This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
JOURNAL OF CHEMICAL PHYSICS VOLUME 118, NUMBER 21 1 JUNE 2003
Vibrational corrections to indirect nuclear spin–spin coupling constantscalculated by density-functional theory
Torgeir A. Ruden, Ola B. Lutnæs, and Trygve Helgakera)
Department of Chemistry, University of Oslo, Box 1033 Blindern, N-0315 Oslo, Norway
Kenneth RuudDepartment of Chemistry, University of Tromsø, N-9037 Tromsø, Norway
~Received 18 November 2002; accepted 5 March 2003!
The indirect nuclear spin–spin coupling constantsnuclear magnetic resonance~NMR! spectroscopy may nowadays be calculated by a variety of electronic-structmethods.1 Until recently, the most popular methods for sucalculations were multiconfigurational self-consistent fie~MCSCF! theory2–12 and the second-order polarizatiopropagator approach~SOPPA!,13–24although some work habeen carried out using coupled-cluster theory.25–29 Lately,density-functional theory~DFT! has become a popular toofor the calculation of spin–spin coupling constants. The fisuccessful implementations are those by Malkin, Malkiand Salahub from 199430 and by Dickson and Ziegler from1996.31 In 2000, Sychrovsky, Gra¨fenstein and Cremer32 andHelgaker, Watson and Handy33 independently presented fullanalytical spin–spin implementations at the hybrid leof DFT, demonstrating that hybrid theory represents a rable and inexpensive method for the calculation of suconstants.
The current status of the theory for the calculationspin–spin coupling constants is now such that the differebetween theory and experiment is often no larger thanvibrational corrections to the couplings, which may constute as much as 10% of the coupling.17,34Therefore, to makefurther progress towards the accurate description of indinuclear spin–spin coupling constants, it has become imtant to develop efficient methods for the calculation of vibtional corrections.
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
f
e
t,
li-h
fee
-
ctr--
The vibrational corrections to molecular properties cbe calculated in several ways. For polyatomic molecules,most common techniques are based on perturbatheory.35–40Although the details may vary, these methodsrequire the calculation of the geometrical derivatives ofmolecular property itself as well as of the potential-enersurface. Usually, no implementation exists for the analytievaluation of property derivatives, which are insteobtained numerically by, for example, finite-difference tecniques, making the calculation of vibrational correctioexpensive. Therefore, to calculate vibrational correctiofor systems containing 10–15 atoms, we must reducemuch as possible the cost of evaluating the molecular prerty at each geometry. This is particularly true for indirenuclear spin–spin coupling constants, whose evaluationgeneral is very expensive.
In view of the low cost and the high accuracy achievby hybrid DFT for the calculation of indirect spin–spin coplings constants, here we shall apply this theory to the cculation of the vibrational corrections to these constants. Pvided DFT yields good results compared to wave-functmethods for small molecules, it will represent a very usemethod for the calculation of vibrationally corrected indirespin–spin coupling constants in large molecules. Heretherefore first apply DFT to the calculation of vibrationcorrections to the nuclear spin–spin coupling constantssmall molecules, comparing these corrections with thoseviously obtained using wave-function methods. Next, weply DFT to the calculation of the vibrationally averaged idirect nuclear spin–spin couplings of benzene, a molectoo big to be treated accurately by non-DFT methods.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
ioingpihu
br
linalag
reon
t
plo–
e
-n
entln-ng tonsen-
andpin
rgy,
r
the
dtorD,ba-q.of
er-
soryln-
ra-ic-
ed-of
qui-
9573J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Vibrational corrections to coupling constants
II. THEORY AND IMPLEMENTATION
In this section, we discuss in some detail the calculatof vibrationally averaged indirect nuclear spin–spin couplconstants. After a review of Ramsey’s theory of spin–scoupling constants in Sec. II A, we describe in Sec. II B tcalculation of vibrational corrections to the spin–spin copling constants as implemented inDALTON.41
A. Ramsey’s theory
The indirect nuclear spin–spin coupling constants cancalculated as derivatives of the electronic energy. We firecall that the nuclear magnetic momentsMK are related tothe nuclear spinsIK as
MK5gK\IK , ~1!
where gK is the nuclear magnetogyric ratio of nucleusK.The normal and reduced indirect nuclear spin–spin coupconstantsJKL and KKL may then be calculated as the totderivatives of the energy with respect to the nuclear mnetic moments,
JKL5hgK
2p
g l
2pKKL5h
gK
2p
g l
2p
d2E
dMKdML. ~2!
In the Born–Oppenheimer approximation, Ramsey’s nonativistic expression for the reduced spin–spin coupling cstantsKKL of a closed-shell molecule is given by42
KKL5^0uhKLDSOu0&12(
sÞ0
^0uhKPSOus&^suhL
PSOTu0&E02Es
12(t
^0uhKFC1hK
SDut&^tuhKFCT1hK
SDTu0&E02Et
. ~3!
