COMMENTS

Comment on Density functional calculation of nuclear magnetic

resonance chemical shifts [J. Chem. Phys. 102, 2806 (1995)]
Vladimir G. Malkin
Institute of Inorganic Chemistry, Slovak Academy of Sciences, SK-84236 Bratislava, Slovakia

Olga L. Malkina
Computer Center, Faculty of Natural Sciences, Comenius University, Mlynska dolina CH-1,
SK-84215 Bratislava, Slovakia

Dennis R. Salahub
Departement de Chimie, Universite de Montreal, C.P. 6128, Succursale Centre-ville, Montreal, Quebec,
Canada H3C 3J7, and Centre de Recherche en Calcul Applique, 5160, boul. Decarie, Bureau 400, Montreal,
Quebec, Canada H3X 2H9

~Received 28 July 1995; accepted 3 October 1995!

@S0021-9606~96!00902-1#
Very recently van Wullen has presented a theoretical
analysis of different approaches for the calculation of NMR
chemical shifts based on density functional theory ~DFT!.1
He presented a lucid comparison of various approaches to
the gauge problem and the role of the current density in
current-density functional theory.
The paper contains one point concerning our recently
developed sum-over-states density functional perturbation
theory ~SOS-DFPT! approach2,3 which should be clarified
and put into proper context, because our method and code
are currently in wide use.
Van Wullen correctly pointed out that @considering the
case of a common gauge origin for all occupied molecular
orbitals ~MOs!# our SOS-DFPT approach is not gauge invariant, even in the limit of a complete basis. This is clearly true
because, in our approach, we changed the expression for the
paramagnetic contribution to the shielding tensor s p while
keeping the expression for the diamagnetic contribution s d
unmodified. Since only the sum of these two terms is gauge
invariant, the final expression in our approach is gauge dependent, even in the limit of a complete basis set. Approaches such as the coupled HartreeFock ~CHF! and uncoupled DFT ~UDFT! methods are free of such a fault, in the
limit of a complete basis set.
However, in practical calculations of chemical shifts one
always has to use a finite basis set and therefore the CHF and
UDFT approaches have the same faultthey, too, are not
gauge invariant ~here we restrict our attention to calculations
with a common gauge origin!. Indeed, this is the reason that
different approaches to avoid this problem have been introduced ~see discussion in Ref. 1, for example!. In order to
judge whether the shortcoming pointed out by van Wullen is
primarily formal in nature or whether a significantly worse
practical dependence of the results on the choice of gauge
origin in comparison with the UDFT method might exist, we
have performed calculations of the shielding constant for the
F2 molecule with both UDFT and SOS-DFPT approaches
using different common gauge origins and the IGLO-III basis set.2,4 The results of these calculations are presented in
Table I.
J. Chem. Phys. 104 (3), 15 January 1996

At first sight, the conclusion is a bit surprising: SOSDFPT is roughly twice as stable with respect to movement of
the gauge origin as UDFT! However, this is easy to rationalize, remembering that the SOS-DFPT approach with the
LOC1 approximation2,3 leads to more shielded results by decreasing the ~absolute! value of s p . Since s p is the major
source of total error and gauge dependence in the calculations, the SOS-DFPT approach improves the results by decreasing this term and, along with it, the corresponding
gauge dependence. For F2 this improvement is about 4050
ppm and that results in decreasing the gauge dependence of
SOS-DFPT compared with UDFT. We have not always
found that SOS-DFPT is more stable than the UDFT approach ~judging by the results for other systems!. For example, results for HNSN21 are given in Table II, which
should be self-explanatory. However, the general conclusion
is essentially the same: SOS-DFPT improves results in comparison with UDFT and this improvement normally is much
bigger than the difference in gauge dependence between
these two methods, when practical basis sets are used.
TABLE I. Absolute shielding constant ~SC! for F1 in the F2 molecule. Shift
with respect to SC calculated with gauge origin on F1 is in parentheses. All
values are in ppm.
Approximation
Gauge origin ~ !

UDFT

SOS-DFPT

*s
F1
F2

2245.2

2198.3

s
*
F1
F2

2247.8
~2.6!

2199.7
~1.4!

s s
* F2
F1

2246.5
~1.3!

2199.0
~0.7!

