Reverse mode automatic differentiation (RMAD) is widely used in deep learning training due to its runtime being independent of the number of training parameters. However, RMAD is limited by its high memory consumption, storing every intermediate value and operation, making it incompatible with commonly employed time-stepping finite difference time domain (FDTD) electromagnetic simulators. To address this issue, a differentiable FDTD simulator is proposed that exploits the time-reversal properties of Maxwell’s equations and removes redundant operations at each timestep, resolving the memory bottleneck. This approach enables the efficient calculation of high-dimensional objective function gradients, expanding the applicability of inverse-design topology optimization.

To meet the growing demand for design automation of metalens with inverse design capability, Synopsys has developed a powerful and user-friendly multi-domain design tool, MetaOptic Designer (MOD), that can handle lenses with subwavelength features. The optimization algorithm employs the well-known adjoint method, which can easily handle millions of design variables. The forward propagation is done by the efficient Fourier transform, also called angular spectrum method. The transfer function of each metalens is characterized by a parametric BSDF databases built with RCWA or FDTD. Application examples include achromatic metalens, large FOV metalens, reflective metalens, chiral hologram, and hybrid optical system.

We report on the design of a tunable, near-infrared (NIR) femtosecond noncollinear optical parametric amplifier (OPA) seeded by gain-managed nonlinear amplified1 parabolic pulses. In our numerical simulations, we achieve signal and idler amplification bandwidths between 1000-1180 nm and 914-1062 nm; the second harmonic of these pulses corresponds directly to the visible spectrum. This appreciable value is due to the high spectral energy density of our seed pulses. Fiber-amplified pulses thus present a method to engineer efficient OPA systems that can operate at high-average powers in the NIR and visible.

A single-shot, spatially resolved wavefront gradient measurement and polarimetry technique is proposed by combining a Shack-Hartmann Wavefront Sensor (SHWFS) and a Star Test Imaging Polarimeter (STIP). By placing the focal plane of the SHWFS at the object plane of the STIP, the gradient and polarization can be retrieved with a position and Stokes retrieval algorithm. The algorithm generates a set of irradiance patterns, using a measurement matrix, from a set of Stokes vectors as system inputs. A normalized cross-correlation is performed with the measured irradiance pattern and each generated irradiance pattern. By searching through this set for the maximum degree of correlation, the Stokes vector and the displacement is retrieved. Using simulated irradiance patterns with noise levels of 5% or less, the algorithm correctly retrieves the Stokes vector and displacement of the measured irradiance pattern and has sub-pixel spread for both. At noise levels of 7.5%, the algorithm demonstrates bias in the Stokes vector retrieval with sub-pixel spread and it correctly retrieves the displacement with sub-pixel spread. At noise levels of 10%, the algorithm demonstrates bias in the Stokes vector retrieval with single-pixel spread and it correctly retrieves the displacement with single-pixel spread in the horizontal direction.

Reflectivity control devices (RCDs) based on polymer dispersed liquid crystals were fabricated for Solar Cruiser, a SmallSat NASA Pathfinder Mission consisting of a 1653 square meter solar sail that would establish an artificial orbit sunward of the L1 Lagrange point for heliophysics observations. Here we describe the optical characterization of these birefringent electro-optic devices including thin film measurements and analysis, hyperspectral bidirectional reflectance distribution function measurements, and radiometric analysis. These measurements demonstrate the promise of RCDs for roll control and momentum management of solar sails.

The lack of accessible tools to design fluid-affected optical systems hampers performance prediction, often causing failures. Modeling such systems must consider optical changes from turbulent media, including pressure, temperature, and density variations, resulting in a non-homogenous refractive index. These variations induce optical aberrations, affecting image clarity. Advanced multi-physics simulations, combining computational fluid dynamics, finite element analysis, and optics modeling, enable designers to predict performance and mitigate potential failures. We showcase the process by analyzing airflow's impact on electro-optic infrared and laser communication systems. This approach allows for early design-stage mitigation strategies. The technology's robustness extends to other fluids and conditions like underwater applications and turbulent flows.

We present simple formulas for the diffraction efficiencies of a binary phase grating that performs array illumination with ultrashort laser pulses. Using scalar diffraction theory, we formulated the efficiencies as a function of pulse spectral width by Fourier-transforming the complex-modulated frequency spectra of diffracted pulses in the far-field region. From the analytical simulations, we found that pulse array uniformity departs from unity as the spectral width increases, or the pulse duration decreases, thereby limiting the attainable split counts. This finding can be considered in the design of gratings for delivering controlled amounts of pulse energies to diffraction orders of interest.

Recently, there has been great interest in photonic inverse design with the goal of creating groundbreaking photonic devices. This has largely been enabled by the adjoint method, which allows for efficient optimization of electromagnetic structures with respect to a large number of degrees of freedom. However, despite considerable progress, inverse design of large-scale 3D structures using full-wave simulations still requires computational time that can be prohibitively long. Here, we demonstrate photonic inverse design enhanced by a hardware-accelerated finite-difference time-domain solver and automatic differentiation. A large-area, CMOS-compatible grating coupler was optimized in a few hours using our methodology.