Optimizing EDELWEISS detectors for low-mass WIMP searches

The physics potential of EDELWEISS detectors for the search of low-mass weakly interacting massive particles (WIMPs) is studied. Using a data-driven background model, projected exclusion limits are computed using frequentist and multivariate analysis approaches, namely, profile likelihood and boosted decision tree. Both current and achievable experimental performances are considered. The optimal strategy for detector optimization depends critically on whether the emphasis is put on WIMP masses below or above $\sim 5\mathrm{GeV}/{\mathrm{c}}^2$. The projected sensitivity for the next phase of the EDELWEISS-III experiment at the Modane Underground Laboratory (LSM) for low-mass WIMP search is presented. By 2018 an upper limit on the spin-independent WIMP-nucleon cross section of ${\sigma }_{\mathrm{SI}}=7\times {10}^{-42}{\mathrm{cm}}^2$ is expected for a WIMP mass in the range $2–5\mathrm{GeV}/{\mathrm{c}}^2$. The requirements for a future hundred-kilogram-scale experiment designed to reach the bounds imposed by the coherent scattering of solar neutrinos are also described. By improving the ionization resolution down to $50{\mathrm{eV}}_{ee}$, we show that such an experiment installed in an even lower background environment (e.g., at SNOLAB) together with an exposure of $1000\mathrm{kg}·\mathrm{yr}$, should allow us to observe about 80 ${}^8\mathrm{B}$ neutrino events after discrimination.

Figure 1

Constraints in the spin-independent (SI) WIMP-nucleon cross section as a function of WIMP mass. Closed contours correspond to signal hints reported by the CDMS-II Si [19] (dashed blue, 90% C.L.); CoGeNT [20] (dashed green, 90% C.L.); CRESST-II [21] (dashed pink, 95% C.L.); and DAMA/LIBRA [22] (dashed purple, 90% C.L.) experiments. Results interpreted as 90% C.L. exclusion upper limits are represented by lines: experimental limits shown are from DAMIC [12] (green), CRESST [13] (pink), SuperCDMS low mass [14] (dotted dark blue), CDMSlite with Ge [15] (dark blue), PandaX-II [5] (brown), LUX combined [4] (turquoise blue), XENON-100 high and low mass [3] (dashed orange and orange), EDELWEISS low mass [16] (red), ZEPLIN-III [17] (blue-green), and DarkSide-50 [18] (pale brown). The region delimited by the yellow dashed line corresponds to the neutrino floor [23].

Figure 2

(a): A FID detector. (b): Cross section of a FID detector as indicated by the dashed purple lines on panel (a). The zone in semitransparency delimited by orange lines indicates the fiducial zone (i.e., where an energy deposit leads to charge collection on fiducial electrodes B and D as schematized on the picture). The color code gives the electric potential map. (c): Charge collection for a surface event (i.e., involving at least one veto electrode).

Figure 3

Event rate for a total exposure of $1\mathrm{kg}·\mathrm{d}$ as a function of the recoil energy. Solid lines correspond to the recoil energy spectra of the different background components as indicated by the color code. The dashed blue line shows the theoretical spectrum of a $6\mathrm{GeV}/{\mathrm{c}}^2$ WIMP with ${\sigma }_{\mathrm{SI}}=\mathrm{4.4.}\times {10}^{-45}{\mathrm{cm}}^2$, which is extremely similar to the one of solar ${}^8\mathrm{B}$ neutrinos represented with a solid gray line.

