Many current and future dark matter and neutrino detectors are designed to measure scintillation light with a large array of photomultiplier tubes (PMTs). The energy resolution and particle identification capabilities of these detectors depend in part on the ability to accurately identify individual photoelectrons in PMT waveforms despite large variability in pulse amplitudes and pulse pileup. We describe a Bayesian technique that can identify the times of individual photoelectrons in a sampled PMT waveform without deconvolution, even when pileup is present. To demonstrate the technique, we apply it to the general problem of particle identification in single-phase liquid argon dark matter detectors. Using the output of the Bayesian photoelectron counting algorithm described in this paper, we construct several test statistics for rejection of backgrounds for dark matter searches in argon. Compared to simpler methods based on either observed charge or peak finding, the photoelectron counting technique improves both energy resolution and particle identification of low energy events in calibration data from the DEAP-1 detector and simulation of the larger MiniCLEAN dark matter detector.

The DEAP-1 \SI{7}{kg} single phase liquid argon scintillation detector was operated underground at SNOLAB in order to test the techniques and measure the backgrounds inherent to single phase detection, in support of the \mbox{DEAP-3600} Dark Matter detector. Backgrounds in DEAP are controlled through material selection, construction techniques, pulse shape discrimination and event reconstruction. This report details the analysis of background events observed in three iterations of the DEAP-1 detector, and the measures taken to reduce them. The $^{222}$Rn decay rate in the liquid argon was measured to be between 16 and \SI{26}{\micro\becquerel\per\kilogram}. We found that the background spectrum near the region of interest for Dark Matter detection in the DEAP-1 detector can be described considering events from three sources: radon daughters decaying on the surface of the active volume, the expected rate of electromagnetic events misidentified as nuclear recoils due to inefficiencies in the pulse shape discrimination, and leakage of events from outside the fiducial volume due to imperfect position reconstruction. These backgrounds statistically account for all observed events, and they will be strongly reduced in the DEAP-3600 detector due to its higher light yield and simpler geometry.