Nonlinear Eigenvalue Problem Techniques in Matrix Analysis

The study of nonlinear eigenvalue problems (NEPs) has matured into a vibrant field at the intersection of numerical analysis, operator theory and applied mathematics. In contrast to the classical linear eigenproblem, NEPs involve a matrix or operator that depends on the eigenparameter in a nonlinear fashion, often as a polynomial, rational or analytic function. Central approaches include linearisation strategies that embed the original problem into a larger linear framework, contour integration techniques that isolate spectral components via complex integration of the resolvent, projection and subspace iteration methods that approximate invariant subspaces, and Newton‐type or fixed‐point schemes that exploit smooth parameter dependence. Advances in the conditioning and stability analysis of NEPs have driven the design of robust solvers capable of handling large‐scale systems arising in vibration analysis, control theory, photonic crystals and fluid–structure interaction. Recent efforts also address rational eigenvalue problems by constructing minimal state‐space realizations and by developing root‐polynomial frameworks that unify the treatment of poles and zeros. Together, these techniques have expanded the scope of matrix analysis into infinite‐dimensional settings, enhanced the computational reliability of eigenvalue solvers, and opened new avenues for high‐performance implementations on emerging hardware architectures.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Contemporary work has refined normalisation strategies for both linear and nonlinear eigenvalue problems. A recent study introduced regularised nonhomogeneous formulations that supplement the eigen-equation with explicit scaling conditions, enabling derivative-free merit functions and accelerated convergence via golden‐section and Newton iterations. This framework yields precise real and complex eigenvalues while simultaneously producing accurate eigenvectors, and it demonstrates marked improvements in speed and robustness over classical approaches.

Advances in infinite-dimensional NEPs have been reported through subspace iteration algorithms for analytic Fredholm valued functions. By exploiting contour integration and the analytic Fredholm theorem, this method approximates eigensubspaces for spectral components of operator-valued functions, effectively extending FEAST-type solvers from finite to infinite dimensions. Numerical experiments confirm its efficacy for polynomial and rational eigenvalue problems within functional settings.

In the realm of rational matrix problems, systematic theory and computation of root vectors have been developed. The approach constructs minimal state-space realizations of rational matrices and then applies staircase algorithms on the associated linearized pencils. This dual theory–algorithm framework handles coalescent poles and zeros, covers singular and full-rank cases, and provides preprocessing steps to remove singularities while retaining the ability to recover original root vectors.

Nonlinear Eigenvalue Problem Techniques in Matrix Analysis publication trend

The graph below shows the total number of articles in nonlinear eigenvalue problem techniques in matrix analysis across all publications each year (not limited to Nature Index journals).

Technical terms

Nonlinear eigenvalue problem: An eigenvalue problem in which the matrix or operator depends nonlinearly on the spectral parameter, typically through polynomial, rational or analytic functions.

Matrix polynomial: A polynomial whose coefficients are matrices, leading to polynomial eigenvalue problems where eigenvalues satisfy det(∑ A_i λ^i)=0.

Linearisation: The process of reformulating a polynomial or rational eigenvalue problem as a larger linear eigenproblem by embedding into an extended matrix pencil.

Contour integration methods: Techniques that compute spectral projectors or eigenpairs by numerically integrating the resolvent (A(λ)^{-1}) around closed contours in the complex plane.

Fredholm operator: An operator on an infinite‐dimensional space that can be written as the identity plus a compact perturbation, yielding analytic dependence on the spectral parameter.

Root polynomial: A vector polynomial that encodes the local multiplicity and structure of eigenvalues and poles in polynomial or rational matrix eigenproblems.

References

1. Regularized Normalization Methods for Solving Linear and Nonlinear Eigenvalue Problems. Mathematics (2023).
2. A Subspace Iteration Algorithm for Fredholm Valued Functions. Mathematical Problems in Engineering (2015).
3. Root vectors of polynomial and rational matrices: Theory and computation. Linear Algebra and its Applications (2023).