Higgs Boson Trilinear Self-Coupling

• Higgs Boson Trilinear Self-Coupling is defined as the coefficient of the cubic interaction in the Higgs potential after electroweak symmetry breaking, with its standard model value set by 3mₕ²/v.
• It plays a crucial role in double-Higgs production where interference effects, especially in gluon fusion, create a non-monotonic dependence on deviations from the SM expectation.
• The coupling is sensitive to loop corrections, renormalization schemes, and BSM modifications, making precise one-loop and multi-loop analyses vital for collider phenomenology.

The Higgs boson trilinear self-coupling is the coefficient of the cubic interaction of the physical Higgs field after electroweak symmetry breaking, and it is the first self-interaction parameter needed for reconstructing the Higgs potential. In the Standard Model tree-level normalization most commonly used for phenomenology, it is defined from the scalar potential by $\lambda_3=\partial^3 V/\partial h^3|_{h=0}$ with $H=(0,(v+h)/\sqrt{2})$, and obeys $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$; deviations are usually expressed through $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$ (Bahl et al., 2023). Because $\lambda_3$ enters Higgs-pair amplitudes, electroweak loop corrections to single-Higgs observables, and the structure of many Beyond-the-Standard-Model scalar sectors, it has become a standard precision target in collider phenomenology and in automated higher-order calculations (1212.5581).

1. Definition, normalizations, and vertex structure

The standard field-theoretic definition starts from the Higgs doublet expanded around the vacuum, $H=(0,(v+h)/\sqrt{2})$, and identifies the cubic interaction of the physical field $h$. With the Lagrangian convention $\mathcal{L}\supset -\lambda_3 h^3/3!$, the Standard Model tree-level relation is $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$, with $v\approx 246\ \text{GeV}$ (Bahl et al., 2023). The same normalization is used in many collider analyses, where the modifier $H=(0,(v+h)/\sqrt{2})$0 parameterizes departures from the Standard Model expectation (1212.5581).

A closely related but distinct convention writes the post-EWSB potential as $H=(0,(v+h)/\sqrt{2})$1. In that convention one has $H=(0,(v+h)/\sqrt{2})$2 as the coefficient in the potential, while the on-shell trilinear amplitude convention remains $H=(0,(v+h)/\sqrt{2})$3; the two are related by combinatorics (Degrassi et al., 2017). A convention-specific exception appears in the coupled-technicolor analysis, which consistently normalizes the cubic coupling as $H=(0,(v+h)/\sqrt{2})$4 and defines $H=(0,(v+h)/\sqrt{2})$5 relative to that choice (Doff et al., 2021). For cross-comparison across the literature, the distinction between the coefficient in the potential, the coefficient in the Lagrangian, and the Feynman-rule normalization is therefore nontrivial rather than merely notational.

Beyond tree level, the object of interest is not just a constant but the renormalized three-point function $H=(0,(v+h)/\sqrt{2})$6. A general one-loop representation is

$H=(0,(v+h)/\sqrt{2})$7

where $H=(0,(v+h)/\sqrt{2})$8 collects counterterms from the tree-level coupling and from the renormalization of masses, the vacuum expectation value, and any additional parameters entering the cubic interaction (Bahl et al., 2023). This makes the phenomenological quantity intrinsically scheme- and kinematics-dependent once higher orders are included.

2. Entry into double-Higgs production and interference structure

The dominant direct probe at hadron colliders is gluon-fusion Higgs-pair production. At amplitude level one may write

$H=(0,(v+h)/\sqrt{2})$9

or equivalently

$\lambda_3^{\text{SM(tree)}}=3m_h^2/v$0

where the three terms encode the pure box contribution, box–triangle interference, and pure triangle contribution (Heinrich et al., 2019). The destructive interference between the box and triangle topologies is the central dynamical feature of the process and is responsible for the characteristic non-monotonic dependence of the total rate on $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$1.

