Prospects for Direct Electron Detectors in Ultrafast Electron Diffraction and Scattering Experiments

This paper demonstrates that hybrid pixel counting detectors, while promising for ultrafast electron diffraction due to their low noise and high frame rates, suffer from severe count losses in pulsed beam experiments that saturate at approximately two electrons per pixel per pulse, necessitating new data handling strategies and detector adaptations to optimize signal-to-noise ratios.

The Gist

Imagine you are trying to listen to a very quiet whisper (the subtle changes in a molecule's structure) happening inside a room that is occasionally filled with a loud, booming voice (the intense laser and electron beams). To hear the whisper, you need a microphone that is incredibly sensitive but doesn't get overwhelmed by the boom.

This paper is about testing a new type of "microphone" for a scientific experiment called Ultrafast Electron Diffraction (UED). In these experiments, scientists shoot tiny, super-fast packets of electrons at a material to see how its atoms are moving. They use a special camera (a Direct Electron Detector) to catch the electrons after they bounce off the material.

Here is the breakdown of what the researchers found, using simple analogies:

1. The Problem: The "Busy Cashier"

The new cameras they tested are called Hybrid Pixel Counting Detectors (HPCDs). Think of each pixel on the camera as a tiny, super-fast cashier at a grocery store.

• How they usually work: When an electron hits a pixel, the cashier counts "1" and moves on. They are amazing because they don't make mistakes (no "noise") and can count very accurately when the line is short.
• The Ultrafast Twist: In these experiments, the electrons don't arrive one by one over a long time. Instead, they arrive in a massive, instantaneous "shower" all at once (like a sudden crowd rushing the checkout counter).
• The Glitch: Because the electrons arrive in a tiny fraction of a second, the "cashier" gets overwhelmed. If two electrons hit at the exact same moment, the cashier only sees one and misses the other. If too many hit, the cashier gets "paralyzed" and stops counting entirely. This is called saturation.

2. The Failed Fix: The "Speedy Re-Trigger"

The camera manufacturers had a feature called "Retrigger Mode."

• The Idea: Imagine the cashier has a special trick. If the line gets too long, they try to estimate how many people are in the crowd by looking at how long the line stays busy, rather than counting each person individually.
• The Reality: The researchers found this trick doesn't work for these ultrafast electron showers. Instead of helping, it made the data messy and full of random errors. It's like the cashier getting so confused by the crowd that they start shouting random numbers.
• The Verdict: They had to turn this feature off and stick to the standard "counting" mode.

3. The Smart Solution: The "Empty Bin" Trick

Since the cashier (pixel) can't count high numbers accurately when the crowd is huge, the researchers invented a clever math trick called P0 Counting.

• The Analogy: Imagine you have a bucket that can only hold one ball. If you throw 100 balls at it, it will just overflow, and you won't know if you threw 10 or 100.
• The Trick: Instead of trying to count how many balls hit the bucket, you count how many times the bucket was completely empty after you threw the balls.
  • If the bucket is empty often, you know you didn't throw many balls.
  • If the bucket is never empty, you know you threw a lot.
• The Result: By counting the "zeros" (the empty buckets), they can use a simple math formula to figure out the average number of electrons that actually hit, even if the camera got saturated. This allowed them to measure much brighter signals than the camera was supposed to handle.

4. The Noise Issue: The "Flickering Lightbulb"

Even with a perfect camera, the experiment has another problem: the source of the electrons isn't perfectly steady. It's like a lightbulb that flickers slightly, making the whole room brighter or dimmer randomly.

• The Question: Does it matter if we take a photo of the flickering light every single millisecond (Shot-to-Shot) or if we just take one long, blurry photo (Integration)?
• The Discovery: They found that it doesn't matter. Whether you try to correct the flicker every single millisecond or just average it out over a minute, the final picture quality (Signal-to-Noise Ratio) is exactly the same.
• Why this is good news: It means scientists don't need to store terabytes of data for every single millisecond. They can just take longer "exposures" and save massive amounts of computer space without losing any quality.

5. The Big Picture: What This Means for Science

• For Weak Signals: These cameras are fantastic for listening to the "whispers" (weak signals like phonon scattering) because they are so quiet and sensitive.
• For Strong Signals: They struggle with the "booms" (bright Bragg peaks in single crystals) because the pixels get saturated.
• The Future: The researchers suggest that for the future, we need cameras that can count multiple electrons at once (like a cashier who can count a whole crowd at once) rather than just counting "1 or 0."

In summary: The team discovered that while these new high-tech cameras are amazing, they get confused by the speed of ultrafast electron pulses. By turning off a "helpful" feature and using a clever "count the emptiness" math trick, they can still get great data. They also proved that scientists can save time and storage space by not over-analyzing every single millisecond of data.

Technical Summary

1. Problem Statement

Ultrafast Electron Diffraction and Scattering (UED(S)) experiments aim to resolve dynamic structural changes in molecules and materials with sub-50 femtosecond resolution. A critical challenge in these experiments is achieving high sensitivity to small changes in electron scattering intensity (often <1%), particularly for weak signals like phonon-diffuse scattering.

While Hybrid Pixel Counting Detectors (HPCDs) (e.g., Dectris Quadro) offer near-zero readout noise and dark counts—making them ideal for low-count regimes—they face severe limitations in ultrafast pulsed beam applications:

• Pulse Pile-up: Unlike continuous wave (CW) beams where electrons arrive randomly, ultrafast experiments deliver electrons in ultrashort pulses (<10 ps). If multiple electrons hit a single pixel within one pulse, the detector's dead time (100 ns) prevents the registration of all counts.
• Saturation: Standard HPCDs saturate at very low doses (approx. 1 electron per pixel per pulse), leading to significant count losses and distorted data.
• Retrigger Mode Failure: Manufacturers offer "instant retrigger" modes to mitigate pile-up in CW beams by measuring signal duration. However, it was unknown if this works for ultrashort pulses, and whether it introduces artifacts.

