Introduction
Heavy neutral leptons (HNLs) are a natural extension of the neutrino sector and an important target of searches for physics beyond the Standard Model (SM). Most searches rely on their mixing with active neutrinos. If further particles exist at a higher scale, however, HNLs can also interact through higher-dimensional operators. Transition dipoles are particularly interesting because they couple a neutral fermion directly to electroweak field strengths, opening production and decay channels even when mixing-induced interactions are small. Their collider phenomenology has recently been explored in a gauge-invariant framework in ref. [1 ] .
The experimental programme spans very different physical regimes. For light sterile states, neutrino scattering, stellar environments and cosmological evolution probe the photon dipole [2 , 3 ] . Meson decays and fixed-target experiments extend the laboratory coverage [4 , 5 ] . At higher masses, collider searches provide a direct probe of the dipole portal, through HNL production in Higgs decays [6 ] and in lepton [7 , 8 , 9 , 10 ] and hadron collisions [1 , 11 ] . The relevant observables range from displaced or radiative decays to charged-lepton and electroweak-boson final states. These channels probe different combinations of couplings and depend on the HNL lifetime, flavour structure and kinematic accessibility.
Precision electroweak measurements address a complementary question.
Even when HNLs are too heavy to be produced directly at a given collider, their virtual effects can modify the decay rates and angular asymmetries of Standard Model particles.
The baseline FCC-ee programme envisages approximately 6 × 10 12 6 \times10^{12} 6 × 1 0 12 Z Z Z bosons [12 ] , enabling precise measurements of a broad set of electroweak observables sensitive to small virtual corrections. Unlike a direct search, this test does not require reconstructing an HNL decay. Its interpretation does require understanding which corrections can be attributed to the dipole portal and which could instead arise from other interactions of the particles that generate it. Comparing direct and indirect searches is therefore useful both for their mass coverage and for interpreting a possible signal.
A consistent calculation must use the electroweak-invariant portal. The hypercharge and weak dipoles involve a Higgs doublet before symmetry breaking; their photon, Z Z Z and W W W interactions are consequently related. The weak field strength also generates vertices with two gauge bosons. These contacts are fixed by the same operator as the three-point dipole, and enter the virtual corrections together with the familiar triangle diagrams. This makes precision tests sensitive to the gauge completion of the portal, rather than only to a photon transition moment.
The effective-theory literature provides several ingredients for this analysis. Reference [13 ] studied the running of Higgs–neutrino operators and the consequences of large dipoles. Gauge and Yukawa anomalous dimensions have been developed further [14 , 15 ] , including numerical evolution in wilson [16 ] .
The FCC-ee study of ref. [17 ] considers the Standard Model effective field theory extended with gauge-singlet right-handed neutrinos (ν S M E F T \nu\mathrm{SMEFT} ν SMEFT ). It explores a broad set of production and decay interactions, with dipoles outside the operator set considered. Our work complements that study by determining the finite electroweak corrections generated by two dipole insertions.
This is a different perturbative order from single-insertion running: it involves both dimension-six thresholds proportional to the heavy mass squared and dimension-eight terms. Work on dimension-eight bases and universal electroweak corrections [18 , 19 , 20 ] and on double dipole insertions in the low-energy effective field theory below the electroweak scale (LEFT) [21 ] supplies the relevant methodological context.
The ultraviolet interpretation is equally important. Explicit loop completions [3 , 6 , 22 ] generally generate interactions besides the dipoles. Reference [23 ] gives a renormalizable-model example and its one-loop matching onto the sterile-neutrino EFT. A symmetry that suppresses neutrino masses or active–sterile mixing [24 ] can still allow operators that modify electroweak gauge propagation and lepton currents. The size of these additional operators determines whether precision data primarily probe the dipoles or a combination of interactions in the completion. Studies combining mixing and dipoles [25 ] illustrate the importance of this distinction. We therefore examine how the interpretation changes when independent ultraviolet contributions are allowed, rather than assuming from the outset that every precision limit applies to every completion.
We take the HNLs to be Majorana fermions. To isolate the flavour-conserving electroweak response, the numerical analysis uses three mass-degenerate states with real, universal, flavour-aligned dipoles and negligible active–sterile mixing. It permits a study of the diagonal response without a simultaneous fit to flavour-changing transitions or a specified model of neutrino masses. Nonaligned flavour structures and their charged-lepton flavour-violation constraints are left to dedicated studies. The radiative stability of the mass and mixing assumptions is discussed explicitly, and the relation of the retained amplitudes to a Dirac HNL is given in § 2.1 .
Our calculation relates the gauge-invariant portal to a correlated pattern of widths and asymmetries. At momenta below the HNL mass, the virtual amplitudes admit a local expansion in interactions among SM fields. Matching expresses these interactions in a common operator basis, connecting the corrections to gauge propagation and lepton couplings with the observables and with possible contributions from heavier particles. The resulting local response spans four physical combinations. Independent ultraviolet coefficients can reproduce all four, so the pole data constrain the dipoles only once these additional contributions are restricted. We quantify the FCC-ee reach under a common dipole-dominance boundary, its sensitivity to correlated ultraviolet effects, and the additional information from radiative Higgs decays. The comparison with the HL-LHC projections of ref. [1 ] illustrates the complementarity of virtual and direct probes within matched flavour and operator conventions.
§ 2 defines the model and conventions. § 3 –§ 5 develop the loop calculation, matching and oblique description. § 6 and § 7 connect these results to pole observables and radiative Higgs decays, respectively. They are followed by the FCC-ee results and their ultraviolet interpretation in § 8 ; the direct-search comparison is presented at its end. The appendices give the technical checks and further comparisons with related operator sectors. Table 7 summarizes the scope of the results.
Table 7. Scope of the results. The assumptions defining the conditional forecasts are specified in § 6 and § 8 .
Gauge-invariant HNL dipole portal
We consider a sterile Majorana fermion N N N of mass M N M_N M N and neglect active–sterile mixing. Its interactions with SM particles are mediated by the dipole portal. Above electroweak symmetry breaking, the Higgs doublet is needed to combine the active lepton and sterile state into a gauge-invariant operator. For one SM lepton flavour, the two independent dimension-six interactions are
O B = g ′ L ‾ σ μ ν N H ~ B μ ν , O W = g L ‾ σ μ ν N τ I H W μ ν I , (1) \mathcal{O}_B = g' \overline{L} \sigma^{\mu\nu} N \widetilde{H} B_{\mu\nu}, \qquad\mathcal{O}_W = g \overline{L} \sigma^{\mu\nu} N \tau^I H W^I_{\mu\nu}, \tag*{(1)} O B = g ′ L σ μν N H B μν , O W = g L σ μν N τ I H W μν I , (1)
with
L d i p = C B O B + C W O W + h . c . , τ I = σ I 2 . (2) \mathcal{L}_{\mathrm{dip}} = C_B \mathcal{O}_B + C_W \mathcal{O}_W + \mathrm{h.c.}, \qquad\tau^I = \frac{\sigma^I}{2}. \tag*{(2)} L dip = C B O B + C W O W + h.c. , τ I = 2 σ I . (2)
Here L L L is the SM left-handed lepton doublet, H H H is the Higgs doublet, and H ~ = i σ 2 H ∗ \widetilde{H} = i\sigma^2 H^* H = i σ 2 H ∗ its conjugate. The couplings g ′ g' g ′ and g g g belong to hypercharge and weak, respectively; B μ ν B_{\mu\nu} B μν and W μ ν I W^I_{\mu\nu} W μν I are their field strengths. The matrices σ I \sigma^I σ I act on weak isospin, whereas σ μ ν \sigma^{\mu\nu} σ μν acts on spinor indices. The coefficients C B C_B C B and C W C_W C W have mass dimension − 2 -2 − 2 .
After electroweak symmetry breaking, d γ d_\gamma d γ and d Z d_Z d Z multiply the dimension-five interactions ν ˉ L σ μ ν N R A μ ν \bar{\nu}_L\sigma^{\mu\nu}N_R A_{\mu\nu} ν ˉ L σ μν N R A μν and ν ˉ L σ μ ν N R Z μ ν \bar{\nu}_L\sigma^{\mu\nu}N_R Z_{\mu\nu} ν ˉ L σ μν N R Z μν , respectively, with their Hermitian conjugates understood. Here A μ ν = ∂ μ A ν − ∂ ν A μ A_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu A μν = ∂ μ A ν − ∂ ν A μ and Z μ ν = ∂ μ Z ν − ∂ ν Z μ Z_{\mu\nu}=\partial_\mu Z_\nu-\partial_\nu Z_\mu Z μν = ∂ μ Z ν − ∂ ν Z μ ; the two-gauge interactions are displayed below. With
A μ = s W W μ 3 + c W B μ , Z μ = c W W μ 3 − s W B μ , (3) A_\mu=s_W W^3_\mu+c_W B_\mu,\qquad Z_\mu=c_W W^3_\mu-s_W B_\mu, \tag*{(3)} A μ = s W W μ 3 + c W B μ , Z μ = c W W μ 3 − s W B μ , (3)
we define the corresponding gauge-basis combinations
d B ≡ c W d γ − s W d Z = g ′ v 2 C B , (4) d_B\equiv c_Wd_\gamma-s_Wd_Z=\frac{g'v}{\sqrt{2}}C_B, \tag*{(4)} d B ≡ c W d γ − s W d Z = 2 g ′ v C B , (4)
d 3 ≡ s W d γ + c W d Z = g v 2 2 C W . (5) d_3\equiv s_Wd_\gamma+c_Wd_Z=\frac{gv}{2\sqrt{2}}C_W. \tag*{(5)} d 3 ≡ s W d γ + c W d Z = 2 2 g v C W . (5)
The charged-current dipole satisfies
d W = 2 d 3 = g v 2 C W = 2 ( c W d Z + s W d γ ) . (6) d_W=\sqrt{2}\,d_3=\frac{g v}{2}C_W=\sqrt{2}\left(c_Wd_Z+s_Wd_\gamma\right). \tag*{(6)} d W = 2 d 3 = 2 g v C W = 2 ( c W d Z + s W d γ ) . (6)
Equations (4) –(6) will be useful for displaying the constraints either in the gauge-invariant ( C B , C W ) (C_B,C_W) ( C B , C W ) basis or in the broken-phase ( d γ , d Z ) (d_\gamma,d_Z) ( d γ , d Z ) basis.
Majorana benchmark and the Dirac limit
The sterile field in eq. (2) is Majorana, N = N c N=N^c N = N c , with L N = 1 2 N ˉ ( i ∂̸ − M N ) N \mathcal{L}_N=\frac{1}{2}\bar{N}(i\not{\partial}-M_N)N L N = 2 1 N ˉ ( i ∂ − M N ) N . Only its right-handed component enters the portal. The factors associated with Majorana contractions are included in the amplitudes and must not be appended to the final Wilson coefficients.
The relation to a Dirac HNL follows from the chiral structure of the propagator connecting two dipole vertices. Consider a Dirac state N D N_D N D with the same mass and dipole coefficients, coupled only through N D , R N_{D,R} N D , R , with its other chiral component carrying no additional interactions. Assigning it lepton number one makes the dipoles lepton-number conserving, as in ref. [2 ] . In both theories the heavy contraction between a portal and its Hermitian conjugate is
P R i ( k̸ + M N ) k 2 − M N 2 + i 0 P L = i k̸ P L k 2 − M N 2 + i 0 . (7) P_R\frac{i(\not{k}+M_N)}{k^2-M_N^2+i0}P_L=\frac{i\not{k}P_L}{k^2-M_N^2+i0}. \tag*{(7)} P R k 2 − M N 2 + i 0 i ( k + M N ) P L = k 2 − M N 2 + i 0 i k P L . (7)
The additional Majorana contraction carries lepton number violation. It does not contribute to the external lepton-number-conserving currents or the bosonic bubbles considered here in the limit of vanishing light-neutrino masses and mixing.
Equivalently, write N D = ( N 1 + i N 2 ) / 2 N_D=(N_1+iN_2)/\sqrt{2} N D = ( N 1 + i N 2 ) / 2 in terms of two degenerate Majorana fields. Their dipole coefficients are C X a = C X ( 1 , i ) a / 2 C_{X a}=C_X(1,i)_a/\sqrt{2} C X a = C X ( 1 , i ) a / 2 , and hence
∑ a = 1 2 C X a C Y a ∗ = C X C Y ∗ , ∑ a = 1 2 C X a C Y a = 0 , X , Y ∈ { B , W } . (8) \sum_{a=1}^{2}C_{X a}C_{Y a}^{*}=C_XC_Y^{*},\qquad\sum_{a=1}^{2}C_{X a}C_{Y a}=0,\qquad X,Y\in\{B,W\}. \tag*{(8)} a = 1 ∑ 2 C X a C Y a ∗ = C X C Y ∗ , a = 1 ∑ 2 C X a C Y a = 0 , X , Y ∈ { B , W } . (8)
The first identity preserves the lepton-number-conserving response; the second cancels the lepton-number-violating one. The normalization of the two Majorana couplings compensates their multiplicity, preserving the strength of the original Dirac interaction. At equal physical dipole normalization and flavour bilinears, the pole amplitudes and the two-dipole radiative Higgs bubbles retained in this work therefore have the same Dirac and Majorana values. For example, three aligned Dirac states give the same conditional pole contours as three aligned Majorana states with the same C B , C W C_B,C_W C B , C W . This statement applies to these lepton-number-conserving amplitudes at quadratic dipole order.
Lepton-number-violating observables do distinguish the two cases. In particular, the Majorana mass naturalness estimate relies on lepton-number violation, and the HL-LHC projections used here exploit same-sign dileptons. Their interpretation is specific to the Majorana scenario, even though the precision amplitudes discussed above also apply to a lepton-number-conserving Dirac HNL with the matched couplings. Conservation of lepton number protects against the Majorana mass term, but does not by itself forbid Dirac Yukawa mixing with the active neutrinos.
Gauge completion and contact interactions
With Y L = − 1 / 2 Y_L=-1/2 Y L = − 1/2 , Y H ~ = − 1 / 2 Y_{\widetilde{H}}=-1/2 Y H = − 1/2 and Y N = 0 Y_N=0 Y N = 0 , both operators in eq. (1) are hypercharge neutral. The contraction L ‾ H ~ \overline{L}\widetilde{H} L H is an S U ( 2 ) L SU(2)_L S U ( 2 ) L singlet, whereas L ‾ τ I H ~ \overline{L}\tau^I\widetilde{H} L τ I H transforms as an adjoint and contracts with W μ ν I W^I_{\mu\nu} W μν I . Thus the unbroken action is gauge invariant before a gauge or a Higgs background is chosen [1 ] .
For D μ = ∂ μ + i g τ I W μ I + i g ′ Y B μ D_\mu=\partial_\mu+ig\tau^I W^I_\mu+ig'YB_\mu D μ = ∂ μ + i g τ I W μ I + i g ′ Y B μ , define the field strength by [ D μ , D ν ] = i g τ I W μ ν I + i g ′ Y B μ ν [D_\mu,D_\nu]=ig\tau^I W^I_{\mu\nu}+ig'YB_{\mu\nu} [ D μ , D ν ] = i g τ I W μν I + i g ′ Y B μν . Its non-Abelian part is − g ϵ I J K W μ J W ν K -g\epsilon^{IJK}W^J_\mu W^K_\nu − g ϵ I J K W μ J W ν K . Consequently O W \mathcal{O}_W O W necessarily contains N ℓ W γ N\ell W\gamma N ℓ W γ , N ℓ W Z N\ell WZ N ℓ W Z and N ν W + W − N\nu W^+W^- N ν W + W − contact interactions; their coefficients are fixed by the same C W C_W C W as the three-point dipoles. To display them explicitly, let N R = P R N N_R=P_RN N R = P R N , σ μ ν = i [ γ μ , γ ν ] / 2 \sigma^{\mu\nu}=i[\gamma^\mu,\gamma^\nu]/2 σ μν = i [ γ μ , γ ν ] /2 , W ± = ( W 1 ∓ i W 2 ) / 2 W^\pm=(W^1\mp iW^2)/\sqrt{2} W ± = ( W 1 ∓ i W 2 ) / 2 , and use the derivative field strengths defined above. Before choosing a scalar background, the complete two-gauge part is
L 2 V = − g 2 C W ϵ I J K ( L ‾ a σ μ ν N R ) ( τ I ) a b H ~ b W μ J W ν K + h . c . (9) \mathcal{L}_{2V}=-g^2C_W\epsilon^{IJK}(\overline{L}_a\sigma^{\mu\nu}N_R)(\tau^I)_{ab}\widetilde{H}_bW^J_\mu W^K_\nu+\mathrm{h.c.} \tag*{(9)} L 2 V = − g 2 C W ϵ I J K ( L a σ μν N R ) ( τ I ) ab H b W μ J W ν K + h.c. (9)
In the background H ~ = ( v / 2 , 0 ) T \widetilde{H}=(v/\sqrt{2},0)^T H = ( v / 2 , 0 ) T , the interactions split as
L 1 V = ν ‾ L σ μ ν N R ( d γ A μ ν + d Z Z μ ν ) + d W ℓ ‾ L σ μ ν N R ( ∂ μ W ν − − ∂ ν W μ − ) + h . c . , (10) \mathcal{L}_{1V}=\overline{\nu}_L\sigma^{\mu\nu}N_R(d_\gamma A_{\mu\nu}+d_ZZ_{\mu\nu})+d_W\overline{\ell}_L\sigma^{\mu\nu}N_R(\partial_\mu W^-_\nu-\partial_\nu W^-_\mu)+\mathrm{h.c.}, \tag*{(10)} L 1 V = ν L σ μν N R ( d γ A μν + d Z Z μν ) + d W ℓ L σ μν N R ( ∂ μ W ν − − ∂ ν W μ − ) + h.c. , (10)
L 2 V ( v ) = 2 i g d W ℓ ‾ L σ μ ν N R W μ − ( s W A ν + c W Z ν ) + 2 i g d W ν ‾ L σ μ ν N R W μ + W ν − + h . c . (11) \mathcal{L}^{(v)}_{2V}=2igd_W\overline{\ell}_L\sigma^{\mu\nu}N_RW^-_\mu(s_WA_\nu+c_WZ_\nu)
+\sqrt{2}igd_W\overline{\nu}_L\sigma^{\mu\nu}N_RW^+_\mu W^-_\nu+\mathrm{h.c.} \tag*{(11)} L 2 V ( v ) = 2 i g d W ℓ L σ μν N R W μ − ( s W A ν + c W Z ν ) + 2 i g d W ν L σ μν N R W μ + W ν − + h.c. (11)
The signs in eq. (11) follow from the declared D μ = ∂ μ + i g τ ⋅ W μ + i g ′ Y B μ D_\mu=\partial_\mu+ig\tau\cdot W_\mu+ig'YB_\mu D μ = ∂ μ + i g τ ⋅ W μ + i g ′ Y B μ convention. For example,
W μ ν 3 = ( ∂ μ W ν 3 − ∂ ν W μ 3 ) + i g ( W μ + W ν − − W μ − W ν + ) , (12) W^3_{\mu\nu}=(\partial_\mu W^3_\nu-\partial_\nu W^3_\mu)+ig(W^+_\mu W^-_\nu-W^-_\mu W^+_\nu), \tag*{(12)} W μν 3 = ( ∂ μ W ν 3 − ∂ ν W μ 3 ) + i g ( W μ + W ν − − W μ − W ν + ) , (12)
W μ ν − = ( ∂ μ W ν − − ∂ ν W μ − ) + i g ( W μ − W ν 3 − W μ 3 W ν − ) . (13) W^-_{\mu\nu}=(\partial_\mu W^-_\nu-\partial_\nu W^-_\mu)+ig(W^-_\mu W^3_\nu-W^3_\mu W^-_\nu). \tag*{(13)} W μν − = ( ∂ μ W ν − − ∂ ν W μ − ) + i g ( W μ − W ν 3 − W μ 3 W ν − ) . (13)
Contracting with the antisymmetric σ μ ν \sigma^{\mu\nu} σ μν gives the factor of two. Changing the covariant-derivative sign requires changing the field-strength and SM gauge-vertex conventions consistently. There are no independent C B C_B C B two-gauge contacts: the hypercharge field strength is Abelian. In particular the W γ W\gamma W γ and W Z WZ W Z contact coefficients are fixed to 2 i e d W 2ied_W 2 i e d W and 2 i g c W d W 2igc_Wd_W 2 i g c W d W in this convention.
These vertices are part of the dimension-six operator, despite their vev-dependent dimension-five broken-phase form. They are not optional higher-order corrections to the single-boson dipole. The necessity of their inclusion in direct multi-boson production was demonstrated in ref. [1 ] ; the same completion supplies the charged contact diagrams in the present lepton-current calculation. Omitting them changes the theory and spoils the gauge relations between amplitudes. An electromagnetic dipole-only description can be appropriate below the weak scale, but cannot replace this completion for electroweak loops. The contacts are also included in the collider study of ref. [10 ] , as noted in ref. [1 ] .
The scalar interactions are obtained by restoring the Higgs and Goldstone fields around the vacuum,
H = ( ϕ + ( v + h + i ϕ 0 ) / 2 ) , H ~ = ( ( v + h − i ϕ 0 ) / 2 − ϕ − ) . (14) H=\begin{pmatrix}\phi^+\\(v+h+i\phi^0)/\sqrt{2}\end{pmatrix},\qquad
\widetilde{H}=\begin{pmatrix}(v+h-i\phi^0)/\sqrt{2}\\-\phi^-\end{pmatrix}. \tag*{(14)} H = ( ϕ + ( v + h + i ϕ 0 ) / 2 ) , H = ( ( v + h − i ϕ 0 ) / 2 − ϕ − ) . (14)
Substitution into the two operators fixes all Higgs and Goldstone vertices, including the Higgs/Goldstone partners of the contact interactions. The neutral-scalar partners of eq. (11) follow by v → v + h − i ϕ 0 v \to v+h-i\phi^{0} v → v + h − i ϕ 0 in its non-Hermitian part, with the conjugate replacement in its Hermitian conjugate. The charged-Goldstone partners are fixed by the second component − ϕ − -\phi^{-} − ϕ − in eq. (9) ; retaining that compact expression specifies them without any additional Wilson coefficient. The v v v -dependent vertices therefore cannot be varied independently:
d γ = v 2 ( g ′ c W C B + 1 2 g s W C W ) , d Z = v 2 ( − g ′ s W C B + 1 2 g c W C W ) (15) d_{\gamma}=\frac{v}{\sqrt{2}}\left(g'c_{W}C_{B}+\frac{1}{2}gs_{W}C_{W}\right), \qquad d_{Z}=\frac{v}{\sqrt{2}}\left(-g's_{W}C_{B}+\frac{1}{2}gc_{W}C_{W}\right) \tag*{(15)} d γ = 2 v ( g ′ c W C B + 2 1 g s W C W ) , d Z = 2 v ( − g ′ s W C B + 2 1 g c W C W ) (15)
C B = 2 g ′ v ( c W d γ − s W d Z ) , C W = 2 2 g v ( s W d γ + c W d Z ) (16) C_{B}=\frac{\sqrt{2}}{g'v}\left(c_{W}d_{\gamma}-s_{W}d_{Z}\right), \qquad C_{W}=\frac{2\sqrt{2}}{gv}\left(s_{W}d_{\gamma}+c_{W}d_{Z}\right) \tag*{(16)} C B = g ′ v 2 ( c W d γ − s W d Z ) , C W = g v 2 2 ( s W d γ + c W d Z ) (16)
These relations reduce the broken-phase couplings to two independent portal coefficients. The same operator structure also simplifies the loop amplitudes. In particular, the antisymmetry of the dipole gives q μ σ μ ν q ν = 0 q_{\mu}\sigma^{\mu\nu}q_{\nu}=0 q μ σ μν q ν = 0 at a three-point vertex. Because the longitudinal part of a gauge propagator is proportional to its momentum at each endpoint, this identity removes that part whenever one endpoint is a single-gauge dipole. It therefore simplifies the direct-current diagrams considered here, as demonstrated in appendix § 12 .§ 7 . Higher-point Slavnov–Taylor identities also involve contact and Goldstone interactions; they cannot be inferred from the transverse three-point vertex alone.
For several lepton flavours the quadratic neutral self-energies generalize as
∣ d B ∣ 2 → ∑ α ∣ d B α ∣ 2 , ∣ d 3 ∣ 2 → ∑ α ∣ d 3 α ∣ 2 , d B d 3 → ∑ α Re ( d B α d 3 α ∗ ) (17) \lvert d_{B}\rvert^{2}\to\sum_{\alpha}\lvert d_{B}^{\alpha}\rvert^{2}, \qquad\lvert d_{3}\rvert^{2}\to\sum_{\alpha}\lvert d_{3}^{\alpha}\rvert^{2}, \qquad d_{B}d_{3}\to\sum_{\alpha}\operatorname{Re}\left(d_{B}^{\alpha}d_{3}^{\alpha*}\right) \tag*{(17)} ∣ d B ∣ 2 → α ∑ ∣ d B α ∣ 2 , ∣ d 3 ∣ 2 → α ∑ ∣ d 3 α ∣ 2 , d B d 3 → α ∑ Re ( d B α d 3 α ∗ ) (17)
Unless stated otherwise, below we display the one-flavour, real-coupling result. Flavour multiplicities are inserted explicitly when constructing the universal benchmark.
One-loop topology. The implemented gauge-invariant action is Hermitian, and its three-point broken-phase couplings reproduce eqs. (4) –(6) . At the one-loop order considered here, adding a propagating scalar between two dipole vertices that already contain two internal fermion lines would increase the loop number to two. Higgs and Goldstone vertices therefore add no such one-loop vacuum-polarization graph. Their role depends on the external states: the Higgs partners are essential for the radiative decays studied in § 7 .
Calculation strategy and model implementation
The gauge-invariant dipole Lagrangian is implemented in FeynRules [26 ] , with explicit electroweak index expansion before exporting the interaction vertices. The model, called dipole_GI_FA, supplies generic and classes files with the vertices for diagram generation with FeynArts [27 ] . The amplitudes are converted to FeynCalc [28 ] for Dirac and Lorentz algebra, external-state projection and Passarino–Veltman reduction. Scalar-integral expansions are evaluated analytically, with Package-X [29 ] used for the scalar checks. Dimensional dependence is retained until pole subtraction and finite-part extraction; the renormalization prescription is specified in appendix § 12 .§ 8 .
Several complementary checks test the normalization and symmetry relations that enter the observable predictions. These include Born-level normalization of the physical projectors, neutral and charged two-point transversality, gauge-basis rotations, Majorana orientations, and reconstruction of the direct currents with external residues. The contact diagrams are included throughout, so these checks apply to the electroweak-invariant portal. We also test selected tree-level matrix elements using the UFO format [30 ] and MadGraph5_aMC@NLO [31 ] ; these tests check the vertex normalizations and coupling dependence. The likelihood and figures are evaluated from the resulting analytic responses using Python numerical routines. The direct-search comparison uses the projections of ref. [1 ] .
We use Matchete [32 ] to check an example of the additional lepton currents that can accompany a UV completion. The example contains a heavy charged vector-like singlet E E E coupled to the SM lepton and Higgs doublets. Integrating it out at tree level yields correlated singlet and triplet currents. Field redefinitions simplify the effective action through dimension six, and the Pauli-matrix identity expresses its current operators in the Warsaw basis. This reproduces the E E E direction in eq. (F.12), providing an independent check of the tree-level current benchmark, distinct from the dipole-loop thresholds.
Model files and reuse for direct searches. The FeynRules implementation and its dipole_GI_FA and dipole_GI_UFO exports describe the same gauge-invariant portal Lagrangian used to relate the indirect effects to the direct-search processes of ref. [1 ] . The Higgs, Goldstone and non-Abelian interactions follow from the gauge-invariant operators, rather than being added as independent dipoles after symmetry breaking. The UFO can therefore also be used for direct-production studies with matched flavour assignments, parameter cards and normalization conventions. The FeynRules source, FeynArts models and UFO implementation used in the associated studies are publicly available, together with usage instructions and reproducible validation scripts, in the public model repository . Version 1.0.0 is archived on Zenodo at doi:10.5281/zenodo.23667622 .
One-loop electroweak vacuum polarizations
A gauge boson can fluctuate into an HNL and a light lepton and then return to a gauge boson, as shown in Figure 1 . The resulting self-energy, or vacuum polarization, changes its propagation and can mix the neutral gauge fields. We derive the common loop function governing these fluctuations and separate its constant and momentum-dependent terms. This organization exposes the correlations among neutral and charged channels and provides the bosonic input for the effective-theory matching in the next section.
Figure 1. Two-dipole vacuum polarizations. The neutral graph has V , V ′ ∈ { γ , Z } V,V' \in\{\gamma,Z\} V , V ′ ∈ { γ , Z } and includes the two Majorana orientations; the charged graph has the normalization in eq. (26) . Squares indicate dipole insertions. The relative group and orientation factors give F W W = d W 2 F 0 / 2 = d 3 2 F 0 F_{WW}=d_W^2F_0/2=d_3^2F_0 F W W = d W 2 F 0 /2 = d 3 2 F 0 for massless charged leptons.
Conventions and transverse decomposition
We define the transverse gauge-boson self-energy by
i Π V V ′ μ ν ( q ) = i ( q μ q ν − q 2 g μ ν ) F V V ′ ( q 2 ) , (18) i\Pi_{VV'}^{\mu\nu}(q) = i\left(q^\mu q^\nu- q^2 g^{\mu\nu}\right)F_{VV'}(q^2), \tag*{(18)} i Π V V ′ μν ( q ) = i ( q μ q ν − q 2 g μν ) F V V ′ ( q 2 ) , (18)
For local bosonic terms the coefficient is that of the quadratic action insertion; see appendix § 11 .§ 5 . We define the corresponding dimension-two function by
Π V V ′ ( q 2 ) = q 2 F V V ′ ( q 2 ) . (19) \Pi_{VV'}(q^2) = q^2 F_{VV'}(q^2). \tag*{(19)} Π V V ′ ( q 2 ) = q 2 F V V ′ ( q 2 ) . (19)
The explicit one-loop calculation is performed in dimensional regularization, D = 4 − 2 ϵ D = 4 - 2\epsilon D = 4 − 2 ϵ . Details of the tensor projection, Passarino–Veltman reduction and Majorana orientations are given in appendix § 11 .§ 2 .
The three-point dipole vertex contains σ μ ν q ν \sigma^{\mu\nu}q_\nu σ μν q ν . Consequently q μ σ μ ν q ν = 0 q_\mu\sigma^{\mu\nu}q_\nu= 0 q μ σ μν q ν = 0 by antisymmetry, so the two-dipole bubble is transverse at each external gauge leg. This fixes its tensor structure but neither its finite constant nor the numerical UV residue.
Neutral sector and universal loop function
The dipole–dipole contributions to the neutral self-energies are controlled by a single scalar function. In the ( A , Z ) (A,Z) ( A , Z ) basis we find
F A A = d γ 2 F 0 , F A Z = d γ d Z F 0 , F Z Z = d Z 2 F 0 , (20) F_{AA} = d_\gamma^2 F_0,\qquad F_{AZ} = d_\gamma d_Z F_0,\qquad F_{ZZ} = d_Z^2 F_0, \tag*{(20)} F AA = d γ 2 F 0 , F A Z = d γ d Z F 0 , F Z Z = d Z 2 F 0 , (20)
where each external gauge field couples through a dipole vertex. The ordinary SM Z ν ν Z\nu\nu Z ν ν vertex instead enters the direct-current triangles discussed below. The exact dipole–dipole loop function is
F 0 ( q 2 ) = 1 2 π 2 ( D − 1 ) q 2 { A 0 ( M N 2 ) [ ( D − 2 ) M N 2 − q 2 ] − ( M N 2 − q 2 ) [ ( D − 2 ) M N 2 + q 2 ] B 0 ( q 2 , 0 , M N 2 ) } . (21) F_0(q^2)=\frac{1}{2\pi^2(D-1)q^2}\left\{A_0(M_N^2)\left[(D-2)M_N^2-q^2\right]-(M_N^2-q^2)\left[(D-2)M_N^2+q^2\right]B_0(q^2,0,M_N^2)\right\}. \tag*{(21)} F 0 ( q 2 ) = 2 π 2 ( D − 1 ) q 2 1 { A 0 ( M N 2 ) [ ( D − 2 ) M N 2 − q 2 ] − ( M N 2 − q 2 ) [ ( D − 2 ) M N 2 + q 2 ] B 0 ( q 2 , 0 , M N 2 ) } . (21)
A nontrivial feature of the q 2 → 0 q^2\to0 q 2 → 0 expansion is that the finite constant term must be obtained before setting D = 4 D=4 D = 4 : a contribution of the form ( D − 4 ) × 1 / ϵ (D-4)\times1/\epsilon ( D − 4 ) × 1/ ϵ survives. The result is
F 0 ( 0 ) = − M N 2 2 π 2 , (22) F_0(0)=-\frac{M_N^2}{2\pi^2}, \tag*{(22)} F 0 ( 0 ) = − 2 π 2 M N 2 , (22)
and
F 0 ′ ( 0 ) = 1 6 π 2 [ 1 ϵ + ln μ P X 2 M N 2 + 11 6 − γ E − ln π ] , (23) F'_0(0)=\frac{1}{6\pi^2}\left[\frac{1}{\epsilon}+\ln\frac{\mu_{\mathrm{PX}}^2}{M_N^2}+\frac{11}{6}-\gamma_E-\ln\pi\right], \tag*{(23)} F 0 ′ ( 0 ) = 6 π 2 1 [ ϵ 1 + ln M N 2 μ PX 2 + 6 11 − γ E − ln π ] , (23)
in the dimensional-regularization normalization of the displayed Package-X scalar integrals, whose scale is denoted by μ P X \mu_{\mathrm{PX}} μ PX . The common M S ‾ \overline{\mathrm{MS}} MS scale used below is μ = μ P X / ( 2 π ) \mu=\mu_{\mathrm{PX}}/(2\pi) μ = μ PX / ( 2 π ) . The measure-dependent constant accompanying the pole is absorbed when converting to the common M S ‾ \overline{\mathrm{MS}} MS convention; the numerator contribution to the finite 11/6 is preserved. Appendix § 11 .§ 6 specifies this conversion. The renormalized slope and all Wilson coefficients below use the same subtraction convention.
Rotating eq. (20) to the ( B , W 3 ) (B,W^3) ( B , W 3 ) basis gives
F B B = d B 2 F 0 , F 33 = d 3 2 F 0 , F 3 B = d 3 d B F 0 . (24) F_{BB}=d_B^2F_0,\qquad F_{33}=d_3^2F_0,\qquad F_{3B}=d_3d_BF_0. \tag*{(24)} F B B = d B 2 F 0 , F 33 = d 3 2 F 0 , F 3 B = d 3 d B F 0 . (24)
Thus the neutral self-energy matrix factorizes:
F ( q 2 ) = F 0 ( q 2 ) ( d B 2 d B d 3 d B d 3 d 3 2 ) = F 0 ( q 2 ) ( d B d 3 ) ( d B d 3 ) . (25) \mathbf{F}(q^2)=F_0(q^2)\begin{pmatrix}d_B^2 & d_Bd_3 \\ d_Bd_3 & d_3^2\end{pmatrix}=F_0(q^2)\begin{pmatrix}d_B\\d_3\end{pmatrix}\begin{pmatrix}d_B&d_3\end{pmatrix}. \tag*{(25)} F ( q 2 ) = F 0 ( q 2 ) ( d B 2 d B d 3 d B d 3 d 3 2 ) = F 0 ( q 2 ) ( d B d 3 ) ( d B d 3 ) . (25)
It therefore has rank at most one (rank one for a nonzero matrix) at any q 2 q^2 q 2 within this contribution. Factorization uses a common loop kernel and one real coupling vector, not gauge invariance alone. A sum over nonaligned flavour vectors, complex relative phases in the CP-even matrix, or independent local matching terms can spoil the rank-one equality.
Charged sector
For one charged-lepton flavour and neglecting the charged-lepton mass, the charged-current result is
F W W ( q 2 ) = d W 2 F s i n g l e ( q 2 ) = d W 2 2 F 0 ( q 2 ) = d 3 2 F 0 ( q 2 ) . (26) F_{WW}(q^2)=d_W^2F_{\mathrm{single}}(q^2)=\frac{d_W^2}{2}F_0(q^2)=d_3^2F_0(q^2). \tag*{(26)} F W W ( q 2 ) = d W 2 F single ( q 2 ) = 2 d W 2 F 0 ( q 2 ) = d 3 2 F 0 ( q 2 ) . (26)
Consequently,
F W W ( q 2 ) = F 33 ( q 2 ) ( m ℓ = 0 ) . (27) F_{WW}(q^2)=F_{33}(q^2)\qquad(m_\ell=0). \tag*{(27)} F W W ( q 2 ) = F 33 ( q 2 ) ( m ℓ = 0 ) . (27)
This equality provides a useful check of the S U ( 2 ) L SU(2)_{L} S U ( 2 ) L normalization, in particular of the relation d W = 2 d 3 d_{W}=\sqrt{2}\,d_{3} d W = 2 d 3 .
