All traditional photoconductors, such as the MIPS Ge:Ga unstressed and stressed detectors, show multiple time constant response. In addition, they can exhibit spontaneous spiking and non-monotonic response characteristics (the ''hook''). These behaviors are indicated qualitatively in Figure 2.12. Although detailed matching of the theory with observed behavior requires use of numerical techniques, the main features in Figure 2.12 can be understood readily. The initial rapid response is due to the generation and recombination of charge carriers in the bulk region of the detector. The ''hook'' response arises through dielectric relaxation effects. The initial rapid migration of free electrons can leave a distribution of charge that reduces the electric field in the bulk and hence reduces the gain of the detector, resulting in a reduction of response after the initial reaction to the signal.
The other effects arise at the detector contacts. In the first order theory of photoconductive response, it is assumed that a charge carrier absorbed at one electrode is ''immediately'' replaced by one emitted at the other to maintain electrical neutrality in the detector volume. However, the necessity to emit a charge carrier to maintain equilibrium is only communicated across the detector at roughly the dielectric time constant (that is, basically the RC time constant of the detector), so in fact the detector space charge adjusts to a new configuration at this relatively slow rate, not ''immediately.'' The result is a slow increase in detector response lasting tens of minutes (MIPS 160 µm detectors) to hours (MIPS 70 µm detectors).
In addition, at the contact junction between conductor and semiconductor, large fields are produced by the migration of charges to reconcile the differing band gap structures. These fields are a result of the required matching of Fermi levels between the metal contact and the semiconductor detector. To control the peak fields, it is necessary to produce a 'graded contact' by adding dopants through ion implanting to increase the electrical conductivity in a thin layer of semiconductor just below the metal contact. Nonetheless, as the illumination level of the detector is changed, these fields must adjust. During this process, the field local to a small region near a contact can accelerate charge carriers sufficiently to impact-ionize impurities in the material, producing a mini-avalanche of charge carriers that appears as a spike on the output signal. Such spikes are indistinguishable from cosmic ray hits, and are removed in pipeline processing. Care in the manufacture of the contacts for a detector can reduce some of these undesirable effects, such as spiking. It is also helpful to operate the detector at a constant bias; the CTIA readouts used in MIPS have this advantage. Figure 2.13 shows that the response characteristics of the MIPS detectors closely follow the qualitative behavior discussed above.
Figure 2.12: Schematic response characteristics of a bulk photoconductor. The signal indicated with a solid line increases from zero to a fixed level at time t2, as indicated in (a). The detector response, in (b), shows a variety of adjustments including ''hook'', spontaneous spiking, and a slow tail as a result of the drastic change in detector conductivity. The signal indicated with a dashed line is of the same size as the one at t2, but the detector was already being illuminated at a level comparable with the signal level, similar to observing a source against a background. The background flux helps suppress the hook and spike response of the detectors, and lessens the impact of the slow response on photometry.
The fast generation-recombination response (fast response) of the detectors is not subject to the problems inherent in the longer term response described above. MIPS takes advantage of this fact by using the scan mirror to modulate (chop) the source signal on timescales of a few seconds, keeping measurements mostly confined to the fast response regime. At very low background levels such as those provided by Spitzer, it is possible to extract the source signal largely from the fast component, thereby mitigating most of the undesirable longer-term response effects. Frequent use of on-board calibration sources (see stimulator discussion in section 4.1.1) additionally allows tracking of the long-term drifts in response.
One potential area where the long-timescale behavior of the Ge:Ga arrays may be of concern is in situations where there is a high background level in the region of sky being observed. When a significant background flux falls on the detectors, the dielectric time constants are reduced because the background illumination produces a steady state concentration of free charge carriers.
Figure 2.13: Measured response of a pixel of the MIPS 70 µm array to a sudden increase in illumination from a dark background. The hook and long-term increase in response are evident, as are 3 cosmic ray hits.
The 24 µm silicon BIB array is much better behaved than the photoconductors used for the two far-infrared channels. The improvement is inherent in the structure of the detector, which separates the functions of photon absorption and maintaining high detector impedance. Consequently, the infrared-active layer is sufficiently heavily doped that it adjusts to new equilibrium conditions rapidly and avoids the 'two time constant' behavior seen with the germanium detectors. The high absorption in this layer allows it to be small and the cosmic ray hit rate is low. Again, because of the high level of doping, cosmic rays have much less of a tendency to modify the solid state properties of the detector and thereby to produce the problematical response characteristics of the germanium devices.
Nonetheless, the silicon array has some non-ideal behavior. Its output signal is subject to 'droop' whereby the output for a pixel is proportional to the photon signal that fell on that pixel plus a signal proportional to the average signal falling over the entire array. The MIPS flight array has a droop coefficient of 0.33, meaning that 33% of the average integrated charge across the array appears in the output of every pixel. In addition, after observing a bright source, there is a latent image (usually at the < 1% level). The latent images decay very slowly, taking up to ~10 minutes. The droop phenomenon is corrected in the calibration pipeline processing, but latent images may leave artifacts in the data delivered to users. Please see Data Features and Artifacts for additional information and examples.