Illuminating the problem of blue verditer synthesis in the early modern English period: chemical characterisation and mechanistic understanding

We present a study into early modern English production of blue verditer, an early synthetic copper-based blue pigment chemically analogous to azurite. Verditers have been identified in numerous wall and easel paintings. While initial documentation occurs in the mid 1500s and production recipes were documented by the 17th c., the synthesis was known to be unreliable. This study replicates historical and recent scientific work on blue verditer and represents a significant advance in our understanding of verditer production and its challenges. Procedures for verditer synthesis are drawn from both 17th c. documentation and 20th c. replication work. The effects of temperature, copper and carbonate sources, solution stirring, copper ion concentration, and atmospheric composition are studied in order to elucidate the mechanism of synthesis and explain its unreliability in early modern refineries. Products are characterised by polarised light microscopy, scanning electron microscopy, Raman spectroscopy, and powder X-Ray diffraction. Rouaite, a green basic copper nitrate, is for the first time confirmed as a product of the refiners’ synthesis and a precursor to blue verditer in laboratory syntheses. This result problematises the blanket identification of green verditer as basic copper carbonate and provides important clues to the mechanism for blue verditer synthesis. Solution chemistry and ion equilibria allow us to explain the route by which rouaite is first formed and then converted to blue verditer. Conditions favouring blue verditer production are also clarified further. Although it is commonly stated that low temperatures are required for blue verditer production, blue verditer is produced here at a range of ambient temperatures. The reaction is found instead to be controlled by solution equilibria and heavily favoured by high partial pressures of carbon dioxide. Alongside archival materials about refining and verditer production, these results are contextualised and explanations for the unreliability of historical synthesis are proposed.


Blue: worth its weight in gold
In the medieval and early modern periods, European painters had few choices when it came to blue pigments.Ultramarine pigment, produced from the rare mineral lazurite (a component of lapis lazuli), was difficult to source and prohibitively expensive for all but the finest work, although it produced the finest blue colour [1,2].Another blue mineral, azurite, produced a good pigment, but was still extremely valuable [3,4].The expense of azurite (and ultramarine) drove painters and scientists to seek out alternative methods of obtaining blue pigments, including by synthetic routes.
Blue and green verditers are synthetic copper-containing pigments.Blue verditer is a copper hydroxycarbonate chemically analogous to azurite (Cu 3 (CO 3 ) 2 (OH) 2 ), while green verditer has long been understood to be chemically analogous to malachite, a copper hydroxycarbonate with alternative stoichiometry (Cu 2 CO 3 (OH) 2 ).This work, however, identifies for the first time a copper hydroxynitrate, rouaite (Cu 2 NO 3 (OH) 3 ), as the main product of laboratory-based replications of the historical verditer production process.Although this work does not directly address the identification of rouaite in historical samples, this is the subject of an additional forthcoming publication.One previous study (of an early seventeenth century miniature painting) has also identified rouaite as a pigment, although its origin is not investigated in detail [5].
Verditers are widely documented in written sources and historical works of art.Due to the expense of alternative blue pigments, the blue colour was much more valuable than the green [6,7].The earliest mention of 'verditer' occurs in 1506-1507.As the origin of the name likely derives from 'verd de terre' , it is assumed that successful production of the blue pigment followed that of the green; the earliest possible mention of blue verditer, 'verditer bis' , 1 occurs in 1537-1538 [8,9].
Blue verditer, identical to azurite when studied by PXRD and Raman spectroscopy, is distinguished from naturally produced azurite by its specific morphology and lack of diverse associate minerals common to natural pigments.Since it is synthesized by precipitation from solution, it exhibits a distinct round shape and consistent, small particle size [6,8,11,18].It appears highly birefringent and lobed under polarized light [6,18].By-products of synthesis such as unreacted calcium carbonate, a synthetic reagent, may be present [19][20][21].Natural calcium carbonate is also a common associate mineral of natural azurite and its presence should not be used to definitively draw conclusions about pigment origin [20,21].
Several common associate minerals of naturally occurring azurite such as earth pigments, cerussite, and kaolinite were also used as pigments or extenders; their presence should not be used to conclusively decide a natural origin.Occurrence of a blue copper carbonate pigment absent of any of these species, however, suggests that the pigment possesses a synthetic origin.Characterisation of impurities and by-products resulting from verditer synthesis may also be useful to conservators in identifying extenders and degraded pigments originally present [11,22].
The interpretation of the original colour of areas painted with verditers is challenging for conservators.Synthesis produced a range of colours from green to blue.Additionally, verditers were often mixed with organic yellow pigments that are difficult to detect and prone to fading.Yellowing of the medium over time may also make a blue pigment appear greener [11,22].Further scientific work on synthetic procedures and degradation reactions is therefore of great value.

