An analytical survey of zinc white historical and modern artists’ materials

This study is the first systematic survey of a large corpus of zinc white (ZnO) artists’ materials. Zinc white is a white pigment developed within the wave of 19th-century technological developments in the paint industry. The composition, particle morphology and size, and luminescence of 49 zinc white samples from artists’ materials were characterized, including three references of known synthesis methods (indirect and direct) and synthesized by the authors (ZnO nanosmoke). The corpus included historical and modern zinc white pigment powders and paint materials from the leading European and American color manufacturers. The study aims to characterize and evaluate the variability of the properties of zinc white and its paint formulations. The reference materials presented properties in agreement with the literature: indirect ZnO exhibited submicron prismoidal blue-luminescent particles of higher purity than direct ZnO, which had larger acicular green-luminescent particles. ZnO nanosmoke presented acicular (tetrapod-like) blue/green-luminescent nanoparticles. Composition, particle morphology, size, and documentary sources suggested a production via the indirect method for the analyzed corpus. However, the luminescence behavior was more complex to interpret. The fundamental emission of ZnO was not always detected, even in pure ZnO powders. Three trends were identified: smaller ZnO particles for the most recent samples; green luminescence connected to larger particle size; fewer trace elements, and of the same type (i.e., lead, sulfur) for historical materials. Another interesting finding was the detection of hydrozincite in some powders, likely a degradation product of ZnO. In terms of methodology, cathodoluminescence proved a valuable tool for pigment identification. The study provides a database of zinc white references for pigment and artwork analysis.


Introduction
Zinc white (zinc oxide, ZnO) is a modern pigment, a product of the technological and industrial development of the 19th century.Though known since antiquity for ointments and as a sub-product in brass manufacturing, it was only at the end of the 18th century that it started to be studied as an alternative to the white par excellence, the toxic lead white [1].At the time, zinc white was praised for not being a risk to human health, not darkening upon exposure to sulfurous gases, and for being permanent.Its covering power in oil is weaker than lead white's [2], which makes it more suitable for transparency effects and to be mixed with other pigments [3].
ZnO is used not only for paint materials (e.g., watercolors, oil colors, pastels, house paint, mastics, acrylics) but also as a vulcanization accelerator for rubber production, as well as for applications in ceramics, electronics, polymers, cosmetics, pharma [4] and nanotechnology [5][6][7][8].As a pigment, it has been used as white as-is (in house paint and fine arts), but also for ground layers (e.g., by Pre-Raphaelites [9] and some 20th-century American artists [10,11]).Still commercialized today, zinc white has been widely replaced by titanium white since the second half of the 20th century and can often be found in titanium white tubes to adjust its tinting strength [12].
It is manufactured by two main methods, both pyrometallurgical [1]: -The French or indirect method, ~ 80% of today's ZnO production [9], which uses metallic zinc as a raw material; -The American or direct method, where zinc ores and a reducing agent (e.g., carbon coke) are heated up to reduce zinc compounds to zinc before oxidation.
Other methods, such as the wet chemical process, are used to produce ZnO for different applications, such as in the textile [4] and rubber industries.
The first successful trials of the synthesis of zinc white were performed by the chemists Jean-Baptiste Courtois (1777-1838) and Louis-Bernard Guyton de Morveau (1737-1816) in Dijon (France) around 1780 after some early tests in Germany [14].In 1796, the English colormaker John Atkinson of Harrington issued a patent for the manufacturing of ZnO in the United Kingdom [14].In 1834, Winsor & Newton first sold Chinese white for watercolors [15].Large-scale production of the pigment for oil colors only started in the second half of the 19th century, when the architectural painter Edme-Jean Leclaire (1801-1872), influenced by the experiments of the French inventor Stanislas Sorel (1803-1871) [16], developed the so-called French or indirect method.He also discovered a manganese-based drier able to overcome the longer drying time of the pigment, the major drawback to its use [14].
Shortly after developing the indirect process, Leclaire collaborated with the leading European zinc manufacturer, the Belgian company Vieille Montagne, which exploited calamine deposits, 1 mainly in Belgium [17].The produced pigments were then sold to the main color merchants of the time, such as Lefranc [18], Sennelier 2 , and Winsor & Newton [15].
From the second half of the 19th century, many technological developments improved pigment manufacturing (e.g., reverberatory 3 and muffle/retort 4 furnaces).However, parameters such as temperature and airflow were still hard to control [19].
In the 1850s, the New Jersey Zinc Company, taking advantage of the discovery of franklinite 5 deposits, patented an alternative method of ZnO manufacturing in the United States, the so-called American or direct method.In 1892, the French process also started being used in the United States, becoming the most widespread production method of ZnO worldwide [1].
At the beginning of the 20th century, scientists realized that zinc white was linked to some degradation issues like chalking of house paint [20,21] and metal soaps [22][23][24][25][26].These phenomena can change the visual appearance of paintings and provoke cracking, delamination, and paint loss in the most severe cases [10,23,24,26,27].Thus, it would be important for conservators to have an easy method to identify zinc white in artworks and pinpoint early warning signals of its degradation [19].
A considerable amount of studies have already been performed on the interaction of zinc white with oils and the formation of metal soaps [3,[28][29][30][31][32][33][34][35], its photoluminescence [20,[36][37][38][39][40], composition [18], rheological [41] and mechanical properties [42], as well as its use and degradation in watercolors [9], oil paint [27,43,44] and grounds [10,11].Moreover, the pigment has been used for restoration since the 19th century [45].It is still on the market today, even though many artists are unaware of its properties, use, and related degradation phenomena [24]. 1 Smithsonite, a zinc carbonate mineral containing some lead. 2 Personal communication. 3 Raw materials are isolated from the fuel, but not the combustion gas. 4 Heat is applied to a vessel that contains the raw materials, so that emitted gas and products are carried away to a separation/collection section. 5A mineral composed of iron, zinc, and manganese oxides.
Different types of zinc white have been identified, but as Eastaugh et al. [19] pointed out, the cause-and-effect relations of these remarks are still not completely understood.They tackled this issue by studying documentary sources and establishing historical synthesis patterns.Other research groups analyzed historical zinc white artists' materials, focusing on specific examples, such as the French brands Lefranc and Ripolin [18,38,[46][47][48].
Our study is the first systematic survey of a large and unique corpus of zinc white historical and modern artists' materials (e.g., pigment powders, paint tubes) from Europe and the United States of America, gathered with the help of museums, cultural institutions, and research centers.This research aims to correlate the synthesis method of zinc white to its properties to identify potential markers for identifying and classifying the pigment and create a reference database for painting conservation and authentication.
The corpus of samples and the methodology used are presented first.The results are then described and discussed in light of historical and modern scientific literature.

