Iron, Light and Prussian Blue

The pigment in a cyanotype was a hundred and thirty years old and on every painter’s shelf before anybody made a photograph with it. What Herschel found in 1842 was not a new substance but a way of making light decide where that substance would appear. The whole process is two reactions, and only the first of them needs light; everything else on this page — the exposure times, the green cast before washing, the print that deepens overnight, the highlights that bleed, the rule about acid — follows from the shape of those two steps and from one awkward fact about the compound they make.

This page answers three questions. What is the iron doing before the ferricyanide ever meets it? What exactly is the blue, given that its composition is a family rather than a formula? And why is it blue at all? None of the formulations is set out here: the seven cyanotype sensitisers in the formulary carry the quantities, the provenance and the function of every ingredient, and this page is what those pages assume you already understand.

The sensitiser is a complex, not an iron salt

Weigh out a classic cyanotype and you handle two solids: an iron compound and potassium ferricyanide. Only one of them is meaningfully light-sensitive, and Herschel established which by the most direct experiment available — leaving the other one out.

His Article 210 is worth reading in his own words, because the modern account has not improved on it. Paper washed with the ammonio-citrate of iron alone is “of a bright yellow colour, and is apparently little, but in reality highly sensitive to photographic action”. Held “for only four or five seconds in the sun (the effect of which is quite imperceptible to the eye)”, then withdrawn into the shade and washed over with ferricyanide, it gives “a considerable deposit of Prussian blue… on the part sunned, and none whatever on the rest”. In Article 211 he ran a graduated exposure series and drew the conclusion that every subsequent account has restated: the photographic influence “being confined to the first stage, and the ferrosesquicyanate acting as a mere precipitant on the nascent compounds resulting from that influence.”

So the ferricyanide is a developer in no sense at all. It is a reagent that reveals, by precipitation, a change light has already made in the iron.

The ligand is the electron donor

Here is the part that the phrase “an iron salt” hides. The photoactive species is not a bare iron(III) ion. It is iron(III) held by a carboxylate ligand — the anion of an organic acid, gripping the metal through its oxygen atoms — and the ligand is what supplies the electron that reduces the iron. Change the ligand and you change the speed, the keeping, the penetration into paper and the character of the print; that is the whole design space of the seven formulations in the formulary.

Two ligands cover all of them.

Citrate, in Herschel’s own process, in the classic sensitiser and in Ware’s Simple cyanotype and Mike’s cyanotype. Oxalate, in the New Cyanotype and in the commercial blueprint papers that ran the drawing-office trade for eighty years.

Ware states the relationship between them plainly on his own website: the chemistry of the citrate is similar in principle to that of the oxalate but more complicated. That sentence is the licence for most of what follows, and also its limit — where the oxalate case is settled and the citrate case is not, this page says so.

⚗Does iron(III) chloride print?

The usual teaching sentence is that iron(III) chloride on paper is not usefully light sensitive, and in practice that is right: it is not in any published cyanotype sensitiser and no source read for this course recommends it. But Herschel’s own text is more careful than the sentence, and the care is instructive. In the same Article 210 he adds that the effect “is not, it should be observed, peculiar to the ammonio-citrate of iron. The ammonio- and potasso-tartrate fully possess, and the perchloride exactly neutralized partakes of the same property: but the experiment is far more neatly made and succeeds better with the other salts.”

Read that as a chemist rather than as an endorsement. Ammonio- and potasso-tartrate are iron(III) carboxylates like the citrate, and behave like it. An “exactly neutralized” iron(III) chloride is no longer iron(III) chloride in any useful sense: neutralise it and the iron hydrolyses to a polymeric hydrous oxide, and what little photochemistry follows is not the chemistry of the chloride ion. The course’s position, and it is an interpretation rather than a measurement: the photosensitivity belongs to the iron-carboxylate complex, Herschel’s marginal result with the neutralised perchloride is consistent with that rather than against it, and the practical statement stands.

The classic sensitiser is not a compound

Ammonium iron(III) citrate has no formula on its encyclopaedia page and no molar mass, and that is not an omission. It is a family of iron-citrate-ammonium complexes of variable composition sold under one name. Ware records an iron content ranging from 14 to 28 per cent by weight across samples, with the colour running from light green to dark brown as that proportion rises, and quotes the chemist’s verdict on such a material: ill-characterised.

The reason is in the preparation, which has not changed since the pharmacopoeias adopted it. Iron(III) hydroxide is precipitated with ammonia, dissolved in citric acid and neutralised — and iron(III) hydroxide is a highly polymeric, often colloidal substance, so the product inherits its indefiniteness. No single formula can be written for the species actually present. One rare investigation, at the Mallinckrodt Chemical Works in 1949, did obtain both the green and brown forms as crystalline solids giving sharp X-ray powder lines, and the authors could still draw no structural inferences from the data, not even unit cell parameters.

Two consequences follow, and they run through the whole part. Two printers working apart cannot be certain they hold the same substance. And no honest molar arithmetic is possible on the sensitiser without a stated iron content, which is why the classic sensitiser page has to reckon the iron-to-ferricyanide ratio through a range rather than a number.

⚗The one iron-citrate complex that has been pinned down

In 1998 Matzapetakis, Raptopoulou, Tsohos, Papaefthymiou, Moon and Salifoglou isolated a monomeric dicitratoferrate(III) salt and determined its structure by X-ray diffraction: (NH₄)₅Fe(C₆H₄O₇)₂·2H₂O. Their title in the Journal of the American Chemical Society claims it as the first mononuclear, water-soluble iron-citrate complex. The structure shows each citrate acting tetrabasic — it has lost four protons, including the one on its hydroxyl group — and terdentate, with one of its three carboxylate groups not coordinated to the iron at all. To analyse it, the authors had to precipitate it out of a highly soluble solution with a large volume of ethanol.

Ware’s insight, twenty years later, is that a photographer does not need the solid. Dissolve iron(III) nitrate into citric acid already present at twice the molar quantity, then bring the pH up with ammonia, and the same anion forms in the beaker. That is the whole of the Simple cyanotype, and it is why a formulation invented in 2019 is chemically better defined than the one in use since 1842.

