Skip to content

Forward Scattering Sum Rules

In a gapped forward amplitude, crossing combines the two cuts and the optical theorem turns the right-cut discontinuity into inclusive production data. Even low-energy Taylor coefficients then become inverse-energy moments of a physical absorptive part; crossing-odd combinations instead give generally signed sum rules unless extra structure supplies a sign.

Required background. Subtracted dispersion relations supplies the contour formula and subtraction count. The optical theorem and cut interpretation supplies the inclusive state sum and its normalization.

The crossing-even forward moment expansion

Section titled “The crossing-even forward moment expansion”

Consider identical massive scalars and first remove all stable one-particle poles below the two-particle threshold. At t=0t=0 define

ν=s2m2,B(ν)=B(ν),ν0=2m2,\nu=s-2m^2, \qquad B(\nu)=B(-\nu), \qquad \nu_0=2m^2,

so the cuts begin at ν=±ν0\nu=\pm\nu_0. Assume real analyticity, no other first-sheet singularities, and B(ν)/ν20B(\nu)/\nu^2\to0 on the large complex arc. The twice-subtracted relation about ν=0\nu=0 is

B(ν)=B(0)+2ν2πν0ImA(2m2+ν+i0,0)ν(ν2ν2)dν.B(\nu)=B(0)+\frac{2\nu^2}{\pi} \int_{\nu_0}^{\infty} \frac{\operatorname{Im}A(2m^2+\nu'+i0,0)} {\nu'(\nu'^2-\nu^2)}\,\mathrm d\nu'.

The factor of two is the combined right- and left-cut contribution. Its sign can be checked by expanding both kernels 1/(νν)1/(\nu'-\nu) and 1/(ν+ν)1/(\nu'+\nu) before combining them. This forward crossing construction and its subtraction dependence are developed in Weinberg 1995, § 10.8, pp. 465–469.

For ν<ν0|\nu|<\nu_0, expand the denominator geometrically:

B(ν)=B(0)+n=1c2nν2n,c2n=2πν0ImA(2m2+ν+i0,0)ν2n+1dν.\begin{aligned} B(\nu)&=B(0)+\sum_{n=1}^{\infty}c_{2n}\nu^{2n},\\ c_{2n}&=\frac{2}{\pi} \int_{\nu_0}^{\infty} \frac{\operatorname{Im}A(2m^2+\nu'+i0,0)} {\nu'^{2n+1}}\,\mathrm d\nu'. \end{aligned}

Thus c2n=B(2n)(0)/(2n)!c_{2n}=B^{(2n)}(0)/(2n)! is an inverse-energy moment. The first undetermined subtraction constant is B(0)B(0); crossing has removed the linear one. More subtractions would leave more Taylor data outside the integral.

With the common relativistic normalization in four dimensions, the forward optical theorem reads

ImA(s+i0,0)=s(s4m2)σtot(s)0,s4m2.\operatorname{Im}A(s+i0,0) =\sqrt{s(s-4m^2)}\,\sigma_{\mathrm{tot}}(s)\ge0, \qquad s\ge4m^2.

Other amplitude normalizations change the positive kinematic factor, not the sign. The moment therefore sums every on-shell final state accessible from the chosen initial state. It is not just the elastic cross section and is not a tree-level identity. Adams and collaborators use exactly this inclusive input to turn a forward contour integral into a positive coefficient Adams et al. 2006, § 4, pp. 14–19, PDF.

Convergence is checked twice:

  • Near threshold, phase space and any threshold singularity must make the weighted integral locally integrable.
  • At infinity, the power ν(2n+1)\nu'^{-(2n+1)} must beat the absorptive growth. The same growth estimate must also justify discarding the original large arc.

A convergent moment does not retroactively justify the contour step: an amplitude could have acceptable real-axis data but uncontrolled growth in other complex directions.

The contour figure can now be read as a sum rule. The right and left lips become the same inclusive spectrum only after crossing; the small pole circles must be removed before a Taylor coefficient is identified.

In the forward crossing-even amplitude, the two oriented cuts combine into one inclusive absorptive integral after explicit poles are removed; subtraction constants remain at the interior point.

