Professional Documents
Culture Documents
Kao 2016
Kao 2016
Kao 2016
To cite this article: Der-you Kao & Mark R. Pederson (2016): Use of Löwdin orthogonalised
Fermi orbitals for self-interaction corrections in an iron porphyrin, Molecular Physics, DOI:
10.1080/00268976.2016.1225992
Article views: 34
Download by: [University of California, San Diego] Date: 12 December 2016, At: 04:52
MOLECULAR PHYSICS,
http://dx.doi.org/./..
a
Mechanical and Aerospace Engineering, The George Washington University, Washington, DC, USA; b Department of Chemistry, Johns Hopkins
University, Baltimore, MD, USA
incorrect fillings in the f-electron series. A discussion of been determined. Consider, for example, the neon atom
this problem can be found in [18]. which is a closed shell with a 2s2p3 valence. Within this
The need for the use of SICs for systems containing shell, one can construct at least three sets of localised
transition-metal ions was first suggested by Lindgren [6] orbitals that automatically satisfy the localisation equa-
in applications to the Cu atoms. The formulation pro- tions due to symmetry considerations. These sets are the
posed by Lindgren introduced the self-interaction at the cartesian set given by {|2s 〉, |2px 〉, |2py 〉,
level of the single-electron Hamiltonian and proposed |2pz 〉}, an explicitly complex spherical harmonic set
|2p +i|2p |2p −i|2p
correcting the error in the interelectronic exchange func- that is given by {|2s, x√(2) y , x√(2) y , |2pz }, and
tional, derived from the free-electron-gas model, in a the standard sp3 hybrids. It was demonstrated in 1988
local-density approximation. In 1981, Perdew and Zunger by Pederson and Lin [20] that the hybrids minimise the
proposed the SIC at the level of the total energy expres- total energy for the Kohn–Sham exchange only func-
sion, so it could be applied to functionals designed to tional. It has been further discussed that this is the only
include correlation and exchange effects [1] and also serve choice that automatically leads to three-fold degenera-
as an ab initio alternative to the LSDA+U method for the cies for the canonical 2p orbitals and it is now recog-
description of strongly correlated systems. nised that this set is the minimising set determined from
Perdew and Zunger posited that the orbital densi- the FLO-SIC formulation [34]. Here, we argue further
ties, ρ i (r) = |φ iσ (r)|2 , could be constructed from a set that since the choice of the nodal features of the min-
of localised orbitals that are similar to the Edmiston– imising set of localised orbitals, in the original PZ func-
Ruedenberg [19] energy-localised orbitals in the Hartree tional, will also depend on the features of the functional
approximation. Later, Pederson et al. [20–22] suggested itself (GGA, LDA, SCAN [35], etc.), the FLO-SIC formu-
the use of localisation equations as a variational means lation would remove this seemingly undesirable trait of
for addressing the unitary non-invariance that existed the original formulation. Based on the considerations of
in the original PZ formulation. This approach has been this paragraph, we argue strongly for the hypothesis that
the primary means for variational SICs as discussed the FLO-SIC formulation must be further considered and
in recent reviews [23–30]. While the original version that arguments for further modification of the PZ for-
of SIC [20–22] only considered real orbitals, several mulation, presented in [34] based on considerations of
researchers [27,31,32] have noted that there is no a pri- results from many authors, may be a better way to further
ori reason to expect that complex orbitals could not lead achieve the nearly nodeless orbital densities discussed in
to lower SIC energies in the original PZ formulation and [33].
have subsequently derived additional formula for finding In Section 2, the FLO-SIC [34] is reviewed. The results
stationary solutions. Pederson and Perdew noted in [33] from a calculation on a spin-unpolarised porphyrin are
that real localised orbitals which minimise the SIC energy reported in Section 3, with additional remarks about this
are stationary with respect to complex unitary transfor- study in Section 4.
