Overcoming the logic of the restrictions option

asked by Marius Retegan (2019/12/15 14:26)

I am trying to overcome the AND-only logic implemented in the restrictions option by tuning the basis set used to express the wavefunctions. The script below, which is a modified version of the ground state ligand-field calculation, uses first the standard way to get the determinants needed to express the wavefunctions, while in the second part I get the wavefunctions in a two-step fashion. This is the output for the expectation values:

Oppldots = NewOperator("ldots", NF, IndexUp, IndexDn)
  #    <E>      <S^2>    <L^2>    <J^2>    <S_z^3d> <L_z^3d> <l.s>    <F[2]>   <F[4]>   <Neg^3d> <Nt2g^3d><Neg^Ld> <Nt2g^Ld><N^3d>
  1   -3.395    1.999   12.000   15.147   -0.905   -0.280   -0.319   -1.043   -0.925    2.189    5.989    3.823    6.000    8.178
  2   -3.395    1.999   12.000   15.147   -0.000   -0.000   -0.319   -1.043   -0.925    2.189    5.989    3.823    6.000    8.178
  3   -3.395    1.999   12.000   15.147    0.905    0.280   -0.319   -1.043   -0.925    2.189    5.989    3.823    6.000    8.178

  #    <E>      <S^2>    <L^2>    <J^2>    <S_z^3d> <L_z^3d> <l.s>    <F[2]>   <F[4]>   <Neg^3d> <Nt2g^3d><Neg^Ld> <Nt2g^Ld><N^3d>
  1   -3.320    1.999   12.000   15.214   -0.916   -0.298   -0.338   -1.039   -0.918    2.167    5.987    3.846    6.000    8.154
  2   -3.320    1.999   12.000   15.214   -0.000   -0.000   -0.338   -1.039   -0.918    2.167    5.987    3.846    6.000    8.154
  3   -3.320    1.999   12.000   15.214    0.916    0.298   -0.338   -1.039   -0.918    2.167    5.987    3.846    6.000    8.154

I would be grateful if someone could tell me what I need to change to get the same expectation values as in the standard calculation. Thank you. Here is the full script.

NiO_LF.Quanty
Verbosity(0)
-- This tutorial calculates the ground-state of NiO within the Ligand-field theory approximation
 
-- in Ligand field theory we approximate the solid by a single transition metal atom d-shell
-- interacting with a non-interacting Ligand shell. (Nowadays in the literature often called
-- a bath) For transition metal oxides one can think of the ligand orbitals as the O-2p
-- orbitals. For NiO there would be six O-2p orbitals and one might expect a cluster of
-- 1 Ni-d shell and 6 O-2p shells (10+36=46 spin-orbitals in total). For a theory where
-- calculation times scale roughly exponential with respect to number of orbitals going from
-- 10 spin-orbitals (crystal-field theory) to 46 spin-orbitals slows thing down a lot.
 
-- There is however a simple optimization one can make to speed up the calculations, without
-- changing the final answer. One can make linear combinations of the O-2p orbitals to form
-- ligand orbitals. Out-off the 36 O-2p orbitals only 10 interact with the Ni-d orbital.
-- (see PRB 85, 165113 (2012) for nice pictures of the ligand orbitals in cubic symmetry or
-- PRL 107, 107402 (2011) and J. Phys. Condens. Matter 24, 255602 (2012) for an example in
-- lower symmetry (TiOCl))
 
-- In ligand field theory we thus have 20 spin-orbitals. 10 representing the Ni-3d shell and
-- 10 representing the Ligand-d shell.
 
-- we again take the ordering to be dn even and up odd
 
NF=20
NB=0
IndexDn_3d={ 0, 2, 4, 6, 8}
IndexUp_3d={ 1, 3, 5, 7, 9}
IndexDn_Ld={10,12,14,16,18}
IndexUp_Ld={11,13,15,17,19}
 
-- we can define the angular momentum operators for the d-shell as we did in crystal-field
-- theory
 
OppSx_3d   =NewOperator("Sx"   ,NF, IndexUp_3d, IndexDn_3d)
OppSy_3d   =NewOperator("Sy"   ,NF, IndexUp_3d, IndexDn_3d)
OppSz_3d   =NewOperator("Sz"   ,NF, IndexUp_3d, IndexDn_3d)
OppSsqr_3d =NewOperator("Ssqr" ,NF, IndexUp_3d, IndexDn_3d)
OppSplus_3d=NewOperator("Splus",NF, IndexUp_3d, IndexDn_3d)
OppSmin_3d =NewOperator("Smin" ,NF, IndexUp_3d, IndexDn_3d)
 
