[CPMD-list] Some words in output files

[Martin Konopka]

On Wed, 4 Aug 2004, Keta Jones wrote:

>  Dear All,
> My out put file of geometry optimization is as
> follows.
> Please help to understand the terms like GEMAX, GNMAX,
> GNORM, CNSTR in just following few lines and also in
> after ward line.I mean these terms also appear after
> every 4/5 or 10 steps as summary.I want to know what
> is the significance of these terms towards the
> geometry optimization.How do we know that our system
> has reached the perfect optimized geometry.
>
> Thanks in advance,
> keta
> ================================================================
>  =                  GEOMETRY OPTIMIZATION
>          =
>
> ================================================================
>  NFI      GEMAX       CNORM           ETOT
> DETOT      TCPU
>  EWALD SUM IN REAL SPACE OVER  1* 1* 1 CELLS
>    1  3.568E-05   2.478E-06     -32.068434
> -3.207E+01     10.58
>    2  4.127E-05   1.016E-06     -32.068434
> -1.064E-07     10.77
>    3  1.768E-05   4.957E-07     -32.068434
> -3.922E-08     10.79
>    4  3.217E-06   2.835E-07     -32.068434
> -4.823E-09     11.19
> ======================================================
>  *** TOTAL STEP NR.     4           GEOMETRY STEP NR.
>     1  ***
>  *** GNMAX=  7.760103E-04                ETOT=
> -32.068434  ***
>  *** GNORM=  4.191947E-04               DETOT=
> 0.000E+00  ***
>  *** CNSTR=  0.000000E+00                TCPU=
> 43.37  ***
>
> ***************************************************
>
> =====
> Keta
> **************************************************************************************


Hi,

GNMAX = the largest (cartesian component) of the forces (-gradients)
on atomic nuclei; must be small for well converged structures;
typically 1.0E-4 or better, even of the order 1.0E-5.

GEMAX = the largest of the gradients on (electronic) wavefunctions;
must be small for well converged ground-state electronic
structure (typically 10-100 times smaller than GNMAX)

GNORM = NORM OF GRADIENT (see gsize.F)
It may have a sense of an average of forces on nuclei.

CNSTR - something relevant in constrained runs - do not know exactly;
better to look into the code.

Small GEMAX and GNMAX are just necessary conditions for well-converged
systems. Experience shows that their values as I show above often mean
that the structure is very close to its ground state or better local
minimum.

Best regards
Martin.

------------------------------------------------------------------------
Martin Konopka                 http://www.ccms.elf.stuba.sk/konopka.html
Department of Physics, CCMS                      tel: +421 (0)2 60291714
Slovak University of Technology (FEI STU)        fax: +421 (0)2 65411483
Ilkovicova 3, 812 19 Bratislava, Slovakia
------------------------------------------------------------------------