Ultraviolet for the Alternative Processes: Sources, Dose, Uniformity and Hazard

A cyanotype sensitiser is bright yellow, and the obvious inference — that a yellow coating must be absorbing blue light — is right about the colour and useless about the exposure. The yellow you see is mostly the ferricyanide, which is not the light-sensitive part. The light-sensitive part is the iron(III) complex, and what it can use is further into the ultraviolet than your eye reaches. So the first thing to establish about a UV source is not how bright it looks. It is what fraction of what it emits the sensitiser can actually absorb, and that fraction turns out to fall off a cliff between 365 nm and 420 nm.

This page settles four things before the UVA unit is built: what wavelengths the processes need, how irradiance and time combine into a dose, what each available source really delivers, and what ultraviolet does to a person — which matters more here than anywhere else in the course, because this is the only source in the darkroom that is doing damage while you feel nothing at all.

What these processes can actually see

Ware puts a floor under the whole family. Working from the Stark–Einstein law — one photon absorbed per molecule transformed — he calculates the maximum possible sensitivity of any “proto-photographic” material, whatever its detailed chemistry, and finds it lies in the near ultraviolet and blue, over roughly 300 to 400 nm, and that it takes about 34 J/m² to form a just-perceptible image. On a photographic scale that is a speed of about 10⁻⁵ ISO: one ten-millionth of an ordinary 100 ISO film.

Two consequences follow immediately, and both shape practice.

These papers are not darkroom materials. Bostick & Sullivan state it for their own kit: the solutions are sensitive to ultraviolet light only and can be handled under normal room lighting, with incandescent lighting usable throughout the process, and only fluorescent lighting to be limited because of its ultraviolet content. That is why you coat cyanotype at a kitchen table under a tungsten lamp and it is why the safelight discipline of the previous three pages does not transfer here. A different discipline replaces it: the coated sheet must be kept away from daylight and away from fluorescent and LED room lighting, whose blue-pumped white spectra reach further towards the violet than tungsten does.

And the exposures are long. Ware’s arithmetic for sunlight is worth following because it is the only end-to-end calculation of its kind in this course’s corpus. The sun at its zenith delivers about 900 W/m² at the Earth’s surface, of which only about 7 per cent is actinic — UVA and blue — which is about 63 W/m² at best; taking account of the sun’s average elevation, the working figure is 30 to 40 W/m² of UVA. Against 34 J/m² for a just-perceptible image, that is about a second. But the exposure scale of a printing-out process like cyanotype is 7 to 8 stops from just-perceptible to full density, a factor of 128 to 256 — so full density in average sunlight lands at 2 to 4 minutes. That is the number every artificial source is competing with.

The bands, as the standards define them

Three letters, and the course uses them in the standards’ sense and no other.

The ultraviolet bands, the LED peaks that are sold for these processes, and where the sensitiser stops using them

• UVB 280-315 nm (280–315 nm) — no source in this course emits here
• UVA 315-400 nm (315–400 nm) — the working band for the iron processes
• 365 nm LED (360–370 nm) — half width about 9 to 10 nm
• 385 nm LED (380–390 nm) — half width about 11 nm
• 395 nm LED (390–400 nm) — half width about 11 nm
• 410 nm: ferricyanide absorbs (405–415 nm) — Ware: lamps peaking here are not effective for cyanotype

Band limits from the CIE designations quoted by ICNIRP; LED peaks and half-widths from the Nichia and Luminus datasheets; the 410 nm note and the absorbed fractions from Ware's Cyanomicon. The coloured strip approximates where the visible spectrum falls and is a reading aid only; the wavelengths in the labels carry the information.

UVA is 315 to 400 nm, UVB is 280 to 315 nm, UVC is 100 to 280 nm, in the CIE designations ICNIRP work from. ICNIRP add one caveat the course repeats rather than smooths over: some specialists follow the same scheme but put the UVA/UVB dividing line at 320 nm rather than 315. Five nanometres sounds like pedantry until you notice that the biological weighting changes by a factor of three across it.

The course’s rule, stated once and applied everywhere after this. Every ultraviolet source in this course is UVA. No UVB, no UVC, and no mercury discharge lamp of any kind. The reason is Ware’s and it is not a matter of taste: there is no advantage in the shorter wavelengths for these processes, because the sensitiser cannot use what it cannot absorb, and there is a great deal of additional risk, because short-wave ultraviolet damages eyes and living tissue rapidly. A control that costs nothing in results is a control the course takes.

Irradiance, dose, and the arithmetic of distance

Two quantities, and the course keeps them apart the way it keeps concentration and dilution apart.

