single one of the 100 billion stars in the diskof our galaxy has ever run into another star.

But that does not mean collisions are uncommon. Jeans’s as-sumptions and conclusion apply to the environs of the sun butnot to other, more exotic parts of the Milky Way. Dense starclusters are a veritable demolition derby. Within these tight knotsof stars, observers in recent years have discovered bodies that are

forbidden by the principles of ordinary stellar evolution—butthat are naturally explained as smashed-up stars. Collisions canmodify the long-term evolution of entire clusters, and the mostviolent ones can be seen halfway across the universe.

A Star-Eat-Star WorldTHE 1963 DISCOVERY of quasars was what inspired skep-tical astronomers to take stellar collisions seriously. Manyquasars radiate as much power as 100 trillion suns. Becausesome brighten or dim significantly in less than a day, their en-ergy-producing regions must be no larger than the distancelight can travel in a day—about the size of our solar system. Ifyou could somehow pack millions of stars into such a smallvolume, astronomers asked, would stars crash? And could thisjostling liberate those huge energies?

By 1970 it became clear that the answer to the second ques-tion was no. Nor could stellar slam dancing explain the nar-row jets that emanate from the central powerhouses of manyquasars. The blame fell instead on supermassive black holes.(Ironically, some astronomers have recently proposed that stel-

lar collisions could help feed material into these holes.)Just as extragalactic astronomers were giving up on stellar

collisions, their galactic colleagues adopted them with a ven-geance. The Uhuru satellite, launched in 1970 to survey the skyfor x-ray-emitting objects, discovered about 100 bright sourcesin the Milky Way. Fully 10 percent were in the densest type ofstar cluster, globular clusters. Yet such clusters make up only0.01 percent of the Milky Way’s stars. For some reason, theycontain a wildly disproportionate number of x-ray sources.

To express the mystery in a different way, consider whatproduces such x-ray sources. Each is thought to be a pair ofstars, one of which has died and collapsed into a neutron staror a black hole. The ex-star cannibalizes its partner and in do-ing so heats the gas to such high temperatures that it releasesx-rays. Such morbid couplings are rare. The simultaneous evo-lution of two newborn stars in a binary system succeeds in pro-ducing a luminous x-ray binary just once in a billion tries.

What is it about globular clusters of stars that overcomesthese odds? It dawned on astronomers that the crowded con-ditions in globulars could be the deciding factor. A millionstars are crammed into a volume a few dozen light-yearsacross; an equivalent volume near the sun would accommo-date only a hundred stars. Like bees in a swarm, these starsmove in ever changing orbits. Lower-mass stars tend to beejected from the cluster as they pick up energy during close en-counters with more massive single and double stars, a processreferred to as evaporation because it resembles the escape ofmolecules from the surface of a liquid. The remaining stars,having lost energy, concentrate closer to the cluster center. Giv-

52 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e N o v e m b e r 2 0 0 2 i s s u e

MAR

K A

. G

ARLI

CK

(p

rece

din

g p

ag

e)

■ This article represents one of those cases in which thetextbooks need to be revised. The conventional wisdomthat stars can never hit each other is wrong. Collisionscan occur in star clusters, especially globular clusters,where the density of stars is high and wheregravitational interactions heighten the odds of impact.

■ The leading observational evidence for collisions is two-fold. Globular clusters contain stars called blue stragglersthat are best explained as the outcome of collisions. Andglobulars contain an anomalously high number of x-raysources—again the likely product of collisions.

Overview/Stellar Collisions

The tranquil night sky masks a universe in which a THOUSAND PAIRS OF STARS COLLIDE EVERY HOUR.


en enough time, the tightly packed stars will begin to collide.Even in a globular cluster, the average distance between

stars is much larger than the stars themselves. But Jack G. Hillsand Carol A. Day, both then at the University of Michigan atAnn Arbor, showed in 1975 that the probability of impact isnot a simple matter of a star’s physical cross section. Becausethe stars in a globular cluster move at a lackadaisical (by cos-mic standards) 10 to 20 kilometers a second, gravity has plen-ty of time to act during close encounters. Without gravity, twostars can hit only if they are aimed directly at each other; withgravity, each star pulls on the other, deflecting its path. The starsare transformed from ballistic missiles with a preset flight pathinto guided missiles that home in on their target. A collision be-comes up to 10,000 times more likely. In fact, half the stars inthe central regions of some globular clusters have probably un-dergone one or more collisions over the past 13 billion years.

Around the same time, Andrew C. Fabian, James E. Pringleand Martin J. Rees of the University of Cambridge suggested thata grazing collision or a very near miss could cause two isolated

stars to pair up. Normally a close encounter of two celestial bod-ies is symmetrical: they approach, gather speed, swing past eachother and, unless they make contact, fly apart. But if one is a neu-tron star or a black hole, its intense gravity can contort the oth-er, sapping some of its kinetic energy and preventing it from es-caping, a process known as tidal capture. The neutron star orblack hole proceeds to feast on its ensnared prey, spewing x-rays.

If the close encounter involves not two but three stars, it iseven more likely to produce an x-ray binary. The dynamics ofthree bodies is notoriously complex and sometimes chaotic; thestars usually redistribute their energy in such a way that thetwo most massive ones pair up and the third gets flung away.The typical situation involves a loner neutron star that comesa little too close to an ordinary binary pair. One of the ordi-nary stars in the binary is cast off, and the neutron star takesits place, producing an x-ray source. The bottom line is thatthree-body dynamics and tidal capture lead to a thousandfoldincrease in the rate at which x-ray sources form in globularclusters, neatly solving the puzzle raised by Uhuru.


ALFR

ED

T.

KAM

AJIA

N

TIDAL CAPTUREBLACK HOLE or neutron star makes an evensmaller target than a normal star. But it canexert powerful tidal forces that bend apassing star out of shape. The distortiondissipates energy and can cause the twobodies to go into orbit. A collision betweenthe two is then just a matter of time, assuccessive close passages rob ever moreorbital energy.

PROCESSES THAT MAKE COLLISIONS MORE LIKELY

GRAVITATIONAL FOCUSINGIN THE COSMIC SCHEME of things, stars aresmall targets for impacts. Each sweeps outa very narrow region of space, and at firstglance it appears that two such regions areunlikely to overlap. But gravity makes starsinto larger targets by deflecting the pathsof any approaching objects. In effect, eachstar actually sweeps out a region manytimes its own size, greatly increasing theprobability of overlap and collision.

EVAPORATIONSTARS IN A GLOBULAR CLUSTER zip aroundlike bees in a swarm. Occasionally three orfour come close to one another. Their closeencounter redistributes energy and canfling one of the stars out of the clusteraltogether. The remaining cluster membershuddle together more tightly. If enoughstars are ejected, the ones left behindbegin to collide. This process typicallyoccurs over billions of years.

EJECTED STAR

DEFLECTED PATH

BLACK HOLE



Crash SceneWHAT HAPPENS WHEN TWO STARS smack into each oth-er? As in a collision involving two vehicles, the outcome de-pends on several factors: the speed of the colliding objects, theirinternal structures and the impact parameter (which specifieswhether the collision is head-on or a sideswipe). Some incidentsare fender benders, some are total wrecks and some fall in be-tween. Higher-velocity and head-on collisions are the best atconverting kinetic energy into heat and pressure, making fora total wreck.

Although astronomers rely on supercomputers to study col-lisions in detail, a few simple principles govern the overall ef-fect. Most important is the density contrast. A higher-density

star will suffer much less damage than a tenuous one, just as acannonball is barely marked as it blows a watermelon to shreds.A head-on collision between a sunlike star and a vastly denserstar, such as a white dwarf, was first studied in the 1970s and1980s by me and my colleagues Giora Shaviv and Oded Regev,both then at Tel Aviv University and now at the Technion-Israel Institute of Technology in Haifa. Whereas the sunlike staris annihilated, the white dwarf, being 10 million times as dense,gets away with only a mild warming of its outermost layers. Ex-cept for an anomalously high surface abundance of nitrogen,the white dwarf should appear unchanged.

The dwarf is less able to cover its tracks during a grazingcollision, as first modeled by me, Regev and Mario Livio of theSpace Telescope Science Institute. The disrupted sunlike starcould form a massive disk in orbit around the dwarf. No suchdisks have yet been shown to exist, but it is possible that as-tronomers might be mistaking them for mass-transferring bi-nary stars in star clusters.

When the colliding stars are of the same type, density andsize, a very different sequence of events occurs. The case of twosunlike stars was first simulated in the early 1970s by AlastairG. W. Cameron, then at Yeshiva University, and Frederick G. P.Seidl, then at the NASA Goddard Institute for Space Studies.As the initially spherical stars increasingly overlap, they com-press and distort each other into half-moon shapes. Temper-atures and densities never climb high enough to ignite dis-ruptive thermonuclear burning. As a small percent of the to-tal mass squirts out perpendicular to the direction of stellarmotion, the rest mixes together. Within an hour, the two starshave fused into one.

It is much more likely that two stars will collide somewhatoff-axis than exactly head-on; it is also more likely that theywill have slightly different rather than identical masses. Thisgeneral case has been studied in detail by Willy Benz of the Uni-versity of Bern in Switzerland, Frederic A. Rasio of North-western University, James C. Lombardi of Vassar College,

Joshua E. Barnes of the University of Hawaii at Manoa andtheir collaborators. It is a beautiful mating dance that ends inthe perpetual union of the two stars.

The object that results is fundamentally different from anisolated star such as our sun. An isolated star has no way of re-plenishing its initial allotment of fuel; its life span is preor-dained. The more massive the star is, the hotter it is and thefaster it burns itself out. Given a star’s color, which indicatesits temperature, computer models of energy production canpredict its life span with high precision. But a coalesced stardoes not follow the same rules. Mixing of the layers of gas dur-ing the collision can add fresh hydrogen fuel to the core, witha rejuvenating effect rather like tossing twigs on a dying camp-

fire. Moreover, the object, being more massive than its pro-genitors, will be hotter, bluer and brighter. Observers who lookat the star and use its color and luminosity to deduce its agewill be wrong.

For instance, the sun has a total life span of 10 billion years,whereas a star twice its mass is 10 times brighter and lasts only800 million years. Therefore, if two sunlike stars merge halfwaythrough their lives, they will form a single hot star that is fivebillion years old at the moment of its creation but looks asthough it must be younger than 800 million years. The lifetimeremaining to this massive fused star depends on how much hy-drogen fuel was thrown to its center by the collision. Usuallythis lifetime will be much shorter than that of each of its par-ents. Even in death the star distinguishes itself. When it dies (byswelling to become a red giant, a planetary nebula and finallya white dwarf), it will be much hotter than other, older whitedwarfs of similar mass.

Got the BluesIN A GLOBULAR CLUSTER, massive merged stars will standout conspicuously. All the members of a globular are born atroughly the same time; their temperature and brightness evolvein lockstep. But a coalesced star is out of sync. It looks preter-naturally young, surviving when others of equal brightness andcolor have passed on. The presence of such stars in the cores of


MICHAEL SHARA wanted to be an astronomer from age seven. Hisearliest interest came from observing binary stars with surplusWorld War II binoculars. Today he is curator and chair of the de-partment of astrophysics at the American Museum of Natural His-tory in New York City. Before joining the museum, he put in 17 yearsat the Space Telescope Science Institute, where he oversaw thepeer-review committees for the Hubble Space Telescope. Shara’sresearch interests include stellar collisions, novae and supernovae,and the populations of stars that inhabit star clusters and galaxies.Nowadays he observes with Hubble and ground-based instruments.

THE

AU

THO

R

If a white dwarf hit the sun, Earth would fly offAND WANDER LIFELESSLY AROUND THE GALAXY.


CR

ED

IT

Having an Impact

STARS COME IN seven basic types, with black holes having thegreatest density and supergiants the least. Our sun is a main-sequence star. This table lists the outcomes of the 28 differentpairings. In many cases, a collision can have more than one possibleoutcome, depending on impact speed, angle and other parameters.The results here assume deeply penetrating collisions at modestspeeds. Two such collisions (yellow) are shown below.

WHITE DWARF STAR takes a monthto penetrate the bloated red giant.It escapes unscathed and spiritsaway some of the giant’s gas. Thegiant, however, falls apart, althoughits core remains intact and becomesanother white dwarf. (The full movieis available at www.ukaff.ac.uk/movies/collisions.mov)

ORDINARY STARS of unequal massstrike off-center. The less massiveone is also denser, so it stays intactfor longer. For an hour, it burrowsinto the larger star. A single, rapidlyspinning star results. Some mass islost to deep space. (The movie is atwww.ifa.hawaii.edu/faculty/barnes/research_stellar_collisions)

1

2

4

5

3

1

2

4

5

3

MAIN SEQUENCE HITS MAIN SEQUENCEWHITE DWARF HITS RED GIANT

MAT

THE

W B

ATE

Un

iver

sity

of

Exet

er (

bot

tom

lef

t);

JOSH

UA

BAR

NE

S U

niv

ersi

ty o

f H

aw

aii

(b

otto

m r

igh

t)


BROWN DWARF

black hole + disk

neutron star or black hole + disk

white dwarf or neutron star

main sequence or brown dwarf

WHITE DWARF

black hole +disk


neutron star or white dwarf

NEUTRON STAR

black hole +disk

neutron staror black hole + disk

BLACK HOLE

black hole

SUPERGIANT

black hole + disk + white dwarf

neutron star or black hole + disk + white dwarf

white dwarf +white dwarf

brown dwarf +white dwarf

main sequence +white dwarf



MAIN SEQUENCE

black hole +disk


white dwarf

main sequence

main sequence

RED GIANT

black hole +disk + white dwarf

neutron star or black hole + disk + white dwarf


brown dwarf +white dwarf

main sequence +white dwarf


BLACK HOLE

NEUTRON STAR

WHITE DWARF

BROWN DWARF

MAINSEQUENCE

RED GIANT

SUPERGIANT



IN THE AFTERMATH of the collision betweenthe sun and a white dwarf, the sun explodesas a giant thermonuclear bomb, leaving agaseous nebula. A small percent of the sun’smass collects in a disk around the whitedwarf, which continues on its way. Earthsurvives, but the oceans and atmosphere boil away. No longer held by the gravity of a central star, the planets all fly off intointerstellar space and wander lifelesslyaround the galaxy.


dense star clusters is one of the most compelling predictionsof stellar-collision theory.

As it happens, Allan R. Sandage of the Carnegie Institutionof Washington discovered in the early 1950s that globular clus-ters contain anomalously hot and bright stars called blue strag-glers. Researchers have since advanced a dozen theories of theirorigin. But it is only in the past decade that the Hubble SpaceTelescope has provided a strong link with collisions.

In 1991 Francesco Paresce, George Meylan and I, all thenat the Space Telescope Science Institute, found that the verycenter of the globular cluster 47 Tucanae is crammed with bluestragglers, exactly where collision theory predicted they shouldexist in greatest number. Six years later David Zurek of theSpace Telescope Science Institute, Rex A. Saffer of VillanovaUniversity and I carried out the first direct measurement of themass of a blue straggler in a globular cluster. It has approxi-mately twice the mass of the most massive ordinary stars in thesame cluster—as expected if stellar coalescence is responsible.Saffer and his colleagues have found another blue straggler tobe three times as massive as any ordinary star in its cluster. As-tronomers know of no way other than a collisional merger tomanufacture such a heavy object in this environment.

We are now measuring the masses and spins of dozens ofblue stragglers. Orsola De Marco of the American Museum ofNatural History in New York City and her colleagues have re-cently detected disks of gas orbiting several blue stragglers—perhaps remnants of their violent births. Meanwhile observersare also looking for the other predicted effects of collisions. Forinstance, S. George Djorgovski of the California Institute ofTechnology and his co-workers have noted a decided lack ofred giant stars near the cores of globular clusters. Red giantshave cross sections thousands of times as large as the sun’s, sothey are unusually big targets. Their dearth is naturally ex-plained by collisions, which would strip away their outer layersand transform the stars into a different breed.

To be sure, all this evidence is circumstantial. Definitiveproof is harder to come by. The average time between collisionsin the 150 globular clusters of the Milky Way is about 10,000years; in the rest of our galaxy it is billions of years. Only if weare extraordinarily lucky will a direct collision occur closeenough—say, within a few million light-years—to permit today’sastronomers to witness it with present technology. The first real-time detection of a stellar collision may come from the gravita-tional-wave observatories that are capturing data. Close en-counters between stellar-mass objects should lead to distortionsin the spacetime continuum. The signal is especially strong forcolliding black holes or neutron stars. Such events have been

implicated in the enormous energy releases associated withgamma-ray bursts.

Collisions are already proving crucial to understandingglobulars and other celestial bodies. Computer simulations sug-gest that the evolution of clusters is controlled largely by tight-ly bound binary systems, which exchange energy and angularmomentum with the cluster as a whole. Clusters can dissolve al-together as near-collisions fling stars out one by one. Piet Hut

of the Institute for Advanced Study in Princeton, N.J., and Al-ison Sills of McMaster University in Ontario have argued thatstellar dynamics and stellar evolution regulate each other bymeans of subtle feedback loops.

The fates of planets whose parent stars undergo close en-counters is another addition to the topic of stellar collisions. Nu-merical simulations by Jarrod R. Hurley of the American Mu-seum of Natural History show that the planets often fare bad-ly: cannibalized by their parent star or one of their planetarysiblings, set adrift within the star cluster, or even ejected fromthe cluster and doomed to tramp through interstellar space. Re-cent Hubble observations by Ron Gilliland of the Space Tele-scope Science Institute suggest that stars in a nearby globularcluster do indeed lack Jupiter-size planets, although the causeof this deficiency is not yet known for sure.

Despite the outstanding questions, the progress in this fieldhas been astonishing. The very idea of stellar collisions wasonce absurd; today it is central to many areas of astrophysics.The apparent tranquillity of the night sky masks a universe ofalmost unimaginable power and destruction, in which a thou-sand pairs of stars collide somewhere every hour. And the bestis surely yet to come. New technologies may soon allow directand routine detection of these events. We will watch as somestars die violently, while others are reborn, phoenixlike, dur-ing collisions.


MAR

K A

. G

ARLI

CK

The First Direct Measurement of the Mass of a Blue Straggler in theCore of a Globular Cluster: BSS 19 in 47 Tucanae. Michael M. Shara, Rex A. Saffer and Mario Livio in Astrophysical Journal Letters, Vol. 489, No. 1, Part 2, pages L59–L62; November 1, 1997.Star Cluster Ecology III: Runaway Collisions in Young Compact StarClusters. Simon Portegies Zwart, Junichiro Makino, Stephen L. W.McMillan and Piet Hut in Astronomy and Astrophysics, Vol. 348, No. 1,pages 117–126; 1999. arxiv.org/abs/astro-ph/9812006Evolution of Stellar Collision Products in Globular Clusters–II: Off-AxisCollision. Alison Sills, Joshua A. Faber, James C. Lombardi, Jr., Frederic A.Rasio and Aaron Warren in Astrophysical Journal, Vol. 548, No. 1, Part 1,pages 323–334; February 10, 2001. astro-ph/0008254The Promiscuous Nature of Stars in Clusters. Jarrod R. Hurley andMichael M. Shara in Astrophysical Journal, Vol. 570, No. 1, Part 1, pages 184–189; May 1, 2002. astro-ph/0201217


Astronomers have recently detectedDISKS OF GAS ORBITING BLUE STRAGGLERS—

remnants of the stars’ violent births.



All the shimmering stars that pierce the

night sky shine because of the same fun-

damental process: nuclear fusion. When

two or more atomic nuclei collide and

fuse into one, they release virtually un-

imaginable amounts of energy. The fu-

sion of one gram of hydrogen, for ex-

ample, liberates as much energy as the

combustion of 20,000 liters of gasoline.

In stars such as the sun, fusion reactions

burn brilliantly for billions of years.


ALFR

ED

T.

KAM

AJIA

N

In these systems, ultradense neutron stars feed on their more sedatecompanions. Such stellar cannibalism produces brilliant outpourings of x-rays and drastically alters the evolution of both stars

By Edward P. J. van den Heuvel and Jan van Paradijs

X-RAYX-RAYB I N A R I E S

X-RAY BINARIES make up twovery different classes ofdouble-star systems. In bothcases, a neutron star lies atthe heart of the x-ray source.Most young x-ray binaries,such as Centaurus X-3 (top),contain a bright blue starhaving 10 to 40 times themass of the sun. Low-massx-ray binaries usuallycontain far older, sunlikestars; in the tiny system 4U 1820-30 (bottom), bothstars must be compactobjects, presumably aneutron star and a larger butless massive white dwarf.