While the first summation is over all singlet statesus& differ-ent from the ground stateu0&, the second is over all triplestatesut&. The energiesE0 , Es , and Et are those of theground state, of the singlet excited states, and of the triexcited states, respectively. In atomic units, the operatorscurring in Eq. ~3! are, respectively, the diamagnetic spinorbit ~DSO! operator, the paramagnetic spin–orbit~PSO! op-erator, the Fermi-contact~FC! operator, and the spin–dipol~SD! operator:
hKLDSO5a4(
i
~r iKT r iL !I32r iKr iL
T
r iK3 r iL
3 , ~4!
hKPSO5a2(
i
r iK3pi
r iK3 , ~5!
hKFC5
8pa2
3 (i
d~r iK !si , ~6!
hKSD5a2(
i
3~siTr iK !r iK2r iK
2 si
r iK5 . ~7!
Here, a is the fine-structure constant,I3 is the three-dimensional unit matrix,r iL
T is the transpose of ther iL vector,and the summations are over the electrons.
Although Eq. ~3! clearly displays the different mechanisms that contribute to the total spin–spin coupling co
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
n
ne-
est
g
-
l--
etc-
-
stants in the conventional formalism of time-independperturbation theory, it is not useful for practicacalculations.33 Instead, the nuclear spin–spin coupling costants are evaluated as second-order properties accordiEq. ~2!, using the standard techniques of linear respotheory.43 In this approach, the closed-shell Kohn–Sham eergy is written asE(MK ,lS ,lT), wherelS andlT contain,respectively, the parameters that represent the singlettriplet variations of the ground state. The reduced spin–scoupling constants can then be calculated as
KKL5d2E
dMK dML5
]2E
]MK ]ML1
]2E
]MK ]lS
]lS
]ML
1]2E
]MK ]lT
]lT
]ML, ~8!
where all derivatives are evaluated for the optimized enefor which lS andlT are zero. The derivatives oflS andlT
with respect toMK are obtained by solving the first-orderesponse equations:
]2E
]lS ]lS
]lS
]ML52
]2E
]lS ]ML, ~9!
]2E
]lT ]lT
]lT
]ML52
]2E
]lT ]ML, ~10!
where the symmetric matrices on the left-hand sides aresinglet and triplet electronic Hessians, respectively.43 The so-lutions to Eqs.~9! and~10! represent the first-order perturbewave functions due to the imaginary singlet PSO operaEq. ~5! and due to the combined real triplet FC and Soperators, Eqs.~6! and ~7!, respectively. By spin symmetrythere is no coupling between the singlet and triplet perturtions. We finally note that the real singlet DSO operator, E~4!, enters the reduced coupling constant in the first termEq. ~8!, which represents an expectation value of the unpturbed reference state.
B. Vibrational corrections to molecular properties
The theory for the calculation of vibrational correctionto molecular properties by second-order perturbation theis well documented.35–40 Here we evaluate the vibrationacorrection to the indirect nuclear spin–spin coupling costants as the zero-point vibrational~ZPV! correction, usingthe approach of Kernet al.35–37
In this approach, the zeroth-order ground-state vibtional wave function is written as a product of harmonoscillator functions in normal coordinates:
X(0)~Q!5F0~Q!5 )K51
3N26
f0K~QK!, ~11!
wherefnK is then’ th excited harmonic-oscillator state of th
K ’ th vibrational normal mode. Next, the first-order grounstate vibrational wave function is expanded in the full setvirtual excitations fromX(0)(Q). Assuming a fourth-orderTaylor expansion of the potential energy-surface about e
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
x-ri
m
ly.
m
e
fo
to
ndthebic
alvessoare
theTheheisivevesible
lcu-st
pinine
allye-ntsin
ble
tect
ofhantheu-
r-ntskerke
eur
ngncegu-
ts.omical
9574 J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Ruden et al.
librium, the only contributions are from single and triple ecitations. The ground-state wave function may then be wten in the form35
X(1)~Q!5 (K51
3N26
@aK1 FK
1 ~Q!1aK3 FK
3 ~Q!#
1 (K,L51
3N26
bKL21 FKL
21 ~Q!
1 (K,L,M51
3N26
cKLM111 FKLM
111 ~Q!, ~12!
where FKLMklm (Q), for example, has been obtained fro
F0(Q) by exciting theK ’ th, L ’ th, and M ’ th modes to thek’ th, l ’ th, andm’ th harmonic-oscillator states, respectiveThe expansion coefficients in Eq.~12! may be calculatedfrom the cubic force constants,
FKLM5d3E
dQK dQL dQM, ~13!
and the harmonic frequenciesvK as follows:
aK1 52
1
4&vK3/2 (
L51
3N26FKLL
vL, ~14!
aK3 52
)
36vK5/2FKKK , ~15!
bKL21 52
1
4vKAvL
FKKL
2vK1vL, ~16!
cKLM111 52
1
12A2vKvLvM
FKLM
vK1vL1vM. ~17!
To determine the ZPV correction to the equilibriuvalue Peq of some molecular propertyP, we consider theexpectation value
^P&5^X(0)1X(1)uPuX(0)1X(1)&, ~18!
whereX(0) andX(1) are given by Eqs.~11! and~12!, respec-tively. ExpandingP in Eq. ~18! in a Taylor series about thequilibrium geometry,
P5Peq1 (K51
3N26dP
dQKQK1
1
2 (K,L51
3N26d2P
dQKdQLQKQL1¯,
~19!
and collecting terms, we obtain the following expressionthe expectation value:
^P&5Peq11
4 (K51
3N261
vK
d2P
dQK2 1&(
K
dPexp
dQK
aK1
AvK
1¯
5Peq11
4 (K51
3N261
vK
d2P
dQK2
21
4 (K51
3N261
vK2
dP
dQK(L51
3N26FKLL
vL1¯ . ~20!
