Sunteți pe pagina 1din 10

2352 Owicki, Lentz, Hagler, and Scheraga

Structure of Liquid Water. 111. Thermodynamic Properties of Liquid Deuterium Oxide’

John C. Owlck1,28Barry R. LentzlZbArnold T. Hagler, and Harold A. Scheraga*2C


Department of Chemistry, Cornell University, Ithaca, New York 74853 and Weizmann hstitute of Science, Rehovoth, lsrael

(Received May 27, 1975)

A statistical thermodynamic treatment of a cluster model for liquid water is adapted to DzO. When com-
pared with the results of the theory for HzO, the principal thermodynamic quantities for DzO are fit with
slightly less accuracy, but the signs of the isotopic differences in these quantities are predicted correctly.
The intermolecular vibrational spectral density of DzO is calculated (as a function of temperature) from
the normal mode vibrational frequencies of the clusters and the distribution of cluster sizes. The theory in-
dicates the occurrence of only very few clusters containing more than six molecules, the same conclusion
reached for HzO. Also, the fraction of loosely bonded noncyclic clusters of DzO increases with increasing
temperature, as was found for HzO. The results suggest that the energy required to break hydrogen bonds
in liquid DzO may be smaller than that for liquid Hz0. The similarities between the theoretical equilibri-
um cluster distributions and fractions of unbroken hydrogen bonds for DzO and HzO imply that these liq-
uids are structurally quite similar. Most of the thermodynamic differences between liquid DzO and HzO
are attributable to isotopic effects on the intra- and intermolecular vibrational frequencies of the water
clusters rather than to differences in the structures of the two liquids.

I. Introduction As an example, consider the isotopic difference in the


In the first paper in this series, Hagler, Scheraga, and heat capacities: AC, = C,(DzO) - C,(HzO) = 2.3 cal/mol
N6methy3 developed a statistical thermodynamic theory of deg a t 4°C.14,15 Since part of the heat capacity of the liquid
liquid water based on a “cluster” model. The model was re- is associated with structural relaxation (hydrogen-bond
fined and improved by Lentz, Hagler, and Scheraga4 (here- rupture),16 this value of AC, may indicate that there are
after referred to as LHS). This paper reports an adaptation differences between the structures of liquid DzO and HzO.
of that theory to DzO. The two main purposes of this study However, the structures of DzO and HzO ice Ih are essen-
are to explore further the validity of the LHS theory and to tially identica1,l’ and yet AC, = 1.6 cal/mol deg for the ices
investigate some of the isotope effects in the liquid water a t O°C.18 The substitution of D for H will, by itself, lower
system. the intermolecular vibrational frequencies, and lead to a
The primary criterion in assessing the validity of the positive contribution to AC, in both phases. Hence, struc-
theory is its ability to reproduce a variety of experimental tural effects are indicated not merely by the sign of AC,
results. A secondary criterion is that the values of the two but by its magnitude. The key question is whether the ad-
adjustable parameters of the theory for DzO should repre- ditional 0.7 cal/mol deg in AC, in the liquids, as compared
sent physically reasonable interactions among clusters and to the solids, represents isotopic differences in structural
not differ too drastically from the values obtained for H20. relaxation in the liquids, or nonrelaxational isotopic effects
Liquid DzO is generally considered to be more highly caused, e.g., by differences in the vibrational spectral den-
structured than liquid HzO at the same temperature, where sity of the two phases. This is a difficult question to resolve
the term “structure” is usually taken to mean the extent of conclusively at the present state of knowledge of these sys-
hydrogen bonding5 in the liquid. Experimental quantities tems.
that have been cited6 to support the view that d20 is more In an argument similar to that given above, Holtzer and
structured than H20 include the heat capacity, viscosity, Emersonlg have shown that current hydrodynamic models
melting point, and temperature of maximum density, all of are inadequate to determine whether there is a structural
which are higher for DzO than for Hz0. component in the difference in the viscosities of liquid Dz0
This question also has been examined in a number of and HzO. Some of the other isotopic differences (e.g., in the
theoretical studies.6-10 For example, using their cluster melting points) that are used in structural arguments also
model for liquid water,11-13 Nemethy and Scheraga6J2 are difficult to interpret, although the high temperature of
found that the energy required to break a hydrogen bond in maximum density of DzO seems to be explained most easily
liquid DzO was 240 cal/mol of bonds greater than in HzO. in terms of structural differences between liquid D20 and
In addition, the fraction of hydrogen bonds ( X H B )was HzO.
found to be slightly greater for DzO than for HzO. Specifi- Of course, the problem arises because conclusions about
cally, AXHB= XHB(DZO)- XHB(HZO)varied from 0.029 structures are drawn from thermodynamic rather than
at 4°C to 0.007 at 6 5 ” C , the highest temperature for which spectroscopic data, A definitive comparison of the struc-
the theory was applicable. tures of D20 and HzO awaits an accurate experimental de-
However, the interpretation of the observed isotopic dif- termination of the O--O, 0.-H (0.-D), and H-H, (D-D)
ferences in the two liquids depends markedly on the model pair correlation functions; a study of this type is presently
employed for the liquid, and it is very difficult to provide in progress.20 At present, however, the best information
an accurate model of liquid water. Hence, although DzO available is the finding of Narten et a1.21that the low-angle
usually is considered to be more structured than HzO, X-ray scattering patterns of HzO and DzO a t 4°C are es-
these structural differences are not established on firm sentially identical. As noted by Frank,22this suggests that
grounds. the structural differences may be much smaller than often
The Journal of Physical Chemistry, Vol, 79, No. ?2, 1975
Thermodynamic Properties of Liquid Deuterium Oxide 2353

