fading and colour change of prussian blue: methods of manufacture and the influence of extenders

NationalGallery Technical BulletinVolume 25, 2004

National Gallery CompanyLondon

Distributed by Yale University Press

This volume of the Technical Bulletin is published with thegenerous support of the Samuel H. Kress Foundation and theAmerican Friends of the National Gallery, London, Inc.

Series editor Ashok Roy

National Gallery Company Limited 2004

All rights reserved. No part of this publication may be transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without the prior permission in writing of the publisher.

First published in Great Britain in 2004 by National Gallery Company LimitedSt Vincent House, 30 Orange StreetLondon wc2h 7hh

www.nationalgallery.co.uk

British Library Cataloguing in Publication DataA catalogue record for this journal is available from the British Library.

isbn 1 85709 320 8issn 0140 7430525045

Senior editor Jan GreenProject manager Tom WindrossEditor Diana DaviesDesigner Tim HarveyPicture research Kim Klehmet Production Jane Hyne and Penny Le Tissier

Printed in Italy by Conti Tipocolor

front coverRaphael, An Allegory (Vision of a Knight) (NG 213), detail of plate 13, page 16.

title pageNicolas Lancret, The Four Times of Day: Morning(NG 5867), detail of plate 1, page 49.

A note on the reproductions

The reproductions of complete paintings from the NationalGallerys collection in this book have been printed fromcolour-correct, high-resolution digital scans made with the MARC II Camera. This process was described in TheMARC II Camera and the Scanning Initiative at the NationalGallery, National Gallery Technical Bulletin, 23, 2002, pp. 7682.

Infrared examinations were performed by Rachel Billinge,Rausing Research Associate in the Conservation Department.Infrared reflectography was carried out using a HamamatsuC2400 camera with an N2606 series infrared vidicon tube.The camera is fitted with a 36mm lens to which a Kodak87A Wratten filter has been attached to exclude visiblelight. The infrared reflectogram mosaics were assembledusing Vips-ip software.

For further information about the software see the Vipswebsite at www.vips.ecs.soton.ac.uk

Prussian blue can be described as the first ofthe modern, totally synthetic pigments, arisingonly as the result of a deliberate chemical reaction;it has no natural equivalent. Its first preparation inaround 1704, by a Berlin colourmaker namedDiesbach, appears to have been accidental (theintended product was a cochineal lake),1 but it isclear that the pigment went into production quitesoon afterwards: it was first announced anony-

mously in 1710 and the earliest occurrences datefrom a few years later.2

The benefits of an intensely coloured inexpen-sive blue pigment, flexible in its handling propertiesand not gritty or difficult to use, were obvious. Itcould easily be manufactured and its supply was notdependent on the vagaries of transcontinental tradeor environmental conditions. The use of Prussianblue spread extremely rapidly, in easel painting,

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 25 | 73

Fading and Colour Change of Prussian Blue: Methods of Manufacture and the Influence of Extenders

jo kirby and david saunders

plate 1 ThomasGainsborough, Mrs Siddons(NG 683), 1785. Canvas,126.4 99.7 cm.

watercolour and for interior decoration and, eventhough doubts about its permanence were alreadybeing expressed by the mid-eighteenth century, itspopularity did not diminish: effectively, there wereno good alternatives until the syntheses of cobaltblue in 1802 and artificial ultramarine in 18267. Itis the pigment of cool, white and blue decorativeschemes, seen in the feathery touches of Watteau orGainsborough, the bravura strokes of Tiepolo, orthe calm, timeless skies of Canaletto. Mixed withwhite, its vibrancy of colour surpasses that of theinsoluble natural dyestuff indigo, the only otherblue pigment available at that time with a similarlyhigh tinting strength. Prussian blue may have agreenish undertone, but this is not always so: thepigment used by Sir Joshua Reynolds for the count-esss dress in Anne, 2nd Countess of Albemarle(NG 1259, painted in 1760), and by ThomasGainsborough in Mrs Siddons (NG 683, dated 1785;

plate 1), is of very good colour. The extraordinarilyhigh tinting strength of Prussian blue is shown inplate 2, which illustrates a sample of a modern,high-quality pigment (sample ALP in table 1, pp.923) mixed with different proportions by weight oflead white. Only one part of Prussian blue mixedwith 200, or even 400, parts of lead white gives lightblue tints very similar to those used for paintingskies. fig. 1 presents reflectance curves for seven ofthe mixtures illustrated in plate 2, showing thecharacteristic maxima for Prussian blue in the 455to 475 nm region.

Indigo and Prussian blue have other features incommon which would have made the properties ofthe new pigment relatively familiar to the painter. Inoil medium, both need to be ground with a largeamount of oil to give a workable paint, and bothcan be used as glazing pigments as they have a lowrefractive index. Generally, however, both are usedmixed with white as their colour is so deep at fullstrength that it is almost black. The blue pigmentthen benefits from the easy handling properties oflead white to allow painterly scumbles and impastopassages to be achieved. Relatively soon after theintroduction of the pigment, however, it wassuspected that the admixture of so much white hadundesirable consequences. Indigo used as an oilpaint already had a poor reputation as far as itslightfastness was concerned;3 when the permanenceof Prussian blue also came to be questioned, thepresence of white pigment or extender was recog-nised as a contributory factor.4

Jo Kirby and David Saunders

74 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 25

plate 2 Prussian blue (Aldrich Chemical Company,FeIII4[FeII(CN)6]3xH2O) mixed with lead white in theproportions 1:5, 1:10, 1:20, 1:40, 1:50, 1:100, 1:200 and1:400 (w/w).

0

10

20

30

40

50

60

70

80

90

100

400 450 500 550 600 650 700 750

Wavelength / nm

Ref

lect

ance

/ %

1:400

1:200

1:100

1:50

1:20

1:10

1:5

Prussian blue: white dilution

fig. 1 Reflectance spectra for Prussian blue / lead whitemixtures.

Factors that may influence colour change inPrussian blue

An earlier article in the Technical Bulletin (1993)gave examples of eighteenth-century paintings inthe National Gallery collection in which the fadingof Prussian blue had been observed, and discussedeighteenth- and early nineteenth-century reports ofthe instability of the pigment to light and thefactors that might influence this.5 For conveniencethese findings are summarised below. The presentaccount describes the experimental investigation ofthese factors, particularly the method of prepara-tion and the presence of white pigment or extender,based on the examination of samples of late eighteenth- and early nineteenth-century pigmentscollected from various sources and of pigmentsprepared in the laboratory.

The original method for preparing Prussian blue,first published in 1724, used dried cattle blood, analkali derived from the detonation of potassiumhydrogen tartrate (potassium hydrogen 2,3-dihy-droxybutanedioate) and potassium nitrate, greenvitriol (iron(II) sulphate) and alum (in this case,probably the potassium salt, aluminium potassiumsulphate). The resulting pale greenish-blue productwas treated with hydrochloric acid to give the deepblue pigment.6 Neither the underlying chemicalreaction producing the pigment, nor the nature ofthe cyano or hexacyanoferrate groups (CN orFe(CN)6) present in the blood, which are essentialparts of the complex, was known. The process wastherefore very difficult to control. As the chemistryof the preparation became better understood,conditions such as the proportions of the startingreagents, their dilution and the temperature atwhich the reaction took place could be varied, withthe intention of giving a product with a particularcolour or range of properties, or of improving theefficiency of the process. By the end of the eigh-teenth century, many recipes for making the pigmentfrom the essential starting materials of animalmatter, an iron salt and alkali (usually potassium-containing) had been published (see table 2, pp. 945,for a selection). With increasing knowledge, theprocedure could be refined and methods standardised.

Even modern methods of manufacture ofPrussian blue are not, however, without problems;the extremely fine particle size and colloidal natureof the pigment make separation of the product fromadsorbed water or impurities, including excess start-ing reagents or contaminants produced during thereaction, particularly hard. These may influence the

permanence of the pigment. Differences in workingproperties and texture depend very much on thetreatment of the precipitated particles, which areonly 0.010.2 m in diameter and, after the pigmenthas been washed, tend to agglomerate immediately.7

Already in the eighteenth century, insufficient wash-ing was known to contribute to hardness of thepigment and other undesirable properties.8 The tiny,colloidal particles of one variety of potassium-containing Prussian blue are particularly easilydispersed (peptised) in aqueous solutions, giving ablue suspension resembling a solution. This varietyis described as soluble and it has long beensuspected of being less permanent than other vari-eties.9 The lattice structures of the two forms of thepigment (insoluble and soluble), with or withoutalkali metal, discussed below, are slightly differentand this factor should also be considered wheninvestigating the permanence. In fact manycommercially available varieties of the pigmentcontain a variable amount of alkali metal, whichuntil after the First World War was most likely tohave been potassium. Sodium salts could also havebeen used, but they tend to give a pigment of infe-rior colour. After the First World War, the shortageof potassium stimulated the use of ammonium saltsin the production of the pigment, and most modernmethods for the manufacture of Prussian blue yieldthe ammonium-containing soluble variety, whichhas a brighter colour and better alkali resistance.10

The deliberate addition of white pigment hasbeen discussed above. The lightfastness of a typicalmodern Prussian blue pigment has been describedas excellent when the proportion of blue pigment ishigh (90% Prussian blue, 10% titanium white),decreasing as the proportion of white pigment isincreased to give very poor durability at 1% blue,99% white.11 The white pigment in these experi-ments is different from the lead white that wouldhave been used in the eighteenth century. By thenineteenth century, zinc white and other whites alsobecame available, all of which have slightly differentoptical properties (for example, as far as the absorp-tion or reflectance of radiation in the ultravioletregion of the spectrum is concerned),12 so it isconceivable that they may affect the lightfastness ofPrussian blue in different ways.

In addition to admixed white pigment, variableamounts of different inert white extenders may bepresent. As well as reducing the intensity of colourof the pigment, they also influence other propertiessuch as the softness or ease of grinding; from thispoint of view their inclusion may be beneficial. The



original method of preparation resulted in the pres-ence of hydrated alumina extender in the pigment,as alum (aluminium potassium sulphate) was one ofthe ingredients. Alumina may still be added nowa-days during the manufacturing process to help thepigment settle, quite apart from its properties as anextender. Other extenders such as starch, calcite orchalk, gypsum and barytes (barium sulphate) couldalso be used. As with the addition of white pigment,the presence of extenders may influence the light-fastness of the Prussian blue and the addition ofwhite pigment to an already extended blue couldincrease its impermanence.

From the middle of the eighteenth centuryonwards it was noted that the change in colour, orfading, of Prussian blue was sometimes reversible. Areversible reduction-oxidation process during whicha form of the pigment is produced is the basis of theproduction of blueprints and cyanotypes.13 It is alsoreported that Prussian blues tend to lose theircolour on storage in a closed container by a processof reduction, due to the lack of oxygen; this isencouraged by the presence of acidity in thesurrounding medium.14 They also tend to bereduced in the presence of oxidisable media.15 Thesechanges are reversed when oxygen (from the air, forexample) is supplied.16 The phenomenon has indeedbeen reported in painted samples of the pigment,17

but not particularly frequently, and the role of thepaint medium has not been investigated previously.

The structure of Prussian blue and the chemistry of its manufacture

Structure

Prussian blue is perhaps the best known of the classof materials described as type II mixed-valence tran-sition metal complexes.18 It is a hydrated iron(III)hexacyanoferrate(II) complex of variable composi-tion: depending on the method of manufacture andthe conditions of precipitation, potassium, sodiumor ammonium ions and a variable amount of watermay be present. The formula for the alkali metal-free (insoluble) pigment is FeIII4[FeII(CN)6]3xH2O(iron(III) hexacyanoferrate(II), where x = 1416). X-ray diffraction and neutron scattering studies haveshown that it has a face-centred cubic lattice struc-ture, where the metal sites are occupied by iron(II)and iron(III) alternately, bridged by the cyanogroups in a regular array to give a three-dimensionalpolymeric network. A quarter of the FeII, C and Nsites are vacant. Mssbauer and other spectroscopic

studies have revealed that low-spin iron(II) is co-ordinated octahedrally to the carbon atoms of sixcyano groups, while high-spin iron(III) is co-ordinatedoctahedrally to the nitrogen atoms, and also to theoxygen of co-ordinated water molecules occupyingthe vacant N sites; the average composition of thismixed environment is FeIIIN4.5O1.5. For the idealstoichiometry FeIII4[FeII(CN)6]314H2O, a unit cellwould, therefore, contain six co-ordinated and eightuncoordinated water molecules, the latter occupy-ing zeolitic-type sites within the cell structure.19

The formula for the potassium-containing(soluble) variety, which has a similar lattice struc-ture, may be written as KFeIII[FeII(CN)6]xH2O,where x is variable, but has been reported to be 5 insome syntheses.20 In this structure, as all the FeII, Cand N sites are occupied, only uncoordinated watercan be present. The potassium ion, K+, acts as acounter ion to complete the neutralisation of thenegative charge of the [FeII(CN)6]4- group, andpotassium ions occupy half the zeolitic-type siteswithin the unit cell taken up by water in the potas-sium-free form.21

The essential structural element of the pigmentframework thus consists of the sequenceFeIIINCFeII and its intense colour arises from anelectronic transition (charge transfer transition)between the two valence states of iron: from the FeII

ion to the FeIII ion, which occurs when light in thered region of the visible spectrum, between 690 and730 nm, depending on the structure, is absorbed.22

Chemistry of manufacture

The methods of making Prussian blue fall into twotypes, direct and indirect. In the direct method,which is a one-step process, a solution of aniron(III) salt, such as ferric chloride, FeIIICl3, ismixed with a solution of a hexacyanoferrate(II) salt,for example K4[FeII(CN)6]. Alternatively, solutionsof an iron(II) salt, such as ferrous sulphate, FeIISO4,and a hexacyanoferrate(III) salt, such as potassiumferricyanide, K3[FeIII(CN)6], are mixed. If the ironsalts are present in excess, the deep blue precipitateformed is a variety of the so-called insolublepigment. It was originally thought that the productof the second reaction, which produces a pigmentknown as Turnbulls blue, was an iron(II) hexa-cyanoferrate(III), but studies by X-ray diffractionand Mssbauer spectroscopy have shown that bothreactions produce iron(III) hexacyanoferrate(II).The molecule contains very little, if any, potassium(although rapid precipitation may cause some to be



brought down as an impurity), but does containwater. If the reagents are mixed in 1:1 molarproportions, or with the hexacyanoferrate in excess,the product is again identical in both cases, beingthe potassium-containing soluble variety men-tioned above, although it may be called solublePrussian blue or soluble Turnbulls blue dependingon the starting materials.23

There are thus four direct preparations, produc-ing either insoluble or soluble Prussian blue orTurnbulls blue. The production of insolublePrussian blue can be summarised as:

4FeIIICl3 + 3K4[FeII(CN)6] FeIII4[FeII(CN)6]3 + 12KCl

and that for preparing soluble Turnbulls blue as:

FeIICl2 + K3[FeIII(CN)6] KFeIII[FeII(CN)6] + 2KCl

The indirect method is a two-stage process, basedon the reaction between an iron(II) salt and a ferrocyanide salt, typically potassium hexacyano-ferrate(II), to give an insoluble white product knownas Berlin white (an iron(II) hexacyanoferrate(II),M2FeII[FeII(CN)6], where M may be ammonium,sodium or potassium, depending on the reagentsused). This product is then treated with a powerfuloxidising agent such as a chlorate or chromate togive a soluble Prussian blue.24 The production ofPrussian blue from iron(II) chloride and ammoniumhexacyanoferrate(II), using hydrogen peroxide asthe oxidant, can be summarised as follows:

FeIICl2 + (NH4)4[FeII(CN)6] FeII(NH4)2[FeII(CN)6] + 2NH4Cl

2FeII(NH4)2[FeII(CN)6] + H2O2 2FeIIINH4[FeII(CN)6] + 2NH3 + 2H2O

The preparation of Prussian blue samples

In order to study the behaviour and characteristicsof Prussian blue observed in eighteenth- and nine-teenth-century paintings in the National Gallerycollection in the context of how the pigment wasprepared at the time, it was necessary to havesamples of eighteenth- and nineteenth-centurypigments for comparison and to study their suscep-tibility to light. These are listed in table 1 anddiscussed below. Unfortunately, although a numberof early and mid-nineteenth-century samples couldbe obtained, very few samples from the eighteenth

century were available. The Prussian blue identifiedin earlier eighteenth-century pictures, such asGainsboroughs Gainsboroughs Forest (CornardWood) (NG 925) of about 1748, or CanalettosVenice: Campo San Vidal and Santa Maria dellaCarit (The Stonemasons Yard) (NG 127), paintedbetween 1726 and 1728 (both discussed in the earlierTechnical Bulletin article),25 is likely to have beenprepared by some variant of the original method,from cattle blood. However, it was not necessarilythe case that any of the collected samples, detailedin table 1, had been so prepared. It was thereforenecessary to prepare this type of Prussian blue inthe laboratory.

The recipe used to make the pigment from theoriginal ingredients was that published by RobertDossie in the first volume of his Handmaid to theArts (London 1758). Although the original 1724recipe, an English version of which Dossie includedin the appendix of the second volume, gave anequally fine colour, he says of the 1758 recipe: Theproportions of the ingredients are more accuratelyadapted to each other.26

In the original 1724 recipe, four ounces (about113 g) each of crude tartar (potassium hydrogentartrate, potassium hydrogen 2, 3-dihydroxybutane-dioate) and dried crude nitre (potassium nitrate,KNO3) were mixed and detonated with charcoal togive an alkali (essentially potassium carbonate,K2CO3). This was powdered while still hot, mixedwith well-dried, powdered ox blood and heated in acovered crucible until the contents ignited.Subsequently it was heated more intensely until itglowed and there was little further flame. Thecontents were ground while still hot, added to twopints (about 1.1 litres) of boiling fresh water, boiledand filtered twice. Eight ounces (about 226 g) ofcrude alum (aluminium potassium sulphate,AlK(SO4)212H2O) were dissolved in two pints ofboiling water; this solution was then added to theiron solution and the mixture added rapidly to thehot alkaline blood extract, giving a greenish or blue-green precipitate. The solution was well stirred untileffervescence had ceased; the mixture was then leftto stand and the precipitate filtered off. On the addi-tion of two or three ounces (about 5785 ml) ofspirits of salt (hydrochloric acid, HCl), the precipi-tated matter turned blue. It was left overnight, thenwashed very thoroughly and dried in gentle heat.

According to the recipe, the strength of calcina-tion of the blood and alkali mixture was importantas it influenced the resulting colour. In addition, asDossie pointed out, the use of more alkali gave a



lighter colour, while more acid would, by more effi-cient removal of the hydrated alumina and otherimpurities formed during the reaction, give a deeperblue. His modified version (see Dossie 1758 (ii) intable 2) used six pounds (about 2.7 kg) of driedblood to two pounds (about 900 g) of pearl ashes(potassium carbonate), four pounds (about 1.8 kg)of alum and two pounds of green vitriol, notcalcined. Four pounds of acid were used to producethe deep blue colour from the greenish precipitate.These proportions and ingredients can be comparedwith those used in other contemporary recipes intable 2; all claimed to give satisfactory results.Blood was not the only form of animal matter:hoofs, horns and other materials were widely used,sometimes with the addition of extra iron in therecipe. For large-scale preparations, many materials,from rotten wood to chimney soot, were tried forreasons of economy, with varying degrees ofsuccess. The emphasis by Carl Hochheimer andothers on the need for care in the initial calcinationand, if necessary, adjustment of the proportions ofingredients during the preparation is thus entirelyunderstandable. The effects on the final colour werealso well understood. In one of Hochheimers recipesfor so-called Paris blue, attributed to the Frenchchemist Philippe Macquer, alum was deliberatelyomitted, resulting in a very deep blue pigment.27 Bythe nineteenth century many contemporary recipesfor preparing Prussian blue have as their startingmaterial potassium hexacyanoferrate(II) (cyano-ferrure de potasse). However, this was still usuallyobtained from natural sources. These might includeanimal matter and iron filings; or perhaps a sourceof carbon, such as logs or charcoal, with nitrogenfrom the air and iron filings.28

The laboratory preparation based on Dossiessecond method was repeated several times, heatingthe dried blood/alkali mixture in a crucible in afurnace and varying the temperature and heatingtime. In this preparation the dried blood was acommercial product supplied for use as a gardenfertiliser. Several trials were carried out using the 3:1ratio suggested by Dossie, but better results wereobtained by increasing the proportion of driedblood. The best results were achieved by heating4050 g of dried blood and 10 g potassium carbon-ate to about 350C. The matter burned with anorange flame, producing extremely noxious fumes,and the door of the furnace was periodically openedbriefly to allow combustion to continue. Thecrucible was then removed, the contents stirred andthe temperature of the furnace allowed to rise

again; this was continued until no further combus-tion took place. The temperature was then raisedbriefly to about 450500C, after which the cruciblewas removed, the contents ground briefly and thenadded to 300 ml of boiling distilled water. Themixture was boiled for 45 minutes then filtered,giving a clear, usually pale yellow, filtrate.

Iron(II) sulphate (10 g) and aluminium potas-sium sulphate (20 g) were dissolved separately indistilled water, mixed and then added rapidly to thehot yellow solution, resulting in the immediateproduction of effervescence, an unpleasant sul-phurous smell and a greenish-blue precipitate. Anyprolonged delay between the production of the paleyellow dried blood extract and the addition of thealum and iron sulphate solutions tended to result inthe formation of brownish iron(III) compounds,spoiling the pigment.

The greenish-blue precipitate was left to settlebefore filtering, and it was observed that the mate-rial became noticeably bluer even on the filter paper.After thorough washing it was treated withhydrochloric acid to remove the hydrated aluminaalso formed during the reaction, resulting in a finesuspension of slightly purplish Prussian blue. Theidentity of the pigment was confirmed by Fourier-transform infrared analysis.29 In one sample (K12ain table 1), the hydrated alumina was allowed toremain as an extender, giving a pale, slightly greenish-blue product.

It is not entirely clear which species are producedby the initial calcination, although the cyano groupsin the final Prussian blue must derive from theextract. If a small amount of the filtrate from thecalcinations was acidified with 2M sulphuric acidand applied to dipyridyl paper (supplied byMacherey Nagel) a red colour was produced, indi-cating the presence of iron(II) ions. However, whendried blood is used in the preparation, it seemsunlikely that free cyanide and free iron(II) are bothpresent in the filtrate, as in the prevailing alkalineconditions these would react to give Prussian blueprior to the addition of iron(II) sulphate.30 Thesecond possibility is that the filtrate contains hexa-cyanoferrate(II), but that the positive dipyridyl testresult is due to dissociation of this ion to give freeiron(II) in the acidified solution applied to thepaper.31 Finally, both free cyanide and hexacyano-ferrate may be present; in alkaline conditions littlefree iron(II) is present so no Prussian blue will beformed, whereas on acidification for the dipyridyltest free iron(II) is produced, giving the positiveresult. Certain recipes published by Hochheimer



and Weber (see table 2) use calcined hoofs, a start-ing material which presumably contained little or noiron, although like others Hochheimer used ironvessels at this stage of the process. Attempts tomimic this method in the laboratory, using collagen-based matter as a starting material, produced noPrussian blue, although a silica rather than ironvessel was used, but this line of investigation wasnot pursued exhaustively.32

The pale colour of the initial product is partlydue to the alumina formed on addition of the solu-tion of iron(II) sulphate and aluminium potassiumsulphate. However, the main reason for the slowdevelopment of the colour is that at least part of theinitial product formed by reaction of the iron(II)sulphate and hexacyanoferrate(II) solutions is thereduced form of Prussian blue, Berlin whiteK2FeII[FeII(CN)6], which gradually oxidises oncontact with air and on acidification of the solutionto give soluble Prussian blue KFeIII[FeII(CN)6].

Although usable quantities of Prussian bluecould be obtained by this method, the yields weresmall compared with the quantities of startingmaterials. (The small yield was also noted by eigh-teenth-century writers: see, for example, Hellot1762 in table 2.) The preparation often failedentirely, particularly if the iron(II) sulphate solutionwas heated, as recommended by Dossie, before itwas added to the filtrate. This gave rise to insolubleiron(III) species (presumably including FeIII(OH)3)which, when added to the filtrate, produced noPrussian blue. It seems likely that although the orig-inal recipe was followed as closely as possible, theresulting Prussian blue was probably purer thaneighteenth-century commercial products, as greatcare was taken to wash the precipitate and todissolve out the alumina (except in the case wherethis was deliberately retained).

For comparison, modern samples were made bythe direct method using either the Prussian blue orTurnbulls blue routes described above, or by theindirect method. In the first preparation by thedirect method for Prussian blue, 2.54 g (0.02 moles)of iron(II) chloride, FeIICl2, were dissolved in 25 mlof water. Excess hydrogen peroxide was added toconvert the iron(II) to iron (III). Separately, 2.84 g(0.01 moles) of ammonium hexacyanoferrate(II),(NH4)4[FeII(CN)6], were dissolved in 25 ml of waterand added slowly to the iron(III)-containing solu-tion with stirring. The blue precipitate that formedwas collected, washed with two 50 ml aliquots ofdistilled water, and dried in a desiccator. This mate-rial is labelled SA in table 1. The preparation was

repeated, but the precipitate was washed four timesrather than twice, to give sample SB.

Sample SC was also prepared by the directmethod for Turnbulls blue, by adding a solutioncontaining 3.29 g (0.01 moles) potassium hexa-cyanoferrate(III), K3[FeIII(CN)6]), in 25 ml of waterto a solution of 2.54 g (0.02 moles) of iron(II) chlo-ride, FeIICl2, in 25 ml of water. The blue precipitatewas again washed twice with 50 ml aliquots ofdistilled water before drying.

Finally, sample SD was prepared by the indirectmethod. A solution of 2.84 g (0.01 moles) of ammo-nium hexacyanoferrate(II), (NH4)4[FeII(CN)6], in 25ml of water was added to a solution of 2.54 g (0.02moles) of iron(II) chloride, FeIICl2, in 25 ml ofwater. A pale blue precipitate formed which becamedark blue on standing, particularly at the surfacewhere the suspension was in contact with air. Anexcess of hydrogen peroxide was added to completethe conversion of the first-formed Berlin white toPrussian blue. The precipitate was then washedtwice with 50 ml of distilled water and dried in adesiccator.

Description of the eighteenth- and nineteenth-century Prussian blues examined

The particle characteristics and composition ofPrussian blues prepared in the laboratory accordingto an eighteenth-century recipe were compared withthose of genuine late eighteenth- and early nine-teenth-century samples (table 1), and also withPrussian blues identified in National Gallery paint-ings (table 3, pp. 969). When examined under theoptical microscope, the laboratory samples werefound to consist of small, irregular fragments oflow refractive index, sometimes with the appearanceof angular flakes. These were occasionally quitelarge, flat and plate-like and sometimes a moretransparent, paler blue (plate 3). Elemental analysisof several of the laboratory-prepared pigments,including K7, K12a and K12b in table 1, revealedthe presence of iron and a variable amount ofaluminium. Potassium was not detected in theseparticular samples, but was present in other prepa-rations not included in this table.

Thorough treatment of the pigment with acidwould be expected to remove most of the hydratedalumina produced during the preparation. In thecase of K7, no aluminium could be detected andrather few translucent angular flakes were present. Inpractice, however, complete removal of the aluminawas not necessarily desired. K12b still contained



some alumina and also a higher proportion of theangular flakes. In K12a, where the alumina had notbeen dissolved out, the particles were rounded, verytranslucent and a pale greenish blue, similar tothose of a traditional lake pigment (plate 4).

In the genuine Prussian blue samples (table 1)and in Prussian blue identified in National Gallerypaintings (table 3), analysis revealed the presence ofpotassium in most cases, along with other elementsthat derive from the extenders where present. Someof the elements detected by Energy Dispersive X-rayanalysis (EDX) may be present because the sampleswere not washed sufficiently thoroughly to removeunreacted starting materials; for example, chlorinefrom FeIICl2 was detected in samples SA and SB.33

The gum was extracted from the paints in water-

colour cakes and the oil from those in paint blad-ders so that the particle characteristics could beexamined for all the samples collected. It was foundthat several pigments resembled the laboratorysamples: they showed translucent, paler-coloured,angular flakes, and also contained aluminium.These included watercolour cakes manufactured byAckermann (AP, AR) and Reeves and Woodyer(MA), powder pigments from the Ozias Humphreywatercolour box and Elizabeth Careys pastel box(OZ, VB), and the two samples of oil paint frombladders (GA, JSM). plates 5 and 6 show typicalexamples from Ozias Humphreys watercolour box(OZ) and Elizabeth Careys pastel box (VB). Alldate from relatively early in the nineteenth century,or possibly from the late eighteenth century. The



plate 3 Prussian blue prepared in the laboratory accordingto an eighteenth-century recipe (sample K12b). Dispersionin odourless kerosene, photographed by reflected light.Original magnification 250; actual magnification 220.

plate 4 Prussian blue prepared in the laboratory according to an eighteenth-century recipe, omitting treatment with acid to remove hydrated alumina (sampleK12a). Dispersion in odourless kerosene, photographed by reflected light. Original magnification 500; actualmagnification 440.

plate 5 Prussian blue from Ozias Humphreys watercolourbox (17421810), London, Royal Academy. The transparentbrownish material is also iron-containing and may havebeen formed during the preparation of the pigment.Dispersion in odourless kerosene, photographed by transmitted light. Original magnification 250; actualmagnification 220.

plate 6 Prussian blue from Elizabeth Careys pastel box (17701831), London, Victoria and Albert Museum.Dispersion in odourless kerosene, photographed by transmitted light. Original magnification 250; actualmagnification 220.

samples from the German pharmacy, which datefrom 1830 or before, might be expected to sharethese features (and some of the characteristic angu-lar flakes were observed in samples RB, RC and RE),but the particles tend to coat or to coagulate aroundthe extender, which makes their own features diffi-cult to assess. On the other hand, several samplesthat probably date from later in the nineteenthcentury had intensely coloured, rounded agglomer-ates similar to the modern Winsor & Newtonsample (WN, plate 7). These included the samplesfrom the Bull paintbox (BCB), the Nicoll samples(NB, NC, NP) and the sample of Turnbulls blue(TB). Deeply coloured clumps of pigment mayexhibit bronzing when examined by reflected light.The phenomenon was most obvious in the Winsor& Newton sample and other later pigments, butwas also seen in one of the laboratory samples (K7)and an Ackermann sample (AP), suggesting that themethod of preparation is not an important factor.

Like the laboratory samples and the earlierpigments described above, Prussian blue identifiedin eighteenth-century National Gallery paintingsoften contains angular, flake- or plate-like particles(plate 8), which may be a lighter blue. Whereinstrumental analysis has been carried out, somealuminium has usually been detected. Thesefeatures appear to be characteristic of the method ofpreparation in which dried blood or other animalmatter and alum are ingredients. The connectionbetween the singular particle shape and the methodof preparation has also been observed in anAmerican eighteenth-century interior decorationscheme.34 The concomitant presence of aluminium

in the particles is a significant feature, however, andit is probable that these qualities, together with theirregularity and variability of the particle morphol-ogy, are indicative of the method of preparation.The Prussian blue identified in some early nine-teenth-century paintings, such as A Horse frightenedby Lightning (NG 4927) by Thodore Gricault,shows the same characteristics. Clearly, animalmatter continued to be the source of the hexacyano-ferrate(II) ion during the early decades of thenineteenth century: Prussian blue with the samefeatures has been identified, for example, in thepaintings of J.M.W. Turner.35

The alumina incorporated in the pigment as aresult of the method of manufacture in effect acts asan extender. Examination of the genuine Prussianblues in table 1 showed that some other sort ofextender had often been added. The materials usedincluded starch, calcium salts, barytes, kaolin and,indeed, alumina itself. As mentioned above, aluminamay be added during the preparation of the modernpigment to assist precipitation, but the principalreasons for the addition of extender would be toreduce the intensity of the colour and to influence itsworking properties. Particularly pale varieties of thepigment were sold under names like Antwerp blue;these contained a great deal of translucent extender.One example, a watercolour cake supplied byAckermann (AA in table 1), is very similar in colourto the laboratory sample in which the hydratedalumina formed was deliberately retained, giving aproduct similar to a blue lake (K12a). However, theparticle characteristics are different: the Antwerpblue does not resemble a blue lake, but contains a



plate 7 Modern sample of Prussian blue, supplied byWinsor & Newton, showing rounded agglomerates ofpigment. Dispersion in odourless kerosene, photographedby transmitted light. Original magnification 500; actualmagnification 440.

plate 8 Gainsborough, Mrs Siddons. Cross-section ofdeep blue stripe on dress. The uppermost dark blue paintlayer consists of Prussian blue with a very little whitepigment added; this is applied over a layer of lead white. A few characteristic light-coloured flakes of Prussian bluecan be seen in the pale blue layer beneath this. Originalmagnification 500; actual magnification 440.

mixture of dark blue particles, larger colourlessparticles of extender and only a few pale bluerounded particles, similar to those seen in K12a.

The tinting strength of Prussian blue is so highthat it is rarely used undiluted; a relatively largeamount of extender could be added by the pigmentmanufacturer or colourman to improve the proper-ties of the material for decorative or printingpurposes, for example, without greatly reducing thestrength of the colour. Where the extender has beenco-precipitated with the Prussian blue, as for exam-ple in the mid-nineteenth-century preparation ofbleu de ciel and bleu Marie-Louise, in which theblue is co-precipitated with barium sulphate, thecolour differs from that obtained by grindingbarium sulphate with blue pigment.36 While paintingout some of the genuine eighteenth- and nineteenth-century samples for the fading experiments describedbelow it was noticed that, compared with the Winsor& Newton pigment or the pure iron(III) hexacyano-ferrate(II) pigment used as a standard for comparison,some of the samples containing extender were verytranslucent even at full strength (RD, RE and RF,from the German pharmacy, are examples).

Some of the early nineteenth-century pigmentsthat contain starch grains (RA, RB; plate 9) mayhave been intended for a particular use, or as low-cost pigments. Starch has also been identified in thePrussian blue in several National Gallery paintings:an example is the Portrait of a Boy (NG 4034), byan unknown French artist, which is probably earlynineteenth century (plate 10). The starch grains areclearly visible in the sample from the blue paint illus-trated in plate 11. While starch-containing pigments

may not have been in the highest quality range, costwas not necessarily the only factor: the artist mayhave liked the particular feel of the paint. It wasquite noticeable that the extenders present in theeighteenth- and nineteenth-century pigments had amarked effect on their handling properties, bothduring grinding and when dispersed in the medium.Some extenders gave a particularly pleasant silkyfeel to the pigment. Constant de Massoul, writing in1797, noted that Prussian blue was difficult to use in



plate 9 Prussian blue from an old German pharmacy near Darmstadt (180030) containing a starch extender,supplied by Professor Dr E.-L. Richter (sample RB).Dispersion in odourless kerosene, photographed by transmitted light. Original magnification 250; actualmagnification 220.

plate 10 French School, Portrait of a Boy (NG 4034), earlynineteenth century. Canvas, 55.9 42.5 cm.

plate 11 French School, Portrait of a Boy (plate 10).Cross-section of dark blue paint of the boys jacket, showing round or oval translucent starch grains within the blue paint layer. Partial evaporation of the odourlesskerosene used to moisten the sample for photography hasrendered the grains easily visible. Original magnification500; actual magnification 440.

watercolour because in drying all the particlesreunite, even after having been well ground uponPorphyry.37 This was found to be so with the labo-ratory samples prepared from dried blood, and withseveral others; the presence of any extenderimproved the grinding markedly.

To offset the greenish undertone of the pigment,recipes for Prussian blue occasionally suggest theaddition of red dyestuff, usually cochineal, in someform (see, for example, Crker 1736 and Secrets1790 (v) in table 2). Particles of red lake wereobserved in the Random and Sneath watercolourcake (RSR in table 1), although insufficient materialwas available for analysis to be possible. As thesered dyestuffs are particularly unstable to light, theirfading would be expected to have an effect on thehue of these Prussian blues, quite apart from anysensitivity to light the blue pigment itself might have.

Colour change in samples of Prussian blue

Experimental

To examine the fading of Prussian blue a number ofthe samples detailed in table 1 were painted out ingum arabic solution as mid-toned watercolourwashes on Silversafe paper; these were not particu-larly pigment-rich, but nor were they characteristicof the very thin washes of Prussian blue sometimesfound in watercolours.

As admixture with a white pigment has beenimplicated in the impermanence of Prussian blue, aset of samples containing different proportions ofblue pigment to lead white was prepared in a linseedoil medium. Lead white was chosen because thiswas the standard white pigment used in the eigh-teenth and nineteenth centuries. However, as onlytiny amounts of some of the genuine eighteenth-and nineteenth-century pigments were available, apreliminary experiment was performed to findappropriate proportions of lead white and Prussianblue. A series of mixtures of pure iron(III) hexa-cyanoferrate(II), FeIII4[FeII(CN)6]3xH2O (sample ALPin table 1), with lead white was painted out. Theseries, seen in plate 2, comprises samples withPrussian blue to lead white proportions (weight toweight) of 1:5, 1:10, 1:20, 1:40, 1:50, 1:100, 1:200,1:400 painted in linseed oil. Three shades, termeddeep (1:10), mid (1:50) and light (1:400), wereselected and used as visual standards against whichto compare mixtures of the Prussian blues with leadwhite. Some samples were also painted as very lightor extremely light by further diluting the lightmixtures with lead white. Because of their scarcity,not all the historic Prussian blues were painted out,and not every shade was prepared for each sample;see table 1 for details. In every case the mixture wasground in cold-pressed linseed oil and painted ontoa gessoed panel primed with lead white in linseedoil, plate 12. To assess the effect of film depth on



plate 12 Test samplesof modern and historicPrussian blues, mixedwith different propor-tions of lead white in oilmedium, before (left)and after (right) expo-sure to light: see table 1for reference and shadecodes. The sequence isidentical for the fadedpanels on the right.

From top (unfaded panels on left, left to right):row 1: ALP (D, M, L3L1), RE (D, M, L2), RD (D)row 2: K12b (L1, L3), MC (L1, L2), K12a (L3),

RF (D, M, L2), RD (M)row 3: K7 (D, M, L1L3), JSM (M, L2, L3), RD (L2)row 4: K12a (D, M,), GA (M, L2, L3), BCB (M, L1L3)

row 5: VB (M, L2, L3), MA (L1, L3), [blank], AP (M, L1, L3)row 6: JSM (L2, at 3 different thicknesses), SB (D, M, L2)row 7: TB (D, M, L2), SC (D, M, L2)row 8: NB (D, M, L3), WN (D, M, L2)row 9: NP (D, M, L2)row 10:NC (D, M, L2)

1

2

3

4

5

6

7

8

9

10

the fading, the very light mixture containingPrussian blue from a paint bladder of 183040(JSM) was painted out at three thicknesses: one,two or three layers.

The samples were all placed in a purpose-builtlight box at the National Gallery.38 Briefly, thesamples were exposed to light from twelve 65WGeneral Electric Artificial Daylight lamps, giving anilluminance at the sample surface of 2022,000 lux;the temperature and relative humidity during theperiod of these experiments were 24.5 1C and364% respectively. The light was not filtered toexclude ultraviolet radiation.

The colour of each sample was measured at theoutset and periodically during light exposure using aMinolta CR221 colour meter. The standard geometryof this device ensures that the surface to be measuredis illuminated at 45 and the light reflected along anaxis perpendicular to the surface is measured. Thisso-called 45/0 geometry ensures that specular reflec-tion (gloss) is excluded from the colour measurementprocess; the sample area is 3 mm in diameter. Thedata from the colour meter were converted toCommission Internationale de lEclairage (CIE) L*,a*, b* co-ordinates using the standard illuminantD65 as reference.39 These co-ordinates refer to thelightness (L*), red-greenness (a*), and yellow-blue-ness (b*) of the sample respectively.

Although dark recovery of colour has beenreported for Prussian blue in very thin washes, asfound in cyanotypes,40 the phenomenon is lessfrequently observed in oil paint films. During thecourse of the experiments described above, severalattempts were made to replicate the recovery of bluecolour reported for early samples of Prussian bluewhen they were stored in the dark after light expo-sure. Duplicate sets of the samples were placed in thedark after exposure to light for 74 Mlux.h, but thechanges in colour that resulted from dark storagewere very small. As a result of these experiments,however, three Prussian blues were selected as promis-ing for a further study of this phenomenon. Samplesof three Prussian blues, ALP, K7 and SD (see table 1),were painted out on a roughened PTFE plate inlinseed oil, a copal varnish dating from 1888, andLiquitex acrylic medium. After initial colour measure-ment, the plate was exposed in the light box for 6.7Mlux.h.41 The colour was measured again and theplate was then placed in a dark metal drawer for twoweeks. At the end of the period of dark storage, theplate was measured and exposed to a further 6.7Mlux.h, then a second period of two weeks in darkstorage and, finally, a third exposure of 6.7 Mlux.h.

All the samples painted out in watercolour faded tosome extent on exposure to light, some showingmarkedly more fading than others. fig. 2 shows thecolour change in two historic samples withoutextender (MA and TB) and five samples of modernmanufacture, one sample of insoluble Prussianblue, FeIII4[FeII(CN)6]3xH2O (ALP), and four typesof soluble Prussian blue, MFeIII[FeII(CN)6]xH2O(where M = K+ for SC and M = NH4+ for SA, SB andSD). It is clear that the insoluble variety fades muchless markedly than the modern soluble samples andthe two representative samples of historic pigments,which from their elemental composition (table 1)can be assumed to be soluble Prussian blues.

The effect of extender can be seen by comparingthe fading of two samples prepared in the labora-tory using Dossies 1758 recipe (fig. 3). One half ofthe material from this preparation (K12) containsundissolved alumina as an extender, while the otherportion was treated with acid to remove most of thealumina. The sample containing alumina (K12a)shows a greater colour change. This is at first sightunexpected as the net colour change for completefading of the paler, alumina-containing samplewould be less than that for complete fading of thedarker sample. This indicates that the rate of fadingof the blue pigment is very greatly increased by thepresence of the alumina extender, as has been previ-ously noted.

Although the degree of colour change seems todepend on the type of Prussian blue and is markedlyaffected by the presence or absence of extender, all thePrussian blue samples showed similar behaviour interms of colour shift. The pattern is illustrated in fig.4: in addition to lightening (which is not shown in thisfigure) each of the samples becomes less blue (increasein b*) and less green (increase in a*) as it fades.

The samples painted out in oil with lead whiteshow less straightforward behaviour, although thesame pattern emerges for many of the samplesexamined. fig. 5 shows the colour change for fiveshades of a mixture of Prussian blue preparedaccording to Dossies 1758 recipe (K7) and leadwhite. In each case, the colour readings for theinitial samples have been excluded as there is a largechange in the a* and b* co-ordinates during the first24 hours of light-accelerated ageing. This is due tocolour changes in the linseed oil medium that areparticularly apparent in the samples containing ahigher proportion of lead white.42 All five shadesshow a gradual increase in b*, corresponding to a



Results and discussion

decrease in blueness. The extremely light and verylight samples concomitantly decrease in greenness(increase in a*), while the deep sample first increasesin greenness before decreasing in greenness. Thepattern for the light and mid-toned samples is evenmore complicated; they first decrease slightly ingreenness, then increase in greenness somewhatbefore consistently decreasing in greenness. Thispattern is seen in many of the early Prussian bluesexamined, for example the sample from ElizabethCareys pastel box from the Victoria and AlbertMuseum (VB) shown in fig. 6. For all the specimensexamined, the extremely light, very light and lightsamples show an increase in a* of approximately 3to 7, while the increase in b* is around 14 to 20 (sothey are becoming less green and less blue). Thefinal colour is near-neutral on the blue to yellow(b*) axis and appears a pale greenish grey; the

lighter the initial colour the yellower the final grey.The sigmoidal plots for the mid-tones show smalloverall increases in a* accompanying larger increasesof approximately 13 to 20 in b*. The deep tonesshow smaller changes in both a* and b*, but stillsufficient to produce a noticeable loss of blueness.

There are no differences between the behaviourshown by samples of Prussian blue mixed with leadwhite and samples of extended Prussian blue mixedwith less lead white to give the same shade. In otherwords, the nature of the diluent (white pigment orextender) has no effect on the fading of the Prussianblue. This is reinforced by the finding that Prussianblues containing different extenders behave verysimilarly, for example the sample of Prussian blueprepared from Dossies 1758 recipe, which isextended with alumina (K12a), and one of thesamples from the German pharmacy extended with



medium

0

5

10

15

20

0 20 40 60 80 100 120 140

Exposure / Mlux. hours

Co

lou

r ch

ang

e (E

)'Insoluble' Prussian blue (ALP)

'Soluble' Prussian blue (SA)

'Soluble' Prussian blue (SB)

'Soluble' Prussian blue (SC)

'Soluble' Prussian blue (SD)

Reeves 180517 (MA)

Turnbull's blue (TB)

watercolour medium

0

5

10

15

20

25

0 20 40 60 80 100 120 140

Exposure / Mlux. hours

Co

lou

r ch

ang

e (E

)

Sample without alumina extender

Sample with alumina extender

fig. 2 Colour change(E) for insoluble andsoluble Prussian bluesin watercolour medium.

fig. 3 Effect of extenderon colour change (E)in samples of Prussianblue prepared accordingto Dossies recipe of1758 (K12a, withhydrated alumina, andK12b, without) in water-colour medium.

barium sulphate (RE). A trial experiment to investi-gate the effect of other white pigments, includingtitanium white, zinc white and barium sulphate, infour different binding media cold-pressed linseedoil, Liquitex acrylic medium, egg tempera or gumarabic gave no indication that the nature of thewhite pigment or the binding medium affected therate of fading of Prussian blue.

There were no marked differences in the fadingof the three test patches comprising a sample ofPrussian blue from a paint bladder of 183040(JSM) mixed with lead white and painted out inone, two or three layers. In each case the increase ina* was between 1.1 and 2.1, while the increase in b*(loss of blueness) was between 12.7 and 13.7, indi-cating that the loss of colour was a surface effect.This confirmed the evidence from cross-sectionsfrom paintings and test samples.

As might have been expected, the Random and

Sneath watercolour cake (RSR) that contains a redlake (probably cochineal) showed a noticeable lossin redness on exposure to light, presumably due tofading of the red lake.

The three samples of Prussian blue that weresubjected to alternate periods in light and darkshowed some small changes in colour. The resultsare presented in fig. 7 for the samples in Liquitexacrylic medium. In each case a light exposure of 6.7Mlux.h resulted in an increase in lightness of 7 to 8units, a decrease in redness of 4 to 6 units and adecrease in blueness of 3 to 4 units (not shown inthe figure). During a period of dark storage somerecovery of the colour was observed, again shown(as First dark) in fig. 7. A second period of lightexposure produced lightening and loss of rednessand blueness comparable to the first light exposure,resulting in colours that were lighter, less red andless blue than those after the initial light exposure.



-20

-15

-10

-5

0

-7 -6 -5 -4 -3 -2 -1 0

a*

b*

Reeves 180517; Museum of London (MA)

VA paint box 17711831 (VB)

Turnbull's blue from BM (TB)

Dossie's 1758 recipe (K12b)

Filled symbols represent initial colours

-25

-20

-15

-10

-5

0

5

-7 -6 -5 -4 -3 -2 -1 0 1 2

a*

b*

Extremely light

Very light

Light

Mid

Deep

Filled symbols represent colours after

initial bleaching

fig. 4 Colour change in foursamples of Prussian blue inwatercolour medium.

fig. 5 Colour change formixtures of Prussian blueprepared according to Dossiesrecipe of 1758 (K7) with leadwhite in oil.

Subsequent dark storage caused some recovery, buton each lightdark cycle it is clear that the colour isnot fully recovered and that subsequent light expo-sure causes further irreversible fading.

Conclusions

Factors affecting the degree of fading

The principal factor in the permanence of Prussianblues of different manufactures and dates appearsto be the amount of extender present in the suppliedpigment. Different methods of preparation onlyaffect permanence in as much as they lead to theincidental introduction of extender, either deliber-ately or due to inadequate purification of theproduct. For example, dissolving out the aluminiumsalt from Prussian blue made from animal blood iseasily carried out in a small-scale laboratory experi-

ment, but is less easily achieved in the industrialpreparation of the pigment.

Adding white pigment to Prussian blue increasesthe colour change observed on light exposure. Thismay be due to increased inter-reflection of light inthe paint layer, leading to greater fading, or maysimply be a visual effect, because the colour changeis more noticeable in the paler blue shades whereless Prussian blue is present in the first place. Thenature of the white pigment does not seem to affectthe fading, nor does it matter whether the white is adeliberate addition to dilute the colour, or a pre-mixed extender added to the Prussian blue by themanufacturer or colourman. When a relatively pureand a much-extended Prussian blue are each mixedwith sufficient lead white to produce shades ofcorresponding lightness, the fading behaviour of thetwo samples is broadly similar. As alumina extenderis present in the light-coloured, translucent, plate-



-30

-25

-20

-15

-10

-5

0

5

-8 -7 -6 -5 -4 -3 -2 -1 0

a*

b*

Extremely light

Very light

Mid

Filled symbols represent colours after

initial bleaching

20

25

30

35

40

45

-12 -10 -8 -6 -4 -2 0 2 4

a*

L*

Sample ALP Sample K7 Sample SD

SD: Initial

K7: Initial

ALP: Initial

ALP: First light

K7: First light

SD: First light

ALP: First dark

K7: First dark

SD: First dark

SD: Second dark

SD: Second lightALP: Second dark

ALP: Second light

K7: Second dark

K7: Second light

fig. 6 Colour change formixtures of Prussian bluefrom the paintbox held atthe Victoria and AlbertMuseum (VB) with leadwhite in oil.

fig. 7 Colour changeaccompanying alternatingperiods of light exposureand dark storage forPrussian blue samples inLiquitex acrylic medium.

like flakes characteristic of the eighteenth- and earlynineteenth-century pigment, these are particularlylikely to exhibit fading, sometimes only at the top ofthe particle. A small pigment particle surrounded bywhite pigment in a typical light blue paint filmwould also be likely to fade.43 Where the particlesform aggregates they are less vulnerable.44

Samples in watercolour medium showed similarcolour shifts on fading to the mixtures of Prussianblue and lead white in oil medium classified as verylight or extremely light. Deeper shades in oilmedium showed smaller changes in colour and morecomplicated colour shifts.

The formula of the pigment has some effect onits permanence; in watercolour, the modern sampleof insoluble Prussian blue faded less than modernor historic samples of soluble Prussian blue.However, the counter ion in the different solublePrussian blues (K+ or NH4+) seems to have littleeffect on the fading.

The reversibility and chemistry of fading

The films that comprised Prussian blue and whitedid not regain their colour when placed in the darkafter considerable exposure to light, nor did thewatercolour samples. The reversible behaviour wasonly observed in films exposed for relatively shortperiods before dark storage. It may be that theshort-term reversible fading is simply an early stageof the long-term loss of colour, but evidence fromother sources seems to indicate that there are twoprocesses involved in the fading of Prussian blue: afairly rapid, reversible loss of colour, and a slowerirreversible colour change. The reversible changecan be induced by light in oil paint films, as demon-strated in the experiments described above, but ismore usually associated with, and is more dramaticin, thin washes on paper, in cyanotypes forexample.45 The slower irreversible fading is seen insamples in all media and results in a final colourthat is best described as yellowish grey.

The chemistry behind these types of colour lossis, however, not completely clear. As alreadymentioned, the strong blue colour is due to thecharge transfer transition between iron(II) andiron(III) in the FeIIINCFeII moiety. Anything thatdisrupts this configuration will lead to a loss or shiftof colour.

Three reactions of the iron(III) hexacyano-ferrate(II) complex have been implicated in colourchange. First, a reduction of Prussian blue to Berlinwhite, K2FeII[FeII(CN)6] (also called Everitts salt),

which can be light-induced, or can be initiated by areducing agent:

FeIII[FeII(CN)6]- + e- FeII[FeII(CN)6]2-

Electrochemical studies of Prussian blue films haveshown that the reduction is a single step process andthat the blue colour returns on oxidation.46 Indeed,the FeII[FeII(CN)6]2- complex is an intermediate in thesynthesis of Prussian blue from iron(II) and hexa-cyanoferrate(II) ions. As described earlier, mixingsolutions containing these two ions yields a pale blueprecipitate (containing both K2FeII[FeII(CN)6] andKFeIII[FeII(CN)6]), which on exposure to air deepensin colour as the Berlin white is slowly oxidised in air,or more rapidly if an oxidising agent is introduced.In the paint film, the situation is not quite the sameas in the electrochemical solution studies, wherethere is a ready supply of counter ions, usuallypotassium. However, a spectroelectrochemical studyof dry films of soluble Prussian blue using in situFourier transform infrared spectroscopy with atten-uated total reflectance (FTIR/ATR) demonstratedthat both reduction and oxidation were possible inthe absence of an electrolyte.47

This redox process is believed to be responsiblefor the lightdark bleaching and recovery of cyano-types. Although the paper or size may provide thereducing agent, the fading and recovery has alsobeen noted on cyanotypes printed on glass fibrepapers, suggesting that an excess of one of thereagents used in the preparation might act as areducing agent.48 In either case, the cycle graduallydecays as the reservoir of reducing agent is usedup, although if a cyanotype is re-sized further reduc-tion can be induced.49

It also seems likely that the reported bleaching ofPrussian blue when it is stored in oxygen-free envi-ronments and its recovery on re-exposure to air, forexample when the dry pigment is stored in anairtight bottle,50 results from Prussian blue to Berlinwhite inter-conversion.

A second possible reaction that would lead toloss of colour in Prussian blue would be the oxida-tion of [FeII(CN)6]4- ion. Electrochemical studies ofPrussian blue films deposited on electrodes haveshown that, unlike the abrupt reduction of Prussianblue to Berlin white, this oxidation proceedsthrough a number of intermediate steps beforeyielding a yellowish product, sometimes referred toas Prussian yellow.51 One intermediate in this reac-tion is the mixed valence compound known asBerlin green. While the oxidation of Prussian blue



to Berlin green is reversible, the subsequent oxida-tion to Prussian yellow is not fully reversible.52

Prussian blue Berlin green

[FeII(CN)6]4-

{[FeIII(CN)6]0.67[FeII(CN)6]0.33}3.33- + 0.67e-

Berlin green Prussian yellow

{[FeIII(CN)6]0.67[FeII(CN)6]0.33}3.33- [FeIII(CN)6]3- + 0.33e-

The very light and extremely light paintsamples examined in this study all yielded fadedproducts that were a yellowish grey, possibly amixture of Prussian yellow and a little unreactedPrussian blue highly diluted with lead white.

A third possible reaction of the hexacyano-ferrate(II) ion is a substitution reaction. Theaquapentacyanoferrate(II) ion is known to beformed by the action of light on Prussian blue:

FeIII[FeII(CN)6]- + H2O h

FeIII[FeII(CN)5 (H2O)] + CN-

The reaction is generally studied in aqueous solu-tions,53 where there is a ready source of water toco-ordinate at the vacant site on the FeII ion; inhydrophobic environments, the water molecule co-ordinating to the iron may be one of theunassociated water molecules occupying a zeoliticsite in the Prussian blue structure. As the aqua-pentacyanoferrate(II) complex is green, a furtherreaction must occur during the fading of Prussianblue paint samples examined in this study, as theirfinal colours are considerably less green than thestarting materials. It is possible that further substi-tution reactions displace further cyanide ligands,causing a shift in colour towards yellow and colour-less, but this has not been studied in any detail.Alternatively, it is known that FeIII[FeII(CN)5 (H2O)]can also undergo redox reactions like those ofPrussian blue to yield, for example, the aquapenta-cyano analogue of Berlin white:54

FeIII[FeII(CN)5 (H2O)] + e- FeII[FeII(CN)5(H2O)]-

As far as easel paintings are concerned, it seems thatif the reversible fading, seen in cyanotypes andwatercolour paintings on occasions (or in the absenceof oxygen), takes place at all, it is at the very begin-ning of the deterioration process. The yellowish-greyappearance of changed Prussian blue-containing

paint films suggests that the long-term colourchanges result from oxidation of the pigment.Perhaps because the appropriate conditions for theregeneration of Prussian blue are not present in thepaint film, these changes seem to be irreversible.There is certainly no shortage of oxygen in a dryingoil paint film; the oxygen would be consumed bothby the oil as it cross links and by Prussian blue as itchanges colour. This would accord with eighteenth-century reports of changes in the colour of thepigment relatively soon after painting.55

Acknowledgements

We would like to thank our colleagues in theScientific Department at the National Gallery whohave supplied us with examples of the use ofPrussian blue in paintings and conducted EDXanalysis of pigment samples: Ashok Roy, MarikaSpring and Rachel Grout, Radcliffe Reseach Fellow,20002. We would also like to thank those whoprovided historical samples of the pigment: DrAndreas Burmester, Doerner Institut, Munich;Professor Dr E.-L. Richter (retired), Institut frTechnologie der Malerei, Staatliche Akademie derBildenden Knste, Stuttgart; Jane Zeuner, Museumof London; MaryAnne Stevens, Royal Academy ofArts, London. Other samples were already in thepossession of the National Gallery ScientificDepartment. In addition, Professor Richter kindlysupplied us with a copy of the 1985 Diplomarbeiton Berliner Blau by Hans Becker. After the publica-tion of the article on the fading of Prussian blue inVolume 14 of the National Gallery TechnicalBulletin (1993), Professor Richter asked us to pointout that it was in fact Mr Becker who made theobservation of the occurrence and fading of thepigment in Watteaus Rcreation italienne (Berlin,Charlottenburg Castle). We are happy to do this.Also as a result of this publication, we have receivedinteresting reports and had useful discussions oncolour change in Prussian blue in eighteenth-centurydecorative schemes and easel paintings with ChiAsai, paintings conservator, Munich, Johannes Preis(of Gebr. Preis GmbH., restorers, Parsberg), then ofthe University of Cologne, and Marc Giudicelli,then of the Straus Center for Conservation, HarvardUniversity Art Museums: our thanks are due tothem. Finally we are particularly grateful to DrMike Ware for useful and stimulating discussions onthe chemistry of colour change in Prussian blue.



Notes and references

1 G.E. Stahl, Experimenta, Observationes, Animad-versiones, CCCNumero, chymicae et physicae, Berlin 1731, pp. 2803.

2 Notitia Coerulei Berolinensis nuper inventi, Miscellanea Berolinensia,I, 1710, pp. 3778; J. Kirby, Fading and Colour Change of PrussianBlue: Occurrences and Early Reports, National Gallery TechnicalBulletin, 14, 1993, pp. 6271, esp. pp. 63, 701.

3 M. van Eikema Hommes, Discoloration in Renaissance and BaroqueOil Paintings: Instructions for Painters, Theoretical Concepts, andScientific Data, PhD Thesis, University of Amsterdam, 2002, pp. 13948.

4 Kirby 1993 (cited in note 2), pp. 65, 67.5 Kirby 1993 (cited in note 2).6 J. Woodward, Praeparatio caerulei Prussiaci ex Germania missa ad

Johannem Woodward, Philosophical Transactions, XXXIII, no. 381,January-February 1724, pp. 1517.

7 J.A. Sistino, Ferriferrocyanide Pigments: Iron Blue, Pigment Handbook,ed. T.C. Patton, New York 1973, Vol. 1, pp. 4017, esp. pp. 4034.

8 C.F.A. Hochheimer, Chemische Farben-Lehre, oder ausfhrlicherUnterricht von Bereitung der Farben zu allen Arten der Malerey, 2 partsin 1 volume, Leipzig 17924, p. 171.