mations but again did not guarantee that the real solu-
tions could not be a saddle point or local maximum. More
recently, a modification of the functional has been made 2. Fermi–Löwdin orbital-based self-interaction
through the use of FLOs for the SIC. correction
While we have stressed the fact that this new version of
In PZ self-interaction-corrected expression for a spin-
SIC has the same functional form as the original Perdew–
polarised system, the localised orbital density can be
Zunger (PZ) SIC but offers formal improvements by con-
represented with both localised {φ i,σ (r)} and canonical
straining the functional to be an explicit density func-
{ψ i,σ (r)} orbitals [20–22]:
tional, it is appropriate to reconsider some past discus-
sions on one of the simplest atoms where the original
formulation leads to multiple stationary solutions for a ρσ (r) = φi,σ (r)2 = ψi,σ (r)2 . (1)
given set of Kohn–Sham orbitals. Except for cases that i i
Based on the complexity associated with the above Table . For the two-orbital case, there are two ways to use Gram–
equations, a new version based upon FLOs [34] which Schmidt to reorthogonalise orbitals but only one way to deter-
mine new orbitals with the Löwdin’s method. The generalisation
satisfies many desirable properties for self-interaction-
to N orbitals is that there would be N! different ways to construct
corrected functionals has been developed. This advan- orthonormal orbitals if Gram–Schmidt is employed.
tageous new formulation [34] of the self-interaction-
Gram–Schmidt Gram–Schmidt Löwdin
corrected density-functional expression, that evades the
requirement for solving the computational demanding φ F F F1 − 21 F1 |F2 F2
localisation equation (Equation (3)), is now discussed. φ F − F |F 〉F F − F |F 〉F F2 − 21 F1 |F2 F1
This expression is explicitly unitarily invariant and size-
extensive [2] because it is based on the FO [36,37], con-
structed from the spin-density matrix, which is explicitly The intermediate Löwdin orbitals (ILO), which struc-
invariant under unitary transformations. Given a trial set turally are similar to molecular orbitals and Bloch
of Kohn–Sham orbitals, a normalised FO can be defined Functions, are eigenstates of the overlap matrix and are
at any point in space aiσ according to therefore constructed from a unitary transformation
on the FOs according to
∗
μ ψμσ (aiσ )ψμσ (r) ρσ (aiσ , r)
Fiσ (r) =
=√ . (4) |Tα = Tα j |Fj (7)
|ψ (a )|2 ρσ (aiσ ))
μ iσ j
Once the FLOs are determined for a given set of Table . Structural and energetic data of flat Dd Fe(II)-porphyrin
descriptors, the energy is minimised by varying these without constraining spin-up and spin-down electrons in the
descriptors. A set of Kohn–Sham orbitals is obtained same orbitals. This is the lowest energy of Ms = electronic con-
figuration, so the relative energy is eV (the first row). The super-
through standard DFT self-consistent field procedure and script on the left of symmetry label is the occupation number and
the positions of descriptors can be determined with the on the right is spin index of occupied electron. The Fe atom is at
methods for optimising molecular geometries. Minimisa- the centre.
tion of the energy in molecules by brute force adjustment E = . eV
of the FO descriptors leads to improved atomisation ener-
gies [2–4,30,34]. However, it was noted that with deriva- Coordinates of atoms (Å)
tives it would be possible to use methods such as conju- N (., ., .)
gate gradients to variationally determine the positions of C (., ., .)
C (., ., .)
FO descriptors. In order to do this, the derivatives of SIC C (., ., .)
energy with respect to the position of FO descriptors is H (., ., .)
H (., ., .)
needed:
Configurations of iron d orbitals (eV)
dE SIC dφk SIC dφk
= Vk φk + φk VkSIC . d↓
z2
.
dam dam dam . d↑ /. d↑ .
k xz yz
. d↓ − .
dφk dφl xy
. d↑ |
=
klk φl − φk , (10) − . |
dam
xy
dam . d↓ /. d↓
xz yz − .
kl
. d↑ − .
z2
with
klk = φl |VkSIC |φk . Due to the fact that the localised
orbitals are constrained to lie in the space of the Kohn– Table . Structural and energetic data of flat Dd Fe(II)-porphyrin
Sham orbitals, this is a generally correct formula and with constraining spin-up and spin-down electrons in the same
there is no requirement that Kohn–Sham orbitals need to orbitals. The relative energy is . eV (the first row). The super-
script on the left of symmetry label is occupation number and on
be self-consistent solutions of a Hamiltonian. A complete
the right is spin index of occupied electron.
derivation of the FO derivatives may be found in [30,34].