Early discussions of verditer synthesis
Early modern synthesis of blue and green verditer was widely understood at the time to be unreliable, with small variations in synthetic procedure resulting in products spanning from green to blue, as shown in Fig. 1 [6,10].Nonetheless, this affordable pigment was evidently widely used, particularly in England [16,[23][24][25]; later treatises note that it was useful for producing vibrant greens when mixed with organic yellow pigments [11, Note that all products are named according to synthesis conditions as outlined in Table 2 26].Its ease of acquisition is reflected in the multiple accounts detailing the procedures for synthesis.
Verditer was produced from copper-water, a byproduct of the refining industries wherein nitric acid (aqua fortis) and copper metal reacted to form copper nitrate.Refiners' operations were often small scale and verditer production tended to be a secondary source of income [7].Christopher Merrit's The Art of Refining (1677-78), describes the process: 'They put into a Tub a Hundred Pound weight of Whiting, and thereon pour the Copper-Water, and stir them together every Day, for some Hours ... when the Water grows pale, they take it out, and set [it] by for farther Use, and pour on more of the Green Water; and so continue till the Verditer be made... ' [27] Other sources include John Harris (early 1700s) and Robert Dossie (1758), providing similar procedures [26,28].Robert Boyle (1669) tells of successful synthesis by heating the solution before reacting it with whiting [29].Boyle may have been intentionally misled by a refiner looking to protect trade secrets [6].The problem of unreliability was paramount for verditer producers, as Merrit notes in his appendix to a translation of an Italian glassmaking treatise published in 1662: 'they can assign no reason, nor can they hit on a certain rule to make constantly their Verditer of a fair Blew, to their great disprofit, the Blew being of manifold greater value than the Green.' [30] Further complicating the discussion of verditer synthesis is evidence of many other syntheses for blue pigments, likely cuprammonium salts, dating back as far as the eighth century [31][32][33][34][35].These syntheses are outside the scope of this work and products have not been detected in many surviving samples, though the possibility of their use should not be ruled out.

Previous scientific research
MacTaggart & MacTaggart studied the products of historical verditer syntheses in 1980, which provided a basis for this work.The green and blue synthetic products were characterized by polarised light microscopy (PLM).They found that the product colour was dependent on the synthesis temperature; the blue product was not formed above 12°C, a result that this work brings into question [6].
This led to the hypothesis that the success of the English verditer industry was due to cold English winters, while syntheses were not successful during warmer months.Historians have cited this work to explain why the English verditer industry succeeded where those in warmer continental Europe failed [8].The mechanism for successful blue verditer synthesis, however, has not yet been elucidated, and has important implications for the historical understanding of pigment production and availability in the early modern period.

Objectives and impact of research
Although previous scientific research has investigated the products of the 'refiners' verditers' synthesis method, the products of this synthetic procedure have not been previously characterised by chemically specific analytic methods.This work combines visual analysis, electron microscopy (SEM), Raman spectroscopy, and Powder X-Ray Diffraction (PXRD) to study the products of verditer synthesis.
The objectives of this work, therefore, is to address these open questions: -What products are formed in the "refiners' verditers" synthesis?-Does the identity of products explain variation in colour?-Is it possible to identify plausible reaction mechanism(s), and explain the dependence on various experimental/environmental variables?-Do the variables controlling product outcomes suggest any historical influences leading to successful or unsuccessful verditer production, and are these in turn supported by primary source material?
The products formed in these reactions are also relevant as copper corrosion products in both industrial and archaeological settings [36][37][38][39][40][41], and as synthetic catalyst precursors for mediation of various chemical contaminants [42][43][44][45][46][47].Their formation and reactivity under a range of reaction conditions, subject to significant debate in the literature, is therefore relevant beyond the area of pigment studies [39,48,49].
The use of 'verditer' throughout the text refers to blue verditer, the synthetic analogue of azurite.Azurite refers to the naturally occurring mineral.'Green verditer' indicates the green synthetic pigment often previously identified as synthetic malachite, instead found here to be rouaite, a synthetic copper nitrate.It is important to note that blue verditer and azurite (and similarly, synthetic malachite and malachite) share the same Raman spectra and powder X-Ray diffraction patterns, and therefore, reference data from the American Mineralogist Crystal Structure Database (AMSCD) for azurite and malachite are referenced throughout the text under the mineral names.The terminology has been chosen to reflect the historical texts and previous scientific work on these materials and to minimise confusion.Table 1 summarises the nomenclature used in this work.

