Arxiv:1911.01427V2 [Hep-Ph] 15 Jun 2020

FERMILAB-PUB-19-538-T

Decaying Sterile Neutrinos and the Short Baseline Oscillation Anomalies

Mona Dentler,1,a Ivan Esteban,2,b Joachim Kopp,3, 4,c and Pedro Machado5,d 1Institut fürAstrophysik, Georg August-UniversitätGöttingen,Germany 2Departament de Fis´ıca Quàntica i Astrof´ısica and Institut de Ciencies del Cosmos, Universitat de Barcelona, Spain 3Theoretical Physics Department, CERN, Geneva, Switzerland 4PRISMA Cluster of Excellence and Mainz Institute for Theoretical Physics, Johannes Gutenberg-UniversitätMainz, Germany 5Fermi National Accelerator Laboratory, Batavia, Illinois, USA (Dated: June 16, 2020) The MiniBooNE experiment has observed a significant excess of electron neutrinos in a muon neutrino beam, in conflict with standard neutrino oscillations. We discuss the possibility that this excess is explained by a sterile neutrino with a mass ∼ 1 keV that decays quickly back into active neutrinos plus a new light boson. This scenario satisfies terrestrial and cosmological constraints because it has neutrino self-interactions built-in. Accommodating also the LSND, reactor, and gallium anomalies is possible, but requires an extension of the model to avoid cosmological limits.

I. INTRODUCTION In this work, we propose a diﬀerent explanation for the MiniBooNE anomaly, and possibly also for the LSND, Many major discoveries in neutrino physics have reactor, and gallium anomalies. In particular, we con- started out as oddball anomalies that gradually evolved sider a sterile neutrino that rapidly decays back into into incontrovertible evidence. In this work, we enter- Standard Model (“active”) neutrinos νa [22–24]. The tain the possibility that history is repeating itself in the MiniBooNE excess is then interpreted as coming from context of the MiniBooNE anomaly. From 2002 to 2019, these decay products. We will see that this scenario re- the MiniBooNE experiment has been searching for elec- quires only very small mixing between νs and νµ, thus avoiding the strong ν disappearance constraints. It also tron neutrinos (νe) appearing in a muon neutrino (νµ) µ beam [1–3],1 and has found a corresponding signal at requires somewhat larger mixing between νs and νe, in 4.8σ statistical signiﬁcance. For some time, the simplest line with the hints from reactor and radioactive source explanation for this signal appeared to be the existence experiments. Finally, we will argue that decaying sterile neutrinos may avoid cosmological constraints because of a fourth neutrino species νs, called “sterile neutrino” because it would not couple to any of the Standard Model the model automatically endows sterile neutrinos with interactions, but would communicate with the Standard self-interactions (“secret interactions” [25, 26]). Model only via neutrino mixing. If νs has small but non- zero mixing with both ν and ν and if the corresponding e µ II. DECAYING STERILE NEUTRINO mostly sterile neutrino mass eigenstate ν is somewhat 4 FORMALISM heavier ( 1 eV) than the Standard Model neutrinos, the MiniBooNE∼ signal could be explained. This explanation would also be consistent with a similar 3.8σ anomaly We extend the Standard Model by a sterile neutrino νs from the earlier LSND experiment [4], and with several (a Dirac fermion) and a singlet scalar φ. The relevant in- reported hints for anomalous disappearance of electron teraction and mass terms in the Lagrangian of the model neutrinos in reactor experiments [5,6] and in experiments are using intense radioactive sources [7,8]. 2 However, the X g ν¯sνsφ mαβ ν¯ανβ . (1) sterile neutrino parameter space consistent with Mini- L ⊃ − − a=e,µ,τ,s

