In this paper
AbstractRare decays with missing energySM predictions and heavy new physicsLight new particles in Complementarity of invisible decaysAn aligned modelResonant dark matterConclusionsAcknowledgementsReferencesWorksPhysics and Astronomy
invisible decays as probes of light new physics
Abstract
Rare decays with missing energy probe heavy particles contributing to and new light states produced on shell. We discuss both possibilities in light of the Belle II excess in B^+\to K^+\Emiss. Fits to the missing-mass distribution favour a two-body contribution near 2.1\,\GeV, while scalar and fermion pairs of masses around -0.6\,\GeV remain viable alternatives. Extracting their couplings would profit from additional measurements, including the B\to K^*\Emiss rate and longitudinal polarization, together with invisible decays which constrain scalar-pair interactions. We discuss how an aligned Higgsed model can realize the single-vector explanation through a vector-like top partner. Loop-induced electroweak mixing makes visible decays unavoidable, requiring an additional invisible channel to evade dimuon searches. We show that a light dark fermion can provide this channel and also account for the observed dark matter abundance through resonant thermal annihilation.
Rare decays with missing energy
Flavour-changing neutral current (FCNC) decays provide stringent tests of the Standard Model (SM) and its extensions. Their loop and CKM suppression allows small new-physics (NP) contributions to compete with the SM amplitude. The transitions are particularly interesting in this respect: they are theoretically cleaner than and experimentally sensitive to any sufficiently light particles that escape detection. Thus, the same measurement can probe heavy virtual particles or the production of new light degrees of freedom.
These decays are also well suited to NP coupled predominantly to the third generation. Existing flavour bounds strongly constrain generic interactions with the light quarks, motivating flavour structures such as minimal flavour violation (MFV) or approximate symmetries [1, 2]. In these scenarios, sizeable third-generation couplings can still induce observable effects in transitions through CKM mixing.
The Belle II collaboration has reported evidence for using 362 fb of data. Combining inclusive- and hadronic-tag analyses, and assuming an SM template for the distribution, they find [3]
This result has a significance of relative to the background-only hypothesis and lies above the SM expectation. Interestingly, the excess is concentrated at low reconstructed missing mass, -5 GeV. Its distribution therefore contains important information beyond the integrated rate. Due to efficiency and missing-mass smearing effects, which change the populations of reconstructed bins, the extracted branching fraction can be influenced by the assumed shape of the distribution. The branching fraction in Eq. (1) is extracted by fitting an assumed signal spectrum, the SM spectrum, rather than by summing independently measured partial rates. Scalar or tensor interactions, and especially light-particle production, can therefore substantially change the extracted branching fraction through their different distributions. While the localized excess is suggestive of light NP resonance production, it could also reflect an unaccounted-for background contribution, a possibility requiring further experimental investigation.
Complementary to , the channel probes additional quark-current structures, while the inclusive rate and fully invisible decays provide complementary, but less stringent, constraints. Potentially, searches with , and final states could test the flavour structure of the interaction [4, 5]. Interestingly, Belle has recently performed a dedicated search for two-body decays [6]. This search sets stringent mass-dependent limits which exclude a two-body invisible-particle interpretation of the Belle II excess in several mass ranges. However, it vetoes the region near favoured by the Belle II spectrum, rather than fitting the narrow backgrounds from decays whose charm decay products are not fully reconstructed.
SM predictions and heavy new physics
At energies below the electroweak scale, the interaction is described by the effective Hamiltonian
where , , and the sum over neutrino flavours is implicit. Only the left-handed operator contributes in the SM, with and . The resulting branching fractions are [7, 8, 9]
Here SD denotes the short-distance contribution. The charged mode also receives a contribution from followed by , included in the second line. Unlike , these decays are free from the nonlocal electromagnetic charm contribution. The main uncertainties in the exclusive predictions arise from the form factors and CKM inputs. Fully invisible decays are instead extremely suppressed in the SM: the two-neutrino rate vanishes in the massless-neutrino limit, and multi-neutrino final states also have negligible rates at present experimental sensitivities.