H = E × t

Dose from irradiance and time

E is the irradiance at the paper, in watts per square metre — the rate at which energy arrives. t is the exposure time in seconds. H is the radiant exposure, or dose, in joules per square metre. Irradiance is a property of the source and the geometry; dose is what the paper responds to.

The geometry half is the inverse square law for anything small enough to count as a point:

E ∝ 1 ÷ d²

Irradiance against distance from a small source

Double the distance and the irradiance falls to a quarter, so the time to reach the same dose goes up four times. This is exact for a point source, approximately right for a compact one, and wrong for an array close to the frame — where the paper sees many emitters at once and the fall-off is much gentler than d⁻², which is precisely why arrays are built close and flat rather than far and bright.

🧮Two worked doses, both from published figures

Platinum-palladium, calculated and confirmed. Ware reports a calculation for an iron-based sensitiser predicting a full-scale image in about two minutes at 50 W/m² at 365 nm, and records that the prediction is borne out in practice. Two minutes at 50 W/m² is a dose of

H = 50 × 120 = 6,000 J/m².

The simple cyanotype, measured. Ware’s own instructions give 5 to 10 minutes under an average 365 nm UVA source. If your unit delivers 50 W/m², that is 15,000 to 30,000 J/m² — two to five times the platinum-palladium figure, which is consistent with his remark that the formation of the Prussian blue image is the less efficient of the two.

What this is for. Not to save you the dose series: your unit is not “an average 365 nm source” until you have measured it. It is so that when your first test strip comes out at twelve minutes rather than ninety seconds, you know immediately whether your unit is in the right order of magnitude or whether something is wrong with it.

Does reciprocity hold?

Over the exposures these processes actually use, the course treats the reciprocity law as holding, and says on what basis.

What the evidence supports. Ware’s dose calculations, for both cyanotype and platinum-palladium, assume Bunsen–Roscoe reciprocity and are borne out in practice at exposures of minutes. That is the range this course prints in, and within it the assumption is doing real predictive work rather than being asserted.

Where he says it fails. When he considers cyanotype in a camera, at exposures of hundreds of hours, he concludes that reciprocity failure would in all likelihood ensure that no image forms at all. So the failure is real and it lies far outside printing exposures.

A distinction the page insists on. Ware also demonstrates, with measurements, that the fading of finished cyanotypes does not obey reciprocity — a print at 50 lux for 2,000 hours shows no detectable fading while the same 100 kilolux-hours delivered in 25 hours of daylight fades it measurably, because aerial re-oxidation regains density as fast as light removes it at low illuminance. That is a different phenomenon in a different material at a different time, and quoting it as evidence about printing exposures would be a category error. The course states it here precisely so that nobody makes that error later.

And the honest gap. No systematic published test of printing-exposure reciprocity for the iron processes across the intensity range a home unit spans — an array at 20 W/m² against summer sun at 35 W/m² — was found in this course’s corpus. If you want to know whether your own unit and the sun agree, the dose series tells you, and it is a worthwhile experiment rather than a formality.

The sources, compared honestly

Ware lists eight in ascending order of cost. The course uses four of them, excludes two, and adds one he does not have.

The sun. Free, spectrally ideal, and completely unrepeatable. It is very nearly a point source, subtending only 0.5 degrees, which makes it the sharpest contact-printing source there is — go back to the penumbra arithmetic and its aspect value of 0.01. Two costs come with that. It moves through its own diameter in about two minutes, so a long exposure is made by a source that is not in one place. And it heats the print, which can discolour highlights over a long exposure. The north summer sky is the diffuse alternative: no shadow problems, an aspect value like a light bed, and about three stops (eight times) weaker than direct sun.

Fluorescent blacklight tubes. For decades the standard answer, and still a good one: even over a large area, cheap per square metre, and available as a luminaire full of 600 mm tubes. Ware notes that tubes sold as insect attractors are suitable while the “super actinic” plant and aquarium variety is not, and that filtered blacklight-blue lamps work but less efficiently. Three costs: output falls with operating hours, the tubes are physically fragile, and they contain mercury, which is a disposal obligation this page states below.

UVA LEDs. Efficient, instantly at full output, controllable by a low-voltage switch, and available at the three peaks the diagram above shows. They are what this course builds with, and the reasons are practical rather than romantic: no warm-up, no ballast, no mercury, and a datasheet.

UV nail lamps. Cheap, small, everywhere, and a genuine hazard. They are designed to put ultraviolet onto a hand held a few centimetres away with no enclosure at all, which is exactly the geometry the control hierarchy below exists to prevent. The course permits one only as an enclosed source inside a box you build, for postcard-sized work, and never as a bench-top lamp.

