Chapter 7.1. Detectors for X-rays

Visual estimation consists of comparing the spot or line to be measured with a series of exposure-calibrated marks similar in shape to the object of measurement, and preferably made with the same specimen and incident beam. Lack of complete similarity and unfavourable background usually cause the error of such measurements to be larger than the optimum contrast threshold of the eye. For a spot area of 1 mm², the latter amounts to roughly 1% or 0.004 density units, a difference that can in fact be detected under favourable circumstances (low density and low background).

If the blackening is measured with a microdensitometer, an accuracy of 0.002 density units up to densities of at least 2 is easily attained. Higher precision is rarely required, as the limiting factors are graininess of the film and irregularities in the emulsion and processing. The grains in processed X-ray film are larger than those produced by visible light, and occur in clusters around each absorbed quantum. The resulting statistical fluctuations may be minimized by appropriate choice of densitometer slit dimensions and scanning speed. If the X-rays are not incident normally on double-coated film, it may be necessary to make corrections for obliquity (Whittaker, 1953; Hellner, 1954).

The commonest types of detector for both powder and single-crystal diffractometry are proportional counters and especially scintillation counters (Section 7.1.4). The detector system consists of the detector itself, a high-voltage power supply, a single-channel pulse-height analyser, and a scaling circuit, as shown schematically in Fig. 2.3.3.5 . For position-sensitive detectors (Section 7.1.3.3) and solid-state detectors (Sections 7.1.4.2 and 7.1.5), multichannel analysers are necessary. The X-ray manufacturers and a number of electronic companies provide complete detector systems, often integrated with the computer data-collection system.

Proportional counters are available in various sizes and gas fillings. A typical detector is a metal cylinder about 2 cm in diameter and 8–10 cm long, with central wire anode and 0.13 mm Be side window, Fig. 7.1.2.1(b). Some have an opposite exit window to transmit the unabsorbed beam and thus avoid fluorescence from the wall. The tube may be filled with Xe to atmospheric pressure for high absorption, and a small amount of quenching gas such as CO₂ or CH₄ is added to limit the discharge. When an X-ray quantum is absorbed, the discharge current is the sum of the Townsend avalanches of the secondary electrons and the gas amplification is about 10⁴. A charge-sensitive preamplifier is generally used. Some proportional counters are filled to several atmospheres pressure to increase the gas absorption. Very thin organic film windows are used for very long wavelengths as in fluorescence spectroscopy. They may transmit moisture, and gas may migrate through them so that flow counters are used to replenish the gas. This requires careful control of the pressure to avoid changes in the counting efficiency.

One variety of position-sensitive detector, in which the photon absorptions in different regions are counted separately, is a special type of proportional counter. The following description applies primarily to one-dimensional detectors for powder diffractometry; two-dimensional (area) detectors are treated in Section 7.1.6.

Position-sensitive detectors (PSD's) are being used in increasing number for various powder-diffraction studies. They have the great advantage of simultaneously recording a much larger region of the pattern than conventional counters. The difference in receiving apertures determines the gain in time. The position at which each quantum is detected is determined electronically by the system computer and stored in a multichannel analyser. There is a digital addition of each incident photon address and the angular address of the diffractometer.

The PSD's are available in short straight form and as longer detectors with curvature to match the diffractometer focusing circle. The short detectors can be used in a stationary position to cover a small angular range or scanned. Göbel (1982) developed a high-speed method using a short (8° window) scanning PSD with 50 µm linear resolution in the diffractometer geometry shown in Fig. 2.3.1.12 (b). He was able to record at speeds of a hundred or more degrees a minute, and patterns with reasonably good statistical precision in several tens of degrees a minute. This is faster than conventional energy-dispersive diffraction and has the advantage of much higher resolution.

The PSD should be selected to match best the diffraction geometry. The detector is sensitive across the 1–2 cm gas-absorption path. If the diffracted rays are not perpendicular to the window, the parallax causes broadening and loss of resolution. This becomes important in the focusing geometries and can be minimized if the diffractometer and specimen focusing circles are nearly coincident. A large loss of resolution would occur in the conventional geometry, Fig. 2.3.1.3 , because only the central ray of a single reflection would be normal to the window. The problem is minimized in powder-camera geometry with a thin rod specimen, Fig. 2.3.4.1 (a), where the entire pattern can be recorded with a long, curved PSD (Ballon, Comparat & Pouxe, 1983); see also Shishiguchi, Minato & Hashizume (1986), Lehmann, Christensen, Fjellvåg, Feidenhans'l & Nielsen (1987), Wölfel (1983), and Foster & Wölfel (1988).

The topics of energy resolution, pulse-height discrimination, quantum-counting efficiency, and linearity are common to proportional, scintillation and solid-state counters, and are treated in Subsections 7.1.4.3.–7.1.4.5.

The most frequently used detector is the scintillation counter (Parrish & Kohler, 1956). It has two elements: a fluorescent crystal and a photomultiplier tube, Fig. 7.1.2.1(c). For X-ray diffraction, a cleaved single-crystal plate of optically clear NaI activated with about 1% Tl in solid solution is used. The crystal is hygroscopic and is hermetically sealed in a holder with thin Be entrance window and glass back to transmit the visible-light scintillations. The size and shape of the crystal can be selected, but is usually a 2 cm diameter disc or a rectangle 20 × 4 × 1 mm thick. A small thin crystal has been used to reduce the background from radioactive samples (Kohler & Parrish, 1955). A viscous mounting fluid with about the same refractive index as the glass is used to reduce light reflection and to attach it to the end of the photomultiplier tube. The crystal and photomultiplier are mounted in a light-tight cylinder surrounded by an antimagnetic foil. The high X-ray absorption of the crystal provides a high quantum-counting efficiency.

