---
title: 'On-Shell Approximation: QFT and Scattering'
url: https://www.emergentmind.com/topics/on-shell-approximation
type: topic
---

# On-Shell Approximation: QFT and Scattering

Searching arXiv for recent and foundational papers on on-shell approximation across the contexts represented in the source material.
On-shell approximation denotes a family of procedures in which dynamical quantities are constrained, evaluated, or matched at physical mass-shell kinematics rather than through fully off-shell Green functions or kernels. In the literature represented here, the term can refer to an on-shell renormalization scheme fixed by pole masses, unit residues, and physical charges; to replacing a momentum-dependent scattering kernel by its on-shell value; to driving an interaction to a momentum-diagonal “on-shell limit”; or to constructing amplitudes directly from on-shell states and factorization [1905.02222] [2303.02675] [1412.2077] [1912.04334]. In some contexts, notably modern amplitude theory, “on-shell” is not an approximation that discards off-shell information but an S-matrix formulation in which only physical states and factorization data are used [1912.04334].

## 1. Core meanings and terminological boundaries

The expression is context dependent. What remains common is the use of physical kinematics—typically \(p^2=m^2\) in relativistic field theory or \(E=\hbar^2 k^2/(2m_r)\) in nonrelativistic scattering—as the organizing principle.

| Context | Operational meaning | Representative papers |
|---|---|---|
| Renormalized QFT | Fix masses at propagator poles, residues to 1, and couplings from physical amplitudes | [1905.02222] |
| Nonrelativistic scattering | Replace off-shell kernels in the \(T\)-matrix equation by their on-shell values | [2303.02675], [2507.20421] |
| Nuclear many-body theory | Drive \(V_\lambda(p,p')\) to a momentum-diagonal limit \(V_{\lambda=0}(p)\delta_{pp'}\) | [1412.2077] |
| Unitarized EFT | Replace \(V(q,Q,P)\) by an on-shell kernel \(V_{\text{on}}(s)\) in the Bethe–Salpeter equation | [1310.5224] |
| Amplitude methods | Build observables from on-shell three- and four-point amplitudes and factorization | [1912.04334], [2411.12798] |
| Loop unitarity methods | Reconstruct nonanalytic loop terms from on-shell intermediate states across physical cuts | [1611.03074], [1609.00714], [2305.01426] |
| Femtoscopy | Replace the exact \(q\)-dependent pair momentum \(P^\mu\) by an on-shell pseudo-momentum \(p^\mu=P^\mu|_{q=0}\) | [2602.02810] |

A recurrent misconception is that “on-shell” always means “drop off-shell effects.” The electroweak on-shell program instead takes the S-matrix as primary and derives tree amplitudes from Poincaré invariance, little-group covariance, factorization, and UV behavior, with no Lagrangian and no Higgs vacuum expectation value [1912.04334]. A separate mathematical usage concerns approximation of \(\delta\)-shell interactions for the Dirac operator, where “shell” refers to support on a surface \(\Sigma\), not to the relativistic mass shell [2309.12911].

## 2. Renormalization-theoretic meaning

In perturbative quantum field theory, an on-shell renormalization scheme fixes renormalization constants by imposing conditions on physical quantities. For an ordinary fermion propagator, the renormalized mass is defined by the pole condition
\[
S^{-1}(p)\equiv \slashed{p}-m_0-\Sigma(\slashed{p})
\quad\Rightarrow\quad
S^{-1}(p)\big|_{\slashed{p}=m}=0,
\]
and the field renormalization is fixed so that the residue at the pole is 1 [1905.02222]. Charge renormalization is defined from a physical amplitude, equivalently from the zero-momentum photon propagator in the Thomson limit.