arXiv:1911.01427v2 [hep-ph] 15 Jun 2020 BooNE and these other anomalies is in severe tension with the non-observation of anomalous νµ disappearance The neutrino flavor eigenstates να are linear combina- [9–19], unless several additional new physics effects are tions of the mass eigenstates νj (j = 1, 2, 3, 4) according invoked concomitantly [20, 21]. to the relation να = Uαjνj, where U is the unitary 4 4 leptonic mixing matrix. The first term in eq. (1) can thus× be rewritten as 2 ∗ g ν¯F νF φ g Us4 ν¯4ν4φ (g U ν¯4νF φ + h.c.) , (2) a [email protected] − − | | − s4 b [email protected] with c [email protected] d [email protected] 3 1 X Here and in the following, when we say neutrino we mean also νF Usiνi . (3) ≡ the corresponding anti-neutrinos. i=1 2 The latter class of experiments is usually referred to as “gallium experiments”, based on the active component of their target ma- We assume initially that the fourth, mostly sterile, mass terial. eigenstate ν4 νs has a mass m4 between (eV) and ' O 2

(100 keV), and that the mass of φ is of the same order, beam in our model can be described by a neutrino den- O but smaller. The last term in eq. (2) will then induce sity matrixρ ˆν (E, x) (a 4 4 matrix in flavor space), the × ν4 νF + φ decays, while the first term is responsible corresponding antineutrino density matrix ρˆ¯ν (E, x), and → for φ νF +ν ¯F decays. When these decays occur in the scalar density function ρφ(E, t). The evolution equa- a neutrino→ beam, they will produce lower-energy neutri- tions are [27, 28], nos at the expense of higher-energy ones, and they may also alter the flavor structure of the beam. In particular, dρˆν (E, t) 1 they can produce excess low-energy νe in a νµ beam, as ˆ m4 ˆ = i[H, ρˆν ] E Γ, ρ + ν [ˆρν , ρφ, E, t] suggested by the MiniBooNE anomaly. dt − − 2 R (4) The phenomenology of the model depends mainly on dρφ(E, t) mφ five new parameters. Besides m4 and mφ, these are the = E Γφρφ + φ[ˆρν , E, t] (5) 2 2 dt − R coupling g and the mixings Ue4 , Uµ4 between ν4 and | | | | νe, νµ. We will assume the mixing with ντ to be zero ˆ 1 2 2 2 where H = 2E diag(0, ∆m21, ∆m31, ∆m41) is the stan- and neglect the complex phases, as these parameters do dard neutrino oscillation Hamiltonian, written here in not play an important role in explaining the MiniBooNE the mass basis, and Γˆ = Γ Πˆ is the decay term, which excess. For practical purposes, it is convenient to quote 4 4 contains the projection operator Πˆ 4 = ν4 ν4 onto the m4Γ4 instead of g, as m4Γ4 appears directly in the lab- fourth, mostly sterile, mass eigenstate| asih well| as the oratory frame decay length E/(m4Γ4). Also, it is conve- decay width Γ4 of ν4 in its rest frame. Similarly, Γφ nient to use the ratio mφ/m4 instead of just mφ because is the rest frame decay width of φ. The functional the ratio measures more directly the kinematic suppres- ν [ˆρν , ρφ, E, t] describes the appearance of the daughter sion in ν4 decays. R neutrinos from ν4 and φ decay. Neglecting the masses of The evolution in energy E and time t of a neutrino ν1, ν2, and ν3, it is given by

Z ∞ lab Z ∞ lab X dΓ (ν4 νkφ) X dΓ (φ νkν¯j) ν [ˆρν , ρφ, E, t] = Πˆ F dE4 ρˆν,44(E4, t) → + Πˆ F dEφ ρφ(Eφ, t) → , (6) R E dEk E dEk 1−x2 k k,j φ4

lab where dΓ (X Y )/dEk are the diﬀerential decay widths for the various decays X Y in the lab frame, and → → xφ4 mφ/m4. The projection operator ≡

3 ∗ ˆ νF νF X UsiUsj ΠF = | ih |2 = P 2 νi νj (7) νF νF Usk | ih | |h | i| i,j=1 k | | isolates the speciﬁc combination of mass eigenstates that appears in ν4 and φ decays, and the integrals run over all parent energies E4, Eφ that lead to daughter neutrinos of energy E. Analogously, φ[ˆρν , E, t] describes the appearance R of scalars from ν4 decay:

