Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2411.18921 · Physical Review Research

Effective temperature in approximate quantum many-body states

Yuqin Chen, Shi‐Xin Zhang

Published Aug 6, 2025• 4 citations
arXivHTMLPDF

Abstract

In the pursuit of numerically identifying ground states of quantum many-body systems, approximate quantum wave function ansatzes are commonly employed. This study focuses on the spectral decomposition of these approximate quantum many-body states into exact eigenstates of the target Hamiltonian, which reflects the intricate physics at the interplay between quantum systems and numerical algorithms. Here, we examine various parametrized approximate quantum states constructed from neural networks, tensor networks, and quantum circuits, employing differentiable programming to numerically approximate ground states and imaginary-time evolved states. Our findings reveal a consistent exponential pattern in the energy eigenbasis decomposition contributions of approximate states across different ansatzes, optimization objectives, and quantum systems, characterized by remarkably small decay rates, i.e., effective temperatures. This finding is counterintuitive for high-fidelity approximate ground states: While the total contribution from excited states can be made sufficiently small, the residual spectral weight does not decay rapidly with energy. This behavior is an intrinsic property of the variational ansatz, independent of the approximation’s overall accuracy. The effective temperature is related to ansatz expressiveness and accuracy and shows phase transition behaviors in learning imaginary-time evolved states. The universal picture and unique features suggest the significance and potential of the effective temperature metric in characterizing approximate quantum states.

DOI
10.1103/zpjv-bm5c

Related papers

Bang-bang algorithms for quantum many-body ground states: A tensor network exploration

Ruoshui Wang, Timothy H. Hsieh, Guifré Vidal

We use matrix product techniques to investigate the performance of two algorithms for obtaining the ground state of a quantum many-body Hamiltonian $H={H}_{A}+{H}_{B}$ in infinite systems. The first algorithm is a generalization of the quantum approximate optimization algorithm and uses a quantum computer to evolve an initial product state into an approximation of the ground state of $H$ by alternating between ${H}_{A}$ and ${H}_{B}$. We show for the one-dimensional (1D) quantum Ising model that the accuracy in representing a gapped ground state improves exponentially with the number of alternations. The second algorithm is the variational imaginary time ansatz, which uses a classical computer to simulate the ground state via alternating imaginary time steps with ${H}_{A}$ and ${H}_{B}$. We find for the 1D quantum Ising model that an accurate approximation to the ground state is obtained with a total imaginary time $\ensuremath{\tau}$ that grows only logarithmically with the inverse energy gap $1/\mathrm{\ensuremath{\Delta}}$ of $H$. This is much faster than imaginary time evolution by $H$, which would require $\ensuremath{\tau}\ensuremath{\sim}1/\mathrm{\ensuremath{\Delta}}$.

arXiv:2208.00271•Physical review. B./Physical review. B•Nov 17, 2022•5 citations

Scalable imaginary time evolution with neural network quantum states

Eimantas Ledinauskas, Egidijus Anisimovas

The representation of a quantum wave function as a neural network quantum state (NQS) provides a powerful variational ansatz for finding the ground states of many-body quantum systems. Nevertheless, due to the complex variational landscape, traditional methods often employ the computation of quantum geometric tensor, consequently complicating optimization techniques. Contributing to efforts aiming to formulate alternative methods, we introduce an approach that bypasses the computation of the metric tensor and instead relies exclusively on first-order gradient descent with Euclidean metric. This allows for the application of larger neural networks and the use of more standard optimization methods from other machine learning domains. Our approach leverages the principle of imaginary time evolution by constructing a target wave function derived from the Schrödinger equation, and then training the neural network to approximate this target. We make this method adaptive and stable by determining the optimal time step and keeping the target fixed until the energy of the NQS decreases. We demonstrate the benefits of our scheme via numerical experiments with 2D J_1-J_2 J 1 − J 2 , Heisenberg model, which showcase enhanced stability and energy accuracy in comparison to direct energy loss minimization. Importantly, our approach displays competitiveness with the well-established density matrix renormalization group method and NQS optimization with stochastic reconfiguration.

arXiv:2307.15521•SciPost Physics•Dec 6, 2023•14 citations

Variational neural and tensor network approximations of thermal states

Sirui Lu, Giacomo Giudice, J. I. Cirac

We introduce a variational Monte Carlo algorithm for approximating finite-temperature quantum many-body systems, based on the minimization of a modified free energy. This approach directly approximates the state at a fixed temperature, allowing for systematic improvement of the expressiveness without accumulating errors from iterative imaginary-time evolution. We employ a variety of trial states—both tensor networks as well as neural networks—as variational for our numerical optimization. We benchmark and compare different constructions in the above classes, both for one- and two-dimensional problems, with systems made of up to N = 100 spins. Our results demonstrate that while restricted Boltzmann machines show limitations, string bond tensor network states exhibit systematic improvements with increasing bond dimensions and the number of strings.

