10 investigation of the grounds of tasset et l’hôte commercially primed canvas used by van gogh in the period 1888 to 1890 johanna salvant, Muriel Geldof, elisabeth ravaud, Luc Megens, charlotte Walbert, Michel Menu and don h. johnson We know from Van Gogh’s correspondence that Tasset et L’Hôte was one of the main suppliers of commercially primed canvas during his last years: from his stay in Arles until his death in 1890. At that time, pre-primed canvas was available to purchase either as individual ready-stretched primed canvas on frames that were usually of standard size, or as a roll of primed canvas sold by the metre, which provided cheaper supports. Tasset et L’Hôte was a relatively small shop established in July 1885 at 31 rue Fontaine in the 9th arrondissement of Paris.1 There is evidence that Van Gogh had bought painting materials from this supplier during his stay in the city, including stretchers and ready-stretched canvas.2 When he arrived in Arles in February 1888, he initially purchased his painting materials locally.3 Unsatisied,4 he tried out diferent qualities of primed canvas from Tasset et L’Hôte between April and August 1888, which seems to have resulted in the decision to use the ‘ordinary’ quality (toile ordinaire), priced at 2.5 francs per metre, which he considered to be well prepared.5 The irst reference to receiving this grade of prepared canvas is on 1 July 1888 (letter 635). References in Van Gogh’s letters between July 1888 and July 1890 reveal at least sixteen further orders for Tasset et L’Hôte primed canvas sold by the roll. For orders between August and November 1888, cross references between the letters indicate that the canvas ordered and received by Van Gogh is the ‘ordinary’ canvas at 2.5 francs.6 For the following orders, he did not specify any particular quality of canvas, but it was most likely the same ‘ordinary’ canvas that he mainly, but not exclusively, used until his death in 1890. Van Gogh usually placed orders for an entire roll (10 metres) or half a roll (5 metres).7 Rolls of primed canvas from Tasset et L’Hôte were prepared as follows. 8 A piece ten metres long was cut from a bolt of raw canvas 100-200 metres long and about 2.10 metres wide.9 The approximately 2.10 × 10 metre canvas was then stretched on a large frame to be individually primed. When the ground was dry, it was removed from the frame and rolled to be sold. Van Gogh would purchase 182 � Van GoGh's studio practice – johanna saLVant et al. We would like to thank conservators ella hendriks and Kathrin pilz from the Van Gogh Museum, amsterdam, for their close co-operation while examing the paintings and samples. We are also grateful to anne distel (former General curator at the Musée d’orsay, paris), arnauld Brejon de Lavergnée (former director of the palais des Beaux-arts, Lille) and torsten Gunnarsson (head of collections at the nationalmuseum, stockholm) for making the study of the paintings and ground samples possible. thanks are due to conservators Luuk struick van der Loeff from the Kröller-Müller Museum, otterlo, natasha Walker from tate Britain, London, Friederike steckling from the Fondation Beyeler, riehan, and sophie eichner from the Kunstmuseum, Basel, for providing us with ground samples from the paintings in their collections. We are also grateful to jean-paul rioux, now retired from the centre de recherche et de restauration des Musées de France, paris (c2rMF), for his earlier work on a group of paintings from French collections, which provided us with the ground samples studied at the c2rMF. Furthermore we are indebted to aviva Burnstock from the courtauld institute of art, London, and inge Fiedler from the art institute of chicago for so generously sharing with us their analytical results on Van Gogh ground layers. We would also like to thank our (former) colleagues Gisela van der doelen, Katrien Keune and Karin Groen from the cultural heritage agency of the netherlands, amsterdam (rce), for performing part of the analysis on the ground layers included in this essay. We are grateful to Kees Mensch from the shell technology centre in amsterdam, for performing with us the seM-edX analysis on the samples from the rce. We acknowledge the european synchrotron radiation Facility, Grenoble (esrF) for provision of synchrotron facilities and we would like to thank juan angel sans tresserras for his assistance in using beamline id18F. synchrotron radiation analyses performed on beamline id18F were funded by grants from esrF (experiment ec-598). We also thank Marine cotte for providing us with the opportunity to perform sr-Ftir analysis on beamline id21 at esrF and for her help and advice in carrying out sr-Ftir and sr-μXrd experiments. We thank Marie radepont for participating in the synchrotron based experiments. We are also very grateful to Lucile Beck and Laurent pichon for their help in achieving piXe analysis and data treatment. Finally, we are indebted to teio Meedendorp and ella hendriks, for their substantial contribution to this research. Fig. 1 Schematic representation of the diferent stages involved in priming the canvas from Tasset et L’Hôte ordered by Van Gogh: from bolt of raw canvas to inished painting. 1 | hendriks and Van tilborgh 2011, p. 95, note 16: according to the information provided by stéphanie constantin from the ‘états cadastraux’ of 1876, the shop measured 20 m2; reff 1999, p. 76. an entire roll of primed canvas or a batch of a few metres and then cut a section from it to a desired format before mounting it on a stretcher for painting (ig. 1). Previous studies based on manual thread counting10 have shown that the canvas supports of many of Van Gogh’s paintings made between 1888 and 1890 share similar characteristic asymmetric thread counts. Two diferent thread 2 | chapter 9 in this volume. see also hendriks and Van tilborgh 2011, p. 93. 3 | Letters 583 and 593. 4 | in letter 593, Van Gogh indicated that, as his local supplier of canvas was reluctant to produce absorbent canvas, he had decided to order canvas from paris or Marseilles. Based on the Van Gogh letters, these counts very likely correspond to the ‘ordinary’ 5 | Letter 666: ‘tasset ordinary canvas, which at 50 centimes was dearer than Bourgeois’s, is very much to my liking and is very well prepared.’ canvas that he regularly ordered from Tasset et L’Hôte from July 1888 onwards. 6 | Letters 658, 680, 687, 700, 710 and 719. countings were identiied in these studies: about 12 × 15 and 12 × 17 threads per cm2. Automated thread counting developed during the past few years and applied to works by Van Gogh has enabled us to group paintings from the same bolt of canvas and, in some cases, from the same roll, when matching weft-thread density patterns are found.11 As diferent rolls were not necessarily primed at the same time, the subsequent variations in the composition of the priming medium make it possible to discriminate between rolls from the same bolt. Study of the ground is therefore an essential step in narrowing down the thread count identiication of canvases from a common bolt to identiication of the rolls. This study aims to characterise the grounds of Van Gogh’s paintings on Tasset et L’Hôte primed ‘ordinary’ canvas in an attempt to identify which paintings have supports from the same roll or batch Van Gogh ordered from 7 | Letters 593, 683, 777, 800, 808, 823 and 863. there is also one example where Van Gogh might have ordered 7 metres of primed canvas (letter 863). 8 | information about commercial canvas priming business practices kindly provided by philippe huyvaert, president of claessens artists’ canvas, Waregem, Belgium, experts in the traditional preparation of artists’ canvas. 