arXiv subjects

Pavel Baláž

Publications and source records attributed to Pavel Baláž.

15 recordsLinked to original sources

arXiv2026

Soft Contrastive Learning for Unsupervised Discovery of Phases of Matter

A machine learning (ML) framework for phase-diagram construction in condensed matter systems with unknown phase structure is presented. The method combines physically motivated descriptors with contrastive representation learning and a two-stage data-generation workflow. First, a sparse set of configurations is generated by Monte Carlo simulated annealing in order to train a contrastive neural network that produces a low-dimensional embedding. Consequently, representative phase configurations are used as initial conditions for gradient-based optimization on a dense parameter grid, enabling efficient generation of refined datasets for high-resolution phase-diagram construction. In order to optimize the workflow, a variant of the method, in which derived descriptors are used as soft labels in the contrastive objective, is examined as well. The framework is demonstrated on a PbZrO3-inspired model potential formulated in terms of coupled order-parameter fields and used as a physically motivated testbed for exploring configuration diversity and phase behavior. The proposed approach provides a physically informed route to ML-assisted phase-diagram construction that integrates data analysis with adaptive dataset refinement.

cond-mat.dis-nn↗

arXiv2026

Learning quantum tomography from incomplete measurements

We revisit quantum tomography in an informationally incomplete scenario and propose improved state reconstruction methods using deep neural networks. In the first approach, the trained network predicts an optimal linear or quadratic reconstructor with coefficients depending only on the collection of (already taken) measurement operators. This effectively refines the undercomplete tomographic reconstructor based on pseudoinverse operation. The second, based on an LSTM recurrent network performs state reconstruction sequentially. It can also optimize the measurement sequence, which suggests a no-free-lunch theorem for tomography: by narrowing the state space, we gain the possibility of more efficient tomography by learning the optimal sequence of measurements. Numerical experiments for a 2-qubit system show that both methods outperform standard maximum likelihood estimation and also scale to larger 3- and 4-qubit systems. Our results demonstrate that neural networks can effectively learn the underlying geometry of multi-qubit states and use it for their reconstruction.

quant-ph↗

arXiv2026

Accelerating the Design of Resorbable Magnesium Alloys: A Machine Learning Approach to Property Prediction

Resorbable magnesium (Mg) alloys are promising candidates for temporary medical devices due to their biodegradability and favorable mechanical properties. To accelerate the design of diluted Mg alloys for implants, we developed a data-driven framework to elucidate the complex relationships between composition, processing, and mechanical properties. The framework screens mechanical properties within biocompatible compositional limits, treating degradation as a design constraint rather than an explicit prediction target. Using a dataset of 410 samples, we trained six different machine learning (ML) models to predict yield strength, ultimate tensile strength, and elongation. Among them, ensemble models, particularly CatBoost, demonstrated high predictive accuracy (R2, YS = 0.950, UTS = 0.916 and El = 0.903). SHapley Additive exPlanation analysis revealed that thermomechanical processing conditions and alloying elements such as Zn, Mn and Gd are the most influential factors governing mechanical behavior in diluted Mg alloys. Validation on the experimental dataset confirmed the models' robustness and generalization capability in capturing process-property relationships. The optimized CatBoost model was further employed to generate predictive property maps visualizing the strength-ductility trade-off as a function of Zn-Mn composition. This work establishes a validated ML framework for rapid in silico screening and targeted design of next-generation resorbable Mg alloys.

cond-mat.mtrl-sci↗

arXiv2025

Learning entanglement from tomography data: contradictory measurement importance for neural networks and random forests

