[Discussions on Crystallography]
[Crystallography World Wide] | [Sci.techniques.xtallography Homepage] | [Web Page Maintainer]
[CCP14 Project] | [Xtal Guide to the Internet ]
The DIFABS Discussion homepage located at http://www.unige.ch/crystal/stxnews/stx/discuss/dis-dif2.htm

Should DIFABS be Banned?

Discussion on sci.techniques.xtallography

(20th April 1994 to 3rd May 1994 - and beyond (18th November 1999)

Latest Update - 18th November 1999

Using DIFABS in Crystals as a diagnostic test

Why bother using DIFABS Analysis?

"You can use DIFABS in test mode at any stage, and it quite simple tells you if there is a systematic trend to the residual as a function of orientation. If the plot is fairly monochrome, then the residuals are not peaking up. If there are bright patches, it means that there is a systematic trend to the residual. Either FO or Fc is systematically wrong. Possible reasons are:

Refer:

Gui WinCRYSTALS 2000 Web and FTP sites

2nd Last Update - 6th April 1999
A Web Based Comparison of Absorption Correction Methods - using the WinGX Suite

The Initial Discussion

Creating a Summary Document of the Discussion

Summary Document

(Included in "L. M. D. Cranswick, J. Silverton and T. Buckner", Global communications , IUCr Newsletter. Vol. 3, (1), 1995, pp. 6-8,10)

Where do you get the DIFABS and XABS2 software?

Comments Based on IUCr Newsletter Article (Vol. 3, (1), 1995, pp. 6-8,10)

Posting from Ton Spek to sci.techniques.xtallography | top 

Newsgroups: sci.techniques.xtallography
From:  spea@faculty.chem.ruu.nl (Ton Spek)
Subject: DIFABS - SHOULD IT BE BANNED ??
Date: 20 Apr 94 13:43:33 GMT
Organization: Physics Department, University of Utrecht, The Netherlands

This is meant to start a (possibly heated) discussion on the use (although according to some abuse) of the program DIFABS (Walker & Stuart, 1983, Acta Cryst., A39, 158-166) as a method to correct for the effects of absorption. Such a discussion is rapidly becoming acute in view of the fact that Acta Cryst is moving to a position of strongly discouraging the publication of structures 'massaged' with this or similar empirical methods. A quick look through recent issues of AC-C shows that in the order of 75% of the papers made use of empirical correction methods .....

The arguments against DIFABS that I know of are :

Supporters of DIFABS offer: so ??? 
  Ton Spek
  National Single Crystal Service Facility
  Utrecht University
  The Netherlands
  spea@xray.chem.ruu.nl

Followup by Paul Boyle | top 

From: "boyle@mantis.chem.unb.ca"
Newsgroups: sci.techniques.xtallography
Subject: RE: DIFABS - SHOULD IT BE BANNED ??
Date: 22 APR 94 11:08:08 AST
Organization: The University of New Brunswick

Well, Ton, I guess I'll take the bait. In general, I am against the use of DIFABS as a routine absorption correction. I think it is bad practice to fit the data to the model rather than fitting the model to the data. I try to use psi scan data to make an empirical absorption correction. While the psi scan method is not perfect, it does have the advantage of being based on some experimental data. One must take care that the psi scan data is not contaminated with Renniger effects.

I'll try to address each of the arguements given below

> The arguments against DIFABS that I know of are :
>
>- Primary data are massaged and not recoverable from the deposited
>  Fo/Fc-tables

I think this is a valid objection,and fits into my comment about fitting the data to the model.

>- Not just absorption effects but also other systematic effects
>  (certainly extinction) that can be fitted with the absorption model
>  are polished

Some Spanish crystallographers presented a poster at a recent ACA meeting about this. Unfortunately, I can't recall the reference at this point. If I recall correctly, their machine had suffered a collision and had affected the quality of the data. DIFABS apparently did a good job of hiding these errors.

>- Crystallographers using DIFABS are lazy.

I don't think personal judgements like these are the most appropriate criterion to apply when deciding a piece of software is scientifically justifiable.

>- Hiding of all kinds of sloppy experimental practice

I think DIFABS has great potential for this. If there are systematic errors in the data due to incorrectly set apetures or other data collection parameters, these are going show up in Fobs. The use of DIFABS assumes all systematic errors are due to absorption, which in the above example is not the case.

>- Correction factor ranges are often unrealistically large

I can't really comment on this.

>- Fudged (comfortably embellished) R-values

I have received some pressure from my structure clients to use DIFABS for the cosmetic improvements of R's. I have resisted such pressure. At one time, I did do a refinement on a structure without an absorption correction and with DIFABS. The R factor went from 0.08 to 0.05. I felt uncomfortable releasing the DIFABS refinement (which I did not). My opinion is that if the data or model as problems it is more honest to let these problems show up in the R factor, ORTEP's, esd's etc. To massage the data into giving a better R factor, is to mislead the readers of the literature (especially non-crystallographers) about the level of confidence they should place on the structure.

>
>        Supporters of DIFABS offer:
>
>- Easy to use (once implemented as a standard procedure)

Too easy to use. I colleague once told me that he knew someone who used DIFABS on structure where this person wanted to investigate difference electron densities. Beacause DIFABS the data to the model and the "standard crystallographic model" is typically spherical atoms undergoing 3 dimensional rectilinear harmonic motion, this person by using DIFABS was wiping out the very features of bonding density being looked for.

>- Reliable in the hands of people who know what they are doing

Possibly. One would have to be sure that the systematic errors were due to only absorption. Does everyone really look at each data point collected to make sure that the scan width didn't truncate any refections, correct apetures settings, contamination from TDS, extinction, Renniger effects and so on? Do "overworked" service crystallographers normally take this level of care?

>- Applicable also in cases that are for various reasons inaccessable
>  for 'exact' correction methods.

This is the only reason that I would use DIFABS. If the crystal was lost before psi scan could be taken, or if there were no suitable refections at high enough values of chi (or kappa).

>- The only practical method for the heavily overloaded service
>  crystallographer.

Is spending a couple of hours taking 3 or 4 psi scan profiles really going to cause a bottle neck? This has not been my experience. I spend more time checking the Laue symmetry and getting a good cell, than I do taking psi scans.

>- Experience with cases that are accessable for both correction methods
>  show that the results obtained with refinements based on data corrected
>  with both methods are nearly indistinguishable (even the ORTEP's).

This is something I have wanted to test for the last couple of years. I haven't been able to come across the right sample yet, however. I would need a highly absorbing crystal, which contained rigid molecules that weren't disordered. Take the data, take psi scans, and do an exact aborption correction and compare all 3 results. I think doing a rigid body analysis on the three structures would be informative. One could compare the values of TLS, deltaU's, libration corrected bond lengths. I think seeing if there is a correlation between the DIFABS 'absorption ellipsoid' and changes in the components of the T L and S tensors derived from the various refinements would be useful. We could answer the question does DIFABS primarily affect the T or L tensor, or both.

As a service crystallographer, I take which samples I get and I have not gotten a crystal which meets the above criterion. I think a compound like SnI4 might work, but I have no facilties to prepare compounds. Anybody want to collaborate on this?

Paul D. Boyle Dept. of Chemistry University of New Brunswick Bag
Service #45222 Fredericton, New Brunswick E3B 6E2 CANADA
boyle@mantis.chem.unb.ca

Followup by Bill Harrison | top 

From: chem91@Elroy.UH.EDU ("Bill Harrison (713)-743-2789")
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: 22 Apr 1994 15:12:23 GMT
Organization: University of Houston


>This is meant to start a (possibly heated) discussion on the use
>(although according to some abuse) of the program DIFABS (Walker &
>Stuart, 1983, Acta Cryst., A39, 158-166) as a method to correct for the
>effects of absorption. Such a discussion is rapidly becoming acute in

Could this be the first flame war in s.t.x? :=) 

>	The arguments against DIFABS that I know of are : 
>
>- Primary data are massaged and not recoverable from the deposited
>  Fo/Fc-tables

Good point. I guess that if DIFABS is used, then "real" Fo and "corrected" Fo should BOTH be deposited? 

