[molpro-user] threshhold for MULTI

[Andy May]

Lydia,

When running optg, the multi thresholds are automatically set tighter.
The value of the thresholds used in multi when running optg is the
minimum of the input thresholds, or an internal default (not documented)
which is tighter than in regular multi-calculation.

This explains why you do not appear to see a change with respect to your
input values, try setting the thresholds very tight, eg.:

gthresh,gradient=1.11d-9,energy=2.22d-9,step=3.33.d-9

and you will see these values are used by optg multi calculations.

Best wishes,

Andy

On 17/11/10 15:13, lydia wrote:
> Thanks, Andy.
> I tried your input, it gives the
>  Convergence thresholds  0.10E-01 (gradient)  0.10E-05 (energy)
> 0.10E-02 (step length); It is the same as the default.
> Then I tried your input plus "optg", it gives
> Convergence thresholds  0.10E-04 (gradient)  0.10E-07 (energy)
> 0.10E-02 (step length) in the out and log files;
> 
> I attached my out and log files for a optmizing and frequencies
> calculation, please help me to check it.
> 
> Thank you very much,
> all the best,
> lydia
> 
> On Wed, Nov 17, 2010 at 8:27 AM, Andy May <MayAJ1 at cardiff.ac.uk> wrote:
>> Lydia,
>>
>> Running with the following trivial input:
>>
>> r=1.85,theta=104
>> geometry={O;
>>          H1,O,r;
>>          H2,O,r,H1,theta}
>> basis=vdz
>>
>> hf
>> multi
>>
>> generates the following output:
>>
>>  Convergence thresholds  0.10E-01 (gradient)  0.10E-05 (energy)
>> 0.10E-02 (step length)
>>
>> which corresponds to what is written in the manual.
>>
>> If this is not what you see, please can you send the output file and
>> I'll take a look.
>>
>> Best wishes,
>>
>> Andy
>>
>> On 16/11/10 21:31, lydia wrote:
>>> Hi everyone,
>>> I just found a problem about the accuracy  in the MULTI  convergency.
>>> The default threshhold for orbital gradient, change of total energy,
>>> size of step is 1.d-2, 1.d-6, 1.d-3, respectively, as shown in p144 of
>>> 2010 version manule.
>>> In my casscf optimzation calculation, I did not give any keyword to
>>> set the accuracy. But the output file show different ones for the
>>> first iteration of MULTI calculation as follows:
>>>  Convergence thresholds  0.10E-01 (gradient)  0.23E-05 (energy)
>>> 0.10E-02 (step length)
>>>
>>> and the log file give another different accuracy as follows:
>>> Convergence thresholds  0.10E-05 (gradient)  0.10E-07 (energy)
>>> 0.10E-02 (step length)
>>>
>>> Does anyone have similar problem? Any suggestions are welcome.
>>> Thanks,
>>>
>>> lydia
>>> _______________________________________________
>>> Molpro-user mailing list
>>> Molpro-user at molpro.net
>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>