9 | roll reconstruction from Van Gogh’s late ‘double-square’ paintings suggests that the tasset et L’hôte roll was about 214 cm wide (see chapter 9 in this volume). the rolls from some suppliers, such as claessens, were 2.10 metres wide, whereas those from others, for example from Lefranc & cie in 1890, were 2 metres wide. see haaf 1987, pp. 17, 56-58. From Van Gogh’s letters, we learn that ‘ordinary’ canvas from Bourgeois is 2 metres wide (letters 631 and 639). seven paintings with the characteristic asymmetric thread count indicative of 10 | ravaud 1999, pp. 59-60; hoermann Lister et al. 2001, pp. 356, 364-65; hendriks and Van tilborgh 2001, p. 151. this quality of canvas from Tasset et L’Hôte have been characterised. These 11 | see chapter 8 in this volume. paintings have been dated from between September 1888 and July 1890 and are 12 | average thread count analysis resulting from automatic thread counting performed on the X-ray by johnson et al. (see chapter 8). Tasset et L’Hôte between September 1888 and July 1890. Grounds from forty- listed in Table 1, together with the average thread counts of the canvas.12 InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 183 184 Table 1 � Van GoGh's studio practice – johanna saLVant et al. Summary of information regarding canvas and ground characteristics obtained from automated thread counting and ground analysis for the forty-seven paintings made on Tasset et L’Hôte primed canvas. canVas Title f no. date Sunlowers F458 Augustine Roulin (La berceuse) Grounds addition of mixture baso4/caco3 in ground formulationg average thread counta (th/cm) (warp × weft) match groupa analysis carried out on ground samplesb Type of grounde lead white phases detected by μxrdf second half of january 1889 11.4 × 16.9 1 seM-edXc, μXrd a h and c F506 january 1889 11.5 × 16.9 1 seM-edXc a Self-portrait with bandaged ear F527 january 1889 11.5 × 17.3 1 seM-edX , μXrd, Ftir a Augustine Roulin (La berceuse) F504 March 1889 11.5 × 16.9 1 seM-edX a La Crau with peach trees in blossom F514 april 1889 11.4 × 17.2 1 seM-edXd, μXrd a h and c The bedroom F483 mid-late september 1889 11.4 × 17.0 1 seM-edX*, μXrd, piXe F’ h and c The woodcutter (after Millet) F670 between 20 september-october 1889 11.5 × 17.1 1 seM-edX, μXrd, Ftir F h and c A pair of leather clogs F607 september-october 1889 11.2 × 17.0 1 seM-edX, μXrd F h and c The garden of Saint Paul’s Hospital (The fall of the leaves) F651 october 1889 11.4 × 16.7 1 seM-edX, μXrd F h and c The garden of Saint Paul’s Hospital F659 november 1889 11.4 × 16.9 1 seM-edX, μXrd F h and c Portrait of Doctor Gachet F754 june 1890 11.5 × 16.6 1 seM-edX*, μXrd, piXe, Ftir F’ h and c yes Marguerite Gachet in the garden F756 1 june 1890 11.5 × 17.1 1 seM-edX*, μXrd, piXe, Ftir F’ h and c yes Roses and anemones in a vase F764 early june 1890 11.4 × 16.5 1 seM-edX*, μXrd F’ h and c Thatched cottages F792 irst half of june 1890 11.4 × 16.9 1 seM-edX*, μXrd, piXe F’ h and c yes Marguerite Gachet at the piano F772 june 1890 11.6 × 17.5 1 seM-edX F Two children F783 june 1890 11.4 × 16.1 1 seM-edX*, μXrd, piXe, Ftir F’ h and c yes Landscape at twilight F770 late june 1890 11.3 × 17.1 1 seM-edX F Wheatield under thunderclouds F778 irst half of july 1890 11.4 × 16.5 1 seM-edX G Farms near Auvers F793 july 1890 11.4 × 16.4 1 seM-edX G Field with haystacks F809 july 1890 11.4 × 17.0 1 seM-edX G Tree roots F816 late july 1890 11.4 × 16.7 1 seM-edX G Eugène Boch (The poet) F462 early september 1888 11.1 × 17.3 20 seM-edX*, μXrd c d h and c h and c yes InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas Sower F451 late november 1888 11.3 × 17.9 20 seM-edX* c Undergrowth F746 irst half of july 1889 11.2 × 17.9 20 seM-edX F Entrance to a quarry F744 mid-july 1889 11.3 × 17.8 20 seM-edX B The bedroom F484 early september 1889 11.5 × 18.1 20 seM-edX B Wheatield with a reaper F618 early september 1889 11.2 × 17.7 20 seM-edX B Self-portrait F627 early september 1889 11.2 × 17.7 20 seM-edX*, μXrd, piXe, Ftir B Pietà (after delacroix) F630 mid-september 1889 11.3 × 17.6 20 seM-edX B The sheep-shearer (after Millet) F634 c. 20 september 1889 11.3 × 18.0 20 seM-edX*, μXrd F” h and c Peasant woman bruising lax (after Millet) F697 mid-september 1889 11.5 × 18.1 20 seM-edX*, μXrd F” h and c Emperor moth F610 late May-early june 1889 12.1 × 15.4 2 seM-edX e The reaper (after Millet) F687 mid-september 1889 11.9 × 15.4 2 seM-edX e Peasant woman binding sheaves (after Millet) F700 mid-september 1889 11.8 × 15.2 2 seM-edX e Olive grove with two olive pickers F587 december 1889 12.1 × 15.9 2 seM-edX* d Noon: rest (after Millet) F686 irst half of january 1890 12.1 × 16.4 2 seM-edX*, μXrd, piXe, Ftir d h and c Snow-covered ield with a harrow (after Millet) F632 second half of january 1890 11.9 × 15.1 2? seM-edX*, μXrd d h Doctor Gachet’s garden F755 late May 1890 12.0 × 14.4 2 seM-edX*, μXrd, piXe d h and c yes Acacia in lower F821 early june 1890 12.0 × 15.6 2 seM-edX* e The bedroom F482 mid-october 1888 11.4 × 16.1 17 seM-edX* d Orchard in blossom F511 april 1889 11.6 × 15.3 17 seM-edX* e The thresher (after Millet) F692 mid-september 1889 11.8 × 15.1 17 seM-edX e The Good Samaritan (after delacroix) F633 early May 1890 11.6 × 16.4 17 seM-edX e? Garden with butterlies F402 spring 1890 11.6 × 15.4 17 seM-edX* d Cows (after jordaens) F822 july 1890 11.6 × 16.4 17 seM-edX*, μXrd, piXe, Ftir G h no The dance hall in Arles F547 mid-december 1888 11.9 × 14.6 none seM-edX* e The sheaf-binder (after Millet) F693 mid-september 1889 11.4 × 15.4 none seM-edX e c key a average thread count and match group determined by automatic thread count analysis of X-radiographs of the paintings (see chapter 8 in this volume) b seM-edX: observation and analysis of the samples by seM-edX; seM-edX*: additional semi-quantitative analysis by seM-edX have been performed on the samples c seM-edX analysis performed at the art institute of chicago (samples from paintings F458, F484 and F506) 185 d seM-edX analysis performed at the courtauld institute of art, London (samples from paintings F514 and F527) e type of ground referred by letter (see table 2) f h: hydrocerussite; c: cerussite g information obtained from piXe analysis h and c no yes Table 2 Diferent types of ground identiied in the Tasset et L’Hôte primed canvas ordered by Van Gogh between 1888 and 1890. single layer ground types double ground types Type of ground number of paintings lead white a 5 main B 5 main secondary c 2 main minor minor d 6 main secondary secondary secondary e 8 or 9 minor secondary secondary main F 15 bottom layer main main (variable) secondary secondary Top layer main secondary secondary secondary bottom layer main main secondary secondary Top layer main secondary secondary main G 5 calcium carbonate barium sulphate lithopone secondary First, a comparison of the grounds of those paintings, resulting from samples analysis, will be presented. Secondly, information gathered from the study of the grounds will be considered in the context of match groups identiied using automatic thread counting, to help discriminate between the diferent batches used by Van Gogh. characTerisTics of The TasseT eT l’hôTe Grounds Ground samples from the group of paintings under investigation were studied using optical microscopy and analysed by scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM-EDX)13 to investigate the number of ground layers and their composition. For many paintings, more than one sample was analysed to ensure that the measurements represented the entire ground. Additional analyses were performed on some samples: semi-quantitative analysis by SEM-EDX14 (indicated with an asterisk in Table 1), particle induced X-ray emission (PIXE), synchrotron radiation micro X-ray difraction (SR-µXRD) and synchrotron radiation fourier transform infrared spectroscopy (SR-FTIR). This made it possible to evaluate the proportion of major and minor chemical elements detected in the grounds, as well as trace elements, to conirm the identiication of the compounds in the grounds and to investigate the type of binder used (Table 1). 