the journal of chemical thermodynamics volume 73 issue 2014 [doi 10.1016_j.jct.2013.12.018] irikura,...
TRANSCRIPT
-
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
1/7
Anharmonic partition functions for polyatomic thermochemistry
Karl K. Irikura
Chemical Sciences Division, National Institute of Standards and Technology, Gaithersburg, MD 20899-8320, USA
a r t i c l e i n f o
Article history:
Available online 4 January 2014
Dedicated to the memory of the lateProfessor Manuel Ribeiro da Silva
Keywords:
Anharmonicity
Density of states
Partition function
Quantum chemistry
Thermochemistry
a b s t r a c t
In quantum chemistry, the computation of anharmonic vibrational frequencies, instead of only harmonic
frequencies, is now convenient and increasingly common. For thermochemistry, one ramification is that
vibrational partition functions may be constructed using anharmonic manifolds. Practical issues are dis-cussed here, and a few examples are provided. Evaluating the accuracy of the results is hindered by the
lack of experimental data for ideal-gas heat capacities at high temperature.
Published by Elsevier Ltd.
1. Introduction
Quantum chemistry is increasingly important in thermochem-
istry. Advances in electronic structure theory now allow precise
predictions (that is, uncertainties of only a few kJ mol
1
) to bemade on molecules as large as guanine and tetracene [1], in the
ideal-gas limit. At the small-molecule extreme, precision for small
clusters of helium is high enough to serve as a physical measure-
ment standard [2]. Although special problems are encountered
for molecules with unusual electronic structure, such as
metal-centered radicals, for typical organic compounds the
accuracy bottleneck has become the contributions of molecular
vibrations.
The vibrational zero-point energy (ZPE) contributes already at
zero temperature. ZPE is a large quantity, on the order of 20 n
kJ mol1 for typical organic compounds, where n is the number
of atoms in the molecule. It is manifested in the atomization en-
thalpy of a molecule, because atoms contain no ZPE. Since atoms
are a common, if not optimal, choice of thermochemical reference
state in quantum calculations, errors in the ZPE can be important.
However, because ZPE cannot be measured (or even defined)
experimentally, its accuracy is problematic to assess[3].In conven-
tional quantum thermochemistry, vibrations are computed in the
harmonic approximation. Thus, ZPE is taken as one-half the sum
of the frequencies. It is often multiplied by an empirical scaling
factor to compensate for bias [4].
Besides ZPE, the other major contribution of vibrations is to
temperature-dependent thermodynamic functions, by way of the
molecular partition function. Standard practice is to make the sim-
plifying approximations that electronic, vibrational, rotational, and
translational degrees of freedom are separable and non-interact-
ing, that the molecular geometry is independent of rotational exci-
tation, and that vibrations are harmonic. This is well-known as therigid-rotor/harmonic-oscillator (RRHO) approximation. Sometimes
the vibrational frequencies are scaled by empirical factors before
computing the thermodynamic functions[5].Unlike the ZPE, ther-
modynamic functions can be measured experimentally. Thus, it is
possible to assess the accuracy of theoretical predictions, at least in
principle.
Deficiencies of the harmonic-oscillator model are well-known, if
seldom addressed in practice. Torsional motion, such as exhibited
by methyl groups, is not harmonic. The classic work by Pitzer[6]
continues to guide development in this area [7]. Other large-
amplitude vibrations, such as inversions and ring-puckering, are
equally problematic [8]. Even for relatively rigid molecules,
vibrations are anharmonic. In the simplest anharmonic model,
vibrational energy levels are conventionally described by two-
mode anharmonicity constants (xij) according to the approximate
equation(1). The xiare harmonic constants, thexijare anharmonicconstants, vi is the quantum number for the ith vibrational mode,
and the sums are over all the vibrational modes in the molecule
Ev1;v2;. . . ;vN Xi
xi vi1
2
Xi
Xj6i
xij vi1
2
vj
1
2
: 1
Simple anharmonic calculations are increasingly available as
keyword options in popular quantum chemistry software.
Although such calculations require far more computer time than
0021-9614/$ - see front matter Published by Elsevier Ltd.http://dx.doi.org/10.1016/j.jct.2013.12.018
Tel.: +1 301 975 2510; fax: +1 301 869 4020.
E-mail address:[email protected]
J. Chem. Thermodynamics 73 (2014) 183189
Contents lists available at ScienceDirect
J. Chem. Thermodynamics
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j c t
http://dx.doi.org/10.1016/j.jct.2013.12.018mailto:[email protected]://dx.doi.org/10.1016/j.jct.2013.12.018http://www.sciencedirect.com/science/journal/00219614http://www.elsevier.com/locate/jcthttp://www.elsevier.com/locate/jcthttp://www.sciencedirect.com/science/journal/00219614http://dx.doi.org/10.1016/j.jct.2013.12.018mailto:[email protected]://dx.doi.org/10.1016/j.jct.2013.12.018http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://crossmark.crossref.org/dialog/?doi=10.1016/j.jct.2013.12.018&domain=pdfhttp://-/?- -
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
2/7
harmonic calculations, they require little or no additional effort by
the user. Thus, one may anticipate that their popularity will in-
crease. The output of such quantum chemistry programs typically
includes a table of the spectroscopic constants for use in equation
(1). The purpose of the present work is to examine the effect upon
thermodynamic functions of including simple vibrational anhar-
monicity as obtained from such programs. The resulting model
may be called the rigid-rotor/anharmonic-oscillator (RRAO)approximation.
