Scientific Examination of Art: Modern Techniques in Conservation and Analysis

Home , International Klein Blue, Maya blue, Zaffre

Scientific Examination of Art

MODERN TECHNIQUES IN CONSERVATION AND ANALYSIS

Washington, D.C. March 19–21,2003

THE NATIONAL ACADEMIES PRESS 500 Fifth Street, N.W. Washington, D.C. 20001

This work includes articles from the Arthur M. Sackler Colloquium on the Scientific Examination of Art: Modern Techniques in Conservation and Analysis held at the National Academy of Sciences Building in Washington, D.C., March 19-21, 2003. The articles appearing in these pages were contributed by speakers and attendees at the colloquium and were anonymously reviewed, but they have not been independently reviewed by the Academy. Any opinions, findings, conclusions, or recommendations expressed in this work are those of the authors and do not necessarily reflect the views of the National Academy of Sciences.

The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished scholars engaged in scientific and engineering research, dedicated to the furtherance of science and technology and to their use for the general welfare. Upon the authority of the charter granted to it by the U.S. Congress in 1863, the Academy has a mandate that requires it to advise the federal government on scientific and technical matters.

International Standard Book Number: 0-309-09625-1 (Book)

Copies of this report are available from the National Academies Press, 500 Fifth Street, N.W., Lockbox 285, Washington, D.C. 20055; (800) 624-6242 or (202) 334-3313 in the Washington metropolitan area; Internet, http: // www.nap.edu.

Cover: “Corner of the Studio” by Antonio Ciocci. Courtesy of Catherine and Wayne Reynolds

The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished scholars engaged in scientific and engineering research, dedicated to the furtherance of science and technology and to their use for the general welfare. Upon the authority of the charter granted to it by the Congress in 1863, the Academy has a mandate that requires it to advise the federal government on scientific and technical matters. Dr. Bruce M. Alberts is president of the National Academy of Sciences.

The National Academy of Engineering was established in 1964, under the charter of the National Academy of Sciences, as a parallel organization of outstanding engineers. It is autonomous in its administration and in the selection of its members, sharing with the National Academy of Sciences the responsibility for advising the federal government. The National Academy of Engineering also sponsors engineering programs aimed at meeting national needs, encourages education and research, and recognizes the superior achievements of engineers. Dr. Wm. A. Wulf is president of the National Academy of Engineering.

The Institute of Medicine was established in 1970 by the National Academy of Sciences to secure the services of eminent members of appropriate professions in the examination of policy matters pertaining to the health of the public. The Institute acts under the responsibility given to the National Academy of Sciences by its congressional charter to be an adviser to the federal government and, upon its own initiative, to identify issues of medical care, research, and education. Dr. Harvey V. Fineberg is president of the Institute of Medicine.

The National Research Council was organized by the National Academy of Sciences in 1916 to associate the broad community of science and technology with the Academy’s purposes of furthering knowledge and advising the federal government. Functioning in accordance with general policies determined by the Academy, the Council has become the principal operating agency of both the National Academy of Sciences and the National Academy of Engineering in providing services to the government, the public, and the scientific and engineering communities. The Council is administered jointly by both Academies and the Institute of Medicine. Dr. Bruce M. Alberts and Dr. Wm. A. Wulf are chair and vice chair, respectively, of the National Research Council.

www.national-academies.org

Arthur M. Sackler, M.D. 1913-1987

Born in Brooklyn, New York, Arthur M. Sackler was educated in the arts, sciences, and humanities at New York University. These interests remained the focus of his life, as he became widely known as a scientist, art collector, and philanthropist, endowing institutions of learning and culture throughout the world. He felt that his fundamental role was as a doctor, a vocation he decided upon at the age of four. After completing his internship and service as house physi- cian at Lincoln Hospital in New York City, he became a resident in psychiatry at Creed-moor State Hospital. There, in the 1940s, he started research that resulted in more than 150 papers in neuroendocrinology, psychiatry, and experimental medicine. He considered his scientific research in the metabolic basis of schizophrenia his most significant contribution to science and served as editor of the Journal of Clinical and Experimental Psychobiology from 1950 to 1962. In 1960 he started publication of Medical Tribune, a weekly medical newspaper that reached over one million readers in 20 countries. He established the Laboratories for Thera- peutic Research in 1938, a facility in New York for basic research that he directed until 1983. As a generous benefactor to the causes of medicine and basic science, Arthur Sackler built and contributed to a wide range of scientific institutions: the Sackler School of Medicine established in 1972 at Tel Aviv University, Tel Aviv, Israel; the Sackler Institute of Graduate Biomedical Science at New York University, founded in 1980; the Arthur M. Sackler Science Center dedicated in 1985 at Clark University, Worcester, Massachusetts; and the Sackler School of Graduate Biomedical Sciences, established in 1980, and the Arthur M. Sackler Center for Health Communications, established in 1986, both at Tufts University, Boston, Massachusetts. His pre-eminence in the art world is already legendary. According to his wife Jillian, one of his favorite relaxations was to visit museums and art galleries and pick out great pieces others had overlooked. His interest in art is reflected in his

philanthropy; he endowed galleries at the Metropolitan Museum of Art and Princeton University, a museum at Harvard University, and the Arthur M. Sackler Gallery of Asian Art in Washington, D.C. True to his oft-stated determination to create bridges between peoples, he offered to build a teaching museum in China, which Jillian made possible after his death, and in 1993 opened the Arthur M. Sackler Museum of Art and Archaeology at Peking University in Beijing. In a world that often sees science and art as two separate cultures, Arthur Sackler saw them as inextricably related. In a speech given at the State University of New York at Stony Brook, Some reflections on the arts, sciences and humanities, a year before his death, he observed: ‘‘Communication is, for me, the primum movens of all culture. In the arts. . . I find the emotional component most moving. In science, it is the intellectual content. Both are deeply interlinked in the humanities.’’ The Arthur M. Sackler Colloquia at the National Academy of Sciences pay tribute to this faith in communication as the prime mover of knowledge and culture.

ORGANIZING COMMITTEE BARBARA BERRIE, Senior Conservation Scientist, National Gallery of Art, Washington, D.C. E. RENÉ DE LA RIE, Head of Scientific Research, National Gallery of Art, Washington, D.C. ROALD HOFFMANN (NAS) (Chair), Frank H. T. Rhodes Professor of Humane Letters, Cornell University JANIS TOMLINSON (NAS), Director of University Museums at the University of Delaware TORSTEN WIESEL (NAS) (Chair), President Emeritus, The Rockefeller University JOHN WINTER, Conservation Scientist, Freer Gallery of Art and Arthur M. Sackler Gallery, Washington, D.C.

Staff KENNETH R. FULTON, Executive Director ALYSSA CRUZ, Program Administrator (from October 2005) MIRIAM GLASER HESTON, Program Officer (until October 2005)

vii

Preface

The study of works of art using scientific methods dates back to the late 18th century but expanded exponentially in the late 20th century. The Sackler conference held March 19-21, 2003, assembled a group of leading conservators and conservation scientists to present and assess recent initiatives providing a unique overview of this important field. Six of the following fourteen papers begin with a key material for cultural artifacts (Venetian pigments, works of art on paper, photographs, stone sculpture, modern paints, and early Chinese jade) and enumerate various means of identification and analysis. Four of the papers start with an advanced analytical method and discuss its applications: infrared reflectography, multi-spectral imaging, Raman microspectroscopy, and quantitative gas chromatography-mass spectrometry. Two papers focus on mechanisms of deterioration—biodeterioration of outdoor stone and disruptions in the surfaces of aged paint films. The breadth of the discourse is well illustrated by the topics listed above and by three summary papers: an overview of the concept of conservation science, a brief history of the evolution of practical conservation techniques and attitudes in the 20th century, and a discussion of the impact of collaborative research among conservators, scientists, and art historians. These fourteen contributions exemplify the wide variety of art materials that challenge the investigative scientist and the increasing sophistication of an array of scientific tools that now aid in the decision making for the important task of the preservation of works of art and cultural heritage.

Dr. Joyce Hill Stoner, Professor and Paintings Conservator Winterthur/University of Delaware Program in Art Conservation

Contents

THE STATE OF THE FIELD Overview 3 John Winter

Material Innovation and Artistic Invention: New Materials and New Colors in Renaissance Venetian Paintings 12 Barbara H. Berrie and Louisa C. Matthew

The Scientific Examination of Works of Art on Paper 27 Paul M. Whitmore

Changing Approaches in Art Conservation: 1925 to the Present 40 Joyce Hill Stoner

An Overview of Current Scientific Research on Stone Sculpture 58 Richard Newman

Biodeterioration of Stone 72 Thomas D. Perry IV, Christopher J. McNamara, and Ralph Mitchell

xii CONTENTS

TECHNIQUES AND APPLICATIONS Analytical Capabilities of Infrared Reflectography: An Art Historian’s Perspective 87 Molly Faries

Color-Accurate Image Archives Using Spectral Imaging 105 Roy S. Berns

Multi-Spectral Imaging of Paintings in the Infrared to Detect and Map Blue Pigments 120 John K. Delaney, Elizabeth Walmsley, Barbara H. Berrie, and Colin F. Fletcher

Modern Paints 137 Tom Learner

Material and Method in Modern Art: A Collaborative Challenge 152 Carol Mancusi-Ungaro

Raman Microscopy in the Identification of Pigments on Manuscripts and Other Artwork 162 Robin J. H. Clark

Paint Media Analysis 186 Michael R. Schilling

A Review of Some Recent Research on Early Chinese Jades 206 Janet G. Douglas

APPENDIXES A Contributors 217 B Program 225 C Participants 228

The State of the Field

Overview

John Winter Freer Gallery of Art and Arthur M. Sackler Gallery Smithsonian Institution Washington, D.C.

This paper introduced a colloquium whose theme was the study of works of art by scientific methods. To present a brief overview of a field where all kinds of works might be studied by any applicable kind of scientific technique is hardly a practical possibility. Rather, I would like to try to give a little more depth to all of this, in terms of both the history and the diversity to be found in studies of these types. One basic problem lies in the conceptual magnitude and diversity of such a field. A “work of art” can mean a human artifact designated as such and made from an enormous variety of materials. Implicitly we are attempting to bring together objects made from rocks and minerals, metals of all kinds, ceramics, organic materials derived from plants and animals, or synthetically created—the list goes on. An artifact may be a complex, partially ordered system with components of diverse chemical nature, as is true of most paintings and many other things, or it may comprise only one type of component. The scale can vary from thumbnail size to that of architecture and monuments. Even the word “art” does not help much, since any familiarity with the field reveals people working with what is usually termed self-conscious art, with decorative art, or with functional objects regarded for the purpose as art. For the most part, scientists who choose to do this kind of research do not seem to trouble themselves overmuch with how artistic the art is. The field overlaps that of archaeological science, which studies archaeological, usually excavated artifacts, although much archaeological science is not concerned with artifacts at all. All these things might be examined using any method from any branch of science that holds the promise of yielding some kind of result. This colloquium will be covering large segments of this whole area,

4 SCIENTIFIC EXAMINATION OF ART

though it would be optimistic to suppose that all possible types of work and kinds of artifacts could possibly be covered in two days.

PEOPLE A word should be entered concerning the scientists who choose to do this kind of work and where they do it. The field can scarcely be said to be overpopulated by practitioners, at least in relation to its overall conceptual scale. Tennent (1997) saw the organizational structure as being in four parts: laboratories in museums, those university departments that take an interest, research institutes (often national research institutes) that have departments established for this purpose, and to a lesser extent the private sector. Many people in the field nowadays are professional research scientists fully committed to this branch of research in the same sense that other scientists will consider themselves fully committed to a particular branch of science. These tend to be found working in the research institutes and in departments of the larger museums, occasionally in universities. The majority of them are scientists who started out in some branch of the mainstream sciences, typically a branch of chemistry or physics or materials science, before moving into the present field. There are now a few, though only a few, who were able to do graduate studies in the field itself. A smaller group of research scientists have their primary interests elsewhere but also take part in cultural properties studies. They tend to be in academic institutions and may work on projects of interest for a short or extended period and then move out of the field again. Then there is a less easily defined group of scholars and professionals who are trained in fields other than the sciences but who perform and apply research to problems in their own field: art historians, conservators, and archaeologists may fall into this category. Most major branches of physical science have much higher numbers of researchers than is the case with us, and modern science has as a consequence a considerable social structure, for want of a better term. Leading scientists form groups and schools of research that interact with one another, perhaps in collaboration, perhaps in competition. This can be on a relatively large scale and may sometimes last for extended periods. It includes direct, informal contact as well as more formalized kinds. In our field this intensity of interaction, which depends on a kind of critical mass of people, is much less. The number of practicing researchers is too small in relation to the number of kinds of things that they might be doing, that is the number and variety of research topics that exist. Since it is at least arguable that the immense success of the twentieth-century scientific endeavor in general was to some extent a result of such social structuring, problems are implied for our own comparatively diluted areas for which it might be difficult to find answers.

OVERVIEW 5

TECHNIQUES AND TERMINOLOGY The scientific methods that we use deserve some comment, though it is difficult to generalize. They have usually been methods of study—of analysis, imaging, accelerated testing, and so forth—taken quite directly from other areas of science and technology. They tend, as a result, to have been optimized for work within some other field. With a few important exceptions, such as one or two dating methods, most techniques were not developed specifically within our own field. This state of affairs, of a conceptually large research field populated by relatively small numbers of researchers using techniques borrowed from elsewhere, led one colleague, Irwin Scollar, (actually with reference to archaeological science) to suggest that this was equivalent to conducting guerrilla warfare using captured weapons (Olin, 1982, p. 102). One of the consequences of the rather complex situation that I have just sketched is terminological: There is no general agreement on what to call this field of study, taken as a whole. There is not even total agreement on what general term to use for the objects of study. Since they may or may not be archaeological, may or may not be historical, and may or may not always be artistic (according to somebody’s definition), such phrases as “cultural heritage,” “cultural property,” and “cultural assets” have come into use but are clumsy when an extension of the terms into studies using scientific methods is required. For the field of study itself we have on the archaeological side, “archaeometry,” “archaeological science,” and “science in archaeology,” which have all been used, and sometimes criticized. These terms are not usually extended to research on works of artistic or historical importance unrelated to archaeology. Here “conservation science” has become prevalent, especially in the United States, though the work may or may not be related to efforts to conserve the objects concerned. “Technical studies of works of art” was in use in the 1930s but is seldom found now. “Technical art history” has appeared, and the parallel to archaeological science would appear to be “art historical science.” All these terms, however, seem to imply subsets of the field as a whole, which awaits its definitive title and therefore perhaps its precise definition.

HISTORY It might help give some depth to the discussions to look briefly at the history of the field. Even an extended look would be partial, since to the best of my knowledge no definitive account is available: Much of the historical spadework remains to be done. We do know that scientific study of antiquities and works of art goes back to the late eighteenth century. Earle Caley (1951) located almost 100 publications dated before 1875 (of which the earliest was late eighteenth century) mostly concerned with archaeological materials, and especially with the analysis of metals. Through the nineteenth century, work on this kind of material was

6 SCIENTIFIC EXAMINATION OF ART

sporadic and mostly conducted by a few individuals concerned with identifying and analyzing archaeological and similar material on the side in laboratories primarily devoted to other purposes. Thus were the origins of one kind of research that continues to the present day: the study of artifacts that we consider archaeological, whether or not systematically excavated. It is now regarded as one segment of archaeological science, the segment concerned with artifacts. Much of this research seems to be done in academic institutions. The driving force is largely archaeological, and although the objects concerned may also be classified as fine art, this is largely coincidental. There is typically freedom to take samples necessary for analysis, and conservation of the objects has not usually been an issue. We might regard this as the archaeological tributary of the research efforts that developed during the twentieth century. The examination of paintings and sculpture appears to go back over a similar time period. Since this paper was delivered, Nadolny (2003) has published a historical study of early analytical work on paintings, which appears to date from ca 1780. We know of analyses of pigments in mural painting by Haslam in 1800 and Humphrey Davy in 1815 (Rees-Jones, 1990), and of work in Munich on easel paintings from 1825 (Miller, 1998). It can be regarded as forming another line of development leading to where we are now. Two of the better-known practitioners were A. H. Church in the late nineteenth century and A. P. Laurie in the earlier part of the twentieth century; both served as professors at the Royal Academy of Arts in London. Much of the motivation for this type of work seems to have been historical interest, with reference being made also to various historical texts. Both connoisseurship and a desire to encourage contemporary artists to use appropriate and durable materials may also have played a part. This kind of research seems mostly to have taken place in the larger museums and in research institutes set up to work with them, occasionally in academic departments. Here conservation of the object is much more of an issue; the taking of samples is more restricted, especially in recent times, and may be forbidden outright. Consequently noninvasive methods have become important. Scientific research devoted to making conservation itself more rational and effective came along a little later than the preceding two tributaries of development, though it can also be traced back to the nineteenth century. The National Gallery in London commissioned reports on the condition of its paintings in the 1850s (Brommelle, 1956), and the British Museum consulted outside scientists on conservation problems well before setting up its own facilities (Watkins, 1997). In 1888 Friedrich Rathgen’s laboratory was set up in the Königlichen Museen in Berlin (Plenderleith, 1998). The years following the First World War saw the founding of conservation departments in a number of places: the British Museum and the National Gallery in London, Le Louvre in Paris, the Fogg Museum at Harvard University, among others. This kind of research has come to overlap extensively the research in the preceding category, the historical investigation of

OVERVIEW 7

the fine arts. It tends to be done in similar places and often by the same people, and similar restrictions on methods of investigating an object usually apply. There is much complexity in the ways that these historical streams have flowed down to contribute to the present state of affairs. There is overlap of major categories, both conceptually and in the sense that the same people may conduct kinds of research that might be looked upon as logically different. Different classes of cultural property also impose their own characteristics on any studies that are conducted on them. Research on large-scale entities (for example, buildings, monuments, and sites) is probably driven very largely by conservation needs, including protection and restoration, but its practitioners might see little in common with the conservation of museum objects.

AESTHETIC CONSIDERATIONS Given this complexity in the study of anything held to be of cultural significance, using many techniques from the sciences, with a number of reasons and motiva- tions driving us, what are the common threads? What kind of conceptual framework is it possible to discern in all this? Before making any attempt to answer we must refer to yet another aspect of the situation. When we say we want to study works of art using the methods of science, we imply that these works have significance quite outside any scientific considerations, and that this significance is the reason for finding them important enough to study. This aspect cannot be ignored. Obviously the practicing conservator can never ignore it, but I suggest that the scientist doing research on works of art cannot ignore it either, even when the research appears to consist entirely of, say, solving problems of analysis and to be quite matter of fact in nature. The distinction to be seen here has been drawn before, perhaps many times. Anything that we call a work of art is being seen by definition from at least two points of view. One point of view sees it as a physical object, the other looks at whatever properties the object has that lead us to say that it is a work of art, and to attach value to it on this basis. Joseph Margolis (1980) defined a work of art as a token embodied in a physical object. Referring to a work as an image conveys much the same idea. When we speak of such aspects of the work as expressiveness, style, symbolism, the meaning of the whole work or parts of it, any emotional feelings (positive or negative) that may be aroused, we are adopting the token or image point of view. Seeing the work as a physical object is, I believe, self-evident in meaning, and doing so is not confined to the research scientist or conservator; however, to study a work of art using scientific methods means scrutinizing it as a physical object to a greater depth and from more points of view than would be done with any other approach. The specification of what should be studied springs from other parts of human culture. Traditional art history adopts the token or image point of view largely, though

8 SCIENTIFIC EXAMINATION OF ART

not entirely. Around the late nineteenth to early twentieth century we see examples of art historians, such as Konrad Fiedler, who saw the final form and the style of a work as the product of the interaction of artists with their materials, and Gottfried Semper, who appeared to see art as the byproduct of handicraft (Hauser, 1985). Although this kind of thing does not represent very much that has endured in art historical concepts, the physical object that embodies the art as a token has meant something in traditional art history. For example, art historians have from time to time taken an interest in workshop organization and procedures in the production of paintings (e.g., Phillips, 2000; Shimizu, 1981). For all this, the conceptual framework of art history has been established very largely in aesthetics and similar considerations. It is reasonable to ask how far this can affect our own interest in the same objects of study and how far there can be intersections in the frames of reference.

ART AND TIME Apart from the fact that we study works of art rather intensively as physical objects, what other commonality can be discerned to make us think that the scientific study of this huge mass of disparate cultural assets can form a coherent subject? One way of looking at it is to say that we study those products of human- kind, defined as cultural assets—or art—along each object’s time axis. Such a concept can be divided into three phases. At one end of the time axis we look at the materials the creator (or creators) of an artifact used and how they used them. Then we can consider what changes have occurred in the product. Finally we assess what is the situation for the artifact in question now and how we can predict and influence its life into the future. We start with the production of the work of art. Art historians talk about the inspiration of the artist, that artist’s vision, the influence of other artists or schools, and so on that results in the creation of the particular thing that we now admire and discuss. The fact remains that no painting or sculpture or anything else springs from somebody’s mind in the fashion of a “thinks” bubble in a cartoon strip. It has to be fashioned from whatever materials were available, using whatever techniques were in use, and these aspects are among the things we are trying to discover about that object. The identity of the artist may or may not be known, and commonly more than one person was involved. We could look on this as investigating the ethnology of the creation of a surviving work. We take account of the historical context and the cultural context in which this process occurred, both of which inevitably had their influences on what was created, which raw materials were used, and on how it all happened. We have a link with human beings who lived in the past—perhaps the recent past, perhaps a more remote past—not just in the sense of the aesthetic concepts or visions they possessed (important as these were) but also in the sense of how they got their hands dirty to make something; ultimately we are investigating not just interesting assemblies of

OVERVIEW 9

pigments, binders, stone, ceramic, wood, or whatever it may be but the real people who created things. On to the second phase: What has happened to our cultural asset since it was made. Any artifact, whether artistic or not, starts to change from that moment. The kinetics of such changes vary rather a lot, but on some time scale changes are happening. We call these deterioration mechanisms, and to me as a chemist they are both extremely interesting and quite difficult to study. An understanding of deterioration mechanisms is important from at least two opposite-facing points of view. If we are concerned with the production of an artifact by bygone persons, we are presumably concerned with what they actually produced, which will have changed to a greater or lesser extent in the meantime. There are some areas where such changes are small enough to be ignored but a great many more where they are not. To project our understanding back to the start of the object’s time axis, we need to talk about what has happened to it. This is true even though many artists may have known well that their creations would change over time and they may have been perfectly content with that. The second reason for understanding deterioration mechanisms is conservation, which one may think of as facing forward rather than backward. Conservators are given the responsibility for stabiliz- ing, treating, and perhaps restoring something that has survived in better or worse condition, and trying to ensure its continued survival into the future. To deal with this rationally they need to know what has been happening chemically and physically to the assembly of materials constituting each object. This links directly to the third phase of our time axis: how to extend it forward as far as possible. The conservator needs to know not only what is there in a material sense but also what is likely to happen with it chemically and physically, possibly after some treatment has been applied. Knowledge of such processes is also needed for any present-day materials that may be used for treatment in the context of the ways in which they are used. Investigations of these complex issues in conservation have become of primary interest in recent years.

IMPLICATIONS FOR THE SCIENTIST To the researcher in this field who was brought up, as many of us were, in some branch of the mainstream sciences, the demands can be challenging. Typically, work to obtain a scientific research degree, possibly followed by a year or two of postdoctoral research, will lead to proficiency in some branch of science taught in universities, probably a subdiscipline of chemistry or physics. The science thus mastered may be applied to situations arising possibly over many types of works of art and cultural heritage generally. Committed professionals in our field may soon find themselves with some research specialty defined in terms of the works of art themselves; my own, for example, happens to be East Asian paintings. The professional researcher then finds that studying the works of art as physical objects within his chosen area, whether limited or broad, requires the application of

10 SCIENTIFIC EXAMINATION OF ART

scientific knowledge and understanding from a number of scientific disciplines, which may be removed from his original area of proficiency. There has been a kind of orthogonal transposition of concepts; rather than specializing in a single scientific discipline in depth, the researcher needs to take a range of basic disciplines and apply them to a class of objects that will themselves be studied in depth. No doubt this happens in other fields of research too, and it is certainly intellectually stimulating. It can also be alarming. Most scientists, I think, are sensitive to the implications of specialization, to the probability of wandering into error when they venture into branches of science other than their own. The physicist John Ziman published a book (1987) some years ago dealing with questions of mobility and career change in the sciences, including the reasons why most scientists tend to be reluctant to change areas of research in which they work. The problem of how to apply selected, specialized areas of science to a further understanding of things that ultimately have to be understood on their own terms is also an intellectual challenge of the field.

CONCLUSION I conclude with a few words about the colloquium that followed. For reasons that I mentioned earlier, describing all aspects—or all important aspects—of the scientific examination of art is impractical. We hope to have organized a fair sampling of what the field is about, in all its variety and complexity. This first day was intended to give fairly broad reviews of progress in at least some of the major areas of work. The second day saw accounts of significant progress in more specific topics. This was intended to give us some realistic perspectives on what has been achieved and what has not been achieved in research, particularly that of the past few years. I think that most of the presentations will fit on the time axis of an object that I suggested as describing the kinds of work done. Some may look at questions of the materials and methods used by the creators of artifacts that we choose to call “art,” some at research on deterioration mechanisms, and others at questions of an object’s present status and the prognostications we may have for its future. In this introductory paper, rather than discussing modern techniques or recent progress, which others will discuss later, I have tried to give some sugges- tion of depth, even (in a sketchy kind of way) historical depth to the subject. I would like to be able to give it some coherence, but I fear that would be claiming altogether too much. Do we really have just one field here, or several smaller fields that happen to overlap here and there? What are the connections between scientific studies and considerations of aesthetics, the original intent behind creating something, and the connections to questions of intended use? This colloquium was never intended to cast light on problems of this nature, but if we have a serious intellectual discipline underpinning what we do, the more fundamental questions implied by its pursuit should at least be recognized to exist.

OVERVIEW 11

REFERENCES

Brommelle, N. 1956. Studies in Conservation 2:176-187. Caley, E. R. 1951. Journal of Chemical Education 28:64-66. Hauser, A. 1985. The Philosophy of Art History. Evanston, Ill.: Northwestern University Press. English version of Philosophie der Kunstgeschichte, Oscar Beck, Munich, 1958, pp. 216, 232-234. Margolis, J. 1980. Art and Philosophy: Conceptual Issues in Aesthetics. Brighton, Sussex: Harvester Press. Miller, B. F. 1998. In Painting Techniques. History, Materials and Studio Practice. Contributions to the Dublin Congress 7-11 September 1998, eds. A. Roy and P. Smith, pp. 246-248. London: Interna- tional Institute for Conservation of Historic and Artistic Works. Nadolny, J. 2003. Reviews in Conservation 4:39-51. Olin, J. S., ed. 1982. Future Directions in Archaeometry. A Round Table. Washington, D.C.: Smithsonian Institution. Phillips, Q. E. 2000. The Practices of Painting in Japan, 1475-1500. Stanford, Calif.: Stanford Univer- sity Press. Plenderleith, H. J. 1998. Studies in Conservation 43:129-143. Rees-Jones, S. 1990. Studies in Conservation 35:93-101. Shimizu, Y. 1981. Archives of Asian Art 34:20-47. Tennent, N. 1997. In British Museum Occasional Papers, 116: The Interface between Science and Con- servation, ed. S. Bradley, pp. 15-23. London: The British Museum. Watkins, S. C. 1997. In British Museum Occasional Papers, 116: The Interface between Science and Conservation, ed. S. Bradley, pp. 221-226. London: The British Museum. Ziman, J. 1987. Knowing Everything about Nothing. Specialization and Change in Scientific Careers. Cambridge: Cambridge University Press.

Material Innovation and Artistic Invention: New Materials and New Colors in Renaissance Venetian Paintings

Barbara H. Berrie National Gallery of Art, Washington, D.C. and Louisa C. Matthew Department of Visual Arts, Union College, Schenectady, N.Y

Sixteenth-century Venetian painters have been regarded as “colorists” since their own time. The phrase “Venetian palette” is used today by art historians to describe the colors used by Renaissance painters of Venice, among whom Titian, Giovanni Bellini, and Tintoretto are the most famous. There is in fact little written consensus about how to define this so-called Venetian palette, but our knowledge is continually expanding thanks to scientific research on these artists’ paintings. One color has always been mentioned as being particularly Venetian: a rich deep orange, used generously by Venetian painters from about 1490. These artists

used the arsenical sulfides yellow orpiment (As2S3) and orange realgar (As4S4) to achieve this color. Until the end of the fifteenth century this pair of minerals had been largely confined to the miniaturists’ palette, but they became so popular in sixteenth century Venetian painting that G. P. Lomazzo remarked in his 1584 treatise “burnt orpiment is the color of gold and it is the alchemy of the Venetian painters” [1]. Artists such as Giovanni Bellini used it abundantly in their paintings; for example, Bellini used it for Silenus’ robe in The Feast of the Gods (1514; reworked by Titian, 1524) (Figure 1). The analytical data we discuss here, while still fragmentary, points to a richness of materials and their innovative use by Venetian artists that is greater than imagined heretofore, and much more than simply the addition of the arsenical minerals. Recently discovered evidence has established that professional color-sellers plied their trade in Venice from the end of the fifteenth century. It appears that they existed here as much as a century earlier than in any other Italian city. These color-sellers were neither apothecaries (“speziali”) nor general grocers from whom artists had purchased their painting supplies throughout the middle ages and

MATERIAL INNOVATION AND ARTISTIC INVENTION 13

FIGURE 1 The Feast of the Gods, Giovanni Bellini and Titian, 1514/1529, oil on canvas, (National Gallery of Art, Washington, D.C. 1942.9.1).

early Renaissance. They were sources who specialized in materials used in the arts and trades that dealt with color and color manufacturing. Some of the most interesting and useful evidence for the existence of professional color-sellers takes the form of inventories of the contents of their shops. The earliest found so far dates to 1534 [2]. Another, longer inventory of a color-seller’s shop dated 1596 has been found and published [3]. Examination of the materials in the 1534 inventory and investigation of their uses, particularly in glass-making and ceramics, coupled with our new analyses, reveal relationships that encompass both tradition and innovation. There is evidence for more cross-fertilization of technological know-how and taste among artisan industries than previously sup- posed. In this paper we will show how the information from the inventories combined with new analytical data has been used to expand our knowledge and understanding of the materials used by painters in Venice and add to the complexity of the definition of the Venetian Renaissance palette.

14 SCIENTIFIC EXAMINATION OF ART

The 1534 inventory lists 102 items; weights or amounts are given but no monetary values. Many of the materials on the inventory have an established connection with the easel painters’ art, including, for example, the pigments azurite, vermilion, lead white, and orpiment. Kermes and brazilwood, organic extracts which were used to make red dyes as well as red paints, are listed. Other items in the “vendecolore” shop that relate to the dyers’ craft include alum for mordanting dyes, galls (for making black dyes), and various resins. The first printed book on dyeing on a commercial scale was published in Venice in 1548, titled The Plichto of Gioanventura Rosetti [4]. It was written not by a dyer but by a technologist, Gioanventura Rosetti, whose intention was to provide information on what might be termed “best practices” to benefit the Venetian Republic. The recipes in the Plichto contain many of the items on both the 1534 and the 1596 inventories, including some usually considered by historians as pigments, including orpiment, vermilion and azurite, which are described in one recipe as mineral dyes (Figure 2). The overlap between painters’ and dyers’ colorants continues to become more apparent.

FIGURE 2 Extract from “The Plictho of Gioanventura Rossetti” first published in Venice in 1548. Translated by Sidney M. Edelstein and Hector C. Borghetty, The MIT Press (1969).

MATERIAL INNOVATION AND ARTISTIC INVENTION 15

The Venetian glass industry, centered on Murano, one of the islands in the Venetian lagoon, was burgeoning in the late fifteenth century. By this time the glassmakers had produced a clear glass called “cristallo” after the rock crystal that had inspired its invention. Large quantities of clear and colored glass were produced for making a wide variety of objects, including tableware, goblets, glasses, and mosaic tesserae. Recipes for richly colored glass, both single-toned and multi- colored to imitate opal and chalcedony, were developed. Special, deeply-colored glass was produced for making false rubies, sapphires, and emeralds that were as intensely and beautifully colored as the real gems. In the first decades of the sixteenth century recipes for glassmaking were being compiled [5]. The Darduin manuscript provides important information on Renaissance glassmaking, and the work of the Florentine, Antonio Neri (died 1614), who wrote L’Arte Vetraria (1612), a compilation of recipes including many of sixteenth-century origin, is an invaluable source. [This recipe book was translated into English by Christopher Merrett in 1662.] For our knowledge of the Venetian glassmaking industry we also owe much to the work of the Muranese, Luigi Zecchin [6]. Materials necessary for glassmaking are found on the 1534 inventory. Recipes for glass indicate that tin and lead were required in large quantities; both of these are on the inventory. Other ingredients include tartar, mercuric chloride, borax, alum, salt, and “tuzia” (zinc oxide), as well as orpiment. These materials are also used by dyers and some by painters. The wide range of materials available at the color-seller’s shop suggests that artisans from many trades that used color went there to obtain their raw materials. The variety available in this one place prompted us to consider whether there was more cross-fertilization among artisans than previously assumed and if we might find some evidence for this in the painting practice of the Venetian artists. We reanalyzed samples from paintings in this light, looking for materials not previously recognized. Samples from several paintings by Venetian Renaissance artists were available from prior studies. They are preserved as cross-sections of the paintings mounted in bioplastic polyester/acrylate resin. For optical microscopy, a Leica DMRX polarizing light (PL) microscope was used with PL fluotar objectives. For fluorescence microscopy the light source was a mercury lamp (100W) and the D and I3 filter packs. Scanning electron microscopy (SEM) was undertaken using a JEOL 6300 equipped with an Oxford Instruments Tetra backscatter detector. For energy dispersive spectrometry (EDS) the SEM was fitted with an Oxford Si(Li) ATW detector (capable of detecting low-energy x-rays) with a resolution at the Mn kα line greater than 130 eV. The cross-sections were usually carbon-coated, but sometimes gold-palladium coatings were used. X-ray powder diffraction patterns were obtained using Philips XRG 3100 x-ray genera- tor with a copper tube. Data were collected on photographic film in a Gandolfi camera (radius 57.3 mm). Line spacings were measured against a calibrated rule and relative intensities estimated by eye.

16 SCIENTIFIC EXAMINATION OF ART

Samples from paintings by the Venetians Lorenzo Lotto (1480-1556) and Jacopo Tintoretto (1519-1594) were among the first to be re-examined. Although the samples are limited in number they already show that the range of materials used to make paint is wider than previously known. Lotto was “rediscovered” in the late nineteenth century, but it took most of the twentieth century for him to become acknowledged as a Venetian painter. Recent research on his painting technique and color palette has helped define his place in the Renaissance [7, 8]. There is little documentary information on Lotto’s early career as an artist, but it is believed that he trained in Venice and spent his first years as an independent artist there. Later, he painted in Bergamo and the Marches. He traveled a good deal, usually within the economic and political orbit of the Venetian Republic, and he returned to the city itself for several periods. Our knowledge of Lotto’s working methods is augmented by the survival of one of his account books in which he documented commissions and expenditures during the years 1538 to 1556 [9]. One particularly valuable section of the account books is an appendix of spese per l’arte (expenditures for art), where he recorded the purchase of painting supplies, among which are notes on pigments he purchased in Venice. Among Lotto’s paintings at the National Gallery of Art in Washington, D.C. is St. Catherine, signed and dated 1522 (Figure 3). St. Catherine’s dress is a glori- ous red, perhaps reminiscent of the color of expensive red cloth worn by some Venetian brides at this time. A cross-section from the sleeve (Figure 4) shows the complicated layering Lotto used to create this color. In the cross-section, we see,

from the bottom, the preparatory layer of gesso (CaSO4.2H2O in glue), used to provide a smooth surface for painting, over which many layers of paint were applied. The first layers of paint are pinks prepared from a mixture of vermilion and lead white. Lying over these are layers of transparent red paint. From fluorescence microscopy (Figure 5) it can be discerned that what appears to be a thick homogeneous paint film is in fact many layers of thin glazes of paint; there appear to be at least six layers. The same painting technique was found in two versions of another composition painted by Lotto in the same year, The Virgin and Child with Saints Jerome and Nicholas of Tolentino [8]. It was shown, using high-performance liquid chromatography, that for the version at the National Gallery, Lon- don, Lotto used both madder and insect lakes. The fluorescence of the lakes in St. Catherine’s dress implies that he used two different lakes here also. Digital dot maps of the distribution of the elements in a sample from St. Catherine obtained using SEM-EDS are shown in Figure 6. The lowest layer of paint contains mercury, confirming that Lotto used vermilion for mixing the light red underpaint. Aluminum is present throughout most of the upper layers of transparent paint glazes. This strongly suggests that the pigment is a dye laked on alumina, the traditional way to prepare insoluble pigments from dyes made from lakes. Unexpectedly, several of the layers of transparent paint contain small, rounded particles, ca. 4-8 microns in diameter. These particles appear to be very pure silica. It is difficult to obtain information on individual particles embedded

MATERIAL INNOVATION AND ARTISTIC INVENTION 17

FIGURE 3 St. Catherine, Lorenzo Lotto, oil on panel, Samuel H. Kress Collection, 1939.1.117.

in paint, owing to the comparatively large interaction volume (the volume being analyzed) in a low-density matrix such as paint made using lake pigments. EDS spectra were obtained at 20 kV and 15 kV accelerating voltage; lowering the voltage was designed to decrease the analysis volume. The spectra (Figure 7) indicate a (rather) pure form of silica; only aluminum is present, and its origin is likely the surrounding particles of red lake. Only silicon and oxygen are significant elements in line scans through the particles. Elements that would indicate this material is a glass, for example, the fluxes sodium and potassium or the stabilizers, calcium and lead, are below detectable limits. Venetian glassmaking required pure silica, which was, in this period, provided by quartzite pebbles from the Ticino River.

18 SCIENTIFIC EXAMINATION OF ART

FIGURE 4 Cross section from a dark fold in the sleeve of St. Catherine (Figure 3) near the bottom edge, photographed in reflected light.

FIGURE 5 The cross-section illustrated in Figure 4, observed using fluorescence microscopy (filter cube: Leitz I3).

MATERIAL INNOVATION AND ARTISTIC INVENTION 19

FIGURE 6 Digital dot maps of the cross-section shown in Figures 4 and 5.

Fifteenth and sixteenth century treatises suggested using crushed marble or crushed travertine as additives to give body to paints [10]. Glass has been described as a drier for paint in Renaissance treatises and has been found in some artists’ red lake paint [11]. However, the presence of silica is unexpected, and this occurrence appears to be the first finding of this material used by Italian Renais- sance painters as an extender or an agent to give body in red lake paints. The major ingredient in Antonio Neri’s recipe for “cristallo” is pebbles “pounded small, serced as fine as flower” [12] (serce is probably a variant of sarce, to sieve through a cloth). This description corresponds to the material in Lotto’s red paint, which was a ground silica.

20 SCIENTIFIC EXAMINATION OF ART

FIGURE 7 Energy dispersive spectrum of small rounded particles in the translucent red paint; obtained at 20 kV.

The artist Jacopo Robusti, called Tintoretto, worked in Venice a few decades later than Lorenzo Lotto. Tintoretto was born in that city in 1519; his father was a member of the “cittadini” class, involved in the dyeing profession. Tintoretto lived and worked in the city throughout his career, and rarely traveled. He established a family workshop that outlived him, and he worked for a wide variety of Venetian patrons. Arguably his most famous surviving work is a series of paintings executed for the Scuola Grande di San Rocco over several decades [13]. The painting Christ at the Sea of Galilee (Figure 8) is attributed to Tintoretto and dated to 1575/80. This picture presents complicated issues in understanding its structure and the artist’s painting technique since the canvas support was assembled from several pieces of fabric that had been used for painting images different from the one we see now. The infrared reflectogram of the painting

MATERIAL INNOVATION AND ARTISTIC INVENTION 21

FIGURE 8 Christ at the Sea of Galilee, Jacopo Tintoretto 1575/1580, oil on canvas, Samuel H. Kress Collection,1952.5.27.

reveals that at some point the largest, central piece of canvas had been used to begin a portrait. The portrait had been sketched out using a wash of dark paint, clearly imaged in the infrared. The x-radiograph reveals that the canvas had also been used for a landscape that is of a different scale from both the portrait and the current image. Tintoretto’s painting techniques have been well studied [14, 15]. An investigation into the materials used for the Gonzaga cycle (1577-1578) showed that the artist employed a diverse palette [16]. Here we restrict the discussion to two pigments found in Christ at the Sea of Galilee that have special relevance to the use of glassy materials for pigments. A cross-section obtained from the sea at the right-hand side of the boat is shown in Figure 9. The bottom layer of the section appears to relate to the landscape observable in the x-radiograph. The pigment is a green, transparent, glassy- appearing pigment. The particle shape and size is similar to that of the blue glass pigment smalt (a potassium silicate colored by small amounts of cobalt). Al- though the term “smalt” is used in English today to describe only a blue glass pigment, reading the contemporary documents shows that artists of the sixteenth century used this term to describe not only blue but also numerous other

22 SCIENTIFIC EXAMINATION OF ART

FIGURE 9 Cross section from the sea near the right hand side edge of the boat. The bottom layer contains a green glassy pigment.

colored glasses, including yellow, white, and green, at least some of which may have been used by painters [17, 18]. The backscatter image of this section is shown in Figure 10. The greenish pigment in the bottom layer appears dark gray, and therefore we can infer it is of low atomic weight. The EDS spectrum of the pigment shows that it has a composition very similar to blue smalt (Figure 11). An anonymous Venetian glassmaker’s recipe book dating to early-mid sixteenth century has recipes for green glass that have the same general composition as blue smalts: “Per fare smalto verde bellissimo. Prendi della zaffera e un po’ di manganese, pestati sottili e ben lavati e di questi prendi 2 libbre, aggiungi 3,5 libbre di pani cristallini e fa fondere in forno.” [To make a beautiful green glass. Take some zaffre (an impure cobalt ore), grind it fine and wash well and of this take 2 lbs, add 3.5 lbs of crystal frit (a potash glass) and melt in the furnace” [5]. This green smalt in Christ at the Sea of Galilee contains an impurity of bismuth. Bismuth has been found in late-fifteenth and early-sixteenth Venetian enamels and in fifteenth century cobalt blue enamels and smalt in a south Ger- man painting [19]. Bismuth is an impurity in the cobalt ore from Germany, and its presence in this pigment suggests that the source of the raw cobalt-containing material, “zaffera,” used for making this glass was from north of the Alps. The spectrum shows that the glass contains iron. Iron can give rise to a yellow glass. Therefore the green color of this pigment might arise from a mixture at the microscopic level of blue and yellow glasses.

MATERIAL INNOVATION AND ARTISTIC INVENTION 23

FIGURE 10 Backscatter electron image of the sample in Figure 9.

FIGURE 11 Energy dispersive spectrum of the green pigment in the bottom layer of the section illustrated in Figure 9.