The relevant speciation also explains why pH is a real variable in an iron sensitiser and not a detail. Ware reports that in the iron(III)/citrate system a photo-inactive monomer predominates between pH 0.5 and 1.5 while a photo-active dimer forms above pH 2, and that two crystalline dinuclear citratoferrate(III) anions have been isolated and their structures determined. Acidify a citrate sensitiser too far and you are not restraining it, you are switching it off.

The photochemical step: an electron moves from the ligand to the iron

The mechanism class has a name, and the name is the mechanism: ligand-to-metal charge transfer. A photon in the near ultraviolet is absorbed by the complex as a whole, and the transition it drives moves an electron out of an orbital that is mostly on the ligand into one that is mostly on the metal. The iron is left as iron(III) with an extra electron — that is, as iron(II) — and the ligand is left one electron short, which for a carboxylate means it falls apart, losing carbon dioxide. The technical word for that loss is decarboxylation, and it is why an iron sensitiser is used once: the ligand is a consumable, not a catalyst.

⚗What is established, and for which ligand

The oxalate case is settled, and the course prints it as an equation. Trisoxalatoferrate(III) photolysis is one of the best-studied reactions in inorganic photochemistry — it is the basis of the standard chemical actinometer — and Ware gives the overall stoichiometry:

2 [Fe(C₂O₄)₃]³⁻ + light → 2 [Fe(C₂O₄)₂]²⁻ + C₂O₄²⁻ + 2 CO₂

Photoreduction of trisoxalatoferrate(III): two iron(II), one released oxalate and two molecules of carbon dioxide

The mechanism behind it, first proposed by Hatchard and Parker, runs in two steps: light makes an oxalate radical anion inside the excited complex, and that radical then reduces a second, unexcited complex. One photon therefore delivers two iron(II), which is why the measured yield exceeds one.

The citrate case is not settled, and the course does not pretend it is. The organic photoproduct has been identified as acetone dicarboxylic acid, CO(CH₂COOH)₂, which may decarboxylate further to acetone. But Ware writes the change as a scheme rather than a balanced equation, and says why in the same sentence: the nature of the iron(II) photoproduct remains unknown.

hν + [Fe₂(C₆H₄O₇)₂(OH₂)₂]²⁻ → 2 [Fe^II] + CO(CH₂COOH)₂ + CO₂

Ware's scheme for the citrate photolysis, written as he writes it

That “[Fe^II]” carries no ligands and no charge because nobody has established what it is. This course will not close the balance by inventing an iron species, and no balanced equation for the citrate photolysis appears anywhere in it. What is established for the citrate is the direction and the cost: light absorbed, iron(III) reduced to iron(II), citrate oxidised, carbon dioxide lost. Everything finer — the exact complex, the stoichiometry, the intermediate — is extrapolated from the oxalate system, and Ware’s own hedge is that the iron(II) photoproduct “is likely to be a citrato- complex”. Read likely as the word he chose.

Why it is slow, why it must be ultraviolet, and why an enlarger will not do

Two numbers govern the speed of any process of this kind, and both are unforgiving.

The first is the quantum yield: the number of molecules changed per photon absorbed. The Stark-Einstein law, which dates from about 1912, sets the book-keeping — one absorbed photon for each molecule transformed — so a quantum yield of one is the ordinary ceiling and anything above it needs a mechanism that gets a second molecule for free. Ware collates the measured values:

Photoactive species	Quantum yield	Where it appears
Trisoxalatoferrate(III), 250–420 nm	about 1.2	New Cyanotype, blueprint papers
Citratoferrate(III), 365 nm at pH 4	about 0.45, falling to 0.28 at 436 nm	classic sensitiser
Hexacyanoferrate(III) on its own	of the order of 0.01	Herschel’s proto-cyanotype

The second is that nothing amplifies. In a silver gelatin material a developer finds a speck of a few silver atoms and reduces a whole crystal of hundreds of millions; here the photons make the product directly and the process stops. Ware works out what that costs by assuming perfect conditions — every photon absorbed by the photoactive species, every absorbed photon doing chemistry, every photoproduct becoming pigment — and arrives at about 32.8 J/m² for a just-perceptible image in the near ultraviolet, which on a photographic scale is a speed near 10⁻⁵ ISO. Real sensitisers come in below that ideal, not above it.

🧮Why nobody prints a cyanotype in a camera or under an enlarger

Ware runs the standard camera-image (Jones-Condit) equation for a mid-grey subject in average sunshine, with the actinic irradiance at 38 W/m², and reduces it to one convenient approximation:

t ≈ 4A² minutes

Ware's order-of-magnitude exposure for an in-camera iron-process image

where A is the f-number and t is the exposure needed for a tone of mid density. At f/8 that is 4 × 64 = 256 minutes, something over four hours, in direct sun, for a mid grey. At f/16 it is seventeen hours.

An enlarger is worse again, because the lamp is not the sun and the light has been through a negative and a lens before it reaches the paper. This is not a slow process that could be pushed; it is a process for which projection is not an available technique. Every print in this part is a contact print the size of its negative.

And the band matters as much as the dose. The absorption that drives the reaction lies in the near ultraviolet and the edge of the blue, which is where the sun still has useful energy and where a tungsten lamp has almost none — hence Bostick & Sullivan’s instruction that the sensitiser can be handled under ordinary incandescent lighting. Worse, the ferricyanide absorbs strongly in the same region while contributing almost nothing, so it acts as an internal filter on the coating. Part XVI’s lesson on ultraviolet sources works this through and carries the numbers: the fraction of incident light actually reaching the photoactive iron falls from 0.65 at 365 nm to 0.03 at 420 nm. A lamp peaking at 410 nm is nearly useless for cyanotype for that reason, and the quantum yield is falling across the same interval.

Where a cyanotype can work, and where its own ferricyanide takes the light

• Useful band (UVA) (320–400 nm) — absorbed fraction reaching the iron complex 0.65 at 365 nm
• Ferricyanide takes it (400–440 nm) — absorbed fraction 0.03 at 420 nm
• No useful chemistry (440–600 nm) — why the sensitiser is handled under tungsten light rather than in the dark
• The pigment absorbs here (600–750 nm) — maximum near 700 nm — the reason the finished print looks blue

Band edges are drawn to show the argument, not measured: the two absorbed-fraction figures are Ware's calculated values at 365 and 420 nm, and the 700 nm figure is the published maximum of the Prussian blue absorption. Nothing between those points was measured for this drawing. The coloured strip approximates where the visible spectrum falls and is a reading aid only; the wavelengths in the labels carry the information. The bands and curves are drawn to show the relationship, not measured.

