-
Adiabatic Fine-Tuning of Neural Quantum States Enables Detection of Phase Transitions in Weight Space
Authors:
Vinicius Hernandes,
Thomas Spriggs,
Saqar Khaleefah,
Eliska Greplova
Abstract:
Neural quantum states (NQS) have emerged as a powerful tool for approximating quantum wavefunctions using deep learning. While these models achieve remarkable accuracy, understanding how they encode physical information remains an open challenge. In this work, we introduce adiabatic fine-tuning, a scheme that trains NQS across a phase diagram, leading to strongly correlated weight representations…
▽ More
Neural quantum states (NQS) have emerged as a powerful tool for approximating quantum wavefunctions using deep learning. While these models achieve remarkable accuracy, understanding how they encode physical information remains an open challenge. In this work, we introduce adiabatic fine-tuning, a scheme that trains NQS across a phase diagram, leading to strongly correlated weight representations across different models. This correlation in weight space enables the detection of phase transitions in quantum systems by analyzing the trained network weights alone. We validate our approach on the transverse field Ising model and the J1-J2 Heisenberg model, demonstrating that phase transitions manifest as distinct structures in weight space. Our results establish a connection between physical phase transitions and the geometry of neural network parameters, opening new directions for the interpretability of machine learning models in physics.
△ Less
Submitted 18 April, 2025; v1 submitted 21 March, 2025;
originally announced March 2025.
-
Exploring Biological Neuronal Correlations with Quantum Generative Models
Authors:
Vinicius Hernandes,
Eliska Greplova
Abstract:
Understanding of how biological neural networks process information is one of the biggest open scientific questions of our time. Advances in machine learning and artificial neural networks have enabled the modeling of neuronal behavior, but classical models often require a large number of parameters, complicating interpretability. Quantum computing offers an alternative approach through quantum ma…
▽ More
Understanding of how biological neural networks process information is one of the biggest open scientific questions of our time. Advances in machine learning and artificial neural networks have enabled the modeling of neuronal behavior, but classical models often require a large number of parameters, complicating interpretability. Quantum computing offers an alternative approach through quantum machine learning, which can achieve efficient training with fewer parameters. In this work, we introduce a quantum generative model framework for generating synthetic data that captures the spatial and temporal correlations of biological neuronal activity. Our model demonstrates the ability to achieve reliable outcomes with fewer trainable parameters compared to classical methods. These findings highlight the potential of quantum generative models to provide new tools for modeling and understanding neuronal behavior, offering a promising avenue for future research in neuroscience.
△ Less
Submitted 13 September, 2024;
originally announced September 2024.
-
QDsim: A user-friendly toolbox for simulating large-scale quantum dot devices
Authors:
Valentina Gualtieri,
Charles Renshaw-Whitman,
Vinicius Hernandes,
Eliska Greplova
Abstract:
We introduce QDsim, a python package tailored for the rapid generation of charge stability diagrams in large-scale quantum dot devices, extending beyond traditional double or triple dots. QDsim is founded on the constant interaction model from which we rephrase the task of finding the lowest energy charge configuration as a convex optimization problem. Therefore, we can leverage the existing packa…
▽ More
We introduce QDsim, a python package tailored for the rapid generation of charge stability diagrams in large-scale quantum dot devices, extending beyond traditional double or triple dots. QDsim is founded on the constant interaction model from which we rephrase the task of finding the lowest energy charge configuration as a convex optimization problem. Therefore, we can leverage the existing package CVXPY, in combination with an appropriate powerful solver, for the convex optimization which streamlines the creation of stability diagrams and polytopes. Through multiple examples, we demonstrate how QDsim enables the generation of large-scale dataset that can serve a basis for the training of machine-learning models for automated tuning algorithms. While the package currently does not support quantum effects beyond the constant interaction model, QDsim is a tool that directly addresses the critical need for cost-effective and expeditious data acquisition for better tuning algorithms in order to accelerate the development of semiconductor quantum devices.