>- Correction factor ranges are often unrealistically large

Good question. What *is* an acceptable correction factor range? 

>- Fudged (comfortably embellished) R-values

DIFABS can have a huge effect with inorganics -- I've seen R-factor reductions from 15% to 5% before, and it *seems* that the anisotropic thermal parameters are *much* more "sensible" (physically realistic) post-DIFABS. So long as you state that DIFABS was used, I don't think it's a problem... ? 

>
>        Supporters of DIFABS offer:
>
>- Applicable also in cases that are for various reasons inaccessable
>  for 'exact' correction methods.

Not just applicable, but perhaps essential. Most of the crystals I do are poorly (un)faceted lumps, and an analytical correction is just not on. I always collect psi-scan data if I can, but sometimes there just aren't any suitable reflections, and you can't tie up the machine for another day(?) just looking for them... 

>- Experience with cases that are accessable for both correction methods
>  show that the results obtained with refinements based on data corrected 
>  with both methods are nearly indistinguishable (even the ORTEP's).

I probably have several data sets I could test this on more systematically... BTW, I've seen some papers which report that psi-scans *and* DIFABS were applied...

Interesting questions!
Bill 

+-------------------------------------------------------------------------+
|   Bill Harrison -- University of Houston, TX --  CHEM91@JETSON.UH.EDU   |
|        "Only the extreme is interesting" -- Magnus Lindberg             |
+-------------------------------------------------------------------------+

Followup by Sean Parkin | top 

From: sp@oedipus.llnl.gov
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: Fri, 22 Apr 94 15:30:52 PDT
Organization: Lawrence Livermore National Laboratory, NCD

I have not used DIFABS, but I have used XABS and XABS2 (I wrote the latter). The program calculates a 4 dimensional scale factor from 24 coefficients that come from Fo^2 and Fc^2 differences. Real improvements in data i.e. the ability to find H atoms after correction for example are often quite dramatic. Assuming DIFABS can also actually improve data, rather than just make it closer to calculated data, and that a more rigorous technique is not practicable, then the method is fine for routine work.

Provided that whoever applies the correction properly understands what they are doing, and the limitations of it, then there shouldn't be a problem. The person who wanted to use it in electron density work, for example, shouldn't be allowed near it. Such programs have been abused in the interests of cosmetics, however, which is unfortunate.

The 24 coefficients generated by XABS2 (12 by XABS) are currently obtained in a standalone program. A better approach would be to refine them at the same time as x,y,z,Uij. The least-squares process would then obtain a fit to 'atomic parameters' and 'crystal parameters'. Then there would then be no need to deposit data other than the dataset itself and the absorption coefficients. Extinction is usually treated this way (and it also uses Fc in its derivation).

Sean Parkin

Followup by Jim Silverton | top 

Newsgroups: sci.techniques.xtallography
From: jvs@helix.nih.gov ("Jim Silverton")
Subject: Difabs
Date: Mon, 25 Apr 94 10:41:54 EDT
Organization: NIH

I must admit that my own instinctive reaction to Difabs is "CHEATING" but is this really true? All the arguments I have seen seem to be instinctive, stands-to-reasons or usable by "experts who know what they are doing".

Has the method ever been tested by refining data with known, introduced, systematic errors and shown not to produce systematic errors in the result or, even better, to increase real accuracy?

Even if the method proves harmless, it seems scientifically unjustifiable to me to use cosmetic changes producing no real improvement in an answer. This is especially true, as others have pointed out, if the deposited observations have been massaged and there is no way to recall the the raw data. I think that real, not cosmetic, improvement in the refined structure must be demonstrated to justify the continued use of Difabs. Preferably, tests would be done by several different laboratories and the results compared.

James V. Silverton

        NIH, Bldg. 10, Room 7N-309, Bethesda MD 20892, USA.

           Phone:301.496.1515           Fax:301.402.3404

Followup from Ton Spek | top 

Newsgroups: sci.techniques.xtallography
Path: dmp.csiro.au!mel.dit.csiro.au!munnari.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yeshua.marcam.com!news.kei.com!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!EU.net!sun4nl!ruuinf!ruunfs.fys.ruu.nl!faculty.chem.ruu.nl!spea
From: spea@faculty.chem.ruu.nl (Ton Spek)
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: 25 Apr 94 15:19:13 GMT
Organization: Physics Department, University of Utrecht, The Netherlands


"pboyle@unb.ca"  writes:
>Well, Ton, I guess I'll take the bait.  In general, I am against the
>use of DIFABS as a routine absorption correction.  I think it is bad
>practice to fit the data to the model rather than fitting the model
>to the data.  I try to use psi scan data to make an empirical
>absorption correction.  While the psi scan method is not perfect, it
>does have the advantage of being based on some experimental data.
>One must take care that the psi scan data is not contaminated with
>Renniger effects.

One of the problems with psi-scan data is unfortunately: How representative are the effects observed in individual psi-scans for what is happening with the complete dataset (an issue well addressed with DIFABS-type correction methods). E.g. One should be very carefull not to use a reflection that suffers from severe extinction. As a routine procedure we use three psi-scans to check the validity of our absorption correction.

>Is spending a couple of hours taking 3 or 4 psi scan profiles really
>going to cause a bottle neck?  This has not been my experience.  I
>spend more time checking the Laue symmetry and getting a good cell,
>than I do taking psi scans.

In our automatic data collection procedure (on the CAD4), three suitable reflections are selected 'on-the-fly' and intensites collected in 10 degree intervals.

  Ton Spek
  National Single Crystal Service Facility
  Utrecht University
  The Netherlands
  spea@xray.chem.ruu.nl

Followup by Nat Alcock | top 

From: msrbb@csv.warwick.ac.uk (N W Alcock)
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: 25 Apr 1994 16:44:31 +0100
Organization: University of Warwick, Coventry, UK

Some points in support of Ton Spek

1. Upper bounds can easily be set on the absorption correction range, by assuming that the minimum and maximum T are represented by rays traversing the shortest and longest dimensions of the crystal - comparison with analytical corrections shows that this is a good estimate. Some data from a couple of numbers of Acta last year for DIFABS corrections

Mu/mm-1	Tmax/Tmin	Est. range	Comment
0.08	1.008		1.02		DIFABS pointless
1.05	1.2		1.37		Under-corrected
4.0	2.16		1.86		Disturbing over-correction
14.9	1.12		6.93		Very worrying
0.09	1.39		1.03		This crystallographer should go back
					to school!

2. If the service crystallographer gets organised, then carrying out accurate face-indexed measurements of the crystal need take no more than 1/2 hour - certainly less than a good set of psi-scan measurements - and this can be done off the diffractometer.

3. If DIFABS does anything at all, it will equalise the mean anisotropic displacement parameters in all directions - fine (probably) if your molecule has no significant anisotropy of motion, but not otherwise. I can see **NO** justification for publishing thermal ellipsoid diagrams after DIFABS has been used.

Nat Alcock

Followup by Ton Spek | top 

Newsgroups: sci.techniques.xtallography
From: spea@faculty.chem.ruu.nl (Ton Spek)
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: 25 Apr 94 19:04:57 GMT
Organization: Physics Department, University of Utrecht, The Netherlands

To be sure: my ranking of methods for absorption correction is

a - analytical (e.g. ABSPSI, XTAL etc)
b - gaussian   (e.g. shelx76, XTAL, etc.)
c - difabs and similar approaches
d - sphere or cylinder approximation (when appropriate)
e - psi-scan corrections
Unfortunately, it is only in the case 'd-sphere' that part of the original data may be recovered (given the radius of the sphere) from the Fo/Fc tables.

