🔬 optics

Mid-infrared temporal ghost imaging via two-photon structured encoding

This paper presents a compact, room-temperature mid-infrared temporal ghost imaging system that utilizes non-degenerate two-photon absorption in silicon to achieve ultrafast, high-sensitivity signal reconstruction across a 2.5–3.8 μm bandwidth, overcoming traditional limitations of phase-matching and alignment while enabling applications in spectroscopy, ranging, and free-space communication.

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

Published 2026-05-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ziyu He, Kun Huang, Huijie Ma, Wen Zhang, 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 Idea: Seeing the Invisible with a "Shadow" Trick

Imagine you are trying to take a photo of a hummingbird's wings, which are moving so fast that your camera's shutter is too slow to capture them. The image would just look like a blur.

In the world of light, scientists face a similar problem with Mid-Infrared (MIR) light. This type of light is invisible to our eyes and is crucial for things like detecting chemicals or seeing through fog. However, the electronic "cameras" (detectors) we have for this light are too slow to catch fast-changing signals. It's like trying to film a race car with a camera that only takes one picture every ten seconds.

This paper introduces a clever trick called Temporal Ghost Imaging (TGI). Instead of trying to take a fast picture directly, the researchers use a "shadow" technique to reconstruct the fast-moving signal using slow equipment.

The Problem with Old Methods

Previously, to see these fast MIR signals, scientists had to use complex setups involving special crystals and precise laser alignment.

The Analogy: Think of this like trying to translate a book from one language to another using a very strict, expensive dictionary that only works if you hold the book at a perfect angle under a specific light. If you move the book even a millimeter, the translation fails. It was fragile, hard to set up, and didn't work for all types of MIR light.

The New Solution: The "Silicon Detective"

The researchers developed a new system that is simpler, more robust, and works across a wide range of colors (wavelengths) in the infrared spectrum. Here is how it works, step-by-step:

1. The Two-Photon "Handshake"

The system uses two beams of light:

Beam A (The Signal): The fast, invisible Mid-Infrared light that carries the information we want to see.
Beam B (The Gating Light): A slower, visible Near-Infrared laser that acts like a "flashlight" or a "gatekeeper."

The magic happens inside a standard silicon detector (the kind used in many everyday electronics). Normally, silicon ignores Mid-Infrared light. But, if you shine the "Gatekeeper" laser and the "Signal" light onto the silicon at the exact same time, they can join forces.

The Analogy: Imagine a heavy door (the silicon detector) that is locked. The Mid-Infrared light is a person trying to push the door open, but they aren't strong enough. The Near-Infrared light is a second person. Alone, neither can open the door. But if they push at the exact same moment, their combined strength opens the door. This is called Non-Degenerate Two-Photon Absorption (ND-TPA).

2. The "Shadow" Coding

Instead of just shining the lasers on, the researchers "code" the Gatekeeper laser with a pattern of on/off flashes (using a Walsh-Hadamard matrix, which is just a fancy way of saying a specific, organized sequence of patterns).

The Analogy: Imagine you are trying to figure out the shape of a hidden object in a dark room. You don't have a fast camera. Instead, you shine a flashlight on the object, but you cover the flashlight with a stencil that has a specific pattern of holes (like a QR code). You take a photo of the shadow the object casts. Then, you change the stencil to a different pattern and take another photo. By comparing all the shadows to the patterns you used, a computer can mathematically reconstruct the shape of the hidden object, even though your camera was too slow to see the object directly.

3. The Result

Because the silicon detector only "opens the door" when both lights are present, the detector effectively records the pattern of the Gatekeeper laser filtered by the shape of the invisible Signal light. By running this process many times with different patterns, the computer can reconstruct the fast-moving MIR signal with incredible detail.

What They Achieved

The paper claims several impressive results with this new "Silicon Detective" system:

Super-Speed: They could reconstruct signals that were 40 times faster than the detector's normal speed limit. It's like a camera that usually takes 1 frame per second suddenly being able to capture 40 frames per second.
Extreme Sensitivity: They could detect incredibly faint signals, as small as 0.05 picojoules per pulse. This is like hearing a whisper in a hurricane.
Efficiency: They could figure out the signal using 80% fewer measurements than usual, thanks to a technique called "compressed sensing." It's like solving a puzzle by looking at only a few key pieces instead of the whole box.
Broad Range: The system worked perfectly across a wide range of infrared colors (from 2.5 to 3.8 micrometers) without needing to be adjusted or realigned. It's a "one-size-fits-all" solution for this part of the spectrum.
Simplicity: Unlike old methods, this system doesn't need expensive crystals or precise alignment. It works at room temperature and is compact.

Why This Matters (According to the Paper)

The authors state that this method opens the door for:

Time-resolved molecular spectroscopy: Watching how molecules move and react in real-time.
High-precision infrared ranging: Measuring distances with extreme accuracy.
High-speed free-space communication: Sending data through the air using infrared light at very high speeds.

In short, the researchers found a way to use a simple, cheap silicon detector to "see" ultra-fast, invisible light by using a clever coding trick and a two-light handshake, bypassing the need for slow, expensive, and fragile equipment.