Heavy NP modifies the Wilson coefficients in Eq. (2). For lepton-flavour-universal vector interactions, it is convenient to define [8]
The and rates and the longitudinal polarization then satisfy
where is the short-distance rate normalized to its SM value and , depending mildly on form factors and the integration range. The opposite interference signs allow right-handed quark currents to enhance while respecting the bound. SM gauge invariance relates these interactions to . Charged-lepton bounds restrict heavy-NP explanations, but tau-dominated couplings can still give sizeable effects [10, 11].
For , vector interactions preserve the SM dependence. Scalar and tensor interactions, however, involve different form factors and can modify the spectrum. A recent Belle II weak effective theory (WET) analysis [12] includes scalar and tensor operators. In the normalization of Eq. (2), these read
Neutrino flavour indices are suppressed and the sum over flavours is implicit. Notice that the tensor bilinear selects an antisymmetric combination of different neutrino flavours [13]. These operators enter Eq. (2) with , and their coefficients vanish in the SM. The normalization absorbs Belle II’s common prefactor into the operators. The Belle II Lagrangian convention differs in its sign convention from the standard one, corresponding to for the same flavour sums. Notice that, the scalar and tensor operators are on a different footing from their vector counterparts, as they violate lepton number and arise from SMEFT operators of dimension seven or higher when only SM fields are retained. Scalar and tensor interactions can also arise from dimension-six operators in the , with a light right-handed neutrino . These operators conserve lepton number for , but require additional light fields, unlike the vector interactions of the SM-field-only dimension-six SMEFT [14]. For active neutrinos, the distorted spectra remain broad across the physical range. A localized contribution such as that from a light on-shell particle cannot be generated by the local heavy-NP operators alone.
Light new particles in
Light NP offers a different explanation of the Belle II excess. New weakly interacting particles can be produced on shell and carry away the missing energy, replacing the neutrino pair of the SM decay. At the -meson scale, their interactions are described by an EFT containing the new fields explicitly. For scalar, fermion and vector fields, denoted by , and , representative interactions are , , and , with coefficients of the appropriate mass dimension [9, 15]. The missing-mass distribution depends strongly on the number of emitted particles. A two-body decay gives . For a narrow on-shell state, this contribution is monochromatic in the true missing mass, while its reconstructed shape and width are depend on the experimental missing-mass resolution. Production of an pair instead yields a continuum above , whose shape also depends on the spin and Lorentz structure of the interaction.
Missing energy can arise from stable particles, long-lived states escaping the detector, or mediators decaying to invisible daughters. For with lifetime , boost and invisible branching fraction , the invisible probability is approximately
where is the detector path length and visible decays within this distance are assumed to be detected. The observed missing-energy rate is therefore the production rate weighted by .
To compare the predicted spectrum with the measured distribution, we need to include the selection efficiency and missing-mass smearing. The reconstructed signal spectrum and binned likelihood take the form
where is the number of produced pairs, accounts for the smearing of reconstructed relative to true , and is the selection efficiency. Integrating over bin gives . The and denote observed and background yields, and constrains the nuisance parameters. In Refs. [9, 15], we reconstructed binned likelihoods using Belle II, BaBar and ALEPH results. While several limits exist from Belle, they only provide integrated limits without a published binned spectra for such a fit. We instead use BaBar’s binned result [16]. Its result is not included, as its expected impact is smaller, and its efficiency information is less detailed than that available for Belle II’s charged- mode. The full Belle II likelihood and simulated joint number densities are now public [12], allowing direct tests of new signal distributions.
These fits favour a single emitted boson near , while fermion pairs near and scalar pairs near - give viable alternative shapes (Figure 1). Vector and spin-3/2 pair production gives a poorer fit for the interactions considered in Ref. [15]. A subsequent analysis using the public Belle II likelihood finds and for its narrow-width benchmark, with a reported preference over the SM in its frequentist test [17].