is assumed. Similar conclusions derive from near-infrared where K1 and K, are force constants, and R1 and R, are ef-
observationsz3on liquid HzO and DzO, where the small ob- fective hard-sphere diameters. The mean librational fre-
served differences (3%)in XHBthroughout the liquid tem- quency for each cluster is obtained with the appropriate
perature range are within the experimental error. In sum- force constant and the geometric mean of the principal mo-
mary, it appears that the question of difference in the de- ments of inertia of the cluster.
gree of “structure” between DzO and HzO remains unre- C. Vibrational Partition Function. Normal mode cluster
solved, although most authors have considered DzO to be vibrational frequencies obtainedz6 with the effective pair
the more structured liquid. potential of Ben-Naim and Stillir~ger?~ (BNS) are inserted
into an Einstein partition function. Following LHS, the li-
11. Model and Form of t h e Partition Function brational frequencies calculated from the BNS potential
Except for isotopic differences, the model and form of are first scaled by a factor of 0.67; a similar procedure was
the partition function used in this study are identical with followed by Weres and Rice28 for H20. The final normal
those of LHS.4 The reader should refer to that paper for a mode vibrational frequencies of DzO are compiled in Table
fuller analysis than the summary presented here. I (see paragraph at end of text regarding miniprint materi-
Liquid DzO is considered to be a mixture of independent al).
clusters of hydrogen bonded molecules; thus, the internal D. Internal Configuration Partition Function. This is
and external degrees of freedom of the clusters can be the partition function for the various ways of arranging j
treated separately. The external variables are themselves hydrogen-bonded water molecules to obtain a j-sized clus-
assumed to be separable into translational and rotational ter. Such arrangements may be accomplished by rearrang-
contributions. The internal degrees of freedom are likewise ing either the oxygen or deuterium atoms of a ~ l u s t e rAs .~
assumed to be separable into independent vibrational and in LHS, one arrangement of oxygen atoms is chosen for
internal configuration (cluster hydrogen bonding geome- each type of cluster except the pentamer, for which two are
try) parts. The maximum term in the canonical partition used. The different possible arrangements of deuterium
function may then be written as atoms (analogous to the hydrogen randomness which pro-
duces the zero point entropy of ice) are evaluated princi-
pally by molecular model building. The choices of the rele-
vant atomic configurations and the energy assigned to each
where the product is taken over different sizes of clusters j ,
N j is the number of clusters of size j present in the maxi- are based on the results of an extended basis set ab initio
quantum mechanical study of water molecule interactions
mum-term equilibrium distribution of clusters, and No is
by Lentz and S ~ h e r a g a With
. ~ ~ one exception, these are
Avogadro’s number.
identical with those used by LHS.
No clusters containing more than eight molecules are
The exception is the energy assigned to a three- or four-
considered, because of the results of LHS for HzO; as in the
coordinated molecule, and it is brought about by a differ-
case of the LHS study, the small contribution of clusters of
ence between the lattice energies of H20 and DzO. The ex-
size seven and eight to the equilibrium cluster distribution
perimental data of Whalley30 indicate that the lattice ener-
tends to support this procedure. Following LHS, only one
gy (Elat) of HzO is -13,400 cal/mol while that for D20 is
basic cluster geometry is included for each cluster size ex-
cept the pentamer, for which two are used. The configura-
-13,700 cal/mol. Lentz and ScheragaZ9obtained a value of
-12,900 cal/mol for the lattice energy of HzO. Given the
tions are as follows: monomer, dimer, “sequential” linear
near identity of the molecular dimensions,”’ ice lattice con-
trimer (central molecule both deuterium donor and accep-
tor), cyclic tetramer, cyclic pentamer, “star” pentamer (te- stants,17 and dipole moments32of D20 and H20, the same
tracoordinated central molecule), cyclic hexamer, cyclic value of Elatshould hold for D20. LHS attributed the dis-
heptamer, and “bridged” octamer [the analog of tricy- agreement between ab initio and experimental results to a
clo(2.2.2)octanel. neglect of third- and higher-body interactions in their com-
A. Translational Partition Function. In their transla- putation based on the ab initio calculations. To make the
tional motion, the clusters are assumed to behave as a mix- theoretical internal configurational energy converge to the
ture of hard spheres, the thermodynamics of which has correct lattice energy in the limit of an infinitely large clus-
been derived by Lebowitz and RowlinsonZ4from Lebow- ter, they added a correction energy of -500 cal/mol for
itz’sZ5solution of the generalized Percus-Yevick equation. each three- or four-coordinated molecule in a cluster. The
Effective hard-sphere diameters for the clusters were ob- same procedure is followed for Dz0 in this study, except
tained by LHS.4 In the translational partition function, at- that the correction is -800 cal/mol for each three- or four-
tractive inter-cluster interactions are treated in an average coordinated molecule. The only clusters which contain
way by the inclusion of an adjustable parameter ( a ) for a these species are the star pentamer and octamer. In the in-
constant background potential energy, which is taken to be ternal configurational partition function, therefore, these
inversely proportional to the experimental molar volume, clusters are slightly more stable relative to the others in
V (i.e., 0: a / V ) . DzO than in HzO. For DzO calculations using Elat =
B. Rotational Partition Function. Qr,,t, is taken as an -13,400 cal/mol (see section IV), the correction energy was
Einstein partition function based on a mean librational fre- restored to -500 cal/mol.
quency for each cluster. As above, an adjustable parameter The reference state for the partition function is ice at 0
( d ) is introduced to treat inter-cluster interactions. d / V = K in its ground intermolecular and intramolecular vibra-
K1 is the rotational force constant for the monomer. The tional states. Since the ab initio quantum mechanical ener-
force constants of larger clusters are constructed so that, gies are taken with respect to separated molecules at rest, it
with the clusters represented as hard spheres, points on is necessary to adjust the zero of energy to the desired ice
their surfaces sweep out equal average distances during li- reference state. The procedure is discussed more fully by
brations. Mathematically, the relationship is LHS. The requisite thermodynamic quantities for DzO at 0
K are the sublimation enthalpy (ivlsub), 11,900 cal/mol;:’:’
KI = K1(Rj/R1lZ (2) the intermolecular vibrational zero point energy (ZPEinte,),
The Journal of Physical chemistry, voi. 79. NO 22, 1975
2354 Owicki, Lentz, Hagler, and Scheraga

3 521 439 338 178 161 83


294 213 162 56 33
142
4 539 478 458 170 169 164
458 318 318 164 114 46
315 213 165
148 148 139
5 595 551 517 206 195 170
506 472 386 167 147 102
346 334 284 95 33 19
265 187 176
138 149 145
5* 705 512 479 199 191 189
408 355 313 123 113 93
294 253 208 70 59 47
196 187 46 44 36

6d 568 515 515 215 208 208


514 514 490 154 154 149
341 343 323 121 68 68
305 305 236 47 28 20
189 169 169
139 139 128

660 603 574 210 204 187


543 501 456 179
154 171
139 165
120
440 393 368
347 340 333 425 93 50
19
300 287 216
165
203 151
193 146
183

694 630 554 212 206 201


550 541 519 197 155
191 169
146
495 476 415 162
128
394 348 336 82
335 318 301 8
36 46 27
43
272 254 217
155
209 200
152 147
183

(a) msle have been sealed as described in the t e x t .

