Quench dynamics via recursion method and dynamical quantum phase transitions
Ilya Shirokov, Viacheslav Khrushchev, Filipp Uskov, Ivan Dudinets, Igor Ermakov, Oleg Lychkovskiy · 2025
Read Quench dynamics via recursion method and dynamical quantum phase transitions by Ilya Shirokov, Viacheslav Khrushchev, Filipp Uskov, Ivan Dudinets, Igor Ermakov, Oleg Lychkovskiy (2025) 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-2503-24362.txt.
Introduction
Ilya Shirokov, Viacheslav Khrushchev, Filipp Uskov, Ivan Dudinets, Igor Ermakov, and Oleg Lychkovskiy wrote this paper in two thousand twenty-five. Its text weighs the strengths and limits of the recursion method for quench dynamics in quantum many-body systems.
The method solves Heisenberg equations in a Lanczos base and shines for correlation functions, since the Lanczos weights act in a universal way that can be stretched. A quench also needs overlap weights against the start state. Those, the team shows, have no universal shape and cannot be stretched with trust. So the sound time span shrinks, by how wild the overlaps run. For tame start states the method still rivals the best, and one symbolic run serves many scans.
Catalog record
| CREATOR | Ilya Shirokov, Viacheslav Khrushchev, Filipp Uskov, Ivan Dudinets, Igor Ermakov, Oleg Lychkovskiy |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-24362
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quench dynamics via recursion method and dynamical quantum phase transitions Ilya Shirokov, Viacheslav Khrushchev, Filipp Uskov, Ivan Dudinets, Igor Ermakov, Oleg Lychkovskiy Abstract The recursion method provides a powerful framework for studying quantum many-body dynamics in the Lanczos basis recursively constructed within the Krylov space of operators. Recently, it has been demonstrated that the recursion method, when supplemented by the universal operator growth hypothesis, can effectively compute autocorrelation functions and transport coefficients at infinite temperature. We extend the scope of the recursion method to far-from-equilibrium quench dynamics. We apply the method to spin systems in one, two, and three spatial dimensions. In one dimension we find an evidence that the convergence of the method is limited in time by the dynamical quantum phase transition. Introduction. Addressing quantum dynamics by means of the recursion method has a long history Mori 1965 ; Dupuis 1967 ; Haydock 1980 ; Mattis 1981 ; Joslin and Gray 1986 ; Roldan et al. 1986 ; Viswanath and Müller 2008 ; Lindner and Auerbach 2010 ; Nandy et al. 2024 . The (Heisenberg version of the) method essentially amounts to constructing an orthonormal Lanczos basis within the Krylov space of an evolving quantum observable and harnessing the simplifications this basis brings about. A new impact has been given to the method by the universal operator growth hypothesis Parker et al. 2019 – a conjecture about a generic form of the evolution generator in the Lanczos basis. Up to now the practical application of the recursion method to quantum many-body systems has been limited almost exclusively to computing infinite-temperature autocorrelation functions and transport coefficients Parker et al. 2019 ; Khait et al. 2016 ; de Souza et al. 2020 ; Yates et al. 2020a ; Yates et al. 2020b ; Yuan et al. 2021 ; Wang et al. 2024 ; Bartsch et al. 2024 ; Uskov and Lychkovskiy 2024 ; Teretenkov et al. 2024 ; Bhattacharyya et al. 2024 ; Füllgraf et al. 2025 . There the method has demonstrated high efficiency. This efficiency is particularly notable in application to two-dimensional systems Uskov and Lychkovskiy 2024 ; Teretenkov et al. 2024 ; Bhattacharyya et al. 2024 , where traditional numerical methods often struggle. Another notable advantage of the recursion method is that it allows a straightforward symbolic implementation, when a single computation covers all parameters of the Hamiltonian Uskov and Lychkovskiy 2024 ; Teretenkov et al. 2024 . Formally, the output of the recursion method is the Heisenberg operator of an observable. This suggests that the scope of the method can be much broader than infinite-temperature correlation functions and should cover, in particular, the out-of-equilibrium dynamics after quantum quenches. Addressing the later in systems of dimension higher than one is known to be highly challenging, with limited body of results obtained to date Schmitt and Heyl 2020 ; Li et al. 2020 ; Dziarmaga 2022 ; Bernier and Agarwal 2025 ; Begušić and Chan 2025 ; Loizeau et al. 2025a ; Loizeau et al. 2025b . Given the success of the recursion method in computing two-dimensional autocorrelation function, it seems tempting to apply it to quench dynamics. Here we undertake exactly this attempt. The paper is organized as follows. We start from briefly introducing the recursion method and specifying it for addressing quench dynamics. Then we apply it to a paradigmatic