A Cu Kα quantum produces about 500 visible photons of average wavelength 4100 Å in the scintillation crystal (which matches the maximum spectral sensitivity of the photomultiplier), but only about 25 will be effective in the photomultiplier operation. High-speed versions with special pulse-height analysers have recently become available; they are linear to about 1% at 10⁵ counts s⁻¹ and can be used at rates approaching 10⁶ counts s⁻¹ (see Rigaku Corporation, 1990).

The detector system is as described in Subsection 7.1.3.1.

The following description applies primarily to the use of solid-state detectors in powder diffractometry. Further details of their operation and their use in energy-dispersive diffractometry are treated in Section 7.1.5.

The most common form of solid-state detector consists of a lithium-drifted silicon crystal Si(Li) and liquid-nitrogen Dewar. A perfect single crystal is used with very thin gold film on the front surface for electrical contact. The first amplifier stage is a field-effect transistor (FET). The unit must be kept at liquid-nitrogen temperature at all times (even when not in use) to prevent Li diffusion and to reduce the dark current when in use. The unit is large and heavy and, if not used in a stationary position, a robust detector arm is required, which is usually counter-balanced. The crystal is made with different-size sensitive areas and the resolution is somewhat dependent on the size of the area. In the detector process, the number of free charge carriers (the electron and electron–hole pairs) generated during the X-ray absorption changes the conductivity of the crystal and is proportional to the energy of the X-ray quantum. Details of the mechanism are given in several books [see, for example, Heinrich, Newbury, Myklebust & Fiori (1981) and Russ (1984)].

Intrinsic germanium detectors have higher absorption than silicon detectors, but they have lower energy resolution and there are more interferences from escape peaks. A mercuric iodide (HgI₂) detector can be operated at room temperature and has high absorption (Nissenbaum, Levi, Burger, Schieber & Burshtein, 1984). They have poorer resolution than Si or Ge detectors but can be improved to FWHM = 200 eV at 5.9 keV by cooling to 269 K (Ames, Drummond, Iwanczyk & Dabrowski, 1983).

A small (about 16.5 × 10 cm), lightweight (3.2 kg) silicon detector with Peltier thermoelectric cooling is available (e.g. Kevex Corporation, 1990). This development has supplanted a number of the methods of collecting powder data. The elimination of the liquid-nitrogen Dewar and the compact size makes it possible to replace conventional detectors and the diffracted-beam monochromator in scanning powder diffractometry. The spectrum is displayed on a small screen and the window of the analyser can be set closely on the energy distribution obtained from a powder reflection to transmit, say, only Cu Kα. The monochromator can be eliminated for a large gain of intensity without loss of pattern resolution. The energy resolution is FWHM 195 eV at 5.9 keV. Elemental analysis can be performed by energy-dispersive fluorescence, and the background can be restricted to the narrow energy window selected. Bish & Chipera (1989) used it to obtain a 3–4 times increase of intensity, the same pattern resolution, and lower tails than with a graphite monochromator and scintillation counter in conventional diffractometry. The major limitation at present is the limited input intensity that can be handled. The limiting (total) count rate is about 10⁴ counts s⁻¹ and the detector becomes markedly nonlinear at 2 × 10⁴ counts s⁻¹. Internal dead-time corrections can extend the range by increasing the counting times.

The pulse amplitudes are proportional to the energy e of the absorbed X-ray quantum so that electronic methods can be used to reduce the background from other wavelengths and sources. The rejection range is limited by the energy resolution of the detector. As noted above, the pulse amplitudes have distributions that vary around the average value A, Figs. 7.1.4.1(a),(b) . The FWHM of the distribution increases linearly with increasing e (eV) and is proportional to , i.e. it improves inversely with . The ratio FWHM/A (expressed in %) is a measure of the energy resolution at a given wavelength; the smaller the ratio the better the resolution. For example, as e increases from 5 to 45 keV, the FWHM approximately doubles while FWHM/e decreases from 5 to 1%. The resolution of proportional counters is about 18% for Cu Kα and somewhat better for high-pressure gas fillings; in scintillation counters, it is about 45%. The solid-state detectors have much better resolution. The best are about 2.4% (145 eV) at 5.9 keV (which is the energy of Mn K X-rays from a radioactive ⁵⁵Fe source used as a standard for calibration).

Figure 7.1.4.1| top | pdf | Calculated pulse-amplitude distributions of Cu Kα and Mo Kα in the form of (a) integral curves and (b) differential curves. Resolution W/A for Cu Kα = 50%. Analyser settings show window between lower level (LL) and upper level (UL). (c) Plateaux of scintillation counter for various wavelengths and fixed amplifier gain. Curves normalized to same intensity at highest voltage. Noise curve is plotted in counts s−1. Curves can be moved to higher or lower voltages by changing amplifier gain. (d) Calculated quantum-counting efficiency (QCE) of scintillation counter as a function of wavelength (top curve) and its reduction when the pulse-height analyser is set for 90% Cu Kα. E.P. is escape peak at lower left.

