Dark matter search in nucleon, pion, and electron channels from a proton beam dump with MiniBooNE

A search for sub-GeV dark matter produced from collisions of the Fermilab 8 GeV Booster protons with a steel beam dump was performed by the MiniBooNE-DM Collaboration using data from $1.86\times {10}^{20}$ protons on target in a dedicated run. The MiniBooNE detector, consisting of 818 tons of mineral oil and located 490 meters downstream of the beam dump, is sensitive to a variety of dark matter initiated scattering reactions. Three dark matter interactions are considered for this analysis: elastic scattering off nucleons, inelastic neutral pion production, and elastic scattering off electrons. Multiple data sets were used to constrain flux and systematic errors, and time-of-flight information was employed to increase sensitivity to higher dark matter masses. No excess from the background predictions was observed, and 90% confidence level limits were set on the vector portal and leptophobic dark matter models. New parameter space is excluded in the vector portal dark matter model with a dark matter mass between 5 and $50\mathrm{MeV}{c}^{-2}$. The reduced neutrino flux allowed to test if the MiniBooNE neutrino excess scales with the production of neutrinos. No excess of neutrino oscillation events were measured ruling out models that scale solely by number of protons on target independent of beam configuration at $4.6\sigma$.

Figure 1

The production of neutrinos in the Booster Neutrino Beamline in on-target running [22].

Figure 2

Feynman diagrams for the production channels relevant for the MiniBooNE dark matter search [20].

Figure 3

Feynman diagrams for the dark matter interactions with nucleons and electrons in MiniBooNE. The $\mathrm{\Delta }$, in the bound-nucleon case, would be observed by its decay products: a pion and a nucleon.

Figure 4

Zoomed-in example of the BNB pulse microstructure as measured by the RWM. The data points come from neutrino-mode ${\nu }_{\mu }$ charged-current interactions in the MiniBooNE detector during 2015–2016. The example RWM trace is plotted by the readout value of the trace.

Figure 5

The production of dark matter in off-target running [20].

Figure 6

(Top) Production of dark matter and neutrinos when the beam hits a thin target. (Bottom) The production of dark matter and suppression of neutrino generation when the beam hits a thick target.

Figure 7

The simulated geometry around the target. Those listed with an asterisk were added for the off-target simulation. The added materials change the neutrino-mode flux by less than a percent.

Figure 8

Comparing CCQE data in off-target mode to three different Monte Carlo predictions for neutrinos interacting in the detector (${\nu }_{\det }$). The dotted line is the output of the nominal off-target beam profile, the dashed line is the nominal profile scale by 1.6, and the solid line is the average of the scrapings (Average CV) used as the final ${\mathrm{\Phi }}_{\mathrm{Off}}$ [41]. $E_{\nu }^{\mathrm{QE}}$ is defined by Eq. (5).

Figure 9

Picture of FRED at the 25-m absorber.

Figure 10

(Top) The off-target neutrino flux seen by the MiniBooNE detector. (Bottom) The off-target/neutrino flux ratio [20].

Figure 11

The ${\theta }_{{\pi }^0}$ vs $p_{{\pi }^0}$ distributions from BooNEG4Beam used for generating dark matter candidate events. The color scale gives the number of pions per delivered POT in each bin.

Figure 12

Illustration showing how the RWM time signature is passed to the detector. Production of neutrinos and dark matter particles are also shown for comparison. Heavy dark matter will arrive later than the neutrinos.

Figure 13

Comparison of simulated and measured CCQE bunch times after applying $\delta t({\sigma }_{\text{inst}})$ and $\delta t_{\text{data}}$ calibrations (see text). Only statistical errors are shown.

Figure 14

Illustration of the timing difference between an electron event and a backward-going photon.

Figure 15

Comparison of the bunch-time shape for different event types, that pass NCE selection cuts, as predicted by the detector simulation.

Figure 16

The expected number of events before and after NCE and $\mathrm{NC}{\pi }^0$ cuts for true NCE, true $\mathrm{NC}{\pi }^0$ and true $\mathrm{NC}{\pi }^{\pm }$ interactions. NCE cuts are just as efficient at detecting true $\mathrm{NC}{\pi }^0$ events at low energy transfer because of the pion absorption in the nucleus.

Figure 17

CCQE, NCE, and $\mathrm{NC}{\pi }^0$ distributions for neutrino, antineutrino and off-target modes. The points are data with statistical errors. The triangles with light lines are constrained predictions with constrained systematic errors. The dotted lines are the total predictions with unconstrained systematic errors. The backgrounds that add up to give the total predictions are given by the stacked histogram.

Figure 18

The (top) visible electron energy $E_{\mathrm{vis}}^e$, (middle) electron angle $\cos {\theta }_e$, and (bottom) bunch time distributions that pass $\nu \text{−}e$ cuts for off-target mode. The prediction was scaled to match the number of data events for $0.9\le \cos {\theta }_e<0.99$. An example dark matter prediction is given (dashed line) to illustrate how forward the resulting electron is expected to be.

Figure 19

The $E_{\nu }^{\mathrm{QE}}$ distribution for events that pass the ${\nu }_e$ oscillation cuts. Data comes from off-target mode.

Figure 20

(Top) The integral number of events that pass NCE, $\mathrm{NC}{\pi }^0$, and $\nu \text{−}e$ selection cuts as a function of $m_{\chi }$. Predictions for the total, $0\pi$, $1{\pi }^0$, and $1{\pi }^{\pm }$ are given by solid, dash-dotted, dotted, and short-dashed lines respectively. The colors of the lines correspond with the bottom figures. (Bottom left) The mean $Q_{\mathrm{QE}}^2$ distribution for NCE selection cuts. (Bottom middle) The mean $p_{\pi }^0$ distribution for $\mathrm{NC}{\pi }^0$ selection cuts. (Bottom right) The mean $E_{\mathrm{vis}}^e$ distribution for $\nu \text{−}e$ selection cuts. All plots are functions of $m_{\chi }$ with $m_V=3m_{\chi }$ and ${\epsilon }^4{\alpha }_D=1\times {10}^{-13}$.

Figure 21

Comparison between signal interactions that pass (a) NCE and (b) $\mathrm{NC}{\pi }^0$ cuts to different dark matter distributions as a function of bunch time. The neutrino interactions are for the neutrino mode while dark matter is for the off-target mode (see text).

Figure 22

The 90% confidence level limit for (a) full nucleon with timing and (b) electron with timing on ${\epsilon }^4{\alpha }_D$ for various combinations of $m_V$ and $m_{\chi }$ using the vector portal dark matter model.

Figure 23

Comparing the full nucleon and electron confidence level results to the elastic nucleon results from Ref. [20]. Also shown is the result when including timing (solid lines) compared to that when not including the timing (dashed lines).

Figure 24

Comparison of the MiniBooNE confidence level limits (solid lines), and sensitivities (dashed lines) to other experiments for (a) $Y$ as a function of $m_{\chi }$ assuming ${\alpha }_D=0.5$ and $m_V=3m_{\chi }$ and (b) in the leptophobic dark matter model with $m_V=3m_{\chi }$. An explanation of vector portal limits lines was given in Refs. [9, 29, 36, 37, 38]. An explanation of the leptophobic limit lines was given in Refs. [8, 34, 35].

Figure 25

90% confidence level in the vector portal dark matter model with (a) $Y$ as a function of $m_{\chi }$ assuming ${\alpha }_D=0.1$ and $m_V=3m_{\chi }$ and (b) ${\alpha }_D=0.1$ and $m_V=7m_{\chi }$. An explanation of the limit lines was given in Refs. [9, 29, 36, 37, 38].