arXiv subjects

Lorenzo Cassola

Publications and source records attributed to Lorenzo Cassola.

4 recordsLinked to original sources

arXiv2026

Space-Charge-Driven Nonlinear Charge Transport in Silicon Reconfigurable Nonlinear-Processing Units

Reconfigurable nonlinear-processing units (RNPUs) are multi-terminal electronic devices that act as computational primitives, exploiting intrinsic nonlinear charge transport combined with electrostatic tunability. Silicon-based realizations provide a scalable and technologically relevant platform toward unconventional computing hardware, yet the physical origin of their room-temperature nonlinearity has remained unresolved. Here, we demonstrate room-temperature operation in both boron- and arsenic-doped silicon RNPUs and show, using temperature- and length-dependent measurements supported by TCAD simulations, that charge transport is governed by space-charge effects. Interface trap states strongly suppress the equilibrium carrier density, while the functional nonlinearity arises from the competition between injected carriers and ionized dopants. The resulting transport evolves from Ohmic to strongly nonlinear and space-charge-limited current regimes, as evidenced by voltage and length scaling. The opposite-polarity background doping is shown to control the onset and strength of the nonlinearity, producing behavior beyond the quadratic dependence of the classical Mott-Gurney law. Agreement between experiment and simulation supports that the spatial distribution of injected carriers and fixed charge governs the electric-field profile and device response. These results establish a space-charge-based framework for RNPUs that does not require disorder or hopping transport, and provide design guidelines for scalable, CMOS-compatible nonlinear computing hardware.

cond-mat.other↗

arXiv2025

In-Materia Speech Recognition

With the rise of decentralized computing, as in the Internet of Things, autonomous driving, and personalized healthcare, it is increasingly important to process time-dependent signals at the edge efficiently: right at the place where the temporal data are collected, avoiding time-consuming, insecure, and costly communication with a centralized computing facility (or cloud). However, modern-day processors often cannot meet the restrained power and time budgets of edge systems because of intrinsic limitations imposed by their architecture (von Neumann bottleneck) or domain conversions (analogue-to-digital and time-to-frequency). Here, we propose an edge temporal-signal processor based on two in-materia computing systems for both feature extraction and classification, reaching a software-level accuracy of 96.2% for the TI-46-Word speech-recognition task. First, a nonlinear, room-temperature dopant-network-processing-unit (DNPU) layer realizes analogue, time-domain feature extraction from the raw audio signals, similar to the human cochlea. Second, an analogue in-memory computing (AIMC) chip, consisting of memristive crossbar arrays, implements a compact neural network trained on the extracted features for classification. With the DNPU feature extraction consuming 100s nW and AIMC-based classification having the potential for less than 10 fJ per multiply-accumulate operation, our findings offer a promising avenue for advancing the compactness, efficiency, and performance of heterogeneous smart edge processors through in-materia computing hardware.

eess.AS↗

arXiv2025

Gradient descent in materia through homodyne gradient extraction

Deep learning, a multi-layered neural network approach inspired by the brain, has revolutionized machine learning. One of its key enablers has been backpropagation, an algorithm that computes the gradient of a loss function with respect to the weights and biases in the neural network model, in combination with its use in gradient descent. However, the implementation of deep learning in digital computers is intrinsically energy hungry, with energy consumption becoming prohibitively high for many applications. This has stimulated the development of specialized hardware, ranging from neuromorphic CMOS integrated circuits and integrated photonic tensor cores to unconventional, material-based computing system. The learning process in these material systems, realized, e.g., by artificial evolution, equilibrium propagation or surrogate modelling, is a complicated and time-consuming process. Here, we demonstrate a simple yet efficient and accurate gradient extraction method, based on the principle of homodyne detection, for performing gradient descent on a loss function directly in a physical system without the need of an analytical description. By perturbing the parameters that need to be optimized using sinusoidal waveforms with distinct frequencies, we effectively obtain the gradient information in a highly robust and scalable manner. We illustrate the method in dopant network processing units, but argue that it is applicable in a wide range of physical systems. Homodyne gradient extraction can in principle be fully implemented in materia, facilitating the development of autonomously learning material systems.

cs.NE↗

arXiv2021

Ultra-Highly Linear Magnetic Flux-to-Voltage response in Proximity-based Mesoscopic bi-SQUIDs

Superconducting double-loop interferometers (bi-SQUIDs) have been introduced to produce magnetic flux sensors specifically designed to exhibit ultra-highly linear voltage response as a function of the magnetic flux. These devices are very important for the quantum sensing and for signal processing of signals oscillating at the radio-frequencies range of the electromagnetic spectrum. Here, we report an Al double-loop bi-SQUIDs based on proximitized mesoscopic Cu Josephson junctions. Such a scheme provides an alternative fabrication approach to conventional tunnel junction-based interferometers, where the junction characteristics and, consequently, the magnetic flux-to-voltage and magnetic flux-to-critical current device response can be largely and easily tailored by the geometry of the metallic weak-links. We discuss the performance of such sensors by showing a full characterization of the device switching current and voltage drop \textit{vs.} magnetic flux for temperatures of operation ranging from 30 mK to $\sim 1$ K. The figure of merit of the transfer function and of the total harmonic distortion are also discussed. The latter provides an estimate of the linearity of the flux-to-voltage device response, which obtained values as large as 45 dB. Such a result let us foresee a performance already on pair with that achieved in conventional tunnel junction-based bi-SQUIDs arrays composed of hundreds of interferometers.

cond-mat.supr-con↗