The electronics include a high-voltage power supply to about 1200 V for scintillation counters and 2000–3000 V for proportional counters, and a single-channel pulse-amplitude discriminator. The latter contains pulse-shaping circuits and the amplifier, and is designed to transmit pulses whose amplitudes lie within the selected range. The lower level rejects all pulses below the selected level and the upper level rejects the higher amplitudes (Figs. 7.1.4.1a,b). The range selected is called the window and determines the pulse amplitudes that will be counted by the scaling circuit.

The multichannel analyser is generally used with solid-state detectors. It may have up to 8000 channels and sorts the pulses from the amplifier into individual channels according to their amplitudes, which are proportional to the X-ray photon energies. The pattern can be stored and displayed on a CRT screen, but nowadays a personal computer with a suitable interface card is normally used in place of the analyser. Various programs are available for peak-energy identification, spectral stripping, intensity determination, and similar data-reduction requirements. The limiting count rate that can be handled by the electronics is determined by the total number of photons striking the detector. Pulse-pileup rejectors are used to stop counting momentarily when another pulse is too close in time to allow the original pulse to return to the baseline voltage. A live-time correction extends the counting period beyond the clock time to compensate for the time the analyser is gated off. About 50 000 counts s⁻¹ is the maximum rate so that the individual powder reflections have a much smaller number of pulses. If good statistical accuracy is required, the count times are, therefore, much longer than in conventional diffractometry.

For a given e, the pulse amplitudes of scintillation and proportional counters increase with increasing voltage (internal gain) and amplifier setting (external gain). The detector must be operated in the plateau region for the wavelength used (Fig. 7.1.4.1c). The counts are measured as a function of the voltage and/or gain, and the plateau begins where there is no further significant increase of intensity. In selecting the operating conditions, one should avoid excessively high voltages and amplifier gains, which may cause noise pulses and unstable operation. Optimum settings can be determined by experiment and from manufacturer's instructions. The average pulse height should be set at about 20–25% of the full range of the pulse-height analyser. Lower settings move the low-energy tail into the noise, and high settings broaden the distribution and may be too wide for the window.

The pulse-amplitude distribution can be measured with a narrow (1–3 V) upper level and increasing the lower level by small equal steps. When making this calibration, it is advisable to keep the incident count rate below 10⁴ counts s⁻¹ to avoid nonlinearity and pulse pileup. A plot of intensity versus lower-level setting shows the distribution, Fig. 7.1.4.1(a). In some electronics, this can be done automatically and displayed on a screen. The window should be set symmetrically around the peak with the window decreasing the characteristic line intensity only a few per cent below that obtained with the lower-level set to remove only the circuit noise. The intensity change can be seen with a rate meter. Narrow windows cause a larger percentage loss of intensity than the decrease in background and, hence, the peak-to-background ratio is reduced. Asymmetric windows are sometimes used to decrease the fluorescence background.

The quantum-counting efficiency E of the detector, its variation with wavelength, and electronic discrimination determine the response to the X-ray spectrum. E is determined by where f_T is the fraction of the incident radiation transmitted by the window (usually 0.013 mm Be) and f_A is the fraction absorbed in the detector (scintillation crystal or proportional-counter gas). E varies with wavelength as shown in Fig. 7.1.4.1(d). The scintillation counter has a nearly uniform E approaching 100% across the spectrum and detects the short-wavelength continuous radiation with about the same efficiency as the spectral lines. The gas-filled counters have a lower E for the short wavelengths and, therefore, may have a slightly lower inherent background; high-pressure gas counters have a higher and more uniform spectral efficiency.

The effectiveness of electronic discrimination with a scintillation counter is shown in Fig. 2.3.5.3 (c) for 50 kV Cu target radiation. The method cannot separate the Kα-doublet components because of their small energy difference, and has little effect on the Kβ peak. The results are greatly enhanced by the addition of a Kβ filter, which removes most of the Kβ peak and a portion of the continuous radiation below the filter absorption edge, Fig. 2.3.5.3 (b). The combination of discrimination and filter produces mainly the Kα doublet, Fig. 2.3.5.3 (d). Spectral analysis of the background of a non-fluorescence powder sample using this method with 50 kV Cu radiation and a scintillation counter shows it to be 50–90% characteristic radiation.

The linearity of the system is determined by the dead-time of the detector, and the resolving times of the pulse-height analyser and scaling circuit. The observed intensity n_obs is related to the effective dead-time of the system τ_eff by the relation The value of τ_eff can be measured with an oscilloscope, or with the multiple-foil method in which a number of equal absorption foils (e.g. Al 0.025 mm or Ni 0.018 mm for Cu Kα) are inserted in the beam one or two at a time. To make certain monochromatic radiation is used, a single-crystal plate such as Si(111), which has no significant second order, and low X-ray tube voltage are employed. The linearity is determined from a regression calculation. A less accurate method is to plot n_obs on a log scale against the number of foils on a linear scale. Recent developments in high-speed scintillation counters have extended the linearity to the 10⁵–10⁶ counts s⁻¹ range.