OppLx_3d   =NewOperator("Lx"   ,NF, IndexUp_3d, IndexDn_3d)
OppLy_3d   =NewOperator("Ly"   ,NF, IndexUp_3d, IndexDn_3d)
OppLz_3d   =NewOperator("Lz"   ,NF, IndexUp_3d, IndexDn_3d)
OppLsqr_3d =NewOperator("Lsqr" ,NF, IndexUp_3d, IndexDn_3d)
OppLplus_3d=NewOperator("Lplus",NF, IndexUp_3d, IndexDn_3d)
OppLmin_3d =NewOperator("Lmin" ,NF, IndexUp_3d, IndexDn_3d)
 
OppJx_3d   =NewOperator("Jx"   ,NF, IndexUp_3d, IndexDn_3d)
OppJy_3d   =NewOperator("Jy"   ,NF, IndexUp_3d, IndexDn_3d)
OppJz_3d   =NewOperator("Jz"   ,NF, IndexUp_3d, IndexDn_3d)
OppJsqr_3d =NewOperator("Jsqr" ,NF, IndexUp_3d, IndexDn_3d)
OppJplus_3d=NewOperator("Jplus",NF, IndexUp_3d, IndexDn_3d)
OppJmin_3d =NewOperator("Jmin" ,NF, IndexUp_3d, IndexDn_3d)
 
Oppldots_3d=NewOperator("ldots",NF, IndexUp_3d, IndexDn_3d)
 
-- And similar we can define the angular momentum operators for the ligand d-shell
 
OppSx_Ld   =NewOperator("Sx"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppSy_Ld   =NewOperator("Sy"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppSz_Ld   =NewOperator("Sz"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppSsqr_Ld =NewOperator("Ssqr" ,NF, IndexUp_Ld, IndexDn_Ld)
OppSplus_Ld=NewOperator("Splus",NF, IndexUp_Ld, IndexDn_Ld)
OppSmin_Ld =NewOperator("Smin" ,NF, IndexUp_Ld, IndexDn_Ld)
 
OppLx_Ld   =NewOperator("Lx"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppLy_Ld   =NewOperator("Ly"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppLz_Ld   =NewOperator("Lz"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppLsqr_Ld =NewOperator("Lsqr" ,NF, IndexUp_Ld, IndexDn_Ld)
OppLplus_Ld=NewOperator("Lplus",NF, IndexUp_Ld, IndexDn_Ld)
OppLmin_Ld =NewOperator("Lmin" ,NF, IndexUp_Ld, IndexDn_Ld)
 
OppJx_Ld   =NewOperator("Jx"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppJy_Ld   =NewOperator("Jy"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppJz_Ld   =NewOperator("Jz"   ,NF, IndexUp_Ld, IndexDn_Ld)
OppJsqr_Ld =NewOperator("Jsqr" ,NF, IndexUp_Ld, IndexDn_Ld)
OppJplus_Ld=NewOperator("Jplus",NF, IndexUp_Ld, IndexDn_Ld)
OppJmin_Ld =NewOperator("Jmin" ,NF, IndexUp_Ld, IndexDn_Ld)
 
-- In order to calculate the total angular momentum in the cluster we can sum the operators
 
OppSx = OppSx_3d + OppSx_Ld
OppSy = OppSy_3d + OppSy_Ld
OppSz = OppSz_3d + OppSz_Ld
OppSsqr = OppSx * OppSx + OppSy * OppSy + OppSz * OppSz
OppLx = OppLx_3d + OppLx_Ld
OppLy = OppLy_3d + OppLy_Ld
OppLz = OppLz_3d + OppLz_Ld
OppLsqr = OppLx * OppLx + OppLy * OppLy + OppLz * OppLz
OppJx = OppJx_3d + OppJx_Ld
OppJy = OppJy_3d + OppJy_Ld
OppJz = OppJz_3d + OppJz_Ld
OppJsqr = OppJx * OppJx + OppJy * OppJy + OppJz * OppJz
 
-- Just like in crystal-field theory we have Coulomb interaction on the Ni-d shell.
-- We again expand the Coulomb interaction on spherical harmonics, where the angular part
-- is solved analytical and the radial part gives three parameters F0, F2 and F4.
-- We here define three operators separately and only later provide parameters
 
OppF0_3d =NewOperator("U", NF, IndexUp_3d, IndexDn_3d, {1,0,0})
OppF2_3d =NewOperator("U", NF, IndexUp_3d, IndexDn_3d, {0,1,0})
OppF4_3d =NewOperator("U", NF, IndexUp_3d, IndexDn_3d, {0,0,1})
 
