Refining with both generic and merohedral twinning in smtbx

[oleg]

OK, I was looking and looking into the code, it seems that now things should be fine (another update should come on the development version). Of course I would love to have the data so that I can play with it myself and see what is going on, but anyway - I might have sorted it out, could you try again please?

Cheers,
o

[Zamu]

¡That was quick! I received the update before I read your answer. Thanks a lot, Oleg, and I am sorry if I am wasting your time. I'll send you a PM with the data.
With the new version the large min. peaks are gone, BASF refines to 'BASF 0.527543 0.321993', which are the values I expected from the refinements with only HKLF5 and only merohedral twinning. Besides, the caching bug seems to be gone, since I don't need to use 'file' and restart OLex2 anymore.
But the agreement factors (both R and wR2) are still much higher than using TWIN alone, without batches. It could be that the HKLF5 is not correct, but I reprocessed the data a couple of times and seems to be OK.

Pablo

[oleg]

OK, it is fixed in the development snapshot now, thanks again for the sample,

Cheers,

o

[Zamu]

Hi Oleg,

Thanks again for all your help. I tested the new version, but I still get quite high wR2 (about 0.3), higher than just using the twin law, event though BASF refines to what seems to be the correct values.

Cheers,

Pablo

[oleg]

This happens because the use of HKLF 5 implies disabling merging... You can add 'MERGE 0' and use shelx (olex2 does not obey this command well yet!) to refine any of the detwinned datasets to observe the trend.

Cheers,

o

[Zamu]

You are right, if I use MERG 0 with the HKLF4 datasets (unmerged during data reduction) it happen just that.
Thanks a lot for all your help

Cheers,

Pablo

[PJC]

I had seen this discussion a while ago and now I have a pseudo-merohedral + non-merohedral twin that I'm trying to deal with. Is this capability included in the current version of Olex2? Is there documentation anywhere or can you summarize what I need to do? (This discussion was a little hard to follow because, apparently, you were just developing this capability)

[johnewarren]

I'm not certain if this will help and I will probably get shot for putting the link here but I noticed they are testing their new beta website: http://wp.olex2.org/olex2/documentation/ and as you can see it has more documentation with it?

[Marc]

I got this reply from George when asking a similar question (apply merohedral twin law to HKLF5 data):

"I would simply write a (trivial) program to read the .hkl file and expand the two components for each observation to four copying the intensities and using the appropriate matrix to transform the indices, and set the twin component numbers 1 to 4 rather than 1 to 2. The first three contributors should have negative twin component numbers and the fourth should have a positive number."

That worked reasonably well for me, R values did increase though as predicted by Oleg

"Hi Marc,

you could give a try to olex2-refine. We did implement this quite a while ago, look here for the relevant discussions:
http://www.x-rayman.co.uk/xforum/viewto ... 122&t=1467

People however find it not much exciting, since using more data increases the R-factor values ,

Cheers,

Oleg."

[PJC]

Thanks for the info Marc. Let me see if I understand:

I need to take my existing hkl 5 file for component 1 and operate on the reflections with the matrix that relates the merohedral twins, converting a batch no. of 1 to 3 and a batch no. of -2 to -4. Then append these new reflections to the existing hkl 5 file. Is that right? And then there will be three BASF twinning parameters?

Have I got it correct?

[Marc]

right, but batch no 3 has to be negative as well. Also due to the peculiarities of shelxl (at least in the 97 version) the batch no of the last refl has to be positive.

Marc

[PJC]

Great - now I just need to figure out scripting in python. Anyone able to help me get my twinning matrix {1 0 1 0 -1 0 0 0 -1} into the hklf5.py script in olex2?

[oleg]

Hello,

if you use olex2.refine - you do not need to do anything, it does support a combination of merohedral and generic twinning, however, if you want to prepare that hkl file for shelxl - use the following as a start:
@py -l
and pick the hkl5.py file from etc/scripts. Modify the file to suit your requirements - then you can change the HKL file against which you are refining,

Cheers,

o

[oleg]

Sorry - just got that you got to the hklf5.py script! So what you need to do is to have this inside the loop over all hkl:

Code: Select all

for hkl in hkl_file:
   # post a transformed reflection with batch number equal -2
   output.append(hkl[0]+hkl[2], -hkl[1], -hkl[2], hkl[3], hkl[4], -2) )
  # post the original reflection with batch number equal to 1
  output.append( hkl[:5] + (1,) )

o

[PJC]

Oleg,

Thanks for the info, but one last thing:

I'm starting from a type 5 hkl file which already has batch nos. of 1 and -2. I think if I do a transformation on a reflection with a batch no. of 1, the new reflection needs to have a batch no. of -3 AND if I transform a reflection with a batch no. of -2, it needs to have a batch no. of -4. Does that sound right? I think I need four batch nos. which will require three BASF twinning parameters. How do I do all that in hkl5.py?