The calculations reported here were performed with the E = . eV
NRLMOL suite of codes [39]. Coordinates of atoms (Å)
relocated to somewhere on the ring. This implies that [6] I. Lindgren, Int. J. Quant. Chem. 5(S5), 411–420 (1971).
the centred iron atom is +2 charged. These observations [7] J.F. Janak, Phys. Rev. B 18(12), 7165–7168 (1978).
imply that the behaviour of FO descriptors coincide with [8] J.P. Collman, J.L. Hoard, N. Kim, G. Lang, and C.A. Reed,
J. Am. Chem. Soc. 97, 2676–2681 (1975).
chemical intuition. [9] H. Goff, G.N. La Mar, and C.A. Reed, J. Am. Chem. Soc.
99, 3641–3646 (1977).
[10] M.S. Liao and S. Scheiner, J. Chem. Phys. 117, 205–219
4. Summary (2002).
In this paper, the FLO-based SIC is applied to the Ms = [11] H. Nakashima, J.-Y. Hasegawa, and H. Nakatsuji, J. Com-
put. Chem. 27(4), 426–433 (2006).
0 Fe(II)-porphyrin to demonstrate that this new theoret-
[12] C.A. Reed, T. Mashiko, W.R. Scheidt, K. Spartalian, and
ical method is viable for large molecules containing tran- G. Lang, J. Am. Chem. Soc. 102, 2302–2306 (1980).
sition metal systems. FOs depend explicitly on the den- [13] E. Sigfridsson and U. Ryde, J. Inorg. Biochem. 91, 101–
sity matrix and are therefore unitarily invariant highly 115 (2002).
localised orbitals. Use of Löwdin symmetric orthonor- [14] S. Paul, F. Amalraj, and S. Radhakrishnan, Synth. Met.
malisation preserves size-extensivity and the transferabil- 159(11), 1019–1023 (2009).
[15] C. She, S.J. Lee, J.E. McGarrah, J. Vura-Weis, M.R.
ity of descriptors. In this study, we show that the SIC is Wasielewski, H. Chen, G.C. Schatz, M.A. Ratner, and J.T.
able to lower the energy and resolve the fractional occu- Hupp, Chem. Commun. 46(4), 547–549 (2010).
pation. The similarity between FLO descriptors and the [16] D. Danilovic, C.L. Lin, T. Yuen, L. Pan, and J. Li, J. Appl.
classical chemical bonds is explored. In addition, aro- Phys. 101(9), 09E103 1–2 (2007).
maticity of Fe(II)-porphyrin ring is interpreted by the [17] E.I. Ioannidis and H.J. Kulik, J. Chem. Phys. 143(3),
034104 1–11 (2015).
hovering FLO descriptors.
[18] A. Svane, L. Petit, Z. Szotek, and W.M. Temmerman, Phys.
Rev. B 76(11), 115116 (2007).
[19] C. Edmiston and K. Ruedenberg, Rev. Mod. Phys. 35,
Acknowledgments 457–465 (1963).
Mark R. Pederson acknowledges the long-standing support [20] M.R. Pederson and C.C. Lin, J. Chem. Phys. 88(3), 1807–
from NRL which allowed development of NRLMOL. Der-you 1817 (1988).
Kao was supported by a fellowship at The George Washing- [21] M.R. Pederson, R.A. Heaton, and C.C. Lin, J. Chem. Phys.
ton University Institute of Nanotechnology. The authors thank 80(5), 1972–1975 (1984).
Drs Torsten Hahn, Jens Kortus, Tunna Baruah, Koblar Jackson, [22] M.R. Pederson, R.A. Heaton, and C.C. Lin, J. Chem. Phys.
and Mr Simon Liebing for many interesting discussions. 82(6), 2688–2699 (1985).
[23] J.P. Perdew, A. Ruzsinszky, J. Sun, and M.R. Pederson, in
Advances in Atomic, Molecular, and Optical Physics, edited
Disclosure statement by E. Arimondo C.C. Lin S.F. Yelin (Burlington Academic
Press, 2015), Vol. 64, Chap. 1, pp. 1–14.
No potential conflict of interest was reported by the authors. [24] K.A. Jackson, in Advances in Atomic, Molecular, and Opti-
cal Physics, edited by E. Arimondo C.C. Lin S.F. Yelin
Funding (Burlington Academic Press, 2015), Vol. 64, Chap. 2,
pp. 15–27.
