Ultraviolet Photography


UV Camera.

The ultraviolet region is expansive, though perhaps not as vast as infrared, and covers the range of about 400nm down to about 10nm wavelength. (A nanometer, or nm, is a billionth of a meter.) Unlike infrared, which can be produced and detected with little effect on objects or people besides a slight warming, ultraviolet ranges from slightly dangerous to very dangerous. Fortunately, the harshest UV wavelengths are blocked by the air, since nitrogen, oxygen, and water vapor are all as clear as ink to wavelengths below about 170-200nm. Our ozone layer blocks UV deeper than about 290nm from reaching us, though artificial UVC lamps (such as the germicidal type) emit a hazardous 253.7nm ray that travels easily through air and destroys bacteria, makes intense sunburns on skin, and causes conjunctivitis - that's where your eyes feel like they have sand in them.

But in our everyday lives, most of us aren't exposed to that kind of danger. It's true that rays even up to 340nm or longer can cause sunburn, if they are intense enough or one is exposed to them for long enough. Fluorescent blacklight tubes in the USA emit at either 350nm or 370nm, in both cases not intensely enough to cause much harm; in the UK only the 370nm type tubes are available. The region from about 340-400nm is commonly referred to as the blacklight region, and a great many things fluoresce when exposed to this kind of light. A great many animals can also see in this range, including cats and dogs (they see to about 360nm), rats and mice (down to about 300nm), birds (limited to anywhere from 360nm for ducks all the way down to about 300nm for small songbirds), reptiles, fish (depending on the species), insects (down to about 280nm), crustaceans, arachnids, etc. Most of these animals see ultraviolet as a primary color.

Plants take advantage of insects' superb UV sensitivity to signal the presence of nectar in their flowers. Some use a UV patterning scheme where the nectar rich center absorbs UV while the outsides of the petals reflect it; others adopt a strategy of strongly absorbing UV all over and being either white or colorful in other wavelengths.


Yellow bell flowers (species Tecoma stans).

Bougainvilleas.

Some kind of yellow composite flower.

Some kind of orangeish composite with strong UV.

Red rose.

Petunias.

Pink and white lantana*.

Dianthus (pinks).

Pinkish red zinnia.

Poinsettia.

Some kind of little white flower.

Some kind of purple flowers with yellow centers.
*The more usual yellow and red lantanas just look very dark, almost black, to this camera.

As you can see, flower colors take advantage of the full range of bees' visual hues. We primates see most flowers as white, yellow, pink, red, or purple. It's as if plants were selecting their bloom colors by turning knobs on some machine, one knob for each color channel, getting creative in so many ways, except they kept turning the red channel all the way up. But it's actually kind of hard for a lifeform to absorb red - very few biochemicals do aside from chlorophyll - so the fact that the flowers usually don't have to means they usually don't. The bright color of a red rose is as lost on most bees as the bright infrared reflectance of leaves is on our eyes. Or, to think of it another way, while we miss out on the UV channel of our four-color world, most insects miss out on the red channel instead.

All of these flower pictures were taken with a Raspberry Pi and a NoIR camera. (See the infrared page for context if you haven't already.) It may seem counterintuitive to use an IR camera to image in the UV range, but the fact is the hot mirror in a regular digital image sensor also blocks some of the UV. But that means any infrared, even the faintest leak through the camera's exterior, can overexpose the image, or create internal reflections, etc, and just generally ruin the photos. So the camera case has to be made as black as possible to all wavelengths the image sensor can detect, and it has to be light tight. Because recall the spectral response of the sensor - here it is again for reference:

The UV range is off the edge of the chart, past 400nm, and it's so inconsequential to the total wavelength response that the chart doesn't bother to include it. Yet this insignificant sliver of light sensitivity is the very thing we're after. The camera uses three glass filters each with intense absorption at the wavelengths we don't want. One puts a sharp cutoff at 550nm, to remove all yellow and red light. It admits a bit of infrared, however. Another filter absorbs strongly in the infrared and red regions, becoming clear between approximately orange and UV. Finally, the filter that makes the magic happen. It's a special piece of glass made of pretty much the same material as the tube of a fluorescent blacklight. But it has to be a certain thickness: too thick and the camera will see only UV; too thin and visible blue and green light will overpower the image. The result is as close to a perfect balance of wavelengths as is available. The image sensor can see blue and green much more easily than UV, so the filters compensate by passing almost all blacklight UV and only a little blue and green.

Only a small percentage of the incoming light makes it to the image sensor, and most of it is at wavelengths where the sensor has a low efficiency. Altogether the camera registers about a thousandth of the available light. So it can pretty much only shoot in daylight or under bright artificial light. But the system of filters works and the device takes pictures in a 350-550nm wavelength range.

It is certainly possible to create CUV images just like we did for CIR. One can simply map green to red, blue to green, and UV to blue.


CUV picture showing how strongly white paint absorbs UV.

But for the flower photos, and for UV photos in general, I decided to create a "bee mode", a channel mapping that, while we cannot know if it is closer to what an insect sees, may at least look more familiar to us. This mode maps green to yellow, blue to cyan, and UV to magenta. It's not exact since our monitors' idea of blue is about 440-460nm while the camera's idea of blue is closer to 425nm, and wavelengths down near 350nm can cycle back into the green channel and look like bees' purple, but the basic idea is that foliage remains greenish yellow and UV tends to show up as purple/pink which is how it often gets represented anyway. (Bees' purple, by the way, is a mixture of UV with green or yellow, and looks pink or red in bee mode, like the fringes of the yellow bell flowers.)


Spectrum of sunlight in bee mode.

UV reptile lamp next to a regular white LED bulb.