Alfred Baron, 2010 (updated 2016)
The following are some relevant points for measuring phonons in single crystals using IXS. Note, especially, that sometimes a sample prepared for neutron scattering, and especially for inelastic neutron scattering (INS), can sometimes be of poor quality for IXS measurements (the constraints are just different for INS) - please be careful.
0. Laue/Transmission vs. Bragg/Reflection Geometry. One sometimes, especially when cutting a crystal from a large piece, can choose to use either a transmission geometry or a reflection geometry (Laue and Bragg, respectively). Generally, if one knows the specific momentum transfer of interest, a reflection geometry will give higher rates (e.g. x2). However, a transmission geometry will usually be more flexible - allowing access to more of reciprocal space. Of course, for a transmission geometry, the sample needs to be thin enough (see 2). Conversely, for a Bragg geometry, the sample must be thick enough so that we do not see scattering from any supporting material (e.g. the copper of the sample holder) behind the sample.
1. Larger (up to a few mm) is almost always easier. We can and do measure ~10 micron samples. But count rate is always an issue, and aligning a small sample is often harder than aligning a larger one, and also, there can be beam shifts, or sample shifts on cooling, etc. 1mm samples are much easier, with, probably, ~2x2 mm³ faces being close to ideal. If larger than ~5 mm (or thicker than 2 mm), consider cutting a smaller piece, or talk to a local contact, to make sure the sample will fit in/on our holders.
2. Know the attenuation length. You should know the absorption length of your sample. This sets the scale for the size and possible interference/shadowing from surface features/roughness. It also sets the length scale for a transmission (Laue geometry) measurement. For transmission measurements, the sample should be 1 to 1.5 absorption lengths thick along the x-ray beam - and not more than 2. You can use http://henke.lbl.gov/optical_constants/atten2.html to estimate the attenuation length.
3. Pre-aligned samples are required. Knowing the alignment of the sample (i.e.: two non-parallel reflections, or two axes of the unit cell, etc, not just one surface normal) at the level of a few (say ±5) degrees is crucial. While we will refine the alignment on the spectrometer, to go searching for Bragg reflections from scratch is a waste of beam time. We can, in some cases take a precession type photo, but always better to know the alignment ahead of time.
4. The sample must be a SINGLE DOMAIN. Be sure your sample is a single crystal without, say, some tiny domains with a different orientation (as can easily appear in flux-grown crystals). The beam size is ~50 microns, and often does not illuminate the full sample, so if present, micro-domains can appear unexpectedly and lead to strange results. You are urged to put your sample on a 4-circle diffractometer and scan over ALL of symmetry allowed reciprocal space to insure no micro-domains are present. Note: TWINNING, should be avoided as this can impact (severely!) careful measurements.
5. Mosaic. The mosaic requirements for IXS are relatively relaxed by x-ray standards, unless you need either very good Q resolution, or want to work very near a Bragg peak, etc. Typically, 0.1 degree mosaic is not a problem. We begin to worry above 0.5 degrees. Of course, smaller mosaic is usually better.
6. Azimuthal Orientation. The azimuthal orientation of the sample (ie: rotation about the scattering vector) has no impact on the momentum transfer measured by the analyzer crystal in the center of the arm. However, it has huge impact on the momentum transfers seen by the other analyzer crystals and so can be very important to make proper use of the analyzer array (esp. for transverse modes). See the discussion in section 4.6 of http://arxiv.org/abs/1504.01098