[CPMD-list] FEMD and Tolerance

[Eung-Gun Kim]

Dear List Subscribers,

When LANCZOS DIAGO is used in combination with FEMD, it seems that TOLERANCE for
each FRIESNER ITERATION is no longer followed.  In the example attached below,
one FRIESNER ITERATION is finishing before reaching any of the set LACZOS
PARAMETERs (CYCLE, NCONV or B2MAX).  This does not happen when plain LANCZOS
DIAGO (without FEMD) is used.  Is there an additional parameter to control FEMD?

Thanks very much for reading.

Regards,

E Kim


Partial input for a band structure calc =======================================

&CPMD
  RESTART COORDINATES DENSITY
  OPTIMIZE WAVEFUNCTION
  FREE ENERGY FUNCTIONAL
  TEMPERATURE ELECTRON
    298
  LANCZOS DIAGONALISATION
  LANCZOS PARAMETER
    4000  8  0  1.D-16
  ENERGYBANDS
  STORE WAVEFUNCTIONS OFF
  MAXSTEP
    1
&END

&DFT
  FUNCTIONAL BLYP
&END

&SYSTEM
  ANGSTROM
  SYMMETRY
    ORTHORHOMBIC
  CELL ABSOLUTE DEGREE
    7.92534743  32.0  22.0   90.0 90.0 90.0
  CUTOFF
    70
  CHARGE
    0
  STATES
    48
  KPOINTS BANDS
    37   0.0  0.0  0.0   0.5  0.0  0.0
    0   0 0 0  0 0 0

=== Partial output ============================================================

 <<37:37<<<<<<<<<<<< LANCZOS DIAGONALIZATION <<<<<<<<<<<<<<<<<<<<
 >> TIME FOR INITIAL SUBSPACE DIAGONALIZATION:               3.24
 >> CYCLE     NCONV        B2MAX        B2MIN     #HPSI      TIME
        1         0    3.240E-08    2.959E-09      8.00     20.88
        2         0    8.211E-09    1.370E-10      7.00     20.89
.....
.....
.....
      291        46    2.971E-16    6.694E-17       .29      1.26
      292        46    2.952E-16    6.694E-17       .29      1.26
      293        47    2.692E-16    6.694E-17       .27      1.21

 EIGENVALUES(EV) AND OCCUPATION:
 K POINT:       1     .000000     .000000     .000000     .027027
     1    -26.2847636     2.000        2    -26.2687872     2.000
     3    -25.4367240     2.000        4    -25.4245011     2.000