Figure 1. Reconstructed missing-mass distribution (left) and profiled mass-dependent likelihood (right) for light-particle contributions, from Refs. [9, 15]. Single-boson emission favours GeV; pair production favours lighter invisible particles.
Complementarity of invisible decays
Using the SM-based branching fraction to constrain light NP can give an incorrect mass or coupling, even after correcting the total efficiency. For a normalized spectrum , one can define an effective signal efficiency , where the outer integral covers the fitted region in reconstructed . Equating the total expected signal yield under hypothesis to that of an SM-shaped template, without using the distribution among bins, gives
The triangles in Figure 2 include this rescaling, with efficiency and smearing, but remain displaced from the rescaled-SM band. The fit also depends on the bin populations and their backgrounds: correcting the effective signal efficiency does not recover the mass and coupling inferred from the full distribution.
Figure 2. Best-fit branching fractions for and longitudinal polarization in the light-particle scenarios of [9], compared with a rescaled SM contribution. Triangles in the panel include efficiency and smearing and are rescaled according to Eq. (9). Their difference from the rescaled-SM band illustrates the dependence of the extracted rate on the shape.
The pseudoscalar and axial-vector quark currents have vanishing matrix elements, but contribute to . Their coefficients are independent in a parity basis, whereas a single chiral operator fixes the relative parity-even and parity-odd couplings. The bound can therefore exclude a chiral interaction allowed when both parity components vary independently.
Fully invisible decays provide particularly useful constraints on scalar-pair interactions [15]. For the vector/axial fermion interactions considered here, the rate is helicity suppressed [9].
The decay also measures the longitudinal fraction through
Here is the angle of the kaon in the rest frame relative to the flight direction in the rest frame. Scenarios giving similar spectra can predict very different rates and polarization [9]. For a bin , the polarization is the ratio of integrated rates, . To compare with the reconstructed angular distribution, the longitudinal and transverse rates must be folded separately with the angular acceptance and missing-mass resolution. This is particularly important for a two-body decay, whose contribution can be diluted over a wide reconstructed bin.
An aligned model
Several UV completions of the interactions discussed above have been considered in the literature, including axion-like particles, scalar portals, non-universal models and dark photons [18, 19, 20]. We discuss one of these, an aligned Higgsed model proposed in [21], which realizes the single-vector explanation of the Belle II excess. The model introduces a vector-like quark and a singlet scalar , with
where breaks the new gauge symmetry and gives . We take and , so that mixes only with the top quark. The resulting -t/ penguins generate with the CKM factor . This top alignment naturally gives an MFV-like flavour structure, with the largest FCNC effects in transitions.
The top partner also induces kinetic mixing between the new vector and hypercharge. Taking this mixing to vanish at a high scale , the leading logarithmic contribution to photon- mixing is
For and , . Mixing in the top sector further induces - mass mixing, which cannot be removed by tuning the hypercharge kinetic-mixing counterterm. Consequently, the couples to charged leptons and decays visibly. Although the minimal model can reproduce the required production rate, it is excluded by LHCb searches for prompt and displaced dimuon resonances [22, 23].
A minimal extension introduces a light vector-like SM singlet with charge , opening the decay
For , this channel can dominate the total width and suppress the dimuon branching fraction sufficiently to evade LHCb. For and , this is achieved with the representative charge ratio . Several light dark fermions with order-one charge ratios could replace this single large charge.
The surviving parameter space in Figure 3 is further constrained by BaBar monophoton searches, direct top-partner production, the modified top Yukawa coupling and electroweak precision observables [25, 26]. The scalar-sector perturbative-unitarity bound assumes a vanishing Higgs-portal quartic at , with its low-energy value generated radiatively, and a scalar mass .
