Config Interaction & Complex-Energy Formalisms

• Configuration interaction with complex-energy formalisms rigorously models electronic resonances by employing non-Hermitian quantum mechanics.
• Advanced CI frameworks integrate complex scaling, CAP, and relativistic corrections to extract precise resonance positions and widths.
• These methodologies enable benchmark-level simulations of atomic and molecular metastable states, reducing discrepancies with experimental data.

Configuration interaction (CI) and complex-energy formalisms constitute foundational methodologies for the quantum mechanical treatment of electronic resonances in atoms, ions, and molecules. While CI is established for bound-state correlation, modern extensions employing complex scaling (CS) and complex absorbing potentials (CAP) enable the rigorous, non-Hermitian representation of metastable (autoionizing) states. These approaches deliver direct access to resonance positions ($E_R$) and widths ($\Gamma$), transcending the limitations of conventional, real-symmetric frameworks.

1. Non-Hermitian Quantum Mechanics and Resonance Theory

Resonance states are quantum states embedded in the continuum, characterized by complex energies $E = E_R - i \Gamma/2$, where $E_R$ is the position and $\Gamma$ is the width, corresponding to the inverse lifetime. Standard Hermitian Hamiltonians cannot capture the outgoing flux of such metastable states; this motivates the use of non-Hermitian extensions. Both complex scaling (CS) and complex absorbing potential (CAP) methodologies render resonance wavefunctions square-integrable by analytic continuation (${\bf r} \to {\bf r} e^{i\theta}$ for CS, addition of $-i\eta W$ for CAP), permitting their direct treatment within CI frameworks (Zaytsev et al., 2019, Damour et al., 2024).

2. Relativistic Configuration Interaction and Complex Scaling (CS-CI)

For atomic and ionic systems, especially those exhibiting relativistic effects, the CI formalism is predicated on the Dirac–Coulomb–Breit (DCB) Hamiltonian: $$H_{\rm DCB} = \sum_{i=1}^N h_D(i) + \sum_{i<j} [V_C(i,j) + V_B(i,j)],$$ where $h_D(i)$ is the one-electron Dirac operator, $V_C$ is the Coulomb interaction, and $V_B$ is the Breit interaction [Eq. (1), (Zaytsev et al., 2019)]. Standard CI expands eigenstates as linear combinations of antisymmetrized configuration-state functions (CSFs): $\Psi_{PJM} = \sum_{r} c_r\,\Phi_r\,,$ leading to the CI secular equation.

Complex scaling is implemented via uniform rotation of all electronic coordinates: $$r \longrightarrow r\,e^{i\theta},\quad \theta\in(0,\tfrac{\pi}{2}),$$ resulting in a non-Hermitian, complex-symmetric scaled Hamiltonian, $H_{\rm DCB}^{(\theta)}$ [Eq. (5)]. Resonance poles are isolated as discrete complex eigenvalues stationary with respect to $\theta$ in an “allowed” window, with physical resonance parameters extracted as

$$E_R = \mathrm{Re}\, E(\theta_{\mathrm{opt}}),\qquad \Gamma = -2\,\mathrm{Im}\,E(\theta_{\mathrm{opt}})$$

[Eq. (9)].

3. Selected Configuration Interaction and Complex Absorbing Potentials (CAP-SCI)

For molecular resonances, selected configuration interaction (SCI; e.g., CIPSI variants) has been adapted to non-Hermitian CAP frameworks (Damour et al., 2024). Starting from the physical electronic Hamiltonian $H_0$, CAP introduces an imaginary, one-body “box” potential: $H(\eta) = H_0 - i \eta W,$ with $\eta > 0$ the CAP strength and $W$ positive semi-definite: $$w(\alpha) = (|\alpha|-\alpha_0)^2 \text{ for } |\alpha| > \alpha_0,\quad 0 \text{ otherwise}.$$

The variational SCI wave function is expanded in determinants as $$|\Psi_{\text{sta}}(\eta)\rangle = \sum_{I \in \mathbb{I}} c_I(\eta)|I\rangle$$, with energies evaluated via a complex-symmetric CI matrix. External determinants are selected by Epstein–Nesbet second-order perturbation theory: $$e^{(2)}_\alpha(\eta) = \frac{|\langle \alpha \Vert H(\eta) \Vert \Psi_{\text{sta}}(\eta) \rangle|^2}{E_{\text{sta}}(\eta) - \langle \alpha \Vert H(\eta) \Vert \alpha \rangle},$$ and ranked by $|e^{(2)}_\alpha(\eta)|$ for balanced convergence of $\mathrm{Re}\,E$ and $\mathrm{Im}\,E$.

