Resonance Sum Rules (RSR)

• Resonance Sum Rules (RSR) are analyticity-driven frameworks that use weighted spectral integrals to isolate resonance contributions from background effects.
• They combine techniques such as OPE-based inverse reconstructions, kernel-designed finite-energy sum rules, and chiral spectral modeling to match theory with resonance observables.
• RSR methodologies are applied across QCD inverse problems, electromagnetic scattering, and many-body theory, offering versatile, consistency-driven approaches to extract resonance parameters.

Resonance Sum Rules (RSR) is a context-dependent label for analyticity-based frameworks in which resonance contributions are isolated, reconstructed, or constrained through weighted spectral relations. In the literature surveyed here, the term is used for inverse reconstructions of hadronic spectral densities from operator-product expansions (OPE), finite-energy sum rules with kernels chosen to suppress unknown continuum contributions, resonance-saturated realizations of chiral and dispersive constraints, and moment sum rules for collective strength functions in many-body theory. The shared structure is the replacement of direct line-shape modeling by contour integrals, spectral moments, or weighted kernels that connect resonance observables to asymptotic or low-energy dynamical input (Li et al., 2020, Nasrallah et al., 2023, Delion et al., 2016).

1. Scope and common analytic structure

Across its different usages, RSR begins from a correlator, scattering amplitude, or strength function with a spectral representation, and then imposes integral constraints that privilege resonance contributions over poorly known background. In QCD correlator applications this usually means matching an OPE to a hadronic spectral density; in finite-energy sum rules it means integrating an amplitude against a polynomial or pinched kernel; in dispersive electromagnetic problems it means relating low-energy constants or helicity amplitudes to weighted cross-section integrals; in many-body theory it means taking moments of a strength function and enforcing commutator identities (Li et al., 2020, Nasrallah et al., 2023, Pascalutsa et al., 2010, Delion et al., 2016).

A compact way to compare the main usages is as follows.

Setting	Core object	RSR role
QCD inverse problem	$\Pi(q^2)$ or $\rho(s)$	Reconstruct resonance and continuum directly from OPE input
Kernel FESR	$\oint ds\, w(s)\Pi(s)$	Eliminate unknown spectral regions with analytic kernels
Chiral spectral modeling	$\rho_V(s)-\rho_A(s)$	Saturate Weinberg-type sum rules with explicit resonances
Dispersive electromagnetic processes	Polarized cross sections or VVCS amplitudes	Constrain resonance contributions through superconvergent moments
Extended RPA	$S_F(\omega)$ and $m_k$	Enforce strength-function sum rules and Goldstone properties

This multiplicity of meanings is not merely terminological. It reflects genuinely different technical constructions. A recurrent source of confusion is therefore the assumption that RSR denotes a single standardized formalism. The literature instead shows a family of related methods linked by analyticity, weighted spectral integration, and an explicit focus on resonance content (Hohler et al., 2012, Lensky et al., 2017, Delion et al., 2016).

2. Inverse-problem QCD RSR without duality thresholds

A particularly explicit usage appears in the vector-channel QCD construction where the two-current correlator is treated as an inverse problem rather than through the conventional SVZ pole-plus-continuum ansatz. In the isovector-vector channel one uses

$$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$

with

$$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$

Instead of imposing a continuum threshold $s_0$ and quark-hadron duality at low energy, the hadronic spectral density is reconstructed from the OPE side by solving a Fredholm equation of the form

$$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$

with $\rho(s)$0 and kernel $\rho(s)$1 (Li et al., 2020).

The OPE input retained there contains the perturbative term, the gluon condensate, the quark condensate contribution, and a dimension-six four-quark condensate term proportional to $\rho(s)$2. The analysis emphasizes that the dimension-six term is crucial: with perturbative and dimension-four terms only, no robust resonance solutions are found, whereas including the $\rho(s)$3 contribution generates nontrivial RSS minima associated with resonances (Li et al., 2020).

The inverse problem is regularized by expanding the continuum as

$\rho(s)$4

imposing the boundary conditions $\rho(s)$5 and $\rho(s)$6, and minimizing

$\rho(s)$7

Convergence checks indicate that truncation at $\rho(s)$8 suffices. A Borel-transformed variant was also examined, but the results with and without the Borel transform are numerically similar, so the Borel transformation is not essential in this inverse formulation (Li et al., 2020).

Within this RSR framework, a series of $\rho(s)$9 resonances emerges as distinct global minima:

$\oint ds\, w(s)\Pi(s)$0

$\oint ds\, w(s)\Pi(s)$1

$\oint ds\, w(s)\Pi(s)$2

$\oint ds\, w(s)\Pi(s)$3

For the $\oint ds\, w(s)\Pi(s)$4, replacing the pole by a Breit-Wigner-type parametrization gives

$\oint ds\, w(s)\Pi(s)$5

with the global minimum at $\oint ds\, w(s)\Pi(s)$6 (Li et al., 2020).

