[Wien] false convergence in structure optimization

[L. D. Marks]

Please read the notes, 
http://www.wien2k.at/reg_user/textbooks/Optimization-Notes.pdf in 
particular 4c & 4d and also section 6.

On Tue, 10 Jan 2006, Jian ZHOU wrote:

> Dear marks,
>
> yes, the total energy and charge density have been converged well:
>  grep :ENE *mini
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074941
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074943
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074977
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074930
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074910
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074845
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074846
> :ENE  : ********** TOTAL ENERGY IN Ry =      -304766.074822
>  grep :DIS *mini
> :DIS  :  CHARGE DISTANCE       0.0002048
> :DIS  :  CHARGE DISTANCE       0.0000333
> :DIS  :  CHARGE DISTANCE       0.0000244
> :DIS  :  CHARGE DISTANCE       0.0000483
> :DIS  :  CHARGE DISTANCE       0.0000717
> :DIS  :  CHARGE DISTANCE       0.0000318
> :DIS  :  CHARGE DISTANCE       0.0000186
> :DIS  :  CHARGE DISTANCE       0.0000016
>
> Thanks.
>
>
> L. D. Marks wrote:
>> I you do "grep -e :" you will see all the relevant information. Most
>> useful is :ENE and :DD7DOG. The first is the energy, the second the
>> trust-region step. The code is working exactly as it should, it's
>> minimizing the energy, taking smaller and smaller trust radii if it
>> cannot find a better point (it says at the end :MIN Probably converged
>> to numerical accuracy). Your problem is that your forces are not quite
>> consistent with your energy, which means that (at least one of) RKMAX,
>> RMT or the force convergence etc are not good enough (which we cannot say).
>>
>> N.B., I think you might be using an older version since I did not see
>> :GG for the gradient, although this might not be in the web version.
>>
>> On Mon, 9 Jan 2006, Jian ZHOU wrote:
>>
>>>
>>> I enclose the outputM.
>>>
>>>
>>> L. D. Marks wrote:
>>>> The "False Convergence" need not mean anything (see the minimization
>>>> notes). Can you attach your case.outputM, since this has the relevant
>>>> information to understand what is happening.
>>>>
>>>> Note: if you have an old email address for me, please note that "nwu"
>>>> has
>>>> been changed to "northwestern".
>>>> -----------------------------------------------
>>>> Laurence Marks
>>>> Department of Materials Science and Engineering
>>>> MSE Rm 2036 Cook Hall
>>>> 2220 N Campus Drive
>>>> Northwestern University
>>>> Evanston, IL 60201, USA
>>>> Tel: (847) 491-3996 Fax: (847) 491-7820
>>>> email: L-marks at northwestern dot edu
>>>> http://www.numis.northwestern.edu
>>>> -----------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Wien mailing list
>>>> Wien at zeus.theochem.tuwien.ac.at
>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>
>>>
>>
>> Note: if you have an old email address for me, please note that "nwu" has
>> been changed to "northwestern".
>> -----------------------------------------------
>> Laurence Marks
>> Department of Materials Science and Engineering
>> MSE Rm 2036 Cook Hall
>> 2220 N Campus Drive
>> Northwestern University
>> Evanston, IL 60201, USA
>> Tel: (847) 491-3996 Fax: (847) 491-7820
>> email: L-marks at northwestern dot edu
>> http://www.numis.northwestern.edu
>> -----------------------------------------------
>>
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

Note: if you have an old email address for me, please note that "nwu" has
been changed to "northwestern".
-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60201, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
http://www.numis.northwestern.edu
-----------------------------------------------