lda vs gga vs mgga
TRANSCRIPT
-
7/29/2019 lda vs gga vs mgga
1/25
Comparison of
exchange-correlation functionals:
from LDA to GGA and beyond
Martin Fuchs
Fritz-Haber-Institut der MPG, Berlin, Germany
Density-Functional Theory Calculations for Modeling Materials and Bio-Molecular Properties and Functions - A Hands-On
Computer Course, 30 October - 5 November 2005, IPAM, UCLA, Los Angeles, USA
-
7/29/2019 lda vs gga vs mgga
2/25
Density-functional theory
DFT is an exact theory of the ground state of an interacting many-particle system:
E0 = minnNEv[n] E0 = minN|Hv| by Hohenberg-Kohn theorem: v(r) v[n(r)]
Total energy density functional (electrons: w(r, r) = 1/|r r|)
Ev[n] = minn|
T +
W| +Z
n(r
)v(r
)d = F[n] +Z
n(r
)v(r
) d
Non-interacting case: w(r, r) 0
F[n] minn|T| = [n]|T|[n] = Ts[n] [n] is a Slater determinant
Interacting case rewritten:
Ev[n] = Ts[n] + Exc[n] + 12
Zn(r)w(r, r)n(r)dd +
Zn(r)v(r) d
XC energy functional: kinetic correlation energy & exchange and Coulomb correlation energy
Exc[n] =: [n]|T|[n] Ts[n] + [n]|W|[n] 12R
n(r)w(r, r)n(r)dd
-
7/29/2019 lda vs gga vs mgga
3/25
Kohn-Sham scheme
Minimization of Ev[n] with n(r) N determines groundstate:
Ev[n]n(r)
= F[n]n(r)
+ v(r) = Ts[n]n(r)
+ vKS([n]; r) =
. . . effective non-interacting system Kohn-Sham independent-particle equations:
h
2
2 + vKS(r)i
i(r) = ii(r) with vKS
(r) =Z
w(r, r
)n(r
)d
+Exc[n]
n(r) + v(r) ,
density n(r) =P
i fi|i(r)|2,
occupancies i < : fi = 1, i = : 0 fi 1, i > : fi = 0 (aufbau principle).
non-interacting kinetic energy treated exactly Ts[n] =RP
i fi|i(r)|2d
classicalelectrostatics (e-n & e-e) treated exactly
onlyneed to approximate Exc[n] and vxc([n]; r) =Exc[n]n(r) determines accuracy in practice!
applied is spin density functional theory: n, treated as separate variables
beware: vKS(r) is a local operator. Direct approximation of whole F[n] possible too: DFT with
beware: generalized Kohn-Sham schemes1 and nonlocal effective potentials, e.g. Hartree-Fock eqs.
1 Seidl, G orling, Vogl, Majewski, Levy, Phys Rev B 53, 3764 (1996).
-
7/29/2019 lda vs gga vs mgga
4/25
Local-density approximation
ELDAxc [n] =
Zn(r)e
homxc (n(r))d
0.1 1.0 10.0 100.0
rs
(bohr)
30.0
20.0
10.0
0.0
exc
(Ry)
local dependence on density or Wigner-Seitz radiusrs = (
43 n)
1/3
XC energy per electron is that of the homogeneous electron gas,exc[n] = e
homxc (n)|n=n(r)
exchange part ehomx (n) known analytically correlation part ehomc (n) known
analytically for rs 0 and rs numericallyexact for 2 < rs < 100 from QMC data
1
. . . as parametrization interpolating over all rsPW91,Perdew-Zunger,VWN2
workhorse in DFT applications to solids
real systems are far from jellium-like homogeneity
. . . why does LDA work at all?
. . . where does it fail?
. . . how to improve beyond it?
1 Ceperley, Alder, Phys Rev Lett 45, 566 (1980).2 Perdew, Wang, Phys Rev B 45, 13244 (1992); Perdew, Zunger (1980); Vosko, Wilk, Nussair (1980).