!A quantum yield for the classic sensitiser is not published, and this page will not imply one

The 0.45 in the table above is for citratoferrate(III) in aqueous solution, from photochemical studies of the speciation. It is not a figure for a coated cyanotype sensitiser, and no source read for this course publishes one. The nearest published quantity for the coating is different in kind: Ware compares the pigment actually formed at maximum density with the sensitiser originally laid down and finds that only about 6 per cent of the sensitiser is converted into image substance. That number folds in the internal filter, the self-masking of the pigment as it forms, incomplete ligand release and the losses in the wash, so it is an efficiency for the whole process and not a quantum yield. Treat the two as answering different questions.

The second step: iron(II) meets hexacyanoferrate(III)

Nothing that follows needs light. Two iron centres in different oxidation states are put next to each other and one electron moves between them, which is the simplest reaction in the whole course.

Fe²⁺ + [Fe(CN)₆]³⁻ → Fe³⁺ + [Fe(CN)₆]⁴⁻

The oxidation states switch: the free iron(II) is oxidised, the hexacyanoferrate(III) is reduced

The products are iron(III) and hexacyanoferrate(II), and they precipitate together. Written as one step, which is how Ware writes it:

Fe²⁺ + [Fe(CN)₆]³⁻ → Fe[Fe(CN)₆]⁻

The overall reaction, whose product is Prussian blue: iron(III) hexacyanoferrate(II)

That formula unit carries a net negative charge, and the charge is balanced by a single cation — potassium from the ferricyanide, or ammonium — sitting in a cavity in the lattice. Written with the cation in place it is the compound the trade calls “soluble” Prussian blue:

K⁺ + Fe²⁺ + [Fe(CN)₆]³⁻ → KFe[Fe(CN)₆]

The same reaction with the counter-ion written in: the potassium-containing variety

⁇Prussian blue, Turnbull's blue, and why the second name survives

The disagreement, as chemistry inherited it. Two routes reach a blue precipitate. Iron(III) with hexacyanoferrate(II) gives what has always been called Prussian blue. Iron(II) with hexacyanoferrate(III) — which is the cyanotype’s route — gives what has always been called Turnbull’s blue, after the Glasgow manufacturing chemists. Because the reactants differ, the products were assumed to differ: iron(III) hexacyanoferrate(II) in one case, iron(II) hexacyanoferrate(III) in the other. Ware calls describing the cyanotype’s pigment as ferrous ferricyanide a common error in all older published accounts of the process, and notes that the misconception persists today.

What settled it, and how. Spectroscopy on the solids, not chemistry in the beaker. Ware cites Mössbauer spectroscopy — Duncan and Wigley in the Journal of the Chemical Society for 1963, and Bonnette and Allen’s isotopically labelled study in Inorganic Chemistry in 1971 — together with Wertheim and Rosencwaig’s photoelectron spectroscopy of inequivalent iron sites in The Journal of Chemical Physics, also 1971. Both techniques report on the oxidation state of each iron site directly. All forms of Prussian blue and of Turnbull’s blue turn out to be, in terms of oxidation states, iron(III) hexacyanoferrate(II): the iron(II) is low-spin and six-coordinate to carbon, the iron(III) is high-spin and coordinated to nitrogen. Ware records that radio-isotope studies further established the mechanism to be electron transfer between the two iron centres, and not the cyanide ligands switching which end they point.

Who holds what, and what the course uses. Ware states the modern position without hedging and keeps the historical name only to record which way the preparation went. He then adds the complication himself: Reguera and co-workers’ Mössbauer study of 1992 finds the two iron centres of Turnbull’s blue in a 3:2 ratio, against 1:1 for the “soluble” form and 4:3 for the “insoluble” one, which requires an extra anion in the lattice for charge neutrality. The course therefore says two things and not one. The oxidation-state question is closed — the cyanotype’s blue is iron(III) hexacyanoferrate(II), the same substance by either route, and this is established by measurement. The composition question is not — Turnbull’s blue is a defect lattice with more vacancies still, and its X-ray powder pattern resembles the “insoluble” form’s. Nothing a photographer can see distinguishes them; a spectroscopist can.

The course cites Ware’s collation of that literature and did not read the 1963, 1971 or 1992 papers directly. The chemistry is not in doubt; the citation is second-hand and is labelled so.

The iron(II) is still wearing its ligands

There is a step hidden between the photolysis and the precipitate, and it is the reason the New Cyanotype has an acid first bath while the classic process has none.

The iron(II) that light makes is not a free ion. It is still held by whatever carboxylate held its iron(III), and Ware’s evidence that this matters is a plainly reportable experiment: irradiate a trisoxalatoferrate solution with ultraviolet, then add it to a ferricyanide solution, and little or no Prussian blue forms. The carboxylate ligands have to come off before the iron can enter the lattice. What the complexed iron(II) can do is hand an electron across, which is the reaction written above, and the redox potentials favour it: +0.356 V for the hexacyanoferrate couple against +0.02 V for the iron oxalate one.

For the oxalate the release is then an acid-base matter. Hydrogen ions compete for the oxalate to make undissociated oxalic acid, and Ware multiplies out the two constants to show that the overall equilibrium lies well to the right, with K = 4.2 × 10⁴. That is what the New Cyanotype’s dilute mineral acid bath is doing, and the formulary page carries the arithmetic.

For the citrate, the same argument is an extension rather than a measurement. The iron(II) photoproduct has not been identified, so nobody can write its dissociation constant; Ware’s own statement is only that it is likely to be a citrato-complex. What the practice tells us is consistent with it: Ware notes that the tonal scale of a classic cyanotype is strengthened by developing first in 1 per cent w/v citric acid rather than in plain water, and that too strong an acid bath fogs the highlights blue. Both readings — ligand release helped by acid, and free iron(II) then reaching ferricyanide in places you did not want it — fit. The course reports the observation and marks the mechanism as inferred.

Soluble and insoluble, and which one is in your print

“Soluble” Prussian blue is the worst-named substance in this part of the course, and the misnomer is the direct cause of the classic cyanotype’s biggest defect.