Metal-halide and mercury arc, excluded. Bostick & Sullivan’s own instructions offer a 1,000 W metal halide bulb as a printing source, and commercial units like the NuArc are the professional standard. The course excludes them for a home darkroom on three grounds: the lamp is at mains potential inside a hot enclosure, the short-wave content demands controls a home cannot verify, and the arc’s output drifts enough that professional units carry a light integrator — an instrument that measures irradiance continuously and accumulates dose rather than counting seconds — precisely because time is not a reliable proxy for exposure with them. That last point is worth keeping: it is the clearest statement in the literature of why dose and time are different quantities.

!What a listing means by 'UV', and how to check

Waveform, who sell a UVA strip, say in their own product description that strips marketed as ultraviolet are frequently near-UV at 405 nm, violet at 420 nm, or a red-and-blue mixture. That is a manufacturer making a claim about its competitors, and the course cites it as that rather than as independent evidence — but it is consistent with what the datasheets do and do not say, and the practical test does not require you to believe anybody.

Ask for four numbers, and refuse the part if they are not published: peak wavelength, spectral half width, radiant flux at a stated forward current, and the thermal derating. Nichia publish all four for the NVSU233B; so do Luminus for the SST-10-UV. A listing that gives you “395-405 nm UV” and a wattage in watts of electrical input has told you nothing about what reaches the paper.

And there is a free test. Ware’s 410 nm point is the one you can see: a lamp whose output is substantially above 400 nm looks distinctly violet and lights up a white shirt. A true 365 nm source looks dim, dull and slightly blue-white, because almost all its output is invisible. A UV lamp that looks impressive is a UV lamp that is wasting most of its power in the visible.

Choosing a wavelength: the internal filter decides it

Here the electrical answer and the photochemical answer point in opposite directions, and the photochemical one wins.

The electrical case for the longer peak. Nichia’s figures for one emitter family at 1,000 mA: 365 nm gives 1,450 mW at 3.85 V; 395 nm gives 1,650 mW at 3.65 V. So the 395 nm part delivers 14 per cent more radiant power for 5 per cent less electrical power — about a fifth better wall-plug efficiency — and the shorter-wavelength parts are usually dearer as well. On efficiency and price alone, 395 nm wins easily.

The photochemical case against it. In a cyanotype coating the hexacyanoferrate(III) acts as an internal filter: it absorbs light that the photoactive iron complex would otherwise have used, and its absorption rises steeply into the violet. Ware calculates the fraction of incident light actually absorbed by the photoactive component for a typical cyanotype coating, and it is 0.65 at 365 nm and 0.03 at 420 nm — a factor of more than twenty. He puts the practical conclusion in one sentence: lamps with a peak output around 410 nm are not effective for cyanotype, because 410 nm is an absorption maximum of ferricyanide.

The quantum yield moves the same way, though less dramatically: 0.45 at 365 nm at pH 4, falling to 0.28 at 436 nm.

⚗Two different quantities, both moving against you

It is worth being clear which is which, because they multiply.

Absorbed fraction is how much of the arriving light the photoactive species gets to see at all, after the ferricyanide has taken its share. That is what falls from 0.65 to 0.03.

Quantum yield is how many molecules are transformed per photon that the photoactive species does absorb. That falls from 0.45 to 0.28.

A source at the wrong end of the band therefore loses twice over. And the loss is specific to the chemistry of that sensitiser: the internal-filter argument is a cyanotype argument, and a platinum-palladium or Van Dyke coating has a different filter and a different answer. The course therefore states the wavelength question as one to be settled per process by measurement, and gives 365 nm as the default because it is the wavelength both Ware’s calculation and his instructions use, and because it is where most UVA fluorescent lamps peak — the atomic mercury line — which is why so much of the alternative-process literature is implicitly written for it.

Uniformity from an array

An array is a set of point sources at a distance, and everything about its evenness follows from that.

Spacing against height. Each emitter’s contribution at the paper falls with distance and with obliquity. Directly above an emitter you get its full contribution plus a little from its neighbours; midway between two you get two partial contributions. The difference — the ripple — dies away as the array is moved further from the paper, and it dies faster the closer together the emitters are. The levers therefore trade against one another in cost: closer emitters mean more strip and more current, a greater height means a bigger box and less irradiance at the paper.

A diffuser costs dose and buys evenness. Any diffusing sheet scatters some light back into the box and absorbs some. In a visible-light box that is cheap. In a UV box it is expensive twice over, because your exposures are already minutes long and because many plastics absorb strongly in exactly the band you are using — ICNIRP record that polymethyl methacrylate and polycarbonate normally do not transmit below about 370 nm, which puts a sheet of ordinary acrylic squarely in the way of a 365 nm array. A UV unit’s evenness is therefore bought with geometry rather than with diffusion, which is the opposite of the contact printer’s answer.

