🔬 optics

Mid-infrared single-photon sub-pixel temporal ghost imaging

This paper presents a high-resolution mid-infrared single-photon temporal ghost imaging system that achieves 40 ps temporal precision at a 3.125 Gbps driving rate by integrating nonlinear structured detection with sub-pixel temporal shifting, thereby overcoming conventional limits imposed by modulation bandwidth and detector jitter.

Original authors: Wen Zhang, Kun Huang, Zhibin Zhao, Huijie Ma, Ziyu He, Jianan Fang, Heping Zeng

Published 2026-05-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wen Zhang, Kun Huang, Zhibin Zhao, Huijie Ma, Ziyu He, Jianan Fang, 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

The Big Problem: The "Slow Camera" Dilemma

Imagine you are trying to take a photo of a hummingbird's wings flapping incredibly fast. If you use a standard camera with a slow shutter speed, you just get a blurry mess. You can't see the details of the wings because the camera is too slow to catch the action.

In the world of light, this is a huge problem for Mid-Infrared (MIR) light. This type of light is special because it carries the "fingerprints" of molecules (useful for sensing gases, heat, or chemicals). However, the cameras (detectors) that can see this light are usually slow, expensive, or need to be kept freezing cold to work. They are like that slow camera: they can't capture fast-moving light signals clearly.

The Solution: "Ghost Imaging" (The Detective's Trick)

Instead of trying to build a faster camera, the researchers used a clever trick called Temporal Ghost Imaging.

Think of it like this: Imagine you are in a dark room trying to figure out what a fast-moving object looks like, but you only have a single, slow light sensor that just tells you "how much light hit me" in total. You can't see the shape.

To solve this, you shine a flashlight at the object, but you flick the flashlight on and off in a specific, random pattern very quickly. Even though your sensor is slow, it records the total amount of light that bounced back during each flicker. By comparing the pattern of the flashlight flickers with the total light recorded, a computer can mathematically reconstruct the shape of the object. It's like solving a puzzle where the pieces are hidden in the shadows.

The Innovation 1: The "Magic Translator" (Nonlinear Detection)

The researchers wanted to use this trick on Mid-Infrared light, but there was a catch: the "flashlight" (the pattern generator) works best with Near-Infrared light, not Mid-Infrared.

They solved this with a Magic Translator. They mixed the slow Mid-Infrared signal with a fast, patterned Near-Infrared laser beam inside a special crystal.

The Analogy: Imagine the Mid-Infrared light is a secret message written in a language no one understands. The Near-Infrared laser is a translator. When they meet in the crystal, the translator converts the secret message into a new language (visible light) that our fast, cheap silicon cameras can understand.
This allows them to use a standard, room-temperature silicon detector (which is fast and cheap) to "see" the Mid-Infrared light.

The Innovation 2: "Sub-Pixel Shifting" (The Micro-Step)

Even with the translator, there was still a limit. The speed at which they could flick the flashlight (the pattern) was limited by their electronics. If they flicked it 3 billion times a second, the best detail they could see was limited to that speed.

The researchers introduced a Sub-Pixel Shifting strategy.

The Analogy: Imagine you are trying to measure the length of a table using a ruler that only has marks every 10 centimeters. You can only guess the length to the nearest 10cm.
Now, imagine you take a measurement, then slide the ruler over by just 1 millimeter (a tiny fraction of a mark) and take another measurement. Then you slide it again and take a third.
By combining all these slightly shifted measurements, you can figure out the table's length with millimeter precision, even though your ruler only had 10cm marks.

In this experiment, they didn't just take one picture; they took many pictures, each time shifting the "flashlight pattern" by a tiny fraction of a time-step. By stitching these together, they achieved a resolution of 40 picoseconds (that's 40 trillionths of a second).

The Results: Seeing the Invisible

The team proved this works by:

Creating a fast signal: They made a Mid-Infrared signal that changed incredibly fast (up to 12.5 billion times a second).
Using a slow detector: They used a standard silicon detector that is usually too slow for this job.
The Outcome: Even though the detector was slow and the electronics driving the pattern were limited to 3.125 billion operations per second, the "sub-pixel shifting" trick allowed them to reconstruct the signal with a precision of 40 picoseconds.

This is a big deal because:

It is faster than the "jitter" (wobble) of the single-photon detectors they used.
It is faster than the speed of the electronics they used to create the patterns.
It works even when the light is so dim that only one photon (a single particle of light) is hitting the detector at a time.

Summary

The paper describes a new way to take "photos" of ultra-fast Mid-Infrared light without needing expensive, super-fast cameras. They did this by:

Translating the invisible light into visible light using a crystal.
Flickering a laser in a pattern to act as a "ghost" camera.
Shifting that pattern by tiny fractions of a second to create a super-detailed image, bypassing the speed limits of their hardware.

This allows scientists to see ultra-fast molecular events and signals with incredible clarity, using equipment that is simpler and more sensitive than what was previously possible.