Synthesis reactions
All reaction procedures studied are summarised in Table 2, and products are named and referred to throughout the text according to the relevant reaction conditions used.

Reference materials
A reference sample of commercially produced blue verditer was purchased from L. Cornellisen & Son and compared to lab synthesized pigment samples.Reference samples of green verditer and natural malachite were obtained from Dr. Andrea Kirkham.

Standard synthesis following procedure of MacTaggart & MacTaggart: Rx_Air_4_int
0.31M Solid copper nitrate trihydrate (Scientific Laboratory Supplies, > 98% purity) was dissolved in deionised water in a wide-mouthed glass jar by stirring with a magnetic bar.While solution was stirred, 0.3M solid calcium carbonate (Acros Organics, 99% extra pure) was added (Table 1).The solution was stirred for seven hours either  Following the first 7 h reaction, the solution was stirred twice in the next 24 h.The solid product was then filtered using a Buchner funnel and filter paper and rinsed with deionised water.Subsequent reactions, referred to in the text as second and third washes, of the same product were then carried out, in which additional new calcium carbonate and copper nitrate solution in the same concentrations and quantities as in the first reaction were added to the product and reacted under the above procedure.This process followed the historical procedure outlined by Merrit in The Art of Refining.[27] The final product was then filtered using a Buchner funnel and filter paper, rinsed with deionised water, and dried prior to analysis.
Additional reaction routes are presented in detail in Additional file 1 and summarised in Table 2.

Thermal control
Reactions were carried out at several temperatures.Hot conditions were created by carrying out reaction on a heated stir plate (Rx_Air_50_int).Temperatures were monitored and observed to fluctuate between 45 and 55 °C.Cold conditions were created by placing the stir plate and vessel inside a refrigerator at 4 °C (Rx_Air_4_int).For reactions carried out at room temperature (Rx_Air_20_ int, cont), we have reported the temperature as 20 °C throughout the text for the sake of consistency.The temperature of the laboratory underwent minor fluctuations between 17 and 21 °C.
Selected reactions at 20 °C and 50 °C were also replicated with thermal control to ± 0.1 °C using a thermometer with PID controller (IKA ETS D5).These are denoted by TC in name.Precise thermal control was not found to alter product identity or proportion.

Alteration of atmospheric composition
Rx_N2_20_cont.Removal of atmospheric CO 2 was carried out by two methods.A glove box containing reagents was purged with nitrogen.Then reagents were combined and allowed to react for seven hours followed by immediate filtration.The second method used an inlet line connected to dry nitrogen and an outlet line into a fume hood to purge the air in the reaction vessel (sealed using parafilm).Calcium carbonate was added quickly following purge and vessel was resealed.Reaction was then carried out followed by immediate filtration of products.
Rx_CO2_20_cont.Creation of a CO 2 atmosphere was carried out by installing an inlet line connected to a CO 2 cylinder and an outlet line into a fume hood in the space above the solution in the reaction vessel (Fig. 2).The chamber was sealed using parafilm and purged with CO 2 for 5 min, after which calcium carbonate was added, the vessel was resealed, and reaction was carried out under CO 2 flow (20 cm 3 /min).The product at an intermediate time point, four hours, was sampled by removal of product suspended in solution from reaction chamber by needle followed by drying and analysis.After 7 h, the gas flow was stopped, the vessel unsealed, and the product filtered immediately.
Rx_CO2var_20_cont.For intermediate levels of CO 2 (9500, 20,000, and 50,000 ppm), multiples of a set volume of CO 2 (corresponding to 3500 ppm as measured using gas balloon and CO 2 monitor) was introduced into a sealed glove box and allowed to equilibrate (as evidenced by CO 2 monitor, RS Pro DT-802).The reaction was then carried out as described inside the glove box.
Rx_SynAir_20TC_cont, int.Selected reactions were also replicated under synthetic air.The ratio of N2:O2 was 80:20 (controlled using flow meters, 160 cm 3 /min: 40 cm 3 /min) with an overall flow rate of 200 cm 3 /min and the same setup as that used for creation of CO 2 atmosphere.