15 The use of synchrotron radiation, compared with the one of laboratory instruments, ofered the appropriate beam size of a few micrometres necessary to perform selective analysis of each ground layer. Among the forty-seven grounds studied, seven types of ground, referred to by the letters A-G, have been identiied based on the detected mineral compounds 186 � Van GoGh's studio practice – johanna saLVant et al. 13 | at the c2rMF, seM-edX analyses were performed using a philips XL30cp scanning electron microscope and a Link isis 300 edssystem. the primary electron beam energy was 20 keV. samples have been carbon coated for analysis and examined in the high vacuum. at the rce, seM-edX analyses were carried out in collaboration with Kees Mensch at the shell technology centre amsterdam, using a jeoL jsM 6490 LV seM and a noran system six edssystem. the primary electron beam energy used was 25 keV. the cross-sections were generally coated with carbon and examined in high vacuum mode. at the courtauld institute of art, London, seM-edX analyses were carried out by aviva Burnstock using a jeoL s100 seM with an oxford instruments light element eds detector. at the art institute of chicago, seM-edX analyses of sample from F484 were performed in the epic facility of nuance center at northwestern university by inge Fiedler using a hitachi s3400-n-ii seM coupled with an oxford instruments eds detector, while seMedX analyses of sample from F506 were carried out by tim B. Vander Wood at MVa scientiic consultants (duluth, Georgia). the primary electron beam energy used was 20 keV and the cross-sections were carbon coated. 14 | regarding the accuracy of the semi-quantitative analysis by edX, haswell et al. 2011 investigated the potential of edX analysis to estimate quantitatively the composition of Van Gogh’s grounds. Based on oil paint reconstructions made using traditional lead white with varying amount of extender (barium sulphate or chalk), the results of this study have shown that edX analysis provides a good estimation of the barium sulphate amount with a relative error of 10% compared to the known composition. however, this study suggests that the chalk fraction in traditional lead white/chalk reconstructions can be overestimated by as much as 30% (relative error) using edX analysis. see haswell et al. 2011. Fig. 2 Backscattered electron image (BEI) of a ground sample from Two children (F783) showing an aggregate of lithopone (circled in red). Spectra corresponding to EDX analysis in the particle show the detection of sulphur, zinc and barium. 1200 1000 S Counts 800 600 Ba Zn 400 Ba 200 Zn Ba 0 0 2 4 6 8 10 12 Energy (keV) and on the number of ground layers (one or two) (Table 2). In paintings where the ground binder was investigated, the drying oil has always been identiied.16 The inorganic compounds detected in the grounds are lead white (2PbCO3. Pb(OH)2 /PbCO3 ), calcium carbonate (CaCO3 ) and/or barium sulphate (BaSO4 ). In many grounds, large round aggregates of about 20-50 μm were also detected. 15 | piXe analyses were performed with the external proton beam of the aGLae (accélérateur Grand Louvre pour l’analyse elémentaire) at the c2rMF. a 3 MeV-proton beam has been used for the experiments. proile of piXe analysis across the different layers of the cross-section (see ig. 6) were made with a step from between 5 to 20 μm (adapted to layer thicknesses), providing several data analysis in each layer of the samples. sr-μXrd and sr-Ftir experiments were carried out respectively at the beamlines id18F and id21 at the esrF. Measurements were taken directly on crosssections, without further preparation of the samples. sr-μXrd analyses were performed in transmission at 28 keV. an additional luorescence detector was placed to simultaneously collect information about the atomic composition. For each sample, a 2-d map of about 100 × 60 μm2 (V × h) was carried out with a pixel size of 2 × 5 μm2 and a dwell time of about 10 seconds/pixel. sr-Ftir analyses were performed with a nicolet continuum microscope coupled to a nexus spectrometer in attenuated total relectance (atr) mode. Beam size was 28 × 28 μm2, with an actual resolution of 7 × 7 μm2. a step of 7 μm was used in both directions to acquire a 2-d map. an average of sixty scans has been recorded for each spectrum in the range of 4000-700 cm-1 with a resolution of 8 cm-1. For further details regarding experimental set-ups and extraction of data for piXe, sr-μXrd and sr-Ftir analysis, see salvant 2012, pp. 97-98, 103-04, 134-35, 193-94 and 233-35. 16 | Ground samples from eight paintings (see table 1) were analysed by Ftir: c-h stretching bands at about 2850 and 2950 cm-1 and c=o ester stretching band at about 1730 cm-1 indicate an oily binder. see salvant 2012, pp. 194-95. Moreover, drying oil was identiied in grounds from F755, F756, F627, F783 and F821 based on Gc-Ms (gas chromatographymass spectrometry) analysis previously performed at the c2rMF. thanks to christine Benoit and jeanpaul rioux for this information. drying oil was also detected in the ground of F451 by Gc-Ms analysis. see hoermann Lister et al. 2001, p. 359, table 1. observed between the sum of the detected barium and zinc amounts and the 17 | Lithopone was sold under different grades in function of the proportion of zinc sulphide. according to Lahousse (1909), the lower grade contains 15 wt% Zns and higher grades up to 50 wt% Zns. in literature from 1959, it is indicated that lithopone was approximately made in equimolar proportion of Zns and Baso4, corresponding to about 30 wt% Zns, and that there were grades of lithopone containing 15 up to 60 wt% Zns. see Lahousse 1909, pp. 4-5; o’Brien 1915, p. 113; Peintures Pigments Vernis 1959; perego, pp. 462-63. detected sulphur amount evaluated by SEM-EDX semi-quantitative analysis 18 | Feller 1986, vol. 1, pp. 58-59. These were found to contain barium, zinc and sulphur, strongly suggesting that they are lithopone, a white pigment resulting from co-precipitating barium sulphate and zinc sulphide (ZnS) in variable proportions (ig. 2).17 It is diicult to distinguish between barium sulphate and lithopone using XRD spectra,18 however, the presence of a signiicant amount of zinc, together with the correlation InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 187 amount of sulphur detected (wt%) 20 Fig. 3 The amount of sulphur as a function of the sum of the amounts of barium and zinc (wt%) in grounds where both barium and zinc have been detected. Data results from semi-quantitative SEM-EDX analysis of twenty-seven ground samples from seventeen paintings with ground types D-G. Each point in the graph corresponds to the average values in each ground layer of each sample. A correlation (correlation co-eicient of 0.98) is observed between the amount of sulphur and the total amounts of barium and zinc, indicating the presence of lithopone in those grounds. 15 10 5 0 0 25 50 75 sum of amounts of barium and zinc detected (wt%) (ig. 3) conirmed the presence of lithopone in many grounds. In the grounds analysed, the ratio of the weight percentage (wt%) of barium detected to that of zinc was generally found to be around 2, suggesting the use of lithopone with about 30 wt% zinc sulphide.19 Moreover, in most of the grounds, yellow particles identiied as ochre were detected, contributing of-white tints of difering degrees to the grounds.20 Twenty-seven paintings have a single ground layer, divided into ive types (A-E), while the remaining twenty paintings have double grounds, falling into two ground types (F and G) (see Tables 1 and 2). Examples of backscattered electron images (BEI) of the diferent ground types are presented in ig. 4. Single layer ground types Grounds of types A-E are single layered (Table 2), with thicknesses ranging between 30 and 140 μm. Apart from ground type E, all are lead-white based and difer according to the nature of the extenders detected. Grounds of type A consist of a layer of lead white with some lithopone. This ground type has been identiied in ive paintings, all painted between January and April 1889 (Table 1). Grounds of type B consist of a layer of lead white extended with a high amount of calcium carbonate.21 This ground type has been found in ive paintings produced over a short period of time, between July and September 1889 (Table 1). Grounds of type C consist of a single layer of lead white extended with a little calcium carbonate and barium sulphate. Two paintings dated between September and November 1888 have been identiied with this ground type.22 188 � Van GoGh's studio practice – johanna saLVant et al. 19 | this corresponds to lithopone sold as ‘red seal’ quality (30 wt% Zns and 70 wt% Baso4), corresponding to lithopone with equimolar proportions of Baso4 and Zns (formula Baso4. Zns). see note 17. however, it is important to note that calculations to provide raw estimation of proportions of Zns and of Baso4 in lithopone do not take into account that the barium detected does indeed originate from two different compounds (barium sulphate and lithopone). 