Characterization of Spurious Charge in SENSEI Skipper-CCDs

This paper characterizes spurious charge in SENSEI Skipper-CCDs, identifying the serial register as the dominant background source during readout and demonstrating that a novel "tri-level" clocking scheme reduces single-electron density by a factor of approximately seven.

The Gist

Imagine you are trying to listen to a single, tiny whisper in a very quiet library. This is what scientists are doing when they use special cameras called Skipper-CCDs to hunt for dark matter or rare neutrino interactions. These cameras are so sensitive they can count individual electrons, like counting grains of sand one by one.

However, there's a problem. Even in a super-quiet library, sometimes the floorboards creak, or a book falls over. In these cameras, these "creaks" are called spurious charge. They are fake signals that look exactly like the tiny whispers the scientists are looking for, but they are actually just noise generated by the camera itself.

Here is what this paper discovered and fixed, explained simply:

1. The Problem: The Camera's Own "Static"

The camera works by moving packets of electrons (the signal) from one pixel to the next, like a bucket brigade passing water down a line. To move the water, the camera uses electrical "clocks" that push and pull the buckets.

The scientists found that the main source of noise wasn't coming from the outside world or the main part of the camera where the image is taken. Instead, the noise was coming from the serial register—think of this as the "conveyor belt" that carries the water buckets to the exit to be counted.

The specific culprit: When the camera stops moving the buckets to count them (a process called "Skipper readout"), it holds the electrical clocks at a steady, low voltage. During this pause, tiny trapped electrons at the edge of the conveyor belt get released and accidentally create new electrons. It's like if, while you were holding a bucket still to measure the water, the bucket itself started leaking or generating new water out of nowhere.

2. The Investigation: Cleaning the Pipes

Before measuring the noise, the scientists had to "clean" the camera. They found that how they cleaned the camera mattered a lot.

• The Old Way: They used a "full purge," which is like flushing the entire system with water to clear out debris.
• The New Discovery: They found that if they only flushed the vertical pipes (leaving the horizontal ones alone), they could remove a specific type of debris that was causing massive noise in the main image area. However, this trick didn't help much with the noise on the conveyor belt during the actual reading process.

3. The Solution: The "Tri-Level" Trick

The scientists realized the noise happened because the conveyor belt was held at a very deep "valley" (low voltage) while it waited to be counted. The trapped electrons were happy to sit there, but when they were released, they caused a splash (noise).

The Fix: They invented a new way to operate the conveyor belt called "Tri-level Clocking."

• Normal Mode: The bucket sits in a deep valley (Low Voltage).
• The Fix: While the camera is counting the water, they gently raise the floor of the valley to a "middle height" (Intermediate Voltage).

The Analogy: Imagine holding a ball in a deep hole. If you let go, it might roll out and cause a mess. But if you raise the bottom of the hole so the ball is just sitting on a flat surface, it's much less likely to roll around and cause trouble. By raising the voltage slightly during the counting phase, they stopped the trapped electrons from causing that splash.

4. The Result: A Quieter Library

By using this "Tri-level" trick, the scientists reduced the fake noise on the conveyor belt by a factor of 7.

• Before: About 29 fake electrons per million pixels.
• After: Only about 4 fake electrons per million pixels.

Summary

This paper is about tuning a super-sensitive camera to be even quieter. They discovered that the camera was making its own noise while it was "pausing" to count the signal. By slightly adjusting the electrical settings during that pause (the Tri-level clocking), they successfully silenced the noise, making the camera much better at hearing the faint whispers of the universe's most elusive particles.

Technical Summary

Technical Summary: Characterization of Spurious Charge in SENSEI Skipper-CCDs

Problem Statement
Skipper Charge-Coupled Devices (Skipper-CCDs) are a leading technology for sub-GeV dark matter searches and coherent elastic neutrino-nucleus scattering due to their single-charge resolution and low background. However, the sensitivity of these detectors is currently limited by low-energy background events, specifically single-electron events and their pileup. While exposure-dependent backgrounds (e.g., environmental radiation, dark current) have been significantly reduced, exposure-independent single-electron events remain a dominant limitation. A primary source of these exposure-independent events is "spurious charge" (also known as clock-induced charge), which is generated during the charge transfer operations required for Skipper readout. Understanding the origin and mitigation of this charge in both the active region and the serial register is critical for improving the sensitivity of future rare-event searches.

Methodology
The authors characterize spurious charge generation in SENSEI Skipper-CCDs using a "pumping" technique. This involves repeatedly clocking the device to generate a controlled amount of spurious charge, allowing the rate to be extracted by measuring the linear dependence of the total charge on the number of transfers or pumps.

