[Wien] FW: Internal Coordinates minimization for Monoclinic

[Peter Blaha]

Only you can answer this, because it depends on what exactly you want to do and
what statements you want to make.

For a statement: I've optimized the atomic positions using the exp. lattice parameters,
all you need to do is   min_lapw

For a statement: I've done a full structural optimization, you need to optimize
all lattice parameters AND atomic positions simultaneously.


Ghosh SUDDHASATTWA schrieb:
> Dear Prof.Blaha, 
> Extending the discussion on internal coordinates minimization and structure
> optimization, I have a tetragonal body centered lattice with 58 atoms per
> unit cell and 8 inequivalent atoms. 
> We can use suitable options in the optimization (c/a). Since, the number of
> atoms are large enough, can we go in for min_lapw with similar arguments you
> mention for monoclinic lattice. 
> I ask this question as my total run time for SCF (-ec 0.0001 -cc 0.0001
> -in1ef) on 32 cpu's is more than 7 days. 
> So, is it advisable to do only min_lapw 
> 
> I am not asking for any shortcuts, Dear users, 
> 
> Suddhasattwa Ghosh 
> 
> 
> 
> -----Original Message-----
> From: wien-bounces at zeus.theochem.tuwien.ac.at
> [mailto:wien-bounces at zeus.theochem.tuwien.ac.at] On Behalf Of Peter Blaha
> Sent: Monday, June 28, 2010 12:40 PM
> To: A Mailing list for WIEN2k users
> Subject: Re: [Wien] FW: Internal Coordinates minimization for Monoclinic
> 
> You are comparing apples and bananas.
> 
> With min_lapw you optimize internal atomic positions, at given (fixed)
> lattice
> parameters. So you can only say, you have an "optimum structure" with
> respect to fixed
> (experimental) lattice parameters.
> Often, this is a VERY good strategy, since experimental a,b,c are MUCH MORE
> accurate
> than theoretical ones, while for atomic positions often the opposite holds.
> Note, however, you CANNOT use -so for force optimization.
> 
> With the first approach, you can try to find optimal a,b,c,gamma lattice
> parameters. However, usually it is also very important to optimize internal
> coordinates simultaneously, because this may also influence a,b,c...
> As mentioned above, theory is usually quite bad on that and for a monoclinic
> structure this gets expensive.
> 
>> I have done the following 
>> 1. Generate the structure file and run a normal SCF with spin orbit
>> coupling. (Since I am working on an heavy element with 18 atoms per unit
>> cell)
>> 2. Now my question is 
>> We can either use option 7 and generate 15 structures and optimize the
>> structure with minimum energy. 
>> By modifying the command in optimize.job 
>> Run_lapw -cc 0.0001 -ec 0.0001 -p -so 
>>
>> Or, we can simply do min_lapw for the structure file we have generated
>> initially. 
>> After running min_lapw, we get the atomic positions with minimum partial
>> forces, which gets saved in the structure file. 
>> Now, using the revised (optimized) atomic coordinates, run a SCF. 
>> By doing so, can we say that we have the structure with minimum total
> energy
>> (by doing only min_lapw and not using option 7) 
>>
>> I hope I am clear enough now. 
>> Thank you 
>> Suddhasattwa 
>>
>> -----Original Message-----
>> From: wien-bounces at zeus.theochem.tuwien.ac.at
>> [mailto:wien-bounces at zeus.theochem.tuwien.ac.at] On Behalf Of Peter Blaha
>> Sent: Monday, June 28, 2010 11:02 AM
>> To: A Mailing list for WIEN2k users
>> Subject: Re: [Wien] FW: Internal Coordinates minimization for Monoclinic
>>
>> Most likely nobody can answer this question, because nobody understands
> your
>> question.
>> At least I did not understand what you are really doing.
>>
>> Did you run    x optimize with option 7 ?
>> Did you modify optimize.job or just run it unmodified ?
>> Did you put min_lapw instead or run_lapw ?
>> Did you find the structure with lowest total energy ?
>>
>>> Coming back to my question, doing a 4-D optimization of a monoclinic 
>>> case by choosing Option 7, can we say with authority that
>>>
>>> 1. Internal coordinates minimization of monoclinic case (with 18 atoms 
>>> per unit cell) keeping the lattice parameters same would be enough to 
>>> know the total energy.
>>>
>>> In any case, after lapw_mini, the optimized coordinates are saved in the 
>>> case.struct, so running the SCf again would give the optimized total 
>>> energy.
>> Why would you run again an scf ? You already have one ?
>>
>>
> 

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------