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
t-
r
To second order in perturbation theory, the ZPV correctionthe property can then be written as
PZPV51
4 (K51
3N261
vK
d2P
dQK2 2
1
4 (K51
3N261
vK2
dP
dQK(L51
3N26FKLL
vL.
~21!
Thus, to calculate the ZPV correction, we need the first adiagonal second derivatives of the property, as well asharmonic frequencies and the semi-diagonal part of the cuforce field. As pointed out in the Introduction, no analyticimplementation exists for the evaluation of these derivatifor the indirect nuclear spin–spin coupling constants,some numerical procedure must be used instead. Thereseveral ways that derivatives can be found numerically.
One approach is to fit an analytic hypersurface toproperty and energy calculated at different geometries.derivatives can then be obtained by differentiation of tfitted surface.14–19,44–48A disadvantage of this approachthat it is difficult to automate and that it becomes expensfor large systems. Alternatively, the necessary derivatimay be calculated numerically, relying as much as posson available analytical derivatives.49 Unlike the fitting ap-proach, this approach is easily automated, making the calation of vibrational corrections straightforward, and at moequally expensive, even for polyatomic systems.
In this paper, we calculate the indirect nuclear spin–scoupling constants using the DFT implementationDALTON.33 Applying the technique described in Ref. 49, thproperty and energy derivatives are calculated numericfrom the highest available analytical derivatives. With rspect to geometrical derivatives, only molecular gradiehave been implemented analytically at the DFT level—particular, no analytical geometry derivatives are availafor the spin–spin coupling constants inDALTON.
Assuming that the number of normal modes is 3N26,we therefore need to carry out 6N211 property and gradiencalculations to determine the ZPV correction to each indirnuclear spin–spin coupling constant. Since the calculationspin–spin coupling constants is much more demanding tthe calculation of molecular gradients, the calculation ofZPV corrections will be completely dominated by the calclation of the property derivatives.
III. CALCULATIONS
In this section, we discuss the calculation of ZPV corections to the indirect nuclear spin–spin coupling constafor a number of small molecules. As advocated by Helgaet al., all calculations have been carried out with the Bec3-parameter Lee–Yang–Parr~B3LYP! functional.33
Having briefly introduced the basis sets in Sec. III A, wexamine in Sec. III B the force fields that are used in ocalculations of vibrationally averaged spin–spin coupliconstants. After an investigation of the basis-set dependeof the ZPV contribution to the indirect spin–spin couplinconstants in Sec. III C, we compare in Sec. III D the calclated ZPV corrections with previously published resulThese ZPV corrections are then in Sec. III E subtracted frexperimentally observed constants to yield a set of empir
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
9575J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Vibrational corrections to coupling constants
Downloaded 23 M
TABLE I. B3LYP and valence-electron CCSD~T! harmonic frequencies compared with experiment (cm21).
equilibrium spin–spin coupling constants, which are subquently used to benchmark the coupling constants calculby different theoretical methods. Finally, the vibrationaaveraged spin–spin coupling constants of benzene arecussed in Sec. III F.
A. Basis sets
The ZPV corrections have been calculated using tsequences of basis sets. The first sequence consists oHuzinaga sets HII, HIII, and HIV50,51 with the polarizationfunctions and contraction patterns of van Wu¨llen and Kut-zelnigg et al.52 These basis sets have been widely usedthe calculation of nuclear shielding constants and indirspin–spin coupling constants.
However, for an accurate calculation of the FC contribtion to the spin–spin coupling constants, it is essential tobasis sets with a flexible inner core.2,20,33To ensure a flexiblecore description, we have used the basis sets HII-su2, Hsu3, and HIV-su4. The postfix ‘‘-sun’’ indicates that thesfunctions in the original basis have been decontracted,that an additional set ofn tight s functions have been addein an even-tempered manner.33 For brevity of notation, weshall here abandon the general notation HX-sun and insteadrefer to these basis sets as sHII, sHIII, and sHIV, resptively. The performance of the different basis sets is exained in Sec. III C.
B. Quality of the B3LYP force field
For an accurate description of vibrational correctionsis necessary to ensure that the quadratic and cubic ffields are calculated to sufficient accuracy. Several studie
ay 2003 to 129.240.80.34. Redistribution subject to A
-ed
is-
othe
rt
-e
I-
nd
c--
itceof
DFT harmonic and anharmonic force fields have shown thin a sufficiently large basis, B3LYP provides a good descrtion of harmonic and anharmonic force fields.53–55In particu-lar, Martin et al. found that, for 13 small molecules, thB3LYP harmonic frequencies have a mean absolute erroonly 30 cm21 relative to experimental harmonifrequencies.54
In Table I, we have listed the B3LYP harmonic frequecies for all molecules included in this study except for etheand benzene, calculated using the same basis sets assubsequent spin–spin calculations. For comparison, we hincluded experimental harmonic frequencies as well asharmonic vibrational frequencies of Martinet al.,54 obtainedusing the valence-correlated coupled-cluster singles-adoubles~CCSD! method with a perturbative triples correction @CCSD~T!#. In their study, Martinet al. found that, rela-tive to experiment, the mean absolute error of the CCSD~T!frequencies are 8 cm21 for the 13 molecules.