Samples
The corpus consisted of 49 samples referred to as zinc white donated by museums, researchers, foundations, a company, or synthesized (some examples in Fig. 1).It covered a significant timeframe, from the end of the 19th century until today, even though it was impossible to date all the samples accurately.In this paper, the samples dating back to the 19th-20th century are indicated as "historical"; those produced or bought in the 21st century are referred to as "modern".
Table 1 below provides an overview of the samples, classified into three categories: • Reference materials of known production methods; • Samples from ZnO manufacturers; • Samples from colormen and paint manufacturers.
The samples were either in the form of powder (referred to as "powder") or of powder ground in a binder (referred to as "paint").Details about the analyzed samples classified by manufacturer/supplier, including their product description, are available in Additional file 1: Table S1.

Reference materials
Three modern materials, all in the form of powder and of known production methods, were used as a reference for the study: • ZnO nanosmoke synthesized by the authors as described in Zhang et al. [49], well-known in terms of size, shape, crystallographic characteristics, optical properties, point defects, and surface reactivity; • ZnO manufactured by the German chemical company Brüggemann, 6 either by direct or indirect method.
Paint mockups were also prepared by grinding these powders in linseed oil.

Samples from ZnO manufacturers
The fifteen samples from ZnO manufacturers were all powders without specifications of their production method.Two of them were modern materials produced by Kremer; the remaining thirteen were historical samples, including different grades (cachets or, argent, blanc, rouge, bleu, 7 in order of purity) from the Société de la Vieille Montagne (Liège, Belgium).Two of them (cachets argent, bleu) were described by the manufacturer as "lab samples", while the other four (two cachet blanc, one cachet or, and one cachet rouge) were from the production facility Valentin-Cocq (Liège, Belgium).In addition, samples of four different grades (Serena witzegel, Grijsegel n°3, Serena roedzegel, Serena roedzegel n°1 8 ) from another historical ZnO manufacturer in the Netherlands, the Maastrichtsche zinkwit Maatschappij, were also investigated.

Samples from colormen and paint manufacturers
The remaining thirty-one samples of the corpus were materials labeled as zinc white from a large variety of colormen (~ 70% historical, from the end of the 19th century until the 1970s, ~ 30% modern).Many of the leading 19th-20th century European and American ZnO manufacturers and colormen (e.g., Lefranc, Sennelier, Ripolin, Winsor & Newton, Maimeri, Grumbacher) were represented (Table 1).Products from these color-makers were used by the foremost 19th-and 20th-century artists, whose artworks are now in different conservation states [11,22,23,[50][51][52][53][54][55].Paint materials were mainly paint tubes except for a few samples: two fragments from Maimeri and Ripolin color charts, a sample of 1950s Lefranc paint from a canvas, and a Lefranc & Bourgeois pastel.
Table 2 summarizes the techniques used with the addressed property, sample preparation, and number of analyzed samples.

Optical microscopy
Each sample was observed as-is, in the form of powder or mixed with a binder, with no sample preparation.A small amount was put on a microscope glass slide and observed with a Nikon Eclipse LV100ND optical microscope using a Nikon Xenon Power Supply XPS-100 Xenon lamp and a Nikon DS-Ri1 camera.Three observation methods were applied: dark-field mode, fluorescence mode under ultraviolet (UV, 330-390 nm), and blue light (450-490 nm).
The NiSS-Element software was used to acquire the images.

Scanning electron microscopy-energy dispersive x-ray spectroscopy
Two protocols were used for imaging zinc white powders or paint materials.
The size of ZnO particles was estimated by measuring about ten particles for each sample in ImageJ.

ZnO powders
The samples, previously observed using OM, were dispersed in isopropanol with the help of an ultrasonic device.A drop of the dispersion was applied on a piece of Si (100) attached to the sample holder by carbon tape.The samples were then left to dry under a fume hood.
The methodology was developed on reference powder samples (i.e., Brüggemann ZnO, indirect and direct) using a FEG SUPRA 40 ZEISS at the Institut des Nano-Sciences de Paris (INSP; Paris, France); the other powder samples were observed using a FEG Supra55VP ZEISS at the IPANEMA laboratory (Gif-sur-Yvette, France).
The InLens detector (secondary electrons) acquired the images at an operating voltage of 5 kV and a 3-4 mm working distance.

Zinc white samples with binder
Samples containing a binder were embedded in araldite, as explained below for cathodoluminescence.The crosssections were polished up to 1 µm and sputter-coated with 0.7 nm of platinum for analysis with a Jeol JFC-2300HR fine coater.Images and analyses were conducted using a JEOL 7800F with two SDD Bruker AXS 6|30 detectors at the Centre de Recherche et Restauration des Musées de France (C2RMF; Paris, France).The images were acquired using secondary (SE) and backscattered electrons (BSE) at an operating voltage of 5 kV and a working distance of 6 mm.
EDX analyses were performed at an operating voltage of 15 kV and an optimal working distance of 9.5 mm.6-12 points were selected based on the BSE images obtained at × 10 000 magnification.The software Esprit (Bruker, version 2.3) acquired and stored data.

