Variational quantum algorithm for generalized eigenvalue problems and its application to the finite element method

Yuki Sato,Hiroshi C. Watanabe,Rudy Raymond,Ruho Kondo,Kaito Wada,Katsuhiro Endo,Michihiko Sugawara,Naoki Yamamoto

DOI: https://doi.org/10.1103/PhysRevA.108.022429

2023-09-27

Abstract:Generalized eigenvalue problems (GEPs) play an important role in the variety of fields including engineering, machine learning and quantum chemistry. Especially, many problems in these fields can be reduced to finding the minimum or maximum eigenvalue of GEPs. One of the key problems to handle GEPs is that the memory usage and computational complexity explode as the size of the system of interest grows. This paper aims at extending sequential quantum optimizers for GEPs. Sequential quantum optimizers are a family of algorithms that iteratively solve the analytical optimization of single-qubit gates in a coordinate descent manner. The contribution of this paper is as follows. First, we formulate the GEP as the minimization/maximization problem of the fractional form of the expectations of two Hermitians. We then showed that the fractional objective function can be analytically minimized or maximized with respect to a single-qubit gate by solving a GEP of a 4 $\times$ 4 matrix. Second, we show that a system of linear equations (SLE) characterized by a positive-definite Hermitian can be formulated as a GEP and thus be attacked using the proposed method. Finally, we demonstrate two applications to important engineering problems formulated with the finite element method. Through the demonstration, we have the following bonus finding; a problem having a real-valued solution can be solved more effectively using quantum gates generating a complex-valued state vector, which demonstrates the effectiveness of the proposed method.

Quantum Physics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to effectively solve the generalized eigenvalue problems (GEPs) in quantum computing. Specifically, the paper proposes a variational quantum algorithm for solving the generalized eigenvalue problems defined by two Hermitian matrices. This problem is very important in fields such as engineering, machine learning, and quantum chemistry, especially when dealing with large - scale systems, traditional classical computing methods face the problems of a sharp increase in memory usage and computational complexity. Therefore, the goal of the paper is to provide a more efficient method to solve these generalized eigenvalue problems by taking advantage of quantum computing.

### Main contributions of the paper

1. **Re - formulation of the generalized eigenvalue problem**:
- The paper first re - formulates the generalized eigenvalue problem as a minimization/maximization problem in the form of a fraction of the expected values of two Hermitian matrices. Then, it is proved that the part of the objective function with respect to single - qubit gates can be analytically minimized or maximized by solving the generalized eigenvalue problem of a 4×4 matrix.

2. **Solution of linear equations**:
- The paper further shows how to transform the system of linear equations (SLE) problem characterized by a positive - definite Hermitian matrix into a generalized eigenvalue problem and solve it using the proposed variational quantum algorithm. This method requires fewer control unitary gates and auxiliary qubits compared to other quantum algorithms.

3. **Application to engineering problems**:
- The paper demonstrates applications through two important engineering problems in the finite element method (FEM): one is to solve the system of linear equations derived from the Poisson equation, and the other is the eigenfrequency analysis of beam structures. Through these demonstrations, the paper estimates the number of qubits required to solve practical problems and finds that for problems with real solutions, quantum gates that generate complex - valued state vectors can be used to solve more efficiently.

### Mathematical formulas

- The general form of the generalized eigenvalue problem is:
\[
Av=\lambda Bv
\]
where \(A\) and \(B\) are Hermitian matrices, and \(\lambda\) and \(v\) are the generalized eigenvalue and generalized eigenvector respectively.

- The Generalized Rayleigh Quotient is defined as:
\[
R(w; A, B):=\frac{w^{\dagger}Aw}{w^{\dagger}Bw}
\]
where \(w\) is an arbitrary unit vector in \(\mathbb{C}^N\).

- The generalized Rayleigh quotient in the quantum state is represented as:
\[
F(\rho):=\frac{\text{tr}(A\rho)}{\text{tr}(B\rho)}
\]
where \(\rho = |\psi(\{q_d\}_{d = 1}^D)\rangle\langle\psi(\{q_d\}_{d = 1}^D)|\).

- The optimization problem of solving single - qubit gates can be transformed into solving the generalized eigenvalue problem of a 4×4 matrix:
\[
S(\rho', A')p_i=\lambda_i S(\rho', B')p_i
\]

### Conclusion

The paper proposes a method for solving the generalized eigenvalue problem based on the variational quantum algorithm. This method not only has theoretical advantages but also shows good performance in practical applications. Especially when dealing with large - scale systems, the high efficiency of quantum computing makes this method have important application prospects. Through specific engineering problem demonstrations, the paper verifies the effectiveness of this method and provides new ideas for further research.