Disordered Systems and Neural Networks
See recent articles
- [1] arXiv:2409.08288 [pdf, html, other]
-
Title: Van der Waals interactions in linear chains with long-range correlated disorderComments: 5 pages, 4 figuresSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Optics (physics.optics)
We present a theoretical study of van der Waals interaction forces in disordered linear molecule chains. We demonstrate that the interaction energy strongly and nonmonotonously depends on the disorder correlation length. Semianalytical expressions for the interaction energy are obtained.
- [2] arXiv:2409.08693 [pdf, html, other]
-
Title: Boolean mean field spin glass model: rigorous resultsSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn)
Spin glasses have played a fundamental role in statistical mechanics field. Purpose of this work is to analyze a variation on theme of the mean field case of them, when the Ising spins are replaced to Boolean ones, i.e. {0,1} possible values. This may be useful to continue building a solid bridge between statical mechanics of spin glasses and Machine Learning techniques. We have drawn a detailed framework of this model: we have applied Guerra and Toninelli's approach to prove the existence of the thermodynamic quenched statistical pressure for this model recovering its expression using Guerra's interpolation. Specifically, we have supposed Replica Symmetric assumption and first step of Replica Symmetry Breaking approximation for the probability distribution of the order parameter of the model. Then, we analyze the stability of the resolution in both assumptions via de Almeida-Thouless line, proving that the Replica Symmetric one well describes the model apart for small values of temperature, when the Replica Symmetry Breaking is better. All the theoretical parts are supported by numerical techniques that demonstrate perfect consistency with the analytical results.
New submissions for Monday, 16 September 2024 (showing 2 of 2 entries )
- [3] arXiv:2409.08320 (cross-list from quant-ph) [pdf, html, other]
-
Title: Hilbert space fragmentation at the origin of disorder-free localization in the lattice Schwinger modelComments: 13 pages, 9 figures + supplementary informationSubjects: Quantum Physics (quant-ph); Disordered Systems and Neural Networks (cond-mat.dis-nn); Quantum Gases (cond-mat.quant-gas); Strongly Correlated Electrons (cond-mat.str-el); High Energy Physics - Lattice (hep-lat)
Lattice gauge theories, the discretized cousins of continuum gauge theories, have become an important platform for the exploration of non-equilibrium phenomena beyond their original scope in the Standard Model. In particular, recent works have reported the possibility of disorder-free localization in the lattice Schwinger model. Using degenerate perturbation theory and numerical simulations based on exact diagonalization and matrix product states, we perform a detailed characterization of thermalization breakdown in the Schwinger model including its spectral properties, the structure of eigenstates, and out-of-equilibrium quench dynamics. We scrutinize the strong-coupling limit of the model, in which an intriguing, double-logarithmic-in-time, growth of entanglement was previously proposed from the initial vacuum state. We identify the origin of this ultraslow growth of entanglement as due to an approximate Hilbert space fragmentation and the emergence of a dynamical constraint on particle hopping, which gives rise to sharp jumps in the entanglement entropy dynamics within individual background charge sectors. Based on the statistics of jump times, we argue that the entanglement growth, averaged over charge sectors, is more naturally explained as either single-logarithmic or a weak power law in time. Our results thus suggest the existence of a single ergodicity-breaking regime due to Hilbert space fragmentation, whose properties are reminiscent of conventional many-body localization within the numerically accessible system sizes.
- [4] arXiv:2409.08352 (cross-list from cond-mat.stat-mech) [pdf, html, other]
-
Title: Learning Generalized Statistical Mechanics with Matrix Product StatesComments: 6 pages, 6 figures + Supplementary material (5 pages, 5 figures)Subjects: Statistical Mechanics (cond-mat.stat-mech); Disordered Systems and Neural Networks (cond-mat.dis-nn); Quantum Physics (quant-ph)
We introduce a variational algorithm based on Matrix Product States that is trained by minimizing a generalized free energy defined using Tsallis entropy instead of the standard Gibbs entropy. As a result, our model can generate the probability distributions associated with generalized statistical mechanics. The resulting model can be efficiently trained, since the resulting free energy and its gradient can be calculated exactly through tensor network contractions, as opposed to standard methods which require estimating the Gibbs entropy by sampling. We devise a variational annealing scheme by ramping up the inverse temperature, which allows us to train the model while avoiding getting trapped in local minima. We show the validity of our approach in Ising spin-glass problems by comparing it to exact numerical results and quasi-exact analytical approximations. Our work opens up new possibilities for studying generalized statistical physics and solving combinatorial optimization problems with tensor networks.