Some of the problems I see with methods a & b are:

Our way to address such problems is to collect and correct additional psi-scan data and analyze their residual intensity fluctuation after correction. Wrong indexing is easily recognized in this way in the form of a correction (as a function of psi) that is out of phase. A similar test can be put in place for empirical correction techniques as well.

I am still hoping that somebody will come up with a nice gadget (preferably) on my diffractometer telling me what my nice cut-to-size oil-covered air-sensitive crystal approximates to in terms of h,k,l & d's..... Up to then it is very tempting to give DIFABS a try ... or can/should we persuade our customers to generate 'journal-quality' as opposed to 'xray-quality' crystals...sothat it is worth the effort to do an exact correction... Programs and computing time are no problem: An analytical correction and a DIFABS run both take only in the order of minutes on my workstation...

What would be realy interesting is when somebody could come up with a goodlooking structure (in terms of the highly valued R-factor) that nonetheless is wrong (geometry etc) due to the application of DIFABS.

In summary: in our setup: standard low-temperature with inert oil covering the air sensitive crystal, DIFABS does a very efficient job in providing the synthetic chemists with an answer (including in many cases hydrogen atom positions picked up from a difference map) as long as deformation densities are not the issue.

Ton Spek

Followup by Richard Harlow, posted by Steven Geib | top 

From: geib+@pitt.edu (Steven Geib)
Newsgroups: sci.techniques.xtallography
Subject: BANNING OF DIFABS
Date: 25 Apr 94 19:56:46 GMT
Organization: University of Pittsburgh

I'm not sure who to send this to in repsonse to the note which I just received but perhaps you can pass the following statement along to whomever is interested in this question of banning DIFABS. I wrote the following statement as a frequent reviewer of ACS journals when I noted that DIFABS was routinely being misused by more chemically-oriented authors. As a result of including this statement with many rejected manuscripts, I have found that the editors of several major ACS journals (apparently) no longer require my services. I understand that they are concerned that a great deal of good chemistry is being held up because of "less-than-perfect" crystal structures. I have much to say about that subject too but I'll save that for later. Needless to say, if I had know earlier that the journals would stop sending me manuscripts because my ethical standards were too high, I would have written this statement much sooner than I did!


				SYSTEMATIC ERRORS

		POLICY ON THE USE OF DIFABS AND SIMILAR PROGRAMS 
		    TO ARTIFICALLY CORRECT THE OBSERVED DATA
		    FOR SYSTEMATIC ERRORS SUCH AS ABSORPTION

Systematic errors in x-ray diffraction data are virtually unavoidable. Absorption effects are probably the best recognized source of systematic errors and the ones which have received the most attention, at least in terms of the number of programs which are available to 'correct' for these errors. Extinction effects are generally a minor source of errors for organic and organometallic compounds, but can be a significant source for inorganic structures. But again, it is a source that almost everyone who has been in business for some time recognizes and can deal with.

There are, however, other sources of systematic errors which can potentially produce effects far more serious than those from absorption but which are generally ignored or improperly corrected for, not that I am particularly happy with all of the present algorithms for absorption corrections. Sources included in this category are, for example, poor diffractometer alignment, poor crystal alignment, the non-uniformity of the x-ray beam on all diffractometers with incident-beam monochromators, and decomposition effects. I have been amazed to find many institutions where routine alignments are never done, or even if they are done, they are completed only with the package of programs available from the manufacturer. Once the computerized instrument alignment is done, very few people check to see that the beam is actually centered on the crystal; on most instruments that I have checked, the beam shines on only part of the crystal and the relative positions of the tube tower, beam collimator, and the circle must all be adjusted. This is made additionally critical because the beam from an instrument with an incident-beam monochromator is non-uniform, generally taking on a pencil shape (with either a vertical or horizontal sense depending on the monochromator orientation) with a uniform beam width of no more than 0.2-0.3 mm. Thus, if a diffractometer is not aligned properly, a major portion of the beam may be just catching an edge of the crystal in one orientation, but only a corner in another. Huge systematic errors may result. Even if the diffractometer is aligned, improper optical alignment of the sample crystal can also lead to serious errors; improper alignment may easily result, for example, if the scope is knocked out of its proper position. Crystals larger than the uniform beam size are also a source of errors. Decomposition effects too are seriously under-estimated by monitoring low-angle 'standard reflections'; experiments show that, on average, a fifty percent drop in high-angle intensities (MoKa, 2t = 50 deg) is accompanied by only a 10% drop in the type of low-angle reflections that are normally used for standards (2t = 10-20 deg). Furthermore, there are wide variations in the effects on individual reflections within any 2t shell such that no correction is ever possible or appropriate!

Programs such as DIFABS have been developed to artificially correct for all systematic errors in the data no matter what their source. Before its development, systematic errors in the intensity data were 'refined' into the anisotropic thermal parameters, and sometimes were so severe that all of the atoms in a molecule would have nearly uniform ellipses elongated in the same direction. Systematic errors were thus sometimes easily recognized by the crystallographer, or failing that, by the reviewers. DIFABS, which attempts to systematically correct all of the 'observed' intensities to better fit the calculated intensities, is an ingenious program to help determine the presence of systematic errors, but is being seriously, and increasingly misused as an 'absorption' correction program. The immediate effect of its use is that poor data, full of various systematic errors, can be made over to look like perfectly reasonable data. Its use in this manner represents a type of scientific fraud in which the 'observed' data is modified to agree with a proposed model, potentially hiding a number of serious systematic errors from reviewers, journal readers and database personnel. The program has become the lazy man's panecea for dealing with systematic errors; there is no need to deal with the root causes of the systematic errors or, in the case of absorption, to attempt a more appropriate correction. For these reasons, I will no longer accept manuscripts for publication in which DIFABS has been used in this manner because I have no real way of evaluating its effects on the data.

My experience with DIFABS, which in fact is quite extensive, has shown that it not only corrects for the systematic errors noted above but also 'corrects' for the anisotropic motion of the heaviest atoms in the structure. If there is only one heavy, it becomes spherical in shape; the light atom ellipsoids are also distorted in the sense that the heavy atom motion is 'subtracted' from each of them. Thus, for the case of a single heavy atom, all of the thermal parameters are distorted. Because DIFABS also allows for a '2-theta' correction, even the absolute values of the thermal parameters may be affected. With multiple heavy atoms, for example a polytungstate, the effects on the thermal parameters are much less dramatic since the heavy atoms are likely to have very different thermal motions to start with. As might be expected from published surveys of multiple intensity-data sets, refinements with and without the application of DIFABS show little change in bond distances and angles when applied to reasonably good data sets. I (or anyone else that I know of) have not made a systematic study of its effect on the refinements of poor data sets with large built-in systematic errors.

DIFABS is a fantastic program when used as a diagnostic tool and, as such, should be a part of every crystallographers bag of tricks. Its use as a routine quick fix for poor intensity data cannot, however, be scientifically justified.

Richard Harlow
harlow@esvax.dnet.dupont.com

Followup by John Huffman | top 

From: huffman@indiana.edu (John C. Huffman)
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Organization: Indiana University
Date: Tue, 26 Apr 1994 00:56:12 GMT

Several interesting and valid points have been made concerning the absorption problem. As stated by Ton, one option is to require that only "publication quality crystals" be submitted--but I doubt that one would make many freinds doing this. We too have problems with the shape of a crystal, expecially when we have one that is quite air sensitive, etc., and we end up with a terribly irregular cyrstal in a glob of grease. One possibility I have toyed with is to come up with some amorphous material of approximately the same absorption coefficient as the crystal in question and shape a sphere around the sample. Presumably one could then use a spherical correction? Let's face it, one normally ignores the fiber, the glue, and miscellaneous other garbage that presumably one should worry about. Lucky for us the problem is overdetermined by a factor of 5:1 to 15:1 and most errors are averaged out to a large extent.

