Tips and Tricks in Crystallography - Diffraction

Date Collection with ID22 and Eiger16M

Fred Dyda (NIDDK): Last week we collected ID data at 50 ms / 0.2 degrees at 30% transmission. With these, 180 degree of data took only 45 s.
      This suggests that one could go faster with wider frames and/or higher transmission.
      John says that the limit in exposure time is 20 ms. This is where the stream server stars to choke in spitting out the .cbf files from the stream.

Fred Dyda (NIDDK): During a survey of data collection practices, it became apparent that most users do not take advantage of the current capabilities of ID22 and the Eiger16M. In particular it has been shown that excellent data can collected using 0.2 or 0.25 degrees / 0.05 sec speed so 200 degrees can be collected in 36 or 50 sec. One would of course have to increase transmission, but this should be possible as most people seem to collect at 80 or 90% attenuation. This should be OK, as all indicate that decay is dependent of totla accumulated does and is independent of dose rate.

Data Completeness from HKL2000 or HKL3000

Zbigniew Dauter (NCI)Sometimes, relatively weak diffraction data processed with HKL2000 or HKL3000 show lower than expected completeness in the high resolution bins, in spite of otherwise high multiplicity and coverage of the reciprocal space. Apparently, this results from the automatic rejection of weak reflections, for which there are no good enough neighbors for building a “standard profile” within the vicinity defined by default value of the “Profile Fitting Radius”. To increase the high-resolution data completeness one can change the default profile fitting radius (in the “Index” window of HKL2000 or HKL3000) to much larger value, e.g. one third of the detector size. The number of “good” reflections, e.g. stronger than 5s(I), within the profile fitting radius can be visualized at the Xdisp window by clicking the “Prof Fit R” option.

Rayonix MX300HS Detector (SER-CAT 22-ID)

Zhongmin Jin (SER-CAT): To support fast speed Rayonix MX300HS detector installed in SER-CAT, new HKL2000, version 705a_Linux, has been installed in several SER-CAT beamline computers: idc24 and idc23 (for ID beamline), bmc83 and bmc93 (for BM beamline). The file can be copied to users data directory from directories of those computers: /usr/local/hklint/MX300HS/ SER-CAT users need to contact HKL Inc. to acquire this new HKL2000 705a version and appropriate licenses to be able to process data collected from Rayonix MX300HS detector in their home laboratories, the older HKL2000 versions do not support this new detector format. Since many users still prefer the older version HKL2000 to process data collected from detectors MAR300 and MAR225, we still keep older version HKL2000_98.699a in all other SER-CAT data processing computers.

The Rigaku Pflipper: For mounting & Dismounting frozen crystals.

Small Molecule Crystal Data Collection

Bernard D SantarsieroYou can collect 5-10deg images, and go ahead and collect 360deg. I usually use 1-3 seconds exposures, depending on the diffraction.  The detector is set as close as possible, to get data to 0.8-0.9A resolution. For a MAR CCD 300mm detector, that's around 90-100mm, with the wavelength set at 0.8A, or around 15.5KeV in energy. You can process it with HKL2000 or XDS easily.

Zbigniew Dauter: I completely agree with Bernard, but I have one additional comment. The range of intensities from small molecule crystals is usually much higher than from protein crystals. It is therefore practically always necessary to collect them in multiple (sometimes even three) passes with different effective exposure time (or beam attenuation) and resolution range. Unfortunately people often forget about it and as a result very many  strongest reflections are overloaded and lost.

Artem Evdokimov: Mounting needle-like crystals for data collection

Recommended Readings on Data Collection:

David Garboczi (NIAID): Our Experiences with the Data Reduction Program XDS
        Let me tell you about our experiences with the data reduction program XDS as an alternative to the more widely used software packages. It is run by a script that is detector-specific and there are examples on Wolfgang Kabsch's web site. We find it to be flexible and easy to use, which allows more experimenting with the raw X-ray data to achieve the best datasets. (Full Article)

Need an image processing software that builds three-dimensional profiles? Try XDS. For an easy start, see INTEGRATE by Kay Diederichs.

Recommended viewing: How long will my crystal last?

Mr. Jerry Alexandratos (NCI): Post-crystallization soaking. Do you have poorly diffracting crystals?  Don't throw them out. Try post-crystallization soaking by adding 1.5-2 M (NH4)2SO4 with or without several percent of PEGs such as 3350. You can also try slightly dessicating drops with crystals.  (I did not note down who mentioned these procedures. My apologies, as well as my thanks for the tips.) 

Dr. Lothar Esser (NCI): Denzo, XDisplayF, Scalepack. In case you are not yet using HKL2000 which is a GUI (graphical user interface) that controls denzo, xdisp and scalepack you should consider getting a copy of it. Officially it is not in general distribution (current version 0.97.638) but since the development of this program is funded by NIH, there is no problem for NIH crystallographers to get a pre-release free of charge. Routine and non-routine data reduction can be done very efficiently if you (or your student) are new to data processing but even more so if you have worked your way through the input files for denzo and scalepack before. Indexing and integrating of data is effortless and so is data reduction.

Is everything done by a push on a button? Well, you still have to manually adjust the error-of-scale factor until the overall χ2 is about 1.0. So for straightforward data sets, i.e. data sets that do not require much adjusting of the error bins, I have been using a utility program that does nothing more than running scalepack until the χ2 is 1 within a user-defined limit. To get the final graphical analyses you may run scalepack from within HKL2000 one more time with the updated reject file and the refined error-of-scale factor. In many cases this saved me a lot of time. (Let me know if you want to have a copy of this program)

Have you ever run xdisp on a remote computer and redirected it to an X display on a SGI? In that case you might have noticed that the fonts of xdisp are too large. In fact some buttons on the zoom window are missing. This problem can be solved by giving the x font server a different path to font files. Try this before you open xdisp on the SGI (the local machine):

set FONT_P=”/usr/lib/X11/fonts/misc/,/usr/lib/X11/fonts/75dpi/”

set FONT_P=”${FONT_P},/usr/lib/X11/fonts/100dpi/,/usr/lib/X11/fonts/Type1/”

set FONT_P=”${FONT_P},/usr/lib/X11/fonts/Speedo/,/usr/lib/X11/fonts/CID”

xset fp $FONT_P

Then open xdisp and it should have all the buttons correctly labeled using a small font. It also affects other programs that use the X font server like netscape. If you are done you can use “xset fp default” to restore the original font path.

Ed Pozharski's python script at the ccp4wiki reads a .sca file and produces a table with both <I/sigmaI> and <I>/<sigmaI> for a user-defined number of shells.

This site is maintained by Dr. Xinhua Ji ( on the NCI-CCR-MCL server (