(b) For a discuseion of the nstare of these mode., sse Table 1


IT1 of LHS.
(0) Normal mode frequenoies of rhe Lsiiesoordinsred s u r
p'"t'er.
(d) Because of eoinp~Ldliionn1 difficulLiCs, the frequencies for
Li
fheas flustern are obtained by scaling the corresponding
H20 freqvanciell In Table 111 of U S . Based on a C~llperi-

son of the IsoIope effect i n cluscsrs for which bath DzO


and HzO ENS normal mode yldyaee wsre avntlabls. the

2800 cal/m01;~~
and the change in intramolecular vibration-
al ZPE upon sublimation (AZPEi,t,,), 1000 cal/m01.~0
X2 = cT I[(Fcalcd,T - Fobsd,T)100/Fobsd,TI2+
[(Ucalcd,T- Uob~d,T)100/Uobsd,T]~)
(3)
111. Experimentally Observed Thermodynamic
Quantities and Computational Procedures x2 measures the sum of the squared percent deviations of
calculated from observed values of the free energy F and
Experimentally observed values of the enthalpy and en- internal energy U over a set of 11 temperatures from 4 to
tropy of DzO throughout the liquid range were obtained 100OC. At every step of the minimization algorithm, for
from Baker's The data compilation of N6methy
each temperature the equilibrium distribution of clusters
and Scheraga6 was used to change Baker's reference state ( N j }was determined by requiring that the free energy be a
to D2O ice a t 0 K. minimum, in accordance with the maximum-term method
The experimental Helmholtz free energy (to be com- used to derive the partition function. As LHS noted, this
pared with the calculated Helmholtz free energy) is taken procedure does not introduce any new adjustable parame-
to be equal to the observed Gibbs free energy. The experi- ters.
mental internal energy (to be compared with the calculated
internal energy) is likewise taken to be equal to the ob-
served enthalpy. IV. Preliminary Results
The constant-pressure heat capacity for D2O throughout As described in section IVA, our calculations have sug-
the liquid range was taken from Eucken and Eigen's gested that it is difficult to obtain physically reasonable re-
study.14 The temperature dependent molar volume was sults by a straightforward best-fit adjustment of a and d
calculated from Kell's equation.35 using the values of Elat and the ZPE's given in section IID.
Following LHS, the two adjustable parameters of the In section IVB, we report our computations that lead to a
theory, a and d , were varied to fit experiment by minimiz- final choice for the values of a, d, Elat, and the ZPE's. In
ing the x2 function: section V, we discuss our principal results which were ob-
The Journal of Physical Chemistry, Vol. 79, No.22, 1975
Thermodynamic Properties of Liquid Deuterium Oxide 2355

TABLE 11: Summary of Theoretical Calculations


Calcu- % rms
lation error
no. Description aa In d b vli< Elatd AHsu< ZPEint,,d AZPEint,,dx; e in F , U
~~

1 H,O; LHS results 5078 -23.90 580 -13,400 11,300 3500 1400 1.47 2.9
2 DzO; aand d adjusted; exptl
Elatand ZPE's 5292 -23.24 593 -13,700 11,900 2800 1000 2.21 3.5
3 D,O; H 2 0 values of a and d ;
exptl Elat and ZPE's 5078 -23.90 426 -13,700 11,900 2800 1000 7.29 6.4
4 DzO; H 2 0 values of a, d, E,,,;
best fit ZPE's. This is the
principal vesult. 5078 -23.90 426 -13,400 11,900 2785 1285 2.27 3.6
5 DzO; H,O values of a, d , Elat;
ZPElnt,, altered 5078 -23.90 426 -13,400 11,900 2500 1000 6.98 6.3
a Units: cal/mol of interaction. Units of d: dyn cm4/radian2molecule. Mean monomer librational frequency in cm-l, calculated by
relating vl,b to d, as described in section IIB. Units: cal/mol. e Reduced ( x ~ ) . ~ ' ' J ~
tained with these final values. by arguments based on anharmonicity in the real cluster
A. Calculations with Elat = -13,700 cal/mol. The mini- interaction potentials, but such arguments are insufficient
mum value of x2 for D20 is obtained with a = 95,300 cal to account for the very large increase in d obtained in cal-
cm3/mol of bond and d = 1.16 X lo6 cal cm3/radian2 mol. culation 2. In fact, the large value of d predicted by the
Physically, this corresponds to an interaction between theory results in the mean D2O monomer librational fre-
water molecules in neighboring clusters of 5290 cal/mol of quency being greater than that for HzO (593 cm-' vs. 580
such interaction (compared to 6850 cal/mol in DzO ice, cm-l), an impossibility. The changes in the adjustable pa-
based on the lattice energy) and a mean monomer libra- rameters (particularly d ) , thus cannot be interpreted sim-
tional frequency of 593 cm-l (compared to a mean libra- ply in terms of their stated roles in the partition function.
tional frequency of 517 cm-l in D20 ice3". Statistically, Rather, they probably reflect compensations for other con-
the fit is characterized by a reduced x 2 parameter value of tributions to the partition function which do not account
2.2137,38 (Table 11,calculation 2). correctly for the effects of isotopic substitution.
The theory predicts F quite accurately (within -4%). One such contribution in which there are large uncer-
The internal energy and entropy are fit somewhat less well tainties involves the lattice energies and vibrational ZPE's
(within 4%), and, since the theory underestimates the of HzO and DzO. Whalley30 has attributed the difference in
magnitude of the temperature dependence of both of these lattice energies calculated from experiment (Elat = -13,700
quantities, C, is underestimated by as much as 40%. This cal/mol for D20 and -13,400 cal/mol for HzO) to anhar-
poor fit is not unexpected, since the calculation of the heat monicities in intramolecular vibrations which cause small
capacity involves the second derivative of the free energy. differences in molecular structure sufficient to affect the
The nature of the fit is similar to that obtained for HzO by multipole interaction energies in ice. There is, however, a
LHS. €airly large uncertainty in the values of the vibrational
The values of a and d are larger by 4 and 94%, respec- ZPE's which Whalley used to compute the lattice energies.
tively, for Dz0 as compared with HzO.* If the values of a Whalley40 recently has published a new analysis of ground
and d for H20 are used in the partition function for D20, state intermolecular vibrational frequencies and sublima-
the calculated values of F, U, and S are all too high, and tion shifts in intramolecular vibrational frequencies for
the value of the reduced x2 parameter rises to 7.29,indicat- H20 only. The Elat and ZPE's calculated with the newer
ing a very poor fit (see Table 11, calculation 3). Thus, the data differ from those previously calculated by several hun-
differences in a and d between DzO and HzO are signifi- dred calories per mole, which is approximately the same
cant. magnitude as the energetic effects of the increased parame-
The adjustable parameters are intended to account for ters. A major source of the high values of a and d for D20
inter-cluster interactions which otherwise have been ne- thus may be errors in the values of Elat and the ZPE's
glected. To the extent that they do, the increases in a and d which were used in the H20 and DzO partition functions.
observed for D20 relative to the values for HzO suggest B. Calculations with Elat = -13,400 callmol. We have
that the average inter-cluster interactions are somewhat sought to obtain more physically meaningful results for
stronger and orientationally considerably more rigid in DzO by the following procedure: (1) fix a and d a t the
DzO than in H2O. However, one might expect that the values for Hz0,39and (2)assume that the lattice energy of
values of a and d for HzO and Dz0 should be approximate- HzO (-13,400 cal/mol) is also correct for DzO. These are
ly This expectation is based on the fact that the reasonable working assumptions.
cluster distributions obtained for the two liquids are quite Elat can be defined4' as
similar (Le., from calculations 1 and 2 of Table 11),imply-
ing that a cluster in Dz0 has, on the average, a set of neigh- Elat = - m s u b - ZPEinter + AZPEintra (4)
bors similar to those of a corresponding cluster in HzO. where all quantities are calculated at 0 K. There are an in-
Then, since the potential energy of interaction between finite number of ways to apportion the 300 cal/mol alge-
water molecules cannot be affected greatly by the substitu- braic increase in Elat of DzO among these three compo-
tion of a deuterium for a hydrogen atom, we are led to the nents. Since m s u b is known relatively precisely, we have
conclusion that the average intercluster interaction should chosen to regard it as fixed (11,900 cal/moP3), thus con-
be approximately the same in Dz0 as in HzO. straining -ZPEinter + AZPEintra to be equal to Elat +
Small increases in a and d for DzO can be rationalized m s u b (= -1500 cal/mol). The best least-squares fit to the