With the above in mind I can state that I always TRY to use the indexed faces for the corrections. To verify the proper indexing, and to locate any gross errors, we have a program that is used to generate the absorption correction program input, and will draw the crystal as mounted on the glass fiber. We can thus input the faces and distances, the diameter of the fiber, and, using a cheap graphics program have a rotating image of the mounted crystal to compare with the one in the microscope. Believe me, this has saved me on several occasions. Once satisfied with the comparison of the real and generated crystal, the absorption correction input is saved for later use. In conjuction with this, I have mounted video cameras on the goniostat's microscope and recorded the images of the crystal. By generating a hard copy of the calculated crystal drawing one can adjust the size to see how well the two match. I hope to be able to include images of the crystals as part of the permanent "raw data" file by the end of this year. The cameras themselves are quite cheap (under $500 for one with an interface card to an IBM PC). In addition to having a permanent record, one has a series of images which can be used to carefully measure the crystal (I use a 1/64" steel ball image as the standard). How do others obtain accurate measurements?

=============================================
John C. Huffman              Senior Scientist
Indiana University Molecular Structure Center
Bloomington, IN 47408-4001                USA
---------------------------------------------
huffman@indiana.edu            (812) 855-6742
=============================================

Followup by Tony Linden | top 

Newsgroups: sci.techniques.xtallography
From: alinden@oci.unizh.ch (Dr. Tony Linden)
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Organization: University of Zurich
Date: Tue, 26 Apr 1994 08:53:14 GMT

Ton Spek and his respondents raise some interesting points about DIFABS. Perhaps we need an IUCr official recommendation on the use of this procedure. Unfortunately, I suspect that this Newsgroup medium is not read by sufficient crystallographers (yet) to canvas a good spectum of opinion.

Having read some of the responses, I will throw in my, perhaps naive, comments.

Firstly, it would be interesting to hear the opinions of the creators of DIFABS, Walker & Stuart in light of the 11 years of experience with their procedure. I do not know of any comments they have made since the publication of their paper. If anyone does know of such comments, I would be pleased to hear them.

Surely when any type of absorption correction is applied, the primary data is altered in such a way that the original raw data cannot be recovered from deposited material. For example, psi scans are usually applied to Fo (or Fo2) during data reduction. Therefore DIFABS is not the only culprit here. Perhaps we need to deposit data containing Fo, Fo(corr) and Fc in the CIF files (as suggested by Bill Harrison). In any event, if laboratories archive their own data properly, the original uncorrected data should always be available from the authors.

Any systematic errors that are of a spherical nature and equally affect all reflections (perhaps there are none of this type) will not be noticed by DIFABS. Also, when DIFABS calculates its absorption profile, the weak reflections and any reflections where Fo and Fc are grossly different are ignored, so that those reflections that contain the greatest systematic, or extinction errors are not influencing the DIFABS correction. Admittedly weaker extinction effects may be mopped up by DIFABS. As Ton Spek points out, psi scans are not infallible if one of the scans happens to hit a reflection suffering from extinction.

Systematic errors often show up in the Goodness-of-fit indicator at the end of the refinement. A misaligned monochromator crystal (the crystal, not the angle) resulted consistently in GoF values in the range 2.0-3.0, although the R-factors were good. DIFABS did not help to reduce GoF by polishing out the systematic error (ie DIFABS does not hide as much as some might think). After the monochromator was correctly set, GoF was found to be in the range 1.0-1.5 for most structures.

It actually takes more refinement time to do a DIFABS correction, if it is done properly, although time spent at the diffractometer is obviously less if crystal or psi scan measurements are not done.

People seem concerned that DIFABS may improve R too drastically (examples given were from 0.08 to 0.05 or from 0.15 to 0.05). While this may give an uneasy feel, we really should compare the improvement in R made on the same structures by other correction methods before discarding such changes as inappropriate. I have seen the R for heavy atom structures undergo similar changes with a numerical correction.

I see that Acta Cryst now frowns upon "Transmission coefficients" greater than 1.1. This is awkward, since DIFABS normalises the correction coefficients (not really a transmission coeff, but the effect is the same) to give a mean of approx 1.0 and Tmax of 1.2 is not uncommon.

Some obviously feel that the thermal parameters are "rounded" or even meaningless after a DIFABS correction. Maybe so, although I have experienced cases where the thermal ellipsoids expanded (ever so slightly) as a result of the correction, so nothing was being "hidden".

One of the biggest dangers for the misuse of DIFABS is applying it inappropriately, or at the wrong stage of the refinement, and this is possibly where the ill-informed worker has fallen into the trap. Since the correction relies on the adequacy of the structural model, it is imperative that the model be as near to correct as possible. This includes having as many of the H-atoms in the model as possible, as well as having any disorder (most important!) sorted out BEFORE the correction is applied. Additionally the correction should be applied to the model with isotropic temperature factors, since after an anisotropic refinement, the thermal parameters will mop up some of the effects of absorption. This does not mean, of course that an anisotropic refinement cannot already have been done, as an aid to finding H-atoms, for example, but the model should be converted back to isotropic before the correction.

The above comments on aniso thermal parameters raises another interesting question. Are the thermal parameters after the DIFABS correction really meaningless, as suggested by Nat Alcock, or is the uncorrected data producing meaningless thermal parameters, because of the (ignored) effects of absorption being mopped up in the guise of thermal motion?

I believe that DIFABS has its place if used appropriately. The difficulty is to educate people, especially the casual non-crystallographer who has a diffractometer in his/her lab as a purely analytical tool, as to when and when not to employ such a procedure. DIFABS should not be used ad hoc, but only when the crystal morphology and mu warrant an absorption correction. For example, people should not blindly accept a large correction for a nice isotropic organic crystal. Further, if the opportunity exists to collect psi scans, this should be done. Both Enraf-Nonius and Rigaku/MSC systems (at least) have this option automated into their data collection programs. Indexing the faces of the crystal is also recommended (perhaps the best procedure), but the problem often occurs that the crystal is ill-faceted, or sealed in a capilliary with a liquid, which makes such measurements impractical - we do not always get perfectly formed crystals delivered to us, do we?

Unusual features in structures are often queried if DIFABS has been employed. In my experience, the unusual feature was not modified in any way by employing uncorrected reflection data, so I do not believe that DIFABS is capable of introducing significant "new features" or wiping out undesirable ones, PROVIDED the program is used properly (see above). Here, I am willing to accept evidence to the contrary.

A final question for thought and comment. In cases where an absorption correction is definitely warranted and psi scans or numerical methods are not practical, what is the best approach?

I am willing to be shot down in flames for my opinions (I am always ready to learn), but at least I hope this will enliven the discussion...

I think that the end result of this discussion will show that some of us need to get together to undertake a systematic study of the effects on all aspects of a structure of the various correction methods, for both weakly and strongly absorbing materials.

Tony.

Dr. Anthony Linden                             Tel:    +41 1 257 4228
Organic Chemistry Institute                    Fax:    +41 1 361 9895
University of Zurich                           E-mail: alinden@oci.unizh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland

Followup by Robert Dewar | top 

From: dewar@cs.nyu.edu (Robert Dewar)
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS - SHOULD IT BE BANNED ??
Date: 26 Apr 1994 10:03:35 -0400
Organization: Courant Institute of Mathematical Sciences

Reading the discussion on DIFABS, it seems hard to escape the obvious conclusion, which is that full disclosure is required. As long as people know that DIFABS has been used, then they can apply their own valud judgments to the resulting data. Cearly it would be valuable to deposit both the original (true or messy, depending on your point of view) data and the DIFABS'ed data (cleaned up or faked, depending on your point of view).

