On several pages of this site, I discuss near-UV (NUV) photography, i.e., photography in the portion of the UV closest to the visible range. There is no "hard" formal definition of the borders of NUV, and in general it is meant as the portion of UV between 400 nm (which is the generally and formally accepted border between the visible and UV ranges) and whatever a consumer digital camera with Bayer sensors can record, once modified by removing its UV-blocking, IR-blocking and antialiasing filter.
With strong broadband UV sources, lenses capable of high UV transmission and cameras particularly suitable to UV photography, it is possible to slightly push the limits of the UV range that can be recorded a little bit beyond what it is reasonable to regard as "near" UV. At this point, it is better to use a more precise definition of the different UV sub-bands. A common and generally accepted subdivision of the UV spectrum is UV-A (400-320 nm), UV-B (320-290 nm) and UV-C (290-200 nm) according to The Free Dictionary. UV further includes wavelengths all the way down to 10 nm (or 4 nm according to some sources), but these are difficult to record even with specialized equipment and, in addition, are quickly absorbed by oxygen and can travel only short distances in the atmosphere. Nitrogen transmits UV down to 125 nm. Below 125 nm, even nitrogen ionizes and quickly absorbs this high-energy UV, and its use is only possible in a vacuum.
For quite a while, there has been some evidence that Bayer sensors are slightly sensitive to UV-B, in addition to UV-A (e.g., see the Photography of the invisible world blog). In addition, Bayer sensors also seem to react in a consistent way to different UV wavelengths, in spite of being produced by multiple companies using different manufacturing processes. In particular, different UV wavelengths are imaged as (false) colors according to the following approximate schema. This schema changes only slightly among different sensors, but is significantly affected by the custom color balance set on a given camera.
The above table contains colors I picked from actual pictures shot with narrow bandpass filters, and is reasonably accurate. However, the perception of these colors may be affected by a variety of factors, including for instance their intensity (e.g., dark yellow is often perceived as orange or brown). In addition, these "colors" intergrade into each other as the wavelength changes. There are no names that I am aware of for colors at these wavelengths, for the simple reason that they cannot be seen by humans, so in this discussion I will use the corresponding visible colors recorded by the Bayer sensor, and qualify their names with quotation marks. It would be entirely possible to invent new names (how about "fuggle", "mue" and "rallow"?) but to keep things simple I will just use easily understood color names.
A practical problem found in imaging UV wavelength at the short end of the UV-A range and into the UV-B range is that the sensitivity of Bayer sensors drops quickly at these wavelengths. Most broadband UV sources, like xenon flash, also produce a substantially lower output at these wavelengths. The "yellow" wavelengths, for instance, require an increase in output power of about 1 stop with respect to "gray/rust" wavelengths. There is a further drop of between 1 and 2 stops for each of the shorter wavelength "colors" in the above table.
It can be noted how both extremities of the above spectrum are recorded as approximately "blue". Then, is there a practical way to distinguish these two comparable colors produced by substantially different wavelengths? How can we be sure that some pictures are really showing a recorded 300 nm radiation instead of a 395 nm one? The simplest and safest solution is to use narrow bandpass filters, which are generally available with passbands of 10 nm. This does indeed show that some of my test subjects are highly reflective around 300 nm (and possibly lower wavelengths). For instance, a clean, non-anodized aluminium surface has an almost flat reflectivity curve across the UV-A and UV-B. However, in my experience, filters with wider transmission bandwidths are more useful, because they allow multiple UV "colors" to be simultaneously recorded.
A major problem in using broadband filters is that "yellow" is not a primary color, but is instead recorded by a Bayer sensor in both the red and green channels. There is no simple way to separate yellow from green in an image recorded by a Bayer sensor, and one has to live with this fact (isolating just the green, on the other hand, is entirely feasible). Using a narrow-band 340 nm filter like the Thorlabs FB340-10 to take an additional, "green-only" image is a more reliable solution.
"Yellow" is best recorded with an Asahi Spectra XRR0340 filter. This "color" is also recorded, albeit to a lower extent and with less clear visual results, with a Baader U filter, as long as the subject reflects plenty of radiation in this band and only a little in the "gray/rust" and "violet".
A possible solution is stacking together two of the filters I use most often, the Baader U (red line, from Diglloyd, in the above diagram) and the Asahi Spectra XRR0340 (violet line, from Asahi Spectra).This produces the combined transmission spectrum approximately shown by the black line. The stacked filters transmit approximately from 345 to 380 nm, which includes some of the long-wavelength "violet" but neither the short-wavelength "blue" nor its long-wavelength counterpart. In other words, if the blue recorded with the XRR0340 disappears by adding the Baader U, it means that this "blue" is produced by wavelengths at or slightly below 300 nm. Likewise, the "blue" transmitted by the Baader U should disappear with both filters stacked.
Since the short-wavelength "blue" is only weakly recorded by a Bayer sensor, in all following pictures I enhanced the blue channel in post-processing.
The above test subject is a cultivated flower (perhaps a Zinnia sp.) frequent throughout the summer in Sweden. I am aware of several other flowers suitable for this type of test, including several large white (in visible light) "daisy" species that should be common in many other places. Violet thistle flowers also usually (but not always) are highly reflective around 300 nm, together with some white orchids. Yellow and orange (in visible light) flowers, on the other hand, typically reflect UV-A but not much UV-B (with exceptions, like the above one).
After enhancing the blue channel in the first two images, it is evident that a strong blue component exists in both images. The distribution of the blue areas, however, differs (it is patchier with the XRR0340. Shooting with both filters stacked produces an image with very little blue (even after enhancing the blue channel in post-processing). This means that the stacked filters remove whatever wavelength is recorded as blue. Since one filter blocks the longest "blue" wavelengths and the other blocks the shortest ones, this means that the "blue" transmitted by the XRR0340 in the test image is short-wavelength, and therefore blocked by the Baader U. In turn, this means that the "blue" in the XRR0340 image is caused by wavelengths around 300 nm. Thus, it is feasible to record these UV-B wavelengths with the equipment used for this test.
The above pictures show a few additional examples of UV-B recorded as blue with the Asahi Spectra XRR0340 filter. There are of course many more instances of images that show no blue component at all with this filter. It can be noted how the distribution of "blue" areas in the above images is often uneven. This may mean that high levels of UV-B in these images are the result of highly directional reflection of radiation from these surfaces, rather than of diffused reflectivity. This behavior tends to agree well with the general observation that UV radiation does not penetrate deep into biological structures, and is instead either absorbed or reflected by their superficial layers.
With suitable equipment, it is possible to use a camera equipped wih a standard Bayer sensor (stripped of its UV- and IR-blocking filter) to record images at wavelengths as low as 300 nm, i.e. in the UV-B range.