The Journal of Physical Chemistry, Vol. 79, No 22, 197.5


2356 Owicki, Lentz, Hagler, and Scheraga

-4WOO 7 20 40

Temperature ?C)
60 80 100

Figure 1. Comparison of the calculated and experimental values of 15


0
I I
20
I 1
40
I I
60
I I
80
I
IO0
the free energy, F,as a function of temperature. The circles are ex- Temperature P C )
perimental values for and the solid line represents the princi-
pal calculated results for D20. The squares and dashed line are the Figure 3. Comparison of the calculated and experimental values of
experimental and calculated free energies, respectively, for H20, the entropy, S, as a function of temperature. The circles are experi-
taken from ref 4. mental values for D20,34and the solid line represents the principal
calculated results for D20. The squares and dashed line are the ex-
perimental and calculated entropies, respectively, for H20, taken
5 0 0 0 0 4 from ref 4.

b
2500 I I I I l I ~ I I
0 20 40 60 80 100
Temperature ("C)
I I I I I I I I I I 1
0 20 40 60 80 IO0
Figure 2. Comparison of the calculated and experimental values of Temperature ("C)
the internal energy, U, as a function of temperature. The circles are
experimental values for D20,34and the solid line represents the prin- Figure 4. Relaxational heat capacity and comparison of the calculat-
cipal calculated results for D20. The squares and dashed line are the ed and experimental values of the heat capacity, C,, as a function
experimental and calculated internal energies, respectively, for H 2 0 , of temperature. The circles are experimental values for D20,14 and
taken from ref 4. the solid lines represent the principal calculated results for D20. The
squares and dashed line are the experimental and calculated heat
capacities, respectively, for H 2 0 , taken from ref 4.
experimental values of F and U (minimizing x2 with re-
spect to ZPEinterwhile keeping a and d fixed at the values
for H2O) was found for ZPEinter = 2785 cal/mol and the value for H20 of -13,400 cal/moPO) is put into
AZPEintra= 1285 cal/mol. These values are not unreason- ZPEinter.The resultant fit to the experimental values of F
able when compared with the values for H2030of 3500 and and U is poor, being only slightly better than that obtained
1400 cal/mol, respectively, although the value (1285) of in calculation 3 (also with the values of a and d for H20,
AZPEintrafor D20 is higher than one would have predicted but with experimental values of Elat and ZPE's). Thus, in-
from isotopic mass effects under the assumption that the creases in AZPEintrafrom 1000 cal/m01~~ are important in
value for HzO is valid. The calculations carried out with fitting theory to experiment, at least under our constraint
this set of parameters represent our principal results and of the equality of Elat for H2O and D2O.
are summarized in Table I1 as calculation 4.
Fixing Elat, ZPEinter,and AZPEintraat their new values,
V. Principal Results
and allowing a and d to adjust to fit the experimental data, The theoretical (calculation 4, Table 11) and experimen-
produces a much smaller change in a and d with respect to tal values of the free energy ( F ) ,internal energy (U),entro-
the values for H20 than that obtained in calculation 2; a py (S),and constant-pressure heat capacity (C,) are pre-
increases by 0.4%,while d increases by 26% (not shown in sented in Table I11 (miniprint), and depicted in Figures
Table 11).This is consistent with our assumption that an 1-4. The LHS results for H2O also are included in the fig-
important cause of the large changes in a and d in calcula- ures. The nature of the fit for our principal results is simi-
tion 2 relative to their values for H2O (calculation 1)is the lar both to that of calculation 2 and to that of LHS for
lack of precision in the values of Elat, ZPEinter, and H20. The theory correctly predicts the signs of the isotopic
AZPEintraused. differences in these thermodynamic quantities throughout
In calculation 4, most of the adjustment in Elat is the liquid temperature range. This is significant, because
achieved by increasing AZPEintra,while ZPEjnteris affected these are relatively small differences between large num-
very little. In order to further delimit the sensitivity of our bers and, therefore, represent a sensitive test of the theory.
results to variations in these parameters, we have per- The theoretical distributions of D2O clusters at 4, 50,
formed a final calculation (no. 5, Table 11) in which all of and 100°C are shown in Figure 5 along with the LHS H2O
the 300 cal/mol increase in Elat (from -13,700 cal/m01~~ to distributions. The results indicate that liquid D2O is com-
The Journal of Physical Chemistry, Vol. 79, No. 22, 1975
Thermodynamic Properties of Liquid Deuterium Oxide 2357

r I I I I I I I

- Ai 1.35 3.40
I IOOOC
-

n
L
U
I

-i
m

0 200 400 600 800


' I 2 3 5 . 4 5 6 7 S
Cluster Species u p (cm-l)

Figure 5. Calculated distributions of the mole fractions of H20 (un- Figure 6. Calculated histograms of the frequency distributions of
filled bars, taken from ref 4) and D20 (hatched bars) molecules in D20, g(q), at
two temperatures. The units of g(vJ are arbitrary, and
various cluster species at three temperatures. The "star" pentamer the histograms have been normalized to the same value at both
is represented by 5" and is presented beside the dimer and trimer temperatures. The smooth curves drawn through the histograms
because it, like them, is a noncyclic cluster species. The data for represent a roughly averaged g(vJ. The numbers over the bracketed
H20 at 4OC were obtained by interpolation. regions are the sums of g(vI)over those regions.