(incidentally, I find it entertaining that Internet allows me to follow what's going on in the field a little bit these days. THe last work I did in crystallography was my Phd thesis work (early automation of direct methods using the programs MAGIC and MAGIA), and it is interesting to see how far the field has come in making the transition from structure determination being quite a challenge to being quite routine at least in simple cases :-)

Robert Dewer

Followup by Jim Silverton | top 

Newsgroups: sci.techniques.xtallography
From: jvs@helix.nih.gov ("Jim Silverton")
Subject: DIFABS
Date: Thu, 28 Apr 94 09:28:35 EDT
Organization: NIH
Given the misgivings expressed by several people concerning the deposition of massaged as opposed to raw data, perhaps it is time for journals like Acta to consider asking for the diffractometer output. An initial reaction might be that the volume of data would get out of hand but the sizes of such files are not as large as one might initially guess. It would not be all that much trouble to set up an FTP server for access over the Internet. After all, the ACS already has one.

To illustrate file sizes, the CAD4.DAT file for a steroid occupied 40 960 bytes. When compressed with gzip the file size was 11 238. Neither would take very long to access and the zipped file is really very small.

Changing the subject! Crystallographers might like to get the low down on themselves by reading a new "non-fiction" book: "The Billion Dollar Molecule" by Barry Worth, Simon and Schuster, 1994, ISBN: 0671723278. I will leave the value as an exercise to the reader!

                          James V. Silverton
        NIH, Bldg. 10, Room 7N-309, Bethesda MD 20892, USA.
           Phone:301.496.1515           Fax:301.402.3404

Followup by Howard Flack | top 

Newsgroups: sci.techniques.xtallography
From: flack@scsun.unige.ch (FLACK Howard)
Subject: Re: DIFABS
Organization: University of Geneva, Switzerland
Date: Thu, 28 Apr 1994 15:12:49 GMT

In article jvs@helix.nih.gov ("Jim Silverton") writes: 

> Given the misgivings expressed by several people concerning the deposition 
> of massaged as opposed to raw data, perhaps it is time for journals like 
> Acta to consider asking for the diffractometer output.
No, Jim, not the native diffractometer output files which often do not contain much of the information that one wants to save and are often in binary with undocumented fields. I suggest the following:

As a by product of a collaboration with people down the road where we had to exchange raw data files coming from different machines, we produced a FORTRAN programme that lets you put the raw data into a standardized form e.g. CIF. Here's some of my blurb about DIFRAC (marked up in HTML) taken from the file on the "Crystallography in Europe" World Wide Web server on URL:

http://www.iucr.org/cww-top/crystal.index.html

(Old address was: http://www.unige.ch/crystal/crystal_index.html)

You can get the programme from me free by e-mail. When I have 5 spare seconds one day, I'll set it up on an anonymous ftp server.

DIFRAC blurb

DIFRAC converts single-crystal diffractometer output files as produced by manufacturers' software into a standardized instrument- independent form consisting of a clear, complete, well-documented record of the sample and the diffraction measurements performed upon it. Information not already available in the manufacturers' diffractometer files is obtained in an interactive question-and-answer session. The software is written in a modular way. Available modules can treat Enraf-Nonius CAD-4, Philips PW1100, Siemens P21 and Stoe DIF4 single-crystal diffractometers, and produce output in CIF (Crystallographic Information File) or SCFS (Standard Crystallographic File Structure) format.

Reference: 'DIFRAC, single-crystal diffractometer output-conversion software' by H.D. Flack, E. Blanc and D. Schwarzenbach (1992), J. Appl. Cryst.25, 455-459.

DIFRAC is distributed by e-mail. Contact flack@scsun.unige.ch for a copy.

H.D.F.

 H. D. Flack

 Laboratoire de Cristallographie   !  Telephone        [+[41] 22] 702 62 49
 University of Geneva              !  Telefax          [+[41] 22] 781 21 92
 24 quai Ernest-Ansermet           !  Telex            ch-42 11 59 siad
 CH-1211 Geneve 4                  !  e-mail(rfc-822)  flack@scsun.unige.ch
 Switzerland                       !

Resend/Followup by Jim Silverton | top 

Newsgroups: sci.techniques.xtallography
From: jvs@helix.nih.gov ("Jim Silverton")
Subject: DIFABS (revision of message re. FTP and deposition of files)
Date: Thu, 28 Apr 94 11:31:51 EDT
Organisation: NIH

Sorry everyone!

Because of a transmission fault the file sizes given in the previous version of this post are quite incorrect. The transmission between the VAX and the UNIX machine failed but I did not get an error message thus the data file was somewhat smaller than reality. I don't think the emphasis is changed by the correct figures and I will post the rest of the message as before. The only changes are the file sizes

Given the misgivings expressed by several people concerning the deposition of massaged as opposed to raw data, perhaps it is time for journals like Acta to consider asking for the diffractometer output. An initial reaction might be that the volume of data would get out of hand but the sizes of such files are not as large as one might initially guess. It would not be all that much trouble to set up an FTP server for access over the Internet. After all, the ACS already has one.

To illustrate file sizes, the CAD4.DAT file for a steroid occupied 289408 bytes. When compressed with gzip the file size was 79839 bytes. Neither would take very long to access and the zipped file is really quite small.

Changing the subject! Crystallographers might like to get the low down on themselves by reading a new "non-fiction" book: "The Billion Dollar Molecule" by Barry Worth, Simon and Schuster, 1994, ISBN: 0671723278. I will leave the value as an exercise to the reader!

                          James V. Silverton
        NIH, Bldg. 10, Room 7N-309, Bethesda MD 20892, USA.
           Phone:301.496.1515           Fax:301.402.3404

Followup by Robet Dewar | top 

From: dewar@cs.nyu.edu (Robert Dewar)
Newsgroups: sci.techniques.xtallography
Subject: Re: DIFABS (revision of message re. FTP and deposition of files)
Date: 29 Apr 1994 10:39:37 -0400
Organization: Courant Institute of Mathematical Sciences

To put some perspective on file sizes and costs, the cost of sufficient disk space to accomodate the compressed raw CAD4.DAT file is about $0.02

just my two cents :-)

Followup by Jim Silverton | top 

Newsgroups: sci.techniques.xtallography
From: jvs@helix.nih.gov ("Jim Silverton")
Subject:  Difabs and absorption corrections
Date: Mon, 2 May 94 10:22:19 EDT
Organistations: NIH

A major problem with analytical absorption corrections is the horrible real facts of life. I often have to carve out an irregular chunk and sometimes also have to seal it up with solvent etc. I would like to grind a sphere but many organic crystals disintegrate completely when I try. I have done a recent count and have found that only one of the last six crystals I have looked at could be said to have a shape accurately describable by crystal faces. Maybe a photographic technique could be developed but I suspect Psi scans will remain popular.

The mathematical or tabular form of an empirical absorption correction like a psi-scan is quite simple and should be deposited in conjunction with structure factor tables, especially if the journals won't take raw data or some modification as suggested by Howard Flack. Incidentally, if Howard's program or something like it were to be required, I think it should be the journal's responsibility to make the program readily and continuously accessible (in other words FTP). However, I would really prefer an ASCII file compressed by some GENERIC program like gzip.

                          James V. Silverton
        NIH, Bldg. 10, Room 7N-309, Bethesda MD 20892, USA.
           Phone:301.496.1515           Fax:301.402.3404

Followup by Howard Flack | top 

Newsgroups: sci.techniques.xtallography
From: flack@scsun.unige.ch (FLACK Howard)
Subject: Re:  Difabs and DIFRAC
Organization: University of Geneva, Switzerland
Date: Tue, 3 May 1994 16:54:37 GMT


In article jvs@helix.nih.gov, jvs@helix.nih.gov ("Jim Silverton") writes:
> Incidentally, if Howard's 
> program or something like it were to be required, I think it should be the 
> journal's responsibility to make the program readily and continuously 
> accessible (in other words FTP).
OK the penny has dropped. DIFRAC is now available by anonymous ftp from our site since I can not imagine a journal providing a service like that. The "instructions" are given below. I did not gzip the files since the programme is supposed to run on all sorts of platforms and the compression programmes seem to be rather system dependent. (But may be you could convince me otherwise). Whats available are ASCII source code files and documentation.

