arXiv subjects

Mathieu Taillefumier

Publications and source records attributed to Mathieu Taillefumier.

12 recordsLinked to original sources

arXiv2025

The CP2K Program Package Made Simple

CP2K is a versatile open-source software package for simulations across a wide range of atomistic systems, from isolated molecules in the gas phase to low-dimensional functional materials and interfaces, as well as highly symmetric crystalline solids, disordered amorphous glasses, and weakly interacting soft-matter systems in the liquid state and in solution. This review highlights CP2K's capabilities for computing both static and dynamical properties using quantum-mechanical and classical simulation methods. In contrast to the accompanying theory and code paper [J. Chem. Phys. 152, 194103 (2020)], the focus here is on the practical usage and applications of CP2K, with underlying theoretical concepts introduced only as needed.

physics.comp-ph↗

arXiv2025

Robustness of deep learning classification to adversarial input on GPUs: asynchronous parallel accumulation is a source of vulnerability

The ability of machine learning (ML) classification models to resist small, targeted input perturbations -- known as adversarial attacks -- is a key measure of their safety and reliability. We show that floating-point non-associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.

cs.LG↗

arXiv2024

Impacts of floating-point non-associativity on reproducibility for HPC and deep learning applications

Run to run variability in parallel programs caused by floating-point non-associativity has been known to significantly affect reproducibility in iterative algorithms, due to accumulating errors. Non-reproducibility can critically affect the efficiency and effectiveness of correctness testing for stochastic programs. Recently, the sensitivity of deep learning training and inference pipelines to floating-point non-associativity has been found to sometimes be extreme. It can prevent certification for commercial applications, accurate assessment of robustness and sensitivity, and bug detection. New approaches in scientific computing applications have coupled deep learning models with high-performance computing, leading to an aggravation of debugging and testing challenges. Here we perform an investigation of the statistical properties of floating-point non-associativity within modern parallel programming models, and analyze performance and productivity impacts of replacing atomic operations with deterministic alternatives on GPUs. We examine the recently-added deterministic options in PyTorch within the context of GPU deployment for deep learning, uncovering and quantifying the impacts of input parameters triggering run to run variability and reporting on the reliability and completeness of the documentation. Finally, we evaluate the strategy of exploiting automatic determinism that could be provided by deterministic hardware, using the Groq accelerator for inference portions of the deep learning pipeline. We demonstrate the benefits that a hardware-based strategy can provide within reproducibility and correctness efforts.

cs.DC↗

arXiv2022

Dynamical scaling as a signature of multiple phase competition in Yb$_2$Ti$_2$O$_7$

$\rm Yb_2Ti_2O_7$ is a celebrated example of a pyrochlore magnet with highly-frustrated, anisotropic exchange interactions. To date, attention has largely focused on its unusual, static properties, many of which can be understood as coming from the competition between different types of magnetic order. Here we use inelastic neutron scattering with exceptionally high energy resolution to explore the dynamical properties of $\rm Yb_2Ti_2O_7$. We find that spin correlations exhibit dynamical scaling, analogous to behavior found near to a quantum critical point. We show that the observed scaling collapse can be explained within a phenomenological theory of multiple--phase competition, and confirm that a scaling collapse is also seen in semi--classical simulations of a microscopic model of $\rm Yb_2Ti_2O_7$. These results suggest a general picture for dynamics in systems with competing ground states.

cond-mat.str-el↗

arXiv2020

CP2K: An Electronic Structure and Molecular Dynamics Software Package -- Quickstep: Efficient and Accurate Electronic Structure Calculations

CP2K is an open source electronic structure and molecular dynamics software package to perform atomistic simulations of solid-state, liquid, molecular and biological systems. It is especially aimed at massively-parallel and linear-scaling electronic structure methods and state-of-the-art ab-initio molecular dynamics simulations. Excellent performance for electronic structure calculations is achieved using novel algorithms implemented for modern high-performance computing systems. This review revisits the main capabilities of CP2k to perform efficient and accurate electronic structure simulations. The emphasis is put on density functional theory and multiple post-Hartree-Fock methods using the Gaussian and plane wave approach and its augmented all-electron extension.

physics.chem-ph↗

arXiv2017

Competing spin liquids and hidden nematic order in spin ice with frustrated transverse exchange

Frustration in magnetic interactions can give rise to disordered ground states with subtle and beautiful properties. The spin ices Ho2Ti2O7 and Dy2Ti2O7 exemplify this phenomenon, displaying a classical spin liquid state, with fractionalized magnetic--monopole excitations. Recently there has been great interest in closely-related "quantum spin ice" materials, following the realization that anisotropic exchange interactions could convert spin ice into a massively-entangled, quantum, spin liquid, where magnetic monopoles become the charges of an emergent quantum electrodynamics. Here we show that even the simplest model of a quantum spin ice, the XXZ model on the pyrochlore lattice, can realise a still-richer scenario. Using a combination of classical Monte Carlo simulation, semi--classical molecular--dynamics simulation, and analytic field theory, we explore the properties of this model for frustrated transverse exchange. We find not one, but three, competing forms of spin liquid, as well as a phase with hidden, spin-nematic, order. We explore the experimental signatures of each of these different states, making explicit predictions for inelastic neutron scattering. These results show an intriguing similarity to experiments on a range of pyrochlore oxides.