△ Less
Submitted 2 August, 2024; v1 submitted 3 April, 2024;
originally announced April 2024.
-
Phase classification using neural networks: application to supercooled, polymorphic core-softened mixtures
Authors:
Vinicius F. Hernandes,
Murilo S. Marques,
José R. Bordin
Abstract:
Characterization of phases of soft matter systems is a challenge faced in many physicochemical problems. For polymorphic fluids it is an even greater challenge. Specifically, glass forming fluids, as water, can have, besides solid polymorphism, more than one liquid and glassy phases, and even a liquid-liquid critical point. In this sense, we apply a neural network (NN) algorithm to analyze the pha…
▽ More
Characterization of phases of soft matter systems is a challenge faced in many physicochemical problems. For polymorphic fluids it is an even greater challenge. Specifically, glass forming fluids, as water, can have, besides solid polymorphism, more than one liquid and glassy phases, and even a liquid-liquid critical point. In this sense, we apply a neural network (NN) algorithm to analyze the phase behavior of a core-softened mixture of core-softened CSW fluids that have liquid polymorphism and liquid-liquid critical points, similar to water. We also apply the NN to mixtures of CSW fluids and core-softened alcohols models. We combine and expand two methods based on bond-orientational order parameters to study mixtures, applied to mixtures of hardcore fluids by Boattini and co-authors [Molecular Physics 116, 3066-3075 (2018)] and to supercooled water by Martelli and co-authors [The Journal of Chemical Physics 153, 104503 (2020)], to include longer range coordination shells. With this, the trained neural network (NN) was able to properly predict the crystalline solid phases, the fluid phases and the amorphous phase for the pure CSW and CSW-alcohols mixtures with high efficiency. More than this, information about the phase populations, obtained from the NN approach, can help verify if the phase transition is continuous or discontinuous, and also to interpret how the metastable amorphous region spreads along the stable high density fluid phase. These findings help to understand the behavior of supercooled polymorphic fluids and extend the comprehension of how amphiphilic solutes affect the phases behavior.
△ Less
Submitted 11 October, 2021; v1 submitted 24 June, 2021;
originally announced June 2021.
-
Core-softened water-alcohol mixtures: the solute-size effects
Authors:
Murilo S. Marques,
Vinicius F. Hernandes,
Jose Rafael Bordin
Abstract:
In a recent work [\textit{J. Mol. Liq}, 2020, \textbf{320}, 114420], using molecular dynamics simulations and a core-softened potential approach, we have shown that adding a simple solute as methanol can "kill" the density anomalous behavior as the LLCP is suppressed by the spontaneous crystallization in a hexagonal closed packing (HCP) crystal near the LLPT. Now, we extend this work in order to r…
▽ More
In a recent work [\textit{J. Mol. Liq}, 2020, \textbf{320}, 114420], using molecular dynamics simulations and a core-softened potential approach, we have shown that adding a simple solute as methanol can "kill" the density anomalous behavior as the LLCP is suppressed by the spontaneous crystallization in a hexagonal closed packing (HCP) crystal near the LLPT. Now, we extend this work in order to realize how longer-chain alcohols will affect the complex behavior of water-alcohol mixtures in the supercooled regime. Besides core-softened (CS) methanol, ethanol and 1-propanol were added to a system of identical particles that interact through the continuous shouldered well (CSW) potential. We observed that the density anomaly gradually decreases its extension in phase diagrams until disappearing with the growth of the non-polar chain and the alcohol concentration, differently from the liquid-liquid phase transition (and the LLCP), which remained present in all analyzed mixtures, in according to \textit{Nature}, 2001, \textbf{409}, 692. For our model, the longer non-polar chains and higher concentrations gradually impact the competition between the scales in the CS potential, leading to a gradual disappearing of the anomalies until the TMD total disappearance is observed when the first coordination shell structure is also affected: the short-range ordering is favored, leading to less competition between short- and long-range ordering and, consequently, to the extinction of anomalies. Also, the non-polar chain size and concentration have an effect on the solid phases, favoring the hexagonal closed packed (HCP) solid and the amorphous solid phase over the body-centered cubic (BCC) crystal.