Variety	Composition	Cation in the lattice?
“Soluble” Prussian blue	KFe[Fe(CN)₆], with variable water	Yes — potassium, or ammonium, in the lattice cavities
“Insoluble” Prussian blue	Fe₄[Fe(CN)₆]₃·14H₂O	No; made with iron(III) in large excess
Turnbull’s blue	Fe₃[Fe(CN)₆]₂·X, X a singly charged anion	Anion, not cation; a more defective lattice again
Prussian white	M₂Fe[Fe(CN)₆], M usually K⁺ or NH₄⁺	Two per formula unit

All of them are highly insoluble: Ware gives a solubility product of about 10⁻⁴¹, and for the “soluble” form specifically 3 × 10⁻⁴¹. So what does “soluble” mean? It means peptisation — the pigment disperses as particles fine enough to pass a filter and to make a blue liquid that looks like a solution and is not one. The tendency is associated with the cations in the lattice, and it is worst for the potassium form.

The classic cyanotype makes exactly that form, because the ferricyanide it is mixed with is the potassium salt. Ware lists the consequence as the fifth of the classic process’s seven shortcomings: a significant proportion of the image substance is peptised and washes out during wet processing, which produces “a serious loss of gradation in the high values, truncating the tonal scale, and conferring an artificially high contrast”. You can watch it happening — the wash water goes blue and keeps going blue.

Two facts about the lattice explain both the problem and its cure. The cavity has a calculated interior radius of 182 pm, against six-coordinate crystal radii of 151 pm for the ammonium ion and 152 pm for potassium, so either fits. And Ware states that replacing potassium by ammonium diminishes the tendency to peptise and gives a blue more resistant to alkali. That single substitution is the reason the New Cyanotype goes to the trouble of crystallising potassium out of its sensitiser, and it is why its exposure scale is about 2.2 where the classic process manages about 0.9.

!One disagreement between good sources, left standing

PubChem carries a water solubility of 6 mg/mL at 25 °C for Prussian blue, from HSDB and DrugBank. Ware is emphatic that every form is highly insoluble and that what looks like a solution is a peptised colloid. The Prussian blue page prints both and reconciles neither, because they may not be measuring the same thing — a filter does not distinguish a dissolved substance from a nanoparticle dispersion. What both explain is the same observation at the sink.

Why Prussian blue is blue

Most short explanations of this are wrong in an interesting way. The colour is not the cyanide, and it is not the iron: it is the pair of irons, in different oxidation states, bridged by cyanide.

One bridge in the Prussian blue lattice

1. 1Iron(II), low spin, bonded to six carbons — the electron donor in the transition; diamagnetic because cyanide is a very strong field ligand at its carbon end
2. 2Cyanide bridge, carbon to iron(II) and nitrogen to iron(III) — it conducts the electron; it is not the source of the colour
3. 3Iron(III), high spin, bonded to six nitrogens — the acceptor; nitrogen is a much weaker field donor, which is why this centre keeps five unpaired electrons
4. 4Lattice cavity with K⁺ or NH₄⁺ — present in the "soluble" variety, absent in the "insoluble" one; which cation sits here decides how readily the pigment peptises

Drawn flat for legibility. The real lattice is face-centred cubic with six bridges at every metal centre, and in the insoluble variety about one quarter of the hexacyanoferrate groups are missing, their places taken by water.

⚗The established spectroscopy, and the loose statements it replaces

Absorb a photon and one electron moves from the low-spin iron(II) to the high-spin iron(III). Because the two ends of the move are the same element in two oxidation states, the transition is called intervalence charge transfer. Ware calls Prussian blue a prototype for the whole class of mixed-oxidation-state compounds, all of them intensely coloured — the class Robin and Day surveyed in their 1967 review of mixed-valence chemistry.

What is measured. A strong, broad absorption band centred near 700 nm, in the red. What white light has left after that is taken out is blue, and intensely so: Ware takes 1.55 × 10⁴ dm³ mol⁻¹ cm⁻¹ as a working mean for the molar extinction coefficient, which is why a very small mass of pigment gives a deep tone. Mortimer and Rosseinsky propose that the position of the maximum characterises the variety — 690 nm for the “soluble” form, 730 nm for the “insoluble” — and Ware notes that different authors put the maximum at somewhat different places, which is what you would expect of a substance whose composition varies.

Why the assignment is believed rather than assumed. A second, weaker band sits near 400 nm and is the same charge transfer ending on a different orbital. The gap between the two, 10,500 cm⁻¹, is the ligand-field splitting for an iron centre surrounded by six nitrogens — a quantity that can be checked independently — and it lands where it should. Robin’s 1962 calculation puts the optical electron 99 per cent localised on the iron(II) centre in the ground state, which is what makes the transition a transfer rather than a delocalised band.

Where this comes from. The spectroscopy in this block is Ware’s Appendix II.5, which collates the primary literature — Robin’s 1962 paper on the colour and electronic configurations of Prussian blue, Robin and Day’s 1967 review, Braterman on the charge-transfer spectra of the iron-group cyanide complexes, Mortimer and Rosseinsky on the two varieties. The course read Ware and not those papers, and names them so that a reader can go further than it did.

What the course refuses. “The cyanide makes it blue” is wrong: the same cyanide ligands are on the ferricyanide, whose solution is yellow, and on the ferrocyanide, whose crystals a European food-additive specification calls lemon-yellow. “Iron compounds are blue” is wrong. Prussian white, which has the same lattice and the same cyanides but two iron(II) centres instead of one of each, is colourless — and Ware gives exactly that reason for it: there is no facile intervalence transition left to make. The mixed valence is the colour. Take it away in either direction and the colour goes with it.

Where the image actually forms

Ask a printer when the picture appears and you will get three answers, all of them partly right, because the image is built in three places.

Three stages, and what the print looks like at each

1. 1Coated and dry: yellow to yellow-greenThe Getty atlas describes the sensitised layer as yellow-greenish. Nothing has reacted; the colour is the iron(III) citrate and the ferricyanide sitting side by side in the paper
2. 2Under the lamp: it prints out, then reversesBlue appears as you watch. Ware’s description of a correct classic exposure is exact: the shadows should be somewhat reversed, giving a solarised look, the highlights distinctly green, and the image "blocked up". A print that looks right in the frame is under-exposed
3. 3In the water: the blue completes and the rest leavesThe unexposed sensitiser dissolves out of the highlights; iron(II) still in the paper meets ferricyanide it had not yet reached; and a real fraction of the pigment washes away with the rest. Twenty minutes face down for the classic process
4. 4On the line: it deepens for hoursAir oxidises the colourless Prussian white made during the over-exposure back to blue. Ware puts the intensification at several hours during drying; a bath of 0.3 per cent hydrogen peroxide does the same thing in half a minute

This is what printing-out means, and it is a different craft from anything in Parts XVIII and XIX. There is no development to time, so there is no exposure to calculate: you look at the print. The contact frame has a split back for exactly this reason.

