🔬 optics

Principle of Entangled-Photon Thermometry for Ultrafast Laser Processing

This paper introduces a quantum-enhanced thermometry method that utilizes polarization anisotropy in entangled photon pairs and machine learning to enable robust, high-speed, remote temperature diagnostics during ultrafast laser processing.

Original authors: Denys Moskal, Jiri Martan, Simon Kos, Miroslav Holecek, Milan Honner, Vladislav Lang

Published 2026-06-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Denys Moskal, Jiri Martan, Simon Kos, Miroslav Holecek, Milan Honner, Vladislav Lang

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 Problem: Trying to Measure Temperature in a Storm

Imagine you are trying to measure the temperature of a piece of metal that is being hit by a super-fast laser. This process is violent. It creates a cloud of plasma (hot gas), sparks, and intense light.

If you try to measure the heat using standard tools (like infrared cameras), it’s like trying to hear a whisper in a rock concert. The "noise" from the plasma and the changing surface of the metal drowns out the actual temperature signal. Standard methods also struggle because the metal’s surface changes color and texture as it heats up, confusing the sensors.

The Solution: The "Quantum Twin" Trick

The authors propose a clever workaround using quantum entanglement. Think of this as having a pair of "quantum twins"—two photons (particles of light) that are magically linked.

The Twins: You create a pair of photons. Let’s call them Signal and Idler. They are born together and share a secret connection: if one is spinning "up," the other is also spinning "up." If one is "down," the other is "down." They always match.
The Mission: You send the Idler photon toward the hot, messy laser spot. You keep the Signal photon in a safe, quiet, clean laboratory room far away from the laser.
The Interaction: When the Idler photon hits the hot metal, it bounces off. The heat of the metal changes how the photon bounces (specifically, its polarization, or "spin direction").
The Remote Readout: Because the twins are linked, the change in the Idler’s spin instantly tells you something about the Signal photon’s spin. You don’t need to look at the messy laser spot to know what happened. You just look at the clean Signal photon in the safe room.

The Analogy: Imagine you and a friend have magic coins. You stay in a quiet library (the Signal arm), and your friend goes to a noisy construction site (the Idler arm). Every time your friend flips their coin, yours flips the same way. If the construction site is so loud you can’t hear your friend, you don’t need to listen to them. You just look at your own coin in the quiet library. If your coin lands on "Heads," you know your friend’s coin also landed on "Heads." You get the information from the chaos without being exposed to the noise.

The Challenge: The Data is "Glitchy"

Here is the catch: Photons are tiny. You don’t get a smooth, continuous temperature reading. You get a stream of binary data: 1 (photon detected) or 0 (no photon).

It’s like trying to understand the mood of a crowd by counting individual people walking through a turnstile. If you only look at one second, the number might be wildly inaccurate just by chance. This is called "shot noise." If you average over a longer time to smooth it out, you lose the speed—you miss the rapid changes in temperature.

The Fix: Teaching a Computer to See the Pattern

To solve this "glitchy data" problem, the researchers didn’t just use simple math averages. They used a Neural Network (a type of artificial intelligence).

Old Way (Sliding Window): This is like looking at the turnstile data in 40-nanosecond chunks. It’s slow and noisy. The temperature reading jumps around wildly, with errors of up to 200°C.
New Way (Neural Network): The AI looks at the entire stream of 1s and 0s. It has learned to recognize the hidden patterns in the "glitches." It can tell the difference between random noise and a real temperature change.

The Analogy: Imagine the binary data is a Morse code message sent over a very static-filled radio. A human (or simple math) might just hear random clicks. But an AI, trained to understand Morse code, can filter out the static and reconstruct the clear message.

The Results

The paper shows that this new method works much better than the old way:

Less Noise: The AI reduced the temperature error from ~200°C down to ~20°C.
Better Clarity: The "Signal-to-Noise Ratio" improved from 10 dB to 26 dB. In simple terms, the useful signal became much louder than the background static.
Faster Response: The AI could track temperature changes with nanosecond precision, whereas the old method was sluggish and lagged behind.

Important Note

The authors emphasize that this is a simulation, not a physical experiment yet. They used real-world data from previous experiments to train their computer model, but they haven’t built the actual device. However, the simulation proves that the principle works: using quantum twins to measure temperature remotely, combined with AI to clean up the data, is a promising way to see what’s happening inside ultra-fast laser processing without being blinded by the heat and light.

