Skip to content

Writing S=1+iTS=1+iT, the operator identity SS=1S^\dagger S=1 is equivalent to TT=iTTT-T^\dagger=iT^\dagger T. Between scattering states, a complete-state insertion turns its right side into a sum over all kinematically allowed on-shell intermediate states. Thus the absorptive part at one perturbative order is fixed by products of lower-order amplitudes, with the same state and phase-space normalization used in the cross section.

Required background. S-Matrix and T-Matrix Normalization supplies the delta-function and state conventions. Hilbert Positivity and Unitary Evolution supplies the operator-level meaning and limitations of unitarity.

Expand

(1iT)(1+iT)=1.(1-iT^\dagger)(1+iT)=1.

It follows that

TT=iTT.T-T^\dagger=iT^\dagger T.

Use relativistically normalized states and define

fiTi=i(2π)4δ(4)(PfPi)Mfi.\langle f|iT|i\rangle =i(2\pi)^4\delta^{(4)}(P_f-P_i)\mathcal M_{fi}.

The first subscript labels the final state and the second the initial state.

Inserting a complete set of asymptotic states XX yields, after the common overall delta function is removed,

i(MfiMif)=X1SXdΦX(Pi)×MXfMXi.\begin{aligned} -i\left(\mathcal M_{fi}-\mathcal M_{if}^*\right) &=\sum_X\frac{1}{S_X}\int \mathrm d\Phi_X(P_i)\\ &\quad{}\times\mathcal M_{Xf}^*\mathcal M_{Xi}. \end{aligned}

Here the sum includes particle species, spins, colors, and every multiplicity. The measure dΦX\mathrm d\Phi_X is the volume-wide labeled invariant measure, while SX=rnr!S_X=\prod_r n_r! divides out permutations of each identical species in the intermediate state. For f=if=i,

2ImMii=X1SXdΦX(Pi)MiX20.2\operatorname{Im}\mathcal M_{ii} =\sum_X\frac{1}{S_X}\int \mathrm d\Phi_X(P_i) \left|\mathcal M_{i\to X}\right|^2\ge0.

The inequality is a forward, diagonal statement. A generic off-diagonal imaginary part need not be positive. Schwartz derives this normalization in Schwartz 2014, § 24.1, printed pp. 452–465, while Weinberg gives the complete-state form and its scattering consequences in Weinberg 1995, § 3.6, printed pp. 147–151.

The statement can also be projected onto a selected set of channels. If PP projects onto retained asymptotic states and Q=1PQ=1-P, then

P(TT)P=iPTPTP+iPTQTP.P(T-T^\dagger)P =iPT^\dagger PT P+iPT^\dagger QT P.

The second term is positive on a forward diagonal matrix element. Omitting open QQ channels while imposing equality with only the first term is therefore not a unitary truncation; it undercounts absorption. Below the first omitted threshold the QQ phase space can vanish, which explains when a reduced channel description can be exact.

Connected pieces and spectator delta functions

Section titled “Connected pieces and spectator delta functions”

The full SS-matrix contains disconnected terms. On both sides of the unitarity relation, identical spectator delta functions must be matched before one isolates the connected amplitude relation. Treating every disconnected product as a new connected channel double counts processes and leaves powers of the spacetime volume.

For the forward connected 222\to2 relation, the identity component has already been separated by S=1+iTS=1+iT. Intermediate states are nevertheless fully inclusive: two-particle, multiparticle, and any other allowed asymptotic channels all contribute. A truncated channel sum gives a valid equality only below the first omitted threshold or within a declared approximation.

The figure previews the forward specialization. Its dashed cut denotes the complete on-shell state sum, not a literal operation on an arbitrary diagram.

The forward amplitude minus its conjugate equals a sum over complete on-shell intermediate states, with the same invariant phase-space measure used for inclusive rates.

Unitarity matches the forward discontinuity to a convention-normalized sum over physical intermediate states. The diagram is schematic: Cutkosky rules provide a later diagram-by-diagram implementation, while the operator identity already requires a complete asymptotic-state basis.

Expand T=n1gnT(n)T=\sum_{n\ge1}g^nT^{(n)}. At order gng^n,

T(n)T(n)=ir=1n1T(r)T(nr).T^{(n)}-T^{(n)\dagger} =i\sum_{r=1}^{n-1}T^{(r)\dagger}T^{(n-r)}.

If a four-point tree amplitude begins at g2g^2, its one-loop imaginary part at g4g^4 is fixed by the phase-space integral of two tree amplitudes. A real tree amplitude below all propagator poles therefore does not violate unitarity; its absorptive part first appears when the perturbative right side has the corresponding order and an open intermediate channel.

For a concrete normalization check, take identical real scalars with Lint=λϕ4/4!\mathcal L_{\mathrm{int}}=-\lambda\phi^4/4!. Above s=4m2s=4m^2, the two-scalar intermediate state contributes 1/2!1/2! and Φ2=ρ/(8π)\Phi_2=\rho/(8\pi), so at order λ2\lambda^2 the forward relation gives

2ImM1-loop(s,0)=12!ρ(s)8πλ2,ImM1-loop=λ2ρ(s)32π.2\operatorname{Im}\mathcal M_{\text{1-loop}}(s,0) =\frac{1}{2!}\frac{\rho(s)}{8\pi}\lambda^2, \qquad \operatorname{Im}\mathcal M_{\text{1-loop}} =\frac{\lambda^2\rho(s)}{32\pi}.

This is the absorptive part of the physical ss-channel bubble in the present normalization. Below threshold ρ\rho is not a physical phase-space factor and the state-sum contribution is zero; analytic continuation of the loop is a separate statement.

This order-by-order identity does not license selective resummation. Adding a width to one propagator while leaving other terms at tree order can spoil the same cancellations used to prove gauge identities. Generalized cuts and positivity bounds require additional structure: continue to generalized unitarity for cut-based reconstruction and to forward-limit positivity bounds only after the analytic and growth hypotheses are in place. The operator-level meaning of unitary time evolution remains with the foundational page linked above.

Assume T=gT(1)+g2T(2)+T=gT^{(1)}+g^2T^{(2)}+\cdots. Derive the anti-Hermitian part of T(2)T^{(2)} and state which intermediate states are present. The check passes only if the complex conjugation reverses initial and final labels and the phase-space measure matches the state normalization.

Solution

The coefficient of g2g^2 in TT=iTTT-T^\dagger=iT^\dagger T is

T(2)T(2)=iT(1)T(1).T^{(2)}-T^{(2)\dagger}=iT^{(1)\dagger}T^{(1)}.

Between states ff and ii, insert the complete asymptotic identity to obtain i(Mfi(2)Mif(2))=XSX1dΦXMXf(1)MXi(1)-i(\mathcal M^{(2)}_{fi}-\mathcal M^{(2)*}_{if})= \sum_X S_X^{-1}\int\mathrm d\Phi_X\, \mathcal M^{(1)*}_{Xf}\mathcal M^{(1)}_{Xi}. The sum contains every kinematically open physical state reachable at first order, including its spin/species sums; SX1S_X^{-1} supplies its identical-particle factor. It is not restricted to the elastic channel unless all others are closed or excluded by exact quantum numbers.

  • Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014, § 24.1, printed pp. 452–465. doi:10.1017/9781139540940.
  • Weinberg, Steven. The Quantum Theory of Fields, Volume I: Foundations. Cambridge University Press, 1995, § 3.6, printed pp. 147–151. doi:10.1017/CBO9781139644167.