The pulse-amplitude distribution may have two or more peaks, even when monochromatic X-rays are used (Parrish, 1966). Absorption of the incident X-rays by the counter-tube gas or scintillation crystal may cause X-ray fluorescence. If this is re-absorbed in the active volume of the counter only one pulse is produced of average amplitude A₁ proportional to the incident X-ray quantum energy (k = constant) However, the gas or crystal has a low absorption coefficient for its own fluorescent radiation, hence, some quanta of the latter of energy e₂ may escape from the active volume of the counter, the amount depending on the geometry of the tube, gas, windows, etc. The average amplitude of the escape pulses is Thus,

The pulse-height analyser discriminates against pulses only on the basis of their amplitudes. When it is set to detect X-rays of energy e₀, it is also sensitive to X-rays of energy . For example, when using an NaI scintillation counter for Cu Kα, e₀ = 8 keV, and for the escape X-rays I Kα, e₂ = 28.5 keV. A pulse-height analyser set to detect X-rays of energy 8 keV is also sensitive to X-rays of energy 36.5 keV, because, from equations (7.1.4.3) and (7.1.4.4), In Figs. 2.3.5.3(c), (d) and 7.1.4.1(d), the escape peak E.P. shows clearly at 0.35 Å, the wavelength of 36.5 keV X-rays. There may be a number of weak escape peaks arising from the stronger powder reflections. In practice, the escape peak should not be confused with a small-angle reflection. It can be tested by reducing the X-ray tube voltage to below the absorption-edge energy of the element in the detector from which it arises.

Detectors may either be true counters in which individual detected photons are counted or they may be integrating devices that generate a signal that is a function of the rate of arrival of photons; this signal may then be digitized for recording purposes.

The choice of a linear or of a two-dimensional, or area, PSD depends on its detection efficiency, linearity of response to incident X-ray flux, dynamic range, and spatial resolution, its uniformity of response and spatial distortion, its energy discrimination, suitability for dynamic measurements, its stability, including resistance to radiation damage, and its size and weight.

Before discussing a few of the many types of PSD individually, we shall examine these criteria in a general way. Some of them have been discussed by Gruner & Milch (1982). Table 7.1.6.1 shows the importance, on a scale of 0 to 3, of some of the factors in different fields of study. Other properties, such as stability and sensitivity, are equally important for all PSD's.

7.1.6.1.4. Spatial resolution

| top | pdf |

The spatial resolution of a PSD is determined by the number and size of resolution or picture elements (pixels) along the length or parallel to the edge of the detector. In most diffraction experiments, the size of the pattern can be scaled by altering the distance of the detector from the sample and what is important is the angular resolution of the detector when placed at a distance where it can `see' the entire pattern. We shall see below that linear PSD's can be made with up to 2000 pixels and that area detectors are mostly limited to fewer than 512 × 512 pixels. The sizes of pixels range from about 10 µm for semiconductor PSD's to about 1 mm for most gas-filled detectors.

The useful number of pixels of a detector is determined by its point-spread function (PSF). This is the relative response as a function of distance from the centre of a point image on the detector. PSF's are not necessarily radially symmetrical and may have to be specified in at least two directions at right angles, for example along and perpendicular to the lines of a television raster scan. The width of the PSF at the 50% level determines the amount of detail visible in a directly viewed image. The accuracy of intensity measurements may depend more critically upon the width of the PSF at a lower level, since a weak spot may be immeasurable when sitting on the `tail' of a very intense one. For various physical reasons, the PSF's of all PSD's, including X-ray film, have appreciable tails.

The spatial resolution of a detector is affected by parallax: when an X-ray beam is absorbed in a thick planar detector at an angle to the normal, the width of the resultant image is smeared out exponentially and its centroid is shifted by an amount . For 8 keV X-rays incident at 45° on a xenon-filled counter, for example, this shift is about 4 mm for a filling pressure of 1 atm and 0.4 mm for a filling pressure of 10 atm. These figures illustrate the desirability of high-pressure xenon (Fig. 7.1.6.2 ) for gas-ionization detectors intended for wide-angle diffraction patterns.

Figure 7.1.6.2| top | pdf | Absorption of 8 keV and 17 keV photons in argon and xenon as a function of pressure in atm × column length in mm.

In all gas-filled counters, whether one-, two-, or three-dimensional, the initial event is the absorption of the incoming X-ray photon in a gas molecule with the emission of a photo-, or alternatively an Auger, electron. The detection efficiency depends on the fraction of the photons absorbed in the gas and this fraction is shown in Fig. 7.1.6.2 as a function of the product of gas pressure and column length for 8 and 17 keV photons on argon and xenon. The ionization energy of noble gases is about 30 eV so that one 8 keV photon gives rise to about 270 electron–ion pairs. With adequately high collecting fields, the electrons acquire sufficient energy to produce further ionization by collision with neutral filling gas molecules; this process is often referred to as `avalanche production' or `gas multiplication'. The factor A by which the number of primary ion pairs is multiplied can be as great as ten to one hundred thousand. Up to a certain value of A, the total amount of ionization remains proportional to the energy of the original X-ray photon. The electrical signal generated at the anode of the counter is due very largely to the movement of the positive ions from the immediate vicinity of that electrode; at the same time, a corresponding pulse is induced on the cathode. The signal can be shaped to produce a pulse with a duration of the order of a microsecond.

In single or multiwire proportional counters, the secondary ionization (avalanche production) takes place in the highest field region, that is, within a distance of a few wire diameters of the anode wire or wires. The electrons are collected on the anode and the positive ions move towards the cathode, with very little spread of the ionization in a direction perpendicular to the field gradient, that is, parallel to the wire direction. It is thus possible to construct position-sensitive devices based on such chambers.