2 Z E/xφ4 lab lab X dΓ (ν4 νkφ) dΓ (¯ν4 ν¯kφ) φ[ˆρν , E, t] = dE4 ρˆν,44(E4, t) → + ρˆ¯ν,44(E4, t) → . (8) R dEφ dEφ E k

With the appearance terms ν [ˆρν , ρφ, E, t] and between the three light neutrino mass eigenstates, the R φ[ˆρν , E, t] defined, the equations of motion (4) electron neutrino flux φe(L, E) appearing in a muon andR (5) can be solved analytically if we neglect matter neutrino beam of energy E after a distance L due to effects. Neglecting furthermore the small mass splittings oscillations and decay is given by

2 2 m4Γ4L m4Γ4L ∆m L νe νF 2 2 − E − 2E 41 2 φe(L, E) = φµ(0,E) Ue4 Uµ4 1 + e 2e cos + Uµ4 | h | i | 2 . (9) | | | | − 2E | | νF νF I | h | i |

Here, νF is the superposition of mass eigenstates into which the ν4 decay (deﬁned in eq. (3)), and the decay integral | i 3

is given by I Z ∞ m Γ L lab − 4 4 X 1 dΓ (ν4 νjφ) E4 = dE4 1 e φµ(0,E4) m4 → I 2 − Γ4 dE E/(1−xφ4) j E4 2 ∞ Eφ/x m Γ L Z Z φ4 φ φ m4Γ4L 1 − E m4Γ4L − mφΓφL + dEφ dE4 1 e φ 1 e E4 m4Γ4L mφΓφL E Eφ − E4 − − Eφ E4 − Eφ lab lab lab 1 X dΓ (ν4 νjφ) ¯ dΓ (¯ν4 ν¯jφ) X 1 dΓ (φ νiν¯j) m4 φµ(0,E4) → + φµ(0,E4) → mφ → . × Γ4 dE dE Γφ dE E4 j i,j Eφ (10)

In the above equations, φµ(0,E) and φ¯µ(0,E) are the scenarios (blue dotted histogram in fig.1)[19], which by initial νµ andν ¯µ fluxes, respectively. A completely anal- themselves already offer an excellent fit as long as only ogous equation describesν ¯e appearance. MiniBooNE data are considered (MiniBooNE quotes a The physical interpretation of eq. (9) is straightfor- χ2 per degree of freedom of 9.9/6.7 [3]). Our model, ward: the first term on the right-hand side describes however, is also consistent with all constraints. Notably, νµ νe oscillations, altered by the removal of neutri- it reproduces the angular distribution of the neutrino in- → nos at energy E due to ν4 decay. In fact, this contribu- teraction products in MiniBooNE because it predicts an tion matches the result of ref. [29], on invisible ν4 decay. actual flux of electron neutrinos instead of attempting to The second term gives the contribution from neutrinos mimic the signal with other particles [30–35]. In particu- 2 generated in ν4 and φ decays. The factor Uµ4 arises lar, the angle between the parent νs and the daughter νe | | because ν4 is the only mass eigenstate that decays. It is suppressed by a large Lorentz boost γ (1 000) [36]. ∼ O describes the amount of ν4 in the νµ beam. The fac- This boost is sufficient to ensure that the daughter neu- 2 2 tor νe νF / νF νF is the probability of the decay trinos enter the MiniBooNE detector, which is a 6 m product| h | toi be | detected| h | i as | an electron neutrino, and the sphere located 500 m from the primary target,∼ under integral controls the energy distribution of the decay essentially the same∼ angle as the parent neutrino would products.I have done. Analytic expressions for the decay widths appearing in eqs. (4) to (10) are given in appendixB. IV. CONSTRAINTS

