Documente Academic
Documente Profesional
Documente Cultură
four-component theory
Miroslav Ilia, Trond Saue, Thomas Enevoldsen, and Hans Jrgen Aa. Jensen
Citation: The Journal of Chemical Physics 131, 124119 (2009); doi: 10.1063/1.3240198
View online: http://dx.doi.org/10.1063/1.3240198
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/131/12?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Relativistic calculation of nuclear magnetic shielding tensor using the regular approximation to the normalized
elimination of the small component. III. Introduction of gauge-including atomic orbitals and a finite-size nuclear
model
J. Chem. Phys. 129, 224103 (2008); 10.1063/1.3028047
A fully relativistic method for calculation of nuclear magnetic shielding tensors with a restricted magnetically
balanced basis in the framework of the matrix DiracKohnSham equationa)
J. Chem. Phys. 128, 104101 (2008); 10.1063/1.2837472
Relativistic effects on nuclear magnetic shielding constants in H X and CH 3 X(X= Br,I ) based on the linear
response within the elimination of small component approach
J. Chem. Phys. 121, 6798 (2004); 10.1063/1.1787495
Calculation of nuclear magnetic shieldings. XIII. Gauge-origin independent relativistic effects
J. Chem. Phys. 110, 131 (1999); 10.1063/1.478089
Calculation of nuclear magnetic shieldings. XII. Relativistic no-pair equation
J. Chem. Phys. 108, 3854 (1998); 10.1063/1.475788
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
Received 24 July 2009; accepted 9 September 2009; published online 29 September 2009
The use of perturbation-dependent London atomic orbitals, also called gauge including atomic
orbitals, has proven efficient for calculations of NMR shielding constants and other magnetic
properties in the nonrelativistic framework. In this paper, the theory of London atomic orbitals for
NMR shieldings is extended to the four-component relativistic framework and our implementation
is described. The relevance of London atomic orbitals in four-component calculations as well as
computational aspects are illustrated with test calculations on hydrogen iodide. We find that the use
of London atomic orbitals is an efficient method for reliable calculations of NMR shielding
constants with standard basis sets, also for four-component calculations with spin-orbit coupling
effects included in the wave function optimization. Furthermore, we find that it is important that the
small component basis functions fulfill the magnetic balance for accurate description of the
diamagnetic shielding and that the role of London atomic orbitals in the relativistic domain is to
provide atomic magnetic balance even in the molecular case, thus greatly improving basis set
convergence. The Sternheim approximation, which calculates the diamagnetic contribution as an
expectation value, leads to significant errors and is not recommended. 2009 American Institute of
Physics. doi:10.1063/1.3240198
I. INTRODUCTION
erties for recent reviews see Refs. 26 and 27. At the twocomponent relativistic level, thus allowing the inclusion of
spin-orbit interaction, schemes for the calculation of magnetic shielding constants in the common gauge origin CGO
have been presented based on the second-order
the
zeroth-order
regular
DouglasKrollHess,2833
approximation,34 the BreitPauli,3543 the normalized elimination of the small component NESC,4447 the
and
the
infinite-order
BaryszSadlejSnijders,33
two-component48 Hamiltonians. Relativistic two-component
methods for calculating NMR shielding tensor with LAOs
included were presented by Fukuda et al.49 and more recently by Hamaya et al.50 and by Autschbach.51 The latter is
based on the regular approximation to the NESC.
At the four-component level, Nakatsuji and co-workers
reported shielding constants calculated at the unrestricted
relativistic HartreeFock HF level with the external magnetic field as a finite perturbation.5254 This approach was
later extended to the four-component relativistic coupledcluster singles-and-doubles CCSD and configurationinteraction singles-and-doubles CISD level by adding the
nuclear magnetic moment as a second finite perturbation.55
The first implementation of four-component LAOs, albeit incomplete to be discussed below, is due to Quiney and
co-workers.5658 They published NMR shielding constants of
the water molecule at the uncoupled HF level. The first full
four-component coupled HF calculations of NMR shielding
tensors of hydrogen halides were reported by Visscher et
131, 124119-1
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-2
Ilia et al.
al.59 These NMR shielding constants were, however, obtained using the computationally simpler CGO method. The
same holds true for the more recent computational approaches in the four-component relativistic DFT framework
proposed by Liu and co-workers6062 and Komorovsk
et al.63
In this paper, we present the theory and the implementation of LAOs for NMR shielding constant calculations based
on the four-component DiracCoulomb Hamiltonian. It is an
extension of the four-component linear response method previously reported by two of us H.J.Aa.J and T.S.,64 which
utilizes quaternion algebra for efficient symmetry handling.