The study utilizes two detector configurations:

1. Surface Setup: A C-module Skipper-CCD housed in a copper tray at a Fermilab surface facility, allowing for flexible modification of clock waveforms and RC filter time constants.
2. Underground Setup: SENSEI@MINOS detectors (using both C-modules and "skinny" modules) located ~100 m underground in the MINOS cavern, surrounded by lead shielding to minimize cosmic ray backgrounds.

The investigation focuses on two specific regions:

• Active Region Spurious Charge (ARSC): Generated during vertical transfers.
• Serial Register Spurious Charge (SRSC): Generated during horizontal transfers and the subsequent Skipper readout phase.

The authors test different initialization procedures, specifically comparing the standard "all-clocks purge" against a "vertical-only purge" (where only vertical clocks are lowered during the electron purge phase). Furthermore, they analyze the dependence of SRSC on clock swing voltage, delay times between transfers, and the duration of the "held-low" phase during Skipper readout.

Key Contributions and Findings

• Mechanism of Generation: The paper supports a model where spurious charge is generated by electrons trapped at the Si–SiO$_2$ interface near channel stops. When gate voltages are high, electrons are attracted to the interface and captured by defects. When voltages drop, these trapped electrons are released and accelerated by potential gradients, inducing impact ionization that generates holes (signal charge).
• Active Region Characterization:
  • The study demonstrates that the initialization procedure significantly impacts ARSC. A "vertical-only purge" (lowering only vertical clocks during the purge phase) drastically reduces ARSC compared to the standard "all-clocks purge" when using large vertical clock swings (10.25 V). This is hypothesized to be due to the draining of electrons from parallel channel stops via a parasitic static induction transistor effect.
  • However, under standard SENSEI operating conditions (2.25 V vertical clock swing), the vertical-only purge provides negligible reduction in ARSC.
  • The measured ARSC rate under standard conditions is $(2.3 \pm 1.1) \times 10^{-8}$ e$^-$/pixel/transfer, which extrapolates to a small contribution to the total single-electron density.
• Serial Register Characterization:
  • The dominant source of SRSC is identified as the parallel channel stops adjacent to the serial register, rather than the serial register channel stop itself. This is evidenced by the fact that SRSC is significantly reduced in the non-prescan region (adjacent to parallel stops) but not the prescan region when using a vertical-only purge with high clock swings.
  • Timing Dependence: The study reveals that the majority of SRSC is not generated during the pixel transfer itself, but during the "dwell time" when the horizontal clocks are held constant at a low voltage while the Skipper amplifier performs non-destructive measurements. The SRSC rate scales logarithmically with the delay time between transfers.
  • At standard SENSEI readout conditions (2.25 V swing), approximately 99% of the SRSC is generated during the Skipper readout dwell time, not the transfer time.

Tri-Level Clocking Scheme
Motivated by the finding that SRSC is generated during the "held-low" phase of Skipper readout, the authors propose and implement a "tri-level" clocking scheme.

• Mechanism: During the pixel transfer, the horizontal clocks operate with the standard full swing ($V_H - V_L$). However, once the charge packet is transferred and Skipper readout begins, the voltage of the "held-low" gate (H2) is raised to an intermediate voltage ($V_M$). This reduces the depth of the potential well, thereby suppressing the electric field that accelerates released trapped electrons and induces impact ionization.
• Optimization: The scheme was optimized to minimize SRSC without introducing significant Charge Transfer Inefficiency (CTI). An intermediate voltage resulting in a potential well depth of 1.2 V was found to be optimal.

Results
Using the SENSEI@MINOS detector with a C-module, the authors measured the single-electron density in the serial register under two conditions:

1. Standard Readout: $(2.9 \pm 0.1) \times 10^{-5}$ e$^-$/pixel/image.
2. Tri-Level Clocking: $(4.0 \pm 0.4) \times 10^{-6}$ e$^-$/pixel/image.

This represents a reduction in serial register single-electron density by a factor of approximately 7.

Significance
The paper concludes that in well-shielded, low-background environments, the dominant contribution to exposure-independent single-electron events in Skipper-CCDs originates in the serial register during the Skipper readout phase. The proposed tri-level clocking scheme offers a promising, hardware-independent method to significantly lower these backgrounds. This improvement is directly applicable to current and future Skipper-CCD experiments, potentially enhancing their sensitivity to rare events such as sub-GeV dark matter interactions and coherent neutrino scattering. Future work will focus on quantifying the relative contributions of active region and serial register spurious charge to the total background and exploring these techniques with next-generation multi-amplifier sensors.