Proportional-counter behaviour is discussed in detail in many standard texts and review articles (Wilkinson, 1950; Price, 1964; Dyson, 1973; Rice-Evans, 1974).

The gas amplification does not have to take place in the same region of the detector as the original absorption. In so-called drift chambers, the primary ionizing event takes place in a low-field region where no avalanching takes place. The electrons drift through a grid or grids into a region where the field is sufficiently high for gas multiplication to occur. The drift field can be made cylindrical in a linear counter (Pernot, Kahn, Fourme, Leboucher, Million, Santiard & Charpak, 1982), or spherical in an area detector (Charpak, 1982; Kahn, Fourme, Bosshard, Caudron, Santiard & Charpak, 1982), centred on the point from which the X-rays diverge, that is on the specimen; the electrons then drift in a radial direction without parallax being introduced (Fig. 7.1.6.3 ).

Figure 7.1.6.3| top | pdf | Spherical drift chamber multiwire proportional chamber (MWPC) (Charpak, 1982; courtesy of G. Charpak).

In many experiments, use is made of the energy discrimination of the detector. The ratio of the full width at half-maximum to the position of the maximum of the pulse-height distribution is given by where N is the number of primary ion pairs produced, F is the Fano factor (Fano, 1946, 1947), which takes into account the partially stoichastic character of the gas multiplication process, and f is the avalanche factor. For proportional counters filled with typical gas mixtures (argon + methane), F = 0.17 and f = 0.65, so that for 8 keV photons , but, in the so-called Penning gas mixtures (e.g. noble gas and ethylene), f can approach zero at a certain field strength. In a wire counter with its rapidly varying field strength, f is small only for a gas amplification of less than 50. The energy resolution for 8 keV photons could then be as low as 6%, but the pulses induced on the cathode wires of a MWPC are then too small to permit a precise localization. This problem has been overcome by using uniform-field avalanching in two regions in tandem, separated by a drift space (Schwarz & Mason, 1984, 1985). The energy information was derived after the first low-gain gas multiplication process (A ∼500): a proportion of the electrons from the first avalanche then drifted into the second avalanche region which boosted the gas gain to more than 10⁵, necessary to give a high spatial resolution.

In an alternative method (Charpak, 1982; Siegmund, Culhane, Mason & Sanford, 1982), the additive avalanche factor f is eliminated by deriving the energy information, not from the collected charge, but from the visible light pulse produced by the individual avalanches of each primary electron.

7.1.6.2.3. Current ionization PSD's

| top | pdf |

For the very highest counting rates, it is necessary to abandon all methods in which individual X-ray photons are counted and instead to measure the ionization current produced by the incident X-rays on either cathode or anode. Fig. 7.1.6.6 shows the principles of a cathode read-out linear PSD. The cathode is divided into strips, each of which is connected to a capacitor and to an input terminal of a CMOS analogue multiplexer. The charge accumulated on each capacitor in a given time period is transferred to a charge-sensitive amplifier when the associated channel is selected by an addressing signal. The output voltage of the amplifier is digitized by means of an analogue-to-digital converter. The complete pattern is scanned by incrementing the addresses sequentially: The resolution is that of the strip spacing (∼0.5 mm) and the principle can be extended to two dimensions (Hasegawa, Mochiki & Sekiguchi, 1981; Mochiki, Hasegawa, Sekiguchi & Yoshioka, 1981; Mochiki, 1984; Mochiki & Hasegawa, 1985). Global count rates in excess of 10⁹ s⁻¹ are possible with this method. Lewis (1994) has published a comprehensive survey of the present status and the future potentialities of gas-filled position-sensitive detectors.

Figure 7.1.6.6| top | pdf | Integrating LPSD. From Mochiki (1984); courtesy of K. Mochiki.

Semiconductor detectors are essentially solid-state ionization chambers in which the incoming photon generates electron–hole pairs instead of electron–ion pairs. Semiconductor detectors are sensitive to the entire electromagnetic spectrum from visible to X-rays, and they can also detect electrons directly. There are, therefore, three possible methods of constructing semiconductor X-ray detectors, each of which has its advantages and disadvantages.

In semiconductor one- or two-dimensional PSD's, the detector and the read-out circuitry are usually integrated on the same chip. In these devices, the pixel size and position are fixed once and for all so that they are geometrically completely stable. Integrated read-out circuitry has a low input capacity and thus an extremely low read-out noise.

All semiconductor X-ray PSD's are derived from imagers for visible light, which are of two basic types: photodiode arrays (PDA's) and charge-coupled devices (CCD's) (Lowrance, 1979; Allinson, 1982; Borso, 1982; Third European Symposium on Semiconductor Detectors, 1984).

In PDA's, the detection of a photon-generated charge takes place in a depletion layer that is formed either in a suitably biased p–n diode or in a metal-oxide-semiconductor (MOS) capacitor. The electron–hole pairs are separated by the field associated with the depletion layer. The individual diodes store charge during the integration period; this is read out into a common video output line via MOS multiplexing switches. This architecture is usually adopted for linear arrays where the switches can be arranged around the periphery of the diodes with a minimum amount of dead space between them and where they can be shielded from the incident X-rays. PDA's tend to suffer from a high fixed-pattern noise due to differences in the performance of individual MOS switches.

