Can we take 3D pictures with single mid-infrared photons?
Converting mid-infrared echoes into visible light with a femtosecond laser gate lets an ordinary silicon camera build detailed 3D images from extremely few photons.
Source
Mid-infrared single-photon 3D imaging
Study at a glance
- Design
- Other — Time-of-flight imaging: 3070 nm pulses illuminate a scene; returning photons are upconverted to 771 nm in a chirped-poled lithium niobate crystal by a delayed femtosecond pump and recorded on an EMCCD
- N
- No sample count; test objects were a coin under a silicon wafer, stacked silicon wafers and a ceramic goldfish
- Population
- Test objects imaged with a lab-built mid-infrared upconversion imaging system
- Outcome
- Depth and lateral resolution, volumetric reconstruction of layered samples, and image recovery at very low signal-to-noise
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The researchers illuminated objects with ultrashort mid-infrared pulses and collected the reflected light in a nonlinear crystal where a synchronized femtosecond pump pulse converted it to near-visible light only when both pulses overlapped in time. Scanning the pump delay sliced the scene by depth, and a megapixel silicon camera recorded each slice. They imaged a coin through a thick silicon wafer, two stacked wafers, and a ceramic goldfish at very low light, using a neighbour-product denoising algorithm for the dimmest data.
What they found
The system resolved a height difference of 30 micrometres on the coin through the wafer, with a lateral resolution of about 60 micrometres over a field roughly 2.5 cm wide. It located every internal interface of the stacked wafers, allowing the refractive index to be inferred from the axial offset. With the denoiser, the goldfish's 3D shape was recovered at a detected signal of only 0.05 photons per pixel per second, far below the background noise.
The limits
What it doesn't show
This is a technique demonstration on a few static, hand-picked objects, not a systematic comparison with mid-infrared OCT on the same samples. Low-light imaging needed long exposures (seconds per frame across hundreds of depth slices), so it is slow for moving scenes. The denoiser assumes smooth, spatially correlated surfaces and could suppress real fine or isolated features; performance on scattering biological tissue was not tested.
Key terms
- Sum-frequency upconversion
- A nonlinear optical process that combines an infrared photon with a pump photon to produce a higher-energy photon that silicon detectors can see.
- Optical gating
- Using a short pump pulse so conversion only happens when signal and pump overlap in time, selecting photons from a narrow depth slice.
- Time-of-flight imaging
- Measuring distance from the time light takes to travel to an object and back.
- Quasi-phase matching
- Periodically reversing a crystal's poling so a nonlinear process stays efficient; a chirped period accepts many angles for wide-field imaging.
- EMCCD
- An electron-multiplying camera sensitive enough to register single photons.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What converts the reflected mid-infrared light into light a silicon camera can detect?
Common questions
Why not just use a mid-infrared camera?
Direct mid-infrared detectors are noisy, often need cryogenic cooling and are slow; converting the light lets a low-noise, fast silicon camera do the detection.
What sets the depth resolution?
The width of the cross-correlation between the pump and signal pulses, not the camera's speed.
Why use mid-infrared at all?
It passes through materials like silicon and scatters less, and molecules have distinctive absorption bands there.
More on Nonlinear optics