Vibrational mode-specific reaction of methane on a nickel surface.
ABSTRACT The dissociation of methane on a nickel catalyst is a key step in steam reforming of natural gas for hydrogen production. Despite substantial effort in both experiment and theory, there is still no atomic-scale description of this important gas-surface reaction. We report quantum state-resolved studies, using pulsed laser and molecular beam techniques, of vibrationally excited methane reacting on the nickel (100) surface. For doubly deuterated methane (CD2H2), we observed that the reaction probability with two quanta of excitation in one C-H bond was greater (by as much as a factor of 5) than with one quantum in each of two C-H bonds. These results clearly exclude the possibility of statistical models correctly describing the mechanism of this process and attest to the importance of full-dimensional calculations of the reaction dynamics.
- [Show abstract] [Hide abstract]
ABSTRACT: Water dissociation on transition-metal catalysts is an important step in steam reforming and the water-gas shift reaction. To probe the effect of translational and vibrational activation on this important heterogeneous reaction, we performed state-resolved gas/surface reactivity measurements for the dissociative chemisorption of D2O on Ni(111), using molecular beam techniques. The reaction occurs via a direct pathway, because both the translational and vibrational energies promote the dissociation. The experimentally measured initial sticking probabilities were used to calibrate a first-principles potential energy surface based on density functional theory. Quantum dynamical calculations on the scaled potential energy surface reproduced the experimental results semiquantitatively. The larger increase of the dissociation probability by vibrational excitation than by translation per unit of energy is consistent with a late barrier along the O-D stretch reaction coordinate.Science 05/2014; 344(6183):504-507. · 31.20 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The laser control of the radiationless decay between the B3u(nπ*) and B2u(ππ*) states of pyrazine using the dynamic Stark effect has been investigated. A vibronic coupling model Hamiltonian in diabatic representation, including potential energy, transition dipole, and static polarizability surfaces as a function of the four most important vibrational modes of the molecule has been parametrized using multi-reference electronic structure calculations. The interaction of the molecule with a strong non-resonant laser pulse has been analyzed in terms of dressed potential energy surfaces. Because of the large polarizability difference between the vibronically coupled B3u(nπ*) and B2u(ππ*) states, the Stark effect induced by the non-resonant laser pulse shifts the conical intersection away from the Franck-Condon region. We have shown, by solving the time-dependent Schrödinger equation for the molecule interacting with a relatively weak pump pulse driving the electronic excitation from the ground state to the B2u(ππ*) state, and a strong non-resonant control pulse, that this control mechanism can be used to trap the wavepacket on the B2u(ππ*) potential energy surface for a much longer time than the natural B2u(ππ*) lifetime.The Journal of chemical physics. 05/2014; 140(19):194309.
- [Show abstract] [Hide abstract]
ABSTRACT: The six-dimensional quantum dynamics calculations for the dissociative chemisorption of HCl on Au(111) are carried out using the time-dependent wave-packet approach, based on an accurate PES which was recently developed by neural network fitting to density functional theory energy points. The influence of vibrational excitation and rotational orientation of HCl on the reactivity is investigated by calculating the exact six-dimensional dissociation probabilities, as well as the four-dimensional fixed-site dissociation probabilities. The vibrational excitation of HCl enhances the reactivity and the helicopter orientation yields higher dissociation probability than the cartwheel orientation. A new interesting site-averaged effect is found for the title molecule-surface system that one can essentially reproduce the six-dimensional dissociation probability by averaging the four-dimensional dissociation probabilities over 25 fixed sites.The Journal of Chemical Physics 11/2013; 139(18):184705. · 3.12 Impact Factor
croscope. Longer modulated waveguides
with shorter modulation periods should sig-
nificantly enhance the signal for several rea-
sons. First, m ? 1 QPM should become
possible by using shorter periods. Second, the
2.5-cm modulated waveguide length corre-
sponds to only 0.3 absorption depths in Ne at
284 eV. Previous theoretical studies have
indicated that 5 to 10 absorption depths are
ideal for generating a maximum signal (32).
Furthermore, because the highest HHG pho-
ton energy scales linearly with laser intensity,
using very reasonable laser parameters (i.e.,
pulsewidth ?10 fs, intensities ?5 ? 1015W
cm?2) and waveguides with 0.1 mm period-
icity, we should be able to generate high-
order quasi–phase-matched light at photon
energies approaching 1 keV.