It also means the print you judge is not the print you get. Between the frame and the dry sheet the image loses substance to the wash and gains it back by oxidation, and the two do not cancel. Ware’s own instruction — expose until it looks blocked up — is a correction for both.

Prussian white, Everitt’s salt, and the print that deepens overnight

The colourless compound in that last stage has four names, all of them in the literature: Prussian white, Berlin white, Everitt’s salt and Williamson’s salt. It is iron(II) hexacyanoferrate(II) — both irons in the same oxidation state — and its formula must carry two singly charged cations for electroneutrality: M₂Fe[Fe(CN)₆], with M usually potassium or ammonium. It has essentially the same cubic lattice as Prussian blue, with a slightly larger cell, and it is colourless for the reason given above. Everything in this paragraph is Ware’s Appendix II.9; the course found no second, independent inorganic-chemistry account of Everitt’s salt to check it against, and its structure has in any case not been fully determined, for want of single crystals.

Two routes make it in a cyanotype. Reducing conditions put it there directly: where the exposure has been heavy, most of the ferricyanide near the image has already been reduced, so the iron(II) arriving late meets hexacyanoferrate(II) rather than hexacyanoferrate(III) and precipitates white instead of blue. And over-exposure reduces blue that has already formed, which is solarisation — the reversal of the shadows that makes a correct cyanotype look wrong.

Herschel had both halves of this in 1842, in a footnote, and his language is worth keeping. He attributes the whitening “to the deoxidation of the precipitated Prussian blue and the formation of the proto-ferrocyanuret of iron; the resumption of colour in the shade, to the re-oxidizement of this compound, which is well known to absorb oxygen from the air with avidity.” Then the observation that puzzled him: “Simple Prussian blue, however, is not whitened by the violet rays. Its state must be peculiar.” Ware supplies the resolution: pure Prussian blue absorbs light and relaxes thermally, returning unchanged, but where an oxidisable impurity is trapped in intimate contact with the lattice the excited state lives long enough to accept an electron. The same photoreduction is why finished cyanotypes fade in light.

⚗Coming back: air, peroxide, and a hydroxide problem

Both re-oxidations are written the same way, and both make alkali.

4 Fe[Fe(CN)₆]²⁻ + O₂ + 2 H₂O → 4 Fe[Fe(CN)₆]⁻ + 4 OH⁻

Aerial reoxidation: what happens on the drying line over several hours

2 Fe[Fe(CN)₆]²⁻ + H₂O₂ → 2 Fe[Fe(CN)₆]⁻ + 2 OH⁻

Ware's 0.3 per cent hydrogen peroxide bath: the same change in about half a minute

Notice the hydroxide on the right of both. Ware makes the point explicitly: all these oxidations generate hydroxyl ions, tending to make the system more alkaline, and since the next section is about what alkali does to Prussian blue, a re-oxidation is best carried out under mildly acidic conditions.

Two practical notes follow. Ware states that the peroxide bath makes no difference to the final densities — it buys you the answer now rather than tomorrow, and nothing else. And under no circumstances is a dichromate bath used for this, although it has been recommended in the past: ammonium dichromate is a listed human carcinogen and Ware records the medical literature on an artist who used a dichromate bath for quilting fabrics.

Solarisation is not simply a loss, and this is the least intuitive thing about printing a cyanotype. A printed-out blue masks itself: pigment already formed absorbs the light that would make more. When the shadows reverse to white, that masking is reduced, more light gets in, more product forms, and after re-oxidation the final density can be higher than it would otherwise have been. Which is why the correct exposure looks over-exposed.

There are two published mechanisms for the reversal and Ware weighs them, preferring Murray’s — that excess iron(II) complex reduces the blue already present, supported by the redox potentials, +0.46 V for the blue/white couple against +0.02 V for the iron complex — over Suzuki’s direct photoreduction of the ferricyanide, whose quantum yield of about 0.01 is far too low to account for what is seen. He then adds a hypothesis of his own from his own experiments, which the New Cyanotype page sets out in full and labels as a hypothesis. For a printer the upshot is the same either way: the reversal is a reduction, not a loss, and it comes back.

The wash is chemistry, not rinsing

Nothing in a cyanotype is fixed. The water bath does three separate jobs, and confusing them is the source of half the faults in the part.

It removes what light did not use. Unreacted iron(III) citrate and unreacted ferricyanide are both freely soluble and both must leave. If they do not, the highlights hold a yellow-brown stain — the colour of the sensitiser itself — and, worse, the print is still a sensitised material: iron and ferricyanide left in the paper will go on reacting and go on responding to light. Ware’s washing instruction for the classic process is twenty minutes, face down, in gently running water or several changes, until the yellow has entirely gone from the highlights.

It completes the blue. Iron(II) made by the light has not necessarily met all the ferricyanide it can reach; the wet paper lets both move. The Getty atlas describes the bath as doing both jobs at once, which is precisely the awkwardness of it.

And it takes some of the image away. This is the part that is not rinsing and not a fault of technique. Ware’s words are that there will be “considerable leaching-out of blue pigment”, and that the extended washing needed to clear the paper will significantly weaken the image and truncate the tonal scale in the high values. It costs the light tones proportionally more than the shadows, because a thin deposit loses a larger fraction of itself, and that is what a short exposure scale is made of.

A slightly acidified first bath helps, and Ware records the practice: develop first in a very dilute weak acid, 1 per cent w/v citric acid rather than plain water, and the tonal scale is strengthened. The mechanism is the ligand-release argument above. His warning in the same sentence is as important as the practice: too strong an acid bath fogs the highlights blue and degrades the paper white. And the Prussian blue page carries the outer limit — the pigment is stable to acids of moderate strength down to about pH 1 or 2, below which there is a risk of releasing hydrogen cyanide.

