[SOLVED] Re: [molpro-user] turning off dkroll

[Benj FitzPatrick]

Kirk,
Thank you so much, that works perfectly.
Benj

---- Original message ----
>Date: Sat, 14 Jun 2008 18:31:01 +0200
>From: Kirk Peterson <kipeters at wsu.edu>  
>Subject: Re: [molpro-user] turning off dkroll  
>To: Benj FitzPatrick <benjfitz at uchicago.edu>
>
>Benj,
>
>try the input I've attached.  It doesn't work quite right on my laptop  
>since I only have 2006.4 there and it suffers
>from the DK bug I mentioned before. I think it might do the trick  
>under 2006.1 however.
>
>-Kirk
>
>________________
>ch3cocl_sp-dkh-avddz-a.inp (2k bytes)
>________________
>
>
>
>
>On Jun 13, 2008, at 8:17 PM, Benj FitzPatrick wrote:
>
>> Dr. Peterson,
>> I tried your suggestion for saving the pre-loop orbitals, but I  
>> think I botched
>> this as well.  If I put basis=... for each HF job in the loop (so  
>> 6-31g* for the
>> first one) it will recalculate the integrals each time, and as soon  
>> as dkroll=1
>> I'm back to the same problem where it will take the starting  
>> orbitals I give it
>> calculates the orbital overlap with the dkh integrals it just  
>> calculated.  On
>> the other hand if I leave out the basis=... lines then it thinks  
>> that it is
>> using the cc-pVDZ-DK basis set the whole time (this shows up in the  
>> orbital
>> overlap again), and my HF energies for the first two calcs in the  
>> loop are not
>> similar to the ones outside the loop.  I attached both attempts
>> at input files, and the one with the 2 at the end is the one where I  
>> try to feed
>> loop HF calcs pre-loop orbitals.
>> Thanks,
>> Benj
>>
>>
>> ---- Original message ----
>>> Date: Fri, 13 Jun 2008 19:29:33 +0200
>>> From: Kirk Peterson <kipeters at wsu.edu>
>>> Subject: Re: [molpro-user] turning off dkroll
>>> To: Benj FitzPatrick <benjfitz at uchicago.edu>
>>>
>>> Benj,
>>>
>>> why do you need to do the dkroll=0 stuff inside the loop anyway?  
>>> Can't
>>> you just save the non-relativistic orbitals
>>> from before the loop to some record and just start from there in the
>>> HF jobs inside the dkhorder loop?  I agree it
>>> should all work anyway, but maybe you can just work around this
>>> problem.  If you could send a whole input, I could
>>> try to reproduce things on my machines here.
>>>
>>> -Kirk
>>>
>>> On Jun 13, 2008, at 7:19 PM, Benj FitzPatrick wrote:
>>>
>>>> Ioannis and Kirk,
>>>> Thank you very much for the input (Kirk, I promise I hadn't
>>>> forgotten your
>>>> recommendation for using the -dk basis sets, I was just trying to
>>>> get something
>>>> working first and hadn't changed much from my starting input
>>>> file :-).  I am
>>>> using 2006.1 with patchlevel 84.
>>>>
>>>> I tried the dkroll=0/int suggestion, but I don't think I implemented
>>>> it properly
>>>> b/c it is still calculating the dk integrals each time.  I put the
>>>> most recent
>>>> version of my input file down below.  The first time through the
>>>> loop everything
>>>> looks fine (no surprise there), but the second time through it
>>>> calculates the