References and Notes
1. R. Schoenlein et al., Science 287, 2237 (2000).
2. D. Attwood, Phys. Today 45, 24 (August, 1992).
3. C. Jacobsen, Trends Cell Biol. 9, 44 (1999).
4. G. Schneider et al., Surf. Rev. Lett. 9, 177 (February,
5. A. McPherson et al., JOSA B 4, 595 (1987).
6. M. Ferray et al., J. Phys. B 21, L31 (1987).
7. J. J. Macklin, J. D. Kmetec, C. L. Gordon, III, Phys. Rev.
Lett. 70, 766 (1993).
8. R. Haight, P. F. Seidler, Appl. Phys. Lett. 65, 517
9. D. Descamps et al., Opt. Lett. 25, 135 (2000).
10. M. Bauer et al., Phys. Rev. Lett. 8702, 5501 (2001).
11. L. Nugent-Glandorf, M. Scheer, D. Samuels, V. Bier-
baum, S. Leone, J. Chem. Phys. 117, 6108 (2002).
12. R. A. Bartels et al., Science 297, 376 (2002).
13. A. Rundquist et al., Science 280, 1412 (1998).
14. A. Paul et al., Nature 421, 51 (2003).
15. I. P. Christov, H. C. Kapteyn, M. M. Murnane, Optics
Express 7, 362 (2000).
16. Z. H. Chang, A. Rundquist, H. W. Wang, M. M. Mur-
nane, H. C. Kapteyn, Phys. Rev. Lett. 79, 2967 (1997).
17. C. Spielmann et al., Science 278, 661 (1997).
18. S. Backus, C. G. I. Durfee, G. A. Mourou, H. C. Kapteyn,
M. M. Murnane, Opt. Lett. 22, 1256 (1997).
waveguide passes through Zr or Ag filters (0.4 ?m) to
block the fundamental light, and the harmonics are
then spectrally dispersed onto an Andor CCD camera
with a Hettrick EUV spectrometer. The energy cali-
bration of the spectrometer was verified by recording
the positions of the Si, B, and C absorption edges, at
99.9 eV, 188.35 eV, and 284 eV, respectively.
20. C. G. Wahlstro ¨m et al., Phys. Rev. A 48, 4709 (1993).
21. K. C. Kulander, K. J. Schafer, J. L. Krause, in Super-
Intense Laser-Atom Physics, B. Piraux, A. L’Huillier, K.
Rzazewski, Eds. (Plenum, New York, 1993), vol. 316,
22. M. V. Ammosov, N. B. Delone, V. P. Krainov, Soviet
Physics JETP 64, 1191 (1986).
23. V. P. Krainov, JOSA B 14, 425 (1997).
24. I. P. Christov, M. M. Murnane, H. C. Kapteyn, Phys.
Rev. Lett. 78, 1251 (1997).
25. I. P. Christov, M. M. Murnane, H. C. Kapteyn, Phys.
Rev. A 57, R2285 (1998).
26. A. Scrinzi, M. Geissler, T. Brabec, Phys. Rev. Lett. 83,
27. Z. Chang et al., Phys. Rev. A 58, R30 (1998).
28. H. J. Shin et al., Phys. Rev. A 63, 053407 (2001).
29. M. Schnu ¨rer et al., Phys. Rev. Lett. 83, 722 (1999).
30. J. Zhou, J. Peatross, M. M. Murnane, H. C. Kapteyn, I. P.
Christov, Phys. Rev. Lett. 76, 752 (1996).
31. M. M. Fejer, G. A. Magel, D. H. Jundt, R. L. Byer, IEEE
J. Quantum Electron. 28, 2631 (1992).
32. E. Constant et al., Phys. Rev. Lett. 82, 1668 (1999).
33. This work was supported by the National Science
Foundation and the Department of Energy and made
use of facilities funded by the W. M. Keck Foundation.
