Grassmann corner transfer-matrix renormalization group approach to one-dimensional fermionic models
Jian-Gang Kong, Zhi Yuan Xie · 2026
Read Grassmann corner transfer-matrix renormalization group approach to one-dimensional fermionic models by Jian-Gang Kong, Zhi Yuan Xie (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-2604-05582.txt.
Introduction
Kong and Xie build a Grassmann tensor-network tool for one-dimensional interacting fermions, squeezing a (1+1)-dimensional network with a corner transfer-matrix renormalization-group routine. The coherent-state path-integral view handles Fermi-Dirac statistics head-on. Tests on the Hubbard model in a magnetic field catch the phase diagram's main features in numbers.
Catalog record
| CREATOR | Jian-Gang Kong, Zhi Yuan Xie |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53027 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.05582
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Grassmann corner transfer-matrix renormalization group approach to one-dimensional fermionic models Jian-Gang Kong, Zhi Yuan Xie Abstract The strongly correlated fermions play a vital role in modern physics. For a given fermionic Hamiltonian system, the most widely used approach to explore the underlying physics is to study the wave function that incorporates Fermi-Dirac statistics, which can be obtained variationally by energy minimization or by imaginary-time evolution. In this work, we develop an accurate tensor network method for one-dimensional interacting fermionic models based on the coherent-state path-integral representation of the fermionic partition function. Employing the coherent-state representation, the partition function is effectively represented as a (1+1)-dimensional anisotropic Grassmann-valued tensor network, and the Grassmann version of the corner transfer-matrix renormalization group algorithm is developed to contract the tensor network and evaluate physical quantities. We validate our method in the one-dimensional fermionic Hubbard model with a magnetic field, where the essential features of the phase diagram in the ( μ , B ) (\mu,B) plane are quantitatively captured. Our work offers a promising approach to interacting fermionic models within the framework of tensor networks. I Introduction Fermionic many-body systems lie at the heart of many fundamental and technologically relevant phenomena in condensed matter physics, such as superconductivity Bednorz1986 , fractional quantum Hall effects Tsui1982 , Mott insulator MIT , quantum spin liquid AndersonQSL , and various topological phases of matter TopoBook . However, accurately simulating fermionic models poses significant challenges in many-body physics, including exponential scaling with system size for exact diagonalization ExpWall , the notorious sign problem in quantum Monte Carlo Troyer2005 , and limited correlations in approximating strongly correlated systems through weak-coupling approaches DFT . Among these numerical methods, although it has its own drawbacks, the tensor network state method PEPS2004 ; Verstraete2008 ; OrusAOP ; CiracRMP ; Xiang2023 ; OrusEPJB ; OrusNRP has attracted increasing attention in recent years due to its advantages in characterizing low-entangled states and handling sign-problematic systems. In particular, by incorporating fermionic exchange statistics into the wave-function representation, the tensor network state has become an important tool for the study of strongly correlated electronic systems. For example, the swap-gate formulation Corboz2009 ; Corboz2010 ; Corboz2010a , probably the most popular fermionic tensor network proposal Barthel2009 ; Kraus2009 ; Shi2009 ; Pi2010 ; Zhang2024 ; Mortier2025 , has been widely utilized for numerical simulations and provided convincing evidences of possible striped states Corboz2011 ; Corboz2014 ; Zheng2017 ; Ponsioen2019 ; Ponsioen2023 as well as superconducting states Corboz2014 ; Ponsioen2019 ; Ponsioen2023 in the two-dimensional (2D) t - J Zhang1988 and Hubbard models Hubbard1963 ; Qin2022 . Nevertheless, based on the graphical projection of the 2D wave function, hosting a three-dimensional structure essentially, to a 2D plane, this swap-gate formulation is not so convenient in some situations, e.g., complex lattice structure where the graphical projection might be too complicated, and finite temperature calculation, which requires the evaluation of the partition function. Using the coherent-state representation, Grassmann tensor networks can describe both the fermionic wavefunction and the path-integral representation of the partition function. In condensed matter physics, Gu et al. proposed the Grassmann tensor network states Gu2010 ; Gu2013 ; Lou2015 for simulating the ground state of fermionic Hamiltonians Gu2013a ; Gu2020 ; Miao2025 ; Liu2025 . Later in the particle physics community, the Grassmann tensor renormalization group was developed Levin2007 ; Yuya2014 ; Takeda2015 ; Sakai2017 ; Yoshimura2018 ; Akiyama2021a ; Akiyama2024 to compute the partition function of relativistic lattice fermionic models. Despite their capability, Grassmann tensor networks are still relatively underexploited compared to their bosonic counterparts, and their potential has not been fully explored. In particular, the corner transfer-matrix renormalization group (CTMRG) method Nishino1996 ; Nishino1996a ; Corboz2014 ; Or2009 ; Fishman2018 ; VCTMRG2022 has demonstrated supreme advantages over many coarse-graining tensor renormalization group methods in contracting 2D bosonic tensor networks, but it has not been utilized so far to showcase its power for Grassmann tensor networks. In this work, we introduce the Grassmannization of the CTMRG method and validate this approach via the one-dimensional fermionic Hubbard model Essler2005 with magnetic field in the path integral formalism Akiyama2021 ; Akiyama2022 . The rest of the paper is structured as follows. In Sec. II , we introduce the one-dimensional Hubbard model and the Grassmann tensor network representation of its partition function in the path integral formalism. In Sec. III , we firstly present a simple example of Grassmann tensor contraction as a warm-up, then the Grassmannization of the CTMRG method will be described in detail. In Sec. IV , we validate our Grassmann tensor network approach in the one-dimensional Hubbard model under a magnetic field. In Sec. V , we summarize and provide a prospect on the Grassmann tensor networks. II Tensor network representation of fermionic partition functions Let us take the one-dimensional fermionic Hubbard model under a magnetic field as a concrete example. The Hamiltonian is given by H=-t\sum_{\langle ij\rangle}\sum_{s=\uparrow,\downarrow}\left(\hat{c}_{is}^{\dagger}\hat{c}_{js}+h.c.