Measure it with the chemistry, not with a meter. The honest instrument for a UV array’s uniformity is a set of identically coated strips exposed simultaneously across the printing area and read after washing, because that measures what the paper receives in the band the paper uses. A broadband UV meter of unknown spectral response measures something adjacent to that. The UVA unit page runs that test.

Heat, ageing and lamp hours

Three drifts, on three different time scales, and each has a different remedy.

Minutes: the junction warms and the output falls. An LED’s radiant flux falls as its junction temperature rises, and the junction rises for as long as the heat is going in faster than the heatsink is taking it out. Nichia give the numbers you need to reason about it: a maximum junction temperature of 130 °C, a thermal resistance from junction to solder point of 3.9 °C/W typical and 5.7 °C/W maximum, and an operating temperature range of −10 to 85 °C. Luminus give 1.4 °C/W for their part. The remedy is a warm-up before the session, not a bigger heatsink — the heatsink stops the emitter failing, the warm-up makes it repeatable.

Hours to years: the tube ages. Fluorescent output falls with operating hours. This course could not source a published decay curve for UVA blacklight tubes from a manufacturer, so it gives no figure and no replacement interval. What it gives instead is the practice that survives not knowing: an hours log on the unit, and a dose series re-established at intervals you choose and record. Ware’s observation about the professional units carrying integrators is the same problem solved with money.

Every session: the paper’s own moisture. Here the course must correct a claim it would be easy to make. Ware is explicit that the prevailing relative humidity has very little effect on the results of printing in cyanotype, but can profoundly affect some of the other iron-based processes. So “keep the humidity constant” is good advice for Van Dyke, kallitype and platinum-palladium, and close to irrelevant for cyanotype. What does matter for cyanotype is the drying: Ware’s method is an hour in the dark, or a uniform stream of warm air at 40 °C for ten minutes — and if you heat-dry, the sheet must rest for half an hour to an hour before exposure, or you lose a little density and gain a little contrast. Direct heat from a hairdryer produces uneven results.

That last point is why a hot exposure unit is a problem beyond its own electronics: a box that warms the frame is drying the paper mid-exposure, and for the processes that care about moisture it is changing the material while it prints it.

What ultraviolet does, stated precisely

The single most important property of ultraviolet as a hazard is that it is invisible and painless while it is doing harm. There is no aversion response, no squint, no discomfort. Every other hazard in this course announces itself.

The bands do different things and the page will not blur them. The acute injuries the word “ultraviolet” usually evokes — photokeratitis, the welder’s flash, sunburn at short exposure — are overwhelmingly UVB and UVC effects, and ICNIRP’s own weighting says so: the relative spectral effectiveness for acute injury is 1.000 at 270 nm, 0.003 at 315 nm and 0.00011 at 365 nm. A UVA photon is roughly nine thousand times less effective at causing the acute injury than a 270 nm photon.

ICNIRP's relative spectral effectiveness across the near ultraviolet, on a log scale

• S(λ) from ICNIRP Table 1

Show the numbers behind this plot

A plot of the base-ten logarithm of ICNIRP's relative spectral effectiveness against wavelength from 300 to 400 nanometres, with the plotted points taken directly from Table 1 of the 2004 guidelines. The curve starts at minus 0.52 at 300 nanometres, which is a weighting of 0.30, and falls very steeply through minus 1.22 at 305 nanometres and minus 1.82 at 310 to minus 2.52 at 315 nanometres, where the UVA band begins and the weighting is 0.003. It then flattens dramatically: from 315 to 400 nanometres the whole further fall is only about two more decades, reaching minus 3.96 at 365 nanometres, where the weighting is 0.00011, and minus 4.52 at 400 nanometres, where it is 0.00003. A vertical guide marks 315 nanometres as the start of the UVA band and a second marks 365 nanometres. The shape carries the argument of the section: almost the entire dynamic range of the acute-injury weighting is spent below 315 nanometres, so within UVA the weighted limit is nearly blind, and it is the separate unweighted limit for the eye that binds instead.

