🔬 optics

Mid-infrared photon counting and resolving via efficient frequency upconversion

This paper demonstrates a highly efficient mid-infrared frequency upconversion detector that achieves 37% overall detection efficiency and, for the first time, enables photon-number-resolving detection up to nine photons at 3 $\mu$ m, thereby opening new avenues for trace spectroscopy and sensitive sensing applications.

Original authors: Kun Huang, Yinqi Wang, Jianan Fang, Weiyan Kang, Ying Sun, Heping Zeng

Published 2026-06-04

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Kun Huang, Yinqi Wang, Jianan Fang, Weiyan Kang, Ying Sun, Heping Zeng

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

Imagine you are trying to listen to a very faint whisper coming from a distant room, but the room is filled with a deafening roar of traffic. Furthermore, your ears are only tuned to hear high-pitched squeaks, not the deep, low rumble of the whisper. This is the challenge scientists face when trying to detect "mid-infrared" light. This type of light is invisible to our eyes and carries a lot of useful information about molecules and gases, but it is very hard to catch with standard sensors, which usually struggle with noise and lack the sensitivity to count individual "packets" of light (photons).

In this paper, a team of researchers from China built a clever "translator" to solve this problem. Here is how they did it, using simple analogies:

1. The Problem: The Wrong Language

Think of mid-infrared light as a person speaking a rare, low-frequency language. The best microphones (detectors) we have are tuned to high-frequency languages (visible light). Trying to record the low-frequency speaker directly with a high-frequency microphone results in static, poor quality, and an inability to hear single words.

2. The Solution: The Frequency Translator

The researchers built a device that acts like a simultaneous interpreter. Instead of trying to build a new microphone for the low-frequency language, they translate the low-frequency whispers into high-frequency squeaks that their existing, super-sensitive microphones can hear perfectly.

The Process: They take the invisible mid-infrared light and mix it with a powerful, synchronized laser beam (the "pump").
The Magic: Through a special crystal (like a mixing bowl for light), the two beams combine to create a new beam. This new beam has a higher frequency, shifting the invisible mid-infrared light into the visible spectrum (specifically, a color we can see).
The Efficiency: Usually, these translators lose a lot of the original message during the switch. However, this team tuned their "mixing" so perfectly that 80% of the original light made it through the translation without being lost. It's like having a translator who repeats 8 out of every 10 words perfectly, rather than just 3.

3. The Result: Counting the Whispers

Once the light is translated into the visible spectrum, they use a standard, high-tech silicon detector (like a super-sensitive camera sensor) to count the photons.

Extreme Sensitivity: Because the translation was so efficient and they filtered out the "traffic noise" (background interference) so well, their system is incredibly quiet. They achieved a sensitivity level that is four times better than the best standard detectors and ten times better than the most advanced, expensive super-cooled detectors currently available.
The "Pixel" Trick: To count how many photons are in a single pulse (not just "is there light?" but "how many?"), they used a special sensor called a Multi-Pixel Photon Counter (MPPC). Imagine a large room with 3,600 tiny tiles (pixels). When a photon hits the room, it randomly lands on one tile. If two photons arrive, they might land on two different tiles. By counting how many tiles light up, the system can tell if 1, 2, or even 9 photons arrived at once.

4. Why This Matters (According to the Paper)

The paper claims this breakthrough allows for:

Counting single photons in the mid-infrared range, which was previously very difficult.
Resolving photon numbers (telling the difference between 1 photon and 9 photons) for the first time at this specific wavelength.
Applications: The authors specifically mention this could help in:
- Trace molecule spectroscopy: Detecting tiny amounts of specific chemicals.
- Sensitive biochemical sensing: Finding biological markers with high precision.
- Free-space communications: Sending data through the air with high security and sensitivity.
- LIDAR: Using light to measure distance with extreme precision.

Summary

Think of this research as upgrading a radio. Instead of struggling to tune into a static-filled, low-frequency station, they built a device that instantly converts that station into a crystal-clear, high-frequency signal that any standard radio can play perfectly. This allows them to hear the faintest whispers of light that were previously impossible to count or distinguish.