Figure 3. Parameter space of the aligned U(1) model with a light dark fermion, from [21]. Left: the region and constraints from flavour, collider and electroweak observables, and perturbative unitarity, for and . Right: dark-matter constraints for , and . The gray curve gives the observed relic abundance calculated with micrOMEGAs [24]; the cyan region is excluded by energy injection into the CMB from late-time annihilation.
Resonant dark matter
The dark fermion can also account for the observed DM abundance if it is stable on cosmological scales. Assuming an initial thermal population, its abundance is depleted by . For , a sufficient annihilation rate can be obtained near . The -wave-dominated thermal average near the narrow resonance is, at leading order in ,
where , and . Here and includes all SM final states. When and kinetic mixing dominates the coupling to SM states, the approximate relic abundance is independent of . The numerical calculation finds a viable region at . Closer to the pole, annihilation can continue after kinetic decoupling; in this regime a calculation assuming a common dark-sector and SM temperature is no longer reliable.
The Planck determination of the DM density requires , with equality if constitutes all the DM. The late-time-annihilation constraint from Planck is parametrized by
where is the energy-deposition efficiency for channel [27]. The direct-detection limit is instead rescaled by for a subdominant relic. The DarkSide-50 constraint in Figure 3 arises from scattering on argon through the predominantly electromagnetic coupling induced by mixing [28]. Electron-scattering bounds also leave the resonant region allowed.
Extensions with additional scalar fields and different charges for the left- and right-handed dark fermions can yield Majorana states with axial-vector couplings and -wave-suppressed annihilation into SM states, potentially relaxing the CMB and direct-detection bounds. Heavier DM can instead annihilate to . In that case, lighter dark states are needed to provide the invisible decays responsible for Belle II.
Conclusions
The Belle II excess can arise from heavy NP in or from new light particles. The missing-mass distribution favours a two-body decay with a boson near , while scalar and fermion pairs remain viable. Their masses and couplings require a differential analysis as different shapes affect both the efficiency and the inferred rate. The public Belle II likelihood now allows these possibilities to be tested directly. The rate and longitudinal polarization distinguish interactions that are degenerate in , while inclusive and fully invisible decays provide complementary constraints. In the aligned model, a vector-like top partner generates the required FCNC coupling, but also induces visible decays excluded by LHCb. Adding a light dark fermion allows a dominant invisible width and the observed DM abundance through resonant freeze-out. Improved missing-mass spectra and polarization will be particularly important in testing these explanations.
Acknowledgements
I thank Patrick D. Bolton, Svjetlana Fajfer and Jernej F. Kamenik for the collaborations on which this contribution is based, and the organizers of Beauty 2026 for the invitation. I also acknowledge financial support from the Spanish Ministry of Science, Innovation and Universities (Agencia Estatal de Investigación MICIU/AEI/10.13039/501100011033) through the Severo Ochoa Centers of Excellence Programme under Grants No. CEX2024-001441-S and No. CEX2023-001292-S, and through Grants No. PID2023-146142NB-I00 and No. PID2023-146220NB-I00.
References
- [1]G. D’Ambrosio et al. “Minimal flavor violation: An effective field theory approach”. In: Nucl. Phys. B 645 (2002), pp. 155–187. DOI: 10.1016/S0550-3213(02)00836-2. arXiv: hep-ph/0207036.DOI
- [2]Riccardo Barbieri et al. “Flavour physics from an approximate U(2)³ symmetry”. In: JHEP 07 (2012), p. 181. DOI: 10.1007/JHEP07(2012)181. arXiv: 1203.4218 [hep-ph].DOI
- [3]I. Adachi et al. “Evidence for B⁺ → K⁺νν̄ decays”. In: Phys. Rev. D 109.11 (2024), p. 112006. DOI: 10.1103/PhysRevD.109.112006. arXiv: 2311.14647 [hep-ex].DOI
- [4]O. Lutz et al. “Search for B → h(*)νν̄ with the full Belle Υ(4S) data sample”. In: Phys. Rev. D 87.11 (2013), p. 111103. DOI: 10.1103/PhysRevD.87.111103. arXiv: 1303.3719 [hep-ex].DOI
- [5]J. Grygier et al. “Search for B → hνν̄ decays with semileptonic tagging at Belle”. In: Phys. Rev. D 96.9 (2017), p. 091101. DOI: 10.1103/PhysRevD.96.091101. arXiv: 1702.03224 [hep-ex].DOI
- [6]M. Abumusa et al. “Search for Feebly Interacting Particles in B Decays with Missing Energy at Belle”. In: Phys. Rev. Lett. 136.23 (2026), p. 231802. DOI: 10.1103/1rl4-j3np. arXiv: 2601.07104 [hep-ex].
