Photonic quantum programmable gate arrays

Watch my talk at CLEO 2020 on the results from our paper here!

Photonic systems have many unique advantages for quantum information processing, but deterministic multi-photon gates are difficult to implement, and complex quantum circuits can be prohibitively large to do with free-space optics since processing is done along the photon path. In this paper, we present an architecture for a photonic quantum programmable gate array (QPGA) which can be dynamically programmed to implement any quantum operation, in principle deterministically and with perfect fidelity. Our photonic integrated circuit architecture consists of a lattice of beamsplitters and phase shifters, which perform rotations on path-encoded photonic qubits, and embedded quantum emitters, which use a two-photon scattering process to implement two-qubit controlled gates deterministically. We show how to exactly prepare arbitrary quantum states and operators on the device, and we apply machine learning techniques to automatically implement highly compact approximations to important quantum circuits. Our design is the first to our knowledge to extend programmable integrated optics to the quantum domain in a manner which is both deterministic and spatially efficient, and ongoing advancements in nanophotonic processors and strongly coupled quantum emitters may allow for feasible near future implementation of our design.

Nanophotonic neural networks

Optical systems are a promising hardware platform for fast and energy-efficient machine learning. A neural network implemented with nanophotonic components requires virtually no energy to operate, produces no waste heat, and can process data at feedthrough rates hundreds of times faster than electronic systems. I developed neuroptica, a popular photonic neural network simulation library. I co-authored a paper addressing architectural and training difficulties with this class of optical devices, and helped to design a new type of physical electro-optic activation function for use in these networks, which has since been experimentally realized.

### Computing photon scattering in arbitrary quantum optical systems

I wrote the scattering module for QuTiP (the Quantum Toolbox in Python). This module uses recent advances in quantum optical theory to solve the generalized problem of computing the scattered state of a driven arbitrary quantum system. I made modifications to the theoretical framework that reduce the space complexity of computing the evolution of a quantum system emitting a fixed number of photons from exponential to polynomial complexity, making the calculations that my module performs tractable. The scattering module was a major feature in the v4.3 release of QuTiP. I also wrote an instructional notebook to demonstrate the capabilities of my module, which is posted on the QuTiP website.

### A distributed simulation framework for quantum networks and channels

SQUANCH (Simulator for Quantum Networks and CHannels) is an open-source Python framework for creating performant and parallelized simulations of distributed quantum information processing. Although it can be used as a general-purpose quantum computing simulation library, SQUANCH is designed specifically for simulating quantum networks, acting as a sort of "quantum playground" to test ideas for quantum networking protocols. The package includes flexible modules that allow you to easily design and simulate complex multi-party quantum networks, extensible classes for implementing quantum and classical error models for testing error correction protocols, and a multi-threaded framework for manipulating quantum information in a performant manner. The whitepaper introducing the framework is available on arXiv.

### Functional testing for the LSST camera system

### Vertex reconstruction with precision timing data

The Phase-II upgrades to the Large Hadron Collider will introduce a variety of new measurement devices to the CMS, including the High-Granularity Calorimeter (HGCAL) and will increase the beam luminosity by an order of magnitude. However, these improvements to luminosity will increase pileup - when multiple collisions happen sufficiently close enough in space and time such that it is difficult to associate the detected particles with their original collision location ("vertex"). A high pileup environment poses particular challenges to distinguishing the two major production mechanisms for the Higgs boson, gluon fusion and vector boson fusion, reducing the accuracy of the current boosted decision tree vertex reconstruction algorithms from >90% to around 30%. Separating the two production mechanisms is of crucial importance for precisely understanding electro-weak symmetry breaking.

Using high precision timing measurements on the order of 10 picoseconds from simulated events in the HGCAL, we designed a vertex reconstruction algorithm that requires only the spatiotemporal arrival coordinates of the detected particles to reconstruct the interaction vertex of a collision with a median resolution of 235 micrometers, approximately 162 times better than the original proof-of-concept algorithm from 2012. To do this, we implemented a set of filters to detect poorly-reconstructed events and designed a helper algorithm capable of inferring likely interaction vertices given the structural "pointing" data of only a single cluster. We authored a CMS analysis note (mirror) detailing the development and results of our algorithm.

### Evolution of the length of day