next previous
Up: Superconducting tunnel junctions

3. The superconducting tunnel junction

3.1. Principles of operation

One of a number of ways of detecting the excess quasiparticles produced as a result of the photoabsorption process is by ensuring that they tunnel from one thin superconducting film tex2html_wrap_inline1141 in which they are created into another tex2html_wrap_inline1143 through a thin insulator (I). To maximise the tunnel probability this insulating barrier must be very thin, of order 1 nm (i.e. only a few atomic layers). This superconductor-insulator-superconductor (or SIS) structure is in essence a superconducting tunnel junction (STJ) or Josephson junction (Giaever 1960; Josephson 1962). A small magnetic field is applied parallel to the junction barrier, as shown in Fig. 2 (click here), so as to suppress the Josephson current which results from the tunnelling of Cooper pairs at zero bias voltage. Applying a bias voltage tex2html_wrap_inline1147 ensures that the only allowed tunnel processes involve the transfer of quasiparticles from one film to the other. The flow of quasiparticles in a time t produces a measurable excess electric current, with the amplitude of the current pulse being directly proportional to the incident photon energy.

  figure286
Figure 2: A schematic of a typical symmetrical tunnel junction deposited on a sapphire substrate together with its orientation in a parallel magnetic field so as to suppress the Josephson current. Such a device was used by Peacock et al. (1996 a,b) to demonstrate single photon counting from tex2html_wrap_inline1151. Back or front illumination is possible

3.2. Multiple tunnelling and tunnel noise

For the case where tex2html_wrap_inline1155 two tunnel processes exist (Gray 1978). In the first process an electron tunnels from film tex2html_wrap_inline1157 to film tex2html_wrap_inline1159 (the direction depends simply on the polarity of tex2html_wrap_inline1161) and effectively a quasiparticle is exchanged, while in the second process an electron tunnels again from film tex2html_wrap_inline1163 to tex2html_wrap_inline1165 while a quasiparticle is exchanged between films tex2html_wrap_inline1167 and tex2html_wrap_inline1169. The combination of these two processes in series leads to an effect known as multiple tunnelling, which can be viewed as equivalent to an amplification of the initial charge tex2html_wrap_inline1171 created in the superconducting film i. If each quasiparticle originally created in the film is transferred across the barrier on average of n times then, in the absence of loss processes, the average total electrical charge detected would be tex2html_wrap_inline1177. In this time-dependent process, recombination, diffusion losses and quasiparticle trapping will all reduce tex2html_wrap_inline1179 (Booth 1987; Verhoeve et al. 1996).

This charge amplification will however degrade the Fano-limited resolution of the STJ by adding in quadrature the variance on this tunnel process, a contribution referred to as the tunnel noise (Goldie et al. 1994). On the assumption of a perfectly symmetrical junction (with equal probabilities of tunnelling from tex2html_wrap_inline1181 to tex2html_wrap_inline1183 and from tex2html_wrap_inline1185 to tex2html_wrap_inline1187) the resolution is given by:
equation301
This limiting resolution for a perfectly symmetrical STJ is also shown in Fig. 1 (click here) for the case n=5. This choice of n is based on the experimental determination by Peacock et al. (1996a) for a symmetrical Nb-based device. Figure 1 (click here) shows that a resolution of tex2html_wrap_inline1193 nm and tex2html_wrap_inline1195 nm is achievable for simple Nb and Al based devices respectively when illuminated by photons of wavelength tex2html_wrap_inline1197. The degradation due to an equivalent amplification through multiple tunnelling is not a basic limitation, since the need for such amplification is dependent on the signal-to-noise ratio and hence on the noise of the readout electronics. Provided the initial charge is sufficiently large compared to the electronic noise then such an amplification is not essential, and the contribution of the tunnel noise to the overall resolution can be reduced. Naturally, this requirement affects the basic design of the STJ and, depending on the materials used, its operating temperature. For these reasons we show the limiting resolution in Fig. 1 (click here) as a band ranging from tex2html_wrap_inline1199 to tex2html_wrap_inline1201, for the case tex2html_wrap_inline1203.

Effects of the quasiparticle lifetime and the role of various recombination mechanisms were considered by Perryman, et al. (1993). For a typical tex2html_wrap_inline1205 device with a film thickness of 100 nm, i.e. a film volume tex2html_wrap_inline1207, the number of thermal quasiparticles can be written:
equation321
where N(0) is the single spin electronic density of states at the Fermi energy. From Eq. (3) a photon of wavelength 500 nm will produce an initial number of quasiparticles tex2html_wrap_inline1211 of order tex2html_wrap_inline1213, tex2html_wrap_inline1215 and tex2html_wrap_inline1217 for Nb, Al and Hf respectively. Combining Eqs. (3) and (6):
equation327
It is reasonable to require that the thermal population is at least an order of magnitude lower than tex2html_wrap_inline1219. This means that, in the absence of additional barrier leakage mechanisms (thermal regime), the operating temperature T has to be less than tex2html_wrap_inline1223, tex2html_wrap_inline1225 and tex2html_wrap_inline1227 for Nb, Al and Hf respectively (i.e. tex2html_wrap_inline1229, 170 or 30 mK) to detect 500 nm photons. In the event of multiple tunnelling, these requirements may be somewhat modified.


next previous
Up: Superconducting tunnel junctions

Copyright by the European Southern Observatory (ESO)
web@ed-phys.fr