What are the contributions mentioned in the paper "The structure of the first coordination shell in liquid water" ?

In this paper, the authors present a survey of the state-of-the-art work in this area.

(Open Access) The Structure of the First Coordination Shell in Liquid Water (2004) | Ph. Wernet

15. J. Higgins, A. Conjusteau, G. Scoles, S. L. Bernasek,

J. Chem. Phys. 114, 5277 (2001).

16. M. J. Murphy, J. F. Skelly, A. Hodgson, B. Hammer,

J. Chem. Phys. 110, 6954 (1999).

17. A. C. Luntz, J. Chem. Phys. 113, 6901 (2000).

18. L. Diekhoner et al., J. Chem. Phys. 115, 9028 (2001).

19. D. C. Jacobs, J. Phys. Condens. Matter 7, 1023 (1995).

20. G. O. Sitz, Rep. Progr. Phys. 65, 1165 (2002).

21. P. R. McCabe, L. B. F. Juurlink, A. L. Utz, Rev. Sci.

Instrum. 71, 42 (2000).

22. H. Hou et al., Science 284, 1647 (1999).

23. M. Gostein, G. O. Sitz, J. Chem. Phys. 106, 7378

(1997).

24. Details are available on Science Online.

25. K. Christmann, O. Schober, G. Ertl, M. Neumann,

J. Chem. Phys. 60, 4528 (1974).

26. Q. Y. Yang, K. J. Maynard, A. D. Johnson, S. T. Ceyer,

J. Chem. Phys. 102, 7734 (1995).

27. J. D. Beckerle, M. P. Casassa, R. R. Cavanagh, E. J.

Heilweil, J. C. Stephenson, Phys. Rev. Lett. 64, 2090

(1990).

28. R. Milot, A. P. J. Jansen, Phys. Rev. B 61, 15657 (2000).

29. L. Halonen, S. L. Bernasek, D. J. Nesbitt, J. Chem. Phys.

115, 5611 (2001).

30. A. Bukoski, I. Harrison, J. Chem. Phys. 118, 9762

(2003).

31. R. D. Beck et al., Science 302, 98 (2003).

32. A. C. Luntz, Science 302, 70 (2003).

33. H. Mortensen, L. Diekhoner, A. Baurichter, A. C. Luntz,

J. Chem. Phys. 116, 5781 (2002).

34. Supported by NSF (grant CHE-0111446) and Tufts

University.

Supporting Online Material

www.sciencemag.org/cgi/content/full/304/5673/992/

DC1

Materials and Methods

2 February 2004; accepted 5 April 2004

The Structure of the First

Coordination Shell in

Liquid Water

Ph. Wernet,

1,2

D. Nordlund,

U. Bergmann,

M. Cavalleri,

M. Odelius,

H. Ogasawara,

1,3

L.Å.Na¨slund,

1,3

T. K. Hirsch,

L. Ojama¨e,

P. Glatzel,

L. G. M. Pettersson,

A. Nilsson

1,3

X-ray absorption spectroscopy and x-ray Raman scattering were used to probe the

molecular arrangement in the ﬁrst coordination shell of liquid water. The local

structure is characterized by comparison with bulk and surface of ordinary hex-

agonal ice Ih and with calculated spectra. Most molecules in liquid water are in two

hydrogen– bonded conﬁgurations with one strong donor and one strong acceptor

hydrogen bond in contrast to the four hydrogen– bonded tetrahedral structure in

ice. Upon heating from 25°C to 90°C, 5 to 10% of the molecules change from

tetrahedral environments to two hydrogen– bonded conﬁgurations. Our ﬁndings

are consistent with neutron and x-ray diffraction data, and combining the results sets

a strong limit for possible local structure distributions in liquid water. Serious discrep-

ancies with structures based on current molecular dynamics simulations are observed.

Experimental studies of the hydrogen-bonded

network structure in water have mainly relied

on neutron and x-ray diffraction and infrared

(IR) spectroscopies (1). Diffraction data from

noncrystalline materials provide radial distribu-

tion functions (RDFs) (2–4) that do not provide

angular correlations needed to uniquely assign

local geometries in water (5–7). A more de-

tailed atomistic picture has been derived theo-

retically by molecular dynamics (MD) simula-

tions (4, 8) that are consistent with diffraction

data. Structural information from IR spectros-

copies generally relies on the correlation be-

tween the O-H stretching frequency and

hydrogen-bond (H-bond) length, which has been

shown to be ambiguous for liquid water (9).

Here, we report an independent experimen-

tal investigation of local bonding configurations

in the first coordination shell of liquid water by

using the near-edge fine structure in x-ray ab-

sorption spectroscopy (XAS), also denoted

XANES and NEXAFS, where a core electron is

excited into empty electronic states. The char-

acter of these states and, hence, the near-edge

fine structure in XAS depends on the chemical

environment, bond lengths, and bond angles

(10). We also obtained the same spectral infor-

mation by using nonresonant x-ray Raman scat-

tering (XRS) involving core excitations (11).

XAS and XRS at the oxygen K-edge (12)

are sensitive to distortions of H-bonds on the

H-sides (donor H-bonds) of the molecules in

the condensed phases of water (11, 13, 14 ).

