Scattering calculations for phenomenology

Take a measured observable: the cumulative thrust fraction in $e^+e^-\to\text{hadrons}$ at center-of-mass energy $Q$,

$$R_\tau(\tau_c;Q) =\frac{1}{\sigma_{\mathrm{had}}(Q)} \int_0^{\tau_c}\mathrm d\tau\, \frac{\mathrm d\sigma}{\mathrm d\tau}, \qquad \tau=1-T.$$

The task is to predict one declared bin of this quantity and compare it with data at the same physical level. Doing that well requires more than an amplitude: the state normalization, perturbative multiplicities, subtraction, phase-space integration, possible factorization and resummation, long-distance corrections, detector treatment, and uncertainty model must form one consistent calculation. Thrust is a useful running example because it is measured, infrared and collinear safe, and sensitive to several separated scales near its two-jet endpoint ALEPH Collaboration 2004, pp. 457–486.

Required background. You should be able to normalize a rate from stable external states using LSZ and tree amplitudes, separate ultraviolet from infrared singularities using loops and regularization, and test an on-shell gauge amplitude using QED and Yang–Mills theory. You should also be able to evolve renormalized parameters with the renormalization group, distinguish matching from running using effective field theory, and apply the soft and collinear tests in the infrared-observables lesson. For a data comparison, you also need basic probability, covariance, and numerical-integration reasoning.

Specify the thrust bin before calculating

For massless final-state three-momenta,

$$T=\max_{\lvert\widehat{\boldsymbol n}\rvert=1} \frac{\sum_i\lvert\boldsymbol p_i\cdot \widehat{\boldsymbol n}\rvert} {\sum_i\lvert\boldsymbol p_i\rvert}.$$

Two back-to-back partons have $\tau=0$, while extra wide-angle radiation moves events to larger $\tau$. The definition and its original use as a jet discriminator are due to Farhi 1977, pp. 1587–1588. A prediction is not specified by writing “thrust at NLO.” At minimum, fix the following:

Item	Running-example choice
Incoming state	Unpolarized $e^+e^-$ at $Q=91.2\ \mathrm{GeV}$
Reported quantity	$R_\tau(0.10;Q)$, with the denominator expanded consistently to the same fixed order
Perturbative content	Massless five-flavor QCD through the first correction relative to the Born rate, matched to an NLL endpoint resummation
Coupling and scales	$\alpha_s^{\overline{\mathrm{MS}}}(\mu_R)$ with central $\mu_R=Q$; correlated variations are retained
Long-distance treatment	A declared hadronization transfer or power-correction model, with its parameters and domain stated
Comparison level	A particle-level bin using exactly the experiment’s stable-particle, photon, acceptance, and normalization definition
Output	Central value, separated theory and numerical variations, comparison covariance, and enough benchmark values to reproduce the bin

This is a prototype, not a universal prescription. At the $Z$ pole, the electroweak input scheme, photon treatment, heavy-quark mass corrections, and the precise experimental definition must be added before quoting a result. The choice $\tau_c=0.10$ also places the example in a region where endpoint logarithms matter but nonperturbative effects cannot simply be ignored.

Take these first three actions:

1. Copy the measured bin definition verbatim into the calculation specification. Fix $Q$, particle or object definitions, cuts, normalization, and whether the published result is detector-level, particle-level, or parton-level.
2. Freeze the normalization and accuracy convention. State the $S$-matrix convention, incoming averages, flux, phase-space measure, coupling scheme, central scales, perturbative powers retained, and whether a normalized ratio is expanded or left unexpanded.
3. Make a contribution table before generating events. List every Born, virtual, real, counterterm, parton-density, and long-distance contribution; attach its final-state multiplicity, coupling power, integration space, measurement weight, and independent check.

These actions prevent a common failure: producing an accurate number for a different observable from the one that was measured.

Normalize amplitudes and count multiplicities

With relativistic state normalization, the weighted partonic rate has the form

$$\sigma[F] =\frac{1}{\mathcal F} \sum_n\frac{1}{S_n} \int\mathrm d\Phi_n\, \overline{\lvert\mathcal M_n\rvert^2}\,F_n.$$

For massless $e^+e^-$ beams in their center-of-mass frame, $\mathcal F=2Q^2$. The bar must mean the same thing in every contribution: average over the four incoming spin states and sum over allowed final spins, colors, and flavors. For the $q\bar q$ and $q\bar qg$ states used below, $S_n=1$ because the labeled particles are not identical copies.