Particle-induced X-ray emission spectroscopy and Ion Beam-Induced Luminescence
The samples were mounted between two polypropylene sheets (FLUXANA) of 6 μm thickness at a 2-3 mm distance to the extracted beam.
The analyses were performed at the Accélérateur Grand Louvre d'Analyses Elémentaires (NewAGLAE, ANR-10-EQPX-22) [54].Four Peltier-cooled SDD detectors were used: one for low energies with a helium flux (1-10 keV), placed at 50° relative to the beam axis, three for high energies, at a relative angle of 45° (1 with a 200 µm Al filter, 2 with 25 µm Co filter) [49].Two areas of 1000 × 1000 µm 2 with a pixel size of 50 × 50 µm 2 were measured for each sample, with a 100,000 total dose (corresponding to a charge of 0.8 µC) using a 3 MeV proton beam.
The used standards were DRN for calibrating the lowenergy detector and two copper-based standards, BS938 and CTIF6, for the high-energy detectors.
Sample luminescence was collected by a 1 mm diameter optical fiber placed at 45° to the beam axis, conducted to a research grade OCEAN OPTICS QE65000 spectrometer.It was measured in the wavelength range 200-1000 nm, with a resolution of 3 nm FWHM (100 µm entrance slit) [56].

X-Ray diffraction
Micro-XRD No specific sample preparation was required; ZnO powders or paint fragments were put as-is on the sample holder.
The experimental setup consisted of a XENOCS GENIX 3D copper source with a high-brilliance X-ray tube (power of 30 W, diameter of 30 µm) and Xenocs FOX 3D optics, which delivers a beam at the length of 1.5418 Å (Cu Kα1-2), at 50 kV, 600 µA, with a pixel size resolution of 200 µm.The diffraction patterns were acquired by a Rigaku R AXIS IV + + imaging plate detector developed by Bede Scientific Instruments Limited.The analyses were carried out at an angle of incidence of about 10°.The acquired 2D data was transformed into a diffractogram by the FIT2D software; crystalline phases were identified by the software QualX 2.0 [57].
High angular resolution-X-Ray diffraction A selection of samples was analyzed at the ID22 beamline at the European Synchrotron Radiation Facility (ESRF) [58][59][60][61].Analyses were performed on samples in borosilicate capillaries (ø = 0.5 mm, WJM-Glas/Müller GmbH); a LaB 6 standard was used as a reference for instrumental resolution.
Measurements were performed at λ = 0.3542 Å (35 keV), thanks to the new Extremely Brilliant Source with a highly monochromatic beam of about 1 × 1 mm 2 and low divergence.Diffracted photons were measured by scanning the 2θ circle over the 0-43° range holding an EIGER2 2 M-W CdTe pixel detector positioned behind thirteen Si (111) analyzer crystals.
A preliminary identification of the present phases was performed using the software QualX mentioned above.A Williamson-Hall analysis was performed by fitting the diffraction peaks through the software WinPLOTR [62] of the FullProf Suite to calculate the average size of the ZnO crystallite in and perpendicular to the (001) plane, the aspect ratio, and the lattice strain of the crystallites, by taking into account the broadening of the diffraction peaks.Details about the procedures are available in the Additional file 1.

Cathodoluminescence
All the samples were embedded in araldite 2020 transparent liquid adhesive epoxy (volume ratio epoxy adhesive/catalyzer of 10:3) to avoid any luminescence from the resin, according to the following protocols: • For ZnO powders, a small amount of sample was mixed with the araldite and applied on a roughened microscope glass slide.The resulting slides were lightly polished with a 1200 grit SiC disc by Buehler (15 µm) to obtain a flat surface; • For zinc white samples containing a binder, small paint flakes were embedded in araldite.The molded samples were sliced to obtain two cross-sections and then polished.
Experimental conditions were fixed at an operating voltage of 11 kV and a current of ~ 390 µA.
The cathodoluminescence system Cambridge Imaging Technology Ltd CL8200 Mk5-1, consisting of a controlled electronics unit and a vacuum chamber associated with an electron gun, was used to study the cathodoluminescence of the samples.Images were acquired with an Olympus BH-2 optical microscope (10 × objective).The exposure conditions were fixed at 3 s with a gain of 5.1x, slightly adjusted for some samples to avoid overexposed images.A Nikon DS-Ri2 camera and the NiSS-Element software were used to acquire and store the images.
An Ocean Optics QE65000 spectrometer acquired luminescence spectra; four accumulated spectra were recorded by the Ocean Optics SpectraSuite software with an exposure of 2 s.The two main luminescence peaks' position, intensity, and width were noted.

Fourier-transform infraRed spectroscopy
No sample preparation was required; a small amount of material was directly placed on the accessory before each measurement.
Samples were analyzed using an FT-IR Frontier Perki-nElmer Spectrometer with an MCT detector.An ATR Goldengate accessory from Specac with a diamond tip or a diamond anvil cell from High-Pressure Diamond Optics, Inc. was used for the measurements, depending on the sample amount available.Spectra were collected in the 4000-400 cm −1 range with a resolution of 4 cm −1 and an accumulation of twenty (ATR Goldengate accessory) or forty (diamond anvil cell) spectra.
Data were acquired and processed using the software Spectrum (PerkinElmer).

Results
This section consists of three parts: results about the composition of the samples are presented first, followed by insights on morphology and size of ZnO particles, and, finally, luminescence properties.An overview of the findings for each sample is provided in Table 3; detailed results per analytical technique are available in Additional file 2: Table S2.

Composition
After a short presentation of the results on the binders, the identified inorganic phases composing the samples are described: the purity and formulations of the ZnO materials are shown first, with major and minor compounds found, followed by the detected trace elements.
Binders 9   FTIR analyses confirmed that all the analyzed paint samples contain oil-based binders except for the Permanent Pigments watercolor tube, made of gum arabic.The Lefranc & Bourgeois pastel had an oil binder.The Lefranc powder contained some wax. 9Zinc carboxylates were found in all the analyzed paint materials.

Inorganic phase(s) of zinc white materials
The inorganic phases of the samples were studied using a combination of results from PIXE and XRD analyses.PIXE measurements were used to quantify the amount of the significant mineral phases composing the samples and trace elements.The results were expressed in weight percent (wt%).XRD allowed the identification of the crystalline phases of the samples.