Clearly, in the Huzinaga-type basis sets, the DFT/B3Lmodel provides a good representation of the harmonic fofield, with mean absolute errors relative to experimentabout 30 cm21. The B3LYP model also compares favorabwith the more expensive CCSD~T!/cc-pVQZ model, whosemean absolute errors are 15 cm21 relative to experiment.
Also, the cubic force field is important for the calculation of ZPV corrections to properties. To examine the quaof the cubic force field, we here compare the calculated Zcorrection to the molecular geometry with available theorical data. To second order in the perturbation, the ZPV crection to the geometry can be calculated using the followformula:35,38
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
oinietrly
heicnp-hth
on-IV
s
nn
HIIla-u-rip-re
bes by
theu-othve
g
r
t
9576 J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Ruden et al.
QKZPV52
1
4vK2 (
L51
NFKLL
vL. ~22!
Since this expression resembles the term in Eq.~21! thatcontains the cubic force constants, it should give a goindication of the error arising from the cubic force fieldthe calculated ZPV corrections to other molecular propertAs seen from Table II, the ZPV corrections to the geomecalculated at the B3LYP level agree well with previouscalculated MCSCF corrections.
C. Basis-set dependence of the ZPV contributionto indirect nuclear spin–spin coupling constants
As seen from Table III, the vibrational corrections to tindirect nuclear spin–spin coupling constants depend notably on the basis set—both when the valence descriptioimproved from HII to HIV and when the inner-core descrition is improved from, say, HII to sHII. However, althougthe couplings change by 5% to 10% in both cases,
TABLE II. B3LYP ZPV corrections to bond distances~pm! and bond angles(°) compared with MCSCF corrections.
B3LYP MCSCF
HII sHII sHIII sHIV
HF r 1.6 1.6 1.5 1.6 1.5a
H2O r 1.5 1.5 1.5 1.4 1.5b
u 20.1 20.1 20.1 20.0 20.1b
H2CO r CO 0.3 0.3 0.3 0.3 0.4b
r CH 1.4 1.4 1.4 1.4 1.4b
uHCH 20.1 20.2 20.1 20.1 20.1b
C2HD r CC 0.4 0.4 0.4 0.4 0.5c
r CD 20.1 20.1 20.2 20.1 20.2c
r CH 20.4 20.4 20.5 20.5 20.5c
uDCC 0.0 0.0 0.0 0.0 0.0c
uCCH 0.0 0.0 0.0 0.0 0.0c
aReference 5.bReference 38.cReference 63.
TABLE III. ZPV corrections to the indirect nuclear spin–spin couplinconstants calculated at the B3LYP level of theory~Hz!.
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
d
s.y
e-is
e
changes are in opposite directions. As a result, the HII cstants are usually closer to the sHIV results than to the Hresults. An exception is1JCH in HCN, where the changeupon the addition of valence and inner-cores orbitals are inthe same direction, giving an sHIV vibrational correctio~5.1 Hz! that is about one third larger than the HII correctio~4.0 Hz!—in all other cases, the differences between theand sHIV corrections are less than 5%. Clearly, in calcutions of ZPV corrections to indirect nuclear spin–spin copling constants, we should not improve the valence desction without simultaneously improving the inner-codescription.
In spite of its good performance, the HII basis shouldused with some care as it sometimes gives good resulterror cancellation. For1JHD , for example, the HII and sHIVvibrational corrections are similar. However, whereassHIV correction is dominated by the anharmonic contribtion, the harmonic and anharmonic contributions are blarge in the HII basis—see Tables IV and V, where we ha
TABLE IV. B3LYP harmonic vibrational contribution to the indirect nucleaspin–spin coupling constants~Hz!.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
ionomnsa
eino
seeathns
io-
ee9%
istr
allled-bu
aththism
inb
s.
-nt
se
it.-
ling
eri-rtedi-nal
bles.
heof
chys-g
arreti-gPV
u n-
g it
9577J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Vibrational corrections to coupling constants
listed separately the harmonic and anharmonic contributto the ZPV corrections, respectively. Clearly, as we go frHII to sHIV, the harmonic and anharmonic contributiochange in opposite directions, leading to an overall smchange in the total vibrational correction.
It is noteworthy that, as we go from sHIII to sHIV, thvibrational corrections change very little—in fact, onlythree cases does the vibrational correction change by mthan 0.1 Hz. This observation indicates that, in most cathe sHIV basis gives vibrational corrections to the nuclspin–spin coupling constants that are within 0.1 Hz ofbasis-set limit of DFT, and that the vibrational correctioobtained with the sHIII basis are also good.
As expected, the change in the vibrational correctupon the addition of tights functions is caused almost entirely by the FC contribution. Indeed, from Table VI, we sthat the FC contribution usually accounts for more than 9of the change in the vibrational correction~in all cases morethan 95%!. Since the calculation of the FC contributionmuch cheaper than the calculation of the remaining conbutions and since the force-field calculation is essentifree, we suggest the following approach for large molecufor the FC contribution, we use sHII, sHIII or sHIV, depening on molecule size; for the SD, PSO, and DSO contritions, we use HII or HIII.
