Transition Edge Sensors

My research focuses on using Terahertz integrated circuit technology to enhance the sensitivity and applicability of detector imaging arrays for submillimeter astrophysics. The industry standard detector for the past decade has been the Transition Edge Sensor (TES) bolometer. These are thin films of metal sputtered onto a released structure in weak thermal contact with the rest of the instrument only through mildly conducting legs (conducting only a few pW/K in our case). We voltage bias our sensors into their transition between normal and supercondcuting states and read their current through SQUID ammeters. These detectors can be monolithically fabricated, the readout can be multiplexed through the SQUIDs, and elecrothermal feedback linearizes their response in a way that leads to simple absolute calibration.
Bolometer thermal circuit: As Popt increases, resistance would tend to rise. However, this response chokes off the bias power Pb, and the resulting electrothermal feedback actually holds these incident powers constant and equal to the out-going power PG. This feedback loop pins the detector temperature and resistance at a constant value in the transition and only breaks down if the incident power heats the device out of its transition or allows it to cool into a latched superconducting state.

These detectors can be made such that the dominant source of noise is that from the incident photons; thus the only practical way to increase sensitivity is to simultaneously receive more electromagnetic modes. One common approach is to receive more spatial modes by packing more pixels into the focal plane of a camera.

Antenna coupled bolometers

Phased Array Antenna coupling

Traditionally, bolometers have been used with horns to match the detectors' optical response to the camera/telescope optics. At Caltech, we have developed an entirely planar equivalent with an integrated, printable, phased array antenna. Whereas technology from circa 2005 allowed for 50 pixels in the progenitor camera BICEP-1, we recently deployed over 1500 dual-polarized pixels between the Keck-array and BICEP-2 cameras, providing 8.5uK √s sensitivity. In November of 2014, we deployed an additional 2500 detectors to the field in the BICEP-3 camera, which will expand our coverage from just 150GHz to 100GHz so we can discriminate CMB photons from galactic foregrounds. The balloon-borne SPIDER flew in January 2015 with a comparable number of pixels as Keck, split between 100GHz and 150GHz. We have used the BICEP-2 and Keck cameras to constrain r<0.07 at 95% confidence. The key behind this rapid scale-up and scientific success has been to use small aperture cameras with just enough resolution to see the 2-degree primordial B-mode peak (an aperture of 26cm at 150GHz), thus making the design scalable to multiple cameras.
Planar phased array antenna. From left to right, picture of device, measured beams (gaussian and tophat fed).

Sinuous Antenna coupling and photometer circuits

At Berkeley, we developed detectors for use in off-axis Gregorian telescopes like Polarbear and SPT. Those telescopes’ larger apertures (2.5m and 10m) provide the resolution to see gravitational lensing effects in the B-mode anisotropies, but by virtue of their size, do not scale as naturally as the small aperture cameras.

Another way to absorb more modes but with such a limited focal plane size is to design each pixel to receive and separate multiple spectral modes. My doctoral thesis developed this technology using a sinuous antenna as the absorber with a contacting lens to increase gain:


You can download my thesis here.

The self similar sinuous antenna is dual-polarized with at least two octaves of continuous bandwidth. We have made two and three channel versions with spectra and beams like these:

Diplexed Pixel. From left to right, picture of device, measured spectra, measured beams.

South Pole Telescope is currently upgrading its camera (SPT-3G) to use these detectors and hopes to be able to measure the sum of neutrino masses Σ mν with a precision of 60meV, which should resolve the mass hierarchy problem between normal and inverted hierarchies.

The figure above shows a sinuous antenna attached to a photometric channelizer that is, like the antenna, self-similar (log-periodic). We expect this technology to find application in a satellite experiment where we are digging deep into galactic foregrounds with multi-component foreground models:
Log periodic channelizer. From left to right, photo of device, effective circuit, and measured response (normalized with load curves)
We may use a small version of a spectrometer in a second flight of SPIDER and this style of sampling might also allow a different experiment to measure Kinetic Sunyaev-Zeldovich to characterize reionization.

Imaging arrays of Spectrometers for reionization studies

Mapping bubble sizes during reionization requires the high sensitivity of an imaging array with high spectral resolving power to precisely measure redshift. We can generalize the above channelizer to circuits with higher Q that exploit weak coupling of microstrip lines.
TES spectrometer. Top: from left to right, layout of the device, FTS measurements of prototype. Bottom: photograph of fabricated device with spectrometer at left and bank of bolometers at right

Above is a prototype with 30 channels with a designed Q~400 and an anticipated noise of 3aW/sqrt(Hz). We ultimately plan to build a camera with 100 pixels and 100 channels each. We have also coupled similar spectrometers to Kinetic Inductance Detectors (KIDs) in the SuperSpec project.

Thermal Kinetic Inductance Detectors (TKIDs)


We are developing TKIDs as a remedy to the expenses associated with SQUID-based multiplexing for TESes as well as the sensitivity challenges experienced with traditional KIDs. KIDs use a common sub-device to absorb optical photons and to act as the sensor in the high-Q circuit, and it can be difficult to simultaneously optimize for both optical efficiency and low noise performance. TKIDs are technically a bolometer where the absorber and resonator are formed in different films, much like TES bolometers. We are developing these for BICEP array's high frequency cameras where detector counts of 7500+ make SQUID multiplexing infeasible.


We are currently studying noise with different recipes and realization. However, these sensors are naturally drop-in-compatible with the antenna-coupled architectures above and can be interrogated with KID-style rfMUX readout, as seen in this pedagogic schematic:


Commercial Readout for laboratory characterization

We develop several low-temperature resonator based detectors at JPL, including TKIDs. To service this R&D, we are also developing a highly flexible readout platform for multi-tone device characterization. The system uses a thin FPGA layer with commercially maintained software for AD/DA between the cryostat/detectors and a computer's PCI bus. The computer then processes fourier transforms of tones in a GPU. Such a scheme avoids the detailed HDL/FPGA programming needed for ROACH systems and instead allows readout algorithms to be customized with commonly used C++ and CUDA. The current lab system uses NI/Ettus' Universal Software Radio Peripheral (USRP) Platform for AD/DA, which provides ~150MHz bandwidth. We are also developing a more sophisticated 1GHz bandwidth version with Xilinx and TI evaluation boards for a potentially deployment-worthy system.