Channel-agnostic finite-temperature phase estimation averaged over variable grids: reconstruction of Green’s function for dynamical mean-field theory Preprint: APS/123-QED

Taichi Kosugi, Hirofumi Nishi, Keito Kasebayashi, Hiroki Takahashi, Yu-ichiro Matsushita · 2026

Read Channel-agnostic finite-temperature phase estimation averaged over variable grids: reconstruction of Green’s function for dynamical mean-field theory Preprint: APS/123-QED by Taichi Kosugi, Hirofumi Nishi, Keito Kasebayashi, Hiroki Takahashi, Yu-ichiro Matsushita (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-29681.txt.

Introduction

For linked electron systems, Taichi Kosugi and four colleagues join quantum and classical steps into one scheme for dynamical mean-field theory. The quantum half uses tweaked phase estimation circuits to pull spectral weights and excitation energies for a one-particle Green's function at finite warmth. It needs no knowledge of which channel fired each time. The classical half rebuilds that function from sampled data by averaging over variable grids and tuning trial terms. Applied to SrVO3 in simulation, the combined QAVG-DMFT scheme checks out as sound.

Catalog record

CREATORTaichi Kosugi, Hirofumi Nishi, Keito Kasebayashi, Hiroki Takahashi, Yu-ichiro Matsushita
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79076
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.29681

