arXiv subjects

Mathieu Luisier

Publications and source records attributed to Mathieu Luisier.

At least 19 recordsLinked to original sources

arXiv2026

Efficient Simulation of Gate Leakage Currents in Ultra-scaled Transistors from First Principles

To ensure high electrostatic integrity in ultra-scaled field-effect transistors (FETs), thin gate dielectric layers are typically needed, potentially giving rise to (strong) leakage currents. Being able to model this effect is therefore crucial. Here, we show how gate leakage currents can be computed at the ab initio level in WS$_2$-hBN FETs by extending the well-established quantum transmitting boundary method (QTBM). Two main challenges are tackled: (i) the calculation of the open boundary conditions (OBCs) for the gate contacts and (ii) the solution of the resulting sparse linear system of equations A x = b where the matrix A possesses an unstructured sparsity pattern. We demonstrate that the proposed method outperforms the state of the art and can be readily applied to simulate transistors with gate lengths exceeding 25 nm.

cond-mat.mtrl-sci↗

arXiv2026

Accelerating SO(2)-Equivariant Materials Models through Low-Precision Emulation

Rotationally equivariant architectures improve the physical fidelity of machine-learned (ML) universal materials models based on graph neural networks (GNNs). However, their scalability is limited by the use of non-trivial, high-precision tensor operations during both training and inference. Adopting the reduced-precision formats of matrix multiplication units (MMUs), e.g., FP8 instead of FP32, provides massive increases in throughput, up to 30x (60x) on NVIDIA Hopper (Blackwell) GPUs, but may lead to drastic loss of accuracy. Here, we introduce tailored 'emulated' low-precision operations under the integer Ozaki scheme to accelerate the 'SO(2)-equivariant' processing layers of GNNs, which dominate the runtime of (1) many state-of-the-art ML interatomic potentials (MLIPs) and (2) most universal ML Hamiltonian (MLH) models. Our framework, FlashSO(2), outperforms TF32, the most accurate native low-precision MMU format, in throughput, accuracy, and memory footprint at once. At the SO(2) layer level, it reaches 40x (7x) the throughput of MALOQ's (UMA's) reference implementation, an MLH (MLIP) model, sustaining up to 236 (232) effective Tflop/s on an NVIDIA GH200, 3.5x the FP32 peak. At the model level, FlashSO(2) helps train MALOQ 3.9x faster than its pure-FP32 version while reducing peak memory by 58% at no loss in accuracy. In molecular dynamics simulations with UMA, the speedup factor is 3.3x, and the peak memory is reduced by 51%. FlashSO(2) thus pushes the speed-accuracy Pareto frontier of equivariant materials models beyond the native low-precision formats and substantially lowers the resources required to exploit their scaling laws.

cs.LG↗

arXiv2026

EmuGEMM: Fused Tensor Core Kernels for Precision Emulation in Matrix Multiplication

Modern GPUs devote an increasing silicon budget to low-precision matrix-multiplication units, widening the precision-throughput gap for scientific computing workloads. Ozaki Schemes I and II offer an alternative by reconstructing high-precision general matrix multiplication (GEMM) from low-precision operations, yet existing implementations leave substantial performance untapped. In particular, intermediate results are repeatedly materialized in global memory, making data movement the dominant bottleneck. We present EmuGEMM, fused integer Tensor Core kernels for NVIDIA Hopper and Blackwell GPUs that eliminate redundant memory round-trips in both Ozaki schemes. Using Scheme I, EmuGEMM sustains up to 1,639 Top/s on Hopper (83% of INT8 peak) and 3,654 Top/s on Blackwell (81%). For large matrices, EmuGEMM surpasses cuBLAS TF32 throughput by up to 1.4x on Hopper and 1.7x on Blackwell, at comparable accuracy. Using Scheme II, EmuGEMM extends to complex arithmetic and outperforms cuBLAS ZGEMM by up to 2.3x on Hopper and 5.5x on Blackwell.