We study the effectiveness of two distinct machine learning techniques, neural networks and random forests, in the quantification of entanglement from two-qubit tomography data. Although we predictably find that neural networks yield better accuracy, we also find that the way that the two methods reach their prediction is starkly different. This is seen by the measurements which arthe most important for the classification. Neural networks follow the intuitive prediction that measurements containing information about non-local coherences are most important for entanglement, but random forests signify the dominance of information contained in occupation measurements. This is because occupation measurements are necessary for the extraction of data about all other density matrix elements from the remaining measurements. The same discrepancy does not occur when the models are used to learn entanglement directly from the elements of the density matrix, so it is the result of the scattering of information and interdependence of measurement data. As a result, the models behave differently when noise is introduced to various measurements, which can be harnessed to obtain more reliable information about entanglement from noisy tomography data.

quant-ph↗

arXiv2025

Unsupervised Machine Learning Phase Classification for Falicov-Kimball Model

We apply various unsupervised machine learning methods for phase classification to investigate the finite-temperature phase diagram of the spinless Falicov-Kimball model in two dimensions. Using only particle occupation snapshots from Monte Carlo simulations as input, each technique, including a straightforward classification based on principal component analysis (PCA), successfully identifies the phase boundary between ordered and disordered phases, independent of the type of phase transition. Remarkably, these techniques also distinguish between the weakly localized and Anderson-localized regimes within the disordered phase, accurately identifying their crossover, which is a challenging task for standard methods. Among the machine learning approaches used, PCA based analysis outperforms more complex methods, such as neural network predictors and autoencoders. These results underscore the effectiveness of simple unsupervised techniques in examining phase transitions and electron localization in complex correlated systems.

cond-mat.str-el↗

arXiv2024

Neural Network Quantum States analysis of the Shastry-Sutherland model

We utilize neural network quantum states (NQS) to investigate the ground state properties of the Heisenberg model on a Shastry-Sutherland lattice using the variational Monte Carlo method. We show that already relatively simple NQSs can be used to approximate the ground state of this model in its different phases and regimes. We first compare several types of NQSs with each other on small lattices and benchmark their variational energies against the exact diagonalization results. We argue that when precision, generality, and computational costs are taken into account, a good choice for addressing larger systems is a shallow restricted Boltzmann machine NQS. We then show that such NQS can describe the main phases of the model in zero magnetic field. Moreover, NQS based on a restricted Boltzmann machine correctly describes the intriguing plateaus forming in magnetization of the model as a function of increasing magnetic field.

cond-mat.dis-nn↗

arXiv2022

Theory of superdiffusive spin transport in noncollinear magnetic multilayers

Ultrafast demagnetization induced by femtosecond laser pulses in thin metallic layers is caused by the outflow of spin-polarized hot electron currents describable by the superdiffusive transport model. These laser-generated spin currents can cross the interface into another magnetic layer and give rise to magnetization dynamics in magnetic spin valves with noncollinear magnetizations. To describe ultrafast transport and spin dynamics in such nanostructures we develop here the superdiffusive theory for general noncollinear magnetic multilayers. Specifically, we introduce an Al/Ni/Ru/Fe/Ru multilayer system with noncollinear Ni and Fe magnetic moments and analyze how the ultrafast demagnetization and spin-transfer torque depend on the noncollinearity. We employ ab initio calculations to compute the spin- and energy-dependent transmissions of hot electrons at the interfaces of the multilayer. Taking into account multiple electron scattering at interfaces and spin mixing in the spacer layer we find that the laser-induced demagnetization of the Ni layer and magnetization change of the Fe layer strongly depend on the angle between their magnetizations. Similarly, the spin-transfer torques on the Ni and Fe layers and the total spin momentum absorbed in the Ni and Fe layer are found to vary markedly with the amount of noncollinearity. These results suggest that changing the amount of noncollinearity in magnetic multilayers one can efficiently control the hot electron spin transport, which may open a way toward achieving fast, laser-driven spintronic devices.