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Channel-agnostic finite-temperature phase estimation averaged over variable grids: reconstruction of Green’s function for dynamical mean-field theory Preprint: APS/123-QED Taichi Kosugi, Hirofumi Nishi, Keito Kasebayashi, Hiroki Takahashi, Yu-ichiro Matsushita Abstract For treating correlated electronic systems on quantum computers, we propose a quantum-classical hybrid scheme for dynamical mean-field theory (DMFT). In the quantum part of the scheme, we use modified quantum phase estimation (QPE) circuits suitable for the one-particle Green’s function (GF) at a finite temperature so that we can extract spectral amplitudes and the excitation energies without knowing the excitation channel invoked at each measurement. In the classical part of the scheme, we adopt an approach that estimates reasonably the GF based on the data collected from the QPE sampling. We dub the approach the QPE averaged over variable grids (QAVG), that may help one to reconstruct the GF via optimization of trial parameters and modeling the probability distributions for various settings of the QPE circuits. We apply the QAVG-DMFT scheme to SrVO 3 to demonstrate its validity via numerical simulations. I Introduction Electronic-structure calculations based on density functional theory (DFT) for classical computers have proven highly successful in providing quantitatively accurate predictions and explanations of the properties of a wide range of electronic systems. In this framework, the complex behavior of interacting electrons is typically reformulated into an auxiliary problem of non-interacting electrons subject to an effective mean-field-like potential, which leads to the well-known Kohn–Sham (KS) equations. Despite these successes, DFT-based methods often fail to capture even qualitative aspects of material properties in strongly correlated systems. To overcome such limitations, numerous approaches have been developed and proposed in the literature. One of those for reliable description of correlated systems are the dynamical mean-field theory (DMFT) in conjunction with DFT, denoted as DFT+DMFT approaches 28 . A typical one of them constructs a second-quantized isolated system from the KS orbitals near the Fermi level and solve a corresponding Anderson impurity model (AIM). DMFT calculations using real quantum computers have been already reported. They include DMFT for the Hubbard–Holstein model 4 by employing the Krylov variational quantum algorithm (KVQA) method 19 ; 20 , DFT+DMFT for a copper oxide 51 by employing the quantum equation of motion (qEOM) method 41 ; 47 , and DMFT for the Hubbard model 18 by employing the cumulant expansion method 60 . While the approaches adopted in these earlier studies were chosen according to their applicability to current quantum hardware, we adopt the quantum phase estimation (QPE) based on the quantum Fourier transform (QFT) 1 ; 2 ; 36 as the central technique in our scheme, as will be described. It is because of the rapid growth in the number of high-fidelity qubits on hardware 5 ; 3 ; 45 and the recent achievements for logical qubits 39 ; 38 ; 50 ; 12 toward the era of fault-tolerant quantum computation (FTQC). Modern DMFT calculations as entirely classical computation often adopt versions of quantum Monte Carlo (QMC) methods as impurity solvers to get the one-particle Green’s function (GF). On the other hand, DMFT with full configuration interaction (FCI) solvers, or equivalently exact-diagonalization solvers, play an important role in the present study since we establish the quantum-classical hybrid scheme based on the FCI-based classical scheme, that introduces a finite number of bath sites whose parameters are updated iteratively in the DMFT feedback loop. The quantum part in our scheme begins with preparation of the Gibbs state as a many-qubit state, which then undergoes modified QPE circuits equipped with excitation partial circuits. The sampled data let us estimate the finite-temperature GF without knowing the excitation channel invoked in each measurement and the energy eigenvalues. For estimating the GF from those data, we extend the QPE averaged over variable grids (QAVG) approach, which was developed originally for a zero-temperature system 25 as post processing, to finite-temperature cases. We optimize trial parameters and model probability distributions so that the sampled data are reproduced accurately. The estimated GF is then fed into the DMFT loop to continue the iterations. We apply the new DMFT scheme to SrVO 3 to demonstrate its validity via numerical simulations. II Methods II.1 One-particle GFs Before going to the DFT+DMFT scheme, we recapitulate the definition and useful expressions for the one-particle GF of a generic many-electron system. Let \mathcal{H} be the second-quantized Hamiltonian of the system composed