Quark Contact Interactions: EFT & Hadron Physics

• Quark contact interactions are effective local operators that model momentum-independent quark scattering, capturing both high-energy compositeness signals and nonperturbative bound-state dynamics.
• In collider physics, these interactions appear as dimension-6 four-quark operators suppressed by a scale, modifying QCD amplitudes through both interference and quadratic contributions.
• Nonperturbative approaches use contact kernels to mimic finite-range gluon exchanges, preserving symmetries and explaining confinement along with dynamical chiral symmetry breaking.

Quark contact interactions are effective local or momentum-independent interactions involving quark fields that arise when shorter-distance dynamics is not resolved explicitly. In collider phenomenology they usually denote dimension-6 four-quark operators suppressed by a scale $\Lambda$, intended to capture quark compositeness or heavy new-physics exchange and tested through jet and dijet production at high momentum transfer. In nonperturbative hadron physics, the same expression denotes symmetry-preserving contact kernels that replace finite-range gluon exchange in NJL- or DSE/BSE-based descriptions of confinement and dynamical chiral symmetry breaking. The two usages share locality in the effective description but differ in kinematic regime, operator content, and physical interpretation (Gao, 2013, Gutiérrez-Guerrero et al., 16 Apr 2026).

1. Effective-theory structure in collider physics

In the collider-EFT usage, quark contact interactions are organized as four-quark operators in an expansion in $1/\Lambda^2$. A standard flavor-symmetric, electroweak-isoscalar basis consists of six operators with definite chirality and color structure, conventionally written as LL singlet/octet, LR singlet/octet, and RR singlet/octet. In compact notation,

$$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$

with $X,Y\in\{L,R\}$, and one convenient normalization is

$$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$

In this convention, the signs of $\lambda_i$ control constructive or destructive interference with QCD (Gao et al., 2012).

The observable effect is a modification of the QCD partonic amplitude by interference and quadratic contact terms,

$$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$

so that hadronic cross sections take the EFT form

$$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$

The interference term is linear in the Wilson coefficients, while the pure-contact contribution is quadratic. Because QCD dijet production is dominated by forward Rutherford-like $t$-channel scattering, whereas contact interactions are more isotropic, the highest sensitivity occurs at large jet transverse momentum or dijet invariant mass and in central angular regions, equivalently at small $\chi=e^{|y_1-y_2|}$ or small $1/\Lambda^2$0 (Gao, 2013).

Experimental papers often adopt a different but equivalent normalization. CMS, following Eichten-type compositeness conventions, writes

$1/\Lambda^2$1

and defines benchmark models $1/\Lambda^2$2, $1/\Lambda^2$3, $1/\Lambda^2$4, $1/\Lambda^2$5, and $1/\Lambda^2$6. In proton-proton scattering, positive $1/\Lambda^2$7 or $1/\Lambda^2$8 corresponds to destructive interference with QCD, whereas negative $1/\Lambda^2$9 gives constructive interference (Collaboration, 2014).

2. Perturbative predictions, NLO QCD, and CIJET

Exact NLO QCD corrections to dijet production induced by quark contact interactions were worked out for the full six-operator basis, including one-loop renormalization, operator mixing, and renormalization-group evolution of the Wilson coefficients. The NLO calculation uses dimensional regularization, explicit virtual amplitudes, real-emission corrections, and either phase-space slicing or Catani–Seymour subtraction to render the cross section infrared finite. The corresponding RG-improved prediction is written as

$$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$0

so that large logarithms $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$1 from Wilson-coefficient running are resummed without double counting (Gao et al., 2012).