⚠Alkaline water is the one thing that destroys the image outright

Prussian blue is decomposed by alkali at pH 10 and above, and by less than that given time. The hydrolysis takes the iron(III) out of the lattice and puts the hexacyanoferrate into solution:

Fe[Fe(CN)₆]⁻ + 3 OH⁻ → Fe(OH)₃ + [Fe(CN)₆]⁴⁻

Alkaline hydrolysis: the blue is gone and a yellow-brown iron(III) hydrous oxide is left behind

The numbers are worse than they look. Ware reports Holtzman’s finding that a buffer at pH 9.4 completely decolourises Prussian blue by irreversible hydrolysis in one to ten minutes, depending on how the pigment was made — and pH 9.4 is the pH of a saturated solution of calcium carbonate, which is the buffer routinely put into archival paper and board. A 0.25 molar sodium carbonate solution at about pH 10.7 destroys the blue of a cyanotype in less than half a minute.

Why “irreversible” is the operative word. Freshly precipitated hydrous iron(III) oxide can be redissolved by dilute acid, and there is some evidence that a freshly bleached cyanotype can be partly restored with dilute acid and added ferrocyanide. But over a few days the hydrous oxide transforms to crystalline forms such as goethite, which dilute acid will not touch, and after that the image cannot be regenerated. Ware calls this an area needing further research; the practical rule is that you have hours, not weeks.

What to do about it. Test your tap water. Where it is alkaline or hard, wash in water acidified with a little citric acid; use unbuffered cotton paper, never a buffered “archival” sheet; and mount and store the finished print in unbuffered board, which is the opposite of the advice every other process in this course gives. Part XX hands that compatibility rule forward, and the negatives and papers lesson carries the paper arithmetic.

The cyanide question, answered properly

A cyanotype sensitiser contains six cyanide groups per hexacyanoferrate ion and the finished pigment contains eighteen per formula unit. That alarms people; it deserves an answer rather than a wave of the hand, and the answer has two halves, both of which have to be given.

The first half: this is complexed cyanide, not free cyanide. The cyanide ions are bound to iron in a low-spin complex of extreme stability, which is why the whole family behaves nothing like a soluble cyanide salt. Photographers’ Formulary put it in one sentence on their own kit instructions: the cyanide groups are chemically bound to the iron atom and are not free to act as a poison. The Environment Agency’s WM3 says the same thing in regulatory language, assigning the hazard statement for liberating a very toxic gas on contact with acid to cyanide salts with the explicit exception of complex cyanides such as ferrocyanides and ferricyanides. Ware adds the most striking evidence: Prussian blue is administered internally as an antidote to thallium and radiocaesium poisoning, because its lattice traps those cations and gives up nothing itself.

The second half: there are conditions under which hydrogen cyanide is released, and they are specific. They are worth learning exactly rather than vaguely, because a vague version leads either to panic or to carelessness.

Condition	What the sources say	Does cyanotype practice meet it?
Strong heat on the solid	The international chemical safety card for potassium ferricyanide, ICSC 1132 of August 2002, states that the substance decomposes on heating and on contact with acids, producing toxic gases including hydrogen cyanide, and directs that it be stored separated from acids	No. Nothing in the process is heated
Concentrated acid	Ware’s own hazard entry: incompatible with concentrated acids, which may release hydrogen cyanide gas. He qualifies the frequently quoted warning in his environmental chapter — the heating has to be very strong and the acid very concentrated to get much hydrogen cyanide, and neither is done in cyanotype	No, for the published processes. The clearing acids are dilute and weak
Prolonged ultraviolet, on hexacyanoferrate in solution or in a wet coating	Ware’s Appendix III.7 gives the established first step of hexacyanoferrate(III) photolysis as a photoaquation that releases hydrogen cyanide directly, and records that the pH rises during irradiation because hydrocyanic acid is weak	Marginally, and it is measurable in one case only — see below

[Fe(CN)₆]³⁻ + 2 H₂O + UV → [Fe(CN)₅(H₂O)]²⁻ + HCN + OH⁻

Ware's first step of hexacyanoferrate(III) photolysis: one cyanide leaves as hydrogen cyanide and water takes its place

That reaction is real, and it is why the third row is not simply “no”. But its quantum yield is of the order of 0.01 against about 0.45 for the citrate complex — per absorbed photon, roughly forty times less productive — so in a normal cyanotype the iron does nearly all of the photochemistry and does it first. Ware states the practical outcome without softening it or exaggerating it: the traditional process, using only ammonium iron(III) citrate and potassium ferricyanide, will not emit any significant, perceptible or harmful amount of hydrogen cyanide during a normal exposure and processing.

The exception proves the mechanism. Herschel’s proto-cyanotype — paper coated with potassium ferricyanide alone, which is the whole of the sensitiser — needs an exposure twenty to fifty times longer, hours instead of minutes, and Ware records that in that case a perceptible odour of hydrogen cyanide, and possibly of cyanogen, can be detected when the printing frame is opened. He adds that the quantity is not life-threatening and that there is no particular reason to make cyanotypes that way. The Herschel 1842 page carries the process; this is why the course teaches it as history.

⚠What the course does with all that, and the classifications behind it

The rule is simple and it is stricter than the evidence strictly requires. No concentrated acid, no darkroom stop bath — which is a far stronger acid than the 1 per cent w/v citric acid first bath described above — no hot solution and no acidified waste bottle ever goes near a hexacyanoferrate tray, bottle or spill. The reason for taking the stricter reading is that the cost of being wrong is asymmetric: the downside of an unnecessary rule is inconvenience, and the downside of a necessary one broken is a toxic gas.

Potassium ferricyanide, GHS classification aggregated by PubChem from the ECHA C&L Inventory and verified for this course on 4 September 2026: signal word Warning, with the health-hazard, irritant and environmental pictograms. There is no harmonised entry. Across 285 reports from 12 notifications, 271 carry hazard codes and 14 — 4.9 per cent — say it meets no criteria at all; among those that classify it, serious eye irritation appears in 62.8 per cent, long-lasting aquatic toxicity in 45.3, harmful if swallowed in 25.3, and suspected reproductive toxicity in 14.7. The course classifies the substance as handled at Level B; the reasoning is on its encyclopaedia page.