This formulation also sharpens a methodological critique of conventional duality-based multi-pole fits. When the standard step-function continuum ansatz is itself treated as an inverse problem, the resulting $\oint ds\, w(s)\Pi(s)$7-$\oint ds\, w(s)\Pi(s)$8 minima are monotonic and do not yield sensible excited-state solutions; triple-pole fits do not produce a reasonable $\oint ds\, w(s)\Pi(s)$9. The paper attributes this to the rigidity of the fixed-height continuum and argues that reconstructing the continuum together with the poles is essential for excited-state spectroscopy (Li et al., 2020).

3. Kernel-designed finite-energy and Gaussian resonance sum rules

A second major RSR tradition uses analytic kernels to suppress unknown spectral regions. In the nucleon case, the method is formulated as a variant of finite-energy sum rules in which the contour relation

$\rho_V(s)-\rho_A(s)$0

is combined with polynomial kernels that vanish at selected energies. For the scalar nucleon structure $\rho_V(s)-\rho_A(s)$1 built from the Ioffe-type current

$\rho_V(s)-\rho_A(s)$2

a practical pinched kernel is

$\rho_V(s)-\rho_A(s)$3

with $\rho_V(s)-\rho_A(s)$4 chosen as the mass of a dominant unwanted resonance in the continuum. The intent is to minimize or eliminate the continuum contribution and render the sum rule resonance-dominated (Nasrallah et al., 2023).

In that application the OPE for $\rho_V(s)-\rho_A(s)$5 is summarized by coefficients

$\rho_V(s)-\rho_A(s)$6

using $\rho_V(s)-\rho_A(s)$7, $\rho_V(s)-\rho_A(s)$8, and $\rho_V(s)-\rho_A(s)$9, with factorization assumed for the four-quark condensate. Stability in the contour radius $S_F(\omega)$0 is central. The resulting masses are

$S_F(\omega)$1

$S_F(\omega)$2

$S_F(\omega)$3

with typical quoted uncertainty of about $S_F(\omega)$4 and broad $S_F(\omega)$5 windows where the moments are flat or maximal (Nasrallah et al., 2023).

A different kernel philosophy underlies Gaussian QCD sum-rules for low-energy scalar mesons. There the hadronic side is

$S_F(\omega)$6

so that scanning $S_F(\omega)$7 at fixed $S_F(\omega)$8 probes a local slice of the spectrum. This is used to test resonance models for the $S_F(\omega)$9 isodoublet and $m_k$0 isotriplet and to connect QCD operators to chiral Lagrangian fields through universal, energy-independent scale factors (Fariborz et al., 22 Jan 2025).

That analysis introduces a background-resonance interference approximation motivated by $m_k$1 and $m_k$2 scattering, leading to extended distorted and generalized Breit-Wigner forms with interference parameters $m_k$3. The Gaussian sum-rules are reported to distinguish clearly between narrow, Breit-Wigner, and interference-improved models, especially for the broad $m_k$4. Among the tested parametrizations, the EGBW model gives the best universality measure,

$m_k$5

with fitted scale factors

$m_k$6

$m_k$7

and small residuals in the energy-independence diagnostics. This suggests, within that framework, that interference effects are not a detail of line-shape modeling but part of the sum-rule consistency conditions themselves (Fariborz et al., 22 Jan 2025).

4. Chiral-partner spectral functions and Weinberg-type constraints

In vector and axial-vector channels, RSR denotes a resonance-based realization of chiral sum-rule constraints. The spectral functions are written as

$m_k$8

with the continuum taken identical in the two channels and the resonance sector built from $m_k$9, $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$0, and their excitations. The key point is that once an identical continuum is imposed, the $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$1 moments are governed almost entirely by resonances, and satisfying the Weinberg-type sum rules forces the inclusion of excited states in both channels (Hohler et al., 2012).

The vacuum Weinberg-type relations are

$$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$2

$$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$3

$$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$4

with $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$5. The analysis concludes that a realistic $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$6 plus identical continuum requires a $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$7 to fit the vector $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$8 data, but that $$J_\mu(x)=\frac{1}{\sqrt{2}}\big(\bar u(x)\gamma_\mu u(x)-\bar d(x)\gamma_\mu d(x)\big),$$9, $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$0, $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$1, and identical continua still violate WSR-0, WSR-1, and WSR-2. Missing axial strength must therefore be added, leading to an inferred excited axial-vector state with

$$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$2

With this addition, the reported discrepancies are about $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$3 for WSR-0 and approximately zero for WSR-1 and WSR-2, whereas WSR-3 remains poorly saturated at about $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$4 because of sensitivity to the chirally breaking four-quark condensates (Hohler et al., 2012).