arXiv:2401.14243•Physical review. B./Physical review. B•Feb 3, 2025•11 citations

Low-temperature Gibbs states with tensor networks

Denise Cocchiarella, Mari Carmen Bañuls

We introduce a tensor network method for approximating thermal equilibrium states of quantum many-body systems at low temperatures. Whereas the usual approach starts from infinite temperature and evolves the state in imaginary time (toward lower temperature), our ansatz is constructed from the zero-temperature limit, the ground state, which can be found with a standard tensor network approach. Motivated by properties of the ground state for conformal field theories, our ansatz is especially well suited near criticality. Moreover, it allows an efficient computation of thermodynamic quantities and entanglement properties. We demonstrate the performance of our approach with a tree tensor network ansatz, although it can be extended to other tensor networks, and present results illustrating its effectiveness in capturing the finite-temperature properties in one- and two-dimensional scenarios. In particular, in the critical one-dimensional case we show how the ansatz reproduces the finite temperature scaling of entanglement in a conformal field theory.

arXiv:2501.08300•Physical Review Research•Jul 29, 2025•6 citations

Unusual Energy Spectra of Matrix Product States

J. Maxwell Silvester, Giuseppe Carleo, Steven R. White

In approximate ground states obtained from imaginary-time evolution, the spectrum of the state-its decomposition into exact energy eigenstates-falls off exponentially with the energy. Here we consider the energy spectra of approximate matrix product ground states, such as those obtained with the density matrix renormalization group. Despite the high accuracy of these states, contributions to the spectra are roughly constant out to surprisingly high energy, with an increase in the bond dimension reducing the amplitude but not the extent of these high-energy tails. The unusual spectra appear to be a general feature of compressed wavefunctions, independent of boundary or dimensionality, and are also observed in neural network wavefunctions. The unusual spectra can have a strong effect on sampling-based methods, yielding large fluctuations. The energy variance, which can be used to extrapolate observables to eliminate truncation error, is subject to these large fluctuations when sampled. Nevertheless, we devise a sampling-based variance approach which gives excellent and efficient extrapolations.

arXiv:2408.13616•Physical Review Letters•Mar 26, 2025•6 citations

Scaling Laws for Neural-Network Quantum States

Riccardo Rende, Alessandro Sinibaldi, Luciano Loris Viteritti +3 more

Scaling laws, the power-law relations between loss, architecture size, and compute observed in modern neural networks, offer a quantitative way to characterize the complexity of a learning problem, with the exponent governing the decay of the loss reflecting how rapidly additional resources translate into improved accuracy, and thus how hard the target is to learn. Whether an analogous framework can characterize the complexity of physical problems remains open. We address this question for Neural-Network Quantum States, a leading variational approach for strongly correlated quantum many-body systems. Using transformer wave functions to approximate ground states of the $J_1$-$J_2$ Heisenberg model on triangular and square lattices with up to $20\times 20$ sites, we find that the $V$-score, a measure of accuracy of a variational state, decays as a power law in training compute. Under an appropriate rescaling of compute, results for different system sizes collapse onto a single curve, analogous to scaling collapse in critical phenomena. The resulting power law is, to a good approximation, independent of the number of sites, showing that the transformer Ansatz is size-consistent for the systems considered. The exponent decreases systematically with frustration, identifying it as a quantitative measure of representational difficulty of the ground state and establishing scaling laws as a general framework for benchmarking variational ansätze.