The gauge-invariant dipole Lagrangian also contains non-Abelian four-point interactions such as N ℓ W γ N\ell W\gamma N ℓ W γ , N ℓ W Z N\ell WZ N ℓ W Z and N ν W W N\nu WW N ν W W . An explicit diagram-by-diagram inspection shows that no additional one-loop two-point topology involving these contact interactions contributes to the W W WW W W self-energy in the absence of active–sterile mixing. Such interactions will, however, become relevant for higher-point amplitudes, including the Zee contact diagrams of appendix § 12 .§ 1 and e + e − → W + W − e^{+}e^{-}\to W^{+}W^{-} e + e − → W + W − .
EFT interpretation and matching
At momenta well below M N M_{N} M N , virtual HNL propagation can be represented by local interactions among SM fields. Matching determines their coefficients by requiring this low-energy description to reproduce the expanded amplitudes of the theory containing N N N . The difference between the coefficients above and below the HNL threshold is the threshold correction calculated here.
There are therefore two effective descriptions in this study. The starting dipole EFT contains the HNL and represents still heavier particles through C B C_{B} C B , C W C_{W} C W and other possible operators. After removing the HNL as a propagating field, the low-energy EFT contains SM fields only. The dipole calculation fixes an additive contribution to its coefficients; the heavier particles can supply independent contributions to the same operators. Keeping these two sources distinct is essential for interpreting the precision reach. We begin with gauge propagation and then consider the direct lepton couplings.
The low-momentum expansion of eq. (24) separates two qualitatively different EFT structures:
F i j ( q 2 ) = d i d j [ M N 2 2 π 2 + q 2 F 0 ′ ( 0 ) + O ( q 4 ) ] , i , j = B , 3. (28) F_{ij}(q^{2})=d_{i}d_{j}\left[\frac{M_{N}^{2}}{2\pi^{2}}+q^{2}F'_{0}(0)+\mathcal{O}(q^{4})\right],\qquad i,j=B,3. \tag*{(28)} F ij ( q 2 ) = d i d j [ 2 π 2 M N 2 + q 2 F 0 ′ ( 0 ) + O ( q 4 ) ] , i , j = B , 3. (28)
Since Π i j = q 2 F i j \Pi_{ij}=q^{2}F_{ij} Π ij = q 2 F ij , the constant in F i j F_{ij} F ij corrects the ordinary gauge kinetic terms, while its slope adds two powers of momentum. These contributions correspond to dimension-six and dimension-eight operators, respectively. Both arise from two insertions of dimension-six dipoles: the factor M N 2 M_{N}^{2} M N 2 supplies the two mass powers needed by the dimension-six coefficient. Thus the dimension of the induced operator and the number of dipole insertions describe different aspects of the expansion. The first term is finite; the slope contains a UV divergence that requires a local counterterm.
Finite dimension-six threshold
The constant term in the transverse form factor modifies the gauge kinetic terms in a Higgs background. It is therefore represented by operators containing two Higgs fields and two field strengths. We use Δ L = + ∑ i C i Q i \Delta\mathcal{L}=+\sum_{i}C_{i}Q_{i} Δ L = + ∑ i C i Q i and the standard Warsaw normalization
Q H B = ( H † H ) B μ ν B μ ν , Q_{HB}=(H^{\dagger}H)B_{\mu\nu}B^{\mu\nu}, Q H B = ( H † H ) B μν B μν ,
Q H W = ( H † H ) W μ ν I W I μ ν , Q_{HW}=(H^{\dagger}H)W^{I}_{\mu\nu}W^{I\mu\nu}, Q H W = ( H † H ) W μν I W I μν ,
Q H W B = ( H † σ I H ) W μ ν I B μ ν . (29) Q_{HWB}=(H^{\dagger}\sigma^{I}H)W^{I}_{\mu\nu}B^{\mu\nu}. \tag*{(29)} Q H W B = ( H † σ I H ) W μν I B μν . (29)
The Warsaw normalization uses a Pauli matrix in Q H W B Q_{HWB} Q H W B , giving twice the isospin contraction of the generator in the dipole operator. With H = ( 0 , v / 2 ) T H=(0,v/\sqrt{2})^{T} H = ( 0 , v / 2 ) T , H † σ 3 H = − v 2 / 2 H^{\dagger}\sigma^{3}H=-v^{2}/2 H † σ 3 H = − v 2 /2 . We identify the displayed transverse coefficient with the quadratic effective-action insertion, as detailed in appendix § 11 .§ 5 . In this convention,
F B B H B ( 0 ) = − 2 v 2 C H B , F 33 H W ( 0 ) = − 2 v 2 C H W , F 3 B H W B ( 0 ) = + v 2 C H W B . (30) F^{HB}_{BB}(0)=-2v^{2}C_{HB},\qquad F^{HW}_{33}(0)=-2v^{2}C_{HW},\qquad F^{HWB}_{3B}(0)=+v^{2}C_{HWB}. \tag*{(30)} F B B H B ( 0 ) = − 2 v 2 C H B , F 33 H W ( 0 ) = − 2 v 2 C H W , F 3 B H W B ( 0 ) = + v 2 C H W B . (30)
Equating these local insertions to the loop form factors therefore gives, for one HNL coupled to one lepton flavour,
Δ C H B = + g ′ 2 M N 2 8 π 2 C B 2 , (31) \Delta C_{HB}=+\frac{g^{\prime2}M_N^2}{8\pi^2}C_B^2, \tag*{(31)} Δ C H B = + 8 π 2 g ′2 M N 2 C B 2 , (31)
Δ C H W = + g 2 M N 2 32 π 2 C W 2 , (32) \Delta C_{HW}=+\frac{g^2M_N^2}{32\pi^2}C_W^2, \tag*{(32)} Δ C H W = + 32 π 2 g 2 M N 2 C W 2 , (32)
Δ C H W B = − g g ′ M N 2 8 π 2 C B C W . (33) \Delta C_{HWB}=-\frac{gg^{\prime}M_N^2}{8\pi^2}C_BC_W. \tag*{(33)} Δ C H W B = − 8 π 2 g g ′ M N 2 C B C W . (33)
The mixed form factor gives a useful physical consequence. Denoting the HNL-loop contribution by F 3 B l o o p F_{3B}^{\mathrm{loop}} F 3 B loop , one obtains
F 3 B l o o p ( 0 ) = − g g ′ v 2 M N 2 8 π 2 C B C W , S ^ = c W s W F 3 B ( 0 ) = + c W s W v 2 Δ C H W B . (34) F_{3B}^{\mathrm{loop}}(0)=-\frac{gg^{\prime}v^2M_N^2}{8\pi^2}C_BC_W,\qquad\widehat{S}=\frac{c_W}{s_W}F_{3B}(0)=+\frac{c_W}{s_W}v^2\Delta C_{HWB}. \tag*{(34)} F 3 B loop ( 0 ) = − 8 π 2 g g ′ v 2 M N 2 C B C W , S = s W c W F 3 B ( 0 ) = + s W c W v 2 Δ C H W B . (34)
The sign of the induced oblique correction is consequently fixed: positive C B C W C_BC_W C B C W gives negative S ^ \widehat{S} S . To translate the loop form factor into a Warsaw coefficient, we use the quadratic action in momentum space, whose second derivative with respect to the gauge fields is the two-point insertion. This accounts for the relative factors of diagonal and mixed field strengths. Equations (31) –(33) then specify the dipole contribution to the bosonic coefficients; the UV boundary terms introduced above are added separately.
Finite lepton-current thresholds
The portal couples to the left-handed SM doublet. With negligible charged-lepton Yukawas, there is no chirality-changing interaction that would turn its direct correction into a right-handed charged-lepton current. Electroweak invariance then organizes the leading direct response into a singlet and a triplet. We use the Warsaw operators [33 ] ,
Q H l ( 1 ) = ( H † i D ↔ μ H ) ( L ‾ γ μ L ) , (35) Q_{Hl}^{(1)}=(H^\dagger i\overleftrightarrow{D}_\mu H)(\overline{L}\gamma^\mu L), \tag*{(35)} Q H l ( 1 ) = ( H † i D μ H ) ( L γ μ L ) , (35)
Q H l ( 3 ) = ( H † i D ↔ μ I H ) ( L ‾ σ I γ μ L ) , (36) Q_{Hl}^{(3)}=(H^\dagger i\overleftrightarrow{D}_\mu^I H)(\overline{L}\sigma^I\gamma^\mu L), \tag*{(36)} Q H l ( 3 ) = ( H † i D μ I H ) ( L σ I γ μ L ) , (36)
where H † i D ↔ μ I H = i H † σ I D μ H − i ( D μ H ) † σ I H H^\dagger i\overleftrightarrow{D}_\mu^I H=iH^\dagger\sigma^I D_\mu H-i(D_\mu H)^\dagger\sigma^I H H † i D μ I H = i H † σ I D μ H − i ( D μ H ) † σ I H . Both triplet currents use the Pauli-matrix normalization. We extract their coefficients from amplitudes with external lepton and Higgs fields and compare the resulting currents with the broken-phase vertex and external-leg contributions shown in Figure 2 ; details are given in appendix § 12 .§ 3 . For one real flavour, negligible charged-lepton Yukawa coupling and the local heavy-mass expansion, this gives
Figure 2. The five one-particle-irreducible (1PI) HNL contributions to Z e e Zee Z ee , organized into three classes. Squares indicate dipole vertices, including the four-point Z e N W ZeNW Z e N W contact in (c); dots denote SM vertices. Fermion lines are schematic and Majorana arrows are suppressed. The sum of (b) and the two contacts, not (b) alone, gives the d W 2 d_W^2 d W 2 term of eq. (B.5). External-leg and input-parameter terms are added separately.
Δ C H l ( 1 ) ( μ ) = M N 2 128 π 2 ( 4 g ′ 2 C B 2 + 3 g 2 C W 2 ) K M S ‾ , (37) \Delta C_{Hl}^{(1)}(\mu)=\frac{M_N^2}{128\pi^2}\left(4g^{\prime2}C_B^2+3g^2C_W^2\right)K_{\overline{\mathrm{MS}}}, \tag*{(37)} Δ C H l ( 1 ) ( μ ) = 128 π 2 M N 2 ( 4 g ′2 C B 2 + 3 g 2 C W 2 ) K MS , (37)
Δ C H l ( 3 ) ( μ ) = − M N 2 128 π 2 ( 4 g ′ 2 C B 2 − g 2 C W 2 ) K M S ‾ , (38) \Delta C_{Hl}^{(3)}(\mu)=-\frac{M_N^2}{128\pi^2}\left(4g^{\prime2}C_B^2-g^2C_W^2\right)K_{\overline{\mathrm{MS}}}, \tag*{(38)} Δ C H l ( 3 ) ( μ ) = − 128 π 2 M N 2 ( 4 g ′2 C B 2 − g 2 C W 2 ) K MS , (38)
K M S ‾ = 1 − 6 ln μ 2 M N 2 . (39) K_{\overline{\mathrm{MS}}}=1-6\ln\frac{\mu^2}{M_N^2}. \tag*{(39)} K MS = 1 − 6 ln M N 2 μ 2 . (39)
At μ = M N \mu=M_N μ = M N the logarithm vanishes. Together with eqs. (31) –(33) , these are the five finite thresholds used in the dimension-six matching. All have mass dimension − 2 -2 − 2 , but arise at second order in the dimension-six portal, M N 2 C 2 M_N^2C^2 M N 2 C 2 . The full coefficients are C i E F T = C i i n d e p e n d e n t + Δ C i C_i^{\mathrm{EFT}}=C_i^{\mathrm{independent}}+\Delta C_i C i EFT = C i independent + Δ C i , where the independent terms encode ultraviolet boundary conditions. We use naive dimensional regularization (NDR) for the open fermion chains. The constant 1 1 1 in K M S ‾ K_{\overline{\mathrm{MS}}} K MS is specific to this prescription together with MS subtraction: a finite scheme change shifts it together with an opposite shift of C i i n d e p e n d e n t C_i^{\mathrm{independent}} C i independent , leaving the full amplitude unchanged. The logarithmic slope is fixed by the one-loop anomalous dimension in this operator convention; the logarithm alone is not an observable. Our dipole-dominance forecasts refer specifically to vanishing independent boundary coefficients in this prescription.
The relative weights of the singlet and triplet follow from their electroweak group contractions. Their common finite normalization contains additional dynamical information: the constant in K M S K_{\mathrm{MS}} K MS includes products of D D D -dependent numerator terms and UV poles. There is no mixed C B C W C_B C_W C B C W term in this unbroken two-insertion bubble because there is no tree-level B B B –W I W^I W I propagator. Mixed coefficients may appear in other topologies or at higher perturbative orders.
For L Z = − Z μ f ˉ γ μ ( g f L P L + g f R P R ) f \mathcal{L}_Z=-Z_\mu\bar{f}\gamma^\mu(g_{fL}P_L+g_{fR}P_R)f L Z = − Z μ f ˉ γ μ ( g f L P L + g f R P R ) f , g Z = e / ( s W c W ) g_Z=e/(s_W c_W) g Z = e / ( s W c W ) , the local direct current shifts are
Δ g ν L = g Z v 2 2 ( − Δ C H l ( 1 ) + Δ C H l ( 3 ) ) , (40) \Delta g_{\nu L}=\frac{g_Zv^2}{2}\left(-\Delta C_{Hl}^{(1)}+\Delta C_{Hl}^{(3)}\right), \tag*{(40)} Δ g ν L = 2 g Z v 2 ( − Δ C H l ( 1 ) + Δ C H l ( 3 ) ) , (40)
Δ g e L = − g Z v 2 2 ( Δ C H l ( 1 ) + Δ C H l ( 3 ) ) . (41) \Delta g_{eL}=-\frac{g_Zv^2}{2}\left(\Delta C_{Hl}^{(1)}+\Delta C_{Hl}^{(3)}\right). \tag*{(41)} Δ g e L = − 2 g Z v 2 ( Δ C H l ( 1 ) + Δ C H l ( 3 ) ) . (41)
These direct shifts already include the external-lepton residues associated with the current matching. The observable response also requires gauge propagation, gauge-field normalization and the conversion to measured electroweak inputs, assembled in § 6 . For the charged current written as − g / 2 W ν + ν ˉ L γ ν e L + h . c . -g/\sqrt{2}\,W_\nu^+\bar{\nu}_L\gamma^\nu e_L+\mathrm{h.c.} − g / 2 W ν + ν ˉ L γ ν e L + h.c. , the direct relative shift is v 2 Δ C H l ( 3 ) v^2\Delta C_{Hl}^{(3)} v 2 Δ C H l ( 3 ) . This relation follows from the triplet operator and does not require a separate charged-current matching coefficient at dimension six.
The two contact diagrams in Figure 2 contain the four-point Z e N W ZeNW Z e N W interaction fixed by the non-Abelian weak field strength. They contribute at the same order as the triangles and are needed for the normalization and symmetry relations of the complete vertex. External-leg residues normalize the asymptotic lepton states, while input corrections express the amplitude in terms of measured electroweak parameters. We combine these contributions using Born-normalized physical couplings and common MS subtraction, as detailed in appendix § 12 .§ 8 .
Dimension-eight counterterms and leading logarithms
Beyond the leading terms discussed above, the momentum expansion probes the curvature of the gauge-boson self-energies. These curvature terms are conventionally described by W W W for weak isospin, Y Y Y for hypercharge and X X X for their mixed component. The parameter S ^ \widehat{S} S introduced above instead measures the leading neutral gauge mixing. Their precise derivative definitions are collected in § 5 .
The divergent q 4 q^4 q 4 terms arise at order C B , W 2 ∼ Λ − 4 C_{B,W}^2\sim\Lambda^{-4} C B , W 2 ∼ Λ − 4 and require dimension-eight counterterms in a consistent EFT treatment. For this purpose we keep a set of operators that reproduces the amplitudes before imposing the external equations of motion, often called a Green basis. It makes the propagator corrections explicit through higher-derivative structures of the form
Q 2 B , H = ( H † H ) ( ∂ μ B μ ν ) ( ∂ ρ B ρ ν ) , Q 2 W , H = ( H † H ) ( D μ W I μ ν ) ( D ρ W ρ ν I ) , Q 2 W B , H = ( H † σ I H ) ( D μ W I μ ν ) ( ∂ ρ B ρ ν ) , (42) \begin{aligned}
Q_{2B,H}&=(H^\dagger H)(\partial_\mu B^{\mu\nu})(\partial^\rho B_{\rho\nu}),\\
Q_{2W,H}&=(H^\dagger H)(D_\mu W^{I\mu\nu})(D^\rho W^I_{\rho\nu}),\\
Q_{2WB,H}&=(H^\dagger\sigma^I H)(D_\mu W^{I\mu\nu})(\partial^\rho B_{\rho\nu}), \tag*{(42)}
\end{aligned} Q 2 B , H Q 2 W , H Q 2 W B , H = ( H † H ) ( ∂ μ B μν ) ( ∂ ρ B ρ ν ) , = ( H † H ) ( D μ W I μν ) ( D ρ W ρ ν I ) , = ( H † σ I H ) ( D μ W I μν ) ( ∂ ρ B ρ ν ) , (42)
up to basis choices, integration by parts and equations of motion [19 , 20 ] . With precisely the definitions in eq. (42) , cancellation of the poles in our convention requires
δ C 2 B , H = + g ′ 2 C B 2 12 π 2 ϵ , δ C 2 W , H = + g 2 C W 2 48 π 2 ϵ , δ C 2 W B , H = − g g ′ C B C W 12 π 2 ϵ . (43) \begin{aligned}
\delta C_{2B,H} &= +\frac{g^{\prime2} C_B^2}{12\pi^2\epsilon},\\
\delta C_{2W,H} &= +\frac{g^2 C_W^2}{48\pi^2\epsilon},\\
\delta C_{2WB,H} &= -\frac{gg'C_BC_W}{12\pi^2\epsilon}. \tag*{(43)}
\end{aligned} δ C 2 B , H δ C 2 W , H δ C 2 W B , H = + 12 π 2 ϵ g ′2 C B 2 , = + 48 π 2 ϵ g 2 C W 2 , = − 12 π 2 ϵ g g ′ C B C W . (43)
The signs and numerical factors in eq. (43) are tied to the operator definitions in eq. (42) ; translating to a different dimension-eight basis requires the corresponding field redefinitions and EOM relations.
A physical prediction for W W W , Y Y Y and X X X therefore requires a boundary condition for the independent dimension-eight Wilson coefficients. Under the minimal assumption that the relevant bosonic dimension-eight coefficients vanish at a UV scale Λ \Lambda Λ , the calculable leading logarithm is obtained from
F 0 ′ ( 0 ) ∣ L L = 1 6 π 2 ln Λ 2 M N 2 . (44) F'_0(0)\big|_{\mathrm{LL}}=\frac{1}{6\pi^2}\ln\frac{\Lambda^2}{M_N^2}. \tag*{(44)} F 0 ′ ( 0 ) LL = 6 π 2 1 ln M N 2 Λ 2 . (44)
The logarithm measures the evolution between the chosen UV boundary and the HNL mass. Equation (44) follows by combining that evolution with the stated vanishing boundary coefficient. A different boundary adds its own local contribution to the same observable response. The logarithm isolates the scale-dependent part of the curvature; finite matching contributes at the same order. The forecast retains both, including the 11/6 constant in the bosonic slope and the finite lepton-current terms given in appendix B.8.
Dimension-six lepton-current poles
The current thresholds of section 4.2 also carry scale dependence. Their UV poles determine the local terms needed to subtract the divergence and the running that connects the boundary scale to M N M_N M N . To make the group structure and normalization explicit, we write the Warsaw operators [33 ] with Pauli matrices in both triplet currents:
Q H l ( 1 ) = ( H † i D ↔ μ H ) ( L ˉ γ μ L ) , Q_{Hl}^{(1)} = (H^\dagger i\overleftrightarrow{D}_\mu H)(\bar{L}\gamma^\mu L), Q H l ( 1 ) = ( H † i D μ H ) ( L ˉ γ μ L ) ,
Q H l ( 3 ) = ( H † i D ↔ μ I H ) ( L ˉ σ I γ μ L ) , (45) Q_{Hl}^{(3)} = (H^\dagger i\overleftrightarrow{D}_\mu^I H)(\bar{L}\sigma^I\gamma^\mu L), \tag*{(45)} Q H l ( 3 ) = ( H † i D μ I H ) ( L ˉ σ I γ μ L ) , (45)
H † i D ↔ μ I H = i H † σ I D μ H − i ( D μ H ) † σ I H . (46) H^\dagger i\overleftrightarrow{D}_\mu^I H = iH^\dagger\sigma^I D_\mu H - i(D_\mu H)^\dagger\sigma^I H. \tag*{(46)} H † i D μ I H = i H † σ I D μ H − i ( D μ H ) † σ I H . (46)
Thus the σ I \sigma^I σ I in Q H l ( 3 ) Q_{Hl}^{(3)} Q H l ( 3 ) must not be replaced by the gauge generator τ I = σ I / 2 \tau^I = \sigma^I/2 τ I = σ I /2 used in O W O_W O W . We use D μ = ∂ μ + i g τ I W μ I + i g ′ Y B μ D_\mu= \partial_\mu+ ig\tau^I W_\mu^I + ig'Y B_\mu D μ = ∂ μ + i g τ I W μ I + i g ′ Y B μ and define physical neutral couplings by L Z = − Z μ f ˉ γ μ ( g f L P L + g f R P R ) f \mathcal{L}_Z = -Z_\mu\bar{f}\gamma^\mu(g_{fL}P_L + g_{fR}P_R)f L Z = − Z μ f ˉ γ μ ( g f L P L + g f R P R ) f . This fixes the sign of the coupling shifts below.
Let C H l a , p o l e C_{Hl}^{a,\mathrm{pole}} C H l a , pole denote the local Wilson coefficient that reproduces the loop UV pole, rather than its cancelling counterterm. Direct four-point matching, with external L ˉ , L , H , H † \bar{L}, L, H, H^\dagger L ˉ , L , H , H † , gives
C H l ( 1 ) , p o l e = − 3 M N 2 64 π 2 ϵ U V ( 4 g B 2 C B 2 + 3 g W 2 C W 2 ) , (47) C_{Hl}^{(1),\mathrm{pole}} = -\frac{3M_N^2}{64\pi^2\epsilon_{\mathrm{UV}}}(4g_B^2C_B^2 + 3g_W^2C_W^2), \tag*{(47)} C H l ( 1 ) , pole = − 64 π 2 ϵ UV 3 M N 2 ( 4 g B 2 C B 2 + 3 g W 2 C W 2 ) , (47)
C H l ( 3 ) , p o l e = + 3 M N 2 64 π 2 ϵ U V ( 4 g B 2 C B 2 − g W 2 C W 2 ) , (48) C_{Hl}^{(3),\mathrm{pole}} = +\frac{3M_N^2}{64\pi^2\epsilon_{\mathrm{UV}}}(4g_B^2C_B^2 - g_W^2C_W^2), \tag*{(48)} C H l ( 3 ) , pole = + 64 π 2 ϵ UV 3 M N 2 ( 4 g B 2 C B 2 − g W 2 C W 2 ) , (48)
δ C H l ( a ) = − C H l ( a ) , p o l e , a = 1 , 3. (49) \delta C_{Hl}^{(a)} = -C_{Hl}^{(a),\mathrm{pole}}, \qquad a = 1,3. \tag*{(49)} δ C H l ( a ) = − C H l ( a ) , pole , a = 1 , 3. (49)
Both expressions have dimension − 2 -2 − 2 : the two coefficients of dimension − 2 -2 − 2 are accompanied by M N 2 M_N^2 M N 2 . These are divergent dimension-six local terms at order Λ − 4 \Lambda^{-4} Λ − 4 , unlike the finite bosonic thresholds in eqs. (31) –(33) . The poles fix the scale dependence of the current operators. Their finite thresholds are given in § 4.2 , while the independent UV terms are specified by a boundary prescription when constructing the phenomenological benchmark.
Electroweak group structure. The relative singlet and triplet weights follow from weak-isospin algebra, independently of the scalar loop integral. The two portal vertices carry weak-isospin indices, and projecting their product onto the current operators separates the singlet and triplet contributions. For H ~ i = ϵ i k H k ∗ \widetilde{H}_i = \epsilon_{ik}H_k^* H i = ϵ ik H k ∗ , ϵ = i σ 2 \epsilon= i\sigma^2 ϵ = i σ 2 , the two relevant S U ( 2 ) SU(2) S U ( 2 ) tensors reduce to
ϵ i k ϵ j l ∗ = 1 2 δ i j δ k l − 1 2 ∑ A σ i j A σ k l A , \epsilon_{ik}\epsilon^*_{jl} = \frac{1}{2}\delta_{ij}\delta_{kl} - \frac{1}{2}\sum_A \sigma^A_{ij}\sigma^A_{kl}, ϵ ik ϵ j l ∗ = 2 1 δ ij δ k l − 2 1 A ∑ σ ij A σ k l A ,
∑ A ( τ A ϵ ) i k ( τ A ϵ ) j l ∗ = 3 8 δ i j δ k l + 1 8 ∑ A σ i j A σ k l A . (50) \sum_A(\tau^A\epsilon)_{ik}(\tau^A\epsilon)^*_{jl} = \frac{3}{8}\delta_{ij}\delta_{kl} + \frac{1}{8}\sum_A \sigma^A_{ij}\sigma^A_{kl}. \tag*{(50)} A ∑ ( τ A ϵ ) ik ( τ A ϵ ) j l ∗ = 8 3 δ ij δ k l + 8 1 A ∑ σ ij A σ k l A . (50)
Using the same index ordering in the two current projections, these identities explain the singlet/triplet ratios − 1 -1 − 1 for B 2 B^2 B 2 and 3 3 3 for W 2 W^2 W 2 . They do not determine the common loop normalization or its sign; those are fixed by the loop integral and a tree-level current-operator benchmark.
In the unbroken two-dipole bubble, the internal gauge propagator is either B B B or W I W^I W I . There is no tree-level B B B –W I W^I W I propagator mixing in the unbroken theory, so this topology produces no C B C W C_BC_W C B C W current pole at dimension six. The cancellation of the corresponding M N 2 M_N^2 M N 2 terms between γ \gamma γ and Z Z Z exchange in the broken calculation is an independent check. This statement is specific to this order and topology; gauge symmetry alone does not forbid mixed coefficients in every amplitude. In particular the mixed bosonic threshold and the dimension-eight vertex residues are nonzero.
Separating the Higgs current from an EOM term. The four-point amplitude in Figure 3 carries independent external Higgs momenta. Its momentum dependence distinguishes a physical current correction from a change of field variables. With all external momenta incoming, p 1 + p 2 + p 3 + p 4 = 0 p_1+p_2+p_3+p_4=0 p 1 + p 2 + p 3 + p 4 = 0 , the bubble momentum can be decomposed as
Figure 3. Schematic unbroken four-point matching. Each square is a dipole insertion with one external Higgs field. The graph displays field content; conjugate insertions and the appropriate external crossings are understood. The local momentum expansion separates the M N 2 M_N^2 M N 2 current poles from higher-derivative terms.
̸ p 1 + ̸ p 3 = 1 2 ( ̸ p 3 − ̸ p 4 ) + 1 2 ( ̸ p 1 − ̸ p 2 ) . (51) \not{p}_1+\not{p}_3=\frac{1}{2}(\not{p}_3-\not{p}_4)+\frac{1}{2}(\not{p}_1-\not{p}_2). \tag*{(51)} p 1 + p 3 = 2 1 ( p 3 − p 4 ) + 2 1 ( p 1 − p 2 ) . (51)
The first structure projects onto the Higgs current. The second is associated with a redundant fermion-derivative operator, for example
R H l ( 1 ) = ( H † H ) L ˉ i D̸ ↔ L → E O M ( H † H ) ( L ˉ Y e H e R + h . c . ) . (52) R_{Hl}^{(1)}=(H^\dagger H)\bar{L}_i\overleftrightarrow{\not{D}}L \xrightarrow{\mathrm{EOM}} (H^\dagger H)(\bar{L}Y_eHe_R+\mathrm{h.c.}). \tag*{(52)} R H l ( 1 ) = ( H † H ) L ˉ i D L EOM ( H † H ) ( L ˉ Y e H e R + h.c. ) . (52)
It vanishes in the charged-lepton Yukawa limit used here, or maps to a Yukawa-type operator when Y e Y_e Y e is retained. The operator is therefore eliminated in the current matching used here. Consequently a lepton self-energy with a constant Higgs background alone is insufficient to determine the two current coefficients; the four-point momentum dependence and the group tensors are required.
After symmetry breaking, the current operators predict
Δ g ν L ( 6 ) = g Z v 2 2 ( − C H l ( 1 ) + C H l ( 3 ) ) , Δ g e L ( 6 ) = − g Z v 2 2 ( C H l ( 1 ) + C H l ( 3 ) ) , g Z = e s W c W . (53) \begin{aligned}
\Delta g_{\nu_L}^{(6)}&=\frac{g_Zv^2}{2}\left(-C_{Hl}^{(1)}+C_{Hl}^{(3)}\right),\\
\Delta g_{e_L}^{(6)}&=-\frac{g_Zv^2}{2}\left(C_{Hl}^{(1)}+C_{Hl}^{(3)}\right),\qquad g_Z=\frac{e}{s_Wc_W}. \tag*{(53)}
\end{aligned} Δ g ν L ( 6 ) Δ g e L ( 6 ) = 2 g Z v 2 ( − C H l ( 1 ) + C H l ( 3 ) ) , = − 2 g Z v 2 ( C H l ( 1 ) + C H l ( 3 ) ) , g Z = s W c W e . (53)
Δ g ν L ( 6 ) , p o l e = 3 e 3 v 2 M N 2 64 π 2 c W 2 s W 3 ϵ U V ( 4 s W 2 C B 2 + c W 2 C W 2 ) , (54) \Delta g_{\nu_L}^{(6),\mathrm{pole}}=\frac{3e^3v^2M_N^2}{64\pi^2c_W^2s_W^3\epsilon_{\mathrm{UV}}}\left(4s_W^2C_B^2+c_W^2C_W^2\right), \tag*{(54)} Δ g ν L ( 6 ) , pole = 64 π 2 c W 2 s W 3 ϵ UV 3 e 3 v 2 M N 2 ( 4 s W 2 C B 2 + c W 2 C W 2 ) , (54)
Δ g e L ( 6 ) , p o l e = 3 e 3 v 2 M N 2 32 π 2 c W s W 3 ϵ U V C W 2 . \Delta g_{e_L}^{(6),\mathrm{pole}}=\frac{3e^3v^2M_N^2}{32\pi^2c_Ws_W^3\epsilon_{\mathrm{UV}}}C_W^2. Δ g e L ( 6 ) , pole = 32 π 2 c W s W 3 ϵ UV 3 e 3 v 2 M N 2 C W 2 .
These coincide with the independently calculated broken-phase residues in appendix § 12 .§ 1 . The absence of C B 2 C_B^2 C B 2 in the electron result follows from C H l ( 1 ) , B 2 + C H l ( 3 ) , B 2 = 0 C_{Hl}^{(1),B2}+C_{Hl}^{(3),B2}=0 C H l ( 1 ) , B 2 + C H l ( 3 ) , B 2 = 0 .
Extended oblique parameters
The same propagator corrections affect several fermion species. Oblique parameters describe this shared part of the electroweak response through a small set of self-energy coefficients: gauge mixing, shifts of mass relations and curvature with momentum. They provide a useful physical summary of the bosonic matching. We use the extended description of ref. [34 ] , including the mixed higher-derivative parameter X X X . This gives a compact physical interpretation of the self-energies before the direct lepton currents are included. Our normalizations are
S ^ = g g ′ Π 3 B ′ ( 0 ) = c W s W F 3 B ( 0 ) , (55) \widehat{S}=\frac{g}{g'}\Pi'_{3B}(0)=\frac{c_W}{s_W}F_{3B}(0), \tag*{(55)} S = g ′ g Π 3 B ′ ( 0 ) = s W c W F 3 B ( 0 ) , (55)
T ^ = Π 33 ( 0 ) − Π 11 ( 0 ) m W 2 , (56) \widehat{T}=\frac{\Pi_{33}(0)-\Pi_{11}(0)}{m_W^2}, \tag*{(56)} T = m W 2 Π 33 ( 0 ) − Π 11 ( 0 ) , (56)
W = m W 2 2 Π 33 ′ ′ ( 0 ) = m W 2 F 33 ′ ( 0 ) , (57) W=\frac{m_W^2}{2}\Pi''_{33}(0)=m_W^2F'_{33}(0), \tag*{(57)} W = 2 m W 2 Π 33 ′′ ( 0 ) = m W 2 F 33 ′ ( 0 ) , (57)
Y = m W 2 2 Π B B ′ ′ ( 0 ) = m W 2 F B B ′ ( 0 ) , (58) Y=\frac{m_W^2}{2}\Pi''_{BB}(0)=m_W^2F'_{BB}(0), \tag*{(58)} Y = 2 m W 2 Π B B ′′ ( 0 ) = m W 2 F B B ′ ( 0 ) , (58)
X = m W 2 2 Π 3 B ′ ′ ( 0 ) = m W 2 F 3 B ′ ( 0 ) . (59) X=\frac{m_W^2}{2}\Pi''_{3B}(0)=m_W^2F'_{3B}(0). \tag*{(59)} X = 2 m W 2 Π 3 B ′′ ( 0 ) = m W 2 F 3 B ′ ( 0 ) . (59)
The sign of mixed quantities is convention dependent and should always be read together with eqs. (3) and (18) .
The finite HNL threshold gives
S ^ = − c W s W M N 2 2 π 2 d B d 3 . (60) \widehat{S}=-\frac{c_W}{s_W}\frac{M_N^2}{2\pi^2}d_Bd_3. \tag*{(60)} S = − s W c W 2 π 2 M N 2 d B d 3 . (60)
The dipole vertices force the two-point insertion to be transverse. Since its scalar form factor is regular at zero momentum, the insertion vanishes there and cannot generate a gauge-boson mass term. Thus, for the dipole–dipole two-point functions considered here,
T ^ = 0 , (61) \widehat{T}=0, \tag*{(61)} T = 0 , (61)
and eq. (27) implies U ^ = V = 0 \widehat{U}=V=0 U = V = 0 in the massless charged-lepton limit.
Separately from the dimension-six thresholds, and imposing C d = 8 ( Λ ) = 0 C_{d=8}(\Lambda)=0 C d = 8 ( Λ ) = 0 for the relevant curvature coefficients in the specified scheme, the leading-logarithmic curvature parameters are
Y L L = m W 2 d B 2 6 π 2 ln Λ 2 M N 2 , (62) Y_{\mathrm{LL}}=\frac{m_W^2d_B^2}{6\pi^2}\ln\frac{\Lambda^2}{M_N^2}, \tag*{(62)} Y LL = 6 π 2 m W 2 d B 2 ln M N 2 Λ 2 , (62)
W L L = m W 2 d 3 2 6 π 2 ln Λ 2 M N 2 = m W 2 d W 2 12 π 2 ln Λ 2 M N 2 , (63) W_{\mathrm{LL}}=\frac{m_W^2d_3^2}{6\pi^2}\ln\frac{\Lambda^2}{M_N^2}
=\frac{m_W^2d_W^2}{12\pi^2}\ln\frac{\Lambda^2}{M_N^2}, \tag*{(63)} W LL = 6 π 2 m W 2 d 3 2 ln M N 2 Λ 2 = 12 π 2 m W 2 d W 2 ln M N 2 Λ 2 , (63)
X L L = m W 2 d B d 3 6 π 2 ln Λ 2 M N 2 . (64) X_{\mathrm{LL}}=\frac{m_W^2d_Bd_3}{6\pi^2}\ln\frac{\Lambda^2}{M_N^2}. \tag*{(64)} X LL = 6 π 2 m W 2 d B d 3 ln M N 2 Λ 2 . (64)
Equivalently,
Y L L = m W 2 g ′ 2 v 2 12 π 2 C B 2 L , W L L = m W 2 g 2 v 2 48 π 2 C W 2 L , X L L = m W 2 g g ′ v 2 24 π 2 C B C W L , L ≡ ln Λ 2 M N 2 . (65) \begin{aligned}
Y_{\mathrm{LL}}&=\frac{m_W^2g'^2v^2}{12\pi^2}C_B^2L,
&
W_{\mathrm{LL}}&=\frac{m_W^2g^2v^2}{48\pi^2}C_W^2L,\\
X_{\mathrm{LL}}&=\frac{m_W^2gg'v^2}{24\pi^2}C_BC_WL,
&
L&\equiv\ln\frac{\Lambda^2}{M_N^2}. \tag*{(65)}
\end{aligned} Y LL X LL = 12 π 2 m W 2 g ′2 v 2 C B 2 L , = 24 π 2 m W 2 g g ′ v 2 C B C W L , W LL L = 48 π 2 m W 2 g 2 v 2 C W 2 L , ≡ ln M N 2 Λ 2 . (65)
The finite threshold can similarly be written as
S ^ = − g 2 v 2 M N 2 8 π 2 C B C W . (66) \widehat{S}=-\frac{g^2v^2M_N^2}{8\pi^2}C_BC_W. \tag*{(66)} S = − 8 π 2 g 2 v 2 M N 2 C B C W . (66)
Rank-one correlations
Equations (62) –(64) imply the exact correlation
X L L 2 = W L L Y L L . (67) X_{\mathrm{LL}}^2=W_{\mathrm{LL}}Y_{\mathrm{LL}}. \tag*{(67)} X LL 2 = W LL Y LL . (67)
This follows directly from the rank-one form of the self-energy matrix and is therefore independent of the choice between the gauge-invariant and broken parameterizations. At leading-logarithmic order in the curvature terms, fixed M N M_N M N and Λ \Lambda Λ also relate S ^ \widehat{S} S to X L L X_{\mathrm{LL}} X LL :
S ^ = − 3 c W s W M N 2 m W 2 ln ( Λ 2 / M N 2 ) X L L . (68) \widehat{S}=-3\frac{c_W}{s_W}\frac{M_N^2}{m_W^2\ln\left(\Lambda^2/M_N^2\right)}X_{\mathrm{LL}}. \tag*{(68)} S = − 3 s W c W m W 2 ln ( Λ 2 / M N 2 ) M N 2 X LL . (68)
Thus this induced two-point contribution occupies a restricted subspace. The relation assumes the stated UV boundary conditions and ln ( Λ 2 / M N 2 ) ≠ 0 \ln\left(\Lambda^2/M_N^2\right)\ne0 ln ( Λ 2 / M N 2 ) = 0 ; independent bosonic Wilson coefficients can violate it. It is not a constraint on the total EFT or on the nonuniversal lepton-current sector.