III. FIT TO MINIBOONE DATA We now discuss the various constraints that an explanation of the MiniBooNE anomaly in terms of decaying To compare the predictions of the decaying sterile neu- sterile neutrinos has to respect. The most relevant con- trino scenario to MiniBooNE data, we evolve the un- straints are also summarized in figs.2 and3. oscillated beam following the formulas given above. We (1) Oscillation null results. Putting MiniBooNE into then follow the fitting procedure recommended by the context with other νe appearance searches, we show in MiniBooNE collaboration (see the data releases accom- fig.3 two slices through the 5-dimensional parameter panying refs. [1,3]), but go beyond it by accounting for space of the decaying sterile neutrino model along the 2 2 the impact of νµ and νe disappearance on the signal and plane spanned by Ue4 and Uµ4 . To produce this fig- background normalization (see Appendix for details). ure, we have used fitting| | codes| from| refs. [9, 12, 19] (based Illustrative results are shown in fig.1, where we have partly on refs. [37–39]). We see that most of the param- chosen parameter values that give an optimal fit to Mini- eter region preferred by MiniBooNE is well compatible BooNE data while being consistent with null results from with the KARMEN short-baseline oscillation search [40] other oscillation experiments, as well as non-oscillation and with the OPERA long-baseline experiment [41]. We 2 constraints. At m4Γ4 = 2.1 eV , most ν4 will have de- have checked that the limits from ICARUS [42–44] and cayed before reaching the detector. The value mφ/m4 = E776 [45] are significantly weaker. 2 2 0.82 implies mild phase space suppression in ν4 decays, All constraints on Ue4 ( Uµ4 ) from νe (νµ) disap- | | | | which tends to shift the νe spectrum to lower energies, in pearance experiments are avoided [17, 19, 46]. This is excellent agreement with the data. Compared to models mostly because in pure oscillation scenarios the number with massless φ [22, 23], our scenario also has the advan- of excess events in MiniBooNE and LSND is proportional 2 2 tage that it allows φ νF ν¯F decays, further boosting the to Ue4 Uµ4 , while in our scenario it is proportional → | | | 2| 2 2 νe flux at low energies. It is therefore favored compared only to Uµ4 as long as Ue4 Uµ4 . Therefore, it | | | | | | to the mφ = 0 case at more than 99% confidence level. agrees well even with the tightest constraints [47, 48] The fit in our model is better than in oscillation-only We can already see from fig.3 that MiniBooNE is also 4

��

� �� |�� = �� (ν ��) |�� = �� (��-ν ��) � � |�μ� = �� |�μ� = �� = �� = �� ν� ��μ �� ν� ��μ �ϕ/�� = �� ϕ/�� = �� ν� �� ν� �� Γ= �� Γ= �� π� ��-�� π� ��-�� Δ→�γ Δ→�γ �� +�(�� ) �� +�(�� )

�� / �� / �� ν� ��(�� )� ν� ��(�� )� ��+ �� + ��

��

�� ν [��] �� ν [��]

FIG. 1. Comparison of MiniBooNE neutrino-mode (left) and anti-neutrino-mode (right) data [3] to the predictions of the neutrino oscillation + decay scenario discussed in this work. We show the expected spectrum at the point which optimally ﬁts MiniBooNE data, while being consistent with all null results (orange histogram with systematic error band; parameters given in the plot). We also show the MiniBooNE-only best point for 3 + 1 oscillations without decay (blue dotted histogram, 2 2 2 2 parameter values ∆m41 = 0.13 eV , |Ue4| = 0.024, |Uµ4| = 0.63).

2 Shaded Hatched noless double beta decay requires m4 Ue4 . 0.2 eV [53]. 2 3 2 | | Ue4 2 10− 1.8 10− This is the reason we always focus on Dirac neutrinos in | |2 · 3 · 3 101 Uµ4 4 10− 1.5 10− this work. ) | mφ | · · 2 0.9 0.9 m4 (4) Neﬀ, a measure for the energy density of rela- (eV 4 Perturbativity tivistic particles in the early Universe (green region in Γ 4

