🔬 applied physics

Limits of Trap-assisted Photomultiplication Gain

This paper demonstrates that while trap-assisted photomultiplication can achieve high apparent quantum efficiencies, its inherent self-limiting nonlinearity and associated fluctuation penalties prevent it from surpassing the intrinsic thermodynamic noise limit of the underlying unity-gain photodiode, rendering single-chord gain comparisons misleading.

Original authors: Ardalan Armin

Published 2026-06-10

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Ardalan Armin

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Magic" Light Detector That Isn't Magic

Imagine you have a light detector (like a camera sensor) that is supposed to count photons (particles of light). In a normal, high-quality detector, one photon hits the sensor, and it releases exactly one electron. You get a 1-to-1 match. This is the best you can theoretically do without "cheating."

However, scientists have been making special detectors (using materials like organic plastics or perovskites) that claim to do something amazing: One photon hits, and the detector spits out 100, 1,000, or even 100,000 electrons. This is called "photomultiplication." It sounds like a superpower that makes the detector incredibly sensitive.

The paper's main argument is: While these detectors do produce a huge electrical signal, they don't actually make the detector fundamentally more sensitive to weak light. In fact, trying to use this "gain" might actually make the measurement noisier and less reliable than a standard detector.

Here is how the paper breaks this down:

1. The Trap and the Gatekeeper (How the Gain Works)

Think of the detector as a factory with a heavy gate at the entrance.

The Trap: When a photon hits the factory, it gets stuck in a "trap" (a hole) near the gate.
The Gatekeeper: This trapped charge acts like a magnet that lowers the heavy gate, making it easy for other workers (electrons/holes) to rush in from the outside.
The Rush: Because the gate is lowered, a flood of workers rushes in. One trapped photon causes a massive flow of current. This is the "gain."

The Catch: The paper explains that this system is self-limiting. The very act of letting the flood of workers in also fills up the trap or knocks the gate back up, eventually stopping the flow. The more you try to amplify the signal, the faster the system "relaxes" or resets. It's like trying to hold a door open with a spring; the harder you push, the harder the spring pushes back.

2. The "Chord" vs. The "Tangent" (Why the Numbers Lie)

Scientists often report a "Gain" number by taking the total current and dividing it by the total light. The paper calls this the "Chord Gain."

The Analogy: Imagine you are driving a car up a very steep, curvy hill. If you measure the average steepness from the bottom to the top (the Chord), it looks like a moderate slope. But if you look at the road right in front of your tires (the Tangent), it might be a vertical cliff.
The Problem: These detectors are incredibly non-linear. Their behavior changes wildly depending on how bright the light is. At very low light, they might look super-sensitive. At slightly higher light, they might look normal or even less sensitive.
The Conclusion: Comparing the "Chord Gain" of two different papers is like comparing the average speed of a car in traffic to the top speed of a race car on a track. They are measuring different things. The paper argues that the only number that matters is the "Tangential Gain" (how the device reacts to a tiny change in light at a specific moment), which is much harder to measure and often much lower than the flashy numbers reported.

3. The Noise Penalty (The Cost of Amplification)

In electronics, when you amplify a signal, you usually amplify the noise (static) along with it.

The Paper's Insight: In these specific detectors, the "gain" isn't free. It requires a constant flow of electricity (bias) to keep the gate lowered. This constant flow creates its own unique type of "static" or noise.
The Thermodynamic Limit: The paper uses physics (thermodynamics) to prove that you cannot create information out of nothing. The detector counts photons first, then amplifies the result. If the initial count was noisy, amplifying it just makes the noise louder.
The Result: You can use this gain to overcome the noise of your readout electronics (like a weak microphone), but you cannot use it to beat the fundamental noise of the light detection itself. The "noise floor" (the quietest signal you can hear) cannot be lowered below the limit of a standard detector, no matter how much you amplify.

4. The "Sweet Spot" (Optimal Operating Point)

The paper calculates that there is a specific "sweet spot" where these detectors work best.

If the light is too dim, the gain mechanism doesn't kick in enough.
If the light is too bright, the system saturates (the gate closes), and the gain drops.
The Trade-off: There is a specific point where the signal is strong enough to be useful, but the noise hasn't gotten out of hand yet. However, even at this perfect point, the detector cannot be more sensitive than a standard detector that doesn't use this trick.

Summary of the Paper's Claims

The "Super-Sensitivity" is an Illusion: The massive "Quantum Efficiency" numbers (like 10,000%) reported in other studies are often misleading because they are measured at specific, non-linear points and don't reflect how the device behaves in real-world, changing light.
Non-Linearity is the Norm: These devices are inherently unpredictable. They don't have a straight line between light in and current out. They can be super-linear, linear, or sub-linear depending on the conditions.
You Can't Cheat Physics: Internal gain acts like a driven amplifier. It can help you hear a signal over a noisy amplifier, but it cannot make the signal itself clearer than the original physics allows. It adds its own noise penalty.
Better Measurement Needed: To truly know how good these detectors are, scientists must measure the noise and the signal at the exact same time and light level. Comparing a signal measured in the light against noise measured in the dark gives a fake, overly optimistic result.

In short: These devices are clever tricks that can boost a signal, but they come with a heavy price tag of noise and non-linearity. They don't break the fundamental laws of physics to create "super-sensitivity," and the way they are currently being compared in scientific literature is often comparing apples to oranges.