Technical Summary: Mid-infrared temporal ghost imaging via two-photon structured encoding

Problem Statement
Temporal ghost imaging (TGI) is a technique capable of reconstructing ultrafast temporal signals beyond the electronic bandwidth limits of conventional photodetectors by correlating structured illumination with integrated detection. While TGI has been successfully demonstrated in the visible and near-infrared (NIR) regimes, extending this capability to the mid-infrared (MIR, 2–20 µm) remains challenging. The MIR region is critical for molecular fingerprinting and applications in chemical sensing and biomedical diagnostics, yet it lacks high-speed modulators and efficient detectors. Existing MIR TGI approaches rely on nonlinear frequency conversion (e.g., difference-frequency generation or second-harmonic generation) in external crystals. These methods are constrained by stringent phase-matching conditions, limited spectral coverage, and the need for complex optical alignment. Furthermore, separating the modulation and detection stages into different components hinders system compactness and scalability.

Methodology
The authors propose and demonstrate a broadband MIR TGI system based on non-degenerate two-photon absorption (ND-TPA) within a silicon photodiode. The system operates on the principle that a temporally encoded NIR pump beam can transfer structured modulation onto a MIR signal directly at the detector surface, bypassing the need for external nonlinear crystals.

The experimental setup utilizes two temporally synchronized beams derived from a common 1030 nm Yb-doped fiber laser:

NIR Pump: A continuous-wave 1550 nm laser is temporally modulated by an electro-optic modulator (EOM) using preprogrammed Walsh-Hadamard sequences. This seed is then amplified via optical parametric amplification in a periodically poled lithium niobate (PPLN) crystal to generate a structured NIR pulse train.
MIR Signal: A tunable MIR source (2.5–3.8 µm) is generated by a fiber-feedback optical parametric oscillator (FOPO). The MIR beam is modulated by an acousto-optic modulator (AOM) to emulate the temporal object to be reconstructed.

Both beams are spatially and temporally overlapped onto a silicon photodiode (Si-PD). Through the ND-TPA process, the simultaneous absorption of an NIR photon and a MIR photon generates a photocurrent proportional to the product of their instantaneous intensities ( $I_{ND-TPA} \propto I_{NIR}(t) \cdot I_{MIR}(t)$ ). This interaction effectively gates the MIR signal with the NIR pattern. The integrated detector output is recorded and correlated with the known encoding patterns to reconstruct the MIR temporal waveform. The system also employs a silicon single-photon counting module (SPCM) for low-light operation.

Key Contributions

Direct Modulation-Detection Integration: The work establishes a unified platform where the silicon detector serves simultaneously as the nonlinear medium for modulation transfer and the sensing element. This eliminates the need for external nonlinear crystals, phase matching, and complex alignment.
Broadband Operation: By leveraging the intrinsic broadband nature of ND-TPA in silicon, the system operates across a wide spectral range (2.5–3.8 µm) without requiring adjustments to the optical configuration or pump wavelength.
High-Fidelity Reconstruction: The system achieves temporal reconstructions that exceed the detector's electronic bandwidth by more than fortyfold.
Compressive Sensing: The implementation supports sub-Nyquist sampling, allowing for accurate waveform recovery with 80% fewer measurements compared to traditional Nyquist-rate acquisition.
High Sensitivity: The system demonstrates detection sensitivity down to 0.05 pJ per pulse in photon-counting mode.

Results

Linearity and Fidelity: Characterization confirmed a bilinear response of the ND-TPA signal to both NIR and MIR intensities, validating the linearity required for high-fidelity reconstruction. The system successfully reconstructed binary temporal objects at modulation rates of 0.1, 1, 2, and 4 Mbps, maintaining accuracy even when the signal frequency far exceeded the 100 kHz bandwidth of the silicon detector.
Compressive Imaging: Using the TVAL3 algorithm, the authors demonstrated that high structural similarity (SSIM > 80%) could be achieved with a sampling ratio as low as 20%, significantly reducing acquisition time.
Photon-Limited Performance: In single-photon counting mode, the system achieved reliable reconstruction at pulse energies as low as 0.05 pJ. The results showed that reconstruction fidelity is governed by the effective photon budget, with higher photon flux suppressing shot noise.
Temporal Resolution: The system successfully reconstructed temporal features with a resolution of 0.25 µs, surpassing the timing resolution limits of conventional direct photon-counting detection by encoding temporal information into statistical correlations rather than relying on electronic timing precision.
Spectral Flexibility: High-fidelity reconstructions were achieved at MIR wavelengths of 2864, 3130, 3430, and 3680 nm, confirming stable operation across the tested band.

Significance
The paper claims that this ND-TPA-enabled MIR TGI system provides a practical, compact, and alignment-free route for fast and sensitive time-domain analysis in the mid-infrared. By removing the constraints of phase matching and external crystals, the approach offers a scalable architecture for MIR detection. The authors position this technology as a promising foundation for applications in time-resolved molecular spectroscopy, high-precision infrared ranging, and high-speed free-space communication. The work underscores the potential of concurrent modulation-detection operations to overcome the bandwidth and sensitivity limitations inherent to conventional MIR detection technologies.