Photon Calibration Techniques for High Resolution Cryogenic Detectors

This paper clarifies the assumptions underlying the standard Poisson-based calibration method for high-resolution cryogenic detectors using monoenergetic photons, analyzes how realistic detector performance violates these assumptions to introduce bias, and evaluates the specific impact of detector parameters on calibration accuracy.

The Gist

The Big Picture: Measuring the Unmeasurable

Imagine you have a super-sensitive scale that can weigh a single grain of sand. Scientists use these "scales" (called cryogenic detectors) to catch tiny particles from space or dark matter. To make sure the scale is accurate, they need to calibrate it.

Usually, they do this by dropping known weights onto the scale. In the world of light, these "weights" are photons (particles of light). If you shine a laser that sends exactly one photon at a time, and the scale reads "1," then two photons read "2," you know your scale is perfect.

The Problem: Many new, high-tech detectors are so sensitive that they can't distinguish between one photon and two. It's like trying to weigh a single grain of sand on a bathroom scale; the needle just wobbles a little bit, and you can't tell if you dropped one grain or two.

Because they can't see the individual "grains," scientists have to use a statistical trick. They shine a light that sends a random number of photons (sometimes 10, sometimes 11, sometimes 12) and look at the average wobble of the needle. They assume the wobble follows a predictable mathematical pattern (like a bell curve) to figure out how much energy one photon actually carries.

The Paper's Discovery: The "Hidden Bias"

The authors of this paper, W. Matava and M.R. Williams, say: "Wait a minute. That statistical trick only works if the scale behaves perfectly."

They argue that in the real world, these detectors are messy. When a photon hits the detector, the energy doesn't always travel the same way to the sensor. Sometimes it gets lost, sometimes it bounces around, and sometimes the sensor reacts differently depending on where the photon hit.

Because of this messiness, the "wobble" (variance) of the needle doesn't match the "average weight" (mean) in the simple way the old math predicts.

The Analogy: The Rainy Day Umbrella Test
Imagine you are trying to measure how much rain falls by holding an umbrella under a sprinkler.

• The Old Method: You assume every drop of water hits the umbrella and falls straight down into a bucket. If you know how many drops the sprinkler tries to shoot, you can calculate how much water is in the bucket.
• The Reality (The Paper's Point): The wind blows some drops away. The umbrella has holes. Sometimes a drop hits the handle and slides off the side. Sometimes it hits the center and goes straight in.
• The Result: If you just count the drops the sprinkler tried to shoot and assume they all made it to the bucket, you will be wrong. You will think the bucket is lighter than it actually is, or that your measuring cup is broken.

The paper calls this error $\delta$ (delta). It's a hidden correction factor that messes up the calibration.

Why Does This Happen?

The authors break down the "messiness" into a few main culprits:

1. The "Lost in Transit" Problem: When a photon hits the detector, it creates a shower of sound waves (called phonons). These waves have to travel through the material to reach the sensor. Some get absorbed by the material itself before they ever reach the sensor.
2. The "Where You Stand" Problem: If a photon hits the sensor right in the middle, it might be very efficient. If it hits near the edge or under a metal wire, it might be very inefficient. If the light source moves around randomly, the detector's efficiency changes randomly.
3. The "Bumpy Road" Problem: Even if the waves make it to the sensor, they might arrive with different amounts of energy, causing the signal to be "noisier" than expected.

What Did They Do?

The authors did two main things:

1. The Math: They wrote new equations that include these messy factors. They showed that if you ignore them, you will underestimate the energy of the particles and think your detector is more precise (sharper) than it really is.
2. The Simulation: They built a computer model to test different scenarios.
   • Scenario A (Good Detectors): If a detector is very well-made (like the older "TES" sensors), the "messiness" is small. The old math is mostly okay, with only a tiny error (less than 10%).
   • Scenario B (Newer Detectors): Newer technologies (like KIDs and qubit sensors) are often less efficient and have more "dead zones" where energy gets lost. For these, the error is huge. Using the old math would give you a completely wrong answer.

The Conclusion: Don't Trust the "Simple" Math

The paper concludes that for the newest, most advanced detectors, the standard way of calibrating them with light is flawed.

• If you use the old method: You might think your detector is seeing a 10 keV particle when it's actually a 12 keV particle. You might think your detector is super sharp when it's actually blurry.
• The Fix: Scientists need to account for the "position dependence" (where the hit happens) and the "collection efficiency" (how much energy actually makes it to the sensor).

The authors suggest that instead of just shining a light and guessing, scientists should either:

1. Use a laser that can be moved to hit specific spots on the detector to map out the "dead zones."
2. Use complex computer simulations to predict exactly how much energy is being lost.

In short: The paper warns scientists that their "ruler" might be bent. If they don't fix the math to account for the bent ruler, their measurements of the universe will be off.

Technical Summary

Technical Summary: Photon Calibration Techniques for High Resolution Cryogenic Detectors

Problem Statement
High-resolution cryogenic detectors, such as Transition Edge Sensors (TESs) and Kinetic Inductance Detectors (KIDs), are increasingly utilized for measuring low-energy phenomena, including dark matter recoils and coherent neutrino-nucleus scattering. Calibrating these detectors typically involves using monoenergetic photons from pulsed laser diodes or LEDs. For detectors with resolutions finer than a single photon, calibration is straightforward: the discrete steps in pulse amplitude corresponding to $N$ and $N+1$ photons can be resolved, allowing for a direct fit of the energy scale.

