Abstract:We propose a new technique to incorporate multiple fermion flavors in the tensor renormalization group method for lattice gauge theories, where fermions are treated by the Grassmann tensor network formalism. The basic idea is to separate the site tensor into multiple layers associated with each flavor and to introduce the gauge field in each layer as replicas, which are all identified later. This formulation, after introducing an appropriate compression scheme in the network, enables us to reduce the size of the initial tensor with high efficiency compared with a naive implementation. The usefulness of this formulation is demonstrated by investigating the chiral phase transition and the Silver Blaze phenomenon in 2D Abelian gauge theories with $N_{\rm f}$ flavors of Wilson fermions up to $N_{\rm f}=4$.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to effectively deal with lattice gauge theories containing multiple fermion flavors in the tensor renormalization group method. Specifically, the paper proposes a new technique that can introduce multiple fermion flavors in the tensor network method. In particular, when dealing with lattice fermions in lattice gauge theories, the Grassmann tensor network formalization method is used. The main innovation of this technique lies in decomposing the site tensor into multiple layers associated with each flavor and introducing the gauge field as a replica in each layer. These replicas will eventually be identified as the same gauge field. By introducing an appropriate compression scheme, this method can significantly reduce the size of the initial tensor, thereby improving computational efficiency.
### Background
When dealing with non - perturbative calculations of fermion systems, it has always been a challenge due to the anti - commutation property of Grassmann variables. In the Monte Carlo method, it is necessary to first integrate out the Grassmann variables to obtain the fermion determinant \(\det M\), which makes the calculation very time - consuming because the size of the matrix \(M\) is proportional to the system size \(V\). Although using the pseudo - fermion technique and an appropriate hybrid Monte Carlo algorithm can reduce the computational cost to \(O(V)\), the calculation is still usually several orders of magnitude slower than the corresponding boson systems. In addition, in many interesting fermion systems, such as finite - density systems, strongly - correlated electron systems, and theories with chiral fermions, the fermion determinant becomes complex, leading to the well - known sign problem in the traditional Monte Carlo method.
### Advantages of the Tensor Network Method
The tensor network method provides a statistical method not based on importance sampling to solve these problems. This method was initially introduced to deal with many - body systems in condensed - matter physics and is mainly applied to ground - state calculations based on the variational principle. However, it can also be directly used to calculate the partition function, and the process is similar to the real - space renormalization group, so it is named the "tensor renormalization group (TRG) method". It is worth noting that this method can calculate the partition function with only a logarithmically growing computational cost. Although the initial TRG method was proposed for two - dimensional boson systems, improved versions have been subsequently developed and extended to higher - dimensional systems and fermion systems, where Grassmann variables are directly processed instead of dealing with the fermion determinant as in the Monte Carlo method.
### Main Contributions of the Paper
The paper proposes a new technique that makes it possible to deal with multiple fermion flavors in the Grassmann tensor network formalization method. The main idea is to decompose the initial tensor into multiple layers associated with each flavor. Since fermions of different flavors interact with the same gauge field, after integrating out the gauge field, the interactions in the flavor direction become non - local. To avoid this problem, the paper introduces the gauge field as a replica in each layer and identifies them as the same gauge field in subsequent steps. Once the system can be described by a tensor network one dimension higher than the original theory, the standard coarse - graining technique can be used to calculate the partition function.
### Application Examples
The paper demonstrates the effectiveness of this method by applying it to Wilson fermions in two - dimensional Abelian gauge theories. First, the chiral phase transition in the case of \(N_f = 2\) was studied. The results show that as \(K\) increases, the results obtained using the \(Z_K\) gauge theory gradually converge to the \(U(1)\) results obtained using the Monte Carlo method. Second, the Silver Blaze phenomenon with up to \(N_f = 4\) at a finite fermion density was studied.
### Summary
The new technique proposed in the paper not only solves the problem of exponential growth of the tensor size when dealing with multiple fermion flavors but also provides an efficient method to calculate the partition function, which is applicable to various fermion systems, especially those that cannot be processed by the Monte Carlo method.