1 July 2003; accepted 3 September 2003
Reaction of Methane on a
Rainer D. Beck,* Plinio Maroni, Dimitrios C. Papageorgopoulos,
Tung T. Dang, Mathieu P. Schmid, Thomas R. Rizzo
of natural gas for hydrogen production. Despite substantial effort in both
experiment and theory, there is still no atomic-scale description of this im-
portant gas-surface reaction. We report quantum state–resolved studies, using
pulsed laser and molecular beam techniques, of vibrationally excited methane
reacting on the nickel (100) surface. For doubly deuterated methane (CD2H2),
we observed that the reaction probability with two quanta of excitation in one
C-H bond was greater (by as much as a factor of 5) than with one quantum in
models correctly describing the mechanism of this process and attest to the
importance of full-dimensional calculations of the reaction dynamics.
The reaction of methane on a nickel catalyst to
form surface-bound methyl and hydrogen is the
rate-limiting step in steam reforming, which is
the principal process for industrial hydrogen
production as well as the starting point for the
large-scale synthesis of many important chem-
icals such as ammonia, methanol, and higher
hydrocarbons (1). Because of its importance,
the dissociation of methane on nickel has been
considered a prototype for chemical bond for-
mation between a polyatomic molecule and a
solid surface, with many experimental and the-
oretical studies directed at elucidating its mech-
anism (2–14). In view of the enormous eco-
nomic importance of this process (15), it would
be desirable to have a reliable theoretical de-
scription that could guide the development of
there is still no atomic-scale picture of the dy-
namics of this important gas-surface reaction.
Molecular beam experiments (4–6) have
firmly established that methane chemisorp-
tion is a direct process that can be activated
with about equal efficiency by both incident
kinetic energy normal to the surface and ther-
mal vibrational energy of the incident meth-
ane. State-resolved reactivity measurements
that used laser excitation of the asymmetric
stretch fundamental vibration (?3) (7) and
first overtone (2?3) (8) of CH4incident on
Ni(100) have confirmed the notion that vibra-
tional energy is similar to translational exci-
tation in its efficiency in promoting this re-
action. Theoretical treatments of methane
chemisorption have included wave packet
simulations with up to nine vibrational de-
grees of freedom (9), reduced-dimensionality
dynamical models with only a single C-H
stretch vibration (10–12), and a greatly sim-
plified statistical model (13). Despite having
diametrically opposed presuppositions, both
dynamical and statistical approaches claim to
reproduce existing experimental data (10,
16), although they make different predictions
about the role of methane vibrational excita-
tion in promoting the reaction. Some dynam-
ical calculations suggest that the reactivity of
vibrationally excited methane on nickel
should depend on the precise nature of the
vibrational mode (9, 17), whereas statistical
models predict the complete absence of such
effects (16). Although the reverse process—
the associative desorption of methane from
transition metal surfaces—seems to deviate
somewhat from a purely statistical model (18,
19), the experimental results reported thus far
do not exclude either approach, because there
is no reported evidence for mode specificity
in the surface reaction of methane.
In contrast, mode-specific reactivity of
methane in the gas phase has been observed.
Yoon et al. (20) have found that when meth-
ane is excited to the symmetric stretch-bend
combination ?1? ?4, it is more reactive with
atomic chlorine (by a factor of 1.9) than when
it is promoted to the nearly isoenergetic an-
tisymmetric combination ?3? ?4. In a simi-
lar study, Kim et al. (21) observed that the
product state distribution for the reaction of
CD2H2with chlorine depends on the initially
Laboratoire Chimie Physique Mole ´culaire, Ecole Poly-
technique Fe ´de ´rale de Lausanne (EPFL), CH-1015 Lau-
*To whom correspondence should be addressed. E-
R E P O R T S
3 OCTOBER 2003VOL 302SCIENCEwww.sciencemag.org
prepared reactant vibrational state. Two
states with similar energy but different C-H
bond stretching amplitudes produce a CD2H
methyl fragment in completely different vi-
brational states. Both studies show that vibra-
tional energy put into specific modes of
methane is not redistributed internally by the
reactive encounter, but instead contributes in
a mode-specific way to promoting the chem-
ical reaction. It remains unknown whether
similar nonstatistical dynamics would occur
in the encounter of a gas-phase molecule with
a solid surface.