Bluegiantstar

Massive x-ray binary(Centaurus X-3)

Neutron star

Low-mass x-ray binary

(4U 1820-30)

Sun

White dwarf starNeutron

star

DETAIL ENLARGED 20 TIMES

w w w . s c i a m . c o m U p d a t e d f r o m t h e N o v e m b e r 1 9 9 3 i s s u e 59COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.


They are not the only source of stellarenergy, however. In 1971 astronomersrecognized a class of bizarre, x-ray-emitting stars, known as x-ray binaries,whose intense emissions require an en-ergy source far more efficient than evenfusion.

Theorists have deduced that theseobjects consist of a normal star orbitinga collapsed stellar corpse, usually a neu-tron star. Neutron stars are so dense thatthe entire mass of the star is squeezedinto what is essentially a single atomicnucleus 20 kilometers across. The starsin these binaries lie so close together thatgas can flow from the normal star to theneutron star. That captured material

forms a rapidly swirling disk whose in-ner edge, just above the neutron star’ssurface, races around at nearly the speedof light. Friction within the disk eventu-ally causes the gas to fall inward, or ac-crete, onto the neutron star. In the pro-cess, violent collisions between particlesheat the gas to temperatures of 10 mil-lion to 100 million kelvins. Under suchincredibly hot conditions, the gas emitstorrents of energetic x-rays. Pound forpound, accretion unleashes 15 to 60times as much energy as does hydrogenfusion.

Astronomers now recognize that ac-cretion powers a rich diversity of astro-physical objects. These range from in-fant stars to quasars, objects about the

size of the solar system that outshine en-tire galaxies, most likely as a result of gasspiraling into a supermassive black hole.X-ray binaries serve as ideal showcasesfor learning in detail how the accretionprocess works. They are bright and rel-atively nearby, residing well within ourgalaxy.

The study of x-ray binaries also pro-vides a glimpse into the life cycle of someof the most exotic and dynamic stellarsystems in the sky. In these stellar duos,one or both members spend some timefeeding off their partner. That transferof material stunningly alters both stars’development. One star may pay for itsgluttony by prematurely ending its life in

a spectacular supernova explosion. Onthe other hand, placid, elderly neutronstars may receive an infusion of rota-tional energy that causes them to be-come a prominent source of rapidlypulsed radio waves.

Rocket to the PulsarsD E S P I T E T H E I R P R O M I N E N C E inthe x-ray sky, x-ray binaries escaped the notice of researchers until the dawnof the space age in the 1960s. Celestial x-rays are absorbed high in the upper at-mosphere, precluding their detectionfrom the ground. The advent of spacetechnology opened up an entirely newfield of investigation by making it possi-ble to loft telescopes above the obscuring

layers of Earth’s turbulent atmosphere.In 1962 Riccardo Giacconi and his

associates at American Science and En-gineering in Cambridge, Mass., placedan x-ray detector on board a rocket anddiscovered the first known celestial x-ray source, Scorpius X-1. (In 2002 Gi-acconi was awarded the Nobel Prize inPhysics for this feat.) The name Scor-pius X-1 indicates that it is the brightestx-ray-emitting object in the constella-tion Scorpius. Scorpius X-1 shinesabout 1,000 times brighter in x-raysthan in visible light. The identity of theobject emitting this radiation was a to-tal mystery.

In the following years, x-ray detec-

tors placed on rockets and very high al-titude balloons revealed a few dozensimilar “x-ray stars.” Astronomers tru-ly began to understand these objectsonly after 1970, when the NationalAeronautics and Space Administrationlaunched Uhuru, the first x-ray satellite,which was designed and built by a teamled by Giacconi. Suddenly, astronomerscould study the x-ray sky around theclock. Within its first few months of ser-vice, Uhuru revealed two intriguing x-ray sources, Centaurus X-3 and Her-cules X-1. Both objects vary in bright-ness in a rapid, extremely regular man-ner: once every 4.84 seconds for Cen-taurus X-3, once every 1.20 seconds forHercules X-1. These sources turned outto be the first of a whole class of pulsedx-ray stars.

The pulses provided a critical clue tothe nature of these objects. In 1967 An-tony Hewish and S. Jocelyn Bell of theUniversity of Cambridge, along withseveral co-workers, discovered pulsars,a class of stars that emit regular blips ofradio emission. After some initial puz-zlement, theorists realized that radiopulsars are swiftly spinning neutronstars whose powerful magnetic fields


JAR

ED

SC

HN

EID

MAN

NEUTRON STAR in the young, massive binary Centaurus X-3 emits pulses of x-rays as it rotates (left).But in the tiny low-mass binary 4U 1820-30, bursts of x-rays occur erratically, when gas collects onthe surface of the old neutron star and undergoes a thermonuclear detonation (right).

0 10 30 0 30

X-ray burst(low-mass x-ray binary)

20 10 20

X-ray pulses(massive x-ray binary)

Time (seconds)Time (seconds)

Rela

tive

X-ra

y In

tens

ity

One star may pay for its gluttony IN A PREMATURE SUPERNOVA EXPLOSION,

but another may receive an energy infusion.


generate a lighthouse beam of radiowaves that flashes by the observer onceeach rotation. The similarly short andconstant variations of the newfound x-ray stars hinted that they, too, were as-sociated with neutron stars.

Another noteworthy trait of Cen-taurus X-3 and Hercules X-1 is that theyexperience regular eclipses, in whichthey dip to a small fraction of their nor-mal brightness. These eclipses provedthat the objects must be binary stars,presumably a neutron star orbiting alarger but much more sedate stellar com-panion that occasionally blocks the neu-tron star from view. Centaurus X-3 hasan orbital period of 2.087 days; for Her-cules X-1, the period is 1.70 days.

The pieces of the puzzle began to fallinto place. The short orbital periods ofthe pulsating x-ray stars demonstratedthat the two stars sit very close to eachother. In such proximate quarters theneutron star can steal gas from its com-panion; the gas settles into a so-called ac-cretion disk around the neutron star.The inner parts of the disk greatly sur-pass the white-hot temperatures on thesurface of the sun (about 6,000 kelvins).As a result, the accretion disk shinesmostly in the form of x-rays, radiationthousands of times as energetic as is vis-ible light. Accretion is so efficient thatsome x-ray binaries emit more than

10,000 times as much energy in x-raysas the sun radiates at all wavelengths.

The x-ray pulsations occur becausethe neutron star has a strong magneticfield whose axis is inclined with respectto its axis of rotation. Close to the neu-tron star, the magnetic field directs theinfalling, electrically charged gas towardthe star’s magnetic poles. There the gascrashes onto the surface, giving rise totwo columns of hot (100 million kel-vins), x-ray-emitting material. As thestar rotates, these columns move in andout of view as seen from Earth, explain-ing the variation in the star’s apparent x-ray flux. Several researchers indepen-dently arrived at this explanation of pul-sating and eclipsing binary x-raysources; indeed, by 1972, it had alreadybecome accepted as the standard modelfor such objects.

Careful timing of the pulsations of x-ray binaries showed that they are notperfectly regular. Instead the period ofpulsation smoothly increases and de-creases over an interval equal to the or-bital period. This phenomenon resultsfrom the motion of the x-ray sourcearound the center of gravity of the bina-ry star system. While the source is mov-ing toward Earth, each pulse travels ashorter distance than the one before andso arrives a minuscule fraction of a sec-ond early; while the source is moving

away from Earth, each pulse arrives asimilar amount late.

The amplitude of this effect revealsthe velocity at which the source movesalong the line of sight to Earth. Centau-rus X-3 swings back and forth at 415kilometers a second. That velocity im-plies that the companion star has atleast 15 times the mass of the sun, typi-cal of a brilliant, short-lived blue star.Since the early 1970s, astronomers haveuncovered about 70 pulsating x-ray bi-naries. In nearly all cases, the compan-ion stars are luminous blue stars havingmasses between 10 and 40 times that ofthe sun.

The bright stars in x-ray binariesshow periodic changes in the frequencyof dark lines, or absorption lines, in theirspectra. These changes, known as Dopp-ler shifts, result from the orbital motionof the visible star around the x-raysource. Radiation from an approachingobject appears compressed, or bluer;likewise, radiation from a receding ob-ject looks stretched, or redder. The de-gree of the Doppler shift indicates thestar’s rate of motion. Because the corre-sponding velocity of the x-ray sourcecan be deduced from the variations ofthe pulse period, one can use Newton’slaw of gravity to derive the mass of theembedded neutron star.

The measured neutron star masses


JAR

ED

SC

HN

EID

MAN

INFALL OF MATTER, or accretion, can be nature’s most efficient mechanismfor generating energy. The energy liberated depends on surface gravity.Matter falling onto the sun (left) attains only a tiny fraction of the velocity

of material accreting onto an ultradense neutron star (right). Frictionconverts kinetic energy into thermal energy; infalling gas in an x-ray binaryreaches 100 million kelvins, causing it to emit energetic x-rays.

1.4 millionkilometers

SunNeutron

star

X-ray

140,000 kilometers a second at impact

Normalstar

620 kilometersa secondat impact

Infallingcomet

20 kilometers

Infallinggas

X-ray binary

Accretion onto the Sun Accretion onto a Neutron Star



fall primarily between 1.2 and 1.6 timesthe mass of the sun, in good agreementwith theoretical expectations. Research-ers have found, much to their excite-ment, that several nonpulsating x-ray bi-naries seem to contain stars having morethan about three solar masses. Currenttheory holds that neutron stars exceed-ing that mass limit will produce a gravi-tational field so intense that it collapseswithout limit. The result is one of na-ture’s most intriguing phenomena: ablack hole, an object whose gravity hascut it off from the rest of the universe.

Burst or Pulse, Not Both AS ASTRONOMERS have found more x-ray binaries, they have come to recog-nize the existence of two distinct popu-lations: those containing large and lu-minous blue stars and those containingmuch older, less massive stars more

akin to the sun. The x-ray binaries thatinclude massive blue stars must be veryyouthful. A star more than 15 times asmassive as the sun squanders its supplyof hydrogen fuel in less than 10 millionyears, a blink of the eye compared withthe 13-billion-year age of the MilkyWay. Hence, the double-star systemsfrom which these x-ray binaries evolvedmust have been born only a few millionyears ago in interstellar gas clouds. Likethese clouds and other young, hot stars,pulsating massive x-ray binaries tend toconcentrate in the plane of the MilkyWay, but not toward the galactic center.

About half the strong x-ray sourcesin our galaxy, including Scorpius X-1,belong to a very different stellar popula-tion. These x-ray binaries concentratepredominantly in the central lens-shapedbulge of the galaxy and in globular clus-ters, dense spherical swarms of stars.

Such regions harbor mostly older stars,those having ages between about five bil-lion and 13 billion years.

In general, these elderly x-ray bina-ries do not undergo regular pulsations.The visible-light spectra of the aged x-ray binaries also appear utterly unlikethose of normal stars. Instead they growsteadily brighter toward the blue end ofthe spectrum; some of their radiationemerges at distinct wavelengths, or col-ors. Theoretical models indicate thatsuch a spectrum would be produced byan inflowing disk of gas heated by in-tense x-rays streaming from inner partsof the disk, just above the neutron star’ssurface.

Emission from the disk almost com-pletely drowns out the light from thecompanion star. That disparity impliesthat the companion must be fairly faint,which in turn indicates that its mass isno greater than that of the sun. Thesedouble-star systems are therefore knownas low-mass x-ray binaries. Solar-massstars remain stable for at least 10 billionyears, consistent with the age of the stel-lar population in which low-mass x-raybinaries reside.

Low-mass x-ray sources undergooccasional extreme flare-ups, or x-raybursts, which have yielded a great dealof information about these systems.Within a few seconds of the beginning ofa burst, the object’s x-ray brightness in-creases by a factor of 10 or more, peaksfor a few seconds to a few minutes andthen decays to the original level in abouta minute. X-ray bursts recur irregularlyevery few hours or so.

Researchers have deduced that thex-ray bursts result from runaway nu-clear fusion reactions in the gas accret-ed onto the surface of a neutron star.Between bursts, new matter flowingfrom the companion star replenishes thenuclear fuel. That steady accretion givesrise to the persistent emission of x-raysseen between the bursts. Despite thespectacular nature of the bursts, low-mass x-ray binaries emit more than 90percent of their total energy duringtimes of quiescence—a testimony to thegreat efficiency of accretion comparedwith fusion.


JAR

ED

SC

HN

EID

MAN

MAGNETIC FIELD of a young neutron star prevents infalling gas from reaching the surface, except at the two magnetic poles. Two hot, x-ray-emitting columns of gas, each about a kilometer across,collect at the poles. The star’s rotation axis is inclined with respect to its magnetic axis, so anobserver perceives regular pulses of x-rays as the magnetic poles rotate in and out of view.

Infallinggas

Magneticfield line

X-ray

Inner edge of accretion disk

Rotation

Magnetic axis


X-ray bursts occur only in low-massbinary star systems and x-ray pulses al-most solely in high-mass ones; not a sin-gle system displays both forms of be-havior. The critical factor responsiblefor this disparity is probably the strengthof the neutron star’s magnetic field.High-mass x-ray binaries must containneutron stars having powerful magneticfields, capable of generating easily de-tectable pulsations. Neutron stars inlow-mass x-ray binaries seem to possess

far weaker fields. This explanation isbolstered by theoretical models indicat-ing that a powerful magnetic field wouldinhibit the nuclear instabilities that pro-duce x-ray bursts.

The disparate characteristics of low-mass and high-mass x-ray binaries un-derscore the very different ways in whichthese systems must have formed andevolved. Almost immediately after thediscovery of high-mass x-ray binaries in1971, workers recognized that such ob-jects represent a normal stage in the evo-lution of close double-star systems inwhich both objects have more than afew times the mass of the sun. The moremassive star quickly consumes its fueland expands into a bloated red giant,whose outer layers spill over onto thecompanion star, exposing the red giant’shelium-rich center. A few hundred thou-sand years later, this helium star ex-plodes as a supernova, shedding much ofits outer mass; its remnant core collaps-es into a neutron star. The neutron starattracts gas from its companion and be-comes a source of x-rays.

The formation of a low-mass x-raybinary involves a more specialized set ofcircumstances. Some of these binariescould have started out as a massive starand a stellar lightweight orbiting eachother. The small companion star wouldhave too little gravity to capture materi-al from the primary star. When the pri-mary annihilates itself as a supernova,much of the system’s mass would escape

into interstellar space. In most cases, thatloss of mass would disrupt the binaryand send the two stars sailing off on sep-arate courses. In the rare instance inwhich the stars remain bound to eachother, they could evolve into a low-massx-ray binary.

There is also a gentler way. If the pri-mary star initially has less than eighttimes the mass of the sun, it will notblow up. Instead it will produce a whitedwarf, a stellar cinder far denser than a

normal star but much less so than a neu-tron star. In a white dwarf the star’sgravity has crushed its constituent atomsinto a soup of electrons and nuclei; awhite dwarf having the mass of the sunwould be about the size of Earth.

As the low-mass star evolves, it willgradually expand; if the two stars are ina close orbit, gas from the low-mass starwill accrete onto the surface of the whitedwarf. The mass of the white dwarf mayeventually exceed a critical value, about1.4 solar masses, and collapse into aneutron star. This kind of quiet collapseejects very little material, so the systemcan remain tightly bound. Later, thestars spiral in closer toward each other,accretion begins and the system becomesa low-mass x-ray binary.

In such binaries, the neutron star’sgravity exerts a strong pull on its muchlarger but less massive companion. Thecombination of gravitational and cen-trifugal forces gives rise to a pear-shapedregion of stability, called a Roche lobe,surrounding the low-mass star. Any ma-terial lying outside the Roche lobe willflow toward the neutron star. The trans-fer of material causes the distance be-tween the two stars to increase if themass-losing star is the less massive of thetwo, as is the case in low-mass x-ray bi-naries. When the size of the orbit in-creases, so does the size of the Rochelobe. Once the lobe grows bigger thanthe companion star, the flow of matterceases and the neutron star stops emit-

ting x-rays. Evidently, some mechanismkeeps feeding gas to the neutron star.

In one class of low-mass x-ray bina-ries—tightly bound systems whose peri-ods are less than about 10 hours—theflow of gas is maintained by a steadyshrinking of the stars’ mutual orbit. Asthe stars orbit, they shed gravitationalwaves that carry off angular momen-tum, which causes the stars to draw clos-er together. That effect negates the ten-dency of mass transfer to move the stars

apart. The stars ultimately settle into aslowly shrinking orbit in which a steadytrickle of gas migrates from the com-panion to the neutron star. In this way,the neutron star accretes about onethousandth of an Earth mass a year, suf-ficient to account for the observed lumi-nosity of many low-mass x-ray binaries(about 3 × 1030 watts).

The brightest x-ray sources in thecentral regions of the galaxy emit about10 times that much energy. These ob-jects constitute a second class of low-mass x-ray binaries that have relativelylong orbital periods of approximatelyone to 10 days. Such leisurely orbits im-ply that the separation between the twostars, as well as the diameter of the nor-mal companion, must be quite large.Here the flow of matter must result fromthe swelling of the companion star as aconsequence of physical changes in itsinterior.

Such changes occur in the later evo-lutionary stages of a sunlike star. Hy-drogen fusion produces helium, whichaccumulates as a dense core; hydrogenfusion takes place in a shell around thiscore. As the star ages, the hydrogen-burning shell migrates outward, causingthe star’s outer envelope to expand andcool. That expansion more than com-pensates for the increasing distance be-tween the stars caused by the transfer ofangular momentum. X-ray binaries hav-ing a period of about five to 10 daysreach equilibrium if they experience a


Pound for pound, accretion unleashes 15 to 60 timesAS MUCH ENERGY AS HYDROGEN FUSION.



mass transfer rate of about five thou-sandths of an Earth mass a year, aboutthe rate required to power the brightsources around the galactic center.

In 1982 a group of researchers—

Ronald F. Webbink of the University ofIllinois, Saul A. Rappaport of the Mass-achusetts Institute of Technology, G. J.Savonije of the University of Amsterdamand Ronald E. Taam of NorthwesternUniversity—investigated the fate of theselow-mass x-ray binaries. Their calc-ulations predict that, regardless of theirinitial traits, these systems always reachthe same evolutionary end point. The gi-ant star soon loses its entire hydrogen-rich envelope; its naked helium core re-mains as a white dwarf containing be-tween 0.25 and 0.45 solar mass. Thestars’ final orbit is extremely circular be-cause of the tens of millions of years oftidal interaction between the neutronstar and its low-mass partner.

After the supply of accreting materi-al dries up, binary star systems no long-er emit detectable amounts of x-rays.The last evolutionary stages nonethelessoffer a fascinating glimpse at what hap-pens to very old neutron stars. Duringthese later phases, the neutron star’smost distinctive emission is in the formof radio waves, not x-rays.

In 1983, while working on the 300-meter radio telescope in Arecibo, PuertoRico, Valentin Boriakoff of Cornell Uni-versity, Rosolino Buccheri of the ItalianNational Research Council in Palermoand Franco Fauci of the University ofPalermo discovered the binary radio pul-sar PSR 1953+29. Its properties closely


LOW-MASS X-RAY BINARY initially consists ofa neutron star pulling material from itscompanion (a). The low-mass star is anelderly subgiant having a dense, inerthelium core. The transfer of mass causesthe stars’ orbit to widen. At the same time,the low-mass star steadily expands andcools as it evolves (b). The neutron stargradually consumes the subgiant’s outerenvelope (c). The exposed helium core (now considered a white dwarf) remains in a circular orbit around the neutron star (d).The rotating neutron star is now a millisecondpulsar that emits pulses of radio waves butno x-rays. (This scenario is based oncalculations originally made by Paul C. Jossand Saul A. Rappaport of M.I.T.)

Accretion disk

Hydrogen-burning shell

Helium core

Millisecondpulsar

Evolution of a Low-Mass X-ray Binary

White dwarf

a

b Time = 45 million years

d

c

100 solarradii

1 solar mass

SUBGIANT STAR

1 solar mass

0.6 solar mass

0.31 solar mass

1.69 solar masses

0.31 solar mass

1.69 solar masses

NEUTRON STAR

Transferof mass

Time = 0

Time = 80 million years

Time = 81 million years

Radio waves

X-rays

1.4 solar masses

JAR

ED

SC

HN

EID

MAN


resemble those of the extinct x-ray bina-ries modeled by Webbink and his col-leagues. The pulsar’s radio signals dis-played no signs of the eclipses or ab-sorption produced by normal stars. Theresearchers recognized that the pulsar’scompanion must itself be a compact ob-ject. Because of its low mass, it is prob-ably a white dwarf.

