2

Click here to load reader

document

Embed Size (px)

Citation preview

Page 1: document

Nature © Macmillan Publishers Ltd 1998

8

Null models are used to assess departuresfrom randomness in other fields such aspalaeobiology10 and the reconstruction ofphylogeny11, so this new analytical proce-dure will be useful beyond community ecol-ogy. Some interlocking statistical and bio-logical concerns remain, however, centredon the problem of interdependencies.

For example, in the Vanuatu analysispairs occurring nine and only nine times,like the twos, are improbably rare, but pairsoccurring ten and only ten times areimprobably common. These opposite direc-tions in deviation of ‘neighbours’ in the fre-quency distribution (and there are others)hint at further structure in the data, struc-ture that may be statistical12,13 or biological.The source of that structure might lie in thefull set of 56 bird species being analysedtogether, rather than just guilds of potentialcompetitors, and in the artificiality (andconvenience) of holding the number ofislands per species constant, for which thereis no biological theory; in contrast thereis good biogeographical theory14 to justifyholding the number of species per islandconstant in the null modelling.

This raises the issue of how a biologicallyrealistic null model can be constructed forcommunities of species in an archipelagowhen the source area and set of species can-not be specified; and of how sequential colo-nization and local adaptation, geographyand history15, can be adequately built intothe null model. In fact, how much biology

should be incorporated into the null model? Here we have echoes of a major issue in

population genetics — whether variation ismaintained by random drift or selection. Weare left, as Diamond2 was, with a challengingproblem, that of determining the relativeinfluences of chance and necessity (system-atic forces, such as competition) in theassembly of biological communities throughtime.Peter R. Grant is in the Department of Ecology andEvolutionary Biology, Princeton University,Princeton, New Jersey, 08544-1003, USA.e-mail: [email protected]. MacArthur, R. H. Geographical Ecology (Princeton Univ. Press,

1972).

2. Diamond, J. in Ecology and Evolution of Communities (eds

Cody, M. L. & Diamond, J. M.) (Harvard Univ. Press, 1975).

3. Sanderson, J. G., Moulton, M. P. & Selfridge, R. G. Oecologia

116, 275–283 (1998).

4. Connor, E. F. & Simberloff, D. Ecology 60, 1132–1140 (1979).

5. Wiens, J. A. The Ecology of Bird Communities, Vol. 1

(Cambridge Univ. Press, 1989).

6. Gotelli, N. J. & Graves, G. R. Null Models in Ecology

(Smithsonian Press, Washington, DC, 1996).

7. Grant, P. R. & Abbott, I. Evolution 34, 332–341 (1980).

8. Colwell, R. K. & Winkler, D. W. in Ecological Communities (eds

Strong, D. R. Jr, Simberloff, D., Abele, L. G. & Thistle, A. B.)

(Princeton Univ. Press, 1984).

9. Harvey, P. H., Colwell, R. K., Silvertown, J. W. & May, R. M.

Annu. Rev. Ecol. Syst. 14, 189–211 (1985).

10.Raup, D. M., Gould, S. J., Schopf, T. J. M. & Simberloff, D. S.

J. Geol. 81, 525–542 (1973).

11.Harvey, P. H., Brown, A. J. L., Maynard Smith, J. M. & Nee, S.

(eds) New Uses for New Phylogenies (Oxford Univ. Press, 1996).

12.Roberts, A. & Stone, L. Oecologia 83, 560–567 (1990).

13.Stone, L. & Roberts, A. Oecologia 91, 419–424 (1992).

14.MacArthur, R. H. & Wilson, E. O. The Theory of Island

Biogeography (Princeton Univ. Press, 1967).

15.Ricklefs, R. E. & Schluter, D. Species Diversity in Ecological

Communities (Univ. Chicago Press, 1996).

news and views

NATURE | VOL 396 | 19 NOVEMBER 1998 | www.nature.com 217

Photonics

An atomic dimmer switchPhilip H. Bucksbaum

On page 239 of this issue1, DoronMeshulach and Yaron Silberbergof the Weizmann Institute show how

atoms can be made to absorb light more orless readily, or not at all, simply by control-ling the phase of light shone at them. Thisjoins a growing list of new ‘coherent control’methods for manipulating the internalquantum dynamics of atoms and moleculesusing the coherence properties of light,rather than its intensity or colour.

