What it looks like to me, Masroor, is that the energy gradients shown
per atom are the gradients that come out of the DFT calculation, and
the max gradient reported in the convergence tests is for the gradient
with rigid motions projected out. In other words, the DFT calculation
produces gradients which are not exactly zero for rigid motions, such
as translation and rotation. They should be zero, but they aren't. The
geometry optimizing code projects out the rigid motions, and only deals
with the gradients for internal degrees of freedom. The gradient max
reported is for the portion of the gradient left after you have
projected out rigid motions.

Hope this helps,
Drew McCormack

On Thursday, September 18, 2003, at 09:59 am, Masroor Ahmad wrote:

> Hi all
> I've rather (possibly) trivial question which I took much longer to
> ask ,in hope (or fear?) that it's not worth asking...
> I'm reproducing below a typical ADF output of a geometry
> step..gradients about individual atomic coordinates and then
> convergence tests:-
> One can see in convergence tests that 'gradient max' has reached
> 0.00000735 hartree/angs but there are plenty of individual atomic
> coordinates (also shown)
> which have a gradient component much greater than this 'gradient
> max'.Is 'gradient max' something different from maximum component of
> overall gradients of all atoms?Thanks....Masroor
> -----------------------------------------------------------------------
> -----------------------------------------------------------------------
> -------------------------------------------------------------
> Energy gradients wrt nuclear displacements
> 2 ==========================================
> 3
> 4 Atom Cartesian (a.u./angstrom)
> 5 X Y Z
> 6 ----------------------------------------
> 7 1 C 0.000019 0.000008 0.000046
> 8 2 N 0.000070 -0.000002 0.000044
> 9 3 C 0.000087 -0.000012 0.000002
> 10 4 C 0.000049 -0.000013 -0.000028
> 11 5 N 0.000006 0.000000 -0.000001
> 12 6 V -0.000059 0.000000 -0.000019
> 13 7 O -0.000031 -0.000005 -0.000031
> 14 8 O -0.000055 -0.000015 -0.000064
> 15 9 O -0.000078 0.000026 0.000024
> 16 10 O -0.000120 0.000005 -0.000039
> 17 11 H -0.000138 0.000013 -0.000016
> 18 12 H -0.000009 0.000022 0.000075
> 19 13 H 0.000084 0.000010 0.000076
> 20 14 H 0.000129 -0.000014 -0.000004
> 21 15 H 0.000046 -0.000022 -0.000067
> 22 ----------------------------------------
> 23
> 24
> 25
> 26 ==========================
> 27 Geometry Convergence Tests
> 28 ==========================
> 29 30 Energy old : -3.88310290
> 31 new : -3.88310382
> 32
> 33 Convergence tests:
> 34 (Energies in hartree, Gradients in hartree/angstr or radian,
> Lengths in angstrom, Angles in degrees)
> 35
> 36 Item Value Criterion Conv.
> Ratio
> 37
> -----------------------------------------------------------------------
> --
> 38 change in energ -0.00000091 0.00000001 NO
> 0.81827850
> 39 gradient max 0.00000735 0.00000700 NO
> 0.69531915
> 40 gradient rms 0.00000337 0.00000467 YES
> 0.91751087
> 41 cart. step max 0.00003128 0.00005000 YES
> 0.76254156
> 42 cart. step rms 0.00000983 0.00003333 YES
> 0.79214666
> 43
> 44 prediction dE : 0.00000000
>
> --
> Your's truly
> Masroor Ahmad
>
> -------------------------------------------------|
> Institute für Anorganische und Analytische Chemie|
> Friedrich-Schiller Universität Jena |
> Carl-Zeiss Promenade 10 |
> D-07743 Jena (Germany) |
> -------------------------------------------------|
> Tel: +49(0)3641-644028 |
> Fax: +49(0)3641-644036 | Mob:
> +49(0)1798880964 |
> -------------------------------------------------|
> Residence: | Apartment 605,Am
> Herrenberg 11 |
> 07745 Jena (Germany) |
> -------------------------------------------------|
>
>
>
>
>
========================================
  Dr. Drew McCormack (Kmr. R153)
  Afd. Theoretische Chemie
  Faculteit Exacte Wetenschappen
  Vrije Universiteit Amsterdam
  De Boelelaan 1083
  1081 HV Amsterdam
  The Netherlands

  Email cormack_at_chem.vu.nl
  Telephone +31 20 44 47623
  Mobile +31 6 483 21307
  Fax +31 20 44 47629
Received on 2003-09-18 11:59:23