Superfast SpinnersONE OF THE MOST surprising as-pects of PSR 1953+29 is its period of ra-dio pulsation: a remarkably swift 6.1milliseconds, or 160 rotations a second.Half a year before the discovery of PSR1953+29, Donald C. Backer of the Uni-versity of California at Berkeley and hisco-workers had found another pulsar,PSR 1937+21, which has a period ofonly 1.6 milliseconds. Astronomers nowrecognize these objects as the prototypesof a category of rapidly spinning neu-tron stars known as millisecond pulsars;70 have been found since.

The inferred history of x-ray binariesmade it clear why these pulsars spin soquickly. In low-mass x-ray binaries (andin many massive x-ray binaries as well),orbital motion prevents matter fromfalling directly onto the neutron star. In-stead it goes into orbit about the star,forming an accretion disk. Materialfrom the disk’s inner edge falls onto theneutron star. During the later stages ofaccretion, that infalling material greatlyspeeds up the star’s rotation.

Nearly all binary radio pulsars pos-sess companions that have evolved intowhite dwarfs or neutron stars. At some


MASSIVE X-RAY BINARY contains a brightblue star and an accreting neutron star (a).The blue star expands until its outerenvelope engulfs both its helium-rich coreand the neutron star (b). The orbital motionof the two stars inside the envelope heatsthe envelope and blows it away, leavingbehind a helium star and a neutron star (c).If the helium star has more than 2.5 solarmasses, it explodes as a supernova (d) andforms a second pulsing neutron star. Theexplosion may disrupt the system (e1);otherwise, the result is two neutron starslocked in a rapid, eccentric orbit (e2). Less massive helium cores do not explode;they end up as white dwarfs in circularorbits about the neutron star.

Evolution of a High-Mass X-ray Binary

a

b Time = 20,000 years

c

e 1

e 2

Helium core

NEUTRON STAR BLUE

GIANTSTAR

HELIUM STAR

25 solarradii

d

15 solar masses1.4 solar masses

SUPERNOVA

Runawaypulsars

Binary pulsar

1.4 solarmasses

Accretion disk

Dispersal of common envelope

Time = 0

Time = 40,000 years

Time = 500,000 years

4 solar masses

X-rays

Radiowaves

1.4 solarmasses

NEUTRON STAR

JAR

ED

SC

HN

EID

MAN



stage, those companions were giantsthat overflowed their Roche lobes anddumped matter onto the neutron stars,increasing the stars’ rate of rotation.During that time, the double stars wouldhave appeared as x-ray binaries. Afterthe companion star lost its outer layersand the accretion process ceased, anaked millisecond pulsar remained.

The power of a pulsar’s radio emis-

sion varies in proportion to the fourthpower of the rate of rotation. Millisec-ond neutron stars can be detected onlybecause they were “spun up” by theircompanions during the x-ray binaryphase. Radio pulsars that acquired theirrapid rotation in this way are now calledrecycled pulsars, a term suggested by V.Radhakrishnan of the Raman ResearchInstitute in Bangalore, India.

Starting in 1987, a number of ob-servers found that globular clusters areincredibly rich hunting grounds for bi-nary and millisecond pulsars. Studies ofglobular clusters have already revealedmore than 60 radio pulsars; 70 percentof these pulsars rotate in less than 10milliseconds, indicating that they are re-cycled. This celestial bounty results fromthe extremely dense nature of globularclusters. In their central regions, theseclusters may contain more than 10,000stars per cubic light-year, a million timesthe density of stars in the sun’s corner ofthe galaxy. Under such crowded condi-tions, neutron stars face good odds ofpassing close to and capturing a stellarcompanion. Globular clusters harbor200 to 1,000 times as many x-ray bina-ries per million stars as does the galaxyas a whole.

The “recycling” model for the originof millisecond radio pulsars, first pro-posed in 1982, was elegantly confirmedin 1998 with the discovery of the firstmillisecond x-ray pulsar in the low-mass x-ray binary SAXJ 1808.4-3658.Rudy Wijnands and Michiel van derKlis of the University of Amsterdam,working with NASA’s Rossi X-ray Tim-ing Explorer Satellite, found that this x-ray source is an accreting neutron starrotating 400 times a second. Since then,four more millisecond x-ray pulses in low-mass x-ray binaries have been discovered.

In addition to the binary radio pul-sars discussed, astronomers have iden-tified another, rarer class of such objectsthat have substantially different charac-teristics. Their orbits often are extreme-ly eccentric, and their companions con-tain 0.8 to 1.4 times the mass of the sun.

These objects probably arose fromhigh-mass x-ray binaries in the follow-ing way. In massive x-ray binaries, ac-cretion causes the two stars to spiral evercloser together (the opposite of the situ-ation for low-mass x-ray binaries). Thatprocess, combined with the swelling ofthe companion star as it evolves, causesthe companion to overflow its Rochelobe completely, engulfing the neutronstar. Frictional drag quickly sends theneutron star spiraling in toward its com-panion. At a certain point, the frictiongenerates so much heat that it drives offthe gaseous hydrogen envelope. Whatremains is the neutron star in a close or-bit around the stripped core of the com-panion, which consists of helium andheavier elements.

If the heavy-element core is suffi-ciently massive, it will later detonateinto a supernova and produce a secondneutron star. The force of the explosionand the precipitous loss of mass causethe stars’ orbit to become elliptical; in many cases, the stars break free entirely to become runaway radio pul-sars. If the orbit survives, the neutronstars follow their eccentric courses al-most forever; over the ages, their orbitswill slowly narrow because of the emis-sion of gravitational waves. One of the most thoroughly studied binary pul-sars, PSR 1913+16, consists of two neu-tron stars that race through a highly el-liptical orbit once every seven hours and45 minutes. This system’s extremeproperties allow it to serve as a sensitivetest bed for many aspects of Einstein’s


JAR

ED

SC

HN

EID

MAN

ROTATION OF NEUTRON STAR is stronglyinfluenced by accretion in an x-ray binary. Theneutron star’s magnetic field defines the inneredge of the surrounding disk of matter, wheregas falls onto the star. When the star is young,its field is strong, so the inner edge of the disk isdistant and comparatively slow-moving (a). Asthe magnetic field decays, the inner edge of theaccretion disk moves inward (b). The star nowaccretes rapidly moving material that causesits rate of rotation to increase. By the time theaccretion ceases, the neutron star may berotating hundreds of times per second (c).

Rotation period:0.005 second

Radio waves

Rotation period:1 second

Neutronstar

b

c

Rotation period:0.1 second

Magneticfield

a

EDWARD P. J. VAN DEN HEUVEL and JAN VAN PARADIJS collaborated on the study of celes-tial x-ray sources from the late 1970s until van Paradijs’s death in 1999. Van den Heuvelreceived his Ph.D. in mathematical and physical science from the University of Utrecht inthe Netherlands. In 1974 he joined the faculty of the University of Amsterdam, where heis now chairman of the astronomy department. He is also a co-founder and the director ofthe Center for High-Energy Astrophysics, operated jointly by the University of Amsterdamand the University of Utrecht. Van Paradijs earned his Ph.D. in astronomy from the Univer-sity of Amsterdam and became a professor of astronomy at the university in 1988.

THE

AU

THO

RS


theory of relativity, as Joseph Taylor ofPrinceton University has beautifullydemonstrated.

Particularly noteworthy is Taylor’sdiscovery that the very precisely mea-sured rate of shrinking of this system’sorbit is exactly as expected from theemission of gravitational waves as pre-dicted by Einstein’s theory. For this find-ing, Taylor and his former student Rus-sell Hulse were awarded the 1993 NobelPrize in Physics.

Studies of binary pulsars have over-turned a long-standing idea about how

neutron stars change over the eons. Sta-tistical analyses of pulsars had led mostastronomers to conclude that a neutronstar’s magnetic field decays withoutany outside assistance and in due timevanishes completely. The existence ofrecycled pulsars proved, however, thatsome magnetic field persists even in ex-tremely old systems. Moreover, thecompanion stars in binary pulsars offera way to determine just how old thosestars are.

Three of the millisecond pulsars haveobservable white dwarf companions,which serve as natural chronometers. Awhite dwarf steadily radiates away theheat left behind from its days as the coreof a red giant star. Over the eons, whitedwarfs grow progressively cooler andredder; the color of a white dwarf there-fore betrays its age.

In 1986, Shrinivas R. Kulkarni ofthe California Institute of Technologymeasured the color of the white dwarfcompanion to PSR 0655+64 and con-cluded it must be at least 500 millionyears old. Using similar reasoning, threesets of researchers—led by J. F. Bell ofthe Mount Stromlo and Siding SpringObservatories; John Danziger of the European Southern Observatory; andCharles D. Bailyn of Yale University—

determined that the white dwarf in the

binary pulsar system PSR J0437-4715is about two billion years old. The pul-sars in these systems must be consider-ably older because they would haveformed long before their companionsevolved into white dwarfs, yet they re-tain substantial magnetic fields—or theycould not be detected.

Magnetic No MoreWORK BY Frank Verbunt, Ralph A.M.J.Wijers and Hugo Burm of the Center forHigh-Energy Astrophysics in the Nether-lands further demonstrates the persis-

tence of neutron-star magnetic fields.The researchers studied three anom-alous, low-mass x-ray binaries that alsoare x-ray pulsars, indicating that theyeach contain a strongly magnetized neu-tron star. No matter how a neutron staroriginates, it always loses at least a fewtenths of a solar mass in the form of neu-trinos. When this happens, the binarysystem widens, shutting off the flow ofgas. Accretion cannot occur until the bi-nary system has shrunk through theemission of gravitational radiation oruntil the companion star has begun toevolve into a giant.

Both these mechanisms require con-siderable time to take effect. This knowl-edge allowed Verbunt to set lower limitsto the ages of the accreting neutron starsin low-mass x-ray binary pulsars. In thecase of Hercules X-1, the strongly mag-netized neutron star is at least 500 mil-lion years old. Neutron stars’ magneticfields evidently do not spontaneously de-cay, at least not on such timescales.

And yet the magnetic fields of almost

all binary radio pulsars are 100 to10,000 times weaker than those of nor-mal, youthful radio pulsars, regardless ofwhether the binary pulsar descendedfrom a high-mass or a low-mass x-ray bi-nary. The weakness of their fields seemsto be attributable to some factor that allbinary pulsars have in common. Themost obvious common factor is accre-tion. In 1986 Taam and one of us (vanden Heuvel) proposed, on observationalgrounds, a link between field decay andaccretion. Theorists then advanced sev-eral models to explain the details.

One model holds that newly accret-ed layers on the surface of a neutron starform an electrically conductive layer thatallows only a small fraction of the star’smagnetic field to reach the outside. An-other possibility, proposed by G. Shrini-vasan of the Raman Research Institute,is that it is the gradual slowing of a neu-tron star’s rotation that causes its mag-netic field to dissipate. Such decelerationoccurs before and during the early stagesof accretion. Once the magnetic field hasweakened below a critical threshhold,the action of accretion reverses the spin-down trend, but that infusion of rota-tional energy cannot restore the field toits original strength.

In any case, there is every indicationthat millisecond radio pulsars will retaintheir fields and continue to pulse untoldbillions of years into the future. Thus it happens that long after their era offlamboyant x-ray emission, x-ray bina-ries settle down to become some of themost steady, unchanging entities in thecosmos.


A New Class of Pulsars. Donald C. Backer and Shrinivas R. Kulkarni in Physics Today, Vol. 43, No. 3, pages 26–35; March 1990.Pulsars Today. Francis Graham Smith in Sky and Telescope, Vol. 80, No. 3, pages 240–244;September 1990.X-Ray Binaries. Edited by W.H.G. Lewin, J. A. van Paradijs and E.P.J. van den Heuvel. Cambridge University Press, 1995.


LONG AFTER THEIR ERA OF FLAMBOYANT emission, x-ray binaries become the most

STEADY ENTITIES IN THE COSMOS.



STARQUAKE ON A MAGNETAR releases a vast amount of magnetic energy—equivalent to the seismic energy of a magnitude 21 earthquake—andunleashes a fireball of plasma. The fireballgets trapped by the magnetic field.


On March 5, 1979, several months afterdropping probes into the toxic atmosphereof Venus, two Soviet spacecraft, Venera 11and 12, were drifting through the inner so-lar system on an elliptical orbit. It had beenan uneventful cruise. The radiation read-ings onboard both craft hovered around anominal 100 counts per second. But at10:51 A.M. EST, a pulse of gamma radia-tion hit them. Within a fraction of a mil-lisecond, the radiation level shot above200,000 counts per second and quicklywent off scale.

Eleven seconds later gamma raysswamped the NASA space probe Helios 2,also orbiting the sun. A plane wave front ofhigh-energy radiation was evidently sweep-ing through the solar system. It soonreached Venus and saturated the PioneerVenus Orbiter’s detector. Within seconds M

AGNE

TARS

Some stars aremagnetized sointensely that they emithuge bursts of magneticenergy and alter thevery nature of thequantum vacuum

By Chryssa Kouveliotou, Robert C. Duncan and Christopher Thompson


the gamma rays reached Earth. They flooded detectors on threeU.S. Department of Defense Vela satellites, the Soviet Prognoz 7satellite and the Einstein Observatory. Finally, on its way outof the solar system, the wave also blitzed the International Sun-Earth Explorer.

The pulse of highly energetic, or “hard,” gamma rays was100 times as intense as any previous burst of gamma rays de-tected from beyond the solar system, and it lasted just twotenths of a second. At the time, nobody noticed; life continuedcalmly beneath our planet’s protective atmosphere. Fortunate-ly, all 10 spacecraft survived the trauma without permanentdamage.

The hard pulse was followed by a fainter glow of lower-en-ergy, or “soft,” gamma rays, as well as x-rays, which steadilyfaded over the subsequent three minutes. As it faded away, thesignal oscillated gently, with a period of eight seconds. Fourteenand a half hours later, at 1:17 A.M. on March 6, another, fainterburst of x-rays came from the same spot on the sky. Over theensuing four years, Evgeny P. Mazets of the Ioffe Institute in St.Petersburg, Russia, and his collaborators detected 16 burstscoming from the same direction. They varied in intensity, butall were fainter and shorter than the March 5 pulse.

Astronomers had never seen anything like this. For want ofa better idea, they initially listed these bursts in cataloguesalongside the better-known gamma-ray bursts (GRBs), eventhough they clearly differed in several ways. In the mid-1980sKevin C. Hurley of the University of California at Berkeley re-alized that similar outbursts were coming from two other ar-eas of the sky. Evidently these sources were all repeating—un-like GRBs, which are one-shot events [see “The Brightest Ex-plosions in the Universe,” on page 92]. At a July 1986 meetingin Toulouse, France, astronomers agreed on the approximatelocations of the three sources and dubbed them “soft gammarepeaters” (SGRs). The alphabet soup of astronomy hadgained a new ingredient.

Another seven years passed before two of us (Duncan andThompson) devised an explanation for these strange objects, andonly in 1998 did one of us (Kouveliotou) and her team find com-

pelling evidence for that explanation. Recent observations con-nect our theory to yet another class of celestial enigmas, knownas anomalous x-ray pulsars (AXPs). These developments haveled to a breakthrough in our understanding of one of the mostexotic members of the celestial bestiary, the neutron star.

Neutron stars are the densest material objects known, pack-ing slightly more than the sun’s mass inside a ball just 20 kilo-meters across. Based on the study of SGRs, it seems that someneutron stars have magnetic fields so intense that they radical-ly alter the material within them and the state of the quantumvacuum surrounding them, leading to physical effects observednowhere else in the universe.

Not Supposed to Do ThatBECAUSE THE MARCH 1979 BURST was so bright, theo-rists at the time reckoned that its source was in our galacticneighborhood, hundreds of light-years from Earth at most. Ifthat had been true, the intensity of the x-rays and gamma rayswould have been just below the theoretical maximum steadyluminosity that can be emitted by a star. That maximum, firstderived in 1926 by English astrophysicist Arthur Eddington, isset by the force of radiation flowing through the hot outer lay-ers of a star. If the radiation is any more intense, it will over-power gravity, blow away ionized matter and destabilize thestar. Emission below the Eddington limit would have been fair-ly straightforward to explain. For example, various theoristsproposed that the outburst was triggered by the impact of achunk of matter, such as an asteroid or a comet, onto a near-by neutron star.

But observations soon confounded that hypothesis. Eachspacecraft had recorded the time of arrival of the hard initialpulse. These data allowed astronomers, led by Thomas LyttonCline of the NASA Goddard Space Flight Center, to triangulatethe burst source. The researchers found that the position co-incided with the Large Magellanic Cloud, a small galaxy about170,000 light-years away. More specifically, the event’s posi-tion matched that of a young supernova remnant, the glow-

70 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e F e b r u a r y 2 0 0 3 i s s u e

DO

N D

IXO

N (

pre

ced

ing

pa

ges

); E

. L.

WR

IGH

T U

niv

ersi

ty o

f C

ali

forn

ia,

Los

Ang

eles

, TH

E C

OB

E P

RO

JEC

T, D

IRB

E A

ND

NAS

A (a

bov

e)

■ Astronomers have seen a handful of stars that put outflares of gamma and x-radiation, which can be millions oftimes as bright as any other repeating outburst known.The enormous energies and pulsing signals implicate thesecond most extreme type of body in the universe (after the black hole): the neutron star.

■ These neutron stars have the strongest magnetic fieldsever measured—hence their name, magnetars. Magneticinstabilities analogous to earthquakes can account for the flares.

■ Magnetars remain active for only about 10,000 years,implying that millions of them are drifting undetectedthrough our galaxy.

Overview/Ultramagnetic Stars

0142+61

2259+586 1900+14

1844–0258

1841–045 1708–40

1806–20 1627–41

MAGNETAR CANDIDATES

TWELVE POSSIBLE magnetars have beendetected in or near our Milky Way galaxy.

1810–197


ing remains of a star that exploded 5,000 years ago. Unless thisoverlap was pure coincidence, it put the source 1,000 times asfar away as theorists had thought—and thus made it a milliontimes brighter than the Eddington limit. In 0.2 second theMarch 1979 event released as much energy as the sun radiatesin roughly 10,000 years, and it concentrated that energy ingamma rays rather than spreading it across the electromagneticspectrum.

No ordinary star could account for such energy, so thesource was almost certainly something out of the ordinary—

either a black hole or a neutron star. The former was ruled outby the eight-second modulation: a black hole is a featureless ob-

ject, lacking the structure needed to produce regular pulses. Theassociation with the supernova remnant further strengthenedthe case for a neutron star. Neutron stars are widely believed toform when the core of a massive but otherwise ordinary star ex-hausts its nuclear fuel and abruptly collapses under its ownweight, thereby triggering a supernova explosion.

Identifying the source as a neutron star did not solve the puz-zle; on the contrary, it merely heightened the mystery. Astrono-mers knew several examples of neutron stars that lie within su-pernova remnants. These stars were radio pulsars, objects thatare observed to blink on and off in radio waves. Yet the March1979 burster, with an apparent rotation period of eight seconds,was spinning much more slowly than any radio pulsar thenknown. Even when not bursting, the object emitted a steadyglow of x-rays with more radiant power than could be suppliedby the rotation of a neutron star. Oddly, the star was significantlydisplaced from the center of the supernova remnant. If it wasborn at the center, as is likely, then it must have recoiled with avelocity of about 1,000 kilometers per second at birth. Such highspeed was considered unusual for a neutron star.

Finally, the outbursts themselves seemed inexplicable. X-rayflashes had previously been detected from some neutron stars,but they never exceeded the Eddington limit by very much. As-tronomers ascribed them to thermonuclear fusion of hydrogen

or helium or to the sudden accretion of matter onto the star. Butthe brightness of the SGR bursts was unprecedented, so a newphysical mechanism seemed to be required.

Spin Forever DownTHE FINAL BURST FROM the March 1979 source was de-tected in May 1983; none has been seen since. Two other SGRs,both within our Milky Way galaxy, went off in 1979 and haveremained active, emitting hundreds of bursts in subsequentyears. A fourth SGR was located in 1998. Three of these fourobjects have possible, but unproved, associations with youngsupernova remnants. Two also lie near very dense clusters of

massive young stars, intimating that SGRs tend to form fromsuch stars. A fifth candidate SGR has gone off only twice; itsprecise location is still unknown.

As Los Alamos National Laboratory scientists Baolian L.Cheng, Richard I. Epstein, Robert A. Guyer and A. CodyYoung pointed out in 1996, SGR bursts are statistically simi-lar to earthquakes. The energies have very similar mathemati-cal distributions, with less energetic events being more common.Our then graduate student Ersin Gögüs of the University of Al-abama at Huntsville verified this behavior for many bursts fromvarious sources. This and other statistical properties are a hall-mark of self-organized criticality, whereby a composite systemattains a critical state in which a small perturbation can triggera chain reaction. Such behavior occurs in systems as diverse asavalanches on sandpiles and magnetic flares on the sun.

