X Development LLC

Jesse Lu

Martin F Schubert

Brian John Adolf

Michael Jason Grundmann

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for designing a multimodal photonic component. In one aspect, a method includes defining a loss function within a simulation space including multiple voxels and encompassing features of the multimodal photonic component. The loss function corresponds to a target output mode profile for an input mode profile, where the target output mode profile includes a relationship between a set of operating conditions and one or more supported modes of the multimodal photonic component at a particular operative wavelength. The initial structure is defined for one or more features, where at least some of the voxels corresponding to features have a dimension smaller than a smallest operative wavelength of the multimodal photonic component, and values for structural parameters for the features are determined so that a loss according to the loss function is within a threshold loss.

Multimodal photonic components

In some embodiments, a computer-implemented method for creating a fabricable segmented design for a physical device is provided. A computing system receives a design specification. The computing system generates a proposed segmented design based on the design specification. The computing system determines two or more loss values based on the proposed segmented design. The computing system combines the two or more loss values to create a combined loss value. The computing system creates an updated design specification using the combined loss value. The generating, determining, combining, and creating actions are repeated until a fabricable segmented design is generated.

Techniques for fusing multiple loss functions in an inverse design process

Methods and systems for designing a photonic computational architecture including a plurality of optical components. At least some of the methods include: defining a loss function within a simulation space composed of a plurality of voxels, the simulation space encompassing the plurality of optical components; defining an initial structure for the photonic computational architecture in the simulation space, at least some of the voxels corresponding to each of the plurality of optical components and having a dimension smaller than an operative wavelength of the computational architecture; determining values for at least one structural parameter and/or at least one functional parameter for each of the plurality of optical components using a numerical solver to solve Maxwell's equations; and defining a final structure of the photonic computational architecture based on the values for the one or more structural and/or functional parameters.

Systems and methods for designing photonic computational architectures

A computer-implemented method for modeling fabrication constraints of a fabrication process is described. The method includes receiving training data including pre-fabrication structures and post-fabrication, training a fabrication constraint model by optimizing parameters of the fabrication constraint model based on the training data to model the fabrication constraints of the fabrication process, receiving an input design corresponding to a physical device, and generating a fabricability metric of the input design via the fabrication constraint model. The fabricability metric is related to a probabilistic certainty that the input design is fabricable by the fabrication process determined by the fabrication constraint model.

Learned fabrication constraints for optimizing physical devices

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for correcting finite floating-point numerical simulation and optimization. Defining a loss function within a simulation space composed of a plurality of voxels each having an initial degree of freedom, the simulation space encompassing one or more interfaces of the component; defining an initial structure for the one or more interfaces in the simulation space; calculating, using a computer system with a finite floating-point precision, values for an electromagnetic field at each voxel using a finite-difference time domain solver to solve Maxwell's equations; and determining, for each voxel, whether to increase a respective numerical precision of respective values representing behavior of the electromagnetic field at the voxel above a threshold precision by the computer system and, in response, assigning one or more additional degrees of freedom to the voxel.

Finite floating-point numerical simulation and optimization

A computer-implemented method for designing a dispersive optical component includes: (i) defining a loss function within a simulation space composed of multiple voxels, the simulation space encompassing optical interfaces of the component, the loss function corresponding to a target dispersion profile for the component including a relationship between a scattering angle and a wavelength of an incident electromagnetic field for different operative wavelengths; (ii) defining an initial structure for the optical interfaces, at least some of the voxels corresponding to each optical interface having a dimension smaller than a smallest operative wavelength of the component; and (iii) determining, using a computer system, a structure for each optical interface using a finite-difference time domain solver to solve Maxwell's equations so that a loss determined according to the loss function is above a specified threshold.

High-dispersion optical components and methods for making the same

A multi-channel photonic demultiplexer includes an input region to receive a multi-channel optical signal including four distinct wavelength channels, four output regions, each adapted to receive a corresponding one of the four distinct wavelength channels demultiplexed from the multi-channel optical signal, and a dispersive region optically disposed between the input region and the four output regions. The dispersive region includes a first material and a second material inhomogeneously interspersed to form a plurality of interfaces that each correspond to a change in refractive index of the dispersive region and collectively structure the dispersive region to optically separate each of the four distinct wavelength channels from the multi-channel optical signal and respectively guide each of the four distinct wavelength channels to the corresponding one of the four output regions.