Only small molecules were considered for investigation here.
Larger molecules [9] were avoided for three reasons. First, the
anharmonic theory included in the quantum software is second-
order vibrational perturbation theory (VPT2), which presumes that
there are only minor deviations from a harmonic potential energy
function. Internal rotation and ring puckering, which are common
in large molecules, have multiple minima that cannot be described
by small adjustments to a harmonic potential. The VPT2 model is
expected to give poor results in such situations. Furthermore, it
is doubtful that such a spectrum can be fitted by equation (1).
Second, the equations of VPT2 feature many denominators that
are energy differences between vibrational levels. Pathological
results may occur when these energy differences approach zero,
which is increasingly likely in larger molecules. Third, low-quality
electronic structure theory produces anharmonic spectra of
dubious quality[10]. Consequently, high-quality calculations are
desired here, which are too expensive for large molecules.
2. Computational methods[11]
2.1. Difficulties with standard formalism
Unfortunately, it is not straightforward to compute anharmonic
partition functions. Harmonic partition functions can be written in
simple closed form, as shown in textbooks on statistical thermody-
namics. Uncoupled anharmonic oscillators can be handled by ana-
lytical approximations[12]or by efficient methods of enumeration
[13], but the levels described by equation (1)must be enumerated
individually. This presents a practical problem because there are
billions of levels even for small molecules. Another complication
is that the anharmonic constants are often negative-valued.
Eventually, the energy given by equation(1)decreases as the quan-
tum numbers increase, which is physically unreasonable (although
in some cases it can be interpreted as bond dissociation). Thus, the
vibrational manifold must be truncated wherever this occurs. This
truncation may lead to physically unreasonable behavior, as exem-
plified below. The molecules in the present study are small enough
to be tractable using a simple Monte Carlo method to estimate the
anharmonic density of states (DOS), starting from a set of harmonic
and anharmonic constants. A more efficient method, based
upon the Wang-Landau classical algorithm [14], has been
described and developed by Basire et al. [15]and by Nguyen andBarker[16].
2.2. Density of states (DOS)
For computational efficiency, energy levels are grouped in en-
ergy bins (width =DE= 2 cm1) instead of listing them individu-
ally. The number of states in the ith bin is the density of states
times the width of the bin, qEiDE. Thus, the vibrational partitionfunction is expressed by equation(2). In practice, the sum cannot
run to infinite energy and is truncated at some cutoff energy, Emax,
so that the sum stops atimax Emax=DE. In other words, the DOS is
taken to be zero at energies above the cutoff. This is in contrast to
the RRHO, in which the DOS increases without bound and the
vibrational partition function is summed analytically to E 1
QvibT Xi
qEiDE expEi=kT: 2
For small problems the density of states, qE, may be evaluatedexactly within the model by exhaustive enumeration of vibrational
levels (often called direct counting). However, thermodynamic
functions converge slowly with the cutoff energy. This requires a
large value ofEmax, which makes enumeration impractically slow.
An approximate Monte Carlo method is described below.Low-lying states contribute most to the partition function. Thus,
the most reliable DOS is obtained by direct enumeration up to
some practical energy, followed by an estimate up to the cutoff en-
ergy,Emax, at which the DOS is truncated. Emaxmay be increased to
verify that it is high enough for thermodynamic functions to be
converged at the temperatures of interest. The Monte Carlo esti-
mate may be improved either by increasing the number of samples
or, equivalently, by constructing it as a composite of estimations
with increasing values ofEmax. This is advantageous because the
computational efficiency decreases as Emax increases.
For ThF4, enumeration was used for energies up to 7000 cm1
followed by Monte Carlo estimation to 15,000 cm1 and then again
to 50,000 cm1. For CF3Br, enumeration was used up to
20,000 cm1
followed by Monte Carlo estimation to 30,000 cm1
and then again to 50,000 cm1. For C2H4 (ethylene), enumeration
was used up to 30,000 cm1 followed by Monte Carlo estimation
to 50,000 cm1. For C2H4O (ethylene oxide), enumeration was used
up to 25,000 cm1 followed by Monte Carlo estimation up to
50,000 cm1.
For reference, harmonic (all x ij= 0) and uncoupled anharmonic
(xij= 0 fori j) DOSs were computed using the SteinRabinovitch
algorithm [13] with a bin size of 2 cm1. The accuracy of this
algorithm is limited only by the accumulation of rounding errors
associated with the energy binning.