\right)+U\sum_{i}\left(\hat{n}_{i\uparrow}-1/2\right)\left(\hat{n}_{i\downarrow}-1/2\right)-\mu\sum_{i}\left(\hat{n}_{i\uparrow}+\hat{n}_{i\downarrow}\right)-B\sum_{i}\left(\hat{n}_{i\uparrow}-\hat{n}_{i\downarrow}\right) (1) The \hat{c}_{is}^{\dagger} and \hat{c}_{is} with s=\{\uparrow,\downarrow\} are the fermionic creation and annihilation operators defined at the i -th lattice site, and \hat{n}_{is}\equiv\hat{c}^{\dagger}_{is}\hat{c}_{is} is the occupation number operator. The first term describes the electrons hopping with energy t on the one-dimensional lattice, the second term is the on-site Coulomb interactions with strength U , the third term includes the chemical potential \mu to control the electron density, and the last term describes the electrons coupled to an external magnetic field B . Following the standard fermionic path integral formalism ColemanBook2015 , the corresponding fermionic action for Eq. ( 1 ) can be expressed as: \displaystyle S\left[\psi,\bar{\psi}\right]=\sum_{n\in\Lambda_{1+1}}\bar{\psi}(n)\left(\psi(n+\hat{\tau})-\psi(n)\right)-\epsilon t\left(\bar{\psi}(n)\psi(n+\hat{\sigma})+\bar{\psi}(n+\hat{\sigma})\psi(n)\right) (2) \displaystyle+ \displaystyle\epsilon U\bar{\psi}_{\uparrow}(n)\psi_{\uparrow}(n)\bar{\psi}_{\downarrow}(n)\psi_{\downarrow}(n)-\epsilon\left(\mu+B+U/2\right)\bar{\psi}_{\uparrow}(n)\psi_{\uparrow}(n)-\epsilon\left(\mu-B+U/2\right)\bar{\psi}_{\downarrow}(n)\psi_{\downarrow}(n)+U/4 In Eq. ( 2 ), \hat{\sigma} ( \hat{\tau} ) denotes the unit vector in the spatial (temporal) direction of the (1+1)-dimensional lattice \Lambda_{1+1} , n\equiv(n_{\sigma},n_{\tau}) is the composite coordinate, and \epsilon stands for the discretization (Trotter) step in the temporal (imaginary-time) direction. The \psi(n)=\left(\psi_{\uparrow}(n),\psi_{\downarrow}(n)\right)^{T} and \bar{\psi}(n)=\left(\bar{\psi}_{\uparrow}(n),\bar{\psi}_{\downarrow}(n)\right) are the two-component Grassmann-valued fields defined on the n -th site. \bar{\psi}_{\uparrow,\downarrow} is the dual of Grassmann variable \psi_{\uparrow,\downarrow} , and they should be treated as strictly independent AtlandQFT . The partition function corresponding to this Hamiltonian can be approximately represented as a 2D Grassmann tensor network Akiyama2021 : \displaystyle Z \displaystyle\approx \displaystyle\int\prod_{n\in\Lambda_{1+1}}d\bar{\psi}(n)d\psi(n)\mathrm{e}^{-S\left[\psi,\bar{\psi}\right]}=\textrm{gTr}\left(\prod_{n\in\Lambda_{1+1}}\mathcal{T}_{\Psi_{\sigma}(n)\Psi_{\tau}(n)\bar{\Psi}_{\tau}(n-\hat{\tau})\bar{\Psi}_{\sigma}(n-\hat{\sigma})}\right) (3) where the integral is over all the Grassmann variables. Similarly, the gTr stands for the Grassmann integral over all the Grassmann variables appearing in the Grassmann tensor network. An illustration is shown in Fig. 1 (a). The local Grassmann tensor \mathcal{T} is defined as: \mathcal{T}_{\Psi_{\sigma}(n)\Psi_{\tau}(n)\bar{\Psi}_{\tau}(n-\hat{\tau})\bar{\Psi}_{\sigma}(n-\hat{\sigma})}=\sum_{I_{\sigma}I_{\sigma}^{\prime}=1}^{16}\sum_{I_{\tau}I_{\tau}^{\prime}=1}^{4}T_{I_{\sigma}I_{\tau}I_{\sigma}^{\prime}I_{\tau}^{\prime}}\Psi_{\sigma}^{p(I_{\sigma})}(n)\Psi_{\tau}^{p(I_{\tau})}(n)\bar{\Psi}_{\tau}^{p(I_{\tau}^{\prime})}(n-\hat{\tau})\bar{\Psi}_{\sigma}^{p(I_{\sigma}^{\prime})}(n-\hat{\sigma}) (4) Here, \Psi_{\sigma}(n) and \Psi_{\tau}(n) are auxiliary Grassmann variables residing on the bonds ( n , n+\hat{\sigma} ) and ( n , n+\hat{\tau} ), respectively. Conversely, \bar{\Psi}_{\sigma}(n-\hat{\sigma}) and \bar{\Psi}_{\tau}(n-\hat{\tau}) are Grassmann variables that introduced on bonds ( n-\hat{\sigma} , n ) and ( n-\hat{\tau} , n ), see Fig. 1 (b). Mathematically, the Grassmann tensor \mathcal{T} in Eq. ( 4 ) represents a multilinear combination of Grassmann variables, and T with commuting elements is called its coefficient tensor. The parity function p yields 0 (1) if the index belongs to the Grassmann-even (odd) sector. For a more detailed introduction to Grassmann tensor operations, such as the Grassmann contraction, fusion, decomposition, etc., interested readers can refer to Refs. Akiyama2024 . For action expressed in Eq. ( 2 ), the coefficient tensor T appearing