[CPMD-list] LSD - KS diag.

[Jochen Blumberger]

Dear CPMD developers,


I encountered a problem for the Kohn-Sham diagonlization with LSD in V3.7.1:

The wavefucntion optimization for a sample of 32 waters gives

  ELECTRONIC GRADIENT:
     MAX. COMPONENT =    5.10300E-06         NORM =    7.29372E-08
  NUCLEAR GRADIENT:
     MAX. COMPONENT =    4.18221E-02         NORM =    1.19544E-02


  TOTAL INTEGRATED ELECTRONIC DENSITY
     IN G-SPACE =                                       256.000000
     IN R-SPACE =                                       256.000000
  TOTAL INTEGRATED SPIN DENSITY                            .000026

  (K+E1+L+N+X)           TOTAL ENERGY =         -548.80201537 A.U.
  (K)                  KINETIC ENERGY =          412.08735410 A.U.
  (E1=A-S+R)     ELECTROSTATIC ENERGY =         -367.57551500 A.U.
  (S)                           ESELF =          404.26149474 A.U.
  (R)                             ESR =           24.02130066 A.U.
  (L)    LOCAL PSEUDOPOTENTIAL ENERGY =         -513.72725987 A.U.
  (N)      N-L PSEUDOPOTENTIAL ENERGY =           54.97660217 A.U.
  (X)     EXCHANGE-CORRELATION ENERGY =         -134.56319678 A.U.
           GRADIENT CORRECTION ENERGY =           -7.53087044 A.U.

  ****************************************************************


but the diagonalization using the Restart file of the optimization run
gives:

TOTAL INTEGRATED ELECTRONIC DENSITY
     IN G-SPACE =                                       256.000000
     IN R-SPACE =                                       256.000000
  TOTAL INTEGRATED SPIN DENSITY                            .000026

  (B+E2+X-V)             TOTAL ENERGY =         -549.88465526 A.U.
  (B)                     BAND ENERGY =          -88.88410491 A.U.
  (E2=I-H-S+R)   ELECTROSTATIC ENERGY =         -499.36486171 A.U.
  (S)                           ESELF =          404.26149474 A.U.
  (R)                             ESR =           24.02130066 A.U.
  (X)     EXCHANGE-CORRELATION ENERGY =         -134.56319678 A.U.
  (V)     EXCHANGE-CORRELATION POTEN. =         -172.92750821 A.U.
           GRADIENT CORRECTION ENERGY =           -7.53087044 A.U.

  ==------------------------------------------------------------==
  ==     NFI=      1      ETOT=   -549.884655   TCPU=     14.14 ==
  == DRHOMAX= 3.550E-09  DETOT=     0.000E+00    THL= 0.000E+00 ==
  ==------------------------------------------------------------==


In the LDA case wavefunction optimization and diagonlization gives the
same total energy which also coincides with the total energy of the LSD 
wavefunction optimization.
Here's the output for LDA diagonalization:

  TOTAL INTEGRATED ELECTRONIC DENSITY
     IN G-SPACE =                                       256.000000
     IN R-SPACE =                                       256.000000

  (B+E2+X-V)             TOTAL ENERGY =         -548.80201716 A.U.
  (B)                     BAND ENERGY =          -88.88413317 A.U.
  (E2=I-H-S+R)   ELECTROSTATIC ENERGY =         -499.36483919 A.U.
  (S)                           ESELF =          404.26149474 A.U.
  (R)                             ESR =           24.02130066 A.U.
  (X)     EXCHANGE-CORRELATION ENERGY =         -134.56321558 A.U.
  (V)     EXCHANGE-CORRELATION POTEN. =         -174.01017083 A.U.
           GRADIENT CORRECTION ENERGY =           -7.53087727 A.U.

  ==------------------------------------------------------------==
  ==     NFI=      1      ETOT=   -548.802017   TCPU=      7.11 ==
  == DRHOMAX= 4.225E-09  DETOT=     0.000E+00    THL= 0.000E+00 ==
  ==------------------------------------------------------------==


The KS energy levels of the LSD and LDA diagonalization are ~identical 
(as they should be) but the total energies are 549.88465526 for LSD and 
  548.80201716 for LDA. The difference comes from the 
EXCHANGE-CORRELATION POTENTIAL: -172.92750821 for LSD, -174.01017083 for 
LDA. Could there be  a bug in the calculation of the total 
exch.corr.Potential in the LSD case ? (In previous versions LSD diag. 
worked well)

many thanks

best regards

Jochen Blumberger