- [5] arXiv:2409.08899 (cross-list from cond-mat.stat-mech) [pdf, html, other]
-
Title: Inferring hyperuniformity from local structures via persistent homologyComments: 18 pages, 10 figuresSubjects: Statistical Mechanics (cond-mat.stat-mech); Disordered Systems and Neural Networks (cond-mat.dis-nn); Soft Condensed Matter (cond-mat.soft); Algebraic Topology (math.AT)
Hyperuniformity refers to the suppression of density fluctuations at large scales. Typical for ordered systems, this property also emerges in several disordered physical and biological systems, where it is particularly relevant to understand mechanisms of pattern formation and to exploit peculiar attributes, e.g., interaction with light and transport phenomena. While hyperuniformity is a global property, it has been shown in [Phys. Rev. Research 6, 023107 (2024)] that global hyperuniform characteristics systematically correlate with topological properties representative of local arrangements. In this work, building on this information, we explore and assess the inverse relationship between hyperuniformity and local structures in point distributions as described by persistent homology. Standard machine learning algorithms trained on persistence diagrams are shown to detect hyperuniformity with high accuracy. Therefore, we demonstrate that the information on patterns' local structure allows for inferring hyperuniformity. Then, addressing more quantitative aspects, we show that parameters defining hyperuniformity globally, for instance entering the structure factor, can be reconstructed by comparing persistence diagrams of targeted patterns with reference ones. We also explore the generation of patterns entailing given topological properties. The results of this study pave the way for advanced analysis of hyperuniform patterns including local information, and introduce basic concepts for their inverse design.
- [6] arXiv:2409.08940 (cross-list from cond-mat.mtrl-sci) [pdf, other]
-
Title: Representing Born effective charges with equivariant graph convolutional neural networksSubjects: Materials Science (cond-mat.mtrl-sci); Disordered Systems and Neural Networks (cond-mat.dis-nn)
Graph convolutional neural networks have been instrumental in machine learning of material properties. When representing tensorial properties, weights and descriptors of a physics-informed network must obey certain transformation rules to ensure the independence of the property on the choice of the reference frame. Here we explicitly encode such properties using an equivariant graph convolutional neural network. The network respects rotational symmetries of the crystal throughout by using equivariant weights and descriptors and provides a tensorial output of the target value. Applications to tensors of atomic Born effective charges in diverse materials including perovskite oxides, Li3PO4, and ZrO2, are demonstrated, and good performance and generalization ability is obtained.
- [7] arXiv:2409.08986 (cross-list from physics.chem-ph) [pdf, html, other]
-
Title: The connection of polaritonic chemistry with the physics of a spin glassSubjects: Chemical Physics (physics.chem-ph); Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech); Quantum Physics (quant-ph)
Polaritonic chemistry has garnered increasing attention in recent years due to pioneering experimental results, which show that site- and bond-selective chemistry at room temperature is achievable through strong collective coupling to field fluctuations in optical cavities. Despite these notable experimental strides, the underlying theoretical mechanisms remain unclear. In this focus review, we highlight a fundamental theoretical link between the seemingly unrelated fields of polaritonic chemistry and spin glasses, exploring its profound implications for the theoretical framework of polaritonic chemistry. Specifically, we present a mapping of the dressed electronic structure problem under collective vibrational strong coupling to the iconic Sherrington-Kirkpatrick model of spin glasses. This mapping uncovers a collectively induced instability in the dressed electronic structure (spontaneous replica symmetry breaking), which could provide the long-sought seed for significant local chemical modifications in polaritonic chemistry. This mapping paves the way to incorporate, adjust and probe numerous spin glass concepts in polaritonic chemistry, such as frustration, aging dynamics, excess of thermal fluctuations, time-reversal symmetry breaking or stochastic resonances. Ultimately, the mapping also offers fresh insights into the applicability of spin glass theory beyond condensed matter systems and it suggests novel theoretical directions such as polarization glasses with explicitly time-dependent order parameter functions.