But why would a neutron star behave like this? The solu-tion emerged from an entirely separate line of work, on radiopulsars. Pulsars are widely thought to be rapidly rotating, mag-netized neutron stars. The magnetic field, which is supportedby electric currents flowing deep inside the star, rotates withthe star. Beams of radio waves shine outward from the star’smagnetic poles and sweep through space as it rotates, like light-house beacons—hence the observed pulsing. The pulsar alsoblows out a wind of charged particles and low-frequency elec-


SOU

RC

ES

FOR

TAB

LE:

CH

RYS

SA K

OU

VELI

OTO

U,

RO

BE

RT

C.

DU

NC

AN A

ND

CH

RIS

TOP

HE

R T

HO

MP

SON

Even

ts p

er 2

0 Da

ys

INTE

NSIT

Y

Minutes

SGR 1900+14 FLARE

Flare

SGR 1900+14 ACTIVITY

20

0

40

60

1980Year

1990 200010 2 3 4 5

GIANT X-RAY FLARE in August 1998 confirmed the existenceof magnetars. It started with a spike of radiation lasting lessthan a second (left). Then came an extended train of pulseswith a period of 5.16 seconds. This event was the mostpowerful outburst to come from the object, designated SGR1900+14, since its discovery in 1979 (right).

0526–660110–72

1048–59

ROTATION PERIOD

(seconds)YEAR OF

DISCOVERYNAME

SGR 0526–66 1979 8.0SGR 1900+14 1979 5.16SGR 1806–20 1979 7.47SGR 1801–23* 1997 ?SGR 1627–41 1998 ?AXP 1E 2259+586 1981 6.98AXP 1E 1048–59† 1985 6.45AXP 4U 0142+61 1993 8.69AXP 1RXS 1708–40† 1997 11.0AXP 1E 1841–045 1997 11.8AXP AXJ1844–0258 1998 6.97AXP CXJ0110–7211† 2002 5.44AXP XTE J1810–197 2003 5.54* N o t o n m a p ; l o c a t i o n n o t k n o w n p r e c i s e l y † A b b r e v i a t e d n a m e



tromagnetic waves, which carry away energy and angular mo-mentum, causing its rate of spin to decrease gradually.

Perhaps the most famous pulsar lies within the Crab Neb-ula, the remnant of a supernova explosion that was observed in1054. The pulsar rotates once every 33 milliseconds and is cur-rently slowing at a rate of about 1.3 millisecond every century.Extrapolating backward, it was born rotating once every 20milliseconds. Astronomers expect it to continue to spin down,eventually reaching a point when its rotation will be too slowto power the radio pulses. The spin-down rate has been mea-sured for almost every radio pulsar, and theory indicates thatit depends on the strength of the star’s magnetic field. From this,most young radio pulsars are inferred to have magnetic fieldsbetween 1012 and 1013 gauss. For comparison, a refrigeratormagnet has a strength of about 100 gauss.

The Ultimate Convection OvenTHIS PICTURE LEAVES a basic question unanswered: Wheredid the magnetic field come from in the first place? The tradi-tional assumption was: it is as it is, because it was as it was. Thatis, most astronomers supposed that the magnetic field is a relicof the time before the star went supernova. All stars have weakmagnetic fields, and those fields can be amplified simply by theact of compression. According to Maxwell’s equations of elec-

tromagnetism, as a magnetized object shrinks by a factor oftwo, its magnetic field strengthens by a factor of four. The coreof a massive star collapses by a factor of 105 from its birththrough neutron star formation, so its magnetic field should be-come 1010 times stronger.

If the core magnetic field started with sufficient strength, thiscompression could explain pulsar magnetism. Unfortunately,the magnetic field deep inside a star cannot be measured, so thissimple hypothesis cannot be tested. There are also good reasonsto believe that compression is only part of the story.

Within a star, gas can circulate by convection. Warm parcelsof ionized gas rise, and cool ones sink. Because ionized gas con-ducts electricity well, any magnetic field lines threading the gasare dragged with it as it moves. The field can thus be reworkedand sometimes amplified. This phenomenon, known as dynamoaction, is thought to generate the magnetic fields of stars andplanets. A dynamo might operate during each phase of the lifeof a massive star, as long as the turbulent core is rotating rapid-ly enough. Moreover, during a brief period after the core of thestar turns into a neutron star, convection is especially violent.

This was first shown in computer simulations in 1986 byAdam Burrows of the University of Arizona and James M.Lattimer of Stony Brook University. They found that temper-atures in a newborn neutron star exceed 30 billion kelvins.

TWO TYPES OF NEUTRON STARS

3B If the newborn neutronstar spins slowly, its

magnetic field, though strongby normal standards, doesnot reach magnetar levels.

5A The old magnetar hascooled off, and much

of its magnetism has decayed away. It emits very little energy.

3A If the newborn neutronstar spins fast enough,

it generates an intensemagnetic field. Field linesinside the star get twisted.

4A The magnetar settlesinto neat layers, with

twisted field lines inside andsmooth lines outside. It mightemit a narrow radio beam.

Age: above 10,000 yearsAge: 0 to 10,000 yearsAge: 0 to 10 seconds

Age: above 10 millionyears

Age: 0 to 10 million years

MAGNETAR

ORDINARY PULSAR

1Most neutron starsare thought to begin

as massive butotherwise ordinarystars, between eightand 20 times as heavyas the sun.

2Massive stars diein a type II

supernova explosion,as the stellar coreimplodes into a denseball of subatomicparticles.

4B The mature pulsar iscooler than a magnetar of

equal age. It emits a broadradio beam, which radiotelescopes can readily detect.

5B The old pulsar has cooled off and no longer

emits a radio beam.

Newbornneutron

star

Age: 0 to 10 seconds

72 S C I E N T I F I C A M E R I C A NCOPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

Hot nuclear fluid circulates in 10 milliseconds or less, carry-ing enormous kinetic energy. After about 10 seconds, the con-vection ceases.

Not long after Burrows and Lattimer conducted their firstsimulations, Duncan and Thompson, then at Princeton Uni-versity, estimated what this furious convection means for neu-tron star magnetism. The sun, which undergoes a sedate ver-sion of the same process, can be used as a reference point. Assolar fluid circulates, it drags along magnetic field lines and givesup about 10 percent of its kinetic energy to the field. If the mov-ing fluid in a newborn neutron star also transfers a tenth of itskinetic energy to the magnetic field, then the field would growstronger than 1015 gauss, which is more than 1,000 times asstrong as the fields of most radio pulsars.

Whether the dynamo operates globally (rather than in lim-ited regions) would depend on whether the star’s rate of rota-tion was comparable to its rate of convection. Deep inside thesun, these two rates are similar, and the magnetic field is ableto organize itself on large scales. By analogy, a neutron starborn rotating as fast as or faster than the convective period of10 milliseconds could develop a widespread, ultrastrong mag-netic field. In 1992 we named these hypothetical neutron stars“magnetars.”

An upper limit to neutron star magnetism is about 1017

gauss; beyond this limit, the fluid inside the star would tend tomix and the field would dissipate. No known objects in the uni-verse can generate and maintain fields stronger than this level.One ramification of our calculations is that radio pulsars areneutron stars in which the large-scale dynamo has failed to op-erate. In the case of the Crab pulsar, the newborn neutron starrotated once every 20 milliseconds, much slower than the rateof convection, so the dynamo never got going.

Crinkle Twinkle Little MagnetarALTHOUGH WE DID NOT develop the magnetar concept toexplain SGRs, its implications soon became apparent to us. Themagnetic field should act as a strong brake on a magnetar’s ro-tation. Within 5,000 years a field of 1015 gauss would slow thespin rate to once every eight seconds—neatly explaining the os-cillations observed during the March 1979 outburst.

As the field evolves, it changes shape, driving electric currentsalong the field lines outside the star. These currents, in turn, gen-erate x-rays. Meanwhile, as the magnetic field moves throughthe solid crust of a magnetar, it bends and stretches the crust.This process heats the interior of the star and occasionally breaksthe crust in a powerful “starquake.” The accompanying releaseof magnetic energy creates a dense cloud of electrons andpositrons, as well as a sudden burst of soft gamma rays—ac-counting for the fainter bursts that give SGRs their name.

More infrequently, the magnetic field becomes unstable andundergoes a large-scale rearrangement. Similar (but smaller) up-heavals sometimes happen on the sun, leading to solar flares.A magnetar easily has enough energy to power a giant flare suchas the March 1979 event. Theory indicates that the first half-second of that tremendous outburst came from an expandingfireball. In 1995 we suggested that part of the fireball wastrapped by the magnetic field lines and held close to the star. Thistrapped fireball gradually shrank and then evaporated, emittingx-rays all the while. Based on the amount of energy released, wecalculated the strength of the magnetic field needed to confinethe enormous fireball pressure: greater than 1014 gauss, whichagrees with the field strength inferred from the spin-down rate.

A separate estimate of the field had been given in 1992 byBohdan Paczynski of Princeton. He noted that x-rays can slip

STRUCTURE OF A NEUTRON STAR can be inferred from theories of nuclear matter.Starquakes can occur in the crust, a lattice of atomic nuclei immersed in a seaof electrons. The core consists mainly of neutrons and perhaps a lump ofquarks. An atmosphere of hot plasma might extend all of a few centimeters.

Quarks?

5 km

Inner crust

Outer crust

Core

Atmosphere

CHRYSSA KOUVELIOTOU, ROBERT C. DUNCAN and CHRISTOPHERTHOMPSON have studied magnetars for a collective 40 years andhave collaborated for the past six. Kouveliotou, an observer,works at the National Space Science and Technology Center inHuntsville, Ala. Besides soft-gamma repeaters, her pets includegamma-ray bursts, x-ray binaries and her cat, Felix; her interestsrange from jazz to archaeology to linguistics. Duncan and Thomp-son are theorists, the former at the University of Texas at Austin,the latter at the Canadian Institute for Theoretical Astrophysicsin Toronto. Duncan has studied supernovae, quark matter and in-tergalactic gas clouds. In his younger days he ran a 2:19 marathonin the 1980 U.S. Olympic trials. Thompson has worked on topicsfrom cosmic strings to giant impacts in the early solar system.He, too, is an avid runner as well as a backpacker.

THE

AU

THO

RS

DO

N D

IXO

N

S C I E N T I F I C A M E R I C A N 73COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

through a cloud of electrons more easily if the charged particlesare immersed in a very intense magnetic field. For the x-raysduring the burst to have been so bright, the magnetic field musthave been stronger than 1014 gauss.

What makes the theory so tricky is that the fields arestronger than the quantum electrodynamic threshold of 4 ×1013 gauss. In such strong fields, bizarre things happen. X-rayphotons readily split in two or merge together. The vacuum it-self is polarized, becoming strongly birefringent, like a calcitecrystal. Atoms are deformed into long cylinders thinner thanthe quantum-relativistic wavelength of an electron [see box onopposite page]. All these strange phenomena have observableeffects on magnetars. Because this physics was so exotic, the the-ory attracted few researchers at the time.

Zapped AgainAS THESE THEORETICAL developments were slowly un-folding, observers were still struggling to see the objects thatwere the sources of the bursts. The first opportunity came whenNASA’s orbiting Compton Gamma Ray Observatory recordeda burst of gamma rays late one evening in October 1993. Thiswas the break Kouveliotou had been looking for when shejoined the Compton team in Huntsville. The instrument thatregistered the burst could determine its position only to withina fairly broad swath of sky. Kouveliotou turned for help to theJapanese ASCA satellite. Toshio Murakami of the Institute ofSpace and Astronautical Science in Japan and his collaboratorssoon found an x-ray source from the same swath of sky. Thesource held steady, then gave off another burst—proving be-yond all doubt that it was an SGR. The same object had firstbeen seen in 1979 and, based on its approximate celestial co-ordinates, was identified as SGR 1806–20. Now its positionwas fixed much more precisely, and it could be monitoredacross the electromagnetic spectrum.

The next leap forward came in 1995, when NASA launched

the Rossi X-ray Timing Explorer (RXTE), a satellite designedto be highly sensitive to variations in x-ray intensity. Using thisinstrument, Kouveliotou found that the emission from SGR1806–20 was oscillating with a period of 7.47 seconds—amaz-ingly close to the 8.0-second periodicity observed in the March1979 burst (from SGR 0526–66). Over the course of five years,the SGR slowed by nearly two parts in 1,000. Although theslowdown may seem small, it is faster than that of any radiopulsar known, and it implies a magnetic field approaching 1015

gauss.More thorough tests of the magnetar model would require a

second giant flare. Luckily, the heavens soon complied. In theearly morning of August 27, 1998, some 19 years after the giantflare that began SGR astronomy was observed, an even more in-tense wave of gamma rays and x-rays reached Earth from thedepths of space. It drove detectors on seven scientific spacecraftto their maximum or off scale. One interplanetary probe, NASA’sNear Earth Asteroid Rendezvous mission, was forced into a pro-tective shutdown mode. The gamma rays hit Earth on its night-side, with the source in the zenith over the mid-Pacific Ocean.

Fortuitously, in those early-morning hours electrical engi-neer Umran S. Inan and his colleagues from Stanford Universi-ty were gathering data on the propagation of very low fre-quency radio waves around Earth. At 3:22 A.M. PDT, they no-ticed an abrupt change in the ionized upper atmosphere. Theinner edge of the ionosphere plunged down from 85 to 60 kilo-meters for five minutes. It was astonishing. This effect on ourplanet was caused by a neutron star far across the galaxy,20,000 light-years away.

Another Magneto MarvelTHE AUGUST 27 FLARE was almost a carbon copy of theMarch 1979 event. Intrinsically, it was only one tenth as pow-erful, but because the source was closer to Earth it remains themost intense burst of gamma rays from beyond our solar sys-


DO

N D

IXO

N

HOW MAGNETAR BURSTS HAPPENThe magnetic field of the star is so strong that the rigid crust sometimes breaks and crumbles, releasing a huge surge of energy.

1Most of the time themagnetar is quiet.

But magnetic stresses are slowly building up.

2At some point the solid crustis stressed beyond its limit.

It fractures, probably into manysmall pieces.

3This “starquake” creates a surging electric current,

which decays and leaves behinda hot fireball.

4 The fireball cools byreleasing x-rays from

its surface. It evaporates in minutes or less.


tem ever detected. The last few hundred seconds of the flareshowed conspicuous pulsations, with a 5.16-second period.Kouveliotou and her team measured the spin-down rate of thestar with RXTE; sure enough, it was slowing down at a ratecomparable to that of SGR 1806–20, implying a similarlystrong magnetic field. Another SGR was placed into the mag-netar hall of fame.

The precise localizations of SGRs in x-rays have allowedthem to be studied using radio and infrared telescopes (thoughnot in visible light, which is blocked by interstellar dust). Thiswork has been pioneered by many astronomers, notably DaleFrail of the National Radio Astronomy Observatory and Shrini-vas R. Kulkarni of the California Institute of Technology. Oth-er observations have shown that all four confirmed SGRs con-tinue to release energy, albeit faintly, even between outbursts.“Faintly” is a relative term: this x-ray glow represents 10 to 100times as much power as the sun radiates in visible light.

By now one can say that magnetar magnetic fields are bet-ter measured than pulsar magnetic fields. In isolated pulsars, al-most the only evidence for magnetic fields as strong as 1012

gauss comes from their measured spin-down. In contrast, thecombination of rapid spin-down and bright x-ray flares pro-vides several arguments for 1014- to 1015-gauss fields in mag-netars. Alaa Ibrahim of the NASA Goddard Space Flight Cen-ter reported yet another line of evidence for strong magneticfields in magnetars: x-ray spectral lines that seem to be gener-ated by protons gyrating in a 1015-gauss field.

One intriguing question is whether magnetars are related tocosmic phenomena besides SGRs. The shortest-duration gam-ma-ray bursts, for example, have yet to be convincingly ex-plained, and at least a handful of them could be flares from mag-netars in other galaxies. If seen from a great distance, even agiant flare would be near the limit of telescope sensitivity. Onlythe brief, hard, intense pulse of gamma rays at the onset of theflare would be detected, so telescopes would register it as a GRB.

In the mid-1990s Thompson and Duncan suggested thatmagnetars might also explain anomalous x-ray pulsars, a classof objects that resemble SGRs in many ways. The one difficultywas that AXPs had not been observed to burst. But, Victoria M.Kaspi and Fotis P. Gavriil of McGill University and Peter M.Woods of the National Space and Technology Center inHuntsville detected bursts from two of the seven known AXPs.One of these objects is associated with a young supernova rem-nant in the constellation Cassiopeia.

Another AXP in Cassiopeia is the first magnetar candidateto have been detected in visible light. Ferdi Hulleman andMarten van Kerkwijk of Utrecht University in the Netherlands,working with Kulkarni, spotted it. Though exceedingly faint,the AXP fades in and out with the x-ray period of a neutronstar. These observations lend support to the idea that it is in-deed a magnetar. The main alternative—that AXPs are ordinaryneutron stars surrounded by disks of matter—predicts too muchvisible and infrared emission with too little pulsation.

In view of these discoveries, and the apparent silence of theLarge Magellanic Cloud burster for nearly 20 years, it appears

that magnetars can change their clothes. They can remain qui-escent for years, even decades, before undergoing sudden peri-ods of extreme activity. Some astronomers argue that AXPs areyounger on average than SGRs, but this is still a matter of de-bate. If both SGRs and AXPs are magnetars, then magnetarsplausibly constitute a substantial fraction of all neutron stars.

The story of magnetars is a sobering reminder of how muchwe have yet to understand about our universe. Thus far we havediscerned at most a dozen magnetars among the countless stars.They reveal themselves for a split second, in light that only themost sophisticated telescopes can detect. Within 10,000 years,their magnetic fields freeze and they stop emitting bright x-rays.So those dozen magnetars betray the presence of more than amillion, and perhaps as many as 100 million, other objects—oldmagnetars that long ago went dark. Dim and dead, these strangeworlds wander through interstellar space. What other phe-nomena, so rare and fleeting that we have not recognized them,lurk out there?


BR

YAN

CH

RIS

TIE

DE

SIG

N

Formation of Very Strongly Magnetized Neutron Stars: Implications forGamma-Ray Bursts. Robert C. Duncan and Christopher Thompson inAstronomical Journal, Vol. 392, No. 1, pages L9–L13; June 10, 1992.Available at http://adsabs.harvard.edu/abs/1992Apj...392L...9DAn X-ray Pulsar with a Superstrong Magnetic Field in the Soft Gamma-Ray Repeater SGR1806–20. C. Kouveliotou, S. Dieters, T. Strohmayer, J. Von Paradijs, G. J. Fishman, C. A. Meegan, K. Hurley, J. Kommers, I. Smith,D. Frail and T. Murakami in Nature, Vol. 393, pages 235–237; May 21, 1998.The Life of a Neutron Star. Joshua N. Winn in Sky & Telescope, Vol. 98, No. 1, pages 30–38; July 1999.Physics in Ultra-strong Magnetic Fields. Robert C. Duncan. Fifth Huntsville Gamma-Ray Burst Symposium, February 23, 2002.Available at arxiv.org/abs/astro-ph/0002442Flash! The Hunt for the Biggest Explosions in the Universe. Govert Schilling. Cambridge University Press, 2002.More information can be found at Robert C. Duncan’s Web site:solomon.as.utexas.edu/magnetar.html


V A C U U M B I R E F R I N G E N C EPolarized light waves (orange) change speed andhence wavelength when they enter a very strongmagnetic field (black lines).

S C A T T E R I N G S U P P R E S S I O NA light wave can glide past an electron (blackcircle) with little hindrance if the field preventsthe electron from vibrating with the wave.

P H O T O N S P L I T T I N GIn a related effect, x-rays freely split in two or merge together. This process is important in fields stronger than 1014 gauss.

D I S T O R T I O N O F A T O M SFields above 109 gauss squeeze electron orbitals into cigar shapes. In a 1014-gauss field, a hydrogen atom becomes 200 times narrower.

EXTREME MAGNETISMMagnetar fields wreak havoc with radiation and matter.



Several years ago astronomers came across a new type of star thatspews out unusually low energy x-rays. These so-called supersoftsources are now thought to be white dwarf stars that cannibalizetheir stellar companions and then, in many cases, explode By Peter Kahabka, Edward P. J. van den Heuvel and Saul A. Rappaport

DAVID AND GOLIATH STARS form a symbiotic binary system: a whitedwarf and a red giant star in mutualorbit. The dwarf, with its intensegravity, is slurping off the outer layersof the giant. The pilfered gas goes intoan accretion disk around the dwarf andeventually settles onto its surface,whereupon it can ignite nuclear fusionand generate a large quantity of low-energy x-rays.