- [7]W. G. Parrott, C. Bouchard, and C. T. H. Davies. “Standard Model predictions for B → Kℓ⁺ℓ⁻, B → Kℓ₁⁺ℓ₂⁻ and B → Kνν̄ using form factors from N_f = 2 + 1 + 1 lattice QCD”. In: Phys. Rev. D 107.1 (2023). [Erratum: Phys.Rev.D 107, 119903 (2023), p. 014511. DOI: 10.1103/PhysRevD.107.014511. arXiv: 2207.13371 [hep-ph].DOI
- [8]Andrzej J. Buras et al. “B → K⁽*⁾νν̄ decays in the Standard Model and beyond”. In: JHEP 02 (2015), p. 184. DOI: 10.1007/JHEP02(2015)184. arXiv: 1409.4557 [hep-ph].DOI
- [9]Patrick D. Bolton et al. “Impact of new invisible particles on B → K⁽*⁾Eₘᵢₛₛ observables”. In: Phys. Rev. D 112.3 (2025), p. 035010. DOI: 10.1103/9rrv-ft75. arXiv: 2503.19025 [hep-ph].DOI
- [10]Lukas Allwicher et al. “Understanding the first measurement of ℬ(B → Kνν̄)”. In: Phys. Lett. B 848 (2024), p. 138411. DOI: 10.1016/j.physletb.2023.138411. arXiv: 2309.02246 [hep-ph].DOI
- [11]Rigo Bause, Hector Gisbert, and Gudrun Hiller. “Implications of an enhanced B → Kνν̄ branching ratio”. In: Phys. Rev. D 109.1 (2024), p. 015006. DOI: 10.1103/PhysRevD.109.015006. arXiv: 2309.00075 [hep-ph].DOI
- [12]Merna Abum usab et al. “Model-agnostic likelihood for the reinterpretation of the B⁺ → K⁺νν̄ measurement at Belle II”. In: Phys. Rev. D 112.9 (2025), p. 092016. DOI: 10.1103/pr66-sd36. arXiv: 2507.12393 [hep-ex].DOI
- [13]Tobias Felkl, Sze Lok Li, and Michael A. Schmidt. “A tale of invisibility: constraints on new physics in b → sνν̄”. In: JHEP 12 (2021), p. 118. DOI: 10.1007/JHEP12(2021)118. arXiv: 2111.04327 [hep-ph].DOI
- [14]S. Rosauro-Alcaraz and L. P. Santos Leal. “Disentangling left and right-handed neutrino effects in B → K(*)νν̄”. In: Eur. Phys. J. C 84 (Apr. 2024), p. 795. DOI: 10.1140/epjc/s10052-024-13104-x. arXiv: 2404.17440 [hep-ph].DOI
- [15]Patrick D. Bolton et al. “Signatures of light new particles in B → K(*)Eₘᵢₛₛ”. In: Phys. Rev. D 110.5 (2024), p. 055001. DOI: 10.1103/PhysRevD.110.055001. arXiv: 2403.13887 [hep-ph].DOI
- [16]J. P. Lees et al. “Search for B → K(*)νν̄ and invisible quarkonium decays”. In: Phys. Rev. D 87.11 (2013), p. 112005. DOI: 10.1103/PhysRevD.87.112005. arXiv: 1303.7465 [hep-ex].DOI
- [17]Lorenz Gärtner et al. “Constraints on invisible B⁺ → K⁺X decays from the Belle II B⁺ → K⁺νν̄ measurement”. In: Phys. Rev. D 114 (2026), p. 032003. DOI: 10.1103/ggtk-gkx4. arXiv: 2602.09666 [hep-ph].DOI