Schematic forward contour, not to scale. Crossing maps the left cut to a physical crossed-channel cut, the optical theorem supplies its inclusive absorptive data, and two subtractions leave B(0)B(0) while determining the even coefficients c2nc_{2n} for n1n\ge1. The dashed arc still requires the stated high-energy bound.

Contour contributionForward sum-rule meaning
right cutinclusive ss-channel production
left cutcrossed-channel production, with crossing matrix or parity retained
pole circlesknown stable exchange terms, removed explicitly
subtraction pointlow-energy constants not fixed at the chosen subtraction order

For particles with charge, flavor, or spin, crossing acts on a vector of amplitudes. In a crossing eigenchannel write

Fη(ν)=ηFη(ν),η=±1.F_\eta(-\nu)=\eta F_\eta(\nu), \qquad \eta=\pm1.

F+F_+ has the even moment expansion above. For FF_-, the ratio F(ν)/νF_-(\nu)/\nu is even, but its right-cut absorptive part is a crossing-weighted combination of physical processes. In ordinary particle–antiparticle examples the resulting integrand contains a difference of inclusive cross sections. Unitarity makes each cross section nonnegative, not their difference. Consequently an odd sum rule can be predictive without being a positivity bound.

If the odd amplitude decreases unusually fast, an otherwise free odd subtraction constant may obey a superconvergent sum rule. That statement requires the stronger falloff explicitly; it must not be inferred from crossing alone.

Write ρ(ν)=ImA(2m2+ν+i0,0)\rho(\nu')=\operatorname{Im}A(2m^2+\nu'+i0,0). Because ρ0\rho\ge0,

c2n>0c_{2n}>0

whenever the chosen channel has nonzero absorptive weight. Cauchy–Schwarz supplies a stronger internal check,

c2n+22c2nc2n+4,c_{2n+2}^{2}\le c_{2n}c_{2n+4},

and support on νν0\nu'\ge\nu_0 gives

0<c2n+2c2nν02.0<c_{2n+2}\le\frac{c_{2n}}{\nu_0^2}.

Indeed, with

dμn(ν)=2πρ(ν)ν2n+1dν,\mathrm d\mu_n(\nu') =\frac{2}{\pi} \frac{\rho(\nu')}{\nu'^{2n+1}}\,\mathrm d\nu',

the three entries are 1dμn\int1\,\mathrm d\mu_n, ν2dμn\int\nu'^{-2}\mathrm d\mu_n, and ν4dμn\int\nu'^{-4}\mathrm d\mu_n. Cauchy–Schwarz gives the first inequality, while ν2ν02\nu'^{-2}\le\nu_0^{-2} gives the second. They are consequences of the positive measure and its support, not additional assumptions. The systematic moment formulation and its finite Hankel constraints are given in Bellazzini et al. 2021, § I.A–B, pp. 2–5, PDF.

Put all spectral weight at one narrow scale M2=2m2+νMM^2=2m^2+\nu_M. Show that c2nνM(2n+1)c_{2n}\propto\nu_M^{-(2n+1)} and that the Cauchy–Schwarz inequality is saturated.

Check

For ρ(ν)=wδ(ννM)\rho(\nu')=w\,\delta(\nu'-\nu_M) with w>0w>0,

c2n=2wπνM2n+1,c_{2n}=\frac{2w}{\pi\nu_M^{2n+1}},

so c2n+22=c2nc2n+4c_{2n+2}^2=c_{2n}c_{2n+4}. A mixture of distinct scales generally moves the sequence into the interior of the moment region.

Continue to the sign theorem: Forward-Limit Positivity Bounds. For nonzero momentum transfer: Fixed-t and Partial-Wave Dispersion.

  • Adams, Allan, Nima Arkani-Hamed, Sergei Dubovsky, Alberto Nicolis, and Riccardo Rattazzi. “Causality, Analyticity and an IR Obstruction to UV Completion.” Journal of High Energy Physics 10 (2006): 014. DOI. Open PDF.
  • Bellazzini, Brando, Joan Elias Miró, Riccardo Rattazzi, Marc Riembau, and Francesco Riva. “Positive Moments for Scattering Amplitudes.” Physical Review D 104 (2021): 036006. DOI. Open PDF.
  • Weinberg, Steven. The Quantum Theory of Fields. Volume I: Foundations. Cambridge: Cambridge University Press, 1995. DOI.