In conclusion, we recommend the sHIV basis for smsystems since it gives vibrational corrections close toDFT basis-set limit. However, very good estimates ofvibrational corrections are obtained also with the sHIII baswhich we advocate for larger systems. For large systesuch as benzene, accurate vibrational corrections to therect nuclear spin–spin coupling constants are obtainedusing sHIII for the FC term and HII for the remaining term
D. Comparison with previously calculated vibrationalcorrections
As seen from Table VII, the B3LYP vibrational corrections to the indirect nuclear spin–spin coupling constaagree well with previous calculations.5,13–19,34,45However,there are two cases of striking differences—the1JNN cou-pling in N2 and the3JHH in C2H2 . In both cases, the DFTvibrational correction does not change with the basis
TABLE VI. Changes in the vibrational corrections to the spin–spin coplings going from the HX basis to the sHX basis at the DFT/B3LYP level oftheory ~Hz!.
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
s
ll
res,re
n
i-ys:
-
llee,s
di-y
s
t,
indicating that the correction is close to the basis-set limWe also note that, for N2 , the calculated SOPPA value constitutes as much as one fourth of the total spin–spin coupconstant. For3JHH in C2H2 , the difference is even larger—infact, the SOPPA~CCSD! correction is an order of magnitudlarger than the B3LYP correction. As the individual contbutions to the vibrational corrections have not been repofor C2H2 in Ref. 17, a comparison of the individual contrbutions is not possible but we note that the other vibratiocorrections to the spin–spin coupling constants in C2H2
agree well with the SOPPA~CCSD! values.For the remaining spin–spin coupling constants in Ta
VII, the DFT corrections are similar to the literature valueThe largest discrepancies occur for H2O, where1JOH differsfrom SOPPA by 24% and from MCSCF by 20%, and for tHF molecule, where the B3LYP vibrational correction238 Hz is bracketed by the MCSCF correction of227 Hzand the experimental correction240 Hz. Although theB3LYP result for HF is close to experiment, we do not attamuch significance to this result since, for this particular stem, B3LYP predicts a much too low equilibrium couplinconstant.
E. Experimental equilibrium values
Once the vibrational corrections to the indirect nuclespin–spin coupling constants have been calculated theocally, we can extract a set of empirical equilibrium couplinconstants from experiment by subtracting the calculated Zcorrections from the experimentally observed couplings:
Jeqemp5Jtot
exp2Jvibcal . ~23!
- TABLE VII. ZPV corrections to indirect nuclear spin–spin coupling costants~Hz!.
B3LYP Other calculations
sHII sHIII sHIV
HD 1JHD 2.7 2.8 2.8 1.8,a 2.0b
HF 1JFH 241.9 238.2 237.7 226.9,c 240d
CO 1JCO 0.7 0.7 0.7 0.8e
N21JNN 0.1 0.1 0.1 0.4e
H2O 1JOH 6.0 5.4 5.2 4.0,f 4.2g
2JHH 0.8 0.9 0.9 0.7,f 0.8g
CH41JCH 5.2 5.3 5.3 5.0,h 4.4i
2JHH 20.6 20.7 20.6 20.7,h 20.6i
C2H21JCC 29.1 210.0 29.3 29.2j
1JCH 5.0 4.6 4.7 4.8j2JCH 22.7 23.0 22.8 23.2j
3JHH 20.0 20.1 20.1 21.2j
aReference 23.bReference 9.cReference 5.dReference 34.eReference 13.fReference 14.gCalculated using the rovibrational numbers from Ref. 10, and correctinwith the temperature dependent part from Ref. 14.
hReference 15.iReference 45.jReference 17.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
III
9578 J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Ruden et al.
TABLE VIII. Calculated and experimental indirect nuclear spin–spin coupling constants~Hz!. The ZPV correction has been calculated at the B3LYP/sHlevel and the empirical coupling constants have been obtained using Eq.~23!.
Such empirical equilibrium coupling constants are listedTable VIII, together with the equilibrium coupling constancalculated by different theoretical methods. The empiriequilibrium values have been obtained by subtractingB3LYP/sHIII vibrational corrections from the experimentvalues listed in the table.
In Table IX, we have made a statistical analysis of terrors of the different theoretical methods relative to theperimental total spin–spin coupling constants and to thepirical equilibrium constants. Somewhat surprisingly, tmean absolute relative error increases for all methods ex
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
le
e--
pt
RAS after the vibrational contributions to the coupling costants have been accounted for. The relative error foundRAS decreases slightly from 11% to 10%.
By contrast, the mean absolute errors and standardviations decrease for all methods except DFT/B3LYP. Treduction in the error is particularly pronounced for tMCSCF model—from 5.8 to 3.3 Hz for the complete activspace self-consistent field~CASSCF! method and from 4.3 to1.6 Hz for the restricted active-space self-consistent fi~RASSCF! method. For SOPPA, CCSD, and CC3, the me
ental
.1.3
7
TABLE IX. Statistics of calculated indirect nuclear spin–spin coupling constants relative to the experimtotal coupling constantsJtot
exp and the empirical equilibrium coupling constantsJeqemp of Table VIII.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
3.
heonnne
teellansbe
tthacoo
fe.n–Cffile
meln
arfotod
ti-haheis
n-v
fo
initn
vempfeoo
wec
on
isin-ib-P/
ndin
t of
–thendofbe-thewithc-
toctedtlynt.h a
mte.ase
peri-rab-er
aleeno aare
Eq.