The first application of our implementation of the theory
presented here was to the xenon dimer and has been reported
in Ref. 65. In Secs. II and III, we present the theory and
details of the computational implementation. In Sec. IV we
report sample calculations of NMR shielding constants for
hydrogen iodide before concluding in Sec. V. Unless otherwise stated, we use in the following SI-based atomic units.
relativistic. It can be argued, however, that all magnetic interactions vanish in the strict nonrelativistic limit. In practice, that is, in conventional nonrelativistic calculations,
one allows the introduction of external magnetic fields B in
an ad hoc manner without any reference to the generating
mechanism, that is, magnetic induction, which is a truly relativistic effect. This allows the successful calculation of magnetic properties within a nonrelativistic framework. A more
complete discussion of this point is found in Ref. 68.
With the vector potentials given above, the Dirac Hamiltonian takes the form
hDB,mK = hD + B hB + mK hmK ,
II. THEORY
1 rK c
hmK = 2
.
c
rK3
A. Hamiltonian
The point of departure is the one-electron Dirac Hamiltonian with perturbations arising from both a uniform external magnetic field B and internal nuclear magnetic moments
mK. These perturbations are introduced through the vector
potential A = AB + KAmK,
hDA = mc2 + + Vnuc = hD + c A,
rG = r RG ,
1 mK rK
.
c2
rK3
1h 1
1d1
1
2
1
2g1,2
1
2d1d2.
p
p
p =
X cX p cp
X=L,L,S,s X
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-3
L
0
L =
,
0
0
UMO
B = Bcq0.
q
L =
L
0
S =
0
0
10
X = L,L,S,S ,
1
hBM B = hD + B rG c M B
2
1
= exp iABR M r hD + B r M c M ,
2
11
showing that the reference to the gauge origin in the operator
rG is replaced with the reference to the center of the orbital
r M . In the LAO-based method, the LAOs Eq. 10 replace
the corresponding basis functions in the basis set expansion
in Eq. 8. It should be noted that, in contrast to the nonrelativistic domain, the relativistic LAOs by themselves do not
describe first-order changes in the atomic orbitals due to an
external magnetic field, as seen from the fact that the Zeeman operator Eq. 5 is no longer proportional to the orbital
angular momentum operator. However, the introduction of
LAOs removes reference to an arbitrary gauge origin in the
calculation of magnetic properties and significantly improves
basis set convergence, as will be shown in the following.
In order to ensure orthogonality of orbitals and thus the
same algebra of creation and annihilation operators at all
values of the magnetic perturbation, one must allow the coefficients cpB to be field dependent
pB = BcpB
14
OMO
B = UMO
BTqpB.
p
q
13
S =
12
1
gBpqrsaparasaq ,
2 pqrs
15
where the tildes on the matrix elements indicate that the field
operators have been expanded in the B-dependent OMOs.
E E i
dE
=
+
.
dA A i A
16
E
=0
17
d 2E
2E i
2E
d E
=
=
+
,
dBdA dB A
i A B B A
18
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-4
Ilia et al.
d E
2E
2E i
=
+
= 0,
dB i
B i i j B
19
d 2E
dmK,dB
20
mK=0,B=0
21
aiaaai
ai
* aa ,
ai
i a
= OMO exp
p
qp
q
= rUMOBTrqBexp qp
23
rq
mK=0,B=0
= Bi hmAi + ihmABi .
25
26
XB =
xB
,
x B*
B
xBai = ai
27
28
The Hessian E2
0 as well as the gradients associated with
properties other than a uniform magnetic field have their
usual form see, for example, Refs. 64 and 67 for explicit
expressions. We follow these references and use the following general form for a property gradient in relativistic fourcomponent linear response:
EX1 =
gX
,
g X*
gXai =
EX
*
ai
X0 = FX ,
= 0 ai aa,H
ai
=0
29
where the first-order derivative with respect to property X of
the Fock matrix in OMO basis appears
FX = hX + L
X ,
pq
pq
pq,jj
X = gX gX .
L
pq,rs
pq,rs
ps,rq
30
FB = hB + LB + TB,h + TB,L ,
0 pq
pqjj
pq
pqjj
pq
j
B=0
d 2E
dBTdmA
In the particular case of a uniform magnetic field, the property gradient is extracted from
22
B di
i
dB
A =
The last term is a relaxation term, whereas the first and second terms together constitute a static term consisting of the
direct GIAO contribution and a reorthonormalization term,
respectively.
