LSZ Reduction and Amputated Distributions

LSZ reduction is licensed only after stable one-particle states and wave operators exist. For fields with nonzero overlap with isolated mass shells, it expresses connected scattering matrix elements as on-shell boundary limits of connected, amputated time-ordered distributions. Here we state the massive scalar formula, track its packet phases and residue normalization, and explain the distributional proof mechanism. Smearing, time ordering, domains, and the order of limits are essential; the full regularity theorem requires the cited scattering framework.

Required background. Wightman functions and spectral support supplies the distributions; Haag–Ruelle construction supplies scattering states; and wave operators and asymptotic fields supplies the $S$-operator.

Helpful background. LSZ reduction: poles, residues, and stable external states gives the practical bridge, while cross sections and decay rates explains the later observable normalization.

Hypotheses before amputation

Consider neutral Hermitian scalar interpolating fields in four-dimensional Minkowski space. For each external species require:

• an isolated stable positive-energy mass shell and a Haag–Ruelle one-particle subspace;
• an interpolating local field $\phi_i$ with nonzero overlap $\langle\Omega|\phi_i(0)|p,i\rangle=Z_i^{1/2}$ in the chosen normalization;
• time-ordered vacuum distributions with the regularity needed for multiplication by wave packets and application of Klein–Gordon operators;
• incoming and outgoing packets with compact, separated velocity supports, followed through the ordered large-time limits;
• connected/truncated parts when the desired matrix element excludes disconnected propagation.

No asymptotic-completeness hypothesis is needed for matrix elements between already constructed scattering states. It is needed only to claim that all physical states are so described.

The original reduction framework is Lehmann, Symanzik, and Zimmermann 1955, pp. 205–225; the bridge from Wightman/Haag–Ruelle assumptions to LSZ is proved in Hepp 1965, pp. 95–111.

The distributional reduction formula

For a scalar connected time-ordered distribution

$$G_c(x_1,\ldots,x_{r+s}) =\langle\Omega,T\{\phi(x_1)\cdots\phi(x_{r+s})\}\Omega\rangle_c,$$

choose relativistically normalized one-particle states and write the incoming and outgoing packets as

$$\begin{aligned} |f\rangle&=\int d\Pi_p\,f(\mathbf p)|p\rangle, &d\Pi_p&=\frac{d^3\mathbf p}{(2\pi)^3\,2E_{\mathbf p}},\\ h_{\mathrm{in}}(x)&=\int d\Pi_p\,f(\mathbf p)e^{-ip\cdot x}, &h_{\mathrm{out}}(x)&=\int d\Pi_p\,g(\mathbf p)^*e^{+ip\cdot x}. \end{aligned}$$

The smooth compact momentum amplitudes have separated velocity supports within each asymptotic state. In the following formula, $h_\ell$ is the incoming or outgoing function appropriate to that leg. With one-particle phases chosen so that the scalar overlap is $+\sqrt{Z_\ell}$, reduction gives

$$\begin{aligned} &\langle g_1,\ldots,g_s;\mathrm{out}\,|\, f_1,\ldots,f_r;\mathrm{in}\rangle_c\\ &\quad= \int\prod_{\ell=1}^{r+s}\mathrm d^4x_\ell\, \left[ \prod_{\ell=1}^{r+s} \frac{i}{\sqrt{Z_\ell}}h_\ell(x_\ell) (\Box_{x_\ell}+m_\ell^2) \right] G_c(x_1,\ldots,x_{r+s}), \end{aligned}$$

The integrals stand for the ordered, smeared limits justified by the asymptotic theorem, rather than an absolutely convergent integral over arbitrary distributions. The factor $i$ on every leg is part of the formula: Fourier transformation sends $\Box+m^2$ to $-(p^2-m^2)$, so $i(\Box+m^2)/\sqrt Z$ becomes $(p^2-m^2)/(i\sqrt Z)$. The outgoing packet is conjugated and has the opposite Fourier phase to the incoming packet. The field-overlap normalization and reduction formula are derived in Srednicki 2006 draft, § 5, pp. 49–56, PDF.

In momentum notation this becomes

$$\begin{aligned} &\left. \prod_{\ell}\frac{p_\ell^2-m_\ell^2}{i\sqrt{Z_\ell}}\, \widetilde G_c \right|_{\text{ordered on-shell boundary values}}\\ &\qquad = i(2\pi)^4\delta^{(4)}(P_f-P_i)\mathcal M_{fi}. \end{aligned}$$

Here $\widetilde G_c$ uses the incoming and outgoing Fourier phases just specified; the remaining packet integrations give the displayed state matrix element. The symbol “on shell” is not a pointwise substitution into an arbitrary distribution. Amputation cancels the isolated simple poles after packet smearing, and the boundary limit is taken in the topology established by the scattering theorem.