Supersoft X-ray Stars


Since the 1930s astronomers have known that ordinary stars shine because of nuclear fu-sion deep in their interior. In the core of the sun, for example, 600 million tons of hydro-gen fuse into helium every second. This process releases energy in the form of x-rays and

gamma rays, which slowly wend their way outward through the thick layers of gas. By the timethe radiation reaches the surface of the star, it has degraded into visible light.

Recently, however, researchers have discovered anew class of stars in which the nuclear fusion takes placenot in the deep interior but in the outer layers just belowthe surface. These stars appear to be white dwarfs—dense, burned-out stars that have exhausted their nuclearfuel—in orbit around ordinary stars. The dwarfs steal hy-drogen gas from their companions, accumulate it on theirsurface and resume fusion. The result is a torrent of x-rays with a distinctive “soft” range of wavelengths; suchstars are known as luminous supersoft x-ray sources. Asthe dwarfs gain weight, they eventually grow unstable,at which point they can collapse into an even denser neu-tron star or explode.

The disruption of white dwarfs has long been con-jectured as the cause of one sort of supernova explosion,called type Ia. With the discovery of the supersoftsources, observers have identified for the first time a classof star system that can detonate in this way. Type Ia su-pernovae have become important as bright “standardcandles” for measuring distances to faraway galaxies andthereby the pace of cosmic expansion. Much of the lin-

gering uncertainty in estimates of the age and the ex-pansion rate of the universe is connected to as-

tronomers’ ignorance of what gives rise to thesesupernovae. Supersoft sources may be one of

the long-sought missing links.The story of the supersoft sources

began with the launch of the Ger-man x-ray satellite ROSAT in

1990. This orbiting observa-tory carried out the first

complete survey of the

sky in soft x-rays, a form of electromagnetic radiationthat straddles ultraviolet light and the better-known“hard” x-rays. Soft x-rays have wavelengths that are onethousandth to one fiftieth those of visible light—whichmeans that the energy of their photons (the unit x-ray as-tronomers prefer to think in) is between about 0.09 and2.5 kiloelectron volts (keV). Hard x-rays have energiesup to a few hundred keV. With the exception of the Na-tional Aeronautics and Space Administration’s orbitingEinstein Observatory, which covered the energy rangefrom 0.2 to 4.0 keV, previous satellites had concentratedon the hard x-rays.

Almost immediately the ROSAT team, led by JoachimTrümper of the Max Planck Institute for ExtraterrestrialPhysics near Munich, noticed some peculiar objects dur-ing observations of the Large Magellanic Cloud, a smallsatellite galaxy of the Milky Way. The objects emitted x-rays at a prodigious rate—some 5,000 to 20,000 times thetotal energy output of our sun—but had an unexpected-ly soft spectrum. Bright x-ray sources generally have hardspectra, with peak energies in the range of 1 to 20 keV,which are produced by gas at temperatures of 10 millionto 100 million kelvins. These hard x-ray sources representneutron stars and black holes in the process of devour-ing their companion stars [see “X-ray Binaries,” on page58]. But the soft spectra of the new stars—with photonenergies a hundredth of those in other bright x-raysources—implied temperatures of only a few hundredthousand kelvins. On an x-ray color picture, the objectsappear red, whereas classical, hard x-ray sources lookblue [see illustration at bottom left of next page].

The reason the supersoft sources had not been rec-ognized before as a separate class of star is that the ear-lier x-ray detectors were less sensitive to low energies. Infact, after the ROSAT findings, researchers went backthrough their archives and realized that two of thesources had been discovered 10 years earlier by Knox S.Long and his colleagues at the Columbia University

and Supernovae

U p d a t e d f r o m t h e F e b r u a r y 1 9 9 9 i s s u e 77COPYRIGHT 2004 SCIENTIFIC AMERICAN, INC.

Astrophysics Laboratory (CAL), using the Einstein Observa-tory. These sources, named CAL 83 and CAL 87, had not beenclassified as distinct from other strong sources in the LargeMagellanic Cloud, although the Columbia team did remarkthat their spectra were unusually soft.

Back of the EnvelopeAT THE TIME, Anne P. Cowley and her co-workers at Ari-zona State University surmised that CAL 83 and 87 were ac-creting black holes, which often have softer spectra than neu-tron stars do. This suggestion seemed to receive support in the1980s, when faint stars were found at the locations of bothsources. The stars’ brightnesses oscillated, a telltale sign of abinary star system, in which two stars are in mutual orbit. In1988 an international observing effort led by Alan P. Smaleof University College London found that the brightness ofCAL 83 fluctuated with a period of just over one day. A sim-ilar project led by Tim Naylor, now at the University of Ex-eter in England, obtained a period of 11 hours for CAL 87.These visible companion stars are the fuel for the hypothesizedblack holes. Assuming they have not yet been decimated, thevarious measurements indicated that they weighed 1.2 to 2.5times as much as the sun.

But the ROSAT observations suddenly made this expla-nation very unlikely. The sources were much cooler than anyknown black hole system. Moreover, their brightness andtemperature revealed their size. According to basic physics,each unit area of a star radiates an amount of energy propor-tional to the fourth power of its temperature. By dividing thisenergy into the total emission of the star, astronomers can eas-ily calculate its surface area and, assuming it to be spherical,


ALFR

ED

T.K

AMAJ

IAN

(p

rece

din

g p

ag

es);

BR

YAN

CH

RIS

TIE

; SO

UR

CE

: P

ETE

R K

AHAB

KA,

ED

WAR

D P

. J.

VAN

DE

N H

EU

VEL

AND

SAU

L A.

RAP

PAP

OR

T (g

rap

hs)

; W

OLF

GAN

G P

IETS

CH

RO

SAT

Tea

m (

bot

tom

lef

t);

SVE

N K

OH

LE U

niv

ersi

ty o

f B

onn

(b

otto

m r

igh

t)

0 0.5 1.0 1.5 2.0 2.5

∞ 2.50 1.25Wavelength (nanometers)

0.83 0.62 0.50

Energy (kiloelectron volts)

SUPERSOFT X-RAY SOURCE

HARD X-RAY SOURCE

Detector efficiency

Inferredsource

spectrum

Detected spectrum

3x

105

degrees

Detector efficiency Inferred

source spectrum

Detected spectrum

3 x 107 degrees

X-RAY COLOR IMAGE (left) shows how a nearby mini galaxy, the LargeMagellanic Cloud, might appear to someone with x-ray vision. A red colordenotes lower-energy (or, equivalently, longer-wavelength) radiation; blue means higher energy (shorter wavelength). Supersoft sources stand

out as red or orange dots; hard x-ray sources look blue. The supersoft starCAL 87 seems green because a cloud of hydrogen alters its true color.(Some red dots are actually sunlike stars in the foreground.) The view israther different from an ordinary photograph of the area (right).

SOFT AND HARD x-ray sources are distinguished by their spectra, asmeasured by the ROSAT orbiting observatory. A typical supersoft source(top) emits x-rays with a fairly low energy, indicative of a comparatively cooltemperature of 300,000 degrees Celsius. A typical hard x-ray source (bottom)is 100 times hotter and therefore emits higher-energy x-rays. In both cases,the intrinsic spectrum of the source (red curves) is distorted by the responseof the ROSAT detector ( gray curves) and by interstellar gas absorption.

CAL 83

CAL 87

RXJ0513.9-6951


its diameter. It turns out that CAL 83, CAL 87 and the otherMagellanic Cloud sources each have a diameter of 10,000 to20,000 kilometers (16,000 to 32,000 miles)—the size of awhite dwarf star. They are therefore 500 to 1,000 times aslarge as a neutron star or the “horizon” at the edge of a stel-lar-mass black hole. When Trümper first described the super-soft sources at a conference at the Santa Barbara Institute forTheoretical Physics in January 1991, several audience mem-bers quickly made this calculation on the proverbial back ofthe envelope.

Some conference participants, among them Jonathan E.Grindlay of Harvard University, suggested that the sourceswere white dwarfs that gave off x-rays as gas crashed ontotheir surface—much as hard x-ray sources result from the ac-cretion of matter onto a neutron star or into a black hole. Oth-ers, including Trümper, his colleagues Jochen Greiner andGünther Hasinger, and, independently, Nikolaos D. Kylafisand Kiriaki M. Xilouris of the University of Crete, proposedthat the sources were neutron stars that had somehow builtup a gaseous blanket some 10,000 kilometers thick. In eithercase, the ultimate source of the energy would be gravitation-al. Gravity would pull material toward the dwarf or neutronstar, and the energy of motion would be converted to heat andradiation during collisions onto the stellar surface or withinthe gas.

Both models seemed worth detailed study, and two of us(van den Heuvel and Rappaport), collaborating with Di-pankar Bhattacharya of the Raman Research Institute in Ban-galore, India, were lucky enough to be able to start such stud-ies immediately. The conference was part of a half-year work-shop at Santa Barbara, where several dozen scientists from

different countries had the time to work together on problemsrelated to neutron stars.

It soon became clear that neither model worked. The su-persoft sources emit about the same power as the brightest ac-creting neutron stars in binaries. Yet gas collisions onto neu-tron stars are 500 to 1,000 times as forceful as the same pro-cess on white dwarfs, because the effect of gravity at thesurface of a neutron star is that much greater. (For bodies ofthe same mass, the available gravitational energy is inverselyproportional to the radius of the body.) Thus, for a dwarf tomatch the output of a neutron star, it would need to sweep upmaterial at 500 to 1,000 times the rate. In such a frenetic ac-cretion flow—equivalent to several Earth masses a year—theincoming material would be so dense that it would totally ab-sorb any x-rays.

Neutron stars with gaseous blankets also ran into trou-ble. Huge envelopes of gas (huge, that is, with respect to the10-kilometer radius of the neutron star) would be unstable;they would either collapse or be blown away in a matter ofseconds or minutes. Yet CAL 83 and CAL 87 had been shin-ing for at least a decade. Indeed, the ionized interstellar gasnebula surrounding CAL 83 took many tens of thousands ofyears to create.

Nuclear PowerAFTER WEEKS OF DISCUSS ING and evaluating models,none of which worked, astrophysicists realized the crucial dif-ference between accretion of material onto neutron stars orblack holes and accretion onto white dwarfs. The former gen-erates much more energy than nuclear fusion of the sameamount of hydrogen, whereas the latter produces much less en-


ALFR

ED

T.

KAM

AJIA

N

COMPACT STARS have colossal escape velocities. A typical white dwarf(left) packs the mass of the sun into the volume of a planet. To break freeof its gravity, an object must travel at some 6,000 kilometers per second.This is also about the speed that a body doing the reverse trip—falling onto

the dwarf from afar—would have on impact. Denser stars, such as neutronstars with the same mass (center), have an even mightier grip. Thedensest possible star, a black hole, is defined by a “horizon” from whichthe escape velocity equals the speed of light (right).

White dwarf

Neutronstar

Stellarblack hole

20 km

20 km

Neutronstar

Escape velocity 150,000 km/s


White dwarf

Earth

Sun

13,000 km


Escape velocity 600 km/s

Escape velocity 11 km/s



ergy than fusion. Of the energy inherent in mass (Albert Ein-stein’s famous E = mc2), fusion releases 0.7 percent. Accretiononto a neutron star, however, liberates more than 10 percent;into a black hole, up to 46 percent before the material disap-pears completely. On the other hand, accretion onto a whitedwarf, with its comparatively weak gravity, liberates onlyabout 0.01 percent of the inherent energy.

Therefore, on white dwarfs, nuclear fusion is potentiallymore potent than accretion. If hydrogen accumulated on thesurface of a white dwarf and somehow started to “burn” (thatis, undergo fusion), only about 0.03 Earth mass would beneeded a year to generate the observed soft x-ray luminosity.Because of the lower density of inflowing matter, the x-rayswould be able to escape.

Stable nuclear burning of inflowing matter would accountfor the paradoxical brightness of the supersoft sources. But isit really possible? Here we were lucky. Just when we were dis-cussing this issue, Ken’ichi Nomoto of the University of Tokyoarrived in Santa Barbara. He had already been trying to an-swer the very same question in order to understand anotherphenomenon, nova explosions—outbursts much less energeticthan supernovae that cause a star suddenly to brighten10,000-fold but do not destroy it. Novae always occur in closebinaries that consist of a white dwarf and a sunlike star. Un-til the discovery of supersoft sources, they were the onlyknown close binaries.

For over a decade, Nomoto and others had been improv-ing on the pioneering simulations by Bohdan Paczynski andAnna Zytkow, both then at the Nicolaus Copernicus Astro-

nomical Center in Warsaw. According to these analyses, hy-drogen that has settled onto the surface of a dwarf can indeedburn. The style of burning depends on the rate of accretion.If it is sufficiently low, below 0.003 Earth mass a year, fusionis spasmodic. The newly acquired hydrogen remains passive,often for thousands of years, until its accumulated mass ex-ceeds a critical value, at which point fusion is abruptly ignit-ed at its base. The ensuing thermonuclear explosion is visibleas a nova.

If the accretion rate is slightly higher, fusion is cyclic butnot explosive. As the rate increases, the interval between burn-ing cycles becomes shorter and shorter, and above a certainthreshold value, stable burning sets in. For white dwarfs ofone solar mass, this threshold is about 0.03 Earth mass a year.In the simulations, fusion generates exactly the soft x-ray lu-minosity observed in the supersoft sources.


ON/OFF EMISSION of supersoft star RXJ0513.9-6951 is a sign that it ispoised between two different modes of behavior. When it shines brightlyin visible light (left graph), its x-ray output (right graph) is low, and viceversa. (The lower x-ray counts are upper limits.) The star is at the border

between a pure supersoft source (which would emit only x-rays) and a white dwarf surrounded by thick gas (which would emit only visiblelight). Slight fluctuations in the rate of gas intake switch the star from one behavior to the other.

17.5

17.0

Visu

al B

right

ness

(mag

nitu

de)

16.5

10–3

Oct. 24, 1994

Dec. 13, 1994

Feb. 1,1995

March 23,1995

Oct. 24, 1994

Dec. 13, 1994

Feb. 1,1995

March 23,1995

10–2

X-Ra

y Br

ight

ness

(cou

nts

per s

econ

d)

10–1

Roche lobe

BR

YAN

CH

RIS

TIE

; SO

UR

CE

: K

LAU

S R

EIN

SCH

Un

iver

sity

of

Göt

tin

gen

(le

ft)

AND

STE

FAN

G.

SCH

AEID

T M

ax

Pla

nck

In

stit

ute

for

Ext

rate

rres

tria

l P

hy

sics

(ri

gh

t)

5b

LIFE CYCLE of a supersoft star (sequence 1–6, on opposite page) beginswith an unequal binary star system and ends with a type Ia supernovaexplosion. The supersoft phase can take one of three forms, depending onthe companion star. If it is an ordinary star in a tight orbit, it can overflowits Roche lobe and cede control of its outer layers to the white dwarf (5a,on opposite page). If the companion is a red giant star of sufficient size, italso overflows its Roche lobe (5b, at right). Finally, if it is a red giant with asmaller size or a wider orbit, it can power a supersoft source with its strongwinds (5c, on opposite page). Not all supersoft sources blow up, butenough do to account for the observed rate of supernovae.


If the rate is still higher, above 0.12 Earth mass a year, theincoming gas does not settle onto the surface but instead formsan extended envelope around the dwarf. Steady burning con-tinues on the surface, but the thick envelope degrades the x-rays into ultraviolet and visible light. Recent calculations haveshown that the radiation is so intense that it exerts an outwardpressure on gas in the envelope, causing part of it to streamaway from the star in a stellar wind.

If the accretion rate hovers around 0.12 Earth mass a year,the system may alternate between x-ray and visible phases.Exactly this type of behavior has been found in the supersoftsource known as RXJ0513.9-6951, which was discovered byStefan G. Schaeidt of the Max Planck institute. It gives off x-rays for weeks at a time, with breaks of several months. Thison/off emission puzzled astronomers until 1996, when KarenA. Southwell and her colleagues at the University of Oxfordnoticed that the visible counterpart to this star fluctuated, too.When the visible star is faint, the x-ray source is bright, andvice versa [see top illustration on opposite page]. The systemalso features two high-speed jets of matter flowing out in op-posite directions at an estimated 4,000 to 6,000 kilometersper second. Such jets are common where an accretion disk

dumps more material on the star than it can absorb. The ex-cess squirts out in a direction perpendicular to the disk, wherethere is no inflowing matter to block it. The outflow velocityis expected to be approximately the same as the escape veloc-ity from the surface of the star. In RXJ0513.9-6951 the in-ferred speed nearly equals the escape velocity from a whitedwarf—further confirmation that the supersoft sources arewhite dwarfs.

Soft-Boiled StarNOT EVERY BINARY SYSTEM can supply material at therates required to produce a supersoft source. If the companionstar is less massive than the white dwarf, as is typically ob-served in nova-producing systems, the fastest that material canflow in is 0.0003 Earth mass a year. This limit is a consequenceof the law of conservation of orbital angular momentum. Asthe small companion star loses mass, its orbit widens and theflow rate stabilizes.

For the rates to be higher, the donor star must have a massgreater than that of the dwarf. Then the conservation of an-gular momentum causes the orbit to shrink as a result of themass transfer. The stars come so close that they begin a gravi-


Pair of ordinarystars burn hydrogenin their cores

One exhausts fuelin core, becomesred giant

Orbit tightens;giant envelopsother star

Giant sheds outerlayers, becomeswhite dwarf

Dwarf steals gasfrom other star,emits soft x-rays

Dwarf reachescritical mass,explodes

BR

YAN

CH

RIS

TIE

ALFR

ED

T.

KAM

AJIA

N

ALFR

ED

T.

KAM

AJIA

N

Accretion disk

White dwarf

Companion star Stellar wind

1 2 3 4 5a

5c

6



tational tug-of-war for control of the outer layers of the donor.Material within a certain volume called the Roche lobe remainsunder the sway of the donor’s gravity, while material beyondit is stripped off by the dwarf. Perversely, the donor abets itsown destruction. While it sheds mass at the surface, theamount of energy generated by fusion in the core remainslargely unaffected. The continued heating from below exertspressure on the outer layers to maintain the original shape ofthe star. This pressure replenishes the material ripped off thedwarf, much as an overflowing pot of soup on a hot burner willcontinue to pour scalding water onto the stove. The situationstabilizes only when the effects of mass loss are felt by the coreitself. For a star originally of two solar masses, the return toequilibrium—and thus the cessation of supersoft emission—

takes seven million years after the onset of plundering. By thistime the star has shrunk to a fifth of its initial mass and becomethe lesser star in the system. The average accretion rate ontothe dwarf in such a case is about 0.04 Earth mass a year.

Following this reasoning, we predicted in 1991 that many supersoft sources would be white dwarfs in tight orbits(with periods of less than a few days) around a companionstar whose original mass was 1.2 to 2.5 solar masses. In fact,CAL 83 and 87 are precisely such systems. Since 1992 orbitalperiods for four more supersoft sources have been measured;all periods were less than a few days. The explanation mayalso apply to a class of novalike binary systems, called V Sagit-tae stars, whose oscillating brightness has perplexed as-tronomers for a century. In 1998 Joseph Patterson of Colum-bia and his collaborators and, independently, Joao E. Steinerand Marcos P. Diaz of the National Astrophysical Laborato-ry in Itajubá, Brazil, demonstrated that the prototype of

this class of stars has the appropriate mass and orbital period.There is one other group of star systems that could give

rise to supersoft sources: so-called symbiotic binaries, in whichthe white dwarf is in a wide orbit about a red giant star. Redgiants are willing donors. Bloated by age, they have relative-ly weak surface gravity and already discharge matter in strongstellar winds. In 1994 one of us (Kahabka), Hasinger andWolfgang Pietsch of the Max Planck Institute discovered a su-persoft symbiotic binary in the Small Magellanic Cloud, an-other satellite galaxy of the Milky Way. Since then, a furtherhalf dozen such sources have been found.

