Andrew V. Knyazev,Merico E. Argentati
Abstract:The Rayleigh-Ritz (RR) method finds the stationary values, called Ritz values, of the Rayleigh quotient on a given trial subspace as approximations to eigenvalues of a Hermitian operator $A$. If the trial subspace is $A$-invariant, the Ritz values are exactly some of the eigenvalues of $A$. Given two subspaces $\X$ and $\Y$ of the same finite dimension, such that $\X$ is $A$-invariant, the absolute changes in the Ritz values of $A$ with respect to $\X$ compared to the Ritz values with respect to $\Y$ represent the RR absolute eigenvalue approximation error. Our first main result is a sharp majorization-type RR error bound in terms of the principal angles between $\X$ and $\Y$ for an arbitrary $A$-invariant $\X$, which was a conjecture in [SIAM J. Matrix Anal. Appl., 30 (2008), pp. 548-559]. Second, we prove a novel type of RR error bound that deals with the products of the errors, rather than the sums. Third, we establish majorization bounds for the relative errors. We extend our bounds to the case $\dim\X\leq\dim\Y<\infty$ in Hilbert spaces and apply them in the context of the finite element method.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the error estimation problem of the Rayleigh - Ritz (RR) method when approximating the eigenvalues of a Hermitian operator \(A\). Specifically, the paper focuses on how to accurately estimate the error between the approximate eigenvalues (i.e., Ritz values) generated by the Rayleigh - Ritz method by using principal angles when given two subspaces \(X\) and \(Y\) of the same dimension, where \(X\) is an \(A\)-invariant subspace.
### Summary of Main Problems:
1. **Error Estimation of the Rayleigh - Ritz Method**:
- The Rayleigh - Ritz method approximates the eigenvalues of a Hermitian operator \(A\) by finding the stationary points of the Rayleigh quotient (called Ritz values) on a given trial subspace.
- If the trial subspace \(X\) is \(A\)-invariant, then the Ritz values are exactly some eigenvalues of \(A\).
- Given two subspaces \(X\) and \(Y\) of the same dimension, where \(X\) is \(A\)-invariant, the absolute variation \(|Λ((P_X A)|_X)-Λ((P_Y A)|_Y)|\) represents the absolute eigenvalue approximation error of the Ritz values with respect to \(X\) and \(Y\).
2. **Majorization Error Bounds**:
- The first main result of the paper is a sharp majorization - type Rayleigh - Ritz error bound, expressed in terms of principal angles, applicable to any \(A\)-invariant \(X\).
- The second main result is to propose a new Rayleigh - Ritz error bound that deals with the product of errors instead of the sum.
- The third main result is to establish a majorization bound for the relative error.
3. **Extension to the Finite Element Method (FEM)**:
- The paper extends these error bounds to the case of \(\dim X\leqslant\dim Y <\infty\) and applies them in the finite element method.
### Formula Representation:
- Let \(X\) and \(Y\) be two subspaces of the same dimension, and \(X\) is \(A\)-invariant.
- Define the cosine square of the principal angles as \(\cos^{2}Θ(X, Y)=Λ((P_X P_Y)|_X)\), where \(Λ((P_X P_Y)|_X)\) are the eigenvalues arranged in ascending order.
- The variation of the Ritz values can be expressed as \(|Λ((P_X A)|_X)-Λ((P_Y A)|_Y)|\).
### Sharp Majorization Error Bound:
\[
|Λ((P_X A)|_X)-Λ((P_Y A)|_Y)|\prec_w (\lambda_{\max}(X + Y)-\lambda_{\min}(X + Y))\sin^{2}Θ(X, Y)
\]
where \(\lambda_{\max}(X + Y)\) and \(\lambda_{\min}(X + Y)\) are the maximum and minimum eigenvalues of the operator \((P_X + P_Y A)|_{X + Y}\) respectively.
Through these formulas and theories, the paper aims to provide a more accurate and effective error estimation method for the Rayleigh - Ritz method, especially for its application in the finite element method.