To test for vibrational mode–specific be-
havior in gas-surface reactions, we performed
quantum state–resolved measurements in a
molecular beam surface science apparatus,
using pulsed laser preparation of vibra-
tionally excited methane molecules incident
on a Ni(100) single-crystal surface (22). Fig-
ure 1 shows a schematic illustration of our
experiment. After a deposition time adjusted
to produce a surface coverage of about 5% of
a monolayer (ML), the products of the disso-
ciative chemisorption of methane were de-
tected as surface-bound carbon by Auger
electron spectroscopy. We quantify the
amount of deposited carbon by recording C
and Ni Auger spectra in a line scan across the
surface and calibrate it by comparison with
the C/Ni Auger ratio for a known 0.5-ML
saturation coverage obtained by extended ex-
posure to a high-energy beam of methane.
We use the doubly deuterated isotopomer of
methane (CD2H2) rather than CH4for these
experiments because of its spectroscopic
properties (23). In CD2H2, combinations of
the two infrared (IR) active C-H stretch fun-
damentals (?1and ?6) form nearly isoener-
getic states with comparable absorption
strength but different nuclear motion, labeled
?20? and ?11? in local mode notation (24).
CD2H2in the molecular beam was excited
into either the ?20? or the ?11? C-H stretch local
mode state (25). A comparison of the surface
carbon Auger signals for incident CD2H2excit-
ed to these two states is shown in Fig. 2. To
containing 20% CD2H2in H2at normal inci-
dence for 15 min at two different positions on
the initially clean Ni(100) surface. For the first
deposition (left-hand peak), the ?20? state of
CD2H2with one unit of angular momentum
(J ? 1) was excited by 120-mJ pulses from an
IR laser tuned to the ? J ? 0 transition at
5879.8 cm?1, and for the second deposition
(right-hand peak), the J ? 1 level of the ?11?
state was prepared using the same IR pulse
energy to excite the corresponding transition at
6000.2 cm?1. We used cavity ring down spec-
troscopy (CRDS) (26) in a separate free-jet
expansion to ensure that the excitation laser
remained resonant with the selected transition
throughout the deposition. CRDS was also used
to determine an upper limit for the rotational
temperature of 9 K for the CD2H2in the beam
and to measure the absolute transition strength.
Although the transition we used to prepare the
?20? level is weaker than that used to excite the
?11? level by a factor of 1.5 ? 0.1, the former
leads to a carbon signal at least three times as
large, indicating clear mode-specific behavior.
Control experiments such as reversing the order
and surface location of the deposition did not
change the result. Experiments under identical
beam conditions but without laser excitation
showed no detectable carbon signal above the
background. To measure the reactivity of the
unexcited methane beam, we increased the flux
by a factor of 80 and used a deposition time of
up to 110 min. Our laser-off measurements
represent only an upper limit for the reactivity
of CD2H2in the vibrational ground state be-
cause of a small fraction of thermally ex-
cited CD2H2in the molecular beam. Both
laser-on and laser-off measurements were
made for a series of incident kinetic energies.
At each energy, the experiment was repeated
up to 10 times. Figure 3 shows the state-
resolved sticking coefficients for CD2H2de-
termined from these measurements.
Fig. 1. Schematic of experimental setup. We prepared methane under collision-free conditions by
expanding it in a mixture with hydrogen gas through a solenoid-actuated pulsed valve operating at
20 Hz. The resulting supersonic jet expansion accelerates the methane molecules to a kinetic
energy controlled by the methane/H2seed ratio and the valve temperature, as determined by
time-of-flight measurements. The free-jet expansion was skimmed, and the resulting molecular
beam was sliced into 30-?s pulses by a chopper wheel. Before the molecular beam impinged on the
sample surface (Ni single crystal 10 mm in diameter, cut within 0.1° of the 100 plane), it was
irradiated by pulsed, tunable IR light to prepare a fraction (1 to 2%) in a specific ro-vibrationally
excited quantum state. Sticking coefficients averaged over all vibrational states of methane
populated in the beam were obtained from the ratio of the carbon coverage to the incident
methane dose. Sticking coefficients for the state-selected methane molecules were calculated from
the difference obtained with the laser on and off, because contributions from the small amount of
thermally excited vibrational states in the beam drop out of this difference (7).
Fig. 2. Surface carbon Auger signal for
identical doses of CD2H2excited to
the ?20? and ?11? vibrational states
incident on a Ni(100) surface at kinet-
ic energy of 41 kJ/mol. The dotted line
indicates the background level of car-
bon accumulated during the deposi-
tion and analysis time.