Extrapolation protocols use the robust “absolute value” PT2 correction: $$aPT2(\eta) = \sum_\alpha |\mathrm{Re}\,e^{(2)}_\alpha(\eta)| + i \sum_\alpha |\mathrm{Im}\,e^{(2)}_\alpha(\eta)|,$$ allowing separate linear fits of $\mathrm{Re}\,E_{\text{sta}}$ vs. $\mathrm{Re}\,PT2$ and $\mathrm{Im}\,E_{\text{sta}}$ vs. $\mathrm{Im}\,aPT2$, converging to the FCI limit.

4. Stabilization and Basis-Balancing Techniques

Besides complex-energy extensions, Hermitian-based indirect strategies such as the stabilization method (SM) and basis-balancing method (BBM) can estimate resonance parameters (Zaytsev et al., 2019). SM scans real basis parameters (e.g., radial grid dilation $\gamma$) for extrema in the real CI eigenvalues; BBM adjusts grid properties so resonances appear midway between continuum-like states. These approaches deliver results that differ from direct CS-CI calculations by $\sim$1–10 meV for atomic resonances, indicating the precision gain offered by explicitly non-Hermitian treatments.

5. Implementation: Basis Construction, Extraction Protocols, and Corrections

For CS-CI, B-spline dual-kinetic-balance bases are used. The construction entails a maximal orbital angular momentum $L_{\max}$ and discrete pseudo-spectrum generation in a finite box, supporting subsequent CI configuration assembly. Nuclear recoil and QED corrections are added:

• The mass-shift operator $H_{\rm MS}$ (Shabaev 1985) is included in first-order perturbation theory.
• Model QED operators $H_{\rm QED}$ are incorporated in separate CI calculations, the difference taken as a correction.

CAP-SCI calculations require careful CAP onset placement (e.g., $(x_0, y_0, z_0)$ coordinates) and basis augmentation (e.g., adding diffuse functions). Choice of orbitals—especially natural orbitals (NOA) for the resonant anion—accelerates convergence, reducing determinant count and extrapolation errors. Semistochastic or deterministic PT2 evaluation and matrix diagonalization algorithms are required for large CI expansions (up to $10^8$ determinants).

6. Numerical Results and Methodological Trends

In the CS-CI study of He-like ions from Boron ($Z=5$) to Argon ($Z=18$), resonance energies for the $2s^2 \,^1S_0$ line increase approximately quadratically with $Z$, while widths grow modestly (from $\sim6.7\times10^{-3}$ to $\sim8.1\times10^{-3}$ a.u.). Metastable states $2p_{1/2}2p_{3/2}\,(^3P_1,\,^3P_2)$ exhibit strong relativistic suppression of widths ($<10^{-6}$ a.u.). Comparison with nonrelativistic and MBPT+CS calculations demonstrates agreement within $\sim10^{-4}$ a.u. in energy and few percent in width (Zaytsev et al., 2019).

For CAP-SCI applied to shape resonances in $\mathrm{N}_2^-$ and $\mathrm{CO}^-$, full CI-quality accuracy is achieved. Notably, high-order correlation shifts CAP-EOM-CCSD results by up to 0.1 eV; e.g., $\mathrm{N}_2^-$ ($^2\Pi_g$) resonates at $E_R=2.449(1)$ eV, $\Gamma=0.391(3)$ eV (CAP-exFCI), compared to $E_R=2.487$ eV, $\Gamma=0.417$ eV (CAP-EOM-EA-CCSD) (Damour et al., 2024).

System	Method	$E_R$ (eV)	$\Gamma$ (eV)
N$_2^-$	CAP-EOM-EA-CCSD	2.487	0.417
	CAP-exFCI	2.449(1)	0.391(3)
CO$^-$	CAP-EOM-EA-CCSD	2.088	0.650
	CAP-exFCI	2.060(8)	0.611(3)

A plausible implication is that full CI correlation is essential to quantitatively explain experiment–the CAP-SCI methodology halves the observed discrepancies with experiment relative to lower-level approaches.

7. Prospects and Future Directions

Complex-energy CI methodologies are progressing toward systematically improvable and benchmark-level treatments of resonances. Refinements include:

• Full CAP parameter extrapolation ($\eta \rightarrow 0$ polynomial fits),
• Orbital optimization in presence of CAP (complex natural or energy-optimized orbitals),
• Application to higher-order resonances (Feshbach, $2p-1h$),
• Integration with complex-basis and further scaling techniques,
• Thorough analysis of basis-set and CAP-form dependencies.

These developments position complex-energy CI approaches, especially CAP-SCI, as central tools for resonance phenomena in both atomic and molecular regimes (Damour et al., 2024).