2.3. Monte Carlo estimation
Each vibrational energy level is specified uniquely by a set ofvibrational quantum numbers, which may be represented as a
point in quantum-number space. We estimate the DOS by Monte
Carlo sampling within an appropriate, fixed volume in quantum-
number space. If a large number of points is sampled (Ntot), and
Ni of those points have energies within the bin Ei, then the DOS
is approximated by equation(3), where Vtot is the volume of the
sampling region. If the volume or the dimensionality is too large,
this strategy is uselessly inefficient, but it is adequate for the
molecules under consideration here. The present procedure has
similarities with that by Barker [17] but was developed
independently.
qEiDE Ni=NtotVtot: 3
The sampling volume is defined by a set of maximum values ofthe quantum numbers. It is represented by inequality(4), wherebiis the boundary (maximum) value for the (non-negative) quantum
number vi. Geometrically, the sampling volume is a right-hypertri-
angle in quantum-number space with a vertex at the origin and a
vertex on the ith axis at bi. Its volume is given by equation (5),
where D is the dimensionality, that is, the number of vibrational
modes. Note that using a rectangular sampling volume would de-
crease computational efficiency by a large factor of (D!). Given a
cutoff energy (Emax), each boundary valuebiis initially determined
from equation(1)while holding vj= 0 (j i). The initial value ofbiis the smaller of: (1) the value ofvifor whichE=Emax; (2) the value
ofvifor which @E=@vi 0. Because many anharmonic constants are
negative-valued, regions of the upper surface of the volume may be
convex instead of flat. To compensate, an inflation factor (/),
184 K.K. Irikura / J. Chem. Thermodynamics 73 (2014) 183189
-
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
3/7
greater than 1 but typically less than 2, is applied to all boundary
values to ensure that the sampling volume includes all points of
interest. The value of/ may be increased to verify that the sam-
pling volume is large enough, but if/ is too large the acceptance
ratio (that is, the computational efficiency) will be impractically
small. Finally, each bi is increased by 1 so that the desired range
of values will be covered after truncating to integers.
Xi
vi=bi 6 1; 4
Vtot D!1Yi
bi: 5
Uniform sampling over the hyper-triangular volume starts with
a set of D random floating-point variables, ui, each taken from a
rectangular distribution over the conventional interval [0, 1). They
are mapped onto floating-point values,yi, using the transformation
in eq. (6) (see Appendix for derivation), starting with i= 1 and
working upward. Truncating gives integer values for the quantum
numbers, vi= [yi] 6 bi. A point is rejected if its energy exceeds Emaxor if@E=@vj 6 0 for any j.
yi bi1 1 ui1=D1i
Yj
-
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
4/7
oxide. The VPT2//CCSD(T)/cc-pVTZ result is higher than the exper-
imental value by 280 cm1. This appears to be a failure of VPT2,
since the harmonic calculation (x14) is only 28 cm1 higher that
the experimental fundamental. Such failures are usually attribut-
able to near-resonances and are sensitive to the rather arbitrary
choice of associated thresholds[22]. In this case, most of the x i,14are positive-valued and several are implausibly large (seeSupple-
mentary material). Note that eq.(1)implies that the fundamental
frequency, mj, for mode j has the value given below.
mj xj1
2Xij
xij 2xjj: 10
As pointed out by a reviewer, it is standard practice among
spectroscopists to perform a de-perturbation analysis to address
pathologies such as that observed for m14for ethylene oxide. How-
ever, no such analysis was done here, for two reasons. First, such
analysis is beyond the skills of non-specialists (including the
present author), who rely upon the quantum chemistry software.
Second, this failure for such a simple molecule is a warning that
VPT2 is questionable, or at least premature, for general application
in thermochemistry.
3.3. ZPE
Anharmonic, uncoupled-anharmonic (also called diagonal-anharmonic), and harmonic values are presented in table 3. The
anharmonic contribution to ZPE is fairly small for these four mol-
ecules, and the off-diagonal contribution is nearly negligible.
3.4. Density of states (DOS)
Figure 2compares DOSs from harmonic (xij= 0), diagonal (xij= 0
fori j), and anharmonic models for ThF4. At low energy, the re-
sults conform to expectations; the diagonal-anharmonic DOS is
slightly larger than the harmonic DOS and the fully anharmonic
DOS is much larger. Above 17,000 cm1, however, the anhar-
monic DOS is the smallest of the three models, even declining in
magnitude above 35,000 cm1. The off-diagonal anharmonicities
have a major effect. For CF3Br the situation is the same qualita-tively, but the crossover occurs near 26,000 cm1 and the decline
begins near 90,000 cm1. For C2H4 the harmonic and uncoupled-
anharmonic models give nearly superimposable DOSs at high en-
ergy. The anharmonic DOS is larger by an increasing factor until
200,000 cm1, where it begins a decline. It crosses the harmonic
and uncoupled DOSs around 240,000 cm1. For C2H4O, the three
models give very similar results (figure 3). The anharmonic DOS
is larger until it crosses the other DOSs near 230,000 cm1. Beyond
that, it is nearly constant until declining gently from around
350,000 cm1. For a thermochemical context, note that
100,000 cm1 1200 kJ mol1 12.4 eV, larger than nearly any
bond dissociation energy and most ionization energies.
There are few measurements of DOS in gas-phase molecules. No
data are available for comparison with the present calculations.However, for formaldehyde (CH2O), is has been reported that the
RRHO and RRAO models predict DOSs that are too small by factors
of 11 and 6, respectively[23].
