First-Order Random Phase Approximation (FRPA)

• FRPA is a beyond-RPA method that adds first-order self-energy, exchange, and vertex corrections to improve standard bubble-level approximations.
• The Faddeev variant employs a decomposition in the 2p1h/2h1p sector to construct an energy-dependent self-energy while avoiding double counting of diagrams.
• Applications in atomic, molecular, and low-dimensional systems demonstrate FRPA’s competitive accuracy with methods like CCSD and ADC(3) for correlation energies and ionization potentials.

FRPA is an acronym used in more than one closely related many-body context. In the cited literature, it denotes both the first-order random-phase approximation, where the random-phase approximation is extended by first-order self-energy and exchange or vertex corrections in response functions and screening, and the Faddeev Random Phase Approximation, a Green’s-function method that couples a single electron to two-particle–one-hole and two-hole–one-particle excitations through RPA phonons (Gangwar et al., 23 Jul 2025, Pascucci et al., 27 Oct 2025, Barbieri et al., 2010). In both usages, the common objective is to move beyond a simple bubble-level RPA while preserving a diagrammatically controlled and computationally tractable framework.

1. Terminological scope and conceptual variants

The dual use of the acronym is structurally important because the two formalisms operate on different primary objects. The first-order response-theory variant corrects the density–density response or proper polarization operator directly. The Faddeev variant constructs an energy-dependent self-energy for the one-body propagator through a Faddeev decomposition of the $2p1h/2h1p$ sector. This suggests that FRPA is best understood as a family of beyond-RPA approximations rather than a single universally fixed algorithm (Gangwar et al., 23 Jul 2025, Barbieri et al., 2010).

Usage of FRPA	Primary object	Representative arXiv sources
First-order random-phase approximation	$\chi(q,\omega)$, $\Pi^*(q,\omega)$, screened interactions	(Gangwar et al., 23 Jul 2025, Pascucci et al., 27 Oct 2025, Berkelbach, 2018, Cances et al., 2011)
Faddeev Random Phase Approximation	$G(\omega)$, $\Sigma^*(\omega)$, $R(\omega)$ in $2p1h+2h1p$ space	(Barbieri et al., 2010, Degroote et al., 2010, Degroote, 2012)

In the first-order usage, standard RPA resums bubble diagrams and FRPA augments that description by adding the leading interaction corrections to the polarizability. In the Faddeev usage, FRPA upgrades the intermediate particle–hole and particle–particle interactions from TDA to RPA and uses Faddeev equations to avoid double counting in the six-point sector. The notational overlap is historical and methodological rather than strictly identity of formal content.

2. First-order response formulations beyond bubble RPA

In one-dimensional metallic quantum wires, the starting point is the exact density–density response $\chi(q,\omega)$, which in standard RPA is approximated by

$$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$

with $\chi_0$ the Lindhard polarizability. FRPA goes beyond this by expanding the full response to first order in the interaction and retaining the direct RPA term together with first-order self-energy and exchange corrections: $\chi(q,\omega)$0 From this response one obtains the static structure factor through the fluctuation–dissipation theorem,

$\chi(q,\omega)$1

and the correlation energy per particle through the adiabatic-connection formula,

$\chi(q,\omega)$2

This formulation was used to compute the structure factor, pair-correlation function, correlation energy, and ground-state energy under six transverse confinement models $\chi(q,\omega)$3 (Gangwar et al., 23 Jul 2025).

In the two-dimensional electron–hole bilayer superfluid, FRPA is formulated at the level of the proper polarization operator,

$\chi(q,\omega)$4

where $\chi(q,\omega)$5 contains all first-order exchange corrections and any surviving direct or self-energy terms. Because one cannot perturb around the noninteracting Fermi sea in a superfluid, the construction follows the modified perturbation method of Nozières and Schrieffer: an auxiliary linearizing Hamiltonian $\chi(q,\omega)$6 is added to $\chi(q,\omega)$7, and $\chi(q,\omega)$8 is treated as the perturbation. At first order in $\chi(q,\omega)$9, all direct Hartree terms cancel by charge neutrality and all self-energy insertions cancel against those already built into $\Pi^*(q,\omega)$0; only the exchange–vertex diagrams survive. The resulting normal and anomalous polarizations $\Pi^*(q,\omega)$1 and $\Pi^*(q,\omega)$2 enter the screened intralayer and interlayer Coulomb interactions through

