On the application of the Rayleigh-Ritz method to a projected Hamiltonian

Basic Information

• Paper ID: 2411.14490
• Title: On the application of the Rayleigh-Ritz method to a projected Hamiltonian
• Author: Francisco M. Fernández (INIFTA, DQT, La Plata, Argentina)
• Classification: quant-ph (Quantum Physics)
• Publication Date: December 31, 2024 (arXiv v5)
• Paper Link: https://arxiv.org/abs/2411.14490

Abstract

This paper applies the classical Rayleigh-Ritz method (RRM) to a projected Hamiltonian operator, which has recently been employed to extend the Rayleigh-Ritz variational principle to ensemble states. Through toy models, it is demonstrated that in most cases, RRM eigenvalues approach the eigenvalues of the projected Hamiltonian from below, though some exceptions exist. The paper also discusses the effects of energy shifts and identity operator projections.

Research Background and Motivation

Problem Background

1. Limitations of Traditional RRM: The classical Rayleigh-Ritz method is one of the most widely used approaches for studying electronic structures of atoms and molecules. Its primary advantage is that RRM eigenvalues converge to the exact energy of physical systems from above.
2. Recent Theoretical Developments: Ding et al. recently proposed a method to extend the Rayleigh-Ritz variational principle to ensemble states, introducing a special Hamiltonian operator on a D-dimensional Hilbert space.
3. Theoretical Gap: Although this operator is impractical for real physical problems, its mathematical properties merit further investigation, particularly regarding the behavior of RRM when applied to projected Hamiltonians.

Research Motivation

• To explore the convergence properties of RRM on projected Hamiltonians, particularly whether the classical property of convergence from above is preserved
• To understand the influence of projection operators on the convergence behavior of variational methods
• To provide theoretical foundations for extending variational principles

Core Contributions

1. First systematic study of RRM application to projected Hamiltonians, revealing convergence behavior different from classical theory
2. Demonstrated that RRM eigenvalues converge from below in most cases, contrasting with traditional convergence from above
3. Provided concrete numerical examples validating theoretical predictions through a one-dimensional particle-in-a-box model
4. Analyzed the effects of energy shifts and identity operator projections, refining the theoretical framework
5. Revealed the interesting phenomenon that N-D zero eigenvalues are given exactly

Methodology Details

Problem Definition

Investigation of convergence properties when the Rayleigh-Ritz method is applied to a projected Hamiltonian operator, where the projected Hamiltonian is defined as:

$H_D = \sum_{k=0}^{D-1} E_k |\psi_k\rangle \langle\psi_k|$

Input: Non-orthogonal basis function set $\{|u_j\rangle\}$ Output: Approximate eigenvalues and eigenvectors Constraints: Basis functions span a larger Hilbert space containing the projected subspace

Theoretical Framework

Construction of Projected Hamiltonian

Consider a Hamiltonian operator $H$ on an infinite-dimensional Hilbert space $\mathcal{H}$, whose projection onto a D-dimensional subspace $\mathcal{H}_D$ is:

$H_D = \sum_{k=0}^{D-1} E_k |\psi_k\rangle \langle\psi_k|$

where $\{|\psi_k\rangle, k=0,1,...,D-1\}$ spans the subspace $\mathcal{H}_D \subset \mathcal{H}$.

Application of RRM

RRM requires computation of matrix elements: $(H_D)_{ij} = \sum_{k=0}^{D-1} E_k \langle u_i|\psi_k\rangle \langle\psi_k|u_j\rangle$

The secular equation is: $|H-WS| = 0$

where $H_{ij} = \langle u_i|H_D|u_j\rangle$ and $S_{ij} = \langle u_i|u_j\rangle$

Technical Innovations

1. Convergence Analysis: Discovery of convergence patterns different from classical RRM, particularly the phenomenon of convergence from below
2. Boundary Condition Analysis: Analysis of special cases through Cauchy-Schwarz inequality to establish upper and lower bounds
3. Exactness of Zero Eigenvalues: Proof that N-D zero eigenvalues are given exactly

Experimental Setup

Model System

One-dimensional particle-in-a-box model: $H = -\frac{1}{2}\frac{d^2}{dx^2}$

Boundary conditions: $\psi(0) = \psi(1) = 0$

Exact solutions:

• Eigenvalues: $E_k = \frac{k^2\pi^2}{2}, k = 1,2,...$
• Eigenfunctions: $\psi_k(x) = \sqrt{2}\sin(k\pi x)$

Basis Function Selection

Non-orthogonal basis function set: $u_i(x) = x^i(1-x), i = 1,2,...$

Experimental Parameters

• Projection dimension: D = 1, 2, 3, 4
• Number of basis functions: N = 1 to 19
• Energy shift: c = -5