3.5. Thermodynamic functions
The values of the thermodynamic functions depend upon the
cutoff energy (Emax) used in the vibrational partition function. A
representative example is shown in figure 4. To obtain converged
results, the required cutoff energy is much larger than kT(in wave
number units k 0.695 cm-1 K1). A higher cutoff energy is
needed for functions at higher temperature. Moreover, the re-
quired cutoff is roughly proportional to the number of atoms in
the molecule. All thermodynamic functions in this study were con-verged by choosing a high cutoff energy ofEmax= 50,000 cm1. The
heat capacity (Cp) is more sensitive toEmaxthan the entropy (S) and
the enthalpy content (DT0Hvib), presumably because it depends
upon the second derivative of the partition function instead of only
the first derivative. The same convergence study for the harmonic
partition functions shows the same qualitative behavior, and con-
verges properly to the analytical harmonic results.
TABLE 1
Lowest and highest harmonic frequencies and anharmonicity constant of largest
magnitude (constants not scaled).
xlow/cm1 xhigh/cm
1 Largest xij/cm1
ThF4 110.3 cm1 594.6 2.26
CF3Br 308.2 1218.6 88.01
C2H4 823.4 3248.7 68.03
C2H4O 816.3 3210.8 106.40
TABLE 2
Vibrational fundamentals from anharmonic theory and from experiment (Refs. [35],
[36],[37], and[38]for ThF4, CF3Br, C2H4, and C2H4O, respectively).
Theory/cm1 Expt/cm1
ThF4, m1 (a1) 589 618 (est.)
m2 (e) 112 121 (est.)
m3 (t2) 532 520m4 (t2) 107 116
CF3Br, m1(a1) 1079 1089
m2 (a1) 761 760
m3 (a1) 360 349
m4 (e) 1193 1210
m5 (e) 545 547
m6 (e) 306 306
C2H4, m1 (ag) 3010 3026
m2 (ag) 1635 1623
m3 (ag) 1342 1342
m4 (au) 1025 1023
m5 (b1g) 3077 3103
m6 (b1g) 1222 1236
m7 (b1u) 945 949
m8 (b2g) 943 943
m9 (b2u) 3096 3106
m10 (b2u) 822 826m11 (b3u) 2997 2989
m12 (b3u) 1435 1444
C2H4O, m1 (a1) 2976 3005
m2 (a1) 1493 1497
m3 (a1) 1261 1270
m4 (a1) 1129 1120
m5 (a1) 878 877
m6 (a2) 3065 3050 to 3073
m7 (a2) 1177 1043 to 1046
m8 (a2) 1090 807 to 851; 1020
m9 (b1) 2985 2978
m10 (b1) 1490 1470
m11 (b1) 1206 1151
m12 (b1) 894 822
m13 (b2) 3184 3065
m14 (b2) 1427 1147
m15 (b2) 853 808
TABLE 3
Vibrational zero-point energies (unscaled) under different approximations.
ThF4 CF3Br C2H4 C2H4O
Harmonic/cm1 1384.0 3191.4 11,157.1 12,640.6
Uncoupled/cm1 1382.9 3197.4 11,142.4 12,625.1
Anharmonic/cm1 1375.8 3177.3 11,007.5 12,667.6
186 K.K. Irikura/ J. Chem. Thermodynamics 73 (2014) 183189
http://-/?-http://-/?-http://-/?-http://-/?- -
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
5/7
4. Discussion
Explicitly considering first-order anharmonicity, as done here,
has variable effects upon computed ideal-gas thermodynamic
functions. The RRAO and RRHO thermodynamic functions (S, Cp,
andDT0Hvib) are tabulated and summarized graphically in theSup-
plementary materialfor the four molecules considered here. Below
1000 K, the anharmonic values are larger than the harmonicvalues, except in the case of C2H4O. Among the four molecules,
the extreme examples are C2H4O (ethylene oxide) and ThF4. For
C2H4O, the anharmonic DOS is nearly identical with the harmonic
DOS (seefigure 3). Consequently, the thermodynamic functions are
nearly equal. The largest difference is for Cp, shown infigure 5. In
contrast, the anharmonic DOS for ThF4 (figure 2) displays trunca-
tion effects as discussed above. The largest difference is again for
Cp. As shown in figure 6, the anharmonic Cp(T) rises faster than
the harmonic function, reaches a maximum near T =700 K, and
then declines sharply. Like the decline in the DOS, this does not ap-
pear physically reasonable. However, an alternative, if unlikely,
interpretation is possible. If the DOS turnover corresponds to disso-
ciation of the molecule, rather than being an artifact of the simple
first-order anharmonic approximation (equation (1)), then the
decline in Cp for ThF4 molecules does not reflect what would be
measured experimentally for a sample of ThF4 together with its
equilibrated dissociation products.