m Short Baseline ﬁg.2). The measured value of Neﬀ is very close to the 100

decay SM value of 3 both at the BBN and recombination

β ∼ eff epochs [54, 55]. Naively, one might expect that this N Free streaming 102 103 104 observation precludes the existence of a fourth neutrino m4 (eV) species with m4 . MeV. In our model, however, the Neff constraint is avoided by the “secret interactions” mech- FIG. 2. Non-oscillation constraints on decaying sterile neu- anism [25, 26]: any small abundance of νs generates a 2 trinos for parameters favored by the global fit without LSND temperature-dependent potential Veff g T , reducing p 2 ∝ (shaded), and by the global fit without the free-streaming the νs–νa mixing by a factor ∆m /(EVeff). Hence, constraint (hatched). the production of νs is suppressed until the temperature drops low enough. For the parameter range that the short-baseline anomalies are pointing to, this can easily be postponed to late times (T MeV), after neutrino- compatible with LSND and with the U 2 range pre- e4 electron decoupling. Consequently, when ν are eventu- ferred by the reactor anomaly, but only| in| a parameter s ally produced, they are produced at the expense of active region that would unacceptably reduce free-streaming of neutrinos, so N does not change any more and con- active neutrinos in the early Universe. We will see be- eff straints are automatically satisfied. More quantitatively, low that this tension can be avoided in extensions of the N constraints are avoided when model. eff (2) Beta decay spectra (purple regions in fig.2 and 4 eff −14 2 m4 black dashed lines in fig.3). Direct searches for sterile (m4Γ4) 2 10 eV , (11) & × eV neutrinos looking for anomalous features in beta decay spectra [49–52] suggest that (0.001 0.01) mixings be- where we have defined tween active and sterile neutrinosO – as− required by Mini- BooNE – are allowed for m4 few keV. m Γ . (m Γ )eff 4 4 . (12) 4 4 2 2 (3) Neutrinoless double beta decay. If neutrinos are ≡ 2 2 2 mφ Us4 ( Ue4 + Uµ4 ) 1 m2 Majorana particles, the non-observation so far of neutri- | | | | | | − 4 5

