Soft-Gluon Corrections in QCD

• Soft-gluon corrections are QCD effects occurring near partonic thresholds due to low-energy gluon emissions, necessitating the resummation of large Sudakov logarithms for accurate predictions.
• They are characterized by process-dependent anomalous dimensions and factorize into hard, soft, and jet functions, with resummation performed in Mellin space.
• These corrections significantly enhance theoretical predictions for LHC observables such as top-quark, heavy boson, and jet production while reducing scale uncertainties.

Soft-gluon corrections are the dominant higher-order QCD effects in hard-scattering processes near partonic threshold, where the invariant mass of unobserved final-state radiation is small. In this regime, real emissions are restricted to soft (low-energy) gluons whose radiation generates large Sudakov logarithms—powers of $\ln(s_4/M^2)$, where $s_4$ is a measure of unresolved radiation and $M$ is a hard scale—which must be resummed for reliable predictions. Soft-gluon corrections are universal, governed by process-dependent anomalous dimensions, and can be systematically resummed and expanded to high orders, providing crucial theoretical improvements in many LHC observables, including but not limited to top-quark, heavy boson, and jet production.

1. Physical Origin and Definition of Soft-Gluon Corrections

In perturbative QCD, the emission of low-energy (soft) gluons from partonic scattering leads to large logarithmic enhancements in the cross section near threshold. The relevant kinematic region is characterized by a threshold variable—$s_4$ in single-particle-inclusive (1PI) kinematics, or $1-z$ in pair-invariant-mass (PIM) or deep inelastic scattering (DIS) kinematics—with $s_4 \to 0$ or $z\to1$. At $n$-th order in $\alpha_s$, the leading singular terms take the form of plus-distributions,

$$\left[\frac{\ln^k(s_4/M^2)}{s_4}\right]_+, \qquad k\leq 2n-1,$$

or analogously $s_4$0 for $s_4$1 (Kidonakis, 2018, Das et al., 2019).

These terms represent enhancements arising from the incomplete cancellation between real and virtual soft-gluon emissions, and are formally singular as $s_4$2, but integrable due to the plus-prescription. The universal structure of these corrections allows their resummation to all orders in $s_4$3.

2. Factorization Structure and Resummation Formalism

The factorization properties of soft-gluon radiation permit the cross section to be refactorized near threshold as

$s_4$4

where $s_4$5 is the Mellin moment conjugate to $s_4$6 or $s_4$7 (Kidonakis, 2018, Forslund et al., 2019, Das et al., 2019, Kidonakis, 2014, Kidonakis et al., 2023). Here,

• $s_4$8 is the hard function, collecting process-dependent virtual corrections,
• $s_4$9 is the soft function, encoding noncollinear soft-gluon emissions and its evolution is governed by the soft anomalous dimension matrix $M$0,
• $M$1 are jet (or collinear) functions for the initial or final colored partons, resumming collinear logarithms.

The renormalization-group (RG) evolution of $M$2 and $M$3 exponentiates the large logarithms. The soft function $M$4 obeys the RG equation

$M$5

where $M$6 is calculable as an expansion in $M$7 and encodes the color and kinematic correlations among external partons (Kidonakis, 2018, Forslund et al., 2019).

3. Mellin-space Resummation and Plus-distribution Expansion

By taking Mellin moments, the threshold logarithms $M$8 become $M$9 and exponentiate, such that in Mellin space, the cross section has the form: $s_4$0 where $s_4$1 are functions of anomalous dimensions and process-dependent constants (Das et al., 2019, Kidonakis et al., 2023).

The inversion back to $s_4$2 space produces the plus-distributions. The fixed-order expansion of the resummed cross section, conducted up to, for example, (N)NNLO or N$s_4$3LO, yields

$s_4$4

with explicit, process-dependent coefficients $s_4$5, typically determined by cusp and soft anomalous dimensions, $s_4$6-function coefficients, and matching constants (Kidonakis, 2014, Kidonakis et al., 2024, Kidonakis et al., 2024).

4. Soft Anomalous Dimension Matrices and Process Dependence

The soft anomalous dimension matrix $s_4$7 plays a central role in determining the structure of soft-gluon corrections. It is computed from the UV poles of renormalized eikonal diagrams and generally admits a perturbative expansion,

$s_4$8

The explicit form of $s_4$9 depends on the color representation and kinematics of the process. For processes with multiple external colored legs, $1-z$0 is a nontrivial matrix in color space, containing logarithms of Mandelstam invariants, mass scales, and, for heavy-quark production, rapidity-dependent terms (Kidonakis, 2018, Forslund et al., 2019, Kidonakis et al., 2024).