TABLE VII: Comparison of Theoretical and


Representative Experimental Results for
Intermolecular Motions in Liquid D20a
Approx positions of
Method vibrational bands, cm'j

This theory 40 160 330 520


Infrared 165 505
Raman' 60 175 3 50 500
Inelastic neutron
scatteringd 50 160 465
a Tabulation of representative experimental results adapted
from Walrafen.45 * Reference 42. c Reference 43. Reference 44. I l l I l l I 1
0 20 40 60 80 IO0

posed primarily of cyclic pentamers and hexamers, with di- Temperolure CC)
mers, trimers, and star pentamers the next most important Flgure 7. Theoretical fraction of hydrogen bonds unbroken, XHB.as
species. Only about 1%of the water molecules are mono- a function of temperature. The dashed line, which is for H20, is
meric. The DzO and H20 distributions are quite similar. taken from ref 4. The solid lines refer to the various D20 calculations
and are numbered as in Table 11.
They are also similar in their temperature dependence; as
was noted by LHS, increasing the temperature increases
the proportion of noncyclic relative to cyclic clusters. and, as before, HzO from LHS) are presented in Table VI11
Tables IV-VI (in miniprint) present the contributions of (miniprint). These include, as functions of temperature,
the various terms of the partition function to the calculated the fraction of unbroken hydrogen bonds ( X H B )hydrogen
,
values of F, U, and S. ZPEinterhas been added to FIC and , and ET)^^ as defined in sec-
bond energies [ ( E I ) , ~(EuIaq,
UHB,the energies derived from the internal configuration- tion VI], the experimental molar volumes (V), and the hy-
al partition function, to adjust the reference zero of energy drogen-bond relaxational heat capacity (Cp,relax = dUH%/
to the bottom of the 0 K ice intermolecular potential well. dT). The relaxational heat capacities are also plotted in
Tables IV-VI also include the isotopic differences calculat- Figure 4.The Xu% data are presented in Figure 7, and the
ed using the LHS HzO data. hydrogen-bond energies are depicted in Figures 8-10.
Histograms representing the spectral density g(v) of the
intermolecular vibrational normal modes of DzO at 4 and VI. Discussion
100°C are presented in Figure 6. These were calculated ac- Our principal results have been obtained by constraining
cording to eq 22 of LHS. Smoothed averages of the data are a and d to have the same values that LHS found for HzO
also presented in Figure 6. The major features of the calcu- and by adjusting ZPEinter(and therefore AZPE,,,,,,) to fit
lated spectra are peaks near 40, 160, 330, and 520 cm-'. the experimental values of F and U of D20, subject to the
There is also a small component at 700 cm-I corresponding constraint that Elat = -13,400 cal/mol (the value for H20).
to the highest frequency mode of the star pentamer. These The parameters a and d retain the values originally as-
results are in fairly good agreement with spectroscopic signed to them in the treatment of H20,4and thus the clus-
studies, the results of which are summarized in Table VII. ter model is extended to D20 through the addition of only
The red shift with increasing temperature which is found one adjustable parameter (Le., ZPEi,,,,).
in infrared spectra44is reproduced by the theory. This can We noted in sections 1V.A and V that calculation 4 (the
be seen by comparison of the sums of g(v) over the brack- principal results), calculation 2, and the LHS calculations
eted librational regions in Figure 6, a t 4 and 100OC. for HzO generate very similar fits to experiment. In all
Quantities related to the structure of the liquids (DzO cases, the major defect of the theory (as noted for HlO by
The Journal of Physical Chemistry, Vol 79, No 22, 1975
2358 Owicki, Lentz, Hagler, and Scheraga

qc
2 2600

1 1 l I I 1 , I I
0 20 40 60 80 100 0 20 40 60 BO I00
Temperature ("C) Temperature (TI

Figure 8. Theoretical hydrogen bond energy as a function of Figure 9. Theoretical hydrogen bond energy, (&Iaq, as a function of
temperature. The dashed line, which is for H20 is taken from ref 4. temperature. The dashed line, which is for H20, is taken from ref 4.
The solid fines refer to the various 1720calculations and are num- The solid lines refer to the various D20 calculations and are num-
bered as in Table 11. For a definition of (€Jag, see the Discussion sec- bered as in Table II.
tion.

LHS) is an inability to account properly for the tempera-


ture dependence of U and s.
We turn now to a discussion of the molecular details of
the thermodynamic differences between D20 and H20, as
given in Tables IV-VI and VIII. The energy data for D2O
and H20 are comparable, because they refer to the same
reference state.
Principal Thermodynamic Functions. The isotopic dif-
ferences in translational and cluster rotational thermody-
namic functions are small compared with those arising
from vibrations. This happens for three important reasons.
First, the number of translational and rotational degrees of
freedom is comparatively small. If one neglects intramolec-
13001
I
0
l l
20
I I
40
a
I I
60
Temperature (TI
I
---__

I
80
I ]
100

ular motions, a cluster of j water molecules has three trans- Flgure 10. Theoretical hydrogen bond energy, ET)^^, as a function
lational, three rotational, and (6.1' - 6) vibrational degrees of temperature. The dashed line, which is for H20, was calculated
of freedom. According to our calculations, most of the using the partition function from ref 4. The solid lines refer to the
water molecules are in clusters with j = 5 or 6, so that vi- various D20 calculations and are numbered as in Table II.
brational modes predominate.
Second, translations and rotations (except for j = 1) are fore is a state function of the system. In some cases, it may
fully excited in the temperature range treated, so that the be desirable to stress a particular contribution to the total
isotope effect on the contributions of these degrees of free- energy change for a given change in state. Then a second
dom to the overall internal energy is minimal. In contrast, part of the definition must be a statement of exactly what
many vibrational modes are excited only weakly, leading to components of the total energy change are to be consid-
a large isotope effect. ered.
Third, isotopic substitution of the hydrogen alters the An example will clarify this. Eisenberg and Kauzmann4'
molecular principal inertial moments much more than the give as one possible definition of the hydrogen bond energy
molecular mass. In the partition function, many vibrational EH-bond = -(1/2)Eiat (= 6700 cal/mol of bond for H2O).
terms are more sensitive to the inertial moments than to The initial state is ice Ih at 0 K; the final state is infinitely
the masses, while the reverse is true for translational and dispersed water molecules again at 0 K. The quantum me-
(roughly) for rotational terms. Comparatively, this en- chanical potential energy of the system is included in
hances the vibrational isotope effects. EH.bond, but Z P E i n t e r and A Z P E i n t r a are not. E H - b o n d iS a
The isotopic differences in Flc and in UHB are interme- measure of the energy of interaction of nonvibrating mole-
diate in size between the vibrational and translational and cules in ice.
rotational differences. They result primarily from the In this paper we consider three definitions of the hydro-
smaller value of AZPEi,tra in DzO. gen-bond energy which are relevant to the liquid state.
Hydrogen-Bond Energy. Next we discuss the hydrogen- These are the theoretical quantities (Eu)aq, and
bond energy in liquid water as well as the effects of isotopic ET)^^ tabulated in Table VI11 and depicted in Figures
substitution on this quantity. I t has been noted46 that 8-10 for both D2O and H20.
there are a variety of ways to define the "hydrogen-bond ( E I )and~~ have been defined by LHS as follows.
energy" in a given system, a fact which can lead to confu- ( E I ) is~ the
~ internal configurational energy increase per
sion in the comparison of various experimentally or theo- hydrogen bond ruptured in transforming ice at 0 K to the
retically determined values. Any definition of this quantity equilibrium liquid at a given temperature and experimen-
should refer to the energy changes during some well-de- tal molar volume.4s (471)~~ includes changes in the magni-
fined physical process which involves the rupture or forma- tudes of potential energies of interaction between mole-
tion of hydrogen bonds. The hydrogen-bond energy there- cules, the effects of the background potential, and the in-