I also put the necessary hypertext links for DIFRAC into the appropriate place in the "Crystallography in Europe" WWW server on URL:

http://www.iucr.org/cww-top/crystal.index.html

which is much less cumbersome to use than direct anonymous ftp

README.1ST

DIFRAC Single-Crystal Diffractometer Output-Conversion Software

By H.D. Flack, E. Blanc and D. Schwarzenbach

Contact: H.D. Flack E-mail: flack@scsun.unige.ch

Reference: J. Appl. Cryst. (1992) 25, 455-459

Distribution: By anonymous ftp

from: ftp://ftp.unige.ch/pub/soft/crystal/difrac/

The description material is in the file [ancill.mat]. Recover this file first and read the text to find out which files you need for your application.

All files are ASCII and unpacked.

Followup by John Huffman | top 

Newsgroups: sci.techniques.xtallography
From: huffman@indiana.edu (John C. Huffman)
Subject: Re: Was DIFABS (hand laser alignment of single xtal diffractometers)
Organization: Indiana University
Date: Tue, 3 May 1994 20:40:34 GMT

In article <199405010918.AA01427@sol.dmp.csiro.au>, lachlan@dmp.csiro.au (Lachlan Cranswick) writes:

|> 
|> [stuff deleted]
|> 
|> I'm not sure if this type of thing is old news to people but 
|> the research XRD lab next door recently had alignment 
|> problems with an re-hashed Siemens AED
|> 3 circle diffractometer.  They used a A$100 pencil laser normally
|> applied to lectures and seminars to realign the diffractometer
|> while it was potentially in a horrible state.
|> 
|> While this diffractometer was/is regularly aligned, during a trip
|> home from having a modern tube tower installed, 
|> the tube tower base  suffered some
|> structural damage.  This threw the factory alignment settings out
|> and was potentially a major job to realign (the factory
|> alignment tools were not available).
|> 
|> Instead a pencil laser was purchased that made the alignment
|> very easy :-  
|> 
|> The tube tower was moved out of the way.
|> 
|> The laser was mounted on a retort stand.
|> 
|>  [more deleted]
|>
|> Lachlan Cranswick  -  CSIRO
I think the use of lasers is not routine simply because there is not enough of a market to entice one of the instsrument makers into making a good laser alignment tool. Because I have quite a few Pickers to align (I believe in periodic checks also!) I had two rather simple- minded laser alignment devices made some time ago. One is attached to the top of what was an X-ray tube and is fixed with a 45 deg mirror so that one can place the laser assembly in the tube housing as though it were an X-ray tube (of course you can't adjust the take-off angle). A second laser is affixed to the track that the scintillation counter fits on on the goniostat. The latter can be used without messing around with the tube. A small (5 mm) front surface mirror mounted on a gnoiometer head, as well as a 1/64" sphere are used in various ways to check the beam path. Believe me, I feel much safer when I do the alignment using this system!

Another trick I use on the Picker systems is a specially built "surface plate" that can be used in conjunction with precision machinist dial indicators and height guages to assist in the preliminary alignment. I have had several specialized alignment tools constructed over the years which make it easy to perform the crude alignment.

For the final alignment we use BeOAc crystals and ruby spheres. We have installed a command for the goniostat which will measure a given reflection in the eight accessible locations and tell the user what adjustments need to be made (see the International Tables--I think they give the formulas?).

In any case, I certainly agree that careful absorption corrections are meaninless if you don't have a grasp on the other errors that are present. Carefully measuring and indexing the faces does no good if the crystal is not fully bathed in a uniform beam.

=============================================
John C. Huffman              Senior Scientist
Indiana University Molecular Structure Center
Bloomington, IN 47408-4001                USA
---------------------------------------------
huffman@indiana.edu            (812) 855-6742
=============================================

Message Sent by Lachlan Cranswick | top 

Newsgroups: sci.techniques.xtallography
From: lachlan@dmp.csiro.au (Lachlan Cranswick)
Subject: TENTATIVE "BANNING OF DIFABS" SUMMARY?
Organization: CSIRO Division of Mineral Products, Melbourne, AUSTRALIA
Date: Wed, 8 Jun 1994 02:30:52 GMT

Hi All,

Recently there are was a discussion about the "BANNING OF DIFABS"

This was a most un-Usenet like discussion. Instead of people using the time to insult and "flame" eachother, the discussion "seemed" to arrive at compatable conclusions.

There was a comment during the DIFABS discussion that this "Newsgroup is not read by sufficient crystallographers (yet) to canvas a good spectum of opinion". Thus, would participants and readers be interested in generating a summary for submition to a conventional paper journal (recommendations?). This would help assist in the wider propogation of the concerns and conclusions. It would also be nice to think that discussions like this could become routine in the not to distant future.

Tentative Summary of DIFABS Discussion.

Do people think the following is a fair summary of the discussion or are there things to add or delete? This is a mish-mash of comments and statements made by participants and no doubt could be put in better categories and priorities.

There is concern over the use and misuse of absorption correction software such as DIFABS to determine crystal structures.

Via discussion, this has resulted in the following actions and conclusions? :-

One of the results from the IUCr CPD Rietveld Refinement Round Robin I, - (R.J. Hill, J Appl. Cryst. (1992) 25, 589-610) follows which is semi relevant :-

"It is appropriate to mention here that an error was discovered in the DBW3.2S code during in-house analysis of the m-ZrO2 data submitted during the course of Part II of the project. This error involved an incorrect assignment of the multiplicity of reflections of the type 0kl in the point group 2/m. Although not producing a catastropic refinement result, the profile agreement indices were larger than normal and the derived crystal structure parameters showed a wider than expected deviation from the accepted values. All participants who used this software wre informed of a cure to the problem and asked to submit a reanalysis of their m-ZrO2 data. This is a clear example of the important role that intercomparison projects can have in highlighting problems with refinement software."

Lachlan.

Lachlan Cranswick  -  CSIRO     _--_|\  lachlan@dmp.CSIRO.AU 
Division of Mineral Products   /      \ tel +61 3 647 0367 
PO Box 124, Port Melbourne     \_.--._/ fax +61 3 646 3223 
3207 AUSTRALIA                       v

Followup by Sean Parkin | top 

From: sp@oedipus.llnl.gov
Newsgroups: sci.techniques.xtallography
Subject: Re: TENTATIVE "BANNING OF DIFABS" SUMMARY?
Date: Thu, 09 Jun 94 11:14:41 PDT
Organization: Lawrence Livermore National Laboratory, NCD

In article <1994Jun8.023052.17411@dmp.csiro.au>, writes:

A few thoughts in response to Lachlan's note:

>       -  More emphasis, if possible, on "customers" providing journal
> quality crystals for analysis.

If the only crystals to ever see a diffractometer were of good quality, current chemistry would be at the state it was at about 30 years ago and structural biology as a science would be non-existent. It would be more appropriate to emphasize the fact that data-collection strategies can be altered to accomodate the crystal at hand. See Hope, H. 'Progress in Inorganic Chemistry' vol. 41 pp 1-19.

>       -  Time effective, accurate face-indexed measurements of the
> crystal can be done off the diffractometer and should be encouraged.

In the few cases that crystals have nice faces.

>       -  DIFABS (and similiar software) is a legitimate(?) package
> in the crystallographer's "back of tricks" - but users should
> be made aware of its abuses and effects (e.g. software correction for
> poorly installed and aligned equipment, inappropriate stages of the
> refinement to be using the correction software).

Diffractometer misalignment is lamentable, but it is not a function of the absorption correction routine used. Banning of DIFABS etc. is not an appropriate response to poorly maintained equipment.

 
>        - Use of such correction software should be explicitly stated 
>              with "real" Fo and "corrected" Fo submitted.

I agree. However, since absorption, extinction, TDS etc. are real physical phenomena, wouldn't it be better to account for them in Fc or Fc^2 ?