20 | other coloured particles (brown, red, blue and/ or black particles), probably added to tone the grounds, have also been detected in several ground samples. some of these were identiied as ochre, ultramarine, carbon black or bone black pigments. 21 | semi-quantitative seM-edX analysis performed on one ground sample from F627 (ground of type B) indicates there are about 87 wt% lead (pb) and about 12 wt% calcium (ca) (average of twenty analyses). 22 | semi-quantitative seM-edX analysis performed on two ground samples from F462 and one from F451 (ground of type c) indicates there are about 97 wt% pb, 0.5 wt% ca and 1 wt% barium (Ba) (based on the averages of two to ive analyses per sample). Type A Type B Type E Type F Type C Type D Type G Grounds of type D consist of a layer of lead white extended with a relatively high amount of calcium carbonate, some barium sulphate and some lithopone.23 Six paintings have this ground type: one painted in October 1888, three from December 1889 to January 1890, and two in spring 1890. Grounds of type E difer from the others in that they are lithopone based. This type also contains some calcium carbonate and barium sulphate with a little lead white.24 Eight, possibly nine,25 paintings have been identiied with this Fig. 4 Backscattered electron images of the diferent types of grounds observed in the forty-seven paintings studied. Single grounds: Self-portrait with bandaged ear (F527) (type A); Self-portrait (F627) (type B); Eugène Boch (The poet) (F462) (type C); Doctor Gachet’s garden (F755) (type D); Emperor moth (F610) (type E). Double grounds: Portrait of Doctor Gachet (F754) (type F); Wheatfield under thunderclouds (F778) (type G). ground type. Three were painted in December 1888 and from April to early June 1889, four in September 1889, and two in May to June 1890. Double ground types Double grounds identiied among the forty-seven paintings studied fall into two types: F and G. The bottom layer is usually thicker (thickness ranging between 30 and 250 μm) than the top layer (between 15 and 60 μm). In grounds of type F, the bottom layer consists of high amounts of lead white and calcium carbonate mixed with some barium sulphate and lithopone. The upper layer has the same constituents as the bottom layer, but with much more lead white and much less calcium carbonate. Two diferent subtypes can be identiied in the large group of paintings with ground of type F, referred to as F’ and F”. The main diference is the amount of lead and calcium detected in the bottom layer: based on the semi-qualitative SEM-EDX analysis of type F grounds for eight paintings, six grounds show about 60 wt% of lead with about 20 wt% of calcium in their bottom layer (type F’ in Table 1), while the two others have InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 23 | semi-quantitative seM-edX analysis performed on ive ground samples from F402, F482, F587, F686 and F755 (grounds of type d) indicates there are about 89 wt% pb, 6 wt% ca, 3 wt% Ba, 1.5 wt% zinc (Zn) and 1 wt% sulphur (s) (based on the averages of two to ive analyses per sample). 24 | semi-quantitative seM-edX analysis performed on two ground samples from F547, four from F821 and one from F511 (grounds of type e) indicates there are about 38 wt% Ba, 20 wt% Zn, 18 wt% s, 12 wt% ca and 8 wt% pb (based on the averages of one to three analyses per sample). 25 | the ground of painting F633 is very similar to grounds of type e, however the bottom part of the ground contains a higher amount of calcium. 189 Fig. 5 190 XRD spectrum of the bottom ground layer of Portrait of Doctor Gachet (F754), showing that the lead white is a mixture of hydrocerussite (red) and cerussite (dark blue). Calcite (calcium carbonate) (green) and barytes (barium sulphate) (pink) have also been identiied. Sphalerite (zinc sulphide) (light purple) may also be present. 180 170 160 150 140 130 lin (counts) 120 110 100 90 80 70 60 50 40 30 20 10 0 6.6 7 8 9 10 11 12 13 14 15 16 17 18 2-Theta - scale more lead (about 73 wt%) and less calcium (about 13 wt%) in their bottom layer (type F” in Table 1).26 The ifteen paintings with type F grounds were painted during two diferent periods: eight between July 1889 and November 1889 at Saint-Rémy (both subtypes F’ and F’’ identiied during that period), and seven in June 1890 at Auvers-sur-Oise (only subtype F’ identiied during that period). In grounds of type G, the bottom layer consists of high amounts of lead white and calcium carbonate, with some barium sulphate and lithopone. The upper layer consists of high amounts of lead white and lithopone with some calcium carbonate and barium sulphate.27 The bottom layer of type G is quite similar to the bottom layer of type F, but the upper layer contains much more lithopone and much less lead white. Grounds of type G were identiied in ive paintings, all painted in July 1890. SR-µXRD analyses were performed on a selection of ground samples from twenty-one paintings of the diferent lead-white based ground types (Table 1) to identify the crystalline phase(s) in the composition of the lead white. They revealed that the lead white in these grounds consists of a mixture of hydrocerussite (2PbCO3.Pb(OH)2 ) and cerussite (PbCO3 ) for nineteen paintings (ig. 5). However, cerussite was not detected in the ground samples from two paintings: Snow-covered field with a harrow (after Millet) (F632) and Cows 190 � Van GoGh's studio practice – johanna saLVant et al. 26 | semi-quantitative seM-edX analysis performed on fourteen ground samples from eight paintings with grounds of type F indicates there are two subtypes of ground of type F (referred as F’ and F”). Ground of subtype F’ (based on the semiquantitative seM-edX analysis of a total of twelve ground samples from F483, F754, F756, F764, F783 and F792): the bottom layer is about 60% wt pb, 20% wt ca, 10 wt% Ba, 5 wt% Zn and 4 wt% s; the upper layer is about 90 wt% pb, 2 wt% ca, 3 wt% Ba, 2 wt% Zn and 0.5 wt% s (based on the averages of three to ive analysis on each layer of each sample). Ground of subtype F” (based on the semi-quantitative seM-edX analysis of a ground sample from F634 and a ground sample from F697): the bottom layer is about 73 wt% pb, 13 wt% ca, 6 wt% Ba, 4 wt% Zn and 1.5 wt% s; the upper layer is about 86 wt% pb, 3 wt% ca, 3 wt% Ba, 5 wt% Zn and 0.5 wt% s (based on the averages of two to four analyses on each layer per sample). 