Purity and formulations of zinc white materials
The amount of ZnO (in wt%) detected for each sample investigated by PIXE is indicated in Table 3.As shown in Fig. 2, most of the analyzed samples were almost pure ZnO (> 98 wt%), in particular the powder samples (including the three reference materials).
The only exceptions were ten powders containing hydrozincite (Zn 5 (CO 3 ) 2 (OH) 6 ), identified by XRD 10 and FTIR.In particular, the compound was found in the 2010 Kremer ZnO powder, while it was not detected on the very same powder back in 2010 with the same setup; the Kremer ZnO powder purchased in 2021 did not contain hydrozincite either.These results suggest that the compound formed over time, probably from converting ZnO upon exposure to air [38,63].Interestingly, it was found only in industrial Vieille Montagne samples, not in the ones referred to by the company as "lab samples".
The Sikkens powder, mostly hydrozincite with some lead-based compound and likely some ZnO, represents a peculiar case.An FTIR spectrum of the sample is available in the Additional file 1: Figure S1.
Interestingly, four samples branded as zinc white 11  proved to be made out of entirely other compounds than ZnO.This is possibly evidence of adulteration, a common issue for artists' materials, particularly zinc white, in the 19th century, as the recurrence of the topic in painting treatises of the time shows [64,65].Those samples, listed at the bottom of Table 3, were not investigated further in this study; only the Sikkens powder mentioned above was investigated further due to the detection of hydrozincite.
Iron and nickel were detected in almost all samples, with values usually lower than 0.1 wt%.Iron is likely ascribed to contamination, while the detection of nickel may be related to the PIXE detector.Lead and sulfur were present in about half of the analyzed samples, almost exclusively historical; calcium, aluminum, and chlorine were found in about a third of the analyzed samples, mostly paint materials, while others were found sporadically.For example, cadmium traces were detected only in the Fezandie & Sperrle zinc white powder and might be related to European zinc ores, as underlined by Holley [66]. 10 HR-XRD used for the quantification, when available. 11A Blockx paint tube, lead-based; Sikkens powders, mostly made of hydrozincite with ZnO and some lead-based minor compound; Permanent Pigments powders, titanium white; a titanium-zinc white jar, a mixture of CaCO 3 , BaSO 4 , TiO 2 , quartz with ZnS and ZnO as minor compounds.
Looking at reference samples, Brüggemann indirect and ZnO nanosmoke powders were the purest of the corpus, with no traces other than iron and nickel (about 200-500 ppm in total).Brüggemann direct ZnO was less pure regarding amount (~ 3200 ppm) and type of trace elements (i.e., sulfur, chlorine, calcium, and iron).The Sennelier powder stood for its extreme purity (less than 2000 ppm of trace elements).Some trace elements were only detected in paint materials (i.e., aluminum, silicon, titanium, strontium, potassium, zirconium, magnesium, and tin; Fig. 3).Aluminum may come from aluminum stearates, a standard paint stabilizer, or aluminum silicates when detected with silicon (i.e., Grumbacher, Vilhelm Pacht, Old Holland, Blockx Ti-Zn white).Strontium likely originated from calcium carbonate or barium sulfate since it is a common impurity of the natural form of barium sulfate [1,67] and was only detected in samples containing these compounds (e.g., Maimeri).Magnesium might come from magnesium carbonate, another standard filler [68].The tin may be a residue from paint tubes made of this material since it was found only in the Vilhelm Pacht sample, dating back to before 1909.Moreover, iron was detected by SEM-EDX only in a grain of the Grumbacher pre-tested, which might support the abovementioned hypothesis that links this element to contamination.
Finally, results obtained by PIXE and SEM-EDX generally agreed, with a few exceptions.The latter technique did not detect certain trace elements (e.g., only zinc was detected in Vilhelm Pacht, Crafint, and Grumbacher 1950s paint tubes).In other cases, such as the Bocour (Ba, S and Ca, Mg, S in some grains) and Lefranc & Bourgeois (Ca, S in some grains) tubes, some elements were only detected by SEM-EDX.The spots analyzed through the two techniques may have been locally richer in some compounds (e.g., driers, additives) due to the complexity and inhomogeneity of their formulations.

Particle morphology and size
This section is dedicated to the characterization of morphology and size of ZnO particles based on SEM observations.A classification upon Williamson-Hall analysis of HR-XRD data for a selection of samples is also presented.Results for reference materials are presented first, followed by results on the rest of the corpus.

Reference samples
The two Brüggemann reference samples produced by indirect and direct processes presented particles of distinct size and morphology (Fig. 4): • ZnO produced by the indirect method had prismoidal particles with hexagonal bases 40-400 nm wide (width W, in red) and 40-500 nm long (length L, in blue), as schematized in Fig. 4c; • ZnO produced by the direct method had mostly acicular particles (i.e., tetrapod-arms grown from a common core), 60-1250 nm wide and 300-5800 nm long, as defined in Fig. 4d.
The third reference, the synthesized ZnO nanosmoke, exhibited tetrapod-like morphology, similar to the direct Fig. 3 Trace elements in zinc white samples belonging to the corpus of the study analyzed by PIXE method ZnO, but at the nanoscale (width < 50 nm, length > 100 nm) [49,69].
The impact of the binder on the characterization of ZnO particle morphology and size was investigated for a couple of powders by comparing them as-is and ground in linseed oil.As shown in Fig. 5 for Brüggemann indirect ZnO powder and the corresponding paint mockup, the size of the particles is comparable.

