arXiv subjects

David Gray

Publications and source records attributed to David Gray.

5 recordsLinked to original sources

arXiv2026

High-Repetition-Rate Projection Multiphoton Lithography for Large-Area Sub-Micron 3D Printing

High-resolution lithographic techniques are often limited by low volumetric throughput, since there is no universal and scalable manufacturing process that can produce 3D metasurfaces. In this work, we demonstrate a high-speed holographic 3D printing platform based on spatiotemporal beam shaping, exceeding the repetition rate while keeping the resolution high. The system integrates a femtosecond laser source with a spectral pulse compressor and a beam shaper to project uniform, axially confined light fields to project patterns directly on the advanced photoresists using a Digital Micromirror Device DMD. We investigate the process window for rapid polymerization, optimizing the photoinitiator choice to eliminate thermal crosstalk at high repetition rates. Using this setup, we achieve a production throughput of more than a million voxels per second with sub-micron resolution below 400 nm. The system's reliability is validated through the fabrication of large-area woodpile-like lattices and uniform micropillar arrays, establishing a workflow for scalable manufacturing of micro-optical components.

physics.optics↗

arXiv2025

Latent space analysis and generalization to out-of-distribution data

Understanding the relationships between data points in the latent decision space derived by the deep learning system is critical to evaluating and interpreting the performance of the system on real world data. Detecting \textit{out-of-distribution} (OOD) data for deep learning systems continues to be an active research topic. We investigate the connection between latent space OOD detection and classification accuracy of the model. Using open source simulated and measured Synthetic Aperture RADAR (SAR) datasets, we empirically demonstrate that the OOD detection cannot be used as a proxy measure for model performance. We hope to inspire additional research into the geometric properties of the latent space that may yield future insights into deep learning robustness and generalizability.

stat.ML↗

arXiv2024

Emerging Ta$_{4}$C$_{3}$ and Mo$_{2}$Ti$_{2}$C$_{3}$ MXene Nanosheets for Ultrafast Photonics

Ultrafast nonlinear optical (NLO) response, fast carrier recovery, broadband absorption, and resistance to radiation and heat make 2D materials promising for photonic technologies. However, low electronic conductivity and carrier concentration limit the performance of semiconducting or semimetallic materials. This work investigates the ultrafast NLO properties and carrier dynamics of Ta$_{4}$C$_{3}$T$_{x}$ and out-of-plane ordered Mo$_{2}$Ti$_{2}$C$_{3}$T$_{x}$ MXenes using Z-scan and pump-probe optical Kerr effect techniques under visible and infrared femtosecond laser pulses. Their NLO response surpasses all previously studied MXenes and most other 2D nanomaterials, attaining exceptionally high third-order susceptibility (\c{hi}(3)) values on the order of 10^{-13} esu. Mo$_{2}$Ti$_{2}$C$_{3}$T$_{x}$ exhibits the strongest NLO response under both excitation regimes, attributed to charge transfer between Mo and Ti layers in the MXene structure. Under visible excitation, the studied MXenes display pronounced saturable absorption, while under infrared excitation, they exhibit strong reverse saturable absorption, resulting in efficient optical limiting. Additionally, pump-probe experiments identify two distinct relaxation processes: a fast one on the sub-picosecond timescale and a slower one a few picoseconds after photoexcitation. Our results indicate that these MXenes are among the strongest NLO materials. They show their great potential for advanced photonic and optoelectronic applications in laser technologies, optical protection, telecommunications, and optical/quantum computing.

physics.optics↗

arXiv2019

Characterizing Inter-Layer Functional Mappings of Deep Learning Models

Deep learning architectures have demonstrated state-of-the-art performance for object classification and have become ubiquitous in commercial products. These methods are often applied without understanding (a) the difficulty of a classification task given the input data, and (b) how a specific deep learning architecture transforms that data. To answer (a) and (b), we illustrate the utility of a multivariate nonparametric estimator of class separation, the Henze-Penrose (HP) statistic, in the original as well as layer-induced representations. Given an $N$-class problem, our contribution defines the $C(N,2)$ combinations of HP statistics as a sample from a distribution of class-pair separations. This allows us to characterize the distributional change to class separation induced at each layer of the model. Fisher permutation tests are used to detect statistically significant changes within a model. By comparing the HP statistic distributions between layers, one can statistically characterize: layer adaptation during training, the contribution of each layer to the classification task, and the presence or absence of consistency between training and validation data. This is demonstrated for a simple deep neural network using CIFAR10 with random-labels, CIFAR10, and MNIST datasets.

stat.ML↗

arXiv2006

A formally verified proof of the prime number theorem

The prime number theorem, established by Hadamard and de la Vall'ee Poussin independently in 1896, asserts that the density of primes in the positive integers is asymptotic to 1 / ln x. Whereas their proofs made serious use of the methods of complex analysis, elementary proofs were provided by Selberg and Erd"os in 1948. We describe a formally verified version of Selberg's proof, obtained using the Isabelle proof assistant.

cs.AI↗