Technical Summary: Mid-infrared Photon Counting and Resolving via Efficient Frequency Upconversion

Problem Statement
The mid-infrared (MIR) spectral region (specifically around 3 µm) is critical for applications ranging from environmental monitoring and molecular spectroscopy to free-space communications. However, achieving single-photon sensitivity in this region remains a significant challenge. Conventional semiconductor detectors (e.g., InSb, MCT) typically exhibit noise-equivalent powers (NEP) limited to the pW/Hz $^{1/2}$ range, far above the single-photon level. While superconducting nanowire detectors offer single-photon sensitivity, they require bulky cryogenic cooling systems and often suffer from reduced detection efficiency (approx. 4%) at MIR wavelengths.

Existing frequency upconversion strategies, which translate MIR signals to visible wavelengths for detection by efficient silicon-based detectors, have historically been limited by low conversion efficiencies (typically below 35%) and severe pump-induced noise. These limitations have prevented the realization of photon-number-resolving (PNR) detection in the MIR, a capability readily available in the visible and near-infrared regimes using silicon multi-pixel photon counters (Si-MPPCs).

Methodology
The authors implemented a high-performance MIR frequency upconversion detector based on a coincidence-pumping configuration and spectro-temporal engineering. The experimental setup comprised three main stages:

Source Preparation and Calibration: A passively synchronized fiber laser system generated dual-color pulses at 1030 nm and 1550 nm. A coherent down-converter (CDC) utilizing a periodically-poled lithium niobate (PPLN) crystal generated a MIR signal at 3069 nm from the 1550 nm seed and 1030 nm pump. This setup allowed for precise calibration of MIR power at the single-photon level by attenuating the near-infrared input, leveraging the superior sensitivity of visible/NIR detectors.
Frequency Upconversion: The MIR signal was spectrally translated to the visible band (771 nm) via sum-frequency generation (SFG) in a second PPLN crystal, pumped by the 1030 nm pulse. The system utilized tight passive synchronization and optimized the spectral and temporal overlap of the signal and pump pulses.
Detection and Filtering: The upconverted visible signal passed through a rigorous filtering stage (short-pass and band-pass filters) to suppress pump-induced parametric fluorescence. The filtered light was detected using two configurations: a silicon avalanche photodiode (Si-APD) for counting and a silicon multi-pixel photon counter (Si-MPPC) for PNR detection.

Key Contributions and Results

High Conversion Efficiency: Through spectro-temporal optimization, the system achieved an intrinsic conversion efficiency of 80% for 3 µm photons. This was validated by measuring the undepletion rate of the MIR signal, confirming the efficiency without relying on cross-wavelength power calibration.
Record Detection Performance: Combined with effective filtering and a high-efficiency Si-APD, the system achieved an overall detection efficiency of 37%. The noise equivalent power (NEP) was measured at 1.8×10 $^{-17}$ W/Hz $^{1/2}$ , representing the highest sensitivity reported for MIR detectors to date. This performance is at least four orders of magnitude better than conventional thermal detectors and significantly superior to state-of-the-art superconducting nanowire detectors in terms of both efficiency and noise.
First Demonstration of MIR PNR: The authors demonstrated, for the first time to their knowledge, photon-number-resolving detection at MIR wavelengths using a Si-MPPC. The system successfully resolved photon numbers up to 9 within a single pulse.
Statistical Fidelity: The detector faithfully reconstructed the Poissonian statistics of MIR coherent states. The system operated at a high repetition rate of 14.6 MHz with a noise probability per pulse of only 0.14% at peak efficiency.

Significance
The paper claims that these achievements open new possibilities for applications requiring both single-photon sensitivity and a large dynamic range with linear response. Specifically, the authors highlight the potential for:

Trace molecule spectroscopy.
Sensitive biochemical sensing.
Free-space communications.
Advanced protocols such as thresholded laser ranging and loss-tolerant optical communication.

By bridging the gap between the high performance of silicon-based visible detectors and the unique capabilities of the MIR region, this work provides a robust platform for future classical and quantum optical technologies operating at room temperature.