However, many emerging detector technologies possess resolutions significantly larger than the energy of a single photon. In these cases, individual photon peaks cannot be resolved. Instead, calibration relies on statistical arguments assuming that the variance of the measured signal scales linearly with the mean signal, derived from Poisson statistics of the photon source. This paper identifies a critical flaw in this standard approach: the implicit assumption that detector resolution (noise) is uncorrelated with the number of detected photons. In reality, factors such as position-dependent phonon collection efficiency and the stochastic nature of phonon transport introduce correlations that violate the assumptions of the standard Poisson-based calibration, potentially biasing the derived energy scale and resolution.

Methodology
The authors develop a mathematical framework to model the calibration process, moving from "classical" assumptions to a more realistic "toy model" of athermal phonon sensors.

1. Classical Derivation: The paper first reproduces the standard derivation where the number of emitted photons ($N$) follows Poisson statistics, and the number of absorbed photons ($N'$) follows a binomial distribution conditional on $N$. Assuming a linear detector response and uncorrelated readout noise, the variance of the measured amplitude ($\sigma^2_I$) is shown to be linearly related to the mean amplitude ($\mu_I$) with a slope corresponding to the single-photon response ($I_0$).
2. Realistic Modeling: The authors introduce a more complex model accounting for the physical processes of phonon generation, transport, and collection. This model incorporates:
   • Fano fluctuations: Variance in the number of athermal phonons generated per photon.
   • Position dependence: Variations in phonon collection efficiency ($Q$) based on where the energy is deposited in the substrate.
   • Energy variance: Fluctuations in the energy of individual athermal phonons.
   • Collection efficiencies: Parameters describing the fraction of energy converted to detectable phonons ($f_1$) and the fraction thermalizing in the sensing element ($f_2$).
3. Derivation of Correction Factor: By applying the laws of total expectation and variance to this realistic model, the authors derive a modified relationship between the mean and variance of the signal. They demonstrate that the slope of the variance-vs-mean plot is no longer simply $I_0$, but rather $I_0(1+\delta)$, where $\delta$ is a correction factor dependent on the detector's physical parameters.
4. Simulation and Estimation: The authors perform Monte Carlo simulations and parameter scans to quantify the magnitude of $\delta$. They vary parameters such as mean collection efficiency ($\mu_Q$), the variance of collection efficiency ($\sigma^2_Q$), phonon energy variance, and the Fano factor ($F$) to determine under which conditions $\delta$ becomes significant.

Key Contributions and Results

• Identification of Systematic Bias: The paper establishes that the standard linear fit used for calibration yields an incorrect calibration constant if the correction factor $\delta$ is non-zero. Specifically, assuming the slope equals $I_0$ leads to an underestimation of the energy deposited in any given event and an underestimation of the baseline energy resolution.
• Quantification of $\delta$: Through parameter scans, the authors find that $\delta$ is generally negligible ($<0.1$) for detectors with high energy collection efficiency and low position dependence. However, $\delta$ becomes significant (potentially $>10\%$) for detectors with:
  • Low average phonon collection efficiency.
  • Significant position dependence (variance in collection efficiency across the detector surface).
• Simulation Verification: Monte Carlo simulations of four distinct detector scenarios confirm the analytical predictions. Detectors with low collection efficiency and high position dependence exhibited the largest deviations from the naive calibration slope, aligning with the theoretical value of $\delta$ calculated via Equation 21.
• Non-Poisson Sources: The authors extend their analysis to non-Poisson photon sources (Appendix B), showing that if the source statistics are not quasi-Poisson, the relationship between mean and variance becomes non-linear, further invalidating standard calibration methods. They propose a method to disentangle these effects by scanning attenuation at a fixed source intensity.

Significance and Claims
The paper argues that the systematic bias introduced by $\delta$ has been largely ignored in the literature regarding the calibration of cryogenic particle detectors. The authors claim that:

1. State-of-the-art detectors (e.g., QETs) with high collection efficiency are likely safe to calibrate using traditional methods, with systematic errors likely below 10%. These detectors often possess the resolution to resolve single photons directly, rendering the statistical method less critical for them.
2. Emerging technologies (e.g., KIDs and qubit-based sensors) are more vulnerable. Due to large dead metal areas and unoptimized quasiparticle trapping, these detectors often suffer from low collection efficiency and high position dependence. For these devices, the naive application of Poisson-based calibration may lead to systematic misestimation of both the energy scale and resolution.
3. Interpretation of Results: Energy resolutions measured via photon calibration for these emerging detectors should be interpreted as "optimistic lower bounds" to the true resolutions.
4. Future Directions: The authors suggest that detailed Monte Carlo simulations of phonon dynamics (e.g., using G4CMP) and experimental techniques such as steerable beam sources to map position dependence are necessary to bound or estimate $\delta$ accurately. They note that existing discrepancies in calibration between optical phonons and gamma rays in the BullKID collaboration may be attributable to the position dependence effects described in this work, rather than just Fano fluctuations.

In conclusion, the paper provides a rigorous statistical correction for photon-based calibration, highlighting that detector non-idealities, particularly position-dependent collection efficiency, can introduce significant systematic errors in the calibration of high-resolution cryogenic detectors.