-
7/29/2019 lda vs gga vs mgga
5/25
Performance of the LDA
structural, elastic, and vibrational properties often good enough crystal bulk lattice constants accurate to within 3%, usually underestimated
bulk moduli somewhat too large, > 10% error not uncommon for d-metals phonons somewhat too stiff
binding energies are too negative (overbinding), up to several eV# cohesive energies of solids, but formation enthalpies often o.k.
# molecular atomization energies, mean error (148 molecules G2 set) 3 eV
activation energies in chemical reactions unreliable# too small/absent, e.g. for H2 on various surfaces (Al, Cu, Si, ...)
relative stability of crystal bulk phases can be uncertain
# SiO2 high pressure phase more stable than zero pressure phase# underestimated transition pressure e.g. for diamond -tin phase transitions in Si & Ge# magnetic phases
electronic structure can be usefully interpreted (density of states, band structures),except for band gaps (a more fundamental issue than LDA!)
-
7/29/2019 lda vs gga vs mgga
6/25
View on XC through the XC hole
Definition of the XC energy by a coupling constant integration of the e-e interaction1:
E([n]; ) = [n]|
T +V +
W|[n] = minn ...
by Hohenberg-Kohn theorem: n(r) v([n]; r) for any 0 1
= 0 non-interacting case/ Kohn-Sham potential, v=0([n]; r) = vKS([n]; r)
= 1 external potential, v=1([n]; r) = v(r)
groundstate energy: E([n]; 1) = E([n]; 0) +
Z10
d E([n]; )
dd
Exc[n] = R1
0[n]|W|[n] d
12 Rn(r)w(r, r
)n(r)dd
X, XX =1
2
Zw(r, r
)n(r)n(r
)[n]|
Pp,q
p(r)
q(r
)q(r)p(r)|[n]
n(r)n(r)dd
XC in terms of the pair correlation function g([n];
r
,r
)
XC in terms of the adiabatic connection integrand
(scaled e-e potential energy)
may distinguish exchange and correlation, e.g. Exc = Ex + Ec
g(r, r)r
r
1 cf. DFT books by Dreizler and Gross & Yang and Parr
-
7/29/2019 lda vs gga vs mgga
7/25
. . . coupling constant averaged XC hole
-integration implies coupling constant averaged pair correlation function
g([n]; r, r) =
Z10
g([n]; r, r)d
Identify the XC energy as
Exc[n] =1
2
Zd n(r)
Zd
n(r
){g([n]; r, r
) 1} w(r, r
)
interpretation: the electron density n(r) interacts with the
electron density of the XC holenxc([n]; r, r
) = n(r){g([n]; r, r) 1}
Pauli exclusion principle & Coulomb repulsion
local density approximation corresponds to
nxc([n]; r, r) = n(r){g
hom(n(r); |r r|) 1}
always centered at reference electron & spherical 0 1 2 3 4 5 6
0.5
0.0
XCh
ole
(n)
exchange
+ correlation
homogeneous electron gas
kF|r r|
(NB: correlation part varies with kF)
-
7/29/2019 lda vs gga vs mgga
8/25
From the XC hole to the XC energy
Focus on Exc[n] = Rn(r)exc([n]; r)d (... component of the total energy) w =1
|rr|= 1|u|
for exc only the angle averaged XC hole matters (and LDA hole is always spherical)
nxc(r, r + u) =Xlm
nxclm(r, u)Ylm(u)
Exc[n] =1
2
Zn(r)
Znxc(r, r + u)
udu d
1
2
Zn(r)
Z0
nxc00(r, u)
uu2dud
for Exc only the system & angle averaged XC hole matters
nxc(u) =1
N
Zn(r)n
xc00(r, u)d
XC energy in terms of averaged XC hole1
Exc[n] =N
2
Z0
nxc(u)
uu2
du = N average XC energy per electron
LDA & GGA approximate average holes rather closely work mostly o.k.
