Scintillation time dependence and pulse shape discrimination in liquid argon

Using a single-phase liquid argon detector with a signal yield of 4.85 photoelectrons per keV of electronic-equivalent recoil energy (keVee), we measure the scintillation time dependence of both electronic and nuclear recoils in liquid argon down to 5 keVee. We develop two methods of pulse shape discrimination to distinguish between electronic and nuclear recoils. Using one of these methods, we measure a background- and statistics-limited level of electronic recoil contamination to be $7.6\times {10}^{-7}$ between 52 and 110 keV of nuclear recoil energy (keVr) for a nuclear recoil acceptance of $50\%$ with no nuclear recoil-like events above 62 keVr. Finally, we develop a maximum likelihood method of pulse shape discrimination based on the measured scintillation time dependence.

Figure 1

Schematic representation of the scintillation cell.

Figure 2

Example of an electronic recoil event from a single PMT, digitized by the eight-bit WFD, sampling at 500 MHz. Four μs of presamples are recorded to measure the baseline.

Figure 3

Schematic of the neutron scattering setup.

Figure 4

Example spectrum of ${}^{57}\mathrm{Co}$ data, along with a simulation done using the Reactor Analysis Tool [30].

Figure 5

Example fit of a single exponential to an average trace to measure the long time constant in liquid argon.

Figure 6

Observed and predicted mean voltage traces for nuclear and electronic recoil events of 80–99 photoelectrons.

Figure 7

Scatter plot of $f_p$ vs energy for tagged electronic and nuclear recoils, where $\xi =90$ ns.

Figure 8

Estimated mean $f_p$ vs energy for both event classes, where $\xi =90$ ns.

Figure 9

Measured electronic recoil contamination obtained by using the prompt fraction method and the multibin method. Also shown are the background estimation and model predictions described in the text. We include a model with the additional noise set to 0 for comparison. No contamination events above 69 keVr were observed by use of either method. The energy axis has been scaled from keVee to keVr by use of a constant nuclear recoil scintillation efficiency of 0.29 as discussed in the text.

Figure 10

Projections of the electronic recoil data from Fig. 7 onto the $y$ axis for 14–15 keVee (top) and 30–31 keVee (bottom) events, fitted by Eq. (11) as discussed in the text.

Figure 11

Estimated ${\sigma }_l^2$ and ${\sigma }_p^2$ parameters from the statistical model plotted against the estimated number of photoelectrons in the late and prompt components, respectively. Each distribution is fitted by a straight line, and the ideal case where ${\sigma }^2=\mu$ is also shown.

Figure 12

Scatter plot of $\ln R_m$ vs energy for both electronic and nuclear recoils.

Figure 13

Projections of Fig. 12 onto the $y$ axis for 14–15 keVee (top) and 30–31 keVee (bottom) events, with Gaussian fits to both the electronic (left) and nuclear (right) recoil distributions.

Figure 14

Monte Carlo estimates of the log-likelihood statistics for events that yield 50 photoelectrons.

Figure 15

Predicted performance of maximum likelihood and prompt ratio PSD methods for a detector yielding 6 photoelectrons/keVee (the energy axis has been scaled from keVee to keVr by use of a constant nuclear recoil scintillation efficiency of 0.29 as discussed in Sec. 3b).