Showing papers by "Viktor A. Podolskiy published in 2022"

Extending plasmonic response to the mid-wave infrared with all-epitaxial composites.

Abstract: Highly doped semiconductor "designer metals" have been shown to serve as high-quality plasmonic materials across much of the long-wavelength portion of the mid-infrared. These plasmonic materials benefit from a technologically mature semiconductor fabrication infrastructure and the potential for monolithic integration with electronic and photonic devices. However, accessing the short-wavelength side of the mid-infrared is a challenge for these designer metals. In this work we study the perspectives for extending the plasmonic response of doped semiconductors to shorter wavelengths by leveraging charge confinement, in addition to doping. We demonstrate, theoretically and experimentally, negative permittivity across the technologically vital mid-wave infrared (3-5 μm) frequency range. The semiconductor composites presented in our work offer an ideal material platform for monolithic integration with a variety of semiconductor optoelectronic devices operating in the mid-wave infrared.

Self phase-matched broadband amplification with a left-handed Josephson transmission line

Abstract: Josephson Traveling Wave Parametric Amplifiers (J-TWPAs) are promising platforms for realizing broadband quantum-limited amplification of microwave signals. However, substantial gain in such systems is attainable only when strict constraints on phase matching of the signal, idler and pump waves are satisfied – this is rendered particularly challenging in the presence of nonlinear effects, such as selfand cross-phase modulation, which scale with the intensity of propagating signals. In this work, we present a simple J-TWPA design based on ‘left-handed’ (negative-index) nonlinear Josephson metamaterial, which realizes autonomous phase matching without the need for any complicated circuit or dispersion engineering. The resultant efficiency of four-wave mixing process can implement gains in excess of 20 dB over few GHz bandwidths with much shorter lines than previous implementations. Furthermore, the autonomous nature of phase matching considerably simplifies the J-TWPA design than previous implementations based on ‘right-handed’ (positive index) Josephson metamaterials, making the proposed architecture particularly appealing from a fabrication perspective. The left-handed JTL introduced here constitutes a new modality in distributed Josephson circuits, and forms a crucial piece of the unified framework that can be used to inform the optimal design and operation of broadband microwave amplifiers.

Optimizing photonic funnels for mid-IR NSOM tips (Conference Presentation)

Abstract: Waveguides with extremely anisotropic metamaterial cores support propagating modes even when their cross section is deeply subwavelength. A funnel-shaped structure with microscale base, nanoscale tip, and hyperbolic core can therefore act as an efficient optical link between the diffraction-limited and subwavelength domains. In this work we analyze the potential applications of this platform for mid-IR near-field scanning optical microscopy. We present the relationship between the shape and composition of the funnels and the resulting compression of light, characterized by both the intensity and the spread of electromagnetic field in the vicinity of the funnel tip.

Hypergrating for Focusing Vortex Beam Below Diffraction Limit

Abstract: Light with an orbital angular momentum can strongly modify optical transition selection rules when beam size is reduced to subwavelength scale. We demonstrated a method for focusing orbital angular momentum beams below the diffraction limit.

Science-informed machine-learning for optical composites (Conference Presentation)

Abstract: Machine learning is widely used for optimization or classification tasks. Unfortunately, extensive labeled datasets are often required for training machine learning models. In this work we demonstrate that incorporating physics-driven constraints into machine learning algorithms can dramatically improve both accuracy and extendibility of resulting models, simultaneously reducing the size of the required training set and enabling training on unlabeled data. Physics-informed machine learning is illustrated on example of predicting optical modes supported by periodic layered composites. The approach can be readily utilized for analysis of electromagnetic modes in composites with 2D periodic geometry or in complex waveguiding structures.

Physics-Guided Problem Decomposition for Scaling Deep Learning of High-dimensional Eigen-Solvers: The Case of Schrödinger's Equation

Abstract: Given their ability to effectively learn non-linear mappings and perform fast inference, deep neural networks (NNs) have been proposed as a viable alternative to traditional simulation-driven approaches for solving high-dimensional eigenvalue equations (HDEs), which are the foundation for many scientific applications. Unfortunately, for the learnedmodels in these scientific applications to achieve generalization, a large, diverse, and preferably annotated dataset is typically needed and is computationally expensive to obtain. Furthermore, the learned models tend to be memory– and compute–intensive primarily due to the size of the output layer. While generalization, especially extrapolation, with scarce data has been attempted by imposing physical constraints in the form of physics loss, the problem of model scalability has remained. In this paper, we alleviate the compute bottleneck in the output layer by using physics knowledge to decompose the complex regression task of predicting the high-dimensional eigenvectors into multiple simpler sub-tasks, each of which are learned by a simple “expert" network.We call the resulting architecture of specialized experts Physics-Guided Mixture-of-Experts (PG-MoE). We demonstrate the efficacy of such physics-guided problem decomposition for the case of the Schrödinger Equation in Quantum Mechanics. Our proposed PG-MoE model predicts the ground-state solution, i.e., the eigenvector that corresponds to the smallest possible eigenvalue. The model is 150× smaller than the network trained to learn the complex task while being competitive in generalization. To improve the generalization of the PG-MoE, we also employ a physics-guided loss function based on variational energy, which by quantum mechanics principles is minimized iff the output is the ground-state solution.