In CCD's (Howes & Morgan, 1979), the sensing elements are always MOS capacitors suitably biased to establish charge-storage volumes. During read out, the charge is transferred in a `bucket-brigade' fashion from one MOS capacitor to the next until it reaches the output. In linear CCD's, this output is at one end of what is essentially an analogue shift register. Most two-dimensional CCD's are frame-transfer devices: at the end of the exposure, the charge is transferred line by line to an identical array of MOS elements; while the next `frame' is exposed in the image array, the contents of the buffer array are transferred, one line at a time, to a single-line buffer from which they are shifted out element by element. Since the transfer circuitry in CCD's is interlaced with the detector elements, they are more difficult to shield and are more subject to radiation damage.

Image converters and television cameras with X-ray-sensitive photocathodes have been reported. These cathodes must be thin, or they must have a low bulk density to allow the photoelectrons to escape (Bateman & Apsimon, 1979; Haubold, 1984).

Television tubes with a beryllium window and a 25 mm diameter lead oxide target are available and experimental 125 mm tubes have been described (Suzuki, Uchiyama & Ito, 1976).

X-ray-sensitive Saticon television camera tubes with an amorphous selenium–arsenic target (Chikawa, Sato & Fujimoto, 1984) have an active diameter of only 10 mm but a limiting resolution (MTF = 5%) of 6 µm. They have an absorption efficiency of 52% for Mo Kα radiation and are used mainly for X-ray topography. Other X-ray-sensitive television camera tubes have been described by Matsushima, Koyama, Tanimoto & Yano (1987), Suzuki, Hayakawa, Usami, Hirano, Endoh & Okamura (1989) and Sato, Maruyama, Goto, Fujimoto, Shidara, Kawamura, Hirai, Sakai & Chikawa (1993). Such tubes are discussed further in Section 7.1.7.

Much development has gone into quantitative measurements with area detectors in which the diffraction pattern is formed on an external phosphor fibre-optically coupled to a low-light-level television camera. Mostly the television camera embodies a demagnifying image intensifier coupled via demagnifying optics to a sensitive television camera tube (Arndt & Gilmore, 1979; Arndt, 1982, 1985; Arndt & In't Veld, 1988; Kalata, 1982, 1985; Gruner, Milch & Reynolds, 1982) or to a CCD or an array of CCD's (Strauss, Naday, Sherman, Kraimer & Westbrook, 1987; Strauss, Westbrook, Naday, Coleman, Westbrook, Travis, Sweet, Pflugrath & Stanton, 1990; Templer, Gruner & Eikenberry, 1988; Widom & Feng, 1989; Fuchs, Wu & Chu, 1990; Karellas, Liu, Harris & D'Orsi, 1992). Camera tubes were frequently read at commercial television rates (625 lines with a field repetition rate of 25 Hz or 525 lines with a field repetition rate of 30 Hz in Europe and in North America and Japan, respectively), leading to pixel rates of about 10 MHz. Successive images were then digitized and their sums stored in memory (see Fig. 7.1.6.7 ). Milch, Gruner & Reynolds (1982) developed a slow-scan method for a silicon-intensifier-target (SIT) tube. This latter method has been facilitated by the development of very large scale integration memory circuits, which have made it possible to construct economical image stores into which the camera can write at a slow rate and which can then be read at normal television rates for display purposes. CCD's are usually operated in this fashion, so that the read–out circuits can have a narrow band-width and produce an excellent signal-to-noise ratio.

Figure 7.1.6.7| top | pdf | Fast-scanning television X-ray detector (after Arndt, 1985).

For very low X-ray intensities in which the probability of the arrival of a photon per pixel per frame period is much less than one, the camera can be operated at normal frame frequencies in a digital mode. Specially designed circuits detect the charge image produced by a single X-ray photon and find the centroid of this image (Kalata, 1982); the events are `counted' in a histogramming memory. The method is capable of some energy discrimination and has a high spatial resolution because the centroid of the image can be found to a high precision.

The use of linear and area detectors has increased markedly in recent years (Arndt, 1988). No new principles for the construction of linear devices have emerged, but more examples of each type have become commercially available.

Many more structures have been determined with area-detector diffractometers. The most commonly used gas-filled detectors at present are the delay-line read out MWPC first described by Xuong et al. (1978), as developed by Hamlin (1985), and a detector using a modification of the coded-anode read out due to Burns (Durbin, Burns, Moulai, Metcalf, Freymann, Blum, Anderson, Harrison & Wiley, 1986; Howard, Gilliland, Finzel, Poulos, Ohlendorf & Salemne, 1987; Derewenda & Helliwell, 1989). Corresponding instruments in the USSR and their use have been described by Anisimov, Zanevskii, Ivanov, Morchan, Peshekhonov, Chan Dyk Tkhan, Chan Khyo Dao, Cheremukhina & Chernenko (1986) and by Andrianova, Popov, Kheiker, Zanevskii, Ivanov, Peshekhonov & Chernenko (1986).

A two-dimensional photon-counting X-ray detector has been described by Collett & Podolsky (1988).

The widespread use of area-detector methods in single-crystal studies, especially for macromolecular material, has been greatly aided by the development of complete software packages to deal with all aspects of data collection and handling. Earlier program packages (Howard, Nielson & Xuong, 1985; Pflugrath & Messerschmidt, 1987; Thomas, 1987) tended to be specific to one particular detector and its associated diffractometer. However, following the initiative of Bricogne (1987) and with financial assistance from EEC funds, a group of scientists, including most of the originators of the earlier packages, are now collaborating in writing, extending, and maintaining a comprehensive device-independent position-sensitive-detector software package.