This constraint can be easily satisfied in the mass range φ, are always sufficiently rare to be consistent with so- allowed by beta decay limits. lar neutrino constraints [29, 72]. Note, however, that we P (5) mν , the sum of neutrino masses. Massive neu- predict the cosmic neutrino background today to consist trinos affect the CMB as well as structure formation, and exclusively of ν1 or ν3, for normal and inverted neutrino this has for instance allowed the Planck collaboration to mass ordering, respectively. P set a limit mν . 0.12 eV [55]. In our model, this (9) Perturbativity (red region in fig.2). Requiring that constraint is easily satisfied because in the interesting the νs–φ coupling constant g in eqs. (1) and (2) is < √4π 2 parameter range with m4 1 eV and m4Γ4 & 1 eV , imposes the bound any ν that are produced in the early Universe will have 4 2 decayed via ν4 ν1,2,3 + (φ ν1,2,3ν¯1,2,3) long before eff 2 m4 (m4Γ4) . 0.25 eV . (14) recombination and→ the onset of→ structure formation. eV (6) Neutrino Free-Streaming (blue region in fig.2 and Similarly to the free-streaming bound, this constraint ap- gray dotted lines in fig.3). Via the mixing with ν , also s plies even for very small mixing when m Γ is fixed. This the light neutrino mass eigenstates ν feel φ-mediated 4 4 1,2,3 bound restricts m in our model to be 100 eV for m Γ interactions and are therefore not fully free-streaming. 4 & 4 4 values large enough to explain the MiniBooNE anomaly. This may put the model in tension with CMB obser- In summary, the sterile neutrino mass range to explain vations, which require that neutrinos should free-stream the MiniBooNE anomaly is between 100 eV and 2.5 keV. from about redshift 105 onwards [56–60].3 This requirement bounds the squared coupling among the lightest 22 neutrino mass eigenstate and the scalar φ, i.e., g Us1 . V. THE LSND AND REACTOR ANOMALIES (Heavier mass eigenstates are not relevant as they| decay| quickly.) Here we are taking ν1 to be the lightest mass As shown in fig.3, decaying sterile neutrinos can si- eigenstate, as favoured by current data. Quantitatively, multaneously fit the MiniBooNE and LSND anomalies, 4 4 but only if cosmological neutrino free-streaming con- eff −10 2 m4 0.1 2 (m4Γ4) . 4 10 eV xφ4 . (13) straints can be avoided (see discussion under point (6) × eV Us1 | | above for possible scenarios). Quantitatively, a parame- 2 −6 ter goodness-of-fit test [73] reveals that LSND is incom- with m4 . 200 eV required for g & 10 [60]. Note that in fig.3, this constraint is present even for very patible with the rest of the data at the 4.7σ level if free- streaming constraints hold. If the free-streaming problem small mixings. This is because, at fixed m4Γ4, small mixings need to be compensated by a large coupling g, is solved by other means, this reduces to 2.1σ, implying strengthening the free streaming constraint. The value of consistency. The best fit to all data including LSND, 2 2 2 but excluding free-streaming is found at m4 = 97 eV, Us1 is fixed in terms of Ue1 and Uµ1 by unitarity, 2 2 2 | | | | | | Ue4 = 0.018, Uµ4 = 0.0015, m4Γ4 = 0.87 eV , assuming the active neutrino mixing angles to be fixed | | | | mφ/m4 = 0.89. at their values from Ref. [66]. 2 Interestingly, at this value of Ue4 , the model can also However, the constraint could be substantially weak- | | ened in extensions of our model, see for instance refs. [67– explain the flux deficit observed in reactor and gallium 70]. A minimalist example is the production of extra experiments [5–8, 19, 74]. We test our model against species of light particles at the expense of the neutrino reactor data by comparing to Daya Bay’s generic flux- sector after neutrino decoupling. These would compen- weighted cross section [75]. To estimate the viable pa- sate for the lack of free-streaming in active neutrinos. rameter space we perform a chi-square-test using the co- (7) SN 1987A. The fact that neutrinos from super- variance matrix given in the same reference. In addition nova 1987A could be observed at Earth without be- we introduce a 2.4% systematic flux normalization error corresponding to the theoretical uncertainty, in accor- ing absorbed through scattering on the cosmic neutrino 2 dance with fig. 28 of ref. [75]. The Ue4 region preferred background constrains neutrino self-interactions [71]. We | | have checked that, due to mixing suppression, these con- by reactor experiments is included in fig.3, and a com- straints are avoided in our scenario. Note that supernova parison of the reactor neutrino spectrum to our model cooling, which is sensitive to non-interacting sterile neu- prediction is shown in fig.4. trinos, does not constrain our model as ν4 and φ quickly decay to lighter neutrinos that remain trapped in the su- VI. DETAILED INVESTIGATION OF THE pernova core. PARAMETER SPACE (8) Decays of SM neutrinos. We have checked that decays of the form ν2,3 ν¯1 +2ν1, mediated by an off-shell → To supplement fig.3 and give the reader a broader overview of the preferred parameter regions of decaying sterile neutrinos, we show in figs.5 and6 additional slices 3 It is noteworthy, though, that some cosmological fits have actu- through the 5-dimensional parameter space. ally found a preference for neutrino self-interactions [56, 61–65] The color coding in the figure is the same as in that could be accommodated in our model. fig.3: the yellow, banana-shaped regions are preferred 6

�� % �� = �� % �� = �� Γ� = � �� Γ� = � �� β

��

�� / �� ϕ/�� = �� ϕ/�� = ��

�

/ ��

��

� �� β �� μ � �� μ � ��

�� | � | �

- ��

�� /� �� |�� |��

2 2 FIG. 3. Allowed values of the squared mixing matrix elements |Ue4| and |Uµ4| (measuring the mixing of νs with νe and νµ, respectively) in the decaying sterile neutrino scenario. We show two representative slices through the 5-dimensional 99% confidence regions. Our fits include MiniBooNE, OPERA, ICARUS, E776, and KARMEN data, as well as constraints from nuclear beta decay spectra and from the requirement of neutrino free-streaming in the early Universe. For the null results from oscillation experiments, the region to the right of the curves is excluded. For the free-streaming constraint, the region to the 2 left of the gray dotted contour is excluded. We also show, as a black rule at the bottom of the plot, the |Ue4| range preferred by the reactor neutrino anomaly. Constraints on νµ disappearance are significantly weaker here than in the 3 + 1 scenario without decay, and are hence not shown. We also do not show a fit including both LSND and cosmology as the goodness of fit 2 2 would be very poor. Note that the global combinations are sensitive to five degrees of freedom, namely m4, |Ue4| , Uµ4, m4Γ4, and mφ/m4; oscillation experiments are sensitive only to the last four of these; beta decay spectra depend on two degrees of 2 2 freedom (m4 and |Ue4| ); reactor experiments depend only on |Ue4| ; and the free-streaming constraint depends only on the parameter combination m4/|Us1|.