The first derivative of the HF energy at any field strength
with respect to a nuclear magnetic moment mK is the expectation value of the hyperfine operator hmK Eq. 5 in terms
of the OMOs at the selected field strength. The shielding
tensor is therefore the first-order derivative of this expectation value with respect to the external magnetic field B
31
32
corresponding to reorthonormalization terms for a oneelectron and a two-electron operator, respectively.10 The
property gradient associated with the nuclear magnetic moment evaluates to
mK
mK
= hai
.
gai
33
The second term of the shielding tensor in the LAO formalism Eq. 26 is an expectation value term
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-5
34
corresponding to the second-order derivative of matrix elements of the Dirac Hamiltonian Eq. 4 in OMO basis with
respect to the external magnetic field and the selected nuclear
magnetic moment. In the nonrelativistic domain, this term
also contains an explicit diamagnetic contribution, which is
absent in the relativistic domain since the Dirac Hamiltonian
Eq. 4 is linear in the nuclear magnetic moments as well as
the external magnetic field. Explicit expressions for the
atomic matrix elements for UMO appearing in Eqs. 31,
33, and 34 are given in the Appendix, whereas explicit
expressions for the first-order connection matrix TB will be
discussed in Sec. II C.
Our working expression, Eq. 26 can be compared to
the previous expressions employed by Quiney and
co-workers5658 in their calculation of the NMR shielding
tensor for the water molecule at the four-component relativistic HF level with LAOs. Quiney et al. calculated the linear
response function at the uncoupled HF level and without
inclusion of the LAO contributions to the property gradient
Eq. 31. Curiously, they employed the opposite sign of the
vector potential in their definition of the four-component
LAO Eq. 10. This does not remove reference to the arbitrary gauge origin in coupled HF calculations and furthermore changes the sign of the expectation value term Eq.
34 of which they calculated only the first one and not the
reorthonormalization term.
A first particular feature of four-component relativistic
calculations is the presence of negative-energy positronic
orbitals. The response of the HF wave function to a pertur++
bation is therefore not only carried by orbital rotations ai
between virtual and occupied positive-energy orbitals but
+
also by orbital rotations ai
involving the virtual negativeenergy orbitals. In the nonrelativistic limit the contribution
from the latter rotations reduces to the diamagnetic
contribution.69
A second feature to consider is that the coupling between
the large and the small components is modified by the introduction of an external vector potential. In our approach to
four-component relativistic calculations, the small component basis set is generated using the kinetic balance relation.
In this work, we have studied two schemes for the kinetic
balance relationthe unrestricted kinetic balance UKB
Ref. 77 and the restricted kinetic balance RKB.78 The
RKB scheme gives a 1:1 ratio between the sizes of the large
and small component basis sets. On the other hand, starting
from a large component scalar basis function of a given
value of the orbital angular momentum l, the extra small
component basis functions generated by the UKB scheme
include, for the atomic problem, the functions needed for
magnetic balance associated with an external magnetic
field, whereas such functions are generated by the RKB
scheme only by inclusion of large component basis functions
of l + 2.62,69,79,80 In this way, UKB provides a better basis set
for the calculation of magnetic properties,81 albeit at the increased risk of introducing linear dependencies.
Alternatively, one may take magnetic balance explicitly
into account in four-component relativistic response calculations of magnetic properties. This approach has been pursued
by Liu and co-workers6062 and Komorovsk et al.63 and
leads to the introduction of perturbation-dependent basis
functions, albeit for solving a different problem than the
gauge origin problem addressed in the present paper. A different approach has been proposed by Kutzelnigg,82 whereby
the contribution from the negative-energy orbitals to the
shielding constant is reduced from Oc0 to Oc4 through a
unitary transformation, involving the full vector potential, of
the Dirac operator. The formulation furthermore leads to perturbation operators similar to those found in the nonrelativistic domain. Numerical studies by Visscher83 and Xiao and
co-workers60,61 have revealed numerical instabilities in the
original formulation, dubbed full field-dependent unitary
transformation FFUT by the latter authors. This problem is
avoided by restricting the unitary transformation to the vector potential associated with the external field only, leading
to the external field-dependent unitary tranformation
EFUT, although the negative-energy orbitals now contribute to Oc2. However, drawbacks of both methods are that
gauge transformations become more complicated and that
the unitary transformation does not commute with the Gaunt
two-electron interaction.84
C. Explicit and implicit orbital connections
35
d
pUMO
BB = pBcq0.
q
dB
36
The symmetric connection sc minimizes the difference between the field-dependent OMOs and the corresponding
UMOs at the same magnetic field strength and leads to TB
= 21 SB, which is independent of the chosen gauge origin.