Zinc white samples of the corpus
Powders and paint materials mainly showed prismoidal morphology (Table 4, Fig. 6a), like the Brüggemann indirect ZnO powder.However, particles with acicular morphology were also observed in some samples, such as the Vieille Montagne powders 12 (Fig. 6b) and Ripolin and Talens paint materials.
All the samples were quite heterogeneous in size, as shown in Fig. 6c (Talens paint tube).
An overview of the observations is provided in Table 3; detailed results are in Additional file 2: Table S2.2.Our findings agree with Johnson's observations [9], who noted their consistency with the optimal particle range of 0.2-0.8μm for paint opacity, durability, and gloss.
Since it was not possible to unequivocally classify the samples based on SEM images, in-depth diffraction studies were performed at the synchrotron radiation facility ESRF in Grenoble (France) to estimate the apparent size, aspect ratio, and strain of the ZnO crystallites.The results of the Williamson-Hall analysis are shown in Fig. 7, where the length along the direction (00 l) has been plotted against the length along (hk0).The values were calculated as described in the Additional file 1.
Figure 7 shows two clusters of samples, with decreasing sizes for the most recent ones (in blue).Vieille Montagne and Lefranc samples are all larger than one micrometer.In this cluster is also the Vilhelm Pacht tube, which dates back to the end of the 19th/beginning of the 20th century.Moreover, samples presenting prismoidal and acicular morphologies had generally larger crystallites (framed in red in Fig. 7).The smaller particles (< 600 nm) cluster included the most recent Lefranc & Bourgeois samples, the Sennelier historical powder, and 20th-century samples like the Maimeri color chart and paint tube.
Results from the Williamson-Hall analysis were generally coherent with SEM observations, but in some cases, the method overestimated 13 or underestimated 14 the size of some particles compared to SEM.In particular, Brüggemann direct ZnO powder and the Talens pristine paint tube presented larger particles in SEM images than the values estimated by Williamson-Hall analysis (Table 3; Additional file 2: Table S2).This shows the limitations of the method, which is unsuitable for assessing the size of crystallites larger than one micrometer and for samples with a broad size distribution.The estimated value is a non-weighted average that may not represent the size distribution.However, the method is still helpful for comparing samples, which is more complex through SEM images.These are high-magnification images of a specific sample area, which might not represent the whole sample.Moreover, it is hard to find a metric based on SEM images to compare samples to each other, especially with such a broad size distribution.
The Williamson-Hall analysis was also a valuable tool to classify paint samples where the ZnO particles were difficult to observe using SEM due to the presence of large agglomerates of particles (e.g., the dried yellowed Talens tube and the Lefranc powder with a binder).
By plotting the projection of the average sizes calculated for each peak over the perpendicular planes ab and bc, it was possible to visualize the crystal geometry of the analyzed samples, as shown in Fig. 8.
This provides a visual representation of the elongated morphology of ZnO particles (aspect ratio > 1) and of the larger crystallite sizes of historical materials, here represented by Vieille Montagne powders and a Lefranc 1930s paint tube (respectively in green, orange, and red in Fig. 8), compared to modern ones (in light and dark blue and black in Fig. 8).Moreover, the highest aspect ratio values were obtained for the synthesized ZnO nanosmoke (2.1) and Vieille Montagne and Lefranc samples (2.2-2.8), which might mirror the presence of acicular particles besides prismoidal ones.
Finally, the calculated strain coefficient was similar for all the analyzed samples (~ 0.01%), meaning that most differences in peaks broadening are related to size rather than strain effects.
A summary of the results of the Williamson-Hall analysis is available in Additional file 2: Table S2

Luminescence
This section is dedicated to the luminescence of the analyzed samples.After a short introduction to the luminescence behavior of ZnO, results of the behavior under UV at optical microscopy will be presented, followed by observations at cathodoluminescence. Finally, findings from IBIL will be briefly presented.
Luminescence is strictly linked to the electronic structure of ZnO.In the case of perfect crystals, the radiative recombination of excitons generated with excitation energy larger than the energy band gap (EBG) results in emission energy equal to the EBG, the so-called near band edge (NBE), or fundamental emission.However, when intrinsic or extrinsic defects are present, they create further energy levels within the band gap, available for excitons to recombine.This results in emission energies lower than the EBG.Therefore, the luminescence behavior directly measures crystal quality (i.e., the presence or absence of defects).The fundamental emission of ZnO is at ~ 3.2 eV (~ 390 nm); Zhang et al. [49,69] attributed green photoluminescence (~ 530-590 nm) to oxygen vacancies, blue/violet photoluminescence (~ 430 nm) to interstitial zinc (due to synthesis in non-equilibrium conditions and rapid quenching, which favors Zn-rich zinc oxide), and yellow photoluminescence (~ 600 nm) to interstitial oxygen-related species.

Photoluminescence under UV at the optical microscope
The three references presented two types of luminescence, blue or green, as illustrated in Fig. 9. Indirect ZnO had grains of weak blue luminescence under UV (Fig. 9a,  b).In contrast, direct ZnO had a bright green luminescence (Fig. 9c, d), and ZnO nanosmoke an overall green (but less bright) luminescence with blue-fluorescing areas (Fig. 9e, f ).
The rest of the corpus presented blue, green, bluegreen, or orange luminescence (Fig. 10).
All the powders and most paint samples presented green luminescence under UV, similar to the Brüggemann direct ZnO powders and the synthesized ZnO nanosmoke.About one-quarter of the paint samples exhibited blue luminescence similar to the Brüggemann indirect ZnO powders.Four paint samples had greenblue luminescence.Some samples presented luminescent grains of different colors and intensities, as shown in Additional file 2: Table S2.4.
Three samples, all from the Netherlands and containing hydrozincite, stood out from the corpus for their orange luminescence (Fig. 10c, d): these are all but one Maastrichtsche powder 15 (i.e., Serena Witzegel, Roedzegel n°1, Grijzegel n°3).The same luminescence was observed on the Sikkens powder, mainly composed of hydrozincite.
Finally, bright blue luminescence was observed when ZnO nanosmoke was ground in linseed oil 16 (Fig. 11).