For the first QCD correction to the cumulative rate, write

$$\begin{aligned} B&=\overline{\lvert\mathcal M_2^{(0)}\rvert^2},\\ V&=2\,\operatorname{Re} \overline{\mathcal M_2^{(0)*}\mathcal M_2^{(1)}},\\ R&=\overline{\lvert\mathcal M_3^{(0)}\rvert^2}. \end{aligned}$$

Here $\mathcal M_2^{(0)}$ is the Born amplitude for $e^+e^-\to q\bar q$, $\mathcal M_2^{(1)}$ is its ultraviolet-renormalized one-loop QCD correction, and $\mathcal M_3^{(0)}$ describes $e^+e^-\to q\bar qg$. Thus $B$ and $V$ are integrated on two-particle phase space, whereas $R$ is integrated on three-particle phase space. For the cumulative bin,

$$F_2=1,\qquad F_3=\Theta(\tau_c-\tau_3).$$

The differential distribution away from $\tau=0$ begins at $O(\alpha_s)$, even though the cumulative rate has an $O(1)$ Born term. Always attach the actual retained powers of $\alpha_s$ to an accuracy label; “NLO thrust” can otherwise refer to different truncations.

Before squaring a gauge amplitude, replace the external gluon polarization in the complete $q\bar qg$ amplitude by its momentum. The result must vanish. Emission from the quark and antiquark must both be present for this Ward check to work. Only after it passes is a covariant polarization sum safe to use.

Make real–virtual cancellation local

Dimensional regularization puts the infrared poles of $V$ in $d=4-2\epsilon$, while the unresolved regions of $R$ are nonintegrable in four dimensions. A local subtraction term $\mathrm d\sigma^A$ must reproduce the real contribution in each singly soft or collinear limit and map $\Phi_3$ to an on-shell, momentum-conserving $\widetilde\Phi_2$. For the running bin,

$$\begin{aligned} \Sigma_{\mathrm{NLO}}(\tau_c) ={}&\int_{\Phi_2} \left[ \mathrm d\sigma^B+\mathrm d\sigma^V +\int_1\mathrm d\sigma^A \right]_{\epsilon=0}\\ &+\int_{\Phi_3} \left[ \mathrm d\sigma^R\,\Theta(\tau_c-\tau_3) -\mathrm d\sigma^A \right]_{\epsilon=0}. \end{aligned}$$

The subtraction term carries the mapped two-parton weight, which is one for $\tau_c>0$. Its integrated poles cancel the infrared poles of the ultraviolet-renormalized virtual term. The minus sign in the three-particle integral and the plus sign of its analytic integral on two-particle phase space are the add–subtract identity, not optional conventions. A complete NLO construction and its measurement-function conditions are given in Catani and Seymour 1997, §§ 2.1 and 7.1–7.2, pp. 297–299 and 343–346.

The implementation should demonstrate all of the following before the final integral is trusted:

• $R/A\to1$ along independently generated soft and collinear trajectories;
• cancellation of every $1/\epsilon^k$ pole between $V$ and $\int_1\mathrm d\sigma^A$;
• exact momentum conservation and on-shell conditions under the map;
• stability under any auxiliary restriction or slicing parameter; and
• agreement with the inclusive result when $\tau_c$ reaches its full physical range.

The last check is especially sharp for a normalized cumulative distribution: $R_\tau$ must approach one at the upper endpoint when numerator and denominator are expanded consistently.

Integrate the finite phase spaces

The subtracted three-particle integrand is finite but can still be sharply peaked. Use phase-space coordinates that resolve soft and collinear regions, apply the same thrust routine to ordinary and mapped momenta, and retain the Monte Carlo variance and correlations between bins. A credible numerical result reports the input parameters, code version, seed policy, sample count, precision, and benchmark phase-space points.