The natural connection nc, on the other hand, minimizes
the difference between the field-dependent OMOs and the
corresponding zeroth-order orbitals and leads to gauge origin
dependent TB = WB.85 However, insertion into Eq. 24
shows that in the nc and in the limit of a complete basis the
reorthonormalization term cancels exactly the direct GIAO
contribution. Moreover, again shown starting from Eq. 24,
in any basis the nc assures that the overlap between any
occupied zeroth- and first-order orbitals is zero, that is,
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-6
Ilia et al.
jBi = 0
37
38
B
Aai = Tai
Bai .
39
A = c* iBhmAi + ihmABci
40
41
* h + h A .
+ Aai
a mA i
i mA a ai
42
i,
ij
ia
Combined with the first-order reorthonormalization condition Eq. 35, this clearly shows that the term involving the
occupied-occupied orbital mixing matrix is independent of
the chosen connection. A further important conclusion is that
since the first term and the total property are clearly connection independent, this must also then hold true for the third
term involving the occupied-virtual block. Compared with
Eq. 39, we then see that if restrictions are introduced on the
virtual orbitals employed in the response calculation, they
must be imposed on the connection matrix as well. In relativistic calculations, rotations +
ai between virtual negativeenergy orbitals and occupied positive-energy ones may be
suppressed and approximated by an expectation value corresponding to the diamagnetic term in the nonrelativistic limit.
+;B
ai
FB
pq
,
2mc2
43
where appears the property gradient of Eq. 31. The reorthonormalization contribution to the property gradient can be
reformulated as
TB,Fai = Trt* Laits + LairtTts.
44
B
B
2 B
*
TB,F+
ai = Tai i + aTai iS 2mc Tai ,
45
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-7
16s12p7d
17s13p7d
21s17p12d1f
22s18p13d2f
27s21p15d4f
28s22p17d5f3g2h
4s1p
5s2p
5s2p1d
6s3p2d
11s6p3d
11s6p3d
Hamiltonian.78 We have compared the results to the corresponding calculations with the standard scalar nonrelativistic
Hamiltonian in the DALTON12,91 program suite, and the results
are identical within numerical accuracy.
The aim of these test calculations is to illustrate and to
emphasize the computational advantages of the use of LAOs
in connection with the rigorous DiracCoulomb based relativistic theory rather than to present very accurate values of
NMR shieldings. This would require inclusion of electron
correlation and is beyond the scope of the present endeavor.
The hydrogen iodide molecule in its experimental geometry 3.040 a.u. Ref. 92 was chosen for this study because
it contains a heavy element where one would expect that
relativistic effects are significant.
NMR shielding constants have been calculated with several decontracted basis sets selected from the literature to
form an approximative sequence of increasing size. They are
shown in Table I. Four gauge origin positions were tested for
the hydrogen iodide molecule coordinates in a.u.: 1 on the
iodine atom 0,0,0 for compatibility with the previous CGO
calculations of Visscher et al.59 2 at the center of mass,
0,0,0.023 952, the typical choice for CGO calculations, 3
on the hydrogen atom 0,0,3.040, and 4 7 a.u. from the
iodine atom 0,0,7. The purpose of the last two positions are
to simulate the effects of LAOs on larger molecules, where
one cannot place the gauge origin close to all atoms. For the
relativistic calculations, we applied a Gaussian finite nucleus
model,93 but in the nonrelativistic LL mode we used the
point nucleus model in order to compare directly to
previous78 and to verifying calculations with the DALTON12,91
program system. All relativistic calculations were done with
the LL LL and SS LL classes of two-electron integrals,
while the SS SS and SL SL Gaunt classes were omitted
L and S denoting large and small component basis functions, respectively, cf. Eq. 9. The used speed of light was
137.035 999 8 a.u. All calculated values of NMR shieldings
are in ppm.
IV. RESULTS AND DISCUSSION
Calculated four-component results for the NMR shielding constants of the HI molecule are collected in Tables II
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-8
Ilia et al.