In \({\cal N}=1\) SQED with higher-derivative regularization plus Pauli–Villars fields, the quadratic matter effective action is parametrized by functions \(G\) and \(J\), and the matter propagator poles occur when
\[
p^2 G^2(p)-m_0^2 J^2(p)=0.
\]
The renormalized mass is therefore
\[
m = m_0\,\frac{J(p)}{G(p)}\bigg|_{p^2=m^2},
\qquad
Z_m\equiv \frac{m_0}{m}
= \frac{G(p)}{J(p)}\bigg|_{p^2=m^2},
\]
while unit residue gives
\[
Z^{-1} = G(p)\,\Big(1+2m^2\frac{\partial}{\partial p^2}\ln\frac{G(p)}{J(p)}\Big)\bigg|_{p^2=m^2}.
\]
The gauge-sector invariant charge \(d\) defines the on-shell coupling by
\[
d(0,m_0/\Lambda,\alpha_0)=\alpha,
\qquad
\beta(\alpha)=\frac{d\alpha}{d\ln m}\bigg|_{\alpha_0,\Lambda=\mathrm{const}.}
\]
In this scheme the exact NSVZ-type relation takes the form
\[
\beta(\alpha)=\frac{N_f\alpha^2}{\pi}\Big(1+\gamma_m(\alpha)\Big),
\qquad
\gamma_m(\alpha)\equiv \frac{d\ln Z_m}{d\ln m}\bigg|_{\alpha_0,\Lambda=\mathrm{const}.}
\]
The paper shows that this relation is valid to all orders in perturbation theory and explicitly finds
\[
\gamma_m(\alpha)=\frac{\alpha}{\pi}-\frac{\alpha^2(3N_f+1)}{2\pi^2}+O(\alpha^3),
\]
hence
\[
\beta(\alpha)=\frac{N_f\alpha^2}{\pi}
+\frac{N_f\alpha^3}{\pi^2}
-\frac{N_f(3N_f+1)\alpha^4}{2\pi^3}
+O(\alpha^5)
\]
in the on-shell scheme [1905.02222].

This usage is sharply distinct from mass-independent schemes such as \(\overline{\text{MS}}\) or \(\overline{\text{DR}}\), where renormalized parameters are subtraction-scale dependent and not directly equal to physical pole masses or Thomson-limit charges. The on-shell scheme is therefore a physically normalized renormalization prescription, not merely a kinematic simplification.

## 3. Scattering-theory and many-body reductions

In nonrelativistic scattering theory, the on-shell approximation usually means replacing the full off-shell dependence of the kernel in the Lippmann–Schwinger equation by its value at the physical external momentum. For identical particles in \(D\) dimensions, the exact s-wave equation is
\[
T_0(k)=V_0(k)+S_D\int_0^\infty \frac{dk''}{(2\pi)^D}
\frac{(k'')^{D-1}}{\frac{\hbar^2 k^2}{m}-\frac{\hbar^2(k'')^2}{m}+i\epsilon}
\,V_0(k,k'')\,T_0(k'',k).
\]
The on-shell approximation sets
\[
V_0(k,k'')\simeq V_0(k,k)\equiv V_0(k),
\qquad
T_0(k'',k)\simeq T_0(k,k)\equiv T_0(k),
\]
which yields the algebraic form
\[
T_0(k)=\frac{1}{\displaystyle \frac{1}{V_0(k)}-C(k)}.
\]
In \(D=3\), this implies
\[
V_0(k)= -\,\frac{4\pi\hbar^2}{m}\,\frac{\tan\delta_0(k)}{k},
\]
and low-momentum expansion gives
\[
g_0=\frac{4\pi\hbar^2}{m}a_s,
\qquad
g_2=\frac{2\pi\hbar^2}{m}a_s^2 r_s
\]
for \(V_0(k)=g_0+g_2 k^2+\dots\) [2303.02675]. A later analytic comparison with square-well and delta-shell potentials found that the accuracy of this approximation improves with increasing momentum and for weaker potentials, and that in the weak-interaction limit it becomes exact at leading order [2507.20421].

In unitarized chiral perturbation theory, the same phrase denotes replacing the full Bethe–Salpeter kernel \(V(q,Q,P)\) by its on-shell value \(V_{\text{on}}(s)\), so that
\[
T(s)=[1-V(s)G(s)]^{-1}V(s)
\]
becomes algebraic rather than integral [1310.5224]. A full off-shell treatment of Nambu–Goldstone boson–\(D\)-meson scattering described the cited lattice-QCD data better than the widely used on-shell approximation, with \(\chi^2/\mathrm{d.o.f.}\approx 0.79\) versus \(\chi^2/\mathrm{d.o.f.}\approx 1.23\) at NLO, but no qualitative difference was found in the light-quark-mass evolution of the scattering lengths, and the \(D_{s0}^*(2317)\) remained qualitatively similar in both schemes [1310.5224].