-- In ligand-field theory the ligand-field interaction is given by three different terms.
-- There is an onsite splitting on the Transition metal d-shell
-- There is an onsite splitting on the Ligand d-shell
-- There is a hopping between the ligand d-shell and the Transition metal d-shell
 
-- These interactions can be seen as effective potentials responsible for the splitting
-- In order to enter these potentials we expand them on renormalized spherical harmonics
-- and add the expansion coefficients to the function NewOperator("CF", ...)
-- We thus need to know the potential expanded on spherical harmonics:
-- Akm = {{k1,m1,Akm1},{k2,m2,Akm2}, ... }
 
-- For specific symmetries we can use the function "PotentialExpandedOnClm" Which for cubic
-- symmetry needs the energy of the eg and t2g orbitals. We here take the potential to be
-- such that we have a 1 eV splitting and later multiply the operator with the actual size
 
-- In crystal-field theory there is only an interaction on the transition metal d-shell
-- In ligand field theory there is an interaction on the transition metal d-shell as well
-- as on the ligand d-shell
 
Akm = PotentialExpandedOnClm("Oh", 2, {0.6,-0.4})
OpptenDq_3d = NewOperator("CF", NF, IndexUp_3d, IndexDn_3d, Akm)
OpptenDq_Ld = NewOperator("CF", NF, IndexUp_Ld, IndexDn_Ld, Akm)
 
-- We want to be able to calculate the occupation of the eg and t2g orbitals, we here use
-- the same operators with potentials of 1 for the eg or 1 for the t2g orbitals to create
-- number operators. (Note that there are many other options to do this)
 
Akm = PotentialExpandedOnClm("Oh", 2, {1,0})
OppNeg_3d = NewOperator("CF", NF, IndexUp_3d, IndexDn_3d, Akm)
OppNeg_Ld = NewOperator("CF", NF, IndexUp_Ld, IndexDn_Ld, Akm)
Akm = PotentialExpandedOnClm("Oh", 2, {0,1})
OppNt2g_3d = NewOperator("CF", NF, IndexUp_3d, IndexDn_3d, Akm)
OppNt2g_Ld = NewOperator("CF", NF, IndexUp_Ld, IndexDn_Ld, Akm)
 
-- We also want to know hom many electrons are in the Ni-d and how many are in the Ligand-d
-- shell. Here the number operators that count them.
 
OppNUp_3d = NewOperator("Number", NF, IndexUp_3d,IndexUp_3d,{1,1,1,1,1})
OppNDn_3d = NewOperator("Number", NF, IndexDn_3d,IndexDn_3d,{1,1,1,1,1})
OppN_3d = OppNUp_3d + OppNDn_3d
OppNUp_Ld = NewOperator("Number", NF, IndexUp_Ld,IndexUp_Ld,{1,1,1,1,1})
OppNDn_Ld = NewOperator("Number", NF, IndexDn_Ld,IndexDn_Ld,{1,1,1,1,1})
OppN_Ld = OppNUp_Ld + OppNDn_Ld
 
-- Besides the onsite energy of the ligand and transition metal d-shell we need to define the
-- hopping between them. We can use the same crystal-field operator, but now acting between
-- two different shells.
 
Akm = PotentialExpandedOnClm("Oh", 2, {1,0})
OppVeg  = NewOperator("CF", NF, IndexUp_3d,IndexDn_3d, IndexUp_Ld,IndexDn_Ld,Akm) +  NewOperator("CF", NF, IndexUp_Ld,IndexDn_Ld, IndexUp_3d,IndexDn_3d,Akm)
Akm = PotentialExpandedOnClm("Oh", 2, {0,1})
OppVt2g = NewOperator("CF", NF, IndexUp_3d,IndexDn_3d, IndexUp_Ld,IndexDn_Ld,Akm) +  NewOperator("CF", NF, IndexUp_Ld,IndexDn_Ld, IndexUp_3d,IndexDn_3d,Akm)
 