cond-mat.str-el↗

arXiv2014

Semi-classical spin dynamics of the antiferromagnetic Heisenberg model on the kagome lattice

We investigate the dynamical properties of the classical antiferromagnetic Heisenberg model on the kagome lattice using a combination of Monte Carlo and molecular dynamics simulations. We find that frustration induces a distribution of timescales in the cooperative paramagnetic regime (i.e. far above the onset of coplanarity), as recently reported experimentally in deuterium jarosite. At lower temperature, when the coplanar correlations are well established, we show that the weath- ervane loop fluctuations control the system relaxation : the time distribution observed at higher temperatures splits into two distinct timescales associated with fluctuations in the plane and out of the plane of coplanarity. The temperature and wave vector dependences of these two components are qualitatively consistent with loops diffusing in the entropically dominated free energy landscape. Numerical results are discussed and compared with the $O(N)$ model and recent experiments for both classical and quantum realizations of the kagome magnets.

cond-mat.str-el↗

arXiv2010

Asymptotically exact trial wave functions for yrast states of rotating Bose gases

We revisit the composite fermion (CF) construction of the lowest angular momentum yrast states of rotating Bose gases with weak short range interaction. For angular momenta at and below the single vortex, $L \leq N$, the overlaps between these trial wave functions and the corresponding exact solutions {\it increase} with increasing system size and appear to approach unity in the thermodynamic limit. In the special case $L=N$, this remarkable behaviour was previously observed numerically. Here we present methods to address this point analytically, and find strongly suggestive evidence in favour of similar behaviour for all $L \leq N$. While not constituting a fully conclusive proof of the converging overlaps, our results do demonstrate a striking similarity between the analytic structure of the exact ground state wave functions at $L \leq N$, and that of their CF counterparts. Results are given for two different projection methods commonly used in the CF approach.

cond-mat.quant-gas↗

arXiv2009

Hall Effect in spinor condensates

We consider a neutral spinor condensate moving in a periodic magnetic field. The spatially dependent magnetic field induces an effective spin dependent Lorentz force which in turn gives rise to a spin dependent Hall effect. Simulations of the Gross-Pitaevskii equation quantify the Hall effect. We discuss possible experimental realizations.

cond-mat.mes-hall↗

arXiv2008

Chiral two-dimensional electron gas in a periodic magnetic field

We study the energy spectrum and electronic properties of two-dimensional electron gas in a periodic magnetic field of zero average with a symmetry of triangular lattice. We demonstrate how the structure of electron energy bands can be changed with the variation of the field strength, so that we can start from nearly free electron gas and then transform it continuously to a system of essentially localized chiral electron states. We find that the electrons near some minima of the effective potential are responsible for occurrence of dissipationless persistent currents creating a lattice of current contours. The topological properties of the electron energy bands are also varied with the intensity of periodic field. We calculated the topological Chern numbers of several lower energy bands as a function of the field. The corresponding Hall conductivity is nonzero and, when the Fermi level lies in the gap, it is quantized.

cond-mat.str-el↗

arXiv2006

Anomalous Hall Effect due to the spin chirality in the Kagomé lattice

We consider a model for a two dimensional electron gas moving on a kagomé lattice and locally coupled to a chiral magnetic texture. We show that the transverse conductivity $σ\_{xy}$ does not vanish even if spin-orbit coupling is not present and it may exhibit unusual behavior. Model parameters are the chirality, the number of conduction electrons and the amplitude of the local coupling. Upon varying these parameters, a topological transition characterized by change of the band Chern numbers occur. As a consequence, $σ\_{xy}$ can be quantized, proportional to the chirality or have a non monotonic behavior upon varying these parameters.

cond-mat.str-el↗

arXiv2002

X-ray Absorption Near-Edge Structure calculations with pseudopotentials. Application to K-edge in diamond and alpha-quartz

We present a reciprocal-space pseudopotential scheme for calculating X-ray absorption near-edge structure (XANES) spectra. The scheme incorporates a recursive method to compute absorption cross section as a continued fraction. The continued fraction formulation of absorption is advantageous in that it permits the treatment of core-hole interaction through large supercells (hundreds of atoms). The method is compared with recently developed Bethe-Salpeter approach. The method is applied to the carbon K-edge in diamond and to the silicon and oxygen K-edges in alpha-quartz for which polarized XANES spectra were measured. Core-hole effects are investigated by varying the size of the supercell, thus leading to information similar to that obtained from cluster size analysis usually performed within multiple scattering calculations.

cond-mat↗