cs.DC↗

arXiv2026

Experimentally constrained modeling of the Pockels response of KNbO3 and KTaNbO3

The soft-mode of electro-optic (EO) metal-oxide perovskites plays a critical role in determining their Pockels responses. This is the case of potassium tantalate niobate (KTN), which exhibits an intrinsic Pockels response 2.5 times larger than that of state-of-the-art barium titanate (BTO), highlighting its potential for high-performance EO applications. By combining ab initio calculations at the density-functional theory level and far-IR measurements, we reveal that the harmonic approximation combined with semi-local exchange-correlation functionals fails to accurately capture the soft transverse optical (TO) Slater mode of both potassium niobate and KTN, which dominates the technologically relevant r51 Pockels coefficient. Replacing the calculated mode frequency with its measured value provides an experimentally constrained approach that substantially improves the predicted Pockels response. For that purpose, the previously unreported TO Slater-mode frequency of KTN is extracted from far-IR reflectivity measurements. The results emphasize the potential of KTN-based EO modulators as an alternative to standard lithium niobate and BTO technologies, with a potentially lower energy consumption and device footprint.

cond-mat.mtrl-sci↗

arXiv2026

Ab initio Modeling of MoS2/Oxide Device Interfaces with Machine Learned Electronic Structures

We introduce a new ab initio approach to simulate semiconductor devices that integrates scalable machine-learned (ML) electronic structure models with an advanced quantum transport (QT) solver. The developed framework enables 10,000X speedups over density functional theory to produce the Hamiltonian matrix of devices made of >20,000 atoms, while offering high prediction accuracy. We use its unique features to investigate MoS2/oxide samples and single-layer MoS2 field-effect transistors, where the surrounding oxide layers, here, HfO2 or Al2O3, are explicitly included into the QT domain. In particular, we reveal that the presence of undercoordinated metal atoms (Hf or Al) close to the semiconductor-oxide interface significantly affects the magnitude of the electronic current and its propagation through MoS2.

cond-mat.mtrl-sci↗

arXiv2026

Elemental Germanium Phase-Change Memory

Phase-change memory (PCM) is a mature technology for fast, scalable, non-volatile data storage, with applications spanning embedded memory, as well as in-memory and neuromorphic computing. PCM predominantly relies on chalcogenide alloys, with $\mathrm{Ge_2Sb_2Te_5}$ (GST) as the industry standard. Yet in these alloys, the individual Ge, Sb, and Te atoms redistribute upon cycling, causing stochastic operation and ultimately device failure. To address this issue, elemental antimony was proposed as a PCM material, but it exhibits a metastable amorphous state that prevents reliable data retention. Moreover, tellurium and antimony can contaminate complementary metal-oxide-semiconductor (CMOS) production lines or act as unintended dopants, restricting manufacturing of PCM to dedicated fabs. Here we introduce elemental germanium (Ge) as a CMOS-native phase-change material that overcomes these fundamental limitations. In a vertical PCM cell architecture, Ge enables sub-nanosecond crystallization (240 ps, 40 times faster than GST), non-volatile data storage with excellent thermal stability ($>$110 °C for 10 years vs. $\sim$87 °C for GST), and a resistance drift coefficient approximately 60% lower than in GST. These results establish pure Ge, a standard semiconductor, as an alternative to chalcogenide phase-change materials, achieving superior performance in key metrics and enabling phase-change memory to be fabricated in standard semiconductor facilities.

cond-mat.mtrl-sci↗

arXiv2026

Simulation of exchange coupling effects in double quantum dot FinFET-like structures

By leveraging a GPU-accelerated Schrödinger-Poisson (SP) solver, we characterize exchange coupling in a hole spin double-qubit device involving a double quantum dot (DQD) system formed inside a 5-gate silicon fin field-effect transistor (FinFET) similar to real experimental structures. The self-consistent SP simulations rely on a finite difference discretization of the 3D volume and on a Luttinger-Kohn 6x6 kp Hamiltonian accounting for magnetic fields and strain distribution. They return the gate-induced confined electronic states and the corresponding electrostatic potential hosting the DQD. These quantities serve as inputs to a two-particle Hamiltonian that is constructed from single-particle Slater determinants through the configuration interaction (CI) method. By diagonalizing this two-particle Hamiltonian, the eigenstates and eigenenergies of the DQD system are obtained, together with their exchange coupling. We show that our simulation framework, using a reduced number of basis states, is capable of reproducing the magneto-electrostatic behavior of the devices of interest, as predicted from theory and observed experimentally. We finally leverage our approach to determine the optimal operating conditions of a two-qubit quantum logic gate implemented in a Si FinFET structure.