This work was supported in part by Office of Naval Research [25] A. Pertsova, C.M. Canali, M.R. Pederson, I. Rungger, and
(ONR) [N00014-16-1-2464]; and The George Washington Uni- S. Sanvito, in Advances in Atomic, Molecular, and Opti-
versity Institute of Nanotechnology. cal Physics, edited by E. Arimondo C.C. Lin S.F. Yelin
(Burlington Academic Press, 2015), Vol. 64, Chap. 3,
pp. 29–86.
ORCID [26] P.M. Dinh, P. Reinhard, E. Suraud, and M. Vincendon, in
Advances in Atomic, Molecular, and Optical Physics, edited
Mark R. Pederson http://orcid.org/0000-0003-2489-4849 by E. Arimondo C.C. Lin S.F. Yelin (Burlington Academic
Press, 2015), Vol. 64, Chap. 4, pp. 87–103.
[27] N. Poilvert, G. Borghi, N.L. Nguyen, N.D. Keilbart,K.
References Wang, and I. Dabo, in Advances in Atomic, Molecular,
and Optical Physics, edited by E. Arimondo C.C. Lin S.F.
[1] J.P. Perdew and A. Zunger, Phys. Rev. B 23(10), 5048–
Yelin (Burlington Academic Press, 2015), Vol. 64, Chap.
5079 (1981).
5, pp. 105–127.
[2] M.R. Pederson, A. Ruzsinszky, and J.P. Perdew, J. Chem.
[28] N. Gidopoulos and N.N. Lathiotakis, in Advances in
Phys. 140(12), 121103 1–4 (2014).
Atomic, Molecular, and Optical Physics, edited by E. Ari-
[3] T. Hahn, S. Liebing, J. Kortus, and M.R. Pederson,
mondo C.C. Lin S.F. Yelin (Burlington Academic Press,
J. Chem. Phys. 143(22), 224104 1–7 (2015).
2015), Vol. 64, Chap. 6, pp. 129–142.
[4] M.R. Pederson, T. Baruah, D. Kao, and L. Basurto,
[29] S. Kümmel, in Advances in Atomic, Molecular, and Opti-
J. Chem. Phys. 144(16), 164117 1–8 (2016).
cal Physics, edited by E. Arimondo C.C. Lin S.F. Yelin
[5] J.P. Perdew and Y. Wang, Phys. Rev. B 45, 13244–13249
(Burlington Academic Press, 2015), Vol. 64, Chap. 7,
(1992).
pp. 143–151.
8 D.-Y. KAO AND M. R. PEDERSON
[30] M.R. Pederson and T. Baruah, in Advances in Atomic, [35] J. Sun, A. Ruzsinszky, and J.P. Perdew, Phys. Rev. Lett. 115,
Molecular, and Optical Physics, edited by E. Arimondo 036402 1–6 (2015).
C.C. Lin S.F. Yelin (Burlington Academic Press, 2015), [36] W. Luken and D.N. Beratan, Theor. Chim. Acta 61(3),
Vol. 64, Chap. 8, pp. 153–180. 265–281 (1982).
[31] S. Klüpfel, P. Klüpfel, and H. Jónsson, Phys. Rev. A. 84(5), [37] W. Luken and J.C. Culberson, Theor. Chim. Acta 66(5),
050501R 1–4 (2011). 279–283 (1984).
[32] S. Lehtola and H. Jónsson, J. Chem. Theory Comput. [38] B.C. Carlson and J.M. Keller, Phys. Rev. 105(1), 102–106
10(12), 5324–5337 (2014). (1957).
[33] M.R. Pederson and J.P. Perdew, K Newsletter Scientific [39] M.R. Pederson, D.V. Porezag, J. Kortus, and D.C. Patton,
Highlight of the Month (February, 2012). <http://www. Phys. Stat. Solidi B 217, 197–218 (2000).
psi-k.org/newsletters/News_109/Highlight_109.pdf >. [40] D. Kao, M.R. Pederson, T. Baruah, T. Hahn, S. Liebing,
[34] M.R. Pederson, J. Chem. Phys. 142(6), 064112 1–7 (2015). and J. Kortus (submitted).