Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/user/socpub/ccd_review/node3.html
Дата изменения: Fri Aug 30 15:55:01 1996
Дата индексирования: Tue Oct 2 11:51:53 2012
Кодировка:
Поисковые слова: с п п п п р п р п п р п р п п п п п п п п п р п р п п п п р п р п п

Contrasting optical and X-ray applications

A visible light photon releases only one electron-hole pair, and consequently many photons must be collected in each pixel for a measurable signal to be produced. Therefore at visible wavelengths CCDs are used as integrating detectors, and typically the astronomical image exposures are several minutes long. Conversely, a single X-ray photon has sufficient energy to form multiple electron-hole pairs through the process of secondary ionization by the primary photoelectron. An average of one electron-hole pair is liberated for each 3.68 eV of photon energy absorbed (this figure varies slightly with temperature). The charge liberated by single X-ray photons ( ~ 100-1000 electrons) is easily detectable if the amplifier noise is low enough. Therefore at X-ray energies CCDs can be used as photon counting detectors, with the measured signal charges proportional to the photon energies.

This use as an imaging X-ray spectrometer requires that no more than one photon is incident on each pixel in any image frame, which imposes a requirement that image collection times be limited to ~seconds for typical astronomical applications. In addition the signal charge must ideally be completely collected within the original pixel, transported to the output node without losses due to imperfect charge transfer efficiency (CTE), and measured without degradation by device readout noise.

Each of these requirements places special demands on the detector structure, and complicates the analysis of the data produced by the device. Charge collection efficiency is a function of the electric field strength at the site of X-ray absorption and is better in devices fabricated from higher resistivity material [Bautz et al, Hopkinson, Lumb et al]. Charge collection efficiency also improves as the pixel size increases. Even in high resistivity devices with large pixels, however, a significant fraction of all X-ray interactions will deposit charge in more than one pixel. A crucial consequence of imperfect charge collection is that not all detectable interactions yield useful spectroscopic information. Since the distribution of charge in a group of adjacent pixels is an indicator of the efficiency of charge collection [Bautz et al], maximum spectral resolution can only be obtained if a mutiplet of pixel values (usually a 3-pixel-by-3-pixel square neighborhood) is analyzed for each event [Lumb and Holland 1988a]. Moreover, as will be illustrated in Section 3.1, one can trade spectral resolution for detection efficiency by varying event analysis parameters. (Analysis of pixel multiplets also provides a means to discriminate between X-ray photon events and background events produced by high-energy particles. See section 3.3)

Charge transfer within a CCD is subject to inefficiencies caused by the trapping of signal charge at discrete sites in the silicon. These may be at crystalline defects or at sites with a defect introduced by a manufacturing or design error. The probability of trapping any signal depends upon many factors such as the temperature, clock rate and previous history of charge passing through the trap. In ground-based optical imaging applications there is often sufficient signal in every pixel due to photons from the night sky background that all traps remained filled, and they have a negligible effect on the image data. For X-ray photon counting applications, there is no such background, and the event arrival rate requirements noted above almost guarantee that traps will depopulate between the arrival of successive signal packets. Charge transfer losses therefore tend to be more severe for X-ray astronomy. The result is that even with a charge transfer inefficiency as low as 1e-5 per pixel, charge packets traversing 1000 pixels will lose 1% of their signal, producing an apparent spatially varying gain function. In principle this is correctable, but the variation of the transfer losses with other parameters, particularly radiation damage, hinder this correction.