*ss
F1
F2

2242.6
~22.6!

2196.9
~21.4!

ss
*
F1
F2

2250.4
~5.2!

2201.1
~2.8!

0021-9606/96/104(3)/1163/2/$6.00

1996 American Institute of Physics

1163

1164

Letters to the Editor

TABLE II. The gauge origin dependence of chemical shifts calculated with
UDFT and SOS-DFPT methods in HNSN21. All values are in ppm. d is the
difference between the absolute values of the shifts due to the different
choice of gauge origins for SOS-DFPT and UDFT. D~s!5s ~SOS-DFPT!
2s~UDFT!.

Nucleus

Gauge origin
on

Approximation
UDFT

SOS-DFPT

D~s!

S
N1
N2
H

0.0
24.0
5.5
5.5

0.0
28.3
3.9
20.6

0.0
4.3
21.6
24.9

50.0
54.3
51.6
56.1

N1

S
N1
N2
H

16.8
0.0
27.3
19.7

18.0
0.0
27.8
19.0

1.2
0.0
0.5
20.7

47.2
48.4
47.9
49.1

chemical shifts, the shortcoming of SOS-DFPT pointed out

by van Wullen has significance only from the theoretical
point of view.
As a consequence, the SOS-DFPT approach with the
IGLO choice of gauge origins2,3 leads to superior results in
comparison with the uncoupled DFT approach ~with the
same or equivalent choice of gauge origins! with essentially
the same computational effort, providing results almost as
good as MP2 for 13C chemical shifts in systems with relatively small correlation effects5 and significantly more reliable for systems with strong correlation.3,6 SOS-DFPT is at
present a widely tested method suitable for calculation of
large systems including biosystems and transition-metal
complexes.2,3,6 11
Chr. van Wullen, J. Chem. Phys. 102 ~7!, 2806 ~1995!.
V. G. Malkin, O. L. Malkina, M. E. Casida, and D. R. Salahub, J. Am.
Chem. Soc. 116, 5898 ~1994!.
3
V. G. Malkin, O. L. Malkina, L. A. Eriksson, and D. R. Salahub, in
Theoretical and Computation Chemistry, edited by J. M. Seminario and P.
Politzer ~Elsevier, Amsterdam, 1995!, Vol. 2.
4
W. Kutzelnigg, U. Fleischer, and M. Schindler, in NMRBasic Principles
and Progress ~Springer-Verlag, Heidelberg, 1990!, Vol. 23, p. 165.
5
Chr. Maerker, P. von R. Schleyer, D. R. Salahub, O. L. Malkina, and V. G.
Malkin, Chem. Phys. Lett. ~to be submitted!.
6
M. Kaupp ~private communication!.
7
M. Kaupp, V. G. Malkin, O. L. Malkina, and D. R. Salahub, Chem. Phys.
Lett. 235, 382 ~1995!.
8
M. Kaupp, V. G. Malkin, O. L. Malkina, and D. R. Salahub, J. Am. Chem.
Soc. 117, 1851 ~1995!.
9
M. Kaupp, V. G. Malkin, O. L. Malkina, and D. R. Salahub, Chem. Eur. J.
~in press!.
10
V. G. Malkin, O. L. Malkina, and D. R. Salahub, J. Am. Chem. Soc. 117,
3294 ~1995!.
11
T. B. Woolf, V. G. Malkin, O. L. Malkina, D. R. Salahub, and B. Roux,
Chem. Phys. Lett. 239, 186 ~1995!.
1
2

N2

S
N1
N2
H

24.2
33.6
0.0
26.4

25.2
34.1
0.0
27.3

1.0
0.5
0.0
0.9

20.3
20.8
21.3
22.2

S
N1
N2
H

16.2
15.7
6.2
0.0

18.0
16.8
7.1
0.0

1.8
1.1
1.1
0.0

0.6
1.2
1.3
2.3

Up to this point, the discussion has been restricted to

calculations using a common gauge origin. However, the
IGLO choice of gauge origins is used throughout SOS-DFPT
calculations ~and recommended for users of the deMonNMR program!. Since, in practice, the IGLO choice of
gauge origins removes the gauge dependence of NMR

J. Chem. Phys., Vol. 104, No. 3, 15 January 1996