In single-color channels or color-singlet production (e.g., Drell–Yan or $1-z$1), $1-z$2 reduces to a scalar. When all particles are massive and all final-state particles are observed, as in $1-z$3 or $1-z$4, a full 1PI-kinematic resummation requires computation of the relevant soft matrices with all mass and kinematic dependence (Kidonakis et al., 2023, Kidonakis et al., 2024).

5. Phenomenological Impact and Numerical Significance

Soft-gluon corrections lead to large and often dominant enhancements in total and differential cross sections, particularly near threshold. For example, in top-pair production at 13 TeV, approximate N$1-z$5LO (aN$1-z$6LO) soft-gluon corrections increase the NNLO result by $1-z$7 and halve the residual scale uncertainty to $1-z$8 (Kidonakis, 2014, Piclum et al., 2018). For single-top and $1-z$9 production, aNNLO or aN$s_4 \to 0$0LO corrections add $s_4 \to 0$1 at NNLO and a further $s_4 \to 0$2 at N$s_4 \to 0$3LO, while reducing scale uncertainties to 2–4% (Kidonakis, 2016, Kidonakis, 2018).

In associated top production ($s_4 \to 0$4, $s_4 \to 0$5, $s_4 \to 0$6, $s_4 \to 0$7), the aNNLO and aN$s_4 \to 0$8LO soft-gluon corrections enhance the NLO rates by $s_4 \to 0$9–$z\to1$0 and significantly reduce theoretical uncertainties. For $z\to1$1 at 13 TeV, the aN$z\to1$2LO prediction is $z\to1$3 pb, in excellent agreement with experimental measurements (Kidonakis et al., 2024). Similar observations hold for $z\to1$4 and $z\to1$5 production (Forslund et al., 2021, Kidonakis et al., 2022).

In W/Z-boson $z\to1$6 distributions and heavy Higgs pair production, the approximate NNLO or N$z\to1$7LO soft terms yield 8–15% enhancements over NLO and shrink scale dependence to the level of a few percent (Kidonakis et al., 2014, Kidonakis et al., 2024).

6. Extensions: Subleading-Power and Medium Effects

Beyond leading-power (eikonal) accuracy, subleading (next-to-eikonal, NE) soft-gluon corrections exhibit partial exponentiation, with systematized diagrammatic and path-integral approaches yielding effective NE Feynman rules (Laenen et al., 2010). These contribute terms suppressed by one power of $z\to1$8 or $z\to1$9 and are necessary for a complete threshold-resummed prediction at subleading power.

In a QCD medium, soft-gluon radiation is modified by screening masses (Debye mass $n$0) and nontrivial thermal averages. For $n$1 in a quark–gluon plasma, corrections to the standard Gunion–Bertsch (GB) formula have been established. The modified soft-gluon yield receives an extra contribution proportional to $n$2, giving a 10–30% increase in soft-gluon multiplicity at $n$3 GeV (Das et al., 2010). This impacts jet quenching observables sensitive to the low-energy sector.

7. Theoretical Uncertainties and Resummation Precision

Soft-gluon resummation significantly reduces theoretical uncertainties in QCD predictions. After inclusion of soft-gluon corrections at (N)NLO or (N)$n$4LO, scale variation becomes a subdominant source of error, at the level of $n$5–$n$6 in most benchmark LHC observables (Kidonakis, 2014, Kidonakis, 2018, Kidonakis et al., 2024). PDF uncertainties are typically at the 2–4% level, becoming the leading uncertainty at high mass or high $n$7.

The soft-gluon approximation, when constructed from all relevant anomalous dimensions and matching to fixed-order results, closely reproduces exact calculations (within $n$8\% for total rates and a few percent for differential distributions). This justifies its use in precision predictions and motivates further extensions to even higher logarithmic and fixed-order accuracy (Das et al., 2019, Kidonakis et al., 2024).

---

References:

• (Das et al., 2010, Laenen et al., 2010, Kidonakis et al., 2014, Kidonakis, 2014, Kidonakis, 2016, Piclum et al., 2018, Kidonakis, 2018, Das et al., 2019, Forslund et al., 2019, Forslund et al., 2021, Kidonakis et al., 2022, Kidonakis et al., 2023, Kidonakis et al., 2024, Kidonakis et al., 2024)