In nuclear many-body theory, the on-shell limit may instead mean diagonalization in momentum space. Under SRG evolution with the Wilson generator,
\[
V_\lambda(p,p')\xrightarrow[\lambda\to 0]{}V_{\lambda=0}(p)\,\delta_{pp'},
\]
so all off-diagonal momentum couplings vanish [1412.2077]. In Hartree–Fock neutron matter, the energy per particle then depends only on diagonal matrix elements \(V_\lambda(k,k)\). For S-wave-only calculations at \(\lambda=1\ \mathrm{fm}^{-1}\), both a separable toy potential and realistic high-precision \(NN\) interactions give a minimum Bertsch parameter in the range \(0.42\text{–}0.45\) at \(k_F\approx 1.1\text{–}1.3\ \mathrm{fm}^{-1}\), while evolution toward \(\lambda\to 0\) reveals strong \(\lambda\)-dependence because induced many-body forces are omitted in the truncated two-body Hartree–Fock treatment [1412.2077].

## 4. On-shell amplitudes, factorization, and loop reconstruction

Modern amplitude theory uses “on-shell” in a stronger sense: amplitudes are built directly from physical external states, little-group covariance, and factorization, rather than from off-shell fields. In the bosonic electroweak sector, the primary object is
\[
\mathcal{M}(p_1,\rho_1,\dots,p_n,\rho_n)
= \delta^4\Big(\sum_a p_a\Big)\,M(p_1,\rho_1,\dots,p_n,\rho_n),
\]
with massless momenta written as
\[
p_{\alpha\dot\alpha}=\lambda_\alpha\tilde\lambda_{\dot\alpha},
\]
and massive ones as
\[
p_{\alpha\dot\alpha}=\epsilon_{JI}\lambda_\alpha^I\tilde\lambda_{\dot\alpha}^J.
\]
Three-point amplitudes are fixed by Lorentz invariance and little-group scaling, and four-point amplitudes are obtained by factorization on all poles plus contact terms fixed by good UV behavior [1912.04334]. In this framework the electroweak relations
\[
e=g\sin\theta_W,\qquad g'=g\tan\theta_W,\qquad \frac{m_W}{m_Z}=\cos\theta_W
\]
and Higgs couplings such as \(e_{WWH}=g\) and \(e_{ZZH}=g/\cos\theta_W\) emerge from consistency of on-shell amplitudes, without introducing a Higgs vacuum expectation value [1912.04334].

A separate on-shell program reconstructs loop amplitudes from physical cuts. The unitarity relation
\[
{\rm Disc}\, T_{fi} = -\sum_n T_{fn}T^\dagger_{ni}
\]
is applied to the \(t\)-channel cut with two on-shell massless intermediate particles, and the nonanalytic part of the loop amplitude is obtained from products of tree-level Compton amplitudes integrated over the physical phase space [1611.03074] [1609.00714]. This yields the long-range terms responsible for \(1/r^2\), \(1/r^3\), and higher tails in electromagnetic and gravitational potentials, while analytic terms are deliberately ignored because they correspond to local contact interactions [1609.00714]. The same strategy extends to mixed electromagnetic–gravitational scattering, where the mixed \(\gamma g\) and \(\gamma\gamma\) cuts reproduce the long-range \(\mathcal{O}(G\alpha)\) potential without summing the full set of mixed Feynman diagrams [2305.01426].

On-shell methods also clarify discontinuous massless limits. In massive supergravity, the massless limit of massive spin-\(3/2\) exchange contains a residual spin-\(1/2\) contribution; on-shell decomposition shows directly that the extra helicity modes do not decouple in the \(m\to 0\) limit, giving the Deser–Kay–Stelle discontinuity as the supersymmetric analogue of the vDVZ effect [2005.14077].

## 5. On-shell matching in effective field theory

Effective-field-theory matching is traditionally performed off shell, at the level of Green functions. That approach requires a Green’s basis containing redundant and evanescent operators, and reduction to a physical basis is often non-trivial, difficult to automate, and error prone [2411.12798]. The on-shell alternative matches directly on physical amplitudes in a physical basis.