A central practical implementation is CIJET 1.0, a Fortran program for single-inclusive jet and dijet cross sections induced by four-quark contact interactions through NLO in QCD. It supports all six chiral/color structures, arbitrary active $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$2, pp or $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$3 beams, single-inclusive jet observables $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$4, and dijet distributions differential in $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$5 and one of $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$6, $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$7, or $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$8. The CI-induced contribution in each analysis bin is parameterized as a polynomial in $$\mathcal{O}^{(1)}_{XY} = \left(\bar q \gamma_\mu P_X q\right)\left(\bar q \gamma^\mu P_Y q\right),\qquad \mathcal{O}^{(8)}_{XY} = \left(\bar q T^A \gamma_\mu P_X q\right)\left(\bar q T^A \gamma^\mu P_Y q\right),$$9 and $X,Y\in\{L,R\}$0, with logarithmic NLO coefficients proportional to $X,Y\in\{L,R\}$1, where all $X,Y\in\{L,R\}$2 coefficients vanish at LO (Gao, 2013).

CIJET also implements a fast-grid interpolation over $X,Y\in\{L,R\}$3, with default grid sizes $X,Y\in\{L,R\}$4 and $X,Y\in\{L,R\}$5 and fourth-order interpolation. Any LHAPDF set may be projected onto the stored weights via the ciconv utility, and cross sections for arbitrary $X,Y\in\{L,R\}$6 and operator choices are then obtained with cixsec. The reported interpolation accuracy is at the $X,Y\in\{L,R\}$7 level relative to direct NLO integration, and there is no $X,Y\in\{L,R\}$8-initiated contribution to CI-induced jet cross sections up to NLO (Gao, 2013).

Phenomenologically, NLO corrections are not a small formal refinement. Representative LHC studies find that they generally reduce the CI-induced cross sections by several tens of percent, especially in large-$X,Y\in\{L,R\}$9 bins, and significantly reduce renormalization- and factorization-scale dependence. For the ratio observable $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$0, the residual theoretical uncertainty from scale variations is about $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$1–$$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$2 in the mass range explicitly studied (Gao et al., 2012).

3. Angular observables and collider constraints

The canonical angular variable is

$$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$3

with $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$4 and $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$5. For massless partons, Rutherford-like QCD scattering gives an approximately flat distribution in $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$6, whereas contact interactions populate low $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$7 more strongly because the scattering is more isotropic (Collaboration, 2014).

ATLAS used early 7 TeV data to measure normalized dijet $$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$8 distributions and the centrality ratio

$$\mathcal{L}_{\rm NP}=\frac{1}{2\Lambda^2}\sum_{i=1}^{6} c_i O_i,\qquad c_i=4\pi\lambda_i.$$9

In the highest mass bin it also defined

$\lambda_i$0

with the fourth-bin upper edge at $\lambda_i$1. For the left-handed isoscalar operator

$\lambda_i$2

ATLAS found agreement with SM QCD and excluded $\lambda_i$3 at $\lambda_i$4 CL for destructive interference, with an expected limit of $\lambda_i$5 (Collaboration, 2010).

CMS later analyzed dijet angular distributions at $\lambda_i$6 TeV using $\lambda_i$7, normalized the distribution within each $\lambda_i$8 bin, included exact NLO CI corrections from CIJET and one-loop electroweak corrections to the SM baseline, and found no significant deviation from NLO QCD+EW. The benchmark LL/RR model was excluded up to $\lambda_i$9 for destructive interference and $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$0 for constructive interference at $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$1 CL. Other NLO limits reached $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$2 and $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$3 for $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$4, $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$5 and $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$6 for $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$7, and $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$8 and $$|\mathcal M|^2 = |\mathcal M_{\rm QCD}|^2 + 2\,{\rm Re}\!\left(\mathcal M_{\rm QCD}\mathcal M_{\rm CI}^\ast\right)+|\mathcal M_{\rm CI}|^2,$$9 for $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$0 (Collaboration, 2014).