For full top-mass-dependent NLO QCD predictions at $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$2, the total cross section shows a pronounced minimum near $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$3, where destructive interference is maximal; the $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$4 spectrum is the most sensitive observable, especially in the low-to-intermediate invariant-mass region (Heinrich et al., 2019). A complementary interference analysis found that, for $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$5 via gluon fusion, the interference structure is almost maximally destructive and nearly independent of collider energy from $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$6 to $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$7, with $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$8 throughout that range (Dicus et al., 2015). This near constancy arises because the hadronic rate is dominated by the threshold region in partonic energy, where the destructive interference is strongest.

Other double-Higgs production mechanisms also depend on the trilinear coupling but with different interference patterns and smaller rates. In vector-boson fusion, the $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$9 dependence is again approximately quadratic, with a minimum around $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$0; in double Higgs-strahlung, the corresponding minimum is around $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$1; and in associated production with top pairs the sensitivity is weaker because the channel is more strongly controlled by the top Yukawa coupling and large QCD backgrounds (1212.5581). This suggests that $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$2 supplies the dominant statistical sensitivity, whereas VBF and $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$3 provide complementary coupling dependence and altered interference structure.

3. Renormalization, momentum dependence, and precise one-loop predictions

Modern calculations treat the trilinear interaction as a renormalized vertex rather than as a fixed constant. The framework anyH3 is a Python library for computing trilinear scalar couplings up to one loop in arbitrary renormalisable quantum field theories from UFO input. It automates on-shell, $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$4, and custom non-minimal renormalization schemes; includes external-leg corrections; optionally retains finite external momenta; and verifies UV-finiteness and decoupling properties across shipped models (Bahl et al., 2023). In that setup, the default evaluation point is zero external momentum, but mixed on-shell/off-shell kinematics can also be studied.

Momentum dependence is not merely formal. In the one-off-shell form factor analysis of $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$5, the renormalized $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$6 develops an imaginary part when thresholds are crossed. In the Standard Model this occurs for $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$7 for top loops, $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$8 and $$\kappa_\lambda\equiv \lambda_{hhh}/\lambda_{hhh}^{(0),\text{SM}}$$9 for gauge-boson loops, and $\lambda_3$0 for Higgs self-loops (Moyotl et al., 2016). In the quoted on-shell scheme, the one-loop correction was decomposed into approximately $\lambda_3$1 from the top-quark triangle, $\lambda_3$2 from Higgs self-interaction loops, and $\lambda_3$3 from the combined $\lambda_3$4 and $\lambda_3$5 contributions, giving a total Standard Model correction of approximately $\lambda_3$6 at the quoted kinematic point (Moyotl et al., 2016).

A distinct one-loop calculation in the zero-momentum approximation, performed in $\lambda_3$7 with $\lambda_3$8 and including top-quark, $\lambda_3$9, $H=(0,(v+h)/\sqrt{2})$0, and Higgs loops while neglecting Goldstones and lighter fermions, obtained

$H=(0,(v+h)/\sqrt{2})$1

to be compared with the tree-level value $H=(0,(v+h)/\sqrt{2})$2 for the input $H=(0,(v+h)/\sqrt{2})$3 and $H=(0,(v+h)/\sqrt{2})$4 (Zhang et al., 16 Oct 2025). The difference between this result and momentum-dependent on-shell form-factor calculations is a concrete illustration that loop-corrected “the trilinear coupling” is not a single universal number without specifying scheme, field renormalization, external kinematics, and particle content.

Finite-momentum effects can nevertheless be phenomenologically modest for representative points. In the THDM-I example implemented with anyH3, one external Higgs leg carries momentum $H=(0,(v+h)/\sqrt{2})$5 while the other two are on shell; the study notes that the integration of the total double-Higgs production cross section peaks around $H=(0,(v+h)/\sqrt{2})$6, and for the benchmark points displayed the momentum-dependent shift does not change the qualitative classification of the points relative to current $H=(0,(v+h)/\sqrt{2})$7 bounds (Bahl et al., 2023). A plausible implication is that zero-momentum approximations can remain useful for broad parameter scans, but point-by-point interpretation near thresholds or resonances requires the full three-point function.