Fig. 3. State-resolved sticking coefficients for
CD2H2in (from top to bottom) the ?20? (}),
?11? (F), and ground (I) vibrational states on
Ni(100) as a function of incident kinetic energy
normal to the surface. The surface temperature
is 473 K.
R E P O R T S
www.sciencemag.orgSCIENCE VOL 3023 OCTOBER 2003
At 41 kJ/mol, the lowest incident kinetic
energy investigated so far, we find that CD2H2
is 5.4 times as reactive when promoted to the
for both states is enhanced by several orders of
magnitude with respect to incident molecules in
the ground vibrational state with the same ki-
netic energy. The relative reactivity of the ?20?
and ?11? states decreases with increasing kinetic
of 80 kJ/mol. At still higher kinetic energy we
accurate determination of the absolute reactivi-
ty becomes increasingly difficult as a result of
the higher reactivity of the ground-state mole-
cules. This decrease in mode specificity is like-
ly due to the increase of the total amount of
available energy relative to the reaction barrier.
As the reaction probability approaches its as-
ymptotic value, the difference between the two
vibrational modes is expected to decrease. On
the other hand, the mode selectivity should be
even larger at lower kinetic energy.
The increased reactivity of the ?20? level
relative to ?11? can be rationalized in terms of
their different vibrational amplitudes: The
former contains two quanta of stretch vibra-
tion in a single C-H bond, whereas the latter
contains one quantum in each of two C-H
bonds. In the gas-phase reaction of CD2H2
with chlorine, the product state distributions
observed by Kim et al. (21) confirm this local
mode description by demonstrating that one
of the two bonds acts as a spectator during the
reaction. The same description also rational-
izes our observation of higher reactivity of
the ?20? state in terms of the larger vibrational
amplitude along the C-H bond relative to
?11?. This difference in reactivity implies that
the C-H bond stretch has a substantial pro-
jection on the reaction coordinate, in agree-
ment with ab initio calculations of the transi-
tion-state structure (27).
Our observation of vibrational mode–
specific gas-surface reactivity has important
implications for theoretical treatments of this
process. Mode-specific reactivity is inconsis-
tent with the existence of a transient phy-
sisorbed complex, as has been suggested in
the statistical model proposed by Harrison
(13). His model assumes complete intramo-
lecular redistribution of the initial vibrational
energy in methane as it transiently resides in
a local “hot spot” and weakly interacts with a
limited number of surface atoms, and it de-
termines rates for desorption and dissociation
according to microcanonical rate theory. As a
result, it predicts a reactivity that scales only
with the total available energy independent of
vibrational mode, which is inconsistent with
our experimental results. In contrast to the
assumptions of this (or any) statistical model,
our observation that CH2D2retains a clear
memory of the initially prepared quantum
state indicates that its interaction with the
metal surface does not induce extensive in-
tramolecular energy redistribution before the
In addition to excluding statistical as-
sumptions, the observation of mode speci-
ficity in the reaction probability provides
guidance for dynamical models. It suggests
that a realistic description of the chemical
dynamics will need to go beyond low di-
mensionality. Calculations including all
nine vibrational degrees of freedom of the
incident molecule are now starting to be-
come feasible (9, 28), and our experimental
results provide stringent tests for such cal-
culations as well as reassurance that efforts
in this direction are warranted.
References and Notes
1. P. L. Spath, M. K. Mann, National Renewable Energy
Laboratory Technical Report NREL/TP-570-27637
2. H. J. Larsen, I. Chorkendorff, Surf. Sci. Rep. 35, 163
3. F. Besenbacher et al., Science 279, 1913 (1998).
4. C. T. Rettner, H. E. Pfnu ¨r, D. J. Auerbach, Phys. Rev.
Lett. 54, 2716 (1985).
5. M. B. Lee, Q. Y. Yang, S. T. Ceyer, J. Chem. Phys. 87,
6. P. M. Holmblad, J. Wambach, I. Chorkendorff,
J. Chem. Phys. 102, 6255 (1995).
7. L. B. F. Juurlink, P. R. McCabe, R. R. Smith, C. L.
DiCologero, A. L. Utz, Phys. Rev. Lett. 83, 868 (1999).