TABLE II. Isotropic proton NMR shielding constants in ppm of hydrogen iodide calculated with nonrelativistic LL, SF, RKB-relativistic, and UKB-relativistic methods; each both with CGO on GO and with LAOs.
The basis sets are specified in Table I. The difference of each value to the corresponding value in the largest
basis set VI is given in parentheses. The experimental value for the 1H shielding constant is 43.86 ppm after
converting the gas-phase shift from Ref. 104 to absolute shielding using CH4 = 30.61 ppm from Ref. 105.
UKB-CGO
GO I
GO CM
GO H
GO 0,0,7
UKB-LAO
UKB-LAOa
RKB-CGO
RKB-CGOb
GO
GO
GO
GO
GO
I
CM
H
0,0,7
I
GO CM
GO H
GO 0,0,7
RKB-LAO
RKB-LAOa,b
SF-CGOb
GO I
SF-LAOa,b
LL-CGOb
LL-LAOb
GO I
II
III
IV
VI
44.03
4.00
45.36
212.58
432.14
45.89
3.45
45.88
44.03
4.00
45.37
213.14
433.43
46.31
3.03
46.31
45.25
2.78
46.46
199.31
400.00
47.03
2.31
47.03
46.09
1.94
47.28
196.85
393.23
47.80
1.54
47.80
47.87
0.16
48.71
153.73
291.61
49.41
0.07
49.40
48.03
37.17
37.93
134.21
260.63
44.69
3.37
46.06
218.44
444.77
39.36
47.19
2.02
39.19
39.96
136.38
262.97
44.25
3.81
45.61
217.18
442.44
40.92
46.49
2.72
42.93
43.68
139.45
265.19
46.16
1.90
47.41
204.24
410.17
44.63
47.93
1.28
45.46
46.21
140.59
264.50
46.33
1.73
47.54
200.50
401.34
47.12
47.92
1.29
47.30
47.84
115.96
205.39
47.99
0.07
48.85
157.12
299.28
48.79
49.27
0.06
29.11
1.84
31.19
0.14
29.58
1.37
31.22
0.17
30.39
0.56
31.10
0.05
30.69
0.26
31.13
0.08
30.91
0.04
31.07
0.02
29.20
1.88
31.32
0.13
29.68
0.40
31.37
0.18
30.53
0.55
31.25
0.06
30.83
0.25
31.28
0.09
31.05
0.03
31.20
0.01
48.68
129.97
236.71
49.34
49.33
47.62
48.03
99.75
167.65
48.06
48.73
133.23
244.17
48.88
49.21
30.95
31.05
31.08
31.19
a
The negative energypositive energy T+ part of the reorthonormalization is neglected holds for UKB-LAO,
RKB-LAO, and SF-LAO approaches, otherwise is present.
b
Diamagnetic term calculated as expectation value of nonrelativistic diamagnetic operator. Otherwise obtained
from the linear response function.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-9
TABLE III. Isotropic 127I iodine NMR shielding constants in ppm of hydrogen iodide calculated with nonrelativistic LL, SF, RKB-relativistic, and UKB-relativistic methods; each both with CGO on GO and with
LAOs. The basis sets are specified in Table I. The difference of each value to the corresponding value in the
largest basis set VI is given in parentheses.
UKB-CGO
GO I
GO CM
GO H
GO 0,0,7
UKB-LAO
UKB-LAOa
RKB-CGO
RKB-CGOb
GO
GO
GO
GO
GO
I
CM
H
0,0,7
I
GO CM
GO H
GO 0,0,7
RKB-LAO
RKB-LAOa,b
SF-CGOb
GO I
SF-LAOb
LL-CGOb
LL-LAOb
GO I
II
III
IV
VI
5673.0
187.4
5675.2
5947.9
6306.1
5905.2
2.0
5905.2
5808.9
51.5
5809.0
5822.2
5839.5
5961.4
54.2
5961.4
5817.8
42.6
5819.1
5977.1
6184.6
5951.6
44.4
5951.6
5852.0
8.4
5852.1
5868.2
5889.4
5921.7
14.5
5921.7
5854.4
6.0
5854.4
5853.2
5851.7
5960.3
53.1
5959.4
5860.4
4753.3
4755.4
5029.3
5388.8
6556.9
217.9
6559.1
6829.8
7185.3
4975.3
6767.6
43.6
4890.6
4890.8
4903.6
4920.4
6690.1
84.7
6690.2
6701.5
6716.3
5027.3
6811.7
0.5
5239.4
5240.7
5399.9
5608.9
6732.9
41.9
6734.1
6890.2
7095.2
5367.1
6845.8
34.6
5278.7
5278.9
5295.7
5317.7
6765.1
9.7
6765.2
6779.5
6798.2
5343.9
6820.1
8.9
5322.8
5322.8
5321.7
5320.2
6769.5
5.3
6769.5
6766.4
6762.4
5422.1
6848.6
37.4
5378.2
5378.2
5377.0
5375.4
6774.8
6295.6
248.2
6465.7
78.2
6472.0
71.8
6509.7
34.2
6490.0
53.8
6540.9
3.0
6540.0
3.8
6542.3
1.6
6540.0
3.8
6540.2
3.7
6543.8
4390.4
148.6
4536.5
2.5
4538.6
0.4
4573.3
43.3
4506.4
32.6
4549.8
10.8
4546.8
7.8
4548.9
9.9
4541.3
2.3
4541.5
2.5
4539.0
5860.4
5859.0
5857.2
5907.2
5908.4
6774.8
6771.6
6767.3
5423.2
6811.2
6543.9
4539.0
a
The negative energypositive energy part T+ of the reorthonormalization is neglected holds for UKB-LAO,
RKB-LAO, and SF-LAO approaches, otherwise is present.