In the broken basis,
Y L L = m W 2 L 6 π 2 ( c W d γ − s W d Z ) 2 , W L L = m W 2 L 6 π 2 ( s W d γ + c W d Z ) 2 , X L L = m W 2 L 6 π 2 ( c W d γ − s W d Z ) ( s W d γ + c W d Z ) , S ^ = − c W s W M N 2 2 π 2 ( c W d γ − s W d Z ) ( s W d γ + c W d Z ) . (69) \begin{aligned}
Y_{\mathrm{LL}}&=\frac{m_W^2L}{6\pi^2}(c_Wd_\gamma-s_Wd_Z)^2,\\
W_{\mathrm{LL}}&=\frac{m_W^2L}{6\pi^2}(s_Wd_\gamma+c_Wd_Z)^2,\\
X_{\mathrm{LL}}&=\frac{m_W^2L}{6\pi^2}(c_Wd_\gamma-s_Wd_Z)(s_Wd_\gamma+c_Wd_Z),\\
\widehat{S}&=-\frac{c_W}{s_W}\frac{M_N^2}{2\pi^2}(c_Wd_\gamma-s_Wd_Z)(s_Wd_\gamma+c_Wd_Z). \tag*{(69)}
\end{aligned} Y LL W LL X LL S = 6 π 2 m W 2 L ( c W d γ − s W d Z ) 2 , = 6 π 2 m W 2 L ( s W d γ + c W d Z ) 2 , = 6 π 2 m W 2 L ( c W d γ − s W d Z ) ( s W d γ + c W d Z ) , = − s W c W 2 π 2 M N 2 ( c W d γ − s W d Z ) ( s W d γ + c W d Z ) . (69)
Within the same leading-logarithmic contribution, the two curvature blind directions are therefore
Y = 0 : d Z = c W s W d γ , W = 0 : d Z = − s W c W d γ . (70) Y=0:\quad d_Z=\frac{c_W}{s_W}d_\gamma,\qquad W=0:\quad d_Z=-\frac{s_W}{c_W}d_\gamma. \tag*{(70)} Y = 0 : d Z = s W c W d γ , W = 0 : d Z = − c W s W d γ . (70)
For the illustrative direction d Z = 0 d_Z=0 d Z = 0 ,
∣ S ^ ∣ Y = 3 M N 2 m W 2 ln ( Λ 2 / M N 2 ) . (71) \frac{|\widehat{S}|}{Y}=\frac{3M_N^2}{m_W^2\ln\left(\Lambda^2/M_N^2\right)}. \tag*{(71)} Y ∣ S ∣ = m W 2 ln ( Λ 2 / M N 2 ) 3 M N 2 . (71)
Using m W = 80.379 m_W=80.379 m W = 80.379 GeV as an illustrative input and Λ = 1 \Lambda=1 Λ = 1 TeV, the equality ∣ S ^ ∣ = Y |\widehat{S}|=Y ∣ S ∣ = Y occurs at M N ≃ 99.66 M_N\simeq99.66 M N ≃ 99.66 GeV. This comparison identifies where the two theory contributions have comparable size. Their relative importance in a precision measurement also depends on the uncertainties and correlations of the observables that probe each oblique direction.
Relation to epsilon combinations
The traditional epsilon variables offer an alternative description of the bosonic contribution to pole observables. In the convention used here, the extended oblique parameters give, for T ^ = U ^ = V = 0 \widehat{T}=\widehat{U}=V=0 T = U = V = 0 ,
δ ϵ 1 = − W + 2 s W c W X − s W 2 c W 2 Y , δ ϵ 2 = − W + 2 s W c W X , δ ϵ 3 = S ^ − W + X s W c W − Y . (72) \begin{aligned}
\delta\epsilon_1&=-W+2\frac{s_W}{c_W}X-\frac{s_W^2}{c_W^2}Y,\\
\delta\epsilon_2&=-W+2\frac{s_W}{c_W}X,\\
\delta\epsilon_3&=\widehat{S}-W+\frac{X}{s_Wc_W}-Y. \tag*{(72)}
\end{aligned} δ ϵ 1 δ ϵ 2 δ ϵ 3 = − W + 2 c W s W X − c W 2 s W 2 Y , = − W + 2 c W s W X , = S − W + s W c W X − Y . (72)
For d Z = 0 d_Z=0 d Z = 0 these reduce to
δ ϵ 1 = 0 , \delta\epsilon_1=0, δ ϵ 1 = 0 ,
δ ϵ 2 = m W 2 s W 2 d γ 2 6 π 2 ln Λ 2 M N 2 , \delta\epsilon_2=\frac{m_W^2s_W^2d_\gamma^2}{6\pi^2}\ln\frac{\Lambda^2}{M_N^2}, δ ϵ 2 = 6 π 2 m W 2 s W 2 d γ 2 ln M N 2 Λ 2 ,
δ ϵ 3 = − c W 2 M N 2 d γ 2 2 π 2 . (73) \delta\epsilon_3=-\frac{c_W^2M_N^2d_\gamma^2}{2\pi^2}. \tag*{(73)} δ ϵ 3 = − 2 π 2 c W 2 M N 2 d γ 2 . (73)
These relations summarize the bosonic part of the pole response. The forecast below uses the pole pseudo-observables directly and also includes the lepton-current corrections, which the oblique parameters alone do not describe.
From matching to electroweak observables
The matching coefficients affect measured quantities in two ways: they change the gauge-boson propagators and they modify the lepton vertices directly. We now combine these effects at fixed electroweak inputs and translate them into partial widths, ratios and asymmetries. This also specifies which measurements carry independent information for the FCC-ee forecast.
Input scheme and aligned-flavour benchmark
We use the measured quantities { α , M W , M Z } \{\alpha, M_W, M_Z\} { α , M W , M Z } to define the electroweak parameters [35 , 36 ] . Holding these inputs fixed is important: a loop correction changes the relation between a measured mass or charge and the parameters in the Lagrangian. The associated parameter shifts contribute to every predicted coupling. The reference parameters obey
s W 2 = 1 − M W 2 M Z 2 , e 2 = 4 π α , v = 2 M W s W e . (74) s_W^2 = 1 - \frac{M_W^2}{M_Z^2}, \qquad e^2 = 4\pi\alpha, \qquad v = \frac{2M_W s_W}{e}. \tag*{(74)} s W 2 = 1 − M Z 2 M W 2 , e 2 = 4 π α , v = e 2 M W s W . (74)
G F G_F G F is predicted, not imposed as an additional input. We take three degenerate HNLs with real, flavour-aligned dipoles C X i a = C X δ i a C_X^{ia} = C_X\delta^{ia} C X ia = C X δ ia , X = B , W X = B,W X = B , W . The bosonic trace has multiplicity n F = 3 n_F = 3 n F = 3 , while the current coefficient of each external lepton has one contribution. In particular,
C H W B u n i v = − 3 e 2 M N 2 8 π 2 s W c W C B C W . (75) C_{HWB}^{\mathrm{univ}} = -3\frac{e^2 M_N^2}{8\pi^2s_Wc_W}C_BC_W. \tag*{(75)} C H W B univ = − 3 8 π 2 s W c W e 2 M N 2 C B C W . (75)
This flavour choice isolates the flavour-conserving electroweak response and avoids introducing additional flavour parameters into the forecast. At this order the aligned matrices obey C X C Y † = C X C Y I 3 C_XC_Y^\dagger= C_XC_Y I_3 C X C Y † = C X C Y I 3 : the response is diagonal and universal. A general flavour texture, and its charged-lepton flavour-violating constraints, require a separate analysis and are not included here.
For comparison, a single HNL with universal real couplings has C X i 1 = C X C_X^{i1}=C_X C X i 1 = C X . It obeys
( C X C Y † ) i i = C X C Y , Tr ( C X C Y † ) = 3 C X C Y , X , Y ∈ { B , W } , (76) (C_XC_Y^\dagger)_{ii}=C_XC_Y,\qquad\operatorname{Tr}(C_XC_Y^\dagger)=3C_XC_Y,\qquad X,Y\in\{B,W\}, \tag*{(76)} ( C X C Y † ) ii = C X C Y , Tr ( C X C Y † ) = 3 C X C Y , X , Y ∈ { B , W } , (76)
just as the three-HNL aligned benchmark. With the same heavy mass and independent local boundary prescription, the two realizations therefore give identical diagonal pole shifts at quadratic order in the dipoles. The one-HNL realization nevertheless has nonzero off-diagonal flavour bilinears. Its flavour-changing observables and allowed parameter space therefore differ from those of the aligned three-HNL benchmark. The common boundary prescription used for the final forecast is specified in § 8 ; it applies consistently to all local terms through the stated order. The matched direct currents and universal input corrections are assembled as in appendix § 12 .§ 8 . Both chiralities of the universal response are included; in particular, δ g e R \delta g_{e_R} δ g e R is not set to zero. No G F G_F G F or W W W -width constraint is added to the pole likelihood.
Linear response of widths and asymmetries
The pole observables measure effective couplings, not Wilson coefficients directly. A partial width measures the sum of squared left- and right-handed couplings, whereas an asymmetry measures their relative weight. Their combination separates changes in normalization from changes in chirality. To compare with precision data, the loop-induced shifts must first be expressed at fixed measured inputs: a correction to a propagator also changes the relation between Lagrangian parameters and those inputs. This is why propagator, vertex and external-residue contributions must be assembled before forming the observable response.
For dimensionless chiral couplings a f L , R = g f L , R / g Z a_{fL,R}=g_{fL,R}/g_Z a f L , R = g f L , R / g Z , define D f = a f L 2 + a f R 2 D_f=a_{fL}^2+a_{fR}^2 D f = a f L 2 + a f R 2 and r f = δ Γ f / Γ f r_f=\delta\Gamma_f/\Gamma_f r f = δ Γ f / Γ f . At linear order in the matched response,
r f = 2 ( a f L δ a f L + a f R δ a f R ) D f , (77) r_f=\frac{2(a_{fL}\delta a_{fL}+a_{fR}\delta a_{fR})}{D_f}, \tag*{(77)} r f = D f 2 ( a f L δ a f L + a f R δ a f R ) , (77)
δ A f = 4 a f L a f R ( a f R δ a f L − a f L δ a f R ) D f 2 . (78) \delta A_f=\frac{4a_{fL}a_{fR}(a_{fR}\delta a_{fL}-a_{fL}\delta a_{fR})}{D_f^2}. \tag*{(78)} δ A f = D f 2 4 a f L a f R ( a f R δ a f L − a f L δ a f R ) . (78)
Universal normalization changes are included in δ a f \delta a_f δ a f . For A F B 0 , f = 3 A e A f / 4 A_{\mathrm{FB}}^{0,f}=3A_eA_f/4 A FB 0 , f = 3 A e A f /4 , the derivative of the forward–backward asymmetry includes both factors: δ A F B 0 , f = 3 ( A f δ A e + A e δ A f ) / 4 \delta A_{\mathrm{FB}}^{0,f}=3(A_f\delta A_e+A_e\delta A_f)/4 δ A FB 0 , f = 3 ( A f δ A e + A e δ A f ) /4 . We evaluate the BSM shifts at the declared reference inputs and add them to the precision-SM prediction. The latter supplies the reference central values and SM uncertainties used in the forecast.
Let f u f_u f u be the SM up-type fraction of the hadronic width, and let B h , B ν , B L B_h,B_\nu,B_L B h , B ν , B L denote hadronic, invisible and total charged-lepton branching fractions. The ratio R ℓ R_\ell R ℓ compares the hadronic and leptonic widths, while R b R_b R b and R c R_c R c are heavy-quark fractions of the hadronic width. The quantity σ h \sigma_h σ h is the hadronic peak cross section; their definitions are given below. At linear order in the induced shifts,
r h = f u r u + ( 1 − f u ) r d , r Z = B h r h + B ν r i n v + B L r ℓ , (79) r_h=f_ur_u+(1-f_u)r_d,\qquad r_Z=B_hr_h+B_\nu r_{\mathrm{inv}}+B_Lr_\ell, \tag*{(79)} r h = f u r u + ( 1 − f u ) r d , r Z = B h r h + B ν r inv + B L r ℓ , (79)
r R ℓ = r h − r ℓ , r σ h = r ℓ + r h − 2 r Z , (80) r_{R_\ell}=r_h-r_\ell,\qquad r_{\sigma_h}=r_\ell+r_h-2r_Z, \tag*{(80)} r R ℓ = r h − r ℓ , r σ h = r ℓ + r h − 2 r Z , (80)
r R b = r d − r h , r R c = r u − r h . (81) r_{R_b}=r_d-r_h,\qquad r_{R_c}=r_u-r_h. \tag*{(81)} r R b = r d − r h , r R c = r u − r h . (81)
These relations allow the same coupling response to be expressed in different observable bases. The likelihood uses one nonredundant basis, with its correlations, so that each measurement contributes once.
Observable content of the likelihood and the role of epsilon parameters
Near the Z Z Z resonance, cross sections and angular distributions are summarized by pole pseudo-observables: widths, ratios and asymmetries extracted after accounting for radiation and nonresonant contributions. Several such quantities share the same measured information, so their correlations are essential when choosing an independent set for a fit. We use the twelve pseudo-observables in § 8.1 , including electron, muon and tau information with the adopted covariance. The defining relations include
σ h 0 = 12 π Γ e Γ h a d M Z 2 Γ Z 2 , R ℓ 0 = Γ h a d Γ ℓ , A F B 0 , f = 3 4 A e A f . (82) \sigma_h^0=\frac{12\pi\Gamma_e\Gamma_{\mathrm{had}}}{M_Z^2\Gamma_Z^2},\qquad R_\ell^0=\frac{\Gamma_{\mathrm{had}}}{\Gamma_\ell},\qquad A_{\mathrm{FB}}^{0,f}=\frac{3}{4}A_eA_f. \tag*{(82)} σ h 0 = M Z 2 Γ Z 2 12 π Γ e Γ had , R ℓ 0 = Γ ℓ Γ had , A FB 0 , f = 4 3 A e A f . (82)
For heavy quarks the ratios instead use the hadronic width in the denominator: R b = Γ b / Γ h a d R_b=\Gamma_b/\Gamma_{\mathrm{had}} R b = Γ b / Γ had and R c = Γ c / Γ h a d R_c=\Gamma_c/\Gamma_{\mathrm{had}} R c = Γ c / Γ had , as in eq. (81) . LEP measurements [37 ] established this pole-observable framework. They also provide a reference for the improvement attainable at FCC-ee, which we quantify in § 8.2 using the same dipole response and UV assumptions for both experiments. The FCC-ee scenario with current SM uncertainties retains the projected FCC-ee experimental precision; it is distinct from the LEP reference. Lepton universality of the signal does not eliminate the experimental correlations between flavours.
Γ Z \Gamma_Z Γ Z includes the hadronic, charged-lepton and invisible widths. The neutrino corrections therefore enter the fit through Γ Z \Gamma_Z Γ Z and the peak cross section. We do not add $\Gamma_{\mathrm{inv}}` or Γ ℓ \Gamma_\ell Γ ℓ as extra independent data derived from the same pole dataset. Likewise, a flavour-averaged leptonic asymmetry would duplicate information already carried by A e A_e A e , A μ A_\mu A μ , A τ A_\tau A τ in the adopted observable vector. The electromagnetic input is the Thomson coupling evolved with SM contributions only, α ˉ Z = α ( 0 ) / ( 1 − Δ α S M ) \bar{\alpha}_Z = \alpha(0)/(1-\Delta\alpha_{\mathrm{SM}}) α ˉ Z = α ( 0 ) / ( 1 − Δ α SM ) ; we use { α ˉ Z , M W , M Z } \{\bar{\alpha}_Z, M_W, M_Z\} { α ˉ Z , M W , M Z } throughout. The dipole contribution to photon propagation remains explicit in the amplitudes. The adopted precision on this input is conditional on controlling additional BSM effects in its experimental extraction.1 No additional G F G_F G F measurement is included in the pole forecast. The measured M W M_W M W constrains the prediction through the input covariance. Charged-current observables such as W W W widths and branching fractions would provide complementary information, but are outside the pole dataset considered here. The joint covariance in eq. (E.1) propagates these input uncertainties without duplicating them.
The ϵ i \epsilon_i ϵ i relations in eq. (72) provide a compact interpretation of the bosonic contribution. The complete dipole response also contains the direct lepton currents C H ℓ ( 1 , 3 ) C_{H\ell}^{(1,3)} C H ℓ ( 1 , 3 ) , so an equivalent epsilon-based analysis would require vertex parameters and a consistently transformed covariance. We instead fit the pole pseudo-observables directly, including both local orders. An epsilon fit derived from the same measurements would re-express this information rather than add an independent dataset.
EFT validity and the heavy-mass expansion
Two scale hierarchies and one perturbative expansion control the interpretation. The hierarchy between electroweak scales and M N M_N M N controls the local approximation; the hierarchy between the HNL and the heavier mediators controls the dipole EFT. Within those descriptions, the number of portal insertions specifies the perturbative order. The expansion used for the pole response requires electroweak momenta and masses to be small relative to the HNL mass. The dipole EFT, in turn, requires the states that generate the portal to lie above the energies and masses being described. A naive estimate of the latter condition, following ref. [1 ] , is obtained by writing
C i = c i Λ i 2 , M N < Λ i ⇒ ∣ C i ∣ < ∣ c i ∣ M N 2 (83) C_i = \frac{c_i}{\Lambda_i^2}, \qquad M_N < \Lambda_i \Rightarrow|C_i| < \frac{|c_i|}{M_N^2} \tag*{(83)} C i = Λ i 2 c i , M N < Λ i ⇒ ∣ C i ∣ < M N 2 ∣ c i ∣ (83)
where c i c_i c i is a dimensionless coupling and Λ i \Lambda_i Λ i is the scale associated with the operator. Setting ∣ c i ∣ = 1 |c_i| = 1 ∣ c i ∣ = 1 gives a reference line against which to compare the inferred coefficients. Its location depends on this coupling assumption; it is a scale diagnostic rather than a unitarity bound. The accuracy of the expansion in electroweak scales over M N M_N M N can be investigated separately using the exact loop functions. A third issue is the truncation in portal insertions: the prediction includes the leading C 2 C^2 C 2 amplitude but not amplitudes with additional dipole insertions. Their size depends on the gauge factors, masses, loop order and independent higher-dimensional coefficients, and cannot be inferred from ∣ C ∣ M N 2 |C|M_N^2 ∣ C ∣ M N 2 alone. In particular, the observable-map linearization test in appendix E does not estimate this portal truncation error. An extra loop with two dipole interactions is expected to be suppressed by a dimensionless factor of order g X 2 ∣ C X ∣ 2 E 4 / ( 16 π 2 ) g_X^2|C_X|^2 E^4/(16\pi^2) g X 2 ∣ C X ∣ 2 E 4 / ( 16 π 2 ) , where E E E is the characteristic energy in the loop. This estimate gives the plausible size of a higher-order correction, but its actual size also depends on the particles and vertices in the diagram. In particular, replacing Higgs fields by their vacuum values replaces some powers of E E E by v v v . A quantitative error estimate therefore requires the omitted amplitudes. This question is separate from whether the couplings of the heavy particles that generate the dipole can be treated perturbatively. Likewise, ∣ c i ∣ > 1 |c_i| > 1 ∣ c i ∣ > 1 by itself is not a breakdown criterion; interpreting such values requires a UV power counting consistent with the assumed hierarchy.
Finite-mass diagnostic
The twice-subtracted two-point remainder R ( t ) = F 0 ( t ) − F 0 ( 0 ) − t F 0 ′ ( 0 ) R(t)=F_0(t)-F_0(0)-tF'_0(0) R ( t ) = F 0 ( t ) − F 0 ( 0 ) − t F 0 ′ ( 0 ) isolates the nonlocal part beyond the constant and slope. Its size at t = M Z 2 t=M_Z^2 t = M Z 2 is
Table 8.
The decrease with M N M_N M N shows how rapidly the two-point function approaches its local form. These ratios isolate one contribution to the full prediction; the corresponding vertex, external-field and input corrections are needed to quantify the error of a complete pole observable. The exact remainder is discussed in appendix C, and the large-mass convergence of the assembled response is tested in appendix C.3.
The same gauge-invariant operators also couple the HNL to the physical Higgs field. In the next section we examine whether radiative Higgs decays supply additional information to the pole observables. Production processes such as W W WW W W and Z H ZH Z H offer further possibilities, discussed later as directions for future studies.
Conditional radiative Higgs decays
We now examine the dipole-induced contributions to h → γ γ h\to\gamma\gamma h → γ γ and h → Z γ h\to Z\gamma h → Z γ in the limit of vanishing light-neutrino masses and mixing. These decays test whether the same gauge-invariant interactions that shift the pole couplings leave an additional signal involving the physical Higgs. Their origin is already contained in the portal Lagrangian: replacing the Higgs vacuum value by the physical field produces the single-Higgs partner of each dipole vertex, with coefficient d V / v d_V/v d V / v . At zero Higgs momentum this is a soft-Higgs insertion: differentiating the vacuum-dependent dipole couplings generates the Higgs vertex, in the spirit of the low-energy theorems of [40 ] . Here the HNL mass is held fixed, so the derivative acts on the dipoles rather than on the heavy mass. The explicit finite-momentum calculation below is needed for the physical decay. Thus the same coefficients probed by pole observables also determine a radiative Higgs signal. We first establish the relation to the local bosonic operators, then include the finite-momentum bubble and its interference with the SM amplitude. A ratio of decay rates isolates the additional information while reducing dependence on Higgs production. The auxiliary two-Higgs projections are described in appendix D.1; the analysis here concerns the neutral two-dipole decay contribution.
Local matching and correlated Higgs couplings
The neutral Higgs expansion H = ( 0 , ( v + h ) / 2 ) T H=(0,(v+h)/\sqrt{2})^T H = ( 0 , ( v + h ) / 2 ) T replaces each three-point dipole by a Higgs-dependent vertex with coefficient d V / v d_V/v d V / v . The local interaction follows by expanding the matched Warsaw operators around the vacuum. We write its coefficients as
Δ L h = h [ c γ γ F μ ν F μ ν + c Z γ Z μ ν F μ ν + c Z Z Z μ ν Z μ ν + c W W W μ ν + W − μ ν ] . (84) \Delta\mathcal{L}_h=h\left[c_{\gamma\gamma}F_{\mu\nu}F^{\mu\nu}+c_{Z\gamma}Z_{\mu\nu}F^{\mu\nu}+c_{ZZ}Z_{\mu\nu}Z^{\mu\nu}+c_{WW}W^+_{\mu\nu}W^{-\mu\nu}\right]. \tag*{(84)} Δ L h = h [ c γ γ F μν F μν + c Z γ Z μν F μν + c Z Z Z μν Z μν + c W W W μν + W − μν ] . (84)
The coefficients c V V ′ c_{VV'} c V V ′ have dimension − 1 -1 − 1 and include the full normalization of the displayed field-strength products, with no implicit factors 1 / 4 1/4 1/4 or 1 / 2 1/2 1/2 . They describe the induced Higgs interactions, whereas the dimension-− 2 -2 − 2 coefficients C B , C W C_B,C_W C B , C W define the portal. At first order in the induced local operators, the neutral rotation gives
c γ γ = v ( c W 2 C H B + s W 2 C H W − s W c W C H W B ) , (85) c_{\gamma\gamma}=v(c_W^2C_{HB}+s_W^2C_{HW}-s_Wc_WC_{HWB}), \tag*{(85)} c γ γ = v ( c W 2 C H B + s W 2 C H W − s W c W C H W B ) , (85)
c Z γ = v [ 2 s W c W ( C H W − C H B ) − ( c W 2 − s W 2 ) C H W B ] , (86) c_{Z\gamma}=v[2s_Wc_W(C_{HW}-C_{HB})-(c_W^2-s_W^2)C_{HWB}], \tag*{(86)} c Z γ = v [ 2 s W c W ( C H W − C H B ) − ( c W 2 − s W 2 ) C H W B ] , (86)
c Z Z = v ( s W 2 C H B + c W 2 C H W + s W c W C H W B ) , (87) c_{ZZ}=v(s_W^2C_{HB}+c_W^2C_{HW}+s_Wc_WC_{HWB}), \tag*{(87)} c Z Z = v ( s W 2 C H B + c W 2 C H W + s W c W C H W B ) , (87)
c W W = 2 v C H W . (88) c_{WW}=2vC_{HW}. \tag*{(88)} c W W = 2 v C H W . (88)
The Wilson coefficients in these equations include the chosen HNL multiplicity. Inserting only their dipole-induced thresholds, for n F n_F n F degenerate aligned HNLs with real couplings, yields
c γ γ = + n F M N 2 4 π 2 v d γ 2 , c Z γ = + n F M N 2 2 π 2 v d γ d Z , (89) c_{\gamma\gamma}=+\frac{n_F M_N^2}{4\pi^2v}d_\gamma^2,\qquad c_{Z\gamma}=+\frac{n_F M_N^2}{2\pi^2v}d_\gamma d_Z, \tag*{(89)} c γ γ = + 4 π 2 v n F M N 2 d γ 2 , c Z γ = + 2 π 2 v n F M N 2 d γ d Z , (89)
c Z Z = + n F M N 2 4 π 2 v d Z 2 , c W W = + n F M N 2 4 π 2 v d W 2 . (90) c_{ZZ}=+\frac{n_F M_N^2}{4\pi^2v}d_Z^2,\qquad c_{WW}=+\frac{n_F M_N^2}{4\pi^2v}d_W^2. \tag*{(90)} c Z Z = + 4 π 2 v n F M N 2 d Z 2 , c W W = + 4 π 2 v n F M N 2 d W 2 . (90)
The neutral local coefficients obey c Z γ 2 = 4 c γ γ c Z Z c_{Z\gamma}^2=4c_{\gamma\gamma}c_{ZZ} c Z γ 2 = 4 c γ γ c Z Z , reflecting the rank-one loop threshold; arbitrary independent UV operators can violate this relation. The coefficients multiplying the displayed field-strength operators are not, by themselves, full h Z Z hZZ h Z Z or h W W hWW hW W amplitudes: input shifts, field normalization and other tensor structures must also be included where relevant.
This result motivates studying h → γ γ h\to\gamma\gamma h → γ γ together with h → Z γ h\to Z\gamma h → Z γ . The former local contribution vanishes on C W = − 2 C B C_W=-2C_B C W = − 2 C B , where d γ = 0 d_\gamma=0 d γ = 0 . The latter also vanishes on C W = 2 ( s W 2 / c W 2 ) C B C_W=2(s_W^2/c_W^2)C_B C W = 2 ( s W 2 / c W 2 ) C B , where d Z = 0 d_Z=0 d Z = 0 and h γ γ h\gamma\gamma hγ γ is generally nonzero. These are zeros of the retained local dipole threshold, not a claim that all Higgs observables become insensitive there. In particular, no sign-definite Z γ Z\gamma Z γ interference follows from this matching alone, since d γ d Z d_\gamma d_Z d γ d Z changes sign between coefficient directions.
For on-shell photons we use e ( 0 ) e(0) e ( 0 ) in d γ d_\gamma d γ and the declared weak-input coupling in d Z d_Z d Z . This convention is also used in the SM radiative amplitudes; a local threshold with a single common coupling must be translated before quoting numerical decay interference.
Interference and perturbative order
With the normalization of eq. (84) , an on-shell CP-even amplitude described by the corresponding total coefficient has
Γ ( h → γ γ ) = m h 3 4 π ∣ c γ γ t o t ∣ 2 , (91) \Gamma(h\to\gamma\gamma)=\frac{m_h^3}{4\pi}\left|c_{\gamma\gamma}^{\mathrm{tot}}\right|^2, \tag*{(91)} Γ ( h → γ γ ) = 4 π m h 3 c γ γ tot 2 , (91)
Γ ( h → Z γ ) = m h 3 8 π ( 1 − M Z 2 m h 2 ) 3 ∣ c Z γ t o t ∣ 2 . (92) \Gamma(h\to Z\gamma)=\frac{m_h^3}{8\pi}\left(1-\frac{M_Z^2}{m_h^2}\right)^3\left|c_{Z\gamma}^{\mathrm{tot}}\right|^2. \tag*{(92)} Γ ( h → Z γ ) = 8 π m h 3 ( 1 − m h 2 M Z 2 ) 3 c Z γ tot 2 . (92)
Here the SM amplitude must be expressed in exactly the same tensor and Lagrangian convention before adding the HNL correction. At the retained order C B , W 2 ∼ Λ − 4 C_{B,W}^2\sim\Lambda^{-4} C B , W 2 ∼ Λ − 4 , its interference gives
δ Γ i Γ i S M = 2 Re ( δ c i c i S M ) , i = γ γ , Z γ . (93) \frac{\delta\Gamma_i}{\Gamma_i^{\mathrm{SM}}}=2\operatorname{Re}\left(\frac{\delta c_i}{c_i^{\mathrm{SM}}}\right),\qquad i=\gamma\gamma,Z\gamma. \tag*{(93)} Γ i SM δ Γ i = 2 Re ( c i SM δ c i ) , i = γ γ , Z γ . (93)
For reference, in the convention above the one-loop SM coefficients are
c γ γ S M = ( − 7.80909 + 0.05273 i ) 10 − 6 G e V − 1 , c_{\gamma\gamma}^{\mathrm{SM}}=(-7.80909+0.05273\,i)\,10^{-6}\ \mathrm{GeV}^{-1}, c γ γ SM = ( − 7.80909 + 0.05273 i ) 1 0 − 6 GeV − 1 ,
c Z γ S M = ( − 2.87048 + 0.00112 i ) 10 − 5 G e V − 1 . (94) c_{Z\gamma}^{\mathrm{SM}}=(-2.87048+0.00112\,i)\,10^{-5}\ \mathrm{GeV}^{-1}. \tag*{(94)} c Z γ SM = ( − 2.87048 + 0.00112 i ) 1 0 − 5 GeV − 1 . (94)
They include W , t , b , c , τ W,t,b,c,\tau W , t , b , c , τ loops.2 At M N = 1 T e V M_N=1\ \mathrm{TeV} M N = 1 TeV , r B = 2 r_B=2 r B = 2 and C B = C W = 0.1 T e V − 2 C_B=C_W=0.1\ \mathrm{TeV}^{-2} C B = C W = 0.1 TeV − 2 , the interference gives ( r γ γ , r Z γ ) = ( − 4.88391 , − 0.72134 ) 10 − 3 (r_{\gamma\gamma},r_{Z\gamma})=(-4.88391,-0.72134)\,10^{-3} ( r γ γ , r Z γ ) = ( − 4.88391 , − 0.72134 ) 1 0 − 3 , where r i = δ Γ i / Γ i S M r_i=\delta\Gamma_i/\Gamma_i^{\mathrm{SM}} r i = δ Γ i / Γ i SM . The squared HNL amplitude is of order Λ − 8 \Lambda^{-8} Λ − 8 and is not included in this truncation. The same-order independent local matching contributions remain separate assumptions, as in the electroweak analysis.
Finite-momentum neutral two-dipole bubble
The finite-momentum decay amplitude is particularly simple because the Higgs attaches to a dipole vertex rather than to a separate neutrino Yukawa interaction. At the other vertex only a gauge boson enters, so its momentum fixes the momentum flowing through the two-point fermion loop. Attaching the Higgs at either end therefore probes the same scalar kernel at the two gauge-boson virtualities. The tensor algebra leading to this result is given in appendix § 14 .§ 2 . For on-shell photons the virtuality is zero, whereas an on-shell Z Z Z probes M Z 2 M_Z^2 M Z 2 . This explains why the two radiative decays have different finite-momentum corrections even though they originate from the same loop function.
Summing the two possible Higgs attachments, with normalization fixed by eq. (90) , gives the on-shell bubble coefficients
δ c γ γ b u b = − n F d γ 2 2 v F 0 r e n ( 0 ) , (95) \delta c_{\gamma\gamma}^{\mathrm{bub}}=-\frac{n_F d_\gamma^2}{2v}F_0^{\mathrm{ren}}(0), \tag*{(95)} δ c γ γ bub = − 2 v n F d γ 2 F 0 ren ( 0 ) , (95)
δ c Z γ b u b = − n F d γ d Z 2 v [ F 0 r e n ( 0 ) + F 0 r e n ( M Z 2 ) ] . (96) \delta c_{Z\gamma}^{\mathrm{bub}}=-\frac{n_F d_\gamma d_Z}{2v}\left[F_0^{\mathrm{ren}}(0)+F_0^{\mathrm{ren}}(M_Z^2)\right]. \tag*{(96)} δ c Z γ bub = − 2 v n F d γ d Z [ F 0 ren ( 0 ) + F 0 ren ( M Z 2 ) ] . (96)
Both reproduce the local threshold. For two on-shell photons the loop coefficient is therefore equal to its local value within this contribution. The Higgs mass enters its tensor kinematics and decay phase space but does not supply a third denominator or an additional scalar-loop invariant in this contact-vertex topology. This property must not be generalized to Higgs triangles with a separate Yukawa vertex or to independent derivative Higgs operators.
Writing z = M Z 2 / M N 2 z=M_Z^2/M_N^2 z = M Z 2 / M N 2 , the remainder is R ( M Z 2 ) = M N 2 r ( z ) / ( 6 π 2 ) R(M_Z^2)=M_N^2r(z)/(6\pi^2) R ( M Z 2 ) = M N 2 r ( z ) / ( 6 π 2 ) . For 0 < z < 1 0<z<1 0 < z < 1 , the function entering the decay amplitude is
r ( z ) = 2 − z − ( 1 − z ) ( 2 + z ) b ( z ) z + 1 − 7 6 z , r(z)=\frac{2-z-(1-z)(2+z)b(z)}{z}+1-\frac{7}{6}z, r ( z ) = z 2 − z − ( 1 − z ) ( 2 + z ) b ( z ) + 1 − 6 7 z ,
b ( z ) = 2 + 1 − z z ln ( 1 − z ) . b(z)=2+\frac{1-z}{z}\ln(1-z). b ( z ) = 2 + z 1 − z ln ( 1 − z ) .
Its derivation and small-z z z expansion are given in appendix § 13 .§ 1 . The Z γ Z\gamma Z γ bubble relative to the local threshold is
δ c Z γ b u b δ c Z γ l o c a l = 1 − z 6 ( ln μ 2 M N 2 + 11 6 ) − r ( z ) 6 . (97) \frac{\delta c_{Z\gamma}^{\mathrm{bub}}}{\delta c_{Z\gamma}^{\mathrm{local}}}
=1-\frac{z}{6}\left(\ln\frac{\mu^2}{M_N^2}+\frac{11}{6}\right)-\frac{r(z)}{6}. \tag*{(97)} δ c Z γ local δ c Z γ bub = 1 − 6 z ( ln M N 2 μ 2 + 6 11 ) − 6 r ( z ) . (97)
This ratio is understood away from a zero of d γ d Z d_\gamma d_Z d γ d Z ; the coefficient itself remains regular and vanishes at that zero. At μ = M N \mu=M_N μ = M N it equals 0.93263, 0.98974 and 0.99745 for M N = 200 M_N=200 M N = 200 , 500 and 1000 GeV, respectively. The corresponding width shifts follow by interference with the SM amplitudes.
The momentum-dependent bubble also produces a UV-divergent local contribution to c Z γ c_{Z\gamma} c Z γ , with residue − n F d γ d Z M Z 2 / ( 12 π 2 v ) -n_Fd_\gamma d_ZM_Z^2/(12\pi^2v) − n F d γ d Z M Z 2 / ( 12 π 2 v ) . Renormalization absorbs this pole into a local Higgs operator. Its finite coefficient describes an additional short-distance contribution, so it must be specified along with the dipole threshold. If its independent renormalized projected contribution is set to zero at Λ \Lambda Λ , the logarithm in eq. (97) is replaced by ln ( Λ 2 / M N 2 ) \ln(\Lambda^2/M_N^2) ln ( Λ 2 / M N 2 ) after combining loop and local running. This is an additional Higgs boundary assumption, not a consequence of setting the Z Z Z -pole local combinations to zero. At 200 GeV the resulting ratios are 0.88460 and 0.82111 for Λ / M N = 2 \Lambda/M_N=2 Λ/ M N = 2 and 5; at 1 TeV they are 0.99553 and 0.99299. Arbitrary independent finite Higgs matching terms can modify these numbers. The photon bubble has no corresponding momentum-dependent UV pole on shell, but independent local h γ γ h\gamma\gamma hγ γ contributions remain possible.