Series	Wavelength, nm	log₁₀ of relative spectral effectiveness S(λ)
S(λ) from ICNIRP Table 1	300.00	-0.52
S(λ) from ICNIRP Table 1	305.00	-1.22
S(λ) from ICNIRP Table 1	310.00	-1.82
S(λ) from ICNIRP Table 1	315.00	-2.52
S(λ) from ICNIRP Table 1	320.00	-3.00
S(λ) from ICNIRP Table 1	325.00	-3.30
S(λ) from ICNIRP Table 1	330.00	-3.39
S(λ) from ICNIRP Table 1	335.00	-3.47
S(λ) from ICNIRP Table 1	340.00	-3.55
S(λ) from ICNIRP Table 1	350.00	-3.70
S(λ) from ICNIRP Table 1	360.00	-3.89
S(λ) from ICNIRP Table 1	365.00	-3.96
S(λ) from ICNIRP Table 1	370.00	-4.03
S(λ) from ICNIRP Table 1	380.00	-4.19
S(λ) from ICNIRP Table 1	390.00	-4.36
S(λ) from ICNIRP Table 1	400.00	-4.52

Points read from Table 1 of the ICNIRP 2004 guidelines. Plotted as a logarithm because a linear axis would show the whole UVA band as a flat line on zero, which is exactly the impression the page is trying to correct.

What UVA does instead. ICNIRP’s own appendix is unusually candid, and the course quotes its shape rather than converting it into reassurance. There is a lack of evidence that the UVA levels met in sunlight or in most indoor work — of the order of 1 to 3 mW/cm² — harm skin or eye. But the hypothesis that UVA may be one causative agent in cataract suggests the need for caution about chronic low-level ocular exposure, and ICNIRP say plainly that experimental threshold data are lacking and that in their absence they recommend a more cautious approach. That is a limit set under acknowledged uncertainty, and a reader is entitled to know that is what it is.

One specific and unusual instruction, worth reading twice. People who have had a cataract operation and received an intraocular lens that was not designed to absorb UVA, and people with no implant at all, are not protected the way an intact crystalline lens protects. ICNIRP say directly that such people should be fitted with UVA protective eyewear if working with sources of UVA radiation. If that is you, the enclosure is not merely good practice — it and the eyewear are the whole of your protection.

The exposure limits, and what they mean for a box in a spare room

ICNIRP set two limits for the unprotected eye within any 8-hour period, and for a UVA source the second is the one that binds.

1. Radiant exposure over 180 to 400 nm, spectrally weighted by S(λ), not to exceed 30 J/m².
2. Total unweighted radiant exposure over 315 to 400 nm, not to exceed 10⁴ J/m².

For skin, the limit for the most sensitive phototypes is the same 30 J/m² weighted figure, stated as a desirable goal rather than an achievable one — ICNIRP note it is hard to meet in sunlight at all.

🧮Why the second limit is the one that matters, and what it works out to

Take a pure 365 nm source, where S(λ) = 0.00011.

Under limit 1, the weighted 30 J/m² corresponds to an unweighted radiant exposure of 30 ÷ 0.00011 = 2.7 × 10⁵ J/m². ICNIRP print exactly that figure in Table 1 as the monochromatic limit at 365 nm.

Under limit 2, the cap is 10⁴ J/m² — twenty-seven times lower.

So limit 2 governs, and it is worth converting into something you can picture. Spread over an 8-hour day (28,800 s):

10⁴ ÷ 28,800 = 0.35 W/m², which is 35 µW/cm².

Compare that with the 30 to 40 W/m² of UVA in average sunlight, or with the tens of watts per square metre a printing array puts on the paper. The array’s working field is four to five orders of magnitude above the all-day limit at the operator’s position. That is not a reason to panic; it is the reason the control is an enclosure rather than a rule about how long to stand nearby.

And the limit the course will not compute for you. Turning an irradiance measured at the side of your box into a permissible standing time needs a calibrated instrument with a known spectral response, and a home unit does not have one. The honest position, which the build page adopts, is therefore to require zero detectable leakage on a fluorescent indicator rather than to invent an acceptable one. A limit you cannot measure against is not a control.

The controls, in order

ICNIRP’s engineering hierarchy and HSE’s own list of control measures under the Control of Artificial Optical Radiation at Work Regulations 2010 say the same things in the same order, and the order is the point: each control removes the possibility of exposure rather than reducing its likelihood.

The control hierarchy for an ultraviolet source, strongest first

1. 1A light-tight enclosureICNIRP name light-tight cabinets and enclosures, absorbing shields and baffles as the key engineering control. The array is never run open on a bench. No viewing window: an observation port must be of tested absorbing material, and the course would rather you had none than an untested one.
2. 2A fail-safe interlockWhere direct access to the source is required, ICNIRP require fail-safe interlocks manufactured, installed, tested and used to agreed technical standards. Opening the lid removes power from the array. HSE list safety interlocks among the control measures to consider.
3. 3Non-reflective interior surfacesMany visibly shiny surfaces reflect ultraviolet well. ICNIRP call for surfaces coated or painted with non-reflective material, which is why the UVA unit is matt black inside and the contact printer is matt white.
4. 4An indicator visible from outsideHSE list safety signs and restricted access. A lamp that is on and invisible is the specific failure mode of this instrument, and an indicator is how it is made visible.
5. 5Eyewear and covered skin, as backup onlyICNIRP place personal protective clothing last, for when neither engineering nor administrative controls are practical. HSE list PPE last too. It is what you wear while servicing the unit, not what makes the unit acceptable.
6. 6Never: the operator's judgement as the primary controlA rule that says do not look at it depends on somebody remembering, in the dark, in a hurry, on the four-hundredth exposure. Every level above this one works whether or not anybody remembers.