Ammonium iron(III) citrate, same source and same verification date: signal word Warning, irritant pictogram, causes skin irritation and serious eye irritation. Aggregated from 1,576 reports across 11 notifications, of which 161 — 10.2 per cent — decline to classify it at all. That one-in-ten dissent is unusually large and is consistent with a substance of variable composition being classified by notifiers who are not describing the same material. The course classifies it Level A.

And two things not to over-read. HSE’s EH40 lists cyanides other than hydrogen cyanide, cyanogen and cyanogen chloride at 5 mg/m³ as CN with a skin notation, but does not name the hexacyanoferrates, and this course could not verify whether that entry is intended to reach a complexed cyanide. EH40’s own introduction states that absence from the list does not indicate that a substance is safe. Your supplier’s current safety data sheet is the authority for the material in your jar, not this page.

What sets maximum density and contrast

Four quantities decide how dark a cyanotype can go and how long a tonal scale it can render, and the experiment two pages on measures three of them for your own materials. They are worth naming here because each one is a different lever.

How much iron(II) the light can make per unit area. This is the exposure, bounded by the sensitiser laid down. Ware’s figures for a typical coating: a well-sized paper takes up about 25 cm³ of sensitiser per square metre, at roughly 0.3 mol/L, giving 7.5 × 10⁻³ mol/m² of iron on the sheet. There is no more iron than that, whatever you do with the lamp.

How much of it becomes pigment. Far less than you would hope. From the extinction coefficient, a maximum density near 1.55 needs a pigment coating weight of about 5 × 10⁻⁴ mol/m² — which, set against the sensitiser laid down, means only about 6 per cent of the sensitiser is converted into image substance even at maximum density. Some of the loss is the internal filter, some is the pigment masking itself as it forms, some is ligands not released, and some is the wash.

How much hexacyanoferrate is there to react with the iron(II). In the classic sensitiser, more iron than ferricyanide: the arithmetic on the classic sensitiser page — the course’s own, from Ware’s figures — puts it between about 1.7 and 2.1 moles of iron per mole of ferricyanide, with the range coming from the unknown iron content of the citrate. The electron transfer is one-to-one, so there is surplus iron(II) with nowhere to go, and it diffuses. That is bleeding: heavily exposed areas staining the highlight beside them blue. Ware lists it as the sixth shortcoming of the classic process and says that the oxalate can be mixed in chemically equivalent amounts to suppress it.

And how much survives the wash. For the classic process this is the dominant term, and it is peptisation. It takes proportionally more from the light tones than the dark ones, so it does not merely weaken the print, it shortens the scale: the top of the curve collapses towards the paper white and what is left is short and contrasty. Hence the numbers you will meet as the part goes on.

Formulation	Exposure scale	Maximum density	What it can print from
Classic cyanotype	about 0.9 — three stops	not published for this formulation; Ware’s worked example takes 1.55 as typical of a cyanotype	A short-scale negative, comparable with what a grade 3 enlarging paper wants
New Cyanotype	about 2.2	Ware calls it “verging on black”, and gives 1.7	The long-scale negative the rest of the alternative-process cluster needs
Simple cyanotype	2.7 down to 1.4, set by how much ammonia goes into the sensitiser	about 1.5	Anything in that range — the one cyanotype whose contrast is chosen at mixing rather than at coating

Read that table as the argument of the whole part in miniature. The classic process is not contrasty because Prussian blue is contrasty, and not because of the ferricyanide. It is contrasty because its image substance is the peptisable potassium form and a lot of the highlight detail goes down the sink. Change the cation and the scale trebles; control the speciation at mixing, as the Simple cyanotype does, and the scale becomes a variable you set rather than one you are given.

Alternative route

This page is a reading lesson and none of it is blocked by having no equipment. The one thing it invites you to do — repeat Herschel’s Article 210 separation, coating a strip with the iron solution alone, exposing it briefly, and raising the blue afterwards in the shade — needs an ultraviolet source, which is why the page is marked as requiring one and why this section is here.

Sunlight is the honest alternative, and for this demonstration it is the better one. Herschel used it and needed four or five seconds. Ware’s arithmetic for average sunshine gives an ultraviolet irradiance of 30 to 40 W/m² and full density in 2 to 4 minutes, so a few seconds for a faint latent change is the right order. Nothing about the chemistry changes; what you lose is control of the source, and the substitute is a written record — date, time, cloud, the appearance at each inspection. Part XVI’s ultraviolet lesson makes the argument in full and applies to sun and lamp alike.

If you have neither sun nor lamp, there is no substitute for the exposure and this page does not pretend otherwise: skip the demonstration and read on. Everything the demonstration shows — that the photochemistry belongs to the iron and that the ferricyanide is a precipitant — is established here from Herschel’s own account and from Ware’s quantum yields, and no later page in this part depends on your having run it.

The hazard has no alternative route either. Ultraviolet is an eye and skin hazard whether it comes from a lamp or from the sky, and a reader printing outdoors has an unenclosed source that cannot be interlocked. Distance, sleeves, a hat and not looking at the source are administrative controls, which are the weakest kind; that is the honest description of printing in sunlight, and it is worth knowing that is what you are doing.

Recap

A cyanotype sensitiser is an iron(III) ion held by a carboxylate ligand, and the ligand is the electron donor. Near-ultraviolet light drives a ligand-to-metal charge transfer: the iron ends as iron(II) and the ligand fragments, losing carbon dioxide. For the oxalate that reaction is written and balanced, with a quantum yield of about 1.2 because one photon reduces two irons; for the citrate the organic product is known, the iron product is not, and the course prints a scheme rather than an equation. Nothing amplifies, which is why the material sits near 10⁻⁵ ISO and why no enlarger and no camera will print it.

The second step needs no light at all. Iron(II) hands an electron to hexacyanoferrate(III), and iron(III) and hexacyanoferrate(II) precipitate together as Prussian blue — the same compound whichever way round you make it, established by Mössbauer and photoelectron spectroscopy, with the name Turnbull’s blue kept only to record the route. Its colour is intervalence charge transfer between the low-spin iron(II) and the high-spin iron(III) across a cyanide bridge, a strong band near 700 nm in the red; remove the mixed valence, as reduction to Prussian white does, and the colour goes.

Everything that then goes wrong is a consequence of one lattice property. The potassium-containing variety the classic process makes peptises, so the wash that clears the highlights also carries image down the sink, which is where its exposure scale of 0.9 comes from. Alkali hydrolyses it outright, which is why an unbuffered paper and an unbuffered mount are chemical requirements rather than preferences. And the cyanide in it is complexed, not free — which is why the process is workable at a bench, and why the rule about acid is absolute anyway.