Cross submissions for Monday, 16 September 2024 (showing 5 of 5 entries )
- [8] arXiv:2212.09775 (replaced) [pdf, html, other]
-
Title: Many-body-localization protection of eigenstate topological order in two dimensionsComments: 9 pages, 6 figures; v2: accepted manuscriptSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Strongly Correlated Electrons (cond-mat.str-el); Quantum Physics (quant-ph)
Many-body localization (MBL) has been proposed to enable and protect topological order in all eigenstates, vastly expanding the traditional ground-state setting. However, for the most intriguing case of two-dimensional (2D) systems with anyons and topology-dependent degeneracies, the dense many-body spectrum challenges studying this MBL protection numerically. Here we use large-scale full-spectrum variational ansätze to demonstrate MBL-protected topological order in the disordered 2D toric code perturbed by magnetic fields. We show that the system has topological local integrals of motion (tLIOMs) for magnetic field strengths below $h_c\approx0.1$ times the toric code coupling scale. Combining tLIOMs with exact diagonalization, we also identify high-energy topological multiplets in the dense many-body spectrum. The phase diagram we find is consistent with toric-code and trivial MBL phases being separated by an intervening thermal phase.
- [9] arXiv:2406.04183 (replaced) [pdf, html, other]
-
Title: Manipulating the Relaxation Time of Boundary-Dissipative Systems through Bond DissipationJournal-ref: Phys. Rev. B 110, 104305 (2024)Subjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Other Condensed Matter (cond-mat.other); Quantum Physics (quant-ph)
Relaxation time plays a crucial role in describing the relaxation processes of quantum systems. We study the effect of a type of bond dissipation on the relaxation time of boundary dissipative systems and find that it can change the scaling of the relaxation time $T_c\sim L^{z}$ from $z=3$ to a value significantly less than $3$. We further reveal that the reason such bond dissipation can significantly reduce the relaxation time is that it can selectively target specific states. For Anderson localized systems, the scaling behavior of the relaxation time changes from an exponential form to a power-law form as the system size varies. This is because the bond dissipation we consider can not only select specific states but also disrupt the localization properties. Our work reveals that in open systems, one type of dissipation can be used to regulate the effects produced by another type of dissipation.
- [10] arXiv:2404.09914 (replaced) [pdf, html, other]
-
Title: Normalizing flows as an enhanced sampling method for atomistic supercooled liquidsJournal-ref: Mach. Learn.: Sci. Technol. 5, 035053 (2024)Subjects: Soft Condensed Matter (cond-mat.soft); Disordered Systems and Neural Networks (cond-mat.dis-nn); Computational Physics (physics.comp-ph)
Normalizing flows can transform a simple prior probability distribution into a more complex target distribution. Here, we evaluate the ability and efficiency of generative machine learning methods to sample the Boltzmann distribution of an atomistic model for glass-forming liquids. This is a notoriously difficult task, as it amounts to ergodically exploring the complex free energy landscape of a disordered and frustrated many-body system. We optimize a normalizing flow model to successfully transform high-temperature configurations of a dense liquid into low-temperature ones, near the glass transition. We perform a detailed comparative analysis with established enhanced sampling techniques developed in the physics literature to assess and rank the performance of normalizing flows against state-of-the-art algorithms. We demonstrate that machine learning methods are very promising, showing a large speedup over conventional molecular dynamics. Normalizing flows show performances comparable to parallel tempering and population annealing, while still falling far behind the swap Monte Carlo algorithm. Our study highlights the potential of generative machine learning models in scientific computing for complex systems, but also points to some of its current limitations and the need for further improvement.