cond-mat.mes-hall↗

arXiv2022

Nonequilibrium dynamics in a spin valve with noncollinear magnetization

We utilize a hybrid quantum-classical equation of motion approach to investigate the spin dynamics and spin-transfer torque in a spin valve under bias voltage. We show that the interplay between localized classical magnetic moments and conduction electrons induces a complex effective exchange coupling between the magnetic layers. This leads to a declination of magnetizations from layers anisotropy axes even in equilibrium. Introducing a finite bias voltage triggers spin currents and related spin-transfer torques which further tilt the magnetizations and govern the relaxation processes of the spin dynamics. Analyzing different scenarios of the applied bias voltage, we show that symmetric and asymmetric voltage drops can lead to relaxation times of the spin dynamics that differ by several orders of magnitude at comparable charge currents. In both cases we observe resonant features, where the relaxation is boosted whenever the chemical potential of the leads matches the maxima in the density of the states of the spin-valve electrons.

cond-mat.mes-hall↗

arXiv2022

Interaction of in-plane magnetic skyrmions with 90$^\circ$ magnetic domain walls: micromagnetic simulations

$90^\circ$ pinned magnetic domain walls can be observed in thin magnetic layers attached to a ferroelectric substrate. The main stabilization mechanism of the noncollinear magnetic texture is the strain transfer which is responsible for imprinting of the ferroelectic domains into the uniaxial anisotropy of the ferromagnet. Here, we investigate by means of micromagnetic simulations how the interfacial Dzyaloshinkii-Moriya interaction influences the $90^\circ$ domain wall structure. It is shown that Dzyaloshinskii-Moriya interaction induces a large out-of-plane magnetization component, strongly dependent on the domain wall type. In particular, it is shown that this out-of-plane magnetization component is crucial for the transport of the in-plane magnetic skyrmions, also known as bimerons, through the magnetic domain walls. Based on the results of micromagnetic simulations, a concept of in-plane magnetic skyrmion valve based on two $90^\circ$ pinned magnetic domain walls is introduced.

cond-mat.mes-hall↗

arXiv2020

Tailoring femtosecond hot-electron pulses for ultrafast spin manipulation

We have measured the hot-electron induced demagnetization of a [Co/Pt]2 multilayer in M(x nm)/Cu(100 nm)/[Co(0.6 nm)/Pt(1.1 nm)]2 samples depending on the nature of the capping layer M and its thickness x. We found out that a Pt layer is more efficient than [Co/Pt]X, Cu or MgO layers in converting IR photon pulses into hot-electron pulses at a given laser power. We also found out that the maximum relative demagnetization amplitude is reached for M(x) = Pt (7 nm). Our experimental results show qualitative agreement with numerical simulations based on the superdiffusive spin transport model. We concluded that the maximum relative demagnetization amplitude, which corresponds to the highest photon conversion into hot-electrons, is an interplay between the IR penetration depth and the hot-electron inelastic mean free path within the capping layer.

cond-mat.mtrl-sci↗

arXiv2019

High-frequency magnon excitation due to femtosecond spin-transfer torques

Femtosecond laser pulses can induce ultrafast demagnetization as well as generate bursts of hot electron spin currents. In trilayer spin valves consisting of two metallic ferromagnetic layers separated by a nonmagnetic one, hot electron spin currents excited by an ultrashort laser pulse propagate from the first ferromagnetic layer through the spacer reaching the second magnetic layer. When the magnetizations of the two magnetic layers are noncollinear, this spin current exerts a torque on magnetic moments in the second ferromagnet. Since this torque is acting only within the sub-ps timescale, it excites coherent high-frequency magnons as recently demonstrated in experiments. Here, we calculate the temporal shape of the hot electron spin currents using the superdiffusive transport model and simulate the response of the magnetic system to the resulting ultrashort spin-transfer torque pulse by means of atomistic spin-dynamics simulations. Our results confirm that the acting spin-current pulse is short enough to excite magnons with frequencies beyond 1 THz, a frequency range out of reach for current induced spin-transfer torques. We demonstrate the formation of thickness dependent standing spin waves during the first picoseconds after laser excitation. In addition, we vary the penetration depth of the spin-transfer torque to reveal its influence on the excited magnons. Our simulations clearly show a suppression effect of magnons with short wavelengths already for penetration depths in the range of 1 nm confirming experimental findings reporting penetration depths below $2\, {\rm nm}$.