1 Perdew et al., J Chem Phys 108, 1552 (1998).
-
7/29/2019 lda vs gga vs mgga
9/25
-
7/29/2019 lda vs gga vs mgga
10/25
Generalized Gradient Approximation for Exc
Gradient expansion of XC energy and (later) hole: nxc([n]; r, u) nxc(r, u = 0) +unxc|u=0 + ...,and generalized to imposing constraints to meet
e.g. nx(r, u) < 0,R
nx(r, u)dr = 1, by real space cutoffs numerical GGA scaling relations and bounds on Exc by analytic approximation to numerical GGA parameter-free GGA by Perdew-Wang PW91
simplified in PBE GGA
contains LDA & retains all its good features
Earlier: analytic model or ansatz + empirical parameter(s)
Langreth-Mehl (C), Becke 86(X) + Lee-Yang-Parr(C) BLYP, ...
Alternative: analytic ansatz + (many) fitted parameters (e.g. fit to thermochemical data)
see accuracy limit of GGA functionals, can be better & worse than PBE
Generic GGA XC functional
EGGAxc [n] = Z n eLDAx (n) FGGAxc (n, s) n=n(r)d, s =
|n|
2kFnn=n(r)
Enhancement factor Fxc(n, s) over LDA exchange: function of density and scaled gradientunderstandinghow GGAs work
Calculations with GGAs are not more involved than with LDA,except that vxc[n; r] = vxc(n,in,ijn)|n(r)
-
7/29/2019 lda vs gga vs mgga
11/25
Differences in present GGAs
PBE
0 1 2 31.0
1.2
1.4
1.6
1.8
rs=100
50
10
5
PBE XC
rs
=0
1
enhancementfactor
scaled gradient
Zhang et al. revised revPBE
0 1 2 31.0
1.2
1.4
1.6
1.8
rs=100
50
10
5
1 Zhang
enhancementfactor
scaled gradient
BLYP
0 1 2 31.0
1.2
1.4
1.6
1.8
rs=100
50
10
5
1 BLYP
enh
ancementfactor
scaled gradient
. . .
WC-PBE Wu & Cohen 05
xPBE Xu & Goddard 04
HCTH Handy et al. 01
RPBE Hammer et al. 99
revPBE Zhang et al. 98PBE GGA Perdew et al. 96
PW91 Perdew & Wang 91
PW91 Perdew & Wang 91
BP Becke & Perdew 88
BLYP Becke & Lee et al. 88. . .
revPBE & BLYP more nonlocal than PBE GGA
molecules: more accurate atomization energies
solids: bondlengths too large, cohesive energies too small
LYP correlation incorrect for jellium
more local GGAs will make lattice constants smaller(e.g. Tinte et al. PRB 58, 11959 (1998)),
but make binding energies more negative
h GGA h l i
-
7/29/2019 lda vs gga vs mgga
12/25
... how GGAs change total energiesAnalysis in terms of selfconsistent LDA density shows how GGAs work:
EGGAtot [n
GGA] = E
LDAtot [n
LDA] + E
GGAxc [n
LDA] E
LDAxc [n
LDA] +O(n
GGA nLDA)2
Spectral decomposition in terms of s:
EGGAxc E
LDAXC =
Z Zs=0
heGGAxc (r) e
LDAxc (r)i (s s(r))dds
0 1 2 3 4
scaled gradient s
0.05
0.00
0.05
energyperatom(
a.u.)
correlation
exchange
PBE rev PBE
X+C
Al fcc
0 1 2 3 4
scaled gradient s
0.5
0.2
0.0
energyperatom
(eV)
PBE
revPBE
differential
integrated
fcc Al
-0.3 (0.5) eV
-0.5 (0.7) eV
only 0 s 4 contribute, similar analysis can be made for n(r)
for more see e.g. Zupan et al., PRB 58, 11266 (1998)
C h i i i GGA
-
7/29/2019 lda vs gga vs mgga
13/25
Cohesive properties in GGA
Bulk lattice constants GGA increase due tomore repulsive core-valence XC.