27 | semi-quantitative seM-edX analysis performed on one ground sample from F822 (ground of type G): the bottom layer is about 53 wt% pb, 16 wt% ca, 17 wt% Ba, 6 wt% Zn and 5 wt% s; the upper layer is about 54 wt% pb, 3 wt% ca, 23 wt% Ba, 10 wt% Zn and 8 wt% s (average of ive analyses on each layer). (after Jordaens) (F822). This suggests either that cerussite is present in very small amounts and distributed very heterogeneously in the lead white used in these two grounds or, more likely, that the lead white used in these two ground formulations consists only of hydrocerussite. Variations in mineral phases found in lead white could be linked to distinct sources of lead white, with diferent manufacturing processes or geographical origins. It might also be the result of diferent storage conditions of lead white powder before being used in ground formulation (for example, the length of storage or luctuating environmental conditions).28 This parameter requires further investigation. PIXE analysis indicates that calcium carbonate and barium sulphate were added as a mixture in some ground formulations of type D and F. Indeed, for all the paintings analysed by PIXE with grounds of type D and F (two and ive paintings respectively, see Table 1), a correlation was found between the amounts of calcium and barium detected in the grounds, suggesting that calcium carbonate was added to the ground formulation as a mixture with a barium compound in all seven grounds.29 The barium compound could be barium sulphate or lithopone, however, no correlation was found between the amounts of calcium and zinc in those grounds, indicating that the barium compound is barium sulphate and not lithopone. Analysis allows us to estimate that the mixture is composed of roughly 90 wt% calcium carbonate to 10 wt% barium sulphate.30 The ingredient based on calcium carbonate and barium sulphate added in the ground formulation could have been prepared in two ways. The irst hypothesis is that it might have been made by co-precipitating calcium carbonate and barium sulphate. Heaton (1947) mentions an extender called ‘blancophone’ made by co-precipitating calcium carbonate and barium sulphate in a proportion of about 30/70;31 unfortunately, the reference does not clarify whether this is weight or molar proportions. In any case, the proportions described for blancophone appear to be the reverse of those for calcium 28 | Welcomme et al. 2007 identiied different compositional groups of lead white in sixteenthcentury paintings by Matthias Grünewald. Lead white consisting of hydrocerussite or cerussite alone, or a mixture of both in different proportions, was detected. these variations could be indicative of different sources of lead white. however, possible evolution of mineral phases in the lead white should be considered, as it has been observed that some environmental conditions of storage can lead to chemical reactions between one mineral phase to another in lead white powder. indeed, under ambient conditions, it has been reported that the equilibrium between hydrocerussite and cerussite in water can shift easily to either side, as it is closely related to partial pressure of co2 in the atmosphere. see Welcomme et al. 2007; Welcomme et al. 2006; Godelitsas et al. 2003. 29 | no mention has been made of a barium compound identiied as an impurity of sedimentary rocks used to produce natural calcium carbonate or of precipated calcium carbonate, which was manufactured in england from around 1850. see Gettens et al. 1997, pp. 204-06 and 210-11; perego 2005, p. 154. less likely. The second hypothesis is that it would have been made by mixing 30 | estimation of proportions in the ingredient based on mixture of caco3 and Baso4 does not take into account that the detected barium also originates from lithopone. together raw compounds in the desired proportions. This second hypothesis 31 | heaton 1947, p. 96. is more likely as relatively large tabular particles of barium sulphate were likely use of natural barium sulphate (barytes) as opposed to a synthetic variety 32 | among the grounds where a correlation between calcium and barium was found, individual particles of barium sulphate were detected in the grounds of F754, F756 F783 and F792 (grounds of type F). (blanc ixe).33 However, an ingredient based on calcium carbonate and barium 33 | see Feller 1986, pp. 55-57. carbonate and barium sulphate detected in our study, making this hypothesis detected in some grounds.32 The form and size of those particles point to the InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 191 Fig. 6 The weight percent of barium as a function of the weight percent of calcium from PIXE analysis. Grounds of type D (F686, F755) are represented with blue triangles; double grounds of type F (F483, F754, F756, F783, F792) with pink diamonds; and double grounds of type G (F822) with green squares. Multiple PIXE analyses were performed for each sample. A proile of analyses across all the ground layer(s) of the cross-section was recorded with a step ranging from 5 to 20 μm in function of thickness of the ground, as schematically shown on a micrograph of the double ground of The bedroom (F483). Each point on the graph represents one analysis in the ground sample. Proile amount of barium (wt%) 30.0 Grounds of type d 22.5 Grounds of type f Grounds of type G 15.0 7.5 0 0 7.5 15.0 22.5 30.0 amount of calcium (wt%) sulphate was not systematically used in Tasset et L’Hôte ground formulations. Grounds of type B contain calcium carbonate without any barium sulphate (Table 2) and no correlation was found between the amounts of calcium and barium detected by PIXE analysis in the type G ground of Cows (after Jordaens) (F822) (ig. 6). Trace elements were investigated by PIXE in a selection of ten grounds from paintings produced on Tasset et L’Hôte primed canvas supports. Three leadwhite based grounds of Van Gogh’s paintings with commercially primed canvas supports from other suppliers were also analysed by PIXE for comparison.34 Traces of copper, nickel and manganese were detected in these grounds. Copper was found in all grounds analysed by PIXE and appeared to be homogeneously distributed, while nickel and manganese, detected in many grounds, seemed to be present only locally in some areas.35 Copper, nickel and manganese were detected in amounts ranging from 0.02 to 0.1 wt%. Higher amounts of manganese (up to 0.2 wt%) were found in two grounds.36 Additional traces of chromium (c. 0.06 wt%) were detected in the three grounds of commercially primed canvas from other suppliers, while traces of chromium were not detected in any of the 192 � Van GoGh's studio practice – johanna saLVant et al. 34 | Père Tanguy (F363), Terrace of a café on Montmartre (F238) and Arles seen from the wheatields (F545). the suppliers of the primed canvas used to achieve these paintings have not been identiied. 35 | copper has been detected in similar amount along the proile of piXe analysis performed across the ground layers. By comparison, manganese and nickel were only detected in some analysis (or above limit detection) along the proile of piXe analysis across the ground layers, suggesting a heterogeneous distribution in grounds where it was detected. 36 | in some analyses, an amount of manganese up to 0.2 wt% has been detected by piXe in the bottom ground layer of F783 and in both bottom and upper ground layers of F822. Tasset et L’Hôte grounds analysed by PIXE. The origins of these trace elements were not identiied. Manganese may originate from the use of raw umber to colour the ground, although no particles of raw umber were detected in the ground samples studied. Traces of manganese and copper could also result from the use of driers.37 Lancelot et al. (1987)38 mentions detecting traces of copper, manganese and chromium in lead white produced in the second half of the nineteenth century, but in amounts ten times lower than those detected in the present study. The variation in the amount of chromium in the Tasset et L’Hôte grounds and those from other suppliers could be related to diferent lead white manufacturing processes. This presumption requires further investigation to identify the origin of the chromium and to conirm this trend in a larger group of works, as it might be speciic to the grounds found on primed canvas rolls sold by Tasset et L’Hôte. Ground Types and canVas maTchinG by Thread counT: idenTificaTion of differenT baTches of primed canVas Study of the