Prof. Niranjan Thatte

Abstract:

Microwave Kinetic Inductance Detector (MKID) devices offer inherent spectral resolution, simultaneous read out of thousands of pixels, and photon-limited sensitivity at optical wavelengths. Before taking observations the readout power and frequency of each pixel must be individually tuned, and if the equilibrium state of the pixels change, then the readout must be retuned. This process has previously been performed through manual inspection, and typically takes one hour per 500 resonators (20 h for a ten-kilo-pixel array). We present an algorithm based on a deep convolution neural network (CNN) architecture to determine the optimal bias power for each resonator. The bias point classifications from this CNN model, and those from alternative automated methods, are compared to those from human decisions, and the accuracy of each method is assessed. On a test feed-line dataset, the CNN achieves an accuracy of 90% within 1 dB of the designated optimal value, which is equivalent accuracy to a randomly selected human operator, and superior to the highest scoring alternative automated method by 10%. On a full ten-kilopixel array, the CNN performs the characterization in a matter of minutes — paving the way for future mega-pixel MKID arrays.

Abstract:

We present detailed simulations of integral field spectroscopic observations of a supernova in a host galaxy at z ∼ 3, as observed by the HARMONI spectrograph on the Extremely Large Telescope, asssisted by laser tomographic adaptive optics. The goal of the simulations, using the HSIM simulation tool, is to determine whether HARMONI can discern the supernova Type from spectral features in the supernova spectrum. We find that in a 3 hour observation, covering the near-infrared H and K bands, at a spectral resolving power of ∼3000, and using the 20×20 mas spaxel scale, we can classify supernova Type Ia and their redshift robustly up to 80 days past maximum light (20 days in the supernova rest frame). We show that HARMONI will provide spectra at z ∼ 3 that are of comparable (or better) quality to the best spectra we can currently obtain at z ∼ 1, thus allowing studies of cosmic expansion rates to be pushed to substantially higher redshifts.

Abstract:

As high pixel count Microwave Kinetic Inductance Detector (MKID) arrays become widely adopted, there is a growing demand for automated device readout calibration. These calibrations include ascertaining the optimal driving power for best pixel sensitivity, which, because of large variations in MKID behavior, is typically performed by manual inspection. This process takes roughly 1 hour per 1000 MKIDs, making the manual characterization of ten-kilopixel scale arrays unfeasible. We propose the concept of using a machine-learning algorithm, based on a convolution neural network (CNN) architecture, which should reliably tune ten-kilopixel scale MKID arrays on the order of several minutes.