All atoms and molecules can absorb light.The absorption usually occurs for light thathas a frequency n at resonance — that iswhere the energy of a single photon in thelight beam, hn, equals the energy differencebetween the ground state and an excitedquantum state of the system. But when anatom is subjected to intense light, such as thatproduced by a laser, nonlinear absorption isalso possible. In nonlinear absorption, twoor more photons pool their energy to excitethe atom, and the sum of the photon energiesequals the excited-state energy difference.The photons must not only have the right

total energy, they must also arrive at thetarget at nearly the same time, so nonlinearoptical effects are usually stronger when thelight is compressed into a short, intensepulse. This is the principle behind severalpractical devices in lasers such as harmonicfrequency converters.

One reason that the arrangement ofphotons is important in a nonlinear processis quantum interference. When a quantumdynamical process can take more than onepath, interference occurs between the differ-ent possible routes. The interference may bedestructive or constructive, slowing or has-

tening the process. An example is the two-photon absorption experiment describedby Meshulach and Silberberg: if the photonsare both present at the same time, then theabsorption may take two paths, correspond-ing to absorption of photon 1 first, followedby photon 2, or the other way around.Coherent control provides a way to adjustthe quantum interferences between thesedifferent paths, affecting the rate of thewhole process.

This is easier to understand if we stoptalking about photons and consider the lightas a classical electromagnetic wave. The lightpulse is a travelling wave with frequency nand electric field amplitude A, so that anatom will experience the light as an oscillat-ing electric field E = Acos(2pnt) where t istime. Figure 1 is a picture of the electric fieldversus time for a typical short laser pulse. Theturn-on and turn-off create a frequencyspread, shown in the spectral density of thispulse given by its Fourier transform (Fig. 2a).

An atom with a resonance at any frequen-cy in this spectrum can be excited by absorb-ing a single photon of light from the pulse.Each frequency present can have differentphases as well as different amplitudes — thatis, the light with frequency n1 might be a sinewave, sin(2pn1t) while the light with n2

might be a cosine wave cos(2pn2t) or anyintermediate phase. In Fig. 1, all the differentfrequencies happen to be cosine waves.

Figure 1 The electric field of a short laser pulse.

Figure 2 Spectral power distributions. a, Thespectrum of the pulse in Fig. 1. b, The two-photon spectrum of the same pulse. c, The two-photon spectrum of a similar pulse, whosespectral components with frequency greaterthan the mean have been reversed in sign. Thetwo-photon nonlinear absorption rate issharpened.

Page 2: document

Nature © Macmillan Publishers Ltd 1998

8

Now if there is no resonance at the prin-cipal frequency, but one at twice that fre-quency, then absorption only occurs forpairs of photons. The atom no longerresponds resonantly to the applied electricfield, but rather to the square of the fieldE2(t)=A2(t)cos2(2pnt), (the probability of asingle photon’s presence is proportional tothe field, so the probability of two being pre-sent at once is proportional to the square).That has a completely different frequencyspectrum (Fig. 2b), one peak of which hap-pens to be at twice the laser frequency — thatis at the position of the nonlinear resonancefor two-photon absorption.

If one now adjusts the various phases ofthe different colours (or frequencies) in theinitial spectrum, but leaves all of their ampli-tudes just the same, then the spectral energydensity of the light field E(t) stays the same,but the spectral density of the nonlinear fieldshown in Fig. 2b can change. For example,let’s reverse the sign of the upper half of thespectrum in Fig. 2a; in other words, trans-form the spectral components above thelaser frequency from cosine waves to minuscosine waves. The power spectrum of E2(t)now looks like Fig. 2c. By manipulatingphases, the nonlinear spectral density canchange so much that any particular resonantfrequency in the spectrum can be made tovanish, and absorption at that frequency willvanish as well.

To achieve this degree of control, theWeizmann group made use of a new tool inoptics called a pulse shaper2 (Fig. 3). It isessentially two spectrometers arrangedback-to-back. The first spectrometer dis-perses the light into its various colours,which are displayed along the exit plane ofthe device. At this exit plane are a series offilters side-by-side, which can attenuate ordelay separate segments of the pulse spec-trum. Meshulach and Silberberg’s filterswere parts of a programmable liquid crystaldisplay, not very different from the one usedin laptop computers, which could adjust the

phase delay of each frequency band. The sec-ond spectrometer then puts the separatedcolours back together again, and the resultis a pulse with the same colours as before,but arranged with different relative phasesand/or amplitudes.

To the original pulse the researchersadded a frequency modulation that changesthe phase, not the amplitude, of any frequen-cy; but its effect on the nonlinear absorptionspectrum was dramatic. As they describe intheir paper, the two-photon resonance couldeven be made to vanish.

