Area detectors - current trends and why you can't have the perfect bucket

There are three factors that prevent you from owning the ideal area detector - the one with infinite count rate, zero dead time and unity detection efficiency out to 25 keV, namely:

• The "Laws of Physics" - don't ask for zero response time and zero noise floor.
• Available materials - it would be nice to have high-quality semiconductor wafers with absorbance coefficients of a few tens of micrometres for Eh = 3-30 keV and a direct bandgap so that only about a dozen electron-hole pairs are created per incident photon; but you've got silicon, gallium arsenide, indium phosphide and a few other esoteric materials with suspect mechanical properties.
• Current technology - the pressures on industry to develop faster 16-bit analogue-to-digital converters, faster and larger mass storage and a myriad of other components are driven by much bigger markets than protein crystallography (PX) instrumentation. We have to use what is available.

PX does not need an ideal detector for routine data collection. It has taken much discussion between practising crystallographers and detector experts to determine what is needed and what can be implemented.

It is a particularly interesting time to consider PX instrumentation, as the demand for quality data grows. The time chart for area detectors is film, wire chambers, image plates and, now, CCDs. Each new wave of detectors has to fight for acceptance by the community due to the slow transformation of a single prototype system into reliable, fully characterized, commercial products and, resistance to change. The current status of area detectors is illustrated by two recent papers in the Journal of Synchrotron Radiation. The first ["A multiple-CCD X-ray detector and its basic characterization", M. Suzuki et al., J. Synchrotron Rad., 6 (1999), 6-18] describes a CCD mosaic system developed for the RIKEN beamline at SPring-8 (Hyogo, Japan). As scintillator-taper-CCD systems have become readily available, the drive has been to increase the aperture dimensions by stacking together single CCD systems to form close-butted mosaics. The RIKEN 4 x 4 CCDs provide a total aperture of 200 mm x 200 mm, giving resolutions out to approximately 2 Å at 12.4 keV. Smaller mosaics are becoming available commercially, but a 4 x 4 mosaic is large. As the paper rightly states, there are clear advantages in using more tapers with smaller demagnification ratios, as this will markedly improve sensitivity. The paper also highlights many of the practical issues necessary to achieve the very high levels of performance now possible with CCD systems. Suzuki et al.'s use of nonstandard CCDs, to reduce cooling requirements, will reduce the saturation charge capacity of the detector and the implications of this on the dynamic range for data collection remains to be seen.

Single CCD systems will continue to revolutionize small-crystal work and the capabilities of the home lab. Whether the trend of today's SR detector becoming tomorrow's home detector will persist depends on the fourth factor, which I forgot earlier: can you afford it? Multiple CCD systems and the pixel detectors are unlikely to fall to a price that the home lab could justify. The future for the home lab, in terms of reasonable cost and appropriate performance, may be with amorphous detectors.

Just as CCD systems are becoming the workhorse for SR PX stations, the long-awaited solid-state pixel detectors are being subjected to immense development efforts by many groups at several SR sites. Their promise is both of photon counting and of continuous read-out.

The second paper ["X-ray powder diffraction with hybrid semiconductor pixel detectors", S. Manolopoulos et al., J. Synchrotron Rad., 6 (1999), 112-115] reports the first application of a silicon pixel detector at an SR facility and the recording of preliminary powder diffraction data. The detecting device employed was small, not designed for SR use and had large dead times but it demonstrates he feasibility of such an approach, that interpixel dead spaces are probably not going to be as big a problem as some suggest, that good spatial resolution is attainable and that the detector survived!

My own views are, for SR facilities, that solid-state pixel detectors will become - after much expense, redesign and assessment - the preferred detector for many applications where crystal dimensions are relatively large (about 0.5 mm) as it will be difficult to reduce individual pixel sizes much below 150 µm. It is not their photon-counting capability, but their high-speed read-out, which will open up opportunities for new dynamic experiments and close the gap between potential sample fluxes and detector count-rate limits.

For the foreseeable future, pixel detectors will not possess unity quantum efficiency across the whole energy range. It is worth rehearsing the arguments for integrating vs. photon-counting detectors - for CCD vs. pixel detector. We can consider the individual CCD pixels as buckets that are filled, over the frame time, by photon-generated electrons. So integrating detectors are where you let the balls (electrons) fall into a bucket for a given time, and just before you empty the bucket so you can count the balls (to calculate the balls caught per second), some demon - who represents the read-out and system noise - throws a random number of balls into the bucket. Ideal photon-counting detectors allow you to count balls individually as they fall into the bucket. For practical solid-state pixel detectors, the problem is a fraction (and you are never certain exactly what the fraction is) of the balls miss the bucket completely - this is particularly so for high-energy balls, which pass straight through the bucket, - or, even more amazingly, a ball may split in two with each half landing in different buckets. These do not get counted at all and this is bad news for low-energy balls. So we will not get the ideal bucket (i.e., detector) but we will with pixel detectors get a better bucket - one that matches closer our specification envelope. Thus the second paper is brief but a flag-planting one as it demonstrates the way forward to a whole new range of buckets and a whole new range of science.