A preprint reports a laboratory system that formed wide-field images in the mid- to long-wave infrared without directly detecting the infrared beam. Using an AgGaS2 crystal and non-collinear phase matching, the experiment reported imaging across 6–10 µm.
The method is called quantum imaging with undetected photons, or QIUP. In the setup, the infrared idler beam passed through the object without direct idler detection, while the signal beam was registered by a CMOS camera. The study asks whether visible-range interference can carry information about an infrared object.
The evidence comes from a benchtop optical demonstration using a thin metal-foil shadow mask and a knife-edge resolution test. The reported figures therefore describe the behavior of that arrangement with those test materials.
Reading the infrared with a visible camera
The optical layout separates the beam that probes the object from the beam the camera reads. In practical terms, the infrared path carries the object interaction, while the signal path carries the interference pattern recorded by the CMOS camera. Image information is reconstructed without direct idler-camera readout.
To build the image, phase-scanned camera images supplied pixelwise fringe-visibility data. Those data were used to reconstruct the object’s complex transmittance, including phase—the component of the optical signal that tracks the wave’s position in its cycle.
The object in the demonstrations was a thin 0.3 mm metal-foil shadow mask. Images were shown at 6, 8 and 10 µm, allowing the mask to be displayed at several infrared wavelength settings.
Resolution was estimated with a knife-edge test. The visibility curves were fitted along concentric arcs with an error function, giving a way to assess how sharply the edge was reproduced at different positions in the image.
A broad image, with conditions attached
At 8 µm, the reported resolution was 297 ± 5 µm. The estimated number of resolvable elements was 8,000 ± 100—roughly, the number of separate details the image could distinguish. The preprint reports the ± values but does not specify how they were derived.
Resolution was also reported as uniform across the image plane when the imaging wavelength range was sufficiently narrow-band. In that operating condition, the analysis reported no radial resolution variation.
The field of view—the image area covered by the measurement—was approximately 700 mm² using a 90% intensity-level definition. The figure is therefore tied to a specific threshold: it describes the area meeting that intensity criterion, not an unqualified boundary of the system.
Together, the reported spatial results describe both local sharpness and overall coverage: the resolution test supplied the detail scale, while the field-of-view calculation supplied the area. The paper presents both under the stated narrow-band and intensity-level conditions.
Faster acquisition came with a trade-off
Speed depended on the acquisition mode. Detailed scans used up to 180 frames at 10 seconds per frame. The preprint also reported single-frame off-axis holography at 10 seconds rather than 30 minutes, with reduced image quality.
Those are not interchangeable measurements. The longer, multi-frame approach was used for detailed scans, while the single-frame approach shortened the reported acquisition at the cost of image quality. The paper describes that loss but does not quantify it.
The infrared power in the idler channel was estimated at 0.2 pW in total, corresponding to 7×10−18 W per camera pixel. The values are estimates, and the supplied analysis reports no uncertainty for them.
For context, the paper compared the result with a BLIP benchmark, a way of describing background-limited detector performance. It reported a noise-equivalent power of approximately 2×10−15 W Hz−1/2 and an estimated blackbody flux density of approximately 2.1×10−10 W. The paper reported QIUP levels at least 2 orders below the BLIP limit.
That comparison should be read as a benchmark calculation under similar parameters. It is not evidence that every scene will produce the same separation from background limits, and the supplied analysis reports no uncertainty for the comparison.
What the experiment does not yet show
The test materials put a clear limit on the claim. The demonstrations used a thin metal-foil mask, and resolution was checked with a knife edge. The experiment therefore shows image formation and edge resolution for those test conditions; it does not establish performance on other kinds of samples.
The reported spectral result is likewise specific: imaging was shown across 6–10 µm with this AgGaS2, non-collinear arrangement. Broader spectral performance was not established by the demonstration.
The comparison with infrared detector noise was also estimated rather than presented as a direct head-to-head trial. The supplied analysis identifies matched benchmarking against conventional infrared detectors and FTIR as an open question, alongside independent replication across crystals and operating conditions.
On that evidence, the strongest supported conclusion is limited: the system can reconstruct images of the supplied test mask and measure an edge under the reported optical conditions. The experiment does not establish biomedical, materials-science or chemical-sensing utility.
Further work would need to test real samples with molecular fingerprint features, examine whether the reported resolution and resolvable-element count are reproducible across crystals and operating conditions, and compare the method directly with conventional infrared cameras and FTIR on matched samples.
Preprint status and data access
The manuscript is a preprint dated 20 Aug 2026.
It reports support from the UK Quantum Technology Hub in Sensing Imaging and Timing and the National Institute for Health and Care Research. The authors declare no conflicts of interest.
Underlying data are not publicly available at this time, but the data-availability statement says they may be obtained from the authors upon reasonable request.
Paper data and sources
Original title: Wide-field mid- to long-wave infrared imaging with undetected photons
Authors: Vladimir Kornienko, Nathan Gemmell, Caiyi Liu et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text