Return to CoolChips Home Page

Development of STJ X-Ray Detectors for Synchrotron Science

F. Carter [1], L. Abra [1], I. Avci [1,2], S. Friedrich [3], B. Neuhauser [1]
1. Cryogenic Electronics Group, San Francisco State University; 2. Tubitak, Turkey; 3. Lawrence Livermore National Laboratory






Abstract
Superconducting tunnel junction (STJ) detectors can offer both the high energy resolution and the fast count rate necessary to detect the energies of individual photons in synchrotron-based soft X-ray spectroscopy. We have developed a full wafer fabrication process for Ta-based STJ X-ray detectors and have carried out a set of experiments to control the growth of Ta in the desired bcc phase, to optimize film stress and uniformity, and to determine Tc for Ta-Al bilayers. Fabricating STJs from high-Z Ta offers higher absorption efficiency than the more traditional Nb STJs and thus reduces line-splitting artifacts and increases the energy range for detector operation. We discuss the fabrication challenges that arise from using Ta and present the preliminary characterization of our devices.

Superconducting tunnel junction (STJ) detectors

A STJ detector consists of two superconducting layers separated by a thin insulating layer called a tunnel barrier. The device is voltage biased so that quasiparticle excitations are induced to tunnel across the barrier rather than recombine. This produces a current pulse and the charge collected is proportional to the energy of the absorbed photon.


Fig. 1 – Adkins representation of STJ. When an incident photon breaks a Cooper pair, it excites quasiparticles (a), which then relax through phonon interactions (b), and diffuse into the trapping layer (c and d) where they tunnel across either directly (e) or through a quasiparticle recycling process known as back tunneling (f ) and then recombine in the base electrode. [O. Drury, PhD Thesis UC Davis, 2007].

Why tantalum?
Current state-of-the art STJs use niobium (Z = 41) with an aluminum trap to increase tunneling probability. Unfortunately, these detectors have an energy threshold above which the top Nb electrode does not completely absorb all the incoming photons; the remainder are absorbed in the bottom electrode. Since the Al traps are not perfectly symmetric, the resulting current pulse creates a line-splitting artifact in the resulting spectrum at an apparent energy roughly 5% above that of the actual photon (Fig. 2).

Tantalum (Z = 73) provides more x-ray absorption at higher energies than niobium and is also more opaque in the low energies. Additionally, the energy gap in a superconducting metal is proportional to the critical temperature. Since bulk Ta has a Tc that is roughly half of the value for bulk Nb, the energy required to break a Cooper pair in Ta will be about half that required for a Nb device; this will ultimately mean an increase in intrinsic energy resolution of nearly a factor of square root of 2.

Fig. 2 – Fluorescence spectrum taken with Nb STJs. Notice the line splitting due to the incomplete absorption in the top electrode [S. Friedrich et. al., J. Electron Spectroscopy, 101, 891-896, (1999)].

Body centered cubic (bcc) vs. tetragonal Ta

Tantalum offers many fabrication challenges over niobium when it comes to constructing STJs. The most vexing is that Ta comes in two phases, and the bcc phase that is useful for STJs is not the natural tetragonal phase for thin-film Ta. Coaxing Ta into the correct phase may be accomplished a number of ways. Growing Ta on a heated sapphire substrate results in the correct phase, but this method is not possible for depositing the counter electrode or wiring layer. A thin (5 nm) niobium seed layer has been shown to nucleate the desired phase in Ta. We have chosen to utilize this method of nucleating Ta as we have considerable experience with Nb thin films. Figure 3 shows experimental data confirming our ability to reliably construct thin films of bcc-Ta.

Fig. 3 – Critical temperature plot for two phases of Ta (left). XRD plot showing phases of Ta in test films (right).

Fabrication Process

We have developed a full-wafer fabrication process for constructing Ta STJs. The tunneling structure is created without removing the wafer from the process chamber which means the process is well controlled and nearly immune to contamination. The mask set produces four single, square, STJ pixels of different sizes; 10, 20, 40, and 80 microns on a side. Additionally there are a number of test structures for measuring Tc, pinhole density, wiring layer interface, and field oxide step coverage.



Fig. 4 – Cross sectional (left) and birds-eye (right) view of fabrication process. Nb layers have been omitted for clarity, and dimensions are not to scale.

Starting with a bare SiOx on Si wafer, the Nb/Ta/Al base electrode is DC sputtered. The Al surface is oxidized in situ to form the tunnel barrier, and then the Al/Nb/Ta counter electrode is sputtered on. The junction region is then defined and the junction walls are anodized. Field oxide is sputtered and windows are etched over the junction. Then the Nb/Ta wiring layer is sputtered and etched. Finally, vias are etched down to the base electrode.

Photos of Ta STJ pixels


Fig. 5 – Microscope photos of: 10 μm2 and 20 μm2 junctions (left), and 80 μm2 and 40 μm2 junctions (right).

Preliminary DC characteristics

The devices are mounted in an adiabatic demagnetization fridge which cools them down to about 70 mK. A custom IV curve tracer is used to current-bias the STJ and the voltages are read out with a standard 4-terminal measurement and a Tektronix digital oscilloscope.

Initial results from the Ta STJ indicate a large amount of sub-gap current (Fig. 6-left) and an unexpectedly low Josephson current. However, the superconducting energy gap (2Δ/e) was measured at 0.7 mV. This is consistent with our expectations for bcc-Ta. As a test of our fabrication process, a Nb STJ was constructed using the same mask set. The IV curve from the Nb STJ (Fig. 6-right) displays a much lower sub-gap current, but also exhibits a troubling lack of Josephson current. One possibility is that the interface between the wiring layer and counter electrode is not pure and is acting as a second tunnel junction in series with the actual device. This could account for the lack of Josephson current and could also account for some of the apparent leakage. Figure 7-right shows a close up view of the pair tunneling region for a 10 micron bcc-Ta STJ. The cusp is a possible indicator of S1-I-S2 tunneling, which is likely, as there are three superconducting metals and an unknown number of superconducting oxides and alloys present in the STJ.



Fig. 6 – IV curve from first bcc-Ta STJ (left). IV curve from Nb STJ fabricated with same mask set (right). Both curves were taken without any magnetic field suppression of the Josephson currents.


Fig. 7 – IV curve from first bcc-Ta STJ (left). Close up of same STJ (right).

Acknowledgements

Thanks to: O. Drury and M. Velazquez for wirebonding and good advice, J. Hamilton for last minute dicing, D. Danielson for beautiful fab drawings, P. Verdone, A. Der, A. Kelly, and F. Lipschultz for keeping the lab running, and S. Burt and F. Ponce for lending a hand when needed.

Return to CoolChips Home Page

Updated 11 May 2010