Harmonic vibrational frequencies are often multiplied by
empirical scaling factors to correct for anharmonicity and for other
deficiencies. Scaling factors have been optimized for thermody-
namic functions[2427]. Unfortunately, because of the lack of data
[24],the experimental benchmarks in such exercises are usually
not from experimental measurements, but from RRHO calculations
based upon experimental vibrational frequencies. Below, we con-
sider whether the results of the present RRAO calculations can be
recovered by using scaled frequencies in the RRHO model.
For each combination of one molecule, one thermodynamic
function, and one temperature, we seek a frequency scaling factor
that causes the RRHO value of the thermodynamic function toequal the RRAO value. A scaling factor can always be found to make
the entropies match, but not necessarily forCpor for DT0H(see Sup-
plementary material for details). The resulting scaling factors are
listed in the Supplementary Material. No single value of the scaling
factor can make the RRHO results conform to the RRAO results,
either at all temperatures for a single thermodynamic function or
for all thermodynamic functions at a single temperature. This is
the expected behavior, but not often acknowledged.
FIGURE 2. Densities of states for ThF4 from harmonic, uncoupled anharmonic, and
coupled anharmonic models.
FIGURE 3. Densities of states for C2H4O (ethylene oxide) from harmonic, uncoupled
anharmonic, and coupled anharmonic models.
FIGURE 4. Ideal-gas heat capacity (Cp) of C2H4O (ethylene oxide) from the RRAO
model, as a function of cutoff (i.e., truncation) energy in the vibrational partition
function. From bottom to top: T/K = 298.15, 400, 500, 700, 900, 1200, 1500, and2000.
FIGURE 5. Ideal-gas heat capacity (Cp) of C2H4O (ethylene oxide) from the RRAOand RRHO models.
K.K. Irikura / J. Chem. Thermodynamics 73 (2014) 183189 187
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?- -
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
6/7
An important question is whether the anharmonic constants,
gained at such great computational expense, yield thermodynamic
functions that are in better agreement with experiment. As shown
infigure 6, and in the Supplementary Material, the largest differ-ence from harmonic theory is for heat capacity (Cp) and at high
temperature. Unfortunately, experimental data are unavailable at
high temperatures. Comparisons with available data, as shown
below, are inconclusive.
For ThF4, a comparison in the reference[21]was made for the
entropy at T =1150 K. The experimental value [28] is
483 4 J K1 mol1. The anharmonic value was 514 J K1
mol1, in worse agreement than the harmonic value of 490 J K1
mol1. The present results are essentially the same, as expected
because the spectroscopic constants from the reference[21]were
used here.
For CF3Br, no experimental data appear to be available at any
temperature despite the technological importance of this
compound.For C2H4, the ideal-gas entropy at T = 169.4 K has been reported
as S = (198.2 0.4) J K-1 mol1 [29]. The present computations
yield anharmonic and harmonic values of (198.2 and 198.1)
J K1 mol1, respectively. These values agree equally well with
the experimental value. For Cp, several measurements have been
made at temperatures up to 473 K [30], but many of these
publications are difficult to obtain. Published values, covering the
temperature range from 178 K to 473 K, are compared with the
calculations infigure 7. The experimental data by Burcik et al., by
Eucken and Parts, and by Haas and Stegeman were taken from
the reference [31] (uncertainties not reported). The data by
Senftleben[32]are higher, for the real gas, and not included in this
comparison. Although the highest-temperature datum appears to
agree better with the RRAO model than with the RRHO model,the weaker agreement at low temperatures suggests that this is
fortuitous.
For C2H4O, the ideal-gas entropy at T =283.66 K has been re-
ported as S= 240.1 J K-1 mol1 (no uncertainty reported) [33].
The RRAO and RRHO models yield (240.1 and 240.2) J K1 mol1,
respectively. These values agree equally well with the experimen-
tal value. For heat capacity, the models are compared in figure 8
with experimental ideal-gas data by Hurly[34]. The RRHO model
is markedly (2 times) closer to experiment than the more sophis-
ticated RRAO model, although both underestimate the heat capac-
ity. As mentioned above, the anharmonic constants associated with
mode 14 are suspect. If those constants are arbitrarily divided by
10 in the RRAO model, the resulting Cp(T) values are 2 times clo-
ser to experiment than the RRHO, but such an arbitrary procedurehas little predictive value.
5. Conclusions
High-temperature experimental measurements of ideal-gas
heat capacities are needed to test the theory more stringently.
However, data below T =500 K suggest that including first-order
vibrational anharmonicity, by means of eq. (1), does not improve
upon the simple harmonic-oscillator model, at least for the small,
rigid molecules studied here. Failures of VPT2 theory can cause
serious errors in vibrational constants, as found here for ethylene
oxide. In the regime of non-physical, negative energy increments,
truncation of the density of states (DOS) leads to implausible re-
sults, as found here for thorium tetrafluoride. An alternative
approximation for the DOS is needed at these higher energies.
Off-diagonal anharmonicity contributes negligibly to ZPE for these
four molecules, but can be important for thermodynamic
functions.
Appendix A
Size of Monte Carlo sampling volume
Equation (5) may be proved by induction on D. For D= 1, the
length of the line segment is V1=b1. For D= 2, the area of theright-triangle isV2=b1b2/2. The volume of the right-hyper-triangle
FIGURE 6. Ideal-gas heat capacity (Cp) of ThF4 from the RRAO and RRHO models. FIGURE 7. Ideal-gas heat capacity (Cp) of C2H4 from the RRAO and RRHO models
and from experimental measurements.