Technical Summary: Principle of Entangled-Photon Thermometry for Ultrafast Laser Processing

Problem Statement
The paper addresses the limitations of current temperature diagnostics in ultrashort laser surface processing (LST). Conventional methods, such as infrared (IR) pyrometry and pump-probe ellipsometry, face significant challenges:

Coupled Dependencies: IR thermal radiation intensity is affected by plasma emission, non-homogeneous temperature distribution, and changes in material emissivity during ablation. Since laser processing modifies optical properties, temperature measurements become coupled to these changing optical properties.
Background Interference: It is difficult to separate IR thermal radiation from plasma glow and scattered light.
Temporal and Spatial Constraints: Pump-probe ellipsometry requires sample movement for time-step measurements and suffers from challenges in separating reflected beams from ablation plasma. Furthermore, classical analysis of single-photon data suffers from strong shot noise or temporal inertia due to the binary, discrete nature of photon detection events.

Methodology
The authors propose a quantum-enhanced approach using polarization-entangled photon pairs generated via spontaneous parametric down-conversion (SPDC) in a type-I configuration. The core principle relies on the strong polarization correlation between the signal and idler photons.

Quantum Setup:
- An entangled pair is generated in the state $|\Psi\rangle = \frac{1}{\sqrt{2}} (|H\rangle_{sg}|H\rangle_{id} + |V\rangle_{sg}|V\rangle_{id})$ .
- The idler photon interacts with the laser-affected material surface. Its reflection is governed by the temperature-dependent complex refractive index ($n + ik$), which modulates the reflectance of p- and s-polarizations.
- The signal photon travels through a remote, undisturbed optical arm.
- A "bucket" detector on the idler side acts as a trigger without resolving polarization. Polarization-resolved detection is performed remotely on the signal side.
Simulation Model:
- The process is simulated using the Qiskit quantum toolkit.
- The time-dependent complex refractive index components, $n(t)$ and $k(t)$ , are modeled based on experimental pump-probe reflectometry data from literature (Winter et al.), interpolated to the 810 nm idler wavelength.
- Temperature is inferred from the polarization anisotropy factor $A(t) = (R_p(t) - R_s(t))/(R_p(t) + R_s(t)$ , derived from coincidence counts of H- and V-polarized signal photons.
Data Processing:
- The output is a binary vector representing H- or V-state detections.
- Two processing methods are compared:
  - Classical Sliding-Window Analysis: Calculates local density of binary states within a temporal window (e.g., 40 ns).
  - Multi-Layer Perceptron (MLP): A neural network regression model that extracts implicit anisotropy information from the photon bitstream to reconstruct temperature dynamics.

Key Contributions

Introduction of a remote, non-invasive thermometry method that decouples the measurement location from the interaction zone, thereby rejecting background noise (plasma, thermal radiation) via coincidence detection.
Demonstration that temperature can be evaluated from relative polarization statistics (H/V ratio) rather than absolute intensity, making the method immune to emissivity drift and surface deformation effects that plague classical pyrometry.
Application of machine learning (MLP) to overcome the limitations of classical sliding-window analysis when processing discrete, noisy single-photon binary data.

Results

Noise Reduction: The MLP-based reconstruction achieved an order-of-magnitude reduction in temperature noise compared to the sliding-window method. The root mean square (RMS) noise was reduced from ~200 °C (sliding-window) to ~20 °C (MLP).
Signal-to-Noise Ratio (SNR): The SNR improved from approximately 10 dB (sliding-window) to 26 dB (MLP).
Temporal Accuracy: The MLP approach improved temporal accuracy to the nanosecond scale, with a temperature rise shift of approximately 8 ns, compared to a 200 ns shift for the 40 ns sliding window.
Robustness: In tests with controlled accidental coincidence noise (up to 95% accidental pairs), the MLP model maintained significantly lower reconstruction errors (RMSE ~300 °C at high noise) compared to the sliding-window method (RMSE ~600 °C at intermediate noise, becoming unstable at high noise).

Significance and Claims
The paper claims that this entanglement-assisted scheme, combined with machine-learning analysis, offers a promising route for undisturbed, background-rejected, high-speed temperature diagnostics in laser–matter interaction studies. The authors emphasize that the method provides a "clear-arm" remote measurement capability, suppressing parasitic effects such as plasma emission and emissivity changes. They note that while the method requires calibration for specific materials, it enables the evaluation of rapid thermal dynamics (e.g., heat accumulation, phase transitions) during femtosecond laser ablation with superior noise performance compared to classical statistical analysis of single-photon events. The work is presented as a simulation-based proof-of-principle, utilizing experimentally validated optical parameters to model realistic material responses.