4. Beyond-the-Standard-Model deformations and non-decoupling mechanisms

The trilinear coupling is exceptionally sensitive to extended Higgs sectors and to heavy states whose masses originate from electroweak symmetry breaking. In the composite-Higgs framework, explicit minimal models predict simple analytic rescalings. For MCHM4,

$H=(0,(v+h)/\sqrt{2})$8

whereas for MCHM5

$H=(0,(v+h)/\sqrt{2})$9

so that the trilinear coupling vanishes at $h$0 and changes sign beyond that point (Grober et al., 2010). In the same models, new $h$1 contact interactions appear and increasingly dominate $h$2 at large $h$3, thereby diluting direct sensitivity to $h$4 even when the total di-Higgs rate is strongly enhanced (Grober et al., 2010).

In weakly coupled extensions, non-decoupling can be equally pronounced. The anyH3 case studies show that in several $h$5-extended models, large mass splittings away from the decoupling limit induce sizable corrections because, when a BSM mass arises entirely through coupling to the SM-like Higgs, one has $h$6, so large masses imply large quartic couplings and hence large loop effects in $h$7 (Bahl et al., 2023). In the real $h$8 triplet extension with $h$9, the difference between on-shell and $\mathcal{L}\supset -\lambda_3 h^3/3!$0 renormalization of the triplet mass is used as a proxy for missing two-loop effects and illustrates a practical one-loop uncertainty estimate (Bahl et al., 2023).

The Inert Doublet Model provides an explicit threshold-driven example. After imposing theoretical constraints, dark-matter bounds, and limits on invisible Higgs decays, the one-loop off-shell vertex correction can exceed $\mathcal{L}\supset -\lambda_3 h^3/3!$1; the paper reports $\mathcal{L}\supset -\lambda_3 h^3/3!$2 at $\mathcal{L}\supset -\lambda_3 h^3/3!$3 for moderate inert masses, and peaks of approximately $\mathcal{L}\supset -\lambda_3 h^3/3!$4 at $\mathcal{L}\supset -\lambda_3 h^3/3!$5 and $\mathcal{L}\supset -\lambda_3 h^3/3!$6 at $\mathcal{L}\supset -\lambda_3 h^3/3!$7 near the $\mathcal{L}\supset -\lambda_3 h^3/3!$8 and $\mathcal{L}\supset -\lambda_3 h^3/3!$9 thresholds, respectively (Falaki, 2023). The mechanism is non-decoupling from heavy inert scalars whose masses are sourced by large quartics rather than by a large inert mass parameter alone.

Supersymmetric singlet extensions exhibit a related pattern. In the real NMSSM, one-loop corrections to effective trilinear couplings can modify Higgs-to-Higgs branching ratios by up to approximately $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$0 and shift $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$1 cross sections by roughly $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$2 to $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$3 relative to predictions based on effective tree-level trilinears, depending on the scenario (Nhung et al., 2013). In the CP-violating NMSSM, the newly computed $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$4 corrections in the gaugeless, zero-momentum limit are smaller than the preceding $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$5 terms but remain phenomenologically relevant; their inclusion in resonant di-Higgs production indicates that missing electroweak higher-order corrections may still be significant (Borschensky et al., 2022).

Fermionic extensions can also be constrained through $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$6. In the Standard Model plus a singlet vector-like top partner mixing with the top quark, the one-loop zero-momentum calculation finds that $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$7 grows rapidly with both the partner mass $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$8 and the mixing $\lambda_3^{\text{SM(tree)}}=3m_h^2/v$9, and, when interpreted with the ATLAS interval $v\approx 246\ \text{GeV}$0 at $v\approx 246\ \text{GeV}$1 CL, yields an upper bound on the singlet top-partner mass of about $v\approx 246\ \text{GeV}$2 under the assumptions of the analysis (Zhang et al., 16 Oct 2025).

5. Direct measurements and collider constraints

Experimental sensitivity is dominated by Higgs-pair production, but both direct and indirect channels already contribute. Run-2 ATLAS-based analyses summarized in the literature quote a combined single- and double-Higgs constraint of $v\approx 246\ \text{GeV}$3 with a $v\approx 246\ \text{GeV}$4 CL interval $v\approx 246\ \text{GeV}$5 under the assumption that new physics modifies only the Higgs self-coupling (Rossi, 2020). A later study cites current ATLAS constraints of $v\approx 246\ \text{GeV}$6 from single- and double-Higgs production at $v\approx 246\ \text{GeV}$7, while another uses the ATLAS interval $v\approx 246\ \text{GeV}$8 observed and $v\approx 246\ \text{GeV}$9 expected at $H=(0,(v+h)/\sqrt{2})$00 CL (Bahl et al., 2023).