Analysis	Observable and dataset	95% CL outcome
ATLAS 7 TeV	Dijet $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$1 and $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$2, $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$3	$$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$4 excluded for LL destructive interference
CMS 8 TeV	Normalized dijet $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$5, $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$6	$$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$7, $$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$8
CMS 8 TeV	Same dataset, NLO CI benchmarks	$$\sigma = \sigma_{\rm SM} + \frac{1}{\Lambda^2}\sigma_{\rm INT} + \frac{1}{\Lambda^4}\sigma_{\rm CI}.$$9, $t$0

A recurrent technical point is that NLO and electroweak corrections matter at the level of the exclusion. CMS found that the NLO limits are systematically weaker than the corresponding LO limits because exact NLO treatment reduces the low-$t$1 enhancement, and that omitting electroweak corrections would lower the observed LO $t$2 limit from $t$3 to $t$4 (Collaboration, 2014).

4. Interference, truncation, degeneracies, and EFT validity

The dominant phenomenology of a contact operator is often controlled by the $t$5 interference term rather than by the pure $t$6 contribution. This makes sign conventions physically consequential and also explains why analyses commonly quote separate constructive and destructive limits. It also motivates the frequent use of an “interference-only” or truncated EFT approximation in which quadratic $t$7 terms are dropped. CIJET supports both treatments directly: one either retains or discards the quadratic terms in the bin-level parameterization (Gao, 2013).

This truncation issue is tied to EFT validity. Since the contact approximation assumes characteristic momentum transfer below $t$8, bounds extracted from the highest kinematic bins can become sensitive to the inclusion of $t$9 terms and to the choice of highest-bin cutoff. A practical recommendation in the CIJET documentation is therefore to report both truncated $\chi=e^{|y_1-y_2|}$0 and full $\chi=e^{|y_1-y_2|}$1 results when limits approach the kinematic reach (Gao, 2013).

Theoretical uncertainties are also structure-dependent. Up to NLO there is no $\chi=e^{|y_1-y_2|}$2 contribution to CI-induced jet production, so the relevant luminosities are predominantly $\chi=e^{|y_1-y_2|}$3 and $\chi=e^{|y_1-y_2|}$4 at high $\chi=e^{|y_1-y_2|}$5, which enhances PDF uncertainty in the tails. In the CMS 8 TeV analysis, PDF uncertainty on the normalized $\chi=e^{|y_1-y_2|}$6 distributions remained at the percent level, but scale variations changed the normalized shapes by less than $\chi=e^{|y_1-y_2|}$7 in the lowest $\chi=e^{|y_1-y_2|}$8 bin and less than $\chi=e^{|y_1-y_2|}$9 in the highest one (Collaboration, 2014).

Single-operator limits can overstate the true reach when several operators are simultaneously present. A symmetry-protected minimal model with two representative four-quark operators,

$1/\Lambda^2$00

was used to show that a two-operator fit to $1/\Lambda^2$01 of 7 TeV pseudo-data gives $1/\Lambda^2$02 and $1/\Lambda^2$03 at $1/\Lambda^2$04 CL, whereas a single-operator treatment would give $1/\Lambda^2$05. This illustrates that interference among different Lorentz and color structures can weaken apparent compositeness bounds (Bazzocchi et al., 2011).

5. Nonperturbative QCD contact interactions

In hadron-structure theory, “contact interaction” denotes something different: a momentum-independent effective gluon exchange used inside Dyson–Schwinger and Bethe–Salpeter equations. A symmetry-preserving realization replaces the gluon kernel by

$1/\Lambda^2$06

or, in equivalent notation,

$1/\Lambda^2$07

and then applies proper-time regularization to preserve Lorentz covariance and the vector and axial Ward–Takahashi identities. The result is a momentum-independent dressed-quark propagator with a constant constituent mass $1/\Lambda^2$08, algebraic meson and diquark Bethe–Salpeter equations, and a framework that reproduces the mass ordering $1/\Lambda^2$09 across light, heavy-light, and heavy sectors while remaining explicitly distinct from collider compositeness operators (Gutiérrez-Guerrero et al., 16 Apr 2026).