Video displays of X-ray and optical images have different features. Although X-ray photon energies are very large, the intensities available in X-ray diffraction are extremely low compared with optical images. Therefore, the photon noise resulting from the statistical fluctuation of the number of photons incident upon an image system gives a detection limit of the image.

Consider the case of defects in a crystal viewed by an imaging system: ν_p photons s⁻¹ mm⁻² are diffracted from the perfect region of a crystal, and are absorbed by the X-ray-sensing layer of the system. An absorbed photon produces a mean number of electrons or visible photons, each of which may be rescattered to produce a mean number of electrons or photons. By repeating s such processes, the mean signal height is obtained from each square-shaped picture element δ × δ mm for t s. The value of δ may be taken as the limiting resolution of the system. Since owing to the large photon energy and the values of , are considered to be less than 100, the photon noise σ_p as a standard deviation of S_p is given by (Arcese, 1964) In order to recognize the defect image in the perfect region, the signal-to-noise ratio (SNR) of the image must be where S_d = signal height of the defect image, C = contrast of the defect image, ; and κ = threshold SNR of 1–5 (Rose, 1948).

Sensitivity is expressed by the intensity required for a certain signal height to the noise height of the imaging system and is proportional to the value of . As is seen from (7.1.7.3), however, the ratio of the signal to the photon noise depends only on the absorption efficiency q. For example, a 100 µm thick silicon-diode array (Rozgonyi, Haszlo & Statile, 1970) and 15 µm thick lead oxide camera tubes (Chikawa, 1974) have similar sensitivities for Mo Kα radiation, but q = 0.15 and q = 0.6, respectively; the former obtains the sensitivity with a higher conversion efficiency of absorbed photon energy and gives a (0.15/0.6)^1/2 = 0.5 times lower SNR.

It is clear from the foregoing discussion that imaging systems should be evaluated by three factors: resolution, integration time for observation, and SNR.

There are two kinds of imaging system (e.g. Green, 1977); one is a direct method in which X-ray input images are converted directly into video signals and the other is an indirect method in which X-ray images are first converted into visible-light images to be observed by the usual electro-optical system. In the latter, phosphor screens are widely used and the visible images can be magnified by a lens. However, the resolution is limited by the thickness of the phosphor screen that is required for a satisfactory absorption efficiency. In contrast, the direct method provides a high resolution.

Various types of indirect method are discussed in Subsections 2.7.5.2 and 7.1.6.5. Here only the direct method will be described in some detail.

A direct method using X-ray TV camera tubes has been used for real-time topography (Chikawa & Matsui, 1994). The construction and operation of an X-ray TV camera tube are similar to those of the conventional video pick-up tube, except for a beryllium window, as shown schematically in Fig. 7.1.7.1. The popular one is a vacuum glass tube which is 15 cm long and 25 mm in diameter. An X-ray sensing photoconductive layer such as Se (McMaster, Photen & Mitchell, 1967), or PbO (Chikawa, 1974) is placed on the inner surface of the window at one end of the tube and is scanned by a narrow electron beam from a cathode on the opposite end to convert image charges on the photoconductive layer into video signals. Therefore, the resolution depends upon the characteristics of the photoconductive layer and the electron-beam size on the layer.

Figure 7.1.7.1| top | pdf | Schematic illustration of an X-ray sensing Saticon camera tube. (a) Schematic representation of the tube. (b) Structure of the target. (c) Concentration (wt%) of As in the target (Se–As photoconductive layer). Since crystalline Se is metallic, As acts to stabilize the amorphous state. (d) Potential in the target. A blocking contact is formed between the X-ray window material and the Se–As alloy layer to prevent holes from flowing into the layer. Incident X-rays form electrons and holes in the layer, and the latter migrate toward the scanning-electron-beam side and contribute to the video signal. On the surface of the Se–As layer, Sb2S3 was evaporated to form another blocking contact that improves landing characteristics of the electron beam. By applying a high voltage on the layer, the holes are accelerated to produce multiplication (avalanche amplification), resulting in a great increase of the signal current.

The camera tube with a PbO photoconductive layer has absorption efficiencies of q = 0.6 for Mo Kα and 0.8 for Cu Kα, using a PbO layer with a resolution-limited maximum thickness of 15 µm (Chikawa, 1974).

Dislocations in a silicon crystal were observed with the PbO camera tube and a high-power X-ray generator (tube voltage: 60 kV, current: 0.5 A) under the conditions ν_p = 4 × 10¹¹ Mo Kα₁ photons s⁻¹ m⁻², C = 0.5 and δ = 30 µm (Chikawa, 1974) in (7.1.7.3). In order to improve the resolution further, q must be as high as possible even when is increased by 10–100 by use of synchrotron radiation, because decreases with improving resolution (decreasing δ). (Note that C is not expected to increase much with improving resolution.) For example, to keep at the same level as the previous case, q should be 0.7, say, for ν_p = 10¹³ photons s⁻¹ m⁻² and δ = 6 µm. However, resolution and detection efficiency are, in general, mutually exclusive; the resolution of X-ray sensing photoconductive or phosphor materials is determined by the sizes of their grains and their sensitivities decrease with decreasing grain sizes. High resolution, without sacrificing detection efficiency, has been obtained with an amorphous Se–As alloy photoconductive layer, which has the advantage that no degradation of resolution occurs on increasing the layer thickness (Chikawa, Sato, Kawamura, Kuriyama, Yamashita & Goto, 1986). A layer with a thickness of 20 µm has an absorption efficiency of q = 0.52 for Mo Kα. The resolution of these tubes is shown with their modulation transfer functions in Fig. 7.1.7.2. The limiting resolution of 6 µm was obtained by making the scanning electron beam narrower with a diode-type electron gun have a barium-impregnated tungsten cathode (DIS type) (Chikawa, 1999). The Se–As layer shows very low lag characteristics (less than 1% after one frame).