canvas vertical and horizontal thread densities by automated thread counting from X-radiographs allows us to group painting supports with similar thread density patterns, referred to as a match group.39 This grouping provides information about painting supports originating from the same bolt of canvas. The information obtained by comparing the ground types for canvases from the same bolt can help narrow down a match group to the diferent rolls or 37 | perego 2005, pp. 663-64. batches (see ig. 1). Diferent rolls or batches thus identiied are here named by 38 | Lancelot et al. 1987, vol. 1, p. 69, table 1. For lead white produced during the second half of the nineteenth century, the amounts of chromium, copper and manganese are respectively 0-35 µg/g, 0-60 µg/g and 0-12 µg/g. For lead white produced earlier (1650-1850), the amount of chromium detected is the same, but amounts detected of copper and manganese are higher (respectively 150-220 µg/g and 70-110 µg/g), similar to the amount detected in our case. perego 2005, p. 95 also mentions that very old lead white could contain copper. juxtaposing a number (referring to the match group/bolt of canvas) to a letter (referring to the ground type). Among all the paintings analysed by automated thread counting,40 canvases with characteristic thread counts from Tasset et L’Hôte (ranging from about 11-12 warp by 14-18 weft threads per cm), presumably of ‘ordinary’ quality, fall into four diferent match groups (see chapter 9, Table 1): match groups 1, 2, 17 and 20.41 To date, match group 1 is by far the largest group, containing sixty-two paintings, of which twenty-one were analysed in the present study (Table 1). Match groups 2 and 20 contain thirty (eight analysed) and twenty-six (ten analysed) paintings respectively. Match group 17 is a relatively small group with currently only eight paintings (six analysed). InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 39 | see chapter 8 in this volume. 40 | about 440 Van Gogh paintings have been analysed using automated thread counting by johnson et al. 41 | average thread count values range from 11.211.6 warp × 16.1-17.5 weft thread per cm for match group 1; 11.8-12.3 warp × 14.4-16.5 weft per cm for match group 2; 11.4-11.8 warp × 14.1-16.4 weft per cm for match group 17; and 11.2-11.6 warp × 17.1-18.6 weft per cm for match group 20. 193 Match group 1: canvas from bolt 1 Paintings of match group 1 have been dated to between January 1889 and July 1890. Although belonging to the same bolt of canvas (bolt 1), three diferent types of ground (A, F and G) have been identiied among those paintings, allowing us to distinguish between at least three diferent batches of primed canvas: one primed with a single lead-white based ground extended with lithopone (batch 1.A) and two primed with double grounds made of lead white extended with calcium carbonate, barium sulphate and lithopone, but with signiicant variations in the proportions of the diferent compounds, especially in the upper layers (batches 1.F and 1.G). Batch 1.A primed canvases were used for Sunflowers (F458), two of the versions of Augustine Roulin (La berceuse) (F506 and F504), Self-portrait with bandaged ear (F527) and La Crau with peach trees in blossom (F514) between January and April 1889 (Table 1). Five other paintings, all executed in the same period, were also probably made from the same batch: three42 show matching weft-thread density patterns as some of the paintings produced from batch 1.A, and two others43 share a matching characteristic priming mark along the edge with two paintings from batch 1.A. Paintings from batch 1.F were produced over a longer period, between September 1889 and June 1890. This suggests they were probably made from more than one batch of primed canvas. Indeed, among the sixty-two paintings currently identiied as originating from bolt 1, none was painted in March and April 1890, while there were one to eleven canvases painted each month before (September 1889-February 1890) and after (May-June 1890) that period. This leads to the hypothesis that there were at least two diferent batches of canvas from bolt 1 primed with the ground of type F: one with evidences of use in September-November 1889 (batch 1.F.1) and another in June 1890 (batch 1.F.2). Supports for four paintings created in July 1890 – Wheatfield under thunderclouds (F778), Farms near Auvers (F793), Field with haystacks (F809) and Tree roots (F816) – come from batch 1.G. Match group 20: canvas from bolt 20 The twenty-six paintings from match group 20 were painted between August 1888 and September 1889. Among the eleven paintings on which ground analysis was performed, three types of grounds were identiied (Table 1): a single leadwhite based ground extended with calcium carbonate (type B), one with both 194 � Van GoGh's studio practice – johanna saLVant et al. 42 | Still life with potatoes (F386), january 1889, Sunlowers in a vase (F455), january 1889, and Orchards in blossom, view of Arles (F515), april 1889, have weft-thread patterns matching F458, F506 and F514 respectively. Matching weft-thread patterns between canvases indicate they come from the same roll, as bolts are only about 2.10 m wide. see chapter 8 in this volume. 43 | Augustine Roulin (La berceuse) (F508) and Still life with a plate of onions (F604), both painted in january 1889. X-rays of F508, F604, F506 and F458 reveal a matching large band (14-18 cm) of thicker primer along the edge. see hoermann Lister et al. 2001, pp. 365-66. F508 and F604 were also found to belong to match group 1 as a result of automated thread counting. calcium carbonate and barium sulphate (type C), and a double lead-white based ground extended with calcium carbonate, barium sulphate and lithopone (type F). These results indicate the use of at least three diferent batches of canvas from bolt 20: batches 20.B, 20.C and 20.F. Batch 20.C was used for Eugène Boch (The poet) (F462) in early September 1888 and Sower (F451) in late November 1888. Moreover, the latter shows a matching weft-thread density pattern with the unexamined paintings Portrait of Camille Roulin (F538), painted in December 1888, and Augustine Roulin (La berceuse) (F505), painted in January-February 1889, suggesting they were also made from batch 20.C. Batches 20.B and 20.F appear to span the same period, between July and September 1889. In July 1889, Undergrowth (F746) was painted from batch 20.F and Entrance to a quarry (F744) from batch 20.B. In September 1889, batch 20.B was used for The bedroom (F484), Wheatfield with a reaper (F618), Self-portrait (F627) and Pietà (after Delacroix) (F630), while The sheep-shearer (after Millet) (F634) and Peasant woman bruising flax (after Millet) (F697) were produced from batch 20.F. Van Gogh’s correspondence indicates that his order for painting materials made on 25 June 1889 (letter 783) was split into two consignments. He received the irst batch of canvas shortly before 6 July (letter 787) and the second a few days later, on 14 or 15 July 1889 (letter 789). These may be batches 20.B and 20.F, which would be consistent with their simultaneous use during JulySeptember 1889. Match group 2: canvas from bolt 2 Examples of paintings from match group 2, made from the same bolt of canvas (bolt 2) between July 1888 and July 1890, have been found between November and December 1888 and June 1890. Ground analysis of eight of these paintings reveals two types of ground (D and E) (Table 1), suggesting that Van Gogh used at least two diferent batches of canvas from bolt 2: batch 2.D commercially primed with a lead-white based single ground extended with calcium carbonate, barium sulphate and lithopone, and batch 2.E commercially primed with a lithopone based single ground. In December 1889-January 1890, canvases from batch 2.D were used for Olive grove with two olive pickers (F587), Noon: rest (after Millet) (F686) and possibly also Snow-covered field with a harrow (after Millet) (F632).44 Doctor Gachet’s garden (F755) was also made from batch 2.D a few months later in May 1890. InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 44 | F632 was painted in the same period (december 1889-january 1890) and it has the same ground type as F587, F686 and F755 from match group 2. however, F632 might be part of match group 2, but there is not strong conirmation. additionally, hydrocerussite and cerussite were detected in the ground of F686, while only hydrocerussite was detected in the ground of F632, which suggest that further batches need to be investigated. 