Jonathan A. Cox

Control of coherent information via on-chip photonic–phononic emitter–receivers

Rapid progress in integrated photonics has fostered numerous chip-scale sensing, computing and signal processing technologies. However, many crucial filtering and signal delay operations are difficult to perform with all-optical devices. Unlike photons propagating at luminal speeds, GHz-acoustic phonons moving at slower velocities allow information to be stored, filtered and delayed over comparatively smaller length-scales with remarkable fidelity. Hence, controllable and efficient coupling between coherent photons and phonons enables new signal processing technologies that greatly enhance the performance and potential impact of integrated photonics. Here we demonstrate a mechanism for coherent information processing based on travelling-wave photon–phonon transduction, which achieves a phonon emit-and-receive process between distinct nanophotonic waveguides. Using this device, physics—which supports GHz frequencies—we create wavelength-insensitive radiofrequency photonic filters with frequency selectivity, narrow-linewidth and high power-handling in silicon. More generally, this emit-receive concept is the impetus for enabling new signal processing schemes.

Control of integrated micro-resonator wavelength via balanced homodyne locking

We describe and experimentally demonstrate a method for active control of resonant modulators and filters in an integrated photonics platform. Variations in resonance frequency due to manufacturing processes and thermal fluctuations are corrected by way of balanced homodyne locking. The method is compact, insensitive to intensity fluctuations, minimally disturbs the micro-resonator, and does not require an arbitrary reference to lock. We demonstrate long-term stable locking of an integrated filter to a laser swept over 1.25 THz. In addition, we show locking of a modulator with low bit error rate while the chip temperature is varied from 5 to 60° C.

Integrated control of silicon-photonic micro-resonator wavelength via balanced homodyne locking

We present a new method for active control of optical micro-resonator modulator and filter wavelength that is insensitive to environmental and optical perturbations and readily integrated on-chip. Experimental results demonstrating precise, long-term filter locking are shown.

Resistive memory device requirements for a neural algorithm accelerator.

Resistive memories enable dramatic energy reductions for neural algorithms. We propose a general purpose neural architecture that can accelerate many different algorithms and determine the device properties that will be needed to run backpropagation on the neural architecture. To maintain high accuracy, the read noise standard deviation should be less than 5% of the weight range. The write noise standard deviation should be less than 0.4% of the weight range and up to 300% of a characteristic update (for the datasets tested). Asymmetric nonlinearities in the change in conductance vs pulse cause weight decay and significantly reduce the accuracy, while moderate symmetric nonlinearities do not have an effect. In order to allow for parallel reads and writes the write current should be less than 100 nA as well.

Neurogenesis deep learning: Extending deep networks to accommodate new classes

Neural machine learning methods, such as deep neural networks (DNN), have achieved remarkable success in a number of complex data processing tasks. These methods have arguably had their strongest impact on tasks such as image and audio processing — data processing domains in which humans have long held clear advantages over conventional algorithms. In contrast to biological neural systems, which are capable of learning continuously, deep artificial networks have a limited ability for incorporating new information in an already trained network. As a result, methods for continuous learning are potentially highly impactful in enabling the application of deep networks to dynamic data sets. Here, inspired by the process of adult neurogenesis in the hippocampus, we explore the potential for adding new neurons to deep layers of artificial neural networks in order to facilitate their acquisition of novel information while preserving previously trained data representations. Our results on the MNIST handwritten digit dataset and the NIST SD 19 dataset, which includes lower and upper case letters and digits, demonstrate that neurogenesis is well suited for addressing the stability-plasticity dilemma that has long challenged adaptive machine learning algorithms.

Silicon photonics platform for national security applications

We review Sandias silicon photonics platform for national security applications. Silicon photonics offers the potential for extensive size, weight, power, and cost (SWaP-c) reductions compared to existing III-V or purely electronics circuits. Unlike most silicon photonics foundries in the US and internationally, our silicon photonics is manufactured in a trusted environment at our Microsystems and Engineering Sciences Application (MESA) facility. The Sandia fabrication facility is certified as a trusted foundry and can therefore produce devices and circuits intended for military applications. We will describe a variety of silicon photonics devices and subsystems, including both monolithic and heterogeneous integration of silicon photonics with electronics, that can enable future complex functionality in aerospace systems, principally focusing on communications technology in optical interconnects and optical networking.

Silicon nanophotonics for scalable quantum coherent feedback networks

The emergence of coherent quantum feedback control (CQFC) as a new paradigm for precise manipulation of dynamics of complex quantum systems has led to the development of efficient theoretical modeling and simulation tools and opened avenues for new practical implementations. This work explores the applicability of the integrated silicon photonics platform for implementing scalable CQFC networks. If proven successful, on-chip implementations of these networks would provide scalable and efficient nanophotonic components for autonomous quantum information processing devices and ultra-low-power optical processing systems at telecommunications wavelengths. We analyze the strengths of the silicon photonics platform for CQFC applications and identify the key challenges to both the theoretical formalism and experimental implementations. In particular, we determine specific extensions to the theoretical CQFC framework (which was originally developed with bulk-optics implementations in mind), required to make it fully applicable to modeling of linear and nonlinear integrated optics networks. We also report the results of a preliminary experiment that studied the performance of an in situ controllable silicon nanophotonic network of two coupled cavities and analyze the properties of this device using the CQFC formalism.

A Signal Processing Approach for Cyber Data Classification with Deep Neural Networks

Abstract Recent cyber security events have demonstrated the need for algorithms that adapt to the rapidly evolving threat landscape of complex network systems. In particular, human analysts often fail to identify data exfiltration when it is encrypted or disguised as innocuous data. Signature-based approaches for identifying data types are easily fooled and analysts can only investigate a small fraction of network events. However, neural networks can learn to identify subtle patterns in a suitably chosen input space. To this end, we have developed a signal processing approach for classifying data files which readily adapts to new data formats. We evaluate the performance for three input spaces consisting of the power spectral density, byte probability distribution and sliding-window entropy of the byte sequence in a file. By combining all three, we trained a deep neural network to discriminate amongst nine common data types found on the Internet with 97.4% accuracy.

Electronic interfaces to silicon photonics

We describe the interface circuits to silicon photonics modulators, optical filters, and detectors that will be required to enable silicon photonics micro-ring and micro-disk devices to be integrated in dense wavelength division multiplexing circuitry.

Wavelength control of resonant photonic modulators via balanced homodyne locking

We present a new, robust method for control of resonant modulator wavelength that is integrated with an on-chip balanced detector. Experimental results demonstrate long-term locking with low bit error rate over greater than 55°C range.