Figure 4

Top panel: Simulated data in the veto ionization energy ($E_{\mathrm{veto}}$) vs. normalized heat energy (${\tilde{E}}_{\mathrm{heat}}$) plane before fiducial selection for a total exposure of $1000\mathrm{kg}·\mathrm{d}$ with standard FID performance: ${\sigma }_{E_{\mathrm{heat}}}=500\mathrm{eV}$, ${\sigma }_{E_{\mathrm{fid}}}=200{\mathrm{eV}}_{\mathrm{ee}}$, and a bias voltage $V_{\mathrm{fid}}=8\mathrm{V}$. The fiducial cut is indicated by the dashed red line. Colored contours correspond to a theoretical $15\mathrm{GeV}/{\mathrm{c}}^2$ WIMP signal (both bulk and surface events), which is used as an illustration. Middle and bottom panels: Same as the top panel but in the fiducial ionization energy ($E_{\mathrm{fid}}$) vs. normalized heat energy (${\tilde{E}}_{\mathrm{heat}}$) plane, before (middle) and after (bottom) fiducial selection.

Figure 5

Comparison of the progression on exclusion limits with the exposure derived either from the likelihood (thick solid lines) or BDT (thick dashed lines) analyses. Three values of the exposure are considered: $10\mathrm{kg}·\mathrm{d}$ (purple), $100\mathrm{kg}·\mathrm{d}$ (orange), and $1000\mathrm{kg}·\mathrm{d}$ (red). Concerning detector configuration, we use the standard bias ($V_{\mathrm{fid}}=8\mathrm{V}$) but expected energy resolutions (${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$). The thin dashed lines correspond to the background-free sensitivity. All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1.

Figure 6

Influence of the bias voltage on the sensitivity at fixed ionization and heat resolutions (${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$). The color code indicates the bias condition rising with a 10 V step from 10 V in purple to 100 V in red. Solid and dashed lines refer to the exclusion limits and to the background-free sensitivities for an exposure of $500\mathrm{kg}·\mathrm{d}$. All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1.

Figure 7

Effect of various experimental parameters on the sensitivity derived from the likelihood analysis for an exposure of $500\mathrm{kg}·\mathrm{d}$ at both 100 V (left panels) and 8 V (right panels). All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1. Top panels: Impact of backgrounds for fixed heat and ionization resolutions (${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$). The solid red line corresponds to the exclusion limit when all backgrounds are included. The dashed blue line indicates the background-free sensitivity. The other limits are derived by considering each background separately as indicated by the color code. Middle panels: Impact of varying heat and ionization energy resolutions considering all background components. Heat energy resolution values are distinguished by their color code: red, orange, and purple for ${\sigma }_{E_{\mathrm{heat}}}=500\mathrm{eV}$, ${\sigma }_{E_{\mathrm{heat}}}=300\mathrm{eV}$, and ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, respectively, while dashed and solid lines correspond to ${\sigma }_{E_{\mathrm{fid}}}=200$ and $100{\mathrm{eV}}_{\mathrm{ee}}$, respectively. The thin dashed lines correspond to the background-free sensitivity, with the same color code as used for ${\mathrm{E}}_{\mathrm{heat}}$. Bottom panels: Impact of different detector designs considering all the background components for fixed heat and ionization resolutions (${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$). The solid purple lines refer to the standard FID detector design and the orange ones correspond to the same design without the ability to read veto channels. The red lines correspond to a FID detector only reading the heat channel, with still two distinct collection biases ($V_{\mathrm{surf}}$ and $V_{\mathrm{fid}}$). Finally the dashed black line refers to a detector in coplanar mode only equipped with one heat channel. The thin dashed purple line corresponds to the background-free sensitivity associated with the standard FID detector design.

Figure 8

EDELWEISS-III projected sensitivities considering that expected R&D upgrades will be achieved either for ionization or heat resolutions, with current LSM setup and background budget. Exclusion limits are derived from both the boosted decision tree and profile likelihood ratio approaches and for the two extreme bias voltage conditions (8 V and 100 V). For both $V_{\mathrm{fid}}$ values, a likelihood analysis gives better limits than a BDT one. Below WIMP masses of $4–5\mathrm{GeV}/{\mathrm{c}}^2$, the best sensitivity is obtained by lowering the thresholds, with the Luke-Neganov boost corresponding to a bias voltage of 100 V, keeping current ionization resolution ${\sigma }_{E_{\mathrm{fid}}}=200{\mathrm{eV}}_{\mathrm{ee}}$ and improving heat resolution to ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$. Official EDELWEISS-III low-mass projected sensitivity is thus given by the solid black line exclusion limit. The background-free sensitivity is shown in thin dashed lines. All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1.

