High-Resolution Gamma-Ray Detection Using Phonon-Mediated Detection

a. Proof of principal device As seen in Fig. 1, we started with a single layer TiNx cm2 proof-of-concept device. TiNx proved to be problematic material both from a fabrication standpoint (there was large variation in Tc over the device) as well as from a phonon absorption standpoint. By measuring the temperature dependent frequency and dissipation response of our device, and comparing to the response from athermal phonon signals induced by particle interactions, we could calculate the overall effencency with which our device transformed phonons into quasiparticles. For our TiNx devices (Tc = 0.5 K) our collection efficiency was less than 1%. We were able to address both problems to some extent by switching to Al films. To make up for the loss in kinetic inductance fraction we moved from 75 nm films to 25 nm films which we found to be optimal signal to noise. The phonon collection efficiency for these films was 7%. This gain in sensitivity allowed us to read out each MKID simultaneously, and preform a position correction as seen in Fig.2. At this point we had recovered an energy resolution within a factor of 2 of our baseline sensitivity, and decided to move ahead with the full focal plane form factor. b. Aluminum three inch device The basic design of the three inch device was initially derived from the SuperCDMS iZIP. The iZIPs are massive Ge particle detectors that are sensitive to both an ionization signal as well as an athermal phonon signal for every particle interaction. Our initial design focused on keeping the ionization readout retivly the same while replacing the highly complex quasiparticle-trap-assisted electro-thermal-feedback transition-edge sensor (QET) phonon detecting elements with our single layer MKID. This set our constraints on metal coverage (5%) MKID size (1mm2) and chargeline to feedline pitch (800 microns). The layout can be seen in Fig.3. The new “folded” MKID was designed to fit into the alloted space while reducing interMKID coupling (neighboring inductor currents should cancel). Coupling was extensively simulated an found to be negligible between MKIDs, but was present in MKIDs adjacent to vertical feedline sections. This affect could be modeled with a simple quadric and removed as seen in Fig.4. Upon testing we found these initial devices to be plagued with very large ripples (“feedline modes”) in the baseline of the transmission through the device. These baseline reflections were similar enough in size to our MKID resonant notches to make standard fitting methods intractable. This feedline mode can be seen in Fig.5. Using a cross correlation technique, we were able to measure the average temperature dependent frequency response of our device and fit it to Mattis-Bardeen theory as can be seen in Fig. 5. This resulted in a kinetic inductance fraction of 0.23 and a gap of 0.178 eV, which are reasonable for a thin film Al device. This procedure is sketched in Fig 6. c. Unit test devices As our Al devices were unusable in their current form, we decided to construct a series of unit test devices to track down the source of our undesirable feedline behavior. The devices tested a variety of things: the basic coplainer waveguide (CPW) to coplainer stripline (CPS) transition (via a device called a balun), the feedline by itself, the feedline and ionization readouts (no MKIDs), and the feedline with only the MKIDs (no ionization readouts). These can be seen in Fig. 7. These devices were tested using Nb films in a 4K 4He fridge. This saved us the time and expense of using our dilution fridge to test the RF properties of our devices. The problem was found to be that the feedline itself was somewhat difficult to fabricate reliably. As my skills in MDL improved, I was able to produce a number of serviceable specimens, which had flat baselines at ~ -10dB S21 over large enough spans of frequency space to be useful. d. Bi-metal devices Our experience with fabricating Nb feedlines had been generally positive: the metal was much more physically robust than Al, and it could go through RF screening in a simple 4K fridge. It's higher Tc also prevented it from being an active source of phonon loss, it's gap is simply too large. We believed that the benefits were worth the extra fabrication steps of a two layer device, the first one of which can be seen in Fig 9. Using just the feedline portion of this mask, we were again able to make a number of devices that had acceptable transmission properties. However, the secondary step of adding MKIDs always rendered the device unusable. We now believe this is due to the ease with which our CPS feeline can couple to, and loose energy to, our holder. The majority of feedlines used on large MKID arrays are of the CPW design. The internal electric fields of CPW lines are mutually opposed and cancel on length scales larger than the feedline gap (usually microns). This is usually sufficient to isolated the line. Field lines in CPS lines, however, only cancel after a half wavelength has been traveled town the line (centimeters in our case). This could easily be a large source of our transmission trouble. At this point we decided to go back to the drawing board and simplify our focal plane design. Instead of a CPS, we used a CPW feedline. We jettisoned the ionization readouts for a phonon-only design. This allowed us to lay our structures out in a Manhattan geometry, easing both simulation and error detection during fabrication. We also increased our minimum feature size (our feedline gap) from 3 to 10 microns, again for ease of error detection. We also re-designed the MKIDs themselves. The mask (as well as a MKID detail) can be seen in figure 10. This device had been simulated, but has yet to be fabricated.