For HNL masses above the Higgs mass the channel h → N ν h\to N\nu h → N ν is kinematically closed.
Production-independent decay contrast
The SM and direct dipolar radiative amplitudes both start at one loop. We include 2 Re ( A S M ( 1 ) ∗ A d i p ( 1 ) ) 2\operatorname{Re}(A_{\mathrm{SM}}^{(1)*} A_{\mathrm{dip}}^{(1)}) 2 Re ( A SM ( 1 ) ∗ A dip ( 1 ) ) . A dipole-induced one-loop input or external-residue correction multiplying A S M ( 1 ) A_{\mathrm{SM}}^{(1)} A SM ( 1 ) is a two-loop amplitude and is not selectively added. This counting does not apply automatically to independent UV operators generated at tree level.
For a fixed production category p p p , the linear signal-strength response is δ μ p i = n p + r i \delta\mu_{pi}=n_p+r_i δ μ p i = n p + r i , where n p n_p n p absorbs production and the total width. Profiling one free n p n_p n p per category leaves the contrast
R H = Γ ( h → Z γ ) Γ ( h → γ γ ) , δ R H R H = r Z γ − r γ γ . (98) R_{\mathrm{H}}=\frac{\Gamma(h\to Z\gamma)}{\Gamma(h\to\gamma\gamma)},\qquad\frac{\delta R_{\mathrm{H}}}{R_{\mathrm{H}}}=r_{Z\gamma}-r_{\gamma\gamma}. \tag*{(98)} R H = Γ ( h → γ γ ) Γ ( h → Z γ ) , R H δ R H = r Z γ − r γ γ . (98)
This construction uses consistently defined, unfolded production categories; it does not identify experimental acceptances between decays. The projected fractional errors on ( σ p B γ γ , σ p B Z γ ) (\sigma_p B_{\gamma\gamma},\sigma_p B_{Z\gamma}) ( σ p B γ γ , σ p B Z γ ) are ( 3.6 , 11.8 ) % (3.6,11.8)\% ( 3.6 , 11.8 ) % for Z H ZH Z H at 240 GeV, ( 13 , 22 ) % (13,22)\% ( 13 , 22 ) % for Z H ZH Z H at 365 GeV and ( 15 , 23 ) % (15,23)\% ( 15 , 23 ) % for fusion at 365 GeV [41 ] . These are rate errors, not errors on absolute partial widths. With independent experimental categories and channels, the combined contrast error is 10.30%. Within-category correlations ρ = ± 0.5 \rho=\pm0.5 ρ = ± 0.5 , considered only as sensitivity tests, change it to 8.37–11.86%; these correlations are not provided by the source.
The combined forecast retains the twelve-observable Z Z Z covariance and adds the decay contrast with its shared SM-input correlations. The contrast derivatives use the six input directions of the source covariance, with the G F → α G_F\to\alpha G F → α conversion applied jointly to the pole and decay observables. This includes the common input uncertainties once. Intrinsic decay uncertainties are common across production categories and independent between the two decay channels. We adopt the upper-end conservative estimates of 1% for γ γ \gamma\gamma γ γ and 1.5% for Z γ Z\gamma Z γ from ref. [42 ] , table 3.3. The aggressive scenario uses 0.4% for γ γ \gamma\gamma γ γ and treats the intrinsic Z γ Z\gamma Z γ theory error as negligible, since the source gives no aggressive projection for that channel. This is an assumption about the future precision, not a statement that the theory error vanishes. The resulting contrast errors are 10.46% (conservative) and 10.31% (aggressive). Retaining instead a 1.5% intrinsic Z γ Z\gamma Z γ error in the aggressive scenario gives 10.42%. The experimental uncertainty dominates in all three cases, so this choice has little impact on the additional information supplied by the Higgs decays.
The SM derivatives are evaluated at one loop with W , t , b , c , τ W,t,b,c,\tau W , t , b , c , τ , holding light masses fixed. The resulting comparison uses a leading response together with a projected SM uncertainty budget and the additional boundary specified above for independent radiative coefficients. A precision-Higgs fit would additionally require higher-order α s \alpha_s α s derivatives and a correspondingly improved amplitude calculation.
Table 1 shows that this complement is modest when future SM theory precision at the pole is achieved. The shared-input correlations with the contrast have magnitude below 4 × 10 − 4 4\times10^{-4} 4 × 1 0 − 4 in the stated covariance model. Both direct radiative contributions vanish for C W = − 2 C B C_W=-2C_B C W = − 2 C B ; the single contrast has a further cancellation direction. Independent finite h γ γ h_{\gamma\gamma} h γ γ and h Z γ h_{Z\gamma} h Z γ terms can absorb its signal. It consequently does not remove the unrestricted UV degeneracy. We therefore present the pole reach separately in the following section, with Table 1 quantifying the additional information supplied by the decay contrast.
Table 1. Largest relative reduction of the conditional coefficient-region radius in an angular scan at M N = 1 M_N = 1 M N = 1 TeV after adding the decay contrast. The pole and combined results use the same joint 95% C.L. construction. The entries give the largest improvement over the scanned coupling directions.
FCC-ee reach and its physical interpretation
The precision programme of FCC-ee can test dipole interactions through their virtual effects, including masses beyond its direct-production reach. We now quantify this reach and identify the assumptions needed to interpret it. The central issue is that the dipoles and additional short-distance interactions can change the same measured couplings. We therefore present the sensitivity under dipole dominance first, then examine how it changes when additional contributions are allowed, and finally compare it with direct HNL searches.
Measurements and benchmark assumptions
Our forecast combines the total Z Z Z width, the hadronic peak cross section, leptonic ratios and asymmetries, and heavy-quark ratios and asymmetries:
( Γ Z , σ h 0 , R e , R μ , R τ , A e , A μ , A τ , R b , R c , A b , A c ) . (99) (\Gamma_Z, \sigma_h^0, R_e, R_\mu, R_\tau, A_e, A_\mu, A_\tau, R_b, R_c, A_b, A_c). \tag*{(99)} ( Γ Z , σ h 0 , R e , R μ , R τ , A e , A μ , A τ , R b , R c , A b , A c ) . (99)
Together, these measurements distinguish changes in the overall decay rate from changes in the relative strengths of left- and right-handed couplings. All three charged-lepton flavours contribute. We use the experimental projections and correlated SM uncertainties supplied with ref. [43 ] , in the input scheme introduced in § 6 .
The theoretical uncertainty of the SM prediction is important because the expected experimental errors are very small. We compare the current SM uncertainties with conservative and aggressive projections for future calculations. In ref. [43 ] , the conservative scenario assumes moderate improvements achievable by extending existing higher-order techniques; the aggressive scenario assumes larger gains requiring advances in those methods. Both include uncertainties in SM predictions and in the theory input needed to extract pole observables. The experimental-only result shows the sensitivity that would be obtained if these SM uncertainties were negligible. Correlations between observables and shared inputs are included throughout; the covariance conversion is detailed in appendix § 15 . Changing the normalization of the electromagnetic input coordinate shifts the quoted coefficient reaches at 1 TeV by less than 0.36%.
To specify dipole dominance quantitatively, we set the independent local UV coefficients to zero at a common scale Λ B \Lambda_B Λ B and evolve them consistently to the HNL threshold. We consider r B = Λ B / M N = 2 , 5 r_B = \Lambda_B/M_N = 2, 5 r B = Λ B / M N = 2 , 5 . The choice of r B r_B r B specifies the interval over which the independent coefficients evolve before the HNL is integrated out. The two benchmarks illustrate this UV dependence, with a wider range discussed in appendix § 15 ; neither r B = 2 r_B = 2 r B = 2 nor the bands between the two curves define an EFT validity boundary or a statistical error. The HNL flavour assignment and the local expansion are those established in § 6 .
We quote joint 95% C.L. regions for an expected dataset centred on the SM. The leading shifts depend on three combinations, C B 2 C_B^2 C B 2 , C B C W C_B C_W C B C W and C W 2 C_W^2 C W 2 . We first construct a joint confidence region for their linear response, then select the points obtainable from two real dipole coefficients. This avoids assuming a nonzero linear dependence on C B , C W C_B,C_W C B , C W at the SM point, where that dependence vanishes. For the fixed-UV benchmark this gives the threshold χ 3 2 ( 0.95 ) ≃ 7.81 \chi^2_3(0.95) \simeq7.81 χ 3 2 ( 0.95 ) ≃ 7.81 . Appendix § 15 explains the construction and its coverage. Axis and diagonal values quoted below are sections through these joint regions.
Sensitivity under dipole dominance
The projected reach is illustrated in Figure 4 and quantified at M N = 1 M_N = 1 M N = 1 TeV in Table 2 . The contours show that FCC-ee is sensitive to both portal coefficients, with different strengths along the axes and the two diagonals. Their shape reflects the relative weights of the hypercharge and weak interactions and the mixed contribution proportional to C B C W C_B C_W C B C W .
Table 2. Sections of conditional joint 95% C.L. regions at M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV , using χ 3 2 ( 0.95 ) ≃ 7.81 \chi^2_3(0.95) \simeq7.81 χ 3 2 ( 0.95 ) ≃ 7.81 , with the common finite boundary and the covariance prescriptions of appendix § 15 . Units are T e V − 2 \mathrm{TeV}^{-2} TeV − 2 ; on each diagonal the entry is ∣ C B ∣ = ∣ C W ∣ \lvert C_B\rvert= \lvert C_W\rvert ∣ C B ∣ = ∣ C W ∣ . All entries are expected sensitivities for SM-centred data. The LEP reference uses four flavour-universal pole observables with their experimental, input and SM theory covariance; the remaining rows use the twelve-observable FCC-ee projection. If Λ U V = Λ B \Lambda_{\mathrm{UV}} = \Lambda_B Λ UV = Λ B , the aggressive C B C_B C B -axis entry at r B = 5 r_B = 5 r B = 5 corresponds to ∣ c B ∣ = ∣ C B ∣ Λ B 2 ≃ 6.85 \lvert c_B\rvert= \lvert C_B\rvert\Lambda_B^2 \simeq6.85 ∣ c B ∣ = ∣ C B ∣ Λ B 2 ≃ 6.85 ; its realization with perturbative loop-generated dipoles requires additional UV assumptions (appendix § 16 .§ 2 ).
Figure 4. Conditional expected 95% C.L. regions with conservative and aggressive SM uncertainty forecasts. Solid and dashed lines denote the two common UV boundary hypotheses.
For C B = C W C_B = C_W C B = C W and r B = 2 r_B = 2 r B = 2 , the reach with current SM uncertainties is 0.5044 T e V − 2 0.5044\ \mathrm{TeV}^{-2} 0.5044 TeV − 2 . It improves to 0.2201 T e V − 2 0.2201\ \mathrm{TeV}^{-2} 0.2201 TeV − 2 for the conservative scenario or 0.1431 T e V − 2 0.1431\ \mathrm{TeV}^{-2} 0.1431 TeV − 2 for the aggressive scenario. The corresponding experimental-only value is 0.0650 T e V − 2 0.0650\ \mathrm{TeV}^{-2} 0.0650 TeV − 2 . The gap between these results makes the physical role of improved SM calculations clear: the experimental sample alone does not determine the attainable dipole sensitivity. The conservative forecast probes coefficients well below 1 T e V − 2 1\ \mathrm{TeV}^{-2} 1 TeV − 2 in several directions. For a uniform rescaling of standard deviations, σ → a σ \sigma\to a\sigma σ → aσ with correlations fixed, the quadratic response gives C r e a c h → a C r e a c h C_{\mathrm{reach}} \to\sqrt{a}C_{\mathrm{reach}} C reach → a C reach . A tenfold reduction of all errors would thus improve the coefficient reach by 10 ≃ 3.16 \sqrt{10} \simeq3.16 10 ≃ 3.16 . The fourth-root scaling applies to the variance, or to inverse luminosity when statistical errors dominate. The actual SM scenarios change different error components by different amounts.
LEP provides a useful measure of the gain in sensitivity. We use the flavour-universal combination ( Γ Z , σ h 0 , R ℓ , A F B 0 , ℓ ) (\Gamma_Z, \sigma^0_h, R_\ell, A^{0,\ell}_{\mathrm{FB}}) ( Γ Z , σ h 0 , R ℓ , A FB 0 , ℓ ) , including the luminosity update and the correlated input and SM theory uncertainties supplied with refs. [38 , 44 ] . To isolate the improvement in precision, we centre both datasets on the SM and apply the same finite dipole response, UV boundary and confidence construction. The LEP entries in Table 2 are therefore expected sensitivities at LEP precision, rather than observed exclusion limits or a global LEP/SLC fit. The covariance prescription is given in appendix § 15 .
At M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV and r B = 2 r_B = 2 r B = 2 , LEP precision probes equal, same-sign coefficients at about 0.92 T e V − 2 0.92\ \mathrm{TeV}^{-2} 0.92 TeV − 2 . FCC-ee lowers this to 0.22 0.22 0.22 or 0.14 T e V − 2 0.14\ \mathrm{TeV}^{-2} 0.14 TeV − 2 , an improvement by factors of 4.2 and 6.4 in the conservative and aggressive SM scenarios. Across the two axes and two diagonals, and for both boundary choices, the gains range from 3.8–4.3 and 5.4–6.5, respectively. These factors compare the four-observable LEP reference with the FCC-ee observable set adopted here, including its separate lepton-flavour and heavy-quark measurements. They quantify the smaller dipole coefficients that can be tested under the same UV hypothesis; the larger LEP values do not, by themselves, establish perturbative realizations at that reach.
With current SM uncertainties, the improvement is only a factor of 1.4–2.2. The substantially greater reach of the future-theory scenarios therefore requires progress in SM predictions as well as the larger experimental sample. In this sense, FCC-ee extends the precision test established at LEP into a region of appreciably weaker dipole interactions, while retaining its dependence on the independent UV contributions discussed below.
Changing the boundary from r B = 2 r_B = 2 r B = 2 to r B = 5 r_B = 5 r B = 5 strengthens the same-sign reach to 0.1614 T e V − 2 0.1614\ \mathrm{TeV}^{-2} 0.1614 TeV − 2 and 0.1069 T e V − 2 0.1069\ \mathrm{TeV}^{-2} 0.1069 TeV − 2 in the conservative and aggressive scenarios, respectively. The boundary affects the finite local coefficients through their running, and hence changes the observable response. We display both choices to expose this theoretical assumption instead of treating one as a unique prediction of a UV completion.
Figure 5 shows how the axis reach evolves with the HNL mass. At fixed dimensionful dipole coefficients the dominant threshold grows as M N 2 C X C Y M_N^2 C_X C_Y M N 2 C X C Y , so the coefficient reach improves approximately as 1 / M N 1/M_N 1/ M N . This is the origin of the favourable mass dependence of the indirect probe. A family of UV models in which the mediator scale grows with M N M_N M N and dimensionless couplings stay fixed instead has a decoupling response, as discussed in appendix § 16 .
Figure 5. Expected 95% C.L. sensitivity bands on the Wilson coefficients C B C_B C B and C W C_W C W as a function of the HNL mass, along the two coefficient axes. Shading connects the two UV boundary hypotheses r B = 2 , 5 r_B=2,5 r B = 2 , 5 . The lower mass endpoint is a presentation choice; the accuracy of the local expansion is discussed in § 6.4 .
The mass scan starts at 500 GeV, where the hierarchy between the HNL mass and electroweak scales is larger than near threshold. Its interpretation at smaller masses requires the finite-mass qualifications in § 6.4 . The radiative Higgs-decay observable of § 7 adds a modest further improvement: at M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV and r B = 2 r_B = 2 r B = 2 , the largest reductions of the coefficient-region radius are about 7% and 1% in the conservative and aggressive scenarios. The contours displayed here therefore present the pole reach, with the separate Higgs contribution quantified in Table 1 .
What changes when other ultraviolet interactions are allowed?
The dipole coefficients predict correlated shifts across the pole observables. However, other particles responsible for generating the portal can also modify the lepton currents or gauge-boson propagation. A precision measurement constrains the sum of these effects. To determine how much of the dipole sensitivity survives, we allow selected additional coefficients to vary and choose their values to give the best agreement with the expected SM data. This is the profiling procedure used below.
Table 3 shows the progression. At M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV , r B = 2 r_B = 2 r B = 2 and with aggressive SM uncertainties, allowing the two independent lepton currents to vary moves the same-sign section from 0.1431 0.1431 0.1431 to 0.2208 T e V − 2 0.2208\,\mathrm{TeV}^{-2} 0.2208 TeV − 2 . Some sensitivity remains because these currents alone cannot reproduce every bosonic effect of the dipoles. Allowing the mixed bosonic coefficient C H W B C_{HWB} C H W B as well removes the bound on the C W C_W C W axis in this local analysis.
Table 3. Effect of profiling independent local contributions at M N = 1 T e V M_N=1\,\mathrm{TeV} M N = 1 TeV , r B = 2 r_B=2 r B = 2 , with aggressive SM uncertainties. The same-sign entries denote ∣ C B ∣ = ∣ C W ∣ |C_B|=|C_W| ∣ C B ∣ = ∣ C W ∣ ; the confidence threshold is recomputed for the surviving response rank. The rank-zero statement allows arbitrary independent UV coefficients in the linearized pole response; it does not establish a perturbative completion realizing an arbitrarily large cancellation.
The disappearance of sensitivity has a direct coupling interpretation. At the fixed pole, all local shifts are described by two universal changes, proportional to weak isospin and electric charge, and two direct changes of the left-handed neutrino and charged-lepton currents. These are the four combinations derived in appendix § 12 .§ 8 . This span can already be realized with C H l ( 1 ) C_{Hl}^{(1)} C H l ( 1 ) , C H l ( 3 ) C_{Hl}^{(3)} C H l ( 3 ) , C H W B C_{HWB} C H W B , C H D C_{HD} C H D at dimension six. Independent local UV operators span all four. When all are free, they can compensate the entire dipole response at the pole and no finite bound on either dipole remains. Additional measurements of the same pole couplings improve their precision but do not identify which interaction produced them. This four-dimensional space of possible coupling shifts is distinct from the three-dimensional dipole signal space used in the statistical construction. The dipole model explores only the patterns generated by its three quadratic monomials; arbitrary UV operators span the larger space and can reproduce those patterns.
This result defines the scope of the indirect reach. The numbers in Table 2 apply when the dipole contribution dominates, and the intermediate cases quantify how specific relaxations weaken that assumption. A dipole interpretation can become more restrictive in a specified UV model, where symmetries and coupling relations correlate the extra contributions rather than leaving them arbitrary.
Ultraviolet models and dipole dominance
The possibility of a dominant dipole signal in a UV completion depends on its particle content, couplings and symmetry structure. A loop-generated dipole typically comes with other effective operators; some can arise already at tree level. The benchmark reach applies when these additional contributions are sufficiently suppressed or constrained by the model. The power counting and examples used for this comparison are given in appendix § 16 .
The UV model of ref. [6 ] contains two colour-singlet vector-like fermions, X E X_E X E (a weak doublet with hypercharge 1 / 2 1/2 1/2 ) and X N X_N X N (a weak singlet with hypercharge 1 1 1 ), and a singly charged scalar ϕ \phi ϕ . All three are odd under a Z 2 Z_2 Z 2 symmetry, while the SM fields and N N N are even. Its photon dipole is proportional to a product of three couplings. A coefficient ∣ C B + C W / 2 ∣ = 0.1 T e V − 2 \left|C_B + C_W/2\right| = 0.1\,\mathrm{TeV}^{-2} ∣ C B + C W /2 ∣ = 0.1 TeV − 2 with mediator masses of 2 2 2 TeV requires a product of about 7.6 × 10 2 7.6 \times10^2 7.6 × 1 0 2 , corresponding to a geometric mean of 9.1 9.1 9.1 . This estimate is already outside a comfortably weakly coupled regime. It illustrates why moving the mediators to a much higher scale does not readily isolate a large loop-generated dipole: preserving the dipole requires increasing the underlying couplings. The example is a scale comparison, not a full constraint on that completion.
The assumption of negligible active–sterile mixing has a related naturalness question. Dipole running can generate a Higgs–lepton–HNL operator, while two Majorana dipoles can induce light-neutrino masses. Near the projected reach, a viable realization therefore needs a mechanism that protects the neutrino mass or compensates the induced contribution. The size of the hierarchy is illustrated by the cutoff estimate in appendix § 16 : for ∣ d γ ∣ = 10 − 5 G e V − 1 \left|d_\gamma\right| = 10^{-5}\,\mathrm{GeV}^{-1} ∣ d γ ∣ = 1 0 − 5 GeV − 1 , M N = 1 M_N = 1 M N = 1 TeV and a 2 2 2 -TeV cutoff, the induced mass is of order 10 10 10 MeV, far above the observed neutrino mass scale. Although this estimate is not a renormalized mass prediction, it makes the need for protection apparent. Such protection does not, on its own, constrain the size of the independent electroweak currents. The quantitative estimates and relevant UV constructions are discussed in the same appendix.
Thus our forecast supplies a target for UV models with controlled additional electroweak effects. We do not identify the zero-boundary benchmark with a generic weakly coupled completion. Its usefulness is that it specifies the dipole contribution and the accuracy with which other effects must be understood to interpret a precision constraint.
How robust is the reach against extra contributions?
Complete freedom in the UV coefficients is the most general case. A model with specified interactions usually predicts a correlated extra response, which can strengthen the signal as well as compensate it. The relevant question is whether that response resembles the dipole pattern across the measured observables, with a reinforcing or opposing sign. A sign comparison of individual Wilson coefficients is insufficient because the measurements weight combinations of couplings.
We compare the full observable responses using their experimental and SM covariance. An extra contribution aligned with the dipole signal strengthens the expected constraint. A sizeable reduction requires both an opposing pattern and a comparable magnitude. For example, if the extra response is at most 20 % 20\% 20% or 50 % 50\% 50% of the dipole response at every coefficient value, the largest possible increases of the coefficient limit are approximately 12 % 12\% 12% and 41 % 41\% 41% , respectively, at the same statistical threshold. These are assumptions about the relative observable shifts; they are not probabilities for a cancellation. The distinction between such relative budgets and fixed UV contributions is derived in appendix § 16 .
Published vector-like-lepton matching gives concrete examples of both behaviours. At M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV , with the aggressive covariance and r B = 2 , 5 r_B = 2, 5 r B = 2 , 5 , nonzero contributions with the charged-singlet E E E and triplet Σ 1 \Sigma_1 Σ 1 patterns reinforce the signal on the tested axes and diagonals. Allowing their strengths to vary freely down to zero therefore cannot weaken the dipole-only constraint in these cases. By contrast, the Σ 0 \Sigma_0 Σ 0 pattern can weaken the same-sign reach by a factor of about 1.5 1.5 1.5 –2.2 2.2 2.2 when the original threshold is held fixed. Some axis directions are nearly degenerate in these restricted examples, so the residual sensitivity there depends strongly on effects outside the test. The representations, current patterns and statistical comparison are specified in appendix § 16 .
The examples demonstrate that the effect of additional UV interactions is controlled by their representation and coupling correlations. Symmetry and mediator quantum numbers can select an opposing direction, while other representations reinforce the dipole signal. Establishing robustness for a particular completion requires its correlated matching. Keeping only the dipole logarithm does not provide an independent lower floor: at a fixed pole momentum, independent local coefficients can compensate the sum of logarithmic and finite terms.
Complementarity with direct HL-LHC searches
Direct production and electroweak precision measurements test different aspects of the portal. The former seeks an on-shell HNL and its decay products; the latter measures the effect of virtual HNLs on ordinary particle couplings. Figure 6 compares the FCC-ee reach with the same-sign dilepton projections of ref. [1 ] for its 13-TeV, 3 a b − 1 3\ \mathrm{ab}^{-1} 3 ab − 1 HL-LHC benchmark. The three panels use that study’s coupling directions A–C, with a common operator normalization. The conversion and flavour correspondence are detailed in appendix § 17 .§ 1 .
Figure 6. Direct and indirect reach in the three benchmark directions. Purple points are the HL-LHC projections shown in figure 6 of ref. [1 ] . FCC-ee bands connect the conditional boundaries Λ B / M N = 2 , 5 \Lambda_B/M_N = 2, 5 Λ B / M N = 2 , 5 for each SM uncertainty scenario; they are not probability bands. Dashed coloured curves below 500 GeV flag local-expansion diagnostics, not a physical threshold. The grey dash-dotted curve marks Λ e f f = M N \Lambda_{\mathrm{eff}} = M_N Λ eff = M N for unit dimensionless dipole coefficients.
Reference [1 ] also presents direct-search projections for FCC-hh and a 10-TeV muon collider. We focus here on HL-LHC as the nearer-term comparison: its data are expected to precede FCC-ee operation. FCC-hh and a high-energy muon collider represent longer-term opportunities with less certain realization timescales, especially for the latter. Their projected reach therefore addresses a different experimental horizon from the comparison made here.
At M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV , the direct sensitivities in A, B and C are ( 7.8 , 10.3 , 51.7 ) × 10 − 6 G e V − 1 (7.8, 10.3, 51.7) \times10^{-6}\ \mathrm{GeV}^{-1} ( 7.8 , 10.3 , 51.7 ) × 1 0 − 6 GeV − 1 in ∣ d γ ∣ \lvert d_\gamma\rvert ∣ d γ ∣ . For the conservative FCC-ee scenario and r B = 2 r_B = 2 r B = 2 , the corresponding sections are ( 7.6 , 7.2 , 22.3 ) × 10 − 6 G e V − 1 (7.6, 7.2, 22.3) \times10^{-6}\ \mathrm{GeV}^{-1} ( 7.6 , 7.2 , 22.3 ) × 1 0 − 6 GeV − 1 . Under the specified UV boundary, the conservative indirect reach is therefore comparable in A and probes coefficients smaller by factors of 1.4 and 2.3 in B and C, respectively. Changing to r B = 5 r_B = 5 r B = 5 gives ( 4.8 , 5.1 , 17.5 ) × 10 − 6 G e V − 1 (4.8, 5.1, 17.5) \times10^{-6}\ \mathrm{GeV}^{-1} ( 4.8 , 5.1 , 17.5 ) × 1 0 − 6 GeV − 1 .3
The mass dependence gives a second source of complementarity. The virtual threshold grows with the HNL mass at fixed dipole coefficients, whereas direct production depends on kinematic accessibility and the production and decay rates. The figure displays the published direct mass interval alongside the indirect curves; its direct endpoints do not define an upper mass reach. The dashed indirect curves at lower masses mark the local-expansion qualification, while the grey guide indicates the mediator-scale estimate for unit dimensionless dipoles. These two features describe different limitations of the EFT and are explained in appendix § 17 .§ 1 .
The complementarity also concerns discovery. A pole deviation can be produced by several UV interactions, whereas on-shell HNL production and measurements of its decay channels would help establish a dipole origin. The comparison here is at matched diagonal flavour response, not a combined likelihood or a flavour-violation analysis. The same-sign dilepton channel used here is particularly favourable for Majorana HNLs because it suppresses the SM background [1 ] . A lepton-number-conserving Dirac HNL does not produce
this signal, but can be sought through opposite-sign leptons or radiative decays. For the latter, the HL-LHC sensitivity is estimated at ∣ d γ ∣ ∼ 3 × 10 − 5 G e V − 1 |d_\gamma| \sim3 \times10^{-5}\ \mathrm{GeV}^{-1} ∣ d γ ∣ ∼ 3 × 1 0 − 5 GeV − 1 in the d W = 0 d_W = 0 d W = 0 benchmark, using N → γ ν N \to\gamma\nu N → γ ν rather than same-sign dileptons [1 , 10 ] . This estimate comes from extrapolating the LHC constraints of ref. [2 ] ; it gives a quantitative reference for a search that does not require lepton-number violation. Although these estimates refer to different coupling directions, the stronger same-sign reach in A and B illustrates the potential advantage of a low-background signature. The radiative channel remains relevant when that signature is absent. By contrast, the lepton-number-conserving virtual response is unchanged when the coupling bilinears are matched as in § 2.1 . The indirect probe thus retains its sensitivity without relying on the low-background same-sign signature.
Information beyond the pole observables
Measurements at different energies and with different external states can distinguish effects that coincide at the Z Z Z pole. Off-pole fermion production probes the momentum dependence, while W W W -pair production also probes charged currents and triple-gauge interactions. In the present model, the latter has an additional tree-level HNL exchange through the charged dipole, with characteristic angular dependence. These channels provide well-motivated extensions of the analysis.
Their interpretation still depends on which independent UV operators are allowed. At any fixed order in a local momentum expansion, free coefficients with the same structures can reproduce that polynomial dependence. Exact mass dependence, additional correlations from a UV model, or new processes can provide further information. The necessary amplitude content and the limits of the local argument are outlined in appendix § 17 .§ 3 . We leave quantitative off-pole and W W W -pair sensitivities to a dedicated study, keeping the present results focused on the pole observables and the radiative Higgs decays.
Conclusions
Electroweak precision measurements can probe the Majorana HNL dipole portal without requiring direct production of the heavy state. We have determined how virtual HNLs modify gauge-boson propagation and lepton couplings, and how these effects appear in the widths and asymmetries measured at FCC-ee. For real, universal, flavour-aligned dipoles, the analysis gives both a quantitative sensitivity target and the conditions under which that target can be interpreted as a constraint on the portal.
The electroweak completion of the dipole interaction is essential to this prediction. Gauge symmetry relates the neutral and charged couplings and requires the two-gauge-boson contact vertices that enter the lepton-current corrections. The same interaction therefore predicts correlated effects in several precision observables and in radiative Higgs decays. Our calculation includes both the leading mass-enhanced correction and the first subleading terms projected onto the pole couplings, with a common treatment of the gauge-boson and lepton contributions.
Under dipole dominance, FCC-ee can probe equal, same-sign coefficients at approximately 0.22 T e V − 2 0.22\ \mathrm{TeV}^{-2} 0.22 TeV − 2 with conservative SM uncertainties and 0.14 T e V − 2 0.14\ \mathrm{TeV}^{-2} 0.14 TeV − 2 with the more ambitious theory projection, for M N = 1 T e V M_N = 1\ \mathrm{TeV} M N = 1 TeV . These values assume that additional local contributions vanish at a common scale twice the HNL mass (Λ B / M N = 2 \Lambda_B/M_N = 2 Λ B / M N = 2 ). Choosing that boundary at five times the HNL mass gives approximately 0.16 0.16 0.16 and 0.11 T e V − 2 0.11\ \mathrm{TeV}^{-2} 0.11 TeV − 2 , respectively. The two variations have different meanings: improved SM predictions increase the experimental programme’s sensitivity to the portal, whereas moving the boundary scale changes the assumed contribution of physics above the HNL mass. Keeping them distinct is necessary when comparing these forecasts with a particular model. Relative to the fourobservable LEP reference evaluated with the same assumptions, the conservative FCC-ee programme tests coefficients about four times smaller; the more ambitious SM projection increases this gain to roughly five–six. Improved SM predictions are essential to realizing this advance.
Larger HNL masses allow smaller dipole coefficients to be probed, with an approximate 1 / M N 1/M_N 1/ M N reach driven by the leading correction proportional to M N 2 C X C Y M_N^2 C_X C_Y M N 2 C X C Y . This mass enhancement describes the response at fixed dipole coefficients, rather than a variation with all parameters of a UV completion held fixed. If the particles that generate the dipoles become heavier together with the HNL, at fixed mass ratios and dimensionless couplings, the dipole coefficients decrease as the inverse mediator mass squared. Their suppression then outweighs the factor M N 2 M_N^2 M N 2 , and the precision effect decreases. Near the electroweak scale, the full mass dependence of both gauge-boson and lepton-current corrections becomes relevant. In the benchmark directions tested here, it changes the coefficient reach by at most 0.15% at 500 GeV and 4.3% at 200 GeV.
The main qualification concerns what pole measurements can identify. At the one-loop order retained here, the local dipole corrections enter four independent combinations of the couplings measured at the Z Z Z pole. Other heavy particles can induce exactly the same four changes through local interactions, already at dimension six. Pole data therefore constrain the sum of these effects and cannot, on their own, establish a dipole origin. If the additional contributions are unrestricted, they can completely cancel the dipole signal. In specific UV models, however, their relative sizes and signs can be correlated by the mediator representations and interactions. The vector-like-lepton examples considered here illustrate the consequence: a charged-singlet contribution reinforces the signal along the tested directions, whereas a weak-triplet contribution can partly compensate it. A large reduction requires an opposing pattern of comparable size across the measured quantities, rather than simply an opposite sign in one Wilson coefficient. The forecast is thus a target for models in which these additional effects are controlled; its realization must also respect perturbative power counting and the smallness of neutrino masses.
Radiative Higgs decays add a modest complementary constraint. At M N = 1 M_N = 1 M N = 1 TeV and Λ B / M N = 2 \Lambda_B/M_N = 2 Λ B / M N = 2 , the ratio of the Z γ Z\gamma Z γ and γ γ \gamma\gamma γ γ widths improves the coefficient-region radius by at most about 7% and 1% in the conservative and aggressive scenarios. The pole measurements therefore provide most of the projected sensitivity under the stated assumptions. Measurements away from the pole and W W W -pair production could supply different information, but their reach requires a dedicated treatment of the additional amplitudes and UV coefficients.
Direct searches provide a particularly useful counterpart. At an HNL mass of 1 TeV, the conservative FCC-ee sensitivities are comparable to or stronger than the HL-LHC projections of ref. [1 ] in the three benchmark directions, under the stated dipole-dominance assumption. Their different mass dependence makes the comparison useful beyond this single benchmark. More fundamentally, observing an HNL and measuring its decay pattern would help distinguish a dipole interaction from other sources of a precision deviation. The HL-LHC comparison uses the favourable Majorana same-sign dilepton channel, whose small SM background enhances the direct reach. A lepton-number-conserving Dirac HNL must instead be sought in other final states, including radiative decays. The available radiative projection is weaker than the same-sign reach in A and B, but uses a different coupling direction; it does not establish a universal loss of direct sensitivity for Dirac HNLs. The indirect response remains unchanged for matched coupling bilinears, preserving the role of precision measurements when the same-sign signature is absent.
Combining direct production with precision measurements would test both the heavy state and the correlated virtual effects of its interactions. Our results identify the correlated electroweak signature of the portal, quantify the precision needed to test it, and specify which additional interactions can obscure its interpretation. Together with direct searches, these predictions provide a framework for testing whether a future HNL signal is accompanied by the virtual effects expected from a gauge-invariant dipole interaction.
Acknowledgments
The author acknowledges the use of GPT-6 and Claude for assistance with literature searches, the development of computational cross-check scripts, and improvements to the clarity of the manuscript. The tools are provided by OpenAI and Anthropic, respectively. The author takes full responsibility for the calculations, interpretation and final text.
Dipole normalization and bosonic loop functions
This appendix supports the model conventions of § 2 and the vacuum-polarization matching of § 3 and § 4 . We specify the contraction and action normalizations, the subtraction convention and the independent scalar check of the finite loop function.
Majorana contractions
The neutral Majorana loop contains two equal orientations. For a single orientation,
F s i n g l e ( q 2 ) = 1 2 F 0 ( q 2 ) , (100) F_{\mathrm{single}}(q^2)=\frac{1}{2}F_0(q^2), \tag*{(100)} F single ( q 2 ) = 2 1 F 0 ( q 2 ) , (100)
An explicit Dirac-trace calculation shows that the two opposite chiral sequences give equal contributions. These are the two contractions needed for the Hermitian transition interaction; the corresponding particle and antiparticle contributions also occur for a Dirac HNL, with both assignments of the transition vertices to the external gauge legs. Keeping just one of these contractions would therefore underestimate the complete lepton-number-conserving amplitude. With the physical coupling bilinears matched as in eq. (8) , the Dirac and Majorana results for the processes considered here agree; there is no additional factor of two in converting between them.
Transverse projection
We use the transverse projector P μ ν = q μ q ν − q 2 g μ ν P^{\mu\nu}=q^\mu q^\nu-q^2g^{\mu\nu} P μν = q μ q ν − q 2 g μν , whose contraction is ( D − 1 ) q 4 (D-1)q^4 ( D − 1 ) q 4 . This standard tensor reduction isolates the scalar function calculated in § 3 . The scalar transverse form factor can therefore be extracted as
F ( q 2 ) = P μ ν Π μ ν ( q ) ( D − 1 ) ( q 2 ) 2 . (101) F(q^2)=\frac{P_{\mu\nu}\Pi^{\mu\nu}(q)}{(D-1)(q^2)^2}. \tag*{(101)} F ( q 2 ) = ( D − 1 ) ( q 2 ) 2 P μν Π μν ( q ) . (101)
Keeping the full D D D dependence through the Passarino–Veltman reduction gives eq. (21) . Setting D = 4 D=4 D = 4 before expanding the scalar integrals would miss the finite contribution in eq. (22) .
Gauge rotation and the charged channel
The neutral amplitudes share the scalar kernel in eq. (20) . Applying the rotation in eq. (3) gives eq. (24) . The charged calculation independently gives F W W = d W 2 F 0 / 2 = F 33 F_{WW}=d_W^2F_0/2=F_{33} F W W = d W 2 F 0 /2 = F 33 in the massless-lepton limit, testing the weak-generator normalization and the charged-dipole vertex together. These relations refer to the two-dipole contributions specified in § 3 .
