Matthias Zebisch wrote:
Dear bb!
What is the optimal wavelength for Sulfur SAD phasing?
Is it 1.9A or should one go below that to reduce absorption/damage.
Also, would the same wavelength be appropriate to maximize anomalous
scattering to position chlorides, calcium, sulfate in already phased
structures?
Thanks in advance,
Matthias
As usual, the "optimum" is a compromise, and depends on the equipment.
I recommend 7 keV (1.77 A) for S-SAD to users at my beamline (8.3.1 at
the Advanced Light Source). This seems to be the best compromise
between the strength of the signal from Sulfur and the absorption in the
air and sample. A good website for calculating the magnitude of
absorption corrections (transmission) is here:
http://henke.lbl.gov/optical_constants/
Cameras that contain less air (He paths like the setups Jim Pflugrath
mentioned) and perhaps thinner crystals will shift the optimum to longer
wavelengths. Beamlines are even being designed to take advantage of
backscattered spots, since at the S edge (5 A), your 2.5 A spots will be
fired back up the beam pipe (lambda = 2*d*sin(theta)). Using the
backscattered geometry is not as crazy as it sounds: the Lorentz factors
are higher, and absorption corrections are both minimized and easy to
calculate. In fact the "reflection geometry" is the way the Braggs did
it (Bragg, James, & Bosanquet, Phil. Mag. 1921a, 1921b, 1922).
http://www.informaworld.com/10.1080/14786442108636225
On the other hand, the current standing record (to my knowledge) for
measuring a weak anomalous signal was done at 0.98 A (Wang et. al. Acta
D 2006): http://dx.doi.org/10.1107/S0907444906038534
In this case, the average anomalous difference was smaller than 0.5%. Z
Dauter was kind enough to make these data available, and I have a copy
of them here:
http://bl831.als.lbl.gov/example_data_sets/index.html#APS22ID
The authors do not claim this as an "optimum" S-SAD wavelength, but I
suspect that using more penetrating radiation reduced the contribution
of errors from the absorption model enough to allow this small signal to
be measured. That is, it is hard to estimate the absorption corrections
to better than a few percent, but if the absorption itself is small,
then errors in absorption corrections are much less significant. For
example, +/-5% for a 2% absorption correction is much less significant
than +/- 1% for a 50% absorption correction.
The trick, as always, is getting the error to be smaller than the signal.
Basically, do this: You can approximate the noise level in your data
with Rmerge or the noise-to-signal ratio (the inverse of Mn(I)/sd).
This will be about 3-5% in favorable cases. Next, estimate the
magnitude of the anomalous signal you are trying to measure with the
Hendrickson-Teeter (Nature 1981) equation, which I paraphrased in a
previous posts:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0903&L=CCP4BB&T=0&F=&S=&P=191973
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0903&L=CCP4BB&T=0&F=&S=&P=195218
For typical S-SAD cases, the anomalous signal or "Bijvoet ratio" will
be ~0.5-1%. Next, divide the noise by the signal and square it. This
is the multiplicity (m) you will need:
m = (noise/Bijvoet_ratio)^2
For example, if your I/sd is 30 and you are trying to measure a 1%
anomalous difference, you will need a multiplicity of not less than (
(1./30) / 0.01 )^2 = 11. Note, however, that if this I/sd = 30 was from
data with a multiplicity of 5, then your "unit" I/sd is actually
30/sqrt(5) = 13.4, and you will need a multiplicity of 55 to measure a
1% anomalous difference. If your anomalous signal is 0.5% in this same
situation, then you will need a multiplicity of 223. Yes, these are big
multiplicities, and that is why S-SAD is usually demonstrated with large
and well-diffracting crystals (less dose per scattered photon). I
stress that the above is a rough estimate of the LOWER limit where the
world's best crystallographers have managed to get something out of the
data. If you are not one of these people, then you will need better
data. If you are one of these people, you will insist on it.
-James Holton
MAD Scientist
