Characterization of the Hamamatsu R12699-406-M4 Photomultiplier Tube in Cold Xenon Environments

TL;DR

The paperCharacterizes the Hamamatsu R12699-406-M4 PMT for cold xenon environments, detailing its SPE response, dark counts, light emission, afterpulsing, and long-term stability, and demonstrates its integration in a small dual-phase TPC. Using the MarmotX facility, SPE analyses reveal gain well above $2 imes 10^{6}$ at −1.0 kV and SPE resolutions around 35–37%, while DCR drops dramatically in GXe and remains modest in LXe; afterpulsing is faster than in conventional PMTs, raising potential signal-overlap concerns. In a separate XAMS TPC setup, the segmented 2×2 anode readout enabled lateral position reconstruction, validating the sensor’s suitability for future LXe detectors with enhanced event localization. Overall, the R12699-406-M4 shows promising performance—high gain, low background, and position sensitivity—in cryogenic xenon, supporting its use in next-generation dark matter experiments and informing further radiopurity and readout optimizations.

Abstract

The Hamamatsu R12699-406-M2 is a $2\times2$ multi-anode 2-inch photomultiplier tube that offers a compact form factor, low intrinsic radioactivity, and high photocathode coverage. These characteristics make it a promising candidate for next-generation xenon-based direct detection dark matter experiments, such as XLZD and PandaX-xT. We present a detailed characterization of this photosensor operated in cold xenon environments, focusing on its single photoelectron response, dark count rate, light emission, and afterpulsing behavior. The device demonstrated a gain exceeding $2\cdot 10^6$ at the nominal voltage of -1.0 kV, along with a low dark count rate of $(0.4\pm0.2)\;\text{Hz/cm}^2$. Due to the compact design, afterpulses exhibited short delay times, resulting in some cases in an overlap with the light-induced signal. To evaluate its applicability in a realistic detector environment, two R12699-406-M2 units were deployed in a small-scale dual-phase xenon time projection chamber. The segmented $2\times2$ anode structure enabled lateral position reconstruction using a single photomultiplier tube, highlighting the potential of the sensor for effective event localization in future detectors.

Figures (14)

• Figure 1: Hamamatsu R12699-406-M4 PMT structure. (a) Electrode structure and electron trajectories in a multianode metal channel dynode PMT. Figure provided by hamamatsuphotonicsk.k.PhotomultiplierTubesBasics2017. (b) Photograph of a Hamamatsu R12699-406-M4 PMT with the first dynode stage visible. The gray crosshair-like feature mitigates charge-up of the photocathode.
• Figure 2: Holder setup for the simultaneous characterization of four Hamamatsu R12699-406-M4 PMTs in MarmotX. Two PMTs each are paired with facing windows at a 3mm distance. (a) Photograph of the assembled setup during preparation for the third data-taking campaign. (b) CAD cutaway rendering of the assembly inside the double-walled cryostat. The blue light is illustrated by the blue hue surrounding the LED.
• Figure 3: (a) Example of an analytical model fit diwanStatisticsChargeSpectrum2020 to a charge spectrum of PMT MA0058 at room temperature and nominal bias voltage of -1000kV. Gaussian components are shown as solid lines, exponentially modified Gaussian contributions as dashed lines. The fit yields a gain of $\mu\approx 2.9\cdot10^6$, occupancy of $\lambda\approx 0.13$ PE/trigger, relative exponentially modified Gaussian contribution $w \approx 3\;\%$, and SPE resolution of $\sigma/\mu\approx 36\;\%$. The fit also gives a peak-to-valley ratio of approximately 2.5. The residuals to the fit, in units of one standard deviation, are shown in the bottom plot. (b) Room-temperature charge spectra of PMT MA0055, for various bias voltages under constant illumination. Markers indicate gain estimates from the model-independent approach.
• Figure 4: R12699-406-M4 PMT gain dependence on the bias voltage and environment. (a) Gain as a function of high voltage fitted with a power law \ref{['eq:gain_hv_power_law']}. (b) Gain measurements across different environments. In Run 2 (squares), expected gain increases upon cool-down were not observed due to systematic bias introduced by LED pulser switching noise.
• Figure 5: Gain evolution during long-term stability tests in LXe (a) Run 1, (b) Run 2, and (c) Run 3. Median gain values for each PMT are indicated by a dotted line. Calibrations recorded within the shaded area in (c) had to be omitted due to a malfunction of the DAQ.