Ref.
9579J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Vibrational corrections to coupling constants
absolute error is reduced from 4.1 to 3.4 Hz, from 3.8 toHz, and from 6.4 to 6.1 Hz, respectively.
Among the different theoretical methods in Table IX, tRASSCF results are closest to the empirical equilibrium cstants with a mean absolute error of 1.6 Hz, the correspoing errors for the CASSCF, SOPPA, CCSD, CC3, aB3LYP methods being 3.3, 3.4, 3.7, 6.1, and 11.7 Hz, resptively.
The relatively poor performance of the coupled-clusmethods is somewhat surprising but arises from the rtively small basis sets used in the calculations—in particuno tight s-functions have been used in the CC3 calculatioClearly, the statistical errors in Table IX cannot directlyused as measures of the intrinsic errors associated withdifferent methods. Indeed, the good performance ofRASSCF method is to some extent a reflection of the fthat, for most of the molecules in our sample, this methhas been applied with great care so as to arrive at the maccurate possible coupling constants, although, for amolecules such as N2, there is still room for improvementAs a very recent investigation of the indirect nuclear spispin coupling constant in BH has shown, the CCSD and Cmethods are capable of very high accuracy—provided suciently large basis sets are used and provided that all etrons ~not just the valence electrons! are correlated in thecalculations.56
The large mean absolute error of DFT in Table IX copared to the wave-function methods is striking. As is wdocumented, the performance of the B3LYP method depecritically on the nature of the coupled nuclei. In particulpoor indirect nuclear spin–spin couplings are obtainedelectronegative atoms such as fluorine, whereas other asuch as hydrogen and carbon are quite well describe33
Thus, for HF, the B3LYP method in Table VIII underesmates the indirect nuclear spin–spin coupling by more t100 Hz. If this molecule is omitted from the statistics, tmean absolute error of B3LYP is reduced to 4 Hz—thatsimilar to the error of the wave-function methods.
Focusing on the mean absoluterelative errors in TableIX, we find that for all methods except RAS, the errorsin-creasewhen we compare with the empirical equilibrium costants instead of the observed total constants. This behais different from that of the mean absolute error, which,all methods except B3LYP, becomessmallerwhen we com-pare with the empirical equilibrium constants, suggestthat the vibrational corrections improve the agreement wexperiment, mostly for the large spin–spin coupling costants. One possible explanation for this behavior are soleffects, since many of the experiments have been perforin solution. In general, however, solvent effects on spin–scoupling constants are rather small, rarely exceeding aHz,57–60 suggesting that the error mostly arises from a pdescription of the electronic system.
F. Experimental equilibrium values for benzene
To illustrate the usefulness of the presented method,have calculated the vibrational corrections to the indirnuclear spin–spin coupling constants of the benzene mecule. In Table X, we have listed the vibrational correctio
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
7
-d-dc-
ra-r,.
heet
dst
w
3-c-
-lds,rms.
n
,
iorr
gh-ntedinwr
etl-
s
calculated with the B3LYP functional, using the sHIII basset for the FC contribution and the HII basis for the remaing contributions. In addition, we have included the equilrium spin–spin coupling constants calculated at the B3LYsHIII by us and at the MCSCF level by Kaski, Vaara, aJokisaari.61 From the experimental indirect nuclear spin–spcoupling constants of Ref. 61, we have obtained a seempirical equilibrium constants by applying Eq.~23!.
The vibrational corrections to the indirect nuclear spinspin coupling constants in benzene are small. In fact,only vibrational correction greater than 1 Hz is the one-boCH correction of 4.8 Hz. Next, we note that inclusionvibrational corrections does not improve the agreementtween theory and experiment. Indeed, only for three often coupling constants in benzene does the agreementexperiment improve with the inclusion of vibrational corretions.
Considering the quality of the vibrational correctionsspin–spin coupling constants, the reason for this unexpebehavior is either that the calculations are not sufficienaccurate or effects of the liquid crystal used in experimeSince the results of Ref. 61 are in good agreement witdetailed liquid-phase investigation by Laatikainenet al.,62
this indicates that the single-point calculations of equilibriuspin–spin coupling constants are not sufficiently accuraThis is also supported by the recently calculated gas-phequilibrium value of1JHC5152.7 Hz.57
Nevertheless, the agreement between theory and exment is much better for B3LYP than for MCSCF, which fothis molecule produces rather poor couplings. The meansolute error is 2 Hz for B3LYP and about four times largfor MCSCF.
G. Conclusions
An automated method for the calculation of vibrationcorrections to indirect nuclear spin–spin couplings has bpresented and applied at the DFT/B3LYP level of theory tnumber of small molecular systems. Our results comp
TABLE X. Indirect nuclear spin–spin coupling constants of benzene~Hz!.
asHIII basis.bSee Reference 61.cObtained by combining the entries in columns 5 and 6 according to~23!.
dHII basis except sHIII for FC.eSee Reference 61 except for the HH couplings. For HH couplings, see62.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
eubedno
heab
tam
enalothoothrriPTr-utivvoca
e
se
ys
d
.
.
tt.
ed
ed
S.
.
gn.
.
