Aerosol formation

Atmospheric particlesfrom organic vapours

Aerosol particles produced over forestedareas may affect climate by acting asnuclei for cloud condensation, but

their composition (and hence the chemicalspecies that drive their production) remainsan open question. Here we show, to ourknowledge for the first time, that thesenewly formed particles (3–5 nm in diam-eter) are composed primarily of organicspecies, such as cis-pinonic acid and pinicacid, produced by oxidation of terpenes inorganic vapours released from the canopy1–4.

Our technique combines aerosol electri-cal-mobility size measurements5 with organ-ic (butanol) vapour growth rates in the cloud chamber of a modified condensation-particle counter6. Particles are introducedinto a condensing flow-tube cloud chamberwhere they are subjected to a supersaturated

organic vapour. During transit through theflow tube, particles nucleate into organiccloud-droplets which are detected by lightscattering on leaving the chamber.

Particles larger than 10 nm grow to thesame final droplet size, whereas smaller par-ticles grow less owing to the stronger Kelvineffect (lower effective supersaturation result-ing from increased surface curvature at thesesizes), resulting in a monotonic link betweeninitial particle size and detected pulse height.In this size region, for a given size, the pulse is also dependent on particle chemical composition, owing to its solubility in theorganic vapour. Therefore, if the initial sizeof the 3–10-nm particles is known, the com-position can be determined as a function ofgrowth in the organic vapour.

Figure 1a gives size-distribution mea-surements made before and during the initial stage of a nucleation event on 2 May2000 at the Hyytiälä forest research station1

over the boreal forest in Finland. Duringthe event, a clear nucleation mode (3–6 nm)

brief communications

NATURE | VOL 416 | 4 APRIL 2002 | www.nature.com 497

is evident, together with a pre-existing particle mode at sizes over 30 nm.

Figure 1b shows the detected pulse heightfrom droplet light scattering in the conden-sation-particle counter for the same periods.Before the event, the pulse height is typical ofthe pre-existing aerosol, with only a singledistinct pulse height (and therefore dropletsize) evident. During the event, additionalsmaller pulses are visible, corresponding topulses generated by the 3–6-nm particles.Also shown are the calculated pulse-heightresponses, based on laboratory calibrationsfor the observed nucleation mode, whichassume particle compositions of ammoniumsulphate, pinic acid and cis-pinonic acid. Thecalculated spectra are generated by taking thelaboratory pulse response for each size andcomposition and scaling it to the concentra-tion measured in the corresponding mobilitysize range (5-nm particle laboratory pulsesare shown as an example in Fig. 1b, inset).

A composition of ammonium sulphatefor the measured nucleation-mode size dis-tribution would not reproduce the observedpulse heights, but the observations agreewell with those expected for pinic acid, andeven better with those for cis-pinonic acid.Given that the growth of other commonatmospheric aerosol inorganic species inthe butanol vapour is very similar toammonium sulphate, this nucleation modecannot be inorganic in composition6.

Although our technique does not directlydemonstrate that the condensing vapourresponsible for producing the new particlesis cis-pinonic or pinic acid, it reveals that an organic vapour with similar solubility in butanol must be responsible for the phenomenon. Further, given that theseevents occur in clean air where the ratio ofbiogenic to anthropogenic volatile organiccompounds (VOCs) is high, and the fact thatbiogenic VOCs are considerably more reac-tive than anthropogenics7, it is unlikely thatthe particle production can be explained byanthropogenic species. Our results indicatethat aerosol formation over forests is drivenby condensable organic vapours.

Anthropogenic activities that result inincreased concentrations of pollutants suchas ozone will influence the conversion ofnatural VOCs into condensable vapours togenerate natural aerosols. Complex feedbackprocesses involving, for example, the coup-ling of emissions, radiative balance, andaerosol and cloud formation, possess un-certainties that must be determined if we areto predict future changes in global climate.Colin D. O’Dowd*†, Pasi Aalto*, Kaarle Hämeri‡, Markku Kulmala*,Thorsten Hoffmann§*Division of Atmospheric Sciences, Department ofPhysical Sciences, PO Box 64, University of Helsinki,00014 Helsinki, Finland†Department of Experimental Physics, National University of Ireland, University Road,

Figure 1 Size distribution of particles and their detection by a PHA (pulse-height analyser) condensation-particle counter. a, Particle size

distributions taken before (04:00 local time) a nucleation event and during the initial stages of the nucleation event (11:30) on 2 May

2000 using a differential-mobility particle sizer. b, Pulse-height spectra measured before and during a nucleation burst. Also shown are

calculated pulse-height spectra for the observed nucleation mode, based on laboratory calibrations for ammonium sulphate, pinic acid

and cis-pinonic acid (for example, 5-nm calibration is shown in the inset, together with 20-nm particles for comparison).