$\Pi^*(q,\omega)$3

or, in the static limit, through the explicit $\Pi^*(q,\omega)$4 screened matrix written in terms of $\Pi^*(q,\omega)$5, $\Pi^*(q,\omega)$6, $\Pi^*(q,\omega)$7, and $\Pi^*(q,\omega)$8 (Pascucci et al., 27 Oct 2025).

A mathematically distinct but conceptually related first-order setting appears for insulating and semiconducting crystals. There, the time-dependent Hartree dynamics—also called the random phase approximation in the physics literature—is expanded to first order about the perfect crystal. The linear term

$\Pi^*(q,\omega)$9

defines the microscopic frequency-dependent polarization operator $G(\omega)$0, and homogenization yields a macroscopic Maxwell–Gauss equation for $G(\omega)$1 (Cances et al., 2011). This places first-order RPA on a rigorous operator-theoretic footing in a periodic setting.

3. Faddeev FRPA as a propagator self-energy method

The Faddeev Random Phase Approximation starts from the exact one-body Green’s function and the Dyson equation,

$G(\omega)$2

or equivalently

$G(\omega)$3

The irreducible self-energy is separated into a static mean-field contribution and a dynamical part that depends on an irreducible polarization propagator $G(\omega)$4 in the $G(\omega)$5 sector: $G(\omega)$6 In molecular notation, the same structure is written with an effective second-order interaction $G(\omega)$7, including the same vertex correction $G(\omega)$8 as in ADC(3), and $G(\omega)$9 truncated to its exact third-order expansion restricted to the subspaces of $\Sigma^*(\omega)$0 and $\Sigma^*(\omega)$1 configurations (Barbieri et al., 2010, Degroote et al., 2010).

The defining step is the Faddeev decomposition of the six-point propagator. Directly solving a Bethe–Salpeter equation for the full $\Sigma^*(\omega)$2 propagator would double-count diagrams. FRPA therefore introduces three partitioned amplitudes according to which pair of lines interacts last. In one compact form,

$\Sigma^*(\omega)$3

while in the molecular derivation the same idea appears as

$\Sigma^*(\omega)$4

This construction generates the intended infinite subset of diagrams once and only once, and preserves Pauli consistency in the $\Sigma^*(\omega)$5 space (Barbieri et al., 2010, Degroote, 2012).

The intermediate phonons are themselves described at the RPA level. In the particle–particle sector one diagonalizes the pp-RPA matrix, while in the particle–hole sector one diagonalizes the ph-RPA matrix. The positive-energy solutions are the RPA phonon frequencies. After solving the Faddeev eigenproblem and constructing the dressed coupling matrix $\Sigma^*(\omega)$6, one either writes the self-energy in pole form,

$\Sigma^*(\omega)$7

or diagonalizes the enlarged single-particle-plus-$\Sigma^*(\omega)$8-plus-$\Sigma^*(\omega)$9 Hamiltonian to obtain one-body poles and residues (Degroote et al., 2010).

4. Relation to ADC(3), CCSD, ring-CCD, and conventional RPA

The tightest relationship in molecular Green’s-function theory is with ADC(3). In Faddeev FRPA, all diagrams through third order are identical between FRPA and ADC(3), but FRPA extends ADC(3) by dressing the particle–hole and particle–particle interaction vertices to RPA order, thereby resumming an infinite class of bubble-chain and ladder diagrams in each channel. If every RPA problem is replaced by its TDA limit, or if the unstable ph channel is treated by ph-TDA only, the method reduces precisely to the standard ADC(3) result (Barbieri et al., 2010, Degroote et al., 2010).

