The STJ has a very high efficiency to the absorption of photons of wavelength
ranging from the ultraviolet to the near infrared. Figure 4 (click here)a illustrates the
absorption efficiency for single photons in Nb, Al and Hf films of thickness
100 nm based on the optical constants of the materials (Weaver & Lynch 1973;
Weaver et al. 1981) and indirectly confirmed for Nb-based
devices through optical
transmission measurements in our laboratory. From 100 nm to 
the
absorption efficiency is over 80 per cent for all three materials. The surface
reflectivity of these films is shown in Fig. 4 (click here)b. While pure Al has
a very high reflectivity from the ultraviolet to the near infrared, typically over
90 per cent, the other two superconductors in Fig. 4 (click here)b have a much improved
capability. Hf in particular has a reflectivity below 30 per cent in
the ultraviolet and
maintains a value below 55 per cent out to 
. For Nb,
where an
optically-sensitive STJ now exists, the reflectivity is below 50 per cent
out to 700 nm. The reflectivity, which affects not only the
superconducting thin film but, depending on the mode of illumination (front
illumination directly onto the top polycrystalline superconducting film or back
illumination through the substrate onto the epitaxial base film), also the
substrate, and substrate to base film interface, can be reduced
through the use of anti-reflection coatings, albeit at the expense of
bandwidth and with device fabrication constraints.
Such coatings are currently under development.
Figure 4:
The theoretical absorption efficiency, 
a) and surface
reflectivity 
b) in a
100 nm film as a function of photon wavelength for Nb, Al and Hf. The
reflectivity can be further decreased through the use of an anti-reflection
coating at the expense of reducing the wavelength coverage. The
overall quantum efficiency of the device for the case of front
illumination is 
.
Further reduction in the electronic noise should lead to resolutions
approaching the theoretical limits shown in Fig. 1 (click here).
The resulting spectroscopic performance is illustrated in Fig. 5 (click here)
(
) applied to emission line spectra from neutral (I) to
quadruply ionised (V) atoms of oxygen and carbon.
The line
strengths (not corrected for abundances) and wavelengths are taken
from Weast (1981). For the ion species III-V the line
strengths should be considered as qualitative estimates of the relative
strengths between different lines from the same atom. The intensity of
the ionised atoms (II-V) relative to the neutral atom evidently
depends on the excitation conditions. Figure 5 demonstrates that even the
Nb device has a significant spectroscopic capability with many of the key
lines of astrophysical interest being clearly resolvable.
The reflectivity of Al junctions is so high that the overall efficiency would limit considerably its usefulness at these wavelengths. While no devices based solely on Hf yet exist, it has formed the basis for some tunnel junction work (Morohashi et al. 1992). Nb- and Ta-based devices are however more standard, requiring only some improvements to achieve the theoretical limited resolutions of Fig. 1 (click here). From Fig. 5 (click here) however it is clear that an excellent spectroscopic capability would exist for devices based on Hf, with key transitions from the same ion as well as different ionic species being possible to resolve and identify.
Figure 5:
The simulated tunnel-limited response 
(
) from either a
pure Nb based STJ (left two diagrams) or Hf based STJ (right two diagrams)
to the principle emission lines from oxygen (top pair)
and carbon (bottom pair). The strongest lines from neutral
and ionised species are marked. Note the photon scale is different in each case
to highlight the various line features. For all these spectra the bin size was
set at 0.1 nm. No abundances are included in these simulations
Another significant strength of superconducting tunnel junctions is the
inherent speed of the device. The time taken to complete the full conversion
of the photon energy into quasiparticles, including their relaxation down to
an energy equal to the bandgap of the absorbing material, is only a few ns.
The quasiparticle confinement time in the film is complex and
depends also on the superconducting material properties, bias voltage and film
thickness. Based on the work of de Korte (1992) the confinement time in a
single film can be written:
where N(0) is the single spin electronic density of states at the Fermi
energy, e the electron charge, 
is the normal state barrier
resistance, 
is the volume of the film i and 
is the bias voltage applied
across the two
films. Thus for a device designed to provide a single tunnel from a 100 nm
thick film of Nb across a barrier having a typical resistivity
cm
, the average confinement time is
of order 100 ns, scaling with
the mean number of transfers n across the barrier. The parameter 
therefore represents approximately the intrinsic time resolution of the
device.
Thus, in theory, superconducting tunnel junctions could be operated as
photon counting detectors at rates of order 
to 
Hz
depending on the device geometry and n. In practice Lumb et
al. (1995) have shown rates as high as 
kHz can be achieved
without any intrinsic
spectroscopic degradation in the device, being limited by the processing speed
of the analogue electronics and substrate noise associated with the X-ray
photon energy used in the experiment. The electronics however will be the
final limitation to the very important photon count rate capability at optical
wavelengths. These rates will still allow very high speed photometry on
objects to be limited primarily by the collecting aperture of the telescope as
well as the observation of fields containing a objects having a wide dynamic
range in intensity.
Since each photon results in an electrical pulse whose amplitude is directly proportional to the photon energy, and a pulse risetime which is characteristic of a single photoabsorption event in the film, the device has intrinsic background rejection capabilities. Background from cosmic rays can be vetoed either by an upper energy threshold discriminator or through the risetime signature.
For a panoramic imaging STJ camera covering a wide field of view a number of
approaches can be considered. It must however be noted that each STJ detector
is independent and requires its own analogue electronics chain including the
preamplifier. This is a considerable limitation, the solution to which
still needs to be fully addressed. Nevertheless, simple limited close-packed
arrays of 
STJs are already under development by our group
(Rando et al. 1996), while
Perryman et al. (1993) already proposed more
complex schemes in an
attempt to reduce the total number of detectors and associated electronic
chains while maintaining field coverage.
Although the count rates that can be handled in photon counting mode are already expected to be very high, there may be applications (for example, at longer wavelengths, for very bright objects, or for large telescope apertures) where even these limits are exceeded. Under these conditions, it is likely that the relevant junctions can be read out in an analogue mode, in which the steady-state current is measured, rather than the pulse corresponding to an individual photon. In these circumstances, the STJ would simply provide the total energy incident on the junction, integrated over wavelength, with information on the energy per photon lost.