Product characterisation Characterisation by Raman spectroscopy
Raman spectra were collected using a Horiba LabRAM HR Evolution confocal Raman spectrometer, with a 50 × microscope objective (Olympus LMPLFLN), a 600 grooves/mm grating, 100 µm pinhole, and a CCD array (Hamamatsu, 1024 × 1024 pixels).532 nm excitation laser (diode-pumped solid-state, Laser Quantum) was selected based on superior S/N ratio with minimal sample damage per previous laboratory work and literature [50].Spectra were collected at 10% or 25% ND filter (approximately 1.25 mW power at surface), 10s acquisitions, and 20 accumulations in the range 100-1800 cm −1 .Fig. 2 CO 2 atmosphere reaction vessel setup; inlet and outlet CO 2 lines are secured in vessel Spectra were processed using OriginPro 8 software.All spectra were fitted to a spline baseline, and anomalous noise peaks were removed in the spectra presented to aid in interpretation.

Characterisation by powder X-Ray diffraction
Powder X-Ray diffraction (PXRD) data were collected on a Malvern Panalytical Empyrean instrument, equipped with an X'celerator Scientific detector using non-monochromated CuKa radiation (l = 1.5418Å).The sample was placed on a glass sample holder and measured in reflection geometry with sample spinning.The data were collected at room temperature over a 2q range of 3-80°, with an effective step size of 0.0167° and a total collection time of 60 min.

Characterisation of powder by scanning electron microscopy/ energy dispersive X-Ray spectroscopy (SEM/EDX)
SEM images were collected on a TESCAN MIRA3 FEG-SEM at magnifications up to 40x.Samples were coated with platinum prior to imaging.Elemental mapping by EDX was carried out using a JEOL JSM-5510LV scanning electron microscope (backscatter electron detector, 20 mm working distance, 15 keV accelerating voltage) coupled to an Oxford instruments INCA EDX detector and software (version 4.01).

Colorimetry
Colorimetry data were collected using a Minolta CR-221 colorimeter with D65 illumination source in the CIELAB colour space.Three data collections were averaged for each sample and standard deviations are presented.

Summary table of synthetic routes and resulting product(s)
The syntheses studied are summarised in Table 2, which details the reaction conditions as well as resulting products for each route.Reactions are named systematically and referenced throughout the text as such.