Experimental Results

Main Results

RRM Results for Classical Hamiltonian

Table 1 shows the upper bound convergence property of classical RRM:

• N=1: $W_1 = 5 > E_1 = 4.934802200$
• As N increases, all eigenvalues converge to exact values from above

RRM Results for Projected Hamiltonian

Case D=1 (Table 2):

• N=1: $W_1 = 4.927671482 < E_1 = 4.934802200$
• Convergence from below to exact values

Case D=2 (Table 3):

• Both non-zero eigenvalues converge from below
• Remaining eigenvalues are exactly zero

Case D=3 (Table 4):

• N=1: $W_1 = 4.988506932 > E_1$ (upper bound)
• N>1: All eigenvalues converge from below

Energy Shift Effects

Table 6 shows results for $E_k = k^2\pi^2/2 - 5$:

• $W_1$ becomes an upper bound
• $W_2$ and $W_3$ maintain lower bound properties

Expectation Value Analysis

Table 5 demonstrates the variation of expectation value $\langle H_D\rangle$ with D:

• As D increases, expectation value approaches $\langle H\rangle = 5$
• Analytical proof: $\lim_{D\to\infty}\langle H_D\rangle = 5$

Key Findings

1. Change in Convergence Direction: Unlike classical RRM, projected Hamiltonian RRM eigenvalues primarily converge from below
2. Exactness of Zero Eigenvalues: N-D zero eigenvalues are given exactly
3. Dimension Dependence: Convergence behavior depends on the relationship between projection dimension D and number of basis functions N

Related Work

Classical Variational Theory

• MacDonald (1933): Established theoretical foundations for RRM upper bound convergence
• Pilar, Szabo & Ostlund: Applications in quantum chemistry

Recent Developments

• Ding et al. (2024): Proposed extension of variational principle to ensemble states
• Previous work by the author: General discussion of RRM

Unique Contributions of This Work

Compared to existing work, this paper is the first to:

• Systematically study RRM behavior for projected Hamiltonians
• Discover the new phenomenon of convergence from below
• Provide concrete numerical verification

Conclusions and Discussion

Main Conclusions

1. Change in Convergence Properties: When RRM is applied to projected Hamiltonians, eigenvalues primarily converge from below, contrasting with the classical upper bound property
2. Exact Zero Eigenvalues: RRM yields N-D exact zero eigenvalues, even though approximate solutions only converge to exact solutions as N→∞
3. Condition Dependence: Convergence direction depends on specific projection dimension and basis function choice

Limitations

1. Non-physical Nature of Model: The studied projected Hamiltonian is impractical for real physical problems
2. Basis Function Restrictions: Results depend on specific basis function choices
3. Incomplete Theoretical Explanation: Lacks rigorous mathematical proof for the convergence-from-below phenomenon

Future Directions

1. Develop rigorous theory for convergence of RRM on projected Hamiltonians
2. Explore possible applications to practical physical systems
3. Investigate properties of other types of projection operators

In-Depth Evaluation

Strengths

1. Theoretical Innovation: Reveals new properties of RRM on projected Hamiltonians, challenging conventional understanding
2. Sufficient Numerical Verification: Systematic calculations across multiple dimensions and parameters validate theoretical predictions
3. Thorough Analysis: Provides not only numerical results but also analytical analysis and theoretical explanations
4. Clear Presentation: Logical structure and rigorous mathematical derivations

Weaknesses

1. Limited Practical Utility: The studied projected Hamiltonian lacks practical physical significance
2. Incomplete Theory: Lacks complete mathematical proof for the convergence-from-below phenomenon
3. Narrow Applicability: Results are highly dependent on specific models and basis function choices

Impact

1. Theoretical Contribution: Provides new perspectives and understanding for variational method theory
2. Methodological Significance: May inspire development of new numerical methods
3. Reproducibility: Provides detailed numerical results facilitating verification and extension

Applicable Scenarios

• Theoretical quantum mechanics research
• Mathematical analysis of variational methods
• Development of projection operator theory
• Benchmark testing for numerical methods

References

1. MacDonald, J. K. L. (1933). Successive approximations by the Rayleigh-Ritz variation method. Physical Review, 43, 830.
2. Ding, L., Hong, C-L., & Schilling, C. (2024). Quantum 8, 1525.
3. Fernández, F. M. (2024). On the Rayleigh-Ritz variational method. arXiv:2206.05122.

Although this paper addresses relatively abstract mathematical problems, the new properties of RRM on projected Hamiltonians it reveals possess significant theoretical value, providing fresh perspectives for understanding the convergence behavior of variational methods.