Useful convergence checks are physical as well as statistical:

• double the sample and verify the uncertainty decreases with the expected sampling law;
• repeat with independent seeds and compare pulls, not just central values;
• integrate a constant over the same phase-space map to test its Jacobian;
• compare an inclusive or analytically known moment; and
• test that tighter soft and collinear approaches improve the subtraction limit rather than destabilize it.

Small Monte Carlo error does not diagnose a missing channel, an inconsistent normalization, or a biased map.

Add incoming-hadron factorization only when it is needed

The $e^+e^-$ example has no incoming parton distributions and no factorization scale $\mu_F$. For a hadron–hadron observable, however, the collider-level prediction begins schematically as

$$\begin{aligned} \sigma_B =\sum_{a,b}\int_0^1\mathrm dx_1\,\mathrm dx_2\, &f_{a/h_1}(x_1,\mu_F)f_{b/h_2}(x_2,\mu_F)\\ &\times \widehat\sigma_{ab,B}(x_1,x_2;\mu_F,\mu_R). \end{aligned}$$

Initial-state collinear poles are absorbed into the parton distributions in a declared factorization scheme. The finite partonic coefficient, the PDFs, their perturbative evolution, and their error sets must use compatible orders and conventions. Residual $\mu_F$ dependence cancels only through the calculated order; it is distinct from $\mu_R$ dependence. A named hard, collinear, or soft momentum region is not by itself a factorization theorem: one also needs operator definitions, overlap treatment, a convolution, and a power-suppressed remainder Collins, Soper, and Sterman 2004, §§ 8–9, pp. 76–95, PDF.

For a Drell–Yan bin, the contribution table would therefore add PDF convolutions, mass-factorization counterterms on the Born phase space, all partonic channels opened at the requested order, and correlated PDF and $\mu_F$ variations. None of these should be copied into the $e^+e^-$ thrust calculation.

Resum only a demonstrated hierarchy, then match

For $\tau_c\ll1$, fixed-order coefficients contain powers of $L=\ln(1/\tau_c)$. When $\alpha_s L^2$ is not small, a finite fixed-order coefficient need not be a reliable approximation. The two-jet factorization of thrust separates a hard scale $\mu_H\sim Q$, jet scales $\mu_J\sim Q\sqrt{\tau_c}$, and a soft scale $\mu_S\sim Q\tau_c$. Renormalization-group evolution between these scales resums the logarithms. The factorization statement and its operator definitions determine this structure Becher, Broggio, and Ferroglia 2015, §§ 4.5–7. For broader continuously global event shapes, the conditions under which the logarithms exponentiate are analyzed in Banfi, Salam, and Zanderighi 2005, §§ 2–3; the names of momentum regions alone do not establish those conditions.

An additive matching formula through fixed order $k$ is

$$R_\tau^{\mathrm{match}} =R_\tau^{\mathrm{res}} +R_\tau^{\mathrm{FO},[k]} -\left[R_\tau^{\mathrm{res}}\right]_{[k]}.$$

Expanding this expression through order $k$ must reproduce the fixed-order result exactly; the subtraction removes double counting. Away from the endpoint, the resummation scales should merge smoothly toward a common hard scale so that resummation turns off. If no parametrically large logarithm or proved factorization supports the resummation, use fixed order and say so.

Keep partonic, hadronic, and detector layers distinct

Layer	Object in the running example	What changes the prediction
Partonic collider calculation	Renormalized and subtracted $e^+e^-\to q\bar q(+g)$ coefficient, optionally resummed and matched	Perturbative order, $\alpha_s$, scales, masses, electroweak inputs, subtraction, and numerical integration
Hadronic or particle level	Thrust evaluated on the experiment’s declared stable particles	Hadronization or power corrections, hadron decays, photon treatment, and the particle definition
Detector level	Reconstructed tracks, clusters, jets, selections, and bin migrations	Efficiency, acceptance, resolution, calibration, backgrounds, and response modeling

A partonic prediction should not be compared directly with reconstructed counts. There are two consistent directions:

• compare a hadron-level prediction with data unfolded to the same declared particle level, using the published covariance and unfolding scope; or
• forward-fold the particle-level prediction through the response and compare with detector-level counts.