Some supersoft sources are harder to recognize because theiraccretion rate varies with time. One source in our galaxy alter-nates between x-ray and visible emission on a cycle of 40 years,as seen on archival photographic plates. A few objects, such asNova Muscae 1983 and Nova Cygni 1992, combine nova be-


BR

YAN

CH

RIS

TIE

; SO

UR

CE

: IC

KO

IB

EN

Un

iver

sity

of

Illi

noi

s

0.3

0.03

0.003

0.6 0.8 1.0 1.2 1.4White Dwarf Mass (solar masses)

Accr

etio

n Ra

te (

Eart

h m

asse

s a

year

)

STEADY BURNING, GASEOUS ENVELOPE AND WIND FORM

STEADY BURNING(SUPERSOFT SOURCES)

CYCLIC, NONEXPLOSIVE BURNING

CYCLIC, EXPLOSIVE BURNING(NOVAE)

STYLE OF NUCLEAR FUSION on the surface of a white dwarf depends on howmassive the dwarf is and how fast it is devouring its companion star (verticalaxis). If the accretion rate is sufficiently low, fusion (which astronomers

misleadingly call “burning”) occurs in spurts, either gently or explosively.Otherwise it is continuous. As shown above, phenomena once thought tobe distinct—such as novae and supersoft sources—are closely related.

PETER KAHABKA, EDWARD P. J. VAN DEN HEUVEL and SAUL A.RAPPAPORT never thought supersoft sources would be explainedby white dwarfs. That insight came about during a workshop thatvan den Heuvel and Rappaport organized on a different topic: neu-tron stars. Two years later these veteran astronomers met Ka-habka, who had discovered many supersoft sources as a memberof the ROSAT team. Today Kahabka is research associate at theUniversity of Bonn in Germany. Van den Heuvel is director of theAstronomical Institute at the University of Amsterdam and the1995 recipient of the Spinoza Award, the highest science awardin the Netherlands. An amateur archaeologist, he owns an exten-sive collection of early Stone Age tools. Rappaport is a physicsprofessor at the Massachusetts Institute of Technology. He wasone of the pioneers of x-ray astronomy in the 1970s.

THE

AU

THO

RS


havior with supersoft emission, which can be explained by ayears-long period of sedate “afterburning” between eruptions.

The Seeds of SupernovaeTHE COMPANION MASSES required of supersoft sourceswith short orbital periods imply that they are relatively youngsystems (compared with the age of our galaxy). Stars of theinferred mass live at most a few billion years and are always lo-cated in or near the youthful central plane of the galaxy. Un-fortunately, that location puts them in the region thick with in-terstellar clouds, which block soft x-rays. For this reason, theobserved population is only the tip of the iceberg. Extrapolat-ing from the known number of supersoft sources, we have es-timated that the total number in our galaxy at any one time isseveral thousand. A few new ones are born every 1,000 years,and a few others die.

What happens as they pass away? The fusion of matter re-ceived from the companion clearly causes the white dwarf to

grow in mass. It could reach the Chandrasekhar limit of about1.4 solar masses, the maximum mass a white dwarf can have.Beyond this limit, the quantum forces that hold up the dwarffalter. Depending on the initial composition and mass of thedwarf, there are two possible outcomes: collapse to a neutronstar or destruction in a nuclear fireball. Dwarfs that either lackcarbon or are initially larger than 1.1 solar masses collapse. Anumber of theorists have analyzed this fate.

White dwarfs that do not meet either of these criteria sim-ply blow up. They may slowly amass helium until they reachthe Chandrasekhar limit and explode. Alternatively, the he-lium layer may reach a critical mass prematurely and igniteitself explosively. In the latter case, shock waves convulse thestar and ignite the carbon at its core. And once the carbonburning begins, it becomes a runaway process in the dense,taut material of the dwarf. Within a few seconds the star isconverted largely into nickel as well as other elements be-tween silicon and iron. The nickel, dispersed into space, ra-dioactively decays to cobalt and then iron in a few hundreddays. Astronomers had already ascribed a kind of explosionto the death of carbon-rich dwarfs—the supernova type Ia.The spectrum of such a supernova lacks any sign of hydro-gen or helium, one of the factors that distinguish it from theother types of supernovae (Ib, Ic and II), which all probablyresult from the implosion and subsequent explosion of mas-sive stars. Type Ia supernovae are thought to be a majorsource of iron and related elements throughout the universe,including on Earth. Four occur every 1,000 years on aver-age in a galaxy such as the Milky Way.

Before supersoft sources were discovered, astronomerswere unsure as to the precise sequence that led to type Ia su-pernovae. The leading explanations implicated either certainsymbiotic stars—in particular, the rare recurrent novae—ormergers of two carbon-rich white dwarfs. But the latter viewis now disputed. Although recently double-dwarf systemswith the necessary mass and orbital period have been discov-ered, calculations by Nomoto and his colleague HadeyukiSaio have shown that such a merger could in many cases betoo gentle to produce a thermonuclear explosion and insteadwould lead to the formation of a neutron star. Supersoftsources and other surface-burning dwarfs seem a good alter-native solution. Their death rate roughly matches the observedsupernova frequency. The concordance makes the luminoussupersoft binary x-ray sources the first firmly identified classof objects that can realistically be expected to end their livesas type Ia supernovae.

This new realization may improve the accuracy of cosmo-

logical measurements that rely on these supernovae to deter-mine distance [see “Surveying Space-time with Supernovae,”by Craig J. Hogan, Robert P. Kirshner and Nicholas B.Suntzeff; Scientific American, January 1999]. Subtle varia-tions in brightness can make all the difference between conflict-ing conclusions concerning the origin and fate of the universe.The worry for cosmologists has always been that slight sys-tematic errors—the product, perhaps, of astronomers’ incom-plete understanding of the stars that go supernova—couldmimic real variations. The implications of the supersoft find-ings for cosmology, however, have yet to be worked out.

When supersoft sources were first detected, nobody ex-pected that the research they provoked would end up unitingso many phenomena into a single coherent theory. Now it isclear that a once bewildering assortment of variable stars, no-vae and supernovae are all variants on the same basic system:an ordinary star in orbit around a reanimated white dwarf.The universe seems that much more comprehensible.


Luminous Supersoft X-ray Sources. P. Kahabka and E.P.J. van denHeuvel in Annual Review of Astronomy and Astrophysics, Vol. 35, pages69–100. Annual Reviews, 1997.SNeIa: On the Binary Progenitors and Expected Statistics. Pilar Ruiz-Lapuente, Ramon Canal and Andreas Burkert in ThermonuclearSupernovae. Edited by Ramon Canal, Pilar Ruiz-Lapuente and Jordi Isern.Kluwer, 1997. Preprint available at xxx.lanl.gov/abs/astro-ph/9609078 Type Ia Supernovae: Their Origin and Possible Applications inCosmology. Ken’ichi Nomoto, Koichi Iwamoto and Nobuhiro Kishimoto inScience, Vol. 276, pages 1378–1382; May 30, 1997. Preprint available atxxx.lanl.gov/abs/astro-ph/9706007


Nuclear burning of inflowing matter WOULD ACCOUNT FOR THE PARADOXICAL BRIGHTNESS of the supersoft sources. But is it really possible?



These paired stellar remnants supplyexquisite confirmations of generalrelativity. Their inevitable collapse

produces what may be the strongest explosions

in the universe

By Tsvi Piran

COLLIDING NEUTRON STARS mark the end of a pattern of stellar evolution thatnow appears to be more likely thanastronomers once thought. More thanhalf the stars in the sky belong to binarysystems; perhaps one in 100 of themost massive pairs will ultimatelybecome neutron star binaries.Gravitational waves given off by thestars as they orbit each other carryaway energy until the stars spiraltogether and coalesce. These mergersgive off radiation that may be detectablefrom billions of light-years away.

Binary Neutron


Even as Bell and Hewish were mak-ing their discovery, military satellites or-biting Earth were detecting the signatureof even more exotic signals: powerfulgamma-ray bursts from outer space. Thegamma rays triggered detectors intendedto monitor illicit nuclear tests, but it wasnot until six years later that the observa-tions were made public; even then, an-other 20 years passed before the bursts’origin was understood. Many peoplenow think gamma-ray bursts are emittedby twin neutron stars in the throes of coalescence.

The discovery of binary neutron starsfell to Russell A. Hulse and Joseph H.Taylor, Jr., then at the University of Mas-sachusetts at Amherst, who began a sys-tematic pulsar survey in 1974. They usedthe Arecibo radio telescope in PuertoRico, the largest in the world, and with-in a few months had found 40 previous-ly unknown pulsars. Among their haulwas a strange source by the name of PSR1913+16 (PSR denotes a pulsar, and thenumbers stand for its position in the sky:19 hours and 13 minutes longitude anda declination of 16 degrees). It emittedapproximately 17 pulses per second, butthe period of the pulses changed by asmuch as 80 microseconds from one dayto the next. Pulsars are so regular thatthis small fluctuation stood out clearly.

Hulse and Taylor soon found that thetiming of the signals varied in a regularpattern, repeating every seven hours and45 minutes. This signature was not new;for many years astronomers have notedsimilar variations in the wavelength oflight from binary stars (stars that are or-biting each other). The Doppler effectshortens the wavelength (and increasesthe frequency) of signals emitted when asource is moving toward Earth and in-creases wavelength (thus decreasing thefrequency) when a source is movingaway. Hulse and Taylor concluded thatPSR 1913+16 was orbiting a companion

star, even though available models ofstellar evolution predicted only solitarypulsars.

The surprises did not end there. Anal-ysis of the time delay indicated that thepulsar and its companion were separat-ed by a mere 1.8 million kilometers. Atthat distance, a normal star (with a radiusof roughly 600,000 kilometers) would al-most certainly have blocked the pulsar’ssignal at some point during its orbit. Thecompanion could also not be a whitedwarf (radius of about 3,000 kilometers),because tidal interactions would haveperturbed the orbit in a way that contra-dicted the observations. Hulse and Tay-lor concluded that the companion to PSR1913+16 must be a neutron star.

This finding earned the two a NobelPrize in Physics in 1993. Astronomershave since mastered the challenge of un-derstanding how binary neutron starsmight exist at all, even as they have em-ployed the signals these strange entitiesproduce to conduct exceedingly fine testsof astrophysical models and of generalrelativity.

Birth from DeathBY ALL THE ASTROPHYSICAL theo-ries that existed before 1974, binary neu-tron stars should not have existed.Astronomers believed that the repeatedstellar catastrophes needed to createthem would disrupt any gravitationalbinding between two stars.

Neutron stars are the remnants ofmassive stars, which perish in a superno-va explosion after exhausting all their nu-clear fuel. The death throes begin when astar of six solar masses or more con-sumes the hydrogen in its center, expandsand becomes a red giant. At this stage, itscore is already extremely dense: severalsolar masses within a radius of severalthousand kilometers. An extended enve-lope more than 100 million kilometersacross contains the rest of the mass. In

StarsAL

FRE

D T

. K

AMAJ

IAN

U p d a t e d f r o m t h e M a y 1 9 9 5 i s s u e 85

In 1967 Jocelyn Bell and Antony Hewish found the first pulsar. Their radio telescopebrought in signals from a source that emitted very regular pulses every 1.34 seconds.After eliminating terrestrial sources and provisionally discarding the notion that thesesignals might come from extraterrestrial intelligent beings, they were baffled. It wasThomas Gold of Cornell University who realized that the pulses originated from arotating neutron star, beaming radio waves into space like a lighthouse. Researcherssoon tuned in other pulsars.


the core, heavier elements such as siliconundergo nuclear fusion to become iron.

When the core reaches a temperatureof several billion kelvins, the iron nucleibegin to break apart, absorbing heatfrom their surroundings and reducing thepressure in the core drastically. Unable tosupport itself against its own gravita-tional attraction, the core collapses. Asits radius decreases from several thou-sand kilometers to 15, electrons and pro-

tons fuse into neutrons, leaving a verydense star of 1.4 solar masses in a volumeno larger than an asteroid.

Meanwhile the energy released in thecollapse heats the envelope of the star,which for a few weeks emits more lightthan an entire galaxy. Observations ofold supernovae, such as the Crab Nebu-la’s, whose light reached Earth in A.D.

1054, reveal a neutron star surroundedby a luminous cloud of gas, still movingout into interstellar space.

More than half the stars in the sky be-long to binary systems. As a result, it isnot surprising that at least a few massivepairs should remain bound together evenafter one of them undergoes a supernovaexplosion. The pair then becomes a mas-sive x-ray binary, so named for the emis-sion that the neutron star produces as itstrips the outer atmosphere from its com-

panion. Eventually the second star alsoexplodes as a supernova and turns into aneutron star. The envelope ejected by thesecond supernova contains most of themass of the binary (since the remainingneutron star contains a mere 1.4 solarmasses). The ejection of such a large frac-tion of the total mass should thereforedisrupt the binary and send the two neu-tron stars flying into space at hundreds ofkilometers per second.

Hulse and Taylor’s discovery demon-strated, however, that some binaries sur-vive the second supernova explosion. Inretrospect, astronomers realized that thesecond supernova explosion might beasymmetrical, thereby propelling thenewly formed neutron star into a stableorbit rather than out into the void. Thesecond supernova also may be lessdisruptive if the second star loses its en-velope gradually during the massive x-ray binary phase. Since then, the discov-ery of other neutron star binaries showsthat other massive pairs have managed tosurvive the second supernova.

In the early 1990s Ramesh Narayanof Harvard University, Amotz Shemi ofTel Aviv University and I, along with E.Sterl Phinney of the California Instituteof Technology, working independently,estimated that about 1 percent of mas-

sive x-ray binaries survive to form neu-tron star binaries. This figure implies thatour galaxy contains a population ofabout 30,000 neutron star binaries. Fol-lowing a similar line of argument, wealso concluded that there should be acomparable number of binaries, yet un-observed, containing a neutron star anda black hole. Such a pair would formwhen one of the stars in a massive pairformed a supernova remnant containing

more than about two solar masses and socollapsed to a singularity instead of aneutron star. Rarer, but still possible intheory, are black hole binaries, whichstart their lives as a pair of particularlymassive stars; they should number about300 in our galaxy.

Testing General RelativityPSR 1913+16 has implications thatreach far beyond the revision of theoriesof binary stellar evolution. Hulse andTaylor quickly realized that their discov-ery had provided an ideal site for testingEinstein’s general theory of relativity.

Although this theory is accepted to-day as the only viable description ofgravity, it has had just a few direct tests.Albert Einstein himself computed theprecession of Mercury’s orbit (the shift ofthe orbital axes and the point of Mer-


ALFR

ED

T.

KAM

AJIA

N

a b

Binary pairX-ray binary

As the two stars approach they tear material from each other,THEN COALESCE WITHIN A FRACTION OF A SECOND.


cury’s closest approach to the sun) andshowed that observations agreed with histheory. A few other tests followed. In1964 Irwin I. Shapiro of Harvard point-ed out that light signals bent by a gravi-tational field should be delayed in com-parison to those that take a straight path.He measured the delay by bouncing ra-dar signals off other planets in the solarsystem. Although general relativitypassed these tests with flying colors, theywere all carried out in the (relativistically)weak gravitational field of the solar sys-tem. That fact left open the possibilitythat general relativity might break downin stronger gravitational fields.

Because a pulsar is effectively a clockorbiting in the strong gravitational fieldof its companion, relativity makes arange of clear predictions about how theticks of that clock (the pulses) will appearfrom Earth. First, the Doppler effectcauses a periodic variation in the pulses’arrival time (the pattern that first alertedTaylor and Hulse).

A “second-order” Doppler effect, re-sulting from time dilation caused by thepulsar’s rapid motion, leads to an addi-tional (but much smaller) variation. Thissecond-order effect can be distinguishedbecause it depends on the square of ve-locity, which varies as the pulsar movesalong its elliptical orbit. The second-orderDoppler shift combines with the gravita-tional redshift, a slowing of the pulsar’sclock when it is in the stronger gravita-tional field closer to its companion.

Like Mercury, PSR 1913+16 pre-

cesses in its orbit about its companion.The intense gravitational fields, howev-er, mean that the periastron—the nadirof the orbit—rotates by 4.2 degrees ayear, compared with Mercury’s perihe-lion shift of a mere 42 arc seconds a cen-tury. The measured effects match the pre-dictions of relativistic theory precisely.Remarkably, the precession and other in-formation supplied by the timing of theradio pulses make it possible to calculatethe masses of the pulsar and its compan-ion: 1.442 and 1.386 solar masses, re-spectively, with an uncertainty of 0.003solar mass. This precision is impressivefor objects 15,000 light-years away.

In 1991 Alexander Wolszczan of theArecibo observatory found another bi-nary pulsar that is almost a twin to PSR1913+16. Each neutron star weighs be-tween 1.27 and 1.41 solar masses. TheShapiro time delay, which was only mar-ginally measured in PSR 1913+16,stands out clearly in signals from the pul-sar that Wolszczan discovered.

Measurements of PSR 1913+16 havealso revealed a relativistic effect neverseen before. In 1918, several years afterthe publication of his general theory ofrelativity, Einstein predicted the existenceof gravitational radiation, an analogue toelectromagnetic radiation. He said mas-sive particles that move with varying ac-celeration emit gravitational waves, smallripples in the gravitational field that alsopropagate at the speed of light.

These ripples exert forces on othermasses; if two objects are free to move,the distance between them will vary withthe frequency of the wave. The size of theoscillation depends on the separation ofthe two objects and the strength of thewaves. In principle, all objects whose ac-celeration varies emit gravitational radi-ation. Most objects are so small andmove so slowly, however, that their grav-itational radiation is utterly insignificant.

Binary pulsars are one of the few ex-ceptions. In 1941, long before the dis-

covery of the binary pulsar, Russianphysicists Lev D. Landau and Evgenii M.Lifshitz calculated the effect of this emis-sion on the motion of a binary. Energyconservation requires that the energy car-ried away by the waves come from some-where, in this case the orbital energy ofthe two stars. As a result, the distance be-tween them must decrease.

PSR 1913+16 emits gravitational ra-diation at a rate of eight quadrillion gi-gawatts, about a fifth as much energy asthe total radiation output of the sun. Thisluminosity is impressive as far as gravi-tational radiation sources are concernedbut still too weak to be detected directlyon Earth. Nevertheless, it has a notice-able effect on the pulsar’s orbit. The dis-tance between the two neutron stars de-creases by a few meters a year, which suf-fices to produce a detectable variation inthe timing of the radio pulses. By care-fully monitoring the pulses from PSR1913+16 over the years, Taylor and hiscollaborators have shown that the or-bital separation decreases in exact agree-ment with the predictions of the generaltheory of relativity.


GE

OR

GE

RE

TSE

CK

(a

bov

e)

MASSIVE BINARY (a) evolves through a sequence of violent events. The heavier star in the pairburns its fuel faster and undergoes a supernova explosion; if the two stars stay bound together,the result is a massive x-ray binary (b) in which the neutron star remnant of the first star stripsgas from its companion and emits x-radiation. Eventually the second star also exhausts its fuel.In roughly one of 100 cases, the resulting explosion leaves a pair of neutron stars orbiting eachother (c); in the other 99, the two drift apart (d). There are enough binary star systems that a typical galaxy contains thousands of neutron star binaries.

ORBITAL PRECESSION, the rotation of the majoraxis of an elliptical orbit, results from relativisticperturbations of the motion of fast-movingbodies in intense gravitational fields. It isusually almost undetectable; Mercury’s orbitprecesses by less than 0.12 of a degree everycentury, but that of PSR 1913+16 changes by4.2 degrees a year.

c

d



The reduction in the distance be-tween the stars can be compared with theother general relativistic effects to arriveat a further confirmation. Just as mea-surements of the orbital decay produce amathematical function relating the massof the pulsar to the mass of its compan-ion, so do the periastron shift and the sec-ond-order Doppler effect. All three func-tions intersect at precisely the same point.

Undetectable CataclysmsAT PRESENT, THE DISTANCE be-tween PSR 1913+16 and its companionis decreasing only slowly. As the gapshrinks, the gravitational-wave emissionwill increase, and the orbital decay willaccelerate. Eventually the neutron starswill fall toward each other at a significantfraction of the speed of light, collide andmerge. The 300 million years until thetwo coalesce are long on a human scalebut rather short on an astronomical one.

Given the number of neutron star bi-naries in the galaxy, one pair shouldmerge roughly every 300,000 years, acosmological blink of the eye. Extrapo-lating this rate to other galaxies impliesthat throughout the observable universeabout one neutron star merger occursevery 20 minutes—frequently enoughthat astronomers should consider wheth-er they can detect such collisions.

To figure out whether such occur-rences are detectable requires a solid un-derstanding of just what happens whentwo orbiting neutron stars collide. Short-ly after the discovery of the first binarypulsar, Paul Clark and Douglas M. Eard-ley, then both at Yale University, con-cluded that the final outcome is a blackhole. Current estimates of the maximummass of a neutron star range between 1.4and 2.0 solar masses. Rotation increasesthe maximal mass, but most models sug-gest that even a rapidly rotating neutronstar cannot be significantly larger than2.4 solar masses. Because the two stars to-gether contain about 2.8 solar masses, col-lapse to a singularity is almost inevitable.