Operationally, the reduction is performed leg by leg. One first smears a field with a Klein–Gordon packet and sends its time support to the appropriate incoming or outgoing end. Integration by parts converts the limiting difference into an insertion of $(\Box+m^2)\phi$ in the time-ordered distribution; contact terms are then organized distributionally. Iteration gives the displayed formula. The packet limit is taken before removing the smearing notation. This proof mechanism explains why multiplying a formal Fourier transform by inverse propagators and substituting numerical on-shell momenta is only shorthand for the properly smeared boundary value.

Changing the interpolating field while preserving its nonzero one-particle projection changes off-shell correlators and residues, but not the reduced matrix element. This invariance is another check that LSZ extracts scattering data rather than a preferred field coordinate.

A massive scalar 2→2 reduction

Choose four wave packets supported near four points of $H_m^+$, with the in and out velocity supports separated as required. Suppose

$$\begin{aligned} \widetilde G_{4,c} ={}&\left[\prod_{\ell=1}^4 \frac{i\sqrt Z}{p_\ell^2-m^2+i0}\right]\\ &\times i(2\pi)^4\delta^{(4)}(P_f-P_i)\mathcal M_{fi}\\ &+\text{terms less singular on the four shells}. \end{aligned}$$

Each external pole contains one vacuum-to-particle overlap $\sqrt Z$. The two-point function has two such overlaps and therefore has residue $Z$; using $iZ$ on each leg of the displayed formula would incorrectly replace four overlaps by eight. Multiplying by $\prod_\ell(p_\ell^2-m^2)/(i\sqrt Z)$ extracts $i(2\pi)^4\delta^{(4)}(P_f-P_i)\mathcal M_{fi}$, and the packet integrals give the matrix element between the Haag–Ruelle states. The practical derivation and diagrammatic evaluation are on LSZ reduction: poles, residues, and stable external states.

Equivalently, one may first divide by full propagators $iZ/(p^2-m^2+i0)$. The resulting amputated four-point kernel is $Z^{-2}$ times the physical transition distribution, so this convention needs a compensating factor $Z^2$. These two amputation conventions give the same answer; their residue factors must not be mixed.

For an independent field-rescaling check, keep the phases of the physical states fixed and denote the signed real overlap by $\kappa=\langle\Omega|\phi(0)|p\rangle$, with $Z=\kappa^2$. Under $\phi\mapsto c\phi$, $c\in\mathbb R\setminus\{0\}$,

$$\kappa\mapsto c\kappa,\qquad G_{N,c}\mapsto c^N G_{N,c},\qquad \prod_{\ell=1}^N\frac1{\kappa_\ell} \mapsto c^{-N}\prod_{\ell=1}^N\frac1{\kappa_\ell}.$$

The reduced matrix element is invariant, including for negative $c$ and odd $N$. Writing only $\sqrt Z\mapsto |c|\sqrt Z$ while leaving the state phases fixed would miss a sign in that odd-leg test. For the four-leg example, the same check is $c^4(c^{-1})^4=1$; a pole product with four residues $Z$ would instead scale as $c^8$ before reduction.

Failure test: unstable and infraparticle legs

An unstable resonance has no isolated real mass-shell projection. A charged QED electron in an infraparticle sector has continuous spectral weight beginning at the mass threshold and no nonzero delta-function residue. In either case the external simple pole assumed above is absent. Assigning a finite $Z$ and applying the formula is not an approximation justified by LSZ; the external-leg limit fails before an amplitude is named.

For QED one instead uses suitably inclusive observables, coherent/dressed asymptotic structures, or detector functionals, each with its own regulator and convergence statement. The existence of an infrared-finite perturbative expression does not retroactively create a Wigner one-electron pole.

Exercises

Explain why disconnected two-point contractions must be removed when extracting the connected $2\to2$ amplitude.

Solution

Disconnected pairings describe independent one-particle propagation and contain products of momentum-conserving delta distributions. After amputation they reproduce identity/no-scattering contributions rather than the connected transition. Passing to $G_{4,c}$ subtracts these pairings and isolates the matrix element of $S-1$.

References

• Hepp, Klaus. 1965. “On the Connection between the LSZ and Wightman Quantum Field Theory.” Communications in Mathematical Physics 1: 95–111. DOI.
• Lehmann, Harry, Kurt Symanzik, and Wolfhart Zimmermann. 1955. “On the Formulation of Quantized Field Theories.” Il Nuovo Cimento 1: 205–225. DOI.
• Srednicki, Mark. Quantum Field Theory. Prepublication draft, 2006. Author’s draft and published-edition errata. Open PDF.