Matrix Random Phase Approximation (mRPA)

• mRPA is a computational framework that recasts Random Phase Approximation into explicit matrix form, enabling systematic analysis of collective excitations and correlations in quantum many-body systems.
• It leverages block-structured matrices and generalized eigenvalue problems to differentiate physical, unstable, and spurious modes, ensuring rigorous stability and duality analyses.
• mRPA underpins scalable numerical methods for diverse applications—from multiorbital Hubbard models to matrix product state simulations—enhancing both computational efficiency and analytical insight.

Matrix Random Phase Approximation (mRPA) is a general mathematical and computational framework that recasts the Random Phase Approximation—a cornerstone linear response theory for quantum many-body systems—into explicit matrix form. This enables efficient algorithmic implementations, rigorous analysis of physical and unphysical modes, and systematic treatment of collective excitations, correlations, and response functions for diverse systems, including matrix product states, multiorbital lattice models, crystals, and finite nuclei. The mRPA formalism subsumes a wide range of applications by expressing the RPA eigenproblem, response matrices, and energy functionals in terms of block-structured matrices and their spectral properties, facilitating both analytic and large-scale numerical studies.

1. Mathematical Formulation of mRPA

The core of mRPA is the generalized eigenvalue problem for the stability matrix $\mathsf{S}$ with an indefinite metric (norm matrix) $\mathsf{N}$, typically of the form $\mathsf{N} = \mathrm{diag}(1, -1)$. The RPA equations are naturally cast as

$$\mathsf{S}\,\mathbf{x}_\nu = \omega_\nu\,\mathsf{N}\,\mathbf{x}_\nu$$

or equivalently as the non-Hermitian eigenproblem $$\mathsf{N}\mathsf{S}\,\mathbf{x}_\nu = \omega_\nu \mathbf{x}_\nu$$. The stability matrix $\mathsf{S}$ encodes the second derivatives of the mean-field energy functional and has symmetry properties ($\mathsf{S} = \mathsf{S}^\dagger$, $\Sigma_x \mathsf{S}^* \Sigma_x = \mathsf{S}$), where $\Sigma_x$ is a “conjugation” block matrix pairing forward and backward amplitude sectors (Nakada, 2016).

The spectrum of $\mathsf{N}\mathsf{S}$ is generally non-Hermitian, yielding physical, unphysical, and spurious modes, organized by precise duality relations (UL- and LR-dualities), and possible Jordan block structures when $\mathsf{N}\mathsf{S}$ is not diagonalizable. Variational principles (e.g., Dirac-Frenkel TDVP for MPS (Kinder et al., 2011)) and linearizations yield mRPA naturally as the systematic expansion around stationary points of the reference manifold, with the resulting matrix equations capturing optimal evolution and linear response.

2. Spectral Properties: Physical, Unphysical, NG Modes and Dualities

The mRPA eigenvalue problem admits a classification:

• Physical modes (Class 1): Real, positive frequencies $\omega_\nu$ with positive $\mathsf{N}$-norms, corresponding to stable collective excitations.
• Negative-norm real modes (Class 2): Real frequencies with negative norm, signaling instability.
• Imaginary (Class 3) and complex quartet (Class 4) modes: Indicating oscillatory instabilities or mean-field saddle points.
• Null (Class 5): Zero frequency, associated with Nambu–Goldstone (NG) or spurious modes due to broken symmetries.

Jordan block structures may emerge, especially for NG modes, but their dimension is at most two in the case of a positive-semidefinite stability matrix, ensuring that all such modes can be separated out via canonical conjugate variables (Nakada, 2016). The biorthogonality, completeness, and expansion properties of the mRPA basis are maintained by rigorous duality constructs (UL for conjugate modes, LR for non-Hermitian biorthogonal pairing) (Nakada, 2016).

3. Construction of Response Functions, Transition Amplitudes, and Green’s Functions

Once mRPA collective modes $\{ \omega_q, X_q, Y_q \}$ are obtained, linear response theory is implemented by expressing the response of any observable to a perturbation as

$$\delta P(\omega) = \Pi_{PQ}(\omega)\, \delta Q(\omega),$$

where the mRPA response matrix is explicitly constructed from the eigenmodes. Poles of $\Pi(\omega)$ at $\pm \omega_q$ yield excitation energies, and residues give transition amplitudes $\langle \Psi_q | P | \Psi_0 \rangle$.