The same construction is simultaneously constrained by vacuum QCD sum rules. Assuming $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$5, the fit yields

$$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$6

with optimized average deviations of $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$7 in the vector channel and $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$8 in the axial channel. In this usage, RSR is therefore a simultaneous resonance fit to data and to OPE-based moment constraints rather than a direct pole-extraction algorithm (Hohler et al., 2012).

At finite temperature, the vector spectral function develops a pronounced low-energy shoulder and the $$\Pi_{\mu\nu}(q)= i\int d^4x\, e^{iq\cdot x}\,\langle0|T\{J_\mu(x)J_\nu(0)\}|0\rangle =(q_\mu q_\nu-q^2 g_{\mu\nu})\Pi(q^2).$$9 peak flattens off. The paper presents this as a possible sign of chiral restoration, but explicitly notes that a definitive statement requires constructing the finite-temperature axial-vector spectral function and checking the finite-temperature Weinberg-type sum rules in both channels (Hohler et al., 2012).

5. Dispersive electromagnetic and Compton-process resonance sum rules

A broad dispersive usage of RSR appears in light-by-light scattering, Compton scattering, and doubly virtual Compton scattering. In forward $s_0$0 scattering, analyticity and crossing yield superconvergent relations for polarized photon-fusion cross sections. The most prominent is

$s_0$1

together with Euler-Heisenberg sum rules

$s_0$2

In the hadronic sector the superconvergence relation is saturated nontrivially by cancellations between pseudoscalar and tensor mesons: $s_0$3 contributes $s_0$4 nb, $s_0$5 contributes $s_0$6 nb, $s_0$7 contributes $s_0$8 nb, while $s_0$9 contributes $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$0 nb and $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$1 contribute $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$2 nb in the narrow-width approximation (Pascalutsa et al., 2010).

For low-energy Compton scattering on arbitrary-spin targets, the resonance sum rules generalize GDH-Weinberg by allowing both mass-degenerate $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$3 intermediate states and narrow resonances with different masses. In one form the relation is

$$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$4

and in the more general case the low-energy side contains additional transition couplings $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$5. A central implication is that $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$6 can occur at tree level without spoiling perturbative unitarity if the necessary transition channels or narrow resonances are present (Grigoryan et al., 2012).

For nuclear Compton scattering, the low-energy photoabsorption integral produces the Thomas-Reiche-Kuhn relation

$$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$7

while a higher-energy constituent-quark-model sum rule matches the resonance-region integral, after Regge-plus-Pomeron subtraction, to a Thomson-like constituent-quark amplitude. Using modern proton data, the extracted $$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$8 pole for the proton is

$$\Pi^{\rm OPE}(Q^2)=\frac{f_V^2}{m_V^2+Q^2}+\int_0^\Lambda ds\,\frac{\rho^h(s)}{s+Q^2},$$9

which differs significantly from the proton Thomson term $\rho(s)$00, whereas for nuclei the extracted fixed pole remains consistent with the Thomson term within errors (Gorchtein et al., 2011).

In unpolarized doubly virtual Compton scattering, new forward sum rules relate low-$\rho(s)$01 slopes of structure-function moments to generalized polarizabilities and off-forward low-energy constants. The empirical proton analysis gives

$\rho(s)$02

in units of $\rho(s)$03, and infers small off-forward coefficients, including

$\rho(s)$04

in the same units. The explicit $\rho(s)$05 contribution to the muonic-hydrogen $\rho(s)$06-$\rho(s)$07 Lamb shift is found to be

$\rho(s)$08

small because the subtraction and inelastic pieces largely cancel (Lensky et al., 2017).

6. Photoproduction finite-energy sum rules and heavy-quark applications

In meson photoproduction, RSR is implemented through fixed-$\rho(s)$09 finite-energy sum rules that match low-energy resonance-region amplitudes to high-energy Regge behavior. For pion photoproduction the scalar invariant amplitudes $\rho(s)$10 satisfy contour relations whose low-energy side is summarized by moments

$\rho(s)$11

while the Regge side reads

$\rho(s)$12

The analysis finds that natural-parity exchanges dominate $\rho(s)$13 and $\rho(s)$14, with the isoscalar non-flip amplitude $\rho(s)$15 particularly large, and that zeros in the low-energy moments are largely moment-independent and map to zeros in Regge residues. The observed pattern is consistent with nonsense wrong-signature zeros in several channels, while the sizeable low-energy moments in $\rho(s)$16 create tension with the high-energy preference for suppressing this amplitude (Mathieu et al., 2018).