△ Less
Submitted 24 June, 2021; v1 submitted 18 February, 2021;
originally announced February 2021.
-
Competing interactions near the liquid-liquid phase transition of core-softened water/methanol mixtures
Authors:
Murilo S. Marques,
Vinicius Fonseca Hernandes,
Enrique Lomba,
J. Rafael Bordin
Abstract:
Water is an unique material with a long list of thermodynamic, dynamic and structural anomalies, which are usually attributed to the competition between two characteristic length scales in the intermolecular interaction. It has been argued that a potential liquid-liquid phase transition (LLPT) ending at a liquid-liquid critical point (LLCP) lies at the core of the anomalous behavior of water. This…
▽ More
Water is an unique material with a long list of thermodynamic, dynamic and structural anomalies, which are usually attributed to the competition between two characteristic length scales in the intermolecular interaction. It has been argued that a potential liquid-liquid phase transition (LLPT) ending at a liquid-liquid critical point (LLCP) lies at the core of the anomalous behavior of water. This transition which has been evidenced in multiple simulation studies seems to be preempted experimentally by spontaneous crystallization. Here, in order to expose the connection between the spontaneous crystallization observed in the supercooled regime in the vicinity of the LLPT, and the density anomaly, we perform extensive Molecular Dynamics simulations of a model mixture of core-softened water and methanol. The pure water-like fluid exhibits a LLPT and a density anomaly. In contrast, our pure methanol-like model does have a LLPT but lacks the density anomaly. Our results illustrate the relation between the vanishing of the density anomaly and an increase in the temperature of the spontaneous crystallization: once this temperature surpasses the LLCP critical temperature, no density anomaly is observed. This peculiar feature illustrates how fine tuning the competitive interactions determine the anomalous behavior of water/alcohol mixtures.
△ Less
Submitted 20 August, 2020;
originally announced August 2020.
-
Structural behavior of a two length scale core-softened fluid in two dimensions
Authors:
Daniel Souza Cardoso,
Vinicius Fonseca Hernandes,
Thiago Pulcinelli Orlandi Nogueira,
José Rafael Bordin
Abstract:
2D competitive systems shows a large variety of solid and aggregate patterns, re-entrant fluid phase, a continuous melting as predicted by the KTHNY theory and, in some cases, waterlike anomalies. In this paper, we employ Langevin Dynamics simulations of a ramp-like core-softened fluid that have two characteristic length scales and all the features mentioned above. Analyzing the fluid phase of thi…
▽ More
2D competitive systems shows a large variety of solid and aggregate patterns, re-entrant fluid phase, a continuous melting as predicted by the KTHNY theory and, in some cases, waterlike anomalies. In this paper, we employ Langevin Dynamics simulations of a ramp-like core-softened fluid that have two characteristic length scales and all the features mentioned above. Analyzing the fluid phase of this system, Bordin and Barbosa [Phys. Rev. E 97, 022604 (2018)] reported the existence of two waterlike anomalous region. Now, we revisit this problem focusing in the low temperature regime and a larger range of densities looking for a relation between the origin of the anomalous behavior, the solid phases and the re-entrant melting. Now, not two, but three anomalous regions were observed. The extra anomalous regions are related to the re-entrant melting of a amorphous solid to a cluster fluid. They are ruled not only by the competition between the characteristic length scales in the potential, but also by extra competitions induced by the variety of particles conformations. These competitions extends from the solid to the fluid phase, reflecting in the structural waterlike anomaly. Our results shade some light in the complex behavior of two length scales competitive potential, and helps to elucidate the relation between the large number of solid phases and the existence of more than one waterlike anomalies region in these systems.
△ Less
Submitted 15 August, 2020;
originally announced August 2020.