In the second case, a simple binned expectation is

$$N_i^{\mathrm{pred}} =\mathcal L\sum_j R_{ij}\epsilon_j\sigma_j^{\mathrm{particle}} +b_i,$$

where luminosity times cross section gives a count and the response, efficiency, and migration probabilities are dimensionless. Applying a detector response to already unfolded data, or applying an inverse response again to an unfolded covariance, double counts the correction.

Assemble uncertainties before inference

The final prediction should separate effects with different meanings:

• perturbative truncation: correlated variations of $\mu_R$ and, for incoming hadrons, $\mu_F$; the envelope is a diagnostic of omitted orders, not automatically a confidence interval;
• resummation and matching: variations of hard, jet, and soft profiles, transition points, and matching prescription subject to the factorization constraints;
• parametric inputs: $\alpha_s$, masses, electroweak inputs, and their covariance;
• parton distributions: PDF replicas or eigenvectors and their correlations for hadronic initial states;
• long-distance modeling: hadronization parameters, power corrections, model dependence, and the range in which the model is applied;
• numerics: integration variance, interpolation, finite precision, and convergence diagnostics; and
• experimental comparison: luminosity, response, backgrounds, calibration, bin correlations, and any unfolding dependence.

Do not combine these automatically in quadrature. Preserve correlations across bins and state which variations are probabilistic. For an unfolded vector $\boldsymbol d$ with a suitable nonsingular covariance $C$, a transparent first comparison is

$$\chi^2 = (\boldsymbol d-\boldsymbol t)^{\mathsf T} C^{-1} (\boldsymbol d-\boldsymbol t).$$

A normalized distribution has an exact sum constraint and therefore a singular full-bin covariance; remove one dependent bin or use the appropriate pseudoinverse in the constrained subspace. For detector-level counts, nuisance-parameter likelihoods are usually more faithful than a Gaussian $\chi^2$, especially for small counts or constrained systematics. The large-sample conditions behind common profile-likelihood approximations are stated in Cowan, Cranmer, Gross, and Vitells 2011, §§ 2–3.

The tangible work product is a prediction vector for the declared bins, its correlation information, a benchmark table for each perturbative component, and a comparison at exactly one physical layer. It should pass normalization, dimension, Ward, pole-cancellation, unresolved-limit, endpoint, scale-merging, and numerical-convergence checks before it supports an inference about $\alpha_s$ or new interactions.

Choose this path for the observable, not for every QFT problem

Use this route when the central question is: What does a specified scattering measurement predict, and with what controlled comparison to data?

• If the main work is constructing a particle, hadron, or nuclear description and deciding which degrees of freedom are relevant, use particle and nuclear physics.
• If the main work is choosing an operator basis, matching a high-scale model, running Wilson coefficients, and estimating EFT truncation, use renormalization and EFT for research.
• If the physics is fixed and the main obstacle is a reproducible numerical implementation, benchmark suite, or numerical validation, use computational field theory alongside this route.

These paths can meet in one project, but their primary outputs differ. Here the output is an observable-level prediction and comparison, not a new model or an operator-basis construction.

Checks and templates

1. Fill a prediction specification

Write a minimally complete specification for the running thrust bin.

Solution

Process: unpolarized $e^+e^-\to\mathrm{hadrons}$ at $Q=91.2\ \mathrm{GeV}$. Observable: $R_\tau(0.10;Q)=\sigma_{\mathrm{had}}^{-1} \int_0^{0.10}\mathrm d\tau\,\mathrm d\sigma/\mathrm d\tau$, with the same stable-particle and photon definition as the selected measurement. Theory: five-flavor massless QCD for the central perturbative coefficient, with declared electroweak inputs and separate heavy-quark corrections. Accuracy: terms through $O(\alpha_s)$ relative to the Born cumulative rate, NLL endpoint resummation, and additive matching whose expansion reproduces that fixed order. Inputs: $\alpha_s^{\overline{\mathrm{MS}}}(\mu_R)$, central $\mu_R=Q$, stated profile scales, and a declared hadronization treatment. Output: the bin value, correlated theory variations, Monte Carlo error, benchmark contributions, and comparison with unfolded particle-level data.

This statement is sufficient to determine which amplitudes, multiplicities, scales, and comparison corrections are required. It still does not supply their numerical values.