Generalized Green’s functions and fluctuation-dissipation relations extend to site-based formulations (e.g., MPS-based systems (Kinder et al., 2011)), with site–site response matrices and correlation functions directly projected from the mRPA eigenbasis. This enables quantification of dynamic correlations and spectral functions with full inclusion of collective mode contributions.

4. Algorithmic Implementations and Computational Scaling

mRPA formulations underpin efficient numerical methods across distinct contexts:

• Stochastic trace evaluation for correlation energies: mRPA recasts the RPA energy as $\mathrm{Tr}[f(A)]$, with $A$ a response matrix, enabling stochastic Hutchinson estimators and Chebyshev polynomial expansions for large systems at $O(N^2)$ or near-linear scaling (Neuhauser et al., 2012).
• Finite Amplitude Method (FAM): mRPA supplies black-box algorithms for constructing and diagonalizing full RPA or QRPA matrices via finite differences, facilitating large-scale spectroscopic calculations (Avogadro et al., 2013).
• Matrix-based response solvers: mRPA formulations for multiorbital models (lattice, Hubbard, DFT) provide matrix equations for susceptibilities, pairing kernels, and more, with diagrammatic corrections systematically incorporated (Altmeyer et al., 2016).

These approaches exploit explicit block structure and matrix operations for scalable, parallelizable implementations, with performance and memory characteristics governed by the size of the active space and underlying symmetries.

5. mRPA in Multiorbital, Crystalline, and Matrix Product State Settings

Matrix RPA generalizes to complex systems with internal degrees of freedom:

• Multiorbital Hubbard models: The mRPA approach yields matrix equations for spin, charge, and orbital susceptibilities, resumming bubble, ladder, and vertex-correction diagrams in the pairing interaction. Vertex corrections unique to multiorbital interactions—absent in scalar RPA—are generated automatically in mRPA, significantly modifying the superconducting pairing kernel (Altmeyer et al., 2016).
• Crystals and dielectric response: mRPA rigorously constructs microscopic polarization matrices indexed by Bloch momentum and reciprocal-lattice vectors, connecting to the macroscopic dielectric tensor via homogenization and the Adler–Wiser formula. Spectral, sum-rule, and analytic properties follow from the block matrix formalism (Cances et al., 2011).
• Matrix Product States: mRPA provides a variational, analytic linear response theory for MPS, enabling dynamical calculations and improved characterization of excitations (spin, charge) in 1D quantum systems at $O(k M^3)$ complexity (Kinder et al., 2011).

6. Statistical and Random Matrix Extensions: Doorway States and Large Configurational Spaces

mRPA is extended to statistical ensembles and random matrix backgrounds:

• Doorway states in mRPA: Coupling of collective states to randomized backgrounds yields non-Hermitian matrix RPA problems with doubled spectra, “cross-energy” couplings, and self-consistent Pastur equations. This produces nontrivial phenomena such as strong mutual attraction of doorway and background states, level shifts, and altered spreading width scaling compared to Hermitian shell-model formalisms (Pace et al., 2014).
• Ensemble averaging and spectral functions: mRPA enables analytic and numeric synthesis of level densities, strength distributions, and resonance behaviors in large configuration spaces, with statistical foundation for resonance characterization in nuclear, mesoscopic, and condensed matter contexts.

7. Practical Consequences, Interpretational Notes, and Generalizations

mRPA constitutes a flexible linear-algebraic foundation for RPA-based quantum many-body theory, supporting:

• Immediate stability analysis from mRPA spectra and norms, systematic removal of spurious modes by projector constructions, and maintenance of completeness in transition operator expansions (Nakada, 2016).
• Consistency of energy-weighted sum rules and commutator identities, contingent on full inclusion of physical, unphysical, and spurious mRPA eigenmodes.
• Algorithmic reduction of generalized eigenproblems to Hermitian form under positive-definite stability conditions, facilitating robust diagonalization and mode selection (Nakada, 2016).
• Adaptability of the mRPA procedures to more refined frameworks, such as inclusion of exchange–correlation effects in TDDFT, and to non-insulators where Fermi surface singularities appear (Cances et al., 2011).

A plausible implication is that future advances in many-body quantum simulations—requiring analytic tractability, large-scale numerical feasibility, and rigorous treatment of correlations—will increasingly leverage mRPA and its variants as foundational building blocks. Extensions to finite temperature, continuum coupling, and more general variational manifolds remain active areas for development.