The Journal of Physical Chemistry, Vol. 79, No.22, 1975


Thermodynamic Properties of Liquid Deuterium Oxide 2359

crease of ZPEintraupon the rupture of a hydrogen bond in the cyclic pentamers and hexamers. Nevertheless, the equi-
the liquid. It does not include translational, rotational, or librium cluster distributions for D2O and HzO are very sim-
(intermolecular) vibrational effects. ( E u ) ~a ~quantity
, sim- ilar. For this reason, the calculated fractions of hydrogen
ilar to (Edaq,refers instead to the conversion of the equi- bonds in D20, XHB,does not differ significantly from that
librium liquid at a given temperature and experimental in HzO, as shown in Figure 7. The uncertainties in the cal-
molar volume to a hypothetical liquid composed entirely of culations make it impossible to determine which liquid (if
monomers (Le., all hydrogen bonds broken) at the same either) is more fully hydrogen bonded. We can, however,
temperature and molar volume. The energy accounting is conclude that the isotopic differences in XHBare small,
done in exactly the same way as it is for For mathe- probably less than 0.01.
matical definitions of these quantities, the reader should There is no inconsistency in the finding that it takes less
consult L H S 4 energy to break a hydrogen bond in liquid D20 than in liq-
The above neglect of translational, rotational, and vibra- uid H2O (calculation 4, Figures 8-10) while, at the same
tional energy changes allows one to focus on the changes in time, D20 may be (slightly) more hydrogen bonded than
the magnitudes of the potential energies of interaction in HzO (calculation 4,Figure 7). The isotopic effects on these
the system. In actual physical processes, however, dynamic quantities are not related in a simple way.
effects (translational, rotational, and vibrational) are im- If one considers “water structure” to refer to the fraction
portant. We can include them by defining the total hydro- of hydrogen bonds unbroken in the liquid and to the equi-
gen-bond energy ET)^^ to be the total internal energy librium distribution of clusters in the liquid, then our re-
change of the system per hydrogen bond ruptured for the sults indicate that DzO and H2O have a very similar struc-
process used to define ( E u ) ~Le., ~ ; ET)^^ is the sum of ture throughout their common liquid temperature range.
( E u ) and
~ ~ changes in the translational, rotational, and vi- This is in accord with near-infrared work of Thomas et aLZ3
brational energies. and the low-angle X-ray scattering study of Narten and
Having defined these hydrogen-bond energies, we now LevyZ1mentioned in section I. This conclusion is relatively
consider the theoretical values that we have obtained for insensitive to variations in a, d , Elat, ZPEint,,, and
them in DzO and H2O. We have included in Figures 8-10 AZPEintrain the range of values likely to be encountered
the results of all the calculations in Table 11. for D2O.
In all three definitions, both the initial and final states Consistent with our finding that the structures of the
are condensed phases, so that the rupture of a hydrogen two liquids are similar, the calculated relaxational heat ca-
bond does not result in as complete a loss of stabilizing in- pacity of D2O is approximately equal to that of H20, as can
teractions with neighboring molecules as it does for be seen in Figure 4.In this theory, the differences between
EH-bond. Hence, hydrogen-bond energies calculated by the heat capacities of D20 and H2O arise not from relaxa-
these definitions are considerably smaller than EH-bond. tional effects, but principally from the lower frequencies of
In contrast to EH-bond, the hydrogen-bond energies intro- the D2O cluster vibrational modes, LHS noted that the
duced above are defined at temperatures other than 0 K theory underestimates the relaxational heat capacity. It is
and, in fact, vary considerably with temperature. For ex- possible that a more correct treatment of this quantity
ample, ET)^^ varies by -20% (-250 cal/mol of bond) over would produce a significant contribution to the isotopic
the liquid temperature range. The temperature depen- differences in the overall heat capacity.
dence of ( E I ) and~ ~ ( E I J )arises
~ ~ primarily from the NBmethy and Scheraga6J1 were able to extend their
changes of the equilibrium cluster distribution and molar theory to explain the temperature variation of the molar
volume with temperature. For ET)^^ there is the addition- volume of liquid water by assigning molar volumes and
al effect of the variation of translational, rotational, and vi- thermal expansion coefficients to the cluster and monomer
brational energies with cluster distribution and tempera- microphases. However, such a procedure would be both dif-
ture. ficult and invalid for the present theory. The reason for
Our calculations suggest that the energy required to this is that here the molar volume of the liquid is deter-
break the hydrogen bonds in the clusters of liquid D20 is mined by complicated packing interactions among nine
somewhat less than the corresponding energy in liquid types of irregularly shaped clusters, which are too small to
H20, as seen in Figures 8-10. The main reason for this in be treated as separate phases. This is the same problem
the case of and (EU)aq is that the change in ZPEintra which prevents us from calculating the radial distribution
upon breaking a hydrogen bond in the liquid is less for D20 function, as noted by LHS. Hence, within this theory, we
than for HzO. There are additional translational, rotation- cannot explain the differences in the temperatures of maxi-
al, and vibrational effects for ET)^^. The differences mum density of D20 and H20 except to say that these dif-
among the curves for different DzO calculations result from ferences must arise from the relatively small isotopic dif-
variations in the parameters listed in Table 11. These con- ferences in the theoretical equilibrium cluster distribu-
clusions, however, must be tempered by the knowledge that tions. A recurrent theme in this work has been that the dif-
the uncertainties in our calculations are not small with re- ferences in thermodynamic properties between DzO and
spect to the calculated isotopic hydrogen-bond energy dif- HzO are caused primarily by isotope effects directly on the
ferences. inter- and intramolecular vibrational frequencies. It should
Water Structure. We now consider the structural differ- be pointed out that, by constraining the values of a, d , and
ences between DzO and HzO. Although we have not dis- Elat for DzO and H2O to be equal (see section IV), we have
played the cluster distributions for calculations other than not allowed for differences arising from isotope effects on
no. 1 (H20) and 4 in Figure 5, those for calculations 2, 3, the molecular geometry and dipole moment. Available evi-
and 5 did not differ greatly from those depicted. The most suggests, however, that these molecular differ-
significant variation occurred for calculations 2 and 3 at ences are small. Our results provide a reasonably satisfac-
high temperatures, where the contribution of the star pen- tory description of the system without considering them.
tamer increased to approximately the same level as that of Finally, LHS noted that this cluster theory has much in