FIGURE 8. Ideal-gas heat capacity (Cp) of C2H4O from the RRAO and RRHO models
and from experimental measurements.
188 K.K. Irikura / J. Chem. Thermodynamics 73 (2014) 183189
-
7/26/2019 The Journal of Chemical Thermodynamics Volume 73 Issue 2014 [Doi 10.1016_j.jct.2013.12.018] Irikura, Karl K. --
7/7
in (D+ 1)-space is the integral (equation(A.1)) over 0; bD1 of the
slice of areaVD, whose verticesaidepend upon the integration var-
iableyD1 by way of equality(4), as shown explicitly in eq. (A.2).
Substitution and integration complete the proof, equation(A.3).
VD1
Z bD10
VDyD1dyD1 1
D!
Z bD10
YDi1
ai dyD1; A:1
ai bi 1 yD1bD1
; A:2
VD1 1
D!
YDi1
bi
Z bD10
1 yD1bD1
DdyD1
1
D 1!
YD1i1
bi: A:3
Probability distribution function
To build a sampling method starting from a rectangular distri-
bution, we want the probability distribution function to satisfy
equation (A.4), where u is from the rectangular distribution. We
want the probability of choosing yD to equal the fraction of the
lower-dimensional right-hypertriangle that constitutes the slicealongy D, as shown in equation(A.5). Combining these two equa-
tions gives an expression for du=dyDthat can be integrated to yield
equation (A.6). To have correspondence between the intervals
u 2 0; 1 and yD 2 0; bD, the integration constant must be c= 1.
Rearranging to expressyD in terms ofu gives equation(A.7).
PyDdyD 1du; A:4
PyDdyD 1
VDVD1 1
yDbD
D1dyD; A:5
u c 1 yDbD
D; A:6
yD bD1 1 u1=D: A:7
Once a value is selected for yD, it constrains the value for yD1because of inequality(4), and so forth for the subsequent y i. This
sequence of constraints is captured by the continued product in
equation (6). The bi may be sampled in any order; the ordering
of the indices in equation(6)is for convenience.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jct.2013.12.018.
References
[1] A. Karton, J.M.L. Martin, J. Chem. Phys. 136 (2012).[2] J.J. Hurly, M.R. Moldover, J. Res. Natl. Inst. Stand. Technol. 105 (2000) 667688.[3] K.K. Irikura, J. Phys. Chem. Ref. Data 36 (2007) 389397.
[4] K.K. Irikura, R.D. Johnson III, R.N. Kacker, R. Kessel, J. Chem. Phys. 130 (2009) 111.
[5] A.P. Scott, L. Radom, J. Phys. Chem. 100 (1996) 1650216513.[6] K.S. Pitzer, J. Chem. Phys. 14 (1946) 239243 .[7] J.J. Zheng, S.L. Mielke, K.L. Clarkson, D.G. Truhlar, Comput. Phys. Commun. 183
(2012) 18031812.[8] G. Reinisch, K. Miki, G.L. Vignoles, B.M. Wong, C.S. Simmons, J. Chem. Theor.
Comput. 8 (2012) 27132724.[9] R.D. Chirico, A.F. Kazakov, W.V. Steele, J. Chem. Thermodyn. 54 (2012) 278
287.
[10] R.L. Jacobsen, R.D. Johnson III, K.K. Irikura, R.N. Kacker, J. Chem. Theor. Comput.9 (2013) 951954.
[11] Certain commercial materials and equipment are identified in this paper inorder to specify procedures completely. In no case does such identificationimply recommendation or endorsement by the National Institute of Standardsand Technology, nor does it imply that the material or equipment identified isnecessarily the best available for the purpose.
[12] L.S. Kassel, Chem. Rev. 18 (1936) 277313 .[13] S.E. Stein, B.S. Rabinovitch, J. Chem. Phys. 58 (1973) 24382445 .[14] F.G. Wang, D.P. Landau, Phys. Rev. E 64 (2001).[15] M. Basire, P. Parneix, F. Calvo, J. Chem. Phys. 129 (2008).[16] T.L. Nguyen, J.R. Barker, J. Phys. Chem. A 114 (2010) 37183730.[17] J.R. Barker, J. Phys. Chem. 91 (1987) 38493854.[18] M.S. Schuurman, S.R. Muir, W.D. Allen, H.F. Schaefer III, J. Chem. Phys. 120
(2004) 1158611599.[19] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman,
G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato,X. Li, H.P. Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M.Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y.Honda, O. Kitao, H. Nakai, T. Vreven, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro,M. Bearpark, J.J. Heyd, E. Brothers, K.N. Kudin, V.N. Staroverov, T. Keith, R.Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar, J.Tomasi, M. Cossi, N. Rega, J.M. Millam, M. Klene, J.E. Knox, J.B. Cross, V. Bakken,C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R.Cammi, C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski,G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, O. Farkas, J.B.Foresman, J.V. Ortiz, J. Cioslowski, D.J. Fox, Gaussian 09, Gaussian Inc,Wallingford, CT, 2010.