one-dimensional mixed-field Ising model. The results are benchmarked by essentially exact numerical simulations. We demonstrate that the recursion method has a convergence domain limited in time by the dynamical quantum phase transition. Then results for the transverse-field Ising model on the square and cubic lattices are presented. We conclude by discussing implications of our results and open questions raised. Recursion method for quench dynamics. Consider an observable given by a self-adjoint Shrödinger operator A . The same observable in the Heisenberg representation reads A_{t}=e^{itH}A\,e^{-itH} , where H is the Hamiltonian of the system. The Heisenberg operator A_{t} evolves within the Krylov space spanned by operators A,{\mathcal{L}}A,{\mathcal{L}}^{2}A,\dots , where {\mathcal{L}}\equiv[H,\bullet] is the superoperator referred to as Liouvillian. A_{t} satisfies the Heisenberg equation of motion \partial_{t}A_{t}=i\,{\mathcal{L}}A_{t} . Within the recursion method Viswanath and Müller 2008 , this equation is addressed in the orthonormal Lanczos basis \{O^{0},O^{1},O^{2},\dots\} constructed as follows: O^{0}=A/\|A\| , A^{1}={\mathcal{L}}O^{0} , b_{1}=\|A^{1}\| , O^{1}=b_{1}^{-1}\,A^{1} and A^{n}={\cal L}\,O^{n-1}-\,b_{n-1}O^{n-2} , b_{n}=\|A^{n}\| , O^{n}=b_{n}^{-1}A^{n} for n\geq 2 . Here a scalar product in the space of operators is introduced according to \big(A|B\big)\equiv{\cal N}\,\tr\big(A^{\dagger}B\big) , where {\cal N} is a normalization constant. For systems of N spins 1/2 and translation-invariant observables considered here it is convenient to choose {\cal N}=N^{-1}\,2^{-N} . The scalar product entails the norm \|A\|=\sqrt{(A|A)} . The coefficients b_{n} are referred to as Lanczos coefficients . In the Lanczos basis the Liouvillian has a tridiagonal form, {\mathcal{L}}=\begin{pmatrix}0&b_{1}&0&0&\dots\\[3.00003pt] b_{1}&0&b_{2}&0&\dots\\[3.00003pt] 0&b_{2}&0&b_{3}&\dots\\[3.00003pt] 0&0&b_{3}&0&\dots\\[3.00003pt] \vdots&\vdots&\vdots&\vdots&\ddots\end{pmatrix}. (1) The solution of the Heisenberg equation for the observable A_{t} can be expressed as A_{t}=\|A\|\,\sum_{n=0}^{\infty}\phi^{n}_{t}\,O^{n}, (2) where \phi^{n}_{t}\equiv(O^{n}|A_{t})/\|A\| are the solution of the system of coupled equations, which can be written in the matrix form as \displaystyle\partial_{t}\phi_{t}= \displaystyle i\,{\mathcal{L}}\phi_{t}. (3) Here \phi_{t}=\|\phi^{n}_{t}\|_{n=0,1,2,\dots} is a column vector and {\mathcal{L}} is the matrix ( 1 ). In practice, one can explicitly compute only a finite number n_{\rm max} of Lanczos coefficients. The rest of the coefficients should be extrapolated. The universal operator growth hypothesis Parker et al. 2019 (see also precursor work Liu and Müller 1990 ; Florencio et al. 1992 ; Zobov and Lundin 2006 ; Elsayed et al. 2014 ; Bouch 2015 ) states that in a generic quantum many-body system Lanczos coefficients b_{n} scale asymptotically linearly with n at large n (with a logarithmic correction for one-dimensional systems). This asymptotic behavior has been confirmed in various many-body models Cao 2021 ; Noh 2021 ; Heveling et al. 2022 ; Uskov and Lychkovskiy 2024 ; De et al. 2024 ; Ermakov 2025 . As a rule, the leading asymptotics can be easily extracted from the first dozen of Lanczos coefficients. However, extrapolating Lanczos coefficients by a linear function is a too rough approximation: the qualitative behavior of the solutions of the Heisenberg equation strongly depends on the subleading asymptotic terms Viswanath et al. 1994 ; Parker et al. 2019 ; Yates et al. 2020a ; Yates et al. 2020b ; Dymarsky and Smolkin 2021 ; Bhattacharjee et al. 2022 ; Avdoshkin et al. 2024 ; Camargo et al. 2023 ; Uskov and Lychkovskiy 2024 ; Dodelson 2025 ; Gamayun et al. 2025 . Guided by the results of ref. Gamayun et al. 2025 , we employ the following extrapolation formula: \displaystyle b_{n} \displaystyle\simeq\frac{\alpha\,n}{\log n}+\gamma+\frac{(-1)^{n}\alpha}{\log n+c} \displaystyle{\rm for}~~1{\rm D}, (4) \displaystyle b_{n} \displaystyle\simeq\alpha\,n+\gamma+\frac{(-1)^{n}\alpha}{\log n+c} \displaystyle{\rm for}~~2,3{\rm D}. (5) The last alternating term is required whenever \varphi^{0}_{t} relaxes to a nonzero value at t\rightarrow\infty Gamayun et al. 2025 , which is generally the case. Parameters \alpha,\,\gamma and c are found by fitting n_{\rm max} explicitly computed Lanczos coefficients. The knowledge of the Heisenberg operator A_{t} of the observable allows one to address the quench dynamics, i.e. compute the time-dependent value \langle A_{t}\rangle\equiv\tr\rho_{0}A_{t} of the observable that evolves after the system is initialized in an out-of-equilibrium state \rho_{0} . Within our approach, we compute \langle A_{t}\rangle according to \langle A_{t}\rangle\simeq\sum_{n=0}^{n_{\rm max}}\phi^{n}_{t}\,\langle O^{n}\rangle, (6) where \langle O^{n}\rangle\equiv\tr\rho_{0}O^{n} . We