The HL-LHC and future hadron-collider programs sharpen this substantially. A Snowmass study of $H=(0,(v+h)/\sqrt{2})$01 in the $H=(0,(v+h)/\sqrt{2})$02, $H=(0,(v+h)/\sqrt{2})$03, and $H=(0,(v+h)/\sqrt{2})$04 channels projects, for the HL-LHC at $H=(0,(v+h)/\sqrt{2})$05 with $H=(0,(v+h)/\sqrt{2})$06, an expected di-Higgs significance of $H=(0,(v+h)/\sqrt{2})$07 and a $H=(0,(v+h)/\sqrt{2})$08 CL upper limit on the production rate of $H=(0,(v+h)/\sqrt{2})$09 times the Standard Model rate. Under an SM di-Higgs signal hypothesis, the same study quotes $H=(0,(v+h)/\sqrt{2})$10 intervals $H=(0,(v+h)/\sqrt{2})$11 at $H=(0,(v+h)/\sqrt{2})$12 CL and $H=(0,(v+h)/\sqrt{2})$13 at $H=(0,(v+h)/\sqrt{2})$14 CL; for the FCC-hh at $H=(0,(v+h)/\sqrt{2})$15 with $H=(0,(v+h)/\sqrt{2})$16, the projected precision on the trilinear coupling is $H=(0,(v+h)/\sqrt{2})$17–$H=(0,(v+h)/\sqrt{2})$18 at $H=(0,(v+h)/\sqrt{2})$19 CL (Taliercio et al., 2022).

A ratio-based approach using

$H=(0,(v+h)/\sqrt{2})$20

emphasizes cancellations of common QCD systematics between double- and single-Higgs gluon-fusion production. In that framework, the combined theoretical uncertainty on the ratio is taken as $H=(0,(v+h)/\sqrt{2})$21, and the HL-LHC expectation without differential fitting is an uncertainty of approximately $H=(0,(v+h)/\sqrt{2})$22 and $H=(0,(v+h)/\sqrt{2})$23 on the self-coupling at $H=(0,(v+h)/\sqrt{2})$24; with $H=(0,(v+h)/\sqrt{2})$25, the trilinear coupling can already be constrained to be positive at $H=(0,(v+h)/\sqrt{2})$26 confidence level (Goertz et al., 2013).

Program or study	Setup	Quoted $H=(0,(v+h)/\sqrt{2})$27 reach
ATLAS Run-2 combination	up to $H=(0,(v+h)/\sqrt{2})$28 at $H=(0,(v+h)/\sqrt{2})$29	$H=(0,(v+h)/\sqrt{2})$30 at $H=(0,(v+h)/\sqrt{2})$31 CL (Rossi, 2020)
ATLAS constraint cited in phenomenology study	$H=(0,(v+h)/\sqrt{2})$32 single- and double-Higgs	$H=(0,(v+h)/\sqrt{2})$33 (Bahl et al., 2023)
ATLAS interval used in VLQ study	$H=(0,(v+h)/\sqrt{2})$34 CL observed/expected	$H=(0,(v+h)/\sqrt{2})$35, $H=(0,(v+h)/\sqrt{2})$36 (Zhang et al., 16 Oct 2025)
HL-LHC projection	$H=(0,(v+h)/\sqrt{2})$37, $H=(0,(v+h)/\sqrt{2})$38	$H=(0,(v+h)/\sqrt{2})$39 at $H=(0,(v+h)/\sqrt{2})$40 CL (Taliercio et al., 2022)
FCC-hh projection	$H=(0,(v+h)/\sqrt{2})$41, $H=(0,(v+h)/\sqrt{2})$42	$H=(0,(v+h)/\sqrt{2})$43–$H=(0,(v+h)/\sqrt{2})$44 precision at $H=(0,(v+h)/\sqrt{2})$45 CL (Taliercio et al., 2022)

The direct-extraction problem is complicated by coupling degeneracies. In gluon fusion, the trilinear interaction multiplies the triangle amplitude, whereas the top Yukawa coupling controls both triangle and box pieces. Two-parameter fits in $H=(0,(v+h)/\sqrt{2})$46 therefore display strong degeneracies unless external information on $H=(0,(v+h)/\sqrt{2})$47 is supplied (Rossi, 2020). This is why multichannel combinations and differential information in $H=(0,(v+h)/\sqrt{2})$48 are structurally more informative than inclusive di-Higgs rates alone.