PHA channel

600 700 800 900 1,000

Nor

mal

ized

co

unts

10–4

10–3

10–2

Non-event

Event

Ammonium sulphate

Pinic acid

cis-pinonic acid

600 700 80010–4

10–3

10–2

Calibration pulses 20nm

Ammoniumsulphate 5 nm

Diameter (nm)

100 101 102 103

dN

/d lo

g D

(cm

–3)

102

103

104

cis-pinonic acid 5 nm

a

b

Before nucleationDuring nucleation

© 2002 Macmillan Magazines Ltd

Galway, Irelande-mail: colin.odowd@cmas.demon.co.uk‡Finnish Institute of Occupational Health,Topeliuksenkatu 41 a A, FIN-00250 Helsinki, Finland§Institute of Spectrochemistry and AppliedSpectroscopy, 44139 Dortmund, Germany1. Kulmala, M. et al. Tellus 53B, 324–342 (2001).

2. Leaitch, W. R. et al. J. Geophys. Res. 104, 8095–8111 (1999).

3. Kavouras, I. G., Mihalopoulos, N. & Stephanou, E. G. Nature

395, 683–686 (1998).

4. Birmili, W., Wiedensohler, A., Plass-Dülmer, C. &

Berresheim, H. Geophys. Res. Lett. 27, 2205–2208 (2000).

5. Mäkelä, J. M. et al. Geophys. Res. Lett. 24, 1219–1222 (1997).

6. Marti, J. J., Weber, R. J., Saros, M. T. & McMurry, P. H.

Aer. Sci. Technol. 25, 214–218 (1996).

7. Hoffmann, T. EUROTRAC Newslett. 21, 12–21 (1999).

Competing financial interests: declared none.

Nanomechanics

Response of a strainedsemiconductor structure

The nanomechanical properties of thinsilicon films will become increasinglycritical in semiconductor devices, par-

ticularly in the context of substrates that consist of a silicon film on an insulating layer(known as silicon-on-insulator, or SOI, sub-strates). Here we use very small germaniumcrystals as a new type of nanomechanicalstressor to demonstrate a surprisingmechanical behaviour of the thin layer of silicon in SOI substrates, and to show that

there is a large local reduction in the viscosityof the oxide on which the silicon layer rests.These findings have implications for the useof SOI substrates in nanoelectronic devices.

We use SOI substrates consisting of ahandle wafer (a thick silicon (Si) layer), athin oxide (a 400-nm thickness of SiO2) anda very thin (10 nm) template layer of crys-talline Si on top of the oxide. The templatelayer is patterned to form micrometre-sized(5–20 mm) mesas (Fig. 1a). About 10 mono-layers of germanium (Ge; total thickness1.6 nm) are deposited by molecular-beamepitaxy at 700 7C. Germanium has a latticeconstant 4% greater than that of silicon.

Figure 1b shows the formation of Genanocrystals (about 10 nm high, with 100-nm bases) that are crystallographicallyin register with the Si template, and ananomalous local bending of the Si templatelayer underneath each individual nanocrys-tal. The curvature underneath the islands isgreater than 0.005 nm11. This new mode oflocal bending (Fig. 2a) of a nanometre-scalethin film is different from the commonlyobserved extended, uniform bending mode(Fig. 2b) that is induced by strained-layerfilm growth on thick Si(001) (refs 1, 2).

Our calculations show that the localbending curvature depends on the nano-crystal’s density and shape (Fig. 2c, d). On athick substrate, local bending is suppressed3,resulting in an overall extended bending thatcan be estimated using Stoney’s formula4 andwhich is independent of nanocrystal densityand shape, as would be the case in a uniformfilm of equivalent thickness.

The local bending mode and large bend-ing magnitude indicate that the Si templatelayer behaves as a ‘free-standing’ layer duringthe growth of Ge nanocrystals, an outcomethat can be achieved if SiO2 acts as a fluidwith substantial viscous flow. The viscosityof SiO2 at 700 7C (the growth temperature) isordinarily much too great for such a largedegree of relaxation to occur. However, thisviscosity can decrease almost exponentiallywith increasing applied shear stress5.