Three things this page could not establish, and says so rather than filling in: no quantum yield is published for the classic cyanotype sensitiser as coated, only for citratoferrate(III) in solution; the identity of the citrate photoproduct is unknown, so its ligand-release chemistry is inferred from the oxalate rather than measured; and the composition of Turnbull’s blue is still open at the level of lattice vacancies, though its oxidation states are not.

Check your understanding

Question 1. Paper is coated with ammonium iron(III) citrate alone, with no ferricyanide, held a few seconds in the sun, then taken into the shade and brushed with potassium ferricyanide solution. What happens?

• Nothing, because the sensitiser was incomplete during the exposure
• A blue image appears where the sun fell and nowhere else
• A uniform blue appears everywhere, because the ferricyanide is the light-sensitive component
• A yellow stain appears where the sun fell, which turns blue only after washing

Show the answer and why

Answer: A blue image appears where the sun fell and nowhere else

This is Herschel’s Article 210, and it is the experiment that decides which component is photosensitive. The light reduces iron(III) to iron(II) in the citrate complex; the ferricyanide, applied afterwards in the shade, is a precipitant and nothing more. Herschel’s own conclusion is that the photographic influence is confined to the first stage.

Question 2. Why does the course publish a balanced equation for the photolysis of trisoxalatoferrate(III) but only a scheme for the citrate?

• The citrate reaction is too fast to follow experimentally
• The identity of the iron(II) photoproduct of the citrate system is not established, so a balanced equation would require inventing a species
• The citrate reaction does not conserve charge, so no equation can be written for it
• Citrate is a mixture, so its reactions are always written as schemes

Show the answer and why

Answer: The identity of the iron(II) photoproduct of the citrate system is not established, so a balanced equation would require inventing a species

The organic side is known - acetone dicarboxylic acid, which may decarboxylate on to acetone - but Ware writes the iron product simply as an iron(II) with no ligands and no charge, because nobody has established what it is. Rule 1 forbids closing the balance by invention, so the course reproduces the scheme and labels the gap.

Question 3. A cyanotype is washed in tap water that turns out to be alkaline, and the blue visibly weakens. Which statements are supported by the sources this page uses? Select all that apply.

• A buffer at pH 9.4 can destroy Prussian blue completely in one to ten minutes
• The damage is a hydrolysis, which strips the iron(III) out of the lattice and puts hexacyanoferrate into solution
• The damage is a reduction, and the colour will return on drying in air
• Acidifying at once may partly restore a freshly bleached image, but the hydrous iron oxide becomes insoluble over a few days

Show the answer and why

Answer: A buffer at pH 9.4 can destroy Prussian blue completely in one to ten minutes, The damage is a hydrolysis, which strips the iron(III) out of the lattice and puts hexacyanoferrate into solution, Acidifying at once may partly restore a freshly bleached image, but the hydrous iron oxide becomes insoluble over a few days

The third is the trap: reduction to Prussian white is reversible and does come back in air, but alkali is a different reaction with a different product. Ware reports Holtzman’s pH 9.4 result, gives the hydrolysis equation, and notes that the hydrous iron(III) oxide transforms over days to forms such as goethite that dilute acid cannot redissolve.

Question 4. Prussian white has the same lattice, the same cyanide bridges and the same two iron atoms per formula unit as Prussian blue. Why is it colourless?

• Its particles are too small to scatter visible light
• Both irons are in the same oxidation state, so there is no intervalence charge transfer to absorb in the red
• The cyanide ligands have been displaced by water
• It absorbs in the ultraviolet instead, which the eye cannot see

Show the answer and why

Answer: Both irons are in the same oxidation state, so there is no intervalence charge transfer to absorb in the red

The colour of Prussian blue is an electron moving from a low-spin iron(II) to a high-spin iron(III) across a cyanide bridge, absorbing near 700 nm. Prussian white is iron(II) hexacyanoferrate(II): both centres are iron(II), so there is no such transition left to make. Oxidise it and the colour returns, which is exactly what happens on the drying line.

Question 5. The classic cyanotype has an exposure scale of about 0.9 while the New Cyanotype reaches about 2.2. What is the main reason?

• Prussian blue formed from oxalate is a different, longer-scale pigment
• The New Cyanotype uses far more ferricyanide, so more pigment can form
• The classic process makes the potassium-containing variety of Prussian blue, which peptises and washes out of the light tones
• The classic sensitiser absorbs less ultraviolet, so its highlights are under-exposed

Show the answer and why

Answer: The classic process makes the potassium-containing variety of Prussian blue, which peptises and washes out of the light tones

The image substance is the same compound in both. What differs is which cation sits in the lattice cavities: the potassium form peptises readily, and the wash that clears the highlights carries a disproportionate share of the thin deposits away, collapsing the top of the curve. Replacing potassium by ammonium diminishes the tendency to peptise, which is the whole reason the New Cyanotype crystallises potassium out of its sensitiser.

Question 6. Which single statement about hydrogen cyanide and cyanotype chemistry is accurate as the sources give it?

• Hexacyanoferrates release hydrogen cyanide continuously in water, which is why gloves are worn
• Hexacyanoferrates cannot release hydrogen cyanide under any conditions, because the cyanide is complexed
• Release needs very strong heating or very concentrated acid, and the photolysis of hexacyanoferrate under prolonged ultraviolet also liberates some, which is detectable in a ferricyanide-only coating but not in a normal cyanotype
• The risk comes from Prussian blue itself, which decomposes to hydrogen cyanide as it dries

Show the answer and why

Answer: Release needs very strong heating or very concentrated acid, and the photolysis of hexacyanoferrate under prolonged ultraviolet also liberates some, which is detectable in a ferricyanide-only coating but not in a normal cyanotype

Both extremes are wrong and the middle is specific. Ware qualifies the frequently quoted warning - the heat has to be very strong and the acid very concentrated, and neither is done in cyanotype - while his own Appendix III.7 gives the photoaquation that releases hydrogen cyanide directly. He records the odour on opening a frame after a proto-cyanotype exposure of hours with ferricyanide alone, and states that a normal cyanotype exposure and processing will not emit any significant, perceptible or harmful amount.

Check my answersClear