>       -  There is no justification for publishing thermal ellipsoid
> diagrams after DIFABS has been used.

This response seems to be the result of emotion. A satisfactory conclusion would come from a comparison of thermal parameters obtained after DIFABS (or XABS2) and after a rigorous face-indexed correction. It is likely that in many cases, that empirically corrected data will produce thermal ellipsoids that are closer to reality than uncorrected data. Does that mean that thermal ellipsoid plots for structures with no absorption correction should be (uncritically) banned ?

 
>       -  Methods for the easy, safe checking ...

Anything that stresses increased respect for the experiment is to be encouraged. There is still no correlation between it and absorption correction algorithms.

Sean Parkin

sp@oedipus.llnl.gov

Followup by Nat Alcock | top 

 
From: msrbb@csv.warwick.ac.uk (N W Alcock)
Newsgroups: sci.techniques.xtallography
Subject: Re: TENTATIVE "BANNING OF DIFABS" SUMMARY?
Date: 10 Jun 1994 12:55:16 +0100
Organization: University of Warwick, Coventry, UK

In article <2t7p5f$g6h@lll-winken.llnl.gov>, sp@oedipus.llnl.gov writes: 

> Lachlan Cranswick  writes:
>>       -  There is no justification for publishing thermal ellipsoid
>> diagrams after DIFABS has been used.
>
>This response seems to be the result of emotion.  A satisfactory
>conclusion would come from a comparison of thermal parameters obtained
>after DIFABS (or XABS2) and after a rigorous face-indexed correction.
>It is likely that in many cases, that empirically corrected data will
>produce thermal ellipsoids that are closer to reality than uncorrected
>data.  Does that mean that thermal ellipsoid plots for structures with
>no absorption correction should be (uncritically) banned ?
>

Seeing that DIFABS works by setting the mean values of the anisotropic displacement parameters to be equal, it follows that ANY systematic variation in anisotropy will be eliminated by its use. One can clearly see in published papers that anisotropic plots for over-DIFABS-corrected structures show notably round atoms. Neglect of a proper absorption correction where the problem is moderate (say Tmin = 0.8; with 10% variation in T) should not have a disastrous effect on the thermal parameters. If severe absorption is neglected, especially if highly anisotropic, the thermal parameters will undoubtedly be meaningless - but the structure should probably not be published in that state at all!

> The small proportion of crystals with measurable faces

One can only use personal experience to judge how big a problem this really is. Of the crystals we see in our lab, I would estimate that at least 85% have well defined and measurable faces. In my view, the comment is not justified.

Nat Alcock, University of Warwick
msrbb@snow.csv.warwick.ac.uk

Followup by Howard Flack | top 

Newsgroups: sci.techniques.xtallography
From: flack@scsun.unige.ch (FLACK Howard)
Subject: Re: TENTATIVE "BANNING OF DIFABS" SUMMARY?
Sender: usenet@news.unige.ch
Organization: University of Geneva, Switzerland
Date: Fri, 10 Jun 1994 13:33:37 GMT

In article 17411@dmp.csiro.au, lachlan@dmp.csiro.au (Lachlan Cranswick) writes: 

>       -  The DIFRAC 
software by Flack and Schwarzenbach for

should read

by Flack, Blanc and Schwarzenbach for

> converting binary diffractometer data from several commercial
should read

converting (possibly binary) diffractometer data .......

H.

Followup by Sean Parkin | top 

From: sp@oedipus.llnl.gov
Newsgroups: sci.techniques.xtallography
Subject: Re: TENTATIVE "BANNING OF DIFABS" SUMMARY?
Date: 10 Jun 94 16:13:26 GMT
Organization: Lawrence Livermore National Laboratory, NCD
I wrote:

> > The small proportion of crystals with measurable faces
In article <2t9kb4$5lb@crocus.csv.warwick.ac.uk>, writes:

 
> One can only use personal experience to judge how big a problem this really
> is.  Of the crystals we see in our lab, I would estimate that at least 85%
> have well defined and measurable faces.  In my view, the comment is not
> justified.
In the last 3 years, I have had direct experience with well over 100 small molecule crystal structures. Exactly one (1) of which had measurable faces. Absorption correction for this crystal was of the rigorous face indexed type.

Sean Parkin
sp@oedipus.llnl.gov

Followup by Lachlan Cranswick | top 

Newsgroups: sci.techniques.xtallography
From: lachlan@dmp.csiro.au (Lachlan Cranswick)
Subject: 2ND TENTATIVE DIFABS SUMMARY
Organization: CSIRO Division of Mineral Products, Melbourne, AUSTRALIA
Date: Fri, 1 Jul 1994 14:10:40 GMT

2ND TENTATIVE DIFABS SUMMARY
A discussion on whether the absorption correction software, DIFABS, should be banned, was started on April 20th by Tom Spek and seems to have reached compatable conclusions. It is hoped a agreed summary can be generated and submitted to an appropriate journal.

The aims of a journal article would be to raise awareness of both the potential applications of this newsgroup, and the DIFABS issue.

There is concern over the use and misuse of absorption correction software such as DIFABS to determine crystal structures.

Via discussion, this has resulted in the following actions and conclusions :-

Where this is not practical, data-collection strategies can be altered to accomodate the crystal at hand. i.e.:-

Use of such correction software should be explicitly stated with the following submitted :-

[which are recommended - 1 or all?]

Methods for the easy, safe checking and alignment of single crystal diffractometers are available and possible with even home made equipment. For instance, using equipment such as cheap hand lasers, it is potentally a trivial and safe job to check that the X-ray beam is bathing the entire crystal. Routine checking of equipment should be encouraged to minimize the collection of poor quality data. Thus minimizing concerns about the abuse of absorption correction software to correct for systematic errors caused by poor alignment and/or partial irradiation of the crystal.

A "round robin" comparing the effects of "various?" absorption correction software vs uncorrected results would be very beneficial.

Comments Please :-

Lachlan.

-- 
Lachlan Cranswick  -  CSIRO     _--_|\  lachlan@dmp.CSIRO.AU 
Division of Mineral Products   /      \ tel +61 3 647 0367 
PO Box 124, Port Melbourne     \_.--._/ fax +61 3 646 3223 
3207 AUSTRALIA                       v

Request by Steve Trail for Location of DIFABS Software | top 

From: st22@cornell.edu (Steve Trail)
Newsgroups: sci.techniques.xtallography
Subject: DiFabs - where can I find it?
Date: Thu, 2 Feb 1995 20:02:32 UNDEFINED
Organization: Cornell University

Is DiFabs available to the general public? If so, where? Any comments on it?

Thanks in advance,

Steve

                      ******************
                      Steven Scott Trail
                      Chemistry Department
                      Cornell University
                      st22@cornell.edu
                      ******************

Followup by John Bridson | top 

From: jbridson@kean.ucs.mun.ca John Bridson
Newsgroups: sci.techniques.xtallography
Subject: Re: DiFabs - where can I find it?
Date: 6 Feb 95 14:35:57 -0230
Organization: Memorial University. St.John's Nfld, Canada

Someone else has already responded with help on getting DiFabs - you also asked for comments. There was an outbreak of grumbling and suspicion a year or so ago - the results are sometimes a little too good to be true perhaps ?? I have used it in difficult structures as a last resort - it has sometimes permitted me to get disordered groups sorted out at the very end of least-squares cycling so I guess it sometimes has a real effect on fit. HOWEVER, for every occasion it has worked I can cite two or three times when the results have been very dubious and I would prefer to use other absorption corrections if at all possible, even if the R values aren't quite as good. The bottom line I guess is use it with suspicion and great caution !