quant-ph↗

arXiv2026

MALOQ: Massively Accelerated Learning of Operators for Quantum Transport

Machine-learned (ML) operator models can be trained to predict density functional theory (DFT) Hamiltonian/density matrices at significantly reduced computational cost, thus extending electronic-structure calculations to previously unfeasible scales. Here, we introduce MALOQ (Massively Accelerated Learning of Operators for Quantum Transport), an application built to train on and predict electronic-structure matrices for systems made of few to 100k atoms, described by large basis sets, and covering a wide range of atomic elements. Based on a state-of-the-art, SO(2)-equivariant backbone architecture, MALOQ provides (i) custom data-processing kernels to handle high-rank Hamiltonian matrix data and (ii) a scalable edge-wise distribution of atomic graph(s). Trained on the largest molecular Hamiltonian datasets available today, it reduces time-per-epoch by over 30% compared to a molecule-wise-distributed framework, and enables inference on material graphs of arbitrary size. We demonstrate scalable training and inference for 3,000-12,000 atoms on the Alps supercomputer, up to 192 GPUs and 256 GPUs, respectively.

cs.LG↗

arXiv2026

Optimizing Semiconductor Device Simulations through Low-Precision Arithmetic

Architectural changes in GPUs, especially the promotion of low-precision computational units, pose significant challenges to traditional, FP64-based high-performance computing (HPC) applications, while also presenting opportunities. Adopting reduced-precision data formats is a promising avenue to exploit the increased throughput capabilities. However, straightforward data conversions may lead to degraded accuracy or even erroneous results. For a given application, only an in-depth analysis of its numerical stability can reveal the potential of low-precision arithmetic. In this work, we consider the open-source quatrex package, a quantum transport solver capable of breaking the sustained FP64 Eflop/s barrier, to illustrate trade-offs between accuracy losses and computational speed-ups when moving from high- to low-precision formats. We use three representative benchmark structures to explore the application's numerical properties. Applying the gained insights to a larger, more realistic system, we achieve up to 51% higher throughput while maintaining accurate results, on 40% fewer HPC resources than the FP64 reference.

cs.CE↗

arXiv2026

20 ps Non-Destructive Read and 1 ns Write Operations at <5 V in Ferroelectric HfO2/ZrO2 Non-Volatile Memories

Achieving low-voltage, nanosecond multi-level programming and non-destructive read-out of ferroelectric non-volatile memories (NVM) is critical for analog in-memory computing architectures relying on ferroelectric capacitive devices (FeCap). We integrate HfO2/ZrO2 ferroelectric nanolayers concurrently in the BEOL of CMOS and on SiO2/Si, achieving nanosecond multilevel switching with programming voltages below 5 V. Partial ferroelectric switching enhances FeCap endurance above 1011 cycles, leading to MemCapacitance (MC) states with non-destructive read-out and 10-year retention. However, experiments reveal the collapse of the MC window for read frequencies above 1 MHz. To overcome this speed limit, we introduce a novel, non-destructive readout methodology. Using electrical pulses with widths down to 20 ps, below the RC time constant of the FeCaps, we enable measurement of the polarization-dependent leakage current, providing ultrafast and non-destructive read operations at only 14 fJ.

physics.app-ph↗

arXiv2026

ADELIA: Automatic Differentiation for Efficient Laplace Inference Approximations

Spatio-temporal Bayesian inference drives environmental and health sciences using latent Gaussian models. Integrated Nested Laplace Approximations (INLA) enable inference for these models at HPC scale but rely on derivative-based optimization over $d$ hyperparameters. State-of-the-art INLA implementations approximate derivatives via central finite differences (FD), requiring $2d{+}1$ evaluations. These evaluations are embarrassingly parallel, but total work and energy grow with $d$, limiting time-to-solution under fixed budgets. Reverse-mode automatic differentiation (AD) computes exact gradients independently of $d$, but its efficient application to INLA's structured-sparse kernels is an open challenge. We present ADELIA, the first AD-enabled INLA implementation with a structure-exploiting multi-GPU backward pass leveraging model sparsity. We evaluate ADELIA on ten benchmark models, including real-world air-pollution monitoring. We achieve $4.2$--$7.9\times$ per-gradient speedups and reliable convergence on production-scale models with up to 1.9M latent variables, where FD struggles. Even when scaled to 16--32 GPUs to match ADELIA's wall-clock time, FD consumes $5$--$8\times$ more energy.