Because the time scale for excitation is much

faster than the molecular (vibrational) motions

in the liquid, these spectroscopies probe the

electronic structure of a distribution of instan-

taneous configurations and thus allow decom-

position in terms of specific H-bond situations

(12). We analyzed the near-edge structures in

the liquid water XA spectrum (the terminology

“XA spectrum” is used for both XAS and XRS)

with the aid of experimental model systems and

calculated spectra. The XA spectra for water

molecules in different H-bonding configura-

tions are depicted in Fig. 1, where ice Ih bulk

and surface spectra are compared with spectra

of bulk liquid water at two temperatures. Bulk

ice Ih is tetrahedrally coordinated, but the exact

H-bonding environment at the ice Ih surface

still raises questions (15, 16 ). However, there is

consensus that a large fraction (50% or more) of

the molecules in the first half bilayer of the ice Ih

surface has one free O-H group, whereas the other

is H-bonded to the second half bilayer. The liquid

water XA spectrum closely resembles that for the

ice surface, but it is very different from that of

bulk ice. We interpret this finding, and our anal-

ysis demonstrates that the molecules in the liquid

are not predominantly four-coordinated.

The spectra in Fig. 1 can be divided into

three main regions: the pre-edge (around 535

eV), the main edge (537 to 538 eV), and the

post-edge (540 to 541 eV). The bulk ice spec-

trum (Fig. 1, curve a) is dominated by intensity

in the post-edge region and shows a weak main-

edge structure. Both the surface ice (Fig. 1,

curve b) and liquid water (Fig. 1, curve d)

spectra have a peak in the pre-edge region, a

dominant main edge, and less intensity com-

pared with bulk ice in the post-edge region.

Termination of the ice surface with NH

(Fig. 1,

curve c) entails a coordination of the free O-H

groups and causes the pre-edge peak to vanish

and the intensity to shift to the post-edge region.

We assign intensities in the pre- and main-edge

regions to water molecules with one uncoordi-

nated O-H group, whereas the intensity in the

post-edge region is related to fully coordinated

molecules. Remarkably, most molecules in

bulk liquid water at room temperature exhibit a

local coordination comparable to that at the ice

surface, with one strong and one non-, or only

weakly, H-bonded O-H group. The contribu-

tion to the spectrum from molecules with four-

fold coordination similar to bulk ice is very

small. Performing the measurements with D

or H

O led to identical spectra within the ex-

perimental resolution, and thus tunneling con-

tributions are not decisive.

Comparison of the XRS spectra of room-

temperature (25°C) and hot water (90°C) (Fig.

1, curve e) shows that heating increases inten-

sities in the pre- and main-edge regions while

decreasing that in the post-edge, but the chang-

es are small compared with the changes ob-

served between ice and the liquid. Figure 1,

curve f, shows the difference spectra of 25°C

water minus ice (17 ) (solid curve) and 90°C

minus 25°C water (circles with error bars). The

latter has been multiplied by a factor of 10 to

Stanford Synchrotron Radiation Laboratory, Post

Ofﬁce Box 20450, Stanford, CA 94309, USA.

BESSY, Albert-Einstein-Strasse 15, D-12489 Berlin,

Germany.

FYSIKUM, Stockholm University, Al-

baNova, S-10691 Stockholm, Sweden.

Department

of Physical Chemistry, Stockholm University,

S-10691 Stockholm, Sweden.

Department of

Chemistry, Linko¨ping University, S-58183 Linko¨p-

ing, Sweden.

Department of Inorganic Chemistry

and Catalysis, Debye Institute, Utrecht University,

Sorbonnelaan 16, 3584 CA Utrecht, Netherlands.

*To whom correspondence should be addressed. E-

mail: nilsson@slac.stanford.edu

R EPORTS

www.sciencemag.org SCIENCE VOL 304 14 MAY 2004 995

match the intensities. The similarity of the

curves, in particular the same isosbestic point at

538.8 eV, supports a two-component structure

in terms of O-H coordination: Configurations

with one uncoordinated or weakly H-bonded

O-H group replace tetrahedral ones in the ice-

liquid phase transition, and heating liquid water

causes similar types of changes, but of about

one-tenth the magnitude. The contributions of

the two different species at room temperature are

80% and 20% for species with only one strong

H-bond and tetrahedrally coordinated, respective-

ly; for 90°C, we find 85% and 15% with uncer-

tainties of ⫾15 to ⫾20% in both cases (12).

We now compare this phenomenological

finding with a detailed theoretical analysis

based on density functional theory (DFT). The

method used provides highly accurate results

for both intensities and energy positions in x-

ray spectra of H-bonded systems (12, 18). Here,

the calculations provide model spectra for var-

ious distributions of bond lengths and angles as

expected to occur in the liquid. To allow for

calculations of a large number of local config-

urations, we use a small model cluster of 11

molecules, where a central molecule is sur-

rounded by the first coordination shell (4 mol-

ecules) plus part of the second shell (6 mole-

cules saturating the dangling O-H groups of the

first shell). Spectra were calculated for the cen-

tral molecule with systematically varied nearest

neighbor H-bond environments.