The relation to coupled-cluster theory is particularly explicit in the formulation based on approximate equation-of-motion ring coupled-cluster doubles. In that setting, RPA excitation energies are identical to those calculated using an approximation to EOM-CCD, provided three approximations are made: the ground-state double-excitation amplitudes are obtained from the ring-CCD equations; the EOM eigenvalue problem is truncated to the single-excitation subspace; and the similarity transformation of the Fock operator is neglected because it corresponds to a dressing of the single-particle propagator not present in conventional RPA. Within that construction, first-order RPA is defined by taking only the first-order solution of the ring-CCD amplitudes,

$R(\omega)$0

and inserting it into the singles-block Hamiltonian. The resulting FRPA secular equation is

$R(\omega)$1

or, in compact form,

$R(\omega)$2

Its excitation energies are the eigenvalues of $R(\omega)$3, and its transition amplitudes are the corresponding eigenvectors (Berkelbach, 2018).

The comparison with CCSD in atomic benchmarks is empirical rather than formal. Light-atom calculations explicitly compare FRPA with CCSD, ADC(3), and experiment, while molecular applications compare equilibrium properties and ionization energies with CCSD(T) benchmarks. A plausible implication is that FRPA occupies an intermediate position: it retains the explicit propagator and screening structure of RPA-based methods while remaining competitive with established coupled-cluster quality indicators in weakly to moderately correlated regimes (Barbieri et al., 2010, Degroote et al., 2010).

5. Applications and numerical benchmarks

For light atoms up to Ar, the Faddeev FRPA was tested by calculating total and ionization energies and comparing with CCSD, ADC(3), and experiment. Even for the two-electron systems He and $R(\omega)$4, the inclusion of RPA effects leads to satisfactory results and therefore does not over-correlate the ground state. The method becomes progressively better for larger atomic numbers, where it gives about $R(\omega)$5 more correlation energy and shifts ionization potentials by $R(\omega)$6–$R(\omega)$7 with respect to ADC(3); the corrections for ionization potentials consistently reduce discrepancies with experiment (Barbieri et al., 2010). In the more detailed benchmark summary, RMS deviations versus experiment excluding Be are $R(\omega)$8 for FRPAc versus $R(\omega)$9 for ADC(3)c and $2p1h+2h1p$0 for CCSD, while for outer-valence ionization energies ADC(3)c / FTDAc has an RMS error $2p1h+2h1p$1 and FRPAc reduces it to $2p1h+2h1p$2 for $2p1h+2h1p$3. In Ar, the $2p1h+2h1p$4 orbital is correctly found to split into a quasiparticle peak plus satellite, and FRPA brings both peak positions and spectroscopic strengths into much better agreement with the measured photoemission spectrum (Barbieri et al., 2010).

For molecules, the first published applications concerned diatomic systems at equilibrium geometry, with additional analysis of the dissociation limit. In a cc-pVDZ basis for systems including $2p1h+2h1p$5, HF, HCl, BF, $2p1h+2h1p$6, $2p1h+2h1p$7, and CO, FRPA ground-state energies and equilibrium bond lengths differ from CCSD(T) by at most $2p1h+2h1p$8–$2p1h+2h1p$9 and from experiment by similar amounts. Vertical ionization energies from FRPAc are typically within $\chi(q,\omega)$0–$\chi(q,\omega)$1 of experiment, on par with or slightly better than ADC(3) (Degroote et al., 2010).

In one-dimensional metallic quantum wires, first-order FRPA was used to examine how transverse confinement modifies correlation effects in the ground state. For the ultrathin wire $\chi(q,\omega)$2 in the high-density limit, the correlation energy for $\chi(q,\omega)$3 (harmonic), $\chi(q,\omega)$4 (cylindrical), and $\chi(q,\omega)$5 (harmonic-delta) approaches

$\chi(q,\omega)$6

in agreement with exact results in this limit, whereas $\chi(q,\omega)$7, $\chi(q,\omega)$8, and $\chi(q,\omega)$9 instead tend to $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$0 (Gangwar et al., 23 Jul 2025). FRPA agrees very well with RPA only at extremely high density, but remains in good agreement with quantum Monte Carlo up to $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$1–1 depending on $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$2. All six confinement models yield a pronounced peak in $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$3 at $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$4, and the sensitivity of $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$5 to the choice of confinement is greatest in the thin-wire, low-density regime, where it can differ by up to $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$6 between the soft models $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$7 and the hard models $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$8 (Gangwar et al., 23 Jul 2025).