The Journal of Physlcal Chemistry, Vol. 79, No. 22, 1975


2360 Owicki, Lentz, Hagler, and Scheraga

common with a continuum point of view: water molecules the miniprinted material from this paper only, Tables I,
are allowed to exist in a large number of states (clusters), 111-VI, and VIII, or microfiche (105 x 148 mm, 24X reduc-
and inter-cluster interactions are found to be fairly strong. tion, negatives) containing all of the miniprinted and sup-
In the present paper, the theory again provides a bridge be- plementary material for the papers in this issue may be ob-
tween the traditional mixture and continuum approaches. tained from the Business Office, Books and Journal Divi-
Although it starts from a mixture premise, it agrees with sion, American Chemical Society, 1155 16th St., N.W.,
continuum theory studies7J0 in the finding that the ther- Washington, D.C. 20036. Remit check or money order for
modynamic differences between liquid D20 and H20 are $4.50 for photocopy or $2.50 for microfiche, referring to
due to rather subtle vibrational effects instead of grosser code number JPC-75-2352.
structural changes in the liquids.
References and Notes
VII. Summary and Conclusions (1) This work was supported by research grants from the National Institute
of General Medical Sciences of the National Institutes of Health, U.S.
This paper reports on the continuing study of the liquid Public Health Service (GM-14312), and from the National Science Foun-
water system by means of a statistical mechanical "cluster" dation (BMS71-00872 A04).
(2) (a) NSF Predoctorai Fellow and NIH Trainee at Cornell University. (b) NIH
model. A theory developed for H203!4is applied to D20 as a Trainee at Cornell University. (c) To whom requests for reprints should
further test of the theory and in order to gain insight into be addressed at Corneil University.
the differences between these liquids. (3) A. T. Hagler. H. A. Scheraga, and G. Nemethy, J. Phys. Chem., 76,
3229 (1972).
Variation of the two adjustable parameters of the theory (4) B. R. Lentz, A. T. Hagler, and H. A. Scheraga. J. Phys. Chem., 78, 1531
to fit experimental thermodynamic quantities produces (1974).
(5) The term "hydrogen bond" is used here for both 0-He-0 and 0-De-0
unexpectedly large changes in the parameters relative to interactions.
their values for H20, and also some physically unreal iso- (6) G. Nemethy and H. A. Scheraga, J. Chem. Phys., 41, 680 (1964).
tope effects at the molecular level. We have attributed this, (7) C. G. Swain and R. F. W. Bader, Tetrahedron, I O , 182 (1960).
(8) M. S. Jhon, J. Grosh, T. Ree, and H. Eyring, J. Chem. Phys., 44, 1465
in large part, to uncertainties in the experimental values of (1966).
the 0 K ice lattice energy (Elat),intermolecular zero-point (9) C. M. Davis, Jr., and D. L. Bradley, J. Chem. Phys., 45, 2401 (1966).
(10) W. A. Van Hook, J. Phys. Chem., 72, 1234 (1968).
energy (ZPE,,t,,), and changes in intramolecular zero-point (11) G. Nemethy and H. A. Scheraga, J. Chem. Phys., 36,3382 (1962).
energy on sublimation (AZPE,,,,,), which are difficult to (12) It has been notedI3 that Nemethy and Scheragal' fit their theory for
obtain accurately. These inaccuracies have created difficul- H20 to an experimental value of the entropy of H20 which did not con-
tain the ap roximately 0.8 cal/mol deg zero point entropy of ice. Their
ties in the application of the theory, and we feel that an ac- f
D20 study, however, utilized the correct entropy: this caused an exag-
curate determination of the isotopic differences in these geration of the experimental thermodynamic isotopic differences. The
theoretically derived differences between the two liquids likewise were
quantities is crucial to a quantitative understanding of the greater than they should have been.
isotopic differences in the liquids. (13) A. T. Hagler, H. A. Scheraga, and G. Nemethy, Ann. N.Y. Acad. Sci.,
204, 51 (1973).
Because of the above-mentioned uncertainties, we have (14) A. Eucken and M. Eigen. Z.Nektrochem., 55, 343 (1951).
obtained our principal results by fixing the adjustable pa- (15) G. S. Kell in "Water a Comprehensive Treatise", Vol. 1, F. Franks, Ed.,
rameters and Elat for D20 a t their values for H20, and by Plenum Press, New York, N.Y., 1972, Chapter 10.
(16) D. Eisenberg and W. Kauzmann, "The Structure and Properties of
varying ZPE,,ter and AZPEintrato obtain the best fit to ex- Water", Oxford University Press, New York. N.Y., 1969, p 174.
perimental thermodynamic quantities subject to the con- (17) K. Lonsdale, Proc. R. SOC.London, Ser. A, 247, 424 (1958).
(18) R. W. Blue, J. Chem. Phys., 22, 280 (1954).
straint on Elst. In this calculation, the free energy (F),in- (19) A. Holtzer and M. F. Emerson, J. Phys. Chem., 73, 26 (1969).
ternal energy ( U ) ,and entropy (S)of D2O are fit fairly well (20) E. Kilmin, personal communication.
throughout the entire liquid temperature range. The error (2 1) A. H. Narten, M. D. Danford, and H. A. Levy, Discuss. Faraday SOC., 43,
97 (1967).
in the predicted heat capacity (C,) is more serious. The (22) H. S. Frank, in ref 15, Chapter 14.
size of the errors is somewhat larger than those obtained by (23) M. R. Thomas, H. A. Scheraga, and E. E. Schrier, J. Phys. Chem., 69,
3722 (1965).
fitting H20 with the same theory, but, significantly, the (24) J. L. Lebowitz and J. S . Rowlinson, J. Chsm. Phys., 41, 133 (1964).
signs of the isotopic differences in these thermodynamic (25) J. L. Lebowitz. Phys. Rev., 133, A895 (1964).
quantities are predicted correctly. We also have calculated (26) Normal mode analysis was carried out identically with that for H20 in B.
R. Lentz, A. T. Hagler, and H. A. Scheraga, J. Phys. Chem., 78, 1844
t,he vibrational spectral density in the intermolecular nor- (1974).
mal mode vibrational region, and have found it to be in rea- (27) A. Ben-Naim and F. H. Stillinger, Jr., in "Water and Aqueous Solutions",
R. A. Horne, Ed., Wiley-lnterscience, New York, N.Y., 1972, Chapter 8.
sonable agreement with vibrational spectroscopic results. (28) 0. Weres and S. A. Rice, J. Am. Chem. Soc., 94, 8983 (1972).
The results of this calculation suggest that the isotopic (29) B. R. Lentz and H. A. Scheraga, J. Chem. Phys., 58, 5296 (1973): 61,
differences in F, U , S, and C, result primarily from the 3493 (1974).
(30) E. Whalley, Trans. Faraday Soc., 53, 1578 (1957).
changes in intra- and intermolecular vibrational frequen- (31) Reference 16, p 4.
cies caused by isotopic changes of the mass and principal (32) T. R. Dyke and J. S. Muenter, J. Chem. Phys., 59, 3125 (1973).
(33) Reference 16, p 101.
moments of inertia of the water molecule. (34) B. L. Baker, US. Atomic Energy Commission Report No. AECU-4738,
We have concluded that the energy required to break a 1959.
hydrogen bond in liquid D2O is somewhat lower than in liq- (35) G. S. Kell, J. Chem. Eng. Data, 12, 66 (1967).
(36) A. J. Leadbetter, Proc. R. SOC.London, Ser. A, 287, 403 (1965).
uid H20. The similarities of the equilibrium cluster distri- (37) The reduced x2.as described by B e v i n g t ~ nis. ~a~measure of the good-
butions and the fractions of hydrogen bonds unbroken for ness of fit after the variance of the experimental data has been taken
into account. We have estimated the variance of ail experimental data
both D20 and H20 imply that the structures of the liquids at 2.5%: a reduced x2 value of about 1.5 or less indicates that it is not
are quite similar. unlikely that the theory provides a valid representation of the data. The
In summary, we find that most of the thermodynamic theory fits D20 less well than H20. for which LHS obtained a reduced x2
value of 1.47.
differences between liquid D2O and H20 are explainable (38) P. R. Bevington, "Data Reduction and Error Analysis for the Physical
directly in terms of isotopic effects on the intra- and inter- Sciences", McGraw-Hill. New York, N.Y., 1969, Chapters 10 and 11.
(39) In addition to the background potential, the a parameter includes a small
molecular vibrational frequencies rather than by differ- contribution from the shift in ZPEintra upon the rupture of a hydrogen
ences in the structures of the two liquids. bond in the liquid (see LHS4).This quantity is isotope dependent, but the
effect should be small enough to neglect at our present level of approxi-
mation.
Miniprint Material Available. Full-sized photocopies of (40) E. Whalley, in "Physics and Chemistry of Ice", E. Whalley, S.J. Jones,