First, the model cluster was tested by calcu-

lating the three known cases of bulk ice, ice

surface, and NH

terminated ice surface (Fig.

2). The calculated spectra based on clusters

with 11 water molecules reproduce the general

shapes and trends seen in the experimental

spectra in Fig. 1: shift of intensity from post to

pre and main edges when going from a fully

coordinated molecule to one with a free O-H

group (compare Fig. 2, curves a and b) and shift

of intensity back to the post-edge region when

the free O-H group is coordinated upon NH

adsorption (compare Fig. 2, curves b and c).

These findings did not depend substantially

on the cluster size: Calculated spectra for iden-

tical nearest neighbor H-bond environments but

now with a total of 44 (Fig. 2, curve a, dashed

line) and 27 (Fig. 2, curves b and c, dashed

lines) water molecules were similar to the spec-

tra for small clusters. Furthermore, thermal

fluctuations and/or vibrational motions did not

affect the calculations considerably; for exam-

ple, superposition of spectra with different in-

tramolecular bonding distances was similar to

that for the mean bonding configuration. The

same holds for intermolecular vibrations (12).

To distinguish possible local configurations

in the liquid, spectra were calculated for the

central molecule in the 11-molecule cluster

Intensity (arb. units)

545540535

Energy (eV)

Fig. 2. Calculated spectra for the model systems

bulk ice Ih. (a) Fully coordinated molecules. (b) Ice

surface (molecules in the topmost surface layer

with one free O-H). (c)NH

-terminated ice sur-

face (molecules with an NH

-saturated O-H).

Solid lines, small clusters; dashed lines, large clus-

ters. Cluster sizes are 11 molecules for (a) to (c),

solid lines; 44 molecules for (a), dashed line; and

27 molecules for (b) and (c), dashed lines. For the

bulk ice cases, the calculated molecules are fully

coordinated to their four nearest neighbors and

surrounded by: (a, solid line) part of the second

coordination shell (6 molecules) and (a, dashed

line) the full second shell (12 molecules) and 27

further molecules. For the ice-surface cases, an

appropriate number of molecules from the bulk

ice clusters were removed [3 and 17 molecules

for (b, solid line) and (b, dashed line), respectively]

to generate conﬁgurations where the calculated

molecule has one free O-H as found on the ice

surface (12, 15, 16). To model the effect of NH

termination, the same molecule was calculated

after saturating the dangling O-H by an NH

molecule (lone pair electrons of the N oriented

toward the free O-H).

Intensity (arb. units)

545540535

Energy (eV)

1.0

0.5

-0.5

-1.0

x10

Fig. 1. (a) Bulk ice Ih (XAS secondary electron

yield). (b) Ice Ih surface (topmost surface

layer, XAS Auger electron yield). (c)NH

terminated ﬁrst half bilayer of the ice Ih

surface (XAS Auger electron yield). (d) Liquid

water at ambient conditions [XAS ﬂuores-

cence yield, taken from (13) and additionally

corrected for saturation effects]. (e) Bulk liq-

uid water at 25°C (solid line) and 90°C

(dashed line) (XRS, spectra normalized to the

same area). (f) Difference spectra: 25°C wa-

ter minus bulk ice (solid curve) and 90°C

water minus 25°C water (circles with error

bars). The latter difference spectrum has

been multiplied by a factor of 10. Note that

XRS and XAS essentially give the same infor-

mation, as indicated by the similarity of the

room-temperature water spectra (e) and (d).

Their slight differences are due to the poorer

energy resolution of XRS. The temperature

effects can best be studied with XRS because

of the larger probing depth and the resulting

easier sample preparation than for XAS. For

details on sample preparation and probing

depths, see (11–13, 15).

Table 1. Relative amounts of local conﬁgurations—

double-donor (DD), single-donor (SD), and non-

donor (ND) conﬁgurations—in liquid water ac-

cording to different MD simulations at 25°C

and 90°C. The values in the ﬁrst column for

each temperature follow from the experimental

observations (EXP) in Fig. 1 (12) and from ﬁtted

spectra (FIT) such as the ones shown in Fig. 5.

The errors are given by the uncertainties relat-

ed to the experimental observations (Fig. 1) and

the calculations (12).

Method

Type EXP ⫹ FIT SPC MCYL CPMD

25°C

DD 15

⫺15

⫹25

70 50 79

SD 80 ⫾ 20 27 41 20

ND 5 ⫾ 5391

90°C

DD 10

⫺10

⫹25

56 39 63

SD 85

⫺20

⫹15

37 47 34

ND 5 ⫾ 5 7 14 3

R EPORTS

14 MAY 2004 VOL 304 SCIENCE www.sciencemag.org996

with systematically varied nearest neighbor H-

bond distances and angles (Fig. 3A, curves a to

h) (12). The spectrum for a configuration with

tetrahedral coordination [linear H-bonds: all

␪⫽0° and 2.75 Å for all O-O distances (1, 2),

Fig. 3A, curve a] shows a maximum close to

540 eV, similar to Fig. 2, curve a. Donor H-

bond distortions were introduced by varying the

O-O distance r and the angle ␪ of the nearest

neighbor (and the two attached molecules; see

inset of Fig. 3A) on only one H-side (Fig. 3A,

curves b to e) and on both H-sides (Fig. 3A,

curves f to h). Intermediate elongations smear out

the maximum close to 540 eV (Fig. 3A, curve b).

Upon further distortion of one donor H-bond, a

pre-edge peak emerges at 535 eV and a main-

edge structure develops at 537 eV, while post-

edge intensities decrease (Fig. 3A, curves c to e).

Additional distortion of the second donor H-bond

(Fig. 3A, curves f and g) further reduces intensi-

ties above 540 eV without considerable change in

the pre- and main-edge regions. Note that varying

r at constant ␪ and varying ␪ at constant r can

yield similar spectra (e.g., Fig. 3A, curves c and e).

Spectra c to e and g to h in Fig. 3A show

that donor H-bond distortions of a certain de-

gree manifest as a distinct pre-edge peak and an

intense main edge in the oxygen K-edge XA

spectrum of water. The analogy of these pre-

and main-edge features to the ones observed for

the ice-surface molecules with one broken do-

nor H-bond (Figs. 1 and 2) suggests denoting

the corresponding distorted H-bonds as “bro-

ken.” More specifically, based on spectral cal-

culations of a large number of different config-

urations in the liquid, we define the cutoff for

H-bonding in terms of O-O distances as the

boundary between zones A and B (black line in

Fig. 3B) (12). For linear H-bonds, we find the

same cutoff as conventionally used in the anal-

ysis of MD simulations (19). For bent H-bonds,

we use a smaller cutoff because of the occur-

rence of pre- and main-edge intensities for large

angular distortions at standard O-O distances

(Fig. 3A, curve e). This difference is not im-

portant, as will be addressed in detail below.

Zones A and B in Fig. 3B can be used to

classify local water configurations: The environ-

ment in bulk ice, e.g., is an AA configuration, with

both nearest neighbor H-bond acceptor molecules

within zone A, whereas that at the ice surface is an

AB configuration. We define three groups of con-

figurations in the liquid: Double-donor (DD) con-

figurations with two intact donor H-bonds are

characterized by XA spectra without pre-edge

peak (Fig. 3A, curves a, b, and f); single-donor

(SD) configurations with one intact and one bro-

ken donor H-bond exhibit a distinct pre-edge peak

and an intense main edge (Fig. 3A, curves c, d, e,

and g); and the nondonor (ND) configuration

with two broken donor H-bonds gives two dis-

tinct peaks (Fig. 3A, curve h) similar to the

spectra of gas-phase water and the species at the

surface of liquid water (20).

The pre-edge intensities for the SD configura-

tions are due to excitations to unoccupied anti-

bonding orbitals that are localized along the inter-

nal O-H bond (Fig. 4). The localization is caused

by breaking one donor H-bond while keeping the

other intact, which entails s-p-rehybridization in

the orbitals close to 535 eV (14). The resulting

increase of p-character translates to an intensity

increase in the pre-edge region of the XA spectra

through the dipole selection rule (12). The orbital

that contributes to the pre-edge intensity closely

matches zone A, as shown in Fig. 4. This com-

parison gives our H-bond criterion its physical

basis and supports our interpretation; the presence

of another molecule inside zone A changes this

orbital and thereby the corresponding spectral fea-

ture, which also explains the sensitivity of XAS

(XRS) to H-bond distortions on the H-side.

The occurrence of the pre-edge feature, fur-

thermore, can be directly correlated with the

energetics in the H-bond interaction. For the

model clusters considered here, we find that

distortions leading to enhanced intensity in the

pre-edge correspond to one donor H-bond los-

ing about 40 to 70% of the individual bond

energy of bulk ice (12). The SD species that

dominate the liquid thus have one strong and

one weak or broken donating H-bond. Consis-

tent with commonly used definitions of

H-bonds in liquid water (and for simplicity), we

use the term “broken” for the distorted donor

H-bond of an SD configuration.

In Fig. 3A, only the calculated spectra of SD

species in curves c to e and g exhibit all features

observed in the experimental liquid water spec-

Intensity (arb. units)

545540535

Energy (eV)

2.75, 0°

3.50, 0°

2.75, 0°

3.10, 0

2.75, 0°

3.50, 35

2.75, 0°

2.75, 50

2.75, 0°

2.75, 20

3.00, 0°

3.50, 0°

3.00, 0°

3.90, 0°

r , θ

2 2

3.8

3.6

3.4

3.2

3.0

2.8

2.6

r (Å)

-50 -25

25 50

(deg)

Fig. 3. (A) Calculated spectra

for a cluster of 11 water mol-

ecules with O-O distances r

(in Å) and angles ␪

(in de-

grees) of the nearest neigh-

bors i ⫽ 1, 2 on the H-sides

(nearest neighbor H-bond ac-

ceptor molecules) systemati-

cally varied as given for each

spectrum. Starting from (a) a

tetrahedral structure with all

four nearest neighbor O-O

distances at 2.75 Å and linear H-bonds, for (b, c) only r

, for

(d) r

and ␪

and for (e) only ␪

are varied to introduce

H-bond distortions on one H-side. For (f) ␪

and r

and for

(g, h) r

and r

are varied to introduce H-bond distortions

on both H-sides. Compared with the cluster calculation for

ice Ih in Fig. 2, curve a, solid line, relative orientations of the

molecules have been changed arbitrarily to better account

for the less ordered local conﬁgurations in the liquid. In

addition, the calculated oscillator strengths are convoluted

with a slightly larger width to account for the larger

vibrational motion of the molecules in the liquid. This

explains the differences between (a) and Fig. 2, curve a,

solid line. (B) The systematic spectral changes allow for a

deﬁnition of zones A and B to denote the positions of the

two nearest neighbor H-bond acceptor molecules. Spectra

show similar features for positions within one zone [com-

pare, e.g., (a) and (b) or (c) to (e) and as veriﬁed by many

further calculations], and the cases shown here can be

considered to be representative. The boundary between the

zones (black line) is used as a cutoff for H-bonding (12)as

based on the occurrence of strong pre- and main-edge

features (similar to the experimental ice surface spectrum)

when the nearest neighbor is outside zone A beyond the

black line (e.g., c to e).

Fig. 4. Connecting electronic and geometric

structures. The contour plot of a typical unoc-

cupied orbital giving rise to the pre-edge inten-

sity in the XA spectrum of a water molecule

with a broken donor H-bond (solid line contour

plot) is compared with our cutoff criterion for

H-bonding (black line in Fig. 3B) (12). Broken

donor H-bonds are characterized by nearest

neighbor O’s outside the shaded area (12). The

orbital contour has been calculated for the

central molecule in a cluster with 11 molecules

(the corresponding XA spectrum is shown in

Fig. 2, curve b, solid line) where the H-bond on

one H-side is broken (SD conﬁguration). Only 4

of the 11 molecules are displayed.

R EPORTS

www.sciencemag.org SCIENCE VOL 304 14 MAY 2004 997

trum: a pre-edge peak close to 535 eV and a

strong main edge close to 537 eV. This result

suggests, consistent with our phenomenological

findings, that asymmetric SD species with one

strong and one broken donor H-bond dominate

the liquid water structure. In a more detailed

analysis, our experimental data were fitted with

sums of calculated spectra. The results are sum-

marized in Table 1. Different combinations of

computed spectra corresponding to different

choices of bond lengths and angles within each

class give similar spectra. All yield a dominant

contribution of SD species both at 25°C and

90°C. We note the close agreement with the

previously described analysis based on model

experimental spectra (12). For 90°C we find the

best fits for a decreased (increased) amount of DD

(SD) species as compared with room temperature,

indicating the breaking of H-bonds with increas-

ing temperature along with the decrease of fully

coordinated molecules (21). In particular, it comes

as no surprise that bulk-ice–like configurations are

virtually absent in water close to the boiling point.

Consistent with the discussion of Fig. 1, we find

that DD configurations are replaced by SD con-

figurations both when heating the liquid and when

melting ice into liquid water. Computed spectra

indicate that XAS is not sensitive to the breaking

of H-bonds on the O-side (acceptor H-bond) (13).

However, symmetry arguments imply that, with a

dominant amount of molecules in SD configura-

tions, most molecules must in addition have a

broken acceptor H-bond. This entails that, within

the H-bond definition given here [Fig. 3 and (12)],

each molecule has on average 2.2 ⫾ 0.5 (2.1 ⫾

0.5) H-bonds at 25°C (90°C).

As mentioned above, XAS cannot distin-

guish between SD configurations with a broken

bond due to elongation (Fig. 3A, curve c) or to

strong bending (Fig. 3A, curve e). However,

these configurations exhibit different nearest

neighbor O-O and O-H distances. We compare

our results with O-O and O-H RDFs from

neutron diffraction data taken at ambient con-

ditions (2, 22). Two XA spectra generated with

different distributions of H-bond lengths and

angles, but both with the same fraction of DD,

SD, and ND species, are shown in Fig. 5A,

curves a and b. Spectra of this type have been

used to estimate the amounts of the different

species (Table 1) that are consistent with the

measured XA spectrum.

The O-O (Fig. 5B) and O-H (Fig. 5C) RDFs

are shown for the two model distributions cor-

responding to the XA spectra in Fig. 5A, curves

aandb(12). The 11-molecule model configu-

rations that can adequately describe the local

structure probed by XAS are obviously not

sufficient for a full description of RDFs beyond

the first coordination shell [⬍3.4 to 3.5 Å in the

O-O RDF (2, 4) and ⬍2.5 to 2.6 Å in the O-H

RDF (2)]. Molecules from the second shell are

not included in the RDFs of models a to c in

Fig. 5 and hence we should not expect agree-

ment for larger distances. The “a” model has a

larger portion of broken H-bonds as a result of

bending, whereas the “b” model has more elon-

gated bonds. Both models reproduce the exper-

imental XA spectrum, but model a clearly ap-

pears to provide a better agreement with the

RDFs within the first coordination shell. Model

b lacks structural contributions in the region 2.9

to 3.3 (1.9 to 2.3) Å and overemphasizes the

contributions around 3.5 and 2.5 Å. It is thus

inconsistent with water density, as illustrated by

the poor agreement with the diffraction data.