24 SCIENTIFIC EXAMINATION OF ART

A yellow pigment is used widely in Christ at the Sea of Galilee. In a cross- section from Christ’s drapery it can be seen mixed with green earth for the sea painted under Christ’s red robe and as an intense yellow layer under the greenish paint of the sea. It was also used, well mixed with green earth and azurite, for the hills in the background. At first glance the pigment appears to be lead tin yellow

type II (Pb(Sn,Si)O3). SEM-EDS clearly indicates that the colorant is an opaque yellow glass composed of particles of lead tin oxide suspended in a glassy matrix. X-ray powder diffraction (XRD) reveals that the yellow opacifier is similar but not identical to the material usually characterized in paintings. The XRD pattern of the pigment is given in Table 1. Although the pattern is very close to that

published for PbSnO3, there are some subtle differences and additional lines not attributable to expected impurities. The compendia of recipes for making glass give several variations for the yellow colorant, which likely cause different hues. It would be interesting to compare the XRD pattern of the colorant and the composition of the glassy matrix of the pigment in this painting with those of enamels on metals and glazes on majolica and relate the results to the contemporary recipes. By comparing the details of these materials we may be able to shed further light on the variety of yellows that was available for the ceramic decorators and used by easel painters to increase the range of their palette. A recent paper differentiates between the production of lead tin yellow pigment and the “raw” material for the production of yellow glass [20]. This difference might be found among the materials used by Venetian artists and craftsmen. Thus the glassy matrix might be important, and this and other differences between glasses and pigments might be the source for the variety of materials and colors that painters used. Many of the materials we find on the 1534 (and the 1596) inventory are materials used by dyers, glassmakers, and glass and maiolica painters. Some of these, including vermilion, kermes, brazilwood, orpiment, and lead white, are expected in paintings by Bellini, Giorgione, and Titian. The re-analysis of samples from pictures by these and other Venetian artists has begun to indicate that the palette they used was enriched by materials that until then had only been used by artisans and artists working in other media. Venetian painters (and others influenced by them) boldly incorporated into their work, to vivid effect, colorants not specifically designed for use in oil paint. We see that artists were using glassy materials and/or “smalti” more often and in greater diversity than we previously thought. Among these materials there appear to be frits and colorants designed for glass-painters and majolica decorators, in addition to the powdered glass, blue smalt and lead tin yellow type II, which have been identified previously. The presence of the professional color-seller in Venice might have been the catalyst and the conduit for the transfer of materials among the arts and contributed to the emergence of the Venetian palette, a palette that cannot be precisely defined, but is characterized by its complexity and diversity of colorants.

MATERIAL INNOVATION AND ARTISTIC INVENTION 25

TABLE 1 d-Spacings and Estimated Intensities of Lines in the Diffraction Pattern of the Glassy Yellow Pigment in Tintoretto’s Christ at the Sea of Galilee

and patterns for PbSnO3 and SnO.

Yellow PbSnO3 SnO Pigment ICDD 17-607 ICDD 24-1342

d I/Imax d I/Imax d I/Imax Angstroms

6.17 18

4.65 4.5* 4.32 4.20* 3.93 3.63* 3.50 w 3.30* 3.25 3.22 12 3.10 100 3.09 100 2.98 20 2.9 80 2.85 2.77 20 2.78 80 2.69 80 2.61* 50 2.63 100 2.46 2.45 2.45 12 2.30* 2.21 10 2.24 10 2.10 5 2.12 10 2.05 2.06 6 1.95 10 1.95 30 1.90 80 1.89 75 1.864 65 1.83 25 1.61 80 1.61 80 1.61 20 1.54 1.52 16 1.23 1.227 29 1.195 1.196 16

*These lines can be attributed to lead white (International Committee for Diffraction Data 13-131).

ACKNOWLEDGEMENTS We are grateful to the Center for Advanced Study in the Visual Arts (National Gallery of Art, Washington, D.C.) where we held a Samuel H. Kress Paired Fel- lowship. We benefited from discussions with members of the scientific research department of The National Gallery, London, and particularly acknowledge stimulating discussions with Jo Kirby-Atkinson.

26 SCIENTIFIC EXAMINATION OF ART

REFERENCES

1. Lomazzo, G.P., Trattato dell’arte della pittura. 1590, Milan: Paolo Gottardo Ponto. 2. Matthew, L.C., ‘Vendecolori a Venezia’: the reconstruction of a profession. The Burlington Maga- zine, 2002. CXLIV (1196): pp. 680-686. 3. Krischel, R., Zur Geschichte des Venezianischen Pigmenthandels - Das Sortiment des Jacobus de Benedictus a Coloribus, in Sonderuch aus dem Wallraf - Richartz - Jahrbuch Band LXIII 2002. 2002, Cologne: Dumont Literatur und Kunst Verlag. pp. 93-158. 4. Rosetti, G., Plictho de l’arte de tentori. 1548. Translated by Sidney M. Edelstein and Hector C. Borghetty, 1969. Cambridge, Massachusetts: The M.I.T. Press. 5. Moretti, C. and T. Toninato, Ricette vetrarie del Rinascimento: Trascrizione da un manoscritto anonimo veneziano. 2001, Venice: Marsilio. 6. Zecchin, L., Vetro e Vetrai di Murano. Vol. 1-3. 1987-1989, Venice: Arsenale. 7. Lazzarini, L., et al., Pittura veneziana: materiali, techniche, restauri. Bollettino d’Arte, 1983. 5: pp. 133-166. 8. Dunkerton, J., N. Penny, and A. Roy, Two paintings by Lorenzo Lotto at the National Gallery. National Gallery Technical Bulletin, 1998. 19: pp. 52-63. 9. Lotto, L., (Libro di spese diverse [1538-1556] con aggiunta di lettere e d’altri documenti.), P. Zampetti, editor. 1969, Venice, Rome. See also Bensi, P., Studi di storia dell’arte, 5 1983-1985, 63. 10. Merrifield, M.P., Medieval and Renaissance Treatises on the Arts of Painting. 1999, Mineola, NY: Dover. p. clii. 11. The Painting Technique of Pietro Vanucci, Called Il Perugino Editors. B. G. Brunetti, C. Seccaroni, A. Sgamellotti, Nardini Editore, 2003. Papers from the conference, 14-15 April, 2003. 12. Merrett, C., The World’s Most Famous Book on Glassmaking ‘The Art of Glass’ by Antonio Neri, M. Cable, editor. 1662, Sheffield: The Society of Glass Technology reprint 2003. (Neri’s book had been first published in Italian in 1612.) 13. Krischel, R., Jacopo Tintoretto. 2000, Cologne: Könemann. 14. Plesters, J. and L. Lazzarini. Preliminary Observations of the Technique and Materials of Tintoretto in Conservation of Paintings and the Graphic Arts. 1972, Lisbon Congress: International Institute for Conservation. 15. Plesters, J. and L. Lazzarini. I materiali e la tecnica dei Tintoretto della scuola di San Rocco, in Jacopo Tintoretto nel quarto centenario della morte. 1994, Venice: Il Polygrafo. 16. Burmester, A. and C. Krekel, “Azurri oltramarini, lacche et altri colori fini”: the quest for the lost colours, in Tintoretto: The Gonzaga Cycle, C. Syre, editor. 2000, Munich: Hatje Cantz Publishers. pp. 193-211. 17. Venturi, A., I due Dossi documenti - prima serie. Archivio Storico dell’Arte Nuovi Documenti, 1892. Anno 5 (Fase VI): pp. 440-443. 18. S. Pezzella, Il trattato di Antonio da Pisa sulla fabricazione delle vetrate artitiche, 1976. Perguia: Umbria Editrice. 19. Darrah, J.A. Connections and Coincidences: Three Pigments. in Historical Painting Techniques, Materials, and Studio Practice. 1995, University of Leiden, the Netherlands: The Getty Conservation Institute. 20. Heck, M., T. Rehren, and P. Hoffmann, The Production of Lead-Tin Yellow at Merovingian Schleitheim (Switzerland). Archaeometry, 2003. 45(1): pp. 33-44.

The Scientific Examination of Works of Art on Paper

Paul M. Whitmore Research Center on the Materials of the Artist and Conservator Carnegie Mellon University Pittsburgh, Pennsylvania

ABSTRACT The scientific examination of works of art on paper utilizes tools from the very simple to state-of-the-art analytical instrumentation, depending in large part on the question that is the objective of the investigation. Identify- ing pigments or paper fibers is straightforward, constrained only by the size of the samples that can be removed for destructive analysis. Inks are more difficult because of the lack of pronounced chemical differentiation between the ink types and because of possible interferences in the analyses from the paper substrate. Paper can be characterized easily to an extent, in identifying a watermark or the risk of deterioration from a high acid content, but the monitoring of the condition and degradation of paper remains an extremely difficult challenge. The assessment of light sensitivity, which is not easy to determine by merely identifying material composition, has been made straightforward by the development of a device that allows rapid, essentially nondestructive fading tests. Those tests are now being exploited to survey groups of objects to determine whether one may make generalizations about their exhibition needs. The further adaptation of nondestructive or micro-scale destructive analytical tools in the study of works of art on paper promises to allow even more extensive investigations of the creation and preservation of these objects.

28 SCIENTIFIC EXAMINATION OF ART

INTRODUCTION The scientific study of works of art on paper shares common objectives with the technical studies of any work of art. Artifacts are examined in order to answer art historical questions about the origin of a work, namely, where, when, and by whom a work was created. The scientific examinations seeking to answer these questions generally require identification of the materials and working methods used to craft the object. Other studies seek to answer basic questions about the care of the artifact: its physical and chemical condition, causes for deterioration, and vulnerability to storage or exhibition conditions. Technical studies of paper-based artifacts tend to resemble the study of paintings, because many paper objects actually are paintings that just happen to be executed on a paper support. Manuscript illuminations, watercolors, litho- graphic prints—these objects could easily be viewed as paintings, amenable to analyses of the colorants, paint media, or layer structure of paints observable in cross-sections. Apart from the occasional thinness of the paint layer itself, as in watercolor paintings, or binder-poor paint layers, such as in pastels, these paper- based paintings can often be studied as one would study any other painting. Despite this similarity, many works of art on paper present special circumstances that constrain analyses or warrant unusual examination techniques. Paper artifacts tend to be small: The sheets were traditionally made in molds that could be manipulated by people, and these sheets were then cut down for use. Thus, books, prints, watercolors, and other paper-based art are relatively small, meant for close-up viewing within an arm’s length. For this reason, analytical methods that require removal of paint samples are often not feasible, for the damage to the artifact can sometimes be visible upon close inspection. Nondestructive tools, particularly optical spectroscopic or imaging techniques, are more widely used to study these objects. Another distinction between paper-based objects and traditional paintings is the use of the paper substrate as part of the image itself. Particularly with such graphic art as drawings and prints but also with printed text or even thinly painted watercolors, the paper substrate is exposed and is part of the image. Thus, the color of the paper and its surface texture are important contributors to the appearance and visual appeal of the object, and study of the paper and its preservation is of great importance. (Occasionally in historical times and more frequently in the twentieth century, paintings too have been created with unpainted canvas as part of the image. For these objects the concern about the appearance and stability of the canvas is of course shared.) A complication in studying objects in which the paper is so intimately associated with the drawing media is the discrimination between the two, so that many analyses must have very small spatial or depth resolution, or contributions to the detected signal from the paper must be subtracted. Paper-based collections in museums are known to pose some of the most

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 29

common preservation problems because many of the artifacts that are now prized were not created as lasting works of art but as more utilitarian objects. Because paper was inexpensive and widely available through much of history, it has seen use for many purposes, a primary one being for communication and recording of information. Some of these artifacts, such as books, were meant to last for a long time, but others, such as newspapers, announcements, or letters, were often not created with posterity in mind. Thus, it is not uncommon for museums and archives to have paper artifacts that are delicate or deteriorating because of their creation with impermanent materials or techniques. Preservation problems are common, particularly with those objects that were not made as art objects. This review will survey the examination techniques of paper-based objects that are used both for art historical investigations as well as for preservation studies. Some of those techniques are routine and can be found in many well- equipped museum laboratories; others are less widely available and have not found widespread use. This survey will conclude with a description of a relatively new tool developed to detect a particular vulnerability, the susceptibility of colored materials to fade from light exposure, and illustrate its use for the study of Japanese woodblock prints.

SURVEY OF EXAMINATION AND MATERIAL IDENTIFICATION TECHNIQUES The most common technical investigation for paintings or colored prints on paper involves identification of the pigments in the paint. For this, the routine analytical tools of polarizing-light microscopy, X-ray diffraction, and elemental analyses by X-ray fluorescence are commonly employed, usually on samples of the paint that have been removed from the artifact. Descriptions of these tools can be found in accounts of painting examinations, or in reference books devoted to pigment identification (Feller, 1986; Roy, 1993; FitzHugh, 1997). Nondestructive techniques can also sometimes be used to identify pigments on paper objects. Open-air X-ray fluorescence is used for elemental analyses of pigments, and Raman spectroscopy and Raman microscopy have been found useful for examining both pigments in paints and dyes in colored paper (Bell et al., 2000; Best et al., 1995). Some pigments have distinctive features in the visible spectrum (Schweppe and Roosen-Runge, 1986; Leona and Winter, 2001), while others, like Indian yellow, can be detected by their peculiar fluorescence observable under ultraviolet light illumination (Baer et al., 1986). Drawing materials can also be studied, although they present some difficulties. Early drawings were created using metal tools or wires as drawing imple- ments (thus the name “metalpoints” for these drawings), and they can be analyzed by measuring the elemental composition of the metals in the lines (by X-ray fluorescence, typically). Inks are more problematic, with the exception of iron gall inks, which can be distinguished by the presence of iron in X-ray fluorescence or

30 SCIENTIFIC EXAMINATION OF ART

in more unusual techniques such as Mössbauer spectroscopy (Rusanov et al., 2002) or PIXE (Budnar et al., 2001). Inks can also be analyzed for the trace elements they contain, introduced in the ink ingredients or as residues from the printing process. Inks in early books (such as a Gutenberg Bible) have been examined for these trace elements by synchrotron-excited X-ray fluorescence in the hope of distinguishing books produced in the early German printing shops (Mommsem et al., 1996). Other organic inks, such as sepia (cuttlefish ink), bistre (from soot), or such black drawing media as charcoal, bone black, lamp black, ivory black, or graphite cannot usually be distinguished by their elemental composition (although bone black is often detected by the presence of phosphorus), nor do the infrared spectra of these inks usually present characteristic features useful for their identification. Polarizing-light microscopy remains a common tool to discriminate between inks on the basis of their particle morphologies. The media used as pigment binders for drawing and painting materials can be identified by analyzing the organic composition of micro-samples. Of the various methods available the most useful are the gas chromatography/mass spectroscopy analyses that have been developed for oils and resins used as paint binders (Mills and White, 2000; Schilling and Khanjian, 1996) and more recently adapted for the study of gums used in watercolors or gouaches (Vallance et al., 1998). In addition to the study of the image-forming materials, the examination of the paper itself is also often a clue to the artifact’s origin. Paper is usually differentiated by its fiber composition, its physical characteristics, and its manufacturing method. The fibers can be studied with optical microscopy, and the plant origin of the component fibers can be determined by appearance or by reaction to certain stains, such as Hertzberg or Graff C stains. The fiber type, length, and heterogeneity can all be distinctive, as can such physical dimensions as sheet thickness. The evidence of manufacture is most easily detected in the pattern left by the papermaking mold, typically a pattern of lines called chain and laid lines in so-called “laid” paper. Watermarks, the decorative patterns often woven into the wire molds or embossed on the cast sheets, are also the most obvious characteristic patterns of the paper manufacture. The evidence of chain and laid lines and watermarks can be captured in any of a number of ways, with transmitted light photography or with beta or soft X-ray radiography, and various image processing tools have been applied to enhance such records (erasing interferences from the printing, for example) and making them more useful for indexing and re- trieval for comparison in a reference database (Brown and Mulholland, 2002). The presence of sizing (a water-resistant finish on the surface of the paper) can be determined by infrared spectroscopy or colorimetric methods (Barrett and Mosier, 1995), and fillers (typically finely ground minerals or clays added for increased opacity of the sheet) can be identified by optical or electron microscopies (Browning, 1969). The fiber, finish, or watermark, along with other historical evidence, can assist in tracing a paper’s origin (Hunter, 1978), yet many challenges remain (Slavin et al., 2001).

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 31

EXAMINATIONS TO ASSESS CONDITION AND PRESERVATION PROBLEMS As with all works of art, preserving works of art on paper focuses on main- taining both the physical integrity of the artifact and its appearance. The physical integrity is derived mainly from the paper sheet itself, and preservation of the sheet’s cohesive strength is of paramount importance. (For books and archival materials that are handled by users, other sheet properties, such as flexibility, are also important, but for works of art that are usually mounted in a frame, the physical stresses are usually merely the tensile stresses from the paper’s weight and from its reaction to temperature and humidity.) The cohesive strength of a sheet of paper is derived from the strength of its constituent fibers and of the bonds between the fibers. Aging tends to reduce the fiber strength, and old weak papers are usually seen to fail from broken fibers rather than by unraveling from weakened interfiber bonds. The reduction of fiber strength is in turn a result of the breakdown of cellulose, the natural polymer of glucose that composes plant fibers. Chemical degradation breaks cellulose chains, which reduces the average molecular weight but more importantly also breaks the connections between the highly crystalline cellulose zones. This progressive rup- ture of the tie chains, the amorphous cellulose chains connecting the crystallites and imparting the cohesive strength to the fiber, is the underlying aging chemistry leading to physical failure of the paper sheet. Unfortunately, there are no analytical tools that can allow detection of such deterioration in a paper artifact without destructive analysis of unacceptably large portions of paper. Typically for degrading polymers, nondestructive tools such as infrared spectroscopy do not have the sensitivity to detect the production of the very small concentrations of new chain ends in the degrading cellulose. Recent studies suggest that production of glucose or xylose residues (Erhardt and Mecklenberg, 1995) or low molecular weight acids (Shahani and Harrison, 2002) may be easier to track as some measure of cellulose reaction, but these techniques have not yet been applied to artifacts. Other efforts to develop micro-scale molecular weight analyses for cellulose have reduced the amount of paper needed (Rohrling et al., 2002), but a recent molecular weight analysis of cellulose in a single paper fiber, while successful, also suggests that such small sample sizes may not be typical of the other fibers or representative of the average molecular weight of larger samples (Stol et al., 2002). Thus, even if the analytical procedure can be adapted, the slow deterioration of the cellulose may not be easily tracked by successive measurements of individual fibers over time. While the deterioration of paper artifacts may be difficult to detect directly, many years of investigation of cellulose degradation have clearly indicated that there are other materials that may be reliable indicators of instability in the paper. Acidity is well established as a catalyst for the hydrolytic breakdown of cellulose, the most important of the known degradation chemistries. Lignin is primarily

32 SCIENTIFIC EXAMINATION OF ART

responsible for the discoloration of groundwood papers, and iron and copper impurities can also act both as acid catalysts for hydrolysis and as catalysts for oxidative breakdown of cellulose. It is much easier to determine the presence of these sensitizing agents in paper than to track the slow deterioration of the cellulose, so the study of artifact materials often does not go beyond a pH measurement, the detection of lignin with phloroglucinol stain or an infrared spectrum, or the analysis for iron or copper impurities by a technique such as electron spin resonance spectroscopy (Attanasio et al., 1995). Iron present not as a paper impurity but as an ink component is also a well-known and easily identifiable risk factor for the preservation of manuscript and print materials. The maintenance of the image-forming materials, particularly the colored paints and inks used to create the image, is another objective of preservation strategies. Light exposure is the most common hazard to the pigments and dyes used on art objects. In contrast to preserving the paper support, in which the deterioration is difficult to monitor and easier to predict by detecting the presence of destabilizing components, the loss of color is easy to monitor by periodic color measurements but difficult to predict. The light stability of a pigment depends not only on the material but also on its preparation, particle size, and prior fading history. None of these is easy to determine from study of the pigments, and until recently the only means to detect light sensitivity was to monitor the damage inflicted by light exposure. Recently a new device has been developed to determine the risk of future fading from light exposure (Whitmore et al., 1999). That device operates as a reflectance spectrophotometer using a very intense focused beam from a xenon lamp as the illumination for the measurement (see Figure 1). By making rapid repeated spectral measurements while the material is illuminated by the intense light, very slight degrees of fading can be detected in light-sensitive materials in only a few minutes (see Figure 2). Because of the high precision of the spectrum acquisition, extremely small amounts of fading are easily recorded, and the test can be stopped before perceptible changes to the art object have been produced. All of the different color areas on a work of art can be tested, and the overall sensitivity of the object can be judged by the fading rate of the most light-sensitive color (see Figure 3). These tests can be used to develop exhibition requirements that are tailored to the needs of the object, with the very light-sensitive objects receiving greater care (less frequent exhibition at lower light levels) so that they do not suffer from fading damage caused by inappropriate display. These same tests, done with filtered illumination, can also be used to test the effectiveness of different lighting in reducing fading rates. By performing the tests in air or under an inert gas, the efficacy of oxygen-free housings for slowing the fading of works of art can also be assessed (see Figure 4). It has been found that this fading test can also be used to identify pigments, not by their elemental or chemical constitution but rather on the basis of their photochemical reaction. Prussian blue, a ferric ferrocyanide complex used in art

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 33

FIGURE 1 Schematic of fading tester. Reprinted from the Journal of the American Institute for Conservation, vol. 38, no. 3, with the permission of the American Institute for Conser- vation of Historic and Artistic Works, 1717 K St., NW, Suite 200, Washington, D.C. 20006.

Bengal Rose

20 Rose Tyrien Magenta E ∆

15 Rose Malmaison

10 CIE Color Difference,

5 Blue Wool 1

Geranium

0 012345 Time (minutes)

FIGURE 2 Fading test results for selected Winsor & Newton gouache paints. “Blue Wool 1” designates fading test results for the ISO Blue Wool no. 1, the most light-sensitive of the standard cloths. Reprinted from the Journal of the American Institute for Conservation, vol. 38, no. 3, with the permission of the American Institute for Conservation of Historic and Artistic Works, 1717 K St., NW, Suite 200, Washington, D.C. 20006.

34 SCIENTIFIC EXAMINATION OF ART

5 BW2

4 E ∆

2 CIE Color Difference,

1 BW3

0 012345 Time (minutes) FIGURE 3 Fading test results for all the different color areas on a Japanese woodblock print (Yoshitoshi, Carnegie Museum of Art No. 89.28.1516). “BW2” and “BW3” desig- nate the degree of color difference produced after five minutes in fading tests of ISO Blue Wool fading standards nos. 2 and 3.

since the early eighteenth century, is known to fade reversibly during light exposure, with the blue color being recovered in a subsequent dark reaction (Ware, 1999). The fading tests of Prussian blue using the tester described above demonstrated this peculiar reversible fading behavior on a cyanotype, an early photographic process used to create blueprints (see Figure 5) (Whitmore et al., 2000). In addition to these fading tests designed to evaluate individual artifacts, current studies are measuring the fading rates of particular colorants in Japanese woodblock prints from different eras, printed at different depths of color, and of varying degrees of prior fading. Results of such a population study will reveal whether there is general consistency or a wide variation in light sensitivity for particular materials. If there is widely varying behavior, fading tests must be performed on each object in order to determine the sensitivity and light exhibition needs. If there are very similar fading rates among the different applications of a pigment, one need not test every object and can instead safely use a rule of thumb to make such judgments of the object’s required care. This formulation of new rules of thumb, based on actual fading sensitivities observed in a large population of objects, will bring a new level of intuition about how to preserve objects. The results of tests on a large number of Japanese woodblock prints indicate that the fading of the colorant dayflower blue (aobana) is very regular, and its sensitivity can probably be safely estimated without individual fading tests (see Figure 6a). By contrast, yellow passages on the Japanese prints vary greatly in their light

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 35

(a) 18

14 ∆ 12

6 CIE Color Difference, E 4

0 0246810 Time (minutes)

(b) 6

5 ∆ 4

2 CIE Color Difference, E

0 0246810 Time (minutes)

FIGURE 4 Fading test results in air (solid lines) and under nitrogen (dashed lines). (a) Results for a gouache paint (Winsor & Newton Rose Bengal), showing slower fading in anoxic environment. (b) Results for ISO Blue Wool cloth no. 1, showing no difference in fading rate in anoxic environment. Reprinted from the Journal of the American Institute for Conservation, vol. 38, no. 3, with the permission of the American Institute for Conser- vation of Historic and Artistic Works, 1717 K St., NW, Suite 200, Washington, D.C. 20006.

36 SCIENTIFIC EXAMINATION OF ART

10 ∆ 8

4 CIE Color Difference, E

0 0.0 0.2 0.4 0.6 0.8 1.0 Cumulative Dose (Million Lux-hours)

FIGURE 5 Reversible fading of Prussian blue under exposure in fading tester. Solid lines are fading measured during light exposure; dashed lines represent period of recovery, and return of blue color (smaller color difference) in the dark. Reprinted from Tradition and Innovation: Advances in Conservation, eds. A. Roy and P. Smith, with the permission of the International Institute for Conservation of Historic and Artistic Works, 6 Buckingham St., London WC2N 6BA, UK.

sensitivity, probably because many different kinds of natural colorants were used in the printing (see Figure 6b). These materials will require individual testing in order to assess their fading risks.

CONCLUSION The scientific examination of works of art on paper utilizes tools from the very simple to state-of-the-art analytical instrumentation, depending in large part on the question that is the objective of the investigation. Identifying pigments or paper fibers is relatively easy, while inks are more challenging because of the lack of pronounced chemical differentiation between the ink types and because of possible interferences in the analyses from the paper substrate. Paper can be characterized easily to an extent, in identifying a watermark or the risk of deterioration from a high acid content, but the monitoring of the condition and degradation of paper—or for that matter, any polymeric material—remains an ex-

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 37

(a) 5 BW2

4 ∆

2 CIE Color Difference, E 1 BW3

0 012345 Time (minutes)

(b) 5 BW2

4 ∆

2 CIE Color Difference, E 1 BW3

0 012345 Time (minutes)

FIGURE 6 Fading test results for (a) 36 dayflower blue (aobana) passages on 25 different Japanese woodblock prints in the collection of the Carnegie Museum of Art; and (b) 55 yellow passages on 48 prints from that collection. “BW2” and “BW3” denote the color change produced in ISO Blue Wool fading standards nos. 2 and 3, respectively, after a five-minute exposure in the fading tester.

38 SCIENTIFIC EXAMINATION OF ART

tremely difficult challenge. The assessment of light sensitivity, which is not easy to determine by merely identifying material composition, has been made straightforward by the development of a device that allows rapid, essentially nondestructive fading tests. Those tests are now being exploited to survey groups of objects to determine whether one may make generalizations about their exhibition needs. The further adaptation of nondestructive or micro-scale destructive analytical tools in the study of works of art on paper promises to allow even more extensive investigations of the creation and preservation of these objects.

REFERENCES

Attanasio, D., D. Capitani, C. Federici, and A. L. Segre. 1995. Archaeometry 37:377-384. Baer, N. S., A. Joel, R. L. Feller, and N. Indictor. 1986. In Artists’ Pigments, vol. 1, ed. R. L. Feller, pp. 17-36. Cambridge, U.K.: Cambridge University Press. Barrett, T., and C. Mosier. 1995. Journal of the American Institute for Conservation 34:173-186. Bell, S. E. J., E. S. O. Bourguignon, A. C. Dennis, J. A. Fields, J. J. McGarvey, and K. R. Seddon. 2000. Analytical Chemistry 72:234-239. Best, S. P., R. J. H. Clark, M. A. M. Daniels, C. A. Porter, and R. Withnall. 1995. Studies in Conserva- tion 40:31-40. Brown, A. J. E., and R. Mulholland. 2002. In Works of Art on Paper, Books, Documents, and Photo- graphs: Techniques and Conservation, eds. V. Daniels, A. Donnithorne, and P. Smith, pp. 21-26. London: International Institute for Conservation. Browning, B. L. 1969. Analysis of Paper. New York: Marcel Dekker. Budnar, M., J. Vodopivec, P. A. Mando, F. L. G. Casu, and O. Signorini. 2001. Restaurator 22:228- 241. Erhardt, D., and M. F. Mecklenberg. 1995. In Materials Issues in Art and Archaeology IV, eds. P. B. Vandiver, J. R. Druzik, J. L. G. Madrid, I. C. Freestone, and G. S. Wheeler, pp. 247-270. Pitts- burgh: Materials Research Society. Feller, R. L., ed. 1986. Artists’ Pigments, vol. 1. Cambridge, U. K.: Cambridge University Press. FitzHugh, E. W., ed. 1997. Artists’ Pigments, vol. 3. Washington, D.C.: National Gallery of Art. Hunter, D. 1978. Papermaking: The History and Technique of an Ancient Craft. New York: Dover. Leona, M., and J. Winter. 2001. Studies in Conservation 46:153-162. Mills. J. S., and R. White. 2000. The Organic Chemistry of Museum Objects, 2nd ed. London: Butterworth-Heinemann. Mommsem, H., T. Beier, H. Dittmann, D. Heimermann, A. Hein, A. Rosenberg, M. Boghardt, E.-M. Hanebutt-Benz, and H. Halbey. 1996. Archaeometry 38:347-357. Rohrling, J., A. Potthast, T. Rosenau, T. Lange, G. Ebner, H. Sixta, and P. Kosma. 2002. Biomacromolecules 3:959-968. Roy, A., ed. 1993. Artists’ Pigments, vol. 2. Washington, D.C.: National Gallery of Art. Rusanov, V., K. Chakarova, and T. Madolev. 2002. Applied Spectroscopy 56:1228-1236. Schilling, M. R., and H. P. Khanjian. 1996. In Preprints of the 11th Triennial Meeting of the ICOM (International Council of Museums) Committee for Conservation, ed. J. Bridgland, pp. 220-227. London: James & James Ltd. Schweppe, H., and H. Roosen-Runge. 1986. In Artists’ Pigments, vol. 1, ed. R. L. Feller, pp. 17-36. Cambridge, U.K.: Cambridge University Press. Shahani, C. J., and G. Harrison. 2002. In Works of Art on Paper, Books, Documents, and Photographs: Techniques and Conservation, eds. V. Daniels, A. Donnithorne, and P. Smith, pp. 189-192. Lon- don: International Institute for Conservation.

THE SCIENTIFIC EXAMINATION OF WORKS OF ART ON PAPER 39

Slavin, J., L. Sutherland, J. O’Neill, M. Haupt, and J. Cowan, eds. 2001. Looking at Paper: Evidence and Interpretation. Ottawa: Canadian Conservation Institute. Stol, R., J. L. Pedersoli, Jr., H. Poppe, and W. T. Kok. 2002. Analytical Chemistry 74:2314-2320. Vallance, S. L., B. W. Singer, S. M. Hitchen, and J. H. Townsend. 1998. Journal of the American Institute for Conservation 37:294-311. Ware, M. 1999. Cyanotype: The History, Science, and Art of Photographic Printing in Prussian Blue. London: Science Museum; Bradford, West Yorkshire, U.K.: National Museum of Photography, Film, and Television. Whitmore, P. M., C. Bailie, and S. Connors. 2000. In Tradition and Innovation: Advances in Conserva- tion, eds. A. Roy and P. Smith, pp. 200-205. London: International Institute for Conservation. Whitmore, P. M., X. Pan, and C. Bailie. 1999. Journal of the American Institute for Conservation 38:395-409.

Changing Approaches in Art Conservation: 1925 to the Present

Joyce Hill Stoner Professor, Winterthur/University of Delaware Program in Art Conservation Winterthur Museum Winterthur, Delaware

ABSTRACT The years between 1925 and 1975 in the United States marked a period of pioneering progress and expansion in the field of art conservation: museums established conservation departments and analytical laboratories; the first art technical journals were published; and professional societies and training programs were established. From 1975 to the present, processes were refined, choices multiplied, and procedures that had once seemed black and white became gray and variable. There was also a hands-off or minimalist movement, increased attention to preventive conservation, and a new role for the conservator as a high-level collaborator. The twenty-first-century conservator should work with museum scientists to understand the strengths and limitations of a vast array of possibilities for instrumental analysis, should collaborate with curators, archivists, archaeologists, architects, and artists, and should understand a vocabulary of technology and connoisseurship that may range from the contents of a shipwreck to Indian miniature paintings. Today’s conservator should understand integrated pest management, light levels, heating, ventilation, and air-conditioning systems, and should be able to speak articulately about the field to audiences ranging from grade school groups to museum and university trustees. Rules are flexing with regard to use and handling of Native American materials in museums, removal of ceremonial substances, and collaboration with living artists. Conservators, who were once lonely advocates for the physical materials of art works and their long-term survival,

CHANGING APPROACHES IN ART CONSERVATION 41

now must look at preservation in a much larger arena, including the cultures of origin and economic survival in the twenty-first century.

EVOLUTION OF THE FIELD OF CONSERVATION FROM 1925 TO 1975 The nineteenth century witnessed a growing collaboration between the fields of art and science. Michael Faraday made analytical and deterioration studies for the National Gallery in London, motivated by an official inquiry into the methods of cleaning paintings. He demonstrated the damaging effect on works of art of sul- phur compounds liberated by coal smoke and gas lighting and showed that deterioration increased during London fogs and high humidity. Louis Pasteur carried out analytical studies of paint in the 1870s. A scientific department was established at the Staatliche Museen in Berlin in 1888, and the British Museum followed suit in 1921. The years between 1925 and 1975 in the United States marked a period of pioneering progress and expansion in the field of art conservation. A special climate of cooperation among scientists, art historians, and restorers developed at the Fogg Art Museum in the late 1920s. One pivotal figure was Edward Waldo Forbes, the director of the Fogg from 1909 to 1944. He realized how misleading the contemporary practice of wholesale retouching of paintings could be. He encouraged technical investigation and X radiography. He was the chairman of the Advisory Committee for the first technical journal, Technical Studies in the Field of the Fine Arts, published by the Fogg from 1932 to 1942. Forbes had approached Francis P. Garvin, the president of the Chemical Foundation, for a donation to finance the publication of Technical Studies. Two significant “Fogg founding fathers” of art conservation were Rutherford John Gettens, the first chemist in the United States to be permanently employed by an art museum, and George L. Stout, the founder and first editor of Technical Studies. Gettens and Stout coauthored Painting Materials: A Short Encyclopaedia, first published in 1942 and reprinted in 1966. This useful compendium is still cited regularly, up to and including our most recent University of Delaware Ph.D. dissertation of 2002 on paint analysis in historic buildings by Susan Buck. Only a few dates and descriptions in the little Gettens and Stout book are now outdated. In 1974 Gettens presented a paper suggesting that we begin a conservation history initiative, and Stout helped launch (after Gettens’s sudden death) our Foundation of the American Institute for Conservation (FAIC) oral history project. We now have more than 150 transcribed interviews with pioneer conservation professionals. Stout lectured about the history of the field from 1925 to 1975 at our American Institute for Conservation meeting in Dearborn, Michigan, in 1976, illustrating his remarks with his own watercolors presenting the evolving profession (see Figure 1). Stout noted that before the Fogg launched its technical laboratory in 1928 featuring the collaboration of art historians, scientists, and practicing conserva-

42 SCIENTIFIC EXAMINATION OF ART

FIGURE 1 Watercolor by George L. Stout illustrating the state of the field of art conservation and restoration in 1925.

CHANGING APPROACHES IN ART CONSERVATION 43

tors that he dubbed “the three-legged stool,” there had been lone “technological investigators” in some museums (see Figure 2). Alan Burroughs (1897-1965) of the Fogg traveled to major museums in Europe with an old Picker X-ray machine in 1926, making landmark X radiographs of Old Master paintings. Burroughs published his findings in 1938 as Art Criticism from a Laboratory. X radiographs represented the paragon of technical investigation for paintings and sculpture for the first half of the twentieth century, accompanied by examination with ultraviolet light. X rays had been discovered in 1895 by Roentgen, and paintings were first x-rayed within a year. Christian Wolters, one of the pioneer conservators of Ger- many, wrote his dissertation on the importance of radiography for art history in 1936. One issue of the Philadelphia Museum Bulletin of 1940 featured a solemn, worshipful photograph of an X-ray unit as its cover image. By 1931 James Joseph Rorimer (1905-1966) of the Metropolitan Museum of Art had published Ultra- violet Rays and Their Use in the Examination of Works of Art. Stout described the 1920s as the great days of Berenson. There was contempt for concern about condition. That was as naughty as to inquire about the digestive system of an opera singer. You didn’t look into those things—it wasn’t proper. And that was very good for the trade. When a dealer sold a picture, he didn’t like to have anyone consider for a moment that the condition of that picture had anything to do with the case. This was something of beauty and your sensitivities for the quality of the beauty were the important matter. Whether it was about to buckle up and begin to give you hell in another couple of years was something never to be considered (see Figure 3).1 There was no luxury of specialization during the early days of the three- legged-stool collaboration at the Fogg. Everyone worked on everything; the conservators and scientists brainstormed, took notes, documented, collected pigments, and painted out samples of paints on the walls. No one was full time; they even did hit-and-run paint chip analysis for the Harvard police. Forbes and later Stout taught young art historians about historical artists’ techniques, insisting that they paint in egg tempera and fresco themselves; these students went off to become curators and directors in museums throughout the United States, bring- ing with them a unique concern for connoisseurship and the physical presence of works of art. In the early 1950s members of the original Fogg team of conservators and conservation scientists were dispersed, largely because of funding issues and the attitude of the administration at that time, according to Richard Buck.2 Gettens next founded the technical laboratory at the Freer Gallery of Art at the Smithsonian Institution in Washington, D.C., in a humble space once used for packing crates, and Richard Buck helped to design the first Regional Conserva- tion Laboratory, the Intermuseum Conservation Association (ICA) at the Allen Art Museum in Oberlin, Ohio, which opened in 1953. The ICA was founded by six major Midwest museums to provide professional and cost-effective art con-

44 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 Watercolor by George L. Stout illustrating the “technological investigator” in museums in the early twentieth century. At the Fogg Art Museum it was Alan Burroughs.

CHANGING APPROACHES IN ART CONSERVATION 45

FIGURE 3 Watercolor by George L. Stout illustrating the “Berensonian” dealer/merchant figure of the 1920s.

46 SCIENTIFIC EXAMINATION OF ART

servation services. There are now 12 regional centers, and in 1997 this group began a consortium known as RAP (Regional Alliance for Preservation) with a website (http://www.rap-arcc.org). Major professional societies and training programs also appeared between 1925 and 1975. The International Institute for Conservation of Historic and Ar- tistic Works (IIC) was incorporated under British law in 1950 as “a permanent organization to co-ordinate and improve the knowledge, methods, and working standards needed to protect and preserve precious materials of all kinds.” Since 1967 when the first meeting on climate control was held in London (see Figure 4), triennial or biennial congresses have been held in international locations on special topics ranging from archaeological conservation (in Stockholm, Sweden, in 1975, with attention to the recovery of the warship Wasa) to library and archive conservation (Baltimore, Maryland, in 2002, with accompanying visits to the paper conservation department of the Library of Congress). Harold Plenderleith (1898-1997), founding member and author of what was once considered the conservation “Bible,” The Conservation of Antiquities and Works of Art: Treat- ment, Repair, and Restoration (1956) noted, “We never envisaged more than 50

FIGURE 4 The first triennial meeting of the International Institute for Conservation, in front of Albert Hall, London, 1967.

CHANGING APPROACHES IN ART CONSERVATION 47

fellows because there weren’t 50 fellows in existence who shared our ideas or were adequately experienced in scientific conservation.” As of June 2002 there were almost 2,400 individual members of this international body, including 301 fellows. (This figure does not fairly represent the international population of conservators, as many now elect to join only their regional groups because of economic concerns and ready availability of key professional information on the World Wide Web; more information about IIC can be found at http:// www.iiconservation.org/.) IIC began publication of its quarterly journal Studies in Conservation in 1952 and sponsored an international abstracting periodical IIC Abstracts in 1955. Tech- nical Studies published by the Fogg from 1932-1942 had also contained abstracts of the international literature, and Gettens compiled Abstracts of Technical Studies in Art and Archaeology (published by the Freer in 1955), a fairly slender volume covering literature published between 1943 and 1952. (Stout and other conservators had been active elsewhere during World War II as officers in the Arts and Monuments initiative to identify and protect cultural heritage.) IIC Abstracts moved to the New York University (NYU) Conservation Center in 1966, and the name of the publication was changed to Art and Archaeology Technical Abstracts (AATA). AATA moved to the Getty Conservation Institute in 1985 (see Figure 5), and became available online in 2002 (http://aata.getty.edu/NPS). The conservation literature vastly expanded between 1975 and 1990. When this author was a

FIGURE 5 The Editorial Board of Art and Archaeology Technical Abstracts, semiannual meeting in 1988, held at the Canadian Conservation Institute.

48 SCIENTIFIC EXAMINATION OF ART

FIGURE 6 The first meeting of the International Institute for Conservation-American Group at the Isabella Stewart Gardner Museum in 1960.

graduate conservation student at the NYU Conservation Center in the 1960s, very few books were available on the topic of art conservation. Current conservation graduate students can readily spend our new Gutmann Foundation grants of $3,000 on books in their specialties; I doubt I could have spent $300 in 1968. The American Group of the IIC began in 1960 (see Figure 6) and became the American Institute for Conservation (AIC) in 1972, publishing a journal and a newsletter and establishing a 501(c)(3) foundation (FAIC). The AIC now has about 3,000 members, including 860 fellows or professional associates (http:// aic.stanford.edu). The National Conservation Advisory Council (NCAC) was organized in 1973, funded by the National Museum Act of the Smithsonian Institu- tion. The NCAC surveyed needs of the field and published useful and colorful booklets and became the National Institute for the Conservation of Cultural Prop- erty (NIC) in 1982. Whereas individual conservators are members of the AIC, the NIC is composed of conservation associations that meet to discuss overarching issues, such as preservation of outdoor sculpture throughout the United States and historic houses and their contents. In 1997 the NIC changed its name to Heritage Preservation and now has 156 institutional members (http:// www.heritagepreservation.org/).

CHANGING APPROACHES IN ART CONSERVATION 49

Major conservation research laboratories were also founded during this period. William J. Young sponsored seminars in 1958, 1965, and 1970 on “The Application of Science in Examination of Works of Art” at the Museum of Fine Arts, Boston; the laboratory there had begun in 1930. In 1950 the National Gal- lery of Art in Washington, D.C. established a fellowship at the Mellon Institute in Pittsburgh. Robert Feller focused on natural and synthetic picture varnishes, color, and the damaging effects of light exposure. In 1976 the research project was reorganized as the Research Center on the Materials of the Artist and Conservator at Carnegie Mellon University. In 1988 Feller retired and Paul Whitmore came to the center to become the director. Whitmore edited an excellent compilation of Feller’s research, Contributions to Conservation Science, published in 2002. Origi- nally established in 1963, principally to provide technical support to the Smithsonian museums in the analysis and conservation needs of the collections, the Conservation Analytical Laboratory (CAL) moved in 1983 to the Museum Support Center in Suitland, Maryland (see Figure 7). In January 1998 the Board of Regents of the Smithsonian Institution voted to rename CAL the Smithsonian Center for Materials Research and Education (SCMRE). The Getty Trust and its various branches is our most significant recent addition; major activities were defined in 1982. In the summer of 1985 the Getty Conservation Institute took up

FIGURE 7 The Conservation Analytical Laboratory building of the Smithsonian Institu- tion (now known as SCMRE, Smithsonian Center for Materials Research and Education).