195 The earliest known example of a paintings made using batch 2.E is Emperor moth (F610) in late May- early June 1889. This batch was also used for The reaper (after Millet) (F687) and Peasant woman binding sheaves (after Millet) (F700) in September 1889, and Acacia in flower (F821) in June 1890. Roses (F749), painted in April-May 1890, probably also used canvas from batch 2.E, as it shows a matching weft-thread density pattern with Emperor moth (F610). It is interesting to note that all the paintings from batch 2.E are made on a small format with very similar dimensions (c. 44 × 33 cm or c. 33 × 24 cm).45 Furthermore, matching weft-thread density patterns have been found in four of these paintings, indicating their supports were cut from nearby areas. Because so few weft-thread density matches were found among the large number of paintings studied by automated thread counting, the fact that matching weftthread density patterns were identiied for these small paintings suggests that batch 2.E was probably smaller in size than the other batch, perhaps only by a few metres. Match group 17: canvas from bolt 17 Based on the ground study of six of the eight paintings currently identiied as part of match group 17, three diferent types of ground (D, E and G) have been identiied (Table 1), leading us logically to the conclusion that there were at least three other batches (17.D, 17.E and 17.G). However, the number of paintings in match group 17 is currently small, providing limited evidence as to batch identiication. This hypothesis will require validation with additional data from more paintings in the future. The bedroom (F482), Garden with butterflies (F402) and The garden of the asylum with dandelions and tree trunks (F676)46 would have been painted from batch 17.D in October 1888 and spring 1890 respectively. Orchard in blossom (F511) and The thresher (after Millet) (F692) would have been produced from batch 17.E between April and September 1889. The Good Samaritan (after Delacroix) (F633) might have also been painted from the same batch later in May 1890. In mid-April 1889, Van Gogh informed his brother that he had placed a new order of canvas with Tasset et L’Hôte and referred to the study Orchard in blossom (F511).47 Batch 17.E used for this painting must therefore have come from a previous order. Batch 17.G would have been used for Cows (after Jordaens) (F822) in July 1890, not long before Van Gogh’s death. 196 � Van GoGh's studio practice – johanna saLVant et al. 45 | sizes of paintings (length by height): F610: 33.5 × 24.5 cm; F687: 44 × 33 cm; F700: 43 × 33 cm; F821: 33 × 24 cm; F749: 33.5 × 24.5 cm. 46 | F676, of april-May 1890, shows a matching weft-thread density pattern with F402, indicating that they were made from the same batch. 47 | ‘i have 6 spring studies, including two large orchards.’ see letter 758, note 3. It is worth noting that batches similar to batches 2.E and 17.E, both in regard to canvas thread count and priming (type E), seem to have been used to paint both The dance hall in Arles (F547) in December 1888 and The sheaf-binder (after Millet) (F693) in September 1889. At least eleven, possibly twelve, diferent batches of Tasset et L’Hôte primed canvas have been identiied among the seventeen canvas orders mentioned in Van Gogh’s correspondence, listed in Table 3. Based on the current evidence regarding the use of the diferent batches (summarised in Table 3) and present knowledge, the following chronological sequence can be suggested: Batch 20.C used in early September 1888-February 1889 Batch 17.D with earliest evidence of use in October 1888 and then later in spring 1890 Batch 1.A used in January-April 1889 Batch 17.E with evidence of use in April and September 1889 and possibly in May 1890 Batch 2.E used in May-September 1889 with evidence of additional use in spring 1890 Batch 20.F used in July-September 1889 Batch 20.B used in July-September 1889 Batch 1.F.1 used in September-November 1889 Batch 2.D used in December 1889-January 1890 with evidence of use also later in May 1890 Batch 1.F.2 used in June 1890 Batch 1.G with evidence of use in July 1890 Batch 17.G with evidence of use in July 1890 A summary of the batches that might have been received for the diferent orders recorded in Van Gogh’s letters is given in Table 3. perspecTiVes on TasseT eT l’hôTe’s acTiViTy and Van GoGh’s pracTice The results of this study raise several points concerning both Tasset et L’Hôte’s activity and Van Gogh’s practice. No evidence has yet been found that Tasset et L’Hôte owned premises large enough to produce painting materials on a large scale,48 suggesting that Tasset et L’Hôte was only a retailer of rolls of primed InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 48 | see hendriks and Van tilborgh 2011, p. 95, esp. note 16. 197 Table 3 Summary of evidence of use for the diferent batches of ‘ordinary’ quality, commercially primed canvas from Tasset et L’Hôte, ordered by Van Gogh between July 1888 and July 1890. The dates of receipt for the seventeen orders as recorded in Van Gogh’s letters are listed in the right-hand column with a note of the probable batch received. bolt 1 1.a bolt 20 1.F.1 1.F.2 1.G 20.c 20.F bolt 2 20.B 2.d bolt 17 2.e 17.d 17.e 17.G date of receipt (with quantity in metres) as recorded in Van Gogh’s letters probable batch received july 1888 before or on 1 july (5 or 10 m) (letter 635); before 31 july (10 m?) (letter 652) august 1888 before or on 9 august (10 m) (letter 658) 20.c september 1888 before 25 september (5 m) (letter 687) 17.d october 1888 9/10 october (10 m?) (letter 700) november 1888 about 9 november (10 m) (letter 719) 1.a and 17.e december 1888 january 1889 February 1889 March 1889 april 1889 mid-april (10 m) (letter 758) 2.e May 1889 june 1889 shortly before 9 june (5 m?) (letter 779) july 1889 shortly before 6 july (5 m?); shortly before 14/15 july (5 m?) (letters 787, 789) 20.F and 20.B september 1889 shortly before 20 september (10 m) (letter 805) 1.F.1 october 1889 on or about 24 october (5 m?) (letter 815) 2.d august 1889 november 1889 december 1889 shortly before or on 7 december (10 m) (letter 824) january 1890 shortly before 13 january (letter 839) February 1890 March 1890 april 1890 May 1890 before 11 May (7 or 10 m) (letter 870) 1.F.2 june 1890 before 3 june (10 m?) (letter 877); 17 june (10 m?) (letter 889) 1.G and 17.G july 1890 198 � Van GoGh's studio practice – johanna saLVant et al. canvas, not a manufacturer. This hypothesis is also consistent with the fact that Van Gogh looked for the price ‘at irst hand’ of another quality of canvas sold by the roll at Tasset et L’Hôte.49 According to the proposed chronological sequence of Van Gogh’s use of diferent batches, it appears that batches from the same bolt were sent over a long period of time (Table 3). For example, batches originating from bolt 1 were received in 1888 and also in 1890.50 Moreover, batches sent to Van Gogh were alternatively made from diferent bolts of canvas over time. All these facts suggest that rolls of primed canvas of ‘ordinary’ quality from Tasset et L’Hôte were not produced to order. This practice can be partly explained by two facts. First, drying of the ground required at least three months,51 therefore, each large frame could produce only about four rolls of primed canvas a year. Second, primed rolls were