01,
.
. C.
r, J.
em.
, J.
9580 J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Ruden et al.
favorably with previous computed and experimentally detmined vibrational corrections to the indirect spin–spin copling constants, the computational cost at the DFT leveling significantly smaller. To illustrate potential anusefulness of the method, we have calculated the vibratiocorrections to the indirect spin–spin coupling constantsbenzene.
Having calculated a set of vibrational corrections to tindirect spin–spin coupling constants, a list of empiricequilibrium spin–spin coupling constants was generatedsubtracting the vibrational correction from the experimencoupling constant. Comparing these empirical equilibriucoupling constants with calculations carried out at differlevels of theory in the literature, we found that, for smmolecular systems, the best indirect spin–spin coupling cstants available in literature are those obtained withRASSCF method. It should be noted, however, that the gRASSCF performance is to some extent due to the factthis method has been applied with great care, so as to aat the most accurate possible coupling constants. The SOand DFT/B3LYP methods perform similarly, although DFfails badly for molecules containing fluorine. The perfomance of coupled-cluster theory is difficult to establish dto basis-set deficiencies. In short, to establish the relaperformance of the different theoretical methods unequically, a more consistent set of calculations needs to beried out for all methods.
ACKNOWLEDGMENTS
The work has received support from the Norwegian Rsearch Council~Program for Supercomputing! through agrant of computer time.
1T. Helgaker, M. Jaszun´ski, and K. Ruud, Chem. Rev.99, 293 ~1999!.2T. Helgaker, M. Jaszun´ski, K. Ruud, and A. Go´rska, Theor. Chem. Acc.99, 175 ~1998!.
3O. Vahtras, H. A˚ gren, P. Jørgensen, T. Helgaker, and H. J. Aa. JenChem. Phys. Lett.209, 201 ~1993!.
4M. Pecul and J. Sadlej, Chem. Phys.234, 111 ~1998!.5P-O. Astrand, K. Ruud, K. V. Mikkelsen, and T. Helgaker, J. Chem. Ph110, 9463~1999!.
6P. Lantto and J. Vaara, J. Chem. Phys.114, 5482~2001!.7J. Guilleme and J. San Fabia´n, J. Chem. Phys.109, 8168~1998!.8M. Jaszun´ski and K. Ruud, Chem. Phys. Lett.336, 473 ~2001!.9O. Vahtras, H. A˚ gren, P. Jørgensen, H. J. A. Jensen, S. B. Padkjaer, anHelgaker, J. Chem. Phys.96, 6120~1992!.
10J. Casanueva, J. San Fabia´n, E. Diez, and A. L. Esteban, J. Mol. Struct565–566, 449 ~2001!.
11J. San Fabia´n, J. Casanueva, E. San Fabia´n, and J. Guillerne, J. ChemPhys.112, 4143~2000!.
12J. Casanueva, J. San Fabia´n, E. Diez, and A. L. Esteban, Chem. Phys. Le361, 159 ~2002!.
13J. Geertsen, J. Oddershede, and G. E. Scuseria, J. Chem. Phys.87, 2138~1987!.
14R. D. Wigglesworth, W. T. Raynes, S. P. A. Sauer, and J. OddershMol. Phys.94, 851 ~1998!.
15R. D. Wigglesworth, W. T. Raynes, S. P. A. Sauer, and J. OddershMol. Phys.92, 77 ~1997!.
16S. P. A. Sauer, W. T. Raynes, and R. A. Nicholls, J. Chem. Phys.115, 5994~2001!.
17R. D. Wigglesworth, W. T. Raynes, S. Kirpekar, J. Oddershede, andA. Sauer, J. Chem. Phys.112, 3735~2000!; ibid. 114, 9192~2001! E.
18B. Bennett, W. T. Raynes, and C. W. Anderson, Spectrochim. Acta A45,821 ~1989!.
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
r---
alf
lyl
tln-edatvePA
ee-r-
-
n,
.
T.
e,
e,
P.
19S. P. A. Sauer, C. K. Møller, H. Koch, I. Paidarova´, and V. Spirko, Chem.Phys.238, 385 ~1998!.
20T. Enevoldsen, J. Oddershede, and S. P. A. Sauer, Theor. Chem. Acc100,275 ~1998!.