b
Diamagnetic term calculated as expectation value of nonrelativistic diamagnetic operator. Otherwise obtained
from the linear response function.
magnetic parts are 973.9 and 4404.4 ppm, respectively. The corresponding UKB-CGO numbers are 970.8 and
4889.6 pm, showing that the RKB and UKB paramagnetic
values are very close to each other, and that the major discrepancy arises from the diamagnetic contribution. This illustrates that the magnetic balance is most important for the
diamagnetic type terms from the +-rotations. This is in
agreement with the general observation that Dyalls modified
Dirac equation,78,96 which makes the number of negativeenergy orbitals equal to the number of positive-energy ones
and in fact defines our RKB scheme, does not change the
quality of the positive-energy orbitals relative to UKB. As an
example of this, the converged HF energy in basis I is
7115.7787 hartrees with RKB and 7115.7768 hartrees
with UKB. For all basis sets, the difference between the
RKB and UKB total energy is less than 0.0001 hartree.
Next, solving the response equation Eq. 19, separately for ++- and +-rotations gives RKB-CGO I
= 5379.4 ppm= 975.4 ppm paramagnetic 4404.0 diamagnetic, which means that ignoring the coupling of the
two classes of rotations, which are on the order of Oc1, in
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-10
Ilia et al.
TABLE IV. Individual parallel and perpendicular contributions to NMR shielding constants of HI. For LAO only: scsymmetric connection,
ncnatural connection. For CGO calculations GO I. Used basis sets I and VI.
RKB-LAO
Basis
UKB-LAO
H
RKB-CGO
UKB-CGO
1.26
0.23
13.85
1932.53
48.80
51.39
5798.33
5800.58
0.35
0.23
4916.40
4890.85
0.70
0.24
5795.80
5797.14
51.45
51.47
4913.57
4886.99
48.80
51.39
48.18
50.22
31.92
2020.10
2.91
3.58
343.61
629.31
42.41
48.39
341.47
624.59
42.41
48.39
1591.06
1667.63
2.91
3.58
1589.75
1663.25
2.91
3.58
45.89
50.11
45.88
47.81
6141.94
6429.88
7386.86
7464.77
42.76
48.15
48.54
47.88
5257.87
5515.44
6503.32
6550.24
43.11
48.14
45.88
47.81
London
I
1.70
5.36
1.70
VI
3.42
5.73
3.42
I
0.00
0.00
0.00
VI
0.00
0.00
0.00
Reorthonormalizationb
,nc
I
10.54
19.35
2.35
,nc
VI
0.41
11.89
3.55
,sc
I
4.80
3.09
12.60
,sc
VI
10242.17
2058.42
12216.98
,nc and ,sc
I
0.00
0.00
0.00
VI
0.00
0.00
0.00
RKB: diamagnetic expectation value/UKB: diamagnetic response
,nc
I
5 789.49
12.33
4 947.89
,nc
VI
5 796.91
5.83
4 906.52
,sc
I
5 789.49
12.33
4 947.51
,sc
VI
5 796.91
5.83
6 883.51
,nc and ,sc
I
5 795.80
51.45
4 913.57
,nc and ,sc
VI
5 797.14
51.47
4 886.99
Paramagnetic response ee
,nc
I
656.17
48.18
654.13
,nc
VI
683.67
50.22
679.22
,sc
I
671.51
31.92
669.47
,sc
VI
10 926.27
2020.09
10 915.66
,nc and ,sc
I
1 591.06
2.91
1 589.75
,nc and ,sc
VI
1 667.63
3.58
1 663.25
Total
I
6 457.90
46.52
5 606.07
VI
6 484.43
49.88
5 585.62
I
7 386.86
48.54
6 503.32
VI
7 464.77
47.88
6 550.24
5.36
5.73
0.00
0.00
6.38
5.61
5.23
131.95
0.00
0.00
A =
i,a
1 rA c
i 2
c
rA3
a i
1
rG c i
2
A =
,
46
where the summation over virtual orbitals a is restricted to
negative-energy ones. Following Sternheim,98 the diamagnetic expectation value expression is obtained by approximating the orbital energy difference a i by 2mc2 and