50 SCIENTIFIC EXAMINATION OF ART

quarters in rented warehouse space at the Marina del Rey, and in 1997 joined other Getty branches at the Getty Center in California, an acropolis in travertine on the hills of Brentwood, representing the apotheosis of the architecture with dedicated space for conservation. Conservation training evolved from apprenticeship to formal education and training programs in London in 1934, Vienna in 1936, Munich in 1938, Rome in 1943, and New York in 1960. The other two U.S. fine arts graduate programs opened in 1970 and 1974. As of 1994 there were at least 50 programs in 30 countries in the fine arts and another 50 in archaeological materials, books, decorative arts, and musical instruments according to a directory co-published by the Getty Conservation Institute. The Conservation Center at the Institute of Fine Arts of New York University accepted its first four graduate students in 1960; we students studied in the basements of the Duke mansion at One East 78th Street. In 1983 the program moved across the street to a converted brownstone with 11 floors of conservation laboratories, classrooms, and libraries; eight students are now accepted each year. Ten students were accepted annually into the Cooperstown Graduate Program in the conservation of historic and artistic works beginning in 1970. This program relocated to Buffalo State College in 1987. In 1974 a third graduate program was sponsored jointly by the University of Dela- ware and the Winterthur Museum. Both the University of Delaware and Buffalo State College have awarded space to their art conservation departments in the flagship buildings of their institutions, another encouraging sign of the architectural status now granted our profession. George Stout summarized the state of conservation in 1975 with another collage of a fence (see Figure 8). The Conserva- tion Education Program at Columbia University accepted students in 1981 for training in library and archives conservation; this program moved to the Univer- sity of Texas in 1992. The conservation training programs also have an association, ANAGPIC (Association of North American Graduate Programs in Conser- vation), including the Master of Art Conservation Program at Queen’s University in Kingston, Ontario, and the Center for Conservation and Technical Studies at Harvard. The students from the programs convene annually and present papers.3

CHANGING STYLES OF CONSERVATION: 1975 TO THE PRESENT The Fogg conservators noted in their interviews that they rarely specialized; there are now nine specialty groups listed by material in each AIC Newsletter: architecture, book and paper, electronic media, objects, paintings, photographic materials, RATS (research and technical studies), textiles, and wooden artifacts. Another group represents the particular concerns of conservators in private practice. There is also an independent group, the Society of the Preservation of Natural History Collections (SPNHC). The International Council of Museums-Committee for Conservation (ICOM-CC) (http://www.icom-cc.org) has at least 18 working groups, including an active group on preventive conservation for conservators

CHANGING APPROACHES IN ART CONSERVATION 51

FIGURE 8 Watercolor by George L. Stout illustrating the state of the field of art conservation in 1975.

52 SCIENTIFIC EXAMINATION OF ART

who specialize in climate and pest control, changing exhibition, loans, and ship- ment. All these spin-off groups are publishing, conducting meetings and workshops, and collaborating with scientists or curators, architects, librarians, and artists. Methods of analysis have become far more sophisticated than just X radiography or examination with ultraviolet light; the Sackler conference of 2003 at the National Academy of Sciences elaborated on many of these excellent resources. I will note that for my own specialty, a professional paper on paintings conservation is now rarely without cross-sections indicating the layered buildup of paint films, dirt, and varnishes shown in normal and in ultraviolet illumination, often with supporting Fourier transform infra red (FTIR) or x-ray diffraction (XRF) data. The landmark article launching this work was Joyce Plesters’s article for Studies in Conservation, “Cross-sections and Chemical Analysis of Paint Samples.” Ashok Roy of the National Gallery, London, noted in his Forbes lecture for the IIC that this 1956 paper is the single most cited reference in the whole of the literature of conservation. By the middle 1970s a hands-off, minimalist, or “less is more” approach appeared, subsequent to our initial excitement regarding many new tools and techniques. For example, in paintings conservation the heat table was first introduced in 1948 in Germany and the United Kingdom, and vacuum pressure was added in 1955 (see Figure 9). About 20 years later, in 1974, a lining moratorium was suggested at a conference in Greenwich, England, reinforced at ICOM-CC

FIGURE 9 The use of a vacuum heat table, a typical treatment of the late 1960s and 1970s (Michael Heslip, paintings conservator at Winterthur Museum in the late 1970s).

CHANGING APPROACHES IN ART CONSERVATION 53

meetings in Venice in 1975 and again at a special 1976 meeting in Ottawa. The problems caused by old and new lining technologies were reexamined. In 1992 I surveyed practicing paintings conservators in the United States; conservators who once said they had formerly lined nearly every painting they treated now reported lining only about 10 percent of their current treatments.4 Teaching treatments became more difficult, as there was now a complex menu of choices—loose linings, drop linings, edge linings, hand linings, humidity treatments, suction tables, local suction platforms, and more highly controlled vacuum tables—and lining is only one of many types of treatments used for paintings. We also have a complex menu of new adhesives and new electronic tools; we have “pharmacists of conservation” who specialize in supplying new spatulas, sampling kits, retouching supplies, and cleaning gel ingredients. There are many new approaches to removing varnishes—new enzymes, gels, and aqueous materials—less toxic and more specific in what they remove, thanks to changes made by conservation scientists or science-friendly conservators such as Richard Wolbers.5 My 1992 survey also revealed that many paintings conservators who used to remove varnishes entirely now may simply thin or reduce them instead, another more conservative approach. A minimally interventive philosophy has now been adopted by most conservators, especially as we revisit our own treatments from 25 years ago and as we watch our materials along with ourselves and our attitudes age. Marjorie Cohn, formerly a conservator of prints and drawings and now a curator and director at the Fogg, noted in the 2001 history issue of the Journal of the Institute of Paper Conservation, “Think first of the high value, both aesthetic and commercial, now attached to the unvarnished or unlined canvas. But ‘unwashed’ is now also a sales point in the catalogue descriptions of old master prints.”6 Our Art Conservation Program faculty members all generally adhere to this minimally interventive philosophy. In furniture conservation new approaches aim to preserve every bit of original material.7 We are now more aware of health hazards for the conservator and for the good health of the environment. Marjorie Cohn continued in her paper with a succinct summary of some other changes: One by one over the past three decades options for fumigation have been deemed too hazardous, until now there are no facilities remaining at my institution, Harvard University, and good housekeeping rather than extermination is the mode. Likewise, carcinogenic solvents have been minimized in the conservators’ repertory, and ventilation and disposal are now the first engineering prior- ity in the design of facilities.8 More interdisciplinary research and publications have appeared in recent years. Cohn was asked by Agnes Mongan of the Fogg to write what she now notes was an “unprecedented essay” for an Ingres catalogue in 1967 and continued that since then more art historians have come to appreciate the potential of technical

54 SCIENTIFIC EXAMINATION OF ART

examination and scientific analysis, and “conservators themselves have realized the essential importance of historical evidence on and in the works of art themselves.”9 George Stout’s three-legged-stool approach is embodied in an increasing number of publications authored by teams of conservators with art historians and scientists, such as Art and Autoradiography: Insights into the Genesis of Paintings by Rembrandt, Van Dyck, and Vermeer from the Metropolitan Museum of Art in 1982 and Examining Velasquez in 1988. There are now paired fellowships sponsored by the National Gallery and the Getty encouraging similar teams to create future publications. Textile conservators may now embrace a “connoisseurship bias” while choosing treatment options.10 Objects conservators may now elect to retain ritual substances such as yak butter that were applied to Tibetan sculptures. We have become more aware of respecting and retaining age value, historical value, and commemorative value, such as the proverbial blood on Lincoln’s shirt. Susan Heald, a textile conservator of the National Museum of the American Indian, gave a memorable talk for our graduate students. Heald worked with the Siletz regalia makers to stabilize pieces before the dance and get them ready for dancing during the dance house dedication ceremony. She noted, “For me this was a career altering experience to see these pieces danced on the night of the dance house dedication—it was very emotional for me and many others.” She described the smoky night atmosphere of the night of the dance she was allowed to observe (see Figure 10).

FIGURE 10 Susan Heald with members of the Siletz community at the National Museum of the American Indian, Smithsonian Institution.

CHANGING APPROACHES IN ART CONSERVATION 55

Other issues and concerns that would probably have never occurred to conservators in the 1950s and 1960s emerged following NAGPRA, the Native Ameri- can Graves Protection and Repatriation Act of 1990, which was discussed extensively by Miriam Clavir, senior anthropological conservator at the University of British Columbia, in her book Preserving What Is Valued: Museums, Conservation, and First Nations (UBC Press, 2002). People from First Nations may be reclaim- ing their sacred materials and reburying them; however, these materials may now have indelible museum registration numbers that must be removed from fragile surfaces or, even more important, be dangerously full of toxic pesticides that were state-of-the-art treatments in museums several decades ago. Glenn Wharton, objects conservator who wrote his dissertation for the Uni- versity College London on “Heritage Conservation as Social Intervention,” reported in the International Council of Museums–Committee for Conservation Preprints on the case of the royal Hawaiian monument Kamehameha the First. The state of Hawaii asked Wharton to remove the brightly colored paints that had been applied to the surface, returning the gilded bronze monument to its original state as part of the NIC’s Save Outdoor Sculpture. He soon learned that the local people intentionally paint Kamehameha in lifelike colors and hold annual cel- ebrations with chants and parades. He convinced the state to support a community-based, two-and-one-half-year participatory conservation project that essentially gave the community a seat at the table in deciding how to conserve the monument. Wharton brought current theory from material culture studies, eth- nography, and a reflexive stance to the study. He noted, “The recognition of cultural relativism and contested meanings embedded in material objects has begun to enter conservation literature.”11 In a landmark conservation conference in 1980 at the National Gallery of Canada, conservators, artists, scientists, and curators discussed issues relevant to the conservation of contemporary art. Several conservators spoke about their collaborations with living artists and the importance of interviewing artists to ascertain their views about materials and addressing damages to their pieces. Consulting and working with artists or collaborating with native Americans have been categorized together as acknowledging the cultures of origin, yet another important new direction for conservation.

CONSERVATION SCIENCE IN THE LAST DECADE The number of conservation scientists and scientific research and analytical laboratories in the United States has increased substantially since the early 1990s. The inclusion of sophisticated scientific data in scholarly conservation conference papers and publications has become standard, enhanced by the students and graduates of active U.S. and U.K. doctoral programs. Conservation methods have been improved by jointly sponsored conservation science research, demonstrating such factors as the impact of solvents on paint films, the absorption of dirt by

56 SCIENTIFIC EXAMINATION OF ART

acrylic paint, and the impact of temperature and humidity changes, with team members from the Getty Conservation Institute, the National Gallery of Art, the Canadian Conservation Institute, and the Tate in Britain. Since the year 2000, Angelica Rudenstine and the Mellon Foundation have taken conservation science another quantum leap forward. In the mid-1990s Mrs. Rudenstine interviewed conservation leaders to determine areas of need within the profession. She identified both conservation of photographic materials and conservation science and has made carefully considered grants available for training, workshops, and internships in those areas. Mrs. Rudenstine embodies a unique case of highly directed and hands-on improvement by the leader of a granting organization. She has funded and thereby created new conservation science positions in graduate training programs and major museum conservation laboratories. She has also imaginatively provided underwriting for talented younger conservation professionals, such as Philip Klausmeyer, who has returned to graduate study in order to enhance the scientific capabilities and instrumentation at his home institution, the Worcester Art Museum. The field of conservation science would not have been able to gainfully absorb the Mellon’s munificence in 1975. Mrs. Rudenstine scrupulously investigates the staff and capacity of an institution before making a grant. The integration of scientific understanding and conservation practice within well-trained individuals or by cooperating teams is a phenomenon of the last decade, and her programs have taken advantage of this moment in conservation history. As I mentioned earlier, the multiplication of approaches, sophistication of scientific research, and explosion of information has not made it easy to be a professor of art conservation. There are no old and outdated methods that we can now delete from the curriculum because someone in the past may have used them on the work we must treat. We now believe that the conservators of the twenty-first century that we are now training must thoroughly know their specialties, including current philosophy, history, literature, ethics, and the material properties and methods of analysis (subjects might range from underwater can- nonballs to ivory miniatures); collaborate with scientists and be able to understand scientific terms and methods; cooperate with allied professionals, including archaeologists, art historians, and the various cultures of origin; understand proper light levels, indoor pollutants, insect life cycles, climate control, emer- gency preparedness, and toxicity; be articulate advocates who write papers, give presentations, and in this time of economic cutbacks, be able to charm politi- cians, foundation heads, and reporters from “Sixty Minutes” if necessary. Per- haps we can now say that conservation has evolved from George Stout’s three- legged stool of the early days of the Fogg Art Museum to a twenty-first-century version of a ten-legged settee.

CHANGING APPROACHES IN ART CONSERVATION 57

NOTES

1. FAIC oral history interview with George Leslie Stout, Richard Buck, and Katherine Gettens by W. Thomas Chase and Joyce Hill Stoner, September 4, 1975. (The FAIC oral history archive is housed at the Winterthur Museum Library, Winterthur, Del.). 2. Ibid. 3. ANAGPIC (Association of North American Graduate Programs in Conservation). North American Graduate Programs in the Conservation of Cultural Property: Histories. Buffalo, N.Y.: 2000. 4. J. H. Stoner. The impact of research on the lining and cleaning of easel paintings. Journal of the American Institute for Conservation 33(1994):131-40. 5. R. C. Wolbers. Cleaning Painted Surfaces: Aqueous Methods. London: Archetype Books, 2000. 6. M. B. Cohn. Change, we hope for the better. The Paper Conservator 25(2000):101-105. 7. M. J. Anderson and M. S. Podmaniczky. Preserving the artifact: Minimally intrusive conservation treatment at the Winterthur Museum. In Wooden Artifacts Group, pp. 1-10. Richmond: Ameri- can Institute for Conservation, 1990. 8. Cohn, op. cit., p. 102. 9. Cohn, op. cit., p. 103. 10. P. Orlofsky and D. L. Trupin. The role of connoisseurship in determining the textile conservator’s treatment options. Journal of the American Institute for Conservation 32(1993):109-118. 11. G. Wharton. Conserving the Kamehameha I monument in Hawai’i: A case study in public conservation. 13th Triennial Meeting, Rio de Janeiro, 22-27 September 2002, ICOM Committee for Conservation I(2003): 203-210.

An Overview of Current Scientific Research on Stone Sculpture

Richard Newman Scientific Research Lab, Department of Conservation and Collections Management Museum of Fine Arts Boston, Massachusetts

ABSTRACT Scientific research on stone sculpture is focused on three major categories: determining sources of raw materials, developing methods of authenticating stone artifacts, and preservation. This paper reviews research in the first two categories. The goal of source determination is identification of the quarry for the stone used in a sculpture. In situ analytical techniques are occasionally applied, but most research involves samples. There are many approaches, ranging from petrography (study of thin sections and quantitative analysis of mineral compositions, usually on the thin sections) to elemental and isotopic analyses on drilled samples. Elemental analyses carried out by instrumental neutron activation, X-ray fluorescence, induc- tively coupled plasma (ICP) techniques, and others often provide the most useful information. Authentication of stone sculpture can focus on materials, techniques by which the materials were worked, and weathering layers. The alteration of a sculptural surface and the buildup of alteration products on that surface in a burial environment can be useful as an indication of age, and as such is often studied in stone sculpture authentication projects.

INTRODUCTION Stone was among the earliest naturally occurring materials to be used to create artifacts because it was readily available and required little processing. The steps

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 59

from procurement to desired final product are few and can be very simple. The current areas of research on stone artifacts to which science has made and continues to make a major impact fall into a few broad categories: determining sources of raw materials; developing methods of authenticating stone artifacts; and preservation. This paper will focus only on the first two categories because of limitations of space. “Rock” is the correct geological term for the raw material. Accord- ing to a definition used by some, “stone” refers to rock that has been intentionally shaped into an artifact. For the remainder of this paper all the material under discussion will be called rock.

DETERMINING SOURCES OF RAW MATERIALS The geographical origin of the rock used to make an artifact is a crucial piece of information in fully understanding the artifact. “Quarry” can be defined as the precise location from which the rock originated. In most cases the rock used in a sculpture is taken from an outcrop or exposure of rock, and the location of that outcrop would be the ultimate desired goal of the source determination. Some outcrops could be quite small, such as a pit in the ground from which some blocks to make small sculptures could be taken. In other cases an outcrop could consist of a very large exposure extending hundreds of yards or more. In the case of a very large quarry, localization of the source to some smaller area within the larger area may be desired. It is possible that a boulder or smaller chunks of rock could serve as source material, and these may have been collected from areas that are distant from the actual geographical location of the quarry. Examples include boulders moved by glaciers to pebbles moved down streambeds. One well-known example is Stonehenge, which some argue was constructed from glacial boulders that were far removed from their original source site. What is the relationship between the quarry location and the location at which the rock was shaped into an artifact? Some artifacts could have been shaped at the quarry, but blocks of rock could also be transported from a quarry to workshops, which could be at distant locations. Pebbles or small rocks could, of course, be easily transported long distances from the points at which they were collected. This is the case, for example, with some of the jade used in ancient China to produce small objects. A very widely traded material in many ancient cultures was obsidian—volcanic glass—to which considerable scientific research on sources has been devoted. We know in the case of marble sarcophagi in Ro- man times that blocks were sometimes partially shaped at the quarry and then shipped to various workshops where the final carving was completed. Determin- ing workshop locations is an aspect of research on rock sculptures that can draw on information about quarry sources.

60 SCIENTIFIC EXAMINATION OF ART

Approaches to Determining Sources Nondestructive methodologies, carried out directly on artifacts without sampling, are particularly valuable in the case of monuments or sculptures on monuments. One technique that has been utilized in recent studies is magnetic susceptibility, which gives a signal based to a great extent on the presence of the mineral magnetite. Sources for Roman-period gray-granite columns in the Rome region have been studied (Williams-Thorpe and Thorpe, 1993), and sources for the gray to yellow-brown sandstone blocks used in different phases of construction of the eighth- to thirteenth-century complexes at Angkor in Cambodia have also been identified (Uchida et al., 2003). For the most part, sourcing of projects involves analysis of samples, which fall into two broad kinds: solid chips and powder. The appropriate type and size of sample depend on many factors related to the rock type, the geology of the potential quarry source or sources, and constraints on sampling of the sculpture(s) being studied. A rock-sourcing project could involve a single artifact or a group of artifacts. Sourcing projects usually begin with a group of artifacts made from a particular type of rock material. The questions to be answered are typically: Do the artifacts all come from one quarry source? What is that quarry source? One phase of the research involves characterizing the material used in the group of artifacts. The goal is to acquire types of data that can characterize the rock material in sufficient detail to determine whether the rock from the various objects in the group being studied can reasonably be concluded to have come from the same source, or is likely to have come from more than one source. If more than one specific source is likely, another aspect of this phase of the research is to group objects that are likely to have come from the different sources suggested by the data. As in sourcing studies of other types of materials used in artifacts (such as ceramics), it is common practice today to use multivariate statistical analysis to evaluate analytical data, thereby establishing potentially related groups of objects on the basis of this kind of evaluation. There are different approaches to utilizing statistics for data evaluation, and some may be more appropriate in certain circumstances than others. Although beyond the scope of this paper, the examples discussed in this paper include most of the current statistical procedures. Evaluation of data from artifacts can establish hypothetical distinct sources or quarries, but at this point in a study it is uncertain whether the groups established is this manner actually represent distinctly separate sources, or whether the artifacts within individual groups actually all originated from a single source. A second phase of research involves characterizing potential quarry sources and then comparing these with results from the artifacts. The goal is to determine the quarries from which the artifact rock came. Assuming that potential quarry sources can be located, systematic sampling of these is crucial in order to fully

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 61

characterize them. In advance it is difficult to establish what would constitute systematic sampling. The nature of the rock being sampled is one factor that needs to be considered, and whether the rock type is present in more than one horizon (or layer) that was formed during distinctly different geological events. A single rock type usually displays compositional variation, particularly if the outcrop is quite large. For multivariate statistical analyses, it is usually stated that the number of source samples should be at least equal to the number of variables (usually elements) being analyzed in each sample. For rock quarries far more extensive sampling than this is usually an absolute necessity unless the outcrop is extremely small. Another aspect that needs to be determined in analytical studies of artifacts (and quarries) is the appropriate sample size. A representative sample of an artifact can be defined as one whose composition reflects the composition of the artifact as a whole. In the case of a block of rock, samples of 3 to 4 milligrams taken from several areas may not have identical compositions, while samples of 1 or 2 grams may. What constitutes the minimum acceptable sample size to truly represent the whole will often vary according to which properties are being analyzed. In studies that involve elemental analysis, certain elements are often found to be good discriminators with very small samples, others may be useful but only if larger samples are studied, and still others may be so unevenly distributed that they have little utility. Research on stable carbon and oxygen isotope ratios in carbonate rocks (discussed in more detail below) has shown that a sample of a few milligrams can be considered to be representative of a large block of the rock.

Solid Chip Samples (Petrography) Solid chip samples are utilized mainly for petrographic analysis, the well- established tool used by geologists to characterize and classify rocks. Geologists typically prepare thin sections that cover most of the surface of a standard petrographic slide (in the United States, 27 x 46 millimeters), but that would require a sample far larger than could be taken from the vast majority of sculptures. Samples that can provide a thin section about 1 centimeter across, taken with sculptors’ chisels or hollow core drills, are often more than adequate for characterizing a rock, unless the rock is relatively coarse grained (with numerous individual mineral grains exceeding about a millimeter in size, for example). The thin sections provide information on the mineral content of the rock, the grain sizes and shapes, and the interrelationships of the different mineral grains (the texture). Geologists have long utilized the electron beam microprobe to quantitatively analyze individual mineral grains in thin sections of rocks. The microprobe can provide excellent quantitative analyses of the major and minor elements in mineral grains that are at least a few micrometers across. This quantitative information further characterizes the mineral in the specific rock being studied. One application of the microprobe focused on a silvery gray schist used in the

62 SCIENTIFIC EXAMINATION OF ART

ancient Gandharan region (present-day northern Pakistan) from about the first to third centuries AD. The specific find sites are unknown for virtually all Gandharan sculptures made from this easily recognized rock. Some small local sources of this type of rock have been identified, but there are undoubtedly many more than are currently known. In a pilot project (Newman, 1992) 6 to 8 grains of chloritoid, a characteristic mineral of the rock, were analyzed in thin sections from 16 sculptures. After considering relative amounts of magnesium, iron, and manganese in the mineral, chloritoid in seven of the sculptures was found to have virtually identical compositions. Separate samples from the top and bottom of one of these sculptures, a 1.5-meter-tall figure, were analyzed to give some sense of the variation in chloritoid compositions in a fairly large block of rock. The spread in the compositions of these grains in the one block was about the same as the spread shown by grains in individual thin sections. As a working hypothesis, it is possible that the rock used in the sculptures with virtually identical chloritoid compositions came from the same quarry, while the other nine sculptures were made from rock taken from other quarries. Although more sculptures could be studied by this method and more extensive chloritoid analyses carried out in each thin section, the conclusions will remain tentative until possible quarry sources can be identified and studied. The research points out a possible relationship between the seven sculptures, but the precise significance of the relationship is uncertain. Were the sculptures produced in the same workshop within a narrow time frame, or produced in a general area in proximity to a certain quarry but at different times, or was material exported to several sites from one quarry? Or are these sculptures even from a single quarry? Another application of microprobe mineral analysis was applied to small scrapings (0.1-1.0 milligram) from predynastic Egyptian basalt vessels (Mallory et al., 1999). It has been estimated that nearly one-quarter of stone vessels from this period were made from basalt. Grains of two minerals were quantitatively analyzed for major and minor elements: plagioclase feldspar and pyroxene. After comparing the results with some analyses of samples from major Egyptian basalt sources, it was concluded that the predynastic sculptural material was all northern Egyptian basalt. Finished objects, or pieces of rock that were shaped in local workshops, were apparently shipped all over Egypt from the source quarry or quarries. A more recently developed tool for individual mineral grain analysis in thin sections is laser ablation microprobe ICP mass spectrometry. Currently, the minimum spot size for this technique is about 20-50 micrometers, meaning that it requires larger grains than does the microprobe, but the technique has the advantage of being able to analyze trace elements more readily than does the microprobe. A recent application focused on basalts in ancient Egyptian sculpture (Mallory-Greenough et al., 1999), using thick polished thin sections. The mineral analyzed, augite (a type of pyroxene), is a major constituent of most basalts. It was found that some trace elements, which can easily be analyzed by the laser probe

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 63

procedure, are far more useful for discriminating between augites in basalts from different quarry sources than are major and minor elements. Samples studied were basalt temper from pottery. A conclusion was that some of the basalt sources utilized during the pharaonic period in ancient Egypt have yet to be identified, since the augite compositions in samples from some artifacts did not correlate well with augite from basalts from the known sources. Both of the techniques just discussed are adjuncts to traditional petrography. There are many rocks, however, that cannot be adequately distinguished by petrography, or that contain minerals that do not show sufficient variations in major, minor, or trace element compositions in order to be suitable for microprobe or laser ablation ICP mass spectrometry. Other analytical techniques that can be applied to solid chip samples, whether prepared as thin sections or not, include Fourier transform infrared spectroscopy and Raman spectroscopy. The latter shows particular promise as an analytical tool, for example, in a project that was concerned with Mesoamerican jadeite sources in Guatemala (Gendron et al., 2002).

Powdered Samples There is a full battery of analytical techniques that are applied to whole-rock samples, that is, drilled (powder) samples that include all of the minerals found in a rock. Most of the analytical techniques currently applied to drilled samples determine elements (in some cases, isotopes), with the exception of X-ray diffraction (XRD). XRD is useful for general mineral identification and, by extension, rock identification in some cases. In sourcing research projects XRD can provide a quantitative estimate of the amounts of different major and minor minerals in a rock. Lacking textural information, XRD is less useful than thin section analysis for specific rock identification and classification. XRD is probably most useful in studying certain monomineralic or nearly monomineralic rocks. Many minerals have variable compositions, and variations in composition can give rise to distinctive XRD patterns. One interesting project studied chert and chalcedony, two microcrystalline varieties of quartz used in New England archaeological sites (Pretola, 2001). XRD patterns were used to identify minor minerals in some of the source materials. In addition, calculated diffraction patterns from crystal structure parameters were fit by computer methods to observed diffraction patterns in order to determine the ratios of quartz to a silica polymorph called moganite in the samples. Powder samples are most commonly used for elemental analysis. The full battery of modern elemental analysis techniques has been applied to rock analysis. Current widely applied techniques include X-ray fluorescence spectrometry, instrumental neutron activation analysis (INAA), ICP optical emission spectroscopy, and ICP mass spectrometry. Some of these techniques are more appropriate

64 SCIENTIFIC EXAMINATION OF ART

for certain elements, or abundances of elements, than others. A given rock, of course, will contain dozens of elements, which are present from major to very low trace levels (parts per billion and less). Not all these elements will be useful in a sourcing project. Which elements are the most useful is not usually known at the outset of a project unless previous work on similar rocks has been carried out. Usually the application of a particular technique or techniques in a discrete research project can easily produce data that is reliable. More difficult sometimes is comparing data acquired by other analytical techniques, or even other research groups using some of the same instrumentation. Use of common standards and frequent equipment calibrations can help to overcome this kind of difficulty, but it can prove to be a limitation in comparing data from different research groups or projects. Both the suite of elements utilized in a sourcing study and the methods by which the data is evaluated are obviously crucial to the conclusions.

Limestone in Medieval France One of the most outstanding research projects yet to be published on sculptural rock sources involves the limestone utilized in French medieval sculptures. The beginning of this massive undertaking goes back to a modest self-contained 1985 project published on a small group of sculptures (French et al., 1985). Nine Romanesque sculptures in four American museum collections that shared certain stylistic features were considered by some scholars to have originated from one monument in southern France. The goal of the project was to determine whether this was the case and the location of the rock source(s). Petrographic examination and neutron activation analysis indicated that the nine were probably made of rock from the same quarry. As a part of this phase of the project the representative sample size for the chosen analytical technique and elements being analyzed was determined, and the fact that a single sample could be considered representative of an entire block of rock (of the size used in the sculptures) confirmed. Possible quarry sites in the general region of origin were selected for sampling and analysis on the basis of the petrographic features of the sculpture samples. Initial neutron activation analysis of a few samples from several different geological formations narrowed down the possibility to one area. From this area over 100 samples of the same geological formation were analyzed, three old quarries in particular being extensively sampled. The neutron activation analyses coupled with petrographic data narrowed down the likely source to two quarries that were about 2.5 kilome- ters apart. Multivariate statistical analysis was carried out on the elemental data. Since that study, the same research group has focused on the characterization of some of the major limestone quarries in the Paris region, a study that to date has involved some 2,300 samples from 300 quarries (Blanc et al., 2002). The research group concluded that 10 to 15 elements were required to best characterize the French limestones, as were many samples from each compositional group

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 65

(quarry), a number that actually ranged from 8 to 40 for the major Paris basin quarries. The extensive database now makes it possible to determine quarry sources for museum artifacts made from limestone that came from this general region of northern France. The database includes samples from some monuments, and this data has enabled the identification of the monument that sculptures in museum collections originally embellished. Examples are the head of an angel, the head of a Virtue, and a choir screen from three separate collections. The trace element patterns of these three objects closely matched the trace element patterns of a group of samples from 25 sculptures on Notre Dame in Paris. The stone on the cathedral and in the three sculptures may have come from one of the ancient quarry sites that the research group has extensively studied, Charenton. This is one of a number of quarries in the Paris region from which one particular type of fine-grained limestone (Upper Lutetian limestone) was taken. Trace element analyses permit rock from many of the different quarries to be distinguished, although all are indistinguishable by petrography. About 1 gram of rock is required for this analysis.

White Marble in Classical Antiquity The most extensively studied general rock type used by any culture or group of cultures in a region are the white marbles of the Mediterranean. Bronze Age sculptors in the Cycladic islands of the Aegean used white marble extensively in Greek and Roman times and well beyond. There are less than a dozen major quarries that supplied rock that was widely exported at different periods. In addition, there were dozens of quarries of mainly regional importance. The majority of the white marbles consist entirely or almost entirely of calcite. Their textures, grain sizes, and minor and accessory mineral compositions provide some useful properties for discrimination, but petrographic properties alone have not been adequate to clearly distinguish most of these marbles. An important breakthrough came in 1972 with a publication by marine geochemists who applied a tool of their trade, stable isotope analysis, to a few samples from several of the major ancient Mediterranean marble quarries (Craig and Craig, 1972). The pilot project showed that simple plots of the ratios of the stable oxygen isotopes (18O and 16O) versus the stable carbon isotopes (13C and 12C), calculated with reference to a standard international reference material, the carbonate fossil Pee Dee belemnite, separated the marble from different quarries. Stable isotope analysis, which is comparatively inexpensive to carry out and requires a sample of only a few milligrams, is now perhaps the most frequently applied technique in sourcing studies of marble sculptures from this part of the world, but time has clouded the picture considerably. A recent publication showed how the isotope fields for the quarries have expanded since the 1972 publication (Gorgoni et al., 2002). It is possible to discriminate between certain quarries with stable isotope data, but many cannot be distinguished from this data alone. Quarry

66 SCIENTIFIC EXAMINATION OF ART

fields were initially defined by drawing an outline around all data points from analyzed samples taken from a quarry, but statistical analysis is now more commonly utilized. In studies of ancient quarries it is not always certain that samples taken from the quarry as it is known today entirely represent the quarry as it may have been known during a much earlier period. For example, it is possible that parts of a quarry may no longer be accessible, or even may have been depleted. In this case, analyses of sculptures that can reasonably be assumed to have been made from material originating in that quarry can be used to define the compositional field, instead of relying solely on newly taken quarry samples. This has recently been done for some of the major Mediterranean marble quarries. Incorporation of sculptural data has in fact somewhat expanded some of the fields. Many other analytical procedures have been applied to the study of these white marbles. Some have been championed by particular research groups and have yet to become widely applied, in spite of their promise. A brief survey of these techniques serves as an example of how there are diverse, often equally useful approaches to a sourcing problem, and more important, how data from more than one technique may be crucial in solving a particular problem. Although petrography was probably not considered a particularly valuable technique by most researchers for many years, within the last few years one particular property that is best determined with a thin section has been shown to be very useful: maximum grain size. Although a large thin section is required to accurately determine this, this parameter does enable some quarries whose isotope fields overlap to be distinguished. In a recent publication (Gorgoni et al., 2002), in fact, isotope maps for fine-grained and coarse-grained marbles were shown (a maximum grain size of 2 millimeters was the dividing point). Elemental analysis has been applied by a number of research groups, which have proposed different elements. Reliable comprehensive databases have yet to be established for many quarries, and elemental analysis is probably best applied in small-scale projects. Certain elements have been shown to be useful in certain either-or questions, where the possible quarries have been narrowed down to two. Two other techniques that have been employed only by a few researchers to date are quantitative cathodoluminescence and electron paramagnetic resonance (EPR). Qualitative cathodoluminescence carried out with a suitably equipped optical microscope and recorded on color film distinguishes certain marbles on the basis of the color they show when bombarded with an electron beam. Quan- titative cathodoluminescence carried out in a scanning electron microscope with an attached spectrophotometer is potentially more valuable. In the ultraviolet/ visible region of the electromagnetic spectrum most marbles show only one or two major peaks: a peak at 620 nanometers, due to manganese (2+) substituting for calcium in the crystalline matrix and a pair of peaks at about 350 and 377 nanometers due to cerium (3+). Ratios of the two can distinguish many marbles

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 67

from one another. The orange color shown by some marbles is related mainly to trace amounts of manganese in the rock. EPR or electron spin resonance (ESR) has been applied for many years by a small number of researchers but in the last few years has come into its own as a viable technique. One recent article evaluated a number of features of the EPR spectra of marbles, features in part attributable to the presence of manganese and iron in the crystalline lattice of calcite (Polikreti and Maniatis, 2002). The study of Mediterranean white marbles has reached such a level of matu- rity that many important quarries have been well characterized and the strengths and limitations of a range of analytical techniques that can be focused on a specific sourcing problem are fairly well understood. In the near future more work will continue to center on characterizing some of the smaller quarries. Multivariate statistical analyses incorporating properties or features from two or more analytical techniques have become common practice in marble provenance studies.

DEVELOPING METHODS OF AUTHENTICATING ROCK ARTIFACTS Authentication of stone artifacts from a scientific point of view usually focuses on one of two lines of evidence: the original materials and manufacturing procedures, and weathering or alteration of the surface after carving or other finishing.

Materials and Working Procedures In this category would obviously be the nature of the rock itself. If a sufficiently large database of information on the rocks that presumably should have been utilized were available, detailed rock analysis may help to build a case for or against the authenticity of a problematic artifact. More often than not, the result of such an analysis even when a large database is available may lead to the conclusion that the material used in an artifact is consistent with the purported period and place of origin, which does not imply that the object is definitely authentic, only that it could be. Working of the rock can also be important. Certain tools or manners of using a tool may be characteristic of a type of stone artifact. Some researchers have carefully compared drill marks produced by known ancient drilling procedures with marks produced by modern drills, and this information could obviously be used in certain authentication studies. Tool marks on ancient gemstones have also been extensively studied (Rosenfeld et al., 2003), with the purpose of determining the tools that were utilized and the way the tools were used. Applications of this type of information to weathered sculptural surfaces is much more difficult since many details have been worn away. There may be residues of tools that can be identified, for example, elements from metal chisels or bits of abrasives used in polishing operations. A study of

68 SCIENTIFIC EXAMINATION OF ART

Roman gemstones found residues of lead and tin metal as well as barite, all of which were probably associated with abrasive polishing procedures (Rosenfeld et al., 2003). It is difficult to quantify the appearance of tool marks on a complex sculpture in such a manner that would allow the data to be readily utilized by other researchers. The use of tool marks in authentication studies may fall more into the realm of technical connoisseurship, where interpretation of significance depends on the experienced eye of a researcher who has carefully examined many tool marks on many artifacts. One example is in a paper published 13 years ago by a sculptor who has studied tool marks on Roman marble sculptures (Rockwell, 1990). He argued that although a skilled modern forger could use the same tools as an ancient Roman sculptor, utilizing them in exactly the same manner would be very difficult, and that the differing manners of use could be distinguished by careful examination of tool marks, when the appropriate marks are still present on a sculpture. A very interesting concluding remark of his was, “In reading literature on fakes and faking I find that no one questions that the faker can, if he wants to, reproduce the technique of the period being faked. It is taken as a given that a good technician can reproduce the technique of any period. I think that this is an assumption about techniques, based on the ignorance of nontechnicians, that deserves serious questioning.”

Surface Alteration One final approach is to look closely at surface alteration. Changes that take place on a newly worked rock surface over time can involve physical as well as chemical factors. Sharp details can become rounded or abraded. Minerals at the surface can chemically break down due to interaction with the atmosphere or burial environment. The surface of an ancient rock sculpture may contain materials that in part arise from interaction of the rock surface with its environment over some period of time. “Weathering layer (layers)” will be used to refer to all such materials. Many ancient rock artifacts will have been buried for some extended period of time, but few if any details are usually known about the burial environment. While it is quite possible that there is some correlation between thickness of a weathering layer and amount of burial time for a given type of rock and given environment, all the information that would be required to establish this is never available. The composition and microstructure of a weathering layer are both useful. Possible variation in both composition and structure from place to place on the surface of an artifact make it prudent to examine multiple samples. Small chips that can be prepared as cross-sections are the single most useful type of sample, since compositional and structural features can both be studied on the same

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 69

sample utilizing such techniques as scanning electron microscopy with attached X-ray spectrometers. On rocks that contain more than one mineral the microscopic structure and composition of a weathering layer will not be identical on exposed surfaces of the different minerals. Some minerals, such as quartz, are very resistant to weathering, while others are much less resistant. The processes by which different minerals break down and the products of those breakdown processes vary from one mineral type to another. Even on monomineralic rocks the nature of the weathering layer will not necessarily be uniform over the entire surface of a sculpture. To have confidence in the application of weathering layer data to an authentication question, the weathering that takes place on the type of rock in question in the (burial) environment in which it is likely to have undergone most of its weathering should be characterized. Geological samples if appropriate or artifacts (or architectural blocks) if appropriate should be examined. A range of deterioration, on a single artifact and on different artifacts, can probably be expected. Rarely can such a database be built, however. Another option is to look for weathering features and by-products that are known to result from the deterioration of certain minerals or types of rocks under general conditions similar to those of a sculpture made from the same material. If the question involves the possible faking of weathering layers, specific analyses that can distinguish the real phenomenon from an artificially produced emulation need to be carried out. Some researchers search for organic materials that might have served as binders used to adhere an artificial patina onto the surface of an artifact. Some organic materials will undoubtedly be present in authentic weathering layers, but the nature of these and their relative abundance can reasonably be expected to be quite different. A wide range of very sensitive techniques for organic analysis are available, such as gas chromatography/mass spectrometry, which can be applied to characterization of organic residues in rock weathering layers. The study of weathering layers on an artifact may require analyses of a number of different kinds. To date, the weathering layers that arise on stone sculptures through long-term burial have not been very extensively studied for many classes of rocks or rock artifacts. Probably the rock to which the most work has been applied is again the white marbles of the Mediterranean region. Marble is sensitive to acidic groundwater. Weathering in a burial environment usually involves solution at the surface, and along grain boundaries, coupled with reprecipitation of calcium carbonate on the sculpture surface. The reprecipitated calcite often incorporates bits of rock from the surface and organic and inorganic materials from the soil. In the case of marble, stable isotope analysis has also been carried out, since interaction between rock and groundwater will result in a change in isotope ratios. This technique can sometimes detect extremely thin weathering layers.

70 SCIENTIFIC EXAMINATION OF ART

Marble made of dolomite instead of the much more common calcite can undergo a very different kind of weathering. Dedolomitization is a phenomenon in which calcite replaces dolomite through interaction with groundwater in some specific circumstances (Doehne et al., 1992). To date, dedolomitization has not been able to be induced in a laboratory setting to more than a minor extent. The presence of a reasonably thick layer of this kind would provide some proof that an artifact has been buried for an extended period of time, at least given our present understanding of the phenomenon. In a study of a dolomite sculpture attributed to the Archaic Greek period, researchers observed a weathering layer that very closely resembled in structure and composition the layers seen on buried dolomite marble that had undergone weathering for nearly two millennia (Newman and Herrmann, 1995). Most rocks are not as simple in mineralogy as marble, nor as reactive to groundwater. The weathering layers on these can be more complex to analyze and interpret. One recent example involves the volcanic tuff utilized in eastern Java during the Majapahit period (1293 to about 1520). In a study of eight statuettes researchers concluded that certain minerals form in small depressions on the sculpture surface during long-term burial (Duboscq, 1989-1990). The depressions are sites where certain minerals that made up the original rock were located (the depressions were formed in geological time, not during burial). The research identified several minerals coating the surfaces in these depressions. It was concluded that these minerals (zeolites, clay, and monazite) should be present in similar depressions on authentic artifacts made from this type of rock. Scanning electron microscopy with energy-dispersive X-ray fluorescence was used to characterize the weathering products, using either cross-sections or, more typically, scrapings from the insides of depressions on the surface of a sculpture. Although the conclusions regarding the value of this type of evidence were too stridently stated given the scope of the project, the research is an example of the use of detailed compositional and structural information associated with weathering. Weathering layers are certainly one of the best pieces of evidence in authentication studies of rock artifacts, but the application of this type of evidence can be said to be in its infancy for the vast majority of rock types. More extensive research will be required to add certainty to conclusions based on this approach. The study of marble weathering layers can be taken as an example of the increasing sophistication applied to the problem as time goes by. A little over 30 years ago, when the use of weathering layers on marble as an authentication tool was first suggested, thin sections examined under a polarizing-light microscope were the only tool applied to their study. Another type of evidence that for many decades has been taken to indicate an ancient weathered surface on a marble artifact are root marks, bits of plant roots that become cemented to the surfaces of sculptures that have been buried. In recent years the very important role of biodeterioration in the weathering of rocks has been the subject of much research, most of it focused on monuments and

AN OVERVIEW OF CURRENT SCIENTIFIC RESEARCH ON STONE SCULPTURE 71

architecture. Some of the molecular characterization techniques that have been applied to the study of biodeterioration in aboveground settings may also be of value in characterizing weathering in burial environments, where biological agents may also be at work.

CONCLUSIONS: THE FUTURE The major categories of research noted in this paper have long been areas of research and will continue to be very important in the future. As examples in this paper have shown, sourcing projects that can be expected to produce worthwhile results are often time consuming. There are many sourcing projects on rock sculptures that remain to be undertaken, projects that can potentially be of great importance to art historians and archaeologists. Authentication of rock sculptures has also long been an active area of research, in which advances are continu- ously made. The future will hold more of the same.

REFERENCES

Blanc, A., L. L. Holmes, and G. Harbottle. 2002. In Interdisciplinary Studies in Ancient Stone, eds. J. L. Herrmann, N. Herz, and R. Newman, pp. 103-109. London: Archetype Publications. Craig, H., and V. Craig. 1972. Science 176:401-403. Doehne, E., J. Podany, and W. Showers. 1992. In Ancient Stones: Quarrying, Trade and Provenance, eds. M. Waelkens, N. Herz, and L. Moens, pp. 179-190. Leuven, Belgium: Leuven University Press. Duboscq, B. 1989-1990. In Majapahit. Paris: Beurdeley & Cie (no pagination). French, J. M., E. V. Sayre, and L. van Zelst. 1985. In Application of Science in Examination of Works of Art, eds. P. A. England and L. van Zelst, pp. 132-141. Boston: The Research Laboratory, Mu- seum of Fine Arts. Gendron, F., D. C. Smith, and A. Gendron-Badou. 2002. Journal of Archaeological Science 29:837-851. Gorgoni, C., L. Lazzarini, P. Pallante, and B. Turi. 2002. In Interdisciplinary Studies in Ancient Stone, eds. J. L. Herrmann, N. Herz, and R. Newman, pp. 115-131. London: Archetype Publications. Mallory, L. M., J. D. Greenough, and J. V. Owen. 1999. Journal of Archaeological Science 26:1261- 1272. Mallory-Greenough, L.M., J. D. Greenough, G. Dobosi, and J. V. Owen. 1999. Archaeometry 41:227- 238. Newman, R. 1992. Archaeometry 34:163-174. Newman, R., and J. Herrmann. 1995. In The Study of Marble and Other Stones Used in Antiquity, eds. Y. Maniatis, N. Herz, and Y. Basiakos, pp.103-112. London: Archetype Publications. Polikreti, K., and Y. Maniatis. 2002. Archaeometry 44:1-21. Pretola, J. P. 2001. Journal of Archaeological Science 28:721-739. Rockwell, P. 1990. In Marble: Art Historical and Scientific Perspectives on Ancient Sculpture, pp. 207- 222. Malibu, Calif.: J. Paul Getty Museum. Rosenfeld, A., M. Dvorachek, and S. Amorai-Stark. 2003. Journal of Archaeological Science 30:227- 238. Uchida, E., O. Cunin, I. Shimoda, C. Suda, and T. Nakagawa. 2003. Archaeometry 48:221-232. Williams-Thorpe, O., and R. S. Thorpe. 1993. Archaeometry 35:185-195.

Biodeterioration of Stone

Thomas D. Perry IV, Christopher J. McNamara, and Ralph Mitchell Division of Engineering and Applied Sciences Laboratory of Applied Microbiology Harvard University Cambridge, Massachusetts

ABSTRACT Stone cultural heritage materials are at risk of biodeterioration caused by diverse populations of microorganisms living in biofilms. The microbial metabolites of these biofilms are responsible for the deterioration of the underlying substratum and may lead to physical weakening and discoloration of stone. Air pollutants in urban environments accelerate biodeterioration by serving as an additional nutrient source for the microorganisms. Current strategies to reduce biodeterioration and repair damage that has already occurred are discussed. Current techniques for assessing microbial populations and their effects are evaluated. Additionally, we describe two new techniques for quantification of these interactions: microcomputer- assisted tomography (microCT) and atomic force microscopy (AFM). The study of biodeterioration of stone cultural heritage materials is as diverse as the sites studied. This review also attempts to address some of the issues facing conservation scientists, including methodology and application.

INTRODUCTION Many cultural heritage materials are at risk of biodeterioration by microorganisms, including metals (e.g., Berk et al., 2001), glass (e.g., Schabereiter-Gurtner et al., 2001), ceramics (e.g., Sand and Bock, 1991), paper (e.g., Fabbri et al., 1997), paintings (e.g., Rubio and Bolivar, 1997), wood (e.g., Bjordal et al., 1999), coatings (e.g., Flemming, 1998), synthetic polymers (e.g., Gu et al., 1998a), and

BIODETERIORATION OF STONE 73

mummified bodies (e.g., Arya et al., 2001). Biodeterioration presents conservation challenges that vary as widely as the types of materials themselves, and discussion of the processes causing degradation of these varied historic materials is not possible here. Consequently, the scope of this paper is limited to discussion of the mechanisms, analytical techniques, and conservation strategies of stone biodeterioration. Biodeterioration plays an important role in the degradation of stone in historic buildings, monuments, and archaeological sites (e.g., Saiz-Jimenez, 1999). Microorganisms that have been demonstrated as the causative agents in deterioration of stone include bacteria (Urzi et al., 1991), Archaea (Rölleke et al., 1998), cyanobacteria, algae (Tomaselli et al., 2000), fungi (Gorbushina et al., 1993), and lichens (Garcia-Rowe and Saiz-Jimenez, 1991). Additionally, stone objects may support novel communities of microorganisms (e.g., alkaliphiles, halophiles, and endoliths) that function in the biodeterioration process (Saiz-Jimenez and Laiz, 2000). Our work focuses on Maya archaeological ruins in the Yucatán, Mexico, which are heavily colonized by microorganisms (see Figure 1). We have isolated copiotrophic (capable of growth on a medium containing a high concentration of organic carbon), alkaliphilic (capable of growing at high pH), oligotrophic (capable of growth on media containing a low concentration of organic carbon), and halophilic (salt-tolerant) bacteria, as well as phototrophic microorganisms and

FIGURE 1 Photograph of a room in the Quadrangle of the Nuns, Uxmal, Yucatán, Mex- ico, showing visible microbial growth and staining.

74 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 Scanning electron micrograph (30,000X) of an unidentified mi- croorganism with an unusual morphol- ogy collected from the surface of an inner room in a temple in the Maya site at Uxmal, Yucatán, Mexico.

fungi from the ruins. Unique organisms that have to date eluded our identification efforts have also been observed (see Figure 2). Microbial biodeterioration of stone occurs as a result of the formation of biofilms (see Figure 3). Biofilms are collections of bacterial cells on surfaces that are maintained by electrostatic forces and/or adhering exopolymers. Biofilm formation begins with the initial adhesion of microorganisms to a surface. Division of attached cells produces microcolonies containing large amounts of exopolymer separated by patchy areas relatively devoid of growth. Production of exopolymer and other exudates is stimulated in response to cellular density by cell-cell signal- ing. The exopolymer matrix is composed mainly of polysaccharides and serves a variety of functions such as providing protection from desiccation, radiation, erosion, and disinfectants, as well as storage of organic carbon and nutrients (Flemming and Wingender, 2001; Costerton et al., 1995). The exopolymer matrix limits the rate of diffusion in microcolonies, resulting in the formation of mi-

croenvironments due to gradients in pH, O2, nutrients, and organic carbon (Rittman et al., 1999).

EFFECTS OF MICROBIAL METABOLITES

Effects of Acids Microbial biodeterioration of stone is widely thought to occur through the action of organic and inorganic acids produced as metabolic by-products (Gu et

BIODETERIORATION OF STONE 75

FIGURE 3 Diagram of a microbial biofilm growing on a stone surface, which highlights the environmental heterogeneity present in attached microbial communities.

al., 2000). Bacteria isolated from the Maya site of Ek’ Balam, Yucatán, Mexico, are capable of producing calcium carbonate-dissolving exudates (Perry et al., 2003). However, not all organic acids produced by microorganisms cause immediate dissolution of stone. For example, oxalic acid may have a protective role by the formation of calcium oxalate on stone surfaces (Di Bonaventura et al., 1999).

Effects of Exopolymers In addition to metabolic acids, biofilm exopolymers may play an important role in deterioration of stone because of their proximity to the stone substratum. Bacteria produce the polymers as biofilm growth is initiated, and adhere directly to the stone (see Figure 4). Bacterial exopolymers are large macromolecules consisting of varied sugar molecules exhibiting several kinds of functional groups (Ford et al., 1991), including acidic carbonyls. These functional groups are often capable of binding cations in solution (Smidsrød and Haug, 1965). For example, negatively charged carboxylic and hydroxyl groups of exopolymeric materials, such as alginic acid, form complexes with the mineral surface and may leach calcium from limestone matrices (Perry et al., 2004). Other polysaccharides with different chemistry may inhibit dissolution (Welch and Vandevivere, 1994). The activity of these molecules and their functional groups in chelation and dissolution of stone is still not understood because of their variability and complexity.

76 SCIENTIFIC EXAMINATION OF ART

FIGURE 4 Scanning electron micrograph (30,000X) showing bacteria growing on a marble surface and producing exopolymer attached to the stone substratum.

Discoloration Microbial pigments frequently cause discoloration of stone. While these metabolites may not cause physical damage, they can cause aesthetic problems. Monuments are particularly susceptible to this form of discoloration. A typical example is the red stain observed on the U.S. Naval Academy’s Tripoli Monument in Annapolis, Maryland (see Figure 5a). Two pigment- producing fungi were isolated from the monument: Epicoccum nigrum and Drechslera sp. Interestingly, in a growth medium of low salt concentration, E. nigrum appeared milky white in color, while in a high calcium concentration medium, the fungus produced a red pigment (see Figure 5b). The red stain may have been an exudate produced to protect the fungus from the stresses of its habitat on the stone, such as ultraviolet radiation or ionic strength (Wheeler and Bell, 1988).

INTERACTION OF MICROORGANISMS WITH AIR POLLUTANTS One of the most serious causes of stone deterioration is urban air pollution caused

BIODETERIORATION OF STONE 77

FIGURE 5 (a) Red pigmentation on the Tripoli Monument in Annapolis, Maryland. (b) Production of pigment by the fungus Epicoccum nigrum in a growth medium containing marble. The fungus was colorless when grown on the same medium without addition of marble.

by fossil fuel combustion. In addition to chemical weathering of stone, pollutants may stimulate microbial biodeterioration. Urban air pollutants are rich in both

nitrogen dioxide (NO2) and sulfur dioxide (SO2). NO2 and SO2 are mainly derived from fossil fuel combustion and are transported by wind or water to stone surfaces (Saiz-Jimenez, 1993). Some chemoautotrophic bacteria obtain energy by oxidizing sulfur and nitrogen compounds to sulfuric and nitric acids. For example, Thiobacillus colonizes weathered surfaces of marble in a polluted area (Mitchell and Gu, 2000). These bacteria, through production of sulfuric acid, cause degradation of acid-sensitive materials, including limestones and concrete (Gu et al., 2000; Gugliandolo and Maugeri, 1990; Sand, 1994). The reaction of stone carbonate with sulfuric acid also causes the formation of gypsum. The contribution of microorganisms to gypsum formation is unknown. Gypsum crystals combine with dust, aerosols, and other atmospheric particles to form black or brown sulfated crusts, which can tarnish the monument’s aesthetic appearance. The composition of these crusts varies and is dependent on the particular air- borne pollutants in individual areas (Saiz-Jimenez, 1993). Furthermore, effects of gypsum formation are not limited to aesthetic problems. While gypsum may temporarily passivate the limestone, the crust may ultimately exfoliate, causing extensive deterioration (Gauri and Holdren, 1981). Nitrifying chemoautotrophic bacteria are able to derive energy by oxidizing nitrogen-containing inorganic substrates (Mansch and Bock, 1996), ultimately resulting in the formation of nitric acid. The effect of biogenic nitric acid attack on stone was investigated and was compared to the effects of a smoggy atmosphere (Mansch and Bock, 1996). Mansch and Bock (1996) indicated that micro-

78 SCIENTIFIC EXAMINATION OF ART

biologically formed nitric acid corrosion was eight times more harmful than the corrosion caused by smog. The relative significance of chemical and biological processes in most locations is unknown. It is probable that there is synergism between the two processes. There is extensive evidence that in urban environments, a wide range of hydrocarbons is adsorbed to historic buildings. The growth of hydrocarbon- utilizing bacteria on these buildings may increase the deterioration rate of the stone (Mitchell and Gu, 2000).

ANALYSIS OF MICROBIAL POPULATIONS AND PROCESSES Microbial growth on stone has traditionally been analyzed using methods that rely on the ability of microorganisms to grow on culture media. In recent years, however, molecular techniques that exploit natural variation in DNA sequences have been developed to enumerate and identify microorganisms. Culture-based and molecular methods each have advantages and disadvantages. For example, despite the simplicity and ease of use of culture-based techniques, they routinely detect less than 1 percent of microorganisms present in environmental samples (Pace, 1997). Very often the physiological state observed in culture does not represent the organism’s activity in situ (Bonin et al., 2001). Molecular techniques avoid the selective bias of culturing but require significant expertise and may introduce other sources of error especially in natural samples containing multiple species’ templates (Thompson et al., 2002). The inability to detect the majority of organisms limits the usefulness of culture-based techniques for studies in which the goal is an accurate description of the microbial community. However, in studies where the goal is a comparison between treatments, analysis of a fraction of the community may be sufficient (Lemke et al., 1997). Further- more, molecular analyses yield little information about the function of organisms. Molecular methods are insufficient for investigating the production of metabolic products and for investigating the effects of these products on mineral dissolution. These data are required if we are to understand the underlying processes of deterioration. In addition to phylogenetic descriptions, the deterioration of stone has been examined using a variety of techniques, which have some utility in the quantification of biodeterioration. These methods include depth measurements using cali- pers, use of reference surfaces, macro-stereophotogrammetry, ion measurements of water runoff, or acid extraction (Winkler, 1986). Scanning electron microscopy (SEM), which is used to observe surface degradation, fails to detect changes below the surface. Nuclear magnetic resonance (NMR) measures changes in pore size distribution within stone (Alesiani et al., 2000) but gives no information about the microorganisms. Both SEM and NMR require destruction of the sample. Acoustic wave velocity provides information about subsurface discontinuities in

BIODETERIORATION OF STONE 79

stone (Papida et al., 2000). However, results generally must be correlated with other measurements, such as change in stone mass. X-ray computed tomography (CT) has been used extensively for the nondestructive visualization of objects in medical research (Berland, 1987), paleontol- ogy (Conroy and Vannier, 1984), soil, and sediments (Phogat and Aylmore, 1989). Recently attempts have been made to adapt CT for use with stone samples from building materials (Jacobs et al., 1995). These standard CT machines are large and have poor resolution. A recently developed small desktop micro computed X-ray tomography (MicroCT) was used to analyze stone samples at high resolution (Ruegsegger et al., 1996). In addition to being nondestructive, MicroCT provided images of the interior and exterior of stone objects, as well as three-dimensional reconstructions of the samples. Marble blocks were exposed to 1 mM sulfuric acid, which is typically associated with acid rain, and the effect was assessed by MicroCT according to McNamara et al. (2002). Exposure to 1mM sulfuric acid caused little change in mass, surface area, or solution pH value. An 18 percent increase in surface area was observed using MicroCT, possibly indicating the formation of a gypsum crust protecting the stone from further dissolution. A three-dimensional reconstruction of horizontal scans can be seen in Figure 6. MicroCT is a useful technique for quantifying the effects of microorganisms and their exudates. MicroCT is also of use for long-term studies, because it is nondestructive and allows for repeated scans on the same sample. Quantitative density measurements can be recorded over time. Atomic force microscopy (AFM) has previously been used to study the kinetics of calcite dissolution when exposed to water and simple acids (e.g., Shiraki et al., 2000). Using a time series of AFM micrographs (see Figure 7), a geometric analysis of surface topographical changes can be used to calculate a microscopic dissolution rate. The AFM chamber can be concurrently used as a flow-through reactor to determine macroscopic dissolution rates (Shiraki et al., 2000) with

FIGURE 6 A three-dimensional reconstruction of microCT planar scans of a marble sample before (a) and after (b) treatment for four days with 1 mM sulfuric acid. The surface shows formation of a gypsum layer.

80 SCIENTIFIC EXAMINATION OF ART

FIGURE 7 An atomic force micrograph of a pit forming in a calcite surface. By monitoring the step velocity of the pit expansion a dissolution rate can be calculated.

methods described in Duckworth and Martin (2003). The AFM may prove useful for quantifying the effect of bacterially produced acids and polymers on stone dissolution (Perry et al., 2004).

STRATEGIES FOR CONTROL OF BIODETERIORATION

Environmental Control Microorganisms can persist in dry environments. Active metabolism, however, requires appropriate levels of relative humidity and temperature. A combination of low humidity and low temperature is the simplest way to control microbial growth, but this treatment may be less effective for control of fungi (Gu et al., 1998b) and is impractical in outdoor situations. Regular cleaning may be the most effective treatment for preventing biofilm formation and subsequent biodeterioration of materials in historic buildings and monuments (Krumbein et al., 1992).

BIODETERIORATION OF STONE 81

Biocides The application of biocides has become a routine practice in the conservation of cultural heritage materials. However, environmental issues have severely limited the number of available effective biocidal chemicals for use in conservation (Bingaman and Willingham, 1994). Biofilm bacteria respond differently to biocides and are generally more resistant than unattached cells (McFeters et al., 1995). Because microorganisms are capable of rapidly acquiring chemical resistance, no one chemical can be relied on for long-term use; frequently several chemicals need to be combined in order to achieve effective eradication of biofilm populations. Biocides are a difficult tool for preservation, because many are too caustic for environmental use, they are not strong enough to discourage microbial growth, or the microorganisms ultimately develop resistance.

Consolidants Consolidants have been used for some time to conserve archaeological stone from biological and chemical weathering (Selwitz, 1992). Consolidation is a means of generating structural strength in disintegrating material and is an artificial means of repairing the damage caused by natural processes (Crafts Council, 1992). The efficacy of consolidants on outdoor stone is controversial, because they can upset the natural saturation and evaporation of moisture from within the stone, often resulting in exfoliation and cracking of stone surfaces (Boyes, 1997). Appli- cation of consolidants is not easily reversible, which is a serious drawback when dealing with ancient monuments. Some consolidants may also discolor as they degrade because of aging, photochemical processes, and oxidation (Biscontin et al., 1976; Gonzalez, 2000). Two of the most common types of consolidants used for monuments and archaeological stone are ethoxysilanes and acrylic resins. Commercial consolidants are susceptible to biodegradation (Koestler et al., 1994). A wide range of consolidants has been tested for effectiveness to protect Maya limestone in Belize (Kumar and Ginell, 1995), and most of the polymers tested proved to be susceptible to microbial degradation under local environmental conditions of the Yucatán. However, guidelines are needed for systematic evaluation of candidate polymers and their suitability in specific applications. Physical conditions for biodeterioration may be particularly favorable in tropical and subtropical regions because of high temperatures and humidity. While problems associated with the use of consolidants for the protection of archaeological stone are numerous, they are only one of the means of preventing the disintegration of stone grains due to exposure and weathering. The addition of biocides to consolidants would help to prevent microbial degradation, increasing the longevity of the treatment. Commercially available and environmentally acceptable biocides could be used as additives in consolidants.

82 SCIENTIFIC EXAMINATION OF ART

CONCLUSIONS Stone cultural heritage materials are constantly at risk of deterioration by microorganisms. This risk is augmented in urban environments, where deposition of pollutants enhances the rate of deterioration. Microbiologists are actively pursu- ing the microorganisms responsible for deterioration and attempting to quantify their effects. The effects of metabolic products, the deterioration processes, and the responsible microorganisms are all being elucidated. This effort, however, requires a multidisciplinary examination of the geology, chemistry, and conservation of these sites.

ACKNOWLEDGEMENTS The authors would like to thank M. Breuker (National Park Service, Lowell, Mass.), M. Muilenberg (School of Public Health, Harvard University, Boston, Mass.), R. Muller (Beth Israel Deaconess Medical Center, Boston, Mass.), and J. Sembrat (Conservation Solutions, District Heights, Maryland.) for their collaboration on aspects of this manuscript. This work was supported in part by a grant from the National Science Foundation (BES-9906337). T. D. Perry is supported by a Samuel H. Kress Fellowhip.

REFERENCES

Alesiani, M., S. Capuani, F. Curzi, and B. Maraviglia. 2000. In International Congress on Deterioration and Conservation of Stone, ed. V. Fassina, pp. 579-585. Venice, Italy: Elsevier. Arya, A., A. R. Shah, and S. Sadasivan. 2001. Current Science 81:793-799. Berk, S. G., R. Mitchell, R. J. Bobbie, J. S. Nickels, and D. C. White. 2001. International Biodeterioration and Biodegradation 48:167-175. Berland, L. L. 1987. Practical CT: Technology and Techniques. New York: Raven Press. Bingaman, W. W., and G. L. Willingham. 1994. International Biodeterioration and Biodegradation 34:387-399. Biscontin, G., S. Frascati, and L. Marchesini. 1976. In The Conservation of Stone, pp. 731-747. Bolo- gna: Centro per la Conservazione delle Sculture all’Aperto. Bjordal, C. G., T. Nilsson, and G. Daniel. 1999. International Biodeterioration and Biodegradation 43:63-71. Bonin, P., J.-F. Rontani, and L. Bordenave. 2001. FEMS Microbiology Letters 194:111-119. Boyes, N. 1997. In Aspects of Stone Weathering, Decay, and Conservation: Stone Weathering and Atmo- spheric Pollution Network Conference (SWAPNET), eds. M. S. Jones and R. Wakefield, pp. 170- 178. London: Imperial College Press. Conroy, G. C., and M. W. Vannier. 1984. Science 226:456-458. Costerton, J. W., Z. Lewandowski, D. E. Caldwell, D. R. Korber, and H. M. Lappinscott. 1995. Annual Review of Microbiology 49:711-745. Crafts Council. 1992. In Adhesives and Coatings, pp. 123-131. London: Conservation Unit of the Museums and Galleries Commission in conjunction with Routledge. Di Bonaventura, M. P., M. D. Gallo, P. Cacchio, C. Ercole, and A. Lepidi. 1999. Geomicrobiology Journal 16:55-64. Duckworth, O. W., and S. T. Martin. 2003.Geochimica et Cosmochimica Acta 67:1787-1801.

BIODETERIORATION OF STONE 83

Fabbri, A. A., A. Ricelli, S. Brasini, and C. Fanelli. 1997. International Biodeterioration and Biodegra- dation 39:61-65. Flemming, H. C. 1998. Polymer Degradation and Stability 59:309-315. Flemming, H. C., and J. Wingender. 2001. Water Science and Technology 43:1-8. Ford, T., J. Black, T. Kelley, R. Goodacre, R. C. W. Berkeley, and R. Mitchell. 1991. Applied Environ- mental Microbiology 57:1595-1601. Garcia-Rowe, J., and C. Saiz-Jimenez. 1991. International Biodeterioration and Biodegradation 28:151- 163. Gauri, K. L., and G. C. Holdren, Jr. 1981. Environmental Science and Technology 15:386-390. Gonzalez, R. F. 2000. In International Congress on Deterioration and Conservation of Stone, vol. 2, ed. V. Fassina, pp. 235-243. Venice, Italy: Elsevier. Gorbushina, A. A., W. E. Krumbein, C. H. Hamman, L. Panina, S. Soukharjevski, and U. Wollenzien. 1993. Geomicrobiology Journal 11:205-221. Gu, J.-D., T. E. Ford, and R. Mitchell. 2000. In Uhlig’s Corrosion Handbook, 2nd ed., ed. R. W. Revie, pp. 477-491. New York: John Wiley. Gu, J.-D., D. B. Mitton, T. E. Ford, and R. Mitchell. 1998a. Biodegradation 9:39-45. Gu, J.-D., M. Roman, T. Eesselman, and R. Mitchell. 1998b. International Biodeterioration and Bio- degradation 41:25-33. Gugliandolo, C., and T. L. Maugeri. 1990. In The Conservation of Monuments in the Mediterranean Basin, ed. F. Zezza, pp. 221-224. Brescia, Italy: Grafo. Jacobs, P., E. Sevens, and M. Kunnen. 1995. Science of the Total Environment 167:161-170. Koestler, R. J., E. Santoro, L. Koepp, M. Derrick, J. Druzik, F. Preusser, A. Rodarte, and N. Valentin. 1994. In Research Abstracts of the Scientific Program: GCI Scientific Program Report, ed. J. R. Druzik, pp. 132-133. Marina del Rey, Calif.: Getty Conservation Institute. Krumbein, W., J. Braams, G. Grote, M. Gross, K. Petersen, V. Schostak, and T. Warscheid. 1992. In The International Conference, ed. R. G. M. Webster, pp. 237-238. Edinburgh: Donhead. Kumar, R., and W. S. Ginell. 1995. In Methods of Evaluating Products for the Conservation of Porous Building Materials in Monuments, pp. 163-178. Rome, Italy: ICCROM. Lemke, M. J., B. J. Brown, and L. G. Leff. 1997. Microbial Ecology 34:224-231. Mansch, R., and E. Bock. 1996. In Microbially Influenced Corrosion of Materials, eds. E. Heitz, H.-C. Flemming, and W. Sand, pp. 167-186. Berlin: Springer. McFeters, G. A., F. P. Yu, B. H. Pyle, and P. S. Stewart. 1995. Journal of Industrial Microbiology 15:333-338. McNamara, C. J., T. D. Perry, M. Breuker, and R. Mitchell. 2002. In Microbially Influenced Corrosion, Corrosion—2002, ed. B. Little, pp. 113-121. Houston: NACE International. Mitchell, R., and J. D. Gu. 2000. International Biodeterioration and Biodegradation 46:299-303. Pace, N. R. 1997. Science 276:734-740. Papida, S., W. Murphy, and E. May. 2000. In 9th International Congress on Deterioration and Conser- vation of Stone, ed. V. Fassina, pp. 609-617. Venice, Italy: Elsevier. Perry IV, T. D., O. W. Duckworth, C. J. McNamara, S. T. Martin, R. Mitchell. 2004. Environmental Science and Technology 38:3040-3046. Perry, T. D., C. J. McNamara, G. Hernandez-Duqué, and R. Mitchell. 2003. In Molecular Biology and Cultural Heritage, ed. C. Saiz-Jimenez, pp. 137-140. Lisse: A. A. Balkema. Phogat, V. K., and L. A. G. Aylmore. 1989. Australian Journal of Soil Research 27:313-323. Rittman, B. E., M. Pettis, H. W. Reeves, and D. A. Stahl. 1999. Water Science and Technology 39:99- 105. Rölleke, S., A. Witte, G. Wanner, and W. Lubitz. 1998. International Biodeterioration and Biodegrada- tion 41:85-92. Rubio, R. F., and F. C. Bolivar. 1997. International Biodeterioration and Biodegradation 40:161-169. Ruegsegger, P., B. Koller, and R. Muller. 1996. Calcified Tissue International 58:24-29. Saiz-Jimenez, C. 1993. Atmospheric Environment, Part B—Urban Atmosphere 27:77-85.

84 SCIENTIFIC EXAMINATION OF ART

Saiz-Jimenez, C. 1999. Geomicrobiology Journal 16:27-37. Saiz-Jimenez, C., and L. Laiz. 2000. International Biodeterioration and Biodegradation 46:319-326. Sand, W. 1994. Werkstoffe und Korrosion 45:10-16. Sand, W., and E. Bock. 1991. International Biodeterioration and Biodegradation 27:175-183. Schabereiter-Gurtner, C., G. Pinar, W. Lubitz, and S. Rolleke. 2001. Journal of Microbiological Meth- ods 47:345-354. Selwitz, C. M. 1992. In Materials Research Society Symposium, vol. 267, pp. 925-941. Los Angeles: Getty Conservation Institute. Shiraki, R., P. A. Rock, and W. H. Casey. 2000. Aquatic Geochemistry 6:87-108. Smidsrød, O., and A. Haug. 1965. Acta Chemica Scandinavica 19:329. Thompson, J. R., L. A. Marcelino, and M. F. Polz. 2002. Nucleic Acids Research 30:2083-2088. Tomaselli, L., G. Lamenti, M. Bosco, and P. Tiano. 2000. International Biodeterioration and Biodegra- dation 46:251-258. Urzi, C., C. Lisi, G. Criseo, and A. Pernice. 1991. Geomicrobiology Journal 9:81-90. Welch, S. A., and P. Vandevivere. 1994. Geomicrobiology Journal 12:227-238. Wheeler, M. H., and A. A. Bell. 1988. In Current Topics in Medical Mycology, vol. 2, ed. M. R. McGinnis, pp. 338-387. New York: Springer-Verlag. Winkler, E. M. 1986. APT Bulletin 18:65-70.

Techniques and Applications

Analytical Capabilities of Infrared Reflectography: An Art Historian’s Perspective

Molly Faries Instituut voor Kunst-en Architectuurgeschiedenis Rijksuniversiteit Groningen Groningen, The Netherlands

Department of the History of Art Indiana University Bloomington, Indiana

ABSTRACT The technique traditionally known as infrared reflectography (IRR) is today only one of many different possibilities for imaging in the infrared. Some cameras developed recently are focal plane arrays based on materials that were formerly classified, and they offer the option of using filters for broad- or narrowband imaging. This paper discusses the application of both traditional and newer cameras to the imaging and interpretation of the layered, pictorial stages of paintings. In many cases infrared can “see through” paint layers to the underdrawing, the layout drawing an artist makes before the application of color; a painting can then be understood in terms of compositional evolution from the layout to the surface. Layers that are opaque to infrared can also register, and may provide equally valuable information about the painting process. The paintings and drawings mentioned in this text derive from different art historical periods, from early Netherlandish paintings of the fifteenth and sixteenth centuries to works by later artists such as Rembrandt and Vincent Van Gogh. Researchers in this field now need to recognize the expanding capabilities of infrared imaging and at the same time define more precisely the appropriate parameters of its use. As an imaging technique, infrared reflectography (IRR) is a method that allows sophisticated types of visual analysis. It accesses various stages in a work of art, particularly those that lie beneath the paint surface. Although it is also possible to obtain different responses of colors at different wave-

88 SCIENTIFIC EXAMINATION OF ART

lengths in the infrared, the provisional identification of pigments that this phenomenon might allow will not be discussed here. Instead, this paper emphasizes infrared’s capacity to “see through” paint, as well as the opposite: infrared’s ability to detect paint layers that remain impervious to infrared light. The interpretation of the infrared imaging of paintings therefore depends on an informed reading of induced opacities and transparencies of paint. The infrared vidicon is the device that has been used traditionally for IRR. This is hardly an emerging technology since it has been in continuous use in art history and conservation for nearly three decades. The method was developed in the late 1960s by the Dutch physicist, J. R. J. van Asperen de Boer, to improve upon the results of infrared photography (van Asperen de Boer, 1970). The advantages of IRR were immediately obvious at the time: The blues and greens that had remained opaque in infrared photography were rendered much more transparent, and underdrawings were revealed to a much greater extent. To date, almost all the results produced by this technique have been interpreted in the field of art history. In this application the IRR equipment rarely stays in the lab, but it is taken into the field (i.e., to museums, churches, and private collections). The infrared vidicon, housed in a high-resolution television camera, is portable, as is the television monitor, where the infrared image called a reflectogram can be viewed and documented by photography or captured to computer. The only additional equipment required is a source of illumination and a sturdy tripod. As a result, what has come to be known as the IRR expedition has been essential in gathering basic information about the working procedures of related artists or artistic groups.

INFRARED IMAGING DEVICES In addition to the vidicon used in IRR, there are now perhaps as many as 15 other possibilities for imaging in the infrared. These range from inexpensive CCD (charge-coupled device) cameras with sensitivity close to that of visible light, some of which have been recommended for use in conservation (Meyer and Raquet, 2002), to more advanced focal plane arrays based on materials that were declassified in the 1990s, such as indium gallium arsenide or platinum silicide. Any of these imaging devices can be used with filters, but since the high-end focal plane arrays do not usually require as much increase in illumination when using filters, they can more easily be adapted to study works of art. If required, imaging can concentrate in narrow-wavelength bands, such as a range near 1100 nanometers (nm) for the visualization of brownish inks that quickly become transparent in infrared, or around 2 microns for the penetration of particularly opaque paints, such as malachite green. In practice multispectral imaging already exists, although it is not clear whether all users of infrared imaging sufficiently distinguish one

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 89

range of infrared from another, or exploit different ranges of infrared in the most appropriate manner. This is hardly surprising, since some of the cameras are so new that no established patterns of use have been developed. This paper makes reference to infrared results from three different types of imaging devices. One is a camera that is sensitive closer to the range of infrared photography or the very near infrared: the MuSIS CCD camera, sensitive to around 1150 nm. (The MuSIS camera has a CCD sensor with 1024 x 768 resolution and a spectral response from 320 to 1150 nm, including two selectable ranges in the near infrared, from 700 to 950 nm and from 950 to 1150 nm.) The second is the “classic” IRR vidicon. (The IRR equipment used in this study is a Grundig 70 H television camera set at 875 lines and outfitted with a Hamamatsu N 214 infrared vidicon that works in a range from 900 to around 1600/2000 nm, a TV Macromar 1:2.8/36 mm lens, and Kodak 87 A filter, with a Grundig BG 12 monitor. Documentation is done with a Canon A-1 35 mm camera, a 50 mm Macrolens, and Kodak Plus X film. Currently the reflectogram negatives are digi- tized and assembled into composites using VIPs and/or Adobe Photoshop.) The third type of camera is the platinum silicide focal plane array: both the Mitsubishi M700 thermal imager and the European AEG Infrarot Module camera. (Both cameras are Stirling-cooled 640 × 480 platinum silicide focal plane arrays with digital and video output, outfitted with Nikon f 3.5, 55 mm lenses, so that the broadband spectral response ranges from 1.1 to 2.5-2.6 µm. For the Mitsubishi camera belonging to the Rijksuniversiteit Groningen, images are captured with Arte software and assembled with Panavue.) More information on the AEG camera has been published elsewhere (van der Weerd et al., 2001). The vidicon and focal plane arrays have a spectral response much further in the infrared, and they can thus be categorized as devices that meet the specifica- tions for infrared reflectography. These cameras all work in the region of infrared that has been determined by theoretical and experimental research to be essential for the optimum transparency of traditional nonsynthetic pigments. The work of J. R. J. van Asperen de Boer, as well as that done later by researchers at the National Gallery of Art in Washington, D.C., has determined that devices must be sensitive in the range around 2 µm—and more generally from 1.5 to 2 µm—for optimum penetration of paint and the visibility of most underdrawings (van Asperen de Boer, 1970; Walmsley et al., 1994).

EXAMPLES The greatest penetration of paint has always been a prime requirement of IRR, along with high resolution of the imaging device. This was the case when the technique was developed with the objective of disclosing underdrawings, and this remains the case today. Although it is to a degree self-evident that these features would facilitate visual interpretation, several examples can further illustrate the types of reasoning that become possible when IRR imaging is optimized.

90 SCIENTIFIC EXAMINATION OF ART

The IRR vidicon was used recently to study the painted wings of an early sixteenth-century altarpiece by Joos van Cleve (now in Warsaw, Museum Naradowe). (The painting was studied in the context of a Rijksuniversiteit Groningen research project, Painting in Antwerp Before Iconoclasm: A Socio- Economic Approach, funded by the Netherlands Organization for Scientific Re- search.) In this case IRR obtains what can be considered almost ideal results (see Figures 1 and 2). The different colors and color mixtures of the painted scenes become almost completely and uniformly transparent. This is true not only of blues and greens, which sometimes maintain more opacity, but also of whites, which if painted thickly, can block the penetration of infrared light. The infrared documents give one the sense of looking directly at the underdrawing without any interference from the paint. The only paint that continues to register has been mixed with dark or black pigments. The underdrawing of the altarpiece registers very clearly, since the underdrawing material, a dark black ink, is able to absorb infrared radiation strongly through the paint layers, as opposed to the white ground, which is highly reflective. The other panels in the altarpiece responded to infrared in a similar way, so that in this case IRR obtained both a complete and consistent record of the underdrawing in all 11 panels of the entire work. The type of underdrawing revealed can be considered representative of the master’s working method. Many color notations were discovered in the underdrawing and since no colored area was obscured by paint, all the notations present were made visible, both for single colors and for color mixtures. These give an insight into the artist’s shop practice. They may have been directives to assistants and/or used to estimate how much of a certain color had to be ground and prepared for painting. Because the overall underdrawing was revealed so clearly in terms of its type and function, it was possible to identify it more as an indicator of workshop routine than of a given artist’s personal drawing style. In this instance the master of the shop took over Albrecht Dürer’s woodcuts as models, probably for the woodcut’s more diagrammatic method of rendering form. This woodcut look has also been found in the underdrawings of other early sixteenth-century Antwerp masters, so that scholars in the field now recognize it as a production method that cuts across different shops. Such a carefully crafted drawing must represent a considerable investment of time and effort, whether or not it represents the work of the master of the shop or an assistant, who would have to be trained to reach a high degree of proficiency. This layout method must have been found useful as a way of facilitating the painting process, and would fit with the streamlining of workshop procedure that we know was occurring in Antwerp at the time. A second example can demonstrate how high-resolution imaging can bring out the character of an underdrawing. Platinum silicide focal plane arrays are capable of very sharp images, equal to and sometimes better than the resolution of the best IRR vidicons (van der Weerd et al., 2001). A camera of this type was used to distinguish the very different underdrawings in two identical Madonna

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 91

and Child compositions from the workshop of the sixteenth-century Dutch painter, Jan van Scorel (Faries et al., 2000). In the digital composites of one of the Madonnas a researcher can obtain a good visual sense of the material of the underdrawing, a dry black chalk, and the artist’s handling of it, which is remarkably free. The second Madonna displays an underdrawing that is entirely different: It seems to have been executed in a hesitant, painstaking manner, and it includes features that are based on fully realized forms of a preexisting model. When the underdrawings of both paintings are compared in their entirety, it becomes evident that there is a complicated compositional change in the first Madonna that is not repeated in the second. The pose of the Christ child was changed significantly between the underdrawing and paint stage of the freely underdrawn Madonna, while the Christ child’s pose in the second Madonna follows the paint surface of the first. These paintings are both products of the same shop. Nonetheless, the infrared comparison provides information about the sequence in which these paintings were made; the first of the Madonnas exhibits a more creative and idiosyncratic manner of working, while the execution of the second version adheres more tightly to predetermined motifs. Many more examples of the interpretation of underdrawings can be found in the several surveys that have been published recently covering the extensive research that IRR has made possible in the last few decades (Faries, 2001). When IRR can be shown to provide clearly legible and reliable results, it gives basic validity to various types of studies. These can range from investigations of individual painters to more general studies of artistic production and painting technology. Infrared results are not always as straightforward as the examples just mentioned, even if the imaging device has optimum resolution and penetration. The condition of a painting can always influence infrared results, and can greatly complicate the reading of infrared documents. (Infrared imaging, in fact, usually provides a great deal of information about the condition of paintings, the types of fills, and the extent of retouching.) Other factors can also have an effect, such as colored grounds and different underdrawing and painting materials. During an infrared examination, underdrawings that are very faint and/or executed in brownish pigments or other colors can become transparent along with the paint. This is a fairly well-known phenomenon, and can occur with any one of the infrared imaging devices currently available. Recent infrared study of a Nativity of Christ from the workshop of Jacob Cornelisz van Oostsanen (Utrecht, Centraal Museum) showed that while the underdrawing could be made visible by the vidicon IRR camera, it was rendered transparent by the broadband imaging of the Mitsubishi focal plane array (see Figures 3 and 4). Obtaining complete and representative material in this case would require a combination or cameras or the use of filters. If the camera allows the use of a variety of filters, one can work from broadband imaging back through ranges that come closer and closer to visible

92 SCIENTIFIC EXAMINATION OF ART

FIGURE 1 Joos van Cleve and workshop. Detail of Saint John the Evangelist in an altarpiece wing, Warsaw, Museum Naradowe (Photo: Micha Leeflang).

light until the underdrawing can be imaged. The drawback, however, is that the closer the imaging approaches visible light, the more opaque are greens, blues, and white, and the more restricted is the view of the underdrawing. Most infrared imaging, in fact, falls between the two extremes mentioned above and the case described here. Ideally IRR will render paint completely transparent and allow the

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 93

FIGURE 2 Same detail as Figure 1. Digital composite of IRR reflectograms at 0.9-1.6/2 µm, showing elaborately worked out underdrawing (IRR and digital composite: Molly Faries).

94 SCIENTIFIC EXAMINATION OF ART

FIGURE 3 Workshop of Jacob Cornelisz van Oostsanen. Detail of Joseph in a Nativity scene (Utrecht, Centraal Museum). Digital composite of IRR reflectograms at 0.9-1.6/2 µm in which underdrawing appears in the sleeves and robe of Joseph (IRR and digital composite: Molly Faries).

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 95

FIGURE 4 Same detail as Figure 3. Digital composite of IRR reflectograms at 1.1-2.5 µm in which the underdrawing becomes partially or totally transparent (IRR and digital composite: Molly Faries).

96 SCIENTIFIC EXAMINATION OF ART

underdrawing to register strongly, while in the least responsive cases most of the paint will remain opaque and it will not be possible to visualize the underdrawing, if indeed it is present at all. Researchers must therefore always keep in mind that visibility of the underdrawing depends on a number of variable factors: the materials and technique of the underdrawing, the overlying pigment, and the spectral response of the detector (van Asperen de Boer, 1970; Walmsley et al., 1994). Although imaging close to the range of visible light has its obvious draw- backs, there is one recent application that exploits the possibilities of the 700- 1150 nm range. This concerns the infrared study of drawings on paper (and could apply equally well to archival documents and perhaps to illuminated manuscripts as well). The development of this examination method has been led by Hans Scholten of the Dutch company, Art Innovation, and should shortly appear in print (Havermans et al., 2003). In this type of study the MuSIS CCD camera captures a high-resolution color image in visible light and an image in the near infrared. These are then used to generate a false color infrared image using the color channels of Adobe Photoshop. Inks such as iron gall inks become transparent in this range of infrared light, but in false color infrared the differences in drawing materials and changes in condition can still be detected. Carbon black inks can be distinguished from iron gall ink; retouches can often be made visible and in some cases the presence of a corrosion product formed by iron gall ink can be detected. This is a promising area for further research, especially since the sensitivity of the MuSIS camera allows the low levels of illumination required for the study of works on paper or parchment. It must be emphasized, however, that this is an application for drawings that are not covered by paint. There seems to be a prevailing notion that infrared reflectography is only useful for revealing underdrawings in paintings. If a researcher suspects that it will not be possible to make underdrawing visible, then the technique is simply ignored. Because infrared can see into the layers making up a painting, it can frequently reveal other equally valuable information about the painting process. Changes are routinely made during the painting process; they can occur during the realization of the original image, or they may reflect changes made to the work after its completion. Many can be detected using such traditional methods as X-radiography, but many are easier to visualize in infrared. In these cases infrared imaging can work in exactly the same way it does when revealing underdrawings: The addition of black to the dark and grey portions of an underlying layer or form will absorb infrared and register. In this way unsuspected modifications and/or particular aspects of painting technique can be revealed. Ironically an imaging device that is optimized for the penetration of paint is still required. A researcher must be certain that infrared has revealed a “true” opacity and not one that is the result of using a camera whose sensitivity is too close to visible light. Selected examples from Faries’s own research can illustrate some of the possibilities of this type of study. One concerns a small Crucifixion panel by an

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 97

anonymous fifteenth-century German master (Houston, Museum of Fine Arts). In the left background of the painting, colors of an underlying painted form show through the paint of the sky (see Figure 5). Further examination of this work revealed that the painting process was surprisingly complex, with some forms underdrawn that were later painted out, and then painted in again. Although X- radiography might seem the method of choice to elucidate these overlapping paint stages, infrared reflectography actually imaged some of the underlying forms in more detail. In X-ray the shape under the background sky could hardly be discerned (either because it was painted extremely thinly or with little dense paint, such as lead white), but in infrared the shape could be identified as a small castle painted on the horizon (see Figures 6 and 7). Because sufficient dark or black pigment was used in its execution, the gate, castle walls, and castle building and tower with crenellations, windows, and varying rooflines could all be distinguished (Faries, 1991). Overpaintings can change not only the appearance of a painting but also its function, and sometimes occur when a new portrait is painted in over the image of an original donor. Other cases indicate a slightly more complicated situation: A work could be supplied with the intention of adding in portraits or coats of arms at a later date. We know now that altarpieces were occasionally ordered or bought on the art market with the middle panel completed but with the wings left blank, painted only with a dark, base color to protect the ground and wood. Infrared, of course, can detect this difference in ground color. In a triptych in the Catharijneconvent Museum in Utrecht, infrared study revealed that the image in the middle panel was built up from a light, reflective ground, while the portraits were added in over a dark underlying layer. The work was completed in two stages, and in its final form the altarpiece represents the work of two unrelated masters: an artist from Antwerp who painted the central scene and supplied the panels of the altarpiece and an Amsterdam master, Dirck Jacobszoon, who executed the portraits that were added to the wings (Faries, 2001). In the same way, infrared can disclose other opacities that are part of developing painting techniques in the fifteenth and sixteenth centuries. In the sixteenth century, painters began to use dark pigments more and more frequently for shading and/or undermodeling. The painter Herri met de Bles is known to have used a gray layer as undermodeling for some of his blues. The infrared study of a landscape by this artist in the Cincinnati Museum of Art clearly singled out some blue drapery as an opaque dark form, while surrounding reds and greens became completely transparent. Further study clarified the infrared results: A cross section showed that the blue pigment, smalt, had been applied over a layer of black particles in a white matrix (Faries and Bonadies, 1998). Blue smalt on its own is known to become transparent in infrared. Developments of this nature eventually lead to the methods of seventeenth-century masters, such as Rubens, Jordaens, and others, and their use of an undermodeling paint stage that is often described as “dead coloring.”

98 SCIENTIFIC EXAMINATION OF ART

FIGURE 5 Anonymous German Master, Crucifixion, Houston, Museum of Fine Arts (Photo: Molly Faries).

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 99

FIGURE 6 X-radiograph detail of Figure 5 (Photo: Molly Faries).

Infrared has not been thought useful for the study of Rembrandt, but in the last few years it has become apparent that we need to look at Rembrandt’s paintings again. The recent IRR study of the painting known as Self-Portrait with Gorget (in the Mauritshuis in The Hague) revealed traces of underdrawing along with undermodeling. The underdrawn shapes and undermodeling have been interpreted by scholars as the result of compositional transfer, that is, a replication of the surface of another closely related portrait of Rembrandt that is in Nuremberg. This unexpected discovery has prompted a lively debate about the attribution of The Hague painting. Some scholars still believe both paintings are by Rembrandt, but others now consider the Nuremberg painting to be the authentic Rembrandt, and The Hague painting the work of a studio assistant or other anonymous seventeenth-century master (see articles by Edwin Buijsen, Jørgen Wadum, and Eric Jan Sluijter in the 2/4 issue of volume 114 [2000] of the journal Oud Holland). Still, at this point in time, little is known about comparable painting practices. The disclosure of the shaped layout of The Hague portrait

100 SCIENTIFIC EXAMINATION OF ART

FIGURE 7 Same detail as Figure 5. IRR reflectogram assembly at .9-1.6/2 µm (IRR and assembly: Molly Faries).

suggests that infrared must be used more systematically in future technical studies of seventeenth-century paintings, and that the undermodeling stage must be taken more seriously. The last examples derive from research in progress (fieldwork done with graduate students at the Rijksuniversiteit Groningen). They concern several paintings by Vincent Van Gogh, although the current findings would in theory apply to any painting with lighter colors on the surface and darker colors in underlying forms. It is now known that nearly a third of Van Gogh’s works were painted on top of previous compositions. Infrared reflectography, however, has rarely been applied to study this phenomenon, since it has generally been assumed Van Gogh’s paint would be too thick to reveal anything except the surface. In one of the paintings studied, Interior of a Restaurant (Otterlo, Kröller- Müller Museum), it was not possible to image any of the underlying composition, but underdrawing was revealed. This took the form of a rectangular border with crossed diagonals, part of the perspective frame that Van Gogh is known to have used to lay out his paintings and drawings. This painting was studied using both

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 101

the MuSIS CCD camera and the Mitsubishi focal plane array. It was found that the underdrawing registered in infrared light accessed by both of the cameras, but the surface colors reflected infrared radiation in a completely different way. Results were different in the second work studied, A Patch of Grass (Otterlo, Kröller-Müller Museum) (see Figure 8). Infrared reflectography was able to image underdrawing as well as dark painted forms at different levels in the painting (see Figure 9). The painted composition on the surface was found to be based on a perspective frame and was easily made visible with the Mitsubishi camera. This camera also disclosed some broad horizontal bands of paint near the center of the painting that have yet to be explained—as well as the darker and lighter shapes of an underlying form. Once the canvas was turned into an upright position, it was easy to recognize the form as a portrait of a woman, similar in appearance to portraits from the period of Van Gogh’s well-known Potato Eaters. The underlying portrait also registers in X-radiography. Both X-ray and infrared reveal the basic shape of the portrait, but the infrared image provides additional information about dark modeling strokes around the mouth and eyes and in the hair. In

FIGURE 8 Vincent Van Gogh, A Patch of Grass, Otterlo, Kröller-Müller Museum. (Photo: Molly Faries).

102 SCIENTIFIC EXAMINATION OF ART

FIGURE 9 Detail of Figure 8. Digital composite of IRR reflectograms at 1.1-2.5 µm; a shadow from the easel appears on the left side (IRR and digital composite: Molly Faries).

this case infrared has imaged at least three unseen stages in the painting: an underlying composition, along with the forms of an as-yet-unidentified intermediate stage, and the perspective frame for the surface landscape, which has been applied over a presumably very thin layer of lead white.

CONCLUSION Infrared reflectography is clearly a versatile technique. This paper has attempted to show how the method can facilitate art historical interpretation and how its

ANALYTICAL CAPABILITIES OF INFRARED REFLECTOGRAPHY 103

application may be extended for study beyond the disclosure of underdrawings. The research discussed here done with both the “classic” infrared vidicon and the newer focal plane arrays would indicate that there is reason to maintain the term “infrared reflectography” as a concept in infrared imaging. For the optimum penetration of paint and visibility of most underdrawings, devices that work beyond 1.5 µm and include the region around 2 µm are required. There are, nonetheless, appropriate applications for devices that are only slightly sensitive in the infrared up to 1.1 µm. Infrared reflectography has always had the advantage of being a portable, noninvasive scanning technique, and with new developments in the computer- ized assembly of reflectograms, it has become easier to document paintings in their entirety. If in coming years X-radiography were also developed as a portable scanning technique, the usefulness of both methods would be enhanced when applied concurrently. With such combined information about the overall evolution of a painting at hand, researchers would be in a better position to refine the questions and selection of methods required for further study, and for any sampling that might be required. One can also envision linking IRR and X-ray results with those from other portable methods that analyze the components of paint, such as X-ray fluorescence. Researchers would then be able to use integrated scanning systems to obtain a dimensional image of the stages of execution coupled with information from materials analysis of the paint surface. The parameters of use for infrared imaging can be determined only by a combination of theoretical study and experiential research. The theoretical work that has already been done gives us a basic guide for infrared applications (van Asperen de Boer, 1970; Walmsley et al., 1994). Such work is essential in modeling the interrelationship of wavelength, pigments, grounds, and underdrawing materials. Nonetheless, reconstructions and test planks cannot replicate all the com- plexities of historical painting techniques in actual works. Nor can they approximate the effects of aging or the varied thicknesses and mixtures of paint. As indicated by the examples discussed in this text, infrared imaging can produce extremely varied results. It is up to the researcher to determine how to adapt and optimize the possibilities of infrared imaging as the examination unfolds. In addition, wavelength response itself is not always an absolute factor. Tests done with some indium gallium arsenide focal plane arrays have shown that the high quantum efficiency of these cameras can improve their performance (van Asperen de Boer, 2003). Systematic comparisons of the same paintings using different infrared imaging devices at different wavelength ranges should therefore be pursued, and this author has already begun to carry out such work. This research could produce a set of reference images that would help in establishing reasonable expectations for researchers and complement the already existing theoretical guidelines for infrared studies. Infrared reflectography is becoming both more sophisticated and more tech- nologically complex. Our current knowledge about the painting techniques of

104 SCIENTIFIC EXAMINATION OF ART

different periods and different artists will undoubtedly need to be expanded. Analytical interpretation in this field depends on the expertise of the researcher, and can even be described as connoisseurship of technical documentation. There is thus every reason to pursue the systematic cataloguing of painting collections as well as larger comparative studies using infrared, and to encourage interdisciplinary research that combines the expertise of technical art historians with that of research-oriented painting conservators and scientists who have a nuanced understanding of art. This text is based on the paper given at the Arthur M. Sackler colloquium, Scientific Examination of Art: Modern Techniques in Conservation and Analysis, held at the National Academy of Sciences, Washington, D.C., March 19-21, 2003. Molly Faries recently (September 2004) began a related research project: “Infra- red Reflectography: Evaluative Studies,” part of the De Mayerne Research Programme on Molecular Studies in Conservation and Technical Studies in Art History funded by the Netherlands Organization for Scientific Research (NWO).

REFERENCES

Faries, M. 1991. Studying Underdrawings: Notes for the Cologne Workshop. Bloomington, Ind., § 1.212. (This reader, under copyright, is used by many who work in the field of technical studies in art history.) Faries, M. 2001. Reshaping the field: The contribution of technical studies. In Early Netherlandish Painting at the Crossroads: A Critical Look at Current Methodologies, ed. M. W. Ainsworth, pp. 70-105. New York: Metropolitan Museum of Art. Faries, M., and S. Bonadies. 1998. The Cincinnati Landscape with the Offering of Isaac by Herri met de Bles: Imagery and artistic strategies. In Herri met de Bles, Studies and Explorations of the World Landscape Tradition, eds. N. E. Muller, B. J. Rosasco, and J. H. Marrow, pp. 73-84. Turnhout: Art Museum of Princeton University in collaboration with Brepols Publishers. Faries, M., L. Helmus, with contributions by J. R. J. van Asperen de Boer. 2000. The Madonnas of Jan van Scorel, Serial Production of a Cherished Motif (exhibition catalog). Utrecht: Centraal Mu- seum. Havermans, J., H. Abdul Aziz, and H. Scholten. Non-destructive detection of iron gall inks by means of multispectral imaging. Restaurator 24(2003) no. 1, pp. 55-60 and no. 2, pp. 88-94. Meyer, M., and M. Raquet. 2002. Digitalfotografie für die Restaurierung Restauro 5:350-355. van Asperen de Boer, J. R. J. 1970. Infrared Reflectography: A Contribution to the Examination of Earlier European Paintings, Ph.D. thesis, University of Amsterdam. van Asperen de Boer, J. R. J. 2003. Slowly towards improved infrared reflectography equipment. In Recent Developments in the Technical Examination of Early Netherlandish Painting: Methodology, Limitations, and Perspectives (M. Victor Leventritt Symposium), ed. M. Faries and R. Spronk, pp. 57-64. Turnhout: Harvard University in collaboration with Brepols Publishers. van der Weerd, J., R. M. A. Heeren, and J. R. J. van Asperen de Boer. 2001. A European 640 x 486 PtSi camera for infrared reflectography. In Colloque XIII pour l’étude du dessin sous-jacent et de la technologie dans la peinture: la peinture et le laboratoire: la peinture et le laboratoire, ed. R. Van Schoute and H. Verougstraete, pp. 231-243. Leuven: Uitgeverij Peeters. Walmsley, E., C. Metzger, J. K. Delaney, and C. Fletcher. 1994. Improved visualization of underdrawings with solid-state detectors operating in the infrared. Studies in Conservation 39:217-231.

Color-Accurate Image Archives Using Spectral Imaging

Roy S. Berns Munsell Color Science Laboratory Chester F. Carlson Center for Imaging Science Rochester Institute of Technology Rochester, New York

ABSTRACT Digital imaging that includes spectral estimation can overcome limitations of typical digital photography, such as limited color accuracy and constraints to a predefined viewing condition or a specific output device. An example includes the use of ICC color management to generate an archive of images rendered for a specific display or for a specific printing technology. A spectral image offers enhanced opportunities for image analysis, art conservation science, lighting design, and an archive that can be used to relate back to an object’s physical properties. The Munsell Color Science Laboratory at Rochester Institute of Technology is involved in a joint research program with the National Gallery of Art in Washington, D.C., and the Museum of Modern Art in New York to develop a spectral-imaging system optimized for artwork imaging, archiving, and reproduction. Progress is being documented at the website www.art-si.org. This paper summarizes the scientific approach.

INTRODUCTION Imaging is an important technique in the scientific examination of art. Its main use has been for visual documentation. Photographs have long been used to document condition before and after transit, microscopic examinations, conservation treatments, and so on. They are used to enable color reproductions in books and from the Internet. Images using materials with spectral sensitivities in

105

106 SCIENTIFIC EXAMINATION OF ART

such non-visible regions of the electromagnetic spectrum as infrared and X ray are equally important to the visible spectrum. Although images are used to record scientific examinations, they are used infrequently as an analytical tool, that is, the amount of colorant in a photographic material would be used to relate to physical properties of the art. In contrast, astronomy, remote sensing, and medicine have exploited this capability for many years. The advent of digital imaging offers increased opportunities to exploit images for the scientific examination of art. A research program is underway at Rochester Institute of Technology to develop an image-acquisition system that records reflection information as a function of wavelength. The system initially is limited to the visible region. This publication will summarize our methodologies and give some performance examples. Full results, documentation, and demonstrations can be downloaded and viewed at www.art-si.org. At the end of this paper are relevant publications written by students, faculty, and staff of the Munsell Color Science Laboratory.

TECHNICAL APPROACH

Complete Sampling—Spectral Measurement A spectrophotometer records spectral reflectance or transmittance for a specific circular aperture; a single color is measured. By analogy a spectral-imaging system records spectral reflectance or transmittance for a projected scene at a specific spatial resolution; many colors are measured. One can envision a number of techniques to disperse light onto a detector plane. The technique we have taken is to couple a monochrome, area-array chaged-couple device detector with a liquid-crystal tunable filter. Successive images are captured, each image centered at a specific wavelength. Typically we capture 31 bands corresponding to 400-700 nm at 10 nm increments. As a measurement device, calibration is necessary. For each band, images are taken of a dark field (to remove fixed-pattern noise), several neutral diffuse papers (to compensate for lighting non-uniformity and optical flare), a pressed polytetrafluoroethylene tablet (to determine optimal exposure time), and a color target made from a number of colorants (to compensate for wavelength and geometry bandwidth). These targets are crucial to achieve acceptable performance. In general, spectroradiometry and imaging have greater uncertainty than contact spectrophotometry. Thus, it is necessary to derive transformations that minimize these uncertainties. A typical transformation is shown in Figure 1. The GretagMacbeth ColorChecker DC and a custom target of blue pigments mixed with titanium white were used to develop the transformation. This figure is a visualization of the matrix transformation from spatially corrected 31-band images to spectral reflectance factor images. The matrix contains 961 coefficients

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 107

0.025

0.02

0.015

0.01 0.04

0.005 0.02

0 0

Coefficient

700

700 Spectral625 reflectance factor (nm)

625 550 550

475 475

400 400

FIGURE 1 Visualization of calibration matrix for 31-band image-acquisition system.

(31 × 31). Ideally the matrix should have dominant diagonal and small off- diagonal coefficients. Figure 2 is an image of the well-known color target, the GretagMacbeth ColorChecker Color Rendition Chart. This independent evaluation target provides a method to benchmark color and spectral accuracy. Typical performance is shown in Figure 3 for these colors. The spectral accuracy was 1.4 percent root-mean-square (RMS) reflectance and an average color accuracy of ∆ 1.5 E00 under daylight (D65) and viewed by the 1931 CIE standard observer.

Subsampling—Spectral Estimation The system described in the previous section performs spectral measurement; there are the same numbers of image bands as wavelengths. The majority of natural and synthesized colorants have large-bandwidth absorption spectra in the visible region. Furthermore, there are not many sharp transitions from high to low reflectance (and vice versa). From a dimensionality reduction perspective it may not be necessary to collect images every 10 nm, that is, sub-sampling may not result in a loss of accuracy. For example, during the 1970s, many spectrophotom-

108 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 GretagMacbeth ColorChecker Color Rendition Chart.

0 1

1 Reflectance factor

0 400 700 400 700 400 700 400 700 400 700 400 700 Wavelength (nm) FIGURE 3 Typical spectral-measurement accuracy for the ColorChecker using a 31-band image-acquisition system (blue lines) compared with a small-aperture contact spectrophotometer (red lines).

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 109

eters used in color technology sampled the visible spectrum in 20-nm-wavelength increments and bandwidth. As the number of direct measurements reduces we are performing spectral estimation rather than spectral measurement. Suppose a painting was created from a single chromatic colorant and white (or paper in the case of a watercolor). Because the concentration of one colorant is being varied, one can measure the light reflection at a single wavelength, usually the wavelength of maximum absorption (minimum reflectance ignoring the white). A single image is captured; differences in gray level relate to differences in colorant concentration. At this wavelength, changes in concentration will result in the greatest change in reflectance (i.e., the greatest image contrast). If we measure the spectral absorption properties of the colorant using a spectrophotometer and determine the relationship between camera signals and concentration and between concentration and spectral reflectance (e.g., Kubelka-Munk theory, Beer’s law), the single image can be used to estimate a 31-band spectral image. This estimation process has also enabled significant data reduction. We need to archive only the single-band image. The spectral reflectances of the colorant and white, the transformation from camera signals to concentration, and between concentration and spectral reflectance are stored in the image tag. This is analo- gous to an ICC input profile except in this case, the profile performs spectral color management. This idea is extended to paintings created with many colorants. Principal component analysis (PCA) is used to define a set of statistical colorants. Because of the spectral properties of colorants in the visible region, the number of statistical colorants (eigenvectors) can vary between 5 and 16. The specific number depends on spectral accuracy requirements and the samples analyzed statistically. In general the imaging system captures the same number of images as the number of statistical colorants. A relationship is determined between the camera signals and statistical colorant amounts (principal components). Concatenating these various steps results in a transformation that relates camera signals to spectral reflectance. Principal component analysis can be interpreted as constraining the spectral outcome of the mathematical transformation, a type of spectral interpolation. With a large enough number of samples, we can eliminate the use of PCA. In its place we derive a direct transformation from camera signals to spectral reflectance. This method uses a singular-value-decomposition-based pseudo-inverse calculation in which several hundred thousand samples are used to estimate several-hundred-transformation coefficients. These many samples are acquired by considering each pixel of an image an individual data point. We have found that these two methods yield similar spectral accuracy. Both techniques are constrained in two ways. The first has to do with the camera. Performance depends on the spectral sensitivities of each camera channel. Optimal filter design has been studied for many years; unfortunately these filters, designed by simulation, cannot be fabricated. The practical solution is to

110 SCIENTIFIC EXAMINATION OF ART

select the best filters from those produced commercially. We have taken this approach. We have also used commercial cameras with color filter area-array sensors. With additional filtration using colored absorption filters, sets of color images are recorded. Three, six, or nine image planes (each triplet is the usual red, green, blue image) are related to three, six, or nine statistical colorants or directly to spectral reflectance. The second constraint is the dependence on a color target. The target is used to derive the mathematical transformation. Ide- ally, the target should have a number of colored patches sampling thoroughly the color gamut of materials to be imaged. The patches should be made from colorants with unique spectral properties. The gloss properties should be consistent. In essence there is an assumption that the color target has spectral properties that encompass those of the art to be imaged. Most commercial targets do not have these ideal properties. Despite these constraints, the method has proven to be nearly equivalent to 31-band spectral imaging. Using a color-filter-array camera and two absorption filters, the average performance for the ColorChecker was 1.6 percent RMS and ∆ 1.2 E00, plotted in Figure 4. The transformation matrix is plotted in Figure 5, derived using the ColorChecker DC. This transformation relates six camera sig-

1 Reflectance factor

0 1

0 400 700 400 700 400 700 400 700 400 700 400 700 Wavelength (nm)

FIGURE 4 Typical spectral-measurement accuracy for the ColorChecker using a two- filter color-filter-array image-acquisition system (blue lines) compared with a small-aperture contact spectrophotometer (red lines).

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 111

20 40

20 10

0 0 Coefficient –20 –10

–40 –20 730 6 680 5 630 580 4 Wavelength, nm –30 530 3 480 Channel 430 2 380 1

FIGURE 5 Visualization of calibration matrix for six-band image-acquisition system, achieved using two absorption colored glass filters and a color-filter area-array image- acquisition system.

nals to 36 wavelengths, totaling 216 matrix coefficients. At each wavelength, there should be at least one peak or valley.

Spectral Advantage A spectral image archive has a number of advantages over many current image archives. Sometimes, an archive is created by digitizing photographs. In other cases direct digitization is used with scanbacks, using repurposed flatbed scanning sensors. Film and scanbacks have spectral sensitivities quite different from the human visual system. As a result these archives require significant visual editing as part of the workflow. Thus, the archive is connected to a particular display, viewing condition, and observer. Color accuracy is limited. Color management principles can be used to reduce the reliance on visual editing. Even so, color accuracy can still be limited. The spectral archive is not subject to these constraints; the result is excellent color accuracy, eliminating the need for visual editing. A non-spectral archive stores three image planes per object, such as RGB

112 SCIENTIFIC EXAMINATION OF ART

TIFF (tagged image file format). For color-managed images tags are used to relate the digital signals to standardized viewing and illuminating conditions (source profiles). Using ICC color management, this includes CIE illuminant D50 and the CIE 1931 standard observer. Thus, the archive is limited to a single observer and illuminant. The spectral archive can be used to relate the digital signals to any observer, viewing, and illuminating condition. This provides tremendous opportunities by enabling an object to be rendered under multiple conditions without re-imaging. Using vision models that account for chromatic adaptation, one can compare an object’s appearance with changes in lighting, providing lighting de- signers with a unique and powerful tool. Many colorants have unique spectral properties within the visible spectrum. Thus, the spectral archive can be used to analyze the colorants used in a work of art. The spectral information can aid conservators in selecting colorants for inpainting (retouching) that result in minimal metamerism. We expect that a combination of spectral imaging and direct small-aperture spectrophotometry can be used to create colorant maps. Printed reproductions are quite useful for scholarly endeavors and during conservation treatments. Color-managed prints are designed to match under CIE illuminant D50 and to be viewed by the 1931 standard observer. By definition the prints are metameric and will only match for this single condition. However, prints are viewed under a variety of conditions. Spectral data can be used to produce prints that better match original objects for these many conditions. Finally, a visible-spectrum archive can be combined with other wavelength regions such as infrared and X ray, aiding in a more complete record on a work of art’s physical properties.

CONCLUSIONS A spectral image archive results in high color accuracy and facilitates the scientific examination of art in the visible region of the electromagnetic spectrum. Two methods of image acquisition have been described: (1) complete spectral sampling and (2) spectral sub-sampling combined with estimation. Each method has advantages and disadvantages. Issues include spectral accuracy, colorimetric accuracy, hardware complexity and cost, software complexity, image capture time, data storage, ease of use, maintenance, and system duplication complexity. One of the research goals is to describe these trade-offs in order to provide museums, archives, and libraries with information to assist them in making practical decisions regarding the incorporation of spectral imaging into their imaging practices.

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 113

ACKNOWLEDGEMENTS This research is supported by the Andrew W. Mellon Foundation; Rochester Institute of Technology; the National Gallery of Art, Washington, D.C.; and the Museum of Modern Art, New York, and would not have been possible without the participation of the students, faculty, and staff of the Munsell Color Science Laboratory.

RELEVANT MUNSELL COLOR SCIENCE LABORATORY (MCSL) PUBLICATIONS

Publications

1994 Vent. D. S. Multichannel analysis of object-color spectra. M.S. Thesis, Rochester Institute of Technology, Rochester, N.Y. 1996 Burns, P. D. and R. S. Berns. Analysis of multispectral image capture. In Proceed- ings of the IS&T/SID Fourth Color Imaging Conference Color Science, Systems, and Applications, pp. 19-22. Springfield, Va.: Society for Imaging Science and Technology. 1997 Burns, P. D. Analysis of image noise in multi-spectral color acquisition. Ph.D. Dissertation, Rochester Institute of Technology, Rochester, N.Y. Burns, P. D. and R. S. Berns. Error propagation in color signal transformations. Color Research and Application 22:280-289. Burns, P. D., and R. S. Berns. Modeling colorimetric error in electronic image acquisition, Proceedings of the Optical Society of America Annual Meeting, pp. 147-149. Washington, DC: Optical Society of America. 1998 Berns, R. S., F. H. Imai, P. D. Burns, and D. Tzeng. Multispectral-based color reproduction research at the Munsell Color Science Laboratory. In Proceed- ings of the International Society for Optical Engineering, vol. 3409, ed. J. Bares, pp. 14-25. Bellingham, Wash.: International Society for Optical Engineering. Imai, F. H. and R. S. Berns. High-resolution multi-spectral image capture for fine arts preservation. In Proceedings of the Fourth Argentina Color Conference, pp. 21-22. Buenos Aires, Argentina: Grupo Argentino del Color.

114 SCIENTIFIC EXAMINATION OF ART

Imai, F. H. and R. S. Berns. High-resolution multi-spectral image archives: a hybrid approach. In Proceedings of the IS&T/SID Sixth Color Imaging Confer- ence Color Science, Systems, and Applications, pp. 224-227. Springfield, Va.: Society for Imaging Science and Technology. 1999 Berns, R. S. Challenges for colour science in multimedia imaging systems. In Colour Imaging: Vision and Technology, eds. L. MacDonald and R. Luo, pp. 99-127. Chichester: Wiley. Burns, P. D. and R. S. Berns. Quantization in multispectral color image acquisition. In Proceedings of the IS&T/SID Seventh Color Imaging Conference: Color Science, Systems, and Applications, pp. 32-35. Springfield, Va.: Society for Imaging Science and Technology. Imai, F. H. and R. S. Berns. A comparative analysis of spectral reflectance reconstruction in various spaces using a trichromatic camera system. In Proceed- ings of the IS&T/SID Seventh Color Imaging Conference: Color Science, Sys- tems, and Applications, pp. 21-25. Springfield, Va.: Society for Imaging Science and Technology. Imai, F. H. and R. S. Berns. Spectral estimation using trichromatic digital cameras. In Proceedings of the International Symposium on Multispectral Imaging and Color Reproduction for Digital Archives, pp. 42-49. Chiba, Japan: Chiba University, Miyake Laboratory . Rosen, M. R. and X. Jiang. Lippmann 2000: A spectral image database under construction. In Proceedings of the International Symposium on Multispectral Imaging and Color Reproduction for Digital Archives, pp. 117-122. Chiba, Japan: Chiba University, Miyake Laboratory. 2000 Berns, R. S. Billmeyer and Saltzman’s Principles of Color Technology, 3rd ed. New York:Wiley. Imai, F. H., R. S. Berns, and D. Tzeng. A comparative analysis of spectral reflectance estimation in various spaces using a trichromatic camera system. Jour- nal of Imaging Science and Technology 44:280-287. Imai, F. H., M. R. Rosen, and R. S. Berns. Comparison of spectrally narrow-band capture versus wide-band with a priori sample analysis for spectral reflectance estimation. In Proceedings of the Eighth Color Imaging Conference: Color Science and Engineering, Systems, Technologies and Applications, pp. 234-241. Springfield, Va.: Society for Imaging Science and Technology. Imai, F. H., M. R. Rosen, R. S. Berns, N. Ohta, and N. Matsushiro. Preliminary study on spectral image compression. In Proceedings of Color Forum Japan 2000, pp. 67-70. Tokyo: Japanese Optics Society, Japanese Illumination Soci- ety, Japanese Color Society, and Japanese Photographic Society.

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 115

Quan, S. and N. Ohta. Optimization of camera spectral sensitivities. In Proceed- ings of the Eighth Color Imaging Conference: Color Science and Engineering, Systems, Technologies and Applications, pp. 273-278. Springfield, Va.: Society for Imaging Science and Technology. Rosen, M. R., M. D. Fairchild, G. M. Johnson, and D. R. Wyble. Color management within a spectral image visualization tool. In Proceedings of the Eighth Color Imaging Conference: Color Science and Engineering, Systems, Technolo- gies and Applications, pp.75-80. Springfield, Va.: Society for Imaging Science and Technology. 2001 Berns, R. S. The science of digitizing paintings for color-accurate image archives: A review. Journal of Imaging Science and Technology 45:305-325. Imai, F. H., M. R. Rosen, and R. S. Berns. Multi-spectral imaging of a van Gogh’s self-portrait at the National Gallery of Art, Washington, D.C. In Proceedings of the IS&T PICS Conference, pp. 185-189. Springfield, Va.: Society for Imag- ing Science and Technology. Imai, F. H., S. Quan, M. R. Rosen, and R. S. Berns. Digital camera filter design for colorimetric and spectral accuracy. In Proceedings of the Third International Conference on Multispectral Color Science, eds. M. Hauta-Kasari, J. Hiltunen, and J. Vanhanen, pp. 13-16. Joensuu, Finland: University of Joensuu Depart- ment of Computer Science. Imai, F. H., M. R. Rosen, D. R. Wyble, R. S. Berns, and D. Tzeng. Spectral reproduction from scene to hardcopy. I: Input and Output. In Proceedings of the International Society for Optical Engineering, vol. 4306, eds. M. M. Blouke, J. Canosa, and N. Sampat, pp. 346-357. Matsushiro, N., F. H. Imai, and N. Ohta. Principal component analysis of spectral images based on the independence of color matching function vectors. In Proceedings of the Third International Conference on Multispectral Color Sci- ence, eds. M. Hauta-Kasari, J. Hiltunen, and J. Vanhanen, pp. 77-80. Joensuu, Finland: University of Joensuu Department of Computer Science. Rosen, M. R., F. H. Imai, X. Jiang, and N. Ohta. Spectral reproduction from scene to hardcopy II: Image processing. In Proceedings of the International Society for Optical Engineering, vol. 4300, eds. R. Eschbach and G. G. Marcu, pp. 33- 41. 2002 Berns, R. S. and F. H. Imai. The use of multi-channel visible spectrum imaging for pigment identification. In Proceedings of the 13th Triennial ICOM-CC Meet- ing, pp. 217-222. London: James & James Ltd.. Berns, R. S. and R. Merrill. Color science and painting. American Artist, 68-70, 72 (January, 2002).

116 SCIENTIFIC EXAMINATION OF ART

Berns, R. S. Visible-spectrum imaging techniques: An Overview. In Proceedings of the 9th Congress of the International Colour Association, Rochester, N.Y. pp. 475-480. SPIE vol. 4421. Bellingham, Wash.: The International Society for Optical Engineering. Imai, F. H. and R. S. Berns. Spectral estimation of oil paints using multi-filter trichromatic imaging. In Proceedings of the 9th Congress of the International Colour Association, Rochester, N.Y. pp. 504-507. SPIE vol. 4421. Bellingham, Wash.: The International Society for Optical Engineering. Imai, F. H., M. R. Rosen, and R. S. Berns. Comparative study of metrics for spectral match quality. In Proceedings of the First European Conference on Color in Graphics, CGIV 2002, Imaging and Vision, pp. 492-496. Springfield, Va.: Society for Imaging Science and Technology. Quan, S. and N. Ohta. Evaluating hypothetical spectral sensitivities with quality factors. Journal of Imaging Science and Technology 46:8-14. Rosen, M. R., F. H. Imai, M. D. Fairchild, and N. Ohta. Data-efficient methods applied to spectral image capture. In Proceedings of the International Congress of Imaging Science, Tokyo, ICIS’02, pp. 389-390. Tokyo: The Society of Photo- graphic Science and Technology of Japan and The Imaging Society of Japan. Rosen, M. R., F. H. Imai, M. D. Fairchild, and N. Ohta. Data-efficient methods applied to spectral image capture. Journal of the Society of Photographic Sci- ence and Technology of Japan 65:353-362. Rosen, M. R., M. D. Fairchild, and N. Ohta. An introduction to data-efficient spectral imaging. In Proceedings of the First European Conference on Color in Graphics, CGIV’2002, Imaging and Vision, pp. 497-502. Springfield, Va.: So- ciety for Imaging Science and Technology. 2003 Berns, R. S., L. A. Taplin, F. H. Imai, E. A. Day, D. C. Day. Spectral imaging of Matisse’s Pot of Geraniums: A case study. In Proceedings of the IS&T/SID Eleventh Color Imaging Conference: Color Science and Engineering, pp. 149- 153. Springfield, Va.: Society for Imaging Science and Technology. Day, D. C. Filter selection for spectral estimation using a trichromatic camera. M.S. Thesis, Rochester Institute of Technology, Rochester, N.Y. Day, E. A. The effects of multi-channel spectrum imaging on perceived spatial image quality and color reproduction accuracy. M.S. Thesis, Rochester Insti- tute of Technology, Rochester, N.Y. Day, E. A., R. S. Berns, L. A. Taplin, and F. H. Imai. A psychophysical experiment evaluating the color accuracy of several multispectral image capture techniques. In Proceedings of the IS&T 2003 PICS conference, pp.199-204. Spring- field, Va.: Society for Imaging Science and Technology. Imai, F. H., D. R. Wyble, R. S. Berns, and D. Tzeng. A feasibility study of spectral color reproduction. Journal of Imaging Science and Technology 47: 543-553.

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 117

Quan, S., N. Ohta, R. S. Berns, and N. Katoh. Heirarchical approach to the optimal design of camera spectral sensitivities for colorimetric and spectral performance, pp. 159-170. SPIE 5008. Bellingham, Wash.: The International Society for Optical Engineering. Quan, S. Evaluation and optimal design of spectral sensitivities for digital color imaging. Ph.D. Dissertation, Rochester Institute of Technology, Rochester, N.Y. Rosen, M. R. Navigating the roadblocks to spectral color reproduction: Data- efficient multi-channel imaging and spectral color management. Ph.D. Dis- sertation, Rochester Institute of Technology, Rochester, N.Y. Sun, Q. Spectral imaging of human portraits and image quality. Ph.D. Disserta- tion, Rochester Institute of Technology, Rochester, N.Y. 2004 Day, E. A., R. S. Berns, L. A. Taplin, and F. H. Imai. A psychophysical experiment evaluating the color and spatial-image quality of several multi-spectral image capture techniques. Journal of Imaging Science and Technology 48:99-110. Mohammadi, M., M. Nezamabadi, R. S. Berns, and L. A. Taplin. Spectral imaging target development based on hierarchical cluster analysis. In Proceedings of the IS&T/SID Twelfth Color Imaging Conference: Color Science and Engineer- ing: Systems, Technologies, Applications, pp. 59-64. Springfield, Va.: Society for Imaging Science and Technology. 2005 Berns, R. S., L. A. Taplin, M. Nezamabadi, Y. Zhao, and Y. Okumura. High- accuracy digital imaging of cultural heritage without visual editing. In Pro- ceedings IS&T Second Image Archiving Conference, in press. Springfield, Va.: Society for Imaging Science and Technology. Berns, R. S., L. A. Taplin, M. Nezamabadi, and M. Mohammadi. Spectral imaging using a commercial color-filter array digital camera. In Proceedings 14th Triennial Meeting The Hague, ICOM Committee for Conservation, in press. Mohammadi, M., M. Nezamabadi, R. S. Berns, and L. A. Taplin, Pigment selection for multispectral imaging. In Proceedings 10th Congress of the Interna- tional Colour Association, in press. Murphy, E. P. A testing procedure to characterize color and spatial quality of digital cameras used to image cultural heritage. M.S. Thesis, Rochester Insti- tute of Technology, Rochester, N.Y. Rosen, M.R., and F.S. Frey. RIT American museums survey on digital imaging for direct capture of artwork. In Proceedings IS&T Second Image Archiving Conference, in press. Springfield, Va.: Society for Imaging Science and Technology.

118 SCIENTIFIC EXAMINATION OF ART

Smoyer, E. P. M., L. A. Taplin, and R. S. Berns. Experimental evaluation of museum case study digital camera systems. In Proceedings IS&T Second Image Archiving Conference, in press. Springfield, Va.: Society for Imaging Science and Technology. Zhao, Y., L. A. Taplin, M. Nezamabadi, and R. S. Berns. Using matrix R method in the multispectral image archives. In Proceedings 10th Congress of the Inter- national Colour Association, in press.

Technical Reports These reports can be downloaded from www.art-si.org and www.cis.rit.edu/ mcsl/research/reports.shtml. 1998 Imai, F. H. Multi-spectral image acquisition and spectral reconstruction using a trichromatic digital camera system associated with absorption filters, parts I- VIII. MCSL Technical Report, August. 2000 Berns, R. S. Direct digital imaging of Vincent van Gogh’s self-portrait—A personal view. MCSL Technical Report, May. Berns, R. S. The science of digitizing two-dimensional works of art for color- accurate image archives—Concepts through practice. MCSL Technical Re- port, May. Imai, F. H. Spectral reproduction from scene to hardcopy: Multi-spectral acquisition and spectral estimation using a trichromatic digital camera system associated with absorption filters. Parts I and II. MCSL Technical Report, October. 2002 Berns, R. S. Phase I final report to the National Gallery of Art, Washington, Art- SI project update. MCSL Technical Report, October. Day, E. A. Colorimetric characterization of a computer-controlled (SGI) CRT display. MCSL Technical Report, April. Day, E. A., F. H. Imai, L. A. Taplin, and S. Quan. Characterization of a Roper Scientific Quantix monochrome camera. MCSL Technical Report, March. Imai, F. H. Simulation of spectral estimation of an oil-paint target under different illuminants. MCSL Technical Report, January. Imai, F. H., L. A. Taplin, and E. A. Day. Comparison of the accuracy of various transformations from multi-band images to reflectance spectra. MCSL Tech- nical Report, Summer. Imai, F. H., L. A. Taplin, D. C. Day, E. A. Day, and R. S. Berns. Imaging at the National Gallery of Art, Washington, D.C. MCSL Technical Report, Decem- ber.

COLOR-ACCURATE IMAGE ARCHIVES USING SPECTRAL IMAGING 119

2003 Imai, F. H., L.A., Taplin, and E. A. Day. Comparative study of spectral reflectance estimation based on broad-band imaging systems. MCSL Technical Report, April. Day, D.C. Spectral sensitivies of the Sinarback 54 camera. MCSL Technical Report, February. Day, D.C. Evaluation of optical flare and its effects on spectral estimation accuracy. MCSL Technical Report, February. 2004 Berns, R. S., L.A. Taplin, M. Nezamabadi, and Y. Zhao. Modifications of a Sinarback 54 digital camera for spectral and high-accuracy colorimetric imaging: Simulations and experiments. MCSL Technical Report, June. Mohammadi, M. and R. S. Berns. Verification of the Kubelka-Munk turbid media theory for artist acrylic paint. MCSL Technical Report, June. Mohammadi, M., M. Nezamabadi, L. A. Taplin, and R. S. Berns. Pigment selection using Kubelka-Munk turbid media theory and non-negative least squares technique. MCSL Technical Report, June. Zhao, Y., L. A. Taplin, M. Nezamabadi, and R. S. Berns, Methods of Spectral Reflectance Reconstruction for A Sinarback 54 Digital Camera. MCSL Tech- nical Report, December.

Multispectral Imaging of Paintings in the Infrared to Detect and Map Blue Pigments

John K. Delaney,1 Elizabeth Walmsley,1 Barbara H. Berrie,1 and Colin F. Fletcher2

SUMMARY Spectral imaging for conservators offers the promise of providing a non-destructive tool for the identification of artists’ materials in situ, as well as determining their spatial distribution in an artwork. In this paper spectral imaging in the reflective infrared (IR) spectral region (0.7 to 2.5 microns) is examined for its potential to discriminate and identify blue pigments in paintings. The blue pigments considered are azurite, indigo, Prussian blue, lapis lazuli, cobalt blue, ultramarine, and thalo blue. Toward this end, visible to shortwave infrared diffuse reflection spectra of the blue pigments in both powder and paint forms were collected to determine the optimal spectral region to discriminate among these pigments. The measured spectra show that these blue pigments have large and varied reflectance in the near infrared (NIR, 0.7 to 1.0 microns) to shortwave infrared (SWIR, 1 to 2.5 microns) as compared to the visible spectral region. The large reflectance variation suggests the ability of broadband multispectral imaging (MSI) (< ~15 spectral bands, bandwidths of a few 100 to a few 10s of nm) to separate and support identification of these blue pigments in situ and to map their distribution. To test this, visible and infrared cameras were equipped with spectral band filters to allow collection of multispectral images of test panels and

1Conservation Division, National Gallery of Art, 2000-B S Club Drive, Landover, MD 20785 2 The Genomics Institute of the Novartis Research Foundation, 10675 John Jay Hopkins Drive, San Diego, CA 92121

120

MULTISPECTRAL IMAGING OF PAINTINGS 121

two paintings known to contain a subset of the blue pigments listed above were imaged. The reflectance spectra of the test panels obtained using cameras were found to correlate to those obtained using the benchtop spectrometers. Multispectral imaging of two paintings by Vincent van Gogh provided reflectance spectra consistent with the presence of the blue pigments Prussian blue, cobalt blue and ultramarine and gave information on the distribution of the pigments in the works by utilizing spectral band ratio images and false color composites. The identification of the pigments was confirmed using air-path X-ray fluorescence spectroscopy and energy dispersive spectrometry. The results show that multispectral imaging, either in numerous spectral bands from the visible to SWIR, or in a judicious selection of bands, can be a powerful tool to aid in pigment identification and distribution in paintings especially when combined with sample- based pigment identification methods such as X-ray fluorescence spectroscopy or analysis of cross-sections.

INTRODUCTION Spectral imaging (or Imaging Spectrometry), the collection of images in separate spectral bands in order to obtain reflection spectra, has been shown to be a powerful tool for geophysical remote sensing (1). The method also provides the spatial distribution (maps) of materials, such as minerals and agricultural crops, across the imaged region. The scientific examination of paintings often begins with the identification of the base set of pigments used by an artist. When identifying and determining the spatial distribution of pigments in a work of art, conservators begin with a visual inspection of the artwork. Subsequently, analytical methods, both destructive and non-destructive, are used in the classification and identification of pigments. Most of these techniques are applied to small areas of the painting either because they are destructive (that is, involve removing a sample from the work for analysis using polarized light microscopy or other techniques) or, if non-destructive, are too cumbersome and expensive to apply across the painting (e.g., air-path X-ray fluorescence spectroscopy, Raman spectroscopy). Although powerful, these techniques provide only localized information, and sampling sites are limited and may be biased. A less precise analytical methodology is used to extrapolate the results of “point analysis” to the rest of the artwork. The identification of pigments in un-sampled regions is often made through visual association. Success often depends on the ability to identify various pigments by their color. This hinders the ability to form general conclusions about the distribution of materials throughout the artwork. This can lead not only to misidentifications but also the possibility that some materials may not be discovered. Consequently, there has been growing interest in the development of methods that do not require sampling and provide information on spatial distribution.

122 SCIENTIFIC EXAMINATION OF ART

One such method is visible reflectance spectroscopy using high-resolution spectra acquired from a limited number of sites in an artwork (2, 3, 4). This methodology is useful for helping to accurately define the color and has been used for pigment identification utilizing spectral reference databases (2, 5). It is especially useful for pigment identification when it is extended to include the infrared region (6, 7, 8). However, the technique suffers from the same limitation of the other non-destructive methods, i.e., small sample sizes and the lack of an analytical methodology to sample the entire painting. Simply extending the number of sample sites over the entire painting would allow, in principle, the determination of the possible pigment composition at each site in the painting. Extending high- resolution visible spectral analysis to large areas of a painting is problematic. The ability to acquire high fidelity visible reflectance spectra (<5 nm band pass) of entire paintings is beyond the resources of conservation laboratories. Using visible spectroscopy alone to identify pigments having similar colors is challenging because of the small inherent reflectance differences among them, which are further reduced and confounded by the presence of scattered light associated with varying pigment particle size, and the collection/illumination geometries. In this paper, we investigate the use of infrared reflectance spectroscopy to aid in separating similar-appearing pigments and provide sufficient spectral data to be an adjunct to pigment identification. One motivation comes from the success of remote sensing reflectance spectral imaging, with its well-developed methodology for the identification of mineral deposits, and because numerous artists’ paints consist of ground minerals. Some success in applying this technique to works of art has already been achieved (6,); Casini et al. have differentiated between two yellow pigments in a painting by Pontormo (7). A well-designed reflectance spectral imaging system requires a compromise between the desired spectral and spatial resolutions. While the goal is to acquire high-quality reflectance spectra with the smallest sample size over the target area of interest (here the entire painting), the signal-to-noise requirements, instrument and data handling complexity force compromise. Increasing spectral resolution requires a reduction in spatial resolution and vice versa. Knowledge of the expected range of artists’ materials to be encountered, their spectral properties, and their particle size can be used to select the spectral and spatial resolutions required for identification and mapping of the pigments. The pigments typically encountered in paintings are either mineral or organic and as such their visible reflectance spectra are determined by broad electronic transitions that account for their color. There are overtones and combinations of vibrational transitions in the near infrared (NIR, 0.7 to 1.0 microns) to shortwave infrared (SWIR, 1 to 2.5 microns). In solids, the vibrational absorption bands have widths ca. 20 nm or wider which determines a limiting spectral resolution required for acquisition of spectra. While measurement of the position of vibrational overtone bands may be useful for unique identification of materials, designing instruments at this level of resolution (i.e., 20 nm) may be unnecessary.

MULTISPECTRAL IMAGING OF PAINTINGS 123

Many minerals have slowly changing, large amplitude reflectance variations, larger than 0.1, in the NIR to SWIR (9), which can be sampled with > 0.1 micron bandwidths. Thus collection of broadband multispectral images (> 0.1 micron bandwidth) appears to be sufficient for material separation and to support pigment identification. The focus of this work was to optimize spatial sampling and utilize broader spectral resolution compared to benchtop instrumentation, that is to use multispectral reflectance imaging. This was applied over a larger spectral region than the visible with the expectation of providing a more robust discrimination tool for pigment identification. This methodology is suitable for conservators who are used to collecting broadband images of paintings in the ultraviolet, visible and infrared (10).

EXPERIMENTAL In this study the reflectance spectra and multispectral images of six blue pigments are measured and analyzed. The pigments include mineral ores (azurite, lapis lazuli); colors manufactured by industrial processes (Prussian blue, thalo blue, cobalt blue, and synthetic ultramarine), and organic materials (indigo). The colorant in lapis lazuli and synthetic ultramarine is chemically the same and is treated as one pigment here. A three-part experimental approach was employed. First, the collection of high-resolution diffuse reflectance spectra of six blue pigments in both pressed powder and paint (linseed oil binder) at expected thickness on a chalk ground. Second, the collection of broadband spectral images of the paint test panels to investigate the correlation between spectral images and high-resolution spectra. Third, the collection of spectral images of paintings to determine the ability of multispectral imaging to separate and identify blue pigments in situ.

Test Panels Test panels of paint were constructed as previously described (10). The panels consist of paintouts of pigments hand-ground in linseed oil for azurite, lapis lazuli, Prussian blue, and indigo. Commercial oil paint was used for cobalt blue (Charbonnel) and thalo blue (Fezandie & Sperrle). The azurite, lapis lazuli, Prus- sian blue, indigo, and thalo blue paints were applied to a white chalk ground and the cobalt blue paint on a gesso ground.

Diffuse Reflection Spectra of Pigments and Paints Diffuse reflectance spectra of seven dark blue pigments, in pressed powder form and from the test panels, were collected using either a Nicolet System 510

124 SCIENTIFIC EXAMINATION OF ART

interferometer spectrometer or a Beckman Instruments UV 5240 dual beam grating interferometer, both equipped with an integrating sphere.

Multispectral Imaging Instrumentation and Procedures The multispectral images were obtained in the visible, near infrared, and shortwave infrared spectral regions using cameras fitted with band pass filters. A Sony XC-77 camera fitted with narrow band pass filters was used to acquire the visible/NIR (0.45-1.0 microns) image sets. The silicon monochrome video charge- coupled device (CCD) camera had a 25 mm f/1.6 lens. The filters (Corion Inc.) were 0.40 micron bandwidth filters in 0.50 micron increments over the range 0.45-1.0 microns. A Mitsubishi M600 PtSi or a Kodak PtSi 310-21X thermal imager was used to collect the SWIR (1.0 to 2.0 microns) image sets. Each camera was fitted with a 55 mm f/1.2 Nikon lens. Images of La Mousmé were collected using 1.2 (0.09 microns FWHH) and 1.6 (0.5 microns FWHH) micron broadband filters. Infrared images of details of the two paintings, the Self-Portrait and La Mousmé, were captured using three broadband filters: 1.1-1.4 microns (As- tronomy J, Barr Associates), 1.5-1.8 microns (Astronomy H, Barr Associates), and 2.0-2.4 microns (Astronomy K, Barr Associates). Diffuse illumination was used to light the scene, typically using a pair of Lowell Tota lights with FDN Q5000T3/4 quartz halogen lamps (500 W, 3200 K) and photographers’ umbrel- las. Black and white Spectralon (Labsphere) diffuse reflection targets (1 inch diameter) were used for in-scene calibration of the camera to reflectance. The low standard was 2-3 percent reflective over the range 0.45-2.4 microns, and the high standard was 98-99 percent reflective over 0.45-2.4 microns. Non-uniformity of the detector response was corrected using a gray card. For each test panel or painting, a set of 8 to 15 images was captured. In the multispectral image sets of the paintings, the paint test panels and Spectralon reflection standards were included in each image. The Spectralon standards were used to convert the digital counts of the image sets to reflectance units. Each multi-spectral experiment was repeated several times. A Macintosh computer with a Perceptics Pixelbuffer card or a Scion AG5 PCI card was used to capture the images. A variety of image processing programs (Scanalytics IP Lab Spectrum, Adobe Photoshop, NIH Image, Research Systems Inc. ENVI 3.4) was used in the analysis, non-uniformity, and calibration of the data sets. Registration and warping of the visible-NIR images with the SWIR image set was performed using ENVI. Spectral image data cubes were constructed and analyzed in ENVI.

X-ray Fluorescence Spectroscopy Non-destructive X-ray fluorescence spectroscopy (XRF) was performed using a Kevex 0750A spectrometer. The air-path instrument was equipped with a

MULTISPECTRAL IMAGING OF PAINTINGS 125

secondary target made from pressed barium chloride. This allowed simultaneous measurement of X-rays from ca. 2 eV to 45 eV. Pigments composed of light elements, for example, ultramarine, cannot be detected using this instrumentation.

Optical Microscopy and Other Analysis of Cross Sections Pigments from the top layer of paint were obtained by gently scraping the surface of the painting using a surgeon’s scalpel. A Leica MPX microscope was used to examine the particles mounted in Cargille MeltMount (refractive index 1.66) on glass slides using polarized light microscopy. The same particles or others could be mounted on a carbon planchet (stub) for examination by scanning electron microscopy (SEM) and energy dispersive spectrometry (EDS). A double- sided sticky carbon tab was used to adhere the particle to the planchet. If the particles could not be examined owing to charging they were coated with carbon. A JEOL 6300 scanning electron microscope with an Oxford Tetra backscattered electron detector was used for SEM. For EDS measurements an Oxford Inca 300 system was used with an Oxford Super ATW Si(Li) detector. Cross-sections were obtained from cracks or areas of loss. The fragments were mounted in Bioplastic® then cut to expose the layer structure of the paint. The sections were polished on SiC grit papers (Micromesh) and examined using optical microscopy and SEM-EDS.

RESULTS AND DISCUSSION

Diffuse Spectra of the Blue Pigments and Paints The six blue pigments have similar reflectance spectra in the visible region with few features. However, the spectra demonstrate larger and more varied changes in reflectance in the reflective infrared (0.7-2.5 microns) (Figure 1). Re- flectance peaks for all the pigments are centered near 0.44 microns, and all display a strong absorption in the red. The mineral samples, azurite and lapis lazuli, have a transition to increased reflectance in the NIR (0.7 to 1 microns), azurite lagging behind lapis lazuli. In the SWIR (1 to 2.5 microns) both pigments have near- constant reflectance. In contrast, the pigments indigo and thalo blue show a rapid rise to high reflectance, peaking at ca. 1.5 microns and decreasing beyond that. Prussian blue’s transition to higher reflectance occurs at about 1.35 microns. Cobalt blue has the most variable reflectance in the IR, becoming highly reflective in the NIR and weakly reflective in the high-energy side of the SWIR, between 1.3 and 1.6 microns, and moderately reflective at wavelengths > 1.6 microns. For all of the pigments the variations in reflectance amplitude through the IR region are large, > 0.1, and occur over spectral regions wider than 0.1 microns. These large, but slowly varying, changes in reflectance are suitable to be followed

126 SCIENTIFIC EXAMINATION OF ART

FIGURE 1 Diffuse reflectance spectra of six blue pigments in powdered form (dark line) and in oil-bound paint (blue line). The powder samples were optically thick. Synthetic ultramarine in powdered form was used instead of powdered lapis lazuli (natural ultramarine). The paint layers were generally 15-25 microns, a thickness often encountered in paintings. The reflectance spectrum of the chalk ground is given in the Prussian blue plot (top solid black curve). Reflectance values derived from multispectral images (solid circles) of the blue paint test panels (Figure 2) were obtained using spectral band pass filters.

using broadband multispectral imaging. However, increased spectral resolution and sampling (~10 nm) would allow the collection of vibrational lines such as the -O—H stretch in azurite that could be useful to characterize certain pigments. In Old Master paintings pigments are bound in organic binders (e.g., drying oils, egg tempera) and applied in thin layers over preparatory ground layers containing calcium carbonate (chalk) or calcium sulfate (gesso). As a result the reflectance spectra of the pigment and corresponding paint made from it may differ for several reasons. To test the influence of the binders on the reflectance spectra

MULTISPECTRAL IMAGING OF PAINTINGS 127

in the IR, diffuse reflectance spectra of the blue paints were collected (Figure 1). The possible effects of the binders on the reflectance spectra are (1) decrease in the amount of first surface reflectance (the intensity of reflectance spectra are reduced), (2) alteration of the depth of “color” from absorption (some scaling), (3) filtering of the pigment reflection by absorption properties of the binder (added spectral lines or shape changes). The spectra of the test panels compared to the powdered pigments show these effects occur. However, the general shape of the reflectance spectra of the pigments in paint is preserved with an offset (white light scatter) and scaling. Additionally, owing to the increasing transparency of the pigments in the SWIR, vibration bands from the preparatory layer may be observed in some cases. Linseed oil as a binder acts to reduce the difference in refractive index at the first interface with air and thus reduces the intensity of the first surface reflectance. For particles larger than the wavelength of light (e.g., azurite) this results in a decrease in the ‘white light’ reflectance, hence the reflectance curves shift down- ward. Since the paints on the test panels do not represent infinitely thick layers, there is also a change in the total absorption and hence a scaling change between powder and paint arises. The binder can act as a spectral filter owing to its own absorption. Linseed oil has little absorption in the visible to infrared, but some of the commercial paint films (e.g., cobalt blue) appear to have additional bands in the IR. High reflectance correlates with low absorption, thus the underlying preparatory ground or underpaint layers may contribute to the reflectance spectra. The contribution of the ground to the reflectance spectrum becomes significant when the paint becomes transparent owing to decreased absorption and scattering in the SWIR. It is dependent on the thickness of the paint film. The addition of absorption features from the ground to the spectrum of Prussian blue at wavelengths beyond 1.6 microns may be noted. The general reflectance signatures are, however, maintained between bulk pressed powder and oil-bound pigments. In the samples painted on a chalk or gesso ground, which itself has a high reflectance in the visible to the SWIR, the paints appear to mimic the scattering of the optically thick powder samples.

Multispectral Imaging of Paint Samples To examine the extent to which diffuse reflectance spectra of pigments can be modeled by using imaging cameras available to conservators, multispectral images of the paint swatches were collected (Figure 2). Two cameras, one sensitive in the visible/NIR and one in the SWIR, each outfitted with bandpass filters, were used to sample the visible/NIR and SWIR spectral regions. For in-scene calibration, the reflectance of the ground and carbon black were used as standards. The reflectance values derived from the multispectral images of the test panels show good agreement with the reflectance spectra measured using the benchtop reflectance spectrometer (Figure 1), demonstrating that the image collection

128 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 Multispectral images of test panels with blue paint swatches. By column, from left to right: Lapis lazuli (natural ultramarine), azurite, Prussian blue, indigo, and thalo blue. Center wavelength of spectral band pass filter, by row, from top to bottom: 0.56, 0.90, 1.2, 1.4, and 1.6 microns. The images show varying changes in the reflectance of the different pigments in the infrared and their increased transparency.

method is adequate. This shows that these blue pigments can be distinguished from each other using only the multispectral images. This ability exists despite the fact that the measurements on the powdered pigment samples were performed using an integrating sphere, whereas the imaging of the panels and the works of art were performed with a narrow collection solid angle. Given the near-Lambertian nature of the samples’ reflectance, and the near-diffuse illumination, the close to superimposition of the spectral data should not be surprising. These results demonstrate that by utilizing multispectral imaging techniques the reflectance of the blue pigments can be characterized adequately to distinguish among them and ideally to identify them. In geophysical remote sensing applications, materials of interest are encountered in optically thick, granular form, like the powdered pigment samples. How- ever, as noted earlier, in works of art the pigments are in thin paint layers. In regions of high reflectance, an increase in the transparency of the paint layer can be observed (compare Figures 1 and 2). The preservation of the spectral reflectance properties between paint and powder pigments and the maintenance of high reflectance of the bound pigments appear to be due to the high reflectance of the ground. A light-absorbing ground or pigment underneath would however alter the reflectance spectra. Thus the effect of layering of thin paint films that occurs in artwork needs to be considered. Since almost all pigments become transparent by 1.5 microns (10), collection of spectral images from 1.5 to 2.5

MULTISPECTRAL IMAGING OF PAINTINGS 129

microns seems to be of little utility for their identification. It is the property of increasing transparency in this region that allows the detection of underdrawings in paintings using infrared reflectography (10).

Multispectral Analysis of Paintings to Characterize Blue Pigments To demonstrate the ability of multi-spectral imaging to discriminate and identify blue pigments in situ, two paintings by Vincent van Gogh in the collection of the National Gallery of Art, Washington, D.C., were examined using the technique. The paintings were selected because they have large regions of bright to dark blue paints. Moreover, results from site-specific analysis techniques such as XRF, polarized light microscopy and SEM-EDS were available, and the paintings contain the pigments Prussian blue, cobalt blue, and ultramarine. The first painting imaged was La Mousmé, by Vincent van Gogh (1888, Chester Dale Collection, 1963.10.151) (Figure 3, visible light image; see also www.nga.gov/search/index.shtm). Van Gogh describes this painting in his own words (11): The portrait of the girl is against a background of white strongly tinged with emerald green, her bodice is striped blood red and violet, the skirt is royal blue, with large yellow-orange dots. The mat flesh tones are yellowish-grey; the hair tinged with violet; the eyebrows and eyelashes are black; the eyes, orange with Prussian blue. A branch of oleander in her fingers for the two hands are showing. Prussian blue has a large change in reflectance at ~1.35 microns where the pigment becomes less absorbing (Figure 1). Other blue pigments do not have as large an absorbance change over this interval. To identify areas of the painting where Prussian blue occurs, a ratio image was created by dividing an image acquired at 1.6 microns by one acquired at 1.2 microns. In the ratio image, passages where Prussian blue is present have a ratio of absorbance >1 in comparison to passages where ultramarine, cobalt blue, or indigo are present. The ratio image of La Mousmé (Figure 3) shows bright (high reflectance ratio) features having higher reflectance at 1.6 than at 1.2 microns, which can be assigned to Prussian blue. The spectral ratio image in the IR indicates that, in the visible image, the dark lines of the chair, the dark lines outlining the flowers, the outline of the girl’s hair and to a lesser extent some daubs of paint on her skirt and some of the stripes on her blouse are painted using Prussian blue. The lack of large reflectance changes in some of the darker and brighter blue regions of the visible image suggests the use of a different blue pigment there. To more confidently assign the regions identified in the band ratio image of La Mousmé, multispectral images in 11 spectral bands from the visible to the SWIR were collected from an 8 × 10-inch section of the painting. Each spectral band image was collected using a visible/NIR or SWIR camera fitted with bandpass

130 SCIENTIFIC EXAMINATION OF ART

FIGURE 3 Vincent van Gogh’s La Mousmé (1888). (Left) Visible light image. (Right) Shortwave infrared spectral band ratio image creaed from two infrared composite images captured at 1.2 microns and at 1.6 microns. The bright areas are regions where the reflectance of the painting is higher at 1.6 microns than at 1.2 microns, and thus indicate both the probable presence of Prussian blue and where it occurs in high concentration within the painting. Each IRR composite comprises a mosaic of 24 flat-field-corrected images.

filters in an arrangement similar to that typically used for collecting infrared reflectograms. The image sets were converted to reflectance units using the in- scene reflectance standards. A multispectral image cube (a 3-D image where z is the spectral band and x,y are spatial locations) was then constructed and the images registered using tie points. The visible color image in Figure 4, generated from an MSI cube using the 0.45, 0.55, and 0.65 micron images, shows a detail of La Mousmé captured with the reference standards and blue test panels. Verification of the reflectance calibration of the MSI cube was determined by comparing the image-derived spectra of “in-scene” blue test panels with the prior high-resolution spectra obtained with the benchtop spectrometer (Figure 4A). Qualitatively these two match (compare Figure 1 with Figure 4A). The image cube spectra derived from the blue test panels were a reasonable fit to the spectra acquired using benchtop instrumentation after scaling and a small amount of translation. Specifically only a small translation for Prussian blue (0.04 offset) and scaling for lapis lazuli (0.87×, 0.03 offset) and cobalt blue (0.75×) (see Figure 4A) were required to get a good match. Since the illumination of the test panels alone, versus in the scene of the painting, are different some transitional and scaling changes are not unexpected. Reflectance spectra derived from the multispectral image cube of the painting itself support the assignment of Prussian blue to bright areas of the ratio

MULTISPECTRAL IMAGING OF PAINTINGS 131

FIGURE 4 Multispectral image analysis of a detail of La Mousmé showing the presence of two different blue pigments. The multispectral image cube consists of 11 spectral bands from the visible through the SWIR. (Left) Visible color composite of spectral images obtained at 0.65, 0.55, 0.45 microns. Along the bottom, from left to right, are black and white reflectance standards (2 and 98 percent Spectralon standards) and blue pigment test panels: cobalt (two swatches), Prussian blue, and lapis lazuli. (Right) A. Plots of reflectance, derived from the multispectral images of the reference panels: high reflectance 98 percent Spectralon (open circles), cobalt blue swatch (solid blue diamonds), Prussian blue (solid black circles), and lapis lazuli (blue squares). The blue and black lines are scaled diffuse spectra. B. Plots of reflectance derived from the multispectral images of the sites 1 to 4 in the detail image. Sites 1 and 2 (black circles and triangles) are in the chair. Sites 3 and 4 are in the blue stripes on the blouse.

image (Figure 3) of La Mousmé. The image-derived spectra of the bright regions in the ratio image that map to the dark blue areas of the visible image most closely resemble the reference spectrum of Prussian blue paint test panel and powdered pigment (Figure 4B). Spectra from two sites on the chair rail (Figure 4B, sites 1, 2) are reasonably well described by the reference spectrum of Prussian blue scaled 0.53. Based on this spectral sensitivity, these areas can be determined to contain Prussian blue. Not all the blue stripes on the blouse are bright in the 1.6/1.2 micron ratio image, suggesting the presence of another pigment (Figure 3 and Figure 4, sites 3, 4). The image-based reflectance spectra show high reflectance in the NIR to SWIR (0.8 to 1.6 microns, Figure 4B, sites 3, 4) that is consistent with the reflectance behavior of indigo or ultramarine and not Prussian or cobalt blue. A scaled diffuse reference spectrum of ultramarine powdered pigment (0.39×,

132 SCIENTIFIC EXAMINATION OF ART

0.17 offset) matches the measured spectra reasonably well. The reflectance spectra thus show that at least two blue pigments were used in this painting. The painting has other passages of the dark blue, for example, the outlines of the buttonholes on the girl’s blouse and the top of the flowers she is holding. Visual inspection of the painting alone might suggest that these could be Prussian blue as well. However, the ratio image does not show the strong reflectance changes associated with Prussian blue. A detail from images obtained at 1.2 and 1.6 microns shows the dark blue outlining the flowers is dark at 1.2 microns and light at 1.6 microns (Figure 5), whereas the dark blue outlining the buttonholes is not apparent in either the image obtained at 1.2 or at 1.6 microns. The reflectance spectra derived from the multispectral image cube show the dark lines of the flowers are similar to Prussian blue (0.53 scaling of reference spectra), but the buttonholes are not (Figure 5). The reflectance spectra of the dark blue outlining

FIGURE 5 Multispectral image analysis of a detail of La Mousmé showing the use of two different dark blue pigments for outlining. The multispectral image cube consists of 11 spectral bands from the visible through the SWIR. (Top left) Visible color composite of spectral images obtained at 0.65, 0.55, 0.45 microns. (Top right) Infrared reflectogram of the detail area obtained at 1.2 microns and (Bottom right) 1.6 microns. Infrared image obtained at 1.6 microns shows the change in reflectance from 1.2 microns for the dark outline of the girl’s flowers, but not the buttonholes. (Bottom left) Plots of reflectance derived from the images of the dark lines outlining the button (blue diamonds and squares) and flower (black circles and triangles). The solid blue and black lines are diffuse reflectance spectra of the paints using the ultramarine and Prussian blue test panels (after scaling).

MULTISPECTRAL IMAGING OF PAINTINGS 133

the buttonhole shows an earlier rise in reflectance than Prussian blue, more like some of the blue stripes on the blouse. A scaled and offset reference spectrum of ultramarine (0.42×, 0.02 offset) does pass though the majority of data in the red and infrared suggesting this blue pigment is present. Confirmation of ultramarine at these sites was provided by SEM/EDS analysis, which also showed the presence of zinc oxide (ZnO). Both the spectra and SEM/EDS are consistent with the pigment in this area being ultramarine. Thus the blue pigments in the La Mousmé painting have been discriminated into Prussian blue and ultramarine blue. The ratio image of the area in combination with the MSI cube suggest that Prussian blue was used in the chair, the outline of the flowers in her hand, and in strokes of the girl’s hair, eyebrows, lips and some stripes of her jacket. The reflectance spectra derived from the image cube suggest that the dark outline of the buttonholes and the blue outlines around the spots on the girl’s skirt were painted using ultramarine. Cobalt blue, a pigment that van Gogh used often, does not appear to have been used for painting La Mousmé. X-ray fluorescence spectroscopy indicated the presence of cobalt in the background of Vincent van Gogh’s Self-Portrait (1889, Collection of Mr. and Mrs. John Hay Whitney, 1998.74.5) (Figure 6, visible light image; see also www. nga.gov/search/index.shtm). A spectral image cube of a region of the painting shows that it can be used to find the locus of cobalt blue. The multispectral image cube of Self-Portrait gives reflectance spectra from the background that are similar to that of the blue cobalt blue test panel (Figure 6). A false color image, made by assigning the infrared image to the red channel shows the distribution of cobalt blue in the painting (Figure 6). Cobalt blue is extensively used in the background and also in the jacket, although spectra from the jacket are complex, indicating a mixture of pigments here.

CONCLUSIONS This report contributes to the body of work that demonstrates the utility of spectral imaging not only for discrimination among pigments but also to determine their spatial distribution within a work of art. While reflectance spectroscopy is a relatively less specific analytical tool than other analytical chemical tools, it is non-destructive and can be readily applied to the entire artwork. Its power lies in helping to define the set of pigments used in the work, and identifying regions of high concentration and thus directing site-specific, more powerful analytical tools such cross-section analysis, XRF, and SEM/EDS for more thorough chemical analysis. Moreoever, in the case of Prussian blue, the reflectance spectra in the infrared may be a more definitive assignment method than XRF given the high tinting strength of Prussian blue. The results here demonstrate that extending the multispectral imaging method to include the infrared

134 SCIENTIFIC EXAMINATION OF ART

FIGURE 6 Multispectral image analysis of a detail of van Gogh’s Self-Portrait (1889). The distribution of cobalt blue is indicated in the false color image by the color red. The multispectral image cube consists of 13 spectral bands from the visible through the SWIR. (Left) Visible light image. (Bottom right) Plots of reflectance derived from the image of the background (solid diamonds). The solid blue and black lines are scaled diffuse reflectance spectra of the cobalt blue paint (blue) and powder (black). (Top right) False color infrared composite of spectral images obtained at 0.90, 0.55, 0.45 microns. This image renders the cobalt blue to appear “red” thus showing the cobalt is present in a large proportion in the background and a smaller proportion in the jacket.

can improve the success of spectral imaging in pigment identification and discrimination given the large and varied reflectance changes in these regions, in particular when the visible spectra of pigments are similar. Large and slowly varying reflectance changes in the infrared allow the utilization of broadband MSI techniques, simplifying the methodology. This study of two paintings demonstrates that multispectral imaging in the visible and NIR regions can be employed as a useful tool in the scientific examination of paintings. We have demonstrated that the optical properties of pigments in the infrared display diagnostic features which can be employed to assign and map pigments, and that these features can be detected using conventional imaging techniques, including modeling of reflectance spectra and ratioing images

MULTISPECTRAL IMAGING OF PAINTINGS 135

obtained at different wavelengths. It is difficult to infer this information from conventional techniques, which rely on a limited number of micro-samples. The advantages of the technique are tempered by the increased complexity of the reflectance spectra owing to particle size variation and increasing transparency in the infrared. Conservators and conservation scientists utilizing imaging systems currently available can apply the power of visible and infrared multispectral imaging to their work.

ACKNOWLEDGEMENTS This research forms part of an ongoing project at the National Gallery of Art on applications of infrared imaging. Over the past ten years, we have benefited from discussions and assistance from many people, including Dr. Jack Salisbury, who gave us access in 1994-95 to his laboratory at Johns Hopkins University in order to collect the diffuse reflectance spectra of the samples, and Mr. Dana D’Aria, who assisted us in the collection of the spectra; Raymond Rehberg, David L Clark, and Rollo E Black, of Eastman Kodak, who provided generous assistance using the Kodak thermal imager; Elizabeth Freeman, Kristi Dahm, Lucy Bisognano, and Laura Rivers, who helped with the image captures; Dr. Lisha Glinsman, who provided the results of XRF data on the Self-Portrait; and Dr. René de la Rie and Mr. Ross Merrill, Chief of Conservation, for their continued interest. An early phase of this research was supported by the Circle of the National Gallery of Art.

REFERENCES

1. A. F. H. Goetz, G. Vane, J. E. Solomon, and B. N. Rock, “Imaging Spectrometry for Earth Femote Sensing,” Science, 228, 1147-1153. 2. Guillaume Dupuis, Mady Elias, and Lionel Simonet, “Pigment Identification by Fiber-Optics Diffuse Reflectance Spectroscopy,” Applied Spectroscopy, 56, (2002), 1329-1336. 3. Otto Hahn, Doris Oltrogge, and H. Bevers, “Coloured Prints of the 16th Century: Non- Destructive Analyses on Coloured Engravings from Albrecht Dürer and Contemporary Artists,” Archaeometry, 46, (2004), 273-282. 4. Mauro Bacci and Marcello Picollo, “Non-Destructive Spectroscopic Detection of Cobalt(II) in Paintings and Glass,” Studies in Conservation, 41(3), (1996), 129-135. 5. Roy S. Berns, Jay Kreuger, and Michael Swicklik, “Multiple Pigment Selection for Inpainting Using Visible Reflectance Spectrophotometry,” Studies in Conservation, 47, (2002), 46-61. 6. Elizabeth Walmsley, John K. Delaney, Barbara H. Berrie, Dana D’Aria, Colin Fletcher, and Jack Salisbury, “Pigment Identification in Artworks by MultiSpectral Imaging in the Near Infrared [abstract],” Final Program of the 34th Annual Eastern Analytical Symposium ’95, Somerset, NJ; section “Imaging for Conservation,” 66. 7. Andrea Casini, Franco Lotti, Marcello Picollo, Lorenzo Stefani, and Ezio Buzzegoli, “Image Spectroscopy Mapping Technique for Non-Invasive Analysis of Paintings, Studies in Conservation, 44 (1999), 39-48.

136 SCIENTIFIC EXAMINATION OF ART

8. Applied Spectroscopy Laboratory of the Institute of Applied Physics “Nello Carrara” of the Italian National Research Council and the Restoration Laboratory of the Opificio delle Pietre Dure, Fiber Optics Reflectance Spectra (FORS) of Pictorial Materials in the 350-1000 nm range database, http: //fors.ifac.cnr.it/index.php. 9. Dana D’Aria and Jack Salisbury, Johns Hopkins University Spectral Library, http://speclib. jpl.nasa.gov. 10. Elizabeth Walmsley, Catherine Metzger, John K. Delaney and Colin Fletcher, “Improved Vi- sualization of Underdrawings with Solid-State Detectors Operating in the Infrared,” Studies in Con- servation, 39, (1994), 217-231. 11. The Complete Letters of Vincent van Gogh, 3 vols. Boston (1958). Letter 518.

Modern Paints

Tom Learner Senior Conservation Scientist Tate London

ABSTRACT Few would argue that oil paint has been the most important type of paint over the last 500 years. The use of oil as the film-forming component of paint—the binding medium—was well established by the start of the fifteenth century, and for many artists oil paints still remain the preferred choice today. However, throughout the twentieth century a wide and varied range of synthetic polymers have been developed, many of which have been used as binding media in modern paints. The introduction of these synthetic binders, most notably acrylic, alkyd, and polyvinyl acetate, has undoubtedly enabled great advances to be made in paint technology, in terms of reduced yellowing, greater flexibility, faster drying times, and in the case of emulsion formulations, the elimination of organic solvents as thinners and diluents. Many artists have utilized these modern paint types, including those that were never intended specifically for artists’ use, and have explored and exploited their distinct handling and optical properties. Establishing the constituents of paint is frequently necessary prior to any kind of conservation treatment and for developing long-term preventive conservation strategies, as well as for technical art historical studies and issues surrounding authenticity. The identification of binding media is particularly important, as this component appears to have the largest influence on many of the properties of the resulting dried paint film. Although noninvasive/nondestructive techniques would clearly be favorable, at present the most useful analysis is obtained from high-sensitivity techniques that

137

138 SCIENTIFIC EXAMINATION OF ART

require the removal of submilligram paint samples. Two analytical techniques—pyrolysis-gas chromatography-mass spectrometry (PyGCMS) and Fourier transform infrared spectroscopy (FTIR)—are now routinely used at Tate to identify and characterize modern paints from works of art. This paper will summarize the three principal classes of synthetic binder and how PyGCMS and FTIR have been utilized to analyze them.

INTRODUCTION Despite the great variety of modern paint formulations (see Figure 1), there are three principal classes of synthetic binder that have been widely used by artists: acrylic, alkyd, and polyvinyl acetate (PVA) (Crook and Learner, 2000; Learner, 2000). The main binder used in the artists’ paint market has been acrylic, although there are two quite distinct forms: acrylic solution, where the acrylic

FIGURE 1 Selection of modern paints.

MODERN PAINTS 139

polymer is dissolved in a mineral spirit or turpentine, and acrylic dispersion (i.e., emulsion), where the acrylic polymer is dispersed in water (with the aid of a surfactant and other additives). The solution form consists of a poly (n-butyl methacrylate) homopolymer, which was developed in the late 1940s, whereas the emulsion form consists of an acrylic copolymer, typically between methyl methacrylate (MMA) and either ethyl acrylate (EA) or n-butyl acrylate (nBA), and only became available in the late 1950s. The two types have quite distinct mechanical properties and exhibit very different sensitivities to organic solvents and water. It is important therefore to be able to distinguish between them analytically. Acrylic binders are also used in the house paint market, but two other important types of synthetic binder—alkyd and PVA—are also widely utilized. Alkyd paints are oil-modified polyester paints, introduced in the late 1930s, although they did not make a significant impact on the paint industry until the late 1950s in Europe and slightly earlier in the United States. Since then, the vast majority of oil-based house paints have incorporated an alkyd resin as the principal binder. Perhaps somewhat surprisingly, they have received only limited use by artists’ colormen. Alkyd resins are produced from three main components: a polyhydric alcohol, a polybasic carboxylic acid, and a source of monobasic fatty acid, which is often added in the form of a drying oil. The polyhydric alcohol (also called “polyol”) and polybasic acid constituents in the vast majority of alkyd house paints are actually limited to just three compounds: glycerol and/or pentaerythritol as the polyol and phthalic anhydride—the dehydrated version of ortho-phthalic acid (1,2-benzenedicarboxylic acid)—as the polybasic acid. PVA has also been used in waterborne polymer emulsions, although it requires some slight modification to lower its glass transition temperature, either by the addition of a plasticizer (common in early formulations) or by copolymerization with a softer monomer (the preferred option since the 1960s). A number of different plasticizing methods have been used for this since the introduction of PVA paints, and PyGCMS is able to differentiate between them. In early emulsion formulations an external plasticizer, such as dibutyl phthalate (DBP), was added and often in appreciable quantities (up to 20 percent by weight) (Martens, 1981, p. 81). The problems caused by these plasticizers migrating out of the paint film were overcome during the 1960s by the copolymerization of PVA with softer monomers, often called internal plasticization. This has been achieved with a variety of other vinyl monomers, including some of the softer acrylates but also commonly achieved with vinyl versatates or VeoVa monomers, which are com-

mercial mixtures of highly branched C9 and C10 vinyl esters manufactured by Shell (Slinckx and Scholten, 1994). The choice of which binder is used in a household emulsion formulation appears to be dependent on such factors as cost (acrylic is more expensive), durability (acrylic is considered more durable and therefore often used for exterior paints), surface finish (acrylic has a superior binding power and is therefore

140 SCIENTIFIC EXAMINATION OF ART

sometimes used for matt paints where less binder is present), and age (PVA emulsions were developed in the 1940s, earlier that the acrylics). The ability to identify the binding medium in paints is often essential for conservation reasons. Since different paint types will respond differently to cleaning solvents and reagents, paint characterization is often needed prior to treatment. It is also necessary when examining the aging properties of paints. Reac- tions such as oxidation, cross-linking or chain-scission all affect the physical and chemical properties of a paint; so understanding the likely reactions is an important consideration. It is, after all, better to prevent deterioration than try to reverse it. Much effort is currently being put into the general understanding of artists’ materials and techniques, in other words, what did an artist use and how? Analysis can also play an important role in authentication issues. Many of the techniques used for traditional medium analysis, such as gas and liquid chromatography, are not totally suited to all these modern paint binders, largely because of their high molecular weights (i.e., they are nonvolatile and frequently insoluble in solvents) and the inability to extract diagnostic components from the polymer matrix. Nevertheless, these polymeric materials can be effectively broken down into volatile fragments through pyrolysis (i.e., heat in the absence of oxygen), and these fragments can consequently be separated and identified by gas chromatography (GC). This technique—pyrolysis-gas chromatography (Py-GC)—has been used since the 1960s by forensic scientists for the identification of synthetic binders in house paints, car paints, and various industrial coatings (Jain et al., 1965; Challinor, 1983; Wheals, 1985) but was not properly assessed by the conservation profession for its capability to identify the synthetic binders used in artists’ painting materials until the 1990s (Sonoda and Rioux, 1990; Stringari and Pratt, 1991; Sonoda, 1998). More recently a wider range of paint binders has been investigated with the many advantages of using a mass spectrometer as a detector (i.e., with PyGCMS) (Learner, 1995a, 2001). Another technique frequently adopted for the analysis of traditional binding media is Fourier transform infrared spectroscopy (FTIR). FTIR is normally used as a comparative technique with the spectra of each unknown material being matched either visually with a library of known standards or through a computer search. Although this technique requires no instrumental modifications to enable the analysis of synthetic polymers, an entire set of new reference standards has to be generated. The technique is also semiquantitative; so it is normally possible to assess the relative proportions of two components in a mixture if the spectra of each are available and adequately different. At Tate, FTIR has been widely used as a nondestructive analytical method (see FTIR section below for details) to be carried out prior to PyGCMS (Learner, 1996). More recently it has also been used in attenuated total reflectance (ATR) mode to examine the migration of surfactants to the surface of acrylic emulsion films as part of an ongoing study into the effects of surface cleaning (Learner et al.,

MODERN PAINTS 141

2002a,b). The use of PyGCMS and FTIR in the analysis of modern paints will now be outlined.

PYROLYSIS-GAS CHROMATOGRAPHY-MASS SPECTROMETRY

Acrylic Solutions Artists’ acrylic solution paints, bound with a poly n-butyl methacrylate (pnBMA) homopolymer resin, produce extremely simple pyrograms consisting of a single peak of nBMA monomer. The acrylic binder undergoes complete depolymerization (a mechanism common to all polymethacrylates [Irwin, 1979]) on pyrolysis.

CH CH3 CH CH 3 3 3 CH . 3

O O O O O O O O O O C H C H C H C H 4 9 4 9 4 9 4 9 C4H9

Figure 2 shows a pyrogram of Paraloid F-10 (Rohm and Haas)—the acrylic binder used in acrylic solution paints—with the mass spectrum of the single peak identified as nBMA. The two most intense ions in the mass spectrum are those of m/z = 69 (loss of the n-butoxy side group) and m/z = 41 (loss of a further C=O),

nBMA

FIGURE 2 Pyrogram of Paraloid F-10 (pnBMA acrylic resin) with mass spectrum of nBMA.

142 SCIENTIFIC EXAMINATION OF ART

and these are both seen with all methacrylate monomers. Strong peaks at m/z = 87 (from the protonation of methacrylic acid) and m/z = 56 (from butene) are also seen. The molecular ion of nBMA (m/z = 142) is extremely weak and often not observed.

Acrylic Emulsions Figure 3 shows the overall pyrogram from Plextol B-500 (Röhm), a p(EA/ MMA) emulsion that has been used in artists’ acrylic paint formulations, and from Spectrum polymer medium (Spectrum), a p(nBA/MMA) acrylic copolymer artists’ product. Also shown in the pyrogram of Plextol B-500 is the mass spectrum from the most intense peak, MMA. The mass spectrum of MMA shows a similar fragmentation pattern to nBMA seen with acrylic solution paints, with

Plextol MMAB-500 (Rohm)

EA trimers sesquimers and dimers

MMA

nBA

trimers sesquimers dimers

FIGURE 3 Pyrogram of Plextol B-500, a p(EA/MMA) emulsion with mass spectrum of MMA (top), and Spectrum polymer medium, a p(nBA/MMA) emulsion (bottom).

MODERN PAINTS 143

MMA

H H 2CH - OC2H5 CH O 2 EA +O + O

C2H5

m/z = 100 (MW) m/z = 55

MMA

nBA

FIGURE 4 Details of early sections of pyrograms of Plextol B-500 (top) and Spectrum polymer medium (bottom) with mass spectra of acrylate components (EA and nBA, respectively).

prominent ions at m/z = 69 and 41, but here the molecular ion (m/z = 100) is clearly visible. Figure 4 shows a detail from the early part of both of these pyrograms with the mass spectrum of the EA and nBA monomers shown, respectively. The ability to separate the EA and MMA monomers, despite their similar retention times, is clearly seen in this detail. The mass spectra of both acrylate monomers are domi- nated by a peak of m/z = 55, corresponding to the loss of the alkoxy side group to + produce a CH2CH.CO fragment ion (as shown for EA). The overall pyrograms (see Figure 3) of both copolymers contain a number of later additional peaks, which are the result of incomplete depolymerization when an acrylate component is present in the polymer. These have been identi-

144 SCIENTIFIC EXAMINATION OF ART

fied as a series of sesquimers, dimers, and trimers of acrylate or acrylate-MMA combinations. Sesquimer is a term meaning 1.5 monomer units (i.e., a molecule consisting of a three-carbon atom backbone with an acrylate/methacrylate group at either end). Some materials labeled acrylic emulsions can actually be copolymers with other monomers, such as styrene (to create styrene-acrylics) or vinyl acetate (to create vinyl-acrylics). Although not shown here, both are readily distinguished by pyrolysis-gas chromatography-mass spectrometry (PyGCMS), by the detection of styrene monomer and acetic acid (see below), respectively.

Polyvinyl Acetate (PVA) Emulsions On pyrolysis, PVA emulsion paints produce principally ethanoic (acetic) acid and benzene by a side group elimination mechanism. Figure 5 shows the overall pyrogram observed from Emultex VV536 (Harco), with the mass spectrum from the intense peak at the start of the pyrogram. This spectrum is mainly that of ethanoic acid (with a molecular ion of m/z = 60, and strong fragment ions at m/z = 43, 45), although the peak of m/z = 78 corresponds to the molecular ion of benzene. It is usually possible to confirm the presence of an emulsion form of PVA (as opposed to a solution form) by the detection of a plasticizer, since pure PVA is slightly too hard to form a continuous film from an emulsion. This particular emulsion, Emultex VV536 (Harco), actually contains both a vinyl versatate resin and a phthalate plasticizer (in the majority of PVA emulsions only one kind is used). The sharp peak at the far right of the pyrogram is identified as DBP, whose mass spectrum (also shown) has a very intense fragment ion of m/z = 149, which is the characteristic fragment ion of all dialkyl phthalates. In the center of the pyrogram is the band of rather broad peaks produced by the VeoVa plasticizer. Although these peaks are not fully resolved, the overall peak pattern does conform to a very distinctive profile.

Alkyds For alkyd paints based on ortho-phthalic acid, phthalic anhydride is the principal peak detected on pyrolysis and therefore used as the diagnostic peak. Figure 5 shows the pyrogram of 75045 alkyd resin (Croda), a typical ortho-phthalic alkyd resin, with the mass spectrum of the very dominant phthalic anhydride peak on the left. The molecular ion (m/z = 148) is clearly seen, with the most intense peak

at m/z = 104 produced from the loss of CO2. The mass spectrum on the right is from palmitic acid (with a molecular ion clearly visible at m/z = 256), normally the most intense fatty acid observed from a dried oil component. The suspected mechanism of phthalic anhydride liberation from the alkyd’s polyester structure is as follows:

MODERN PAINTS 145

O O O O O + O O O

More recently work has been carried out to assess the advantages of carrying out an in situ methylation step at the time of pyrolysis, which appears to give a quantitative method of analysis (Cappitelli et al., 2002). This is being investigated to see whether oil type can be obtained reliably.

acetic acid + benzene VeoVa DBP

phthalic anhydride

palmitic acid

FIGURE 5 Top: Pyrogram of Emultex VV536, a PVA emulsion with mass spectrum of ethanoic acid/benzene (left) and dibutylphthalate plasticizer (right). Bottom: Pyrogram of Croda 75045 alkyd resin with mass spectrum of phthalic anhydride (left) and palmitic acid (right).

146 SCIENTIFIC EXAMINATION OF ART

FOURIER TRANSFORM INFRARED SPECTROSCOPY There are many ways of introducing a sample to a Fourier transform infrared spectroscopy (FTIR) instrument. The main technique currently employed at Tate is to compress the sample in a diamond cell and make the measurement through an infrared microscope, although a beam condenser seems to give equally good results (Learner, 1995b). Although spectra obtained from a diamond cell are arguably inferior to those from a KBr disc, the use of the diamond cell has three major advantages. First, the technique is nondestructive, which permits the sample to be retrieved and then reanalyzed by a complementary technique. Second, there is no sample preparation necessary. The diamond cell simply compresses the sample to a sufficiently reduced thickness for reliable transmission spectra to be obtained. Use of a diamond cell can be problematic with hard and brittle materials, but fortunately the majority of twentieth-century paints are fairly soft materials, which permit easy compression in the cell. Third, this soft nature of most synthetic polymers used in paints (in particular, the acrylics) makes grinding them into KBr powder very difficult. The main disadvantage of the diamond cell is the possibility of pressure effects on the spectrum, although the actual pressures used in the diamond cell are not thought to be particularly high. FTIR is an excellent way of obtaining information quickly about the basic chemical class of a binding material. For homogeneous samples, such as certain synthetic varnishes, this is relatively straightforward. Figure 6 shows the FTIR

FIGURE 6 FTIR spectra of unpigmented media. From top to bottom: acrylic solution (pink line), pEA/MMA-type acrylic emulsion (black line), PVA emulsion (blue line), alkyd resin (green line), and nitro-cellulose (red line).

MODERN PAINTS 147

spectra of five different synthetic binders, all obtained from films of unpigmented media. These are (shown from top to bottom) an acrylic solution, an acrylic emulsion, a PVA emulsion, an alkyd resin, and a nitrocellulose resin (one of the other less common types of modern paint binder). The FTIR spectra of paints are much more complicated as each additional component of the paint formulation, in particular the pigment(s) and extender(s), will exhibit their own individual vibrations and absorb the infrared radiation at those characteristic frequencies. Sometimes this can be a distinct disadvantage, especially if the spectrum from a particular pigment completely dominates the spectrum, thereby in effect masking out the absorptions from the binding media. However, in some instances the absorptions of the various components are so characteristic that it may be possible to sort out the individual bands by visual methods. Here the presence of overlapping bands can be turned into an advantage, as information can be gathered from all the individual components from a paint sample from a single analysis. An example of how FTIR can successfully identify each of the three main components in a paint is given in Figure 7, which shows four overlaid spectra. The pink curve is the overall spectrum, obtained from an acrylic emulsion paint:

FIGURE 7 Overall FTIR spectrum of Hyplar Hansa yellow medium acrylic emulsion paint (Grumbacher 1994 range) (pink line), with spectra of individual principal components: PY1 azo yellow pigment (black line), pEA/MMA binder (red line), chalk extender (green line).