6. Indirect probes and outstanding theoretical issues

Indirect constraints arise because $H=(0,(v+h)/\sqrt{2})$49 enters single-Higgs production and decay through electroweak loops. In the $H=(0,(v+h)/\sqrt{2})$50-framework used in the ATLAS dissertation, single-Higgs production cross sections and decay widths depend on $H=(0,(v+h)/\sqrt{2})$51 through a universal Higgs wavefunction factor

$H=(0,(v+h)/\sqrt{2})$52

together with process-dependent coefficients $H=(0,(v+h)/\sqrt{2})$53 and $H=(0,(v+h)/\sqrt{2})$54 for production and decay modes (Rossi, 2020). This is the basis for combined $H=(0,(v+h)/\sqrt{2})$55 fits and for the statement that indirect sensitivity can meaningfully improve direct bounds on $H=(0,(v+h)/\sqrt{2})$56.

Dedicated single-Higgs channels illustrate the size of the effect. In Higgs-plus-jet production at the LHC, the NLO electroweak correction proportional to $H=(0,(v+h)/\sqrt{2})$57 was computed to be $H=(0,(v+h)/\sqrt{2})$58 for the total cross section, and the corresponding corrections to the invariant-mass and Higgs-transverse-momentum distributions are described as almost flat and similar in size (Gao et al., 2023). In charged-current VBF single-Higgs production at the LHeC, the one-loop $H=(0,(v+h)/\sqrt{2})$59 dependence yields a parton-level $H=(0,(v+h)/\sqrt{2})$60 CL interval $H=(0,(v+h)/\sqrt{2})$61 for $H=(0,(v+h)/\sqrt{2})$62, broadening to $H=(0,(v+h)/\sqrt{2})$63 when approximately $H=(0,(v+h)/\sqrt{2})$64 signal survival after cuts and all backgrounds are included (Li et al., 2019).

Precision electroweak observables provide an additional indirect route. Two-loop diagrams containing an anomalous trilinear Higgs coupling shift $H=(0,(v+h)/\sqrt{2})$65 and $H=(0,(v+h)/\sqrt{2})$66, and the resulting bounds were found to be competitive with those from Higgs-pair production. In the combined fit quoted in that work, the best-fit value is $H=(0,(v+h)/\sqrt{2})$67, with $H=(0,(v+h)/\sqrt{2})$68 at $H=(0,(v+h)/\sqrt{2})$69 CL and $H=(0,(v+h)/\sqrt{2})$70 at $H=(0,(v+h)/\sqrt{2})$71 CL (Degrassi et al., 2017). The same paper argues that, at two loops, the anomalous-$H=(0,(v+h)/\sqrt{2})$72 treatment is equivalent to modifying the scalar potential by a tower of $H=(0,(v+h)/\sqrt{2})$73 operators, with the effects entering precision observables exclusively through gauge-boson self-energies (Degrassi et al., 2017).

Several theoretical issues remain persistent. Zero-momentum extractions can be gauge- and scheme-dependent beyond leading order, as emphasized in the singlet-vector-like-top study, which notes that the quoted number should be regarded as scheme- and gauge-dependent within the zero-momentum $H=(0,(v+h)/\sqrt{2})$74 setup (Zhang et al., 16 Oct 2025). One-loop calculations in automated frameworks can estimate missing higher orders through renormalization-scheme variation, but dedicated two-loop electroweak and QCD corrections are still needed for precision fits in many models (Bahl et al., 2023). A plausible implication is that the phenomenological quantity being constrained experimentally is increasingly a renormalized, kinematics-dependent vertex form factor embedded in a complete process amplitude, rather than a single model-independent constant.