cond-mat.dis-nnCross-listedarXiv:2606.02794•Jun 3, 2026•9 citations
In the pursuit of numerically identifying ground states of quantum many-body systems, approximate quantum wave function ansatzes are commonly employed. This study focuses on the spectral decomposition of these approximate quantum many-body states into exact eigenstates of the target Hamiltonian, which reflects the intricate physics at the interplay between quantum systems and numerical algorithms. Here, we examine various parametrized approximate quantum states constructed from neural networks, tensor networks, and quantum circuits, employing differentiable programming to numerically approximate ground states and imaginary-time evolved states. Our findings reveal a consistent exponential pattern in the energy eigenbasis decomposition contributions of approximate states across different ansatzes, optimization objectives, and quantum systems, characterized by remarkably small decay rates, i.e., effective temperatures. This finding is counterintuitive for high-fidelity approximate ground states: While the total contribution from excited states can be made sufficiently small, the residual spectral weight does not decay rapidly with energy. This behavior is an intrinsic property of the variational ansatz, independent of the approximation’s overall accuracy. The effective temperature is related to ansatz expressiveness and accuracy and shows phase transition behaviors in learning imaginary-time evolved states. The universal picture and unique features suggest the significance and potential of the effective temperature metric in characterizing approximate quantum states.
We use matrix product techniques to investigate the performance of two algorithms for obtaining the ground state of a quantum many-body Hamiltonian H=HA+HBH={H}_{A}+{H}_{B}H=HA​+HB​ in infinite systems. The first algorithm is a generalization of the quantum approximate optimization algorithm and uses a quantum computer to evolve an initial product state into an approximation of the ground state of HHH by alternating between HA{H}_{A}HA​ and HB{H}_{B}HB​. We show for the one-dimensional (1D) quantum Ising model that the accuracy in representing a gapped ground state improves exponentially with the number of alternations. The second algorithm is the variational imaginary time ansatz, which uses a classical computer to simulate the ground state via alternating imaginary time steps with HA{H}_{A}HA​ and HB{H}_{B}HB​. We find for the 1D quantum Ising model that an accurate approximation to the ground state is obtained with a total imaginary time $\ensuremath{\tau}$ that grows only logarithmically with the inverse energy gap $1/\mathrm{\ensuremath{\Delta}}$ of HHH. This is much faster than imaginary time evolution by HHH, which would require $\ensuremath{\tau}\ensuremath{\sim}1/\mathrm{\ensuremath{\Delta}}$.
The representation of a quantum wave function as a neural network quantum state (NQS) provides a powerful variational ansatz for finding the ground states of many-body quantum systems. Nevertheless, due to the complex variational landscape, traditional methods often employ the computation of quantum geometric tensor, consequently complicating optimization techniques. Contributing to efforts aiming to formulate alternative methods, we introduce an approach that bypasses the computation of the metric tensor and instead relies exclusively on first-order gradient descent with Euclidean metric. This allows for the application of larger neural networks and the use of more standard optimization methods from other machine learning domains. Our approach leverages the principle of imaginary time evolution by constructing a target wave function derived from the Schrödinger equation, and then training the neural network to approximate this target. We make this method adaptive and stable by determining the optimal time step and keeping the target fixed until the energy of the NQS decreases. We demonstrate the benefits of our scheme via numerical experiments with 2D J_1-J_2 J 1 − J 2 , Heisenberg model, which showcase enhanced stability and energy accuracy in comparison to direct energy loss minimization. Importantly, our approach displays competitiveness with the well-established density matrix renormalization group method and NQS optimization with stochastic reconfiguration.
We introduce a variational Monte Carlo algorithm for approximating finite-temperature quantum many-body systems, based on the minimization of a modified free energy. This approach directly approximates the state at a fixed temperature, allowing for systematic improvement of the expressiveness without accumulating errors from iterative imaginary-time evolution. We employ a variety of trial states—both tensor networks as well as neural networks—as variational for our numerical optimization. We benchmark and compare different constructions in the above classes, both for one- and two-dimensional problems, with systems made of up to N = 100 spins. Our results demonstrate that while restricted Boltzmann machines show limitations, string bond tensor network states exhibit systematic improvements with increasing bond dimensions and the number of strings.
We introduce a tensor network method for approximating thermal equilibrium states of quantum many-body systems at low temperatures. Whereas the usual approach starts from infinite temperature and evolves the state in imaginary time (toward lower temperature), our ansatz is constructed from the zero-temperature limit, the ground state, which can be found with a standard tensor network approach. Motivated by properties of the ground state for conformal field theories, our ansatz is especially well suited near criticality. Moreover, it allows an efficient computation of thermodynamic quantities and entanglement properties. We demonstrate the performance of our approach with a tree tensor network ansatz, although it can be extended to other tensor networks, and present results illustrating its effectiveness in capturing the finite-temperature properties in one- and two-dimensional scenarios. In particular, in the critical one-dimensional case we show how the ansatz reproduces the finite temperature scaling of entanglement in a conformal field theory.
In approximate ground states obtained from imaginary-time evolution, the spectrum of the state-its decomposition into exact energy eigenstates-falls off exponentially with the energy. Here we consider the energy spectra of approximate matrix product ground states, such as those obtained with the density matrix renormalization group. Despite the high accuracy of these states, contributions to the spectra are roughly constant out to surprisingly high energy, with an increase in the bond dimension reducing the amplitude but not the extent of these high-energy tails. The unusual spectra appear to be a general feature of compressed wavefunctions, independent of boundary or dimensionality, and are also observed in neural network wavefunctions. The unusual spectra can have a strong effect on sampling-based methods, yielding large fluctuations. The energy variance, which can be used to extrapolate observables to eliminate truncation error, is subject to these large fluctuations when sampled. Nevertheless, we devise a sampling-based variance approach which gives excellent and efficient extrapolations.
Scaling laws, the power-law relations between loss, architecture size, and compute observed in modern neural networks, offer a quantitative way to characterize the complexity of a learning problem, with the exponent governing the decay of the loss reflecting how rapidly additional resources translate into improved accuracy, and thus how hard the target is to learn. Whether an analogous framework can characterize the complexity of physical problems remains open. We address this question for Neural-Network Quantum States, a leading variational approach for strongly correlated quantum many-body systems. Using transformer wave functions to approximate ground states of the J1J_1J1​-J2J_2J2​ Heisenberg model on triangular and square lattices with up to 20×2020\times 2020×20 sites, we find that the VVV-score, a measure of accuracy of a variational state, decays as a power law in training compute. Under an appropriate rescaling of compute, results for different system sizes collapse onto a single curve, analogous to scaling collapse in critical phenomena. The resulting power law is, to a good approximation, independent of the number of sites, showing that the transformer Ansatz is size-consistent for the systems considered. The exponent decreases systematically with frustration, identifying it as a quantitative measure of representational difficulty of the ground state and establishing scaling laws as a general framework for benchmarking variational ansätze.