8. M. P. Schmid, P. Maroni, R. D. Beck, T. R. Rizzo,
J. Chem. Phys. 117, 8603 (2002).
9. R. Milot, A. P. J. Jansen, Phys. Rev. B 61, 15657 (2000).
10. A. C. Luntz, J. Chem. Phys. 102, 8264 (1995).
11. M.-N. Carre ´, B. Jackson, J. Chem. Phys. 108, 3722
12. Y. Xiang, J. Z. H. Zhang, J. Chem. Phys. 118, 8954
13. V. A. Ukraintsev, I. Harrison, J. Chem. Phys. 101, 1564
14. J. Higgins, A. Conjusteau, G. Scoles, S. L. Bernasek,
J. Chem. Phys. 114, 5277 (2001).
15. I. Maxwell, Stud. Surf. Sci. Catal. 101, 1 (1996).
16. A. Bukoski, I. Harrison, J. Chem. Phys. 118, 9762
17. L. Halonen, S. L. Bernasek, D. J. Nesbitt, J. Chem. Phys.
115, 5611 (2001).
18. K. Watanabe, M. C. Lin, Y. A. Gruzdkov, Y. Matsumoto,
J. Chem. Phys. 104, 5974 (1996).
19. H. Mortensen, L. Diekho ¨ner, A. Baurichter, A. C. Luntz,
J. Chem. Phys. 116, 5781 (2002).
20. S. Yoon, S. Henton, A. N. Zivkovic, F. F. Crim, J. Chem.
Phys. 116, 10752 (2002).
21. Z. H. Kim, H. A. Bechtel, R. N. Zare, J. Am. Chem. Soc.
123, 12714 (2001).
22. M. P. Schmid, P. Maroni, R. D. Beck, T. R. Rizzo, Rev.
Sci. Instrum., 74, 4110 (2003).
23. J. L. Duncan, M. M. Law, Spectrochim. Acta A 53, 1445
24. M. S. Child, L. Halonen, Adv. Chem. Phys. 57, 1
25. Of the two possible linear combinations of the ?20?
state, we excited the antisymmetric one, ?20?–.
26. K. W. Busch, M. A. Busch, Eds., Cavity Ringdown
Spectroscopy, vol. 720 of ACS Symposium Series
(American Chemical Society, Washington, DC, 1999).
27. H. Yang, J. L. Whitten, J. Chem. Phys. 96 (1992).
28. G. J. Kroes, A. Gross, E. J. Baerends, M. Schieffler, D. A.
McCormack, Acc. Chem. Res. 35, 193 (2002).
29. Supported by Swiss National Science Foundation
grants 200020-100058/1 and by the EPFL.
9 July 2003; accepted 19 August 2003
Identification of a Plant Nitric
Oxide Synthase Gene Involved in
Fang-Qing Guo, Mamoru Okamoto, Nigel M. Crawford*
Nitric oxide (NO) serves as a signal in plants. An Arabidopsis mutant (Atnos1)
was identified that had impaired NO production, organ growth, and abscisic
in Atnos1 mutant plants resulted in overproduction of NO. Purified AtNOS1
protein used the substrates arginine and nicotinamide adenine dinucleotide
phosphate and was activated by Ca2?and calmodulin-like mammalian endo-
thelial nitric oxide synthase and neuronal nitric oxide synthase, yet it is a
AtNOS1 encodes a distinct nitric oxide synthase that regulates growth and
hormonal signaling in plants.
Nitric oxide (NO) functions as a signal and a
cytotoxic agent in many physiological and
immunological processes in animals (1–3)
and is synthesized by nitric oxide synthase
(NOS) (2, 4–6). There are three known iso-
forms of NOS [neuronal (nNOS), inducible
(iNOS), and endothelial (eNOS)], and each
contains a heme-oxygenase and a flavin re-
ductase domain. Related NOS genes have
been described in many eukaryotic species
from vertebrates to fungi (6, 7). Bacteria also
encode NOS proteins that are smaller, con-
taining only the oxygenase domain, which is
similar in sequence and biochemical activity
to the mammalian NOS oxygenase (8–10).
In plants, NO serves as a signal in hor-
monal and defense responses (3, 7, 11–14).
The source of NO synthesis in plants has
Section of Cell and Developmental Biology, Division
of Biological Sciences, University of California at San
Diego, La Jolla, CA 92093–0116, USA.
*To whom correspondence should be addressed. E-
R E P O R T S
3 OCTOBER 2003VOL 302SCIENCEwww.sciencemag.org