In-Situ Inverse Design of a Plasma Metamaterial Beam Steering Device

TL;DR

This work demonstrates in-situ inverse design of a plasma metamaterial beam-steering device by directly tuning 91 plasma elements on hardware via Bayesian optimization guided by measured scattering parameters. The approach yields orders-of-magnitude improvements in port isolation compared with in-silico designs, including ~50 dB isolation in narrowbands and usable isolation across broad 4–12 GHz spans, revealing the limitations of purely simulated models while offering a pathway for rapid, reconfigurable EM devices. The study also outlines how in-situ optimization can inform plasma-model refinement, surrogate modeling, and future enhancements such as 3D hardware, magnetized plasmas, and more advanced optimization strategies. Overall, the PMM paradigm with in-situ optimization offers a compelling route to fully programmable, high-performance electromagnetic devices that adapt in real time to different operating requirements.

Abstract

Inverse design is a commonly used methodology for creating devices that manipulate electromagnetic (EM) waves by algorithmically modifying device parameters to achieve a desired functionality. Utilizing plasma, a dynamically tunable medium, allows the optimization of the design process to be conducted directly on the experimental hardware (in-situ). A key advantage of this method is the creation of devices that are inherently switchable and dynamically reconfigurable. Bayesian optimization is used to tune the plasma density of 91 independent discharges that make up a plasma metamaterial (PMM) device to steer incoming EM waves to desired exit waveguides. Measurements were conducted in an automated loop where a vector network analyzer records the PMM transmission characteristics for each device setting. By relying only on measured scattering parameters, this gradient-free approach is robust to experimental drift and noise and does not require complex full-wave models. Significant performance improvements over traditional simulation-based (in-silico) inverse design are demonstrated, with in-situ Bayesian optimization achieving up to 10,000x higher isolation between ports than the best in-silico design at the same target frequency. This work also presents guidelines for applying Bayesian optimization to noisy, high-dimensional physical systems.

Figures (7)

• Figure 1: Illustration of the inverse-design workflow showing the complementary roles of in-silico and in-situ optimization.
• Figure 2: Experimental setup: (left) schematic of the PMM control system, and (right) photograph of the physical implementation of the 91-element plasma array.
• Figure 3: The quasi-experimental mapping for plasma frequency tuning, showing three fitted voltage-frequency mappings (thick dashed curves) that bound the low-, mid-, and high-density cases encoded by $k$ in Equation 2. Each of the thin colored curves plotted underneath the fitted function show the BOLSIG+-based estimate of plasma frequency for a specific combination of gas pressure and temperature (e.g. 220 Pa and 300 K) resulting from a sweep across possible values.
• Figure 4: PMM FDFD simulation results at 4, 6, and 8 GHz: steering comparing in-silico and in-situ parameters, showing stronger steering with the in-silico parameters. The left panels for each parameter set plot E-field intensity $|E_z|^2$ in the domain and the right panels show the relative permittivity in the domain. The discharges with lowest permittivity correspond to those with the highest plasma density.
• Figure 5: PMM experimental transmission spectra for the standard beam steering objective at 4, 6 and 8 GHz for the in-silico parameters (left) and the in-situ parameters (right) along with the objective evolution over time (center). The operating frequency is denoted by a vertical dashed black line. The objective values are in arbitrary units. For each target frequency, in-situ optimization improves isolation over the in-silico parameters by a factor of up to 10$^3$ at 6 GHz and 10$^4$ at 8 GHz on a power scale.