arXiv subjects

A. Ustyuzhanin

Publications and source records attributed to A. Ustyuzhanin.

9 recordsLinked to original sources

arXiv2026

Machine Learning the Strong Disorder Renormalization Group Method for Disordered Quantum Spin Chains

We train machine learning algorithms to infer the entanglement structure of disordered long-range interacting quantum spin chains by learning from the strong disorder renormalisation group (SDRG) method. The system consists of $S=1/2$-quantum spins coupled by antiferromagnetic power-law interactions with decay exponent $α$ at random positions on a one-dimensional chain. Using SDRG as a physics-informed teacher, we compare a Random Forest classifier as a classical baseline with a graph neural network (GNN) that operates directly on the interaction graph and learns a bond-ranking rule mirroring the SDRG decimation policy. The GNN achieves a disorder-averaged pairing accuracy close to one and reproduces the entanglement entropy $S(\ell)$ in excellent quantitative agreement with SDRG across all subsystem sizes and interaction exponents. RG flow heat maps confirm that the GNN learns the sequential decimation hierarchy rather than merely fitting final-state observables. Finite-temperature entanglement properties are incorporated via the SDRGX framework through a two-stage strategy, using the zero-temperature GNN to generate the RG flow and sampling thermal occupations from the canonical ensemble, yielding results in agreement with both numerical SDRGX and analytical predictions without retraining.

cond-mat.dis-nn↗

arXiv2025

Inverse Design of Broadband Antennas for Terahertz Devices Based on 2D Materials

Terahertz (THz) technology, a cornerstone of next-generation high-speed communication and sensing, has long been hindered by impedance mismatch challenges that limit device performance and applicability. These challenges become particularly pronounced when ultrasensitive two-dimensional (2D) materials are employed as the device substrate in the THz range, further complicating their integration into real-world applications. Furthermore, conventional antenna designs often fail to provide adequate matching across the extensive THz spectrum. In this work, we tackle these challenges using a procedural generation algorithm to design THz broadband antennas that satisfy specific performance criteria. Namely, the developed inverse design methodology enables customization for the target impedance value, bandwidth, and contact topology requirements. The proposed antenna achieves an improvement of up to 40\% in power transfer efficiency compared to traditional bow-tie antennas under realistic operating conditions. High-fidelity electromagnetic simulations validate these results, confirming the design's practicality for THz applications. This work addresses critical limitations of existing antenna designs and advances the feasibility of high-frequency applications in both communication and sensing.

physics.app-ph↗

arXiv2024

Local-time formula for dissipation in solid ionic electrolytes

When ions move through solids, they interact with the solid's constituent atoms and cause them to vibrate around their equilibrium points. This vibration, in turn, modifies the potential landscape through which the mobile ions travel. Because the present-time potential depends on past interactions, the coupling is inherently non-local in time, making its numerical and analytical treatment challenging. For sufficiently slow-moving ions, we linearize the phonon spectrum to show that these non-local effects can be ignored, giving rise to a drag-like force. Unlike the more familiar drag coefficient in liquids, the drag takes on a matrix form due to the crystalline structure of the framework. We numerically simulate trajectories and dissipation rates using both the time-local and non-local formulas to validate our simplification. The time-local formula dramatically reduces the computational cost of calculating the motion of mobile particle through a crystalline framework and clearly connects the properties of the material to the drag experienced by the particle.

cond-mat.mtrl-sci↗

arXiv2023

Directional Sensitivity of the NEWSdm Experiment to Cosmic Ray Boosted Dark Matter

We present a study of a directional search for Dark Matter boosted forward when scattered by cosmic-ray nuclei, using a module of the NEWSdm experiment. The boosted Dark Matter flux at the edge of the Earth's atmosphere is expected to be pointing to the Galactic Center, with a flux 15 to 20 times larger than in the transverse direction. The module of the NEWSdm experiment consists of a 10 kg stack of Nano Imaging Trackers, i.e.~newly developed nuclear emulsions with AgBr crystal sizes down to a few tens of nanometers. The module is installed on an equatorial telescope. The relatively long recoil tracks induced by boosted Dark Matter, combined with the nanometric granularity of the emulsion, result in an extremely low background. This makes an installation at the INFN Gran Sasso laboratory, both on the surface and underground, viable. A comparison between the two locations is made. The angular distribution of nuclear recoils induced by boosted Dark Matter in the emulsion films at the surface laboratory is expected to show an excess with a factor of 3.5 in the direction of the Galactic Center. This excess allows for a Dark Matter search with directional sensitivity. The surface laboratory configuration prevents the deterioration of the signal in the rock overburden and it emerges as the most powerful approach for a directional observation of boosted Dark Matter with high sensitivity. We show that, with this approach, a 10 kg module of the NEWSdm experiment exposed for one year at the Gran Sasso surface laboratory can probe Dark Matter masses between 1 keV/c$^2$ and 1 GeV/c$^2$ and cross-section values down to $10^{-30}$~cm$^2$ with a directional sensitive search.