Technical Summary: Mid-infrared Single-Photon Sub-pixel Temporal Ghost Imaging

Problem Statement
The mid-infrared (MIR) band is critical for applications ranging from molecular spectroscopy to free-space optical communication due to its unique molecular fingerprint signatures and atmospheric transparency. However, high-speed, high-sensitivity MIR detection remains a significant challenge. Conventional MIR detectors (e.g., HgCdTe, InSb) are limited by long carrier lifetimes and thermal noise, restricting their bandwidth to below the GHz level and degrading room-temperature sensitivity. While advanced devices like superconducting nanowire single-photon detectors (SNSPDs) offer single-photon sensitivity, they often require cryogenic operation or suffer from limited wavelength coverage.

Temporal Ghost Imaging (TGI) has emerged as a solution to bypass the bandwidth limitations of detectors by reconstructing temporal profiles through correlation with pre-encoded patterns. However, existing TGI schemes face two primary bottlenecks:

Resolution Limits: The temporal resolution is fundamentally constrained by either the modulation speed of pre-programmed patterns or the fluctuation scale of random sources.
Source Constraints: Most MIR TGI implementations rely on structured illumination, which requires modulating the MIR source itself. This is impractical when the MIR source cannot be modulated (e.g., natural emissions or ultrafast fluorescence) or when studying intrinsic temporal structures.
Detector Jitter: Even with frequency upconversion, the effective resolution of single-photon TGI is often limited by the timing jitter of the detectors.

Methodology
The authors propose a high-resolution MIR single-photon computational TGI system that integrates nonlinear structured detection with sub-pixel temporal shifting.

Nonlinear Structured Detection: Instead of modulating the MIR source, the system employs a structured detection approach. A pre-programmed near-infrared (NIR) pump beam acts as a temporally optical gate. This pump is combined with the MIR signal (at 3.4 µm) in a periodically poled lithium niobate (PPLN) crystal to drive Sum-Frequency Generation (SFG). The upconverted light (at 0.8 µm) carries the product of the MIR signal and the NIR gate intensity ( $I_{up} \propto I_s \cdot I_p$ ), effectively sampling the MIR waveform. This upconverted signal is captured by a room-temperature silicon detector or a single-photon counting module (SPCM), bypassing the need for fast MIR detectors.
Sub-pixel Temporal Shifting: To overcome the resolution limit imposed by the modulation speed, the authors extend the concept of sub-pixel ghost imaging from the spatial to the temporal domain. Rather than increasing the modulation rate, the pre-programmed NIR pump pattern is shifted relative to the fixed MIR temporal object in fractional increments ( $\Delta T/k$ ).
Reconstruction: By performing $k$ successive measurements with these fractional shifts, the system generates a rectangular measurement matrix. The temporal waveform is reconstructed using a least-squares pseudoinverse algorithm. This process decouples the achievable temporal resolution from the modulation speed, making it dependent instead on the precision of the temporal stepping.

Key Results
The system was experimentally validated under both high-flux and single-photon-level conditions:

High-Flux Performance: For binary temporal objects, the system demonstrated that sub-pixel shifting could effectively double or quadruple the sampling density. A 20 Mbps probe rate successfully reconstructed a 40 Mbps object with a resolution of 12.5 ns, surpassing the Nyquist limit of the modulation rate alone.
Single-Photon Sensitivity: The system operated at an incident flux of 0.02 photons/bit. Using a Geiger-mode SPCM with a 717 ps timing jitter, the system successfully reconstructed MIR temporal objects at rates up to 12.5 Gbps.
Resolution Breakthrough: By employing eight encoding sets with a 40 ps step size, the system achieved an effective temporal resolution of 40 ps. This result is significant because it surpasses both the intrinsic modulation speed limit of the pump pattern (3.125 Gbps) and the timing jitter limit of the single-photon detector (717 ps).

Significance and Claims
The paper claims to establish a versatile route for ultrafast MIR waveform reconstruction that addresses the dual bottlenecks of detector jitter and modulator bandwidth. The primary contributions are:

Decoupling Resolution from Speed: The sub-pixel shifting strategy allows for high temporal precision (40 ps) without requiring high-speed modulators or detectors, shifting the constraint to the precision of delay stepping.
Structured Detection for Unmodulatable Sources: By utilizing nonlinear structured detection rather than structured illumination, the method enables the characterization of MIR sources that cannot be modulated, such as natural emissions or ultrafast fluorescence dynamics.
Room-Temperature Single-Photon MIR Imaging: The integration of frequency upconversion with computational TGI enables high-resolution, single-photon-sensitive MIR detection at room temperature, overcoming the thermal noise and cryogenic requirements of traditional MIR single-photon detectors.

The authors conclude that this paradigm opens new opportunities for high-resolution infrared sensing and quantum photonics, with potential extensions to long-wave infrared and terahertz regimes using suitable nonlinear crystals.