Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/user/socpub/ccd_review/node9.html
Дата изменения: Fri Aug 30 15:55:16 1996
Дата индексирования: Tue Oct 2 11:53:31 2012
Кодировка:
Поисковые слова: рер р р р р р р р р р р р р р

Back Illumination

The technique of back-illumination was first described by [Shortes et al], and has been widely investigated in order to improve blue and UV sensitivities of CCDs. The method requires that the CCD substrate be etched from the rear towards the depletion layer boundary. The device can be illuminated through this rear surface, which has a much thinner dead layer than the electrode side.

In practice, the processing of devices for back illumination is complex. If left untreated, the rear surface has free charge states which produce a small depletion layer which drifts signal charge to the rear surface, where the surface recombination velocity is high. Consequently, photo-electrons generated near this surface do not give rise to a measurable signal, yet these are precisely the photon events which back-illumination was designed to detect more readily. There are a number of methods for treating the back surface to prevent the recombination phenomenon, by producing an electric field to drive signal charge towards the front-surface collection potential wells. These methods include:

• Back surface charging by UV flooding [Janesick et al 1985]
• Flashgate deposition [Janesick et al 1986]
• Biased flashgate [Janesick et al 1989a]
• Ion implantation [Stern et al]

One issue to be considered in choosing the appropriate technology is that of stability. The first method requires intermittent re-charging of the rear surface in order to maintain the accumulated condition. Deposition of ultra-thin metallic gates may also create a varying performance if the metal diffuses into the silicon with time. Any further changes caused by temperature, signal flux or ionising radiation will also tend to produce variations in response.

It should be noted that for X-ray spectroscopic applications, not only is a high detection efficiency required, but also the complete collection of all signal electrons in a photo-generated charge packet must be ensured. Variation in physical parameters of the surface may therefore also give rise to a variable spectroscopic response.

A fully depleted condition is necessary to reduce the proportion of inter-pixel split events. The ion-implant and biased flash gates are appropriate technologies for this mode, as their rear surfaces act like equi-potential planes which maintain the surface in an accumulated state while the device is in full depletion. However the flash oxide quality may not be sufficiently uniform to maintain the surface potential correctly at all points on the surface. The depletion edge may then reach the rear surface states in some places, and in this case, the dark current generation rate is increased at the rear surface, and furthermore those electrons generated are more efficiently swept towards the collection site: the ion implant method should be superior in suppressing rear surface dark current.

To obtain reasonable detection efficiencies at energies as low as 300-400eV, particularly if compromised by light filter thicknesses of (see below), requires that the silicon dead layers are no more than a few hundred thick. This is feasible for the ion implant technology, whilst the thickness of flash gates may be ~ 10's . There is clearly a trade-off between the biased flash gate detectors and ion-implant devices for sensitivity versus operational stability.

The requirement for the minimization of split events demands that radial spread of charge occurring during the drift through the depletion layer is also minimised, but thinner depletion layers conflict with the requirement to increase the depletion layer thickness for enhanced hard X-ray response. Figure 5 represents the potential profile in a back illuminated CCD and indicates that the events generated nearest the back surface are initially subject to the lowest drift field. In addition, any events in the very thin ( m) implant layer may partially recombine, so that the enhanced soft X-ray detection efficiencies may be obtained only at the cost of impaired spectroscopic information.

Figure 6 shows the pulse heights observed in a back-illuminated EEV device [Bailey et al 1991] with a 0.2 m implant layer. Examination of this figure shows that at the lower energies (shorter absorption lengths) an increasing fraction of the events are detected in a low energy tail, rather than in the full energy photo-peak. These are the events which experience splitting and recombination near the rear surface. The ability to derive spectral information is hampered by this effect.

The corresponding detection efficiency data are presented in Figure 7, and compared with the predicted data for a 20 m thick depletion layer. Above 2keV, the efficiency is limited by the transparency of the depletion layer. In the soft X-ray region, conventional CCDs have negligible response below the O absorption edge at ~ 530eV, whilst this device extends the response to ~300eV. EEV are investigating further improvements by reducing the implant layer thickness through the use of an excimer laser instead of a ruby laser anneal. This should reduce the dead layer from 0.25 m to 0.05 m.

It should be noted that there is a requirement to inhibit the visible and UV signal from cosmic sources which the CCD is also efficient in detecting. For some stellar sources with L/L this requires significant attenuation. The detection of any optical photons introduces an offset in X-ray energy scale at the contaminated pixel locations, and also an additional noise source due to the shot noise on photon flux. The design of light filters depends in principle on the optics throughput and point response function as well as the CCD efficiency. In practice, aluminum is the preferred material, and to minimize pinhole effects a thickness of as much as 1000 is required. This aluminum layer might be deposited directly on the CCD electrodes or onto a mesh or plastic support such as Lexan. Whilst this form of filter offers rejection of light, it also dominates the soft X-ray response roll-off.