of n_{\mathrm{sorb}} spin orbitals. In this subsection, we assume that the effect of chemical potential, or equivalently the Fermi level, is already taken into account in the one-body part of \mathcal{H}. The one-particle GF in frequency domain is written as \displaystyle G_{mm^{\prime}}(z) \displaystyle=G_{mm^{\prime}}^{(e)}(z)+G_{mm^{\prime}}^{(h)}(z) (1) for a complex frequency z. m and m^{\prime} are indices of spin orbitals. G^{(e)} and G^{(h)} are the electron and hole parts, respectively, given by 7 \displaystyle G_{mm^{\prime}}^{(\xi)}(z) \displaystyle=\frac{1}{\mathcal{Z}}\sum_{\lambda_{0}}e^{-\beta E_{\lambda_{0}}}G_{mm^{\prime}}^{(\xi)}(z|\lambda_{0})\ (\xi=e,h), (2) where \mathcal{Z}\equiv\mathrm{Tr}e^{-\beta\mathcal{H}} is the partition function for an inverse temperature \beta. The summation in Eq. ( 2 ) is taken over each energy eigenstate |\Psi_{\lambda_{0}}\rangle as an initial state of transition, whose energy eigenvalue is E_{\lambda_{0}}. It is reasonable, however, in a practical calculation to introduce a threshold \varepsilon_{\mathrm{B}} for the Boltzmann factors e^{-\beta E_{\lambda_{0}}} such that only the excitations whose Boltzmann factors are larger than \varepsilon_{\mathrm{B}} are incorporated into the GF for avoiding too many excitation channels. \displaystyle G_{mm^{\prime}}^{(e)}(z|\lambda_{0}) \displaystyle\equiv\sum_{\lambda}\frac{\langle\Psi_{\lambda_{0}}|a_{m}|\Psi_{\lambda}\rangle\langle\Psi_{\lambda}|a_{m^{\prime}}^{\dagger}|\Psi_{\lambda_{0}}\rangle}{z-(E_{\lambda}-E_{\lambda_{0}})} (3) is the Lehmann representation 56 of the partial GF responsible for the electron excitations from |\Psi_{\lambda_{0}}\rangle. a_{m}^{\dagger} and a_{m} are the creation and annihilation operators, respectively, for an electron at the m th spin orbital. The summation in the equation above is taken over the energy eigenstates whose electron number is more by one than |\Psi_{\lambda_{0}}\rangle. \displaystyle G_{mm^{\prime}}^{(h)}(z|\lambda_{0})\equiv\sum_{\lambda}\frac{\langle\Psi_{\lambda_{0}}|a_{m^{\prime}}^{\dagger}|\Psi_{\lambda}\rangle\langle\Psi_{\lambda}|a_{m}|\Psi_{\lambda_{0}}\rangle}{z-(E_{\lambda_{0}}-E_{\lambda})} (4) is the partial GF responsible for the hole excitations, where the summation is taken over the energy eigenstates whose electron number is less by one than |\Psi_{\lambda_{0}}\rangle. The partial GF for \xi excitation clearly satisfies G^{(\xi)}_{m^{\prime}m}(z)=G^{(\xi)}_{mm^{\prime}}(z^{*})^{*}. The density of states (DOS) of a many-electron system for the hole and electron parts are experimentally obtained via photoelectron spectroscopy (PES) and inverse PES processes, respectively 11 ; 34 . They are to be compared under certain assumptions with the calculated DOS defined as 26 ; 13 ; 24 ; 37 ; 21 \displaystyle\rho(\omega)=-\frac{1}{\pi}\mathrm{Imtr}G(\omega+i\delta), (5) where \omega is an excitation energy and \delta is an infinitesimally small real constant. For the \xi excitation between an initial state |\Psi_{\lambda_{0}}\rangle and a final state |\Psi_{\lambda}\rangle, we define the transition matrix B^{(\lambda_{0}\to\lambda,\xi)} via \displaystyle B_{mm^{\prime}}^{(\lambda_{0}\to\lambda,e)}\equiv\frac{e^{-\beta E_{\lambda_{0}}}}{\mathcal{Z}}\langle\Psi_{\lambda_{0}}|a_{m}|\Psi_{\lambda}\rangle\langle\Psi_{\lambda}|a_{m^{\prime}}^{\dagger}|\Psi_{\lambda_{0}}\rangle (6) and \displaystyle B_{mm^{\prime}}^{(\lambda_{0}\to\lambda,h)}\equiv\frac{e^{-\beta E_{\lambda_{0}}}}{\mathcal{Z}}\langle\Psi_{\lambda_{0}}|a_{m^{\prime}}^{\dagger}|\Psi_{\lambda}\rangle\langle\Psi_{\lambda}|a_{m}|\Psi_{\lambda_{0}}\rangle. (7) These matrices are clearly Hermitian: B_{m^{\prime}m}^{(\lambda_{0}\to\lambda,\xi)}=B_{mm^{\prime}}^{(\lambda_{0}\to\lambda,e)*}. Since we will work with QPE sampling in the present study, we define the spectral matrix S^{(\xi)} via \displaystyle S^{(\xi)}_{mm^{\prime}}(\varepsilon) \displaystyle\equiv\sum_{\lambda_{0},\lambda}B_{mm^{\prime}}^{(\lambda_{0}\to\lambda,\xi)}\delta_{E_{\lambda}-E_{\lambda_{0}},\varepsilon}, (8) where the Kronecker delta is for the excitation energy from |\Psi_{\lambda_{0}}\rangle to |\Psi_{\lambda}\rangle. The argument \varepsilon is a possible discrete value represented by the ancillae used in the QPE circuit. For a case where a large number of ancillae are used in the QPE circuit, we can write the electron GF approximately as \displaystyle G_{mm^{\prime}}^{(e)}(z) \displaystyle\approx\sum_{\varepsilon}\frac{S^{(e)}_{mm^{\prime}}(\varepsilon)}{z-\varepsilon} (9) and the hole GF as \displaystyle G_{mm^{\prime}}^{(h)}(z) \displaystyle\approx\sum_{\varepsilon}\frac{S^{(h)}_{mm^{\prime}}(\varepsilon)}{z+\varepsilon}. (10) II.2 DFT+DMFT using a quantum computer Here we describe our scheme for a DFT+DMFT calculation based on QAVG. The flowchart of the scheme is provided in Fig. 1 . We assume that

Read the complete text