Cathodoluminescence
The analyzed samples were classified based on the observed luminescence color, supported by the collected spectra (Figs.12,13).Additional file 2: Table S2.4 contains details about peak position and intensity.
Two types of luminescence were observed: • Blue luminescence, as for the ZnO produced by the indirect method and the synthesized ZnO nanosmoke (brighter than the previous one), with an intense peak at ~ 390 nm; the latter also presented a shoulder at ~ 430 nm and a larger band at ~ 530 nm (Fig. 12a, c); • Green luminescence for ZnO produced by the direct method, with a large intense band at ~ 530 nm (Fig. 12b).
The peak at 390 nm will be indicated as NBE (Near Band Edge), and the large band at ~ 530 nm as GL (Green Luminescence).
The results obtained for the entire corpus are described in Table 5, compared to the luminescence observed under UV.In blue modern samples, in black historical ones.White background for samples in form of powder, gray for samples in form of paint materials 15 Maastrichtsche Serena Roedzegel did contain hydrozincite (detected by XRD and FTIR), but exhibited green luminescence. 16The same results were observed when using two types of linseed oil.
All the samples in the form of powders exhibited green luminescence (Fig. 13a) except for Kremer, which was blue-fluorescing.On the other hand, paint tubes were mostly blue-luminescent (Fig. 13b), except for the greenluminescent Lefranc, Ripolin, and Talens samples.Blueluminescent American paint tubes (i.e., Grumbacher, Permanent Pigments, Bocour) and the modern Lefranc & Bourgeois paint tube also had green-fluorescing grains, which would explain the appearance of the GL band in the corresponding spectra (Fig. 13c).
Green luminescence was generally more intense than blue luminescence; thus, exposure conditions had to be slightly adjusted for some samples, such as Vieille Montagne, Sennelier, direct ZnO powders, and Talens and Craftint paint tubes.
The color of the luminescence observed at cathodoluminescence did not always match that observed at OM under UV (Table 5), as in the case of ZnO nanosmoke (Fig. 14a, b), the Talens paint tube (Fig. 14c, d) and the Dutch powders.
Finally, the presence of a binder affected the intensity of the peaks but not their position, as shown on mockups prepared with Brüggemann indirect and direct ZnO and linseed oil (Fig. 15).

Ion beam-induced luminescence
IBIL analyses were also performed on the samples analyzed by PIXE.As for photoluminescence under UV, the general trend of IBIL results did not perfectly agree with CL findings.Additional file 2: Table S2. 4 shows details about peak position and intensity.
Since the perceived color comes from the ratio of microscopic green and blue-fluorescing particles [9], these differences may arise from diverse sources: imperfect exposure conditions; inhomogeneity of some samples, especially paint tubes 17 ; different excitation/ The different excitation/emission paths may be explained by the fact that electrons and protons are massive and carry more momentum than photons; this may lead to different selection rules for cathodo-and ionoluminescence than UV-photoluminescence.The high Fig. 7 Size trend of the samples based on results of the Williamson-Hall analysis.Reference samples are indicated with a star.In blue modern samples, in black historical ones.Framed in red are samples also containing acicular particles.Values larger than 1000 nm are not reliable anymore energy of electrons and protons may generate multiple excitation paths, resulting in relaxation pathways with corresponding cathodo-and ionoluminescence signatures different than those detected by photoluminescence.The higher energy of these particles may also lead to a more efficient excitation process, providing additional intermediate states.

Discussion
The discussion of the results is organized into four parts.Insights on colormen practices will be provided first.Then, our findings will be discussed to make a hypothesis on the synthesis process used to manufacture the samples, followed by the relevant trends found in the corpus with a focus on luminescence behavior.

ZnO manufacturers and colormen practices
The variety of brands forming the corpus allowed us to highlight some of their practices.
Concerning ZnO manufacturers, the comparison Vieille Montagne-Maastrichtsche zinkwit Maatschappij underlines the high purity of powders manufactured by the two companies, especially of the first.Purity is one of the main properties colormen look for in their raw materials.Thus, our findings on Sennelier and Vieille Montagne powders attested to the continuous high-quality provisioning of the first and the purity of the products provided by the latter. 19owever, the complex formulation of some paint tubes like Sennelier's and Maimeri 's (Table 3; Additional file 2: Table S2) confirms the practice of mixing zinc white with other compounds such as barium sulfate [70], lithopone [71], calcium carbonate, zinc sulfide [38] or other pigments like titanium dioxide [72], to modify its optical and mechanical properties or to reduce costs.On the contrary, other paint samples, such as Lefranc and Lefranc & Bourgeois, did not contain any other compound than ZnO with a binder, coherently with the results of Casadio et al. [48] on an artist-grade Lefranc paint tube.However, differences were observed even for the same manufacturer.The dried yellowed Talens paint tube presented larger particles than the pristine one.Did ZnO come from different batches of raw materials even though the tubes were sold in the same box?
Our results also provided information on possible adulteration due to the mismatch between products' labels and their actual content.This is relevant information to

The hypothesis of the indirect process
The findings for pigment powders matched the high purity standards for ZnO produced by the indirect method (i.e., ZnO > 99 wt%) [1].In the same way, the ZnO particles of the analyzed powders were similar to the ZnO indirect reference in terms of morphology and size.These results suggest that all the investigated powder materials of the corpus might have been produced per indirect method, which is also supported by a historical booklet from La Société Vieille Montagne, stating that the company only made ZnO by indirect method to provide their customers with the purest products [73].

Properties trends
While composition, particle morphology and size, and historical sources pointed out the hypothesis of the indirect process, the observed green photo-and cathodoluminescence (i.e., wide, intense GL band at ~ 530 nm) of powders and most paint samples were not coherent with the results obtained for Brüggemann indirect ZnO (i.e., blue photo-and cathodoluminescence). Already Artesani et al. [39] associated samples with both blue-and green photoluminescence to the indirect method, thus to ZnO produced under a Zn vapor-rich environment or in non-equilibrium conditions (i.e., elevated pressure and temperature), for which interstitial zinc is the primary intrinsic defect [39], as found in ZnO nanosmoke [49].Relying only on the NBE peak to identify zinc white may thus be misleading in some cases.The luminescence behavior will be discussed further in the next paragraph.
The second trend identified in the corpus consists of the smaller size of ZnO particles for the most recent samples.Results on Vieille Montagne materials support this hypothesis: while powders produced in the laboratory presented prismoidal submicrometer particles, the ones made at the Valentin-Cocq site had both prismoidal and acicular particles larger than one μm.Less controlled large-scale synthesis conditions may thus have affected their size.This means older samples produced via the indirect method might have larger particles and more defective structures because of poorly controlled synthesis conditions (i.e., growth time, temperature, and atmosphere).This would explain the corpus' significant size distribution, varied morphology, and luminescence behavior.Similar remarks were also reported by, for example, Hageraats et al. [25] and Capogrosso et al. [18] in some Lefranc samples.
As Zhang et al. [49,69] showed, synthesis conditions are crucial in determining ZnO color and properties.The tetrapod-like ZnO nanosmoke used as a reference in this study can acquire hexagonal prism morphology if synthesized in an oxygen-rich atmosphere [69], the morphology most observed in the corpus.They also showed that Chemical Vapor Synthesis (i.e., dynamic conditions under gas flow, controlled temperature, and pressure) leads to smaller, more homogeneous particles (length < 50 nm) [69].This might prove the link of morphology and size of ZnO nanoparticles to oxygen pressure during synthesis or, more generally, to the control of synthesis conditions, A third trend concerns trace elements, possibly helpful in distinguishing historical from modern ZnO, at least for samples in the form of powders.While historical samples (e.g., Vieille Montagne, Dutch powders) consistently contained traces of lead and sulfur, modern ones included other elements like calcium and chlorine.This could be related to the fact that historical zinc white was likely produced using the same raw materials (i.e., zinc ores).In contrast, modern zinc white has a more complex supply chain, employing minerals from other parts of the world or recycled zinc, which could be contaminated.

