From a user:>
I have a trace of the crystal scanned at the bromine absorbance edge. The black lines are before collection and the green lines are after collection.

Can you tell me if the lower black and lower green lines are excited at a remote wavelength?
Also, I am getting a refined occupancy of about 0.7 for the bromine atom. If I set occupancy to 1.0, I always get negative fo-fc electron density. Can you tell me if the reduction from black line to green line in this plot could explain a loss of Br during data collection? The bromine is attached to a phenyl imidazole ring. Thanks!

James Holton Replies:
	
Yes, Bromine and other halides popping off of modified bases and other conjugated systems is one of the "fastest" known rad dam reactions (Olieric et al. 2006: http://dx.doi.org/10.1107/S0907444907019580).  The change in your XANES spectrum looks pretty much exactly like what they saw in the Olieric paper.  There is one caveat: you can also get changes in the XANES spectrum just by rotating the crystal, but as long as your scans are done at the same "phi" value, you are comparing apples to apples.

  This is probably because both Br and the conjugated system of your phenyl imidazole ring are both fairly good leaving groups.  The measured half-dose for this reaction was 1 MGy, or about 1 minute of shutter-open time at 8.3.1.  Like any rad dam reaction however, it is the photons that do the damage, not the exposure time.  It also doesn't really matter what wavelength you use.  Unless your Br concentration in the crystal is more than ~200 mM the effect of collecting at the peak vs remote will be less than a factor of two.  Better to just cut your exposure time in half and get complete data faster.  The impact of exposure time on your resolution limit is surprisingly weak.  Better to collect a lot of weak frames than a few really strong ones.  Have your tried zero-dose extrapolation in XDS?

So, you can cite Olieric et al (2006) as a sound reason for occupancy refinement on your Br atom.  Also, be very careful to use the proper f' value in your refinement.  Near the peak there can be as much as 10 fewer electrons in the Br than you will predict using the standard wavelength-independent form factors.  How you do this depends on which refinement program you are using.

I'm not really sure what you mean by "excited at a remote wavelength".  The X-axis on your graph is the wavelength, or rather 12398.42/lambda.  If you are asking which spectrum to use to calculate f' and f' for refinement, then I'm afraid the answer is both.  Your data set will have different f' and f" for different phi values.  So, the only things you can do are process both spectra and take the average, or try to get away with "early" data only.  That is, cut out frames at the end of your data set and see if that improves your Br occupancy.