Combining XAS and diffraction results therefore

suggests that the H-bonds are predominantly bro-

ken by bending rather than by elongation.

Finally, our results can be compared with

two classical MD simulations using the flexible

SPC (23) and MCYL (24 ) pair potential energy

models and to an ab initio Car-Parrinello MD

(CPMD) simulation (25) [for details of the sim-

ulations, see (12)]. The sampling of the distri-

bution of structures is similar to that in the

XAS/XRS experiment, i.e., the instantaneous

geometries at a sequence of time steps are

analyzed. Using our criterion for H-bonding

(12), we have determined the amounts of DD,

SD, and ND configurations (Table 1). The

MCYL potential yields the largest amount of

SD, but all of the simulations deviate substan-

tially from experimental results. Note that with

our spectroscopically determined electronic

structure criterion, SPC (CPMD) at 25°C still

gives an average of 3.3 (3.6) H-bonds per mol-

ecule, similar to earlier theoretical studies re-

porting 3.5 H-bonds per molecule (19). For

example, with the numbers in Table 1 for SPC

at 25°C, we determine that (0.7 ⫻ 4) ⫹ (0.27 ⫻

2) ⫹ (0.03 ⫻ 0) ⫽ 3.34 H-bonds per molecule.

An XA spectrum generated based on the distri-

bution of species obtained from our SPC MD

simulation at 25°C is shown in Fig. 5A, curve c.

Clearly the agreement with the experimental

spectrum is very poor: no pre-edge intensity

(SD species) and too much intensity around 540

eV (fully coordinated DD species). The corre-

sponding O-O (O-H) RDFs (12) are depicted in

Fig. 5B (C), curve c, as solid lines. We note that

the SPC simulation, in comparison with exper-

imental results, exaggerates contributions at r-

values characteristic for tetrahedral/near tetra-

hedral configurations (in the maxima) in both

the O-O and O-H RDFs. However, the O-O

(O-H) RDFs from SPC and our model a are

rather similar at 2.9 to 3.3 (1.9 to 2.3) Å. The

poor agreement with the experimental XA

spectrum is hence attributed to too few species

with large H-bond angles (SD species) in the

SPC simulation. Similarly, the other two MD

simulations shown in Table 1 cannot create an

XA spectrum that fits the data. The MD simu-

lations are, however, consistent with our results

in predicting the decrease (increase) of DD

(SD) species with increasing temperature.

The large amount of SD species can also

explain femtosecond-IR studies indicating

two different O-H groups in liquid water: one

“strongly” and one “weakly” H-bonded (26 ).

545540535

Energy (eV)

XAS

XRS

2.52.01.5

r (Å)

(r)

3.53.02.5

r (Å)

(r)

ABC

Fig. 5. Calculated XA spectra and RDFs (solid lines) for three models—(a), (b), and (c)—are

compared with (A) experimental XA spectra and RDFs (B and C) at 25°C. Spectra and RDFs have

been generated simultaneously for the central molecule in clusters consisting of 11 molecules, for

a total of 14 clusters (12). A sampling of the possible DD, SD, and ND conﬁgurations for the

calculated molecule in the liquid is guaranteed by using a variety of nearest neighbor O-O distances

and H-bond angles. The RDFs were calculated for the ﬁrst coordination shell only. Experimental

RDFs [open circles in (B) and (C)] were derived from neutron diffraction (2). The calculated XA

spectra were broadened witha1eVGaussian for better comparison with experimental XRS data

[open circles in (A)] taken from Fig. 1, curve e. Models (a) and (b) use a balance of DD:SD:ND ⫽

10:85:5. Model a (b) has more (less) angularly distorted H-bonds and less (more) elongated ones.

Model (c) (solid line) represents the SPC simulation (Table 1) with a balance DD:SD:ND ⫽ 70:27:3,

as determined with our criterion for H-bonding (12). For comparison, the dashed lines in B and C,

curve c, show the O-O and O-H RDFs calculated from the full SPC simulation. There is good

agreement up to 3.2 Å (O-O) and 2.4 Å (O-H), validating our approach to characterize the ﬁrst shell

in the liquid. Only model (a) or similar balances with a dominant fraction of SD species (Table 1)

give good ﬁts when simultaneously considering our experimental XRS data and the RDF curves.

R EPORTS

14 MAY 2004 VOL 304 SCIENCE www.sciencemag.org998

The orientational relaxation dynamics were

shown to be directly connected with the

strength of the two H-bond groups, where the

weak H-bonds relax much faster than the

strong H-bonds. According to our results, the

number of strong H-bonds in the liquid is

substantially smaller than expected, which

may seem in contradiction with the small heat

of melting compared with the heat of subli-

mation for ice. However, quantum chemical

calculations have shown that each bond in the

proposed SD configurations is stronger than

the average bond in four-fold coordination

because of anticooperativity effects (27, 28).