Figure 7.1.7.2| top | pdf | Square-wave modulation transfer function (MTF) measured for the DIS-type, conventional-type Se–As and PbO camera tubes. This function shows the magnitude of brightness modulation in the output image obtained for an input image with a square-wave intensity distribution and is defined by where and are the maximum and minimum intensities of the input image, respectively, and and are those of the output. DIS stands for diode-operation impregnated-cathode Saticon.

This type of camera tube was developed primarily for TV broadcasting use and named `Saticon' as the acronym for the components of the photoconductor, Se, As and Te. (Te enhances the sensitivity to red light.)

Such camera tubes have excellent linearity. The output signal current is proportional to the incident X-ray intensity up to 4 µA. The amplifier is always improving in signal-to-noise ratio, and, at present, the noise level of video amplifiers is ∼0.2 nA; it is equivalent to an input intensity of ∼10¹¹ Mo Kα photons m⁻² s⁻¹ in the US/Japan standard scanning system (30 frames s⁻¹).

Therefore, it is important to increase conversion efficiency . With increasing voltage applied on the photoconductive layer, the signal current is saturated (in Fig. 7.1.7.1, all the holes produced by an incident photon are collected), and then increases again further by avalanche amplification, as shown in Fig. 7.1.7.3; holes accelerated by a strong electric field cause their multiplication (Sato, Maruyama, Goto, Fujimoto, Shidara, Kawamura, Hirai, Sakai & Chikawa, 1993). It was referred to as `HARP' (high-gain avalanche-rushing amorphous photoconductor). Together with the signal current, the dark current also increases as shown in Fig. 7.1.7.3. By allowing it to increase to the noise level of the video amplifier, an order of magnitude higher SNR can be achieved. Consequently, individual X-ray photons can be imaged as spots with a size resulting from the point-spread function, unless they are absorbed near the back surface of the photoconductive layer. For X-ray detection, a thick HARP layer should be employed with a very high applied voltage, and stable operation with avalanche amplification was confirmed for a 25 µm thick HARP layer. These pilot tubes were fabricated with a conventional electron gun and had a resolution of about 25 µm.

Figure 7.1.7.3| top | pdf | Typical example showing dependence of the avalanche amplification on the electric field applied on the amorphous photoconductor (HARP). The solid and dashed lines show the amplification of signal current by visible light and dark current, respectively, for a 2 µm thick HARP layer. The right-hand scale shows the current (in nA). The open circles are the data for an 8 µm thick HARP layer by X-ray irradiation. Note that the thicker layer has higher amplifications owing to longer paths for running of holes.

In general, avalanche amplification results in degradation of spatial resolution and has been used for zero-dimensional detection such as solid-state detectors. To make two-dimensional detection, isolation of each picture element is required. For the HARP, however, no appreciable degradation of resolution due to avalanche amplification was confirmed with a DIS-type tube having an 8 µm thick HARP layer by using visible light through a glass window. X-ray sensitive DIS-type tubes are now commercially available.

The resolution δ and integration time t should be selected appropriately according to experimental requirements (Chikawa, 1980). For example, when topographic images of a single dislocation in silicon were observed with t = 1/30 s by synchrotron radiation, their contrast and SNR were found to be C = 0.5 and R_I = 20 for δ = 30 µm, and C = 1 and R_I = 8 for δ = 6 µm. Since the SNR is desired to be 100, the integration time should be as large as possible unless images of moving objects are degraded. Digital image processing (Heynes, 1977) enables one to adjust the integration time easily. As an example, a noise reducer (McMann, Kreinik, Moore, Kaiser & Rossi, 1978; Rossi, 1978) is shown in Fig. 7.1.7.4. The video signal is sampled and digitized by the A/D converter and the digital video is sent to the adder and thence to the memory. Image information in the memory is continually sent both to the adder through the multiplier for combination with incoming data and to the display through the D/A converter. The weighting of new to old data is made by changing the factor k of the multiplier in the range . For k = 0, the original input image is displayed. In the range , a sliding summation of successive frames is displayed, and the SNR is improved by a factor of . The factor k can be adjusted automatically by detecting the difference between successive frames.

Figure 7.1.7.4| top | pdf | Principles of a noise reducer.

Using a HARP tube, the SNR can be improved without integration of the amplifier noise by image processing, and topographs were displayed with an intensity of  photons s⁻¹ m⁻² by a conventional X-ray generator.

Acquisition of extremely low intensity images, dramatic improvements in SNR via frame integration, and isolation and enhancement of selected-contrast ranges are possible by digital image processing (Chikawa & Kuriyama, 1991).