Zero-momentum threshold
At q 2 = 0 q^2 = 0 q 2 = 0 the neutral threshold matrix is
F ( 0 ) = − M N 2 2 π 2 ( d B 2 d B d 3 d B d 3 d 3 2 ) . (102) F(0) = -\frac{M_N^2}{2\pi^2}\begin{pmatrix} d_B^2 & d_B d_3 \\ d_B d_3 & d_3^2 \end{pmatrix}. \tag*{(102)} F ( 0 ) = − 2 π 2 M N 2 ( d B 2 d B d 3 d B d 3 d 3 2 ) . (102)
Its determinant vanishes. The eigenvalues are
λ 1 = 0 , λ 2 = − M N 2 2 π 2 ( d B 2 + d 3 2 ) , (103) \lambda_1 = 0,\qquad\lambda_2 = -\frac{M_N^2}{2\pi^2}(d_B^2 + d_3^2), \tag*{(103)} λ 1 = 0 , λ 2 = − 2 π 2 M N 2 ( d B 2 + d 3 2 ) , (103)
so the gauge-field combination orthogonal to ( d B , d 3 ) (d_B, d_3) ( d B , d 3 ) receives no threshold from this loop contribution.
Self-energy and Warsaw normalization
Let P μ ν = q μ q ν − q 2 g μ ν P^{\mu\nu} = q^\mu q^\nu- q^2 g^{\mu\nu} P μν = q μ q ν − q 2 g μν , and let K μ ν K^{\mu\nu} K μν be the quadratic action kernel from Δ L = + ∑ i C i Q i \Delta\mathcal{L} = +\sum_i C_i Q_i Δ L = + ∑ i C i Q i . Fourier transformation gives F μ ν ( q ) F μ ν ( − q ) = − 2 V μ ( q ) P μ ν V ν ( − q ) F_{\mu\nu}(q)F^{\mu\nu}(-q) = -2V_\mu(q)P^{\mu\nu}V_\nu(-q) F μν ( q ) F μν ( − q ) = − 2 V μ ( q ) P μν V ν ( − q ) . A diagonal c F F c_F F c F F interaction has kernel − 4 c P -4cP − 4 c P , whereas a mixed c F G c_F G c F G interaction has mixed kernel − 2 c P -2cP − 2 c P . In standard Warsaw normalization, H † σ 3 H = − v 2 / 2 H^\dagger\sigma^3H = -v^2/2 H † σ 3 H = − v 2 /2 , hence
K B B = − 2 v 2 C H B P , K 33 = − 2 v 2 C H W P , K 3 B = + v 2 C H W B P . (104) K_{BB} = -2v^2 C_{HB}P,\qquad K_{33} = -2v^2 C_{HW}P,\qquad K_{3B} = +v^2 C_{HWB}P. \tag*{(104)} K B B = − 2 v 2 C H B P , K 33 = − 2 v 2 C H W P , K 3 B = + v 2 C H W B P . (104)
The action-normalized loop insertion has Π = K \Pi= K Π = K in this dictionary. It gives eq. (30) and the threshold signs in eqs. (31) –(33) . The positive coefficient of X μ ν † X μ ν X_{\mu\nu}^{\dagger}X^{\mu\nu} X μν † X μν in eq. (D.5) provides a consistent Higgs-background check. For the higher-derivative operators in eq. (42) , the same choice gives local contributions
F B B = − v 2 q 2 C 2 B , H , F 33 = − v 2 q 2 C 2 W , H , F 3 B = + v 2 q 2 2 C 2 W B , H . (105) F_{BB} = -v^2q^2 C_{2B,H},\qquad F_{33} = -v^2q^2 C_{2W,H},\qquad F_{3B} = +\frac{v^2q^2}{2}C_{2WB,H}. \tag*{(105)} F B B = − v 2 q 2 C 2 B , H , F 33 = − v 2 q 2 C 2 W , H , F 3 B = + 2 v 2 q 2 C 2 W B , H . (105)
The counterterms in eq. (43) cancel the positive loop poles of eq. (23) in this dictionary.
Scalar-integral subtraction convention
The unrenormalized scalar integrals in eq. (23) display the combination 1 / ϵ − γ E − ln π 1/\epsilon- \gamma_E - \ln\pi 1/ ϵ − γ E − ln π . At the same numerical scale this differs from the conventional Δ ‾ M S = 1 / ϵ − γ E + ln ( 4 π ) \overline{\Delta}_{\mathrm{MS}} = 1/\epsilon- \gamma_E + \ln(4\pi) Δ MS = 1/ ϵ − γ E + ln ( 4 π ) by − ln ( 4 π 2 ) -\ln(4\pi^2) − ln ( 4 π 2 ) . This is an integration-measure convention, distinct from the finite contributions generated by D D D -dependent numerator factors. Explicitly, with μ P X = 2 π μ \mu_{\mathrm{PX}} = 2\pi\mu μ PX = 2 π μ ,
1 ϵ − γ E − ln π + ln μ P X 2 M N 2 = Δ ‾ M S + ln μ 2 M N 2 . (106) \frac{1}{\epsilon} - \gamma_E - \ln\pi+ \ln\frac{\mu_{\mathrm{PX}}^2}{M_N^2} = \overline{\Delta}_{\mathrm{MS}} + \ln\frac{\mu^2}{M_N^2}. \tag*{(106)} ϵ 1 − γ E − ln π + ln M N 2 μ PX 2 = Δ MS + ln M N 2 μ 2 . (106)
Converting the scalar integrals and local subtraction terms together gives the M S ‾ \overline{\mathrm{MS}} MS -normalized finite slope [ ln ( μ 2 / M N 2 ) + 11 / 6 ] / ( 6 π 2 ) [\ln(\mu^2/M_N^2) + 11/6]/(6\pi^2) [ ln ( μ 2 / M N 2 ) + 11/6 ] / ( 6 π 2 ) used in the phenomenological response. The direct scalar quadrature in appendix A.7 uses this normalization from the outset. Equation (B.40) then evolves the independent coefficients from the chosen UV boundary. A consistent change of the integration measure transforms the subtraction convention and those coefficients together, preserving the physical prediction.
Numerical scalar-integral check of the finite bosonic kernel
We also evaluate the scalar integrals directly by Feynman-parameter quadrature, independently of the Package-X closed forms. In units M N 2 = μ 2 = 1 M_N^2 = \mu^2 = 1 M N 2 = μ 2 = 1 and an M S ‾ \overline{\mathrm{MS}} MS -normalized measure, define
A ( ϵ ) = e γ E ϵ Γ ( ϵ ) 1 − ϵ , B ( z , ϵ ) = e γ E ϵ Γ ( ϵ ) ∫ 0 1 d x [ x ( 1 − ( 1 − x ) z ) ] − ϵ . (107) A(\epsilon) = \frac{e^{\gamma_E\epsilon}\Gamma(\epsilon)}{1-\epsilon},\qquad B(z,\epsilon) = e^{\gamma_E\epsilon}\Gamma(\epsilon)\int_0^1 dx\,[x(1-(1-x)z)]^{-\epsilon}. \tag*{(107)} A ( ϵ ) = 1 − ϵ e γ E ϵ Γ ( ϵ ) , B ( z , ϵ ) = e γ E ϵ Γ ( ϵ ) ∫ 0 1 d x [ x ( 1 − ( 1 − x ) z ) ] − ϵ . (107)
Inserting these in eq. (21) with D = 4 − 2 ϵ D = 4 - 2\epsilon D = 4 − 2 ϵ and subtracting the pole gives the dimensionless finite function
lim ϵ → 0 { 3 [ ( D − 2 − z ) A − ( 1 − z ) ( D − 2 + z ) B ] ( D − 1 ) z − z ϵ } = − 3 + 11 6 z + r ( z ) . (108) \lim_{\epsilon\to0}\left\{\frac{3[(D-2-z)A-(1-z)(D-2+z)B]}{(D-1)z}-\frac{z}{\epsilon}\right\}=-3+\frac{11}{6}z+r(z). \tag*{(108)} ϵ → 0 lim { ( D − 1 ) z 3 [( D − 2 − z ) A − ( 1 − z ) ( D − 2 + z ) B ] − ϵ z } = − 3 + 6 11 z + r ( z ) . (108)
At z = − 1 z=-1 z = − 1 , − 0.1 -0.1 − 0.1 , 10 − 4 10^{-4} 1 0 − 4 , 0.01 0.01 0.01 , 0.1 0.1 0.1 , 0.5 0.5 0.5 , 0.9 0.9 0.9 , quadrature at 65-digit precision and linear extrapolation in ϵ \epsilon ϵ agree with this expression to better than 10 − 19 10^{-19} 1 0 − 19 in absolute units. Independent small-z z z extrapolation gives − 3 -3 − 3 and 11 / 6 11/6 11/6 for the constant and slope. This tests the finite scalar evaluation, including evanescent-times-pole terms, and the nonlocal remainder; it shares the tensor reduction in eq. (21) . The fermion bubble contains no internal gauge propagator, so a second gauge choice does not provide a distinct test of this kernel. A separate generic-R ξ R_\xi R ξ check of the direct current diagrams and external residues is described below. The scalar quadrature and the Matchete tree-level matching [32 ] have different scopes and are not independent recalculations of the finite lepton-current coefficients.
Lepton currents and the local pole response
The direct current contribution has two ingredients: the vertex correction and the normalization of the external lepton states. The latter is determined by the residue of the lepton propagator at its pole. This appendix gives both ingredients of the matching in § 4 , together with the identities that connect them. Their combination with gauge propagation and measured inputs produces the pole response used in § 6 . The Ward and gauge-parameter identities test this assembly. The complementary checks, including those discussed in the other appendices, are summarized in Table 4 .
Table 4. Complementary consistency checks. Their assumptions and numerical precision are specified in the corresponding subsections. Quadrature, expansion tests and Ward identities probe different properties; numerical agreement does not exclude an error common to the compared amplitudes.
Lepton self-energies and neutral-current UV residues
Conventions and external-leg factors
We give the UV residues for one lepton doublet ( ν e , e ) L (\nu_e,e)_L ( ν e , e ) L , with massless external leptons and real dipole coefficients. The calculation keeps M N M_N M N , m W m_W m W , m Z m_Z m Z and the external Z Z Z momentum squared q 2 q^2 q 2 distinct. The neutral-current residues combine the 1PI vertices, external-lepton wave-function renormalization (WFR) and SM input-parameter counterterms. Local EFT counterterms cancel the remaining poles. The finite observable response also includes gauge-field normalization, neutral mixing and input shifts, as assembled in appendix § 12 .§ 8 .
The tree-level calibration of the FeynArts/FeynCalc (FC) projector gives
Γ f L , t r e e F C = − g f L , g e L = − e ( c W 2 − s W 2 ) 2 s W c W , g ν L = e 2 s W c W . (109) \Gamma^{\mathrm{FC}}_{fL,\mathrm{tree}}=-g_{fL},\qquad g_{eL}=-\frac{e(c_W^2-s_W^2)}{2s_Wc_W},\qquad g_{\nu L}=\frac{e}{2s_Wc_W}. \tag*{(109)} Γ f L , tree FC = − g f L , g e L = − 2 s W c W e ( c W 2 − s W 2 ) , g ν L = 2 s W c W e . (109)
Accordingly the physical-coupling loop residue is Δ g f L p o l e = − Γ f L F C , U V \Delta g^{\mathrm{pole}}_{fL}=-\Gamma^{\mathrm{FC,UV}}_{fL} Δ g f L pole = − Γ f L FC , UV . The minus sign here converts conventions; it is not UV subtraction. A cancelling EFT counterterm has the further opposite sign, δ g f L E F T , c t = − Δ g f L p o l e \delta g^{\mathrm{EFT,ct}}_{fL}=-\Delta g^{\mathrm{pole}}_{fL} δ g f L EFT , ct = − Δ g f L pole , in the same physical coupling convention. We reserve Δ g \Delta g Δ g for a loop-induced shift and δ C \delta C δ C for a Wilson counterterm.
Writing Σ f ( p ) = p̸ P L Σ f L ( p 2 ) \Sigma_f(p)=\not{p}P_L\Sigma_{fL}(p^2) Σ f ( p ) = p P L Σ f L ( p 2 ) , the calibrated self-energy coefficient extracted with FeynCalc is defined by C f F C = − Σ f L C^{\mathrm{FC}}_f=-\Sigma_{fL} C f FC = − Σ f L . The poles are
Σ e L U V ( p 2 ) = d W 2 8 π 2 ϵ U V ( 6 M N 2 − 2 p 2 + 3 m W 2 ) , (110) \Sigma^{\mathrm{UV}}_{eL}(p^2)=\frac{d_W^2}{8\pi^2\epsilon_{\mathrm{UV}}}(6M_N^2-2p^2+3m_W^2), \tag*{(110)} Σ e L UV ( p 2 ) = 8 π 2 ϵ UV d W 2 ( 6 M N 2 − 2 p 2 + 3 m W 2 ) , (110)
Σ ν L U V ( p 2 ) = 1 8 π 2 ϵ U V [ 2 d γ 2 ( 3 M N 2 − p 2 ) + d Z 2 ( 6 M N 2 − 2 p 2 + 3 m Z 2 ) ] . (111) \Sigma^{\mathrm{UV}}_{\nu L}(p^2)=\frac{1}{8\pi^2\epsilon_{\mathrm{UV}}\!}\left[2d_\gamma^2(3M_N^2-p^2)+d_Z^2(6M_N^2-2p^2+3m_Z^2)\right]. \tag*{(111)} Σ ν L UV ( p 2 ) = 8 π 2 ϵ UV 1 [ 2 d γ 2 ( 3 M N 2 − p 2 ) + d Z 2 ( 6 M N 2 − 2 p 2 + 3 m Z 2 ) ] . (111)
The right-handed self-energy vanishes in this limit. The external-leg factors used in the amplitude assembly are
δ Z e L = Σ e L U V ( 0 ) = 3 d W 2 ( 2 M N 2 + m W 2 ) 8 π 2 ϵ U V , \delta Z_{eL}=\Sigma^{\mathrm{UV}}_{eL}(0)=\frac{3d_W^2(2M_N^2+m_W^2)}{8\pi^2\epsilon_{\mathrm{UV}}}, δ Z e L = Σ e L UV ( 0 ) = 8 π 2 ϵ UV 3 d W 2 ( 2 M N 2 + m W 2 ) ,
δ Z ν L = Σ ν L U V ( 0 ) = 3 [ 2 M N 2 d γ 2 + ( 2 M N 2 + m Z 2 ) d Z 2 ] 8 π 2 ϵ U V . (112) \delta Z_{\nu L}=\Sigma^{\mathrm{UV}}_{\nu L}(0)=\frac{3[2M_N^2d_\gamma^2+(2M_N^2+m_Z^2)d_Z^2]}{8\pi^2\epsilon_{\mathrm{UV}}}. \tag*{(112)} δ Z ν L = Σ ν L UV ( 0 ) = 8 π 2 ϵ UV 3 [ 2 M N 2 d γ 2 + ( 2 M N 2 + m Z 2 ) d Z 2 ] . (112)
There are two external legs, each supplying δ Z L / 2 \delta Z_L/2 δ Z L /2 , so the net coupling contribution is g f L δ Z f L g_{fL}\delta Z_{fL} g f L δ Z f L and the FC coefficient is its negative. These external-leg factors should not be identified with the negative of the Wilson poles in eq. (49) .
The five HNL 1PI diagrams consist of two N ν W N\nu W N ν W triangles, one W N W W N W W N W triangle containing the SM Z W W ZWW Z W W vertex, and two N W N W N W contact diagrams. The contact vertices arise from the non-Abelian field strength in O W O_W O W . Keeping only the triangle with Z W W ZWW Z W W would not implement the operator’s gauge completion. The complete left-handed 1PI pole is
Γ e L 1 P I , F C , U V = e q 2 d W d Z 2 2 π 2 s W ϵ U V − e c W d W 2 48 π 2 s W ϵ U V ( 36 M N 2 − 5 q 2 + 18 m W 2 ) . (113) \Gamma^{\mathrm{1PI,FC,UV}}_{eL}=\frac{eq^2d_Wd_Z}{2\sqrt{2}\pi^2s_W\epsilon_{\mathrm{UV}}}-\frac{ec_Wd_W^2}{48\pi^2s_W\epsilon_{\mathrm{UV}}}(36M_N^2-5q^2+18m_W^2). \tag*{(113)} Γ e L 1PI , FC , UV = 2 2 π 2 s W ϵ UV e q 2 d W d Z − 48 π 2 s W ϵ UV e c W d W 2 ( 36 M N 2 − 5 q 2 + 18 m W 2 ) . (113)
The right-handed 1PI form factor is zero for m e = 0 m_e=0 m e = 0 . Equation (B.5) uses a projector with γ μ P L ↦ 1 \gamma^\mu P_L\mapsto1 γ μ P L ↦ 1 and γ μ P R ↦ 0 \gamma^\mu P_R\mapsto0 γ μ P R ↦ 0 . No additional Majorana multiplicity is to be applied after the listed diagrams have been summed.
The WFR and parameter pieces in the same convention are
Γ e L W F R , F C , U V = + 3 e ( c W 2 − s W 2 ) d W 2 16 π 2 c W s W ϵ U V ( 2 M N 2 + m W 2 ) , (114) \Gamma^{\mathrm{WFR,FC,UV}}_{eL}=+\frac{3e(c_W^2-s_W^2)d_W^2}{16\pi^2c_Ws_W\epsilon_{\mathrm{UV}}}(2M_N^2+m_W^2), \tag*{(114)} Γ e L WFR , FC , UV = + 16 π 2 c W s W ϵ UV 3 e ( c W 2 − s W 2 ) d W 2 ( 2 M N 2 + m W 2 ) , (114)
Γ e L p a r a m , F C , U V = − e 48 π 2 s W 3 c W ϵ U V ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) . (115) \Gamma^{\mathrm{param,FC,UV}}_{eL}=-\frac{e}{48\pi^2s_W^3c_W\epsilon_{\mathrm{UV}}}(2d_Z^2m_Z^2-d_W^2m_W^2). \tag*{(115)} Γ e L param , FC , UV = − 48 π 2 s W 3 c W ϵ UV e ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) . (115)
Here the on-shell input relations used in the calculation are
m W 2 = c W 2 m Z 2 , δ e e ∣ U V = 0 , δ s W 2 ∣ U V = c W 2 ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) 12 π 2 ϵ U V . (116) m_W^2=c_W^2m_Z^2,\qquad\left.\frac{\delta e}{e}\right|_{\mathrm{UV}}=0,\qquad\left.\delta s_W^2\right|_{\mathrm{UV}}=\frac{c_W^2(2d_Z^2m_Z^2-d_W^2m_W^2)}{12\pi^2\epsilon_{\mathrm{UV}}}. \tag*{(116)} m W 2 = c W 2 m Z 2 , e δ e UV = 0 , δ s W 2 UV = 12 π 2 ϵ UV c W 2 ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) . (116)
Differentiating g e L g_{eL} g e L with respect to s W 2 s_W^2 s W 2 gives e / ( 4 s W 3 c W 3 ) e/(4s_W^3c_W^3) e / ( 4 s W 3 c W 3 ) and reproduces eq. (B.7), including the FC sign. The vanishing charge pole follows from the absence of a divergent constant in F A A F_{AA} F AA and from Π A Z ( 0 ) = 0 \Pi_{AZ}(0)=0 Π A Z ( 0 ) = 0 in this sector.
Substituting
d γ = e v 2 ( C B + C W / 2 ) , d Z = e v 2 ( − s W c W C B + c W 2 s W C W ) , d W = e v 2 s W C W , (117) d_\gamma=\frac{ev}{\sqrt{2}}(C_B+C_W/2),\qquad d_Z=\frac{ev}{\sqrt{2}}\left(-\frac{s_W}{c_W}C_B+\frac{c_W}{2s_W}C_W\right),\qquad d_W=\frac{ev}{2s_W}C_W, \tag*{(117)} d γ = 2 e v ( C B + C W /2 ) , d Z = 2 e v ( − c W s W C B + 2 s W c W C W ) , d W = 2 s W e v C W , (117)
we obtain the final residue
Γ e L F C , U V = − 3 e 3 v 2 M N 2 32 π 2 c W s W 3 ϵ U V C W 2 + e 3 v 2 q 2 192 π 2 c W s W 3 ϵ U V ( 17 c W 2 C W 2 − 24 s W 2 C B C W ) − e 3 v 2 m W 2 192 π 2 c W 5 s W 3 ϵ U V ( 9 c W 4 C W 2 − 4 c W 2 C B C W + 4 s W 2 C B 2 ) . (118) \Gamma_{e_L}^{\mathrm{FC,UV}}=-\frac{3e^3v^2M_N^2}{32\pi^2c_Ws_W^3\epsilon_{\mathrm{UV}}}C_W^2
+\frac{e^3v^2q^2}{192\pi^2c_Ws_W^3\epsilon_{\mathrm{UV}}}\left(17c_W^2C_W^2-24s_W^2C_BC_W\right)
-\frac{e^3v^2m_W^2}{192\pi^2c_W^5s_W^3\epsilon_{\mathrm{UV}}}\left(9c_W^4C_W^2-4c_W^2C_BC_W+4s_W^2C_B^2\right). \tag*{(118)} Γ e L FC , UV = − 32 π 2 c W s W 3 ϵ UV 3 e 3 v 2 M N 2 C W 2 + 192 π 2 c W s W 3 ϵ UV e 3 v 2 q 2 ( 17 c W 2 C W 2 − 24 s W 2 C B C W ) − 192 π 2 c W 5 s W 3 ϵ UV e 3 v 2 m W 2 ( 9 c W 4 C W 2 − 4 c W 2 C B C W + 4 s W 2 C B 2 ) . (118)
The leading mass term agrees with − Δ g e L ( 6 ) , p o l e -\Delta g_{e_L}^{(6),\mathrm{pole}} − Δ g e L ( 6 ) , pole in eq. (54) .
Neutrino vertex and independent cross-check
The two HNL 1PI graphs are N ν Z N\nu Z N ν Z triangles with conjugate dipole orientations. The independently normalized projector confirms
Γ ν L 1 P I , F C , U V = e d Z 2 q 2 4 π 2 s W c W ϵ U V , Γ ν R 1 P I , F C , U V = 0. (119) \Gamma_{\nu_L}^{\mathrm{1PI,FC,UV}}=\frac{e d_Z^2q^2}{4\pi^2s_Wc_W\epsilon_{\mathrm{UV}}},\qquad\Gamma_{\nu_R}^{\mathrm{1PI,FC,UV}}=0. \tag*{(119)} Γ ν L 1PI , FC , UV = 4 π 2 s W c W ϵ UV e d Z 2 q 2 , Γ ν R 1PI , FC , UV = 0. (119)
There is no M N 2 M_N^2 M N 2 or m Z 2 m_Z^2 m Z 2 in this 1PI pole. This is an explicit result of the loop reduction, not a consequence of gauge symmetry alone, and it must not be inferred by rescaling the charged-lepton projector correction. The remaining terms are
Γ ν L W F R , F C , U V = − 3 e 16 π 2 s W c W ϵ U V [ 2 M N 2 d γ 2 + ( 2 M N 2 + m Z 2 ) d Z 2 ] , (120) \Gamma_{\nu_L}^{\mathrm{WFR,FC,UV}}=-\frac{3e}{16\pi^2s_Wc_W\epsilon_{\mathrm{UV}}}\left[2M_N^2d_\gamma^2+(2M_N^2+m_Z^2)d_Z^2\right], \tag*{(120)} Γ ν L WFR , FC , UV = − 16 π 2 s W c W ϵ UV 3 e [ 2 M N 2 d γ 2 + ( 2 M N 2 + m Z 2 ) d Z 2 ] , (120)
Γ ν L p a r a m , F C , U V = + e ( c W 2 − s W 2 ) 48 π 2 s W 3 c W ϵ U V ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) . (121) \Gamma_{\nu_L}^{\mathrm{param,FC,UV}}=+\frac{e(c_W^2-s_W^2)}{48\pi^2s_W^3c_W\epsilon_{\mathrm{UV}}}\left(2d_Z^2m_Z^2-d_W^2m_W^2\right). \tag*{(121)} Γ ν L param , FC , UV = + 48 π 2 s W 3 c W ϵ UV e ( c W 2 − s W 2 ) ( 2 d Z 2 m Z 2 − d W 2 m W 2 ) . (121)
The parameter term follows from ∂ g ν L / ∂ s W 2 = − e ( c W 2 − s W 2 ) / ( 4 s W 3 c W 3 ) \partial g_{\nu_L}/\partial s_W^2=-e(c_W^2-s_W^2)/(4s_W^3c_W^3) ∂ g ν L / ∂ s W 2 = − e ( c W 2 − s W 2 ) / ( 4 s W 3 c W 3 ) . After the substitutions in eq. (B.9), the sum is
Γ ν L F C , U V = − 3 e 3 v 2 M N 2 64 π 2 c W 3 s W 3 ϵ U V ( 4 s W 2 C B 2 + c W 2 C W 2 ) + e 3 v 2 q 2 32 π 2 c W 3 s W 3 ϵ U V ( c W 2 C W − 2 s W 2 C B ) 2 + e 3 v 2 m Z 2 384 π 2 c W 3 s W 3 ϵ U V [ − 9 c W 4 C W 2 + 4 c W 2 ( 11 s W 2 − 2 c W 2 ) C B C W + 4 s W 2 ( 2 c W 2 − 11 s W 2 ) C B 2 ] . (122) \Gamma_{\nu_L}^{\mathrm{FC,UV}}=-\frac{3e^3v^2M_N^2}{64\pi^2c_W^3s_W^3\epsilon_{\mathrm{UV}}}\left(4s_W^2C_B^2+c_W^2C_W^2\right)
+\frac{e^3v^2q^2}{32\pi^2c_W^3s_W^3\epsilon_{\mathrm{UV}}}\left(c_W^2C_W-2s_W^2C_B\right)^2
+\frac{e^3v^2m_Z^2}{384\pi^2c_W^3s_W^3\epsilon_{\mathrm{UV}}}\left[-9c_W^4C_W^2+4c_W^2(11s_W^2-2c_W^2)C_BC_W+4s_W^2(2c_W^2-11s_W^2)C_B^2\right]. \tag*{(122)} Γ ν L FC , UV = − 64 π 2 c W 3 s W 3 ϵ UV 3 e 3 v 2 M N 2 ( 4 s W 2 C B 2 + c W 2 C W 2 ) + 32 π 2 c W 3 s W 3 ϵ UV e 3 v 2 q 2 ( c W 2 C W − 2 s W 2 C B ) 2 + 384 π 2 c W 3 s W 3 ϵ UV e 3 v 2 m Z 2 [ − 9 c W 4 C W 2 + 4 c W 2 ( 11 s W 2 − 2 c W 2 ) C B C W + 4 s W 2 ( 2 c W 2 − 11 s W 2 ) C B 2 ] . (122)
All of its M N 2 M_N^2 M N 2 term comes from WFR. Orthogonality of the neutral gauge rotation gives
d γ 2 + d Z 2 = d B 2 + d 3 2 = g ′ 2 v 2 2 C B 2 + g 2 v 2 8 C W 2 . (123) d_\gamma^2+d_Z^2=d_B^2+d_3^2=\frac{g'^2v^2}{2}C_B^2+\frac{g^2v^2}{8}C_W^2. \tag*{(123)} d γ 2 + d Z 2 = d B 2 + d 3 2 = 2 g ′2 v 2 C B 2 + 8 g 2 v 2 C W 2 . (123)
This explains the cancellation of C B C W C_BC_W C B C W in that term and connects the broken-phase calculation to the unbroken current matching. Together the M N 2 M_N^2 M N 2 parts of eqs. (B.10) and (B.14) reproduce both coefficients in eqs. (47) –(48) .
Figure 7. Neutrino 1PI triangle (left; the conjugate orientation is understood) and the self-energy supplying WFR (right). Their UV mass dependence is qualitatively different: only the external-leg contribution supplies the M N 2 M_N^2 M N 2 term in the neutral-current residue.
Dimension-eight interpretation of the residues
The second and third lines of the charged-lepton residue scale as v 2 g 2 C 2 v^2g^2C^2 v 2 g 2 C 2 and v 2 m W 2 C 2 v^2m_W^2C^2 v 2 m W 2 C 2 ; the corresponding neutrino terms scale as v 2 g 2 C 2 v^2g^2C^2 v 2 g 2 C 2 and v 2 m Z 2 C 2 v^2m_Z^2C^2 v 2 m Z 2 C 2 . Since m V 2 ∼ g 2 v 2 m_V^2\sim g^2v^2 m V 2 ∼ g 2 v 2 , these are higher-derivative or additional-Higgs dimension-eight structures at double insertion order. They must not be absorbed into the M N 2 C 2 M_N^2C^2 M N 2 C 2 coefficients of Q H l ( 1 , 3 ) Q_{Hl}^{(1,3)} Q H l ( 1 , 3 ) . General dimension-eight renormalization at order Λ − 4 \Lambda^{-4} Λ − 4 requires such double-insertion terms [45 ] . This framework organizes the double-insertion renormalization; the expressions above give the HNL-specific coefficients.
The two neutral vertices determine the combinations of residues relevant to the pole response. Resolving a complete nonredundant dimension-eight fermionic basis would require additional Green functions and EOM reduction. The independent finite current coefficients require their own boundary conditions; the common prescription used here is specified in appendix § 12 .§ 8 .
For the observable calculation the renormalized couplings enter, in the massless limit, through
Δ Γ f Γ f S M = 2 Re [ g f L ∗ Δ g f L r e n + g f R ∗ Δ g f R r e n ] ∣ g f L ∣ 2 + ∣ g f R ∣ 2 , (124) \frac{\Delta\Gamma_f}{\Gamma_f^{\mathrm{SM}}}
=
\frac{2\operatorname{Re}\left[g_{fL}^{*}\Delta g_{fL}^{\mathrm{ren}}+g_{fR}^{*}\Delta g_{fR}^{\mathrm{ren}}\right]}{\lvert g_{fL}\rvert^2+\lvert g_{fR}\rvert^2}, \tag*{(124)} Γ f SM Δ Γ f = ∣ g f L ∣ 2 + ∣ g f R ∣ 2 2 Re [ g f L ∗ Δ g f L ren + g f R ∗ Δ g f R ren ] , (124)
with the required input and normalization effects included in Δ g r e n \Delta g^{\mathrm{ren}} Δ g ren . A zero right-handed 1PI dipole vertex does not imply vanishing universal input shifts for right-handed charged leptons. The finite response entering the forecast is given in appendix § 12 .§ 8 .
Ultraviolet poles and counterterms
For compactness define K = 3 M N 2 / ( 64 π 2 ϵ U V ) K=3M_N^2/(64\pi^2\epsilon_{\mathrm{UV}}) K = 3 M N 2 / ( 64 π 2 ϵ UV ) , B = g ′ 2 C B 2 B=g'^2C_B^2 B = g ′2 C B 2 and G = g 2 C W 2 G=g^2C_W^2 G = g 2 C W 2 . Table 5 distinguishes the loop pole from its cancelling local term in the same convention. The quantities Γ f F C , U V \Gamma_f^{\mathrm{FC,UV}} Γ f FC , UV refer precisely to the partial assembly specified in appendix § 12 .§ 1 . SM parameter counterterms already included there must not be added a second time. For self-energies and vertex rows the table specifies cancellation at the Green-function level; only rows with a named C i C_i C i are Wilson-coefficient counterterms.
Table 5. Table 5: Ultraviolet poles and cancelling local terms for one flavour and real couplings. For the neutral gauge-basis entry i , j ∈ { B , 3 } i,j\in\{B,3\} i , j ∈ { B , 3 } . A finite threshold is not an absent effect, and a cancelling local term is not an independently fixed renormalized Wilson coefficient.
The bosonic dimension-eight Wilson counterterms, in precisely the Green basis of eq. (42) , remain those of eq. (43) . At coefficient level their loop-pole counterparts are
C 2 B , H p o l e = − g ′ 2 C B 2 12 π 2 ϵ U V , C 2 W , H p o l e = − g 2 C W 2 48 π 2 ϵ U V , C_{2B,H}^{\mathrm{pole}}=-\frac{g'^2C_B^2}{12\pi^2\epsilon_{\mathrm{UV}}},
\qquad
C_{2W,H}^{\mathrm{pole}}=-\frac{g^2C_W^2}{48\pi^2\epsilon_{\mathrm{UV}}}, C 2 B , H pole = − 12 π 2 ϵ UV g ′2 C B 2 , C 2 W , H pole = − 48 π 2 ϵ UV g 2 C W 2 ,
C 2 W B , H p o l e = + g g ′ C B C W 12 π 2 ϵ U V . (125) C_{2WB,H}^{\mathrm{pole}}=+\frac{gg'C_BC_W}{12\pi^2\epsilon_{\mathrm{UV}}}. \tag*{(125)} C 2 W B , H pole = + 12 π 2 ϵ UV g g ′ C B C W . (125)
Electromagnetic gauge invariance fixes the transverse tensor structure but still allows a UV pole multiplying M N 2 M_N^2 M N 2 in F i j F_{ij} F ij , or M N 2 q 2 M_N^2q^2 M N 2 q 2 in the self-energy. Its absence here follows from cancellation between the terms of the loop function. Thus the dimension-six bosonic threshold is finite, whereas its momentum-dependent slope requires the dimension-eight counterterms. These are distinct consequences of the explicit amplitude and of gauge symmetry.
Finite current checks: Zee and Zνν
The chiral projector is normalized by P L [ a γ μ P L ] = a P_L[a\gamma^\mu P_L]=a P L [ a γ μ P L ] = a and P L [ a γ μ P R ] = 0 P_L[a\gamma^\mu P_R]=0 P L [ a γ μ P R ] = 0 ; the broken-phase amplitudes are compared with the unbroken current matching in the same coupling-sign convention. For the pure C W 2 C_W^2 C W 2 sector define
F e = e d W 2 M N 2 16 π 2 s W c W K M S . (126) \mathcal{F}_e=\frac{e d_W^2 M_N^2}{16\pi^2s_Wc_W}K_{\mathrm{MS}}. \tag*{(126)} F e = 16 π 2 s W c W e d W 2 M N 2 K MS . (126)
The local dimension-six terms are Δ g e L = − F e \Delta g_{e_L}=-\mathcal{F}_e Δ g e L = − F e and Δ g ν L = − F e / 2 \Delta g_{\nu_L}=-\mathcal{F}_e/2 Δ g ν L = − F e /2 , as follows directly by substituting eqs. (37) –(38) into eqs. (40) –(41) . The electron result combines 1PI and wave-function contributions. In the neutrino channel, the zero-momentum 1PI contribution vanishes:
F ν L 1 P I ( 0 ) ∣ C W 2 = 0. (127) F_{\nu_L}^{\mathrm{1PI}}(0)\big|_{C_W^2}=0. \tag*{(127)} F ν L 1PI ( 0 ) C W 2 = 0. (127)
The cancellation holds exactly in M N / m Z M_N/m_Z M N / m Z at q 2 = 0 q^2=0 q 2 = 0 . Its compatibility with the generic linear R ξ R_\xi R ξ gauge follows from appendix § 12 .§ 7 ; it does not imply a vanishing vertex at nonzero q 2 q^2 q 2 . The dimension-six current threshold in this channel comes from WFR.
The cancellation can be checked at the level of scalar integrals, retaining the nontrivial C 0 C_0 C 0 term. Put a = M N 2 a=M_N^2 a = M N 2 , b = m Z 2 b=m_Z^2 b = m Z 2 , K = 1 + Δ + ln ( μ 2 / b ) K=1+\Delta+\ln(\mu^2/b) K = 1 + Δ + ln ( μ 2 / b ) and L = ln ( a / b ) L=\ln(a/b) L = ln ( a / b ) , with the common regulated pole Δ \Delta Δ left arbitrary. The scalar integrals are
A a = a ( K − L ) , A b = b K , B 0 a = K − L (128) A_a=a(K-L), \qquad A_b=bK, \qquad B_{0a}=K-L \tag*{(128)} A a = a ( K − L ) , A b = b K , B 0 a = K − L (128)
B 0 b = K , B a b = K − a L a − b , C 0 a b = − L a − b . (129) B_{0b}=K, \qquad B_{ab}=K-\frac{aL}{a-b}, \qquad C_{0ab}=-\frac{L}{a-b}. \tag*{(129)} B 0 b = K , B ab = K − a − b a L , C 0 ab = − a − b L . (129)
The full scalar bracket multiplying the 1PI prefactor is
B ν = 2 A a − 3 A b + ( 3 a − 2 b ) B 0 a + 2 b B 0 b + ( − 5 a + 3 b ) B a b + ( 4 a b − 2 b 2 ) C 0 a b = 0. (130) \mathcal{B}_\nu=2A_a-3A_b+(3a-2b)B_{0a}+2bB_{0b}+(-5a+3b)B_{ab}
+(4ab-2b^2)C_{0ab}=0. \tag*{(130)} B ν = 2 A a − 3 A b + ( 3 a − 2 b ) B 0 a + 2 b B 0 b + ( − 5 a + 3 b ) B ab + ( 4 ab − 2 b 2 ) C 0 ab = 0. (130)
Both the pole and finite parts cancel, with a regular limit at a = b > 0 a=b>0 a = b > 0 .