At tree level the matching condition is
\[
\mathcal{M}^{(0)}_{\mathrm{EFT}}=\mathcal{M}^{(0)}_{\mathrm{full}}.
\]
At one loop the proposal is
\[
\mathcal{M}_{\mathrm{EFT}}^{(0)}
=
\mathcal{M}_{\mathrm{full}}^{(1),\mathrm{hard}}
+\mathcal{M}_{\mathrm{full}}^{(1),\mathrm{soft}|_{\mathrm{UV}}}
-\mathcal{M}_{\mathrm{EFT}}^{(1),\mathrm{soft}|_{\mathrm{UV}}},
\]
so the hard-region contribution gives the usual local matching, while the soft UV pieces account for the delicate cancellation of non-local terms and implicitly retain evanescent effects [2411.12798]. The numerical implementation uses rational on-shell kinematics, ensuring an exact analytic solution despite the numerical procedure. In this way one needs only a physical basis. The method can reduce a Green’s basis to an arbitrary physical one, translate between physical bases, renormalize effective Lagrangians directly in terms of a physical basis, and perform finite matching including evanescent contributions [2411.12798].

This usage is important conceptually because it shows that on-shell methods are not confined to amplitude bootstrap or unitarity cuts. They also provide a basis-management strategy for EFTs in which field-redefinition redundancies are never introduced explicitly.

## 6. Validity, corrections, and limitations

The reliability of an on-shell approximation depends on what has been put on shell and what has been neglected. In femtoscopy, the approximation is kinematic: with
\[
p_1^\mu=(1+\alpha)P^\mu+q^\mu,\qquad
p_2^\mu=(1-\alpha)P^\mu-q^\mu,
\]
one defines the pseudo average momentum
\[
p\equiv P\big|_{q=0},
\qquad
p^2=\left(\frac{m_1+m_2}{2}\right)^2,
\]
and then approximates \(P^\mu\approx p^\mu\) and \(\alpha\approx\alpha_0\) for all \(q\) [2602.02810]. The exact equal-time correlation function
\[
C(q;P^\mu)=
\sum_s w_s
\frac{\displaystyle \int d^3r \int \frac{d^3k}{(2\pi)^3} D_{q,s}(r,k)\,S(r,k;P^\mu)}
{\displaystyle \int d^3r\,S(r,q;P^\mu)}
\]
thereby becomes \(C(q;p^\mu)\). The first on-shell corrections appear at order \(q^2\),
\[
C(q)=\bigl(1-q^2\varepsilon'\bigr)\,C_{\rm on-shell}(q)
+q^2\int d^3r\,K(r,q)S'(r)
+O\Bigl(\frac{q^3}{m^2}\Bigr),
\]
and for angle-averaged correlations the first-order smoothness contributions vanish by symmetry [2602.02810]. In the blast-wave examples quoted there, the corrections are at or below the percent level for pp correlations and deuteron coalescence, while for coalescence the on-shell approximation is essentially exact because \(q=0\) by construction [2602.02810].

In scattering theory, the limitations are different. The s-wave on-shell approximation is exact in the \(k\to 0\) limit of the algebraic \(T\)-matrix derived in the cited work, but its quantitative accuracy depends on low momentum, short-range interactions, and weak off-shell dependence [2303.02675]. The direct comparison with exact square-well and delta-shell solutions shows that stronger potentials worsen the approximation, whereas weaker interactions improve it and make the approximation exact at leading order [2507.20421]. In UChPT, by contrast, keeping full off-shell terms can improve lattice-data fits, which shows that the standard on-shell factorization of the Bethe–Salpeter kernel is not innocuous when precision at several quark masses is required [1310.5224].

A broad conclusion suggested by these examples is that “on-shell approximation” is most stable when the neglected structure is either kinematically suppressed, perturbatively weak, or removable by scheme choice. It becomes more delicate when off-shell momentum dependence feeds directly into nonperturbative resummation, induced many-body forces, or precision extrapolations across scales [1412.2077] [1310.5224] [2602.02810].

Source: https://www.emergentmind.com/topics/on-shell-approximation