21S. P. A. Sauer, I. Paidarova, and V. Spirko~unpublished!.22P. F. Provasi, G. A. Aucar, and S. P. A. Sauer, J. Chem. Phys.115, 1324
~2001!.23J. Oddershede, J. Geertsen, and G. E. Scuseria, J. Phys. Chem.92, 3056
~1988!.24L. B. Krivdin, S. P. A. Sauer, J. E. Peralta, and R. H. Contreras, Ma
Reson. Chem.40, 187 ~2002!.25A. A. Auer and J. Gauss, J. Chem. Phys.115, 1619~2001!.26S. A. Perera, H. Sekino, and R. J. Bartlett, J. Chem. Phys.101, 2186
~1994!.27S. A. Perera, M. Nooijen, and R. J. Bartlett, J. Chem. Phys.104, 3290
~1996!.28H. Sekino and R. J. Bartlett, Chem. Phys. Lett.225, 486 ~1994!.29A. Wu, D. Cremer, A. A. Auer, and J. Gauss, J. Phys. Chem. A106, 657
~2002!.30V. G. Malkin, O. L. Malkina, and D. R. Salahub, Chem. Phys. Lett.221,
91 ~1994!.31R. M. Dickson and T. Ziegler, J. Phys. Chem.100, 5286~1996!.32V. Sychrvosky, J. Gra¨fenstein, and D. Cremer, J. Chem. Phys.113, 3530
~2000!.33T. Helgaker, M. Watson, and N. C. Handy, J. Chem. Phys.113, 9402
~2000!.34S. M. Bass, R. L. DeLeon, and J. S. Muenter, J. Chem. Phys.86, 4305
~1987!; J. S. Muenter and W. Klemperer,ibid. 52, 6033~1970!.35C. W. Kern and R. L. Matcha, J. Chem. Phys.49, 2081~1968!.36W. C. Ermler and C. W. Kern, J. Chem. Phys.55, 4851~1971!.37B. J. Krohn, W. C. Ermler, and C. W. Kern, J. Chem. Phys.60, 22 ~1974!.38P-O. Astrand, K. Ruud, and P. Taylor, J. Chem. Phys.112, 2655~2000!.39M. Toyama, T. Oka, and Y. Morino, J. Mol. Spectrosc.13, 193 ~1964!.40W. T. Raynes, J. Geertsen, and J. Oddershede, Int. J. Quantum Chem52,
written by T. Helgaker, H. J. Aa. Jensen, P. Jørgensen,et al.42N. F. Ramsey, Phys. Rev.91, 303 ~1953!.43J. Olsen and P. Jørgensen, J. Chem. Phys.82, 3235~1985!.44T. A. Ruden, W. Klopper, and M. Quack~unpublished!.45B. Bennett, W. T. Raynes, and W. Anderson, Spectrochim. Acta, Part A45,
821 ~1989!.46A. J. Russel and M. A. Spackman, Mol. Phys.84, 1239~1995!.47A. J. Russel and M. A. Spackman, Mol. Phys.88, 1109~1996!.48A. J. Russel and M. A. Spackman, Mol. Phys.90, 251 ~1997!.49T. A. Ruden, P. R. Taylor, and T. Helgaker, J. Chem. Phys.~to be pub-
lished!.50S. Huzinaga,Approximate Atomic Functions~University of Alberta, Edm-
onton, 1971!.51S. Huzinaga, J. Chem. Phys.42, 1293~1965!.52W. Kutzelnigg, U. Fleischer, and M. Schindler,NMR—Basic Principles
and Progress~Springer Verlag, Berlin, 1990!, Vol. 23, p. 165; C. vanWullen, Ph.D. thesis, Ruhr-Universita¨t, Bochum, 1992.
53A. Miani, E. Cane´, P. Palmieri, A. Trombetti, and N. C. Handy, J. ChemPhys.112, 248 ~2000!.
54J. Martin, J. El-Yazal, and J. P. Franc¸ois, Mol. Phys.86, 1437~1995!.55M. Rosenstock, P. Rosmus, E. A. Reinsch, O. Treutler, S. Carter, and N
Handy, Mol. Phys.93, 853 ~1998!.56A. A. Auer, J. Gauss, and M. Pecul, Chem. Phys. Lett.368, 172 ~2003!.57K. Ruud, L. Frediani, R. Cammi, and B. Mennucci, Int. J. Mol. Sci.~to be
published!.58I. Ando and G. A. Webb, Org. Magn. Reson.15, 111 ~1981!.59K. V. Mikkelsen, K. Ruud, and T. Helgaker, J. Comput. Chem.20, 1281
~1999!.60P.-O. Astrand, K. V. Mikkelsen, P. Jørgensen, K. Ruud, and T. Helgake
Chem. Phys.108, 2528~1998!.61J. Kaski, J. Vaara, and J. Jokisari, J. Am. Chem. Soc.118, 8879~1996!.62R. Laatikainen, J. Ratilainen, R. Sebastian, and H. Santa, J. Am. Ch
Soc.117, 11006~1995!.63K. Ruud, unpublished material.64R. E. Wasylishen, J. O. Friedrich, S. Mooibroek, and J. M. MacDonald
Chem. Phys.83, 548 ~1985!.65J. O. Friedrich and R. E. Wasylishen, J. Chem. Phys.83, 3707~1985!.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
es
9581J. Chem. Phys., Vol. 118, No. 21, 1 June 2003 Vibrational corrections to coupling constants
66N. M. Sergeyev, N. D. Sergeyeva, Y. A. Strelenko, and W. T. RaynChem. Phys. Lett.277, 144 ~1997!.
67K. J. Friesen and R. E. Wasylishen, J. Magn. Reson.41, 189 ~1980!.68G. Dombi, P. Diehl, J. Lounila, and R. Wasser, Org. Magn. Reson.22, 573
~1984!.
Downloaded 23 May 2003 to 129.240.80.34. Redistribution subject to A
,69J. O. Friedrich and R. E. Wasylishen, Can. J. Chem.65, 2238~1987!.70K. Jackowski, M. Wilczek, M. Pecul, and J. Sadlej, J. Phys. Chem. A104,
9806 ~2000!.71J. Kaski, P. Lantto, J. Vaara, and J. Jokisaari, J. Am. Chem. Soc.120, 3993
~1998!.
IP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp