Non-destructive techniques for in situ studies of easel paintings: development and application of a confocal X-ray micro-fluorescence system

Universidad de Sevilla Departamento de Física Atómica, Molecular y Nuclear Non-destructive techniques for in situ studies of easel paintings: development and application of a confocal X-ray micro-fluorescence system PhD Thesis by Kilian Laclavetine Doctoral program of Nuclear Physics Supervisors Prof. Miguel Ángel Respaldiza Galisteo D. Francisco Ager Sevilla, 2015 Universidad de Sevilla Departamento de Física Atómica, Molecular y Nuclear Técnicas no destructivas para el estudio in situ de pinturas de caballete: desarrollo y aplicación de un sistema de micro-fluorescencia de rayos X confocal Tesis doctoral presentada por Kilian Laclavetine Programa de doctorado de Física Nuclear Dirigida por Prof. Miguel Ángel Respaldiza Galisteo D. Francisco Ager Sevilla, 2015 1 A mes parents, Suzy et Jean François, Pour leur soutien constant, Pour leur amour éternel, Pour m’avoir permis d’être un rêveur. 2 3 “Los hombres desaparecemos unos tras otros, y las generaciones se suceden como las olas de polvo que levanta el viento en los caminos; pero las acciones y la memoria de cada uno de los que producen males o bienes, deben quedar en los demás como recuerdo imborrable, para que sirvan de estímulo o de escarmiento, y para que los que nos sucedan sigan o se desvíen de este, o de aquel camino…” Diccionario universal de historia y geografía: v. I, p. III-IV. México, Tipografía de Rafael, Librería de Andrade, 1853-1856. 4 Declaration of the author of this dissertation: Aware of legal responsibility for making untrue statements I hereby declare that I have written this dissertation myself and all the contents of the dissertation have been obtained by legal means. Declaration of the thesis Supervisor: This dissertation is ready to be reviewed. Disclaimer: Reference to any companies or specific commercial products was not intended as an advertisement. 5 PREFACE This work was supported by the Junta de Andalucía project P09-HUM4544 named: “Análisis de la Colección Pictórica de los Siglos XV-XVI del Museo de Bellas Artes de Sevilla Mediante Técnicas Nucleares no Destructivas” leaded by Prof. Miguel Ángel Respaldiza Galisteo. During this PhD thesis, several foreign internships have been made: The study of “Martirio de San Hipólito” at the National Museum of History in Chapultepec was made in September 2011 in Mexico (supported by AUIP grant) while the study of the “Virgen de las Uvas” was made in October 2013 at the Metropolitan Cathedral of Mexico (supported by CONACyT CB2012-179601 project). The study of experimental paint multilayer samples by infrared reflectography and the study of stratigraphic micro-samples by optical microscopy were made at the Institute of Aesthetic Research of the National Autonomous University of Mexico (IIE-UNAM) in January 2013. Finally, the CXRF study of experimental paint multilayer samples at the Center for Research and Restoration of Museums of France (C2RMF), Museum of Louvre in Paris, was made between September and December 2012 with the supervision of Dr. Thomas Calligaro. 12 2.1. μXRF-CONCHA system, CXRF device for in situ analysis .......................... 55 2.1.1. Description of the set-up .............................................................................. 55 2.1.2. Set-up characterization and tune-up............................................................. 58 2.1.3. Data treatment .............................................................................................. 64 2.2. LouX3D system and CXRF set-up at AGH University of Science and Technology .................................................................................................................... 64 2.3. 2D motion device at the Fine Arts Museum of Seville ................................... 70 2.4. IRR motion systems and mosaic production software ................................... 72 2.4.1. IRR images capturing methods .................................................................... 72 2.4.2. Mosaic production software ......................................................................... 74 2.4.3. Stitching methods for images obtained with the 2D motion device ............ 75 2.4.4. Stitching methods for images obtained with the Gigapan system ............... 76 2.5. XRF setups used during Alonso Vazquez Project .......................................... 78 2.5.1. Description of the X-Panda set-up ............................................................... 79 2.5.2. Description of the “Museum XRF device” and the SANDRA system. ....... 82 2.5.3. Data acquisition and processing ................................................................... 85 CHAPTER 3. IRR AND CXRF STUDIES OF EXPERIMENTAL PAINT MULTILAYERS .............................................................................................................. 87 3.1. Making of experimental paint multilayers...................................................... 87 3.1.1. Preparation of the wood support .................................................................. 87 3.1.2. Application of the ground layer ................................................................... 88 3.1.3. Application of the priming layers ................................................................ 89 3.1.4. Application of the paint layers ..................................................................... 90 3.2. Comparison of the capabilities of the IRR cameras....................................... 92 3.2.1. IR reflectivity classification ......................................................................... 93 3.2.2. Si-CCD camera issue ................................................................................... 93 3.2.3. Specific reflectivity of several inorganic pigments ..................................... 94 13 3.2.4. Discrimination of inorganic pigments with the same color ......................... 95 3.2.5. Influence of overlapping layers of different inorganic pigments................. 97 3.2.6. Conclusions of the IRR study ...................................................................... 98 3.3. Multilayered samples CXRF study .................................................................. 99 3.3.1. Thin foil samples study with µXRF-CONCHA .......................................... 99 3.3.2. Experimental paint multilayers study with µXRF-CONCHA ................... 100 3.3.3. Intercomparison of three CXRF systems ................................................... 105 3.3.4. Conclusions of the CXRF study ................................................................ 108 CHAPTER 4. ALONSO VÁZQUEZ PAINTINGS STUDY ................................. 109 4.1. Alonso Vázquez and their artworks .............................................................. 109 4.1.1. Vida del pintor (Life of the artist) .............................................................. 109 4.1.2. “Sagrada Cena” .......................................................................................... 111 4.1.3. “Virgen del Valle” ..................................................................................... 111 4.1.4. La serie sobre la vida de San Pedro Nolasco (“Nolasco” series) ............... 114 4.1.4.1. “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” .......... 114 4.1.4.2. “San Pedro Nolasco redimiendo cautivos” ............................................ 114 4.1.5. “Inmaculada” ............................................................................................. 115 4.1.6. “Martirio de San Hipólito” ......................................................................... 116 4.1.7. “Virgen de las Uvas” ................................................................................. 117 4.1.8. Summary .................................................................................................... 119 4.2. Methodology .................................................................................................... 120 4.2.1. UV-induced visible fluorescence photography and IRR ........................... 120 4.2.2. XRF ............................................................................................................ 120 4.2.3. CXRF spectrometry ................................................................................... 121 4.2.4. Other techniques applied............................................................................ 121 4.3. Alonso Vázquez paintings study .................................................................... 122 14 4.3.1. UV-induced visible fluorescence photography .......................................... 122 4.3.2. IRR ............................................................................................................. 124 4.3.4. Inorganic pigments study with XRF .......................................................... 127 4.3.5. Accuracy of the methodology applied to the treatment of the XRF data .. 129 4.3.6. 1D analysis with XRF ................................................................................ 133 4.3.7. CXRF spectrometry ................................................................................... 137 4.4. Synthesis of the corpus studied ...................................................................... 140 4.4.1. Preparatory drawings ................................................................................. 140 4.4.2. Pentimenti .................................................................................................. 140 4.4.3. Pigments used on artworks produced in Seville. ....................................... 143 4.4.4. The Mexican artworks ............................................................................... 143 4.4.5. CXRF results .............................................................................................. 145 4.4.6. SEM-EDX results ...................................................................................... 146 4.5. Conclusions of the Alonso Vazquez’s artworks study ................................. 147 CONCLUSIONS ............................................................................................................ 151 APPENDIX ..................................................................................................................... 155 I. Specifications of the micro-focus X-ray source .................................................... 155 II. Specifications of the two polycapillary lenses of the µXRF-CONCHA device ... 156 III. Resolution of SDD detectors ............................................................................. 158 IV. Specifications of the X-123SDD model detector from Amptek ....................... 159 V. Treatment of the data obtained with CXRF and XRF setups using PyMca software 160 VI. Specifications of the two polycapillary lenses of the LouX3D device ............... 162 VII. "Louvre0" acquisition software of LouX3D set-up (in French) ......................... 164 VIII. Specifications of X-ray tube Mini-X model from Amptek ............................... 173 IX. In-depth scan results with the three CXRF devices .......................................... 174 X. Color scale classification of XRF peak area values by order of magnitude ......... 180 15 XI. Brief report of each artwork analyzed ............................................................... 182 i. “San Pedro Nolasco despidiéndose de Jaime I el Conquistador” ..................... 183 ii. “San Pedro Nolasco redimiendo cautivos” ....................................................... 185 iii. “Sagrada Cena” .............................................................................................. 187 iv. “Inmaculada” ................................................................................................. 190 v. “Virgen del Valle” ............................................................................................. 192 vi. “Martirio de San Hipólito” ............................................................................ 194 vii. “Virgen de las Uvas” ..................................................................................... 196 INTERNATIONAL CONGRESSES AND AWARD ................................................. 199 BIBLIOGRAPHY .......................................................................................................... 201 16 17 INTRODUCTION Archaeometry is a field of discipline developed during the last century, which consists in the application of sciences (physics, chemistry, geology, biology, etc.) to cultural heritage studies of archaeological artifacts and artworks. On the one hand, archaeometry was brought about by the advent of new technology yet, on other hand, by the natural evolution of archaeology, art history and other historical sciences. Indeed, those research areas need to test the veracity of the bibliography, in order to obtain new information and clues confirming or refuting the hypotheses presented within and to revisit what we know about our past. The key resides in trans-disciplinary research, which allows for studying new perspectives and hypotheses and for testing old ones. Following from that, within the world of physics applications there are basically two main areas of study: dating techniques (14C, K/Ar, thermoluminescence, etc.) and material characterization techniques, sometimes called archaeophysics (Fig. 1). In the case of easel painting characterization studies, the main goals of interdisciplinary collaboration with art historians and conservator-restorers is to characterize the artist’s palette and technique, and locate previous restoration interventions and reintegration zones in order to help conservator-restorers work and to confirm the authorship of some paintings, on the basis of material and aesthetical study [1]. Archaeometry Physics, Chemistry, Mathematics, Biology, Geology, Computing, etc. Cultural Heritage applied Physics (Archaeophysics) Material characterization techniques Dating techniques (14C, K/Ar, Thermoluminiscence, etc.) XRF, IRR, CXRF, UV-Visible, OM, SEM-EDX PIXE, RBS, XRD, SXRF, Raman, FTIR, LA-ICP-MS, Resistivity, etc. Applied to Archaeology, Anthropology, History, Art History, Restoration, Conservation, Linguistics, etc. Fig. 1. Archeometry and several techniques involving applied physics. 18 The contribution of this thesis to achieve these goals is focused on the complementary application of two spectroscopic techniques: - X-ray fluorescence spectroscopy (XRF) [2], - Confocal X-ray micro-fluorescence (CXRF) [3]. Along with two imaging techniques: - UV-induced visible fluorescence photography [4], - Infrared reflectography (IRR) [4, 5]. All these techniques have the advantage of being non-destructive and non-invasive to allow the analysis of unique and fragile artifacts. In the last twenty years, these nondestructive techniques have been highly demanded within the cultural and historical heritage world whose main objective is the long-term preservation and integrity of artwork. Some of those techniques (XRF, UV, and IRR) have been applied for several decades in the study of paintings, providing universally accepted results. In the case of IRR it is noteworthy that it remains undeveloped at a very basic level of diagnosis [6]. This work proposes to adapt IRR to the study of large artworks and delve deeper into the analysis of pigments. Moreover, the CXRF is a relatively new technique with the first application and subsequent publication of a pictorial heritage happening as recently as 2009 [7]. It is one of the most recent non-invasive techniques in the field of cultural heritage studies. The confocal X-ray micro-fluorescence is gaining considerable interest because it can provide compositional and spatial information that are typically obtained using standard micro-invasive and/or micro-destructive techniques (e.g. scan electron microscope in combination with energy dispersive X-ray spectroscopy (SEM-EDX)), without the need for sampling [8, 9]. In particular, CXRF extends the capabilities of the micro X-ray fluorescence to the depth profiling of multilayered materials. Due to the novelty of this developing technique, its abilities remain to be clearly defined. This work focuses not only on the methodological application of these four techniques, but also presents the mechanical engineering and computing developments made during this PhD. A large effort has been made to provide several in situ scientific setups (e.g. µXRF-CONCHA and X-Panda systems) which seek to study artwork directly in places of exhibition, so that artwork is not moved unnecessarily. Internationally renowned and well-funded laboratories, such as the Centre de recherche et de restauration des musées de France (C2RMF) in the Louvre Museum of Paris [10, 11], usually grant permission to move artwork for study (requiring large funding for insurance rates, the capacity to solve complex administrative requirements, the necessary funds to cover the high cost of transporting delicate artwork to adequate facilities prepared for the reception and storage of artwork during the study, etc.). However, in many other cases, the development of in situ studies is crucial. Those scientific setups developed during this PhD made possible an international collaboration between the University of Seville, the Fine Arts Museum of Seville and the Universidad Nacional Autónoma de México (UNAM) about to research artwork by Alonso Vázquez, an important painter from Seville [12]. Vázquez was active in the second half of the 16th century and in the first decade of the 17th century. In 1603 he travelled to New Spain (modern-day Mexico) as a member of the Viceregal Court. There, he left paintings as well, though regrettably his works, discovered on the other side of the Atlantic, are decontextualized and unsigned. The main goal of this research is to 19 characterize the technique used by Vázquez and to confirm the authorship of some paintings in Mexico, using material study. We begin by explaining the different techniques applied in this thesis, and the function of the main components. Then, we will present the work done in the instrumental development and characterization of each setup and the design and operation of the corresponding acquisition programming software. We will discuss the different components of each portable device, their capabilities and their advantages for the study of paintings. Next, we will detail the results of the intercomparison of three CXRF setups, the LouX3D system at the C2RMF laboratory [11], the CXRF set-up at AGH University of Science and Technology in Krakow [13] and the µXRF-CONCHA system developed at the Centro Nacional de Aceleradores (CNA) in Seville. The aim is to check the reliability of the new Seville device and to compare the performance of each CXRF setup. Finally, we will describe the methodology applied in the corpus study of seven pieces of artwork attributed to the painter Alonzo Vázquez. We will illustrate with emblematic examples the contribution of each technique in this global study. Finally, we will compare the artworks attributed to Vázquez during his period in Mexico and his previous period painting in Seville. The technique of the artist and his palette in terms of inorganic pigments will be shown amongst other information provided by the archaeometric investigation. In conclusion, the attribution by historian of art of five of those paintings will be discussed based on the material study. 20 21 Chapter 1. Spectroscopic and multispectral imaging techniques applied to in situ studies of easel paintings The first chapter of this thesis presents the theory behind four different techniques for material characterization of artwork. It also illustrates the experimental development incurred for the UV-induced visible fluorescence photography and part of the infrared reflectography. 1.1. X-ray fluorescence spectroscopy (XRF) 1.1.1. XRF principals X-rays are part of the electromagnetic spectrum and were discovered and described by W.C. Röntgen in 1895 [14, 15]. Only a portion of this spectral range, between 1 and 100 keV, is used in energy-dispersive X-ray spectroscopy (EDS). One of the EDS method is named X-ray fluorescence (XRF), is a non-destructive characterization method based on the interactions of X-rays with matter [16]. Those interactions can be divided into two categories: dominant photoelectric absorption and scattering (cf. Chapter 1.2.7). More specifically, XRF uses the emission produced by photoelectric effect of characteristic X- rays of elements composing a material. This secondary X-ray emission is produced when a material is irradiated by a primary X-ray beam [17]. This technique needs an X-ray beam to illuminate the object of interest and an X-ray detector to collect the signal emitted by the irradiated item. For this thesis, an X-ray tube and a silicon drift detector (SDD) were used. The emitted X-rays are specific to each element that exists in nature, thus allowing for their identification and as a result a deduction of the likely constituents of the analyzed material (Fig. 2) [18]. This noninvasive technique allows users to execute qualitative and semi-quantitative multi-elemental analysis in a short period of time. This method is highly sensitive and able to detect even trace or minor elements. Quantitative analysis may also be performed with high accuracy [19] and allows for the analyzing of artwork without physical contact or causing visible damage to the studied areas with respect to the experimental conditions usually employed in this case study [20]. In principle, this technique permits the identification of all chemical elements, except for hydrogen and helium, which do not produce X-ray fluorescence. However, other light elements may be difficult to measure due to several reasons (the matrix effect on the item studied, absorption by the air or absorption by the detector window), which limit the study of all elements with a higher atomic number than 13 (Aluminum, Al). Nevertheless, the analysis can be extended to sodium (Na, Z = 11) using thin polymer detector window and helium atmosphere between the sample and the X-ray detector [21, 22]. In higher energy range, the limitations are related with energy of exciting radiation (intensity of primary beam for energies above the absorption edge of element) and detector thickness. 28 1.1.3. SDD X-ray detector This paragraph describes the operation of the silicon drift detector (SDD). The main characteristics of the X-ray detector are: - Active area; - Quantum efficiency; - Energy resolution; - Peak to background ratio; - Maximum count-rate. The active area determines the solid angle of the radiation emitted from the sample to be registered and defines the total detection efficiency. Quantum efficiency depends on two factors: at higher energies it grows in conjunction with detector thickness and at low energies is connected to entrance window thickness. Energy resolution depends on the total noise produced in the detector volume and by the detector electronics. Consequently, the output resolution (expressed in FWHM) can be written as [32]: 22 det 2selectronicectoroutput FWHMFWHMFWHM  (4) The detector volume noise is connected to the Poisson statistics of charge production process, which occurs in detector volume and increases in relation to the energy of detected radiation [32]: EFFWHM ector  35.2 det  (5) where F is the Fano factor, which describes the deviation from Poisson statistics, ε is the mean energy required to produce an electron-hole pair and E is the energy of incident photons. Electronic noise can be caused by thermally generated leakage currents in the detector volume. Thus, the resolution degrades in accordance with detector size. It can also be introduced at the first stage of pulse processing. This effect is avoided by cooling the input transistor of the preamplifier along with the detector. Fig. 8. Components of a SDD detector [33]. The silicon drift detector (SDD) is widely used nowadays in XRF applications. The SDD concept was introduced in 1984 by Gatti and Rehak [34] and in 1985, Rehak et al. 29 proposed an SDD optimized for X-ray spectroscopy [35]. Later, in 1987, Kemmer et al. improved its design significantly [36]. Even nowadays, investigation in the area of SDD is still very active [37-40]. The SDD is highly sensitive to 1 to 30 keV X-rays. The lower limit of detection is due to the incapacity of the radiation with an energy below 1 keV to pass through the Be detector window (Fig. 8) [39]. The upper limit of detection is fixed by the silicon crystal thickness. The SDD provides better performance than conventional planar detectors by using a special type of photodiode, functionally similar to a planar photodiode but with a specific electrode structure. The small surface of the anode provides the major advantage, meaning a much lower capacitance than a planar diode with the same area, resulting in better energy resolution and higher count rates. Also, it presents less electronic noise than an equivalent planar detector, especially for short measurement times. It provides the SDD with better energy resolution for moderate count rate and even better for high count rate. Its great advantage is that the low energy peaks can be measured with high count rate and only a small loss in resolution. The SDD photodiode cathode is composed of a p-type silicon cathode and an n-type silicon anode (Fig. 9). X-ray photons enter the photodiode and generate electron-hole pairs. Around the very small central anode, a series of concentric derivation electrodes which are negatively polarized create an electric field. That focuses the electrons in the central anode, which is directly connected to the FET (field-effect transistor) (Fig. 9). Optimal resolution is obtained by stabilizing the system and cooling it electrically (using the Peltier effect) to less than 15 ºC. Moreover, with this design, the active volume of the diode can be expanded, thus adding more derivation electrodes without modifying the small size of the area of the central anode, i.e. without changing the input capacitance. Due to its unique electrode structure, a very high active area is available without degradation of performance. Incident X-rays Radial drift field Edge of active volume Inner Drift Ring P+ (-20V) Outer Drift Ring P+ (-100V) Signal Anode N+ (FET) Cathode Thin P+ (-50V) Fig. 9. Scheme of the Amptek SDD design [33]. Besides, the electronics of the signal processing is almost identical to that of planar silicon detectors. As presented in Fig. 10, the anode is connected to a charge sensitive preamplifier and to the pulse processor, which detects the pulse and measures its amplitude. The bottom right of Fig. 10 illustrates the pulse form in each step of the signal process. 30 Incident X-ray SDD Detector electrons ions Cathode (-) Anode Electrods (-) Preamplifier I (t) CF Shaping Amplifier Reset Circuit Low Frequency Feedback Detector Bias Supply Fig. 10. Scheme of the SDD electronics of the signal processing [33]. 1.2. Confocal X-ray fluorescence (CXRF) spectrometry 1.2.1. CXRF spectrometry principals and historical background As explained previously, X-ray primary radiation is employed to induce the emission of characteristic X-ray radiation in XRF. This characteristic X-ray emission is not only generated superficially but also emerges from extended depths (several tens to hundreds of micrometers) below the surface of the material. This phenomenon is exploited by confocal μ-XRF [1]. CXRF extends the potential of the technique known as micro X-ray fluorescence (μ- XRF) to depth profile of multilayer materials. The μ-XRF technique consists of the placement of a polycapillary X-ray lens at the exit of micro-focus X-ray tube, with 45º between the excitation and detection channels and 90º between the excitation channel and the surface of the sample (Fig. 11). Meanwhile, in CXRF, an additional polycapillary lens at the entrance of the X-ray detector is employed to limit the space from which X-ray photons can be detected (Fig. 11) [41]. In the case of CXRF, the excitation channel forms an angle of 45° with respect to the surface of the irradiated object and 90° with respect to the detector channel. A study from Bjeoumikhov et al. determined that this angle configuration procures the best depth resolution [42]. Several µXRF devices are commercially available (Artax [43] or M6 JetStream [44] from Bruker, for example). Nevertheless, at present, CXRF setups are not commercially available. Confocal micro-X-ray fluorescence is a non-destructive technique that can provide depth-resolved information about elemental composition and it has proven its suitability in the investigation of stratified structures in microscales [3, 45-47]. A probing microvolume (called confocal volume) is defined by the overlap of the foci of both X-ray lenses [11]. By scanning materials through this micro-volume, one can obtain intensity profiles reflecting the local elemental composition variation versus depth or versus a lateral coordinate (with a resolution in the range of 3 to 10 μm for most CXRF setups) [1]. The maximum depth of analysis is limited by absorption effects (cf. Chapter 1.2.7) and usually does not exceed 1 mm in low-Z matrices or tens of microns in heavy-Z matrices. 31 SDD detector Sample X-ray Tube 45º µXRF Polycapillary half lens or Poly-CCC SDD detector Analyzed micro-volume Sample X-ray Tube Depth profile 90º CXRF Fig. 11. Configuration of µXRF (a) and CXRF technique (b). CXRF is associated with another technique named Synchrotron X-ray fluorescence (SXRF). In SXRF, a synchrotron monochromatic beam is used for the excitation channel instead of a polychromatic micro focus X-ray tube beam (CXRF). The idea is similar but the results obtained present differences. Indeed, Kanngießer et al. [48, 49] compared the sensitivity of table-top and synchrotron setup. These two techniques can be used to perform confocal studies. Nevertheless, this work only focuses on CXRF study. In the late 1980s the quick development of capillary optics took place. The transport abilities of capillaries were presented by Rindby in 1986 [50]. The use of capillary optics opened up the possibility to create efficient laboratory micro-XRF setups. The first laboratory spectrometers utilizing capillary optics were presented in 1989 by Carpenter [51] and Rindby [52]. In 1990 Kumakhov [53, 54] introduced the polycapillary X-ray lens as a new method of focusing X-rays. Due to the high intensity gain it became the widest spread X-ray focusing device. The configuration of the confocal micro-XRF system was decided upon in 1993 by Gibson and Kumakhov [41]. In 2000, Ding et al. presented the first experiment with a CXRF setup for three dimensional elemental analyses (Fig. 12) [8]. Up until now, several CXRF setups have been presented worldwide. At the International Atomic Energy Agency (IAEA) [55] and Vienna University of Technology, both in Austria [56], at Beijing Normal University in China [8], at C2RMF (Centre de recherche et de restauration des musées de France) in France [11], at the Technical University of Berlin, Germany [48], at Osaka City University in Japan [57, 58], at AGH University of Science and Technology in Poland [59] and at Los Alamos National Laboratory in New Mexico, USA [60, 61]. CXRF is a relatively novel technique and may appear under different names in scientific publications. This fact may difficult the visibility of the technique. Sometimes, the name used focuses on the 3D ability of the technique for example “3D micro-X-ray fluorescence” [9] or “confocal 3D XRF” [7]. Acronyms such as 3D-XRF will be not used in this work as it results in a misuse of language. On one hand, CXRF can provide depthresolved information (3D analysis) but other X-ray emission based techniques can do it as well as: - classic XRF and μXRF [21, 62-66], - PIXE and RBS [64], - μPIXE [67, 68]. (a) (b) 32 On the other hand, in the study of paintings, CXRF is mostly used to obtain depth profiles (1D study) or 2D profiles [3, 69, 70]. 3D study of paintings can be very timeconsuming and does not provide additional information respect to 2D analysis. Fig. 12. Diagram of the experimental setup of the first CXRF system [8]. 1.2.2. CXRF application in Cultural Heritage Most of the published confocal studies in Cultural Heritage were performed in synchrotron installations (SXRF) [27, 45-47, 71, 72]. The significance of confocal micro X-ray fluorescence has increased quickly in the last decade. The numerous applications have proven that this method is now a well-established tool for non-destructive imaging of three-dimensional distributions of chemical elements within sampling. This method became especially useful for testing multi-layered systems in micrometer scale. In the study of artwork, it is becoming a strong non-destructive alternative to the classical scanning electron microscopy supplemented with energy-dispersive X-ray spectroscopy (SEM-EDX) applied to cross section micro-samples [73]. In 2003, the first publication of depth analysis on 18th century Mughal miniature paintings by means of SXRF by Kanngießer et al. took place [9]. In 2005, Kanngießer et al., again, published the first CXRF study of experimental paint layers [48]. In 2009, the first CXRF depth analysis applied in Cultural Heritage of Japanese lacquerware ‘Tamamushi-nuri’ was published by Nakano and Tsuji [7]. In 2012, the first CXRF study of Renaissance paintings was published by Reiche et al. [11]. In this study, the CXRF depth profiles obtained on the paintings were related to lateral CXRF scans on crosssections. The cross sections were prepared from samples taken next to the analyzed points, in order to better evaluate the performance of CXRF depth profiling for the analyses of paint layers. The cross-sections were also studied by optical microscope (OM) and SEM-EDX in order to compare the results to those obtained by CXRF [11]. Beside the analysis of paint layers, the study of other objects of Cultural Heritage was also 33 performed. For example, Mantouvalou et al. [49] studied Medieval reverse paintings on glass and Dead Sea scrolls [74]. Moreover, CXRF was also used in other interesting fields: - analysis of biological samples as seeds or little fish [58, 69, 75, 76], - studies of solid/liquid interfaces [76], - studies of the elemental distribution in pharmaceutical tablets [77, 78], - density measurements in ultra-low density materials [79]. Some other applications are also possible, such as the 3D elemental imaging of microSD card [80] or the depth elemental imaging of multi-layered automotive paint fragments and microchips [69]. 1.2.3. Micro-focus X-ray tube Due to the use of X-ray optics for focusing the beam, CXRF needs a micro-focus X- ray tube as an excitation source. Indeed, those micro-focus low-power X-ray tubes (with a typical spot size of about 50×50 μm2) make the combination with polycapillary lenses possible. The first micro-focus X-ray tube was developed for X-ray radiography in 1953 [81]. The first micro-focus X-ray tube with an electromagnetic focalization system was described by Arndt et al. in 1998 [82]. Those tubes may also be employed for XRF. One of the most common uses is for the so-called macro-XRF technique. Using a small spot size and high intensity beam, this technique consists in obtaining the cartography of paintings, for example, with an exposure time per point that is extremely short (10 seconds) which allows the scanning of a large area with a small step (1 mm) [10, 83]. From the X-ray optics point of view the brilliance is a much more important parameter than total photon flux. The brilliance is a measure of quality of the X-ray source [84, 85]. Brilliance is related to the concentration of photons and takes four factors into account: - Number of photons produced per second (NP), - The angular divergence of the photons (θP), or how fast the beam spreads out, - The cross-sectional area of the beam (AP), - The photons falling within a bandwidth (BW) of 0.1% of the central wavelength or frequency. The resulting formula is: )(*001,0*)(*)( ).( ).mrad.mms( 22 1 2-2-1- photonsBWmmAmrad sphotonsN Brilliance PP P    (6) The greater the brilliance, the more photons are concentrated on a spot. The increase of brilliance can be achieved by increasing the total power of the X-ray tube or by decreasing the size of the focal spot. The reduction of the focal spot size can be achieved by the focusing of the electron beam by magnetic lenses. Increasing the total power of the X-ray tube enhances the heat load on the anode which is limited by thermal conductivity and cooling efficiency. Consequently, the higher power load reduces the durability of the tube. In a typical micro-focus side-window X-ray tube with total power of 50 W, the focal spot with an area of 100 µm2 can be produced. In this case, the anode power load can 34 reach values greater than 1 kW/mm2 meaning that the brilliance can be higher despite the low total power of the tube. For this reason, low power X-ray tubes have become very popular. Moreover, low power tubes are more compact than high power tubes as the size of the cooling system as well as the high voltage power supply are smaller. Their small size makes them particularly suitable for transportable setup. The design of a micro-focus side-window X-ray tube with electromagnetic focalization is illustrated in Fig. 13. X-ray photons are generated by electron irradiation of a target producing the X-ray emission by bremsstrahlung radiation and X-ray fluorescence of the anode. The X-ray tube generates an X-ray beam by a process involving an electron beam produced by an electron gun, accelerated by electrical potential, and electromagnetically focused before colliding with a target. The electron gun consists of a Wehnelt electrode and an indirectly heated activated dispenser cathode. The advantage of the dispenser cathode is that it is mechanically stable and, due to the lower power consumption and operating temperature, it has a greater lifetime than heated filament cathodes. The electrons are accelerated from the cathode, held at a high negative potential, towards the grounded anode. They pass through a hole in the anode before being electromagnetically focused [86]. Micro-focus low-power X-ray tube with high brilliance coupled to a polycapillary lens allows to obtain a reduced beam focal spot size with high intensity. For this reason, both elements are essential for CXRF. Cathode Wehnelt grid electrode Anode assembly Electromagnetic axial focusing lens Electromagnetic quadrupole lens Exit of X-ray beam X-ray shutter Anode target Fig. 13. Schematic drawing of an electromagnetically focused micro-focus X-ray tube [86]. 35 1.2.4. Polycapillary lenses and polycapillary collimators The main goal of focusing optics is to capture the radiation emitted in a large solid angle and concentrate it in a very small spot. Polycapillary lens is one kind of X-ray focusing device. It consists of large amounts of hollow glass channels (from several thousands to hundreds of thousands, each working as a waveguide) (Fig. 14). It collects photons in a large solid angle transporting them by multiple reflections inside every channel and focusing the beam in a small spot with a higher density radiation. Bent capillaries transport the radiation and also change the direction of the beam. The radiation emitted from the source point is transported by capillaries which have openings that are oriented towards the source. In the case of polycapillary full-lens, the arrangement of channels changes from divergent into convergent, so that the radiation can be refocused into a focal spot as shown in Fig. 15. Another group of focusing devices is called polycapillary collimators. They can be made of bent (polycapillary half-lenses) or straight channels (polycapillary conical collimators or poly-CCC). The main role of polycapillary half-lenses is to make the divergent radiation beam parallel whereas poly-CCC is able to guide and collimate divergent radiation (see Fig. 15 b and c). They are used in front of the X-ray detector to complete the CXRF configuration. Fig. 14. Microchannel makeup of polycapillary optics [87]. Fig. 15. Schemes of polycapillary full-lens (a), polycapillary half-lens (b) and polycapillary conical collimator (poly-CCC) (c) [13]. 36 There are several important parameters that describe the performance of the X-ray optics: - The size of the output beam or focal spot size S defines the achievable spatial resolution of the spectrometer. The X-ray beam formed by the focusing device reveals uneven distribution of intensity across the beam. Usually it can be approximated by a Gaussian function. In this case the size of the beam is defined by the Full Width at Half Maximum (FWHM) of the Gaussian. - The gain is a useful parameter which is used to describe the X-ray lens. It is a dimensionless factor that defines the increase of beam intensity in comparison with a pinhole collimator with an internal diameter equivalent to this of the polycapillary lens. This parameter is defined as the number of X-ray photons by unit of surface, by unit of time (Appendix II and VI). - The working distance or output focal distance f defines the distance between the optics and the focal spot. For short working distances the sample has to be put very close to the optics. This parameter is crucial for depth profiling of a sample – for short working distances the maximum accessible probing depth and the security of the measurement are reduced. In the Cultural Heritage study, the selection of the working distance is critical. 1.2.5. X-ray optic principals Total External Reflection (TER) occurs when X-ray radiation strikes the internal surface of a capillary at an angle which is smaller than the critical angle (θc) (Fig. 16, a). Because it is a very small number (10-5-10-6) in the X-ray energy range, the angle of total reflection is very small (several mrad). The X-ray photons are then transported by reflection along the length of the tube without any losses. Meanwhile, those X-rays that do not fulfill this condition are absorbed into the walls of the capillary (Fig. 16, b). θc is a function of the density of reflective material and the energy of the incident radiation. For borosilicate glass, θc is determined by the X-ray energy through the formula [88]: )( 30 )( keVE mrad c  (7) The size of the output beam S can be roughly estimated by the following equation [88]: )()()(2)( mdmradmmfmS outc   (8) where f is the working distance (or output focal distance) and dout is the channel diameter of the capillary at the output end of the optic, which is very small compared with S and can be neglected. Combining the equations 7 and 8 and neglecting dout gives: )( )(60 )( keVE mmf mS   (9) 37 Because the critical angle (θc) changes with energy (eq. 7), for lower energies (higher θc), more photons can be accepted and thus emitted. For this reason capillary works as a low-pass filter. According to equation 9, the focal spot size S is energy dependent. Consequently, for a determined working distance f, S is smaller for higher energies. This property will explain the variation of the depth resolution with X-ray energy (cf. Chapter 1.2.6). The energy dependence of transmission of X-rays through the polycapillary with bent channels results connected to the curvature. The curvature radius R is related to the critical angle (θc) as followed [89]: 2 22   R d c (10) where d is the channel diameter. θc Capillary surface b a Fig. 16. Conditions of Total External Reflection. X-rays with incidence angles θ smaller than the critical angle (θc) are reflected (a). On the contrary, X-rays are absorbed (b). The great advantages of these devices are the simple manufacturing process (low cost) and broad band transmittance. Because of its compact size, focalizing properties and ability to produce high-flux of polychromatic beams, they have become very popular especially for polychromatic X-ray tube purposes. The greatest advantage of polycapillaries is the acceptance of a large solid angle of emitted radiation. 1.2.6. Depth resolution and depth profile As previously explained, a probing micro-volume is defined by the overlap of the foci of both polycapillary lenses (Fig. 17). The characterization of the CXRF setup is achieved by defining the depth resolution [49, 57, 70, 90]. The depth resolution is represented as a function of the X-ray energy. The FWHM of several depth profiles of elemental thin films are represented versus the respective X-ray energy. It is obtained measuring depth profile of a series of thin foils. The foils had to be thinner than the depth resolution in order to prevent the absorption effect (cf. Chapter 1.2.7). Then, the FWHM of the depth profile of the Kα or Lα radiation of those elements (in µm) is represented versus the energy of emission of the chemical element composing the foil (in keV) (Fig. 18). For every in-depth scan measured, the estimated error applied to the FWHM values is half of step size. The depth resolution is energy dependent, being smaller for higher energies (Fig. 18). This means that, in multilayered samples, the depth resolution is slightly different depending of the chemical element excited by the X-rays. 44 Finally, certain pigments also have specific fluorescence signatures, such as zinc white (which has been used in ground preparations since the 19th century) which appears yellow. Certain pigments and binders can fluoresce with very precise colors under UV, however this identification technique must be used very carefully as many factors can influence the fluorescence color observed, such as impurities that can may even cause fluorescence to diminish or disappear [100]. In conclusion, based on UV-induced visible fluorescence photography, a specific interest can be done on some fluorescence observed areas during located analysis (XRF, CXRF, etc.) in order to define their chemical specificities. Fig. 22. Photography of a detail of “San Pedro Nolasco despidiéndose de Jaime I el Conquistador” (a) and the same detail in UV-induced visible fluorescence photography (b). 1.4. Infrared reflectography (IRR) Infrared (IR) radiation was first detected by astronomer Sir William Herschel in 1800. He discovered that many molecules have both vibrational and rotational resonances in the infrared and, as a result, are good absorbers. Therefore, IR is often mistakenly referred to as heat waves. Since the J.R.J. Van Asperen de Boer thesis, infrared reflectography (IRR) has become a classic image registration method in the technical examination of art [5]. This technique reveals the presence of preparatory drawings made of carbon black (infrared no-reflective material) and the artist’s process of execution and changes in the composition (pentimenti) as well. It also provides information that helps to deduce the nature of some pigments present on the artwork’s surface [6, 101]. Indeed, images under IR radiation have been employed, for many years, during the material study of artwork, mainly in the restoration field, but also as an important tool from the history of art perspective. The IRR technique permits users to visualize a painting by registering its response in the infrared spectrum, invisible to the human eye. It has the advantage of being both nondestructive and noninvasive. Infrared radiation extends from the nominal red edge of the visible spectrum at 700 nm to 1 mm. At present, different IRR cameras are available with detectors which are able to record a specific part of the IR spectra (Fig. 23). The reflection intensity of the IR light by a painting registered by an IRR camera depends not only on the wavelength range of the IR radiation recorded by the sensor camera but also on the thickness of the painting layer as well as the kind of materials present and its quantities. (a) (b) 45 During the present studies, the excitation of the paintings were achieved with two halogen light reflectors (800W with intensity regulation control) placed on either side of the artwork. To achieve the most homogeneous illumination, in the case of Seville, two tripods were placed at a distance of 1 m from each side of the 2D motion device (cf. Chapter 2.3), both at an angle of 45º with respect to the surface of the painting (Fig. 24). Those conditions were applied for all artworks studied in order to homogenize the results obtained. Gamma Rayos X Rayos UV Visible Infrarrojo Microondas Infrarrojo cercano Infrarrojo medio Infrarrojo lejano 10-8 200 400 760 106 Longitud de onda (nm) 760 1500 5600 106 Cámara InGaAs XenICs 900 1700 Cámara Vidicón Hamamatsu 760 1800 Cámara Phase One Fig. 23. Specific part of the IR spectra registered by each IRR camera. Fig. 24. Position of the tungsten lights in order to obtain the most homogeneous illumination. 46 During this work, three different technologies of IRR cameras were used: - two InGaAs cameras, the Xeva-XS-512 model in Seville and the more recent Xeva-1.7- 640 model in Mexico (both from Xenics) (Fig. 25), - a Si-CCD camera (model Power Phase from Phase One) (Fig. 26), - an infrared vidicon camera (model C2741-03 with a 25 mm tube diameter (model N2606-06) from Hamamatsu Photonics K.K.) (Fig. 26). The difference between both Xenics cameras is the resolution. The Xeva-XS-512 has a resolution of 320×256 pixels while the resolution of the Xeva-1.7-640 is 640×512 pixels. To define the IR range of each camera, the sensor response division scheme has been chosen [102]. This scheme divides the band based on the response of various detectors. In this case, near-infrared ranges from 0.7 to 1.0 µm (0.7 µm being the approximate end of the response from the human eye to 1.0 µm being that of silicon) while short-wave infrared is defined as 1.0 to 3 µm (1.0 µm being cut-off of silicon to 3 µm being cut-off of the MWIR atmospheric window). The last division belongs to InGaAs, which covers up to 1.8 µm (Table 2). Now, the different technologies of IRR cameras are presented and explained. Fig. 25. The Xeva-XS-512 model in Seville (a) and the Xeva-1.7-640 model in Mexico (b). Fig. 26. Si-CCD camera (model Power Phase) (a) and infrared vidicon camera (model C2741-03) (b), both in Mexico. (a) (b) (a) (b) 47 Table 2. Classification of the different technologies of IRR cameras based on the sensor response division scheme. Technology type Responsivity range (nm) Sensor response division scheme Si-CCD 300-1000 near-infrared Vidicon 400-2200 near-infrared and short-wave infrared InGaAs 900-1700 short-wave infrared 1.4.1. Si-CCD detector The charge-coupled device (CCD) was invented in 1969 at AT&T Bell Labs by Willard Boyle and George E. Smith [103] who were awarded the Nobel Prize in Physics in 2009 [104]. In a CCD for capturing images, there is a photoactive region and a transmission region. Fig. 27 illustrates the electrode structure defining an individual CCD sense element. Image generation can be divided into four stages: charge generation through photon interaction with the device's photosensitive region, collection and storage of the liberated charge, charge transfer, and finally charge measurement. When an image is projected through a lens onto the two-dimensional capacitor array (the photoactive region), each capacitor accumulates an electric charge proportional to that of the light intensity present in that location. Once the array has been exposed to the image, a control circuit causes each capacitor to transfer its contents to its neighbor (operating as a shift register). The last capacitor in the array dumps its charge into a charge amplifier, which converts the charge into voltage. By repeating this process, the controlling circuit converts the entire contents of the array in the semiconductor to a sequence of voltages. In a digital device, these voltages are then sampled, digitized, and usually stored in memory. During readout, the collected charge is subsequently shifted along the transfer channels under the influence of voltages applied to the gate structure. The bucket brigade for rainfall measurement analogy aids in visualizing the concept of serial readout of a CCD. It is very well illustrated (Fig. 28) by K.R. Spring, T.J. Fellers and M.W. Davidson [105]. Fig. 27. Electrode structure defining an individual CCD sense element [106]. 48 Fig. 28. The bucket brigade for rainfall measurement analogy aids in visualizing the concept of serial readout of a CCD [106]. Fig. 29. Image capture by a Si-CCD camera with an IR cut-off filter (a) and without IR cutoff filter (b). The last image is redder than the normal one. (a) (b) 49 In the case of the CCD Power Phase camera, this technology works with scanning back. A digital scan back uses one array per RGB component. This translates to a triple sensor matrix layered vertically in a holder. The main advantage of using this technology to analyze artwork is the extremely high quality images obtained. This translates to a tremendously accurate color reproduction, as every pixel is measured individually for red, green and then blue, and allows to print in very large sizes without loss of detail. The downside of capturing images using a Power Phase camera is the amount of time needed to achieve a scan (from several seconds to minutes). A second inconvenience is the necessity to work connected to a computer due to the huge file sizes (hundreds of megabytes or even gigabytes) each scan produces. During this study, the Power Phase camera used had a sensor size of 7000×7000 pixels with a resolution of about 36 megapixels, 12 bit. It consisted of a trilinear CCD camera which was able to register a wide spectrum from ultraviolet to near-infrared including visible light (from 300 to 1000 nm). Consequently, in order to obtain visible photography, specific cut-off filters had to be used which prevented IR and UV radiations from reaching the CCD and thus appearing in the captured image. In the case of IRR image captures (cf. Chapter 3.2), it is important to note that a UV cut-off filter but not an IR cut-off filter were used allowing IR radiation to reach the CCD. The image captured in this way was redder than a normal one (Fig. 29). Nevertheless, no visible cut-off filter was used during this study meaning that both visible spectrum radiations and nearinfrared radiations are registered on the image. This imaging method will be referred as visible-IR imaging. Generally, in IRR, the images are represented using a grey scale established depending of the wavelength range measured by the detector. So, after capturing the image, the computing process consisted in a conversion of the visible-IR image in a grey scale to make it appears IRR image-like. Nevertheless, registering only the near-infrared radiations might be achieved using a visible cut-off filter to prevent the visible spectrum radiations from reaching the CCD. 1.4.2. InGaAs detector InGaAs photodiodes were invented in 1977 by Pearsall [107]. Indium gallium arsenide (InGaAs) is a ternary alloy composed of indium, gallium and arsenic. It is used as a semiconductor with applications in electronics and optoelectronics which provide excellent photosensitivity in the IR. Among them, its principal application is as an infrared detector. The spectral response of an InGaAs photodiode is shown in Fig. 30. Infrared InGaAs detectors offer a specific spectral range which varies depending of the fraction of indium in the ternary compound. The second important innovation was the engineering of InGaAs focal-plane arrays (FPAs) present in InGaAs cameras. An InGaAs FPA consists of a two-dimensional photodiode array (PDA) (itself comprising an indium phosphide (InP) substrate, an InGaAs absorption layer, and an ultrathin InP cap) that has been indium bump bonded to a readout integrated circuit (ROIC) (Fig. 31). As in a CCD, in an InGaAs FPA, the twodimensional array detects incident light and then generates and collects charge. The ROIC clocks and converts the collected charge to voltage and then transfers the resulting signal to off-chip electronics. 50 Fig. 30. The spectral response of an InGaAs photodiode [108]. InGaAs absorption layer Ultrathin InP cap Indium bump bonding Connection to camera preamplifier Readout integrated circuit InP substrate Antireflective coating Fig. 31. An InGaAs FPA is a two-dimensional photodiode array (InP substrate, InGaAs absorption layer, ultrathin InP cap) indium bump bonded to a readout integrated circuit [109]. While Si-CCD cameras cannot offer sensitivity beyond 1100 nm (Fig. 32), InGaAs cameras provide sensitivity over short-wave infrared range due to its lower bandgap. Nevertheless, this bandgap is also responsible for much higher dark current (thermally generated signals) compared to those produced by Si-CCDs. As a result, scientific InGaAs FPA cameras require deep cooling (i.e. down to -90°C) in order to minimize this unwanted noise source. 51 Si-CCD InGaAs Fig. 32. InGaAs FPAs provide better QE above 900 nm than Si-CCD technology [109]. Due to a very complex manufacturing process, almost every InGaAs FPA has a small percentage of defective pixels classified as non-uniform, dark, and bright. A non-uniform pixel demonstrates greater or lesser photosensitive response than the surrounding pixels. Its response, though linear, differs from the average response demonstrated by the rest of the device’s array. A dark pixel, meanwhile, is unresponsive to light and appears black in an image. Finally, a bright pixel is a pixel that saturates independently of incident light and appears as a white spot in an image (Fig. 33). While pixel defects are inevitable when dealing with InGaAs FPA devices, it is necessary to apply a defect-correction algorithm built into the camera’s acquisition software or hardware. Fig. 33. The occurrence of bright pixels with the Xeva-XS-512 model in Seville. 52 Finally, the reflectograms measured by InGaAs cameras are small images and the equipment must be placed near the painting in order to give good results. Depending on the size of the artwork in question, the result can be tens to thousands of images that must then be integrated into a single image. The complexity of obtaining complete IR images of large format easel paintings increases proportionally with the number of reflectograms needed. Precise control of lighting is also necessary to ensure an accurate capture of the painting under the same conditions throughout the reflectrographic study [110]. In this perspective, two different capture methods were used: a 2D motion device and a rotating motion device (cf. Chapter 2.4). 1.4.3. Infrared Vidicon detector A Vidicon tube is a video camera tube design in which the target material is a photoconductor (Fig. 34). The Vidicon was developed in the 1950s at RCA by P. K. Weimer, S. V. Forgue and R. R. Goodrich. The infrared Vidicon is composed of two main components [111]. Firstly, there is a photoconductive target - an electrode that stores a pattern of charges corresponding to the input image. It is the PbO-PbS pyroelectric photoconductive target that is sensitive to a broad range of wavelengths (400 to 2200 nm) (Fig. 35). While PbS is sensitive to infrared, PbO provides high sensitivity in the visible region. This precise combination of two substances gives the tube its spectral response characteristics. Usually, the IR-Vidicon is used for infrared only, with a visible cut-off filter. Secondly, there is the electrode section made of an electron gun, an accelerating electrode, two focusing and deflecting electrodes and a field mesh (Fig. 34). The electron gun directs the thermionic electrons from the cathode into a tight beam, which is further narrowed by passage through a small aperture. The electron beam has the high current density needed to read the stored image efficiently. The various electrodes and coils accelerate the beam, focus and deflect it onto the target. The coils aim at changing the trajectory of the electron beam through the electron gun in order to scan the entire surface of the target. The optimum angle at which the beam should strike the target is 90º, so a collimation lens is provided between the focusing electrode and field mesh. When scanned by the electron beam, the photoconductive target generates a signal current. The fluctuating voltage coupled with a video amplifier is used to record the input image. The electrical charge produced by an image will remain in the faceplate until it is scanned or until the charge dissipates. 53 FOCUSING COIL DEFLECTION COIL ALIGNMENT COIL SIGNAL ELECTRODE LENS LIGHT FACEPLATE TRANSPARENT CONDUCTIVE FILM PHOTOCONDUCTIVE TARGET BASE HEATER CATHODE (K) ACCELERATING ELECTRODE (G2) BEAM CONTROL ELECTRODE (G1) BEAM FOCUSING ELECTRODE (G3) FIELD MESH (G4) Fig. 34. Scheme of an infrared Vidicon detector. 400 600 800 1000 1200 1400 1600 1800 2000 2200 WAVELENGTH (nm) 1000 100 10 1 0 SENSITIVITY (nA/μW) N2606-06 N2606 N2606-01 N2606-02 Fig. 35. Typical spectral response of infrared Vidicon detector model N2606-06 [111]. 60 In total, five lateral scans were performed. Each scan was composed of 30 spectra taken with 10 μm step size and a counting time of 5 s per point. The evolution of the lateral diameter size of the beam versus Z distance is presented in Fig. 42. The smallest FWHM measured were 49 ± 5 μm. The lateral beam profile in the focal spot position is presented in Fig. 43 as well as its derivative. 0100 200 300 400 500 45 50 55 60 65 70 75 FWHM Fe K Gauss of FWHM FWHM (m) Position Z (m) Fig. 42. Evolution of the lateral diameter size of the beam versus Z distance. 050 100 150 200 250 300 0.0 0.2 0.4 0.6 0.8 1.0 Area of Fe K Derivative of Fe K lateral profile Gauss of derivative Relative intensity (a.u.) Lateral position (m) Fig. 43. Lateral beam profile in the focal spot position and derivative of the profile. 61 Once the focal distance is found, the position is pinpointed with the two laser diodes. Then, the impact point of the X-ray beam is visualized using a scintillator in a dark room environment before pinpointing the laser diodes to the exact location. Thereafter, the CMOS camera coupled to its zoom imaging lens is adjusted according to its working distance (35 mm ± 1 mm) using the bronze standard for observation. Then, the half lens is fastened to the SDD (Fig. 37) and the optimal position of the half lens had to be found in order to finish the installation of the CXRF set-up. To align the confocal geometry, a series of measures positioning the bronze standard in the focus of the primary X-ray beam were done modifying the position of the detector-half lens looking at the maximum count rate. At the maximum count rate corresponds the optimal position of the optics, and thus the smaller depth resolution. Once the confocal geometry is established, the characterization of the CXRF set-up is achieved. Then, the lateral resolution of the confocal volume has to be determined using the same “1_Working_distance_Adjustment” LabVIEW routine. It scans horizontally over a copper grid (model T200-CU, from Electron Microscopy Sciences). It is a grid with 125 μm pitch, 113 μm hole and 12 μm bar (Fig. 44). A lateral profile is performed measuring the FWHM from the Cu Kα at the focal distance (Fig. 45). The scan was composed by 79 spectra taken with 5 μm step size and a counting time of 120 s per point. The results in Table 4 show a lateral resolution about approximately 34 ± 3 µm. Fig. 44. Scheme of a copper grid. 0100 200 300 400 0 5000 10000 15000 20000 Area Cu K Intensity (counts) Distance (m) 1 2 3 Fig. 45. Lateral profile of a copper grid in the focal spot position. 62 Table 4. Results of lateral profile of a copper grid in the focal spot position. Bar FWHM (μm) Xc (μm) Distance between bars (μm) 1 34 ± 3 69 ± 3 124 ± 3 2 34 ± 3 193 ± 3 122 ± 3 3 34 ± 3 315 ± 3 Consequently, depth profile of a series of thin foils (1 m thick) of different chemical elements (Ti, Cr, Fe, Ni, Cu, Zn, Zr, Ag, Au and Pb) (from Goodfellow) is measured to define the depth resolution of the CXRF set-up (Fig. 46). From now on, the depth profile of the 1 µm Cu thin foil will be used as a reference so as to verify the calibration of the set-up every time. To achieve this experiment, “2_Depth_profile_Acquisition” LabVIEW routine is used (Fig. 47). It performs spectra acquisition of the depth profile controlling the OWIS motion. The program input consists of the name of the folder to save the spectrum files, the number of analyzed points, the pitch distance and the measurement time. It starts recording the spectrum at the initial position so that it performs a loop including two steps, changing the Z position and then recording the spectrum. The difference with “1_Working_distance_Adjustment” routine resides in the control of OWIS motor instead of Standa motor. Each scan was composed by 30 spectra taken with 10 μm step size and a counting time of 120 s per point. The evolution of the depth resolution versus secondary fluorescence energy is presented in Fig. 46 (repetition of Fig. 18). 2 4 6 8 10 12 14 16 18 0 10 20 30 40 50 60 70 80 FWHM (K or L) (m) Energy (keV) XRF CONCHA Fig. 46. Depth resolution of the µXRF-Concha set-up as a function of the emitted energy (repetition of Fig. 18). 63 Fig. 47. Front panel of “2_Depth_profile_Acquisition” LabVIEW routine. Besides, using the same “2_Depth_profile_Acquisition” LabVIEW routine, a series of Cu thin foils (from Goodfellow) of different thicknesses (1, 5, 10, 15, 20, 25, 30, 34, 40, 45, 50, 75 and 100 µm) were analyzed to determine the capability of the CXRF technique in measuring the thickness of copper foils and also to illustrate the absorption effect (cf. Chapter 1.2.7). Each scan was composed by 30 spectra taken with 10 μm step size and a counting time of 120 s per point. The evolution of the height of the Gaussian curve versus thickness of foils is presented in Fig. 48. The results show that it may be possible to determine the thickness of a copper single layer thinner than 20 µm. Beyond 20 μm, the self-absorption effect in a copper foil prevents the secondary X-ray fluorescence from escaping the sample and reaching the detector. The total counts registered by the detector reach a limit value that remains constant for all copper foils thicker than 20 μm. Consequently, the information about thickness is lost. This phenomenon can be modeled in the case of this kind of simple sample and the equation giving the relation between Gaussian height and thickness can be achieved easily. Nevertheless, for more complex unknown samples, such as cultural heritage samples, thickness estimation is a challenge. In this way, simulations have been performed by Pawel Wrobel, in a collaboration study, on two experimental paint multilayers made during this PhD thesis (about experimental samples, see Chapter 3.1) [117]. The results show that, in order to simulate the thickness of each layer, density and composition must be known. So, the ability to measure the thickness of a multilayered sample based only on CXRF results without any previous knowledge of the density and composition of each layer remains impossible. Another focus can be proposed on the measurement of thickness layers. Nondestructive techniques such as conventional transmission computer microtomography (µ-CT) [118], micro-beam X-ray fluorescence computed tomography (µ- XRFCT) [119] or in situ nuclear magnetic resonance (NMR) [120-122] can be considered in order to offer complementary information to CXRF study. 64 020 40 60 80 100 50000 100000 150000 200000 250000 Gaussian height of Cu K Height (count) Thickness (m) 050 100 150 200 250 0 1x1011 2x1011 3x1011 4x1011 5x1011 6x1011 7x1011 Area Cu K Gauss of counts Area Cu K (counts) Distance (m) Gaussian Heigh Fig. 48. Evolution of the height of the Gaussian curve versus the thickness of Cu foils (error included in the points). 2.1.3. Data treatment Each depth profile produces a several tens of spectra that have to be analyzed. All CXRF data were processed with the free software PyMca. It was developed by the "Software Group of the European Synchrotron Radiation Facility (ESRF)" and distributed under GNU General Public License [123]. The PyMca Toolkit is a collection of tools for visualization and analysis of energy-dispersive X-ray fluorescence data. The version used here was "PyMca Application Version 4.4.1". This software proposes a lot of advantages. It allows automatic analysis of a group of spectra and permits a visualization of the sum of spectra which allows for better and faster recognition of all the chemical elements presents. Each tool, like the data analysis routines, can be used independently. Nonetheless, its manipulation is uneasy therefore the methodology used to make the complete treatment of the data is explained in Appendix V. This appendix presents the use of PyMca software and its sub-programs to analyze depth or lateral profiles scans (1D) respectively obtained with all CXRF setups used during this investigation. 2.2. LouX3D system and CXRF set-up at AGH University of Science and Technology Intercomparison of three confocal micro X-ray fluorescence systems will be made in Chapter 3.3.3 for the non-destructive characterization of experimental paint layers. In this 65 third chapter the results obtained using three CXRF systems are presented: the LouX3D at the C2RMF laboratory in Paris [11], the CXRF set-up at AGH University of Science and Technology in Krakow [59] and, the µXRF-CONCHA previously presented. The CXRF set-up at AGH University of Science and Technology in Krakow was designed and characterized by Pawel Wrobel in his PhD thesis. The LouX3D set-up was designed and firstly characterized by Ina Reiche and Katarina Müller. As part of the intercomparison study, the LouX3D was newly characterized using the same thin foil standards and the same methodology applied to the µXRF-CONCHA system in Seville (cf. Chapter 2.1.2). Hereafter, the main features of this device and the results obtained are presented. The LouX3D system consists of a micro-focus X-ray source (IMOXS, IfG) [112, 113] and a SDD Rontec X Flash (series 4, model 4010, Bruker AXS) (Fig. 49, 50 and Table 5). It is complete with a full lens at the exit of the X-ray source and a Poly-CCC lens at the entrance of the detector (both from IfG) (see specifics in Appendix VI). Détecteur Microscope Tube de rayons X Caméra Webcam SDD detector Microscope X-ray tube Fig. 49. General view of the LouX3D set-up. 66 Détecteur Policapillaire (full lens) du tube de rayons X Caméra Policapillaire (poly-CCC) du détecteur Webcam SDD detector Full-lens of X-ray tube Poly-CCC of SDD detector Fig. 50. Head analysis of the LouX3D device. Table 5. Characteristics of the LouX3D device. CXRF set-up LouX3D - Paris Excitation channel Detection channel X-ray tube anode Rh Manufacturer/model IfG/iMOXS MFR Manufacturer/ model of detector Bruker AXS/ SDD Rontec X Flash Maximum power (W) 30 Maximum voltage (kV) 50 Si crystal thickness (μm) 450 Maximum intensity (μA) 600 Detector active area (mm2) 10 Polycapillary lens Full lens Polycapillary lens Poly-CCC Focus spot size (μm) 29-45 Focus spot size (μm) 40 (Cr Kα and Fe Kα) Focal distance (mm) 11 ± 0.1 Focal distance (mm) 6.2 67 The X-ray source is similar to the one used with the µXRF-CONCHA. The only difference is that the X-ray tube works with a rhodium anode. It can be operated with maximum voltage and current of 50 kV and 600 µA respectively and provides a 50 µm2 beam spot at the entrance of the full lens which is directly fastened to the exit of the X-ray source. This beam is focused with the polycapillary X-ray full lens. The lens has a nominal working distance of 11 ± 0.1 mm and a nominal focal spot diameter of 29 µm FWHM at Rh-Kα energy (20.17 keV). Finally, the tube disposed of an integrated shutter module for switching on and off the X-ray output. The active area of the SDD Si crystal is 10 mm2 for an active thickness of 450 µm. The head of the SDD is interlocked to the Poly-CCC which has a nominal working distance of 6.2 mm and a nominal focal spot diameter of 40 µm FWHM at Fe-Kα energy (6.40 keV). The X-ray source and the X-ray detector are positioned at 45° with respect to the surface normal of the analyzed sample. A microscope (with a primary magnification from 2.5x to 10x) from Opto is placed in front of the sample allowing the observation of its surface. Finally, a webcam permits preview visualization of the sample and safety control during the study. All these components (X-ray tube, detector, microscope and webcam) are fastened to a motionless tabletop support made by IfG. In this device, the sample support in front of the CXRF head set-up is motorized, allowing the movement in the XYZ directions. The sample support is moved by three 100 mm mid-travel steel linear stage motors from Newport (from Newport Micro-Controle Klinger, model UTS100). These motors have a 200 N maximum horizontal load capacity (approximately 20 kg) and a 50 N maximum vertical load capacity (approximately 5 kg) for a uni-directional repeatability of 1 µm. These are the limitations, which have to be considered when studying artwork. X-ray tube, detector, microscope and the three motion motors are connected to a computer via USB. The device is controlled by a software called "Louvre0" developed by IfG [112]. The software version used is dated October 19, 2007. It allows single measurements, 1D, 2D and 3D analysis. The use of the "Louvre0" acquisition software is detailed in Appendix VII. The characterization of the LouX3D set-up was achieved by measuring the same series of thin foils (1 µm thick) of different chemical elements (Ti, Cr, Fe, Ni, Cu, Zn, Zr, Ag, Au and Pb) with the same condition as with the µXRF-CONCHA to define the depth resolution of the device (Fig. 51). Consequently, each scan was composed by 30 spectra taken with 10 μm step size and a counting time of 120 s per point. In the case of the data from the CXRF set-up at the AGH University of Science and Technology, the depth resolution was measured by Pawel Wrobel [13]. The characterization was obtained by measuring a different series of thin foils of different chemical elements (Table 6). Each scan was obtained with 50 kV, 1 mA and it is composed by spectra taken with 2.5 μm step size. The evolution of the depth resolution versus secondary fluorescence energy of the three CXRF setups is presented in Fig. 51. As mentioned before, all the CXRF data were processed with the PyMCA software. The methodology is slightly different to that used with the µXRF-CONCHA device and it will be explained with more detail in Appendix V. 68 Table 6. Series of thin foils used to measure the depth resolution in the case of the AGH CXRF device. Chemical element Thickness (μm) Counting time (s) Ti 6 1 Mn 15 1 Fe 5 1 Co 8 1 Ni 7.5 1 Cu 7 1 Zn ~0.025 1 Zr 66 2.5 Sn 250 3 Pb 120 3 Au ~0.025 30 2 4 6 8 10 12 14 16 18 0 10 20 30 40 50 60 70 80 FWHM (K or L) (m) Energy (keV) XRF CONCHA LouX3D AGH Fig. 51. Depth resolution of the three CXRF setups as a function of the emitted energy. 69 Finally, Table 7 summarizes the specifications of each CXRF set-up with respect to the intercomparison of the three devices presented later in Chapter 3.3.3. Table 7. Characteristics of the three CXRF setups. CXRF set-up μXRF-CONCHA - Seville LouX3D - Paris AGH - Krakow Excitation channel X-ray tube anode W Rh Mo Manufacturer/model IfG/iMOXS MFR IfG/iMOXS MFR XOS/Powerflux PF X-Beam Maximum power (W) 30 30 50 Maximum voltage (kV) 50 50 50 Maximum intensity (μA) 600 600 1000 Polycapillary lens Full lens Full lens Full lens Focus spot size (μm) 17-29 29-45 13 (Mo Kα) Focal distance (mm) 7.3 ± 0.1 11 ± 0.1 4.4 ± 0.3 Detection channel Manufacturer/model of detector Amptek/Super 123SDD Bruker AXS/SDD Rontec X Flash Ketek/VITUS-SDD (AXAS-V) Si crystal thickness (μm) 500 450 450 Resolution at 5.9 keV (eV) 138 NC 139 Active area (mm2) 25 10 10 Polycapillary lens Half-lens Poly-CCC Poly-CCC Focus spot size (μm) 19-36 40 (Cr Kα and Fe Kα) 28 (Fe Kα) Focal distance (mm) 6.5 ± 0.1 6.2 4.4 ± 0.3 76 Autopano Giga. Comparing the three final assemblies, Autopano Giga provides a clearer picture in order to deduce the nature of some pigments present on the artwork’s surface. The automatic brightness compensation results in a balanced image where the IR reflection contrasts can be interpreted more accurately. Fig. 59. Mosaic assembly result obtained with VARIM (a), Autopano Giga (b) and ICE (c). 2.4.4. Stitching methods for images obtained with the Gigapan system In this example, the “Virgen de las Uvas” attributed to Alonso Vazquez from the Metropolitan Cathedral of Mexico was studied. 30 images for the mosaic assembly were captured within a distance of 110 cm between the painting and the IRR camera. In the case of ICE software, the “rotating motion” mode was selected for the camera motion and 30% for feature matching search radius. The “Projection Perspective” option was selected in order to correct the deformation. In the following figures, the results obtained with each software without correction (Fig. 60) and with correction (Fig. 61) are shown. VARIM software (Fig. 60) had some difficulties in stitching the more distorted part of the picture (left and right side of the paintings) producing an artificial duplication of some parts of the mosaic. Although VARIM proposes several correction options, it was impossible to use them in order to obtain the mosaic assembly without the deformation effect. In this case, ICE and Autopano Giga produced the best results. This last one also has the advantage of allowing exposure compensation in areas with too much contrast. Consequently, it eases the observation of contrasting areas on the IRR image of the artwork to aid in the detection of certain pigments (Fig. 61). (a) (b) (c) 77 Fig. 60. Mosaic assembly result obtained with VARIM (a) and ICE (b). Fig. 61. Mosaic assembly result obtained with ICE with projection perspective (a) and with Autopano Giga with correction effect (b). (a) (b) (a) (b) 78 2.4.5. Comparison of the efficiency of the software results Consequently, the efficiency of the software in the final assembly of IRR images was evaluated, comparing them by means of different variables (like time of acquisition, digital correction features, etc.), and creating a ranking based on a numerical scale (Table 8). Table 8. Efficiency comparison and rating. Software VARI M Autopano Giga ICE Total time for 2D motion assembly (16 images) 7 min 2 min 2 min Assembly speed (1 = slow, 3 = fast) 1 3 2 Total time for rotation motion assembly (30 images) 16 min 3 min 8 min Assembly speed (1 = slow, 3 = fast) 1 3 2 Total time for rotation motion assembly with correction (30 images) X 5 min 13 min Assembly speed (1 = slow, 3 = fast) 1 3 2 Facility of use (1 = difficult, 3 = easy) 2 1 3 Possibility to work with the original PNG image format without necessity of conversion (1 = no, 3 = yes) 3 1 3 Necessity to modify mosaic assemblies with Photoshop (1 = necessary, 2 = eventually for publication, 3 = not necessary) 2 3 2 Photos exposure compensation (1 = hazardous changes, 2 = no change, 3 = controlled modification) 2 3 1 TOTAL (average of 10) 5.7 8.1 7.1 The ICE program is the best in terms of quality/time. It provides enough information for the preliminary interpretation of the technical aspects of the artwork. It is especially suitable in order to control the quality acquisition process during field work, where time is often limited. Autopano Giga software provides the highest quality mosaics, but needs previous conversion of PNG photos using other commercial software, such as Adobe Photoshop. It would be suitable for the production of higher quality material for publication. VARIM provides high quality results, nonetheless, its manual mode performance is very time-consuming. It is a very efficient freeware for photos taken without regard to distance, as a case in point, when the camera is used with a basic tripod. Finally, Microsoft ICE and Autopano Giga are very efficient for the correction of angular deformations. This deformation is the consequence of the rotated motion used by Gigapan system while acquiring images. 2.5. XRF setups used during Alonso Vazquez Project The archaeometry research group of the Centro Nacional de Aceleradores at the University of Sevilla (CNA-US) has developed wide expertise in in situ cultural heritage analysis by the use of XRF portable systems. This technique has been used to study a wide variety of artefacts and museum objects as well as artworks of several kinds: statues 79 [126, 127], easel paintings [128, 129], coins [130], bronzes [131], gold jewelry [132], glassware [133], etc. During the Alonso Vazquez project, the XRF study of the different artworks in Seville and Mexico were managed through the use of three different XRF devices. In this section, the X-Panda set-up will be detailed. The two others setups will be also briefly described: an old one designed as the “Museum XRF device” from the CNA and the SANDRA system from Mexico. 2.5.1. Description of the X-Panda set-up During this PhD thesis, a new portable XRF system adaptable to the 2D motion system from the Fine Arts Museum of Seville has been designed for in situ study of paintings. X- Panda is a Spanish acronym for "Equipo de rayos X Portátil de Análisis No Destructivo para el Arte" meaning “Portable XRF set-up for the non-destructive analysis of art” (Fig. 62). The set-up consists of an X-ray source (Mini X model, Amptek) and a SDD (model X123SDD, Amptek) (more specifics respectively in Appendix VIII and IV) [33]. It is completed with two laser diode modules and a webcam for the positioning and inspection of the surface sample. SDD detector X-ray tube Equipo X-Panda Webcam Fig. 62. General view of the X-Panda device. 80 Fig. 63. Isopower curve of the Mini-X 4W X-ray tube (From Amptek, 2011). The X-ray source is composed by a low-power 4W X-ray tube with tungsten anode, a 500 μm Be window and an internal power supply unit. It can be operated with maximum voltage and current of 40 kV and 200 µA (Fig. 63). Its maximum operating temperature is 50°C. To avoid reaching this temperature, it is possible to use two small fans placed around the Mini X run by USB. The X-ray tube is controlled via USB with “Mini-X Controller Software” (version 1.0.0.8, copyright © Amptek Inc.). This software allows the turning on and off of the X-ray tube and setting the working voltage and the current passing through the filament. The diameter of the primary X-ray beam is reduced using an in-house designed tungsten collimator made from a tungsten rod from Goodfellow and processed by the Mecanica de Precision company [134] with a 2 mm diameter hole. A 1 mm thick aluminum filter is placed at the exit of the collimator, fixed by a grey plastic piece in order to reduce the low energy of the continuous bremsstrahlung radiation and almost entirely remove the tungsten L peaks from the anode. If necessary, another plastic piece can be used fixing a 1 mm thick aluminum filter together with a thick tungsten collimator (with a 1 mm diameter hole) providing a smaller incident X-ray beam. The latter has never been used during this study. Two laser diode modules (Micro LDM Laser Diode Modules model, Edmund Optics) pinpoint the working distance from the set-up to the sample (Fig. 64). This working distance has been fixed to 10 ± 0.1 mm. Consequently, the surface of the irradiated sample is a disc of 2.9 ± 0.3 mm diameter and the beam aperture from the exit of the collimator has been estimated to be 2,5 ± 0.5 ° [135]. The active area of the SDD Si crystal is 7 mm2 with an active thickness of 450 µm. It is sealed with a 25,4 µm Be window. The resolution of the detector measured is 148 eV for the Mn-K line (FWHM of the peak at 5.9 keV). More details on the characteristics of the X-123SDD model detector are provided in Appendix IV. To complete the description of the device, the geometry of irradiation-detection channels has to be described. Like the other two XRF devices, irradiation is normal to the surface of the analyzed sample and the detection angle is 45° (Fig. 64). This arrangement permits the study of any kind of material with greater penetration, making it less sensitive to the heterogeneity of the surface but more sensitive to the matrix effects. 81 Finally, a CMOS color camera from Woxter (model i-Cam 60) allows the saving of images from the position of each measurement, controlling the safety of the study, observing the irradiated region of the sample and selecting the region of interest (Fig. 63). The camera is attached to the side of the XRF system and can be easily dismantled or moved if its presence hinders the analysis. All of these components (X-ray tube, SDD, laser diode modules and webcam) are fastened to a support designed and conceived at the workshop of Physics Faculty from US by Jose Luis Benjumea Acevedo. The support can be fastened to the 2D motion system together with the Z linear motion from OWIS presented previously (cf. Chapter 2.3) or it can be coupled to a manual XYZ motion system [135]. Finally, a removable L-shaped bracket can be fastened to the support in order to be able to study small samples easily (like standards used for calibration) (Fig. 64). The idea behind the design of this device is to provide a robust set-up allowing for evolution and versatility. Each component can be removed quickly and easily. The device only takes ten to fifteen minutes to be fully assembled and ready to operate. 45° SDD detector X-ray tube Removable L- shaped bracket Fig. 64. Detail of the X-Panda device. An in-house LabVIEW routine called “XRF_Acquisition” can be used when X-Panda works in connection with the 2D motion system (Fig. 65). It allows the acquisition of punctual spectrum, 1D or 2D analysis. “XRF_Acquisition” controls the 2D motion device and saves each image and spectrum. The user has to choose the name of the folder receiving the files, the number of spectra taken horizontally and vertically, the type of path processes (serpentine or zigzag), the distance between each spectrum and the acquisition time. First of all, the user has to move the 2D motion device to the initial position and has to place the X-Panda at the working distance from the sample using OWISoft (Version 2.71 by OWIS GmbH) which controls the OWIS motion motor. Then, “XRF_Acquisition” calls two sub-programs, moving the X-Panda device and saving the spectrum and the webcam image. First, the main program creates a folder with the name chosen by the user. Then, the main program performs a loop between the two subprograms. The first sub-program saves the spectrum and the image. Then, “2D_Motion_Control” moves the device to the next position. At the end of the loop, “2D_Motion_Control” moves the device back to its original position. Like with 82 “1_Working_distance_Adjustment” LabVIEW, Amptek SDD is controlled with FW5 firmware. Finally, the evolution, in real time, of the detected spectral signals is available in logarithmic or linear scale. Fig. 65. Front panel of “XRF_Acquisition” LabVIEW routine. 2.5.2. Description of the “Museum XRF device” and the SANDRA system. Now, the “Museum XRF device” and the SANDRA system will be briefly presented. The detailed characteristics of the three XRF setups are shown in Table 9. The “Museum XRF device” used in Seville consists of a W X-ray source (RX38 model, EIS S.L.) and an SDD detector (Fig. 66). It is complete with a positioning system made by two laser diode modules [129]. SDD detectorX-ray tube SDD detector X-ray tube Articulating arm boom Fig. 66. “Museum XRF device” during “Virgen del Valle” study (a) and SANDRA setup during “Martirio de San Hipólito” study (b). (a) (b) 83 In Mexico, SANDRA is a Spanish acronym for Sistema de Analisis No Destructivo por RAyos X meaning System for Non Destructive Analysis using X-rays. This portable X-ray system was developed at the Physics Institute of the UNAM (IFUNAM, Mexico). For this study, it was composed by a 75W Mo X-ray tube (XTF5011 model, Oxford Instruments) and a Si-PIN detector (XR-100-CR model, Amptek) (Fig. 66). The X-ray tube is powered by a high voltage power supply XLG50P100 from Spellman and manually controlled by a console. The maximum operating temperature is 50°C. To avoid reaching this temperature, three small fans and a main fan at the bottom of the X-ray tube provide the necessary cooling. For safety reasons, the temperature is monitored by a thermocouple. Finally, a security LED turns on when the X-ray tube is working and a shutter can stop the X-ray beam at any time. The X-ray beam diameter is determined by a lead collimator. The beam diameter is inferior to 1.2 mm. Two laser diode modules pinpoint the working distance from the setup to the sample. This working distance has been fixed to 8 mm. To improve the data recorded and to observe the irradiated region on the sample, a webcam with medium resolution is used. The head of the device is mounted on an XYZ manual motion support allowing movements about 3 cm in each direction to smoothly reach the region of interest on the surface of the sample. The SANDRA system is mounted on an articulating arm boom stand ensuring high mobility and flexibility (Fig. 66). This XRF device can not only be used for analysis on vertical and horizontal planes but also can be tilted as necessary. For more details about the SANDRA set-up, please refer to [136]. 84 Table 9. Characteristics of the three XRF setups. XRF set-up X-Panda - Seville “Museum XRF device” - Seville SANDRA - Mexico Excitation channel X-ray tube anode W W Mo Manufacturer/model Amptek/ Mini-X EIS S.L./RX38 Oxford Instruments/ XTF5011 Maximum power (W) 4 10.5 75 Maximum voltage (kV) 40 30 50 Maximum intensity (μA) 200 350 1500 Be window thickness (μm) 500 150 125 Collimator W W Pb Collimator inner diameter size (mm) 1.5 1 1 Filter Al Al None Filter thickness (mm) 1 1 None Beam diameter spot (mm) 2.9 ± 0.3 3 < 1.2 Working distance (mm) 10 ± 0.1 12 8 Detection channel Active area (mm2) 7 7 6 Si crystal thickness (μm) 450 450 500 Resolution (eV) 148 140 180 Be window thickness (μm) 25,4 127 0.5 Manufacturer/model Amptek/ 123SDD EIS S.L. Amptek/XR-100-CR 85 2.5.3. Data acquisition and processing With both XRF setups, the signals processed by the digital pulse processor are acquired via USB and displayed on the laptop thanks to the ADMCA software (ADMCA Analog and Digital Acquisition, version 2.0.0.0, copyright © 1998-2010, Amptek Inc.) (Fig. 67). This software allows controlling the onset and duration of data acquisition and the visualization in real time of the detected spectral signals presented in logarithmic or linear scale. During this study, ADMCA software was used to record the data and perform a qualitative assessment of the elemental composition of the analyzed area, reading the energy at which the different peaks appear. Prior to this it is necessary to perform an energy calibration recording a spectrum belonging to a composition’s known standard (bronze standard in this case). Once the spectrum is obtained, the energy value of some of the characteristic peaks from the standard has to be entered in order to calibrate the energy scale, allowing the software to assign an energy value to each channel of the multichannel. This procedure has to be performed each time a new session of ADMCA is started. This previous analysis of the same standard also allows for checking the stability of the set-up. It was done for the three XRF devices at the beginning and at the end of each day of study. Finally, all the spectra measured with these three XRF devices have been analyzed with PyMCA. The methodology is slightly different than that which was used with the CXRF data and is detailed in Appendix V. Fig. 67. Graphical user interface of ADMCA software showing a spectrum in logarithmic scale. 92 3.2. Comparison of the capabilities of the IRR cameras Since the PhD thesis of J.R.J. Van Asperen de Boer, infrared reflectography (IRR) has become a classic image registration method in the technical examination of art [5]. This technique reveals the presence of preparatory drawings made of carbon black (infrared no-reflective material) as well as the artist’s process of execution and his changes in the painting composition (pentimenti). Although little explored by an inherent issue due to the complexity of an artwork, it also provides information that helps to deduce the nature of some pigments present on the artwork’s surface [6, 101]. At present, different kinds of IRR cameras are commercially recording a specific part of the IR spectrum depending on the detector technology used. The intensity and the wavelengths of the IR reflected by an artwork depend on the thickness of the paint layers and the kind of materials present (pigments, binding medium, etc.) as well as its quantity. If the IR spectrum recorded is different between the IRR cameras, the image obtained in a comparative study may show differences. During this PhD, three different technologies of IRR cameras were used: - a Si-CCD camera Power Phase model from Phase One, - a Vidicon camera C2741-03 model with N2606-06 tube from Hamamatsu Photonics K.K., - two InGaAs cameras with Xeva-XS-512 model (used in Seville) and, the most recent Xeva-1.7-640 model (in Mexico); both from Xenics. The characteristics of each system were previously explained (cf. Chapter 1.4). As detailed in Chapter 1.4, the sensor response division scheme has been chosen to define the IR range of each camera (Table 13) [102]. It is important to note that both Si- CCD and Vidicon cameras register visible spectrum and Si-CCD also registers UV light. While the Vidicon camera from Hamamatsu uses a visible radiation cut-off filter, the Si- CCD camera from Phase One only works with a UV cut-off filter. Like no visible cut-off filter was used during the study, both visible spectrum radiations and near-infrared radiations participate in the image. This imaging method will be referred as visible-IR imaging. After that, there is a computing process converting the image into a grey scale image to make it appear like an IRR image. These considerations will be important to understand and interpret the results. Table 13. Classification of the different technologies of IRR cameras based on the sensor response division scheme. Camera type Responsivity range (nm) Sensor response division scheme Si-CCD 300-1000 near-infrared Vidicon 400-2200 near-infrared and short-wave infrared InGaAs 900-1700 short-wave infrared 93 Consequently, the experimental paint multilayers previously presented were observed with the four IRR cameras in order to compare the results obtained. The IRR images were obtained under controlled lighting conditions, length and aperture of the cameras. The objectives were to: - evaluate the performance of the different cameras, - define the kind of information obtained by each camera, - set the specific IR reflectivity of each material, - evaluate the accuracy of a first diagnosis about the nature of the pigments presents based on IRR images; to do so, discrimination of different pigments with the same color was tried, - identify the influence of overlapping layers of different materials. 3.2.1. IR reflectivity classification At the beginning of this study, it was important to define a simplified, accurate and more objective classification for the description of images obtained by IRR which will be the foundation of the methodology proposed to differentiate the pigments present in the experimental samples. This classification is divided into three categories based on the characteristic IR reflectivity by the pigments: - Transparent materials, which reflect a specific part of the IR spectrum, - Opaque materials, which do not reflect IR radiation, - Shiny materials, which reflect the most part of the IR spectrum reaching the saturation of the detector. Thereafter, this variation of IR reflectivity will be used to evaluate pigments nature. 3.2.2. Si-CCD camera issue The first result concerns the images obtained with Si-CCD camera. As mentioned before, the image is a combination of IR and visible spectrum information. For example, in Fig. 74, the three red pigments are distinguishable from the yellow one based on observed visible color contrast consideration. The visible spectrum contribution distorts the IR contrast and renders the quality evaluation scale proposed to discriminate pigments useless. With this device, the observation of drawings underneath the pigment is limited not only by shiny and opaque pigments but also by the color of the pigments which can mask the drawing signal (Fig.74 and Table 14). In order to solve this problem and suppress the visible spectrum information from the image, it would be necessary to use the Si-CCD camera coupled with a visible light cut-off filter. Meanwhile, the results obtained cannot be compared with those provided by Vidicon and InGaAs cameras. Table 14. Description of R5 sample. Support name Priming layer 1st pigment layer 2nd pigment layer 3rd pigment layer R5 Hematite Hematite Vermilion Lead-tin yellow 94 Fig. 74. Differences between the images obtained respectively with the Power Phase (near-infrared) (a) and with the Xeva-1.7-640 (short-wave infrared) (b). 3.2.3. Specific reflectivity of several inorganic pigments Excluding the Si-CCD camera, specific IR reflectivities are observable with the other cameras. Regardless of the kind of priming (white or red), or the type of IR equipment (Vidicon or InGaAs), pigments studied show the same features on IR images (Table 15). There is one exception with yellow lead-tin. It looks shiny with the Vidicon camera while it looks transparent with InGaAs camera (Fig.75 and 77). This can be explained by the fact that yellow lead-tin is made of a material which reflects short-wave infrared better than other materials. The Vidicon camera registers a wider IR spectrum than InGaAs camera, and consequently the intensity of the IR reflection registered by the Vidicon detector is higher than that of the InGaAs detector. As a result, for all short-wave infrared reflecting materials, the images obtained with those two detectors show a difference of contrast. The Vidicon image will be shinier than the InGaAs one. This observable difference can be helpful to discriminate two categories of materials: those which reflect only near-infrared and those which reflect short-wave infrared as well like yellow leadtin. There are two possibilities in order to obtain identical results between Vidicon and InGaAs cameras: - Use the Vidicon camera together with a short-wave infrared cut-off filter, - Lighting the artwork only with near-infrared radiation. Table 15. Table of characteristics IR reflectivity. Pigments Vidicon C2741-03 Xeva-XS-512 Xeva-1.7-640 White lead Transparent Transparent Transparent Vermilion Shiny Shiny Shiny Hematite Transparent Transparent Transparent Yellow lead-tin Shiny Transparent Transparent Yellow ochre Transparent Transparent Transparent Malachite Opaque Opaque Opaque Azurite Transparent Transparent Transparent (a) Hematite Hematite Lead-tin yellow Vermilion (b) 95 3.2.4. Discrimination of inorganic pigments with the same color Regarding our selection of pigments, the results obtained with two different yellow pigments (yellow lead-tin and yellow ochre) and two different red pigments (hematite and vermilion) can be compared. The objective is to observe if IRR images allow discrimination of those same color pigments. Regardless of the type of IRR camera, lead-tin yellow is brighter than yellow ochre. This effect hinders the observation of preparatory drawing (Fig.75 and 77). In the case of white priming, the contrast between both pigments is low (Fig.75 and 76) thus obstructs the distinction of both pigments while, in the case of red priming, the contrast is apparent (Fig.77 and 78). Fig. 75. Lead-tin yellow with white priming acquire with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). Fig. 76. Yellow ochre with white priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). Fig. 77. Lead-tin yellow with red priming acquire with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). (a) (a) (a) (b) (b) (b) (c) (c) (c) 96 Fig. 78. Yellow ochre with red priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). In the case of the red pigments, regardless of the type of IRR camera, the vermilion is brighter than the hematite thus hindering, in the case of the vermilion, the observation of preparatory drawing. (Fig.79 and 81). Moreover, regardless of the kind of priming, the difference is apparent between the two pigments (Fig.79 to 82), so it is possible to discriminate them from IRR images. Fig. 79. Vermilion with white priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). Fig. 80. Hematite with red priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). (a) (a) (a) (b) (b) (b) (c) (c) (c) 97 Fig. 81. Vermilion with red priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). Fig. 82. Hematite with white priming acquired with Hamamatsu (a), Xeva-XS-512 (b) and Xeva-1.7-640 (c). 3.2.5. Influence of overlapping layers of different inorganic pigments Several observations can be made about the influence of different overlapping layers. First of all, with all IRR cameras, juxtaposing up to three paint layers with "transparent" pigments did not prevent the observation of the preparatory drawing (Fig.83). Secondly, if one “opaque” pigment layer composed the multilayer, preparatory drawings were not observable (Fig.84). As a result, when IRR studies looked at the preparatory drawings in artwork, it is expected that the areas containing malachite, for example, do not allow the detection of underdrawings. On the other hand, the juxtaposition of paint layers using “shiny” pigments raises the brightness observed in the IRR images and also prevented any observation of preparatory drawings (Fig.85). Fig. 83. Three-layer system with lead white, azurite and hematite with red priming acquired with Xeva-1.7-640. (a) (a) (b) (b) (c) (c) 98 Fig. 84. Bilayer with malachite under white lead (a) and three-layer with yellow ochre above lead white and malachite (b) both with red priming acquired with Xeva-1.7-640. Fig. 85. Bilayer with yellow lead-tin and white lead (a) and three-layer with vermilion above yellow lead-tin and white lead (b) both with red priming acquired with Xeva-1.7-640. 3.2.6. Conclusions of the IRR study The three technologies which were used to do IRR (Si-CCD, Vidicon and InGaAs) allowed the observation of preparatory drawings made in carbon black. To obtain results comparable to those provided by InGaAs and Vidicon cameras, the Si-CCD camera will have to be used together with a UV-visible light cut-filter. A classification linked to IR reflectivity of material was established and applied to several inorganic pigments. White lead, hematite, yellow ochre and azurite are transparent. Malachite is opaque and vermilion is shiny. Yellow lead-tin proved to reflect near-infrared and short-wave infrared radiations. The pigments classified as opaque and shiny hindered the observation of preparatory drawings. Regardless of the type of IRR camera used (Vidicon or InGaAs), lead-tin yellow proved to be brighter than yellow ochre and vermilion showed to be brighter than hematite. So, it is possible to discriminate between those different inorganic pigments with the same color using IRR images. The study of the influence of overlapping layers of different inorganic pigments gives yields information about the best way to observe preparatory drawings. The observation of preparatory drawings should be focused on the location on the artwork presenting only “transparent” pigments. Considering these initial results, a methodology for the preliminary assessment of pigments used in ancient paintings can be proposed. The observation of a first global image of the artwork obtained with an IRR camera allows to focus on the area where detection of preparatory drawing is possible. Diagnosis of pigments used by the artist is also achievable. Obtaining a unique image of the entire artwork is a crucial first step in the methodology proposed in this project for the study of old paintings. Indeed, it (a) (a) (b) (b) 99 enhances the strategy applied in the analysis by other techniques such as X-ray fluorescence (XRF), Fourier transform infrared spectroscopy (FTIR) [138], Raman spectroscopy [139], confocal X-ray fluorescence spectrometry (CXRF), etc. 3.3. Multilayered samples CXRF study In theory, confocal micro X-ray fluorescence (CXRF) provides compositional and spatial information without sampling (which is not the case with SEM-EDX sampling). CXRF extends the capabilities of the micro X-ray fluorescence to in depth of profiling multilayered materials, as previously explained (cf. Chapter 1.2). In this part of the study, its functionality when studying controlled multilayered samples was tested. In order to do so, the depth resolution the µXRF-CONCHA set-up was measured as a function of the energy (see Fig. 46, Chapter 2.1.2). Besides, the depth resolution (FWHM) depends on the chemical composition of the sample. Moreover, the shape of the depth profile depends on the composition, density and thickness of the sample. The results obtained with the µXRF-CONCHA set-up will be compared below with those obtained using two other CXRF systems. The aim of these measurements is to compare the composition and sequence information obtained by analyzing the same experimental paint layers using the three different setups outline in this thesis in order to determine if some differences inherent to those existing between the three devices appear. 3.3.1. Thin foil samples study with µXRF-CONCHA The success of in-depth profiling of multilayered materials depends on several phenomena. To illustrate them, several experiments were performed. First, a superposition of three reference thin layers (Cr, Ni and Pb, 1 µm thick, from Goodfellow) was analyzed. The layers were separated by organic material (kapton polyimide film used in analog photography) with a thickness around 200 μm, superior to the depth resolution of the CXRF device (Fig. 86). The scan was obtained with 10 μm step size and a counting time of 120 s per point. The three layers in the depth profile were observable without any trouble (Fig. 86). The CXRF measurement objective was completed by determining the composition and sequence of the layers. 0200 400 600 0 20000 40000 60000 80000 XRF CONCHA Cr K Ni K Pb L (*10) Intensity (counts) Depth (m) Surface to depth Fig. 86. Depth scan of a superposition of three reference thin layers (Cr, Ni and Pb, 1 µm thick). 100 3.3.2. Experimental paint multilayers study with µXRF-CONCHA However, for more complex unknown samples, such as cultural heritage samples, it might be more difficult to differentiate the layers. To test this assertion, experimental paint multilayers were analyzed. Paintings work on a layer-by-layer system basis. First, the artist prepares the ground layer, then the priming layer. He then adds one or several pigment layers, usually by melting several pigments (organic and inorganic) together and finally, but not always, applying a layer of varnish on the artwork. The interests of CXRF study of antique paintings resides in the definition of the chemical composition of every layer composed by inorganic pigments, with particular interest in the ground layer, the priming layer and the pigment layers. Previously, cross sections of these samples were also studied using optical microscopy in order to evaluate the thickness of each layer and to check the accuracy of the CXRF analyses. The preparation and measurements of these cross sections were then made at the IIE of the UNAM with the invaluable help of Víctor Santos. The average thickness of the layers was measured with AxioVs40 software (version 4.8.2.0 copyright © 2006-2010 Carl Zeiss MicroImaging GmbH) and is summarized in Tables 16 and 17. Table 16. Range of thickness of each layer of the experimental paint multilayers (error estimated ± 1 μm). Support name Priming layer 1st pigment layer 2nd pigment layer 3rd pigment layer Thickness (µm) B1 White lead White lead Malachite Yellow ochre 12-25 6-31 8-17 B2 White lead Lead-tin yellow Hematite Azurite 13-27 12-39 5-10 9-24 B3 White lead White lead Lead-tin yellow Vermilion 18-51 7-12 6-10 B4 White lead Hematite Azurite White lead 28-77 5-23 10-30 8-19 B5 White lead Hematite Vermilion Lead-tin yellow 30-50 3-7 8-14 10-25 R1 Hematite White lead Malachite Yellow ochre 19-39 9-38 12-38 8-20 R2 Hematite Lead-tin yellow Hematite Azurite 6-14 8-15 4-7 11-36 R3 Hematite White lead Lead-tin yellow Vermilion 6-17 5-16 10-31 11-15 R4 Hematite Hematite Azurite White lead 7-30 10-42 8-20 R5 Hematite Hematite Vermilion Lead-tin yellow 10-16 5-10 25-46 Table 17. Range of thickness of each layer of the experimental paint multilayers (error estimated ± 1 μm). Support name Priming layer 1st square 2nd square 3rd square 4th square Thickness (µm) B6 White lead Malachite Azurite Vermilion Yellow ochre 20-56 21-35 11-32 12-17 6-11 R6 Hematite Malachite Azurite Vermilion Yellow ochre 6-31 12-37 30-55 9-20 7-11 101 As illustrated in Fig.87 and Table 18, the CXRF depth profile resulted in the correct information about the sequence and composition of the layers. All the in-depth scans were composed of 40 spectra taken with 5 μm step size and a counting time of 120 s per point. Surface Length (µm) 2nd paint layer - Vermilion (Hg, S) 8-14 1st paint layer - Red Ochre (Fe, Ti) 3-7 Priming - Lead White (Pb) 30-50 Ground (Ca, K) >250 Support –Wood 050 100 150 200 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 XRF CONCHA Hg L Fe K Pb L Intensity (counts) Depth (m) Fig. 87. Depth profile of an experimental paint multilayer. Table 18. Results of depth profile of the experimental paint multilayer. Layer FWHM (μm) Xc (μm) Distance between Xc (μm) Hg Lα 43 ± 3 98 ± 3 2 ± 3 Fe Kα 41 ± 3 100 ± 3 9 ± 3 Pb Lα 37 ± 3 109 ± 3 108 3.3.4. Conclusions of the CXRF study In conclusion, the µXRF-CONCHA set-up allows to carry out the compositional analysis with depth resolution in a nondestructive way. In the study of old paintings, it may provide complementary results to the classical extraction of stratigraphic microsamples. This technique also gives access to depth compositional profiles of the pictorial layers for the purpose of identifying the inorganic pigments palette and techniques used by the artist. However, the data interpretation is not trivial. The absorption effect is the principal limitation of this technique. Results obtained on experimental paint multilayered samples allowed us to propose a specific methodology to study original old paintings whose chemical constitution would be unknown and more complex. In order to simplify the interpretation of a particular depth profile, the characterization of each specific layer must be achieved using a specific area on the painting. The principal problem when considering 16th century paintings is the use of pigments composed of heavy elements (e.g. Pb or Hg) which masked the signals given by the ground layer or the priming layer. As proposed, the first depth profile should be focused in a painting area free of heavy elements. To select this area, the results obtained previously with IRR and XRF should be used. Once ground and priming layers have been clearly defined by CXRF, the second step of the study should focus on pigment layers containing heavy elements. Finally, the CXRF analysis ends with study of interesting details such as pentimenti or a succession of pigments layers. All this information can help to specify or compare the technique of the artist, the choices made during his or her execution, to detect the presence of original painting layers under restoration interventions or to evaluate its condition for conservation. There exists another limitation inherent to the depth resolution. It has proven to be difficult to differentiate layers defined by the same chemical element when the layers are closer or thinner than the depth resolution. It is possible to reduce the depth resolution using polycapillary lenses with lower working distance, which is a safety issue when dealing with invaluable artwork. Finally, the lack of organic material information might be emphasized. In the future, it would be interesting to develop and evaluate the fiber optic mid-Fourier transform infrared spectroscopy (FTIR) [140]. Depending on its usefulness and the technological possibilities, the design of a combined CXRF-fiber optic mid-FTIR setup working in situ, could be considered. In light of the CXRF intercomparison study, it is possible to compare results obtained by different setups. 109 Chapter 4. Alonso Vázquez paintings study In this chapter the study of seven paintings attributed to the artist Alonso Vázquez is presented. First, the painter and the seven artworks in question are presented. Then, the methodology applied is explained and each technique used in the study is illustrated with emblematic examples. In conclusion, a summary of the study of all paintings is proposed. 4.1. Alonso Vázquez and their artworks Alonso Vázquez was a painter from Seville (Spain) who was one of the first Spanish artists to work in the New Spain in 1603. This part of the thesis is written in Spanish as required to claim the international PhD title. At the end of this chapter a synthesis of the seven artworks studied is available in English (cf. Chapter 4.1.8). Sevilla se benefició de la importancia que adquirió el comercio con América. En las artes, afluyeron numerosos maestros foráneos en especial flamencos e italianos y muchas obras penetraron en la ciudad fundamentalmente a través de la actividad comercial del puerto del Guadalquivir. Esos maestros y esas obras contribuirán a configurar al peculiar desarrollo de la escuela pictórica sevillana [141]. La creación de riqueza facilitó la posibilidad de financiar numerosas actividades constructivas, escultóricas y pictóricas. Durante la segunda mitad del siglo XVI, se empieza a constatar la actividad de artistas locales dotados de cierto talento creativo. En la última generación de artistas sevillanos del siglo XVI destaca la presencia de Alonso Vázquez quien practicó una pintura de carácter manierista que, en su momento, tuvo una gran aceptación [142]. 4.1.1. Vida del pintor (Life of the artist) Su nombre completo era según las fuentes Alonso Vázquez Perea [143]. No obstante, no usaba su segundo apellido, lo que hace difícil distinguirlo en los documentos oficiales de homónimos. No se conoce la fecha ni el lugar de nacimiento de este pintor [144]. Lo más probable, debido a lo que se sabe de él con certeza, es que fuera originario de la ciudad de Sevilla. Se puede tener en cuenta que, en el preámbulo de su testamento, Vázquez se considera “natural de la ciudad de Sevilla” [145]. Se casó con Doña Inés de Mendoza el 25 de septiembre de 1588 en la parroquia de Omnium Sanctorum de Sevilla [145]. Tuvo con ella ocho hijos. En el momento de su testamento, les nombra como Diego, doña María de Vargas, Antonio y Alonso mientras cuatro habian ya fallecido: un varón (Juan) y tres mujeres (Francisca, Francisca y Melchora) [145]. No se sabe con quién pudo realizar su formación [144]. El historiador del arte Juan Miguel Serrera define su estilo dentro del Manierismo tardío [146]. El estilo manierista se caracterizó por el predominio de colores vivos artificiales y la representación de la figura humana desproporcionada y en posiciones poco naturales, pero siguiendo en cierto modo "la bella maniera", es decir, la habilidad técnica de los grandes maestros (Leonardo Da Vinci, Rafael Sanzio y Miguel Ángel). Serrera insiste también en su pertenencia al círculo 110 artístico del pintor flamenco Martín de Vos, proponiendo que las obras de Vázquez estén basadas en sus estampas [147]. El historiador del arte José Rogelio Ruiz Gomar Campos describe la tendencia de Vázquez a pintar cuerpos enormes y musculosos. Dice de él que “es un buen dibujante y anatomista que gusta de los escorzos; es poseedor de un eficaz colorido y revela fruición por imprimir a sus ropajes artificiosos pliegues” [148]. En su campo de especialidad como pintor se incluía el policromar retablos y esculturas [147] junto a su aptitud como decorador de techos y policromador de rejas [145]. La carrera artística de Alonso Vázquez se divide en dos etapas: - en Sevilla hasta 1603, con el año 1588 como primera referencia al artista, - en México desde 1603 hasta su muerte en 1607. De la etapa sevillana se conocen gran número de referencias documentales y se conservan numerosas obras, unas documentadas y otras firmadas y fechadas [147]. La primera noticia de su trabajo en Sevilla data de 1588 mientras pintaba la “Sagrada Cena” de la Cartuja de Santa María de Las Cuevas [146]. Durante más de 15 años, Vázquez desarrolló una intensa labor, pintando al servicio de la mayoría de las órdenes religiosas y de los hospitales pertenecientes al Arzobispado sevillano, así como de mercaderes y nobles sevillanos [147]. Fue reconocido como uno de los maestros más importantes y representativos del último cuarto del siglo XVI sevillano junto con Vasco de Pereira y Francisco Pacheco [146] cuando fue contratado con ellos para pintar obras del túmulo levantado en 1598 en la catedral de Sevilla con motivo de las exequias del rey Felipe II [149]. Alonso Vázquez partió de Cádiz a la Nueva España con el virrey Juan de Mendoza y Luna, Marqués de Montesclaros, el 29 de junio de 1603 dejando su esposa y sus hijos en Sevilla [150]. A partir del 26 de octubre de 1603 [150], día en que el virrey y su séquito entraron en la ciudad de México, Alonso Vázquez realizó una serie de trabajos al servicio del virrey marqués de Montesclaros. De esas obras realizadas en México, la mayoría se han perdido o no se han identificado [144]. De hecho, se tiene pocas referencias documentales de su etapa mexicana [145, 147]. Basándose en la “Inmaculada” del Hospital de Jesús (México) y el “Martirio de San Hipólito” atribuidas a Alonso Vázquez, Serrera califica sus obras mexicanas como una prolongación de su etapa sevillana dentro del Manierismo tardío [147]. Alonso Vázquez falleció por causas naturales en la mañana del 13 de abril de 1607 en la ciudad de México [145]. Había permanecido en territorio mexicano 3 años, 7 meses y 1 semana [145]. Alonso Vázquez no fue el primer pintor español que trabajó en la Nueva España, pero sí fue de los primeros (según las fuentes bibliográficas, el primer pintor hubiera sido el andaluz Cristóbal de Quesada en 1538 o Juan de Najara o Nájera [151]). En esos pintores españoles, primeros llegados a Nueva España, se está buscando, hoy en día, los orígenes del arte pictórico colonial mexicano. Por lo tanto, el estudio de las obras de Alonso Vázquez resulta clave para los historiadores del arte para poder evaluar su influjo real sobre la evolución del arte en México. Pero antes de poder hacer ese trabajo, resulta ya necesario determinar qué obras de ese pintor existen todavía y efectuar un trabajo de identificación imprescindible, ya que no están firmadas. El estudio realizado resulta de gran interés para poder comparar las obras mexicanas atribuidas a Vázquez con las de su anterior etapa sevillana. El análisis comparativo permitirá definir con mayor precisión las técnicas pictóricas y la paleta de ese pintor con aportes significativos para el estudio de la pintura Novohispana y la influencia de las 111 escuelas pictóricas españolas y europeas en el México colonial. A continuación, presentamos las siete obras estudiadas. 4.1.2. “Sagrada Cena” La “Sagrada Cena” o “Ultima Cena” es un óleo sobre lienzo que mide 318 × 402 cm (Fig. 94) [142]. Según el historiador del arte Enrique Valdivieso, “la composición de esta obra parece derivar de varias estampas flamencas, de las que el artista ha extraído diferentes elementos, fundiéndolos después con notoria habilidad” [144]. Destaca en esta escena la última cena de Jesús junto a los apóstoles. Se puede apreciar los detalles de los alimentos y de las vajillas que se disponen sobre la mesa y en el suelo del cenáculo. Es una pintura realizada en 1588 para el refectorio de la Cartuja de Santa María de las Cuevas de Sevilla [147, 152, 153]. Si su atribución es cierta [144], se trata de la obra más temprana conocida de Alonso Vázquez. Se conserva desde 1840 en el Museo de Bellas Artes de Sevilla [142]. 4.1.3. “Virgen del Valle” La “Virgen del Valle” o “Virgen del Pozo Santo” es un óleo sobre tabla que mide 213 × 123,5 cm (Fig. 95) [154]. La obra representa a la Virgen con el Niño en su regazo mientras es coronada por dos ángeles y acompañada por otros dos que portan instrumentos musicales. A sus pies se narra la milagrosa intervención de la Virgen para conseguir que salga indemne un niño del pozo al que había caído y en un segundo plano se presenta un paisaje de un valle [155]. En 1978, Valdivieso la atribuyó a Alonso Vázquez. Esa atribución fue corroborada por Serrera quien identificó esa obra como la de la “Virgen del Valle”. La “Virgen del Valle” fue la pintura centrica del retablo que Doña Beatriz Pérez encargó a Vázquez el 2 de julio de 1597 para la iglesia del Convento de Santa María de Jesús de Sevilla [146]. La obra perdió su titulacion original al desmembrar el retablo y fue presentada como “Virgen del Pozo Santo”. El hecho que esa obra fuera inicialmente parte de un retablo está confirmado con los daños observados durante su última restauración en 1999. Se notificó una intervención anterior en la que se modificó el formato original de la tabla de medio hexágono a rectángulo [155]. Si esa atribución es correcta, la obra fue ejecutada en 1597 según el contrato [146]. Fue comprada por el Deán Manuel López Cepero a raíz de la desamortización de los bienes eclesiásticos [146] y, finalmente, donada en 1860 por sus herederos al Cabildo de la Catedral de Sevilla, su ubicación actual [154]. Esa obra fue restaurada en 1999 por el Instituto Andaluz del Patrimonio Histórico con motivo de la exposición “Velázquez y Sevilla” del cual se tiene el informe de la investigación efectuada [155]. 112 Fig. 94. “Sagrada Cena”. Fig. 95. “Virgen del Valle”. 113 Fig. 96. “San Pedro Nolasco despidiéndose de Jaime I El Conquistador”. Fig. 97. “San Pedro Nolasco redimiendo cautivos”. 114 4.1.4. La serie sobre la vida de San Pedro Nolasco (“Nolasco” series) Pacheco relata en su “Arte de la pintura” que fray Juan Bernal, prior del convento de la Merced Calzada de Sevilla (actual edificio del Museo de Bellas Artes de Sevilla), lo contrató junto a Alonso Vásquez en 1600 para realizar una serie pictórica sobre la vida de San Pedro Nolasco y San Ramón Nonato destinadas a decorar el claustro principal [142, 144, 152]. Sólo se conservan dos de las obras que Vázquez pudo haber realizado. Permanecen en el Museo de Bellas Artes de Sevilla y representan a “San Pedro Nolasco redimiendo cautivos” y “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” [144]. 4.1.4.1. “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” Se trata de un óleo sobre lienzo que mide 208 × 250 cm (Fig. 96) [142]. Del conjunto de obras estudiadas, es la única firmada por Alonso Vázquez y fechada en el año 1601 [152]. La obra se divide horizontalmente en dos planos. En el superior, a la izquierda, aparece San Pedro Nolasco despidiéndose de Jaime I el Conquistador mientras que, a la derecha, se abre en la lejanía un paisaje marítimo con un puerto [156]. En el plano inferior, a la derecha, dos personajes sentados a una mesa cubierta con un mantel rojo parecen estar concluyendo un trato mientras que, en el ángulo inferior derecho, San Pedro Nolasco supervisa la maniobra de un grupo de porteadores [156]. 4.1.4.2. “San Pedro Nolasco redimiendo cautivos” Se trata de un óleo sobre lienzo que mide 206 × 250 cm (Fig. 97) [142]. Al contrario que el precedente, éste no está ni firmado ni fechado. Por lo tanto se propone una fecha de ejecución entre 1600 (año del contrato) y 1603 (año de su viaje a México). Esta escena podría tratarse de una de las redenciones que San Pedro Nolasco efectuó en el Norte de África en la que se exalta a la Orden de la Merced como redentora de cautivos [142]. Se puede diferenciar tres planos distintos en la obra. En la parte superior izquierda, se representa el momento en el cual San Pedro Nolasco concluye el trato económico con un Sarraceno. La escena principal se desarrolla en el plano central con San Pedro Nolasco y el mismo Sarraceno supervisando al grupo de los cautivos representados en la zona inferior izquierda [142]. 115 4.1.5. “Inmaculada” La “Inmaculada” o “Virgen Inmaculada” es un óleo sobre lienzo que mide 283 × 206 cm (Fig. 98) [147]. El año de ejecución de la obra y su lugar de origen son desconocidos [147]. Sólo se sabe que fue desamortizada en 1840 y permanece en el Museo de Bellas Artes de Sevilla [142]. En esa obra, la Inmaculada aparece triunfante sobre el dragón demoniaco, acompañada de ángeles. En la parte inferior, está representado un paisaje donde se advierten los símbolos de las advocaciones marianas [142]. Fue atribuida estilísticamente a Alonso Vázquez por Serrera. La única referencia documental que se tiene a día de hoy sobre una Inmaculada pintada por Vázquez está mencionada por Palomero Páramo. Se trata de una pintura que interesaba el misionero agustino fray Pedro Solier [145]. Se refiere a un cuadro que debió de ser pintado en torno a 1605 en México del cual dice que ignora su destino final, que, “en caso de aparecer, resolvería el problema de atribución que hoy tienen las Vírgenes apocalípticas del Museo de Sevilla y del Hospital de Jesús, de México”. Fig. 98. “Inmaculada”. 116 4.1.6. “Martirio de San Hipólito” Se trata de un óleo sobre lienzo que mide 103 × 73 cm (Fig. 99) [147]. La obra conmemora la caída de la capital azteca México-Tenochtitlán ocurrida el 13 de agosto de 1521, justo el día de San Hipólito [157]. La composición se divide en dos planos. En el fondo se aprecia la plaza de una iglesia ocupada por soldados que portan escudos decorados con águilas bicéfalas doradas. En el centro del cuadro, San Hipólito yace de cabeza, sus pies están atados y va arrastrado por un caballo que corre en medio de observadores civiles y militares [157]. El santo mártir está flanqueado del lado izquierdo por un indio americano vestido con taparrabos y un tocado simple de plumas en el cabello atado. No se trata de una representación de un azteca sino una versión simbólica del indio salvaje de piel rojiza [157]. Al lado derecho del Santo, está representado Hernán Cortés hincado y con las manos unidas en gesto de oración [157]. En el frente del pilar que aparece detrás de la figura orante de Hernán Cortés quedan restos de una leyenda, de la que sólo son identificables el nombre de un mes y las cifras correspondientes a un día y año: el 13 de agosto de 1526, día de San Hipólito y de la conquista de la ciudad de México por Hernán Cortés [147]. Finalmente, encima del pilar, la escultura de un ángel sostiene un medallón de forma ovalada sobre el que se lee la inscripción “Salve Gloria” [157]. Serrera identifica esa obra como la que componia el triptico del altar mayor de la iglesia del Hospital de Jesús de la ciudad de México fundado por Hernán Cortés. Esa pintura estuvo desubicada a partir de 1657 y, hoy en día, está conservada en el Museo Nacional de Historia (MNH) del Castillo de Chapultepec en la ciudad de México [147]. Serrera destaca que debido a sus restauraciones es imposible precisar la autoría del “Martirio de San Hipólito” y que un estudio técnico contribuiría a confirmar tales suposiciones [147]. Si su atribución es correcta, esa obra fue realizada entre los años 1605 (año del contrato) y 1607 (año de su muerte) y la imagen de Hernán Cortes sería una de sus representaciones más antiguas hasta ahora conocida [147]. Esa obra tiene otra historia debido a su valor simbólico. Durante el siglo XIX, estuvo atribuida al artista Rodrigo de Cifuentes [151]. No obstante, Rodrigo de Cifuentes nunca existió y resultó ser un invento del conde de la Cortina, cuando, en 1853, escribió su biografía en el “Diccionario Universal de Historia y Geografía” [158]. En aquella época, en que la nación mexicana estaba creciendo y el nacionalismo se desarrollaba, el personaje de Rodrigo de Cifuentes encarnó el primer gran artista mexicano de la Nueva España en la propaganda nacionalista. El historiador del arte Pedro Ángeles expone las razones de la invención histórica de ese personaje fundamental para defender la idea del origen de la pintura mexicana [151]. Durante ese proceso de propaganda, se le atribuyeron varias obras de arte y hasta se falsificaron algunas cubriendo firmas originales con la supuesta suya o añadiéndola simplemente [157]. En 1983, el “Martirio de San Hipólito” fue restaurado y se eliminó la firma de Cifuentes por considerarla un falso histórico [157]. Esta acción radical borró un aspecto muy importante en la historia del objeto. Más allá del estudio del pintor Alonso Vázquez, en el caso del “Martirio de San Hipólito”, se plantean otras preguntas: - ¿La representación de Hernán Cortés es contemporánea a la representación de San Hipólito? - ¿Las inscripciones ubicadas en la zona donde está representado Hernán Cortés son contemporáneas al resto de la obra o fueron añadidas posteriormente? 117 - ¿Quedan restos de alguna firma del pintor original (o sea, Alonso Vásquez) escondida en el momento de la falsificación del siglo XIX? - ¿Quedan restos de la firma falsa de Rodrigo de Cifuentes? 4.1.7. “Virgen de las Uvas” La “Virgen de las Uvas” o “Virgen apocalíptica” o “Inmaculada Concepción” es un óleo sobre lienzo que mide 145 x 103 cm (Fig. 100) [159]. El año de ejecución de la obra es desconocido. Solo se sabe que permanece hoy en día en la Sala Capitular de la Catedral Metropolitana de la ciudad de México [159]. La obra representa a la Virgen con el Niño en su regazo mientras es coronada por dos ángeles. El Niño sostiene uvas en su mano derecha. En la parte inferior, está representado un paisaje. La historiadora del arte Nelly Sigaut atribuye estilísticamente esa obra a Alonso Vázquez [159]. Posteriormente dice: “Asumo el riesgo y no oculto que Ruiz Gomar me comunicó verbalmente que Serrera vio la obra de la catedral de México y desechó una posible atribución a Alonso Vázquez.” 124 Table 23. Net peak areas (in counts) of the elements detected by XRF in different restoration interventions found in selected points showed in Fig. 103. Points Ti-K Co-K Zn-K Ba-L Presence of restoration interventions 207 1528 ± 39 n.d. 9765 ± 99 112 ± 11 Ti and Zn 208 14 ± 4 43 ± 7 988 ± 31 124 ± 11 209 42 ± 6 5 ± 2 963 ± 31 81 ± 9 210 77 ± 9 22 ± 5 882 ± 30 30 ± 5 211 65 ± 8 53 ± 7 838 ± 29 87 ± 9 212 70 ± 8 220 ± 15 1553 ± 39 43 ± 7 213 6427 ± 80 n.d. 29160 ± 170 308 ± 18 Ti and Zn 214 101 ± 10 n.d. 2190 ± 47 114 ± 11 Zn 215 25 ± 5 25 ± 5 977 ± 31 114 ± 11 216 52 ± 7 39 ± 6 1042 ± 32 87 ± 9 217 100 ± 10 13 ± 4 824 ± 29 181 ± 13 222 60 ± 8 265 ± 16 1575 ± 40 60 ± 8 234 4 ± 2 136 ± 12 949 ± 31 133 ± 12 235 104 ± 10 115 ± 11 861 ± 29 64 ± 8 236 12 ± 3 18 ± 4 1746 ± 42 104 ± 10 237 79 ± 9 222 ± 15 2173 ± 47 18 ± 4 Zn 238 61 ± 8 188 ± 14 1623 ± 40 99 ± 10 4.3.2. IRR The first use of IRR in the study of paintings consisted in the observation of the carbon black preparatory underdrawings. Indeed, carbon black is visible in IRR images due to its non-IR reflective property. In the IRR mosaic of the “San Pedro Nolasco redimiendo cautivos”, the underdrawings, done by the artist in order to guide him through the painting execution of the arched architecture and stairs, is visible (Fig. 104). As mentioned before, IRR can also identify the presence of restored areas, when material with different IR reflectivity from the original material has been used (Fig. 103). Because of this ability, it is complementary to the observation by UV-induced visible fluorescence photography. Fig. 104. Underdrawings of the arched architecture and stairs visible on the IRR mosaic of the “San Pedro Nolasco redimiendo cautivos” (b). (b) (a) 125 Fig. 105. IRR detail of the “Virgen del Valle” obtained with a manual tripod system. The little deviations and modifications of distance that occure during the capture of the IRR images hinder the stitching process. The advantage of using an automatic device (e.g. Gigapan) instead of a manual tripod system in order to obtain a full mosaic of a great artwork is obvious as shown in the example of the “Virgen del Valle” (Fig. 105). The difficulty in using a tripod is too high for it to be chosen as a systematic method. It should be employed only in cases where an automatic device is unavailable. In the most difficult case studies (confined space, in situ studies, etc.), the Gigapan system is the most appropriate (cf. Chapter 2.4.1). The concept of differentiating inorganic pigments using their different IR reflectivity properties is difficult to apply to artworks. The reason is that the artist may have used a mixture of carbon black or ochre to darken the tone of the pigment. In consequence, the resulting IR reflection of the mixture is different than the IR reflection of a pure pigment. Because of this, the characteristic IR reflection of the pigment is not consistent In the detail of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” (Fig. 106 and Table 24), the artist used, on one side, only vermilion for bright red (XRF points 15 and 34) and, secondly, red ochre to obtain a dark red (XRF points 33, 39 and 40). XRF point 15 presents a higher reflection rate according to the properties of the vermilion. Meanwhile, XRF point 34 is not as evident as the painter probably used a mixture that includes carbon black. The XRF points revealing a mixture of ochre with vermilion (33, 39 and 40) present similar IR reflection properties to ochre. In the same painting, the difference between lead-tin yellow (XRF points 14 and 17) and yellowish ochre (XRF points 23, 25 and 26) is most obvious in IRR. This is due to the fact that, usually, the artist did not mix lead-tin yellow with any other pigment to get a brighter yellow (Fig. 106 and Table 24). It has been observed (Chapter 3.2.3) that azurite presents a higher IR reflection rate than malachite. However, the IR reflection properties of other copper-based green pigments (e.g. copper resinate) have not been studied in this work. In another detail of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” (Fig. 107 and Table 25), azurite (detected in XRF points 50, 80 and 98) presents the IR 126 reflection property expected. However, this property is modified when azurite is underlay with black iron oxide in order to obtain dark blue (XRF point 99). XRF point 84 is more difficult to interpret. The presence of Cu detected by XRF probably belongs to a blue sublayer (representing the sea) which was then covered by the upper layer (representing the railing) composed by white lead mixed with carbon black. This hypothesis will be confirmed by CXRF results (cf. Chapter 4.3.7). Finally, the copper-green pigment observed in XRF point 82 presents a higher IR reflection rate than pure malachite. Due to the large amount of Cu observed in XRF, this different IR reflection could correspond to another kind of green copper-based pigment. . . . . 15 34 33 39 40 . . . . . .. 17 14 26 25 23 . Fig. 106. IRR details of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador”. XRF detection of vermilion is marked in bright red while red ochre is marked in dark red (a). Lead-tin yellow is marked in bright yellow while yellowish ochre is marked in orange (b). Table 24. Net peak areas (in counts) of the elements detected by XRF in selected points showed in Fig. 106. Pigments Characteristic elements Points Fe-K Sn-K Hg-L Pb-L Lead-tin yellow Pb, Sn 14 2904 ± 54 2298 ± 48 4017 ± 63 570600 ± 760 17 9476 ± 97 2892 ± 54 13790 ± 120 605720 ± 780 Yellow ochre Fe 23 38490 ± 200 903 ± 30 3990 ± 63 299530 ± 550 25 57610 ± 240 301 ± 17 1215 ± 35 70610 ± 270 26 22700 ± 150 281 ± 17 3040 ± 55 262650 ± 510 Vermilion Hg 15 2511 ± 50 1280 ± 36 64680 ± 250 548930 ± 740 34 29160 ± 170 306 ± 17 60140 ± 250 275640 ± 530 Red ochre Fe 33 70050 ± 270 423 ± 21 201 ± 14 75850 ± 280 39 59390 ± 240 218 ± 15 7883 ± 89 31430 ± 180 40 69920 ± 260 217 ± 15 7468 ± 86 34690 ± 190 (a) (b) 127 . .50 84 82 80. . .99 98 . . Fig. 107. IRR details of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador”. XRF detection of azurite is marked in blue while green Cu-based pigment is marked in green. Table 25. Net peak areas (in counts) of the elements detected by XRF in selected points showed in Fig. 107. Pigments Characteristic elements Points Fe-K Cu-K Azurite Cu 50 6657 ± 82 194770 ± 440 80 6264 ± 79 298620 ± 550 84 7446 ± 86 180680 ± 430 99 3100 ± 56 158630 ± 400 Azurite underlay with black iron oxide Cu, Fe 98 31240 ± 180 318760 ± 570 Cu-based pigment Cu 82 4168 ± 65 458170 ± 680 4.3.3. 4.3.4. Inorganic pigments study with XRF XRF was used to study the palette of the artist. A novel methodology was applied to the treatment of the XRF data in order to compare the results measured. In fact, the studies of the seven artworks were made using three different XRF setups (cf. Chapter 2.5). Table 26 presents the conditions used for each device. 128 First, the presence of modern pigments resulting from restoration interventions (Ti, Zn, Co and Ba) was detected. As for the original pigments, 24 categories linking a pigment or pigment mixture to a certain color palette were defined (Table 27) [160, 161]. Table 26. Conditions used for each XRF device. XRF device Time per spectrum (s) Voltage (kV) Intensity (µA) X-Panda 150 30 130 “Museum XRF device” 300 33.4 80 SANDRA 120 45 150 Table 27. Description of the 24 categories defined for XRF study. Color palette Pigments Characteristic elements White Lead white Pb Flesh tones Lead white with vermilion Pb, Hg Lead white with lake Pb Yellow Lead-tin yellow Pb, Sn Massicot Pb Red Vermilion Hg Mixture of vermilion with ochre Hg, Fe Red ochre Fe Purple lake / Blue Azurite Cu Azurite underlay with black iron oxide Cu, Fe Azurite underlay with umber Cu, Fe, Mn Organic blue pigment / Green Bright green Cu-based pigment Cu Mixture of Cu-based pigment with ochre Cu, Fe Organic green pigment Fe Ochre Yellow ochre Fe Ochre mixed with vermilion Fe, Hg Dark color Black iron oxide Fe Mixture of ochre with Cu-based pigment Fe, Cu Umber Fe, Mn Mixture of umber with Cu-based pigment Fe, Mn, Cu Brown obtained with Cu-based pigment Cu Organic black pigment / White was obtained with white lead (Pb). When mixed with vermilion, the artist gets flesh tones. Also, flesh tones obtained with lake were observed (only presence of Pb was detected). Yellow was obtained with lead-tin yellow (Pb and Sn) or massicot (Pb). The methodology distinguishes four kinds of red, three kinds of blue, two kinds of green, and two kinds of ochre. Finally, the pigments with hues from dark ochre to black through gray and blue-gray were separated into 6 categories. In the case of the “Virgen de las Uvas”, the presence of organic green and blue pigments was noted. 129 4.3.5. Accuracy of the methodology applied to the treatment of the XRF data To determine the accuracy and efficiency of the methodology, radar charts were drawn. The homogeneity of each group of pigments has been verified from the XRF study of the “Sagrada Cena” done with the X-Panda setup. The ability to compare artworks has been tested in the case of the blue category comparing XRF spectra of each artwork obtained with different XRF setups. The radar charts represent the net peak areas (in counts) of the six characteristic elements detected by XRF (Pb-L, Ca-K, Mn-K, Fe-K, Cu-K, Hg-L). This type of plot is a graphical method of displaying multivariate data in the form of a two-dimensional chart of several variables represented on axes starting from the same point. The maximum of each axis corresponds to the maximum value measured by one XRF set-up on one painting. Each spectrum is represented by a color area depending of the value of the net peak areas of the six elements selected. If the pigment category is well defined, the area resulting by the superposition of all spectra has to be homogeneous. Consequently, each pigment shows a specific pattern depending of its chemical composition. In the case of the “Sagrada Cena” study, the maximum value chosen for each axis corresponds to the maximum value registered between all the XRF spectra measured in the painting. Lead white (Pb) and every organic pigment mixed with lead white (e.g. organic black and red lake) exhibit similar patterns as they are all characterized by XRF by the presence of Pb (Fig. 108). Mixing lead white with vermilion, the artist gets flesh tones (Fig. 109). Red categories: the mixture of vermilion with ochre (Hg, Fe) presents more Fe than the vermilion (Hg) as expected (Fig. 110). Indeed, the mixture of vermilion with ochre (Hg, Fe) and the ochre mixed with vermilion (Fe, Hg) have the same pattern (Fig. 110). The patterns of yellow ochre (Fe) and of black iron oxide (Fe) are also similar (Fig. 111). The azurite (Cu) and the azurite underlay with black iron oxide (Cu, Fe) present similar patterns, with a higher content of Fe in the second, as expected (Fig. 112). The brown obtained with azurite (Cu) is similar to azurite (Cu) while the mixture of ochre with Cubased pigment (Fe, Cu) is similar to the azurite underlay with black iron oxide (Cu, Fe) (Fig. 112 and 113). “Umber” (Fe Mn) and the mixture of umber with Cu-based pigment (Fe, Mn, Cu) show similar patterns, with a higher content of Cu for the second, as expected (Fig. 114). The bright green Cu-based pigment (Cu) and the mixture of Cubased pigment with ochre (Cu, Fe) present similar patterns, with a higher content of Fe for the second (Fig. 115). 130 Fig. 108. Radar charts of the lead white (19 spectra) (a), the organic black (32 spectra) (b) and, the red and purple lakes (7 spectra) (c) categories. Fig. 109. Radar charts of the flesh tones category (Lead white with vermilion) (36 spectra). Fig. 110. Radar charts of vermilion (5 spectra) (a), the mixture of vermilion with ochre (24 spectra) (b) and the ochre mixed with vermilion (18 spectra) (c) categories. 131 Fig. 111. Radar charts of the yellow ochre (4 spectra) (a) and the black iron oxide (12 spectra) (b) categories. Fig. 112. Radar charts of the azurite (23 spectra) (a) and the azurite underlay with black iron oxide (16 spectra) (b) categories. Fig. 113. Radar charts of the brown obtained with Cu-based pigments (10 spectra) (a) and the mixture of ochre with Cu-based pigment (21 spectra) (b) categories. 132 Fig. 114. Radar charts of the umber (13 spectra) (a) and the mixture of umber with Cu-based pigment (7 spectra) (b) categories. Fig. 115. Radar charts of the bright green Cu-based pigment (19 spectra) (a) and the mixture of Cu-based pigment with ochre (6 spectra) (b) categories. For the blue category comparison, for each artwork, XRF spectra were obtained using different XRF setups. In this case, the maximum value chosen for each axis of the radar charts corresponds to the maximum value registered between all the XRF spectra measured in each painting with each setup. Five of the six artworks (Fig. 116 to 118) show similar patterns with high content in Cu and Pb, and low content in Mn and Hg with the exception of the “Virgen del Valle” which presents a higher content in Hg due to the fact that the artist applied blue pigment above flesh tones (Hg, Pb) (XRF points 25, 33 and 34) (Fig. 117 and 188). Finally, the pattern of the “Virgen de las Uvas” is noticeably different (Fig. 118). This is because of the use, in this painting, of a blue organic pigment instead of azurite (i.e. indigo). 133 Fig. 116. Radar charts of “Bright blue” category of “San Pedro Nolasco redimiendo cautivos” (3 spectra) (a) and “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” (4 spectra) (b). Fig. 117. Radar charts of “Bright blue” category of “Virgen del Valle” (8 spectra) (a) and “Martirio de San Hipólito” (4 spectra) (b). Fig. 118. Radar charts of “Bright blue” category of “Sagrada Cena” (23 spectra) (a), “Inmaculada” (8 spectra) (b) and “Virgen de las Uvas” (12 spectra) (c). 4.3.6. 1D analysis with XRF In the case of the “Sagrada Cena”, three 1D analyses with XRF using the 2D motion device at the Fine Arts Museum of Seville have been made. The objective was to test the device's ability to obtain 1D profile in areas that present interesting contrasts in IRR. All spectra of the lateral profiles were recorded during 10 s with 30 kV for 130 µA. The first lateral profile illustrates the technique applied by the artist to paint the shadow of the apostle’s leg on the gray floor. The IRR image shows an interesting 140 020 40 60 80 100 120 140 160 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000 XRF CONCHA Fe K (*10) Cu K Intensity (counts) Depth (m) Fig. 129. Depth profile of the railing of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador”. 4.4. Synthesis of the corpus studied The study synthesis of the seven artworks is presented below. The artist’s painting technique and the inorganic pigments of his palette are shown amongst other information provided through the archaeometric investigation. 4.4.1. Preparatory drawings The IRR study revealed very few preparatory drawings carried out by carbon black. All these drawings were used in architectural areas such as stairs, grille (Fig. 104), windows or elements such as a table (Fig. 130). They are all geometric elements difficult to produce without drawing them first. 4.4.2. Pentimenti Faces and hands of the characters show very few corrections. The paint application is sharp; the borders of the bodies and clothes were improved using a dark color mixture (Fig. 59, 128, 131 and 132). More pentimenti are observed on decorative items such as dishes, cutlery, pots, etc. In the case of the “Sagrada Cena”, the presence of numerous pentimenti is probably unique due to the size and complexity of this painting. The modification of design of the calyx (Fig. 133) or the disappearance of a knife (Fig. 134) can be shown as examples. Besides pentimenti, the IRR mosaic reveals the different elaboration steps of some parts of the artwork. For example, the artist first painted the entire dish before he made the lamb, starting with his body, finishing with his feet (Fig. 133). 141 Fig. 130. Details of the IRR image of the “Sagrada Cena” showing the preparatory drawings of the window (a) and the modification of the table border (b). Fig. 131. Details of the IRR (a) and visible (b) images of the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador” showing how the border of the arms and hands were corrected using a dark mixture. 142 Fig. 132. Details of the IRR image of the “San Pedro Nolasco redimiendo cautivos”. Fig. 133. Details of the IRR (a) and visible (b) images of the “Sagrada Cena” showing the modification of design of the calyx and the different steps of construction used to represent the dish with the lamb. Fig. 134. Details of the IRR (a) and visible (b) images of the “Sagrada Cena” showing the disappearance of a knife. 143 4.4.3. Pigments used on artworks produced in Seville. From the XRF studies, based on the methodology proposed (cf. Chapter Erreur ! Source du renvoi introuvable.), a synthesis of the pigments detected on each artwork is presented (Table 33). The modern pigments used for restoration interventions were zinc white (Zn), titanium white (Ti) and cobalt blue (Co). White barium sulfate (Ba) was detected in the case of the “Inmaculada”. The most complex palettes were detected in the “Sagrada Cena” (with 20 different pigments and mixtures). It is a logical conclusion as it is the largest artwork. Comparing the five artworks from Seville, the palette of inorganic pigments used is very similar. The different tones were obtained by applying the same mixtures with some variations from one painting to another. For example, the dark blue was obtained by a layer of azurite underlay with umber on the “Inmaculada”. Some brown color was obtained only with copper-based pigments most likely mixed with a black organic pigment on the “Sagrada Cena”, the “San Pedro Nolasco redimiendo cautivos” and the “Virgen del Valle”. The use of organic pigments was suggested by the absence of inorganic elements able to explain the color observed in the artwork. The use of an organic black pigment (as lamp black for example) in the “Sagrada Cena” and in the “Virgen del Valle” was detected. This organic black pigment may have been used in the other artworks. However, the XRF study did not reveal any areas on those paintings where the absence of an inorganic element suggested its use. Purple lake (as carmine for example) was observed on all of the artworks (with the exception of “San Pedro Nolasco redimiendo cautivos”). It is possible that a similar lake served to produce some flesh tones on the “Sagrada Cena” and the “Virgen del Valle”. Finally, both the “Inmaculada” and the “Sagrada Cena”, attributed to Alonso Vázquez by Serrano, present a palette of pigments and mixtures coherent with those observed on his confirmed artworks carried out in Seville. In the case of the “Sagrada Cena”, IRR observations are also coherent while the “Inmaculada” displays too much deterioration and restoration interventions to obtain any useful observation from the IRR mosaic. 4.4.4. The Mexican artworks The pigments’ palette detected on the “Martirio de San Hipólito” is consistent with the artworks from Seville, while the palette on the “Virgen de las Uvas” differs drastically. The blue and green colors cannot be explained by the use of inorganic Cu-based pigments as observed on the artworks from Seville. The future Mexican SEM-EDX study of the micro-samples extracted from this painting may confirm the presence of organic pigments (as indigo for the blue color and a mixture of indigo with ochre for the green color for example). In the case of “Martirio de San Hipólito”, IRR images revealed the pentimenti of the pediment above Hernán Cortes as well as the angel and the shield. The pigments and mixtures detected are similar to those from the rest of the painting. It suggests that the representation of Hernán Cortés, as well as the inscriptions, are contemporary with the representation of San Hipólito. If Serrera’s attribution is correct, this portrait may be one of the oldest known from the Spanish conqueror. Finally, IRR images did not show remains of the signature of Alonso Vásquez nor of the fictional Rodrigo de Cifuentes. 144 Table 33. Synthesis of the pigments detected on the seven artworks. Color palette Pigments Characteristic elements Inmaculada Sagrada Cena Nolasco, cautivos Nolasco, Jaime I Virgen, Valle San Hipolito Virgen, Uvas White Lead white Pb 1 1 1 1 1 1 1 Lead white with vermilion Pb, Hg 1 1 1 1 1 1 Lead white with lake / 1 1 1 1 Lead-tin yellow Pb, Sn 1 1 1 1 Massicot Pb 1 1 1 Vermilion Hg 1 1 1 1 1 1 Mixture of vermilion with ochre Hg, Fe 1 1 1 1 1 Red ochre Fe 1 1 1 Purple lake / 1 1 1 1 1 1 Azurite Cu 1 1 1 1 1 1 Azurite underlay with black iron oxide Cu, Fe 1 1 1 1 1 Azurite underlay with umber Cu, Fe, Mn 1 Organic blue pigment / 1 Bright green Cu-based pigment Cu 1 1 1 1 1 1 Mixture of Cu-based pigment with ochre Cu, Fe 1 1 1 Organic green pigment Fe 1 Yellow ochre Fe 1 1 1 1 1 1 1 Ochre mixed with vermilion Fe, Hg 1 1 1 Black iron oxide Fe 1 1 1 1 1 1 Mixture of ochre with Cu-based pigment Fe, Cu 1 1 1 1 1 Umber Fe, Mn 1 1 1 1 1 1 Mixture of umber with Cu-based pigment Fe, Mn, Cu 1 1 1 1 1 Brown obtained with Cu-based pigment Cu 1 1 1 Organic black pigment / 1 1 1 1 15 20 15 13 15 16 12 Zinc white Zn Titanium white Ti Cobalt blue Co White barium sulfate Ba Number of categories Restoration Green Flesh tones Yellow Red Ochre Dark color Blue 145 4.4.5. CXRF results From the nine in-depth scans performed at the Fine Arts Museum of Seville (Table 28), one of the objectives was to define the priming and, eventually, the ground preparation of each painting. As example, Figure 135 and Table 34 show the in-depth profile obtained on a blue area of the “Inmaculada”. The results show a first layer of dark blue, azurite underlay with black iron oxide (Fe and Cu), sustained by a Pb-based layer and a red priming layer (Fe). 050 100 150 200 0 10000 20000 30000 40000 50000 60000 XRF CONCHA Cu K Pb L (*10) Fe K (*10) Intensity (counts) Depth (m) Fig. 135. Depth profile of the “Inmaculada”. Table 34. Results of depth profile presented Fig. 134. Layer Xc (μm) Distance between Xc (μm) Fe Kα 105± 3 1 ± 3 Cu Kα 106± 3 6 ± 3 Pb Lα 112 ± 3 15 ± 3 Fe Kα 127 ± 3 Table 35. Synthesis of the information obtained with CXRF about priming and ground layers. Artwork title Priming layer Ground layer “San Pedro Nolasco despidiéndose de Jaime I” Fe Ca “Sagrada Cena” Fe or Pb ND “Inmaculada” Fe ND The other results obtained were presented in the Chapter 4.3.7 and are summarized in Table 35. The three paintings show the presence of a red priming layer. In the case of the “Sagrada Cena”, a second kind of priming layer was detected. This second Pb-based priming is used in the area with the representation of the apostles. Based on the observation of the IRR image (Fig. 132), it is possible to suggest that this technique, which uses two different kinds of priming layer, is also present in the “San Pedro Nolasco redimiendo cautivos”. Nevertheless, the CXRF results were not conclusive on this last artwork and more CXRF analyses have to be done in the future to answer this question. 146 4.4.6. SEM-EDX results Micro-sampling was done on three artworks. The IAPH restorers took six samples during the restoration intervention of the “Virgen del Valle” in 1999. Five and six samples were extracted respectively on the “Martirio de San Hipólito” and on the “Virgen de las Uvas” by Mexican researchers from LDOA (IIE-UNAM). The results report from IAPH is available. The results of the SEM-EDX study of the “Martirio de San Hipólito” samples were communicated in 2012 [157] while the study of the “Virgen de las Uvas” is still in progress and results are not yet available. The XRF study of the “Virgen del Valle” includes analysis of the areas next to the sampling made by the restorers of the IAPH, in order to compare the XRF results with the SEM-EDX results from the IAPH (Table 36). The IAPH report does not propose interpretation of the results. However, the similarity of the information obtained for Val2, Val3 and Val4 can be observed. The IAPH report concludes that the white ground preparation consists of white calcium sulphate mixed with animal glue [162]. The thickness of the layer is from 100 to 450 µm. The reddish/pink priming layer is a mixture of lead white, calcite and ochres with a thickness from 15 to 60 µm. Table 36. Comparison of the results obtained with SEM-EDX and XRF studies of “Virgen del Valle”. Observed color SEM-EDX study XRF study IAPH sample name Characteristic elements Spectrum name Characteri stic elements Pigment or mixture determined Dark green Val1 Pb, Fe 51 Pb, Cu Cu-based pigment (Cu) Flesh tones Val2 Pb, lake 43 Pb, Ca Lead white with lake (Pb) Purple Val3 Pb, Ca, lake 62 Pb, Ca Purple lake (Pb) Blue Val4 Pb, Ca, Cu 50 Pb, Cu Azurite (Cu) Dark ochre Val5 Ca, Pb, Fe 27 Pb, Ca, Mn, Fe Umber (Mn, Fe) Dark brown Val6 Ca, Pb, Fe, Cu 60 Pb, Ca, Cu Brown obtained with azurite (Cu) Table 37. Comparison of the results obtained with SEM-EDX and XRF studies of “Martirio de San Hipólito”. Observed color SEM-EDX study XRF study LDOA sample name Characteristic elements Spectrum name Characteristic elements Pigment or mixture determined White M1 Pb 59 Pb Lead white Red M2 Red lake, Pb 35 Fe Red ochre Brown red M3 Lamp black, Fe 64 Fe, Mn Umber Dark color M4 Lamp black, Pb 67 Pb Organic black pigment Green M5 Cu 3 Cu Cu-based pigment The XRF study of the “Martirio de San Hipólito” includes analysis of the areas next to the sampling made by researchers from LDOA in order to compare the XRF results with 147 the SEM-EDX results obtained (Table 37). The similarity of the results for M1, M4 and M5 can be observed. The LDOA results report concludes that the dark red ground preparation consists of red clay (iron oxides with quartz, feldspars, igneous rock fragments with iron and titanium, calcite, silica and aluminum as impurities) mixed with hematite and barium sulfate or barite. The presence of barite in red ground preparation has already been identified in various examples of New Spanish paintings [163] as well as in 17th French paintings [164]. The thickness of the layer is more or less 100 µm. The dark gray priming layer is a mixture of the previous ground preparation with white lead oxide and copper presenting a thickness from 20 to 50 µm. The LDOA report points out that literature does not mention any previous examples of the use of Cu-based pigments in the priming layer during the 17th century [157]. Finally, in the case of “Martirio de San Hipólito”, the chemical information obtained by SEM-EDX about the ground preparation and the priming layers is similar to the CXRF in-depth profiles results obtained at the Fine Arts Museum of Seville and to the SEMEDX study made by the IAPH during the restoration intervention of the “Virgen del Valle”, supporting the authorship attribution to Alonso Vázquez (Table 38). Table 38. Summary of the results obtained with SEM-EDX and CXRF studies. Artwork title Study Ground preparation Ground color Priming Priming color Sagrada Cena CXRF ND ND Fe or Pb Red or white Virgen del Valle SEM-EDX (IAPH) Ca, S White Pb, Ca, Fe Reddish/pink San Pedro Nolasco despidiéndose de Jaime I CXRF Ca White Fe Red San Pedro Nolasco redimiendo cautivos CXRF In progress La Inmaculada CXRF ND ND Fe Red Martirio de San Hipólito SEM-EDX (LDOA) Fe, Ca, Ba, S Dark red Fe, Pb, Ca, Ba, Cu, S Dark gray Virgen de las Uvas SEM-EDX (LDOA) In progress 4.5. Conclusions of the Alonso Vazquez’s artworks study Alonso Vázquez was among the first Spanish painters to work in New Spain. Therefore, the study of the Mexican paintings of Alonso Vázquez is important for the Art History in order to assess their real influence on the evolution of art in Mexico. Nonetheless, before being able to carry out this work, it is necessary to identify artworks of this painter that still exist and to do an unpredictable identification study for those that remain unsigned. The present study was used to compare the artworks attributed to Vázquez during his period in Mexico and his previous stay in Seville. The comparative analysis defined the pictorial techniques and palette of this artist. The information about the seven artworks is summarized in Table 39. For the material study of these paintings, the artworks with 148 confirmed authenticity were taken as reference. They are from the series of “San Pedro Nolasco”. A comparative study of other paintings allowed to discuss their attribution from the material point of view. The methodology used to analyze the seven artworks in Seville and Mexico was similar in order to normalize the results and be able to compare them. The study was carried out following an interdisciplinary methodology developed for the analysis of paintings which consisted of: - Photographic record in the reflected visible light spectrum and UV-induced visible fluorescence photography, - IRR study, - XRF study, - CXRF study (in the case of the artworks from the Fine Arts Museum of Seville), - OM and SEM-EDX studies of micro-samples mounted in cross sections (in the case of Mexican artworks). The accuracy and efficiency of the methodology applied to XRF data treatment and interpretation was verified. Its ability to compare artworks was also demonstrated. From the XRF studies, based on the methodology proposed, a synthesis of the pigments detected on each artwork was presented. The technique of the artist and his palette in terms of inorganic pigments was shown amongst other information provided by the archaeometric investigation. The IRR study revealed very few preparatory drawings made of carbon black. All these drawings were used in architectural areas (e.g. stairs, grille, windows) or elements such as the table. They are all geometric elements difficult to produce without drawing them first. Faces and hands of the characters show very few corrections. The paint application is sharp, the border of the bodies and clothes were corrected using the same mixture of dark pigments used in the decorated adjacent parts. More pentimenti are observed on decorative items such as dishes, cutlery, pots, etc. In XRF, the most complex palettes were detected in the “Sagrada Cena” and the “San Pedro Nolasco despidiéndose de Jaime I El Conquistador”, followed by “San Pedro Nolasco redimiendo cautivos” and the “Inmaculada”. It is a logical conclusion as they are the greatest artworks. Comparing the five artworks from Seville, the palette of inorganic pigments used is very similar. The different tones were obtained by applying the same mixtures with some variations from one painting to another. The use of organic pigments was suggested by the absence of inorganic elements able to explain the color observed in the artwork. The use of an organic black pigment in the “Sagrada Cena” and in the “San Pedro Nolasco redimiendo cautivos” was detected. This organic black pigment may have been used in the other artworks. Purple lake was observed on all the artworks. It is possible that the same or similar lake served to produce some flesh tones on the “Sagrada Cena”. Finally, both the “Inmaculada” and the “Sagrada Cena”, attributed to Alonso Vázquez by Serrano, present a palette of pigments and mixtures coherent with those observed on his confirmed artworks done in Seville. Finally, the results obtained from the two Mexican paintings were compared to the five artworks analyzed in Seville. The palette of pigments detected on the “Martirio de San Hipólito” is consistent with the artworks from Seville, while the palette on the “Virgen de las Uvas” differs drastically. The blue and green colors cannot be explained by the use of inorganic Cu-based pigments as observed on the artworks from Seville. The future Mexican SEM-EDX study of the micro-samples extracted from this painting may confirm the presence of organic pigments. In the case of “Martirio de San Hipólito”, IRR images 149 revealed the pentimenti of the pediment above Hernán Cortes as well as the pentimenti of the angel and the shield. The pigments and mixtures detected are similar to those from the rest of the painting. It suggests that the representations of Hernán Cortés and San Hipólito are contemporary. If Serrera’s attribution is correct, this portrait may be one of the oldest known of the Spanish conqueror. To conclude this study, Table 39 presents the new synthesis of the corpus excluding “Virgen de las Uvas” by material study and confirming the attribution of four paintings including “Martirio de San Hipolito” as the first Mexican painting attributed to Alonso Vazquez by both material and art history studies. Table 39. Summary about the seven artworks. Artwork title Alternative title Year of execution Authenticity Alonso Vazquez paintings Sagrada Cena Ultima Cena 1588 Attributed by material and art history studies Virgen del Valle Virgen del Pozo Santo 1597 Attributed to an altarpiece with contract and by material study San Pedro Nolasco despidiéndose de Jaime I El Conquistador None 1601 Contract and artwork signed San Pedro Nolasco redimiendo cautivos None 1600-1603 Contract La Inmaculada Virgen Inmaculada Unknown Attributed by material and art history studies Martirio de San Hipólito None 1605-1607 Attributed to an altarpiece with contract and by material study Unknown artist painting Virgen de las Uvas Virgen apocalíptica or Inmaculada Concepción Unknown Excluded by material study