Melvyn B. Davies of Caltech, WillyBenz of the University of Arizona, Fried-rich K. Thielemann of the Harvard-Smithsonian Center for Astrophysics andI have simulated the last moments of aneutron star binary in detail. The two ob-jects are very dense and so behave effec-tively like point masses until they arequite close to each other. Tidal interac-tion between the stars becomes impor-tant only when they approach to within30 kilometers, about twice the radius ofa neutron star. At that stage, they beginto tear material from each other—abouttwo tenths of a solar mass in total. Oncethe neutron stars touch, within a tiny

fraction of a second they coalesce. Thematter torn from the stars before the col-lision forms a disk around the centralcore and eventually spirals back into it.

What kinds of signals will this se-quence of events generate? Clark andEardley realized that the colliding starswill warm up to several billion kelvins.They figured that most of the thermal en-ergy would be radiated as neutrinos andantineutrinos, much as it is in a superno-va. Unfortunately, these weakly interact-ing, massless particles, which escapefrom the dense neutron star much moreeasily than do photons, are almost unde-tectable. When supernova 1987A ex-ploded, the three detectors on Earthcaught 21 neutrinos out of the 5 × 1046

joules of radiation. Although the burstexpected from a binary neutron starmerger is slightly larger than that of a su-pernova, the typical event takes place


GE

OR

GE

RE

TSE

CK

BINARY PULSAR SIGNALS are affected by relativistic phenomena. (Eachillustration shows one of the effects whose combination produces theobserved timing of the pulses.) The Doppler effect slows the rate at which

pulses reach an observer when the pulsar is moving away from Earth in itsorbit and increases the rate when the pulsar is moving toward Earth (a).The second-order Doppler effect and the gravitational redshift (b) impose

Observer

Obse

rver

FIRST-ORDER DOPPLER EFFECT SECOND-ORDER DOPPLER EFFECT

ba

TSVI PIRAN has studied general relativ-ity and astrophysics for 30 years. He isa professor at the Hebrew University ofJerusalem, where he received his Ph.D.in 1976. He has also worked at the Uni-versity of Oxford, Harvard University,Kyoto University and Fermilab. Piran,with Steven Weinberg of the Universityof Texas, established the JerusalemWinter School for Theoretical Physics.

THE

AU

THO

R


much farther away than the mere150,000 light-years of SN 1987A. To de-tect one merger a year would requirepicking up signals one sixteen-millionththe intensity of the 1987 event. Becausecurrent neutrino detectors must monitorinteractions within thousands of tons ofmaterial, it is difficult to imagine the ap-paratus that would be required. Further-more, supernovae are 1,000 times morefrequent than are neutron star collisions.Even if we detected a neutrino burst fromtwo neutron stars, it is unlikely we wouldbe able to distinguish it among the su-pernova neutrino bursts.

Before it emits its neutrino burst, theneutron star binary sends out a similarlyenergetic (but not quite so undetectable)train of gravitational waves. During the15 minutes before coalescence, the twostars cover the last 700 kilometers be-tween them, and their orbital periodshrinks from a fifth of a second to a fewmilliseconds. The resulting signal is justin the optimal range for terrestrial grav-itational-wave detectors.

An international network of such de-tectors has been built in the U.S. and inItaly. The American Caltech-M.I.T. teamhas detectors for the Laser InterferometerGravitational-Wave Observatory (LIGO)near Hanford in Washington State andnear Livingston, La. The French-Italian

team has its VIRGO facility near Pisa inItaly. The first detectors were designed tobe able to detect neutron star mergers upto 70 million light-years away; currentestimates suggest that there is only oneevent per 100 years up to this distance.Researchers have proposed to improvetheir instruments dramatically over timeto be able to detect neutron star mergersas far away as three billion light-years—

several hundred a year.

High-Energy PhotonsFOR SEVERAL YEARS after the dis-covery of PSR 1913+16, I kept wonder-ing whether there was a way to estimatewhat fraction of the coalescing stars’binding energy is emitted as electromag-netic radiation. Even if this fraction istiny, the binding energy is so large thatthe resulting radiation would still beenormous. Furthermore, photons aremuch easier to detect than neutrinos orgravitational waves, and so mergerscould be detected even from the most dis-tant parts of the universe.

In 1987 J. Jeremy Goodman ofPrinceton University, Arnon Dar of theTechnion-Israel Institute of Technologyand Shmuel Nussinov of Tel Aviv Uni-versity noticed that about a tenth of apercent of the neutrinos and antineutri-nos emitted by a collapsing supernova

core collide with one another and anni-hilate to produce electron-positron pairsand gamma rays. In a supernova the ab-sorption of these gamma rays by thestar’s envelope plays an important role inthe explosion of the outer layers.

In 1989 David Eichler of Ben GurionUniversity of the Negev, Mario Livio,then at the Technion, David N. Schrammof the University of Chicago and I specu-lated that a similar fraction of the neutri-nos released in a binary neutron starmerger would also produce electron-posi-tron pairs and gamma rays. The collidingneutron stars, however, have no envelopesurrounding them, and so the gammarays escape in a short, intense burst.

Gamma-ray bursts might arise froma more complex mechanism. The diskthat forms during the neutron star merg-er falls back onto the central coalescedobject within a few seconds, but duringthat time it, too, can trigger emissions. In1992 Bohdan Paczynski of Princeton,Narayan of Harvard and I suggested thatthe rotation of the disk could intensifythe neutron-star magnetic fields entan-gled in the disk’s material, causing giantmagnetic flares, a scaled-up version ofthe flares that rise from the surface of thesun. These short-lived magnetic distur-bances could generate gamma-ray burstsin the same way that solar flares produce


a similar variation because the pulsar’s internal clock slows when it movesmore rapidly in its orbit closest to its companion (longer arrow). Mostsubtle is the Shapiro time delay, which occurs as the gravitational field of

the companion bends signals passing near it (c). The signals travel fartherthan if they took a straight-line path (d) and so arrive later. This effect isundetectable in PSR 1913+16 but is clear in other systems.

Observer

Observer

SHAPIRO TIME DELAY

c d



gamma rays and x-rays. The large vari-ability in the observed bursts implies thatboth mechanisms may be at work.

Puzzle UnscrambledHAD IT NOT BEEN FOR the LimitedTest Ban Treaty of 1963, we would nothave known about these bursts fordecades. No one would have proposed asatellite to look for them, and had such aproposal been made it would surely havebeen turned down as too speculative. Butthe U.S. Department of Defense launcheda series of satellites known as Vela, whichcarried omnidirectional x-ray and gam-ma-ray detectors to verify that no onewas testing nuclear warheads in space.

These spacecraft never detected a nu-clear explosion, but as soon as the firstsatellite was launched it began to detectentirely unexpected bursts of high-energyphotons in the range of several hundredkiloelectron volts. Bursts lasted betweena few dozen milliseconds and about 30seconds. The lag between the arrival timeof the bursts to different satellites indi-cated that the sources were outside the so-

lar system. Still, the bursts were kept se-cret for several years, until in 1973 RayW. Klebesadel, Ian B. Strong and Roy A.Olson of Los Alamos National Labora-tory described them in a seminal paper.Theorists proposed more than 100 mod-els in the next 20 years; in the late 1980sa consensus formed that the bursts origi-nated on neutron stars in our own galaxy.

A minority led by Paczynski arguedthat the bursts originated at cosmologi-cal distances. In the spring of 1991 theCompton Gamma Ray Observatory,which was more sensitive than any pre-vious gamma-ray satellite, was launchedby NASA. It revealed two unexpectedfacts. First, the distribution of burst in-tensities is not homogeneous in the waythat it would be if the bursts were near-by. Second, the bursts came from allacross the sky rather than being concen-trated in the plane of the Milky Way, asthey would be if they originated in thegalactic disk. Together these factsdemonstrate that the bursts do not orig-inate from the disk of our galaxy. A live-ly debate still prevails over the possibili-

ty that the bursts might originate fromthe distant parts of the invisible halo ofour galaxy, but as the Compton obser-vatory collects more data, this hypothe-sis seems less and less likely. It seems thatthe minority was right.

In the fall of 1991, I analyzed the dis-tribution of burst intensities, as did Pa-czynski and his colleague Shude Mao.We concluded that the most distantbursts seen by the Compton observatorycame from several billion light-yearsaway. The cosmological origin of gam-ma-ray bursts was indeed confirmed sev-eral years later.

A Rare LaboratoryPSR 1913+16 HAS been a great toolfor astrophysics, but it will disappearfrom our sight by about 2020. Accord-ing to the general theory of relativity, apulsar’s spin should precess and its di-rection in space should vary. A significantprecession will carry the beam away fromus altogether. In 1998 Michael Kramerof the University of California at Berkeleyfinally determined that the pulsar’s beamis pointing toward Earth only 60 years outof every 300. Our distant offspring will beable to see it again around the year 2260.

Several other binary pulsar systemshave been discovered, bringing the totalknown to seven. Most remarkable isJ0307-3039, found in December 2003 byMarta Burgay, a Ph.D. student at the Uni-versity of Bologna in Italy, and an inter-national team of researchers. A fewweeks later Andrew Lyne of the Univer-sity of Manchester in England and thesame team reported the detection of thesecond pulsar in the system, making it thefirst binary system with two observed pul-sars orbiting each other.

This system is even more remarkablethan PSR 1913+16. With a separation ofonly 800,000 kilometers—about twicethe distance between Earth and the


GE

OR

GE

RE

TSE

CK

GAMMA-RAY BURSTS may also result from a mechanism similar to the one that powers solar flares. Themagnetic field from the coalesced neutron stars is amplified as it winds through the disk of materialthrown out during the merger. The field accelerates charged particles until they emit gamma rays.

Burst

Magnetic field lines

NASA’S SWIFT SATELLITE, TO LAUNCH THIS FALL, may determine whether a partner in a neutron star merger

MORE NATURALLY ENDS AS A SUPERNOVA OR A BLACK HOLE.


moon—the stars orbit each other in justover two hours. With an orbital veloci-ty of 600 kilometers a second, relativis-tic effects are more pronounced than inany other known binary. The periastronshift, for example, is as large as 16.9 de-grees a year (four times faster than PSR1913+16). Because the line of sight toEarth passes almost within the binary or-bital plane, pulses from the pulsars passclose to each other and we can observethe Shapiro time delay. Indeed, the puls-es pass so close that we observe aneclipse. So far Einstein’s general theory ofrelativity has superbly passed the nu-merous tests that J0307-3039 provides.The effect of gravitational-wave emissionon the orbital motion has not yet beenobserved, but researchers expect to mea-sure it within a year. The predicted spinprecession period is around 70 years.Not surprisingly, the discoverers calledthis system “a rare laboratory for rela-tivistic gravity.”

We may also soon have more infor-mation about gamma-ray bursts. InFebruary 1997 an Italian-Dutch team,headed by Enrico Costa of the Nation-al Research Council in Rome, used theBeppoSAX satellite to pinpoint thesource of the burst GRB 970228 (eachburst is denoted by the date it is ob-

served). The x-ray telescope revealed anx-ray afterglow that persisted for sever-al days. This was followed by opticaland radio afterglows lasting months.

The identification of the exact burstposition enabled detailed examinationsthat were impossible before. In April1998 Titus Galama of the University ofAmsterdam and an international teamdiscovered supernova 1998bw withinthe error box of GRB 980425. Furtherevidence for association between a su-pernova signal, emerging within thelight curve of the afterglow of the burstGRB 030329, convincingly confirmedthis association.

When a massive star consumes its nu-clear fuel, its core collapses, usuallyforming a neutron star. The stellar enve-lope that surrounds the core explodes asa supernova. But according to the col-lapsar model, suggested by Stan Woosleyof the University of California at SantaCruz, some collapsing stellar cores forma black hole. A black hole that is fed by

an accretion disk produces aburst in a process that is remark-ably similar to the one expectedin the merger model describedearlier. The essential difference isthat the black hole is now withinthe exploding stellar envelope.

Bursts belong to two distinctclasses: long (those lasting morethan two seconds) and short(those less than two seconds). Sofar afterglow has been seen onlyfrom long bursts, and it is gener-ally accepted that the bursts areassociated with the supernovadeath of a massive star. With nodetectable afterglow, however,little is known about the shortbursts, and they remain as mys-terious as ever.

The collapsar model cannotaccount for short bursts. On theother hand, the natural outcome

of a neutron star merger is a short burst.But can we confirm this? NASA’s Swiftsatellite, scheduled to launch in the fallof 2004, should help. It is designed toovercome current technical difficultiesin detecting afterglow from shortbursts, but it could turn out that shortbursts do not have a detectable after-glow. In that case, hope is not lost. Neu-tron star mergers emit unique gravita-tional-radiation signals. With the inter-national network of gravitationaldetectors operational, we expect that—if the merger model is correct—these de-tectors, or at least an upgraded LIGOdetector, will eventually find a gravita-tional-wave merger signal that coincideswith a gamma-ray burst. If we can de-tect the unique gravitational-wave sig-nal of spiraling neutron stars in coinci-dence with a gamma-ray burst, astro-physicists will have opened a newwindow on the final stages of stellarevolution, one that no visible-light in-struments can hope to match.


LIG

O

LIGO INTERFEROMETERS, such as this one at the Hanford Observatory in Richland, Wash., should be able todetect the gravitational radiation of colliding neutron stars from a distance of billions of light-years. If thesesignals arrive at the same time as gamma-ray bursts, a decades-old mystery may be solved.

Was Einstein Right? Putting General Relativity to the Test. Clifford M. Will. Basic Books, 1988.

LIGO: The Laser Interferometer Gravitational-Wave Observatory. Alex Abramovici et al. in Science,Vol. 256, pages 325–333; April 17, 1992.

Probing the Gamma-Ray Sky. Kevin Hurley in Sky and Telescope, Vol. 84, No. 6, pages 631–636;December 1992.




A PICTURE LIKE THIS could not have been drawn with any confidence a decade ago, because no one had yet figured out what causes gamma-raybursts—flashes of high-energy radiation that light up the sky a couple of times a day. Now astronomers think of them as the ultimate stellarswan song. A black hole, created by the implosion of a giant star, sucks indebris and sprays out some of it. A series of shock waves emits radiation.


Every time a gamma-ray burst goes off, a black hole is born

By Neil Gehrels, Luigi Piro and Peter J. T. Leonard

Early in the morning of January 23, 1999, a robotic

telescope in New Mexico picked up a faint flash of

light in the constellation Corona Borealis. Though just

barely visible through binoculars, it turned out to be

the most brilliant explosion ever witnessed by hu-

manity. We could see it nine billion light-years away,

more than halfway across the observable universe. If

the event had instead taken place a few thousand light-

years away, it would have been as bright as the midday

sun, and it would have dosed Earth with enough radi-

ation to kill off nearly every living thing.

The flash was another of the famous gamma-ray

bursts, which in recent decades have been one of as-

tronomy’s most intriguing mysteries. The first sighting

of a gamma-ray burst (GRB) came on July 2, 1967,

from military satellites watching for nuclear tests in

space. These cosmic explosions proved to be rather dif-

ferent from the man-made explosions that the satellites

UniverseExplosions

in the

The Brightest


were designed to detect. For most of the37 years since then, each new burst hadmerely heightened the puzzlement.Whenever researchers thought they hadthe explanation, the evidence sent themback to square one.

The monumental discoveries of thepast several years have brought astrono-mers closer to a definitive answer. Before1997, most of what we knew aboutGRBs was based on observations fromthe Burst and Transient Source Experi-ment (BATSE) onboard the ComptonGamma Ray Observatory. BATSE re-vealed that two or three GRBs occursomewhere in the observable universe ona typical day. They outshine everythingelse in the gamma-ray sky. Althougheach is unique, the bursts fall into one oftwo rough categories. Bursts that last lessthan two seconds are “short,” and thosethat last longer—the majority—are

“long.” The two categories differ spec-troscopically, with short bursts havingrelatively more high-energy gamma raysthan long bursts do. The January 1999burst emitted gamma rays for a minuteand a half.

Arguably the most important resultfrom BATSE concerned the distribution ofthe bursts. They occur isotropically—thatis, they are spread evenly over the entiresky. This finding cast doubt on the pre-vailing wisdom, which held that burstscame from sources within the MilkyWay; if they did, the shape of our galaxy,or Earth’s off-center position within it,should have caused them to bunch up incertain areas of the sky. The uniform dis-tribution led most astronomers to con-clude that the instruments were picking upsome kind of event happening through-out the universe. Unfortunately, gammarays alone did not provide enough infor-

mation to settle the question for sure. Re-searchers would need to detect radiationfrom the bursts at other wavelengths. Vis-ible light, for example, could reveal thegalaxies in which the bursts took place,allowing their distances to be measured.Attempts were made to detect these burstcounterparts, but they proved fruitless.

A Burst of ProgressTHE FIELD TOOK a leap forward in1996 with the advent of the x-ray space-craft BeppoSAX, built and operated bythe Italian Space Agency with the partic-ipation of the Netherlands Space Agen-cy. BeppoSAX was the first satellite to lo-calize GRBs precisely and to discovertheir x-ray “afterglows.” The afterglowappears when the gamma-ray signal dis-appears. It persists for days to months,diminishing with time and degradingfrom x-rays into less potent radiation, in-cluding visible light and radio waves. Al-though BeppoSAX detected afterglowsfor only long bursts—no counterparts ofshort bursts have yet been identified—itmade follow-up observations possible atlast. Given the positional informationfrom BeppoSAX, optical and radio tele-scopes were able to identify the galaxiesin which the GRBs took place. Nearly alllie billions of light-years away, meaningthat the bursts must be enormously pow-erful. Extreme energies, in turn, call forextreme causes, and researchers began to

94 S C I E N T I F I C A M E R I C A N U p d a t e d f r o m t h e D e c e m b e r 2 0 0 2 i s s u e

MAR

K A

. G

ARLI

CK

(p

rece

din

g p

ag

es);

EN

RIC

O C

OST

A In

stit

ute

of

Spa

ce A

stro

ph

ysi

cs a

nd

Cos

mic

Ph

ysi

cs,

CN

RAN

D T

HE

BE

PP

OSA

X TE

AM(l

eft)

; P

AUL

J. G

RO

OT

Un

iver

sity

of

Amst

erd

am

(ri

gh

t)

■ For three decades, the study of gamma-ray bursts was stuck in first gear—

astronomers couldn’t settle on even a sketchy picture of what sets off these cosmic fireworks.

■ Over the past seven years, however, observations have revealed that bursts are the birth throes of black holes. Most of the holes are probably created when a massive star collapses, releasing a pulse of radiation that can be seen billions of light-years away.

■ Now the research has shifted into second gear—fleshing out the theory andprobing subtle riddles, especially the bursts’ incredible diversity.

Overview/Gamma-Ray Bursts

X-RAYS: Eight hours after a burst went off on February 28, 1997,astronomers using the BeppoSAX satellite—including one of theauthors (Piro)—saw an x-ray afterglow for the first time. Thesecond image was taken a couple days later, by which time thex-rays had faded by a factor of 20.

VISIBLE LIGHT: A comparably quick reaction by astronomers onLa Palma in the Canary Islands allowed the same afterglow to beseen in visible light. Over the next week, the light dimmed to onesixth its original brightness, and as it did so, the surroundinggalaxy slowly became apparent.

Eight hours

X-RAYS VISIBLE LIGHT

Three days 21 hours Eight days

A (VERY) WARM AFTERGLOW


associate GRBs with the most extremeobjects they knew of: black holes.

Among the first GRBs pinpointed byBeppoSAX was GRB970508, so namedbecause it occurred on May 8, 1997. Ra-dio observations of its afterglow provid-ed an essential clue. The glow varied er-ratically by roughly a factor of two dur-ing the first three weeks, after which itstabilized and then began to diminish.The large variations probably had noth-ing to do with the burst source itself;rather they involved the propagation ofthe afterglow light through space. Just asEarth’s atmosphere causes visible star-light to twinkle, interstellar plasma caus-es radio waves to scintillate. For this pro-cess to be visible, the source must be sosmall and far away that it appears to usas a mere point. Planets do not twinkle,because, being fairly nearby, they looklike disks, not points.

Therefore, if GRB970508 was scin-tillating at radio wavelengths and thenstopped, its source must have grown froma mere point to a discernible disk. “Dis-cernible” in this case means a few light-weeks across. To reach that size, the sourcemust have been expanding at a consider-able rate—close to the speed of light.

The BeppoSAX and follow-up obser-vations have transformed astronomers’view of GRBs. The old concept of a sud-den release of energy concentrated in afew brief seconds has been discarded. In-deed, even the term “afterglow” is nowrecognized as misleading: the energy ra-diated during both phases is comparable.The spectrum of the afterglow is charac-teristic of electrons moving in a magneticfield at or very close to the speed of light.