approximating the sum over negative-energy orbitals by a
resolution of identity. We consider these two approximations
separately. We find that the replacement of the orbital energy
difference by a common energy denominator changes the
diamagnetic contribution by 318.2 ppm to 4724.3 ppm,
2 i,p
1 rK
c2
rK3
1
1 rK
i 2
2 i,p+
c
rK3
1
rG i
2
1
rG i ,
2
47
where the first term has a sum over all orbitals, whereas the
sum in the second term is restricted to all positive-energy
orbitals. In the second term the Dirac matrices of the magnetic operators couple the large and small components of
positive-energy orbitals, and this term therefore goes to zero
in the nonrelativistic limit with the small components. The
Sternheim approximation therefore becomes exact in the
nonrelativistic limit. However, ignoring the second term at
the finite value of the speed of light changes the diamagnetic
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-11
A scheme for GIAOs for use in four-component relativistic calculations of second-order magnetic properties has
been implemented and tested. The same good convergence as
in nonrelativistic calculations of calculated NMR shielding
constants with basis set size was observed for the London
orbital perturbation dependent basis sets. It can be rationalized by the fact that the introduction of LAOs in the relativistic domain provides atomic magnetic balance even for the
molecular problem. The extra small component functions
then needed for magnetic balance can be obtained by taking
magnetic balance explicitly into consideration. For an exter-
Here the derivations of matrix elements of the derivatives of the Dirac Hamiltonian Eq. 4 over LAOs are
given, as we have not been able to find them in literature. For
completeness we also give relevant derivatives of the twoelectron Coulomb term although they are the same as in
nonrelativistic work and can be found in Ref. 10, for example.
In the LAO basis the matrix elements of the one- and
two-electron operators are given by
h = M hN
=
M exp
i
B R MN r
2
1
hD + B rN c
2
+
1
r c
mK K r3
c2 K
K
N ,
A1
g = M PgNQ
= M P exp
i
B R MN r1
2
+ R PQ r2 g NQ ,
A2
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-12
Ilia et al.
1
i
hB = Q MNM rhDN + M rN c N ,
2
2
hmK =
1
M
c2
K
hBm
=
rK c
rK3
N ,
rK c
i
Q MN M r
2c2
rK3
A3
A4
N ,
i
gB = M PQ MNr1 + Q PQr2gNQ,
2
A5
A6
Z MN
Z MN
Y MN
X MN
Y MN
X MN .
0
A7
These matrix elements in AO basis have all been implemented in the HERMIT module in DIRAC04. In the linear response module, they are transformed to the corresponding
matrix elements in UMO basis appearing in Eqs. 31 and
34 using Eq. 13.
F. London, J. Phys. Radium 8, 397 1937.
Aa. E. Hansen and T. D. Bouman, J. Chem. Phys. 82, 5035 1985.
3
H. F. Hameka, Mol. Phys. 1, 203 1958.
4
J. A. Pople, Proc. R. Soc. London, Ser. A 239, 541 1957.
5
J. A. Pople, Proc. R. Soc. London, Ser. A 239, 550 1957.
6
H. F. Hameka, Rev. Mod. Phys. 34, 87 1962.
7
R. Ditchfield, D. P. Miller, and J. A. Pople, J. Chem. Phys. 53, 613
1970.
8
R. Ditchfield, Mol. Phys. 27, 789 1974.
9
K. Wolinski, J. F. Hinton, and P. Pulay, J. Am. Chem. Soc. 112, 8251
1990.