by MiniBooNE, the large dark red ones by LSND; the Exclusion limits from oscillation experiments do not de- 2 orange regions at low Ue4 correspond to a global fit to pend on m4 for m4 eV. | | MiniBooNE, OPERA, ICARUS, E776, KARMEN, nu- Comparing fig.5 with mφ/m4 = 0.5 and fig.5 with clear beta decay spectra, and cosmological free-streaming mφ/m4 = 0.9, we see that it becomes in general more constraints; bright red regions show instead a global difficult to fit all experiments at smaller mφ/m4. The fit to MiniBooNE, LSND, OPERA, ICARUS, E776, reason is that, at small mφ/m4, the active neutrinos pro- KARMEN, and nuclear beta spectra, but excluding duced in ν4 and φ decays have a harder spectrum. This the free-streaming constraint. Solid lines indicate con- in particular makes it more difficult to explain the Mini- straints from OPERA (blue), ICARUS (purple), KAR- BooNE low-energy excess. In fact, for even smaller values MEN (cyan), E776 (green), nuclear beta decay spectra of mφ/m4, and in particular for nearly massless φ (as con- (black dashed), and free-streaming in the early Universe sidered in refs. [22, 23]), the MiniBooNE-preferred region (black dotted). The region to the right of the lines is would disappear completely from the plots. excluded.

We observe that, at smaller values of m4Γ4, the allowed parameter regions from short-baseline oscillations (Mini- VII. CONCLUSIONS BooNE, LSND, KARMEN) shift towards larger values of U 2 and U 2. In this case, only a small fraction e4 µ4 In summary, we have shown that scenarios in which of neutrinos| | decays| | before reaching the detector, making the SM is extended by a sterile neutrino that has a de- the phenomenology more similar to that of 3 + 1 mod- cay mode to active neutrinos can well explain the Mini- els without decay. Strong constraints from beta decay BooNE anomaly without violating any constraints. An spectra and from cosmology imply that a good global fit 2 explanation of the LSND and reactor/gallium anomalies cannot be achieved at m4Γ4 1 eV . is possible if the model is extended to avoid constraints Regarding the dependence of the fit on m4, we note on neutrino free-streaming in the early Universe. The that smaller values of m4 are favored by beta decay spec- preferred mass of the sterile neutrino is of order few hun- tra, but disfavored by cosmology, in agreement with fig.2. dred eV. 7

10 42 0.20 × − orations’ ﬁt as described in the supplemental material to ] . Data ref. [1], and using the data released with ref. [3]. In par- Huber-Mueller ﬁss

/ 0.15 Huber-Mueller w decay ticular, we consider the following three eﬀects, which are \ Model parameters: Theoretical Error relevant in a ﬁt to a 3+1 scenario, but are not encoun- MeV

/ 0.10 m4 =97eV

2 2 m4Γ =0.87eV tered in a 2-ﬂavor ﬁt.

[cm mφ /m4 =0.89 0.05 2 2 Ue4 =1.8 10−

Flux-Weight. Cross Sect. | | 2 · 3 1. Normalization of the νµ νe oscillation sig- Uµ4 =1.5 10− → | | · nal. To predict the number of expected νe events 1.15 from νµ νe oscillations for a given set of oscilla- . 1 00 tion parameters,→ the initial ν ﬂux must be known. 0.85 µ