[20] J.F. Stanton, J. Gauss, M.E. Harding, P.G. Szalay, A.A. Auer, R.J. Bartlett, U.Benedikt, C. Berger, D.E. Bernholdt, Y.J. Bomble, L. Cheng, O. Christiansen, M.Heckert, O. Heun, C. Huber, T.-C. Jagau, D. Johsson, J. Juslius, K. Klein, W.J.Lauderdale, D.A. Matthews, T. Metzroth, L.A. Mck, D.P. ONeill, D.R. Price, E.Prochnow, C. Puzzarini, K. Ruud, F. Schiffmann, W. Schwalbach, S. Stopkowicz,A. Tajti, J. Vzquez, F. Wang, J.D. Watts, J. Almlf, P.R. Taylor, T. Helgaker, H.J.A.
Jensen, P. Jrgensen, J. Olsen, A.V. Mitin, C. van Wllen, CFOUR: Coupled-Cluster techniques for Computational Chemistry, a quantum-chemicalprogram package, Johannes Gutenberg-Universitt, Mainz, 2010.
[21] K.K. Irikura, J. Phys. Chem. A 117 (2013) 12761282.[22] V. Barone, J. Chem. Phys. 122 (2005) 110 .[23] W.F. Polik, D.R. Guyer, C.B. Moore, J. Chem. Phys. 92 (1990) 34533470 .[24] D.F. DeTar, J. Phys. Chem. A 111 (2007) 44644477.[25] J.P. Merrick, D. Moran, L. Radom, J. Phys. Chem. A 111 (2007) 1168311700 .[26] M.L. Laury, M.J. Carlson, A.K. Wilson, J. Comput. Chem. 33 (2012) 23802387.[27] C. Cervinka, M. Fulem, K. Ruzicka, J. Chem. Eng. Data 57 (2012) 227232.[28] K.H. Lau, R.D. Brittain, D.L. Hildenbrand, J. Chem. Phys. 90 (1989) 11581164.[29] C.J. Egan, J.D. Kemp, J. Am. Chem. Soc. 59 (1937) 12641268.[30] J. Smukala, R. Span, W. Wagner, J. Phys. Chem. Ref. Data 29 (2000) 10531121 .[31] J. Chao, B.J. Zwolinski, J. Phys. Chem. Ref. Data 4 (1975) 251262 .[32] H. Senftleben, Z. Angew. Phys. 17 (1964) 8687.[33] W.F. Giauque, J. Gordon, J. Am. Chem. Soc. 71 (1949) 21762181 .[34] J.J. Hurly, Int. J. Thermophys. 23 (2002) 667696.[35] R.J.M. Konings, D.L. Hildenbrand, J. Alloy Compd. 271 (1998) 583586.[36] T. Shimanouchi, J. Phys. Chem. Ref. Data 6 (1977) 9931102.[37] T. Shimanouchi, Tables of Molecular Vibrational Frequencies, Consolidated
Volume I, US GPO, Washington, DC, 1972.[38] A. Zoccante, P. Seidler, M.B. Hansen, O. Christiansen, J. Chem. Phys. 136 (2012)
204118-1204118-12.
JCT 13-419
K.K. Irikura / J. Chem. Thermodynamics 73 (2014) 183189 189
http://dx.doi.org/10.1016/j.jct.2013.12.018http://refhub.elsevier.com/S0021-9614(13)00483-7/h0005http://refhub.elsevier.com/S0021-9614(13)00483-7/h0010http://refhub.elsevier.com/S0021-9614(13)00483-7/h0010http://refhub.elsevier.com/S0021-9614(13)00483-7/h0015http://refhub.elsevier.com/S0021-9614(13)00483-7/h0020http://refhub.elsevier.com/S0021-9614(13)00483-7/h0020http://refhub.elsevier.com/S0021-9614(13)00483-7/h0020http://refhub.elsevier.com/S0021-9614(13)00483-7/h0025http://refhub.elsevier.com/S0021-9614(13)00483-7/h0030http://refhub.elsevier.com/S0021-9614(13)00483-7/h0035http://refhub.elsevier.com/S0021-9614(13)00483-7/h0035http://refhub.elsevier.com/S0021-9614(13)00483-7/h0040http://refhub.elsevier.com/S0021-9614(13)00483-7/h0040http://refhub.elsevier.com/S0021-9614(13)00483-7/h0045http://refhub.elsevier.com/S0021-9614(13)00483-7/h0045http://refhub.elsevier.com/S0021-9614(13)00483-7/h0050http://refhub.elsevier.com/S0021-9614(13)00483-7/h0050http://refhub.elsevier.com/S0021-9614(13)00483-7/h0060http://refhub.elsevier.com/S0021-9614(13)00483-7/h0060http://refhub.elsevier.com/S0021-9614(13)00483-7/h0065http://refhub.elsevier.com/S0021-9614(13)00483-7/h0070http://refhub.elsevier.com/S0021-9614(13)00483-7/h0075http://refhub.elsevier.com/S0021-9614(13)00483-7/h0075http://refhub.elsevier.com/S0021-9614(13)00483-7/h0080http://refhub.elsevier.com/S0021-9614(13)00483-7/h0085http://refhub.elsevier.com/S0021-9614(13)00483-7/h0090http://refhub.elsevier.com/S0021-9614(13)00483-7/h0090http://refhub.elsevier.com/S0021-9614(13)00483-7/