Na
NaC
l
Al C Si Ge
SiC
AlAs
GaA
s
Cu
5
0
5
erro
r(%)
PW91GGA
inconsistent GGA
LDA
expt.
Cohesive energies GGA reduction mostlyvalence effect.
Na
NaC
l
Al C Si Ge
SiC
AlAs
GaA
s
Cu
0.5
0.0
0.5
1.0
1.5
erro
r(eV)
PW91GGA
inconsistent GGALDA
expt.
for comparison of LDA, GGA, and Meta-GGA see Staroverov, Scuseria, Perdew, PRB 69, 075102 (2004)
b i H C (111)
-
7/29/2019 lda vs gga vs mgga
14/25
... energy barriers: H2 on Cu(111)
barrier to dissociative adsorption
- PW91-GGA 0.7 eV- LDA
-
7/29/2019 lda vs gga vs mgga
15/25
Phase transition -tin - diamond in Si
diamond-structure
-tin structure
a a
c
diamond: a0 = 10.26 bohr, semiconductor
-tin: c/a = 0.552, metallic
Spectral decomposition
in terms of n
0 1 2 3 4
Density parameter rs
[bohr]
0.0
0.5
1.0
1.5
2.0
Rela
tivevolume/r
s
-tin
diamond
Phase transition tin diamond in Si
-
7/29/2019 lda vs gga vs mgga
16/25
Phase transition -tin - diamond in Si
Gibbs construction: Et + ptVt = E
diat + ptV
diat , p
expt = 10.3 . . . 12.5 GPa
0.6 0.7 0.8 0.9 1.0 1.1
relative volume V/Vexp
0.0
0.2
0.4
0.6
energy(eV/atom)
tin diamond
LDA
GGA
0.2 0.3 0.4 0.5 0.6
energy change (eV/atom)
5
10
15
20
25
transitionpressure(GPa)
LDA
PBE
revPBE
PW91
BP
BLYP
expt
GGA increases transition pressure,inhomogeneity effect use of LDA-pseudopotentials insufficient
Moll et al, PRB 52, 2550 (1995); DalCorso et al, PRB 53, 1180 (1996); McMahon et al, PRB 47, 8337 (1993).
Performance of PBE GGA vs LDA
-
7/29/2019 lda vs gga vs mgga
17/25
Performance of PBE GGA vs. LDA
atomic & molecular total energies are
improved
GGA corrects the LDA overbinding:average error for G2-1 set of molecules:
+6 eV HF
1.5 eV LSDA0.5 eV PBE-GGA
0.2 eV PBE0 hybrid0.05 eV goal,
better cohesive energies of solids
improved activation energy barriers in
chemical reactions (but still too low) improved description of relative stability of
bulk phases
more realistic for magnetic solids
useful for
electrostatic
hydrogen bonds
# GGA softens the bonds
increasing lattice constants
decreasing bulk moduli no consistent improvement
LDA yields good relative bond energies forhighly coordinated atoms,
e.g. surface energies, diffusion barriers on surfaces
GGA favors lower coordination (larger gradient!),not always enough where LDA has a problem,
e.g. CO adsorption sites on transition metal surfaces
significance of GGA ?
# GGA workfunctions for several metals turn out somewhatsmaller than in LDA
one-particle energies/bands close to LDA
# Van der Waals (dispersion) forces not included!
Comparing LDA and (different) GGAs gives an idea about possible errors!