Followup by Lawrence Henling | top 

From: lmh@celia.caltech.edu (Henling, Lawrence M.)
Newsgroups: sci.techniques.xtallography
Subject: Re: Absorption correction programs
Date: 22 Feb 1994 12:08 PDT
Organization: California Institute of Technology

In article <94052.111949RXB116@psuvm.psu.edu>,  writes...
>I'm looking for a (FORTRAN) program that could help me correct my
>data for absorption effects. The data was obtained for a single

I have the following DIFABS program, but have never used it:

larry henling
lmh@shakes.caltech.edu

In the subsequent e-mail messages are files for running DIFABS; including the source code of the latest version (V9.0) which uses an improved model for the absorption surface. For details, please read the comments at the beginning of the program.

I have sent a short description of the files that are required and how the Fo^2/Fc^2 file may be generated if using the SHELXL93 refinement program. The roundabout procedure previously required for SHELX-76 and SHELXTL-PLUS is no longer required.

Although the routines for the calculation of spherical polar angles are general for any four-circle diffractometer data collection geometry, in some instances extra code may be required in routine ANGLES for the calculation of Eulerian angles from the orientation matrix and reflection indices. Despite testing these routines as thoroughly as possible, I should state that my personal experience is only with Syntex P21, Siemens P4 and CAD4 in the bisecting position. Therefore, I suggest that you check that the Eulerian angles calculated by the program are consistent with those output by your diffractometer, and that the calculated polar angles are consistent with the direction senses given in Fig. 1 and equations (A1) in the 1983 Acta Cryst. paper.

I believe that the program code conforms fully to the Fortran-77 standard and that although developed on a VAX computer does not contain any of the extra facilities provided by VAX/VMS Fortran (except those calls highlighted at the beginning of the program). The following files are being sent:

                  1) Fortran source of DIFABS V09.0        DIFABS93.FOR
                  2) Test input parameter file             DIFABS.DAT
                  3) Test input Fo^2/Fc^2 data file        SROAC.FCF
                  4) Test output corrected Fo^2 data file  SROAC.ABS
                  5) VAX/VMS job-control command file      DIFABS.COM
                  6) Output listing file                   DIFABS.LIS
                  7) Integration with SHELXL93 text        SHELXL93.TXT

The test data set is that of SROAC with generated absorption errors, as mentioned in the Acta Cryst. papers. For information on when and how to use this method of absorption correction, I refer you to the 1983 Acta Cryst. paper.

Query by Karl Byriel | top 

From: chkennard@uqvax.cc.uq.oz.au (Colin Kennard, University of Queensland)
Newsgroups: sci.techniques.xtallography
Subject: Re: Absorption correction programs - where do you get it
Date: 23 Feb 94 11:46:32 AEST
Organization: The University of Queensland

Sorry to be dumb but the files never arrived here at UQ, are they available by FTP or could they be posted directly to me please

Thanks
Karl

**************************************************************************

Karl A. Byriel                               ,-_|\   
Department of Chemistry,                    /     \Brisbane
The University of Queensland                \_,-._/  
QLD.  4072  Australia.                           v  

email: byriel@chem.chemistry.uq.oz.au  IP : 130.102.84.15
phone: 61-7-365 4300                   fax: 61-7-365 4299

***************************************************************************

Final Followup by Lawrence Henling | top 

From: lmh@celia.caltech.edu (Henling, Lawrence M.)
Newsgroups: sci.techniques.xtallography
Subject: Absorption corrections
Date: 23 Feb 1994 21:41 PDT
Organization: California Institute of Technology

[I sent copies of DIFABS to all who requested it; if not received please send e-mail and I will try again.]

We do not use DIFABS here as some people don't like it.

Instead, we use either an analytical absorption calculation or an empirical psi-scan correction.

The analytical absorption calculation is the Gaussian quadrature method of Busing and Levy (Acta Crys, 10, 180, 1957). It involves measuring the phi and chi angles and centroid-face distance for all faces and then integrating over an 8x8x8 grid.

When the faces cannot be measured (eg, obscured by grease in a capillary), we measure psi-scans for 6 reflections at 10 degree intervals. The plotted relative intensities are used to generate a correction curve based on phi which is then applied to the data.

The second method should not be as good as the first (dependent on the crystal shape), yet we have found that the psi-scan method is often slightly better, based on the goodness of fit for merging symmetry-equivalent reflections. It will be interesting to compare our calculations with those of the program recently posted.

larry henling
lmh@shakes.caltech.edu

Where do you get the DIFABS and XABS2 software?

Followup by Tim Weakley based on IUCr Newsletter Article (Vol. 3, (1), 1995, pp. 6-8,10) | top 

From:  TWEAKLEY@OREGON.UOREGON.EDU(Timoth J R Weakley)
Newsgroups: sci.techniques.xtallography
Subject: DIFABS
Date: 8 May 1995 23:29:36 -0000
Organisation: U. of Oregon

I've only just read the feature article in the latest IUCr Bulletin and learned

I'd like to add my twopennyworth of comments for what they're worth, particularly as Acta Cryst. seems to be rather hostile to DIFABS. By the way, has Nigel Walker been asked for or offered comments?

I assume the crystal was centered, the cell was well determined,a full set of independent reflections was collected, the scan speed was not excessive, background intensities were properly measured, etc.; also, that the spacegroup is correct, the atoms have been properly identified by element, and so on. Then if the DIFABS calculation has been run after isotropic refinement of the complete structure (excluding H atoms) the following should in my experience be true.

  1. For fibre-mounted crystals, the range of `transmission factors' output by DIFABS correlates well with the max. amd min. transmission in a smoothed and averaged psi-scan curve based on several reflections.

  2. If mu for such a crystal is large, the application of DIFABS leads to a substantion reduction in residuals and in sigmas of coordinates and derived dimensions after further isotropic refinement, and gets rid of a lot of (chemically) meaningless residual electron density peaks near heavy atoms.

  3. For a fibre-mounted crystal of low mu, DIFABS does not lead to any significant reduction in residuals, GOF, and sigmas.

  4. For a capillary-mounted crystal of moderate mu, DIFABS often leads to substantial improvement as in 2, above? It seems to be compensating for absorption/scattering by glass and grease, etc. All right, so the "improvement" may be cosmetic; but in that case, why is it that the post-DIFABS model always seems more acceptable on chemical grounds, if the physics is so suspect?

Further comments.

  1. Psi-scans: sometimes there are no adequately strong reflections in the chi range 80-100 deg.; if the crystal happens to have two short axes roughly normal to the phi-axis, there may be no useable reflections at all.

  2. Absorption corrections based on the measured distances of identified bounding faces of the crystal from a common interior point:

    • (i) For highly-absorbing thin plates or elongated needles, the corrections are highly sensitive to errors in linear measurement, which may worsen the data instead of improving it (them?). Have you tried measuring a distance of 0.02 mm recently, even under a good microscope?

    • (ii) The available crystals may all be irregular fragments which cleave badly or shatter when you try to cleave them to roughly equidimensional shape. The crystallographer has to use what he gets, and what the client supplies may be a gunge of material clinging to the interior wall of a rotovap vessel, which decomposes or refuses to come out of solution when you attempt to recrystallize it.

Tim Weakley, Chemistry, U. of Oregon 
TWEAKLEY@OREGON.UOREGON.EDU

Louis Farrugia - Comparing Absorption Correction Methods using the WinGX Suite - 6th April 1999 | top

Using DIFABS Analysis in CRYSTALS package | top

Why bother using DIFABS Analysis?

"You can use DIFABS in test mode at any stage, and it quite simple tells you if there is a systematic trend to the residual as a function of orientation. If the plot is fairly monochrome, then the residuals are not peaking up. If there are bright patches, it means that there is a systematic trend to the residual. Either FO or Fc is systematically wrong. Possible reasons are:

Refer:

Gui WinCRYSTALS 2000 Web and FTP sites

Go Direct!
Crystallography Newsgroup (sci.techniques.xtallography) Homepage
To Crystallography World Wide
To Crystallography World Wide: News
Discussions on Crystallography
Cuban Crystallographic Software Library
Rietveld FAQs
Rietveld (and related) Software for PC and Macs
18th November 1999 - Volunteer Maintainer