In the two-dimensional electron–hole bilayer superfluid, the first-order corrections to screening were found to be modest. At very low density, strong cancellations between normal and anomalous components make screening of the interactions negligible not only within RPA but also with the first-order corrections included. As density increases, the normal–anomalous cancellation weakens and screening becomes increasingly significant; the first-order corrections amplify the normal–anomalous difference but only at large momenta exchanged in the two-particle scattering, so their effect on the interactions remains modest. Up to the density of maximum superfluid gap, RPA is therefore an excellent approximation for screening and for effective electron–hole pairing in this system (Pascucci et al., 27 Oct 2025).

6. Instabilities, rigorous status, and domain of validity

A recurring limitation is the appearance of RPA instabilities in small gaps, near-degeneracies, or bond breaking. In neutral Be, the $$\chi_{\rm RPA}(q,\omega)=\frac{\chi_0(q,\omega)}{1 - V(q)\,\chi_0(q,\omega)}\,,$$9–$\chi_0$0 near-degeneracy forces a ph-RPA instability; in that case the authors fell back on ph-TDA only for that channel, recovering precisely the standard ADC(3) result, and the full FRPAc result reverts to ADC(3)c accuracy (Barbieri et al., 2010). In molecular dissociation, stretching $\chi_0$1 beyond approximately $\chi_0$2 causes the ph spin-triplet channel to develop a zero or complex mode because the Hartree–Fock reference becomes unstable to triplet dissociation. The FRPA equations then become singular. Replacing only the unstable triplet-RPA phonon by its TDA counterpart yields a mixed FRPA/TDA that remains finite but becomes too close to pure FTDA; neither FRPA nor FTDA recovers the correct full-CI dissociation limit, and a fully self-consistent Green’s-function treatment allowing fragmentation of single-particle strength is required to restore the atomic limit (Degroote et al., 2010).

The two-site Hubbard molecule gives a transparent reduced-dimensionality illustration of the same phenomenon. Standard FRPA reproduces the exact two-electron ground-state energy perfectly up to the RPA triplet instability at $\chi_0$3, then breaks down because no real RPA poles exist. The thesis summary identifies practical cures: replacing the unstable channel by TDA, dressing the single-particle propagator in the Bethe–Salpeter kernels, or pursuing SCRPA, which enforces the killing condition at the price of requiring the density matrix (Degroote, 2012). These observations delimit the regime in which FRPA can be used as a black-box method.

At the same time, the formal status of first-order RPA is unusually strong in crystals. For the nonlinear time-dependent Hartree equation in a defective crystal, existence, uniqueness, conservation of generalized trace, convergence of the Dyson series in the defect-state space, and positivity properties of the linear-response operator $\chi_0$4 are established rigorously, and the macroscopic dielectric equation is derived from the first-order approximation by homogenization (Cances et al., 2011). This rigorous formulation does not remove the physical limitations associated with unstable reference states, but it clarifies the operator-theoretic meaning of first-order RPA response.

Taken together, the literature supports a stable core interpretation. FRPA is a controlled beyond-RPA framework that either adds first-order self-energy and exchange corrections to response functions or incorporates RPA phonons into a Faddeev-resummed propagator self-energy. In the atomic and molecular setting, it combines the nonperturbative screening of RPA phonons with full third-order perturbation theory and a Pauli-consistent $\chi_0$5 configuration mixing; in reduced-dimensional electron systems, it captures high-density limits, confinement dependence, and screening corrections beyond the simple bubble approximation (Barbieri et al., 2010, Gangwar et al., 23 Jul 2025). Its principal weaknesses are precisely those expected of any RPA-based construction: sensitivity to unstable reference states, degraded behavior near dissociation and near-degeneracy, and the need for more self-consistent or fragmented propagators when collective and static correlation effects become inseparable.