Finite-current subtraction and consistency checks
The physical-measure result, after stripping the common mass and Dirac prefactor, contains
K b a r e = − 6 ϵ + 1 + 6 γ E − 6 ln ( 4 π ) − 6 ln μ 2 M N 2 , (131) K_{\rm bare}=-\frac{6}{\epsilon}+1+6\gamma_E-6\ln(4\pi)-6\ln\frac{\mu^2}{M_N^2}, \tag*{(131)} K bare = − ϵ 6 + 1 + 6 γ E − 6 ln ( 4 π ) − 6 ln M N 2 μ 2 , (131)
K b a r e + 6 Δ ‾ = 1 − 6 ln μ 2 M N 2 , Δ ‾ = 1 ϵ − γ E + ln ( 4 π ) . (132) K_{\rm bare}+6\overline{\Delta}=1-6\ln\frac{\mu^2}{M_N^2}, \qquad\overline{\Delta}=\frac{1}{\epsilon}-\gamma_E+\ln(4\pi). \tag*{(132)} K bare + 6 Δ = 1 − 6 ln M N 2 μ 2 , Δ = ϵ 1 − γ E + ln ( 4 π ) . (132)
The finite constant is obtained from the full integral after M S ‾ \overline{\mathrm{MS}} MS subtraction, including the numerator dependence on D D D . The identities
C 1 + C 3 = M N 2 g 2 C W 2 32 π 2 K M S ‾ , (133) C_1+C_3=\frac{M_N^2g^2C_W^2}{32\pi^2}K_{\overline{\mathrm{MS}}}, \tag*{(133)} C 1 + C 3 = 32 π 2 M N 2 g 2 C W 2 K MS , (133)
− C 1 + C 3 = − M N 2 ( 4 g ′ 2 C B 2 + g 2 C W 2 ) 64 π 2 K M S ‾ (134) -C_1+C_3=-\frac{M_N^2(4g^{\prime2}C_B^2+g^2C_W^2)}{64\pi^2}K_{\overline{\mathrm{MS}}} \tag*{(134)} − C 1 + C 3 = − 64 π 2 M N 2 ( 4 g ′2 C B 2 + g 2 C W 2 ) K MS (134)
reconstruct the direct charged-lepton and neutrino shifts. In particular,
δ g ν L c u r r e n t = − g Z M N 2 ( d B 2 + d 3 2 ) 16 π 2 K M S ‾ . (135) \delta g_{\nu_L}^{\rm current}=-\frac{g_ZM_N^2(d_B^2+d_3^2)}{16\pi^2}K_{\overline{\mathrm{MS}}}. \tag*{(135)} δ g ν L current = − 16 π 2 g Z M N 2 ( d B 2 + d 3 2 ) K MS . (135)
The two neutrino vertex diagrams are combined before the zero-momentum limit, so their sum has no spurious 1 / q 2 1/q^2 1/ q 2 singularity.
For a threshold Δ C i = A i [ 1 − 6 ln ( μ 2 / M N 2 ) ] \Delta C_i=A_i[1-6\ln(\mu^2/M_N^2)] Δ C i = A i [ 1 − 6 ln ( μ 2 / M N 2 )] , the derivative with respect to ln μ \ln\mu ln μ is − 12 A i -12A_i − 12 A i at this order. Running and local contributions must compensate the scale dependence in a complete prediction; varying only this logarithm is not such a prediction. For example, a vanishing independent coefficient at μ = M N \mu=M_N μ = M N leaves the threshold Δ C i ( M N ) = A i \Delta C_i(M_N)=A_i Δ C i ( M N ) = A i below the heavy-state mass. The forecast instead specifies the common boundary at Λ B \Lambda_B Λ B , as in eq. (B.40). Running of dimension-six Higgs–neutrino operators, as studied in [13 ] , provides context but is not a direct calculation of these quadratic dipole thresholds.
Projector normalization and contact-diagram identities
For massless external states, the left-handed projector must separate the two chiral currents. With p 1 ⋅ p 2 = q 2 / 2 p_1\cdot p_2=q^2/2 p 1 ⋅ p 2 = q 2 /2 , its normalization is fixed by:
P L [ a γ μ P L ] = a , P L [ a γ μ P R ] = 0. (136) P_L[a\gamma^\mu P_L]=a, \qquad P_L[a\gamma^\mu P_R]=0. \tag*{(136)} P L [ a γ μ P L ] = a , P L [ a γ μ P R ] = 0. (136)
The trace projection uses the spinor ordering of the physical external states. For a decay chain u ˉ ( p 1 ) M μ v ( p 2 ) \bar{u}(p_1)M^\mu v(p_2) u ˉ ( p 1 ) M μ v ( p 2 ) , the corresponding spin-summed trace contains ̸ p 1 M μ ̸ p 2 \not{p}_1M^\mu\not{p}_2 p 1 M μ p 2 ; the Born normalization alone does not fix this ordering for a general loop chain. The tree amplitude in eq. (B.1) fixes the overall sign relating the projected coefficient to the physical coupling.
The normalized projectors reconstruct the unbroken current poles from the broken-phase amplitudes. The neutrino 1PI pole has no M N 2 M_N^2 M N 2 or m Z 2 m_Z^2 m Z 2 term, while its external residue supplies the leading mass term. The corresponding finite contributions are discussed in appendix § 12 .§ 3 .
For the charged-lepton mass term, W N W WNW W N W denotes the triangle with one internal HNL and two charged W W W lines, while contact 1 and contact 2 place the four-point dipole vertex at the two ends of the charged line in Figure 2 . The superscripts FC and UV denote the FeynCalc sign convention and the ultraviolet pole, respectively; ∣ M N 2 |_{M_N^2} ∣ M N 2 selects the term proportional to the squared HNL mass. Their contributions are
Γ W N W ∣ M N 2 F C , U V = + 3 e c W d W 2 M N 2 8 π 2 s W ϵ U V , \Gamma_{WNW|M_N^2}^{\mathrm{FC,UV}}=+\frac{3 e c_W d_W^2 M_N^2}{8\pi^2 s_W\epsilon_{\mathrm{UV}}}, Γ W N W ∣ M N 2 FC , UV = + 8 π 2 s W ϵ UV 3 e c W d W 2 M N 2 ,
Γ c o n t a c t , 1 F C , U V ∣ M N 2 = Γ c o n t a c t , 2 F C , U V ∣ M N 2 = − 9 e c W d W 2 M N 2 16 π 2 s W ϵ U V , \left.\Gamma_{\mathrm{contact,1}}^{\mathrm{FC,UV}}\right|_{M_N^2}=\left.\Gamma_{\mathrm{contact,2}}^{\mathrm{FC,UV}}\right|_{M_N^2}=-\frac{9 e c_W d_W^2 M_N^2}{16\pi^2 s_W\epsilon_{\mathrm{UV}}}, Γ contact , 1 FC , UV M N 2 = Γ contact , 2 FC , UV M N 2 = − 16 π 2 s W ϵ UV 9 e c W d W 2 M N 2 ,
Γ W N W + c o n t a c t s F C , U V ∣ M N 2 = − 3 e c W d W 2 M N 2 4 π 2 s W ϵ U V . (137) \left.\Gamma_{WNW+\mathrm{contacts}}^{\mathrm{FC,UV}}\right|_{M_N^2}=-\frac{3 e c_W d_W^2 M_N^2}{4\pi^2 s_W\epsilon_{\mathrm{UV}}}. \tag*{(137)} Γ W N W + contacts FC , UV M N 2 = − 4 π 2 s W ϵ UV 3 e c W d W 2 M N 2 . (137)
Their sum reproduces the M N 2 M_N^2 M N 2 piece of eq. (B.5). This diagram-level identity illustrates why the non-Abelian completion cannot be dropped.
At order M N 2 M_N^2 M N 2 the two broken-phase residues invert to
C H l ( 1 ) , p o l e = − Δ g ν L p o l e + Δ g e L p o l e g Z v 2 , C H l ( 3 ) , p o l e = Δ g ν L p o l e − Δ g e L p o l e g Z v 2 . (138) C_{Hl}^{(1),\mathrm{pole}}=-\frac{\Delta g_{\nu_L}^{\mathrm{pole}}+\Delta g_{e_L}^{\mathrm{pole}}}{g_Zv^2},\qquad C_{Hl}^{(3),\mathrm{pole}}=\frac{\Delta g_{\nu_L}^{\mathrm{pole}}-\Delta g_{e_L}^{\mathrm{pole}}}{g_Zv^2}. \tag*{(138)} C H l ( 1 ) , pole = − g Z v 2 Δ g ν L pole + Δ g e L pole , C H l ( 3 ) , pole = g Z v 2 Δ g ν L pole − Δ g e L pole . (138)
Only the M N 2 M_N^2 M N 2 terms are inserted in these equations. Retaining the q 2 q^2 q 2 or electroweak-mass pieces would incorrectly fold dimension-eight structures into a dimension-six matching relation.
Electromagnetic Ward identity and scale dependence
The unbroken electromagnetic symmetry provides a check on the relative normalization of the charged triangle, the two non-Abelian contact diagrams and the external-electron residue. We obtain their photon vertices from the corresponding neutral-current scalar amplitudes by replacing g c W → e g c_W \to e g c W → e in the V W W VWW V W W coupling and 2 g c W d W → 2 e d W 2g c_W d_W \to2e d_W 2 g c W d W → 2 e d W in the contact couplings. The neutral-dipole triangle vanishes in the zero-momentum limit. In the FC convention of eq. (B.1), the required identity is
Γ e L γ , 1 P I , F C ( 0 ) + e Σ e L ( 0 ) = 0. (139) \Gamma_{e_L}^{\gamma,\mathrm{1PI,FC}}(0)+e\Sigma_{e_L}(0)=0. \tag*{(139)} Γ e L γ , 1PI , FC ( 0 ) + e Σ e L ( 0 ) = 0. (139)
Both terms change sign on conversion to physical coupling shifts. The UV pole cancels algebraically for arbitrary M N M_N M N and m W m_W m W :
Γ e L γ , 1 P I , F C , U V ( 0 ) = − 3 e d W 2 8 π 2 ϵ U V ( 2 M N 2 + m W 2 ) . (140) \Gamma_{e_L}^{\gamma,\mathrm{1PI,FC,UV}}(0)=-\frac{3 e d_W^2}{8\pi^2\epsilon_{\mathrm{UV}}}\left(2M_N^2+m_W^2\right). \tag*{(140)} Γ e L γ , 1PI , FC , UV ( 0 ) = − 8 π 2 ϵ UV 3 e d W 2 ( 2 M N 2 + m W 2 ) . (140)
For the finite parts, evaluated in the same MS convention, we test the three scale pairs
( M N , μ ) ∈ { ( 1000 , 1000 ) , ( 200 , 500 ) , ( 5000 , 2000 ) } G e V , (141) (M_N,\mu)\in\{(1000,1000),\ (200,500),\ (5000,2000)\}\ \mathrm{GeV}, \tag*{(141)} ( M N , μ ) ∈ {( 1000 , 1000 ) , ( 200 , 500 ) , ( 5000 , 2000 )} GeV , (141)
with m W = 80.379 G e V m_W=80.379\ \mathrm{GeV} m W = 80.379 GeV . At 100-digit precision the cancellation residual, normalized to ∣ e Σ e L f i n ( 0 ) ∣ |e\Sigma_{e_L}^{\mathrm{fin}}(0)| ∣ e Σ e L fin ( 0 ) ∣ , is below 1.1 × 10 − 35 1.1\times10^{-35} 1.1 × 1 0 − 35 at q 2 = 10 − 30 G e V 2 q^2=10^{-30}\ \mathrm{GeV}^2 q 2 = 1 0 − 30 GeV 2 and below 1.1 × 10 − 45 1.1\times10^{-45} 1.1 × 1 0 − 45 at q 2 = 10 − 40 G e V 2 q^2=10^{-40}\ \mathrm{GeV}^2 q 2 = 1 0 − 40 GeV 2 . The decrease with q 2 q^2 q 2 also checks that the zero-momentum limit is approached smoothly. This calculation uses the neutral-current scalar reductions with the photon coupling replacements; it does not constitute a separate generation of photon diagrams. It tests the finite contact–triangle–residue cancellation, including the finite terms inherited from D D D -dimensional numerator algebra. It does not constrain the transverse q 2 q^2 q 2 coefficient of the charged vertex.
We also differentiate the finite scalar amplitudes with respect to ln μ 2 \ln\mu^2 ln μ 2 . For each of the four electron vertex groups, the neutrino vertex and the exact external-lepton residue, the derivative equals the corresponding coefficient of 1 / ϵ U V 1/\epsilon_{\mathrm{UV}} 1/ ϵ UV . These identities hold before the heavy-mass expansion and therefore check the scale-dependent terms retained at both local orders. For C B = 0 C_B=0 C B = 0 , the dimension-six direct current shifts satisfy Δ g ν L / Δ g e L = 1 / 2 \Delta g_{\nu L}/\Delta g_{eL}=1/2 Δ g ν L /Δ g e L = 1/2 ; this relation does not include the universal input corrections. Substitution of the physical dipoles in the charged-lepton pole also reproduces the numerator 17 = 12 + 5 17=12+5 17 = 12 + 5 in the q 2 C W 2 q^2 C_W^2 q 2 C W 2 term. In the physical coupling convention, the momentum-dependent poles in eqs. (B.5) and (B.11) equal the L B L_B L B derivatives of the corresponding finite terms in eqs. (B.41)–(B.42). The same relation holds for the mass terms after including the external residues; the common boundary therefore preserves the logarithmic slopes fixed by the scalar poles.
Gauge-parameter independence of the direct currents
We compute the HNL contributions to Z → e + e − Z \to e^+e^- Z → e + e − and Z → ν ν ˉ Z \to\nu\bar{\nu} Z → ν ν ˉ , together with the charged-lepton and neutrino self-energies, in a generic linear R ξ R_\xi R ξ gauge and at ξ = 1 \xi=1 ξ = 1 . These amplitudes receive five, two, one and two diagrams, respectively, including the two non-Abelian contact diagrams in the charged-lepton vertex. We retain the D D D -dimensional fermion chains and generic real dipole coefficients in the massless-light-lepton limit. No Goldstone or ghost diagrams contribute to this inventory at one loop with two dipole insertions, vanishing active–sterile mixing and massless light leptons. Ghosts have no fermion vertex, while a Goldstone-containing dipole vertex also contains a gauge field; it does not replace a three-point N ℓ V N\ell V N ℓ V vertex in these one-loop current and self-energy graphs. Mixing-induced Goldstone couplings are absent in the stated limit. This statement is restricted to these amplitudes and perturbative order. The reference amplitudes are nonzero and quadratic in the dipoles; setting all gauge parameters to one recovers the reference before conversion and contraction.
For every diagram, the generic-gauge minus Feynman-gauge difference vanishes algebraically before integration. The relevant gauge parameters are ξ W \xi_W ξ W for the charged-lepton vertex and self-energy, ξ Z \xi_Z ξ Z for the neutrino vertex, and ξ A , ξ Z \xi_A,\xi_Z ξ A , ξ Z for its self-energy. The cancellation follows from the transverse single-gauge dipole vertex,
k μ σ μ ν k ν = 0 , (142) k_\mu\sigma^{\mu\nu} k_\nu= 0, \tag*{(142)} k μ σ μν k ν = 0 , (142)
which removes longitudinal propagator contributions. Each internal gauge line in the listed diagrams has at least one endpoint on a single-gauge dipole vertex. This also holds for the contact diagrams: the other end of the internal gauge line is a pure dipole, so transversality does not require the two-gauge contact vertex itself to be separately transverse. Thus the equality holds for both the poles and finite parts, and differentiating the equal self-energies also gives equal external residues. The retained bosonic vacuum-polarization bubbles contain no internal gauge propagator; their input and external-gauge-boson corrections are constructed from those same two-point functions. Together these checks test gauge-parameter independence of the retained pole ingredients.
Assembly and independent local directions
We assemble the local pole response at one loop with two real dipole insertions, retaining terms of orders M N 2 M_N^2 M N 2 and M N 0 M_N^0 M N 0 . We assume aligned flavours and neglect light-lepton Yukawa couplings and active–sterile mixing. The input parameters are e , M W , M Z e,M_W,M_Z e , M W , M Z . We describe the bosonic derivative operators through their corrections to gauge propagators. Using the gauge-field equations of motion would express the same effects as current interactions; including both representations simultaneously would count them twice. The resulting physical response can be written as
δ a f χ = A T T f χ 3 + A Q Q f + δ f χ , ν L δ a ν d i r + δ f χ , e L δ a e d i r . (143) \delta a_{f\chi}=A_T T^3_{f\chi}+A_Q Q_f+\delta_{f\chi,\nu_L}\,\delta a^{\mathrm{dir}}_\nu+\delta_{f\chi,e_L}\,\delta a^{\mathrm{dir}}_e . \tag*{(143)} δ a f χ = A T T f χ 3 + A Q Q f + δ f χ , ν L δ a ν dir + δ f χ , e L δ a e dir . (143)
Here the last two terms denote the corresponding universal flavour classes, not an additional sum over three heavy states. Bosonic closed loops carry the heavy-state multiplicity of the benchmark; an aligned external-flavour current does not.
Independent current checks. The two-lepton–two-doublet amplitudes reduce algebraically to the dipole self-energy master integral after aligning loop momenta and including the appropriate coupling factors. A charged component, not used to determine the two neutral leading coefficients, verifies the triplet relation. The tree-level fermionic Z Z Z vertex fixes the coupling sign, and the Z h G 0 ZhG^0 Z h G 0 vertex fixes the momentum convention of the derivative scalar current. Here G 0 G^0 G 0 is the neutral would-be Goldstone boson in the Higgs doublet. Comparing these two vertices ensures that the fermion and scalar currents use the same action and Fourier conventions. With the resulting Wilson dictionary, δ a ν ( 6 ) = − v 2 ( C H l ( 1 ) − C H l ( 3 ) ) / 2 \delta a^{(6)}_\nu=-v^2(C^{(1)}_{Hl}-C^{(3)}_{Hl})/2 δ a ν ( 6 ) = − v 2 ( C H l ( 1 ) − C H l ( 3 ) ) /2 and δ a e ( 6 ) = − v 2 ( C H l ( 1 ) + C H l ( 3 ) ) / 2 \delta a^{(6)}_e=-v^2(C^{(1)}_{Hl}+C^{(3)}_{Hl})/2 δ a e ( 6 ) = − v 2 ( C H l ( 1 ) + C H l ( 3 ) ) /2 reproduce the Born-normalized loop shifts. The Wilson dictionary and action-to-amplitude map are fixed together by these Born calibrations.
The doublet fields used in this check are auxiliary matching sources, not Higgs observables. In particular, putting v = 0 v=0 v = 0 in a two-doublet source cannot determine operators containing four doublet fields. Keeping instead the terms of order v 2 p v^2p v 2 p in the neutral sources provides two independent checks. Let S = H † H S=H^\dagger H S = H † H , K μ = i H † D ↔ μ H K_\mu=iH^\dagger\overleftrightarrow{D}_\mu H K μ = i H † D μ H , J μ = L ˉ γ μ L J_\mu=\bar{L}\gamma_\mu L J μ = L ˉ γ μ L , and
Q t = S J μ K μ , Q ν t = J ν H , μ K μ , J ν H , μ = S J μ − ( H † σ I H ) J μ I 2 . (144) Q_t=SJ_\mu K^\mu,\qquad Q_{\nu t}=J_{\nu H,\mu}K^\mu,\qquad J_{\nu H,\mu}=\frac{SJ_\mu-(H^\dagger\sigma^I H)J^I_\mu}{2}. \tag*{(144)} Q t = S J μ K μ , Q ν t = J ν H , μ K μ , J ν H , μ = 2 S J μ − ( H † σ I H ) J μ I . (144)
In the neutral background J ν H = S ν ˉ L γ ν ν L J_{\nu H}=S\bar{\nu}_L\gamma_\nu\nu_L J ν H = S ν ˉ L γ ν ν L . For L N = ln ( μ 2 / M N 2 ) L_N=\ln(\mu^2/M_N^2) L N = ln ( μ 2 / M N 2 ) the independently extracted finite combinations in this representative are
C t = e 4 C W 2 128 π 2 s W 4 ( 7 − 6 L N ) , (145) C_t=\frac{e^4 C_W^2}{128\pi^2s_W^4}(7-6L_N), \tag*{(145)} C t = 128 π 2 s W 4 e 4 C W 2 ( 7 − 6 L N ) , (145)
C t + C ν t = e 4 ( c W 2 C W − 2 s W 2 C B ) 2 256 π 2 s W 4 c W 4 ( 7 − 6 L N ) . (146) C_t+C_{\nu t}=\frac{e^4(c_W^2C_W-2s_W^2C_B)^2}{256\pi^2s_W^4c_W^4}(7-6L_N). \tag*{(146)} C t + C ν t = 256 π 2 s W 4 c W 4 e 4 ( c W 2 C W − 2 s W 2 C B ) 2 ( 7 − 6 L N ) . (146)
They reproduce the constant dimension-eight direct shifts − v 4 C t / 4 -v^4C_t/4 − v 4 C t /4 and − v 4 ( C t + C ν t ) / 4 -v^4(C_t+C_{\nu t})/4 − v 4 ( C t + C ν t ) /4 , respectively. The coefficients of q 2 q^2 q 2 are fixed by the independently normalized neutral vertices. The neutral action reconstructs both constants and both q 2 q^2 q 2 coefficients; it is not claimed to determine every off-shell dimension-eight Wilson coefficient. In obtaining Eq. (B.38), the D D D dependence multiplies the ultraviolet poles before the finite part is taken. The summed vacuum integral is proportional to ∫ d D ℓ / [ ℓ 2 ( ℓ 2 − M N 2 ) ] \int d^D\ell/[\ell^2(\ell^2-M_N^2)] ∫ d D ℓ / [ ℓ 2 ( ℓ 2 − M N 2 )] and is infrared convergent at the origin.
Soft regions and normalization. A loop containing heavy and light particles can receive contributions from momenta of order M N M_N M N and from momenta of order the electroweak scales. The first region determines the local threshold; the second is reproduced by light-field diagrams in the low-energy theory. For the orders considered here, the chiral structure makes this separation particularly simple. For the lepton-number-conserving open chains, the chiral projectors remove the heavy-mass numerator. The soft expansion therefore starts as
ℓ ℓ 2 − M N 2 = − ℓ M N 2 + O ( M N − 4 ) . (147) \frac{\ell}{\ell^2-M_N^2}=-\frac{\ell}{M_N^2}+O(M_N^{-4}). \tag*{(147)} ℓ 2 − M N 2 ℓ = − M N 2 ℓ + O ( M N − 4 ) . (147)
The associated tree-level light-field interaction has the field content L ˉ L H 2 X 2 D \bar{L}LH^2X^2D L ˉ L H 2 X 2 D and canonical dimension ten. With no vertex growing with M N M_N M N , a light-region loop in dimensional regularization cannot turn its C 2 / M N 2 C^2/M_N^2 C 2 / M N 2 coefficient into a term of order M N 0 M_N^0 M N 0 or M N 2 M_N^2 M N 2 . This power counting has been checked against the heavy propagators and chiral chains of the current, residue and auxiliary amplitudes. In the two-point bosonic loop the heavy-mass numerator likewise vanishes by the odd gamma trace. The argument is restricted to the stated lepton-number-conserving approximation; it does not remove the soft terms needed for an exact low-mass prediction or apply to Majorana lepton-number-violating amplitudes.
For each direct current the assembly is δ a f d i r = δ a f v e r t e x + a f S M σ f \delta a_f^{\mathrm{dir}}=\delta a_f^{\mathrm{vertex}}+a_f^{\mathrm{SM}}\sigma_f δ a f dir = δ a f vertex + a f SM σ f . Here a f S M = T f 3 − s W 2 Q f a_f^{\mathrm{SM}}=T_f^3-s_W^2Q_f a f SM = T f 3 − s W 2 Q f is the dimensionless SM chiral coupling and σ f \sigma_f σ f is the net fractional external-fermion residue in our self-energy convention. Each external leg contributes half of this residue, so the pair produces a f S M σ f a_f^{\mathrm{SM}}\sigma_f a f SM σ f . The Wilson-derived direct shifts already contain this contribution. The universal pole residue and input shifts enter only through A T , A Q A_T,A_Q A T , A Q ; adding EOM-induced bosonic currents on top of the unreduced propagator representation would double count them. These statements also account for the light-reducible external-leg and propagator attachments relevant to eq. (B.35).
Common finite prescription. We use the NDR prescription of the open-chain calculation, with D D D retained through the Laurent separation and the specified physical external-state projections. Changing the prescription for γ 5 \gamma_5 γ 5 requires a consistent finite conversion of the renormalized action and its independent coefficients; differences between bare diagrams alone do not define that conversion. For a threshold T i ( L N ) T_i(L_N) T i ( L N ) linear in L N L_N L N , a common independent boundary at L B = ln ( Λ B 2 / M N 2 ) L_B=\ln(\Lambda_B^2/M_N^2) L B = ln ( Λ B 2 / M N 2 ) gives
C i t o t a l = T i ( L N ) + ∂ T i ∂ L N ( L B − L N ) + C i i n d e p e n d e n t ( Λ B ) = T i ( L B ) + C i i n d e p e n d e n t ( Λ B ) . (148) C_i^{\mathrm{total}}=T_i(L_N)+\frac{\partial T_i}{\partial L_N}(L_B-L_N)+C_i^{\mathrm{independent}}(\Lambda_B)=T_i(L_B)+C_i^{\mathrm{independent}}(\Lambda_B). \tag*{(148)} C i total = T i ( L N ) + ∂ L N ∂ T i ( L B − L N ) + C i independent ( Λ B ) = T i ( L B ) + C i independent ( Λ B ) . (148)
This rule applies to both orders in the heavy-mass expansion and to both current and bosonic coefficients. The logarithmic slopes agree with the poles, and the subtraction-scale dependence cancels at the retained order. Independent finite ultraviolet coefficients are not predicted by the dipole threshold.
The finite functions entering eq. (B.35) can be written compactly in terms of s = s W s=s_W s = s W , c = c W c=c_W c = c W , L = L B L=L_B L = L B , q = M Z 2 q=M_Z^2 q = M Z 2 , and D Z = c 2 C W − 2 s 2 C B D_Z=c^2C_W-2s^2C_B D Z = c 2 C W − 2 s 2 C B . For each external lepton flavour, the direct terms are
δ a e d i r = e 2 v 2 M N 2 C W 2 64 π 2 s W 2 ( 6 L − 1 ) + e 4 v 4 C W 2 512 π 2 s W 4 ( 6 L − 7 ) + e 2 v 2 q 1152 π 2 s 2 [ 72 s 2 C B C W ( 2 L + 3 ) − c 2 C W 2 ( 102 L + 241 ) ] , (149) \delta a_e^{\mathrm{dir}}=\frac{e^2v^2M_N^2C_W^2}{64\pi^2s_W^2}(6L-1)+\frac{e^4v^4C_W^2}{512\pi^2s_W^4}(6L-7)
+\frac{e^2v^2q}{1152\pi^2s^2}\left[72s^2C_BC_W(2L+3)-c^2C_W^2(102L+241)\right], \tag*{(149)} δ a e dir = 64 π 2 s W 2 e 2 v 2 M N 2 C W 2 ( 6 L − 1 ) + 512 π 2 s W 4 e 4 v 4 C W 2 ( 6 L − 7 ) + 1152 π 2 s 2 e 2 v 2 q [ 72 s 2 C B C W ( 2 L + 3 ) − c 2 C W 2 ( 102 L + 241 ) ] , (149)
δ a ν d i r = e 2 v 2 M N 2 ( 4 s 2 C B 2 + c 2 C W 2 ) 128 π 2 s 2 c 2 ( 6 L − 1 ) + e 2 v 2 D Z 2 1024 π 2 s 4 c 4 [ e 2 v 2 ( 6 L − 7 ) − 16 s 2 c 2 q ( 2 L + 3 ) ] . (150) \delta a_\nu^{\mathrm{dir}}=\frac{e^2v^2M_N^2(4s^2C_B^2+c^2C_W^2)}{128\pi^2s^2c^2}(6L-1)
+\frac{e^2v^2D_Z^2}{1024\pi^2s^4c^4}\left[e^2v^2(6L-7)-16s^2c^2q(2L+3)\right]. \tag*{(150)} δ a ν dir = 128 π 2 s 2 c 2 e 2 v 2 M N 2 ( 4 s 2 C B 2 + c 2 C W 2 ) ( 6 L − 1 ) + 1024 π 2 s 4 c 4 e 2 v 2 D Z 2 [ e 2 v 2 ( 6 L − 7 ) − 16 s 2 c 2 q ( 2 L + 3 ) ] . (150)
No flavour multiplicity multiplies these direct terms. The universal terms instead contain the sum over the three aligned HNLs. Define
( R A A , R A Z , R Z Z , R W W ) = 3 ( d y 2 , d y d z , d z 2 , d 3 2 ) , (R_{AA},R_{AZ},R_{ZZ},R_{WW})=3(d_y^2,d_yd_z,d_z^2,d_3^2), ( R AA , R A Z , R Z Z , R W W ) = 3 ( d y 2 , d y d z , d z 2 , d 3 2 ) ,
f 0 = − M N 2 2 π 2 , f 0 ′ = L + 11 / 6 6 π 2 , f Z = f 0 + q f 0 ′ , f W = f 0 + M W 2 f 0 ′ , f_0=-\frac{M_N^2}{2\pi^2},\qquad f_0'=\frac{L+11/6}{6\pi^2},\qquad f_Z=f_0+qf_0',\qquad f_W=f_0+M_W^2f_0', f 0 = − 2 π 2 M N 2 , f 0 ′ = 6 π 2 L + 11/6 , f Z = f 0 + q f 0 ′ , f W = f 0 + M W 2 f 0 ′ ,
δ = R Z Z f Z − R W W f W 2 , x = R A A f 0 2 − c 2 s 2 δ , y = R A A f 0 2 + δ , \delta=\frac{R_{ZZ}f_Z-R_{WW}f_W}{2},\qquad x=\frac{R_{AA}f_0}{2}-\frac{c^2}{s^2}\delta,\qquad y=\frac{R_{AA}f_0}{2}+\delta, δ = 2 R Z Z f Z − R W W f W , x = 2 R AA f 0 − s 2 c 2 δ , y = 2 R AA f 0 + δ ,
u = s c ( y − x ) + R A Z f Z , z = R Z Z 2 ( f Z + q f 0 ′ ) . (151) u=sc(y-x)+R_{AZ}f_Z,\qquad z=\frac{R_{ZZ}}{2}(f_Z+qf_0'). \tag*{(151)} u = sc ( y − x ) + R A Z f Z , z = 2 R Z Z ( f Z + q f 0 ′ ) . (151)
Then the two universal coefficients are
A T = c 2 x + s 2 y − z , A Q = − s 2 y − s c u + s 2 z . (152) A_T = c^2x+s^2y-z,\qquad A_Q=-s^2y-sc\,u+s^2z. \tag*{(152)} A T = c 2 x + s 2 y − z , A Q = − s 2 y − sc u + s 2 z . (152)
Here x , y x,y x , y encode the input shifts, u u u the neutral mixing, and z z z the Z Z Z -pole residue. These expressions use the SM-evolved electromagnetic input defined in § 6 . Together with δ a f = A T T f 3 + A Q Q f + δ a f d i r \delta a_f=A_TT_f^3+A_QQ_f+\delta a_f^{\mathrm{dir}} δ a f = A T T f 3 + A Q Q f + δ a f dir , they give the full local response through M N 0 M_N^0 M N 0 ; right-handed direct lepton terms vanish in the massless-lepton limit. The separate radiative-decay analysis is described in § 7.4 .
Gauge-invariant ultraviolet profiling at the pole. The four-dimensional response above can be spanned by independent, gauge-invariant local ultraviolet coefficients. This is more restrictive than simply introducing four arbitrary shifts of measured couplings. Besides Q H l ( 1 , 3 ) Q_{Hl}^{(1,3)} Q H l ( 1 , 3 ) , consider Q H W B Q_{HWB} Q H W B and the dimension-eight representative
Q B ∂ H = ( H † H ) ( ∂ μ B μ ν ) ( ∂ ρ B ρ ν ) . (153) Q_{B\partial H}=(H^\dagger H)(\partial_\mu B^{\mu\nu})(\partial^\rho B_{\rho\nu}). \tag*{(153)} Q B ∂ H = ( H † H ) ( ∂ μ B μν ) ( ∂ ρ B ρ ν ) . (153)
The latter may be eliminated using gauge equations of motion, provided all induced current and contact operators are retained. Here it is kept in the propagator representation, without adding its EOM images again. Define b W B b_{WB} b W B as the constant off-diagonal W 3 W^3 W 3 –B B B transverse form factor and b B b_B b B as the coefficient of q 2 q^2 q 2 in the B B BB B B form factor. These coordinates absorb the nonzero vev and normalization factors of the two Wilson coefficients; b W B b_{WB} b W B is dimensionless and b B b_B b B has mass dimension − 2 -2 − 2 . In the fixed ( e , M W , M Z ) (e,M_W,M_Z) ( e , M W , M Z ) input prescription, including the pole residue, their responses are
\begin{pmatrix} A_T \\ A_Q \end{pmatrix}_{\mathrm{UV}}
=
\begin{pmatrix}
c_W/s_W & -M_Z^2/2 \\
0 & s_W^2M_Z^2/2
\end{pmatrix}
\begin{pmatrix} b_{WB} \\ b_B \end{pmatrix}. \tag*{(154)}\end{pmatrix}
The determinant is M Z 2 s W c W / 2 ≠ 0 M_Z^2s_Wc_W/2\ne0 M Z 2 s W c W /2 = 0 . Together with the two independent lepton-current shifts this gives rank four. Writing d f ≡ δ a f d i r d_f\equiv\delta a_f^{\mathrm{dir}} d f ≡ δ a f dir for f = ν , e f=\nu,e f = ν , e , any dipole-induced local response can consequently be cancelled by choosing
b B = − 2 A Q s W 2 M Z 2 , b W B = − s W c W ( A T + A Q s W 2 ) , ( d ν , d e ) U V = − ( d ν , d e ) . (155) b_B=-\frac{2A_Q}{s_W^2M_Z^2},\qquad
b_{WB}=-\frac{s_W}{c_W}\left(A_T+\frac{A_Q}{s_W^2}\right),\qquad
(d_\nu,d_e)_{\mathrm{UV}}=-(d_\nu,d_e). \tag*{(155)} b B = − s W 2 M Z 2 2 A Q , b W B = − c W s W ( A T + s W 2 A Q ) , ( d ν , d e ) UV = − ( d ν , d e ) . (155)
This cancellation precedes the observable map and covariance weighting. With unrestricted coefficients in these directions, profiling therefore removes all dipole sensitivity of the retained local pole response, including in the presence of a nonzero experimental residual. Two free lepton currents alone span rank two; adding only Q H W B Q_{HWB} Q H W B gives rank three and does not generally remove the subleading charge response.
The same four directions are already spanned by dimension-six Warsaw operators. For Q H D = ∣ H † D μ H ∣ 2 Q_{HD}=|H^\dagger D_\mu H|^2 Q H D = ∣ H † D μ H ∣ 2 , the tree-level mass relations are
M W 2 = g 2 v 2 4 , M Z 2 = ( g 2 + g ′ 2 ) v 2 4 ( 1 + v 2 2 C H D ) . (156) M_W^2=\frac{g^2v^2}{4},\qquad
M_Z^2=\frac{(g^2+g'^2)v^2}{4}\left(1+\frac{v^2}{2}C_{HD}\right). \tag*{(156)} M W 2 = 4 g 2 v 2 , M Z 2 = 4 ( g 2 + g ′2 ) v 2 ( 1 + 2 v 2 C H D ) . (156)
Holding e e e , M W M_W M W , M Z M_Z M Z fixed gives δ s 2 = − c W 2 v 2 C H D / 2 \delta s^2=-c_W^2v^2C_{HD}/2 δ s 2 = − c W 2 v 2 C H D /2 for the Lagrangian mixing angle. Consequently, in the dimensionless coupling convention of eq. (B.35),
\begin{pmatrix} A_T^{(6)} \\ A_Q \end{pmatrix}_{\mathrm{UV}}
=
v^2
\begin{pmatrix}
c_W/s_W & (1-2s_W^2)/(4s_W^2) \\
0 & 1/4
\end{pmatrix}
\begin{pmatrix} C_{HWB} \\ C_{HD} \end{pmatrix}. \tag*{(157)}\end{pmatrix}
Its determinant is v 4 c W / ( 4 s W ) ≠ 0 v^4c_W/(4s_W)\ne0 v 4 c W / ( 4 s W ) = 0 . The two independent C H l ( 1 , 3 ) C_{Hl}^{(1,3)} C H l ( 1 , 3 ) directions then complete the rank-four response. Thus the identifiability result at the fixed Z Z Z pole does not rely on an independent dimension-eight coefficient. This statement concerns linearized pole couplings; the momentum dependence and correlations with other observables distinguish these operator choices.
The degeneracy applies to the retained local pole response. Off-pole, charged-current and direct-production observables probe additional structures, while UV correlations or finite priors restrict the allowed compensation. The boundary variations in eq. (B.40) therefore represent different hypotheses, not a probability distribution.
Finite-mass dependence and the local expansion
Section 6.4 distinguishes the heavy-mass expansion from the validity of the dipole EFT. Here we give the exact two-point remainder, specify the ingredients of a finite-mass pole prediction and compare the assembled amplitudes with their local limit. These results explain the mass qualifications on the sensitivity curves.