Figure 9

Projected sensitivities for a large exposure of $50000\mathrm{kg}·\mathrm{d}$ in the context of a hundred-kg-scale EDELWEISS-like experiment with strongly improved background levels and R&D upgrade performance achieved, with baseline resolutions of ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$ in heat and ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$ in ionization. Limits are computed using a likelihood analysis at 8 V (purple) and 100 V (black) assuming a suppression of the heat-only background (solid line), and no more neutron background associated with a reduction of the Compton background by a factor of 10 (thick dashed line). The background-free sensitivity is shown in thin dashed lines. All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1.

Figure 10

Projected sensitivities computed using a likelihood analysis at 8 V for a large exposure of $50000\mathrm{kg}·\mathrm{d}$ in the context of a hundred-kg-scale EDELWEISS-like experiment, considering the R&D upgrade performed on heat baseline resolution, with ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$, and assuming different conditions of background reduction. The dotted-dashed purple line upper limit, which is obtained with ${\sigma }_{E_{\mathrm{fid}}}=100{\mathrm{eV}}_{\mathrm{ee}}$, considering a reduction of heat-only background by a factor of 100, no more neutron background, and a reduction of the Compton background by a factor of 10, obtained by putting upgraded EDELWEISS detectors in a high-purity-level dedicated environment, has already been shown in Fig. 9. Its purpose is to guide the eye. The projected sensitivities in solid and thin dashed lines shown in this figure are obtained assuming a more drastic improvement on ionization resolution, with ${\sigma }_{E_{\mathrm{fid}}}=50{\mathrm{eV}}_{\mathrm{ee}}$. Solid line upper limits correspond to different background reductions: the blue one is obtained with the same background conditions as for the dotted-dashed purple line, whereas the orange one is performed by removing all backgrounds, including surface events, except the one due to ${}^8\mathrm{B}$ neutrinos. The thin dashed blue line represents the background-free sensitivity. All upper limits, contours, and regions plotted in gray in this figure are those described in Fig. 1.

Figure 11

Simulation of ${}^8\mathrm{B}$ coherent neutrino-nucleus scattering signal and background events, assuming the current EDELWEISS-III background budget, with the exception of no more heat-only background, no neutron events, and a reduction of the Compton background by a factor of 10. These simulations are computed for an exposure of $1000\mathrm{kg}·\mathrm{yr}$, at 8 V, with ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$ and ${\sigma }_{E_{\mathrm{fid}}}=50{\mathrm{eV}}_{\mathrm{ee}}$ baseline energy resolutions. The separation of the ${}^8\mathrm{B}$ signal (medium gray dots) from background events is clearly shown in the plane of fiducial ionization energy versus normalized heat energy, in ${\mathrm{keV}}_{ee}$.

Figure 12

BDT distribution simulated for a ${}^8\mathrm{B}$ neutrino signal, obtained for an exposure of $1000\mathrm{kg}·\mathrm{yr}$, at 8 V, with a heat baseline resolution of ${\sigma }_{E_{\mathrm{heat}}}=100\mathrm{eV}$ and ionization baseline resolution of ${\sigma }_{E_{\mathrm{fid}}}=50{\mathrm{eV}}_{\mathrm{ee}}$. Colored distributions are associated with the different background models. The expected ${}^8\mathrm{B}$ neutrino signal is represented in gray. The BDT distribution of the simulated data in the ROI corresponds to the black dots. Above a BDT score cut of 0.5, a clear signal of 78 ${}^8\mathrm{B}$ neutrino events is obtained after discrimination.