The paper addresses the lack of understanding regarding how HPCDs perform under these specific ultrashort pulse conditions and seeks to establish robust data handling strategies to maximize Signal-to-Noise Ratio (SNR).

2. Methodology

The authors utilized a 90 keV Radio-Frequency (RF) compressed ultrafast electron scattering instrument at McGill University, coupled with a Dectris Quadro HPCD.

• Experimental Setup:

  • Source: 260 nm UV pulses generate electron bunches (0 to $10^7$ electrons) with pulse durations <10 ps at repetition rates up to 1 kHz.
  • Samples: Polycrystalline monoclinic $VO_2$ and a 10 nm-thick polycrystalline Platinum (Pt) film on a SiN membrane.
  • Modes: The detector was tested in both Normal Mode (binary 0/1 counting) and Retrigger Mode.

• Data Analysis Strategies:

  • P0 Counting Method: A statistical estimator derived from the fraction of pixels registering zero counts ($P_0$). Since the detector cannot distinguish between 1 and $k$ electrons (where $k>1$), the authors use the Poisson distribution relationship $\lambda = -\ln(P_0)$ to estimate the true mean electron dose ($\lambda$) per pixel per pulse, correcting for saturation.
  • Noise Characterization: The team measured the Power Spectral Density (PSD) of both the laser pump and electron source to quantify source noise (bunch charge fluctuations) versus counting noise.
  • Normalization: They compared various data normalization strategies, ranging from shot-to-shot (normalizing each single pulse) to long-time integration (normalizing after averaging many pulses), to determine the optimal approach for SNR.

3. Key Contributions

1. Characterization of HPCD Limitations in Pulsed Regimes: The study definitively shows that HPCDs saturate at $\approx$2 electrons per pixel per pulse in ultrafast experiments. Crucially, it demonstrates that the Retrigger Mode is ineffective for ultrashort pulses, introducing massive overcounting artifacts and random noise, rendering it unsuitable for UED(S).
2. Development of the P0 Counting Method: The authors introduce and validate a robust statistical method to recover accurate electron dose estimates from binary (0/1) data. This extends the effective dynamic range of the detector by more than an order of magnitude (up to $\approx$7 EPP) while maintaining quantitative accuracy.
3. Noise Budget Analysis: The paper provides a complete model for measurement uncertainties, separating shot noise (counting statistics) from source noise (bunch charge fluctuations). They quantify that source noise contributes significantly (~40%) to total variance even at moderate doses.
4. Normalization Strategy Optimization: The authors prove that for UED(S) experiments, shot-to-shot normalization offers no significant SNR advantage over longer integration windows, provided the source noise is multiplicative and stationary. This allows for data integration before readout, drastically reducing storage requirements without sacrificing data quality.

4. Key Results

• Detector Performance:
  • In Normal Mode, the detector behaves as a binary counter. At low doses (<1 EPP), it matches Poisson statistics. At high doses, it saturates, but the $P_0$ method accurately recovers the true mean intensity.
  • In Retrigger Mode, the detector fails catastrophically for ultrashort pulses. At high fluxes, it registers unphysical values (e.g., counts >100) and fails to register zero counts, introducing random noise.
• Dynamic Range Extension: Using the $P_0$ method, the relative uncertainty in dose estimation remains below 3% up to 7 EPP, whereas simple summation diverges rapidly due to saturation.
• Noise Sources:
  • Source Noise: The electron source exhibits excess noise ($\approx$0.65% relative intensity noise) beyond the fundamental shot noise limit, likely due to mechanical/magnetic instabilities and laser-to-UV conversion noise.
  • Normalization Impact: Simulations and experiments on Pt films showed that SNR is effectively independent of the normalization time window (from 1 ms to 1 min). Shot-to-shot normalization does not improve SNR over integrating multiple shots before normalization.
• Validation: The theoretical uncertainty model (combining $P_0$ statistics and source noise) was validated against experimental data from $VO_2$ pump-probe experiments. The model accurately predicted the experimental standard error across the entire scattering vector range.

5. Significance and Implications

• Operational Guidelines: The paper provides a critical "user manual" for HPCDs in ultrafast science. It advises against using retrigger modes and recommends the $P_0$ counting method for any experiment involving high electron doses or single crystals.
• Data Efficiency: By demonstrating that integration before normalization is valid, the authors suggest that routine UED(S) experiments can integrate multiple shots into a single frame. This reduces data storage needs by orders of magnitude (from terabytes to gigabytes) without compromising SNR.
• Future Detector Development: The results highlight the need for next-generation detectors with charge-integrating ASICs rather than threshold-based counting. Such detectors would be able to resolve multiple electrons within a single ultrashort pulse, overcoming the current saturation limit and enabling high-sensitivity studies of single-crystal Bragg peaks.
• Sensitivity Limits: The work clarifies that while HPCDs are excellent for weak signals (phonon-diffuse scattering) and polycrystalline samples, they currently impose a strict limit on the electron bunch charge usable for single-crystal studies due to the saturation of Bragg peaks.

In summary, this paper bridges the gap between detector hardware capabilities and the specific demands of ultrafast electron scattering, providing the statistical tools and operational strategies necessary to extract maximum sensitivity from current HPCD technology while outlining the path for future hardware improvements.