h0090http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0105http://refhub.elsevier.com/S0021-9614(13)00483-7/h0110http://refhub.elsevier.com/S0021-9614(13)00483-7/h0115http://refhub.elsevier.com/S0021-9614(13)00483-7/h0115http://refhub.elsevier.com/S0021-9614(13)00483-7/h0120http://refhub.elsevier.com/S0021-9614(13)00483-7/h0120http://refhub.elsevier.com/S0021-9614(13)00483-7/h0125http://refhub.elsevier.com/S0021-9614(13)00483-7/h0130http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0140http://refhub.elsevier.com/S0021-9614(13)00483-7/h0145http://refhub.elsevier.com/S0021-9614(13)00483-7/h0150http://refhub.elsevier.com/S0021-9614(13)00483-7/h0155http://refhub.elsevier.com/S0021-9614(13)00483-7/h0160http://refhub.elsevier.com/S0021-9614(13)00483-7/h0160http://refhub.elsevier.com/S0021-9614(13)00483-7/h0165http://refhub.elsevier.com/S0021-9614(13)00483-7/h0170http://refhub.elsevier.com/S0021-9614(13)00483-7/h0170http://refhub.elsevier.com/S0021-9614(13)00483-7/h0175http://refhub.elsevier.com/S0021-9614(13)00483-7/h0180http://refhub.elsevier.com/S0021-9614(13)00483-7/h0195http://refhub.elsevier.com/S0021-9614(13)00483-7/h0195http://refhub.elsevier.com/S0021-9614(13)00483-7/h0195http://refhub.elsevier.com/S0021-9614(13)00483-7/h0195http://refhub.elsevier.com/S0021-9614(13)00483-7/h0180http://refhub.elsevier.com/S0021-9614(13)00483-7/h0175http://refhub.elsevier.com/S0021-9614(13)00483-7/h0170http://refhub.elsevier.com/S0021-9614(13)00483-7/h0165http://refhub.elsevier.com/S0021-9614(13)00483-7/h0160http://refhub.elsevier.com/S0021-9614(13)00483-7/h0155http://refhub.elsevier.com/S0021-9614(13)00483-7/h0150http://refhub.elsevier.com/S0021-9614(13)00483-7/h0145http://refhub.elsevier.com/S0021-9614(13)00483-7/h0140http://refhub.elsevier.com/S0021-9614(13)00483-7/h0135http://refhub.elsevier.com/S0021-9614(13)00483-7/h0130http://refhub.elsevier.com/S0021-9614(13)00483-7/h0125http://refhub.elsevier.com/S0021-9614(13)00483-7/h0120http://refhub.elsevier.com/S0021-9614(13)00483-7/h0115http://refhub.elsevier.com/S0021-9614(13)00483-7/h0110http://refhub.elsevier.com/S0021-9614(13)00483-7/h0105http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0100http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0095http://refhub.elsevier.com/S0021-9614(13)00483-7/h0090http://refhub.elsevier.com/S0021-9614(13)00483-7/h0090http://refhub.elsevier.com/S0021-9614(13)00483-7/h0085http://refhub.elsevier.com/S0021-9614(13)00483-7/h0080http://refhub.elsevier.com/S0021-9614(13)00483-7/h0075http://refhub.elsevier.com/S0021-9614(13)00483-7/h0070http://refhub.elsevier.com/S0021-9614(13)00483-7/h0065http://refhub.elsevier.com/S0021-9614(13)00483-7/h0060http://refhub.elsevier.com/S0021-9614(13)00483-7/h0050http://refhub.elsevier.com/S0021-9614(13)00483-7/h0050http://refhub.elsevier.com/S0021-9614(13)00483-7/h0045http://refhub.elsevier.com/S0021-9614(13)00483-7/h0045http://refhub.elsevier.com/S0021-9614(13)00483-7/h0040http://refhub.elsevier.com/S0021-9614(13)00483-7/h0040http://refhub.elsevier.com/S0021-9614(13)00483-7/h0035http://refhub.elsevier.com/S0021-9614(13)00483-7/h0035http://refhub.elsevier.com/S0021-9614(13)00483-7/h0030http://refhub.elsevier.com/S0021-9614(13)00483-7/h0025http://refhub.elsevier.com/S0021-9614(13)00483-7/h0020http://refhub.elsevier.com/S0021-9614(13)00483-7/h0020http://refhub.elsevier.com/S0021-9614(13)00483-7/h0015http://refhub.elsevier.com/S0021-9614(13)00483-7/h0010http://refhub.elsevier.com/S0021-9614(13)00483-7/h0005http://-/?-http://dx.doi.org/10.1016/j.jct.2013.12.018http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-