>>>> dkh integrals for 6-31g*, runs the HF, and crashes during the
>>>> calculation of the
>>>> dkh integrals for 6-311g** b/c of a singularity in the overlap
>>>> matrix (this is
>>>> one of the reasons I wanted to turn dkroll off).  Please let me know
>>>> what I'm
>>>> doing wrong.
>>>> Thanks,
>>>> Benj
>>>>
>>>> ...
>>>> basis=6-31G*;
>>>> {hf;
>>>> wf,40,1,0}
>>>> basis=6-311G**;
>>>> {hf;
>>>> wf,40,1,0;}
>>>> basis,default=cc-pvdz-dk;
>>>> {hf;
>>>> wf,40,1,0}
>>>> eref1=energy
>>>>
>>>> dkhp=1
>>>> i=0
>>>> do dkhorder=2,6,1
>>>> i=i+1
>>>> dkroll=0
>>>> basis=6-31G*;
>>>> int
>>>> {hf;
>>>> wf,40,1,0;
>>>> start,atdens}
>>>> basis=6-311G**;
>>>> int
>>>> {hf;
>>>> wf,40,1,0}
>>>> basis,default=cc-pvdz-dk
>>>> dkroll=1
>>>> int
>>>> dkho=dkhorder
>>>> order(i)=dkhorder
>>>> {hf;
>>>> wf,40,1,0}
>>>> edkh(i)=energy
>>>> eref(i)=eref1
>>>> ediff(i)=edkh(i)-eref(i)
>>>> enddo
>>>>
>>>> ---- Original message ----
>>>>> Date: Thu, 12 Jun 2008 20:27:33 -0400
>>>>> From: Ioannis Kerkines <ikerkin at emory.edu>
>>>>> Subject: Re: [molpro-user] turning off dkroll
>>>>> To: molpro-user at molpro.net
>>>>>
>>>>> Dear Benj,
>>>>>
>>>>> I used to have a similar problem once in the past and I solved it  
>>>>> by
>>>>> adding dkroll=0 AND explicitly asking to recalculate the integrals
>>>>> (using the "int" command I think).
>>>>>
>>>>> I did this in an older version of Molpro though, I am not 100% sure
>>>>> if
>>>>> this will work for you in this version.
>>>>>
>>>>> Best regards,
>>>>> Ioannis
>>>>>
>>>>> Benj FitzPatrick wrote:
>>>>>> Hello,
>>>>>> I was interested in seeing the effect of DKHO on the relativistic
>>>>>> calculation
>>>>>> and wanted to put this into a loop.  What do I need to do in order
>>>>>> to compare
>>>>>> these results with ones where I only compute one DKH energy in the
>>>>>> job and run a
>>>>>> lot of jobs?  Namely, I want to turn dkroll off (and dkroll=0
>>>>>> doesn't appear to
>>>>>> do it) b/c I have some preliminary HF calcs (some of the radicals
>>>>>> I'll be
>>>>>> studying don't converge unless I start with a small basis set)
>>>>>> with smaller
>>>>>> basis sets and I don't want them doing DKH calculations.  I gave a
>>>>>> portion of
>>>>>> the input file below.
>>>>>> Thanks,
>>>>>> Benj FitzPatrick
>>>>>> University of Chicago
>>>>>>
>>>>>> basis=6-31G*;
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> basis=6-311G**;
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> basis,default=aug-cc-pvdz,cl=avdz+d;
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> eref1=energy
>>>>>>
>>>>>>
>>>>>> dkhp=1
>>>>>> i=0
>>>>>> do dkhorder=2,6,1
>>>>>> i=i+1
>>>>>> !I need to turn off dkroll here
>>>>>>
>>>>>> basis=6-31G*;
>>>>>> {hf;
>>>>>> wf,40,1,0;
>>>>>> start,atdens}
>>>>>>
>>>>>> basis=6-311G**;
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> dkroll=1
>>>>>>
>>>>>> basis,default=aug-cc-pvdz,cl=avdz+d;
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> dkho=dkhorder
>>>>>> {hf;
>>>>>> wf,40,1,0}
>>>>>> enddo
>>>>>>
>>>>>>
>>>>>>
>>>
>> <ch3cocl_sp-dkh-avddz-a.inp><ch3cocl_sp-dkh-avddz-a2.inp>
>