Luminescence of zinc white
As mentioned above, luminescence was not directly correlated to other features of zinc white.Two hypotheses are presented here to describe the observed trends and their connection to other zinc white properties.
Therefore, a first proposition correlates the GL to larger ZnO particles.Samples with larger particles and GL might once have had a less intense GL band, typical of the indirect method, and turned green-luminescent over time.Johnson [9] demonstrated that blue-fluorescing particles can shift to green-fluorescing particles over time via oxygen removal.Moreover, the role of oxygen on the luminescence behavior of ZnO was already highlighted by Morley-Smith [37], who observed a relation between the GL and the ZnO particle size and related it to the oxygen at the surface of ZnO particles.[69] recently showed that the GL band indicates a more defective material rich in oxygen vacancies.However, they attributed their presence to the oxygen pressure conditions used in the synthesis rather than to the size and morphology of ZnO particles.This leads to a second hypothesis: historical indirect ZnO with larger particles may exhibit more intense GL after manufacturing due to less controlled synthesis conditions.
The ratio of blue-to-green-fluorescing particles has been correlated to the photo-activity of ZnO, being higher for blue-fluorescing particles (e.g., higher chalking 20 rates observed for house paint) [9,37], prevalent type in ZnO produced via the indirect method.To solve the issue, back in the 1950s, Durham Chemicals developed a ZnO indirect grade of larger green-luminescent particles (i.e., lower oil absorption and reactivity, higher reducing power) that could resist chalking [20], which proves the existence of different types of indirect ZnO.Hageraats et al. [74] also observed different types of zinc white in painting materials.This information is particularly relevant in light of the results of Hermans et al. [35], for example, who showed that zinc whites of different reactivity can be observed in paintings and potentially affect paint performance, especially in certain environmental conditions, such as upon use of solvents for conservation treatments, which are a trigger for the crystallization of zinc soaps.Moreover, most documented cases of severe degradation due to zinc white involve20th-century rather than 19th-century artworks [23,27,75]; however, this may not be representative of the use and degradation of zinc white, since the pigment may have been employed more extensively from the end of the 19th century.
Nevertheless, it is essential to consider other factors, such as pigment-binder interaction, aging, and the presence of other compounds.In our corpus, barium sulfate-either alone (e.g., Bocour paint tube) or with calcium carbonate and zinc sulfide (e.g., Maimeri paint tube)-did not modify the cathodoluminescence response of ZnO samples.Hydrozincite did not seem to affect color and spectral features observed at cathodoluminescence either (e.g., in Kremer powders, which both presented bluish CL with no other spectral characteristics).Still, it might have played a role in the orange luminescence observed at the OM under UV light for some Dutch powders since all the samples exhibiting an orange luminescence under UV contained this compound (Table 5).Hydrozincite might affect the photoluminescence of zinc white only when present above a specific concentration since other ZnO samples containing some hydrozincite presented green luminescence (e.g., Maastrichsche Serena Roedzegel powder).However, the contribution of other compounds to the photoluminescence response of these samples cannot be excluded either.Further studies on these materials and hydrozincite will be performed to confirm these preliminary hypotheses on the orange photoluminescence.Finally, optical cathodoluminescence, a technique widely used geology [76], more rarely to study paint materials [77][78][79], proved an appropriate, easy-to-handle tool for identifying zinc white vs. other whites and zincbased pigments.Our results add up to the research of Palamara et al. [77], the only systematic study of white pigments by cathodoluminescence presented until now, to the authors' knowledge.They showed applications of SEM-CL aimed at building a database of the behavior of pigments at cathodoluminescence.In a complementary way, our study demonstrates a more accessible method to assess chemical differences through visual information (i.e., a large field image at the OM) and spectral features (spectroscopy).Once further reference databases are developed, the method could be of great value, especially