The Journal of Physical Chemistry, Vol. 79, No. 22, 7975


Energy Parameters in Polypeptides 2361

and L. W. Gold, Ed., Royal Society of Canada, 1973, p 73. (45) G. E. Walrafen in "Water, A Comprehensive Treatise", Vol. I, F. Franks,
(41) Reference 16, p 138. Ed., Plenum Press, New York, N.Y.. 1972, Chapter 5.
(42) D. A. Draegert. N. W. B. Stone, B. Curnutte, and D. Williams, J. Opt. (46) Reference 16, p 137.
SOC.Am., 56, 64 (1966). (47) Reference 16, p 139.
(43) G. E. Walrafen, J. Chem. Phys., 40, 3249 (1964). (48) The difference in temperature between the initial and final states is not
(44) K. E. Larsson and U. Dahlborg, J. Nucl. Energy, Parts A B , 16, 81 important, because in this theory the relevant energies (e.g., €bt) are
(1962). taken to be temperature independent.

Energy Parameters in Polypeptides. VII. Geometric Parameters, Partial Atomic Charges,


Nonbonded Interactions, Hydrogen Bond Interactions, and Intrinsic Torsional Potentials
for the Naturally Occurring Amino Acids'

F. A. Momany,2aR. F. McGuire,2bA. W. Burgess,2cand H. A. Scheraga*2d


Department of Chemistry, Cornel1University, ithaca, New York 14853 (Received January 3 1, 1975;
RevisedManuscript Received July 10, 1975)

Empirical interatomic potentials are developed for calculating the energetically most favored conforma-
tions of polypeptides and proteins. Geometric parameters, partial atomic charges, nonbonded interaction
energies, hydrogen bond energies, and intrinsic torsional potentials are determined for each of the natural-
ly occurring amino acids. The geometric parameters were obtained from a survey of the recent structural
literature; the bond lengths and bond angles of a given type of residue appear to be similar from structure
to structure. The partial atomic charges (overlap normalized) for every atom of each amino acid were ob-
tained by the CNDOI2 method. Parameters of the nonbonded and hydrogen-bonding potentials were
taken from previous calculations on crystals of small molecules; however, the coefficient of the nonbonded
repulsive term for two atoms separated by three bonds, the central one being the bond about which rota-
tion can take place (1-4 interactions), was reduced by a factor of 2 to make the repulsive force constants
compatible with those computed from Hartree-Fock and Thomas-Fermi-Dirac repulsive potentials. In-
trinsic torsional potentials were introduced, where necessary, to reproduce experimental internal rotation
barriers. Interaction arrays for interatomic pairwise interactions were defined in order to distinguish hy-
drogen bonding, and 1-4, from 1-5, 1-6, etc., interactions. Procedures are described for carrying out con-
formational energy calculations on polypeptides and proteins, using these empirical potentials. Computer
programs for performing these calculations are available.

I. Introduction menta, and later by using MO-LCAO u charges of Del Re48


and MO A charges from Huckel theory.4a Intrinsic torsional
Recent investigations in this and other laboratories3-12 terms were included, with rotational barriers derived from
have emphasized the need for accurate interatomic poten- model A t this stage of development, hydro-
tials to determine polypeptide and protein conformations. gen bonding was treated by a modification of the Lippin-
The present paper provides the details of the set of poten- cott-Schroeder potential function.48 Using this set of po-
tials presently in use in this laboratory. tentials, calculations were carried out to investigate the
In order to obtain a perspective on the present status of conformational properties of single amino-acid residues,
potential functions, we begin with a short summary of the homopolymers, and cyclic peptide^.^ In a few cases [cyclo-
evolution of these empirical potentials. The earliest investi- (GlysProz) and cyclohexaglycyl],4d flexible geometry (i.e.,
gations of polypeptide conformation were carried out with bond stretching and bond angle bending) was allowed by
hard-sphere repulsive potentials, using the van der Waals using appropriate force constants for these motions. When
contact distance as an overlap criterion; these potentials ef- applied to proteins (e.g., lysozyme,13 a-lactalbumin,14 and
fectively allowed for interatomic excluded volume effects, r~bredoxin'~), united-atom potentials were used for -CH,
and provided rough estimates of the allowed regions of con- -CH2, and -CH3 groups to reduce the number of interac-
formational tions that had to be computed.
Next, these potential functions were improved by replac- In the present paper, we report the results of our reex-
ing the hard sphere by a Lennard-Jones 6-12 potential. amination of the problem of interatomic potentials, where-
The attractive (dispersion) term was calculated from atom- by we have obtained an improved self-consistent set. First,
ic polarizabilities, and the position of the minimum of the the coefficients of the repulsive interatomic potential have
6-12 potential was taken to be the van der Waals contact been refined by treating the packing of molecular crystals.9
d i ~ t a n c e . ~ Electrostatic
",~ energies were also included by In that work, it was found that the position of the mini-
estimating partial atomic charges, first from bond mo- mum of the interatomic potential was not the van der

The Journal of Physical Chemistry, Vol. 79, No. 22, 1975

S-ar putea să vă placă și