Order from ICNIRP 14/2007 section 10.2 and HSE's List 3. The bottom two entries are where most home-built units actually sit, which is why the top two are the build's acceptance criteria.

Risk groups, and why they help less than they look. IEC and CIE classify lamps into exempt and Risk Groups 1 to 3 by the exposure duration at which a limit is exceeded at a reference distance — 20 cm for a lamp that is not for general lighting. For the unweighted UVA lens limit, the boundaries are 1,000 s for exempt, 300 s for Risk Group 1 and 100 s for Risk Group 2; anything shorter is Risk Group 3. HSE treat any exempt or Risk Group 1 lamp, LEDs included, as a source needing no further action, and any Risk Group 3 lamp as hazardous.

Three cautions, and ICNIRP raise all three themselves. The classification is a property of the lamp, not of your exposure, which depends on your actual distance and time. The 20 cm reference distance is unrealistic for many real installations. And the CIE exempt category for the unweighted UVA limit was based on a 1,000-second integration where ICNIRP integrate over 8 hours, so a lamp can be “exempt” and still take you past the ICNIRP limit if you stand at 20 cm for more than about sixteen minutes. A risk group on a box is useful information and it is not a permission.

Ozone, and the folklore it replaces

“UV makes ozone” is repeated widely enough to have become a reason people fear UVA sources. The wavelength dependence is what the folklore leaves out, and it is decisive.

ICNIRP state it directly: very intense UVC sources, particularly of wavelengths less than 230 nm, may produce hazardous concentrations of ozone and nitrogen oxides — and this is why many germicidal lamps are now made with quartz envelopes that block below about 230 nm. The mechanism is short-wave photons splitting molecular oxygen, and the low-pressure mercury lamp is the classic culprit because it emits strongly at 254 nm and some quartz envelopes also transmit its 185 nm line.

230 nm is 135 nanometres below the bottom of the UVA band. A UVA LED at 365 or 395 nm, with a spectral half width of about 10 nm, emits nothing remotely near it. A blacklight fluorescent tube is a mercury discharge with a phosphor that converts the 254 nm emission to longer wavelengths, behind an envelope chosen to block the short wavelengths — which is the same construction as an ordinary fluorescent lamp, and ICNIRP report that detailed spectral analysis of general lighting fluorescent lamps finds UVB and UVC emissions extremely low because of the envelope’s attenuation below 320 nm.

So the course’s position is: ventilation in a UVA unit is a thermal control, not an ozone control. Where ICNIRP call for ventilation to exhaust ozone, they are talking about UVC. If you ever smell the sharp, sweet smell of ozone near a lamp, you have a lamp that is not what you were told it was, and the response is to stop using it rather than to open a window.

⚠Mercury, and the one waste obligation this page creates

If you use fluorescent blacklight tubes, you own mercury.

Under the Environment Agency’s WM3 guidance, 20 01 21* fluorescent tubes and other mercury-containing waste is an absolute hazardous entry in the List of Waste, and WM3 states that the vast majority of fluorescent tubes from any source are likely to be similar to domestic types and fall under it. Absolute hazardous means the entry applies regardless of how much mercury a given tube contains: there is no threshold for you to assess.

In practice, in England and Wales that means tubes go to a household waste and recycling centre or a council hazardous waste service, never in the general bin and never broken up. A broken tube releases mercury vapour into a small room, which is a reason to store spares horizontally in their sleeves and to handle a unit full of glass tubes as though it were full of glass tubes. Local regulation governs disposal, and this page gives no jurisdiction-specific instruction; the disposal caveat is the course’s standing statement on that.

An LED array creates no such obligation, and that is a real and unglamorous advantage of building rather than buying second-hand.

Alternative route

If you will never own a UV unit, this page is still the one that makes sunlight usable. Everything in it applies: the sun is a source with an irradiance, a spectrum and a geometry, and the only thing you lose by using it is control of the first of those. Ware’s 2 to 4 minutes for full density in average sunlight is your starting point; a written exposure record — date, time, cloud, the print’s appearance at each inspection — is your substitute for a timer; and the split-back frame is what turns a variable source into a repeatable result, because you judge the print rather than the clock.

