[CPMD-list] Discrepancy between MORTENSEN and TUCKERMAN

[Eung-Gun Kim]

Dear Salomon,

Thanks very much for kindly replying to my posted message.

The system I tested is a rod-like molecule aligned along a body diagonal.  I
used the body diagonal orientation in order to accommodate this 18 A long
molecule within the maximum box size (25x25x25) I could run without G-Vec
ordering errors.  This molecular orientation reduces the spatial dimensions of
the molecule down to 10 A in all three directions, leaving about 7.5 A space
between the molecule and the closest box surface.

I've attached a complete input and output geometry file from a MORTENSEN run for
your reference.  The input geometry is very close to a structure that I get from
PBC or TUCKERMAN.  In the output geometry, however, you will find MORTENSEN has
changed the torsion angle at a phenyl-phenyl bond by almost 180 deg.

Thanks again for your help and also the suggestion on the use of CONVERGENCE
ADAPT.

Regards,

Eung-Gun


Quoting Salomon Billeter <ZRLSRB at ch.ibm.com>:

>
>
>
>
> Dear Eung-Gun,
>
> is there also a difference between single-point energies (same geometry,
> same or random initial wavefunction)? What system do you calculate? I
> cannot reproduce such a large difference (i.e., I get the same numbers for
> both Poisson solvers for your settings with ethylene up to 10^-5 atomic
> units and very parallel evolution of the optimization). It's possible that
> you are converging to two slightly separated local minima. Did you make
> sure there is enough empty space in each direction?
> Just a hint for easier convergence: If you are using quite a tough
> convergence criterion such as 10^-6 AU and have a healthy system, I'd
> strongly recommend using CONVERGENCE ADAPT (e.g. 0.02) and (if not
> restarting from a good wavefunction) CONVERGENCE INITIAL (e.g. 0.02 times
> the gradient on ions you expect for the first step).
> Could you please tell us more about your system?
>
> Best regards,
> Salomon
>
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: bfo.inp
Type: application/octet-stream
Size: 2515 bytes
Desc: not available
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20041207/31c43c60/attachment.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: GEOMETRY.xyz
Type: chemical/x-xyz
Size: 5849 bytes
Desc: not available
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20041207/31c43c60/attachment.xyz