- [18]Wolfgang Altmannshofer et al. “Light New Physics in B → K(*)νν̄”. In: (Nov. 2023). arXiv: 2311.14629 [hep-ph].arxiv.org/abs/2311.14629
- [19]Chandan Hati et al. “QCD axion, color-mediated neutrino masses, and B⁺ → K⁺ + Eₘᵢₛₛ anomaly”. In: Phys. Rev. D 111.1 (2025), p. 015038. DOI: 10.1103/PhysRevD.111.015038. arXiv: 2408.00060 [hep-ph].DOI
- [20]E. Gabrielli et al. “Explaining the B⁺ → K⁺νν̄ excess via a massless dark photon”. In: (Feb. 2024). arXiv: 2402.05901 [hep-ph].arxiv.org/abs/2402.05901
- [21]Patrick D. Bolton, Jernej F. Kamenik, and Martín Novoa-Brunet. “Dark light shining on B → K(*)Eₘᵢₛₛ”. In: (Dec. 2025). arXiv: 2512.16999 [hep-ph].
- [22]Roel Aaij et al. “Search for hidden-sector bosons in B⁰ → K*⁰μ⁺μ⁻ decays”. In: Phys. Rev. Lett. 115.16 (2015), p. 161802. DOI: 10.1103/PhysRevLett.115.161802. arXiv: 1508.04094 [hep-ex].DOI
- [23]R. Aaij et al. “Search for long-lived scalar particles in B⁺ → K⁺χ(μ⁺μ⁻) decays”. In: Phys. Rev. D 95.7 (2017), p. 071101. DOI: 10.1103/PhysRevD.95.071101. arXiv: 1612.07818 [hep-ex].DOI
- [24]G. Alguero et al. “micrOMEGAs 6.0: N-component dark matter”. In: Comput. Phys. Commun. 299 (2024), p. 109133. DOI: 10.1016/j.cpc.2024.109133. arXiv: 2312.14894 [hep-ph].DOI
- [25]J. P. Lees et al. “Search for Invisible Decays of a Dark Photon Produced in e⁺e⁻ Collisions at BaBar”. In: Phys. Rev. Lett. 119.13 (2017), p. 131804. DOI: 10.1103/PhysRevLett.119.131804. arXiv: 1702.03327 [hep-ex].DOI
- [26]Albert M Sirunyan et al. “Measurement of the Higgs boson production rate in association with top quarks in final states with electrons, muons, and hadronically decaying tau leptons at √s = 13 TeV”. In: Eur. Phys. J. C 81.4 (2021), p. 378. DOI: 10.1140/epjc/s10052-021-09014-x. arXiv: 2011.03652 [hep-ex].DOI
- [27]N. Aghanim et al. “Planck 2018 results. VI. Cosmological parameters”. In: Astron. Astrophys. 641 (2020). [Erratum: Astron.Astrophys. 652, C4 (2021)], A6. DOI: 10.1051/0004-6361/201833910. arXiv: 1807.06209 [astro-ph.CO].DOI
- [28]P. Agnes et al. “Low-Mass Dark Matter Search with the DarkSide-50 Experiment”. In: Phys. Rev. Lett. 121.8 (2018), p. 081307. DOI: 10.1103/PhysRevLett.121.081307. arXiv: 1802.06994 [astro-ph.HE].DOI