Initial results of syntheses: detection of an unexpected copper nitrate alongside blue verditer
Previous work reports the formation of a variety of products ranging in colour from green to blue [6].This range was observed in repetitions of syntheses in this work as well, with most results under standard reaction conditions producing an intermediate blue-green product.
Characterisation of the product by Raman spectroscopy confirmed the presence of blue verditer (Cu 3 (CO 3 ) 2 (OH) 2 ) with characteristic bands at 400, 831, and 1090 cm −1 .Representative Raman spectra for samples synthesized under cold conditions with intermittent stirring are presented in Fig. 3. Unreacted calcium carbonate was also observed.In addition to blue verditer, a green basic copper nitrate was identified, rather than the expected malachite (Cu 2 CO 3 (OH) 2 ).
Careful analysis of Raman spectra as well as PXRD (Fig. 4) revealed that this compound was rouaite (Cu 2 NO 3 (OH) 3 ), the metastable monoclinic polymorph of gerhardtite that is most commonly observed in synthetic samples [51].PXRD analysis additionally showed malachite present as a minor component, as well as unreacted calcium carbonate.Rouaite exhibits characteristic Raman bands at 404, 454, and 1046 cm −1 (marked * in Fig. 3a), and can therefore be readily differentiated from verditer and synthetic malachite.Verditer, on the other hand, shows characteristic Raman bands at 400, 831, and 1090 cm −1 (marked + in Fig. 3a).
Microscopic analysis of samples, on the other hand, showed similarity between samples composed of primarily rouaite those composed of primarily verditer (Fig. 5).Examination of synthesized verditer/rouaite mixtures by polarised light microscopy, used in previous studies of these experiments to characterise blue and green verditer, showed structures in agreement with earlier studies of verditers as well as commercial verditer: small, lobed crystals that are highly birefracting [6].
Rouaite, with chemical formula Cu 2 (NO 3 )(OH) 3 , has not been identified in reference synthetic green verditer samples studied previously (Additional file 1: Fig. S1), and has not been identified in historical wall painting samples to date, although in a later publication Mactaggart and Mactaggart suggested that the green synthetic product may be gerhardtite (rather than malachite) [53].Rouaite has also been noted as a common impurity in the synthesis of other mixed-metal hydroxycarbonates, suggesting that its presence in this synthesis should not be unexpected [42].
Product morphologies were studied further by scanning electron microscopy of coated particles.Rouaite shows a distinct tabular morphology, and is readily differentiated from verditer, which appears as smaller intersecting plates forming roughly spherical aggregates (Fig. 6).The appearance of rouaite observed here is consistent with samples synthesised by precipitation published elsewhere [49].The appearance of verditer is consistent with that of various reference samples that have been studied previously as part of ongoing work on its characterisation as well as results published elsewhere [6,18,19,54,55].
Products were more blue following subsequent reactions (second, third wash) of the same product in which additional unreacted carbonate and copper nitrate solution were added to product and reacted.Figure 7 shows the products resulting from one, two, and three washes with corresponding CIEL*a*b* colour values.After subsequent washes, there is a shift to more negative b* (more blue) and less negative a* (less green).While this multistep procedure was initially studied due to its presence in verditer recipes, these results led to a crucial hypothesis about the synthetic mechanism.
The result suggests that the formation of verditer proceeds via a gradual displacement following initial formation of rouaite.The shift in relative verditer proportion was confirmed by PXRD of products following one, two, and three subsequent reactions (Fig. 7).Integrated peak area ratios presented in the inset table in Fig. 8 confirm that, following subsequent washes, azurite increases in abundance relative to rouaite in the product.This is consistent with visual and colorimetric examination of products (Fig. 7 and Additional file 1: Table S2), allowing the use of colour as a proxy for relative phase composition.
The conversion of rouaite to verditer is dependent upon the replacement of nitrate anions with carbonate anions.It was therefore proposed that the synthesis may be influenced by the concentration of carbonate anions in solution and therefore the carbonate-carbon dioxide equilibrium as shown in the following reactions: (1) This equilibrium is dependent on factors such as partial pressure of carbon dioxide (pCO 2 ) and solution pH.
The conversion of rouaite to verditer may be analogous to 'hydroxy double replacement reactions' in which anions in a layered salt structure are loosely bound and therefore readily exchanged.Rouaite has been used as a precursor to synthesise organometallic catalysts in which nitrate anions are replaced by organic anions [49,56,57].
While nitrate anions are loosely bound in layered hydroxy salts (Fig. 9a), the larger negative charge of carbonate anions means that they are not readily exchanged once bound within the crystal structure.Figure 9b shows the crystal structure and bonding in azurite/verditer, in which carbonate anions are multiply bound to nearby copper ions; it is apparent that this crystal structure is radically different, a possible explanation for the irreversibility observed.
Several other copper minerals (including langite, botallackite, and paratacamite) as well as organometallic catalysts form via facile ion exchange with rouaite, in which loosely bound intercalated nitrate anions are exchanged for loosely bound chloride and sulphate anions [49].The original crystal habit, or shape, is often maintained in the product [48,49,58].
It is also possible that conversion occurs via dissolution and recrystallization [49].The structural similarity of azurite to rouaite, both in terms of lattice parameters and bonding of carbonate anion, is low, as observed in Fig. 9, suggesting that dissolution and recrystallisation is more likely than anion exchange [49,59].In this case, the structural and electronic differences between rouaite and azurite still hold as an explanation for the irreversibility for conversion.

