MESA+ Institute for Nanotechnology

This thesis describes the development of low-energy gap superconducting tunnel junctions (STJs) for use as photon detectors, with as a main goal the improvement of the energy resolution in both the optical and the x-ray energy domain.

A new model for the photon detection process with STJs is presented, which includes the full energy dependence of all the quasiparticle processes occurring in the junctions. This model allows for the calculation of the time- and energy-dependent quasiparticle distribution from the moment of generation of the quasiparticles by the photon absorption process until the end of the current pulse, when all the quasiparticles have disappeared. The exact knowledge of the quasiparticle energy distribution in the junctions is of increasing importance for the lower energy gap junctions, as the quasiparticle relaxation rate is approximately proportional to the cube of the energy gap of the superconductor. As a consequence, energy down-conversion of quasiparticles in lower-TC superconductors becomes much slower and the bias energy gained by the quasiparticles due to successive tunnel and back-tunnel events leads to a very broad energy distribution of the non-equilibrium quasiparticle population. Two effects related to the broad quasiparticle energy distribution, which cannot be explained with an energy-independent Rothwarf-Taylor approach, are on one hand the proportion of charge lost due to cancellation tunnel events and on the other hand quasiparticle multiplication. When a quasiparticle has energy above the bias energy level, it can undergo a tunnel event against the bias and annihilate a charge from the current pulse. In this way a certain percentage of the charge output is lost, which can be as large as 80 % for lower energy gap junctions with fast tunnelling. When a quasiparticle has energy larger than 3Dg, it will release a phonon of energy larger than 2Dg, when relaxing down to the gap energy Dg. This released phonon can in turn break a Cooper pair and create two additional quasiparticles. In this way the quasiparticle population increases drastically, which has a strong effect on the measured tunnel currents as well. This mechanism, called quasiparticle multiplication, is typical for lower energy gap, low loss junctions.

Illustrations of the energy-dependent kinetic equations model simulating the response of Ta- and Nb-based junctions show that the quasiparticle energy distribution converges quickly to a “quasi-equilibrium” distribution. The distribution is called to be in quasi-equilibrium in the sense that the normalised distribution is invariable and only the total number of quasiparticles diminishes because of the different quasiparticle loss channels. This quasi-equilibrium distribution shows a step-like structure, with local maxima occurring at multiples of the bias energy because of the energy gain due to subsequent tunnel and back-tunnel events. Even in the relatively larger energy gap junctions based on Ta and Nb the average quasiparticle energy lies well above the energy gap of the superconductor and the main condition of the Rothwarf-Taylor approach is therefore not justified.

The fabrication processes for three different types of STJs are presented in this thesis, based on Vanadium-Aluminium, Aluminium, and Molybdenum-Aluminium electrodes respectively. The V-based junctions were intended as a process route development vehicle for the fabrication of the lower TC junctions based on Al and Mo. The reason for this is the possibility to operate the V-based junctions at 300 mK in a 3He sorption cooler with a very fast turn-around time. The Al and Mo based junctions on the other hand have to be operated in an Adiabatic Demagnetisation Refrigerator at temperatures below 100 mK, which has a much longer turn-around time. The progress made with V-based junctions, while varying certain parameters of the processing steps common to all three junctions, can then be directly transferred to the lower energy gap junctions.

The fabricated V-Al based junctions are of good quality with a normal resistivity rnn approximately equal to 1.2 mW cm2 and a dynamical resistivity rd in the bias range approximately equal to 1.1 W cm2, which corresponds to a quality factor Q = rd/rnn of ~106. The Josephson current suppression pattern is very regular, indicating the good homogeneity of the insulating barrier separating the two electrodes of the junctions. 6 keV photon detection experiments could be performed with V-based junctions having side lengths varying between 7 and 30 mm. The responsivity was shown to be very low of the order of 600 e-/eV and independent of the device size of the junctions. Simulations with the energy dependent kinetic equations model show that the number of localised trapping states is very large, about one and two orders of magnitude larger than in similar Nb and Ta based junctions respectively. This large number of quasiparticle trapping states is believed to be related to the strong reactivity of V with oxygen and the metallic nature of some of the oxides forming small islands in the superconductor with a locally suppressed energy gap. The 6 keV energy resolution of the junctions is 80 eV full width at half maximum (FWHM) for the smallest device sizes and increases to approximately 900 eV FWHM for the 30 mm side length devices. The reason for this poor energy resolution is variation of the responsivity as a function of absorption position over the area of the detector. The work on V-based junctions was discontinued, because they are not well suited as photon detectors. Nevertheless, the main goal for these junctions was achieved, as they were mainly intended as a process route development vehicle for the lower energy gap junctions based on Al and Mo.

High quality single pixel Al STJs were fabricated with side lengths varying between 10 and 70 mm. The normal resistivity of these junctions is ~7 mW cm2 and the dynamical resistance in the bias domain is 1.9 W cm2, corresponding to a quality factor of approximately 2.7 105. The Josephson current suppression pattern is very regular with a pronounced minimum for an applied magnetic field equal to 50 Gauss, allowing the successful suppression of the zero bias Josephson currents. Optical photon detection experiments could be performed with the Al based junctions. The responsivity of the devices is very large of the order of 105 e-/eV and proportional to the area of the detector. Responsivity and pulse decay time of the Al detectors show a strong photon energy non-linearity in the optical domain, indicating the small number of localised quasiparticle trapping states. Simulations with the kinetic equation model reveal that the number of traps in the Al junctions is only of the order of 7000 states, which is a factor 3 and 30 lower than in comparable Ta-Al and Nb-Al junctions respectively. The number of trapping states does not depend on the device size, which is a very strong indication that the trapping states are located in the Nb that forms the contacts to the top and base electrodes. The bias voltage dependence of the responsivity of the junctions shows an increase with increasing bias voltage, the effect being stronger for the larger junctions with the longer quasiparticle loss times. This dependency is related to the broad quasiparticle energy distribution. The lower cancellation currents and stronger quasiparticle multiplication at higher bias voltage result in a higher responsivity. These effects could be successfully simulated with the energy dependent model. The energy resolution of the Al junctions includes a spatial broadening contribution, probably because all the loss sites are located at the positions of contact of the Nb leads to the top and base electrodes. This spatial broadening contribution limits the intrinsic resolving power l/Dl for 500 nm optical photons to approximately 17, well below the theoretical value of approximately 30. Nevertheless, the capabilities of Al STJs as optical photon detectors were demonstrated. Further work on Al based junctions will include the replacement of the Nb leads by Ta, which should homogenise the responsivity over the area of the detector and as a consequence increase the resolving power of the detector. In addition, coupling of the Al junctions to x-ray absorbers is planned in order to take advantage of the good theoretical energy resolution of Al STJs in the x-ray energy domain as well.

For the fabricated Mo-Al based junctions there still exists a problem with the edge structure created by the base etch step. The top Mo film is etched 600 nm further than the base Mo film, resulting in a step like structure at the edges. The Al film in between the two Mo layers does not form a vertical edge structure, which damages the aluminium oxide insulating layer. As a consequence the junctions show large perimeter related leakage currents of 1.25 mA per mm of perimeter length. No photon detection experiments could be performed with these junctions as the large leakage currents prevent stable biasing of the STJ. Further work on Mo based junctions, in particular on the base etch procedure, will have to be performed in order to reduce the leakage currents in these devices.