GRB990123—the January 1999burst—was instrumental in demonstrat-ing the immense power of the bursts. Ifthe burst radiated its energy equally in alldirections, it must have had a luminosi-ty of a few times 1045 watts, which is1019 times as bright as our sun. Althoughthe other well-known type of cosmic cat-aclysm, a supernova explosion, releasesalmost as much energy, most of that en-ergy escapes as neutrinos, and the re-mainder leaks out more gradually thanin a GRB. Consequently, the luminosityof a supernova at any given moment is a

tiny fraction of that of a GRB. Evenquasars, which are famously brilliant,give off only about 1040 watts.

If the burst beamed its energy in par-ticular directions rather than in all direc-tions, however, the luminosity estimatewould be lower. Evidence for beamingcomes from the way the afterglow ofGRB990123, among others, dimmedover time. Two days into the burst, therate of dimming increased suddenly,which would happen naturally if the ob-served radiation came from a narrow jetof material moving at close to the speedof light. Because of a relativistic effect, theobserver sees more and more of the jet asit slows down. At some point, there is nomore to be seen, and the apparent bright-ness begins to fall off more rapidly [see il-lustration on next page]. For GRB-990123 and several other bursts, the in-ferred jet-opening angle is a few degrees.Only if the jet is aimed along our line ofsight do we see the burst. This beamingeffect reduces the overall energy emittedby the burst approximately in proportionto the square of the jet angle. For exam-ple, if the jet subtends 10 degrees, it cov-ers about one 500th of the sky, so the en-ergy requirement goes down by a factor

of 500; moreover, for every GRB that isobserved, another 499 GRBs go unseen.Even after taking beaming into account,however, the luminosity of GRB990123was still an impressive 1043 watts.

GRB-Supernova ConnectionONE OF THE MOST interesting dis-coveries has been the connection be-tween GRBs and supernovae. Whentelescopes went to look at GRB980425,they also found a supernova, designatedSN1998bw, that had exploded at aboutthe same time as the burst. The proba-bility of a chance coincidence was one in10,000. A more firm case is the associa-tion of GRB030329 with SN2003dh.This GRB was localized by NASA’s sec-ond High Energy Transient Explorersatellite (HETE-2), launched in October2000. The temporally and spatially co-incident supernova was discovered viaground-based optical observations, andthe broad spectroscopic features ofSN2003dh are basically identical tothose of SN1998bw.

A link between GRBs and supernovaehas also been suggested by the detectionof iron in the x-ray spectra of severalbursts. Iron atoms are known to be syn-

CO

RN

ELI

A B

LIK

(g

rap

h);

AN

DR

EW

FR

UC

HTE

R S

pa

ce T

eles

cop

e Sc

ien

ce I

nst

itu

te A

ND

NAS

A


Brightest gamma-ray burst yetrecorded went off on January 23,1999. Telescopes tracked itsbrightness in gamma rays (bluein graph), x-rays (green), visiblelight (orange) and radio waves(red). At one point, the rate ofdimming changed abruptly—atelltale sign that the radiationwas coming from narrow jets ofhigh-speed material. About twoweeks into the burst, after thevisible light had dimmed by afactor of four million, the HubbleSpace Telescope took a pictureand found a severely distortedgalaxy. Such galaxies typicallyhave high rates of starformation. If bursts are theexplosions of young stars, theyshould occur in just such a place.

Gamma rays

Robotic-telescope measurement

X-rays

Change in dimming rate

Visible light

Radio

0.1

10–9

10–12

10–15

10–18

10–21

Inte

nsity

(w

atts

per

squ

are

met

er)

1

0 1 2

Time (days)

Time (minutes)

10

16 days 59 days

FADING AWAY



thesized and dumped into interstellarspace by supernovae explosions. If theseatoms are stripped of their electrons andlater hook up with them again, they giveoff light at distinctive wavelengths, re-ferred to as emission lines. Early, margin-al detections of such lines by BeppoSAXand the Japanese x-ray satellite ASCA in1997 have been followed by more solidmeasurements. Notably, NASA’s Chan-dra X-ray Observatory detected iron linesin GRB991216, which yielded a directdistance measurement of the GRB. Thefigure agreed with the estimated distanceof the burst’s host galaxy.

Additional observations further sup-port the connection between GRBs andsupernovae. An iron-absorption featureappeared in the x-ray spectrum of GRB-990705. In the shell of gas around an-other burst, GRB011211, the EuropeanSpace Agency’s X-ray Multi-Mirror satel-lite found evidence of emission lines fromsilicon, sulfur, argon and other elementscommonly released by supernovae.

Although researchers still debate thematter, a growing school of thoughtholds that the same object can produce,in some cases, both a burst and a super-nova. Because GRBs are much rarer thansupernovae—every day a couple of GRBsgo off somewhere in the universe, as op-posed to hundreds of thousands of su-pernovae—not every supernova can beassociated with a burst. But some mightbe. One version of this idea is that super-novae explosions occasionally squirt outjets of material, leading to a GRB. In mostof these cases, astronomers would see ei-ther a supernova or a GRB, but not both.If the jets were pointed toward Earth,light from the burst would swamp lightfrom the supernova; if the jets wereaimed in another direction, only the su-

pernova would be visible. In some cases,however, the jet would be pointed justslightly away from our line of sight, let-ting observers see both. This slight mis-alignment would explain GRB980425.

Whereas this hypothesis supposesthat most or all GRBs might be related tosupernovae, a slightly different scenarioattributes only a subset of GRBs to su-pernovae. Roughly 90 of the bursts seenby BATSE form a distinct class of theirown, defined by ultralow luminositiesand long spectral lags, meaning that thehigh- and low-energy gamma-ray pulsesarrive several seconds apart. No oneknows why the pulses are out of sync. Butwhatever the reason, these strange GRBsoccur at the same rate as a certain type ofsupernova, called Type Ib/c, which occurswhen the core of a massive star implodes.

Great Balls of FireEVEN LEAVING ASIDE the question ofhow the energy in GRBs might be gener-ated, their sheer brilliance poses a para-dox. Rapid brightness variations suggestthat the emission originates in a small re-gion: a luminosity of 1019 suns comesfrom a volume the size of one sun. Withso much radiation emanating from sucha compact space, the photons must be sodensely packed that they should interactand prevent one another from escaping.The situation is like a crowd of peoplewho are running for the exit in such apanic that that nobody can get out. Butif the gamma rays are unable to escape,how can we be seeing GRBs?

The resolution of this conundrum, de-veloped over the past several years, is thatthe gammas are not emitted immediately.Instead the initial energy release of the ex-plosion is stored in the kinetic energy ofa shell of particles—a fireball—moving atclose to the speed of light. The particlesinclude photons as well as electrons andtheir antimatter counterpart, positrons.This fireball expands to a diameter of 10billion to 100 billion kilometers, bywhich point the photon density hasdropped enough for the gamma rays toescape unhindered. The fireball then con-verts some of its kinetic energy into elec-tromagnetic radiation, yielding a GRB.

The initial gamma-ray emission is


JUAN

VE

LASC

O

Relativity plays tricks on observers’view of jets from gamma-ray bursts.

1Moving at close to the speed oflight, the jet emits light in narrow

beams. Some beams bypass the observer.

2As the jet slows, the beams widen, so fewer of them bypass the observer.

More of the jet comes into view.

3Eventually beams from the edges reach the observer. The

entire jet is now visible. Data reveal this transition.

BEAM LINES

Central engine

Jet

Light beam

Observer

NEIL GEHRELS, LUIGI PIRO and PETER J. T. LEONARD bring both observation and theory tothe study of gamma-ray bursts. Gehrels and Piro are primarily observers—the lead scien-tists, respectively, of the Compton Gamma Ray Observatory and the BeppoSAX satellite.Leonard is a theorist, and like most theorists, he used to think it unlikely that the burstswere bright enough to be seen across the vastness of intergalactic space. “I have to admitthat the GRBs really had me fooled,” he says. Gehrels is head of the Gamma Ray, CosmicRay and Gravitational Wave Astrophysics Branch of the Laboratory for High Energy Astro-physics at the NASA Goddard Space Flight Center. Piro is a member of the Institute of SpaceAstrophysics and Cosmic Physics of the CNR in Rome. Leonard works for Science Systemsand Applications, Inc., in support of missions at Goddard.

THE

AU

THO

RS


most likely the result of internal shockwaves within the expanding fireball.Those shocks are set up when faster blobsin the expanding material overtake slow-er blobs. Because the fireball is expandingso close to the speed of light, the timescalewitnessed by an external observer is vast-ly compressed, according to the principlesof relativity. So the observer sees a burstof gamma rays that lasts only a few sec-onds, even if it took a day to produce. Thefireball continues to expand, and eventu-ally it encounters and sweeps up sur-rounding gas. Another shock wave forms,this time at the boundary between thefireball and the external medium, andpersists as the fireball slows down. Thisexternal shock nicely accounts for theGRB afterglow emission and the gradual

degradation of this emission from gam-ma rays to x-rays to visible light and, fi-nally, to radio waves.

Although the fireball can transformthe explosive energy into the observed ra-diation, what generates the energy to be-gin with? That is a separate problem, andastronomers have yet to reach a consen-sus. One family of models, referred to ashypernovae or collapsars, involves starsborn with masses greater than about 20to 30 times that of our sun. Simulationsshow that the central core of such a stareventually collapses to form a rapidly ro-tating black hole encircled by a disk ofleftover material.

A second family of models invokes bi-nary systems that consist of two compactobjects, such as a pair of neutron stars

(which are ultradense stellar corpses) or aneutron star paired with a black hole. Thetwo objects spiral toward each other andmerge into one. Just as in the hypernovascenario, the result is the formation of asingle black hole surrounded by a disk.

Many celestial phenomena involve ahole-disk combination. What distinguish-es this particular type of system is thesheer mass of the disk (which allows fora gargantuan release of energy) and thelack of a companion star to resupply thedisk (which means that the energy releaseis a one-shot event). The black hole anddisk have two large reservoirs of energy:the gravitational energy of the disk andthe rotational energy of the hole. Exactlyhow these would be converted into gam-ma radiation is not fully understood. It is

JUAN

VE

LASC

O


BURSTING OUTFormation of a gamma-ray burst could begin either withthe merger of two neutron stars or with the collapse of a massive star. Both these events create a black holewith a disk of material around it. The hole-disk system, inturn, pumps out a jet of material at close to the speed oflight. Shock waves within this material give off radiation.

MERGER SCENARIO

HYPERNOVA SCENARIO

Neutron stars

Black hole

Centralengine

Fasterblob

Slowerblob

Blobs collide (internal shock wave)

Gamma rays

Jet collides withambient medium (external shock wave)

X-rays,visiblelight,radiowaves

Massivestar

Disk

Preburst

Gamma-ray emission

Afterglow



possible that a magnetic field, 1015 timesmore intense than Earth’s magnetic field,builds up during the formation of thedisk. In so doing, it heats the disk to suchhigh temperatures that it unleashes a fire-ball of gamma rays and plasma. The fire-ball is funneled into a pair of narrow jetsthat flow out along the rotational axis.

Because the GRB emission is equallywell explained by both hypernovae andcompact-object mergers, some otherqualities of the bursts are needed to de-cide between these two scenarios. The as-sociation of GRBs with supernovae, forexample, is a point in favor of hyper-novae, which, after all, are essentiallylarge supernovae. Furthermore, GRBsare usually found just where hypernovaewould be expected to occur—namely, inareas of recent star formation withingalaxies. A massive star blows up fairlysoon (a few million years) after it is born,so its deathbed is close to its birthplace. In

contrast, compact-star coalescence takesmuch longer (billions of years), and in themeantime the objects will drift all over thegalaxy. If compact objects were the cul-prit, GRBs should not occur preferential-ly in star-forming regions.

Although hypernovae probably ex-plain most GRBs, compact-star coales-cence could still account for the poorlyunderstood short-duration GRBs. More-over, additional models for GRBs are stillin the running. One scenario producesthe fireball via the extraction of energyfrom an electrically charged black hole.This model suggests that both the imme-diate and the afterglow emissions areconsequences of the fireball sweeping upthe external medium. Astronomers havecome a long way in understanding gam-ma-ray bursts, but they still do not knowprecisely what causes these explosions,and they know little about the rich vari-ety and subclasses of bursts.

All these recent findings have shownthat the field has the potential for an-swering some of the most fundamentalquestions in astronomy: How do starsend their lives? How and where are blackholes formed? What is the nature of jetoutflows from collapsed objects?

Blasts from the PastONE OUTSTANDING question con-cerns the dark, or “ghost,” GRBs. Of theroughly 80 GRBs that have been local-ized and studied at wavelengths otherthan gamma rays, about 90 percent havebeen seen in x-rays. In contrast, onlyabout 50 percent have been seen in visi-ble light. Why do some bursts fail toshine in visible light?

One explanation is that these GRBs liein regions of star formation, which tend tobe filled with dust. Dust would block vis-ible light but not x-rays. Another intrigu-ing possibility is that the ghosts are GRBsthat happen to be very far away. The rel-evant wavelengths of light produced bythe burst would be absorbed by inter-galactic gas. To test this hypothesis, mea-surement of the distance via x-ray spectrawill be crucial. A third possibility is thatghosts are optically faint by nature. Cur-rently the evidence favors the dust expla-nation. High-sensitivity optical and radioinvestigations have identified the proba-ble host galaxies of two dark GRBs, andeach lies at a fairly moderate distance.

Another mystery concerns a class ofevents known as the x-ray-rich GRBs, orsimply the x-ray flashes. Discovered byBeppoSAX, later confirmed by reanaly-sis of BATSE data and currently ob-served by HETE-2, these bursts represent20 to 30 percent of GRBs. They give offmore x-radiation than gamma radiation;indeed, extreme cases exhibit no de-tectable gamma radiation at all.

One explanation is that the fireball isloaded with a relatively large amount ofbaryonic matter such as protons, makingfor a “dirty fireball.” These particles in-crease the inertia of the fireball, so thatit moves more slowly and is less able toboost photons into the gamma-rayrange. Alternatively, the x-ray flashescould be typical GRBs with jets that arepointing just out of our view, so that only C

OR

NE

LIA

BLI

K


STARS SPEND MOST OF THEIR LIVES in the relatively unexciting main-sequenceevolutionary phase, during which they casually convert hydrogen into helium in theircores via nuclear fusion. Our sun is in this phase. According to basic stellar theory,stars more massive than the sun shine more brightly and burn their fuel more quickly.A star 20 times as massive as the sun can keep going for only a thousandth as long.

As the hydrogen in the core of a star runs out, the corecontracts, heats up and starts to fuse heavier elements, suchas helium, oxygen and carbon. The star thus evolves into agiant and then, if sufficiently massive, a supergiant star. Ifthe initial mass of the star is at least eight times that of the sun,the star successively fuses heavier and heavier elementsin its interior until it produces iron. Iron fusion does notrelease energy—on the contrary, it uses up energy. Sothe star suddenly finds itself without any useful fuel.

The result is a sudden and catastrophic collapse. Thecore is thought to turn into a neutron star, a stellar corpsethat packs at least 40 percent more mass than the suninto a ball with a radius of only 10 kilometers. Theremainder of the star is violently ejected into space in apowerful supernova explosion.

There is a limit to how massive a neutron star canbe—namely, two to three times as massive as the sun. If it isany heavier, theory predicts it will collapse into a black hole. It can be pushed overthe line if enough matter falls onto it. It is also possible that a black hole can beformed directly during the collapse. Stars born with masses exceeding roughly 20solar masses may be destined to become black holes. The creation of these holesprovides a natural explanation for gamma-ray bursts. —N.G., L.P. and P.J.T.L.

THE DESTINIES OF MASSIVE STARS

Main-sequencephase

Supergiantphase

Explosion

Blackhole


the less collimated and less energetic x-rays reach us. Finally, most flashes mightcome from very distant galaxies—evenmore distant than the galaxies proposedto explain the ghost GRBs. Cosmic ex-pansion would then shift the gamma raysinto the x-ray range, and intergalactic gaswould block any visible afterglow. In fact,most of these x-ray flashes do not have adetectable visible-light counterpart, afinding consistent with this scenario. Ifeither x-ray flashes or ghost GRBs are lo-cated in extremely distant galaxies, theycould illuminate an era in cosmic histo-ry that is otherwise almost invisible.

The next step for GRB astronomy is toflesh out the data on burst, afterglow andhost-galaxy characteristics. Observersneed to measure hundreds of bursts of allvarieties: long and short, bright and faint,heavy in gamma rays or x-rays, burstswith visible-light afterglows and thosewithout. Currently astronomers are ob-taining burst positions from HETE-2 andthe Interplanetary Network, a series ofsmall gamma-ray detectors piggybackingon planetary spacecraft. The Swift mis-sion, scheduled for launch in the fall of2004, will offer multiwavelength obser-vations of hundreds of GRBs and their af-terglows, making automatic onboard x-ray and optical observations. A rapid re-sponse will determine whether the GRB

has an x-ray or visible afterglow. The mis-sion will be sensitive to short-durationbursts, which have barely been studiedthus far.

Another goal is to probe extreme gam-ma-ray energies. GRB940217, for exam-ple, emitted high-energy gamma rays formore than an hour after the burst, as ob-served by the Energetic Gamma Ray Ex-periment Telescope instrument on theCompton Gamma Ray Observatory. As-tronomers do not understand how suchextensive and energetic afterglows can be produced. The Italian Space Agency’sAGILE satellite, scheduled for launch in2005, will observe GRBs at these high en-ergies. The supersensitive Gamma-RayLarge Area Space Telescope mission, ex-pected to launch in 2007, will also be keyfor studying this puzzling phenomenon.

Other missions, though not designedsolely for GRB discovery, will also con-

tribute. The International Gamma-RayAstrophysics Laboratory, launched onOctober 17, 2002, is detecting about 10GRBs a year. The Energetic X-ray Imag-ing Survey Telescope, to launch a decadefrom now, will have a sensitive gamma-ray instrument capable of detecting thou-sands of GRBs.

The field has just experienced a seriesof breakthrough years, with the discoverythat GRBs are immense explosions oc-curring throughout the universe. Burstsprovide us with an exciting opportunity tostudy new regimes of physics and to learnwhat the universe was like at the earliestepochs of star formation. Space- andground-based observations over the com-ing years should allow us to uncover thedetailed nature of these most remarkablebeasts. Astronomers can no longer talk ofbursts as utter mysteries, but that does notmean the puzzle is completely solved.


Classes of Gamma-Ray BurstsEXPLANATION FOR PECULIAR

PROPERTIES

Not applicable

Extremely distant,obscured by dust,

or intrinsicallyfaint

Extremely distantor weighed downby extra particles

Does not occur ina star-forming

region, so ambientgas is less dense

and externalshocks are weaker

HYPOTHETICALCENTRALENGINE

Energeticexplosion ofmassive star



Merger of pair of compact

objects

AFTERGLOWVISIBLE

EMISSION

AFTERGLOWX-RAY

EMISSION

INITIALGAMMA-RAY

EMISSION

TYPICAL DURATION OF

INITIAL EMISSION(SECONDS)

20

20

30

0.3

PERCENTAGEOF ALL

BURSTS

25

30

25

20

BURST CLASS(SUBCLASS)

Long (normal)

Long (ghosts or

dark)

Long (x-ray-rich orx-ray flashes)

Short

? ?

Observation of X-ray Lines from a Gamma-Ray Burst (GRB991216): Evidence of Moving Ejectafrom the Progenitor. Luigi Piro et al. in Science, Vol. 290, pages 955–958; November 3, 2000.Preprint available at arxiv.org/abs/astro-ph/0011337

Gamma-Ray Bursts: Accumulating Afterglow Implications, Progenitor Clues, and Prospects. Peter Mészáros in Science, Vol. 291, pages 79–84; January 5, 2001. arxiv.org/abs/astro-ph/0102255

Blinded by the Light. Stan Woosley in Nature, Vol. 414, pages 853–854; December 20, 2001.

The Biggest Bangs: The Mystery of Gamma-Ray Bursts, the Most Violent Explosions in the Universe. Jonathan I. Katz. Oxford University Press, 2002.

Flash! The Hunt for the Biggest Explosions in the Universe. Govert Schilling. Cambridge University Press, 2002.



Documents

The Secret Lives of Stars - Physics and Astronomyphysics.bgsu.edu/~layden/TMP/A321/Articles/SciAm/secret_stars_2004.pdfTHE SECRET LIVES OF STARS 3Letter from the Editor YOUTH 4The