Technical Summary: Limits of Trap-assisted Photomultiplication Gain

Problem Statement
Trap-assisted photomultiplication (PM) in photodiodes, particularly those based on organic semiconductors, halide perovskites, and nanocrystals, can exhibit apparent external quantum efficiencies (EQE) far exceeding unity ( $10^2$ – $10^5$ %). However, the literature suffers from three critical issues regarding the characterization and interpretation of these devices:

Misreporting of Specific Detectivity ( $D^*$ ): Standard $D^*$ calculations assume a linear relationship between input power and output current and rely on shot-noise formulas derived for equilibrium conditions. In PM photodiodes, which operate in a driven, nonequilibrium regime, both responsivity and noise are intensity-dependent. Extrapolating values measured at high illumination to low-light conditions, or using dark noise to normalize illuminated signals, leads to physically misleading and systematically overestimated sensitivity metrics.
Inadequacy of Chord Gain: The "chord gain" (ratio of total current to total photon rate at a single intensity) is frequently used as a figure of merit. However, because these devices are inherently nonlinear, the chord gain differs from the tangential (differential) gain. Comparing chord gains across different studies conflates devices operating in qualitatively different regimes (superlinear, linear, or sublinear).
Conceptual Ambiguity on Sensitivity Limits: It remains unclear whether internal gain fundamentally enhances the intrinsic detectivity of a photodiode or merely redistributes signal and noise as a driven stochastic amplifier. The role of dark current and the thermodynamic cost of sustaining the gain state are not fully integrated into existing sensitivity models.

Methodology
The authors employ a minimal analytical framework adapted from the classical theory of barrier photoconductivity (Petritz, Rose). The model is built on a single gain-active trapped-charge state that couples photogenerated carriers to contact injection under reverse bias.

Core Mechanism: A single trapped charge electrostatically modulates the injection barrier for the opposite carrier type, enabling amplified current flow. The injection current depends exponentially on trap occupancy ( $I \propto e^{\beta n}$ ).
Dynamics: The model treats trap occupancy and injection as coupled stochastic processes. It accounts for photoinduced filling, thermal emission, and injection-assisted relaxation (where injected carriers recombine with the trap, relaxing the gain state).
Generalization: While the core model uses a single trap level, the authors generalize the framework to include an exponential distribution of trap states (energetic disorder) and bimolecular recombination to capture realistic current-intensity characteristics.
Noise Analysis: A Langevin approach is used to treat fluctuations in trap filling, thermal emission, and injection-assisted relaxation as independent Poisson processes. The analysis derives the power spectral density (PSD) of current noise and the local, small-signal noise-equivalent power (NEP).

Key Contributions and Results

Self-Limiting Nonlinearity: The gain mechanism is intrinsically self-limiting. The injection process that amplifies the current simultaneously accelerates the relaxation of the gain-enabling state. This produces an inherently nonlinear, operating-point-dependent response.
- Depending on the distribution of trap energies and recombination order (monomolecular vs. bimolecular), the device can exhibit superlinear, linear, or strongly sublinear responses.
- The "chord gain" is not a meaningful device descriptor; only the tangential (differential) gain at a specific operating point governs sensitivity.
Noise and Fluctuation Penalty: Internal gain introduces a strictly non-negative fluctuation penalty derived from the dissipative dynamics sustaining the gain state.
- The total noise includes contributions from photogeneration, thermal emission, and a gain-specific injection shot noise term arising from the bias-sustained carrier flux.
- Unlike ideal noiseless amplifiers, trap-assisted gain is a driven stochastic process. The injection shot noise cannot be eliminated without eliminating the gain itself.
Local Detectivity ( $D^*_{loc}$ ) and Thermodynamic Limits:
- A single global $D^*$ is ill-defined for these devices. The appropriate metric is a local, small-signal detectivity evaluated at the specific operating point where both signal and noise are measured.
- The local detectivity exhibits a finite optimum at a specific gain-state parameter ( $x^* > 3$ ), where injection-assisted relaxation exceeds thermal emission.
- Fundamental Limit: The paper proves that internal gain cannot improve the intrinsic detectivity of the underlying unity-gain photodiode. Gain acts downstream of the primary photon-to-charge conversion and cannot alter the statistics of the photoevents. While gain can suppress downstream readout noise (improving system-level SNR in readout-noise-limited systems), it cannot reduce the fundamental noise floor set by the primary detection process. In fact, for physically realizable devices, the input-referred NEP of a PM photodiode is strictly greater than or equal to that of the unity-gain device ( $D^*_{gain} \leq D^*_{intrinsic}$ ).
Reinterpretation of Experimental Data: The paper argues that reported high $D^*$ values in PM photodiodes often result from inconsistent measurement protocols (e.g., referencing illuminated signals to dark noise). A self-consistent measurement requires simultaneous noise and responsivity evaluation at the same illuminated operating point.

Significance
The paper provides a unified theoretical treatment connecting gain, noise, nonlinearity, and bandwidth in trap-assisted photodiodes. Its primary significance lies in establishing a thermodynamic bound on sensitivity: internal gain is equivalent to driven stochastic amplification. It offers a practical means of suppressing readout noise but does not enhance the intrinsic sensitivity of the photodetection process.

The work corrects the community's understanding of how to characterize these devices, emphasizing that:

Chord gains are not transferable figures of merit.
$D^*$ must be reported as a local, small-signal quantity at a defined operating point.
The pursuit of higher gain does not inherently lead to better intrinsic detectivity; rather, optimizing sensitivity requires minimizing dark injection currents (contact engineering) and operating at the specific point where the trade-off between responsivity and injection noise is optimized.

The authors conclude that the "gain" in these devices is not a route to surpassing fundamental limits but a mechanism that redistributes signal and noise within a dissipative system, with an irreducible fluctuation penalty that grows with the gain itself.