cs.DC↗

arXiv2026

Multiscale Modeling of Metal/Oxide/Metal Conductive Bridging Random Access Memory Cells: from Ab Initio to Finite Element Calculations

We present a multiscale simulation framework to compute the current vs. voltage (I-V ) characteristics of metal/oxide/metal structures building the core of conductive bridging random access memory (CBRAM) cells and to shed light on their resistance switching properties. The approach relies on a finite element model whose input material parameters are extracted either from ab initio or from machine-learned empirical calculations. The applied techniques range from molecular dynamics and nudged elastic band to electronic and thermal quantum transport. Such an approach drastically reduces the number of fitting parameters needed and makes the resulting modeling environment more accurate than traditional ones. The developed computational framework is then applied to the investigation of an Ag/a-SiO2/Pt CBRAM, reproducing experimental data very well. Moreover, the relevance of Joule heating is assessed by considering various cell geometries. It is found that self-heating manifests itself in devices with thin conductive filaments with few-nanometer diameters and at current concentrations in the 10s-microampere range. With the proposed methodology it is now possible to explore the potential of not-yet fabricated memory cells and to reliably optimize their design.

cond-mat.mtrl-sci↗

arXiv2026

Conductance-dependent Photoresponse in a Dynamic SrTiO3 Memristor for Biorealistic Computing

Modern computers perform pre-defined operations using static memory components, whereas biological systems learn through inherently dynamic, time-dependent processes in synapses and neurons. The biological learning process also relies on global signals - neuromodulators - who influence many synapses at once depending on their dynamic, internal state. In this study, using optical radiation as a global neuromodulatory signal, we investigate nanoscale SrTiO3 (STO) memristors that can act as solid-state synapses. Via diverse sets of measurements, we demonstrate that the memristor's photoresponse depends on the electrical conductance state, following a well-defined square root relation. Additionally, we show that the conductance decays after photoexcitation with time constants in the range of 1 - 10 s and that this effect can be reliably controlled using an electrical bias. These properties in combination with our device's low power operation (< 1pJ per optical pulse) and small measurement variability may pave the way for space- and energy-efficient implementations of complex biological learning processes in electro-optical hardware.

cs.ET↗

arXiv2026

Acceleration of Atomistic NEGF: Algorithms, Parallelization, and Machine Learning

The Non-equilibrium Green's function (NEGF) formalism is a particularly powerful method to simulate the quantum transport properties of nanoscale devices such as transistors, photo-diodes, or memory cells, in the ballistic limit of transport or in the presence of various scattering sources such as electronphonon, electron-photon, or even electron-electron interactions. The inclusion of all these mechanisms has been first demonstrated in small systems, composed of a few atoms, before being scaled up to larger structures made of thousands of atoms. Also, the accuracy of the models has kept improving, from empirical to fully ab-initio ones, e.g., density functional theory (DFT). This paper summarizes key (algorithmic) achievements that have allowed us to bring DFT+NEGF simulations closer to the dimensions and functionality of realistic systems. The possibility of leveraging graph neural networks and machine learning to speed up ab-initio device simulations is discussed as well.

cond-mat.mtrl-sci↗

arXiv2026

Machine-Learned Hamiltonians for Quantum Transport Simulation of Valence Change Memories