Conclusions
The study of a unique, varied corpus of historical and modern zinc white artists' materials provided information on the variability of the properties of the pigment and colormen practices for zinc white formulations.The corpus presented some unavoidable biases and intrinsic limitations but still represents a good reference of zinc white properties for Heritage professionals.
Common additives, proportions, and combinations were identified for various color manufacturers and compared to their labeling, providing information about zinc white materials, their properties, and possible manufacturers' and artists' choices.
Hydrozincite was detected in some ZnO powders, even as the main component in the Sikkens powder; it was not found in any paint sample.The compound is likely a degradation product formed over time in powders exposed to air.It was identified by XRD and FTIR, which have different sensibilities to the compound but provide coherent results.In our corpus, the presence of hydrozincite had no impact on the cathodoluminescence response of ZnO samples, but it might be a cause of the orange luminescence observed under UV.Further studies are required to confirm if hydrozincite or other compounds are responsible for this behavior.Since crystallographic research on hydrozincite has not been published since the 1960s, the compound will be the object of another study.
The results for indirect and direct ZnO references agreed with the literature.The samples' high purity (i.e., a reduced amount and type of trace elements), prismoidal morphology, and submicron particle size match the characteristics of the ZnO synthesized via the indirect method.Acicular particles were sometimes observed in historical samples in addition to prismoidal ones, with a large particle size distribution for the whole corpus.
The luminescence behavior was indeed not straightforward to interpret.Both blue-and green-fluorescent materials were observed.The Near-Band-Edge peak at ~ 390 nm was not always detectable, even in pure historical powders without binders.This raises questions on the practical use of this feature for identifying zinc white.We formulated two hypotheses to explain that either blue-luminescent particles might have turned greenluminescent over time or some (less controlled) synthesis conditions promoted the formation of larger green-luminescent particles.
Modern materials had smaller ZnO particles than historical ones, containing some recurring trace elements, such as lead and sulfur.
Photo-, cathodo-and ionoluminescence results did not perfectly match.Cathodoluminescence proved a valuable complementary tool for studying pigments and paint materials.To the authors' knowledge, this technique, combined with optical microscopy, was used for the first time on such a large corpus of paint materials.Photoluminescence seems more sensitive than cathodoluminescence to the presence of compounds other than ZnO (e.g., probably hydrozincite), so comparing the two techniques can be informative of compositional variaof the sample.This study represents a solid database in which synthesis routes, intrinsic properties, and luminescence behavior of zinc white were discussed as a guideline for identifying and classifying zinc white.The presented findings will be used as a reference to study samples and cross-sections from 19th-20th century paintings to survey the extent and modalities of the use of zinc white, its properties, and its relation to painting degradation.
Archival and historical research on zinc white, its production, and use among 19th and 20th centuries colormakers, merchants, and artists will be deepened in an art historical review [80].

Fig. 2
Fig. 2 Composition of the zinc white samples belonging to the corpus of the study.Reference samples are indicated with a star; in blue are modern samples, and in black are historical ones.White background for samples in the form of powder, gray for samples in the form of paint

Fig. 4
Fig. 4 SEM images (secondary electrons, 5 kV, working distance ~ 6 mm) of ZnO powders by Brüggemann (at the top): a indirect method; b direct method; schema of the main ZnO morphologies (at the bottom): c prismoidal shape of hexagonal basis; d acicular shape.The length (L) and width (W) measured on SEM images are also indicated

Fig. 5
Fig. 5 SEM images (secondary electrons, 5 kV, working distance ~ 6 mm) of ZnO powder by Brüggemann: a as-is; b ground in linseed oil

Fig. 8 Fig. 9
Fig. 8 Representation of the crystal geometry-(a*, c*) cross section of ZnO crystallites based on average crystal sizes computed by Williamson-Hall analysis of Brüggemann indirect ZnO powder (blue), ZnO nanosmoke powder (light blue), white seal Vieille Montagne ZnO powder (green), silver seal Vieille Montagne ZnO powder (orange), Lefranc 1930s paint tube (red), Lefranc & Bourgeois paint tube (black) Fig. 9 Images at the OM (20x, under visible light on the left, under UV on the right) of the reference samples: a-b Brüggemann indirect ZnO; c-d Brüggemann direct ZnO; e-f ZnO nanosmoke

Fig. 10 Fig. 11
Fig. 10 Images at the OM (20x, under visible light at left and UV at right) of a-b Vieille Montagne white seal zinc white powder; c-d Maastrichtsche zinkwit mij.Ejsden zinc white powder, "Roodzegel n°1"; e-f Lefranc & Bourgeois zinc white contemporary paint tube; g-h Lefranc 1930s paint tube

Fig. 13
Fig. 13 Cathodoluminescence images (10x, ~ 390 µA, 11 kV; at the top) of a Vieille Montagne white seal zinc white powders; b Lefranc & Bourgeois zinc white paint tube; c Lefranc & Bourgeois modern zinc white paint tube.The corresponding spectra are shown at the bottom of the Figure Color of the luminescence under UV and at cathololuminescence. Reference samples are indicated with a star.In blue modern samples, in black historical samples.Samples containing hydrozincite are underlined.White background for samples in form of powder, gray for samples in form of paint a Zinc white and titan-zinc whiteb The Sikkens powder made of hydrozincite also exhibited orange luminescence under UV light and blue luminescence at CL

Fig. 14
Fig. 14 Images at the OM (10x) under UV (on the left) and at cathodoluminescence (on the right) of a, b ZnO nanosmoke powders; c, d Talens paint tube

Fig. 15
Fig. 15 Cathodoluminescence images (10x) and spectra of Brüggemann of a indirect ZnO powder; b direct ZnO powder; c indirect ZnO powder ground in linseed oil; d direct ZnO powder ground in linseed oil

Fig. 16
Fig. 16 Size trend of samples analyzed at the ESRF (Williamson-Hall analysis) classified by growing size along the (l00) direction (in orange).In comparison, the direction (hk0) is also shown (in blue).With a star are indicated reference samples, in blue are modern samples, and in black are historical ones.Green and blue dots indicate the color observed using cathodoluminescence

Table 1
Summary of the samples in the corpus of analysisBrands are listed in alphabetical order.The date of production of the samples is indicated between brackets in the column on the right when available

Table 2
Summary of the techniques used to characterize the corpus of samples a Detailed in Additional file 2: TableS2.0 b High-angular Resolution X-Ray Diffraction at the ESRF

Table 3
Synthesis of the analytical results obtained on the studied samples

of different composiƟons than zinc white
White background for samples in the form of powder, gray for samples in the form of paint.N/A indicates that the analysis was not performed a From PIXE and XRD.When no PIXE was available an "Y" indicated the presence of ZnO Classification from SEM data (small if both diameter and lengths < 500 nm; medium if at least one dimension > 500 nm; large if at least one dimension > 1000 nm); numerical values from Williamson-Hall analysis (HR-XRD data): in blue the estimated length along (hk0), in orange the estimated length along (00 l); the indicated maximum is 1000 nm, apart from larger particles (i.e., KF Chemicals; VM cachet bleu, blanc et or; Lefranc paint tube) 73.9 wt% Zn calculated by PIXE.XRD and FTIR revealed that the sample is mostly made of hydrozincite b 1-10 wt%, listed in decreasing order; from PIXE and XRD c < 1 wt%, listed in decreasing order; from PIXE d From SEM e f Vieille Montagneg

Table 5
Comparison of photoluminescence under UV and cathodoluminescence of zinc white samples