Germanium Detector Resolves Individual Electron-Hole Pairs

Conventional high-purity germanium detectors provide excellent gamma-ray spectroscopy, but electronic readout noise has held low-energy thresholds above roughly 30 electron-hole pairs, or about 100 eV. That is a serious constraint for low-mass dark matter searches and coherent neutrino scattering, where the relevant deposits can lie below 100 eV. The authors report a cryogenic Ge device that measures ionization calorimetrically through the phonons produced as charge carriers drift in an applied field, reaching single-electron-hole-pair resolution.

Core Contribution

The central design choice is to combine a point-contact-inspired electrode geometry with a NbSi transition-edge sensor (TES). Rather than reading charge directly with conventional electronics, the detector amplifies the energy released during carrier drift through the Neganov–Trofimov–Luke effect and collects the resulting athermal phonons in the TES. This shifts the low-energy limit toward the Ge band-gap scale rather than the noise floor of a charge preamplifier.

The geometry is also intended to preserve information that ordinary single-channel phonon measurements lose. The TES is most sensitive to NTL phonons generated in the high-field region near the point contact, while a Ge NTD thermistor measures a broader phonon response. Comparing those channels gives an independent estimate of ionization efficiency, opening a route to distinguish electron recoils, nuclear recoils, and non-ionizing events.

Technical Approach

The device uses an aluminum electrode over the bottom, lateral, and part of the top Ge surfaces, connected to a field-shaping ring around the TES by a 16-µm-wide line. Simulations at $V_{NTL}=100$ V place more than 80% of the potential drop within 1 mm of the TES; peak electric fields exceed $10^5$ V/cm. The authors’ argument is practical: concentrating the voltage drop where the TES has high athermal-phonon sensitivity makes charge drift visible as a localized calorimetric signal.

Figure 1 shows both the device layout and the simulated field concentration beneath the TES. The simulation establishes the intended collection geometry, but the experimental case rests on the resolved laser and background spectra rather than on field modeling alone.

At 50 V NTL bias, laser-induced signals form discrete Gaussian-like peaks corresponding to integer electron-hole-pair populations. The paper also uses the 10.37 keV K-shell, plus L- and M-shell, lines from activated $^{71}$Ge to extend the TES calibration beyond the low-energy linear regime. For low-energy events, the TES signal is treated as an ionization probe through $E_{TES}\approx\epsilon_{NTL}\,Q\,V_{NTL}$, where $Q$ is the number of collected pairs.

Results and Analysis

The reported spectra show individual 1-, 2-, and 3-pair events in raw TES data at 50 V, and the 50 V laser spectrum resolves peaks from 0 to 5 electron-hole excitations. That is the key experimental result: the detector is not merely extrapolated to single-pair sensitivity from a noise estimate; a one-pair peak is observed under laser calibration.

The dual-channel dataset is more consequential for low-background use. In 11 h of data at 70 V, triggered offline with the NTD-Ge channel, the authors observe a defined electron-recoil band containing $^{71}$Ge M- and L-shell lines, a non-ionizing population near the horizontal axis, and an intermediate population attributed to charge trapping. Gamma events have a TES-inferred-ionization-to-NTD signal ratio near 1/3, while non-ionizing events lie near zero. Figure 4 makes the discrimination claim credible at the demonstrated exposure, although it is not yet a nuclear-recoil calibration.

The evidence supports single-pair detection and separation of identifiable background populations, but it does not establish a deployed sub-100 eV nuclear-recoil search. Charge collection also degrades during a 6-hour, 50 V run: the mean generated pairs per laser pulse fell from $\mu=1.72$ to $\mu=1.36$ in the final hour. The authors attribute this to space-charge accumulation and regenerate the detector after several hours, a cycle lasting about 30 minutes. This operational instability is a material constraint for long exposures, even though the high-field region near the TES appears comparatively resilient.