2. Check the NLO multiplicities and subtraction signs

List the contributions through the first QCD correction and identify where their infrared poles cancel.

Solution

$B=\lvert\mathcal M_2^{(0)}\rvert^2$ and $V=2\operatorname{Re}(\mathcal M_2^{(0)*}\mathcal M_2^{(1)})$ live on $\Phi_2$. The real term $R=\lvert\mathcal M_3^{(0)}\rvert^2$ lives on $\Phi_3$. A local approximation $A$ is subtracted as $R F_3-AF_2$ on $\Phi_3$ and integrated analytically over the unresolved one-particle variables. Its integral is added as $V+\int_1 A$ on $\Phi_2$. Thus the same counterterm enters with a minus sign at real multiplicity and a plus sign at Born multiplicity. The explicit infrared poles cancel in $V+\int_1A$, while $R-A$ is integrable in four dimensions. There is no initial-state mass-factorization counterterm for an $e^+e^-$ beam.

3. Expand the normalization and test the endpoint

Let the cumulative numerator and inclusive denominator be

$$\Sigma=\Sigma_0+\alpha_s\Sigma_1+O(\alpha_s^2), \qquad D=D_0+\alpha_s D_1+O(\alpha_s^2).$$

Expand $R_\tau=\Sigma/D$ consistently and state the full-range check.

Solution

Using $(D_0+\alpha_sD_1)^{-1} =D_0^{-1}-\alpha_sD_1D_0^{-2}+O(\alpha_s^2)$ gives

$$R_\tau =\frac{\Sigma_0}{D_0} +\alpha_s\left( \frac{\Sigma_1}{D_0} -\frac{\Sigma_0D_1}{D_0^2} \right) +O(\alpha_s^2).$$

When the cumulative bin covers the full physical thrust range, $\Sigma_0=D_0$ and $\Sigma_1=D_1$, so $R_\tau=1+O(\alpha_s^2)$. Failure of this check indicates an omitted channel, inconsistent measurement weight, or mismatched normalization. The matched prediction has a second independent check: expanding it through the retained fixed order must give this same result because $R^{\mathrm{res}}-[R^{\mathrm{res}}]_{[k]}$ starts beyond order $k$.

Continue into specialist calculations

Use Loop Integrals and Reduction when the virtual amplitude or master-integral checks are the bottleneck. Continue to Infrared Structure and Factorization for the proof and implementation of soft, collinear, overlap, and resummation structure, then to Infrared-Safe and Precision Observables for full measurement and phase-space calculations. Hadron-initiated work also needs Perturbative QCD and Partons; precision fits and nuisance-parameter interpretation continue in Precision Standard Model Observables and Inference.

References

• ALEPH Collaboration. “Studies of QCD at $e^+e^-$ Centre-of-Mass Energies between 91 and 209 GeV.” European Physical Journal C 35 (2004): 457–486. DOI.
• Banfi, Andrea, Gavin P. Salam, and Giulia Zanderighi. “Principles of General Final-State Resummation and Automated Implementation.” Journal of High Energy Physics 03 (2005): 073. DOI.
• Becher, Thomas, Alessandro Broggio, and Andrea Ferroglia. Introduction to Soft-Collinear Effective Theory. Lecture Notes in Physics 896. Cham: Springer, 2015. DOI.
• Catani, Stefano, and Michael H. Seymour. “A General Algorithm for Calculating Jet Cross Sections in NLO QCD.” Nuclear Physics B 485 (1997): 291–419; erratum 510 (1998): 503–504. DOI.
• Collins, John C., Davison E. Soper, and George Sterman. “Factorization of Hard Processes in QCD.” In Perturbative Quantum Chromodynamics, edited by A. H. Mueller, 1–91. Singapore: World Scientific, 1989. arXiv PDF.
• Cowan, Glen, Kyle Cranmer, Eilam Gross, and Ofer Vitells. “Asymptotic Formulae for Likelihood-Based Tests of New Physics.” European Physical Journal C 71 (2011): 1554. DOI.
• Farhi, Edward. “Quantum Chromodynamics Test for Jets.” Physical Review Letters 39 (1977): 1587–1588. DOI.