Owing to the nature of the Curved Image Plate camera and the interaction between beamline, camera hardware, software, and the scientific users, detailed performance levels cannot be accurately predicted on paper alone. Instead, this section highlights a number of aspects of the system performance as a whole, as determined fi*om actual use.
1. Intensity and statistical quality
Absolute intensities recorded depend, of course, on the samples, the beam strength, quality of the alignment of the sample within the beam etc. A representative 5 minute exposure of NIST 640b silicon powder in a 0.5mm glass capillary tube at 0.6920Â gave a (111) peak intensity of 16000 against a background of 400 (primarily fi*om the capillary). After integration, the figures remain much the same, but owing to averaging over -700 pixels per 26 step, the peak signal to background noise (statistical and random) ratio improves, and is of the order 5000:1. The full width half maximum peak widths are approximately 0.08° in 20. Saturation intensity of the reader is 100000 counts.
2. Experimentally determ ined accuracy of the 20 scale
Estimates of the best precision in the 20 scale can be gauged by looking at the scatter in the found positions for silicon standards relative to their ‘known’ positions. Estimates of accuracy in the system as a whole, and the degree to which precision is reduced over time, are obtained by calibrating the axis using one standard exposure, then repeating that standard several scans later and comparing the two.
With the 350mm radius camera, under typical conditions, for the silicon peaks in the range 0-60° at 0.6920Â the RMS deviation of individual peak positions compared to the best fit can be expected to be around 0.0025° for exposures within one or two scans o f the scanner calibration. The deviation typically remains better than 0.0035° for 15 scans either side of a calibration, worsening to -0.007° at 30 scans away firom a scanner calibration. It is recommended that recalibration of the scanner be performed at least every 20 to 25 scans to maintain precision.
The accuracy of the 20 scale is affected primarily by mechanical variations in the sample position relative to the geometric arc centre from exposure to exposure, but possibly also due to a small drift in the X-ray wavelength fi*om the monochromator over the life of the
synchrotron electron beam. Comparison between the overall 20 scale calibrated using a silicon standard following a synchrotron refill, and a recalibration just prior to the beam dump amount typically to an error of the order 0.02° in 50° (a not-quite linear scale error of about 0.04%). Between adjacent exposures, with the sample being removed, replaced, and re-exposed, scale errors are typically slightly less than this. Performance (when recalibration is carried out at the recommended intervals) is summarised in table II.E.1.
Arc Max 26 Equivalent step size* Peak typical typical
radius 88pm scan 176pm scan FWHM precision accuracy
350mm 64° 0.014° 0.028° 0.08° ±0.003° 0.03%
185mm 120° 0.027° 0.055° 0.16° ±0.007° 0.06% (est)
Table ILE.1: CIP camera performance. Zero point errors depend on the user, but with care are normally less than 70% of the equivalent step size. * angle subtended by one pixel in the centre of the image plate.
3. Round Robin feedback
As part of their Research Computing Initiative, the EPSRC funds a number of Collaborative Computing Projects (CCP) which are designed help academia to develop, maintain and distribute computer programs and methods of use to researchers. One of these projects, CCP 14, is for Single Crystal and Powder Diffraction (‘Freely Available Crystallographic Software for Students and Academia’). During 1998, this project organised a ‘Structure Determination by Powder Diffractometiy Round Robin’, the aim of which was to compare efficiency of methods for structure determination from powder diffractometiy (SDPD). Two samples were chosen, an inorganic, cobaltamine
[Co(NH3)gCOjN03.H20, and achromycin (or tetracycline) hydrochloride
C22H24N20g.HCl, a pharmaceutical. For each, cell parameters and possible space groups, together with several powder diffraction datasets were distributed to the participants. To supplement the Cu-Ka Bragg-Brantano diffractogram for the tetracycline, we were asked to provide a CIP capillaiy pattern (exposed at A, = 0.6920Â).
Seventy teams downloaded the complete datasets, and 4 submitted solutions for the tetracycline. Only two of these ‘solutions’ proved in any way complete, one was from the conventional data, and one was from the CIP dataset. Prof. W.I.F. David of the Rutherford Appleton Laboratory solved and refined the orthorhombic structure (P2j2j2j) from the CIP data using the Global Optimisation Method and extracted 100 structure factors by the Pawley method. The final analysis yielded lattice parameters of
a = 10.98029(7)A, b = 12.85242(7)A, c = 15.73303(8)A, with 7 ^ = 2.92%.
57 atomic positions were refined, including 24 hydrogens (located to a mean error of
0.2Â).
Comparing with the single-crystal reference solution, the results fix)m the CIP data were twice as accurate as those submitted by a different contestant working fi*om the
conventional data (see table n.E.2).
GIF Prof. David Cu - Bragg Brentano Participant 4 Cu - Bragg Brentano Organisers Mean/A Min /A Max /A Mean/A Min /A Max /A Mean/A Min /A Max /A Atom Number C 22 0.080 0.038 0.136 0.191 0.034 0.374 0.173 0.083 0.278 H 24 0.217 0.054 0.549
not found not found
0 8 0.083 0.062 0.104 0.105 0.045 0.156 0.133 0.093 0.177 N 2 0.080 0.071 0.089 0.253 0.246 0.259 0.268 0.261 0.274 Cl 1 0.044 0.027 0.014
Table n.E.2: Structure solution of tetracycline hydrochloride from powder diffractometiy; displacement of atom positions compared to single-crystal data. The data from the Image Plate is highlighted.
The organising team also produced a solution from the conventional data (and have since solved the CIP data, though the results have not been disseminated). Had the CIP been used with a longer wavelength, these results might well have been improved further still (at the wavelength used most of the useful information in the pattern was compressed into
the region below 25° 20).
Although the small number of respondents makes it difficult to differentiate between advantages due to analysis techniques, or raw data, this refinement proves that the CEP data is more than capable of providing structure solutions for, and high quality refinements of, complex organic molecules.