stored not only at the manufacturer, but also at Tasset et L’Hôte’s, depending on the volume of sales. When an order came in the colour merchant could select a roll in either stock. Data from this study fosters the conclusion that the batches of primed canvas Van Gogh purchased from Tasset et L’Hôte were primed by the same manufacturer. Indeed, examples of grounds with similar formulations (types E, D, F and G) were found on the diferent bolts and evidence of use of the same type of mixture of calcium carbonate and barium sulphate was found in the priming of some canvases from both bolts 1 and 2. Furthermore, it is striking to note that paintings from match groups 2 and 17, despite originating from diferent bolts,52 show very similar primed canvases, both regarding distribution of warp and weft thread counts and ground types (D and E). These facts suggest that these two bolts might have been obtained from the same weaver within a short period of time and that some rolls of canvas cut from these bolts were probably primed, if not simultaneously, then over a very short period of time (batches 2.D and 17.D; batches 2.E and 17.E). This study of grounds has revealed the variation over time in the manufacturing process to produce rolls of primed canvas sold by Tasset et L’Hôte. Interestingly, despite these sometimes signiicant variations, the rolls were presumably sold under the same quality category according to the letters.53 As indicated earlier, the chronological use of canvas batches by Van Gogh probably does not relect the order of manufacturing, however, it is interesting to note that there seems to be an evolution from single to double grounds in the rolls of ‘ordinary’ quality canvas received by Van Gogh. The irst evidence of use of a batch with a double ground is July 1889; before then only batches with InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 49 | Letter 635: ‘now, for his canvas at 4.50, i’ll probably be able to ind out the price by the piece at irst hand.’ Van Gogh is referring to the canvas sold by the roll from tasset et L’hôte. 50 | From the preserved letters, there is no record of Van Gogh receiving any orders from tasset et L’hôte between december 1888 and april 1889, suggesting that batch 1.a with evidence of use in early 1889, was received before december 1888. 51 | For more on drying times, see Mérimée 1830, p. 244; Bouvier 1832, pp. 517-32; Labreuche 2011, p. 231. Mérimée indicates a drying time of two to three months in summer and ive to six months in winter, Bouvier of eight to twelve months, and Labreuche of three months. 52 | Based on the thread density variations and current data from automated thread counting, canvas from match groups 2 and 17 originate from at least two distinct bolts of canvas, although very similar. 53 | however, it is interesting to note that examples of overlap in the quality of canvas sold as ‘ordinaire’ and ‘ordinaire étude’ at Bourgeois aîné, showing very similar thread counts, have been reported. see chapter 9, note 45. 199 single grounds have been identiied. Many batches presumably received later have double grounds (batches 1.F.1, 1.F.2, 1.G, 17.G). It has been suggested that commercially primed canvas might have been available with diferent priming thicknesses, providing two distinct types of surface: à grain (single ground) and à lisse (two or three ground layers).54 These elements suggest that irst batches of canvas from Tasset et L’Hôte Van Gogh received were with à grain preparation, while the batches he received later were mainly with à lisse preparation. This change from à grain canvas to à lisse canvas could be the result of a request from Van Gogh for supports with a less textured surface.55 It might also be the case that Tasset et L’Hôte sold a slightly diferent variety of ‘ordinary’ canvas. Despite the distinct types of ground formulation observed over time between the diferent batches/rolls, the lead-white based grounds remain quite similar in formulation. This suggests that workers formulated the ground using diferent raw materials at their disposal without following precisely the same procedure, but sometimes with particular instructions, such as making a formulation based on lithopone (batches 2.E and 17.E). Products based on lead white, lithopone and the mixture of calcium carbonate and barium sulphate previously identiied, were some of these raw materials. Variations in ground formulations might also orginate from the use of diferent qualities of these raw materials. 56 Comparison of Tasset et L’Hôte grounds with the commercially primed grounds from other suppliers found in Van Gogh’s paintings57 indicates that the use of lithopone in ground formulation is speciic to Tasset et L’Hôte grounds.58 It is interesting to note that lithopone has been used as the main component in the majority of ground types identiied and as a secondary compound or in high amounts in the upper layers of the double grounds of the paintings Van Gogh produced shortly before his death. When present in large amounts, it is likely that lithopone was used as a substitute for the more expensive lead white pigment.59 Knowledge of the diferent batches of primed canvas Van Gogh ordered from 55 | as far as we know, Van Gogh did not comment on these different types of grounds in his letters. it could be interesting in future research to look into the inluence of the variation of grounds on the visual aspect and conditions of Van Gogh’s paintings in order to relate these analytical indings to the paintings. Besides a systematic study of this speciic aspect of the actual paintings (for example by comparing raking light details of the tacking margins), this could also be based on the study of reconstructions. 56 | For example, lead white could be purchased under a variety of names and as quite different products, especially regarding the nature of the illers and the proportions of lead white. see Bomford et al. 1990, p. 48; carlyle 2001, pp. 512-15; perego 2005, pp. 95-96; stols-Witlox 2011. 57 | For more about the commercial grounds found in Van Gogh’s paintings, see Van tilborgh and Vellekoop 1999, pp. 21-22; hoermann Lister et al. 2001, pp. 354-69; hendriks and Geldof 2005; hendriks and Van tilborgh 2011, pp. 90-117; and salvant 2012, pp. 211-51. see also chapter 9 in this volume. and by the fact that the same batch is used to produce paintings from diferent 58 | Church at Auvers (F789), painted in june 1890 on a commercially primed canvas, also has a lithopone based ground. however, thread count of its canvas (16.3 × 17 thread/cm2) differs from the asymmetric thread count of the ‘ordinary’ quality canvas from tasset et L’hôte, and the cusping on all four sides indicates it was primed on a small stretcher. these considerations might indicate that the canvas used to paint F789 was from tasset et L’hôte. periods, as exempliied by the paintings from batch 17.D. This later practice is 59 | Lahousse 1909, p. 5. Tasset et L’Hôte, along with the other information such as his correspondence, helps us in dating his paintings. Nevertheless, this study highlights the fact that batches were not used in the order they were received, as illustrated for instance by the simultaneous use of batches 20.B and 20.F in July to September 1889 200 54 | see callen 2000, p. 32. interestingly, callen also indicates that rolls of ‘toile ordinaire with a jaune tinted preparation’ from Lefranc were available with one, two or three ground layers. � Van GoGh's studio practice – johanna saLVant et al. not surprising, as the choice of batch would be highly dependent on the painting format required. One might expect that large or medium formats would be more likely to be cut from new batches, while smaller formats could be made from the remnants of older batches. Such practice demonstrates that data about batches needs to be interpreted with care. InvestIgatIon of the grounds of tasset et L’hôte commercIaLLy prImed canvas 201