148 SCIENTIFIC EXAMINATION OF ART

Hansa yellow medium artists’ acrylic color (Grumbacher). The other three curves are reference curves taken from each individual constituents.

• Red curve: The frequencies of the C-H stretching bands at 2986 cm–1 and 2955 cm–1, the overall profile of the C-H stretching region, C=O stretching at 1732 cm–1, and skeletal vibrations at 1179 cm–1 are all indicative of a p(EA/MMA) acrylic emulsion. • Grey curve: The sharp peaks seen at 1667 cm–1, 1602 cm–1, 1562 cm–1, 1508 cm–1, 1296 cm–1, 1140 cm–1, 953 cm–1, and 774 cm–1 are all present as strong absorptions in the spectrum of the pigment PY1, one of the common organic monoazo yellow pigments. The profile of absorptions in the region between 3000 cm–1 and 3300 cm–1, with peaks at 3098 cm–1, 3145 cm–1, 3181 cm–1, and 3243 cm–1 is also highly diagnostic of pigment PY1. • Green curve: the two absorptions at 2520 cm–1 and 1799 cm–1 are immediately indicative of the presence of chalk (calcium carbonate). Although relatively weak absorptions, these two wave numbers are normally found to be completely separated from the absorptions from all the binding media, pigments, and other extenders. The two very sharp peaks at 877 cm–1 and 713 cm–1 confirm the presence of chalk as extender and the strong and very broad absorption between 1400 cm–1 and 1500 cm–1 is also clearly visible.

Recently, attenuated total reflectance (ATR), a reflective mode of FTIR, has proved useful at identifying the differences at the surface of a paint film, compared with its bulk properties (as measured in transmission mode). This mode is showing great potential for following chemical surface changes on a paint film with age and after certain conservation treatments, such as cleaning. Figure 8 shows three stacked spectra measured with ATR. The top spectrum is from the upper surface of an unpigmented acrylic medium (Golden) that has been cast on a glass slide and been left for approximately five years. The middle spectrum is from the lower surface of the same sample, after removal from the slide. There are clear differences between the two. The peak assignments on each of these spectra that are placed to the left of the relevant peak are characteristic of a p(EA/MMA) acrylic emulsion. All peak labels that are placed to the right of their peaks are indicative of polyethylene glycol, a common class of surfactant. In this example the PEG has gathered at the upper surface of the paint film, a phenomenon that could have significant ramifications for a painting’s appearance and its change with age and/or cleaning.

CONCLUSIONS AND LOOKING AHEAD It is possible to identify, characterize, and differentiate the principal classes of synthetic binders used in modern paints with a combination of pyrolysis-gas chromatography-mass spectrometry (PyGCMS) and Fourier transform infrared

MODERN PAINTS 149

acrylic medium front acrylic medium rear 0.8 poly(ethylene glycol) standard 1107 0.6 2890 1149 1725 0.4 1343 2952 961 0.2

0.0 1149 -0.2 1725 -0.4 Absorbance

-0.6 2952

-0.8 1107

-1.0 2891 1342 -1.2 961 -1.4

4000 3000 2000 1000 Wavenumbers (cm-1) FIGURE 8 FTIR-ATR spectrum of unpigmented pEA/MMA acrylic medium (Golden 1993 range): upper surface (red line), lower surface (after removal from glass slide support) (black line), and polyethylene glycol (PEG) reference spectrum (blue line).

spectroscopy (FTIR). However, there still remain a great many analytical needs for modern paints, including

• improved quantitative methods of medium analysis; • analytical techniques for organic pigments and the additives added to paint formulations; • surface analysis methods for chemical, physical, and optical changes on aging and conservation treatments; and • high spatial resolution techniques to analyze individual layers from layered paint structures.

REFERENCES

Cappitelli, F., T. Learner, and O. Chiantore. 2002. An initial assessment of thermally assisted hydrolysis and methylation—gas chromatography/mass spectrometry for the identification of oils from dried paint films. Journal of Analytical and Applied Pyrolysis 63:339-348. Challinor, J. 1983. Forensic applications of pyrolysis gas chromatography. Forensic Science Interna- tional 21:269-285. Crook, J., and T. Learner. 2000. The Impact of Modern Paints. London: Tate Gallery Publishing. Irwin, W. 1979. Analytical pyrolysis—an overview. Journal of Analytical and Applied Pyrolysis 1:3-25.

150 SCIENTIFIC EXAMINATION OF ART

Jain, N., C. Fontain, and P. Kirk. 1965. Identification of paints by pyrolysis gas chromatography. Journal of the Forensic Science Society 5:102-109. Learner, T. 1995a. The analysis of synthetic resins found in twentieth century paint media. In Resins Ancient and Modern, eds. M. Wright and J. Townsend, pp. 76-84. Edinburgh: Scottish Society for Conservation and Restoration. Learner, T. 1995b. The use of a diamond cell for the FTIR characterisation of paints and varnishes available to twentieth century artists. Postprints: IRUG2 Meeting, pp. 7-20, available at http:// www.irug.org/documents/1Learner.pdf . Learner, T. 1996. The use of FTIR in the conservation of twentieth century paintings. Spectroscopy Europe 8(4):14-19. Learner, T. 2000. A review of synthetic binding media in twentieth century paints. The Conservator 24:96-103. Learner, T. 2001. The analysis of synthetic paints by pyrolysis-gas chromatography-mass spectrometry (PyGCMS). Studies in Conservation 46:225-241. Learner, T., O. Chiantore, and D. Scalarone. 2002a. Ageing studies of acrylic emulsion paints. Pre- prints of the 13th Triennial meeting of the ICOM Committee for Conservation, Rio de Janeiro, pp. 911-919. London: James and James. Learner, T., M. Schilling, and R. de la Rie. 2002b. Modern paints: A new collaborative research project. Conservation. The Getty Conservation Institute Newsletter 17(3):18-20, available at http: //www.getty.edu/conservation/resources/newsletter/17_3/news_in_cons1.html. Martens, C. 1981. Waterborne Coatings. New York: Van Nostrand Reinhold. Slinckx, M., and H. Scholten. 1994. Veova9/(meth)acrylates, a new class of emulsion copolymers. Journal of the Oil and Colour Chemists’ Association 77:107-112. Sonoda, N. 1998. Application des méthodes chromatographiques a la caractérisation des peintures alkydes pour artistes. Techne 8:33-43. Sonoda, N., and J.-P. Rioux. 1990. Identification des matériaux synthétiques dans les peintures modernes. 1. Vernis et liants polymères. Studies in Conservation 35:189-204. Stringari, C., and E. Pratt. 1991. The identification and characterization of acrylic emulsion paint media. In Saving the 20th Century: The Conservation of Modern Materials, ed. D. Grattan, pp. 411-439. Ottawa: Canadian Conservation Institute. Wheals, B. 1985. The practical application of pyrolytic methods in forensic science during the last decade. Journal of Analytical and Applied Pyrolysis 8:503-514.

APPENDIX EXPERIMENTAL CONDITIONS

Pyrolysis-Gas Chromatography-Mass Spectrometry FOM-4LX Curie point pyrolysis unit mounted directly onto the injection port of a Hewlett-Packard 5890 gas chromatograph and interfaced to a Finnigan MAT Incos 50 quadrupole mass spectrometer. Pyrolysis conditions: 610°C for eight seconds. Pyrolysis chamber kept at 200°C. GC injection port kept at 180°C. BPX-5 (SGE) nonpolar column: 25 meters long, 0.32 µm internal diameter and 0.1 µm film thickness. Temperature program: 40°C held for two minutes, then ramped at 10°C/min–1 to 350°C and held for two minutes. The transfer line was kept at 250°C. Incos 50 utilized EI ionization at 70 eV, and scanned from mass 35- 500 every second.

MODERN PAINTS 151

Fourier Transform Infrared Spectroscopy Transmission work carried out on a Nicolet Avatar 360 instrument with SpectraTech IR Plan microscope. Sample held in a diamond cell and 128 scans were averaged at 4 cm–1 resolution. ATR work carried out on a Nicolet Magna IR 560 instrument and Nicolet Nic Plan IR microscope with a Spectra-Tech ATR objective with zinc selenide crystal and purged with dry air. Two hundred scans were averaged at 4 cm–1 resolution.

ACKNOWLEDGEMENTS This work was made possible by the support of the Tate Gallery and the Leverhulme Trust, and the generosity of the FOM Institute (which loaned the PyGCMS instrument). The FTIR was purchased with a grant from the Clothworkers’ Foundation in London. Herant Khanjian at the Getty Conserva- tion Institute carried out the ATR measurements during the author’s guest scholarship there in 2001. The author is extremely grateful to all those involved.

Material and Method in Modern Art: A Collaborative Challenge

Carol Mancusi-Ungaro Associate Director of Conservation and Research Whitney Museum of American Art Director, Center for the Technical Study of Modern Art Harvard Univesity Art Museums

Recently I reacquainted myself with an illuminating interview of Jasper Johns by one of the more important interviewers of American artists in our time, David Sylvester. I knew the interview well. However, this particular rereading occurred as I was reexamining in a more thoughtful way Johns’s encaustic works in the Whitney Museum of American Art, including Three Flags (1958), White Target (1957), and Double White Map (1965). Despite my familiarity with the discussion, there was something that the artist said in the 1965 interview that gave me pause in 2003 and forced me to reconsider the nature of our collective professional charge to elucidate and care for works of art. In addressing the now famous paintings of flags and letters, Sylvester asked Johns about the objects with which he begins. Johns, seeking clarification, asked, “The empty canvas?” “No,” replied Sylvester, “Not only the empty canvas: well, the motif, if you like, such as the letters, the Flag and so on, or whatever it may be.” Johns said soberly, “I think it’s just a way of beginning.” Clearly surprised, Sylvester persisted, “In other words the painting is not about the elements with which you have begun.” Johns explained, “No more than it is about the elements which enter it at any moment. Say, the painting of a flag is always about a flag, but it is no more about a flag than it is about a brush-stroke or about a colour or about the physicality of the paint, I think.” Struck by the candor of this artist, who had obviously given enormous thought to the role of materiality in his art, I eagerly awaited clarification of his view, which came a few sentences later. He explained, “What I think this means is that, say in a painting, the processes involved in the painting are of greater certainty and of, I believe, greater meaning than the referential aspects of the painting. I think the processes involved in the

152

MATERIAL AND METHOD IN MODERN ART 153

painting in themselves mean as much or more than any reference value that the painting has.” “And what would their meaning be?” asks Sylvester. “Visual, intellectual activity, perhaps recreation,” answers Johns.1 Is Johns saying that the making of the work of art is its most relevant aspect? If so, does that mean that we can understand the painting only if we elucidate the process? Is the artist by implication suggesting that a conservator and a museum scientist must readily be at hand with explanations in order for a viewer to comprehend fully a work of art? As tantalizing as we may find that proposition, I believe it oversimplifies what the artist had in mind. Indeed, the cited passage may elicit diverse interpretations of Johns’s view, and his attitude may not necessarily be shared by other artists. In our context Johns’s comments focus attention on a relevant distinction that shapes the way we think about art. Clarifying his thoughts, the artist explained further, “And I think the experience of looking at a painting is different from the experience of planning a painting or of painting a painting. And I think the statements one makes about finished work are different from the statements one can make about the experience of making it.”2 Conservators, museum scientists, and art historians usually come upon the finished work of art. Although some of us may occasionally be a part of the making, generally our roles crystallize once the work of art is complete. After Pollock puts down his stick, Rothko retires his brush, Newman takes off his painting hat, and Johns exhausts his interest, the work of art moves away from the maker and into a realm of the viewer. As researchers interested in how substance and process affect the visual statement, conservators seek to enrich the aesthetic experience through elucidation of the process, while art historians considering primarily what is seen may posit and assess that information in a cultural context. The third collaborative component is the museum scientist who may not only affirm the nature of materials present but through analytic review may also reconfigure historical perspective. From different points of reference that shape different types of statement, as predicted by Johns, each inevitably seeks to resolve the visual and intellectual activity of the process insofar as it affects the meaning of the work of art. The conservation of the Rothko Chapel paintings, which engaged over 20 years of my professional life, provides a personal case in point. Created between 1964 and 1967 by Mark Rothko for a chapel he designed in Houston, Texas, the predominantly black and plum, so-called black-form, paintings had begun to develop a whitening on their surfaces less than five years after their installation in 1971. Over time the whitish films developed into crystals that gathered into dis-

1D. Sylvester. Interview. Jasper Johns Drawings (London: Arts Council of Great Britain, 1974), pp. 13-14. 2Ibid., p. 18.

154 SCIENTIFIC EXAMINATION OF ART

tinct patterns on the surfaces of the paintings. The patterns interfered with the unified, monochromatic nature of the paintings and certainly bore no relation to the final scheme for the chapel as determined by art scholars and as presumably intended by Rothko. For this reason conservators and conservation scientists were called upon to treat what was widely considered an inexplicable condition problem. In the literature, art historians and critics had focused on the dark palette and sharp contours of the chapel paintings as opposed to the bright amorphous coloration of his earlier paintings and even his earlier commissions, namely the Seagram murals of 1958-1959 and the Harvard panels of 1962. Although a resonance certainly existed among the seven black-form paintings and the seven plum paintings that comprised the whole of the chapel, no mention was made in the literature of the facture of the paint or the particular physical properties that shaped it. Rothko, who died before the chapel was consecrated, had been interviewed throughout his career, but none of the discussion focused on the materiality of these enigmatic paintings. My investigation began with the customary conservator’s question of how the paintings were made. To that end we sought and located one of Rothko’s assistants for the project, and he and I painted out simulations, using the same materials and processes that the artist had employed. Although our simulations certainly did not recall the originals, the material effect was close enough to confirm the ingredients of the mixture as whole eggs, tube oil paint, damar resin, and turpentine. Through analysis of the whitening conducted by local scientists at the Shell Oil Company, we were able to attribute the exudate to the migration of fatty acids from the paint, and ultimately we devised a treatment that enabled its removal. The strange patterns of rectangles formed by the exudate were explained by differential amounts of egg in a day’s mixture or by the buildup of media in consecutively layered forms. Ultimately, working drawings provided by the Rothko Foundation offered an astonishing correspondence to the patterns of whitening and thereby confirmed the relationship between the development of the condition and the unfolding of the creative act. What had begun as a conservator’s customary question of “what” were we looking at ended up providing insight into Rothko’s creative process and an explanation of “why” he chose to employ the particular materials that he did. In sum, the technical study offered information that had much broader ramifications for the history of the art. For example, a thoughtful appraisal of the drawings in graphite on black paper indicated that they were more than recordings of process. Throughout, what distinguished the line was not color but reflection. Indeed, in certain light the drawing was hardly visible. Rothko could have used white chalk instead of graphite on the black paper, but he did not. Rather, by his choice of materials the artist acknowledged that differences in reflectivity could be as legible as differences in contrast. Thought of in this way the studies fortified our developing notion of Rothko’s keen regard for nuances of surface as docu-

MATERIAL AND METHOD IN MODERN ART 155

mented also by the variable reflectance of the plum borders and black forms.3 Reference to his earlier work confirmed that the artist had been engaged with these issues throughout his career, but had brought them to fruition in part by eliminating vivid color in the chapel’s paintings. The technical analysis had informed the process that in turn directed the treatment and affected our regard for Rothko’s later work. Although statements about the experience of making art differ from statements about the finished work, they should inform each other. In the world of modern art this discourse could begin with closer scrutiny of its most common descriptive rubric, namely, “mixed media.” This term, which abounds not only on museum label copy but also in catalogues, is as familiar to postwar art scholars as “oil on canvas” is to those who study old-master painting. Walter Hopps, founding director of the Menil Collection, predicted the emergence one day of, in his terms, a “mixed media morass.”4 That era has arrived, but in some ways it is not a new phenomenon, considering that “oil on canvas” is the official description of Sir Joshua Reynolds’s Captain Robert Orne (1756), as well as Willem de Kooning’s Door to the River (1960). Given the visual range of these works of art and the investigative capability of our technological age, one overriding descriptive term seems woefully inadequate for both old-master painting and modern art. In a recent interview Wayne Thiebaud mentioned that he added Zec, a brand name for a substance that added girth to his oil paint, in order to create the creamy “icings” on his cakes. Willem de Kooning apparently did not add Zec, but we know from technical investigations that he did add vegetable oils to his media in order to achieve carefully sought-after working properties and effects that we value in his paintings.5 It is encouraging that recent studies of paintings by Pollock, de Kooning, Jacob Lawrence and Mark Rothko, among others, have identified materials by scientific analysis in the context of technique and have thereby broadened our understanding. Such analysis has also debunked prevailing myths, such as de Kooning’s alleged use of mayonnaise in his paint, and undoubtedly will substan- tiate others. These exemplary studies have also offered insight into the intellectual activity of the artist and that contribution about the making has broadened our understanding of the seen. Once that information becomes an integral part of how a work of art is discussed in the literature at large, we may begin to confront

3C. Mancusi-Ungaro. Nuances of surface in the Rothko chapel paintings. Mark Rothko: The Chapel Commission (Houston: The Menil Collection, 1996), pp. 27-28. 4Private conversation with the author. June 1988. 5S. Lake. The challenge of preserving modern art: a technical investigation of paints used in selected works by Willem de Kooning and Jackson Pollock. MRS Bulletin, January 2001, Volume 26. pp. 56-60.

156 SCIENTIFIC EXAMINATION OF ART

the disorder of the “mixed media morass” that confounds our scholarship. The importance of analytic review in this process cannot be overestimated. Generally, technical questions about modern art cannot be framed in the context of what we know about practice from artists’ treatises or the precepts of a guild system. Rather, they are often shaped by anecdotal information provided by artists or their assistants. Inevitably we have had to rely on art historical precedent and our eyes to assess the credibility of the information. That is not a bad combination, but it is not enough at a time when analytic confirmation is possible. Precise technical information may surface, but depending upon the source and context, its certainty may not be assured. In 1982 I wrote a short technical note about Yves Klein’s materials in a catalogue for a posthumous retrospective exhibition of the artist’s work. I based my information on interviews conducted in Paris with Klein’s former associates and on a patent that the artist had secured in 1960 for “International Klein Blue,” his preferred painting medium. The mixture consisted of dry pigments in polyvinyl acetate and industrial solvents, formulated by Rhône-Poulenc. Klein’s description of the medium in the patent actually differed from that provided by the company and seemed incompatible with the working properties necessary for its sundry applications. Nonetheless, the reason for my note was not to draw attention to a possible error in Klein’s application for a patent but rather to try to describe how his choice of material permitted widely diverse processes. Despite my conclusion that “though quantifiable, this quintessence of unencumbered color owes its vitality and beauty to the magic of the artistic endeavor—a factor that can never be measured or duplicated”—there arose a concern that I had demystified Klein’s art by describing its making.6 Considering that allegation with regard to our work, I am reminded of an esteemed engineer, Peter Rice, who once spoke about the role of Iago in Othello. He said, “Iago, if you remember, destroys the love of Othello and Desdemona by rational argument, by applying reason all the way through to every act which, particularly, Desdemona undertakes. And in the eyes of many, the Iago role is the role given to the engineer in modern life and in modern architecture of actually reducing by reason, to destroy or to undermine the kind of unreasonable and soaring ideas that architects may have.”7 I suspect the same could be said of museum scientists who decipher the material ambiguity of works of art. Admit- tedly the danger is there when the scientist is given a chip of paint in isolation and is asked to identify it. Out of context, devoid of its visual significance not to mention its role in artistic creation, the sample may be reduced to a fact that may not undermine the “soaring ideas” but certainly does little to enhance them.

6C. C. Mancusi-Ungaro. A technical note on IKB. Yves Klein (Houston: Institute for the Arts, Rice University, 1982), pp. 258-259. 7P. Rice. RIBA Royal Gold Medal speech. Arup Journal (winter 1992/1993), p. 20.

MATERIAL AND METHOD IN MODERN ART 157

Providing the data is one thing but explaining it in context is quite another. Cross-sections certainly help explicate the process in specific areas, and simulations can inform the overall technique; however, anyone who has observed an artist looking at a cross-section or tried to make a simulation of a work of art knows that deconstruction of substance and process is informative but far from art. There is the intangible element of the artist’s intent in manipulating tangible material that must be considered. Although associates or even studio assistants of an artist may not comprehend this factor, artists invariably do, because they appreciate the complexity of their undertaking. In part because we share a natural affinity for how substance and process affect the visual statement, I began interviewing artists in front of their work over 12 years ago. I intentionally chose an open interview style that was captured on film because in the presence of the works of art, artists invariably reveal through approach and reaction their relationships to the materials. Alternatively, museums often ask artists to complete questionnaires about technique when a work of art is acquired. If returned, these forms can impart important information. When asked about the materials she used in For the Light (1978-1979), for instance, Susan Rothenberg carefully listed the various media she had employed: Liquitex gesso-ground, Liquitex Matte Medium, and LeFranc and Bourgeois Flashe (vinyl paint made in France). She further noted, “All 3 used in conjunction with matte medium for both gesso + flashe.”8 When asked on the following page about the subject of the work, the ideas expressed and the circumstances under which it was executed, she replied with an emphatically drawn explanation point and question mark. The point is, of course, as Johns postulated years earlier, there is greater certainty about the processes than about the referential aspects. It is not so much a question of relative importance as it is of relative surety. From the outset it was clear to me that my questions would inevitably reflect the concerns of my own time and might therefore not provide answers to the problems that might confront future conservators. What I had hoped to document was not merely a discussion of materials and technique but, more than that, a solid sense of the artists’ concerns about what they were looking at and its future preservation. Naturally, artists’ relationships to their materials and thoughts about the future care of the art are as varied as their personalities. For instance, James Rosenquist may be concerned about the sinking in of his oil medium over time, while Brice Marden worries more about the proper treatment of a localized damage to one of his monochromatic works. The artists’ concerns may be narrow or broad in scope. Yet, inevitably their involvement adds another dimension to the investigation by posing questions unimagined by researchers and thereby enrich- ing the pursuit in unexpected ways.

8Questionnaire statement by S. Rothenberg. Whitney Museum of American Art archive. July 1, 1979.

158 SCIENTIFIC EXAMINATION OF ART

After the Menil Collection acquired Ed Kienholz’s John Doe (1959), the artist came to the museum to discuss the piece. Having carefully surveyed the surface of the sculpture, he opened the drawer and removed a portion of flute pipe from within. I had noted powdery debris in the bottom of the drawer, but I had not initially associated it with the industrial ductwork (Figure 1). Eventually Kienholz explained that the flute pipe represented the mannequin’s male private part and that the dust in the drawer was what remained of its head that had been fashioned from a rubber mask. Evidently the original Halloween mask had totally disintegrated over 30 years, but without the artist’s intervention I doubt we would have known about the change. Archival photographs of John Doe had unfortunately only documented the sculpture with the contents of the drawer locked within. Obligingly Kienholz took the appendage with him to Los Angeles and returned months later with a completed part fashioned out of a new rubber mask (Figure 2). This albeit extreme example raises two issues concerning preservation. First, instances of disintegrating industrial materials unfortunately comprise many important works of modern art, among them John Chamberlain’s foam sculptures of the 1970s. The challenge to preserve the physicality of these objects is enormous. Since the unstable material is central to the works of art and the sculptures cannot be properly viewed encased, it seems the only reasonable course is to restrict periods of exhibition as well as to require proactive storage containment. In this scenario more rigorous research might focus on the object in storage so that storage rooms become de facto laboratories wherein technical solutions are executed without regard for exhibition parameters or other customary restrictions. This approach to a limited degree has been adopted in some institutions, but it should become standard practice. The second issue concerns the broader philosophical question of what to do with the sculpture in the future when the current replacement disintegrates. The more expedient approach, of course, would be to replace the mask yet again, as did Kienholz. Once the artist has died, however, it is unlikely that anyone would be eager to refashion a new part without the artist’s hand. A conservator and a scientist’s approach might be to make a mold of the current form and then cast it in a more permanent material. An art historian could rightly object to the idea of a cast form replacing a found object because it counters Kienholz’s notion of materiality. One wonders if replacement parts are ever appropriate? I had to confront that question once when I made a new white wedding dress to replace the discolored and irreparably stained original on Kienholz’s Jane Doe (1960). I was asked to treat the work in this way by a curator because the aged dress offered a tawdry view of marriage that countered the deceased artist’s expressed intent. These two replacements, undertaken as treatments, were couched in terms that reflect the central importance of the artist’s intent—an elusive concept that defies quantification yet rests at the heart of our collective pursuit.

MATERIAL AND METHOD IN MODERN ART 159

FIGURE 1. Detail of John Doe, showing powdery remnants of original mask. Source: Ed Kienholz, John Doe, 1959, The Menil Collection. Photograph taken by Carol Mancusi- Ungaro. Permission for photograph granted by Nancy Reddin Kienholz.

160 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 Detail of restored John Doe showing mask fashioned as male anatomy. Source: Ed Kienholz, John Doe, 1959, The Menil Collection. Photograph taken by Carol Mancusi- Ungaro. Permission for photograph granted by Nancy Reddin Kienholz.

MATERIAL AND METHOD IN MODERN ART 161

When the artist is still alive, the complicated but key questions regarding intent are often more readily identified. A recent exchange involving two sculptures of a man and a woman by Kiki Smith at the Whitney Museum of American Art comes to mind. Dated 1990, they had been intermittently on exhibition for over a decade but had also spent a fair amount of time in storage. Recent observa- tion revealed patterns of white crystals in the beeswax that seemed to mimic the battens that held the pieces secure in their crates. Upon close examination the disfigurement and its probable cause were obvious to the scientist who sampled the material and the conservator who had begun to consider treatment options. In an impromptu interview the artist offered a totally unexpected assessment of the objects’ physical state. It seems what disturbed her most was not the exudate that commanded our attention, which she summarily dismissed, but rather a reddened pallor that had overcome the male figure. In her view the red wax that underlay the uppermost visible layer and had been used to offer skin tone had somehow become dominant. That condition problem, which had eluded us, far outweighed any other in terms of importance to her. By affirming the certainty of process, which Johns had observed, Smith not only left us with a better understanding of the nature of the problem but also of the work of art itself. Without her intervention would the fundamental alteration in the material have even elicited a question from us? Thoughts about the artist’s intent affect what conservators do with the facts that museum scientists uncover. A discussion of how the material is used and to what artistic end is as important as, if not more important than, what the material is. Analytic investigation is crucial, as are other types of review that take into account art history, criticism, and connoisseurship. All play a part in affirming artistic intent, especially after the artist has died. From their particular perspective scientists offer valuable insight in this debate by evaluating statements about the finished work in light of statements about the experience of making. Beyond naming the material and thinking logically in terms of questions and answers, they bring to the discussion diverse patterns of thinking. That contribution affects the tensions between “reason and intuition, certainty and uncertainty, delibera- tion and spontaneity,” the precise qualities that shape our reasoned comprehen- sion of the illogical artifacts of human expression in our care.9 It is only through intense collaboration among the distinct but related disciplines that consider works of art that we can attempt to frame and pose the relevant technical questions. By digesting the experience of looking as well as the experience of making, we can assign meaning to the chips of paint that are analyzed, offer definition to the “mixed media morass,” and discern the artist’s intent as it relates to materiality. Only in collaboration can we begin to offer the indeter- minate work of art the rigorous yet insightful review it deserves.

9A. Lightman. Art that transfigures science. New York Times, March 15, 2003, p. B9-10.

Raman Microscopy in the Identification of Pigments on Manuscripts and Other Artwork

Robin J. H. Clark Christopher Ingold Laboratories University College London London

The identification of pigments on manuscripts, paintings, enamels, ceramics, icons, polychromes, and papyri is critical in finding solutions to problems of restoration, conservation, dating, and authentication in artwork, and many techniques (molecular and elemental) have been used for this purpose. Raman microscopy has emerged, thanks to recent advances in optics and detectors, as perhaps the most suitable of these techniques on account of its high spatial (≤ 1 µm) and spectral (≤ 1 cm–1) resolution, its specificity, its excellent sensitivity by way of charge coupled device (CCD) detectors, and the fact that many artifacts may be analyzed in situ. Long-needed links between the arts and the sciences in this area are now rapidly being developed. The Raman effect was first detected in 1928 and rapidly became, and then remained for the following 20 years, the basis of the key technique for providing vibrational information on molecules in all states of matter, and of ions and lattice structures (Raman and Krishnan, 1928). However, the technique was then very slow, as it involved the use of a mercury arc as radiation source and photographic plates for detection. With the introduction in the 1960s of lasers as monochromatic polarized light beams of high irradiance and of semiconductors (e.g., GaAs) as detectors, the ability to detect weak Raman signals from materials of all kinds increased markedly. The technique came to be applied to a wide variety of chemical problems and increasingly to those amenable to analysis with a microsampling configuration. In the past 25 years, and especially in the last decade, the technique has been increasingly applied with great effect to the analysis of micrometre-sized particles using optical microscopes interfaced with existing spectrometer systems. These developments have revolutionized the usage of the

162

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 163

technique, which is seen to have many advantages over infrared (IR) microscopy and other techniques for phase identification. The intrinsic weakness of the Raman effect (Long, 2002) in the absence of resonance effects (Clark and Dines, 1986), and the moderate sensitivity of even semiconductors or multichannel intensified diode arrays as detectors were still not ideal features of Raman microscopy for many applications. However, the introduction of CCD detectors over the past decade has made it feasible to detect and identify even poor Raman scatterers of micrometre dimensions and, by change of excitation line, even many fluorescent materials. These technical advances have opened up many new areas to which Raman microscopy could make a major contribution, pigment identification being one of these. It is now recognized that the technique combines the attributes of high reproducibility and high sensitivity with that of being nondestructive. Moreover, the technique can be applied in situ, an important consideration for manuscript study. It also has high spatial resolution (≤ 1 µm) and high spectral resolution (≤ 1 cm–1), features that are of obvious value for establishing the composition of pigment mixtures and even of binders on works of art. Conservators and restorers need to be concerned with pigment identification for at least four reasons.

1. To decide whether (a) all restoration should be carried out with the original pigment and not with alternatives of similar hue; this is important since some alternatives might be liable to react with contiguous pigments with deleterious visual effects or (b) restoration with a different pigment might be desirable owing to instability of the original one or because it is more desirable to restore with carefully documented and easily identifiable nonoriginal pigment, possibly modern. 2. To identify any degradation products of pigments and to suggest possible treatments whereby degradation processes may be prevented, arrested or reversed. 3. It is increasingly likely that auction houses will be required to assess scien- tifically any works of art that they intend to offer for sale. One way of checking for obvious forgeries is to establish the palette and to check that no pigments of inappropriate dates of first manufacture or usage are present. Such analysis is extremely important in view of the immense prices—often > $1 million—for which medieval manuscripts and codices can currently be sold, usually without any scientific validation of the pigments present. 4. It is essential that conclusions on artwork as to the date, school, artist, etc., be drawn not only on paleographical, philological, and stylistic grounds but also on scientific grounds based upon an experimentally established palette. The arts community has been in general slow to accept this point.

This article begins with a brief comment on the range of pigments traditionally used to illuminate artifacts throughout time, and then moves on to a discus-

164 SCIENTIFIC EXAMINATION OF ART

sion of various case histories in which knowledge of pigments present and of their degradation products has proved to be of interest to both the arts and the science communities.

THE PIGMENTS The colors of inorganic pigments arise in most cases from ligand field, charge transfer or intervalence charge transfer transitions, and in the case of metals— commonly for silver or gold—from specular reflectance (Clark, 1995, 2002; Jorgensen, 1962). The depth of color of inorganic pigments is related both to the molar decadic absorption coefficients of electronic transitions in the visible region of the spectrum and to the dimensions of the pigment particles. These dimensions affect the balance between diffuse reflectance, which is controlled by the absorption coefficients and bandwidths, and specular reflectance, which is controlled by complementary factors (Clark, 1964). Early artists were well aware of these effects in practice, and with a restricted palette could often achieve a wide range of hues by selective control of particle size. Listings of standard inorganic pigments are given in reviews (Clark, 1995, 2002), books (Mayer, 1972; Feller, 1986; Roy, 1993; FitzHugh, 1997; Thompson, 1956; Gettens and Stout, 1966; Wehlte, 1975), and proprietary literature. By way of illustration, the commonly used blue inorganic pigments are listed in Table 1, together with their chemical names, formulas, provenance, and an indication as to the nature of the electronic transitions principally responsible for the blue color in each case. Of course, the identification of blue pigments alone would only be of restricted value for dating purposes, and so extensive studies of all the pigments on a work of art are essential in order for it to be possible to estimate the date of production of any particular piece. Many different organic dyes have been extracted from plants through the ages for use on illuminations of all sorts, notably indigo from woad for blue, alizarin from madder for red, weld from the weld plant (related to mignonette) for yellow, crocetin from saffron, and gamboge from gum resin—both also for yellow. In addition, organic dyes were extracted from marine life (e.g., Tyrian purple [6,6'-dibromoindigo] from mollusks and sepia from cuttlefish); from animal life (e.g., Indian yellow from cow urine); from insects (e.g., carmine from cochineal or kermes beetles); and others from lichens (Clark, 1995). Since the isolation by W. H. Perkin in 1856 of mauveine, the first synthetic dye, many hundreds of other organic dyes have been synthesized, and this has greatly extended the nature of the palette to which artists have had access. The main scientific concerns in the examination of a work of art are

• to identify each pigment, its crystal structure and, if possible, its place of origin and to identify the pigment medium; • to assess whether it is feasible to restore any degradation to the paintwork

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 165

TABLE 1 Commonly Used Blue Inorganic Pigments

Pigment Chemical Name Formula Datea Transitionb

Azurite Basic copper(II) 2CuCO3. min. LF carbonate Cu(OH)2

Cerulean blue Cobalt(II) stannate CoO.nSnO2 1821 LF

Chinese blue Barium BaCuSi4O10 ca 480 BC LF copper(II) silicate

Cobalt blue Cobalt(II)-doped CoO.nAl2O3 ca 1550 LF alumina glass Ming dynasty

Egyptian blue Calcium copper(II) CaCuSi4O10 ca 3100 BC LF silicate

Fluorite (and Calcium fluoride CaF2 min. Trapped antonozite) (purple) electrons?

Lazurite (from Sodalite + sulfur Na8[Al6Si6O24]Sn min. 1828 CT - - lapis lazuli) radical anions S3 , S2

Manganese blue Barium manganate(V) Ba(MnO4)2 + BaSO4 1907 LF sulfate

Maya blue Palygorskite/ Mg5(Si,Al)8O20 Mayan Mie c indigo/nano- (OH)2.8H2O, etc. scattering material π π d Phthalocyanine Copper(II) Cu(C32H16N8) 1936 - * blue /Winsor phthalocyanine blue

Posnjakite Basic copper(II) CuSO4. min. LF sulfate 3Cu(OH)2.H2O

Prussian blue Iron(III) Fe4[Fe(CN)6]3. 1704 IVCT hexa-cyanoferrate 14-16H2O Fe(II)/ Fe(III)

Smalt Cobalt(II) silicate CoO.nSiO2 Earlier than 1500 LF

Vanadium blue Vanadium(IV)-doped ZrSiO4(V(IV)) 1950? LF zircon

Verdigris Basic copper(II) 2Cu(O2CCH3)2. Corrosion LF acetate Cu(OH)2 product

Vivianite Iron(II,III)- phosphate Fe3P2O8.8H2O min. IVCT Fe(II)/ Fe(III)

aThe pigment is specified to be a mineral (min.) and/or the date of its first manufacture is listed. bLF = ligand field; CT = charge transfer; IVCT = intervalence charge transfer transition. cThe origin of the color is uncertain. (José-Yacamán M., L. Rendon, J. Arenas, and M. C. Serra Puche. 1996. Science 273:223-227.) dπ-π* = electric-dipole-allowed charge transfer transition of the phthalocyanine ring system.

166 SCIENTIFIC EXAMINATION OF ART

or fabric, given the nature of the chemical processes likely to be involved; • to consider whether knowledge of the identity of the pigments present on a work of art would give an indication as to date of production and hence the authenticity, school, and/or artist; and • to identify the correct measures needed to preserve a work of art from the effects of heat, light, and gaseous pollutants, and from contiguous or underlying pigments, dyes, or inks.

THE RAMAN MICROSCOPE In a Raman microscope the incident laser beam is brought to a focus by the objective onto each different pigment grain in turn on the manuscript under study (see Figure 1). The Raman scattering is collected by the same objective and then directed by a beamsplitter in the optical path to the monochromator and the detector. Although the use of a beamsplitter reduces the overall efficiency of the system, as does the use of a pinhole as a spatial filter, both devices restrict the amount of unwanted scattered light collected from outside the focus of the laser beam. The benefit of a pinhole is that it ensures a good confocal arrangement, thereby providing spatial resolution as a function of sample depth. The overall efficiency of detection of the Raman signal from the older Raman spectrometers was still relatively poor in the 1980s, because it was based on double

FIGURE 1 Schematic representation of a Raman microscope (Renishaw RM 1000), which employs notch filter assemblies and a CCD detector.

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 167

or triple spectrometers with a large number of optical surfaces and on multichannel intensified diode arrays, which are relatively insensitive detectors, albeit better than photographic plates. However, two devices have improved matters greatly: two-dimensional CCDs with up to 80 percent quantum efficiency of detection and holographic notch filters, which are wavelength specific and have the property of blocking out the unwanted Rayleigh scattering. These filters possess a cutoff that readily permits approach to within 50 cm–1 or less of the excitation line. This has the consequence that monochromators with the high dispersion previously required to filter out the Rayleigh line can be replaced by a spec- trograph system with a single grating only, leading to much greater efficiency of throughput. The introduction of notch filters eliminates the need for separate pinholes in the optical path. The benefits of CCD detector/notch filter spectrographs are that (1) low-powered air-cooled lasers can be used, which both reduce the costs of purchase and operation and make the entire system portable (since elaborate and fixed water-cooling systems for the lasers are not required); (2) the spectrographs are much lighter and much easier to realign than earlier double or triple grating systems; and (3) the time required to acquire significant data is greatly reduced, in some cases even to seconds. Such spectrographs can also be used for remote Raman microscopy in which a probe head assembly both delivers the excitation beam to the sample and collects the scattered radiation from the sample by means of fiber optics. They may be appropriate for the study of heavy items that need support from a specially designed cradle, notably large codices that cannot fit safely onto the microscope stage, or for archaeological studies of murals, cave paintings, etc. It is now also possible to collect data taken from many different sample points on an inhomogeneous surface to produce a Raman spectral map. The method involves direct two-dimensional imaging of an inhomogeneous surface by analyzing one or more of the Raman bands characteristic of a given component on the surface. It offers intriguing further opportunities for the study of very small (approximately 1 mm2) areas of artwork on stamps, maps, and writing (so as to be able to follow iron gall ink diffusion, paper damage, etc.), and in many other fields (e.g., that of identification of the precise whereabouts and proportions of active ingredients in pharmaceutical tablets). ν It is usually desirable to have a wide range of excitation lines (frequency 0) available in order to search for the most enhanced Raman spectrum, bearing in mind the often opposing effects on the scattering intensity of ν4 (the fourth power of the frequency of the scattered light), absorption, resonance, and possible photochemical and/or thermal degradation of the sample. Fluorescence from certain materials is often best avoided by use of a Nd/YAG (yttrium aluminum garnet) laser operating at 1064 nm, albeit with significant loss in spatial resolution and scattering intensity. By use of equipment such as that described above, libraries of Raman spectra of common inorganic, mineral and earth pigments, and organic pigments and dyes have been compiled (Griffith, 1987; Bell et al., 1997; Burgio

168 SCIENTIFIC EXAMINATION OF ART

TABLE 2 Comparison of Different Techniques for Pigment Identification

Compound specificity Technique and type of information Sensitivity

Raman excellent, molecular gooda IR excellent, molecular fair PLM fair, molecular fairb UV/VIS poor, molecular good LIBS good, elemental excellent XRF good, elemental goodc XPS good, elemental goode PIXE/PIGE good, elemental excellentg SEM/EDX good, elemental goode XRD excellent, molecular fairb

Raman, visible laser Raman microscopy; IR, mid-infrared reflectance microscopy; PLM, polarised light microscopy; UV/VIS, ultraviolet/visible reflectance spectroscopy or fibre optic reflectance spectroscopy (FORS); LIBS, laser-induced breakdown spectroscopy; SEM/EDX, scanning electron microscopy with Be-windowed energy dispersive X-ray detection; XRF, X- ray fluorescence spectroscopy; XPS, X-ray photoelectron spectroscopy, also called electron spectroscopy for chemical analysis (ESCA); PIXE/PIGE, external beam proton-induced X-ray emission/proton-induced γ-ray emission; XRD, powder X-ray diffraction. aSensitivity can be excellent under resonance conditions. bOnly possible with crystalline materials. cOnly atoms with atomic number Z ≥14 (Si) without an evacuated sample chamber.

and Clark, 2001; Bikiaris et al., 2000; Vandenabeele et al., 2000), as well as of some pigment media (binders, gums, resins, etc.), and these are now widely available for reference purposes. Many other techniques have been and are used for pigment identification, and estimates of their strengths and weaknesses vis-à-vis Raman microscopy have been given (Cilberto and Spoto, 2000; Pollard and Heron, 1996; Brundle et al., 1992; Bousfield, 1992; Smith and Clark, 2002b) and are summarized in Table 2. Raman microscopy is considered by many to be the best single technique for this purpose and is extremely effective when used in conjunction with other complementary techniques, such as polarized light microscopy (PLM), infrared microscopy (IR), X-ray fluorescence (XRF), X-ray diffraction (XRD), particle- induced X-ray emission (PIXE), or laser-induced breakdown spectroscopy (LIBS). Further techniques such as laser-induced fluorescence (LIF) have occasionally been used, as have nuclear ones, such as nuclear reaction analysis (NRA) and Rutherford back scattering (RBS), at laboratories in which a cyclotron source is available. Studies of Western and Eastern manuscripts, painting cross-sections, ceramics, papyri, icons, polychromes, and other artifacts (stamps, coins, etc.) as well as degradation and corrosion products are now discussed with particular reference

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 169

Immunity to interference Spatial resolution In situ analysis Portable

good excellent (< 1 µm) yes yes poor good (~ 20 µm) yes yes excellent good (~ 10 µm) no yes fair good (~ 10 µm) yes yes good good (~ 20 µm) yes no good faird (~ 1 mm) yes yes good fairf (~ 1 mm) yes no good fair (< 1 mm) yes no good excellent (< 1 µm) noh no good fair (< 0.1 mm) no no

dX-ray escape depth varies from nm to mm dimensions depending on the material and the element being detected. Can lead to loss of spatial resolution and to interference from sub-surface layers. eAll atoms with Z ≥ 5 (B). fDepth of sample is normally confined to several nanometres. gAll atoms with Z ≥ 11(Na). A proton beam (via PIXE) generates less bremsstrahlung background radiation than other X-ray techniques, providing greater sensitivity. PIXE can also be modified with different detectors to perform nuclear reaction analysis (NRA) and Rutherford back-scattering (RBS) studies. hEnvironmental SEM can analyse small, intact artefacts.

to those carried out in London. Raman microscopy also has wide application to many other fields of study, as discussed by Corset et al. (1989) and Turrell and Corset (1996).

WESTERN AND EASTERN MANUSCRIPTS The Anglo-Saxon manuscripts in the British Library, one of the world’s foremost collections, have been well studied from a paleographical standpoint but only slightly in respect of the materials used and the methods followed in their construction. Conclusive identification allows the correct materials to be used for the conservation of any object. Most of the pigments present on a large number of Western and Eastern manuscripts have now been identified at University College London, the more important items being cited below.