Thus, the large number of weakened/broken

H-bonds in the liquid leads to only a small

change in energy. A recently developed

quantum chemical model (1, 27 ) that propos-

es the predominance in the liquid phase of

two hydrogen–bonded water molecules in

ring conformations is consistent with our re-

sults. Water is a dynamic liquid where H-

bonds are continuously broken and reformed

(29). The present result that water, on the

probed subfemtosecond time scale, consists

mainly of structures with two strong

H-bonds, one donating and one accepting,

nonetheless implies that most molecules are

arranged in strongly H-bonded chains or

rings embedded in a disordered cluster net-

work connected mainly by weak H-bonds.

References and Notes

1. R. Ludwig, Angew. Chem. Int. Ed. Engl. 40, 1808 (2001).

2. A. K. Soper, Chem. Phys. 258, 121 (2000).

3. A. H. Narten, H. A. Levy, Science 165, 447 (1969).

4. T. Head-Gordon, G. Hura, Chem. Rev. 102, 2651 (2002).

5. P. G. Kusalik, I. M. Svishchev, Science 265, 1219 (1994).

6. A. K. Soper, J. Chem. Phys. 101, 6888 (1994).

7. A. A. Chialvo, P. T. Cummings, J. Phys. Chem. 100,

1309 (1996).

8. F. H. Stillinger, Science 209, 451 (1980).

9. J. Stenger, D. Madsen, P. Hamm, E. T. J. Nibbering, T.

Elsaesser, Phys. Rev. Lett. 87, 027401 (2001).

10. J. Sto¨hr, NEXAFS Spectroscopy (Springer-Verlag, Ber-

lin, 1992).

11. U. Bergmann et al., Phys. Rev. B 66, 092107 (2002).

12. Details of the materials and methods, and supporting

analysis of the experimental data, are available at

Science Online.

13. S. Myneni et al., J. Phys. Condens. Matter 14, L213

(2002).

14. M. Cavalleri, H. Ogasawara, L. G. M. Pettersson, A.

Nilsson, Chem. Phys. Lett. 364, 363 (2002).

15. D. Nordlund et al., unpublished data.

16. A. Glebov, A. P. Graham, A. Menzel, J. P. Toennies, P.

Senet, J. Chem. Phys. 112, 11011 (2000).

17. The ice spectrum (Fig. 1, curve a) has been broadened

to the same instrumental resolution as for XRS and

normalized to the same area before subtraction.

18. M. Nyberg, M. Odelius, A. Nilsson, L. G. M. Pettersson,

J. Chem. Phys. 119, 12577 (2003).

19. A. Luzar, D. Chandler, Phys. Rev. Lett. 76, 928 (1996).

20. K. R. Wilson et al., J. Phys. Condens. Matter 14, L221

(2002).

21. The sharper main edge for hot water (537 eV, Fig. 1, curve

e, dashed line) is accounted for by a different balance

inside the group of SD species, where more of the H-bond

acceptor molecules have distances closer to the AB bound-

ary, i.e., have a larger r

and/or ␪

but still remain in zone

A (more Fig. 3A, curve g, and less curve c, e.g.). Accordingly,

the intensity decrease above the isosbestic point with

temperature can be mainly assigned to the loss of DD

conﬁgurations with tetrahedral and near tetrahedral envi-

ronments by 5 to 10% and an increase of SD conﬁgura-

tions with H-bond acceptor molecules being closer to the

AB boundary.

22. The neutron O-O RDF in (2) is very similar to the

most recently derived O-O RDF from x-ray diffrac-

tion in (4).

23. K. Toukan, A. Rahman, Phys. Rev. B 31, 2643 (1985).

24. G. C. Lie, E. Clementi, Phys. Rev. A. 33, 2679 (1986).

25. P. L. Silvestrelli, M. Parrinello, J. Chem. Phys. 111,

3572 (1999).

26. S. Woutersen, U. Emmerichs, H. J. Bakker, Science

278, 658 (1997).

27. F. Weinhold, J. Chem. Phys. 109, 373 (1998).

28. L. Ojama¨e, K. Hermansson, J. Phys. Chem. 98, 4271

(1994).

29. C. J. Fecko, J. D. Eaves, J. J. Loparo, A. Tokmakoff, P. L.

Geissler, Science 301, 1698 (2003).

30. Supported by the Swedish Foundation for Strategic Re-

search, Swedish Natural Science Research Council, and U.S.

National Science Foundation grant CHE-0089215. Gener-

ous grants of computer time at the Swedish National

Supercomputer Center and the Center for Parallel Com-

puting are gratefully acknowledged. Portions of this re-

search were carried out at the Stanford Synchrotron Radi-

ation Laboratory, a national user facility operated by Stan-

ford University on behalf of the U.S. Department of Energy,

Ofﬁce of Basic Energy Sciences. Use of the Advanced

Photon Source (APS) was supported by the U.S. Depart-

ment of Energy, Basic Energy Sciences, Ofﬁce of Science,

under contract No. W-31-109-ENG-38. Biophysics Collab-

orative Access Team (BioCAT) is a National Institutes of

Health–supported Research Center RR-08630. The Ad-

vanced Light Source (ALS) is supported by the Director,

Ofﬁce of Science, Ofﬁce of Basic Energy Sciences, Materi-

als Sciences Division, of the U.S. Department of Energy

under Contract No. DE-AC03-76SF00098 at Lawrence

Berkeley National Laboratory. Assistance by the APS, ALS,

and the Swedish national laboratory MAX-lab staff is

gratefully acknowledged. We thank J. B. Hastings for his

valuable comments and discussions and S. P. Cramer for

making the XRS spectrometer available.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1096205/DC1

Materials and Methods

SOM Text

Figs. S1 to S11

Tables S1 and S2

References and Notes

29 January 2004; accepted 25 March 2004

Published online 1 April 2004;

10.1126/science.1096205

Include this information when citing this paper.