Effect of temperature: results inconsistent
Initial work focused on the possible thermal dependence proposed by MacTaggart & MacTaggart and widely accepted as an explanation for the superior quality of English blue verditer over French verditer produced in the early modern period [6,8,9,24].Reactions at ambient room temperature (approximately 20 °C) produced mixtures of rouaite and verditer in varying proportions (Fig. 10, Additional file 1: Table S2).Intermittent stirring slightly favoured verditer formation compared to continuous stirring, but this is inconsistent as well.
Although the variability of results suggests that temperature may play a minor role in the successful production of blue verditer, it did not control the synthetic outcome and does not explain why the reaction succeeds or fails.A cooler climate (therefore colder reaction temperatures) does not explain the production of a blue product rather than a green one.
The effect of raising the temperature to 45-55 °C was then evaluated.This produced primarily rouaite with  traces of malachite, suggesting that there is a temperature threshold above which verditer production is disfavoured, although this is well above the range of historical ambient air temperatures in England (not exceeding an average of 20 °C in summer and not falling below an average of 3 °C in winter during the sixteenth and seventeenth centuries) [61,62].
The effects of temperature are still not fully explained.There was significant variability in the hue and relative proportion of verditer/rouaite of products produced at the same temperature and reaction conditions (Fig. 10).It was therefore proposed that temperature may play a role in influencing another factor not previously identified.As solution equilibria may be perturbed by temperature changes, and the rouaite to verditer conversion reaction is dependent on ion concentrations, the influence of carbon dioxide partial pressure (pCO 2 ) and therefore carbonate ion concentration was investigated.

Effect of atmospheric carbon dioxide concentration: is control over bicarbonate equilibrium the key?
Variation in the atmospheric pCO 2 , an equilibrium sensitive to environmental factors such as temperature, was therefore proposed to influence the carbonate ion availability in solution and therefore verditer production.Initially, extreme conditions were used to study the reaction under very low (nitrogen atmosphere) and very high pCO 2 values (100% CO 2 at 1 atm).Some verditer formed in very low abundance alongside mainly rouaite under a nitrogen atmosphere (approximately zero pCO 2 ) and under a synthetic air atmosphere (80% N 2 , 20% O 2 ).The absence of CO 2 in the atmosphere disfavours the presence of carbonate in solution and favours production of carbon dioxide.Raman and PXRD data confirm the presence of primarily rouaite, as presented in Fig. 11.This result, though not historically relevant, is extremely useful in understanding the mechanism of synthesis.It suggests that successful verditer synthesis requires high carbonate concentration in solution, beyond that generated by dissolution of calcium carbonate, while rouaite formation occurs regardless.We note that reaction under synthetic air produces very slightly more verditer (Fig. 11b), which may suggest that stability and reactivity of rouaite are dependent on oxygen levels as well, although further work is necessary to comment on this conclusively.
Extremely pure verditer was observed after reaction under 100% CO 2 at 1 atm.This product was identified by Raman and PXRD analysis (Fig. 12).Rouaite is present after 4 h reaction time (sampled by removal of product suspended in solution from reaction chamber by needle) and absent by 7 h; this suggests that the mechanism of formation of verditer is not altered, but rather, the conversion of rouaite to verditer is greatly facilitated by increased pCO 2 .Visually, the product is a deep blue colour (Fig. 2).
At lower, but still elevated relative to atmospheric, concentrations of CO 2 (9500, 20,000, 50,000 ppm), however, rouaite was still the major product of the synthesis based on PXRD analysis at room temperature after 7 h as shown in Fig. 13.It is therefore likely that elevated atmospheric CO 2 alongside additional factors influencing carbonate concentration would be necessary to definitively control synthesis.