Nonlocal two-point remainder
The nonlocal part of the two-point function provides a quantitative diagnostic of the heavy-mass expansion. Define the twice-subtracted remainder of eq. (21) ,
R ( t ) = F 0 ( t ) − F 0 ( 0 ) − t F 0 ′ ( 0 ) , z = t / M N 2 . (158) R(t) = F_0(t) - F_0(0) - tF_0'(0), \qquad z = t/M_N^2. \tag*{(158)} R ( t ) = F 0 ( t ) − F 0 ( 0 ) − t F 0 ′ ( 0 ) , z = t / M N 2 . (158)
The constant and linear local terms, including their UV and finite-scheme dependence, cancel. For 0 < z < 1 0 < z < 1 0 < z < 1 the expression gives
R ( t ) = M N 2 6 π 2 r ( z ) , (159) R(t) = \frac{M_N^2}{6\pi^2}r(z), \tag*{(159)} R ( t ) = 6 π 2 M N 2 r ( z ) , (159)
r ( z ) = 2 − z − ( 1 − z ) ( 2 + z ) b ( z ) z + 1 − 7 6 z , (160) r(z) = \frac{2-z-(1-z)(2+z)b(z)}{z}+1-\frac{7}{6}z, \tag*{(160)} r ( z ) = z 2 − z − ( 1 − z ) ( 2 + z ) b ( z ) + 1 − 6 7 z , (160)
b ( z ) = 2 + 1 − z z ln ( 1 − z ) , (161) b(z) = 2+\frac{1-z}{z}\ln(1-z), \tag*{(161)} b ( z ) = 2 + z 1 − z ln ( 1 − z ) , (161)
r ( z ) = z 2 2 + 3 z 3 20 + z 4 15 + z 5 28 + O ( z 6 ) . (162) r(z) = \frac{z^2}{2}+\frac{3z^3}{20}+\frac{z^4}{15}+\frac{z^5}{28}+O(z^6). \tag*{(162)} r ( z ) = 2 z 2 + 20 3 z 3 + 15 z 4 + 28 z 5 + O ( z 6 ) . (162)
Since F 0 ( 0 ) = − M N 2 / ( 2 π 2 ) F_0(0)=-M_N^2/(2\pi^2) F 0 ( 0 ) = − M N 2 / ( 2 π 2 ) , the ratio is R / F 0 ( 0 ) = − r ( z ) / 3 R/F_0(0)=-r(z)/3 R / F 0 ( 0 ) = − r ( z ) /3 . The expression agrees algebraically with the Package-X result. In particular, its constant and linear terms before subtraction are proportional to − 3 + z [ 1 / ϵ − γ E − ln π + ln ( μ P X 2 / M N 2 ) + 11 / 6 ] -3+z[1/\epsilon-\gamma_E-\ln\pi+\ln(\mu_{\mathrm{PX}}^2/M_N^2)+11/6] − 3 + z [ 1/ ϵ − γ E − ln π + ln ( μ PX 2 / M N 2 ) + 11/6 ] in the scalar-integral convention of appendix § 11 .§ 6 ; subtracting them reproduces eqs. (C.3)–(C.4). This also avoids discarding the finite effects of evanescent factors in the local terms.
The numerical values at t = M Z 2 t=M_Z^2 t = M Z 2 are given in § 6.5 .
The summed neutrino vertex contains a UV pole proportional to q 2 ( c W 2 C W − 2 s W 2 C B ) 2 q^2(c_W^2C_W-2s_W^2C_B)^2 q 2 ( c W 2 C W − 2 s W 2 C B ) 2 . Its cancellation of spurious 1 / q 2 1/q^2 1/ q 2 terms and its checked zero-momentum limit do not remove the need for momentum-dependent local counterterms at the pole. A finite-mass prediction must recombine the vertex, WFR and input terms with consistently renormalized bosonic and dimension-eight contributions. The same boundary prescription must cover the dimension-six currents and the higher-derivative local terms. A boundary imposed at Λ \Lambda Λ is transported to M N M_N M N by the corresponding running.
Finite-mass pole amplitudes
The scalar functions A 0 , B 0 , C 0 A_0,B_0,C_0 A 0 , B 0 , C 0 can be evaluated with their full mass and momentum dependence while keeping the perturbative order fixed. This removes the expansion in electroweak scales over M N M_N M N , but leaves the independent UV coefficients of the dipole EFT to be specified. The relevant truncation parameters are the ratios appearing in each amplitude, rather than a universal estimate based on v 2 / M N 2 v^2/M_N^2 v 2 / M N 2 alone.
For example, the universal two-point function is already known exactly in eq. (21) . A useful organization of its renormalized loop part is
F 0 , l o o p r e n ( t ; μ ) = − M N 2 2 π 2 + t b l o o p ( μ ) + M N 2 6 π 2 r ( t / M N 2 ) , (163) F^{\mathrm{ren}}_{0,\mathrm{loop}}(t;\mu)=-\frac{M_N^2}{2\pi^2}+t\,b_{\mathrm{loop}}(\mu)+\frac{M_N^2}{6\pi^2}r(t/M_N^2), \tag*{(163)} F 0 , loop ren ( t ; μ ) = − 2 π 2 M N 2 + t b loop ( μ ) + 6 π 2 M N 2 r ( t / M N 2 ) , (163)
where r r r is the unexpanded function in eqs. (C.3)–(C.4) and b l o o p b_{\mathrm{loop}} b loop is the renormalized slope in the declared subtraction convention. This decomposition is exact for this loop below its N N N –massless-particle threshold, not just a truncation at t 2 t^2 t 2 . The full two-point matrix also contains independent local terms,
F i j r e n ( t ) = d i d j F 0 , l o o p r e n ( t ) + F i j l o c a l ( t ; μ ) . (164) F^{\mathrm{ren}}_{ij}(t)=d_i d_j F^{\mathrm{ren}}_{0,\mathrm{loop}}(t)+F^{\mathrm{local}}_{ij}(t;\mu). \tag*{(164)} F ij ren ( t ) = d i d j F 0 , loop ren ( t ) + F ij local ( t ; μ ) . (164)
The local matrix contains the independent UV contributions and need not share the gauge-basis alignment of the dipole term. Its finite boundary and the subtraction convention must therefore be specified even when the loop function is evaluated without a mass expansion.
The analogous physical neutral-current correction must combine all retained pieces at the pole,
δ g f r e n ( M Z 2 ) = δ g f 1 P I ( M Z 2 ) + δ g f e x t e r n a l f i e l d s + δ g f i n p u t s + δ g f l o c a l , (165) \delta g_f^{\mathrm{ren}}(M_Z^2)=\delta g_f^{\mathrm{1PI}}(M_Z^2)+\delta g_f^{\mathrm{external\ fields}}+\delta g_f^{\mathrm{inputs}}+\delta g_f^{\mathrm{local}}, \tag*{(165)} δ g f ren ( M Z 2 ) = δ g f 1PI ( M Z 2 ) + δ g f external fields + δ g f inputs + δ g f local , (165)
with the same sign, coupling normalization and subtraction convention. The schematic external-field term includes both fermion and gauge-field normalization and neutral mixing where applicable. The input term uses the chosen { α , M W , M Z } \{\alpha,M_W,M_Z\} { α , M W , M Z } scheme with self-energies evaluated at the appropriate momenta. The full mass dependence of a pole observable requires these contributions together; the exact neutrino vertex or two-point function alone is insufficient.
At double dipole insertion the dimension-eight local terms enter at the same order in the portal expansion as the loop. Their finite boundary conditions are part of a physical finite-mass prediction. Below we test both the asymptotic expansion and its effect on the projected reach at the masses represented in the figures.
The heavy-mass expansion of the pole amplitudes is controlled by ratios such as M Z 2 / M N 2 M_Z^2/M_N^2 M Z 2 / M N 2 and M W 2 / M N 2 M_W^2/M_N^2 M W 2 / M N 2 . For example, M Z 2 / M N 2 M_Z^2/M_N^2 M Z 2 / M N 2 decreases smoothly from about 0.21 at 200 GeV to 0.033 at 500 GeV. There is no physical transition at 500 GeV: the line-style change in Figure 6 flags the more cautious interpretation of the low-mass extrapolation. Neither that reference mass nor a small two-point remainder alone bounds the error of all current and input contributions; Table 6 quantifies their combined effect in representative directions. This heavy-mass expansion issue is distinct from the ultraviolet coefficient-scale guide in the same figure.
Table 6. Finite-mass corrections on the two axes and two diagonals. Each entry is the largest value among these directions, r B = 2 , 5 r_B = 2, 5 r B = 2 , 5 , and the conservative and aggressive SM covariance scenarios, with the SM-evolved electromagnetic input used throughout.
The low-mass curves in the direct-search comparison illustrate the local prediction under these qualifications. A fit below M Z M_Z M Z would also require the causal continuation and possible real Z → N ν Z\to N\nu Z → N ν final states, beyond shifts of light-fermion widths.
Heavy-mass expansion of the assembled pole response
We compare the local response with the unexpanded scalar amplitudes, combining the Born-normalized electron and neutrino vertices, one net external-fermion residue for each pair of external legs, and the universal vacuum-polarization and input corrections. The exact and expanded expressions use the same M S ‾ \overline{\mathrm{MS}} MS convention and the same input scheme. No independent local UV coefficients are added to either side of this comparison. Writing
δ a f e x = δ a f ( 6 ) + δ a f ( 8 ) + R f , (166) \delta a_f^{\mathrm{ex}}=\delta a_f^{(6)}+\delta a_f^{(8)}+\mathcal{R}_f, \tag*{(166)} δ a f ex = δ a f ( 6 ) + δ a f ( 8 ) + R f , (166)
where the two local terms scale as M N 2 M_N^2 M N 2 and M N 0 M_N^0 M N 0 at fixed dipoles, we evaluate M N = 1 , 2 , 5 , 10 , 20 , 50 M_N = 1, 2, 5, 10, 20, 50 M N = 1 , 2 , 5 , 10 , 20 , 50 TeV and μ / M N = 1 / 2 , 1 , 2 \mu/M_N = 1/2, 1, 2 μ / M N = 1/2 , 1 , 2 , with C B = 0.3 T e V − 2 C_B = 0.3\ \mathrm{TeV}^{-2} C B = 0.3 TeV − 2 and C W = 0.2 T e V − 2 C_W = 0.2\ \mathrm{TeV}^{-2} C W = 0.2 TeV − 2 . The calculation uses 65-digit arithmetic. For the nonvanishing leading responses, the endpoint logarithmic slopes of ∣ δ a f e x / δ a f ( 6 ) − 1 ∣ |\delta a_f^{\mathrm{ex}}/\delta a_f^{(6)} - 1| ∣ δ a f ex / δ a f ( 6 ) − 1∣ lie between − 2.003 -2.003 − 2.003 and − 1.997 -1.997 − 1.997 . The labels f = ν L , e L , e R f = \nu_L, e_L, e_R f = ν L , e L , e R refer to the left-handed neutrino and the left- and right-handed charged-lepton couplings. For each of these couplings and each of the three choices of μ / M N \mu/M_N μ / M N , the omitted remainder R f \mathcal{R}_f R f is smaller than 8 × 10 − 4 8 \times10^{-4} 8 × 1 0 − 4 of the subleading local term δ a f ( 8 ) \delta a_f^{(8)} δ a f ( 8 ) at the sampled masses from 5 to 50 TeV. The leading e R e_R e R response vanishes identically, so only the subleading comparison is used in that channel. The exact self-energy expansion also reproduces the finite combination 7 − 6 ln ( μ 2 / M N 2 ) 7 - 6\ln(\mu^2/M_N^2) 7 − 6 ln ( μ 2 / M N 2 ) , while scalar quadrature of the bosonic kernel recovers its 11 / 6 11/6 11/6 slope constant.
The asymptotic agreement tests the expansion of the same underlying amplitudes, rather than an independent calculation of the diagrams. For the masses shown in the reach plots, a more direct measure is
ϵ O = ∥ δ O e x − δ O l o c ∥ V ∥ δ O l o c ∥ V , ∥ x ∥ V 2 = x T V − 1 x . (167) \epsilon_O = \frac{\lVert\delta O_{\mathrm{ex}}-\delta O_{\mathrm{loc}}\rVert_V}{\lVert\delta O_{\mathrm{loc}}\rVert_V}, \qquad\lVert x\rVert_V^2 = x^T V^{-1}x. \tag*{(167)} ϵ O = ∥ δ O loc ∥ V ∥ δ O ex − δ O loc ∥ V , ∥ x ∥ V 2 = x T V − 1 x . (167)
Table 6 compares the assembled responses with the same inputs and common UV boundaries. It includes the full scalar mass dependence of the lepton currents and external residues, as well as the bosonic functions. The small differences at 500 GeV and 1 TeV support the local approximation in the displayed benchmark directions. The larger effect at 200 GeV motivates retaining the distinct line style for the low-mass extrapolation; none of these reference masses is a physical threshold.4
Higgs-background relations and radiative-decay kernels
The radiative-decay analysis in § 7 uses the Higgs partners of the dipole vertices. We give the auxiliary two-Higgs structures that distinguish derivative terms, the tensor reduction relating the decay bubble to the two-point kernel, and its soft-Higgs limit. These results concern the neutral two-dipole contribution, rather than a matching of the complete Higgs sector.
Independent Higgs-background structures
Two independent derivative structures. Independent Higgs momenta distinguish derivative interactions that reduce to the same structure in a constant Higgs background. The CP-even h h V V hhVV hhV V amplitude can be resolved onto five off-shell tensors in D = 4 − 2 ϵ D = 4 - 2\epsilon D = 4 − 2 ϵ , keeping the external virtualities independent. The separately calculated γ Z \gamma Z γ Z and Z Z ZZ Z Z amplitudes obey the neutral coupling relations. For X μ ν γ = h F μ ν X^\gamma_{\mu\nu}=hF_{\mu\nu} X μν γ = h F μν , define
C γ = n F e 2 ( 2 C B + C W ) 2 16 π 2 , L N = ln μ 2 M N 2 . (168) C_\gamma=\frac{n_F e^2(2C_B+C_W)^2}{16\pi^2},\qquad L_N=\ln\frac{\mu^2}{M_N^2}. \tag*{(168)} C γ = 16 π 2 n F e 2 ( 2 C B + C W ) 2 , L N = ln M N 2 μ 2 . (168)
The local finite result, through four powers of external momenta, is represented in these projections by
Δ L h h γ γ = C γ [ M N 2 4 X μ ν γ X μ ν γ + 1 2 ( L N 6 − 1 36 ) ( ∂ ρ X μ ν γ ) ( ∂ ρ X μ ν γ ) − 1 2 ( 2 L N 3 + 5 9 ) ( ∂ μ X μ ν γ ) ( ∂ ρ X ρ ν γ ) ] . (169) \Delta\mathcal{L}_{h h\gamma\gamma}=C_\gamma\left[\frac{M_N^2}{4}X^\gamma_{\mu\nu}X^{\mu\nu\gamma}+\frac{1}{2}\left(\frac{L_N}{6}-\frac{1}{36}\right)(\partial_\rho X^\gamma_{\mu\nu})(\partial^\rho X^{\mu\nu\gamma})-\frac{1}{2}\left(\frac{2L_N}{3}+\frac{5}{9}\right)(\partial_\mu X^{\mu\nu\gamma})(\partial_\rho X^\rho{}_{\nu\gamma})\right]. \tag*{(169)} Δ L hhγ γ = C γ [ 4 M N 2 X μν γ X μν γ + 2 1 ( 6 L N − 36 1 ) ( ∂ ρ X μν γ ) ( ∂ ρ X μν γ ) − 2 1 ( 3 2 L N + 9 5 ) ( ∂ μ X μν γ ) ( ∂ ρ X ρ ν γ ) ] . (169)
The coefficient includes the three flavour copies of the aligned benchmark through n F = 3 n_F=3 n F = 3 . The charged W + W − h h W^+W^-hh W + W − hh kernel agrees in all five CP-even projections, in the massless charged-lepton and vanishing Yukawa limit. Its normalization is obtained by replacing C γ C_\gamma C γ with n F e 2 C W 2 / ( 16 π 2 s W 2 ) n_F e^2 C_W^2/(16\pi^2s_W^2) n F e 2 C W 2 / ( 16 π 2 s W 2 ) . This equality holds after reduction, not term by term in the integrands.
For the electroweak composite field
X μ ν = ( g ′ C B B μ ν 1 + g C W 2 σ I W μ ν I ) H ~ , (170) \mathcal{X}_{\mu\nu}=\left(g' C_B B_{\mu\nu}\mathbf{1}+\frac{g C_W}{2}\sigma^I W^I_{\mu\nu}\right)\widetilde{H}, \tag*{(170)} X μν = ( g ′ C B B μν 1 + 2 g C W σ I W μν I ) H , (170)
a compatible local representative is
Δ L r e p = α 0 X μ ν † X μ ν + α 1 ( D ρ X μ ν ) † D ρ X μ ν + α 2 ( D μ X μ ν ) † D ρ X ρ ν , (171) \Delta\mathcal{L}_{\mathrm{rep}}=\alpha_0\mathcal{X}^{\dagger}_{\mu\nu}\mathcal{X}^{\mu\nu}+\alpha_1(D_\rho\mathcal{X}_{\mu\nu})^\dagger D^\rho\mathcal{X}^{\mu\nu}+\alpha_2(D_\mu\mathcal{X}^{\mu\nu})^\dagger D_\rho\mathcal{X}^{\rho}{}_{\nu}, \tag*{(171)} Δ L rep = α 0 X μν † X μν + α 1 ( D ρ X μν ) † D ρ X μν + α 2 ( D μ X μν ) † D ρ X ρ ν , (171)
α 0 = n F M N 2 8 π 2 , α 1 = n F ( 6 L N − 1 ) 144 π 2 , α 2 = − n F ( 6 L N + 5 ) 36 π 2 . (172) \alpha_0=\frac{n_FM_N^2}{8\pi^2},\qquad\alpha_1=\frac{n_F(6L_N-1)}{144\pi^2},\qquad\alpha_2=-\frac{n_F(6L_N+5)}{36\pi^2}. \tag*{(172)} α 0 = 8 π 2 n F M N 2 , α 1 = 144 π 2 n F ( 6 L N − 1 ) , α 2 = − 36 π 2 n F ( 6 L N + 5 ) . (172)
The two-point combination is recovered: α 1 + α 2 / 2 = − n F ( 6 L N + 11 ) / ( 144 π 2 ) \alpha_1+\alpha_2/2=-n_F(6L_N+11)/(144\pi^2) α 1 + α 2 /2 = − n F ( 6 L N + 11 ) / ( 144 π 2 ) . The independent Higgs momenta separate the two derivative coefficients, which a two-point function alone cannot distinguish. Equations (D.2)–(D.5) describe these two-field-strength projections. Fixing additional non-Abelian structures would require higher-point amplitudes beyond this sector.
Finite rational terms are essential even in this CP-even calculation. If the projected vector is P / ϵ + F P/\epsilon+F P / ϵ + F and its Gram matrix is G ( D ) = G 4 − 2 ϵ G 4 ′ + ⋯ G(D)=G_4-2\epsilon G'_4+\cdots G ( D ) = G 4 − 2 ϵ G 4 ′ + ⋯ , the finite coefficients obey
G 4 c f i n = F + 2 G 4 ′ c p o l e . (173) G_4c_{\mathrm{fin}}=F+2G'_4c_{\mathrm{pole}}. \tag*{(173)} G 4 c fin = F + 2 G 4 ′ c pole . (173)
Keeping only the four-dimensional Gram matrix before subtracting the pole would change the finite constants.
The finite h Z e e hZee h Z ee projections with generic couplings provide an additional check of the current convention. They do not fix coefficients invisible to the pole projections. A complete non-Abelian higher-point matching and a Z H ZH Z H production likelihood lie outside this analysis.
Tensor reduction of the neutral Higgs bubble
Let S μ ( k ) = [ γ μ , k̸ ] S^\mu(k)=[\gamma^\mu,\not{k}] S μ ( k ) = [ γ μ , k ] , with irrelevant common dipole normalization suppressed. The tensor integral multiplying the CP-even trace has the form I α β = A g α β + B p α p β I_{\alpha\beta}=Ag_{\alpha\beta}+Bp_\alpha p_\beta I α β = A g α β + B p α p β . The Clifford identities
γ α S ν ( k ) γ α = ( D − 4 ) S ν ( k ) , (174) \gamma_\alpha S^\nu(k)\gamma^\alpha=(D-4)S^\nu(k), \tag*{(174)} γ α S ν ( k ) γ α = ( D − 4 ) S ν ( k ) , (174)
p̸ S μ ( p ) p̸ = − p 2 S μ ( p ) (175) \not{p}S^\mu(p)\not{p}=-p^2S^\mu(p) \tag*{(175)} p S μ ( p ) p = − p 2 S μ ( p ) (175)
imply
I α β tr [ S μ ( p ) γ α S ν ( k ) γ β ] = [ ( D − 4 ) A − p 2 B ] tr [ S μ ( p ) S ν ( k ) ] . (176) I_{\alpha\beta}\operatorname{tr}[S^\mu(p)\gamma^\alpha S^\nu(k)\gamma^\beta]=[(D-4)A-p^2B]\operatorname{tr}[S^\mu(p)S^\nu(k)]. \tag*{(176)} I α β tr [ S μ ( p ) γ α S ν ( k ) γ β ] = [( D − 4 ) A − p 2 B ] tr [ S μ ( p ) S ν ( k )] . (176)
Thus the same scalar kernel that appears when the gauge momenta coincide also multiplies the physical Higgs tensor. The factor ( D − 4 ) A (D-4)A ( D − 4 ) A must not be dropped before integration: its UV pole supplies a finite contribution, already retained in F 0 F_0 F 0 . The CP-odd trace cancels in the real, CP-even coupling sum considered here. Each gauge Ward identity follows from contracting its commutator with its own momentum.
Soft-Higgs relation
At fixed M N M_N M N and Wilson coefficients, d γ d_\gamma d γ is proportional to v v v and the neutral fermion-bubble kernel is independent of v v v . Consequently
− 1 4 ∂ ∂ v [ n F d γ 2 F 0 ( 0 ) ] = − n F d γ 2 F 0 ( 0 ) 2 v . (177) -\frac{1}{4}\frac{\partial}{\partial v}\left[n_F d_\gamma^2 F_0(0)\right] = -\frac{n_F d_\gamma^2 F_0(0)}{2v}. \tag*{(177)} − 4 1 ∂ v ∂ [ n F d γ 2 F 0 ( 0 ) ] = − 2 v n F d γ 2 F 0 ( 0 ) . (177)
This reproduces the zero-momentum two-dipole Higgs-bubble coupling. The relation applies to this bubble contribution, with the heavy mass held fixed, and is not a low-energy theorem for the entire Higgs amplitude.
Statistical implementation of the FCC-ee forecast
The forecast in our analysis compares the dipole-induced shifts with the correlated experimental and SM uncertainties. This appendix specifies the input-coordinate conversion, the confidence construction and their numerical checks, supporting the results in § 8 .
Observable set and correlated uncertainties
We use the twelve nonredundant pole pseudo-observables defined in § 8.1 . The projections and correlated SM covariance scenarios are taken from the public dataset accompanying ref. [43 ] . This set includes all three charged-lepton flavours. The corresponding A F B A_{\mathrm{FB}} A FB and epsilon combinations contain overlapping information and are not added as independent measurements. The BSM correction is the linear derivative of the tree pole observables with respect to the matched chiral couplings; the precision SM uncertainty model is additive. Missing higher-order BSM electroweak corrections are not computed by this procedure.
The public covariance is supplied jointly with an electromagnetic input coordinate in the { G F , M W , M Z } \{G_F, M_W, M_Z\} { G F , M W , M Z } scheme. We replace G F G_F G F by that measured coordinate rather than adding a second parametric-error matrix. For the joint residual vector ( δ O , δ α s r c ) (\delta O, \delta\alpha_{\mathrm{src}}) ( δ O , δ α src ) , define
K = ( I , − r ) , r i = ∂ O i / ∂ G F k s r c ∂ Δ α / ∂ G F , V α M W = K ( V e x p + V S M ) K T . (178) K = (I,-r), \qquad r_i = \frac{\partial O_i/\partial G_F}{k_{\mathrm{src}}\,\partial\Delta\alpha/\partial G_F}, \qquad V_{\alpha M_W} = K(V_{\mathrm{exp}}+V_{\mathrm{SM}})K^T. \tag*{(178)} K = ( I , − r ) , r i = k src ∂ Δ α / ∂ G F ∂ O i / ∂ G F , V α M W = K ( V exp + V SM ) K T . (178)
where the forecast uses the local affine convention δ α s r c = k s r c δ Δ α \delta\alpha_{\mathrm{src}} = k_{\mathrm{src}}\delta\Delta\alpha δ α src = k src δ Δ α and k s r c = α r e f , s r c 2 / α ( 0 ) k_{\mathrm{src}} = \alpha_{\mathrm{ref,src}}^2/\alpha(0) k src = α ref , src 2 / α ( 0 ) with α r e f , s r c = 0.00780832 \alpha_{\mathrm{ref,src}} = 0.00780832 α ref , src = 0.00780832 . This convention is an explicit covariance-transfer assumption. The central SM-evolved coupling is α ˉ Z − 1 = 128.9468 \bar{\alpha}_Z^{-1} = 128.9468 α ˉ Z − 1 = 128.9468 , using Δ α S M = 0.05903 \Delta\alpha_{\mathrm{SM}} = 0.05903 Δ α SM = 0.05903 . The source coordinate has experimental uncertainty 4.81098 × 10 − 8 4.81098 \times10^{-8} 4.81098 × 1 0 − 8 ; including the source SM covariance gives marginal uncertainties ( 7.429 , 3.905 , 2.016 ) × 10 − 7 (7.429, 3.905, 2.016) \times10^{-7} ( 7.429 , 3.905 , 2.016 ) × 1 0 − 7 in the current, conservative and aggressive scenarios. These are uncertainties of α s r c \alpha_{\mathrm{src}} α src , whose full correlations are propagated by K K K , rather than additional independent errors on α ˉ Z \bar{\alpha}_Z α ˉ Z . We do not identify the source central value with our running-α \alpha α central value: the expected residuals are recentered at the SM. The underlying SM Jacobians are obtained from ref. [38 ] . The G F G_F G F error direction cancels and the shared-input derivatives obey the chain rule in the new scheme. The uncertainty of the electromagnetic measurement is included once; the G F G_F G F measurement is not retained as another observable.
The normalization of the input coordinate is fixed by separating the nonparametric and shared-input contributions to the public covariance. The source Jacobian reconstructs the remaining joint parametric part. The maximum residual, normalized to the product of marginal uncertainties, is below 2 × 10 − 6 2 \times10^{-6} 2 × 1 0 − 6 . A consistent rescaling of the source coordinate, its covariance and its Jacobian leaves the forecast unchanged. As a diagnostic, changing only the normalization to our running-α \alpha α reference changes the four sections at 1 TeV in Table 2 by less than 0.36% across the three theory scenarios. We use the source normalization throughout; this comparison tests the coordinate convention and is not included as an additional uncertainty.
The finite dipole charge condition is also specified by this choice of input. With photon inverse propagator q 2 [ 1 + F A A ( q 2 ) ] q^2[1+F_{AA}(q^2)] q 2 [ 1 + F AA ( q 2 )] and Π A Z ( 0 ) = 0 \Pi_{AZ}(0)=0 Π A Z ( 0 ) = 0 , the parameter shift that holds the Thomson charge fixed is
δ e e ∣ d i p = F A A ( 0 ) 2 , Δ F = Re [ F A A ( M Z 2 ) − F A A ( 0 ) ] = 3 d q 2 M Z 2 6 π 2 ( L B + 11 6 ) (179) \left.\frac{\delta e}{e}\right|_{\mathrm{dip}} = \frac{F_{AA}(0)}{2}, \qquad\Delta F = \operatorname{Re}[F_{AA}(M_Z^2)-F_{AA}(0)] = \frac{3d_q^2 M_Z^2}{6\pi^2}\left(L_B+\frac{11}{6}\right) \tag*{(179)} e δ e dip = 2 F AA ( 0 ) , Δ F = Re [ F AA ( M Z 2 ) − F AA ( 0 )] = 6 π 2 3 d q 2 M Z 2 ( L B + 6 11 ) (179)
through the local order retained. Thus Δ F \Delta F Δ F remains in the renormalized photon propagation. An ideal effective charge defined from photon exchange at M Z M_Z M Z would instead absorb this difference, shifting the fixed-input response by ( T f 3 − s W 2 Q f ) Δ F / 2 (T_f^3-s_W^2Q_f)\Delta F/2 ( T f 3 − s W 2 Q f ) Δ F /2 . Comparing these two input prescriptions with the same covariance changes the coefficient reach by less than 0.96% at 500 GeV and 0.26% at 1 TeV on the axes and diagonals, for both boundaries and the conservative/aggressive scenarios. This comparison isolates photon propagation, not the additional process-dependent corrections to an experimental extraction.5
The current, conservative and aggressive SM scenarios describe different levels of theoretical calculation accuracy, with the correlations of the public forecast. We represent these projected uncertainties by Gaussian nuisance parameters. The coverage statement below is conditional on this statistical model; a fixed unknown theoretical bias would require a different treatment.
LEP precision as a reference. The LEP reference in Table 2 uses the flavour-universal pole data distributed with ref. [38 ] , including the luminosity correction of ref. [44 ] . The experimental standard deviations of ( Γ Z , σ h 0 , R ℓ , A F B 0 , ℓ ) (\Gamma_Z,\sigma_h^0,R_\ell,A_{\mathrm{FB}}^{0,\ell}) ( Γ Z , σ h 0 , R ℓ , A FB 0 , ℓ ) are 0.0023 GeV, 0.0325 nb, 0.0247 and 0.0010, respectively. Their correlations with the inputs are retained. If C C C denotes the joint measurement covariance of observables O O O and inputs p p p , the covariance of the pole residuals is
V L E P = C O O + J C p p J T − J C p O − C O p J T + V S M , J = ∂ O S M ∂ p . (180) V_{\mathrm{LEP}} = C_{OO} + J C_{pp} J^T - J C_{pO} - C_{Op}J^T + V_{\mathrm{SM}}, \qquad J = \frac{\partial O_{\mathrm{SM}}}{\partial p}. \tag*{(180)} V LEP = C O O + J C pp J T − J C pO − C O p J T + V SM , J = ∂ p ∂ O SM . (180)
Here p = ( Δ α , M W , M Z , M H , m t , α s ) p=(\Delta\alpha,M_W,M_Z,M_H,m_t,\alpha_s) p = ( Δ α , M W , M Z , M H , m t , α s ) , with the same central values as in the FCC-ee response; V S M V_{\mathrm{SM}} V SM contains the current SM theory uncertainties. The dipole response follows from the same coupling shifts as the FCC-ee forecast, with δ A F B 0 , ℓ = ( 3 / 2 ) A ℓ δ A ℓ \delta A_{\mathrm{FB}}^{0,\ell}=(3/2)A_\ell\delta A_\ell δ A FB 0 , ℓ = ( 3/2 ) A ℓ δ A ℓ in the flavour-universal limit. Both response matrices have rank three. Centring the reference on the SM removes dependence on fluctuations in the observed LEP central values, so that the comparison measures expected precision rather than the relative strength of observed limits.
Perturbative accuracy and running. The forecast combines precision SM predictions with a leading one-loop portal response. To specify its perturbative accuracy, we distinguish higher-order SM improvements from corrections to the BSM amplitude. One-loop electroweak SMEFT predictions and input-scheme dependence are studied in refs. [47 , 48 ] ; loop sensitivity at Tera-Z is explored in refs. [49 , 50 ] . Recent two-loop and NLL studies [51 , 52 ] address higher perturbative orders than the portal contribution considered here. The gauge and Yukawa contributions to the dimension-six νSMEFT running are available in refs. [14 , 15 ] , with an implementation in wilson [16 ] . Inserting one-loop running of C B , C W C_B,C_W C B , C W into our one-loop quadratic threshold changes that contribution at two-loop order. We therefore omit this effect at fixed one-loop order. Large logarithms would instead require a consistent renormalization-group improvement of the threshold calculation.
Common boundary and confidence construction
A common boundary prescription gives an unambiguous meaning to dipole dominance: every independent local coefficient is set to zero at the same scale Λ B \Lambda_B Λ B , in the same subtraction scheme. Equation (B.40) then combines its evolution with the HNL threshold. We compare r B = Λ B / M N = 2 , 5 r_B = \Lambda_B/M_N = 2,5 r B = Λ B / M N = 2 , 5 . Running of the dipoles inside the one-loop response is beyond the fixed order used here. The two choices are UV hypotheses; shading between them is neither a confidence interval nor an estimate of the omitted heavy-mass terms.
For illustration, the same-sign section ∣ C B ∣ = ∣ C W ∣ |C_B|=|C_W| ∣ C B ∣ = ∣ C W ∣ at M N = 1 M_N=1 M N = 1 TeV varies with the common boundary as follows, in T e V − 2 \mathrm{TeV}^{-2} TeV − 2 :
Table 9.
The reach strengthens with increasing r B r_B r B because the longer evolution interval enhances the induced local response. This dependence concerns the UV boundary, while the experimental precision is unchanged. Identifying the boundary scale with a mediator mass requires the power counting of a specified completion, particularly when that mass is close to M N M_N M N . Similarly, finite constants and independent Wilson coefficients must be transformed together under a change of subtraction scheme. Their combined prediction is scheme independent to the retained order; varying only the constants at a fixed zero boundary defines a different physical hypothesis.
At fixed M N M_N M N , r B r_B r B the observable shift is
δ O = B t , t = ( C B 2 , C B C W , C W 2 ) T . (181) \delta O = Bt,\qquad t = (C_B^2,C_BC_W,C_W^2)^T. \tag*{(181)} δ O = B t , t = ( C B 2 , C B C W , C W 2 ) T . (181)
The 12 × 3 12\times3 12 × 3 matrix B B B contains the response of each observable to the three quadratic monomials, at the specified mass and boundary. After weighting by the inverse covariance and projecting out any profiled ultraviolet directions, the linear signal space has rank k k k . We construct a Gaussian-mean confidence ellipsoid in this space and intersect it with the surface generated by the two dipole coefficients. For the expected SM dataset, the boundary is
t T B T V − 1 B t = χ k 2 ( 0.95 ) , χ 3 2 ( 0.95 ) ≃ 7.81. (182) t^T B^T V^{-1} Bt = \chi_k^2(0.95),\qquad\chi_3^2(0.95)\simeq7.81. \tag*{(182)} t T B T V − 1 B t = χ k 2 ( 0.95 ) , χ 3 2 ( 0.95 ) ≃ 7.81. (182)
The rank is that of the linear space of quadratic monomials, not the number of dipole parameters. Their physical image satisfies t 2 2 = t 1 t 3 t_2^2=t_1t_3 t 2 2 = t 1 t 3 , t 1 , t 3 ≥ 0 t_1,t_3\ge0 t 1 , t 3 ≥ 0 , but the three-component Gaussian-mean region is constructed before imposing this relation. Intersecting that region with the physical surface preserves inclusion of the true point whenever it belongs to the Gaussian-mean region. This provides a direct coverage argument without relying on a regular two-parameter tangent plane at the SM origin. The value 5.99 often used for two-parameter contours is instead the 95th percentile of a chi-square distribution with two degrees of freedom. It applies to a regular two-parameter likelihood-ratio construction, not automatically to any plot with two axes. Here the observable shifts are quadratic in C B , C W C_B,C_W C B , C W , and their first derivatives vanish at the SM point. The usual regularity argument for that likelihood-ratio prescription therefore does not apply there. Our threshold 7.81 follows from the three-dimensional Gaussian-mean region specified above; it is not an approximation to the two-degree-of-freedom threshold. A different construction on the physical surface would need its own coverage calibration. The numerical dependence on the chosen threshold is weak: C r e a c h ∝ ( χ c r i t 2 ) 1 / 4 C_{\mathrm{reach}} \propto\left(\chi_{\mathrm{crit}}^2\right)^{1/4} C reach ∝ ( χ crit 2 ) 1/4 , so thresholds of 5.99 and 3.84 would rescale the sections by 0.936 and 0.837, respectively, at fixed response and covariance. These factors describe the threshold dependence only; they do not assign coverage to another construction or make the direct-search likelihood equivalent to the precision forecast.
The inverse covariance here denotes its action after any specified nuisance projection. The statistic measures distance from the expected SM mean, rather than from a best fit constrained to the quadratic surface. This choice is useful at C B = C W = 0 C_B = C_W = 0 C B = C W = 0 , where the ordinary coupling Jacobian vanishes and the regularity assumptions of a two-parameter Wilks construction fail. With known Gaussian covariance the construction has nominal pointwise coverage; 200000 Gaussian trials for each of ranks 1, 2, 3 give coverages 0.95006, 0.94966, 0.94964. Axis and diagonal values are sections of one joint region, not separately profiled 1D intervals.