9 A.H. Church, The Chemistry of Paints and Painting, 3rd edn, London1901, pp. 21316. Mortimer and Rosseinsky point out that the solubleform is actually thermodynamically less soluble than the insolubleform: see R.J. Mortimer and D.R. Rosseinsky, Iron HexacyanoferrateFilms: Spectroelectrochemical Distinction and ElectrodepositionSequence of Soluble (K+-containing) and Insoluble (K+-free)Prussian Blue, and Composition Changes in PolyelectrochromicSwitching, Journal of the Chemical Society Dalton Transactions, 1984,pp. 205961.

10 H. Holtzman, Alkali Resistance of the Iron Blues, Industrial andEngineering Chemistry, 37, 1945, pp. 85561.

11 Sistino 1973 (cited in note 7), pp. 4045. Barbara Berrie, citing DegussaChemische, Vossen Blue Pigments and Blue Pigment Compounds,Teterboro, NJ, 1974, gives a change in permanence rating from eight(permanent) when pure to four when mixed with ten parts of titaniumwhite: B.H. Berrie, Prussian Blue, Artists Pigments: A Handbook ofTheir History and Characteristics, Vol. 3, ed. E. West FitzHugh,Washington, DC 1997, pp. 191217.

12 See, for example, W.A. Kampfer, Titanium Dioxide, in PigmentHandbook 1973 (cited in note 7), pp. 136, esp. p. 2; H.B. Stephenson,Zinc Oxide and Leaded Zinc Oxide, in the same volume, pp. 3751;E.J. Dunn, White Hiding Lead Pigments, in the same volume, pp. 6584. The three reflectance curves may also be compared in H.Khn, Zinc White, Artists Pigments: A Handbook of Their Historyand Characteristics, Vol. 1, ed. R.L. Feller, Cambridge/Washington, DC1985, pp. 16986, esp. p. 173.

13 M. Ware, A new blueprint for cyanotypes, Ag+ Photographic, 7, 1995,pp. 7481.

14 Sistino 1973 (cited in note 7), p. 405; A Review of Literature in theForeign Paint and Varnish Field, Paint, Oil and Chemical Review, 25June 1936, pp. 1114, esp. p. 13; Berrie 1997 (cited in note 11), p. 197.

15 Berrie 1997 (cited in note 11), p. 197.16 Sistino 1973 (cited in note 7), p. 405; Berrie 1997 (cited in note 11), p. 197.17 Kirby 1993 (cited in note 2), p. 65.18 P. Day, Introduction to Mixed-Valence Chemistry, in Mixed-Valence

Compounds: Theory and Applications in Chemistry, Physics, Geologyand Biology, ed. D.B. Brown, Dordrecht/ London 1980, pp. 324; A.G. Sharpe, The Chemistry of Cyano Complexes of the TransitionMetals, London 1976, pp. 1216.

19 F. Herren, P. Fischer, A. Ludi and W. Halg, Neutron Diffraction Studyof Prussian Blue, Fe4[Fe(CN)6]3xH2O: Location of Water Moleculesand Long-Range Magnetic Order, Inorganic Chemistry, 19, 1980, pp. 9569; A. Ludi, Descriptive Chemistry of Mixed-ValenceCompounds, in Mixed-Valence Compounds 1980 (cited in note 18),pp. 2547, esp. pp. 3940; Berrie 1997 (cited in note 11), pp. 199202. Itis likely that the Fe3+ ions are in fact hydrolysed and present as[Fe(OH)]2+: see C.A. Lundgren and R.W. Murray, Observations on theComposition of Prussian Blue Films and Their Electrochemistry,Inorganic Chemistry, 27, 1988, pp. 9339.

20 D.R. Rosseinsky, J.S. Tonge, J. Berthelot and J.F. Cassidy, Site-transferConductivity in Solid Iron Hexacyanoferrates by DielectricRelaxometry, Voltammetry and Spectroscopy, Journal of the ChemicalSociety Faraday Transactions I, 83, 1987, pp. 23143.

21 H.J. Buser, D. Schwarzenbach, W. Petter and A. Ludi, The Crystal

Structure of Prussian Blue: Fe4[Fe(CN)6]3xH2O, Inorganic Chemistry,16, 11, 1977, pp. 270410.

22 The wavelength of maximum absorption, max, will depend on the typeof Prussian blue. According to Mortimer and Rosseinsky 1984 (cited innote 9), max for soluble Prussian blue is 690 nm (14500 cm-1), whilethat for insoluble Prussian blue is 730 nm (13700 cm-1). The differencecan be attributed to the different ligand field splitting in the twocomplexes: in the former the Fe(III) centre is surrounded by six nitrogenatoms, while in the latter it is surrounded by both nitrogen and oxygenatoms with an average environment of FeIIIN4.5O1.5. Robin gives avalue of 14100 cm-1 for soluble Prussian blue (M.B. Robin, The Colorand Electronic Configuration of Prussian Blue, Inorganic Chemistry, 1,2, 1962, pp. 33742).

23 Sharpe 1976 (cited in note 18), pp. 1213.24 See, for example, Inorganic Pigments: Manufacturing Processes,

ed. M.H. Gutcho, Park Ridge 1980, pp. 4339; N. Heaton, Outlines ofPaint Technology, 3rd edn, London 1948, pp. 1548; J.S. Remington andW. Francis, Pigments: Their Manufacture, Properties and Use, London1954, pp. 949.

25 Kirby 1993 (cited in note 2), p. 64.26 R. Dossie, The Handmaid to the Arts, 2 vols, London 1758, Vol. I,

pp. 778; Vol. 2, pp. 38791.27 Hochheimer 1792 (cited in note 8), pp. 13372, esp. p. 144 onwards;

20722: for Paris blue see p. 142. The difficulties that could arise duringmanufacture are described by J.A. Weber, Chemische Erfahrungen beymeinem und anderen Fabriken in Deutschland, Neuwied 1793, pp. 13465, 189200. It is notable that Hochheimer refers to the use ofan iron vessel for the calcination and this is likely to have been a generalpractice: see, for example, W. Nicholson, A Dictionary of Chemistry, 2vols, London 1795, Vol. 1, p. 226, cited by Sally Carew-Reed, LewisBerger & Sons (17661960): An English colour manufactory, 2 vols,3rd year project, Department of Conservation & Technology,Courtauld Institute of Art, 1997: Prussian blue. The annotated copyshe cites is in the archives of Lewis Berger & Sons, who manufacturedPrussian blue from 1766.

28 J.-F.-D. Riffault Deshtres, A.-D. Vergnaud and C.-J. Toussaint,Nouveau manuel complet du fabricant de couleurs et de vernis, newedn revised by F. Malepeyre and E. Winckler, 2 vols (Encyclopdie Roret Manuels Roret), Paris 1884, Vol. 1, pp. 237307. Berger & Sons seemto have used prussiate of potash, which they prepared themselves,from about 1834: Carew-Reed 1997 (cited in note 27). The methods ofmaking the pigment, particularly in Germany, are summarised byH.K.L. Becker, Berliner Blau, Diplomarbeit am Institut fr Technologieder Malerei der Staatlichen Akademie der bildenden Knste zuStuttgart, 1985, pp. 937. In 1837 a medal was awarded to LewisThompson by the Society of Arts for his introduction of a method ofmaking potassium ferrocyanide by roasting charcoal with potash andiron filings at red heat: Ures Dictionary of Arts, Manufactures andMines, ed. R. Hunt, London 1860, p. 545 (cited by M. Ware,Cyanotype: The History, Science and Art of Photographic Printing inPrussian Blue, London 1999, p. 22).

29 Fourier transform infrared analysis was performed by Jennifer Pilc.30 In the laboratory, an alkaline solution containing free iron(II) and free

cyanide ions rapidly produced a dark coloured precipitate containing amixture of Prussian blue and iron(III) hydroxide.

31 When an alkaline solution of K4[FeII(CN)6] is applied to a dipyridylpaper a positive result is not obtained. However, if the solution is acidi-fied before it is applied to the paper, a positive result is obtained;obviously some FeII is formed by acid-initiated breakdown of the hexa-cyanoferrate(II) ion. Indeed, if a solution of K4[FeII(CN)6] is acidifiedand left for some hours, a soluble Prussian blue is formed, presumablyby the reaction of the small amount of free FeII produced with theremaining hexacyanoferrate(II). No Prussian blue is formed from analkaline solution of K4[FeII(CN)6].

32 L.J.M. Coleby, A History of Prussian Blue, Annals of Science, 4, 1939,pp. 20611.

33 The laboratory-prepared samples and collected samples were examinedby Marika Spring. Prussian blue was identified in paintings during thecourse of routine examination by Ashok Roy, Marika Spring andRachel Grout, by SEMEDX or (for earlier analyses) laser microspec-tral analysis (LMA), or by Catherine Higgitt using FTIR. In some casesthe pigment was identified microchemically (it reacts rapidly withalkali, giving a brown product).

34 F.S. Welsh, Particle Characteristics of Prussian Blue in an HistoricalPaint, Journal of the American Institute for Conservation, 27, 1988,pp. 5563.



35 J.H. Townsend, The Materials of J.M.W. Turner: Pigments, Studies inConservation, 38, 1993, pp. 23154.

36 Riffault Deshtres et al. 1884 (cited in note 28), Vol. 1, p. 44. GeorgesHalphen describes how the director of a factory at Dugny tested theeffects of extender pigments, including different forms of bariumsulphate, zinc white and lead white (cruse), on the colour of Prussianblue; the differences were apparently quite marked: G. Halphen,Couleurs et vernis, Paris 1895, pp. 1517.

37 C. de Massoul, A Treatise on the Art of Painting and the Compositionof Colours, London 1797, p. 182.

38 D. Saunders and J. Kirby, A comparison of light-accelerated ageingregimes in some galleries and museums, The Conservator, 25, 2001, pp. 95103.

39 Commission Internationale de lEclairage, Recommendations onuniform colour spaces, colour difference equations, psychometriccolour terms, Supplement No. 2 to CIE Publication No.15 (E-2.3.1),1971/(TC-1.3), 1978.