Beyond GGA: orbital dependent XC functionals
-
7/29/2019 lda vs gga vs mgga
18/25
Beyond GGA: orbital dependent XC functionals
Kohn-Sham non-interacting system make i[n] density-functionals: n vKS[n; r] i
LDA and GGA are explicit density-functionals Exc[n]
Implicit density-functionals formulated in terms of i[n]? more flexible for further improvements
self-interaction free: Exc[n]|N=1 = 0,vxc(|r| ) =
1r ,
nonspherical nxc(r, r) ... ?
Start with exact exchange :
Exchange-energy Ex[{i}] =1
2
ZPi,j i(r)j(r
)i(r)j(r)
|r r|dd
v same as in Hartree-Fock, but 2
2 vKS(r)|i(r) = ii(r) KS and HF orbitals different!
Groundstate ...
in KS-DFT: optimized effective potential OEP method, local KS potential
in Hartree-Fock: variation with respect to orbitals yield HF eqs., nonlocal effective potential
EHF0 EKSEXX0
Hybrid functional: mixing exact exchange with LDA/GGA
-
7/29/2019 lda vs gga vs mgga
19/25
Hybrid functional: mixing exact exchange with LDA/GGA
XC = EXX + LDA correlation is less accurate (underbinds) than XC = LDA for molecules
challenge: correlation functional that is compatible with exact exchange?
hybrid functionals tointerpolateadiabatic connection, Uxc[n]
Exc[n] =
Z10
Uxc[n]() d Uxc[n]() = [n]|W|[n] UHartree[n]
molecular dissociation: Uxc = molecules - atoms ... looks like
couplingstrength
Uxc()10
Ex
Ec
Tc
=0 = KSKS system
=1physical system
LDA,
GGA
good
LDA,GGA
too negative
. . . hybrid functionals
-
7/29/2019 lda vs gga vs mgga
20/25
. . . hybrid functionalsHow they are defined:
use exact exchange for = 0and alocal functional for = 1
Hybrid functional = interpolation
Ehybxc = E
GGAxc + a nEx EGGAx o
mixing parameter a = 0.16 . . . 0.3from fitting thermochemical data
a = 1/4 by 4th order perturbation theory PBE0 = PBE1PBE
B3LYP 3-parameter combination ofBecke X-GGA, LYP C-GGA, and LDA
molecular dissociation energies on averagewithin 3 kcal/mol 0.1 eV(but 6 times larger errors can happen)
How hydbrids and GGAs work:
Adiabatic connection N2 2N
0 0.5 1
coupling strength
-6
-4
-2
0
2
U
xc
()
(eV)
exchange only
PBE GGA
PBE0 hybrid
"exact"
XC contributions to binding:
(eV) Eb Ex Ec
PBE -10.58 -1.82 -2.49PBE0 -9.62 -1.01 -2.49
exact -9.75 1.39 -5.07
... X and C errors tend to cancel!
Becke, J Chem Phys 98, 5648 (1993); Perdew et al, J Chem Phys 105, 9982 (1996).comprehensive comparison: Staroverov, Scuseria, Tao, Perdew, J Chem Phys 119, 12129 (2003)
XC revisited: role of error cancellation between X and C
-
7/29/2019 lda vs gga vs mgga
21/25
XC revisited: role of error cancellation between X and C
see e.g. Baerends and Gritsenko, JCP 123, 062202 (2005)
... or: when GGAs or hybrids are not enough
... orbital dependent functionals
-
7/29/2019 lda vs gga vs mgga
22/25
... orbital dependent functionals
Meta-GGAs
XC energy functional
EMGGAxc [n] =
Zn(r)e
MGGAxc (n,n, ts)
n(r)
d
kinetic energy density of non-interacting electrons ts(r) =12
Pocci fi|i(r)|
2
OEP or HF style treatment ofEMGGAxc [i]
i(r)= ...