Ratio to Pred. It is obtained in situ using MiniBooNE’s own sam-

(Huber-Mueller) 2 3 4 5 6 7 8 Neutrino Energy [MeV] ple of νµ events. Note, however, that in a 3+1 model, the measured νµ flux will be reduced by an 2 amount Uµ4 due to νµ νs oscillations. (This FIG. 4. Comparison of the reactor anti-neutrino spectrum ∼ | | → predicted in the decaying sterile neutrino scenario discussed effect is unimportant in a 2-flavor model, where 2 in this work (blue) to the standard Huber–Mueller prediction the deficit is only of order sin 2θµe, where θµe is (orange-dashed) [76, 77] and to Daya Bay data (black data the effective 2-flavor mixing angle.) We account points with error bars) [75]. For model parameters motivated for this effect by first computing the expected νe by the MiniBooNE and LSND anomalies, a flux deficit con- signal based on the unoscillated MiniBooNE flux, sistent with the reactor anomaly can be accommodated. (See and then diving it by the νµ survival probability in text for details, and for a discussion of how possible cosmo- each bin. logical constraints can be avoided.) The impact of this change in normalization is illus- trated in the top panels of fig.7. The colored region ACKNOWLEDGMENTS in panel (a) of this figure shows our reproduction of the official MiniBooNE fit, which is shown as black We would like to thank André de Gouvea, Frank contours. In panel (b), we have included the change Deppisch, Matheus Hostert, Bill Louis, and Michele in normalization for the signal. Maltoni for very helpful discussions. We are partic- 2. Oscillations of the ν backgrounds. Part of ularly grateful to Thomas Schwetz for sharing his fit- e the MiniBooNE background is constituted by the ting codes, from which some of our own codes took in- intrinsic ν contamination in the beam. In a 2- spiration. We also thank Andréde Gouvea, O. L. G. e flavor fit, this contribution to the total event rate Peres, Suprabh Prakash and G. V. Stenico for sharing 2 is only modified by a factor of order sin 2θ , but their draft on a similar neutrino decay solution to the µe in the full 3+1 framework, it is reduced by a factor SBL anomalies [24]. We thank Miguel Escudero and of order U 2 instead. The impact of this modifi- Samuel J. Witte for sharing their cosmological constraint. e4 cation to| the| background sample is shown in fig.7 IE would like to thank CERN and Johannes Guten- (c). berg University Mainz for their kind hospitality. This work has been funded by the German Research Founda- 3. Oscillations of the νµ sample. The fit described tion (DFG) under Grant Nos. EXC-1098, KO 4820/1–1, in the supplemental material to ref. [1] which we are FOR 2239, GRK 1581, and by the European Research following includes also MiniBooNE’s sample of νµ Council (ERC) under the European Union’s Horizon events. This is necessary to properly account for 2020 research and innovation programme (grant agree- systematic uncertainties which are correlated be- ment No. 637506, “νDirections”). Fermilab is oper- tween the two samples. But of course, in a 3+1 ated by Fermi Research Alliance, LLC under contract scenario, the νµ sample suffers from νµ disappear- 2 No. DE-AC02-07CH11359 with the United States De- ance into νs, proportional to Uµ4 . (Once again, partment of Energy. IE acknowledges support from the in a 2-flavor model, only a much| | smaller fraction 2 FPU program fellowship FPU15/03697, the EU Net- sin 2θµe will disappear, which is usually negligi- ∝ works FP10ITN ELUSIVES (H2020-MSCA-ITN-2015- ble.) The impact of including νµ disappearance is 674896) and INVISIBLES-PLUS (H2020-MSCA-RISE- shown in panel (d) of fig.7. 2015-690575), and the MINECO grant FPA2016-76005- C2-1-P. We see that including the effect of 3+1 oscillations on the normalization in the control regions and on the background prediction reduces the significance of the Mini- Appendix A: Impact of Oscillations on the BooNE anomaly, though it remains above 3σ. These Background Prediction in MiniBooNE effects are thus unable to fully explain the MiniBooNE anomaly, but they could well be part of an “Altarelli In this appendix, we briefly discuss our fit to Mini- cocktail” of several effects conspiring to lead to the large BooNE data, and in what ways it differs from the collab- observed excess [78]. 8

�� % �� = �� % �� = �� % �� = �� Γ� = �� Γ� = �� Γ� = �� ϕ/�� = �� ϕ/�� = �� ϕ/�� = ��