Multi-channel integrated photonic wavelength demultiplexer

Methods for designing a mode-selective optical device including one or more optical interfaces defining an optical cavity include: defining a loss function within a simulation space encompassing the optical device, the loss function corresponding to an electromagnetic field having an operative wavelength within the optical device resulting from an interaction between an input electromagnetic field at the operative wavelength and the one or more optical interfaces of the optical device; defining an initial structure for each of the one or more optical interfaces, each initial structure being defined using a plurality of voxels; determining values for at least one structural parameter and/or at least one functional parameter of the one or more optical interfaces by solving Maxwell's equations; and defining a final structure of the one or more optical interfaces based on the values for the one or more structural and/or functional parameters.

Systems and methods for designing optical devices having mode selective facets

A two-channel photonic demultiplexer includes an input region to receive a multi-channel optical signal, two output regions, each adapted to receive a corresponding one of two distinct wavelength channels demultiplexed from the multi-channel optical signal, and a dispersive region including a first material and a second material inhomogeneously interspersed to form a plurality of interfaces that collectively structure the dispersive region to optically separate each of the two distinct wavelength channels from the multi-channel optical signal and respectively guide the first distinct wavelength channel to a first output region and the second distinct wavelength channel to the second output region when the input region receives the multi-channel optical signal. At least one of the first material or the second material is structured within the dispersive region to be schematically reproducible by a feature shape with a pre-determined width.

Two-channel integrated photonic wavelength demultiplexer

A system, apparatus, and method for optimizing structural parameters of a physical device are described. The method includes receiving an initial description of the physical device describing the structural parameters within a simulated environment. The method further includes performing a simulation of the physical device in response to an excitation source to determine a performance metric of the physical device. The simulation environment includes one or more absorbing boundaries for attenuation of an output of the excitation source during the simulation. The method further includes recording attenuated field values of the simulated environment associated with the attenuation during the simulation. The method further includes determining a loss metric based on a difference between the performance metric and a target performance metric, backpropagating the loss metric using the attenuated field values, and generating a revised description of the physical device by updating the structural parameters to reduce the loss metric.

Physical device optimization with reduced memory footprint via time reversal at absorbing boundaries

Systems and methods for designing a hybrid neural network comprising at least one physical neural network component and at least one digital neural network component. A loss function is defined within a design space composed of a plurality of voxels, the design space encompassing one or more physical structures of the at least one physical neural network component and one or more architectural features of the digital neural network. Values are determined for at least one functional parameter for the one or more physical structures, and the at least one architectural parameter for the one or more architectural features, using a domain solver to solve Maxwell's equations so that a loss determined according to the loss function is within a threshold loss. Final structures are defined for the at least one physical neural network component and the digital neural network component based on the values.

Methods for designing hybrid neural networks having physical and digital components

At least one machine-accessible storage medium that provides instructions that, when executed by a machine, will cause the machine to perform operations. The operations comprise configuring a simulated environment to be representative of a physical device based, at least in part, on an initial description of the physical device that described structural parameters of the physical device. The operations further comprise performing a physics simulation with an artificial intelligence ('AI') accelerator. The AI accelerator includes a matrix multiply unit for computing convolution operations via a plurality of multiply-accumulate units. The operations further comprise computing a field response in response of the physical device in response to an excitation source within the simulated environment when performing the physics simulation. The field response is computed, at least in part, with the convolution operations to perform spatial differencing.

Physics simulation on machine-learning accelerated hardware platforms

A method of optimizing structural parameters of a physical device includes: receiving an initial description of the physical device that describes the physical device with an array of voxels that each describe one or more of the structural parameters; performing a time-forward simulation of a field response propagating through the physical device and interacting with the voxels in a simulated environment, wherein the field response is influenced by the structural parameters of the voxels; generating field response values describing the field response at each of the voxels for each of a plurality of time steps; encoding the field response values to generate compressed field response values; storing the compressed field response values; decoding one or more of the compressed field response values to extract regenerated field response values; and generating a revised description of the physical device having a structural parameter optimized.

Compressed field response representation for memory efficient physical device simulation

Embodiments of techniques for inverse design of physical devices are described herein, in the context of generating designs for photonic integrated circuits (including a multi-channel photonic demultiplexer). In some embodiments, an initial design of the physical device is received, and a plurality of sets of operating conditions for fabrication of the physical device are determined. In some embodiments, the performance of the physical device as fabricated under the sets of operating conditions is simulated, and a total performance loss value is backpropagated to determine a gradient to be used to update the initial design. In some embodiments, instead of simulating fabrication of the physical device under the sets of operating conditions, a robustness loss is determined and combined with the performance loss to determine the gradient.

Growing community of inventors

Hollister, CA, United States of America

Jesse Lu