10
T. Helgaker and P. Jrgensen, J. Chem. Phys. 95, 2595 1991.
11
S. T. Epstein, J. Chem. Phys. 58, 1592 1973.
12
K. Ruud, R. Kobayashi, P. Jrgensen, K. L. Bak, and H. J. Aa. Jensen, J.
Chem. Phys. 100, 8178 1994.
13
J. Gauss, Chem. Phys. Lett. 191, 614 1992.
14
J. Gauss, J. Chem. Phys. 99, 3629 1993.
15
J. Gauss, Chem. Phys. Lett. 229, 198 1994.
16
J. Gauss and J. F. Stanton, J. Chem. Phys. 102, 251 1995.
17
J. Gauss and J. F. Stanton, J. Chem. Phys. 104, 2574 1996.
18
G. Schreckenbach and T. Ziegler, J. Phys. Chem. 99, 606 1995.
19
A. M. Lee, N. C. Handy, and S. M. Colwell, J. Chem. Phys. 103, 10095
1995.
20
G. Rauhut, S. Puyear, K. Wolinski, and P. Pulay, J. Phys. Chem. 6310,
6310 1996.
21
J. R. Cheeseman, G. W. Trucks, T. A. Keith, and Mi. J. Frisch, J. Chem.
Phys. 104, 5497 1996.
22
T. Helgaker, P. J. Wilson, R. D. Amos, and N. C. Handy, J. Chem. Phys.
113, 2983 2000.
23
P. Pyykk, Chem. Rev. Washington, D.C. 88, 563 1988.
24
P. Pyykk, Chem. Phys. 74, 1 1983.
25
N. C. Pyper, Chem. Phys. Lett. 96, 204 1983.
26
J. Vaara, Phys. Chem. Chem. Phys. 9, 5399 2007.
27
L. B. Casabianca and A. C. de Dios, J. Chem. Phys. 128, 052201 2008.
28
C. C. Ballard, M. Hada, H. Kaneko, and H. Nakatsuji, Chem. Phys. Lett.
254, 170 1998.
29
H. Fukui and T. Baba, J. Chem. Phys. 108, 3854 1998.
30
T. Baba and H. Fukui, J. Chem. Phys. 110, 131 1999.
31
R. Fukuda, M. Hada, and H. Nakatsuji, J. Chem. Phys. 118, 1015 2002.
1
2
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18
124119-13
1986, p. 1.
H. Sellers, Int. J. Quantum Chem. 30, 433 1986.
76
T. Helgaker and P. Jrgensen, in Methods in Computational Molecular
Physics, edited by S. Wilson and G. H. F. Diercksen Plenum, New York,
1992.
77
T. Saue, K. Fgri, T. Helgaker, and O. Gropen, Mol. Phys. 91, 937
1997.
78
L. Visscher and T. Saue, J. Chem. Phys. 113, 3996 2000.
79
M. Pecul, T. Saue, K. Ruud, and A. Rizzo, J. Chem. Phys. 121, 3051
2004.
80
W. Kutzelnigg and W. Liu, J. Chem. Phys. 131, 044129 2009.
81
A. F. Maldonado and G. A. Aucar, Phys. Chem. Chem. Phys. 11, 5615
2009.
82
W. Kutzelnigg, Phys. Rev. A 67, 032109 2003.
83
L. Visscher, Adv. Quantum Chem. 48, 369 2005.
84
K. G. Dyall and K. Fgri, Jr., Introduction to Relativistic Quantum
Chemistry Oxford University Press, New York, 2007.
85
J. Olsen, K. L. Bak, K. Ruud, T. Helgaker, and P. Jrgensen, Theor.
Chim. Acta 90, 421 1995.
86
X. Gonze, Phys. Rev. A 52, 1096 1995; 52, 4591E 1996.
87
M. Krykunov and J. Autschbach, J. Chem. Phys. 123, 114103 2005.
88
HERMIT, a molecular integral program, written by T. Helgaker, P. R. Taylor, and K. Ruud.
89
H. J. Aa. Jensen, T. Saue, and L. Visscher with contributions from V.
Bakken, E. Eliav, T. Enevoldsen, T. Fleig, O. Fossgaard, T. Helgaker, J.
Lrdahl, C. V. Larsen, P. Norman, J. Olsen, M. Pernpointner, J. K. Pedersen, K. Ruud, P. Salek, J. N. P. van Stralen, J. Thyssen, O. Visser, and
75
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.120.228.135 On: Fri, 20 Jun 2014 15:26:18