The construction of the Hamiltonian matrix \textbf{H} is an essential, yet computationally expensive step in \textit{ab-initio} device simulations based on density-functional theory (DFT). In homogeneous structures, the fact that a unit cell repeats itself along at least one direction can be leveraged to minimize the number of atoms considered and the calculation time. However, such an approach does not lend itself to amorphous or defective materials for which no periodicity exists. In these cases, (much) larger domains containing thousands of atoms might be needed to accurately describe the physics at play, pushing DFT tools to their limit. Here we address this issue by learning and directly predicting the Hamiltonian matrix of large structures through equivariant graph neural networks and so-called augmented partitioning training. We demonstrate the strength of our approach by modeling valence change memory (VCM) cells, achieving a Mean Absolute Error (MAE) of 3.39 to 3.58 meV, as compared to DFT, when predicting the Hamiltonian matrix entries of systems made of $\sim$5,000 atoms. We then replace the DFT-computed Hamiltonian of these VCMs with the predicted one to compute their energy-resolved transmission function with a quantum transport tool. A qualitatively good agreement between both sets of curves is obtained. Our work provides a path forward to overcome the memory and computational limits of DFT, thus enabling the study of large-scale devices beyond current \textit{ab-initio} capabilities

cond-mat.dis-nn↗

arXiv2026

Parallel Quadratic Selected Inversion in Quantum Transport Simulation

Driven by Moore's Law, the dimensions of transistors have been pushed down to the nanometer scale. Advanced quantum transport (QT) solvers are required to accurately simulate such nano-devices. The non-equilibrium Green's function (NEGF) formalism lends itself optimally to these tasks, but it is computationally very intensive, involving the selected inversion (SI) of matrices and the selected solution of quadratic matrix (SQ) equations. Existing algorithms to tackle these numerical problems are ideally suited to GPU acceleration, e.g., the so-called recursive Green's function (RGF) technique, but they are typically sequential, require block-tridiagonal (BT) matrices as inputs, and their implementation has been so far restricted to shared memory parallelism, thus limiting the achievable device sizes. To address these shortcomings, we introduce distributed methods that build on RGF and enable parallel selected inversion and selected solution of the quadratic matrix equation. We further extend them to handle BT matrices with arrowhead, which allows for the investigation of multi-terminal transistor structures. We evaluate the performance of our approach on a real dataset from the QT simulation of a nano-ribbon transistor and compare it with the sparse direct package PARDISO. When scaling to 16 GPUs, our fused SI and SQ solver is 5.2x faster than the SI module of PARDISO applied to a device 16x shorter. These results highlight the potential of our method to accelerate NEGF-based nano-device simulations.

cs.DC↗

arXiv2025

Channel-last gate-all-around nanosheet oxide semiconductor transistors

As we move beyond the era of transistor miniaturization, back-end-of-line-compatible transistors that can be stacked monolithically in the third dimension promise improved performance for low-power electronics. In advanced transistor architectures, such as gate-all-around nanosheets, the conventional channel-first process involves depositing dielectrics directly onto the channel. Atomic layer deposition of gate dielectrics on back-end-of-line compatible channel materials, such as amorphous oxide semiconductors, can induce defects or cause structural modifications that degrade electrical performance. While post-deposition annealing can partially repair this damage, it often degrades other device metrics. We report a novel channel-last concept that prevents such damage. Channel-last gate-all-around self-aligned transistors with amorphous oxide-semiconductor channels exhibit high on-state current ($>$ 1 mA/$μ$m) and low subthreshold swing (minimum of 63 mV/dec) without the need for post-deposition processing. This approach offers a general, scalable pathway for transistors with atomic layer deposited channel materials, enabling the future of low-power three-dimensional electronics.

cond-mat.mtrl-sci↗

arXiv2025

Accelerated Spatio-Temporal Bayesian Modeling for Multivariate Gaussian Processes

Multivariate Gaussian processes (GPs) offer a powerful probabilistic framework to represent complex interdependent phenomena. They pose, however, significant computational challenges in high-dimensional settings, which frequently arise in spatial-temporal applications. We present DALIA, a highly scalable framework for performing Bayesian inference tasks on spatio-temporal multivariate GPs, based on the methodology of integrated nested Laplace approximations. Our approach relies on a sparse inverse covariance matrix formulation of the GP, puts forward a GPU-accelerated block-dense approach, and introduces a hierarchical, triple-layer, distributed memory parallel scheme. We showcase weak scaling performance surpassing the state-of-the-art by two orders of magnitude on a model whose parameter space is 8$\times$ larger and measure strong scaling speedups of three orders of magnitude when running on 496 GH200 superchips on the Alps supercomputer. Applying DALIA to air pollution data from northern Italy over 48 days, we showcase refined spatial resolutions over the aggregated pollutant measurements.

stat.CO↗