astro-ph.IM↗

arXiv2021

Learning velocity model for complex media with deep convolutional neural networks

The paper considers the problem of velocity model acquisition for a complex media based on boundary measurements. The acoustic model is used to describe the media. We used an open-source dataset of velocity distributions to compare the presented results with the previous works directly. Forward modeling is performed using the grid-characteristic numerical method. The inverse problem is solved using deep convolutional neural networks. Modifications for a baseline UNet architecture are proposed to improve both structural similarity index measure quantitative correspondence of the velocity profiles with the ground truth. We evaluate our enhancements and demonstrate the statistical significance of the results.

cs.LG↗

arXiv2020

Allen: A high level trigger on GPUs for LHCb

We describe a fully GPU-based implementation of the first level trigger for the upgrade of the LHCb detector, due to start data taking in 2021. We demonstrate that our implementation, named Allen, can process the 40 Tbit/s data rate of the upgraded LHCb detector and perform a wide variety of pattern recognition tasks. These include finding the trajectories of charged particles, finding proton-proton collision points, identifying particles as hadrons or muons, and finding the displaced decay vertices of long-lived particles. We further demonstrate that Allen can be implemented in around 500 scientific or consumer GPU cards, that it is not I/O bound, and can be operated at the full LHC collision rate of 30 MHz. Allen is the first complete high-throughput GPU trigger proposed for a HEP experiment.

physics.ins-det↗

arXiv2018

Discovery potential for directional Dark Matter detection with nuclear emulsions

Direct Dark Matter searches are nowadays one of the most fervid research topics with many experimental efforts devoted to the search for nuclear recoils induced by the scattering of Weakly Interactive Massive Particles (WIMPs). Detectors able to reconstruct the direction of the nucleus recoiling against the scattering WIMP are opening a new frontier to possibly extend Dark Matter searches beyond the neutrino background. Exploiting directionality would also prove the galactic origin of Dark Matter with an unambiguous signal-to-background separation. Indeed, the angular distribution of recoiled nuclei is centered around the direction of the Cygnus constellation, while the background distribution is expected to be isotropic. Current directional experiments are based on gas TPC whose sensitivity is limited by the small achievable detector mass. In this paper we present the discovery potential of a directional experiment based on the use of a solid target made of newly developed nuclear emulsions and of optical read-out systems reaching unprecedented nanometric resolution.

astro-ph.CO↗

arXiv2017

Deep learning for inferring cause of data anomalies

Daily operation of a large-scale experiment is a resource consuming task, particularly from perspectives of routine data quality monitoring. Typically, data comes from different sub-detectors and the global quality of data depends on the combinatorial performance of each of them. In this paper, the problem of identifying channels in which anomalies occurred is considered. We introduce a generic deep learning model and prove that, under reasonable assumptions, the model learns to identify 'channels' which are affected by an anomaly. Such model could be used for data quality manager cross-check and assistance and identifying good channels in anomalous data samples. The main novelty of the method is that the model does not require ground truth labels for each channel, only global flag is used. This effectively distinguishes the model from classical classification methods. Being applied to CMS data collected in the year 2010, this approach proves its ability to decompose anomaly by separate channels.

physics.data-an↗

arXiv2017

LHCb trigger streams optimization

The LHCb experiment stores around $10^{11}$ collision events per year. A typical physics analysis deals with a final sample of up to $10^7$ events. Event preselection algorithms (lines) are used for data reduction. Since the data are stored in a format that requires sequential access, the lines are grouped into several output file streams, in order to increase the efficiency of user analysis jobs that read these data. The scheme efficiency heavily depends on the stream composition. By putting similar lines together and balancing the stream sizes it is possible to reduce the overhead. We present a method for finding an optimal stream composition. The method is applied to a part of the LHCb data (Turbo stream) on the stage where it is prepared for user physics analysis. This results in an expected improvement of 15% in the speed of user analysis jobs, and will be applied on data to be recorded in 2017.

cs.DC↗