A more detailed hardware description

Old System

The system was originally supplied with MSC/Yale mirrors (see "toroidal mirrors" section of Arndt & Bloomer (1999) and in Garman & Sweet (2007)), but a 2012 upgrade changed the optics from those mirrors to the Xenocs FOX 2D 25-25P multilayer optic which got us about 3-fold in intensity gain and rather less CuKβ in our CuKα X-rays. Nevertheless a larger focal spot inherently begets a larger (or less brilliant) X-ray beam at the sample.

The Rigaku RAXIS-IV++ detector (see link for Rigaku Journal article on RAXIS-IV, see also Dauter and Wilson (1994)) was a dual image plate system where the X-ray signal was accumulated in a Eu³⁺ doped barium halide detector and read out by laser - a lot of moving parts in this detector but well-engineered and quite reliable until right at the point where we retired it. Two image plates attached to a flexible metal belt allowed data collection to continue on one plate (held flat) while the data was being read out by a laser scanning over the other plate (held on a cylindrical surface). Moving laser and laser optics made for a relatively slow readout time (~90 seconds) which was nevertheless a substantial improvement over previous versions of IP detectors (8 minutes and 3 minutes) - far slower than a CCD or CMOS detector. Features included a large active area of 300 x 300 mm, small pixel size (we used 100µ) and high dynamic range. Nevertheless the detector wasn't very sensitive with a detective quantum efficiency (DQE) of around 20%. Large area and relatively moderate intrinsic noise led it to dominate the macromolecular crystallography field for many years and its production was only recently ceased. Many still use one.

New System

Optics: We have a Varimax HF optic. Osmic was one of the earlier companies involved in making multilayer optics: see the "caveat emptor" paper by Yang, Courville and Ferrara back from the days of Molecular Structure Corporation before Rigaku acquired both MSC and Osmic - and obviously Rigaku's VariMax line of multilayer optics are descendents of the original Osmics. Multilayers are synthetic 2D crystals on an elliptically-curved substrate where the spacing between the layers in these "crystals" varies along the length of the optic (See Optics section of Garman and Sweet paper, and technical discussion via the Rigaku Journal paper by Shimizu & Omote). Elliptical multilayers with graded spacing are current state of the art for home source optics with the ability to capture more of the X-rays out of the source while delivering a parallel or focused X-ray beam to the source, depending on hardware design. We opted for Rigaku Varimax HF optics (vendor site) with the Arc)Sec specification (better polishing) that delivers a beam that is 189µm at FWHM at the sample. You can buy an optic with a smaller beam at the sample (VHF model) at the expense of a more divergent beam, or a larger beam with less divergence (HR model) - yet another example of no free lunch. The optic is "double bounce" where the X-ray beam bounces off two orthogonal optical elements. Their action as a 2D diffraction grating also helps to substantially reduce the CuKβ component to the desired CuKα X-rays. Other companies also offer graded multilayer optics of comparable design, including the supplier of our older multilayer (Xenocs).

Goniostat: We have a Rigaku AFC-11 partial 4-axis goniostat to drive the crystal and detector around. Two of the four axes are used for crystal positioning (φ and χ), one is used for the data collection scans (ω) and one is used to position the detector (2θ). There's also a motorized detector distance track. This sort of goniostat is desired because of the potential for high resolution, and also to handle the smaller active area of the X-ray detector (see below) - more crystal permutations are necessary to achieve desired data completeness. The relatively light detector allows movement from +5° to -90° in 2θ and the limit on high resolution data is 0.83 Å due to the CuKα wavelength of 1.54 Å. Your protein crystals will not diffract that far, but small molecule samples routinely do, especially when mounted on a powerful rotating anode generator. The φ axis rotates through 360°, ω through a 180° range either side of the 2θ axis, χ from 0 to 60° and 2θ from +5 to -90°. Detector to sample distance is in the range 30 to 290 mm. The AFC-11 is relatively slow and not all that exciting from a hardware perspective and superceded by a faster goniostat after Rigaku's acquisition of Oxford Diffraction.

Detector: Dectris Pilatus3 R 300K (vendor site, more technical data). The 300K has three of the modules, one more than the base 200K detector offered by Rigaku. It's a photon-counting Hybrid Pixel Array Detector (HPAD) with direct detection of X-ray photons in the Si detection layer followed by charge transfer via Indium bump-bonds to an underlying CMOS chip. The 300K uses room-temperature water cooling, and a dry air/nitrogen supply to suppress humidity. The active area is smaller than the RAXIS-IV++ (84 mm x 106 mm versus 300 mm x 300 mm) and the pixels are larger at 172 µm but there are a host of upsides to this detector: minimal electronic noise; fast readout time (7 millisec); high dynamic range; single pixel point spread function; very high DQE (~98% at CuKα); light weight. With the exception of the relatively small size and the obvious boundaries between the modules the detector is exceptional and probably accounts for the majority of the improvement over the old system - perhaps as much as 10x over the 17 year-old RAXIS IV++ we had. Being relatively small and light it also enables us to collect very high resolution data on e.g. small molecule samples to around 0.83 Å resolution. CCD and CMOS detectors of older designs had limited dynamic range and noise issues - the Pilatus is better than any other detector I've encountered on a home source.