If you cannot see the print-out judgement — a reader with low vision, or working in conditions where opening the frame is awkward — the substitute is a dose series established once against a fixed geometry, so exposure becomes a number of minutes at a known distance from a known source rather than a visual judgement. That is exactly what the built unit provides, and it is the strongest argument for building one.

No route around the hazard. There is no version of this subject in which the controls are optional, and that is why they are stated on a lesson page rather than only on the build. A reader who prints in sunlight has an unenclosed UVA source of 30 to 40 W/m² and cannot interlock it; what they have instead is distance, a hat, sleeves and the sense not to look at the sky. Those are administrative controls, and they are the weakest kind — which is the honest reading of printing outdoors, and it is worth knowing that is what you are doing.

Recap

The iron processes work in a narrow band that your eye barely reaches. Their maximum possible sensitivity lies over about 300 to 400 nm, and in a cyanotype coating the ferricyanide acts as an internal filter that takes the useful absorbed fraction from 0.65 at 365 nm to 0.03 at 420 nm — which is why a lamp peaking at 410 nm is nearly useless for the process and why the 395 nm emitter’s better electrical efficiency does not settle the question.

Dose is irradiance times time, and the two are not interchangeable with anything else. Sunlight gives about 30 to 40 W/m² of UVA, enough for full density in 2 to 4 minutes; published figures put a platinum-palladium exposure at about two minutes at 50 W/m² at 365 nm and a simple cyanotype at five to ten. Reciprocity is treated as holding over printing exposures, on the evidence that calculations assuming it work, and as failing at camera-length exposures — and the reciprocity failure Ware measures for the fading of finished prints is a different phenomenon that must not be borrowed as evidence here.

The hazard is unlike every other one in this course because nothing tells you it is happening. UVA is thousands of times less effective than UVB at acute injury — 0.00011 against 1.000 in ICNIRP’s weighting — which is why the binding limit for the eye is not the weighted 30 J/m² but the separate unweighted 10⁴ J/m² over 315 to 400 nm, about 0.35 W/m² averaged across a working day, four to five orders of magnitude below what an array puts on the paper. The controls therefore run enclosure, interlock, matt interior, indicator, and only then eyewear, with the operator’s judgement never first. And ozone belongs to wavelengths below about 230 nm, which no UVA source reaches — so ventilation in a UV box is there for the heat.

Check your understanding

Question 1. A 395 nm LED delivers 1650 mW at 3.65 V while a 365 nm part from the same family gives 1450 mW at 3.85 V. Why does the course still default to 365 nm for cyanotype?

• Because 365 nm LEDs are cheaper
• Because the fraction of light the photoactive iron complex actually absorbs falls from about 0.65 at 365 nm to 0.03 at 420 nm, so radiant watts at the longer wavelength are largely intercepted by the ferricyanide acting as an internal filter
• Because 395 nm is outside the UVA band as the CIE define it
• Because the shorter wavelength penetrates the paper further

Show the answer and why

Answer: Because the fraction of light the photoactive iron complex actually absorbs falls from about 0.65 at 365 nm to 0.03 at 420 nm, so radiant watts at the longer wavelength are largely intercepted by the ferricyanide acting as an internal filter

The electrical comparison is real and it favours 395 nm by about a fifth on wall-plug efficiency. It is simply answering a different question. What reaches the sensitiser is the incident power multiplied by the fraction the photoactive species absorbs, and Ware calculates that fraction for a typical cyanotype coating as 0.65 at 365 nm and 0.03 at 420 nm - so the useful power can fall faster than the emitted power rises. Note the scope: this is a cyanotype argument, because it depends on that coating's particular internal filter. Another iron process has a different answer, which is why the course says the wavelength question is settled per process by measurement.

Question 2. For a pure 365 nm source, which ICNIRP limit for the eye binds, and roughly what continuous irradiance does it correspond to over an 8-hour day?

• The weighted 30 J/m² limit, giving about 0.001 W/m²
• The unweighted 10⁴ J/m² limit over 315 to 400 nm, giving about 0.35 W/m²
• The weighted limit, giving about 9.4 W/m²
• Neither: UVA is unregulated below 400 nm

Show the answer and why

Answer: The unweighted 10⁴ J/m² limit over 315 to 400 nm, giving about 0.35 W/m²

At 365 nm the weighting S(λ) is 0.00011, so the weighted 30 J/m² corresponds to 2.7 x 10⁵ J/m² of actual radiant exposure - the figure ICNIRP print in Table 1 for that wavelength. The separate unweighted cap of 10⁴ J/m² over 315 to 400 nm is twenty-seven times lower and therefore governs. Divided over 28,800 seconds that is 0.35 W/m², or 35 µW/cm² - four to five orders of magnitude below the 30 to 50 W/m² an array puts on the paper, which is why the control is an enclosure and not a rule about standing time.