This hadronic CI framework has been extended to flavor-dependent couplings. In a $1/\Lambda^2$10 NJL model, vacuum polarization generates flavor-dependent four-quark couplings $1/\Lambda^2$11 and scalar–pseudoscalar splittings $1/\Lambda^2$12. Neutral-sector mixing terms $1/\Lambda^2$13 arise from flavor-symmetry breaking and scale as powers of constituent-mass differences, $1/\Lambda^2$14 with $1/\Lambda^2$15. These mixings enter meson Bethe–Salpeter kernels, can induce mixed gap equations after rotation to the fundamental basis, and are supplemented by sixth-order $1/\Lambda^2$16-breaking quark–antiquark interactions that reduce in mean field to additional effective four-quark mixings (Braghin, 13 Jan 2025).

A related SU(4) formulation constructs flavored contact-interaction couplings from one-loop vacuum polarization in background constituent-quark currents. In that approach a single mass scale determines the flavored couplings, and with realistic isospin breaking $1/\Lambda^2$17 one obtains strong-interaction mass splittings

$1/\Lambda^2$18

The same setup yields meson masses and decay constants in good agreement with reference values for $1/\Lambda^2$19, $1/\Lambda^2$20, $1/\Lambda^2$21, and $1/\Lambda^2$22, while the $1/\Lambda^2$23 comes out about $1/\Lambda^2$24 lighter than experiment, illustrating the limitations of a momentum-independent kernel in heavy quarkonium (Braghin et al., 12 Aug 2025).

6. Scope, limitations, and adjacent developments

The collider four-quark EFT has a sharply delimited scope. CIJET includes only four-quark contact interactions; it does not include electroweak or quark–lepton contact terms, heavy resonance propagator effects beyond the EFT limit, threshold or jet-radius resummation, or parton-shower and hadronization corrections. Its predictions are parton-level jet cross sections with parton-level jet algorithms (Gao, 2013). The hadronic contact-interaction framework has the opposite limitation: it is designed for low and intermediate momentum transfer, is nonrenormalizable, depends on infrared and ultraviolet regulators, and produces form factors that are too hard at large $1/\Lambda^2$25 because anomalous dimensions and realistic momentum dependence are absent (Gutiérrez-Guerrero et al., 16 Apr 2026).

The contact-interaction logic also propagates into broader EFT analyses. Third-generation four-quark operators such as $1/\Lambda^2$26 and $1/\Lambda^2$27 enter Higgs phenomenology at one or two loops, affecting $1/\Lambda^2$28, $1/\Lambda^2$29, $1/\Lambda^2$30, $1/\Lambda^2$31, and $1/\Lambda^2$32; in some cases current single-Higgs data constrain these operators competitively with top-sector fits, and they can degrade the extraction of the Higgs trilinear self-coupling if omitted from global SMEFT analyses (Alasfar et al., 2022). Likewise, for effective dark-matter models with a heavy $1/\Lambda^2$33-channel mediator coupled to quarks and to a Dirac fermion $1/\Lambda^2$34, integrating out the mediator necessarily induces both a $1/\Lambda^2$35 operator and a $1/\Lambda^2$36 contact operator. Using 7 TeV LHC limits, it was shown that for mediator masses $1/\Lambda^2$37 and perturbative couplings, the translated dijet contact-interaction bound can be stronger than the direct monojet bound on the dark-matter operator (Dreiner et al., 2013).

Taken together, these developments show that quark contact interactions are not a single theory but a family of effective descriptions. In collider physics they provide a controlled parameterization of heavy short-distance dynamics in the high-scale tail of jet production; in hadronic QCD they supply symmetry-preserving local kernels for bound-state calculations. The technical commonality is locality in the effective description. The physical content, however, is fixed by regime: quark compositeness and heavy mediator exchange at high energy, or nonperturbative confinement and chiral dynamics at low energy.