40 Ware 1999 (cited in note 28), pp. 1206.41 An exposure of 6.7 Mlux.h is equivalent to around twelve years in a

museum under the lighting conditions recommended for oil paintings(200 lux during 56 viewing hours per week), or to exposure outsidefor one month during typical spring or autumn conditions inLondon.

42 The initially reversible yellowing and bleaching of a fresh drying oilfilm, due to conjugation of unsaturated bonds in the oil, is well-knownand is particularly apparent in white or light-coloured paint. This effectwill be discussed in more detail, from the point of view of colour meas-urement, in a forthcoming publication.

43 R.M. Johnston-Feller, R.L. Feller, C.W. Bailie and M. Curran, TheKinetics of Fading: Opaque Paint Films pigmented with Alizarin Lakeand Titanium Dioxide, Journal of the American Institute forConservation, 23, 1984, pp. 11429.

44 O. Hafner, Lightfastness of Organic Pigments as a Function of TheirParticle Size Distribution, Journal of Paint Technology, 47, 1975, pp. 649.

45 Ware 1999 (cited in note 28), pp. 1206; P.M. Whitmore, C. Bailie andS.A. Connors, Micro-fading tests to predict the result of exhibition:Progress and prospects, Tradition and Innovation: Advances inConservation; Contributions to the Melbourne IIC Congress, 1014October 2000, eds A. Roy and P. Smith, London 2000, pp. 2005.

46 Mortimer and Rosseinsky 1984 (cited in note 9); P.A. Christensen, A. Hamnett and S.J. Higgins, A Study of Electrochemically grownPrussian Blue Films using Fourier-transform Infra-red Spectroscopy,Journal of the Chemical Society Dalton Transactions, 1990, pp. 22338;A. Dostal, B. Meyer, F. Scholz, U. Schrder, A.M. Bond, F. Marken andS.J. Shaw, Electrochemical Study of Microcrystalline Solid PrussianBlue Particles Mechanically Attached to Graphite and Gold Electrodes:Electrochemically Induced Lattice Reconstruction, Journal of PhysicalChemistry, 99, 1995, pp. 2096103.

47 P.J. Kulesza, M.A. Malik, A. Denca and J. Strojek, In Situ FT-IR/ATRSpectroelectrochemistry of Prussian Blue in the Solid State, AnalyticalChemistry, 68, 14, 1996, pp. 24426.

48 Cyanotypes printed on glass plates coated with alumina or silica, glassfibre papers, or ceramics also show reversible fading, despite theabsence of the oxidisable groups present in cellulose papers. The cyano-types are prepared from solutions containing a citrate, which might actas a reducing agent for the reversible fading of Prussian blue. Ware 1999(cited in note 28), p. 125, and personal communication, Dr Mike Ware,31 January 1997.

49 When a cyanotype on paper that has been exposed to lightdark cyclesuntil the reversible fading no longer occurs is re-sized with a solution ofcritic acid, ammonium citrate or ammonium oxalate, the reversiblefading is again observed on lightdark cycling. Personal communica-tion, Dr Mike Ware, 31 January 1997.

50 A Review of Literature in the Foreign Paint and Varnish Field 1936(cited in note 14). A study of the effect of an oxygen-free environmenton the rate of fading of various pigments found that, unlike most of theother pigments tested, Prussian blue lost colour more rapidly whenoxygen was excluded: J.S. Arney, A.J. Jacobs and R. Newman, Theinfluence of oxygen on the fading of organic colorants, Journal of theAmerican Institute for Conservation, 18, 1979, pp. 10817. Eder (J.M.Eder, History of Photography, trans. E. Epstean, New York 1978, p. 130) gives an account of Desmortiers observations on fading in theabsence of air in the context of dyeing fabrics. Desmortiers emphasisesin his Recherches sur la dcoloration spontane du bleu de Prusse, 1801,that fabrics must have free access to the air in order to reverse the

fading and restore the blue colour of the fabric: Ware 1999 (cited innote 28), p. 114.

51 Mortimer and Rosseinsky 1984 (cited in note 9); Christensen et al.1990, Dostal et al. 1995 (both cited in note 46).

52 Ware 1999 (cited in note 28), pp. 1478, and references therein; the reac-tion path for the formation of Prussian yellow from Prussian blue viaBerlin green seems to differ for the insoluble and soluble forms ofPrussian blue: D.R. Rosseinsky and A. Glidle, EDX, Spectroscopy, andComposition Sudies of Electrochromic Iron(III) Hexacyanoferrate(II)Deposition, Journal of The Electrochemical Scociety, 150, 9, 2003, pp.c6415.

53 Sharpe 1976 (cited in note 18).54 V.B. Nechitayilo, V.I. Styopkin, Z.A. Tkachenko, Y.G. Goltsov, V.P.

Sherstyuk and V.V. Zhilinskaya, Structure and electrochromic proper-ties of ferric aquapentacyanoferrate a new analogue of Prussian blue,Electrochimica Acta, 40, 1995, pp. 25014.

55 Kirby 1993 (cited in note 2), pp. 645.



Supplier

John Nicoll: Celestial blue

John Nicoll: Prussian blue

John Nicoll: Blue

Made from [(NH4)4FeII(CN)6] + FeIII (FeIICl2 + H2O2)

As SA (above), washed twice with water

Made from [K3FeIII(CN)6] + FeIICl2

Made from [(NH4)4FeII(CN)6] + FeIICl2 + H2O2

Aldrich Chemicals (pure FeIII4[FeII(CN)6]3)

E.-L. Richter: Prussian blue, barytes, gypsum

E.-L. Richter: Prussian blue

E.-L. Richter: Prussian blue, gypsum

E.-L. Richter: Prussian blue, kaolin, quartz

E.-L. Richter: Prussian blue, barytes

E.-L. Richter: Prussian blue, gypsum

E.-L. Richter: Prussian blue, barytes

A. Burmester: Doerner Institut Good Berlin blue from Weimard

A. Burmester: Doerner Institut Berlin blued

The Royal Academy: from Ozias Humphreys watercolour box

The Royal Academy: Ackermann cake from Reeves box

The Royal Academy: Random & Sneath cake from Reeves box

The Royal Academy: Chinese blue from Bull box

Winsor & Newton alkali ferric ferrocyanide

Ackermann Prussian blue

Ackermann Antwerp blue


TABLE 1 Details of the early Prussian blue samples and pigments prepared in the laboratory


Date

mid-19th century

mid-19th century

mid-19th century

n/a

180030

180030

180030

180030

180030

180030

180030

11.4.1855

30.4.1842

17421810

17961827

17961827

Extendera

CaSO4: Ca, S, Fe, (K)

hydrated alumina: Fe, K, Al, (S)

BaSO4: Ba, S, Fe, Si, (Al), (K)

none: Fe, Cl

none: Fe, Cl

none: Fe, K, (Al), (Cl)

none: Fe, (Al), (S)

none: Fe, (S)

starchc: Ba, Ca, Fe, S, (K)

starchc: Fe, K, S, (Al)

Fe, K, (Al), (S)

Al, Si, K, Fe, (Ti or Ba)

Al, Ba, Ca, Fe, S, K, Si, (P)

Al, Fe, Ba, Ca, K, S, Si, (Na), (P)

Ba, Fe, S, K, Si, (Al)

Al, Fe, K, S, Si, (P)

Fe, (Ca), (Cu), (K)

Fe, Al, K, S, (Cu), (Si)

Fe, Al, K

Al, Ca, K, Fe, S, (Cu), (P)

Fe, Al, K, Si, (S)

Fe, K, (Al)

Fe, Al, K, (P), (S), (Si)

Al, K, S, Fe, Si, (K), (P)

Ref.

NC

NP

NB

SA

SB

SC

SD

ALP

RA

RB

RC

RD

RE

RF

RG

BA

BB

OZ

AR

RSR

BCB

WN

AP

AA

Shades in oilb

LMD

L2MD

L2MD

L2MD

L2MD

L3L2LMD

L2MD

LMD

L2MD

L3L2LM

L2MD

L3LM

Supplier

The Museum of London: Reeves & Woodyer cake

The Museum of London: Reeves box cakee

The Museum of London: Reeves & Inwood powdere

The Museum of London: Brown of Oxford Street box

The Victoria & Albert Museum: Elizabeth Careys box

The Victoria & Albert Museum: Elizabeth Careys box

Gainsboroughs House: paint bladder

J.S. Mills: paint bladder

The British Museum: Turnbulls blue

Made using a recipe of 1758





Date

180517

18th century

1797?

1791

17701831

17701831

early 19th century

183040

Extendera

Fe, Al, K, (S)

not analysed

not analysed

not analysed

lead white: Pb, Al, Fe (Cu)

none: Al, Fe, S (K), (Si)

Fe, K, (Al), (S), (Si)

Fe, K

Fe, K, (Al), (S)

none: Fe, (S)

hydrated alumina: Al, S, Fe

none: Fe, (Al)

Ref.

MA

MB

MC

ME

VA

VB

GA

JSM

TB

K7

K12a

K12b

Shades in oilb

L3L

L2L

L3L2M

L3L2M

L3L2LM

L2MD

L3L2LMD

L3LM

L3L

Notes

a Determined by EDX analysis. Results indicate the elements detected:strong peaks underlined, minor elements in brackets. In italics = inmounting medium.

b D = deep, M = mid, L = light, L2 = very light, and L3 = extremely light.c Detected by FTIR.d gutes Berliner Blau von Weimar am 11 Apr 1855 gerieben, Berliner

Blau vom 30.Apr 1842 e Despite label, pigment is indigo.

Professor Richters samples are from an old German pharmacy near Darmstadt: E.L. Richter and H. Hrlin, A Nineteenth-CenturyCollection of Pigments and Painting Materials, Studies in Conservation, 19,1974, pp. 7682.

Alum

8

8

6

16

2

8

8

6

4

4

4

4

4

4

4



TABLE 2 Ingredients in eighteenth-century Prussian blue recipes and their proportions

Notes

Brown in his experiments on the methodfound the yield of alkali from thedetonation of the tartar and nitre to be4 parts and

fading and colour change of prussian blue: methods of manufacture and the influence of extenders

Documents