Techniques of robust inverse design that account for manufacturing variabilities due to operating conditions
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.
1. A non-transitory computer-readable medium having logic stored thereon that, in response to execution by at least one computing device, causes the at least one computing device to perform actions for generating a design of a physical device, the actions comprising:
receiving an initial design of the physical device;
determining a plurality of sets of operating conditions for fabrication of the physical device;
simulating performance of the physical device as fabricated using the initial design under each set of operating conditions of the plurality of sets of operating conditions;
determining a total performance loss value based on results of the simulations under each set of operating conditions of the plurality of sets of operating conditions;
backpropagating the total performance loss value to determine a gradient corresponding to an influence of changes in the initial design on the total performance loss value; and
revising the initial design of the physical device based at least in part on the gradient.
2. The computer-readable medium of claim 1 , wherein determining the plurality of sets of operating conditions for fabrication of the physical device includes stochastically sampling values for each operating condition from predetermined ranges of values for each operating condition.
3. The computer-readable medium of claim 1 , wherein determining the plurality of sets of operating conditions for fabrication of the physical device includes:
uniformly sampling values for each operating condition; and
using combinations of the uniformly sampled values as the plurality of sets of operating conditions.
4. The computer-readable medium of claim 3 , wherein determining the total performance loss value includes determining a linear combination of performance loss values for the simulations of each set of operating conditions.
5. The computer-readable medium of claim 1 , wherein the actions further comprise:
determining sets of operating conditions that vary each operating condition separately; and
using the sets of operating conditions with separately varied operating conditions to determine a sensitivity toward each operating condition.
6. The computer-readable medium of claim 5 , wherein the actions further comprise determining a plurality of sets of operating conditions to be used for determining the total performance loss value by performing a non-linear sampling of each operating condition based on the determined sensitivities.
7. The computer-readable medium of claim 5 , wherein determining the total performance loss value includes determining a non-linear combination of performance loss values for the simulations of each set of operating conditions based on the determined sensitivities.
8. The computer-readable medium of claim 1 , wherein the actions further comprise determining a perturbation model for calculating a robustness loss value, and wherein determining the perturbation model includes using at least the simulated performance of the physical device as fabricated under each set of operating conditions to determine the perturbation model.
9. The computer-readable medium of claim 8 , wherein using at least the simulated performance of the physical device as fabricated under each set of operating conditions to determine the perturbation model includes using states within the physical device calculated during the simulation to determine the perturbation model.
10. The computer-readable medium of claim 8 , wherein determining the perturbation model for calculating the robustness loss value includes using a model of a fabrication process to determine an effect of perturbations in the operating conditions on a performance loss value based on the results of the simulations.
11. The computer-readable medium of claim 8 , wherein determining the total performance loss value includes combining the robustness loss value with the performance loss value.
12. The computer-readable medium of claim 1 , wherein the operating conditions include at least one of ambient temperature, erosion, dilation, waveguide thickness, structure out of plane, sidewall angle, surface roughness, misalignment, optical aberrations, and material imperfections.
13. The computer-readable medium of claim 1 , wherein the actions further comprise repeating the actions of simulating performance of the physical device as fabricated, determining the total performance loss value, backpropagating the total performance loss value, and revising the initial design until the total performance loss value converges to a minimum.
14. The computer-readable medium of claim 1 , wherein the actions further comprise transmitting the revised initial design to a fabrication system for fabrication.
15. A system comprising at least one computing device configured with logic that, in response to execution by the at least one computing device, cause the system to perform actions for generating a design of a physical device, the actions comprising:
receiving an initial design of the physical device;
determining a plurality of sets of operating conditions for fabrication of the physical device;
simulating performance of the physical device as fabricated using the initial design under each set of operating conditions of the plurality of sets of operating conditions;
determining a total performance loss value based on results of the simulations under each set of operating conditions of the plurality of sets of operating conditions;
backpropagating the total performance loss value to determine a gradient corresponding to an influence of changes in the initial design on the total performance loss value; and
revising the initial design of the physical device based at least in part on the gradient.
16. The system of claim 15 , wherein determining the plurality of sets of operating conditions for fabrication of the physical device includes stochastically sampling values for each operating condition from predetermined ranges of values for each operating condition.
17. The system of claim 15 , wherein the actions further comprise:
determining sets of operating conditions that vary each operating condition separately; and
using the sets of operating conditions with separately varied operating conditions to determine a sensitivity toward each operating condition.
18. The system of claim 17 , wherein the actions further comprise determining a plurality of sets of operating conditions to be used for determining the total performance loss value by performing a non-linear sampling of each operating condition based on the determined sensitivities.
19. The system of claim 17 , wherein determining the total performance loss value includes determining a non-linear combination of performance loss values for the simulations of each set of operating conditions based on the determined sensitivities.
20. The system of claim 17 , wherein the actions further comprise determining a perturbation model for calculating a robustness loss value, and wherein determining the perturbation model includes using at least the simulated performance of the physical device as fabricated under each set of operating conditions to determine the perturbation model.