1
2
exc
tsi
TPSS: Tao, Perdew, Stavroverov, Scuseria, Phys Rev Lett 91, 146401 (2003): + vx finite
PKZB: Perdew, Kurth, Zupan, Blaha, Phys Rev Lett 82, 2544 (1999): XCnon-empirial, LDA limitVan Voorhis, Scuseria, J Chem Phys 109, 400 (1998): XChighly fitted, no LDA limit
Colle, Salvetti, Theoret Chim Acta 53, 55 (1979): C, no LDA limit BLYP GGA
Becke, J Chem Phys 109, 2092 (1998): XC + exact exchangefitted
TPSS accomplishes a consistent improvement over (PBE) GGA
PKZB improved molecular binding energy, but worsened bond lengths in molecules & solids
hybrid functionals on average still more accurate for molecular binding energies
TPSS provides sound, nonempirical basis for new hybridsTPSSh
next step: MGGA correlation compatible with exact exchange?
ACFDT XC: including unoccupied Kohn-Sham states
-
7/29/2019 lda vs gga vs mgga
23/25
C XC c ud g u occup ed Ko S a states
Adiabatic connection: KS system = 0 physical system = 1
Exc[n] = R1
0 |Wee|d U[n]
Fluctuation-dissipation theorem:
Wxc() =: |Wee| =1
2
Ze2
|r r|
Z0
(iu, r, r) du n(r)(r r
)
drdr
...using dynamical density response
From noninteracting Kohn-Sham to interacting response by TD-DFT
0(iu, r, r) = 2P
i,j
i
(r)
j
(r)
j
(r)
i
(r)
iu(ji) i[n] . . . KS eigenvaluesi([n], r) ...KS orbitals
(iu) = 0(iu) + 0(iu) Khxc (iu) (iu) Dyson equation6dim.
. . . using Coulomb and XC kernel from TD-DFT
In principle ACFDT formula gives exact XC functional
In practice starting point for fully nonlocal approximations
RPA: Khxc = |r r|1 and zero XC kernel ... yields exact exchange and London dispersion
forces combine with XC kernels, hybrids with usual XC functionals, split Coulomb interaction . . .
Status of RPA type functionals
-
7/29/2019 lda vs gga vs mgga
24/25
yp
Molecular dissociation energies
RPA for all e-e separations
H 2 N 2 O 2 F 2 Si 2 HF CO CO
2
C 2H 2
H 2O
-2
-1
-0.5
0
0.5
1
2
errorin
dissociation
energy
(eV)
RPA+
RPA
PBE0 hyb
PBE GGA
... no consistent improvement over GGA
... not too bad, without X and C error cancellation
... better TD-DFT XC kernels needed!
Furche, PRB 64, 195120 (2002),
Fuchs and Gonze, PRB 65, 235109 (2002)
Stacked benzene dimer
(Van der Waals complex)short-range: GGA + long-range: RPA
3 3.5 4 4.5
Separation (A)
4
3
2
1
0
1
2
3
4
5
In
teractionenergy(kcal/mol)
CCSD(T)
MP2
vdWDF
GGA(revPBE)GGA(PW91)
... includes dispersion forces!
Dion et al., PRL 92, 246401 (2004)
Summary
-
7/29/2019 lda vs gga vs mgga
25/25
y
LDA & GGA are de facto controlled approximations to the average XC hole
GGA remedies LDA shortcomings w.r.t. total energy differences but may also overcorrect(e.g. lattice parameters)
still can & should check GGA induced corrections for plausibility by
. . . simple arguments like homogeneity & coordination
...results fromquantum chemicalmethods (Quantum Monte Carlo, CI, . . . )
. . . depends on actual GGA functional
hybrid functionals mix in exact exchange (B3LYP, PBE0, ... functionals)
orbital dependent, implicit density functionals:exact Kohn-Sham exchange, Meta-GGA & OEP method,functionals from the adiabatic-connection fluctuation-disspation formula
Always tell what XC functional you used,e.g. PBE-GGA (not just GGA)... helps others to understand your results
... helps to see where XC functionals do well or have a problem