-- Once all operators are defined we need to set parameters
 
-- We follow the energy definitions as introduced in the group of G.A. Sawatzky (Groningen)
-- J. Zaanen, G.A. Sawatzky, and J.W. Allen PRL 55, 418 (1985)
-- for parameters of specific materials see
-- A.E. Bockquet et al. PRB 55, 1161 (1996)
-- After some initial discussion (some older papers use different definitions) the
-- energies U and Delta refer to the center of a configuration
-- The L^10 d^n   configuration has an energy 0
-- The L^9  d^n+1 configuration has an energy Delta
-- The L^8  d^n+2 configuration has an energy 2*Delta+U
--
-- If we relate this to the onsite energy of the p and d orbitals we find
-- 10 eL +  n    ed + n(n-1)     U/2 == 0
--  9 eL + (n+1) ed + (n+1)n     U/2 == Delta
--  8 eL + (n+2) ed + (n+1)(n+2) U/2 == 2*Delta+U
-- 3 equations with 2 unknowns, but with interdependence yield:
-- ed = (10*Delta-nd*(19+nd)*U/2)/(10+nd)
-- ep = nd*((1+nd)*U/2-Delta)/(10+nd)
--
-- note that ed-ep = Delta - nd * U and not Delta
-- note furthermore that ep and ed here are defined for the onsite energy if the system had
-- locally nd electrons in the d-shell. In DFT or Hartree Fock the d occupation is in the end not
-- nd and thus the onsite energy of the Kohn-Sham orbitals is not equal to ep and ed in model
-- calculations.
--
-- note furthermore that ep and eL actually should be different for most systems. We happily ignore this fact
--
-- We normally take U and Delta as experimentally determined parameters. Especially
-- core level photo-emission is sensitive to these parameters and can be used to determine
-- the starting point of these models. (see the work of Bockquet et al as referenced above)
 
-- number of electrons (formal valence)
nd = 8
-- parameters from experiment (core level PES)
U       =  7.3
Delta   =  4.7
-- parameters obtained from DFT (PRB 85, 165113 (2012))
F2dd    = 11.142
F4dd    =  6.874
tenDq   =  0.56
tenDqL  =  1.44
Veg     =  2.06
Vt2g    =  1.21
zeta_3d =  0.081
Bz      =  0.000001
 
-- turning U and Delta to onsite energies (Including the transformation from U to F0)
 
ed      =(10*Delta-nd*(19+nd)*U/2)/(10+nd)
eL      = nd*((1+nd)*U/2-Delta)/(10+nd)
F0dd    = U+(F2dd+F4dd)*2/63
 
-- and our Hamiltonian is the sum over several operators
 
Hamiltonian =  F0dd*OppF0_3d + F2dd*OppF2_3d + F4dd*OppF4_3d + zeta_3d*Oppldots_3d + Bz*(2*OppSz_3d + OppLz_3d)
            + tenDq*OpptenDq_3d + tenDqL*OpptenDq_Ld + Veg * OppVeg + Vt2g * OppVt2g
            + ed * OppN_3d + eL * OppN_Ld
 
Hamiltonian_CT = tenDqL*OpptenDq_Ld + Veg * OppVeg + Vt2g * OppVt2g + ed * OppN_3d + eL * OppN_Ld
-- we now can create the lowest Npsi eigenstates:
Npsi=3
-- in order to make sure we have a filling of 8 electrons we need to define some restrictions
StartRestrictions = {NF, NB, {"1111111111 0000000000",nd,nd}, {"0000000000 1111111111",10,10}}
 
psiList = Eigensystem(Hamiltonian, StartRestrictions, Npsi)
oppList={Hamiltonian, OppSsqr, OppLsqr, OppJsqr, OppSz_3d, OppLz_3d, Oppldots_3d, OppF2_3d, OppF4_3d, OppNeg_3d, OppNt2g_3d, OppNeg_Ld, OppNt2g_Ld, OppN_3d}
 
-- print of some expectation values
print("  #    <E>      <S^2>    <L^2>    <J^2>    <S_z^3d> <L_z^3d> <l.s>    <F[2]>   <F[4]>   <Neg^3d> <Nt2g^3d><Neg^Ld> <Nt2g^Ld><N^3d>");
for i = 1,#psiList do
  io.write(string.format("%3i ",i))
  for j = 1,#oppList do
    expectationvalue = Chop(psiList[i]*oppList[j]*psiList[i])
    io.write(string.format("%8.3f ",expectationvalue))
  end
  io.write("\n")
end
 
-- Alternative way to get the wavefunctions.
psiList1 = Eigensystem(Hamiltonian, StartRestrictions, Npsi, {{"Restrictions", StartRestrictions}})
psiList1 = Hamiltonian_CT * psiList1
Eigensystem(Hamiltonian, psiList1, {{'ExpandBasis', false}})
 
io.write("\n")
-- print of some expectation values
print("  #    <E>      <S^2>    <L^2>    <J^2>    <S_z^3d> <L_z^3d> <l.s>    <F[2]>   <F[4]>   <Neg^3d> <Nt2g^3d><Neg^Ld> <Nt2g^Ld><N^3d>");
for i = 1,#psiList1 do
  io.write(string.format("%3i ",i))
  for j = 1,#oppList do
    expectationvalue = Chop(psiList1[i]*oppList[j]*psiList1[i])
    io.write(string.format("%8.3f ",expectationvalue))
  end
  io.write("\n")
end
You could leave a comment if you were logged in.
Print/export