Eta photoproduction exhibits the same dual structure. Fixed-$\rho(s)$17 FESR relate low-energy $\rho(s)$18 amplitudes to Regge residues $\rho(s)$19, and the analysis argues that the absence of a dip near $\rho(s)$20-$\rho(s)$21 in $\rho(s)$22 photoproduction is explained by the dominance of the isovector natural-exchange helicity-flip amplitude $\rho(s)$23, which lacks the required NWSZ. By contrast, the corresponding dip in $\rho(s)$24 photoproduction is associated with the $\rho(s)$25 contribution in $\rho(s)$26. In the conservative Regge model with $\rho(s)$27, the beam asymmetry $\rho(s)$28 is predicted to remain close to $\rho(s)$29 for $\rho(s)$30, $\rho(s)$31, and $\rho(s)$32 GeV, with structure near the NWSZ region (Collaboration et al., 2016).

Heavy-quark applications show the same logic in a different guise. For the couplings between bottomonium $\rho(s)$33 states, the isovector resonances $\rho(s)$34, and a pion, analyticity and unitarity imply superconvergent sum rules for the form factor describing $\rho(s)$35. Eliminating the inaccessible $\rho(s)$36 term yields, for each $\rho(s)$37 resonance,

$\rho(s)$38

With currently available data, the sums through $\rho(s)$39 leave a residual of about $\rho(s)$40 MeV for $\rho(s)$41 and about $\rho(s)$42 MeV for $\rho(s)$43, so the two sum rules cannot be satisfied simultaneously unless production above $\rho(s)$44 shows a considerable dissimilarity between the yields of $\rho(s)$45 and $\rho(s)$46 (Voloshin, 2016).

7. Alternative usages, limitations, and current frontiers

In many-body theory the term RSR is used in a broader moment sense. For a Hermitian one-body operator $\rho(s)$47, the strength function

$\rho(s)$48

has moments

$\rho(s)$49

with the energy-weighted sum rule

$\rho(s)$50

Within standard RPA, self-consistent RPA, and renormalized RPA, preservation of this double-commutator identity is tied to the correct description of collective modes and, when continuous symmetries are broken, to the appearance of Goldstone modes. SCRPA and rRPA were shown to preserve the EWSR and Goldstone properties while incorporating correlated occupation numbers beyond the quasi-boson approximation (Delion et al., 2016).

A more specific nuclear example concerns dipole excitations in calcium and zirconium isotopes. There, ground-state correlations in phRRPA reduce both isoscalar and isovector energy-weighted sums, leading to apparent EWSR violation. For example, in $\rho(s)$51Ca the phRRPA exhausts $\rho(s)$52 of the isoscalar EWSR and gives $\rho(s)$53 in the isovector channel, whereas including particle-particle and hole-hole transitions in an effective RRPA restores the sums to $\rho(s)$54 and $\rho(s)$55, respectively. The same extension increases the ratio of pygmy to giant dipole energy-weighted strength in several isotopes (Hung et al., 2016).

Methodological limitations vary across the field but recur in recognizable forms. In correlator-based QCD RSR they include OPE truncation, factorization violation, width modeling, and continuum regularization. In photoproduction FESR they include subthreshold continuation, exchange degeneracy assumptions, and factorization-breaking effects. In chiral spectral modeling the major unresolved source of uncertainty is the four-quark condensate sector, reflected by the poor saturation of WSR-3. A plausible implication is that RSR often gains predictive power by shifting modeling freedom from a single continuum threshold to a richer set of analyticity and spectral-shape constraints, but it does not remove systematic uncertainty (Li et al., 2020, Hohler et al., 2012, Mathieu et al., 2018).

Conventional SVZ one-pole sum rules provide a contrasting perspective on how strongly resonance estimates can depend on continuum assumptions. In the light scalar isosinglet $\rho(s)$56 channel, however, analytic bounds independent of the perturbative threshold can still be derived:

$\rho(s)$57

These bounds were obtained by extremizing the Borel mass formula itself and are presented as intrinsic to the OPE structure rather than to any chosen $\rho(s)$58 (Afonin, 2016).

A more radical frontier extends RSR into the complex energy plane. By introducing a contour that reaches the second Riemann sheet and using a conformal mapping with a Gaussian kernel, a 2025 study extracted resonance poles and residues in the square-well potential. In that testbed the extracted $\rho(s)$59-wave and $\rho(s)$60-wave pole positions agreed with analytic values within $\rho(s)$61, and the residues within $\rho(s)$62. This suggests a possible route from real-axis sum rules to direct second-sheet pole determination, although the result was presented as a proof of concept rather than a mature hadronic application (Ou-Yang et al., 29 Sep 2025).

Taken together, these developments show that Resonance Sum Rules are best understood as a family of analyticity-driven resonance constraints rather than a single formalism. Their common objective is to turn spectral information into resonance parameters, weighted sum rules, or consistency relations while keeping the resonance sector explicit. Their technical realization, however, depends strongly on the problem: inverse OPE reconstruction, kernel-engineered FESR, chiral spectral saturation, dispersive cross-section integrals, or strength-function moments in correlated many-body theory.