Question 3. Someone tells you their UVA nail lamp must be safe because it makes no ozone smell. What is wrong with the reasoning?

• Ozone is odourless, so the absence of a smell proves nothing
• Ozone production is associated with wavelengths below about 230 nm, so a UVA source would not produce it whether or not it were hazardous to eyes and skin — the test is unrelated to the hazard
• Nail lamps do produce ozone but at concentrations below the odour threshold
• Ozone is produced by all ultraviolet sources including UVA ones

Show the answer and why

Answer: Ozone production is associated with wavelengths below about 230 nm, so a UVA source would not produce it whether or not it were hazardous to eyes and skin — the test is unrelated to the hazard

The two things are not connected. ICNIRP attribute hazardous ozone and nitrogen oxide production to very intense UVC sources, particularly below 230 nm, which is why germicidal lamps now use envelopes blocking below about that wavelength. A UVA emitter at 365 or 395 nm with a 10 nm half width emits nothing near 230 nm, so it will never smell of ozone regardless of how much ultraviolet it is putting into a hand held five centimetres away. Using ozone as a safety indicator for UVA is testing for a hazard that band does not have and ignoring the one it does.

Question 4. A lamp is labelled "exempt" under the IEC and CIE risk-group scheme. What does that entitle you to conclude?

• Nothing further is needed: HSE list exempt and Risk Group 1 lamps among sources requiring no action
• That at the 20 cm reference distance the relevant limits are not exceeded within the standard's integration times — but for the unweighted UVA limit that integration is 1000 s, not ICNIRP's 8 hours, so standing at 20 cm for more than about sixteen minutes can still exceed the ICNIRP limit
• That the lamp emits no ultraviolet at all
• That the lamp is safe at any distance for any duration

Show the answer and why

Answer: That at the 20 cm reference distance the relevant limits are not exceeded within the standard's integration times — but for the unweighted UVA limit that integration is 1000 s, not ICNIRP's 8 hours, so standing at 20 cm for more than about sixteen minutes can still exceed the ICNIRP limit

The first option quotes HSE correctly and is still the wrong conclusion to draw on its own, which is the trap. ICNIRP raise the objection themselves: the CIE exempt group for the unweighted UVA limit was based on a 1000-second integration where ICNIRP integrate over 8 hours, and the 20 cm reference distance is unrealistic for many installations. A risk group is a property of the lamp; your exposure is a property of your distance and your time. It is useful information and it is not a permission.

Question 5. Why does this page insist that the reciprocity failure Ware measures for cyanotypes is not evidence about printing exposures?

• Because his measurements were made on fabric rather than paper
• Because those measurements concern the fading of finished Prussian blue prints under gallery and daylight illumination, where aerial re-oxidation regains density as fast as light removes it at low illuminance — a different material undergoing a different reaction over days rather than minutes
• Because reciprocity failure cannot occur in an iron-based process
• Because the fading measurements were made at a different wavelength

Show the answer and why

Answer: Because those measurements concern the fading of finished Prussian blue prints under gallery and daylight illumination, where aerial re-oxidation regains density as fast as light removes it at low illuminance — a different material undergoing a different reaction over days rather than minutes

It is the same words describing two unrelated things. In the fading experiments the image already exists and light is destroying it, while a competing dark reaction rebuilds it - which is precisely why 100 kilolux-hours delivered over 2000 hours at 50 lux does nothing and the same exposure in 25 hours of daylight fades the print measurably. Printing exposure is the formation of the image over minutes, and there Ware's calculations assume Bunsen-Roscoe reciprocity and are borne out. Carrying a result across that boundary would be the kind of category error the course's fact-and-interpretation rule exists to catch.

Question 6. You are told to keep the relative humidity constant in the room where you print. For which processes is that advice load-bearing?

• All of the iron-based processes equally
• Cyanotype especially, because Prussian blue formation depends on water
• Some of the other iron-based processes, but not cyanotype, for which Ware reports that prevailing humidity has very little effect on printing results
• None: humidity affects only the wet processing

Show the answer and why

Answer: Some of the other iron-based processes, but not cyanotype, for which Ware reports that prevailing humidity has very little effect on printing results

The advice is right for the family and wrong for its best-known member, which is exactly the kind of generalisation the course tries not to make. Ware states that prevailing relative humidity has very little effect on the results of printing in cyanotype but can profoundly affect some of the other iron-based processes. What does matter for cyanotype is the drying history: an hour in the dark, or warm air at 40 °C for ten minutes followed by a rest of half an hour to an hour, because a heat-dried sheet used immediately loses a little density and gains a little contrast.

Check my answersClear