From the bending curvature, and hencethe bending stress, we estimate5 that the viscosity of SiO2 can be reduced by three to five orders of magnitude in the regionsbeneath the bent Si layer below the Genanocrystals. The relaxation time for SiO2

flow is then reduced by a few orders of mag-nitude, to well within the deposition time ofabout 150 s. Thus, the large bending stressin the Si layer greatly enhances the viscousflow of SiO2, which in turn helps to increasethe bending of the Si layer, because the Sifilm can then behave as a free-standing film.

The local stressor on the thin Si templatelayer of SOI substrate modifies both themechanical properties of the Si layer and itselectronic properties, providing a uniquemethod for electronic (band) engineeringon a nanometre scale. For these reasons,

local stressors in SOI substrates could alsobecome a significant issue for the semicon-ductor industry, which is increasingly usingsuch substrates to manufacture devices.Feng Liu*, Paul Rugheimer†, E. Mateeva‡,D. E. Savage†, M. G. Lagally†*Department of Materials Science and Engineering,University of Utah, Salt Lake City, Utah 84112, USA†Department of Materials Science and Engineeringand Physics, University of Wisconsin, Madison,Wisconsin 53706, USAe-mail: lagally@engr.wisc.edu‡Colorado School of Mines, Golden, Colorado 80401, USA

1. Floro, J. A., Chason, E., Twesten, R. D., Hwang, R. Q. &

Freund, L. B. Phys. Rev. Lett. 79, 3946–3949 (1997).

2. Floro, J. A. et al. Phys. Rev. B 59, 1990–1998 (1999).

3. Johnson, H. T. & Freund, L. B. J. Appl. Phys. 81,

6081–6090 (1997).

4. Stoney, G. G. Proc. R. Soc. Lond. A 82, 172–178 (1909).

5. Rafferty, C. S., Borucki, L. & Dutton, R. W. Appl. Phys. Lett. 54,

1516–1518 (1989).

Competing financial interests: declared none.

brief communications

498 NATURE | VOL 416 | 4 APRIL 2002 | www.nature.com

Figure 2 Different bending modes of thin silicon-template layer in

SOI substrates induced by growing germanium nanocrystals.

a, Local bending mode that occurs only when the density of

nanocrystals (dotted domes) is low. b, Extended bending mode

that occurs at high nanocrystal densities. For a thin Si layer, a

transition from local to extended bending occurs with increasing

nanocrystal density. c, Some parameters used to calculate the

bending induced by a pyramidal nanocrystal acting as a local

stressor. ti is the thickness of the Ge nanocrystal, t s is the thick-

ness of the Si membrane, l is the base dimension of the

nanocrystal, em is the misfit strain in the nanocrystal, Z is the nor-

mal to the membrane, and L is the dimension of the bent region of

membrane underneath the nanocrystal. d, The calculated bending

curvature, K, of a free-standing, 10-nm Si layer as a function of

the equivalent film thickness, tGe, normalized to the Si-layer thick-

ness, tSi. Arrow indicates the curvature (local thickness, local

bending) that corresponds to the observed 10-nm Ge nanocrystal

height; this is consistent with the data in Fig. 1b.

Figure 1 Anomalous local bending of the thin silicon-template layer

in silicon-on-insulator (SOI) substrates induced by the growth of ger-

manium (Ge) nanocrystals. a, Mesa structures patterned and etched

on a SOI substrate wafer. b, Transmission electron microscope

image showing a Ge nanocrystal and the bent Si-template layer

underneath. The Ge nanocrystals, which grow in register with the Si

lattice, have a 4% greater lattice constant than the Si, and thus

cause it to bend locally. The local shear stress is sufficient to reduce

the viscosity of the oxide underneath the thin Si layer so as to allow

the oxide to flow locally. Scale bars, 20 mm (a) and 50 nm (b).

Si

Glue

SiO2

Ge

a

b

ts

l

L

Z

b

c

d

0.0 0.2 0.4 0.6 0.8 1.00.000

0.001

0.002

0.003

0.004

tGe/tSi

εm

K (n

m–1

)

a

t i

© 2002 Macmillan Magazines Ltd