Discussion
The concentration of carbon dioxide in the air above the open reaction vessel was found to significantly affect the composition of products.Verditer formation was facilitated by running reaction under 100% CO 2 at 1 atm, and suppressed by purging with N 2 , removing atmospheric CO 2 as well as any evolved during the reaction.Although verditer formation was suppressed above 45°C, typical ambient temperature fluctuations corresponding to climatic variation in seventeenth century England (4°C for cold synthesis up to 17-20 °C, laboratory room temperature) did not influence the results.When verditer formation was suppressed, rouaite formed as an insoluble metastable product and was isolable.The conversion of verditer to rouaite was not observed, suggesting that Fig. 8 PXRD patterns of Rx_Air_4_int after first, second, and third washes.Data are normalized to the intensity of the rouaite peak at 12.8, and peak intensities of verditer bands at 17.2 and 17.4 are observed to increase following the second and third washes.Inset table provides the integrated peak area ratio A(rouaite)/A(azurite) determined by integrating the peak areas (12.0-13.25 2θ for rouaite, 16.5-18.02θ for azurite) from the non-normalised patterns for each sample.The integrated peak area ratios similarly show increasing azurite content relative to rouaite following subsequent washes verditer is significantly more stable than rouaite and the reverse reaction is unlikely to occur.
Perturbations to the bicarbonate-carbon dioxide equilibrium significantly affect the composition of products, and may explain the unreliability of historical verditer synthesis [6,29,30].There are several competing reactions occurring in solution, with reversible reactions indicated by double headed arrows: Formation of rouaite: Rouaite to azurite: Rouaite to malachite: (3) Azurite to malachite: Bicarbonate equilibrium: Dissolution: Carbon dioxide generation, favourable at low pCO 2 (here, when system is run under N 2 ), competes with the formation of blue verditer and synthetic malachite from rouaite in solution.At high pCO 2 , the concentration of carbonate ions in solution increases; subsequently, rouaite is converted to verditer.Carbonate concentration is also dependent on the solubility of CaCO 3 , which is low in water and decreases as temperature increases (6) [63].Higher pCO 2 (here, running the reaction at 20,000 ppm, 50,000 ppm, or 1atm CO 2 ), on the other hand, increases CaCO 3 solubility [63].It should be noted that the temperature dependence of CaCO 3 solubility is strongly influenced by the temperature dependence of CO 2 solubility, which decreases at lower temperatures [63,64].High pCO 2 values are reported in previous analysis of natural azurite formation [58] and thermodynamic calculations [65], so this result is not unexpected.However, the CO 2 pressure required to attain complete conversion of rouaite to azurite in this work is much lower than those reported for formation of azurite from Cu(NO 3 ) 2 •3H 2 O [65].
There is some evidence from previous studies of patina formation that rouaite is formed readily at high temperatures (> 100 °C) [37,38,40].Although the conditions of patina formation on a surface are significantly different to those of inorganic synthesis studied here, rouaite's stability as temperature is increased may help to explain the formation of pure rouaite at > 45 °C.
It is significant that verditer formation was not observed with a highly soluble carbonate, Na 2 CO 3 , which is also a stronger base.It is relevant to note that others Fig. 11 a PXRD results of Rx_N2_20_cont (black) and Rx_SynAir_20TC_cont (orange) with verditer, rouaite, and malachite patterns provided for comparison [52].The two conditions produce very similar patterns which correspond closely with reference for rouaite.b A smaller region of the PXRD pattern is shown, with grey regions highlighting very weak peaks in samples 1-3 corresponding to verditer [52] have synthesized azurite (CaCO 3 , RT, pCO 2 = 5.0 MPa for 48 h) and malachite (Na 2 CO 3 , 353K, pure CO 2 atmosphere) analogues from carbonates and copper nitrate, but at significantly higher pCO 2 values [65].
Malachite was not readily produced at the copper concentrations studied.Whilst malachite is more thermodynamically stable at low pressures of CO 2 , these calculations assume equilibrium conditions [39,48,66].This is a disequilibrium system, so we are wary of relying on these calculations to explain the products observed.We observed that synthetic malachite forms at lower Cu 2+ concentrations, and in combination with rouaite and blue verditer when solutions were left undisturbed, suggesting that there may be some diffusion/mixing condition affecting its formation (see Additional file 1).It is not observed in most samples regardless of colour.
The mechanism of conversion from rouaite to verditer is also of significant interest, both in explaining the results and for work developing copper nitrate and carbonate catalysts.Conversion may occur via anion replacement or dissolution and recrystallisation.The crystal morphology observed does not clearly support one route over the other.Well-formed tabular rouaite crystals are present and do not appear to be disintegrating (Fig. 6), while verditer crystals form as rosettes or intersecting discs rather than tabular crystals.The relative abundance of rouaite decreases following subsequent reactions, however, suggesting that conversion is occurring.The dissimilarity in structure of the crystals and in the samples observed by SEM suggests that the likeliest explanation is dissolution-recrystallization, but further work would be useful to confirm this.