The passage from coupling shifts to measured widths and asymmetries introduces a further, separate approximation. For example, a partial width contains ( a f S M + δ a f ) 2 (a_f^{\mathrm{SM}} + \delta a_f)^2 ( a f SM + δ a f ) 2 : the linear observable response retains 2 a f S M δ a f 2a_f^{\mathrm{SM}}\delta a_f 2 a f SM δ a f and omits ( δ a f ) 2 (\delta a_f)^2 ( δ a f ) 2 . Evaluating the full tree-level expressions with the same calculated coupling shifts tests this approximation without calculating any new amplitudes. For the four coefficient sections at 1 TeV, both boundaries and the conservative and aggressive scenarios, the difference from the linear response is below 7 × 10 − 5 7 \times10^{-5} 7 × 1 0 − 5 of the signal, measured with the covariance-weighted norm. Linearizing the observable map is thus numerically adequate for these benchmarks. This comparison does not estimate corrections to δ a f \delta a_f δ a f itself, such as higher-order portal amplitudes or terms omitted from the heavy-mass expansion.
The entries in Table 2 follow from the information matrix B T V − 1 B B^{T}V^{-1}B B T V − 1 B with the threshold χ 3 2 ( 0.95 ) \chi^2_3(0.95) χ 3 2 ( 0.95 ) . Independent reconstruction from the response and covariance matrices agrees with all 32 FCC-ee entries to the displayed precision; the LEP reference uses the covariance specified above.
Ultraviolet interpretation: power counting and correlated responses
A UV interpretation of the FCC-ee reach must address two issues: the conditions under which dipoles dominate the pole observables, and the effect of additional interactions. To develop the interpretation of § 8 , we first discuss the size of the relevant coefficients and the stability of negligible mixing. We then express reinforcement and compensation in observable space and illustrate them with heavy-lepton representations.
Radiative stability and ultraviolet power counting
The tree/loop classification in weakly coupled UV completions is discussed in refs. [53 , 54 , 55 ] and implemented for selected one-loop SMEFT operators in the SOLD dictionary [56 ] . Strong-sector power counting, such as SILH [57 ] , rests on different dynamical assumptions. These classifications organize coefficient sizes according to the underlying dynamics; probabilities require additional assumptions on the model parameters.
Neutrino-mass naturalness constrains large Dirac magnetic moments in ref. [58 ] . Explicit constructions include the leptoquark model in ref. [3 ] , the model accompanying Higgs production in ref. [6 ] , and the one-loop matching example of ref. [23 ] . Symmetry-based separation of mass and magnetic moments is discussed in ref. [24 ] ; ref. [22 ] instead uses electroweak representations to select the weak dipole contribution while suppressing the associated mass diagram. These mechanisms motivate suppressed mixing or selected dipole directions. Establishing dipole dominance in the pole observables also requires the currents and bosonic operators generated by the same completion. The interplay with mixing at GeV masses and the wider UV phenomenology are discussed in refs. [25 , 59 ] .
Radiative stability of negligible mixing. The negligible-mixing benchmark requires a radiatively consistent boundary condition or a protecting symmetry. The operator C L N H ( L ˉ H ~ N ) ( H † H ) C_{LNH}(\bar{L}\tilde{H}N)(H^\dagger H) C L N H ( L ˉ H ~ N ) ( H † H ) is generated by dipole running [13 ] . In the convention
L ⊃ − Y N L ˉ H ~ N + C L N H ( L ˉ H ~ N ) ( H † H ) + h . c . , \mathcal{L} \supset-Y_N\bar{L}\tilde{H}N+C_{LNH}(\bar{L}\tilde{H}N)(H^\dagger H)+\mathrm{h.c.}, L ⊃ − Y N L ˉ H ~ N + C L N H ( L ˉ H ~ N ) ( H † H ) + h.c. ,
the Dirac mass is
m D = v 2 ( Y N − v 2 2 C L N H ) , ∣ θ ∣ ≃ ∣ m D ∣ M N . (183) m_D=\frac{v}{\sqrt{2}}\left(Y_N-\frac{v^2}{2}C_{LNH}\right), \qquad|\theta|\simeq\frac{|m_D|}{M_N}. \tag*{(183)} m D = 2 v ( Y N − 2 v 2 C L N H ) , ∣ θ ∣ ≃ M N ∣ m D ∣ . (183)
For a single Majorana state, or independent real aligned copies without cancellations, the seesaw contribution is ∣ m ν ∣ ≃ ∣ m D ∣ 2 / M N |m_\nu|\simeq|m_D|^2/M_N ∣ m ν ∣ ≃ ∣ m D ∣ 2 / M N . Requiring this contribution below an illustrative 0.1 e V 0.1\ \mathrm{eV} 0.1 eV gives
∣ m D ∣ ≲ 0.32 M e V M N ( 1 T e V ) , ∣ θ ∣ ≲ 3.2 × 10 − 7 ( 1 T e V ) M N . (184) |m_D|\lesssim0.32\ \mathrm{MeV}\sqrt{\frac{M_N}{(1\ \mathrm{TeV})}}, \qquad|\theta|\lesssim3.2\times10^{-7}\sqrt{\frac{(1\ \mathrm{TeV})}{M_N}}. \tag*{(184)} ∣ m D ∣ ≲ 0.32 MeV ( 1 TeV ) M N , ∣ θ ∣ ≲ 3.2 × 1 0 − 7 M N ( 1 TeV ) . (184)
Two transition dipoles can also induce a lepton-number-violating mass. The cutoff power counting discussed in ref. [3 ] gives
∣ δ m ν ∣ N D A ∼ ∣ μ ν ∣ 2 Λ U V 2 M N 16 π 2 = ∣ d γ ∣ 2 Λ U V 2 M N 4 π 2 , μ ν = 2 d γ . (185) |\delta m_\nu|_{\mathrm{NDA}}\sim\frac{|\mu_\nu|^2\Lambda_{\mathrm{UV}}^2M_N}{16\pi^2}=\frac{|d_\gamma|^2\Lambda_{\mathrm{UV}}^2M_N}{4\pi^2}, \qquad\mu_\nu=2d_\gamma. \tag*{(185)} ∣ δ m ν ∣ NDA ∼ 16 π 2 ∣ μ ν ∣ 2 Λ UV 2 M N = 4 π 2 ∣ d γ ∣ 2 Λ UV 2 M N , μ ν = 2 d γ . (185)
For ∣ d γ ∣ = 10 − 5 G e V − 1 |d_\gamma|=10^{-5}\ \mathrm{GeV}^{-1} ∣ d γ ∣ = 1 0 − 5 GeV − 1 , M N = 1 T e V M_N=1\ \mathrm{TeV} M N = 1 TeV and Λ U V = 2 T e V \Lambda_{\mathrm{UV}}=2\ \mathrm{TeV} Λ UV = 2 TeV , this estimate is of order 10 M e V 10\ \mathrm{MeV} 10 MeV , far above sub-eV masses. The estimate is sensitive to the cutoff, order-one factors and local mass counterterms. Its role is to quantify the mass-protection problem: a small renormalized mixing does not suppress this separate contribution. A viable realization near the conditional reach needs mass protection or cancellations as well as control of the EWPO operators. Approximate lepton number with pseudo-Dirac partners changes the Majorana/flavour benchmark and requires the coupling decomposition of § 2.1 and a separate treatment of lepton-number-violating observables; it cannot be assumed to protect the three independent real Majorana states used here.
Dipole dominance in ultraviolet completions
The relative size of dipole loops and independent local interactions can be expressed in terms of the physical mediator scale Λ U V \Lambda_{\mathrm{UV}} Λ UV , which need not equal the boundary scale Λ B \Lambda_B Λ B :
C X = c X Λ U V 2 , D 6 = u 6 Λ U V 2 , D 8 = u 8 Λ U V 4 . (186) C_X=\frac{c_X}{\Lambda_{\mathrm{UV}}^2}, \qquad D_6=\frac{u_6}{\Lambda_{\mathrm{UV}}^2}, \qquad D_8=\frac{u_8}{\Lambda_{\mathrm{UV}}^4}. \tag*{(186)} C X = Λ UV 2 c X , D 6 = Λ UV 2 u 6 , D 8 = Λ UV 4 u 8 . (186)
For a representative quadratic combination c 2 c^2 c 2 , the threshold-to-UV ratios scale as
∣ Δ D 6 ∣ ∣ D 6 ∣ ∼ ∣ κ 6 ∣ c 2 16 π 2 r U V 2 ∣ u 6 ∣ , ∣ Δ D 8 ∣ ∣ D 8 ∣ ∼ ∣ κ 8 ∣ c 2 16 π 2 ∣ u 8 ∣ , r U V = Λ U V M N . (187) \frac{|\Delta D_6|}{|D_6|}\sim\frac{|\kappa_6|c^2}{16\pi^2r_{\mathrm{UV}}^2|u_6|}, \qquad\frac{|\Delta D_8|}{|D_8|}\sim\frac{|\kappa_8|c^2}{16\pi^2|u_8|}, \qquad r_{\mathrm{UV}}=\frac{\Lambda_{\mathrm{UV}}}{M_N}. \tag*{(187)} ∣ D 6 ∣ ∣Δ D 6 ∣ ∼ 16 π 2 r UV 2 ∣ u 6 ∣ ∣ κ 6 ∣ c 2 , ∣ D 8 ∣ ∣Δ D 8 ∣ ∼ 16 π 2 ∣ u 8 ∣ ∣ κ 8 ∣ c 2 , r UV = M N Λ UV . (187)
The κ i \kappa_i κ i include gauge and flavour factors, finite terms and logarithms; mixed dipoles replace c 2 c^2 c 2 by the relevant product. These ratios compare coefficients in the same operator convention and do not replace observable-level cancellation tests.
For illustration, with ∣ κ 6 ∣ = ∣ c ∣ = 1 |\kappa_6|=|c|=1 ∣ κ 6 ∣ = ∣ c ∣ = 1 , a tree-sized ∣ u 6 ∣ = 1 |u_6|=1 ∣ u 6 ∣ = 1 gives ratios 1.58 × 10 − 3 1.58\times10^{-3} 1.58 × 1 0 − 3 and 2.53 × 10 − 4 2.53\times10^{-4} 2.53 × 1 0 − 4 at r U V = 2 , 5 r_{\mathrm{UV}}=2,5 r UV = 2 , 5 . For ∣ u 6 ∣ = 1 / ( 16 π 2 ) |u_6|=1/(16\pi^2) ∣ u 6 ∣ = 1/ ( 16 π 2 ) the ratios become 1 / 4 1/4 1/4 and 1 / 25 1/25 1/25 . Dominance therefore requires, parametrically, ∣ c ∣ ≳ 4 π r U V / ∣ κ 6 ∣ |c| \gtrsim4\pi r_{\mathrm{UV}}/\sqrt{|\kappa_6|} ∣ c ∣ ≳ 4 π r UV / ∣ κ 6 ∣ or ∣ c ∣ ≳ r U V / ∣ κ 6 ∣ |c| \gtrsim r_{\mathrm{UV}}/\sqrt{|\kappa_6|} ∣ c ∣ ≳ r UV / ∣ κ 6 ∣ , respectively. A loop-generated dipole can introduce an additional suppression. Weakly coupled power counting therefore generally favours independent UV terms over a double insertion of loop-generated dipoles. Dipole dominance would require a mechanism suppressing those terms, such as a selection rule or correlated matching. Assessing such a mechanism requires the full set of operators contributing to the pole response.
If one additionally identifies Λ U V = Λ B \Lambda_{\mathrm{UV}}=\Lambda_B Λ UV = Λ B , the aggressive sections at 1 TeV correspond to dimensionless dipoles ( ∣ c B ∣ , ∣ c W ∣ , ∣ c + ∣ , ∣ c − ) ≃ ( 1.74 , 0.79 , 0.57 , 0.73 ) (|c_B|,|c_W|,|c_+|,|c_-)\simeq(1.74,0.79,0.57,0.73) ( ∣ c B ∣ , ∣ c W ∣ , ∣ c + ∣ , ∣ c − ) ≃ ( 1.74 , 0.79 , 0.57 , 0.73 ) for r B = 2 r_B=2 r B = 2 and ( 6.85 , 3.11 , 2.67 , 3.17 ) (6.85,3.11,2.67,3.17) ( 6.85 , 3.11 , 2.67 , 3.17 ) for r B = 5 r_B=5 r B = 5 . The r B = 5 r_B=5 r B = 5 scenario thus probes coefficients larger than unity at this mass. Their viability depends on the matching couplings, multiplicities and loop expansion of a specified UV completion; the coefficient size alone supplies no universal perturbativity boundary.
An explicit scale estimate follows from the approximately degenerate mediators X E , X N , ϕ X_E,X_N,\phi X E , X N , ϕ in appendix A of ref. [6 ] . Its leading matching gives
α X N Λ 2 ≃ e g L g X g N 192 π 2 M m e d 2 . \frac{\alpha_{XN}}{\Lambda^2}\simeq\frac{e g_L g_X g_N}{192\pi^2 M_{\mathrm{med}}^2}. Λ 2 α X N ≃ 192 π 2 M med 2 e g L g X g N .
In our normalization this corresponds to
∣ g L g X g N ∣ ≃ 192 π 2 M m e d 2 ∣ C B + C W / 2 ∣ . (188) |g_Lg_Xg_N|\simeq192\pi^2M_{\mathrm{med}}^2|C_B+C_W/2|. \tag*{(188)} ∣ g L g X g N ∣ ≃ 192 π 2 M med 2 ∣ C B + C W /2∣. (188)
For the illustrative coefficient ∣ C B + C W / 2 ∣ = 0.1 T e V − 2 |C_B+C_W/2|=0.1\,\mathrm{TeV}^{-2} ∣ C B + C W /2∣ = 0.1 TeV − 2 and M m e d = 2 T e V M_{\mathrm{med}}=2\,\mathrm{TeV} M med = 2 TeV , the required product is about 7.6 × 10 2 7.6\times10^2 7.6 × 1 0 2 , or a geometric-mean coupling 9.1. This is already outside a comfortably weakly coupled regime; increasing M m e d M_{\mathrm{med}} M med strengthens that conclusion. The geometric mean summarizes the required coupling product; it does not impose equal couplings, since the model assumes a hierarchy among them. Its matching also produces O L N H O_{LNH} O L N H . The estimate exposes the difficulty of obtaining a large dipole with perturbative mediators, while an exclusion of the UV model would require its complete correlated EWPO response.
The apparent mass enhancement is also conditional. At fixed dimensionful C X C_X C X , the threshold grows as M N 2 C X C Y M_N^2C_XC_Y M N 2 C X C Y and the leading coefficient reach improves approximately as 1 / M N 1/M_N 1/ M N . In a family of UV theories with fixed dimensionless c X c_X c X and fixed r U V r_{\mathrm{UV}} r UV , instead, M N 2 C X C Y = c X c Y / ( r U V 4 M N 2 ) M_N^2C_XC_Y=c_Xc_Y/(r_{\mathrm{UV}}^4M_N^2) M N 2 C X C Y = c X c Y / ( r UV 4 M N 2 ) decreases. There is no contradiction with decoupling in that limit.
Blind directions in explicit multi-field UV completions, including finite matching and running, are exhibited in ref. [60 ] . Positivity of W , Y W,Y W , Y in universal theories follows under the assumptions of ref. [61 ] ; it cannot be applied as an unrestricted prior to our full response, which includes direct lepton currents. The four directions constrained in the universal analysis of ref. [34 ] concern its own enlarged bosonic parameter space, not a derivation of our four-combination map.
Geometry of compensating contributions
An additional interaction weakens the forecast only if it changes the measured quantities in a sufficiently opposing direction. The covariance provides the appropriate metric for comparing these changes, giving a simple description of the examples in section 8.5.
Magnitude and orientation of the additional response. Let d \mathbf{d} d denote the dipole-induced shifts of the twelve pseudo-observables and u \mathbf{u} u the sum of all additional contributions, in the same input scheme. For the SM-centred forecast define
∥ d ∥ V 2 = d T V − 1 d , r = ∥ u ∥ V ∥ d ∥ V , η = d T V − 1 u ∥ d ∥ V ∥ u ∥ V . (189) \|\mathbf{d}\|_V^2=\mathbf{d}^{T}V^{-1}\mathbf{d},\qquad r=\frac{\|\mathbf{u}\|_V}{\|\mathbf{d}\|_V},\qquad\eta=\frac{\mathbf{d}^{T}V^{-1}\mathbf{u}}{\|\mathbf{d}\|_V\|\mathbf{u}\|_V}. \tag*{(189)} ∥ d ∥ V 2 = d T V − 1 d , r = ∥ d ∥ V ∥ u ∥ V , η = ∥ d ∥ V ∥ u ∥ V d T V − 1 u . (189)
Here r r r compares the sizes of the two observable patterns, while η \eta η measures their alignment after accounting for measurement uncertainties and correlations. Parallel patterns have η = 1 \eta= 1 η = 1 and opposite patterns have η = − 1 \eta= -1 η = − 1 . For nonzero responses the change in the quadratic statistic is exactly
R ≡ χ d i p + e x t r a 2 χ d i p 2 = 1 + r 2 + 2 r η . (190) R \equiv\frac{\chi^{2}_{\mathrm{dip+extra}}}{\chi^{2}_{\mathrm{dip}}} = 1 + r^{2} + 2r\eta. \tag*{(190)} R ≡ χ dip 2 χ dip + extra 2 = 1 + r 2 + 2 r η . (190)
A negative overlap is necessary, but not sufficient, to weaken the forecast: weakening requires η < − r / 2 \eta< -r/2 η < − r /2 . In particular, η ≥ 0 \eta\ge0 η ≥ 0 or r ≥ 2 r \ge2 r ≥ 2 guarantees R ≥ 1 R \ge1 R ≥ 1 . The orientation is defined by the complete extra response in the SM-centred forecast. It incorporates the relative precision and correlations of all measured quantities.
Relative and fixed budgets. First consider a relative budget that scales with the dipole response at every coefficient value. If ∥ u ∥ V ≤ ϵ ∥ d ∥ V \lVert\mathbf{u}\rVert_{V} \le\epsilon\lVert\mathbf{d}\rVert_{V} ∥ u ∥ V ≤ ϵ ∥ d ∥ V is assumed, minimization over its direction gives
χ min 2 = max ( 1 − ϵ , 0 ) 2 χ d i p 2 . (191) \chi^{2}_{\min} = \max(1-\epsilon,0)^{2}\chi^{2}_{\mathrm{dip}}. \tag*{(191)} χ m i n 2 = max ( 1 − ϵ , 0 ) 2 χ dip 2 . (191)
Along a fixed dipole ray the response scales quadratically with the coefficient magnitude. At an unchanged reference threshold the maximum increase of the upper limit is therefore ( 1 − ϵ ) − 1 / 2 (1-\epsilon)^{-1/2} ( 1 − ϵ ) − 1/2 for ϵ < 1 \epsilon< 1 ϵ < 1 . For example, budgets of 20% and 50% allow increases of approximately 12% and 41%, respectively. These budgets are assumptions on observable shifts, not probabilities or estimates of natural UV Wilson coefficients. For ϵ ≥ 1 \epsilon\ge1 ϵ ≥ 1 no nonzero residual is guaranteed in the unrestricted response space.
A different assumption fixes the extra response independently of the dipole magnitude. Let C 0 C_{0} C 0 be the nominal limit on a chosen ray, k = ∣ C ∣ / C 0 k = \lvert C\rvert/C_{0} k = ∣ C ∣ / C 0 , and d 0 = d ( C 0 ) \mathbf{d}_{0} = \mathbf{d}(C_{0}) d 0 = d ( C 0 ) , so that ∥ d 0 ∥ V = χ c r i t 2 \lVert\mathbf{d}_{0}\rVert_{V} = \sqrt{\chi^{2}_{\mathrm{crit}}} ∥ d 0 ∥ V = χ crit 2 and d ( C ) = k 2 d 0 \mathbf{d}(C) = k^{2}\mathbf{d}_{0} d ( C ) = k 2 d 0 . For the fixed antiparallel contribution u = − ϵ 0 d 0 \mathbf{u} = -\epsilon_{0}\mathbf{d}_{0} u = − ϵ 0 d 0 , the acceptance condition is
∣ k 2 − ϵ 0 ∣ ≤ 1 , max ( 0 , ϵ 0 − 1 ) ≤ k ≤ 1 + ϵ 0 . (192) \lvert k^{2} - \epsilon_{0}\rvert\le1,\qquad\sqrt{\max(0,\epsilon_{0}-1)} \le k \le\sqrt{1+\epsilon_{0}}. \tag*{(192)} ∣ k 2 − ϵ 0 ∣ ≤ 1 , max ( 0 , ϵ 0 − 1 ) ≤ k ≤ 1 + ϵ 0 . (192)
Thus the upper endpoint increases by 1 + ϵ 0 \sqrt{1+\epsilon_{0}} 1 + ϵ 0 : 1.095 and 1.225 for ϵ 0 = 0.2 , 0.5 \epsilon_{0} = 0.2, 0.5 ϵ 0 = 0.2 , 0.5 . For a fixed contribution with ϵ 0 > 1 \epsilon_{0} > 1 ϵ 0 > 1 the origin is excluded and the allowed interval is detached from it. In contrast, profiling over a fixed ball ∥ u ∥ V ≤ ϵ 0 ∥ d 0 ∥ V \lVert\mathbf{u}\rVert_{V} \le\epsilon_{0}\lVert\mathbf{d}_{0}\rVert_{V} ∥ u ∥ V ≤ ϵ 0 ∥ d 0 ∥ V gives the same upper endpoint but retains the origin, since u = 0 \mathbf{u} = 0 u = 0 is allowed. These fixed budgets differ from the relative budget above; none of these geometric rescalings assigns a probability to a UV model.
To recover a point k > 1 k > 1 k > 1 times the nominal coefficient limit requires R ≤ k − 4 R \le k^{-4} R ≤ k − 4 . Necessary conditions are
∣ r − 1 ∣ ≤ k − 2 , η ≤ − 1 − k − 4 . (193) \lvert r - 1\rvert\le k^{-2},\qquad\eta\le-\sqrt{1-k^{-4}}. \tag*{(193)} ∣ r − 1 ∣ ≤ k − 2 , η ≤ − 1 − k − 4 . (193)
The second condition follows by optimizing r r r as well. A factor-two weakening therefore requires contributions of comparable size with a strongly opposing orientation. A symmetry or mediator representation can enforce that orientation. Only four local combinations enter the universal pole response, even though twelve pseudo-observables are measured. These diagnostic rescalings use the original threshold; recalibrated 95% confidence regions would require including the extra nuisance parameters in the statistical construction.
Finite terms, logarithms and UV boundary conditions. In the fixed prescription the local dipole response can be separated as d = F + K log ( Λ B 2 / M N 2 ) \mathbf{d} = \mathbf{F} + \mathbf{K}\log(\Lambda_{B}^{2}/M_{N}^{2}) d = F + K log ( Λ B 2 / M N 2 ) . The dipole contribution to K \mathbf{K} K is fixed by the EFT, whereas additional UV operators can also contribute to running. At a fixed pole momentum, the logarithm of scales contributes to the same local coupling combinations as the finite term. Free boundary coefficients can consequently compensate their sum. The split between finite terms and logarithms also depends on the scheme and reference scale. A prediction retaining only the logarithmic response therefore corresponds to a particular boundary assumption, rather than to a minimum signal shared by all completions.
Heavy vector-like leptons provide concrete examples in which gauge quantum numbers fix the relative changes of neutral-current couplings. Their tree-level currents can reinforce or compensate a dipole signal, depending on the representation. The matching of ref. [62 ] fixes their relative signs. Translating to our convention L Z = − ( g / c W ) Z μ f ˉ γ μ ( a f L P L + a f R P R ) f \mathcal{L}_Z = -(g/c_W) Z_\mu\bar{f}\gamma^\mu(a_{fL}P_L+a_{fR}P_R)f L Z = − ( g / c W ) Z μ f ˉ γ μ ( a f L P L + a f R P R ) f , write ( δ a ν L , δ a ℓ L , δ a ℓ R ) = x p (\delta a_{\nu L},\delta a_{\ell L},\delta a_{\ell R})=x\,\mathbf{p} ( δ a ν L , δ a ℓ L , δ a ℓ R ) = x p , with x = v 2 ∣ λ ∣ 2 / M m e d 2 ≥ 0 x=v^2|\lambda|^2/M_{\mathrm{med}}^2 \ge0 x = v 2 ∣ λ ∣ 2 / M med 2 ≥ 0 . The patterns are
Table F.12.
The E E E -singlet result also follows from tree-level matching with Matchete [32 ] . Integrating out E E E in L S M + E ˉ ( i D̸ − M E ) E − ( y E ℓ ˉ L H E R + h . c . ) \mathcal{L}_{\mathrm{SM}}+\bar{E}(i\not{D}-M_E)E-(y_E\bar{\ell}_L H E_R+\mathrm{h.c.}) L SM + E ˉ ( i D − M E ) E − ( y E ℓ ˉ L H E R + h.c. ) gives, for one lepton flavour, the current interaction
− ∣ y E ∣ 2 2 M E 2 i [ H i ∗ ( D μ H ) j − ( D μ H ) i ∗ H j ] ( ℓ ˉ L j γ μ ℓ L i ) . (194) -\frac{|y_E|^2}{2M_E^2}i\left[H_i^*(D_\mu H)^j-(D_\mu H)_i^*H^j\right](\bar{\ell}_{Lj}\gamma^\mu\ell_L^i). \tag*{(194)} − 2 M E 2 ∣ y E ∣ 2 i [ H i ∗ ( D μ H ) j − ( D μ H ) i ∗ H j ] ( ℓ ˉ L j γ μ ℓ L i ) . (194)
Using δ i l δ j k = ( δ i j δ k l + σ i j I σ k l I ) / 2 \delta_{il}\delta_{jk}=(\delta_{ij}\delta_{kl}+\sigma^I_{ij}\sigma^I_{kl})/2 δ i l δ j k = ( δ ij δ k l + σ ij I σ k l I ) /2 gives
C H l ( 1 ) = C H l ( 3 ) = − ∣ y E ∣ 2 4 M E 2 , ( δ a ν L , δ a ℓ L , δ a ℓ R ) = v 2 ∣ y E ∣ 2 M E 2 ( 0 , 1 / 4 , 0 ) . (195) C_{H l}^{(1)}=C_{H l}^{(3)}=-\frac{|y_E|^2}{4M_E^2},\qquad(\delta a_{\nu L},\delta a_{\ell L},\delta a_{\ell R})=\frac{v^2|y_E|^2}{M_E^2}(0,1/4,0). \tag*{(195)} C H l ( 1 ) = C H l ( 3 ) = − 4 M E 2 ∣ y E ∣ 2 , ( δ a ν L , δ a ℓ L , δ a ℓ R ) = M E 2 v 2 ∣ y E ∣ 2 ( 0 , 1/4 , 0 ) . (195)
Field redefinitions also generate a charged-lepton Yukawa operator. It belongs to the effective action but leaves the current projection unchanged in the light-lepton limit used here. This establishes the current normalization used in the response comparison.
Here N s N_s N s denotes an additional neutral mediator, not an identification with the dipolar HNL already included. Separate aligned copies for each lepton flavour realize the universal diagonal patterns. We minimize the quadratic statistic over the nonnegative strength of one pattern at a time, independently of the dipole coefficients. The resulting displacement isolates the effect of that correlated current sector on the dipole forecast.
At M N = 1 M_N=1 M N = 1 TeV with the aggressive covariance and both r B = 2 , 5 r_B=2,5 r B = 2 , 5 , the E E E and Σ 1 \Sigma_1 Σ 1 patterns do not reduce the statistic on either axis or either diagonal. A fixed nonzero contribution therefore tightens the expected constraint on these sections; if its magnitude is profiled freely down to zero, the minimum remains at zero additional contribution. The Σ 0 \Sigma_0 Σ 0 pattern instead allows a same-sign diagonal weakening of approximately a factor 1.5 1.5 1.5 –2.2 2.2 2.2 at the unchanged threshold.6 The N s N_s N s pattern gives a more moderate effect on the opposite-sign diagonal. More importantly, N s N_s N s on the C B C_B C B axis and Σ 0 \Sigma_0 Σ 0 on the C W C_W C W axis are nearly degenerate with the dipole response in this restricted test. The resulting small residual is particularly sensitive to additional contributions. Gauge-related charged-current measurements, mediator loops and other constraints would therefore be essential for assigning an
exclusion to these nearly degenerate cases. In particular, the optimal strengths illustrate the geometry of compensation in pole observables. Their realization in a viable mediator model would also require consistency with mixing and direct-production constraints. The pole comparison already establishes that the representation can select either a reinforcing or a compensating direction.
In the dipole-generating E + ϕ E+\phi E + ϕ model of ref. [11 ] , the charged-lepton current is correlated with the same Yukawa vector that enters one contribution to the dipole; the latter also involves an antisymmetric flavour coupling and an HNL coupling. For the dipole source sharing this Yukawa coupling, the E E E tree current is much larger in the heavy-mediator regime considered. The unequal magnitudes prevent compensation between these two contributions. The full response additionally includes scalar contacts, other thresholds and mixing effects. A loop involving a different, larger coupling can compete with a small tree contribution without invalidating perturbation theory.
The examples identify reinforcing and compensating responses fixed by mediator representations. Assigning probabilities to these possibilities would require a specified UV model and a measure on its parameters.
Direct-search conventions and extensions beyond the pole
This appendix follows the discussion in § 8.6 and § 8.7 . We first give the coupling and flavour conversion used for the HL-LHC comparison, then distinguish related searches and summarize what off-pole and charged-current measurements could add. No additional production likelihood is included in the present forecast.
Coupling and flavour conversion
The conversion below fixes the normalization and flavour convention of Figure 6 .
The direct curves use the same-sign dilepton projections of ref. [1 ] . The observable comparison requires the same operator normalization and diagonal flavour response on both sides. The coordinate dictionary follows directly from the gauge-invariant operators. For C W = k C B C_W=kC_B C W = k C B ,
C B = 2 d γ e v ( 1 + k / 2 ) , C W = k C B . (196) C_B=\frac{\sqrt{2d_\gamma}}{ev(1+k/2)}, \qquad C_W=kC_B. \tag*{(196)} C B = e v ( 1 + k /2 ) 2 d γ , C W = k C B . (196)
The benchmark directions of ref. [1 ] are
A C B = 0 B k = 2 s W 2 c W ( c W − 2 ) ( d W = d Z ) C k = 2 s W 2 c W 2 ( d Z = 0 ) D C W = 0 (197) \begin{aligned}
A \quad& C_B=0 \\
B \quad& k=\frac{2s_W^2}{c_W(c_W-\sqrt{2})}\quad(d_W=d_Z) \\
C \quad& k=\frac{2s_W^2}{c_W^2}\quad(d_Z=0) \\
D \quad& C_W=0 \tag*{(197)}
\end{aligned} A B C D C B = 0 k = c W ( c W − 2 ) 2 s W 2 ( d W = d Z ) k = c W 2 2 s W 2 ( d Z = 0 ) C W = 0 (197)
The photon-blind direction k = − 2 k=-2 k = − 2 cannot be parameterized by d γ d_\gamma d γ . A change from σ I / 2 \sigma^{I/2} σ I /2 to σ I \sigma^I σ I in the portal operator also requires a factor-of-two coefficient conversion.
For the comparison in Figure 6 , both curves can be interpreted in the single-HNL, flavour-universal benchmark of ref. [1 ] . At quadratic dipole order the diagonal pole response depends on ( C X C Y † ) i i (C_X C_Y^\dagger)_{ii} ( C X C Y † ) ii and its flavour trace. For real universal coefficients, C X = C X ( 0 ) ( 1 , 1 , 1 ) T C_X=C_X^{(0)}(1,1,1)^T C X = C X ( 0 ) ( 1 , 1 , 1 ) T and the three-HNL aligned matrix C X = C X ( 0 ) I 3 C_X=C_X^{(0)}I_3 C X = C X ( 0 ) I 3 give the same diagonal entries and trace. Thus the bosonic factor remains three in both cases, and each diagonal current is unchanged. The relevant multiplicity is therefore the flavour trace, which includes both heavy-state and active-flavour couplings. Off-diagonal flavour amplitudes differ and their constraints are not included in this comparison.
The HL-LHC sensitivities are taken from Figure 6 of ref. [1 ] , over the mass interval shown there. The HL-LHC criterion S / S + B = 2 S/\sqrt{S+B}=2 S / S + B = 2 and our joint 95% C.L. region are different statistical constructions. The separate UV-scale guide assumes unit dimensionless coefficients: Λ e f f = [ max ( ∣ C B ∣ , ∣ C W ∣ ) ] − 1 / 2 \Lambda_{\rm eff}=[\max(|C_B|,|C_W|)]^{-1/2} Λ eff = [ max ( ∣ C B ∣ , ∣ C W ∣ ) ] − 1/2 . Above the guide, keeping the mediator scale above M N M_N M N requires larger dimensionless coefficients; this is not a model-independent exclusion of EFT validity. For direct production the partonic energies must also remain within the EFT domain. The low-mass line-style change has a different purpose, described in § 6.5 .
The comparison is restricted to the active–sterile dipole and the stated flavour benchmark. The relation to other direct searches is summarized next.
Related operators and matching tools
Several useful comparisons involve different operators or production regimes. The dimension-five sterile–sterile moments of refs. [63 , 64 ] and the see-saw portal studied at Higgs factories [65 ] cannot be mapped directly onto our dimension-six active–sterile coefficients. In particular, their quoted large cutoff reach is not a bound on C B , C W C_B,C_W C B , C W . The active–sterile searches through meson decays, including NA62, probe the low-mass regime [4 , 5 ] . The gauge-invariant dimension-six portal at 3- and 10-TeV muon colliders is studied in ref. [9 ] , including single-flavour benchmarks. Comparison with our curves requires matching the flavour assumptions and collider setting. Other active–sterile searches are presented in refs. [2 , 8 , 10 , 11 ] . The Majorana transition moments between active neutrino flavours studied in ref. [66 ] probe lepton-number and flavour violation and are a distinct operator sector from the active–sterile portal here. Broader sterile-neutrino EFT phenomenology is covered by refs. [67 , 68 , 69 , 70 ] . Mixing-induced EW fits and boson-decay corrections in seesaw scenarios, refs. [71 , 72 ] , are distinct from our negligible-mixing benchmark.
Earlier one-loop electroweak analyses of top and bosonic effective operators [73 , 74 , 75 ] provide context for renormalization and indirect sensitivity, rather than the portal-specific matching coefficients. The parameter-sensitivity measure of ref. [76 ] addresses a different question from our covariance-weighted response comparison.
Automated off-shell matching with matchmakereft [77 ] and the SOLD dictionary [56 ] can assist UV-model matching. Applying them to the present nonrenormalizable portal would require its explicit implementation, the retained operator order and a compatible finite prescription; these tools have not been used as an independent validation of our two-insertion result.
Energy dependence and ultraviolet degeneracies
Away from the Z Z Z pole, energy dependence and new scattering structures can distinguish interactions that have the same pole response. Their usefulness depends on the local UV coefficients allowed to vary.
Measurements at distinct energies can separate the constant and slope of a local amplitude. If both polynomial coefficients also have free UV contributions, those terms can compensate the dipole polynomial at every energy. The additional leverage therefore comes from restrictions on the UV coefficients, new structures in the scattering amplitude, or mass dependence that is not captured by that local polynomial.
At M N = 1 M_N = 1 M N = 1 TeV the expansion parameters s / M N 2 s/M_N^2 s / M N 2 are 0.0576 and 0.133 at 240 and 365 GeV; ( s − M N 2 ) / M N 2 (s - M_N^2)/M_N^2 ( s − M N 2 ) / M N 2 is about 0.0493 and 0.1249. These ratios characterize the hierarchy underlying the local expansion. At M N = 200 M_N = 200 M N = 200 GeV, s / M N 2 = 1.44 , 3.33 s/M_N^2 = 1.44, 3.33 s / M N 2 = 1.44 , 3.33 , so this local argument cannot be used to forecast those runs. A numerical off-pole analysis requires the complete e + e − → f f ˉ e^+e^- \to f\bar{f} e + e − → f f ˉ amplitudes, including boxes, photon interference and contact terms, together with the enlarged coefficient space and an experimental covariance.
Prospects for a charged-current test with W W W pairs
The process e + e − → W + W − e^+e^- \to W^+W^- e + e − → W + W − probes charged currents and triple-gauge interactions in addition to the neutral pole couplings. Doubly resonant W W W data can lift flat directions in restricted SMEFT fits [78 ] ; this does not by itself establish that the four-direction degeneracy found here is removed.
The charged dipole also permits tree-level t t t -channel HNL exchange with two insertions and an exact propagator proportional to ( t − M N 2 ) − 1 (t-M_N^2)^{-1} ( t − M N 2 ) − 1 . With negligible active–sterile mixing it probes C W C_W C W , with no analogous tree contribution on the pure C B C_B C B axis. At quadratic dipole order the rate depends on its interference with the SM amplitude, whereas the squared HNL amplitude is quartic. The electron-flavour coupling enters without a closed-loop flavour multiplier.
Angular distributions at 240 and 365 GeV could separate this exchange from shifts of the W e ν W e\nu W e ν , Z e e Zee Z ee , W W Z WWZ W W Z and W W γ WW\gamma W W γ interactions. The relevant information is the part of that distribution which cannot be reproduced by the allowed UV coefficients. Establishing its reach requires the full gauge-related interference, W W W decays and widths, and correlated experimental uncertainties. This explains the potential complementarity of W W W pairs while delimiting the pole and decay forecast presented here.