This dependence of quantum processeson the coherence of the driving radiation hasalready been used to control molecular disso-ciation and ionization pathways3, current insemiconductors4, and the direction of photo-emission5. Ultrafast pulse shapers, whichmake it possible to manufacture very compli-cated phase distributions, have been used intime-domain coherent control experiments,where the control is achieved by manipulat-ing the internal motion of a molecule, toguide it to a particular set of final states6. Pulseshapers have also been used to shape quan-

tum wave packets in atoms7. The new use ofcomputer-controlled programmable pulseshapers in nonlinear interference experi-ments adds further capabilities to the field ofcoherent control. This could facilitate differ-ent chemical reactions in applications wherelaser chemistry is important, such as photo-lithography. A shaped nonlinear spectrumcould also aid or inhibit atomic motion inmolecules or in solids, where the energies ofthe two photons subtract rather than add(Raman interference).Philip H. Bucksbaum is in the Department ofPhysics, University of Michigan, Ann Arbor,Michigan 48109-1120, USA.e-mail: [email protected]. Meshulach, D. & Silberberg, Y. Nature 396, 239–242 (1998).

2. Weiner, A. M., Heritage, J. P. & Kirschner, E. M. J. Opt. Soc. Am.

B 5, 1563–1572 (1988).

3. Gordon, R. J. & Rice, S. A. Annu. Rev. Phys. Chem. 48, 601–641

(1997).

4. Dupont, E., Corkum, P. B., Liu, H. C., Buchanan, M. &

Wasilewski, Z. R. Phys. Rev. Lett. 74, 3596–3599 (1995).

5. Yin, Y. Y., Elliott, D. S. & Shehadeh, R. Chem. Phys. Lett. 241,

591–596 (1995).

6. Bardeen, C. J. et al. J. Phys. Chem. A 101, 3815–3822 (1997).

7. Weinacht, T. C., Ahn, J. & Bucksbaum, P. H. Phys. Rev. Lett. 80,

5508–5511 (1998).

news and views

NATURE | VOL 396 | 19 NOVEMBER 1998 | www.nature.com 219

Figure 3 A pulse shaper, capable of imposingselective phase changes on a laser pulse.

The major histocompatibility complex(MHC) is unambiguously part of theimmune system. Generally speaking

the immune system operates outside theblood–brain barrier, and this has been corre-lated with the belief that MHC glycoproteinsare not expressed in the normal central ner-vous system. But this comforting correlationis now challenged by the radical observationsof Carla Shatz and colleagues, published inNeuron1. They not only demonstrate, for thefirst time, considerable amounts of surface-expressed MHC glycoproteins in the opticsystem, but they also show that theseproteins may be involved there in synapticremodelling during development.

Shatz and co-workers were searching forgenes that are differentially regulated in aregion of embryonic cat brain called the lat-eral geniculate nucleus (LGN). They lookedspecifically between days 42 and 52 of gesta-tion, when retinal input from the two eyessegregates into eye-specific zones. This activ-ity-dependent repatterning of synaptic con-nectivity is blocked by tetrodotoxin, so theauthors compared transcription in the LGNin the presence and absence of this toxin.They found just one messenger RNA thatwas differentially expressed, and this turnedout, most surprisingly, to be the product of aknown feline class I MHC gene.

Class I molecules usually present pep-tides derived from intracellular pathogens(such as viruses and bacteria) to the immune

system. In the brain, most normal, restingcells have vanishingly low levels of MHCclass I (ref. 2). Now, however, Shatz andcolleagues show, through elegant in situhybridizations using feline class I MHCprobes, that spatio-temporal expression ofboth the mRNA and protein correlatesclosely with synaptic remodelling in thedevelopment of binocular vision.

Shatz et al. demonstrate that expressionof class I MHC within the LGN is highest inthe late fetal and early neonatal stages ofdevelopment. This period is well after axonsfrom the retina first reach the LGN, but iswhen their terminal processes segregate intoeye-specific zones and their morphology isfine-tuned (Fig. 1a, overleaf). Segregationand fine-tuning depend on neuronalactivity3,4, but it is the pattern rather than thelevel of activity that is important. In the latefetal period, populations of retinal ganglioncells spontaneously fire short bursts of actionpotentials and then remain silent5. Althoughneurons in the same place tend to fire at thesame time within each retina, there is no cor-relation between firing patterns in the twoeyes. Tetrodotoxin blocks the eventual segre-gation of inputs from the two eyes in thebrain, so this simultaneous firing seems to bea kind of link among the ganglion cells ofeach eye. Thus, if ‘neurons that fire together,wire together’, the pattern of spontaneousactivity will ensure segregation of input fromthe two eyes in the LGN.

Developmental neurobiology

First class way to develop a brainJonathan Howard and Ian Thompson