Conclusions
The results of this work represent a significant development in our understanding of English verditer synthesis as practiced during the early modern period, as well as of the synthesis and stability of basic copper nitrates and carbonates more generally.Synthetic conditions were examined to understand the influence of conditions that may have been relevant to the success or failure of verditer production historically as well as to clarify sequence of reactions and byproducts formed.Although modern laboratory conditions cannot perfectly replicate those of pre-industrial production (and documentation of historical conditions is incomplete), it is nonetheless valuable to study verditer production using a combination of modern methods and attention to historical documentation [67].
The production of rouaite, an unusual basic copper nitrate previously unidentified as a verditer synthesis product, suggests that, in early modern English verditer production, green byproducts may have been composed of primarily rouaite and only occasionally malachite depending on synthetic conditions.Rouaite therefore may be a marker in historical samples for pigment produced by the refiners' synthesis.Rouaite has not been widely identified in historical samples to date, although it is relevant to note that it was detected in a miniature painting dated to 1617, used in combination with massicot to paint a green doublet [5].Its presence in wall painting samples approximately dated to the period in which the refiners' synthesis is the subject of a forthcoming publication, in preparation.
The difficulty of detecting light elements using elemental analysis and its visual similarity to other green pigments in cross section may explain its sparse identification thus far.It is also possible that rouaite is highly susceptible to degradation over time; future work should address this issue.The identification of rouaite also challenges the widespread assumption that green verditer Fig. 12 a Raman spectra of Rx_CO2_20_cont sampled at 4 h (black) and 7 h (red).Reference Raman spectra of verditer (blue) and rouaite (green) are provided, and the strongest rouaite band at 1046 cm −1 is marked in the 4 h spectrum (*) [52].b The PXRD pattern of the 7 h product is compared to the reference PXRD pattern for azurite/verditer, confirming the identity of the product as verditer with no detection of rouaite pigment is comprised of the more stable basic copper carbonate, analogous to malachite.
The dependence of verditer colour on reaction temperature was found to be much weaker in this study than previously reported.Blue products were synthesized at ambient temperatures within the range of historical climatic variation in England.The success of English producers compared to those in continental Europe cannot be explained solely by direct climatic differences.Alternative explanations for the unreliability of the synthesis were therefore considered.
The synthesis was studied under a range of atmospheric gas concentrations, temperatures, concentrations, carbonate sources, and water sources.Factors influencing the concentration and availability of carbonate ions in solution were found to control the conversion of rouaite to blue verditer.
Most significantly, variation in the partial pressure of carbon dioxide during the reaction was found to affect the identity of the resulting product.At low concentrations of CO 2 , blue verditer formation is suppressed, and the resulting product is primarily At high concentrations of CO 2 , the sole product is blue verditer.This effect is rationalised by an examination of studies on the stability of copper nitrate and carbonate and the complex chemical equilibria at play during this synthesis.
Although it is highly speculative, potentially significant variation in pCO 2 due to the location of synthesis and factors such as ventilation, indoor fires, and exhalation may have contributed to the observed unreliability of verditer synthesis.Documentation of refining procedures and workspaces in London and Sheffield, where verditer production is recorded, indicate that refining was done indoors in spaces often also containing smelting furnaces; this would have increased pCO 2 during synthesis, and these conditions may have been seasonally and regionally dependent [7,69,70].Other historical factors influencing the carbonate equilibrium, such as local variations in water mineral content, may be of interest for further research.
This work represents a significant step in the understanding of verditer synthesis, and a novel contribution to literature regarding pigment synthesis and reproduction of historical recipes.Rouaite is reported for the first time as a major byproduct of blue verditer synthesis replications in the laboratory.Access to modern chemical analysis allowing the conclusive detection of rouaite Fig. 13 PXRD patterns of Rx_CO2var_20_cont (black, 100% CO 2 at 1 atm; red, 9500 ppm; blue, 20,000 ppm; yellow, 50,000 ppm) at 20 °C, with continuous stirring.The dominant product for all lower pCO 2 values is rouaite.Inset shows zoomed in region of verditer peaks, emphasizing minimal verditer production at all values studied was critical to this work, showing the value of species and phase specific methods for researchers in heritage science.

Fig. 1
Fig. 1 Verditer synthesis under varying experimental conditions produces a range of colours due to varying proportion of two mineral species, verditer and rouaite (Cu 2 (NO 3 )(OH) 3 ); a Rx_Air_50_int.b Rx_Air_4_S1.c Rx_N2_20_cont. d Rx_Air_20_S3.e Rx_Air_4_int, three washes.F Rx_Air_20_int. g Rx_Air_20_int.h Rx_CO2_20_cont.Note that all products are named according to synthesis conditions as outlined in Table2

Fig. 9 Fig. 10
Fig.9 Crystal structures and information for a-c rouaite and d-f azurite.Nitrate ions in rouaite are singly bound to copper ions and located in intercalated layer.Carbonate anions in azurite are multiply bound to copper above and below.Structures visualised using VESTA (Version 3.5.8)[60]and data from the American Mineral Society Crystal Database[59]

Table 1
Summary of nomenclature

Table 2
Summary of synthetic routes studied