cond-mat.mtrl-sci↗

arXiv2019

Domain wall dynamics due to femtosecond laser-induced superdiffusive spin transport

Manipulation of magnetic domain walls via a helicity-independent laser pulse has recently been experimentally demonstrated and various physical mechanisms leading to domain wall dynamics have been discussed. Spin-dependent superdiffusive transport of hot electrons has been identified as one of the possible ways how to affect a magnetic domain wall. Here, we develop a model based on superdiffusive spin-dependent transport to study the laser-induced transport of hot electrons through a smooth magnetic domain wall. We show that the spin transfer between neighboring domains can enhance ultrafast demagnetization in the domain wall. More importantly, our calculations reveal that when the laser pulse is properly focused on to the vicinity of the domain wall, it can excite sufficiently strong spin currents to generate a spin-transfer torque that can rapidly move the magnetic domain wall by several nanometers in several hundreds of femtoseconds, leading to a huge nonequilibrium domain wall velocity.

cond-mat.mes-hall↗

arXiv2018

Static properties and current-induced dynamics of pinned $90$ degree magnetic domain walls under applied fields: an analytic approach

Magnetic domain walls are pinned strongly by abrupt changes in magnetic anisotropy. When driven into oscillation by a spin-polarized current, locally pinned domain walls can be exploited as tunable sources of short-wavelength spin waves. Here, we develop an analytical framework and discrete Heisenberg model to describe the static and dynamic properties of pinned domain walls with a head-to-tail magnetic structure. We focus on magnetic domain walls that are pinned by 90$^\circ$ rotations of uniaxial magnetic anisotropy. Our model captures the domain wall response to a spin-transfer torque that is exerted by an electric current. Model predictions of the domain wall resonance frequency and its evolution with magnetic anisotropy strength and external magnetic field are compared to micromagnetic simulations.

cond-mat.mes-hall↗

arXiv2018

Transport theory for femtosecond laser-induced spin-transfer torques

Ultrafast demagnetization of magnetic layers pumped by a femtosecond laser pulse is accompanied by a nonthermal spin-polarized current of hot electrons. These spin currents are studied here theoretically in a spin valve with noncollinear magnetizations. To this end, we introduce an extended model of superdiffusive spin transport that enables to treat noncollinear magnetic configurations, and apply it to the perpendicular spin valve geometry. We show how spin-transfer torques arise due to this mechanism and calculate their action on the magnetization present, as well as how the latter depends on the thicknesses of the layers and other transport parameters. We demonstrate that there exists a certain optimum thickness of the out-of-plane magnetized spin-current polarizer such that the torque acting on the second magnetic layer is maximal. Moreover, we study the magnetization dynamics excited by the superdiffusive spin-transfer torque due to the flow of hot electrons employing the Landau-Lifshitz-Gilbert equation. Thereby we show that a femtosecond laser pulse applied to one magnetic layer can excite small-angle precessions of the magnetization in the second magnetic layer. We compare our calculations with recent experimental results.

cond-mat.mes-hall↗

arXiv2011

Current-induced dynamics of composite free layer with antiferromagnetic interlayer exchange coupling

Current-induced dynamics in spin valves including composite free layer with antiferromagnetic interlayer exchange coupling is studied theoretically within the diffusive transport regime. We show that current-induced dynamics of a synthetic antiferromagnet is significantly different from dynamics of a synthetic ferrimagnet. From macrospin simulations we obtain conditions for switching the composite free layer, as well as for appearance of various self-sustained dynamical modes. Numerical simulations are compared with simple analytical models of critical current based on linearized Landau-Lifshitz-Gilbert equation.

cond-mat.mes-hall↗