Asphalt Volcanism and

Chemosynthetic Life in the

Campeche Knolls, Gulf of Mexico

I. R. MacDonald,

* G. Bohrmann,

E. Escobar,

F. Abegg,

P. Blanchon,

V. Blinova,

W. Bru¨ckmann,

M. Drews,

A. Eisenhauer,

X. Han,

K. Heeschen,

F. Meier,

C. Mortera,

T. Naehr,

B. Orcutt,

B. Bernard,

J. Brooks,

M. de Farago´

In the Campeche Knolls, in the southern Gulf of Mexico, lava-like ﬂows of solidiﬁed

asphalt cover more than 1 square kilometer of the rim of a dissected salt dome at

a depth of 3000 meters below sea level. Chemosynthetic tubeworms and bivalves

colonize the sea ﬂoor near the asphalt, which chilled and contracted after discharge.

The site also includes oil seeps, gas hydrate deposits, locally anoxic sediments, and

slabs of authigenic carbonate. Asphalt volcanism creates a habitat for chemosyn-

thetic life that may be widespread at great depth in the Gulf of Mexico.

Salt tectonism in the Gulf of Mexico hy-

drocarbon province controls the develop-

ment of reservoirs and faults that allow oil

and gas to escape at the sea floor (1). More

than 30 years ago, investigators studying

the Gulf’s abyssal petroleum system (2)

photographed an asphalt deposit (3) among

salt domes in the southern Gulf of Mexico.

During exploration of the Campeche

Knolls, about 200 km south of the photo-

graphed site (Fig. 1, A and C), we have now

found numerous, deeply dissected salt

domes with extensive slumps and mass

wasting at depths of 3000 m or greater.

Massive, lava-like flow fields of solidified

asphalt, evidently discharged at tempera-

tures higher than the ambient bottom water

(4°C), have been colonized by an abundant

chemosynthetic fauna.

The Campeche Knolls are salt diapirs ris-

ing from an evaporite deposit that underlies

the entire slope region (4) and hosts the

Campeche offshore oil fields (5 ). Numerous

reservoir and seal facies have also been at-

Physical and Life Sciences Department, TexasA&M

University–Corpus Christi, 6300 Ocean Drive, Corpus

Christi, TX 78412, USA.

Fachbereich 5 Geowissen-

schaften, University Bremen, D-28334 Bremen, Germa-

ny.

Universidad Nacional Autonoma de Mexico, Insti-

tuto de Ciencias del Mar y Limnologı`a, Apartado Postal

70-305, Mexico 045510, D.F. Mexico.

Instituto de Cien-

cias del Mar y Limnologı`a, Apartado Postal 1152, Can-

cu´n, D.F. Mexico.

IFM-GEOMAR, Leibniz-Institut fu¨r

Meereswissenschaften, D-24148 Kiel, Germany.

Second

Institute of Oceanography, State Oceanic Administra-

tion, Zhejiang 310012, China.

Universidad Nacional

Autonoma de Mexico, Instituto de Geoﬁsica, Mexico,

04510, D.F. Mexico.

University of Georgia, 220 Marine

Sciences Building, Athens, GA 30602, USA.

TDI-Brooks

International, Inc. 1902 Pinion, College Station, TX

77845, USA.

Aurensis, SA, San Francisco de Sales 38,

28003, Madrid, Spain.

*To whom correspondence should be addressed. E-

mail: imacdonald@falcon.tamucc.edu

R EPORTS

The Structure of the First Coordination Shell in Liquid Water

Figures

Citations

Water as an Active Constituent in Cell Biology

Structure and nanostructure in ionic liquids.

Attosecond Physics

Effect of ions on the structure of water: structure making and breaking.

Hydrophilic interaction chromatography.

References

Water: From Clusters to the Bulk

The radial distribution functions of water and ice from 220 to 673 K and at pressures up to 400 MPa

Effect of environment on hydrogen bond dynamics in liquid water.

Ultrafast hydrogen-bond dynamics in the infrared spectroscopy of water.

Molecular-dynamics study of atomic motions in water

Related Papers (5)

The inhomogeneous structure of water at ambient conditions

Comparison of simple potential functions for simulating liquid water

The radial distribution functions of water and ice from 220 to 673 K and at pressures up to 400 MPa

The missing term in effective pair potentials

Generalized Gradient Approximation Made Simple

Frequently Asked Questions (1)

Q1. What are the contributions mentioned in the paper "The structure of the first coordination shell in liquid water" ?