UV holographic microscopy offers a different way to approach a specialized imaging problem: identifying protein crystals and distinguishing them from salt crystals. Instead of relying on a conventional lens-based microscope, the demonstrated system records light-interference patterns on a nearby image sensor and processes them computationally. The result is a portable imaging platform built from a UV LED, a CMOS sensor and a single-board computer—an engineering approach of interest wherever compact instrumentation and image processing can replace some traditional optical hardware.
Why distinguish protein crystals from salt?
During protein crystallization, salt crystals can form alongside the desired protein crystals. Telling the two apart helps crystallographers identify samples suitable for further analysis. Conventional systems commonly combine bright-field microscopy with UV-induced fluorescence. Under UV illumination, most protein crystals absorb light and emit fluorescence, associated with tryptophan residues, whereas most salt crystals do not show the same response.
These microscopes can provide useful imaging, but lens-based UV systems require optics designed for ultraviolet wavelengths, as well as UV illumination and suitable cameras. The equipment can therefore be bulky and expensive. Like other lens-based microscopes, the approach also involves a trade-off between field of view and resolution: increasing one can limit the other.
How UV holographic microscopy works
The alternative uses on-chip holographic imaging. A 280 nm UV LED illuminates a crystal sample positioned approximately 300–400 micrometres from the active area of a decapped CMOS image sensor. Light scattered by the crystals interferes with the background illumination. The sensor captures the resulting in-line hologram, which contains information about the sample even though the system has no imaging lens.
A computer then digitally back-propagates the recorded hologram to reconstruct an image. In this case, processing uses the angular spectrum approach to model the propagation of light. The reconstruction reveals amplitude contrast: protein crystals appear darker than the background because they absorb UV light strongly, while the tested salt crystals show no comparable, statistically significant contrast.
This is an important distinction in how the instrument generates an image. Rather than forming a conventional image directly through a lens, the sensor records an interference pattern and software calculates a view of the sample from that data. The optical layout is simplified, but image formation depends on both the sensor capture and the reconstruction process.
What the demonstrated platform includes
The reported prototype combines a 280 nm UV LED and band-pass filter with a consumer-grade CMOS sensor connected to a Raspberry Pi single-board computer. The filter narrows the LED output around the target wavelength. The sample sits close to the sensor, allowing the system to use the sensor’s active area as a field of view greater than 10 mm2.
Because the sensor is intended for visible-light imaging, the researchers adapted its raw data for UV capture: red and blue pixel values were replaced with averages from neighbouring green pixels, which were more responsive to the UV illumination. The holograms were then processed to produce amplitude reconstructions. The reported processing took about one minute on a standard desktop computer, so the demonstrated workflow was not presented as real-time imaging.
Sample preparation also matters. The crystal chambers used UV-transparent fused silica and ACLAR crystallization covers. Materials in the optical path must suit coherent imaging: irregularities can introduce background distortions. In the tests, ACLAR produced only faint background modulation that did not affect image quality.
Testing the contrast approach
The platform was evaluated with several protein crystals, including proteinase K, a RING1B complex and maltose-binding protein, alongside salts such as sodium chloride, ammonium sulfate, lithium acetate and lithium sulfate. In reconstructed images, the protein crystals appeared darker than the background, while the salt crystals did not show the same contrast. Mixed samples containing protein and salt crystals were also imaged within the same field of view.
For comparison, samples were imaged using a conventional dual-mode UV microscope. The fluorescence response and the holographic amplitude contrast provided complementary ways to distinguish the tested crystal types. The reported contrast differences were statistically significant in the evaluated experiments. The authors also note that phase-retrieval methods could further reduce twin-image artefacts in holographic reconstructions, though doing so would require additional computation or measurements.
Engineering considerations for compact imaging
For engineers evaluating computational imaging, the work demonstrates a broader design principle: a sensing task may not require a complete conventional optical train if the sample, illumination, detector and processing are designed together. Removing imaging lenses can support a smaller, simpler platform, but it shifts some of the design work into sensor selection, illumination control, sample-to-sensor spacing, material compatibility and reconstruction software.
- Illumination: The demonstrated contrast depends on 280 nm UV light and a band-pass filter that blocks LED side-band emissions.
- Sample positioning: The short distance between sample and sensor is central to capturing useful in-line holograms.
- Materials: UV transmission and low optical irregularity help avoid unwanted background distortion.
- Processing: Digital back-propagation extracts sample information, with more advanced reconstruction potentially requiring added computation or measurements.
These details are specific to the reported protein-crystal application; the system should not be treated as a general-purpose industrial inspection instrument. Its significance lies in showing how a narrowly defined imaging requirement can be addressed with accessible components and computational reconstruction rather than bulky UV optics.
Practical takeaway
UV holographic microscopy distinguished the tested protein crystals from salt crystals by converting UV absorption into amplitude contrast in digitally reconstructed images. For crystallography, the approach offers a portable alternative to conventional lens-based UV microscopy, while retaining important dependencies on UV-compatible materials, close sample placement and post-capture processing. More broadly, it provides a useful example for technical decision-makers considering compact imaging architectures: simplifying optics can be effective when the illumination, sensor and computation are engineered as one system.