Western Manuscripts Early Raman studies of a Paris Bible of ca 1275 in Latin rapidly revealed the ease with which most inorganic pigments may be distinguished, even with Raman microscope systems of a 1980s design (Best et al., 1992, 1993). Spectra could

170 SCIENTIFIC EXAMINATION OF ART

FIGURE 2 Magnified (100 times) portion of a dark grey-black part of an illuminated letter R on a German choir book (sixteenth century) showing grains of different pigments, all identifiable by Raman microscopy (Best et al., 1992). Reproduced with permission, Elsevier.

readily be obtained from a pigment grain of 1-2 µm diameter, even when adjacent to other grains of different composition (see Figure 2), this is not an uncommon situation in artwork, since many artists choose to mix pigments in order to obtain shades of color not otherwise available. Other studies established the palettes of Latin manuscripts (Burgio et al., 1997a), a German choir book (Burgio et al., 1997b), German manuscripts (Burgio et al., 1997b), and illuminated plates from the Flora Danica (Burgio et al., 1999a). Of particular interest is the Skard copy of the Icelandic Book of Law, ca 1360, which was shown to have been richly illuminated, albeit not with either of the lead pigments commonly used in Europe at

that time (i.e., white lead [2PbCO3.Pb(OH)2] and red lead [Pb3O4] [Best et al., 1995]) but with bone white (Ca3PO4) and vermilion (HgS)/red ochre (Fe2O3), respectively, possibly owing to the lack of lead ores in Iceland. Extensive studies have now been carried out on many manuscripts and codices in the British Li- brary, the Victoria and Albert Museum, the Museum of London, the Beinecke

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 171

FIGURE 3 Illumination on the prologue page of the King George III version of the Guten- berg Bible held in the British Library (Chaplin et al., 2002a, 2005). Reproduced with permission, Wiley and the American Chemical Society.

Library, and elsewhere, the more recent work having the benefit of detection by CCD of the Raman scattered light from the pigments. One of the most notable of the recent studies is that in which the palettes of seven different Gutenberg Bibles were established (i.e., the King George III in the British Library, the ones at Eton College [Windsor] and Lambeth Palace [Lon- don], and two in France and two in Germany [Chaplin et al., 2002a, 2005]). These are brilliantly illuminated codices, the red, green, blue, white, and black pigments on the King George III Bible consisting of vermilion (HgS), lead tin

yellow Type I (Pb2SnO4), carbon black, azurite (2CuCO3.Cu(OH)2), malachite (CuCO3. Cu(OH)2), verdigris (approximately 2Cu(CH3CO2)2.Cu(OH)2), chalk (CaCO3), gypsum (CaSO4.2H2O), and lead white (2PbCO3.Pb(OH)2), in agreement with instructions given in the accompanying model book (a situation that is far from being always the case). The illuminations on the British Library and

172 SCIENTIFIC EXAMINATION OF ART

Eton College Bibles are similar to one another, consisting of interwoven flora and fauna around the columns of printed text (see Figure 3). Those on the Lambeth Palace Bible differ for the major illuminations, which consist of geometric patterns in blue, white, and gold but are similar for the minor ones. Also recently completed has been a detailed Raman study (Brown and Clark, 2004a) of the Lindisfarne Gospels (valued at perhaps $40 million), which represent to many the pinnacle of artistic achievement of manuscript illumination. Considered to have been created around 715 by Eadfrith in honor of St. Cuthbert, who was bishop of Lindisfarne in Northumbria during the period 685-687, the major pages display fantastic complexity of zoomorphic and interlace ornament, by contrast to the simplicity of the evangelist portraits. The most important result of the early pigment analyses of the Gospels by light microscopy was thought to be the apparent identification (Roosen-Runge and Werner, 1960) of lazurite (along with some indigo) on the evangelist portraits of St. Mark, f. 93v (folio page 93, verso) and St. Luke, f. 139, a result which would indicate the earliest known usage of lazurite on an Anglo-Saxon manuscript. Since the presence of non- indigenous materials on a work of art of known origin is regarded as indicating that a trade route between the source of the material and the place of construction of the work existed at its date of construction, the existence is implied at that very early date of a trade route to Northumbria from the Badakshan mines in Afghani- stan, then the only source of lazurite. The Gospels were recognised even in ca 715 to be very prestigious, and so there is little doubt that the most impressive and expensive blue pigment would have been used thereon, if available. However, Raman studies (Brown and Clark, 2004a) led to the identification of indigo alone, both on f. 93v as well as on f. 139 (see Figure 4), there being no evidence at all for lazurite. This suggests that trade in lazurite to Northumbria was most unlikely to have begun by the early eighth century. Several other substantial studies of Anglo- Saxon (Brown and Clark, 2004b,c) and Carolingian (Clark and van der Weerd, 2004) manuscripts have also recently been published.

Eastern Manuscripts Studies quickly revealed that the palette of Eastern manuscripts was not in general greatly different from that of Western ones, except in respect to some plant and animal extracts (e.g., Indian yellow [euxanthic acid,

MgC19H16O11.5H2O, from cow urine]). Included in early studies were ones on various Persian manuscripts: “Anatomy of the Body,” a nineteenth-century copy of an earlier manuscript, and “Poetry in Praise,” sixteenth century (Ciomartan and Clark, 1996); three very rare sixteenth-century copies of the Qazwini manuscript “Wonders of Creation and Oddities of Existence,” a late thirteenth-century encyclopedic work in Arabic but of Indian style (Clark and Gibbs, 1998a); a Qu’ran section, Iran or Central Asia, thirteenth century, eastern Kufic script (Clark and Huxley, 1996); a Byzantine/Syriac Gospel lectionary, Iraq, thirteenth

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 173

FIGURE 4 Initial page of the Gospel of St. Luke, f. 139, in the Lindisfarne gospels, ca 715 (Brown and Clark, 2004a). Reproduced with permission, Wiley.

174 SCIENTIFIC EXAMINATION OF ART

century (Clark and Gibbs, 1998b); manuscript and textile fragments from Dunhuang, northwest China, tenth century (Clark et al., 1997a); Thai, Javanese, Korean, Chinese, and Uighur manuscripts (Burgio et al, 1999b); and a precious sixteenth-century Turkish manuscript (Hurev U Sirin) (Jurado-Lopez et al., 2004). The serious and widespread blackening of the white areas on the valuable lectionary referred to above (Clark and Gibbs, 1998b) is considered to arise from the conversion of white lead, either alone or in admixture with colored pigments, into black lead(II) sulfide by hydrogen sulfide or other sulfur-containing species. Hydrogen sulfide could arise from atmospheric pollutants, from bacteria, from degradation of adjacent pigments, or from pigments on the reverse side of the page containing the illumination in question. This blackening is a widespread problem in artwork, the reversal of which has posed as many questions as answers. Treatment of lead(II) sulfide with alkaline hydrogen peroxide generates lead(II) sulfate, which is white, and this superficially at least, appears to solve the problem. Whether the resulting pigment will remain permanently attached to the manuscript has yet to be established.

The photochemical conversion of the red pigment realgar (As4S4) to yellow pararealgar (also As4S4) by light optimally in the wavelength range 530-560 nm is of great interest. The conversion occurs naturally in sunlight and was apparently recognized in Mesopotamia even as early as 1220, when pararealgar was applied to the above lectionary as a yellow pigment in addition to, and in distinctly differ-

ent places from, the much more common yellow pigment orpiment, As2S3 (Clark and Gibbs, 1998b).

PAINTINGS It may be possible to remove samples from watercolor or oil paintings in such a way that the lacunae are not discernible to the naked eye, a procedure that is rarely if ever permitted for manuscripts. Alternatively, the pigment might be able to be sampled from under the frame of a painting or, in the case of codices, from offsets transferred to the opposite page. A further possibility is to sample pigment cross-sections, a procedure with the advantages of requiring neither the removal of the work from its permanent location nor, equally importantly, the removal of delicate and valuable scientific equipment from a laboratory to a library. Studies of cross-sections are of considerable importance for paintings and icons when depth profile analyses are required. Many Raman studies on pigments removed from paintings have now been made in order to complement those made by infrared spectroscopy and other techniques. For example, studies of Titian and Veronese paintings at the National Gallery in London have allowed the characterization of two distinct types of lead

tin yellow, Type I, Pb2SnO4, and Type II, PbSn0.76Si0.24O3, which was shown to have a defect pyrochlore structure (Clark et al., 1995). These two pigments, in use

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 175

FIGURE 5 Painting “Young Woman Seated at a Virginal” under study by Raman microscopy to identify the pigments present, and leading to evidence consistent with its attribution to Vermeer, ca 1670 (Burgio et al., 2005). Reprinted with permission, American Chemical Society.

at different periods of history, have very similar colors, but they are readily distin- guishable by their Raman spectra. Some paintings are sufficiently small that they may be examined in situ under a Raman microscope. Figure 5 shows a painting—considered by art historians to be from the late 17th century and Dutch —of a young woman with red ribbons in her hair, a pearl necklace, and a cream-coloured skirt beneath a yellow shawl. The pigments (vermilion, lazurite, and lead tin yellow Type I, particularly the last two) identified by Raman microscopy and other techniques are consistent with the attribution of this painting to Vermeer. In consequence of these identifications and other considerations, when it was auctioned at Sotheby’s on July 8, 2004, it realised US$30 million (Burgio et al., 2005).

176 SCIENTIFIC EXAMINATION OF ART

CERAMICS The effective use of Raman microscopy in the study of ceramics was demonstrated in 1997 on the glaze from buried fragments of medieval faience/majolica excavated from the abandoned village of Castel Fiorentino in southern Italy. The studies showed for the first time that lapis lazuli is stable at the firing temperature of the glaze and that it had been used as a pigment in Italian glaze; the brown/ black pigment used was manganese(IV) dioxide (the latter identified at the time by photoelectron spectroscopy) (Clark et al., 1997b,c). Red-brown shards of medieval pottery from many sites in Italy have, not surprisingly, been identified to be pigmented with iron oxides and yellow shards with hydrated iron(III) oxides (Clark and Curri, 1998). Raman studies of many different ochres used in wall paintings have also been carried out (Bikiaris et al., 2000). A Raman study (Clark and Gibbs, 1997) of the pigments on Egyptian faience of the XVIIIth dynasty (ca 1350 BC) recovered in 1891 from the Nile Valley by Sir William Flinders Petrie and held at the Petrie Museum, University College Lon- don, has revealed that the red shards are pigmented with red ochre/red earth and

the brilliant yellow shards with lead antimony yellow (Pb2Sb2O7), a very early synthetic pigment dating back to ca 1500 BC (see Figure 6). The latter gives a very clear and distinctive Raman spectrum that is identical to that of a contemporary sample of this pigment. Similar studies of shards from an ancient (4300-2800 BC)

FIGURE 6 Lotus leaf shard from El Amarna; the pigment used is lead antimony yellow, Pb2Sb2O7 (Clark and Gibbs, 1997). Reproduced with permission, Wiley.

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 177

site in Xishan, Henan, China, indicate that anatase, hematite, and magnetite were among the pigments used to decorate pottery found at that site (Zuo et al., 1999). Raman studies have been made by Colomban et al. (2001) of the palettes characteristic of the Sèvres Factory, one relating to colored glazes/enamels on bisque (1050 °C) and the other on moufle (850 °C) painting colors. Other recent studies by the same group relate to Vietnamese porcelain and celadon glazes (Liem et al., 2002; Faurel et al., 2003). It is now clear that Raman microscopy is a very valuable technique for compositional and provenance studies on ceramics from sites of archaeological interest (van der Weerd et al., 2004b), this area having recently been reviewed by Smith and Clark (2004).

PAPYRI Egyptian papyri supposedly dating from the thirteenth to the first centuries BC and brought to London for auction were recently shown to be illuminated not only with mineral pigments but also with the modern pigments phthalocyanine blue (1935) and green (1936), a Hansa yellow (ca 1950), ultramarine blue (1828)— the synthetic form of lazurite, red organic lakes (probably β-napthols, ca 1939), and synthetic anatase (1923) (Burgio and Clark, 2000). Moreover, the pigments had been painted directly onto the papyri, with no intervening ground layer of mineral pigments. The papyri are clearly modern. An authentic papyrus from the Petrie Museum at University College London had no modern pigments on its

illuminations, only carbon, orpiment (As2S3), malachite, and Egyptian blue (CaCuSi4O10), the earliest synthetic pigment (ca 3000 BC). Such discoveries highlight the urgent need for proper scientific evaluation of items offered for sale or auction; indeed purchasers increasingly expect this prior to purchase.

ICONS AND POLYCHROMES The combined application of Raman spectroscopy and LIBS has proved to be very effective for the identification of pigments at different depths below the surfaces of icons and polychromes. LIBS is an atomic emission technique in which an intense nanosecond laser pulse onto the surface of the sample results in the formation of plasma which, upon being allowed to cool, emits radiation characteristic of the elements present. The technique has high sensitivity and selectivity, and only a minute amount of material is consumed during each pulse (Anglos et al., 1997). Successive pulses probe deeper into the artwork, so that depth profiling becomes possible. A continuous-wave laser beam can be used as a Raman probe at each depth and so complementary molecular information can also be obtained. Recent combined LIBS and Raman studies of a nineteenth-century Russian icon (see Figure 7) have revealed the identities of the pigments found in the upper layers to be white lead, zinc oxide (ZnO, for repair purposes), vermilion, and red earth, etc., all above a silver foil. Below this was found the white ground consisting

178 SCIENTIFIC EXAMINATION OF ART

FIGURE 7 Nineteenth-century Russian icon of St. Nicholas held in Greece on which studies by LIBS and Raman microscopy enabled depth profile analyses of the pigments to be made (Burgio et al., 2000). Reproduced with permission, Society for Applied Spectros- copy.

mainly of gypsum immediately above the wood (Burgio et al., 2000). Similar studies of a rococo polychrome, a fragment from a gilded altarpiece in a church in Escatrón in Spain, has been carried out (Castillejo et al., 2000), and these reveal the power of the combined application of the two techniques for stratigraphic analyses of the pigmentation on artwork. An extensive study of cross-sections from two post-Byzantine icons from Chalkidiki, Greece, have revealed the current state of preservation of these icons, any damage thereto, and details of the pigments and materials used in the original paintings and in their overpaintings (Daniilia et al., 2002). Similar details are revealed in cross-sections from a monastic habit on the icon of St. Athanasios the Athonite (see Figure 8).

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 179

FIGURE 8 A. Cross-section of a monastic habit painted on a Greek icon; photography with a microscope in reflected light. B. Spectra of pigments taken with a Raman microscope. Identification: (a) underlayer: caput mortuum, lead white, azurite, red lake, and yellow ochre; (b) highlight: lead white and grains of caput mortuum; (c) varnish; (d) overpainting: ultramarine blue, minium, lithopone, and carbon black (Daniilia et al., 2002). Reproduced with permission, Wiley.

PHILATELY Raman microscopy has recently been shown to have potential for establishing whether postage stamps are authentic or forgeries by way of offering an effective, rapid, and nondestructive way of identifying the pigments and dyes used in the inks, paper, and cancel marks. Thus the rare and valuable Hawaiian Missionary stamps (1851) from the Tapling Collection at the British Library were shown to

have been printed using Prussian blue, Fe4[Fe(CN)6]3.14-16H2O, as the blue pig-

180 SCIENTIFIC EXAMINATION OF ART

FIGURE 9 Hawaiian Missionary stamp (left) of 1851 showing (right) the blue printing (Prussian blue, Fe4[Fe(CN)6]3.14-16H2O) on the surface of the stamp (top left) and particles of ultramarine blue (bottom right) within the paper fibers (Chaplin et al., 2002b). Reproduced with permission, Wiley.

ment (see Figure 9). In addition, the paper fibers of the stamps were shown to have ultramarine blue particles interspersed between them, to act as an optical brightener. Distinctions between genuine and forged or reproduction stamps can be drawn on the basis of the pigments used (Chaplin et al., 2002b). Similar studies of the earliest Mauritian stamps (1847) have been carried out, notably on an extremely rare one penny (1d, orange-red, used) stamp, a rare 2d (blue, unused) stamp, as well as on a reproduction stamp (1905), early forgeries, and Britannia-type stamps (1858-1862), in order to identify the pigments used.

For the Britannia-type stamps the pigments used were red lead (Pb3O4) on the 1d stamp, Prussian blue on the 2d stamp, chrome green—a mixture of Prussian blue

and chrome yellow (PbCrO4)—on the 4d stamp, and vermilion (HgS) on the 6d (orange) stamp. The technique has great potential for expertising, i.e. distinguishing between, genuine and forged or reproduction, stamps (Chaplin et al., 2004).

WALL PAINTINGS Many studies have now been carried out on the pigments used to illuminate colored frescos, whose palettes are much more restricted than those of manuscripts and paintings. By contrast with scriptoria, libraries, and museum collections, where pigment degradation has usually only been slight, wall paintings and frescos often show obvious signs of overexposure to environmental extremes of temperature and humidity. Moreover, the effects of microbial, fungal, and lichen colonization on exposed frescos can be severe (Perez et al., 1999). Many more Raman studies of the pigments and pigment degradation products on frescos and on wall paintings in caves are likely to be made with the effective development of mobile Raman systems (Clark and Gibbs, 1998a).

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 181

PIGMENT DEGRADATION PROBLEMS Manuscripts are subject to problems arising from the degradation of pigments, particularly those that are copper-, arsenic-, or lead-based. Degradation products of pigments can be identified by Raman microscopy, enabling each such process to be revealed and its progress monitored. Verdigris is an umbrella term used for any green or blue corrosion product resulting from the action of atmospheric agents on copper, in particular acetic acid and sometimes formic acid. It can be regarded as a disfiguring product of the corrosion of objects made of copper alloys or as a pigment made deliberately by corrosion of copper or by the conversion of a copper compound. Moreover, it dissolves in many oils and resins to form copper resinates that can themselves be dissolved in size or gelatin to form copper proteinates (Scott et al., 2001). Raman microscopy has recently been applied to the problem of characterizing these chemically very similar compounds. Many black minerals are used, or may have been used, as pigments at differ-

ent periods of time, notably carbon black, chromite (FeCr2O4), covellite (CuS), galena (PbS), ilmenite (FeTiO3), magnetite (Fe3O4), plattnerite (PbO2), pyro- lusite (MnO2), tenorite (CuO), and silver glance (Ag2S). Both galena and plattnerite may develop on artwork as degradation products of other lead- containing pigments, notably white lead, the net result being a gross disfigurement of the manuscript or painting. The identification may now be carried out using modern Raman microscopes and excitation lines of low power, both di-

rectly and by way of the recognition of their oxidation products (i.e., PbO.PbSO4, 3PbO.PbSO4, and 4PbO.PbSO4) (Giovannoni et al., 1990; Burgio et al., 2001). Reversal of black PbS to white PbSO4 is one option sometimes used for restoration of the intended effects of the artist, but this procedure is by no means fully accepted by conservators. Detailed Raman studies have also been carried out on millimetre-sized single crystals of PbS, a model material for quantum dot research, under resonance Raman conditions, and the phonon modes identified and assigned (Smith et al., 2002b). In Figure 10 a detail of the illumination (a set of dividers) in volume 1, f. 33v, of the Jamnitzer Manuscript reveals that severe degradation of lead white to lead sulfide can readily be demonstrated in situ by Raman microscopy, most particularly in the highlights (now black) (Smith et al., 2002a). Copper pigments, such as

azurite, likewise rapidly degrade to covellite in the presence of gaseous H2S (Smith and Clark, 2002a), as is also easily shown by Raman microscopy (see Figure 11). Raman microscopy may also be applied to the study of the corrosion of metals (Martens et al., 2003), including the main encrustments on ancient bronzes derived from copper, tin, and lead. Thus bronze artifacts from Chinese tombs of the Eastern Han dynasty (25-220) have been shown to include among their corro-

sion products Cu2O (cuprite), CuCO3.Cu(OH)2, PbO, PbCO3, and PbSO4 (McCann et al., 1999). Two-dimensional mapping provided information on the

182 SCIENTIFIC EXAMINATION OF ART

FIGURE 10 Detail of the illumination in vol. 1, f. 33v, of the Jamnitzer manuscript at the Victoria and Albert Museum, London, in which the original highlights of lead white have degraded to lead sulfide (black) (Smith et al., 2002a). Reproduced with permission, Inter- national Institute for Conservation.

FIGURE 11 Raman spectra of covellite, CuS, azurite, 2CuCO3.Cu(OH)2, and blackened azurite following brief exposure of the azurite to H2S vapor (Smith and Clark, 2002a). Reproduced with permission, Elsevier.

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 183

spatial extent of each corrosion product. Many extensions to this work are being planned with the vast collections of degraded metals in museums in order to understand the nature and causes of the degradation. The most recent studies of this sort have led to the Raman-based identification of iron oxide impurities (magnetite, iron-deficient magnetite, and hematite) in early industrial-scale pro- cessed platinum (e.g., in the platinum metal constituting the three-ruble Russian coin of 1837) (van der Weerd et al., 2004a). Modern Raman spectrometers may now be capable of identifying iron gallotannate inks on manuscripts, a notoriously difficult task. Recent studies on the Vinland Map (Beinecke Library, Yale University) have shown that the frag- mented black ink lines that define the map consist of carbon black rather that iron gallotannate and that the yellow-brown background to the lines, but not else-

where, contains nearly pure anatase (TiO2) (Brown and Clark, 2002). This material had earlier been shown to have a particle size (approximately 0.15 µm) and particle size distribution that is characteristic of the synthetic ca 1920 product (McCrone, 1988). The Raman results thus confirm that the map dates from the twentieth not the early fifteenth century, and thus that it is not pre-Columbian; this is in complete agreement with the analysis of Towe (1990).

CONCLUSION Raman microscopy is now established to be a key technique for the identification of pigments on works of art, largely because of its high spatial and spectral resolution, excellent sensitivity and specificity, and because it can be applied to an object in situ. Difficulties may arise on occasions with certain organic pigments, supports, and binders that fluoresce, that are photosensitive, or that fail to yield a Raman spectrum owing to their small particle size, high dilution, or poor scattering efficiency. The use of other techniques in conjunction with Raman microscopy then becomes essential in order to effect full pigment characterization. Re- mote laser Raman microscopy (Clark and Gibbs, 1998a) will increasingly be used for the study of objects unable to be moved from their place of exhibition. The whole area is one in which the arts and the sciences can coordinate with great effect.

ACKNOWLEDGEMENTS The author is most grateful to the members of his group in this field, most recently Drs. K. L. Brown, L. Burgio, T. D. Chaplin, S. Firth, A. Jurado-Lopez, G. D. Smith, and J. van der Weerd, and to Renishaw PLC, the Engineering and Physical Sciences Research Council, the European Union, and the British Library for their support of this research.

184 SCIENTIFIC EXAMINATION OF ART

REFERENCES

Anglos, D., S. Couris, and C. Fotakis. 1997. Applied Spectroscopy 51:1025-1030. Bell, I. M., R. J. H. Clark, and P. J. Gibbs. 1997. Spectrochimica Acta Part A 53:2159-2179. Best, S. P., R. J. H. Clark, and R. Withnall. 1992. Endeavour, New Series 16:66-73. Best, S. P., R. J. H. Clark, M. A. M. Daniels, and R. Withnall. 1993. Chemistry in Britain 118-122. Best, S. P., R. J. H. Clark, M. A. M. Daniels, C. A. Porter, and R. Withnall. 1995. Studies in Conserva- tion 40:31-40. Bikiaris, D., Sister Daniilia, S. Sotiropoulou, O. Katsimbiri, E. Pavlidou, A. P. Moutsatsou, and Y. Chryssoulakis. 2000. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 56:3-18. Bousfield, B. 1992. Surface Preparation and Microscopy of Materials. Chichester, U.K.: Wiley. Brown, K. L., and R. J. H. Clark. 2002. Analytical Chemistry 74:3658-3661. Brown, K. L., and R. J. H. Clark. 2004a. Journal of Raman Spectroscopy 35:4-12. Brown, K. L., and R. J. H. Clark. 2004b. Journal of Raman Spectroscopy 35:181-189. Brown, K. L., and R. J. H. Clark. 2004c. Journal of Raman Spectroscopy 35:217-223. Brundle, C. R., C. A. Evans, and S. Wilson. 1992. Encyclopaedia of Materials Characterization. Boston: Butterworth-Heinemann. Burgio, L., and R. J. H. Clark. 2000. Journal of Raman Spectroscopy 31:395-401. Burgio, L., and R. J. H. Clark. 2001. Spectrochimica Acta Part A 57:1491-1521. Burgio, L., D. Ciomartan, and R. J. H. Clark. 1997a. Journal of Raman Spectroscopy 28:79-83. Burgio, L., D. Ciomartan, and R. J. H. Clark. 1997b. Journal of Molecular Structure 405:1-11. Burgio, L., R. J. H. Clark, and H. Toftlund. 1999a. Acta Chemica Scandinavica 53:181-187. Burgio, L., R. J. H. Clark, and P. J. Gibbs. 1999b. Journal of Raman Spectroscopy 30:181-184. Burgio, L., R. J. H. Clark, T. Stratoudaki, D. Anglos, and M. Doulgeridis. 2000. Applied Spectroscopy 54:463-470. Burgio, L., R. J. H. Clark, and S. Firth. 2001. Analyst 126:222-227. Burgio, L., R. J. H. Clark, L. Sheldon, and G. D. Smith, 2005. Analytical Chemistry 77:1261-1267. Castillejo, M., M. Martin, D. Silva, T. Stratoudaki, D. Anglos, L. Burgio, and R. J. H. Clark. 2000. Journal of Molecular Structure 550:191-198. Chaplin, T. D., R. J. H. Clark, D. Jacobs, K. Jensen, and G. D. Smith. 2002a. In Raman Spectroscopy, eds. J. Mink, G. Jalovszky, and G. Kereszbury, pp. 823-824. Chichester, U.K.: Wiley. Chaplin, T. D., R. J. H. Clark, D. Jacobs, K. Jensen, and G. D. Smith. 2005. Analytical Chemistry 77. Chaplin, T. D., R. J. H. Clark, and D. R. Beech. 2002b. Journal of Raman Spectroscopy 33:424-428. Chaplin, T. D., A. Jurado-Lopez, R. J. H. Clark, and D. R. Beech. 2004. Journal of Raman Spectroscopy 35:600-604. Cilberto, E., and G. Spoto, eds. 2000. Modern Analytical Methods in Art and Archaeology. New York: Wiley. Ciomartan, D., and R. J. H. Clark. 1996. Journal of the Brazilian Chemical Society 7:395-402. Clark, R. J. H. 1964. Journal of Chemical Education 41:488-492. Clark, R. J. H. 1995. Chemical Society Reviews 24:187-196. Clark, R. J. H. 2002. In Handbook of Vibrational Spectroscopy, eds. J. M. Chalmers and P. R. Griffiths, pp. 2977-2992. Chichester, U.K.: Wiley. Clark, R. J. H., and M. L. Curri. 1998. Journal of Molecular Structure 440:105-111. Clark, R. J. H., and T. J. Dines. 1986. Angewandte Chemie-International Edition in English 25:131-158. Clark, R. J. H., and P. J. Gibbs. 1997. Journal of Raman Spectroscopy 28:99-103. Clark, R. J. H., and P. J. Gibbs. 1998a. Journal of Archaeological Science 25:621-629. Clark, R. J. H., and P. J. Gibbs. 1998b. Analytical Chemistry 70:99A-104A. Clark, R. J. H, and K. Huxley. 1996. Science and Technology for Cultural Heritage 5:95-101. Clark, R. J. H., and J. van der Weerd. 2004. Journal of Raman Spectroscopy 35:279-283. Clark, R. J. H., L. Cridland, B. M. Kariuki, K. D. M. Harris, and R. Withnall. 1995. Journal of the Chemical Society, Dalton Transactions 2577-2582.

RAMAN MICROSCOPY IN THE IDENTIFICATION OF PIGMENTS 185

Clark, R. J. H., P. J. Gibbs, K. R. Seddon, N. M. Brovenko, and Y. A. Petrosyan. 1997a. Journal of Raman Spectroscopy 28:91-94. Clark, R. J. H., M. L. Curri, and C. Laganara, 1997b. Spectrochimica Acta Part A 53:597-603. Clark, R. J. H., M. L. Curri, G. S. Henshaw, and C. Laganara. 1997c. Journal of Raman Spectroscopy 28:105-109. Colomban, P., G. Sagon, and X. Faurel. 2001. Journal of Raman Spectroscopy 32:351-360. Corset, J., P. Dhamelincourt, and J. Barbillat. 1989. Chemistry in Britain 612-616. Daniilia, Sister, D. Bikiaris, P. Gavala, R. J. H. Clark, and Y. Chryssoulakis. 2002. Journal of Raman Spectroscopy 33:807-814. Faurel, X., A. Vanderperre, and P. Colomban. 2003. Journal of Raman Spectroscopy 34:290-294. Feller, R. L., ed. 1986. Artists’ Pigments, vol. 1. Cambridge: Cambridge University Press. FitzHugh, E. W., ed. 1997. Artists’ Pigments, vol. 3. Oxford: Oxford University Press. Gettens, R. J., and G. L. Stout. 1966. Painting Materials. New York: Dover. Giovannoni, S., M. Matteini, and S. Moles. 1990. Studies in Conservation 35:21-25. Griffith, W. P. 1987. In Advances in Spectroscopy, vol. 14, eds. R. J. H. Clark and R. E. Hester, pp. 119- 186. Chichester, U.K.: Wiley. Jorgensen, C. K. 1962. Absorption Spectra and Chemical Bonding. Oxford: Pergamon. Jurado-Lopez, A., O. Demko, R. J. H. Clark, and D. Jacobs. 2004. Journal of Raman Spectroscopy 35:119-124. Liem, N. Q., N. T. Thanh, and P. Colomban. 2002. Journal of Raman Spectroscopy 33:287-294. Long, D. A. 2002. The Raman Effect. Chichester, U.K.: Wiley. Martens, W., R. L. Frost, J. T. Kloprogge, and P. A. Williams. 2003. Journal of Raman Spectroscopy 34:145-151. Mayer, R. 1972. The Artist’s Handbook. London: Faber and Faber. McCann, L. I., K. Trentleman, T. Possley, and B. Golding. 1999. Journal of Raman Spectroscopy 30:121-132. McCrone, W. C. 1988. Analytical Chemistry 60:1009-1018. Perez, F. R., H. G. M. Edwards, A. Rivas, and L. Drummond. 1999. Journal of Raman Spectroscopy 30:301-305. Pollard, A. M., and C. Heron. 1996. Archaeological Chemistry. Cambridge: Royal Society of Chemistry. Raman, C. V., and K. S. Krishnan. 1928. Nature 121:501. Roosen-Runge, H., and A. E. A. Werner. 1960. In Evangeliorum Quattuor Codex Lindisfarnensis, vol. 2, ed. T. D. Kendrick, pp. 263-272. Lausanne: Urs Gras. Roy, A., ed. 1993. Artists’ Pigments, vol. 2. Oxford: Oxford University Press. Scott, D. A., Y. Taniguchi, and E. Koseto. 2001. Reviews in Conservation 2:73-91. Smith, G. D., and R. J. H. Clark. 2002a. Journal of Cultural Heritage 3:101-105. Smith, G. D., and R. J. H. Clark. 2002b. Reviews in Conservation 2:92-106. Smith, G. D., and R. J. H. Clark. 2004. Journal of Archaeological Science 31:1137-1160. Smith, G. D., A. Derbyshire, and R. J. H. Clark. 2002a. Studies in Conservation 47:250-257. Smith, G. D., S. Firth, R. J. H. Clark, and M. Cardona. 2002b. Journal of Applied Physics 92:4375-4380. Thompson, D. V. 1956. The Materials and Techniques of Painting. New York: Dover. Towe, K. M. 1990. Accounts of Chemical Research 23:84-87. Turrell, G., and J. Corset, eds. 1996. Raman Microscopy: Developments and Applications. London: Academic Press. van der Weerd, J., T. Rehren, S. Firth, and R. J. H. Clark. 2004a. Materials Characterisation 53: 63-70. van der Weerd, J., G. D. Smith, S. Firth, and R. J. H. Clark. 2004b. Journal of Archaeological Science 31:1429-1437. Vandenabeele, P., L. Moens, H. G. M. Edwards, and R. Dams. 2000. Journal of Raman Spectroscopy 31:509-517. Wehlte, K. 1975. The Materials and Techniques of Painting. New York: Van Nostrand Reinhold. Zuo, J., C. Xu, C. Wang, and Z. Yushi. 1999. Journal of Raman Spectroscopy 30:1053-1055.

Paint Media Analysis

Michael R. Schilling Senior Scientist The Getty Conservation Institute

Pigments and organic binding media are the two principle components of paint. Whereas pigments impart color to paint, it is the role of the organic binding media to bind together the grains of pigment and adhere them to the work of art. Synthetic polymers are the binding media of choice for most of today’s commercial paints. Nevertheless, the continued use by contemporary artists of such natural products as egg, milk, animal hides, vegetable oils, plant gums, waxes, and natural resins (which were the only binding media available from antiquity through the end of the nineteenth century [Kühn, 1986]) attests to their durability, versatility, and working properties that artists value. Conservation scientists are often called upon to analyze organic binding media and pigments in painted works of art. Knowledge obtained from the study of artists’ materials and techniques enriches our understanding of the history of art, informs the decisions of conservators who must develop appropriate conservation treatments, and reveals compositional changes in artists’ materials brought about by age, weathering, and environmental factors. Although many instrumental analysis techniques now exist for identifying organic substances, several key factors limit the actual number of techniques that are suitable for identifying organic binding media. To begin, typical samples removed from paintings weigh in the range of 1 to 50 micrograms; in many instances the medium simply may be present below instrumental detection limits. Mixtures of organic binding media may present problems of overlapping signals. Physical aging and pigment interferences may complicate data interpretation by changing the original composition. Moreover, it is extremely difficult, if not impossible, to resolubilize some organic binding media (such as egg tempera) once they have become dried into

186

PAINT MEDIA ANALYSIS 187

paint films. For those instrumental techniques that are capable of detecting organic binding media, application of simple qualitative analysis limits the extent to which the analytical test results may be interpreted. Quantitative gas chromatography-mass spectrometry (GC-MS) is one of the few analytical techniques capable of overcoming the myriad of problems associated with identification of natural organic binding media in painted works of art. In research at the Getty Conservation Institute, quantitative GC-MS procedures were developed for identifying organic binding media based on proteins, oils, and plant gums. The procedures were validated on test paints that were subjected to two types of artificial aging (six weeks at 80°C and 500 hours in a Weather-O- Meter light exposure chamber at 50 percent RH and 50°C). The present study illustrates the utility of quantitative GC-MS in the study of paintings by two prominent American artists.

A TECHNICAL STUDY OF PAINTINGS BY JACOB LAWRENCE Jacob Lawrence was known for his simplified, brilliant graphic forms that depict African American history and experience (Steele, 2000). Throughout his long career he favored working in various water-based organic binding media commonly referred to as tempera: casein, egg, plant gums, and animal glue (Mayer, 1940). It is known that Lawrence mixed some of his own tempera paints from artists’ recipe books, whereas in many of his later works he used commercially available tube colors. Manufacturers often add materials to tube colors, in addition to the binding media, to modify the working properties of the paints and stabilize the mixtures. These additives include glycerol, seed oils (such as linseed, poppy, and walnut), natural resins (dammar, rosin), phthalate plasticizers, and sugar. From these lists it is quite clear that Lawrence’s paint media may be complicated mixtures of many substances (Steele and Halpine, 1993). It should also be noted that it is nearly impossible to differentiate these tempera media based solely on the appearance of the painted surface, yet because this is sometimes the only means available to museum registrars when cataloguing their collections, these records are sometimes erroneous. Recently a technical study of samples from a number of Lawrence’s paintings (see Table 1) was undertaken to learn more about his painting technique, check the accuracy of museum archival records, and contribute to a catalogue raisonné of Lawrence’s paintings (Schilling et al., 2000). Pigments were identified in the paint samples using polarized-light microscopy. Some samples were tested using Fourier transform infrared microspectrometry (FTIR) to identity the paint components. To test for proteinaceous media in paint samples, amino acids were liberated by acid hydrolysis and analyzed by quantitative GC-MS in the form of (tert-butyldimethylsilyl) derivatives (see Appendix A for experimental details) (Simek et al., 1994; Columbini et al., 1998; Schilling and Khanjian, 1996a). The quantitative

188 SCIENTIFIC EXAMINATION OF ART

TABLE 1 Jacob Lawrence Paintings Analyzed in This Study

The Metropolitan Museum of Art, New York Blind Beggars, 1938 National Museum of American Art, Washington, D.C. Painting the Bilges, 1944 New Jersey, 1946 Men Exist for the Sake of One Another, 1958 Library, 1960 Hirshhorn Museum and Sculpture Garden, Washington, D.C. African Gold Miners, 1946 Vaudeville, 1951 The Cue and the Ball, 1956 Magic Man, 1958 Playing Card (Joker) or (King), 1962 Harriet and the Promised Land No.10, 1967 In a Free Government, 1976 Worcester Art Museum The Checker Players, 1947 The Museum of Modern Art, New York Sedation, 1950 Private Collection Struggle Series No.11: Informers Coded Message, 1955 Ordeal of Alice, 1963 National Gallery of Art, Washington, D.C. Street to Mbari, 1964 Daybreak-A Time to Rest, 1967 Merril C. Berman Collection Students with Books, 1966 Jacob and Gwen Knight Lawrence Collection Other Rooms, 1975

results for the alkyl- and imino-substituted amino acids (the so-called “stable” amino acids) were normalized to 100 mole percent. The quantitative yields for the other amino acids are often unreliable due to pigment interferences in the hydrolysis and/or derivatization procedures, or due to aging (Halpine, 1992; Ronca, 1994; Schilling and Khanjian, 1996b); these amino acids were excluded from the final dataset. Samples tested for plant gum media were hydrolyzed in trifluoroacetic acid, and the monosaccharides were analyzed as O-methyloxime acetate derivatives (see Appendix B for details) (Murphy and Pennock, 1972; Neeser and Schweizer, 1983). For comparative purposes the monosaccharide dataset, excluding glucose and fructose, were normalized to 100 weight percent.

PAINT MEDIA ANALYSIS 189

Test Results for Jacob Lawrence Paintings Table 2 lists the quantitative stable amino acid test results for the Lawrence paint samples with those of several common proteinaceous and plant-gum- binding media included for reference (the reference data originated from in- house tests and from published sources [Schilling et al., 1996]). Carbohydrate compositions for selected Lawrence samples and various reference materials are listed in Table 3. Using the method of correlation coefficients, the quantitative stable amino acid composition for each sample was compared to those of the common binding media in order to find the closest match, as listed in Table 4 (Anderson, 1987); the same method was employed for identification of plant gums (see Table 3). Most samples correlated very closely either to glue, egg, casein, or gum arabic. In two paint samples, however, there were indications in the test data that two proteinaceous media were present. This situation may arise either because the artist intentionally mixed two binding media together in the paint, or because the paint sample was contaminated with medium from a second paint layer (this last situation occurs frequently in samples from egg tempera paintings that have ground layers mixed with glue). And so, for the two Lawrence samples, simple algebraic equations were used to find the most likely pair of proteinaceous media that gave the closest correlations to those in the paint samples (Schilling and Khanjian, 1996c). Thus, the red from Blind Beggars had a 0.99 correlation to a mixture (1:2) of casein and glue, whereas the green from Sedation had a 0.95 correlation to a mixture (1:3) of casein and glue. In general, good agreement was evident between the analytical findings and the medium attributions in the museum archives. One notable exception was the detection of glue as the medium of Playing Card, which had been previously misidentified in the archives as plant gum. Another exception was Street to Mbari, which had been assessed visually as having a gouache medium. GC-MS was useful for detecting components in the paints that were unrelated to the protein or plant gum media. Some samples revealed the presence of commercial paint additives, such as Struggle Series Number 11. The brown paint contained egg medium plus high amounts of glycerol, gallic acid, and rosin; this formulation is consistent with an artists’ tube color (Steele, 2000; Steele and Halpine, 1993; Schilling et al., 2000). Moreover, a few samples showed evidence of biodeterioration of the paint medium. For instance, oxalic acid (a common byproduct of microbial activity [Matteini, 1998]) was detected in the dark paint from The Checker Players. It may be that the relatively poor quality of the correlation for the medium in this paint to the reference materials (0.91 to casein, 0.74 to egg) may be due in part to the effects of biodeterioration.

190

ALA VAL ILE LEU GLY PRO HYP % Protein

Concentration (mole percent)

12.4 16.8 11.1 21.8 8.8 28.9 0.2 3

13.5 10.8 7.8 13.7 28.8 20.9 4.6 4

16.1 14.9 11.3 22.1 13.1 22.3 0.2 7

24.3 19.7 14.2 23.3 15.9 2.3 0.3 9 21.5 16.3 13.5 24.0 16.4 8.3 0.0 3

14.0 13.1 9.0 16.8 25.6 21.0 0.5 5

23.1 18.3 13.2 22.4 15.4 6.7 0.9 6

20.6 18.3 16.4 25.7 16.3 2.7 0.0 2

14.9 3.5 2.1 4.3 51.6 13.1 10.4 12 14.2 2.7 1.6 3.2 55.8 12.3 10.2 18

22.2 16.3 12.3 21.9 17.2 10.2 0.0 4

22.4 20.8 15.7 25.8 9.5 5.3 0.5 6

27.2 19.5 14.1 21.7 13.3 3.7 0.4 2 20.5 16.1 12.9 22.7 21.3 6.4 0.1 9

22.2 16.2 12.4 24.1 14.4 10.8 0.0 12

24.0 17.0 12.9 26.1 17.9 2.1 0.0 6

17.1 5.5 3.6 6.9 52.6 10.8 3.5 11

18.7 3.7 2.0 3.5 52.4 12.6 7.2 14 17.8 3.1 1.7 3.2 51.2 13.2 9.7 13

25.0 17.8 13.1 23.5 10.4 8.7 1.5 16

22.0 15.1 12.2 22.3 20.6 7.8 0.0 8

Red Paint & ground Yellow Ochre Brown Light black Blue & ground Brown Blue Blue & paper Red & fibers Black & paper Green & paper Yellow & blue Blue Brown Yellow Brown & ground Yellow Blue

Ground

Sample

Blind Beggars Checker Players Daybreak Library

Magic Man Men Exist Ordeal of Alice Parade Playing Card (Joker) Playing Card (King)

Sedation Street to M’Bari

Struggle Series #11 Students and Books

Vaudeville

TABLE 2 Stable Amino Acid Compositions of Jacob Lawrence Paint Samples

Painting

PAINT MEDIA ANALYSIS 191

A STUDY OF WILLEM DE KOONING’S PAINTINGS FROM THE 1960S AND 1970S Willem de Kooning was born in the Netherlands in 1904, and immigrated to the United States in 1926. He trained in guild and craft traditions of wood graining, gilding, marbleizing, lettering, and sign painting; he also spent time as a house painter and a commercial artist. These experiences gave him a thorough mastery of materials and a craftsman’s skills (Lake et al., 1999). De Kooning routinely exploited unconventional materials for his pictures. Historical and anecdotal records report that he mixed house paint, safflower cooking oil, water, egg, and even mayonnaise into his artists’ paints to achieve the desired appearance and texture. During the course of creating a painting, he scraped the painted canvas at day’s end, and repeatedly reworked it; for this process to be successful, soft, slow-drying paints were required. Such paints were abundant in the 1940s and 1950s, when oil was the typical medium in house paints. Unfortunately for de Kooning, alkyd paint formulations became popular in the retail trade industry in the 1960s and 1970s; their fast-drying properties were incompatible with his chosen technique. In paintings from this period, sources document his use of Bellini Bocour artists’ tube colors, which contained heat-bodied linseed oil, to which he occasionally added safflower oil and water (Lake et al., 1999). Although his methods and materials have been well documented, it was not clear what his actual practices were at specific times in his career. Moreover, there was concern that his unusual paint formulations could negatively affect the long- term stability of his paintings. The paintings executed during the 1960s and 1970s, in particular, are problematic for conservators, with passages that remain soft and sticky. Such paint surfaces are easily deformed when touched and they readily pick up surface dust. To learn more about de Kooning’s materials and techniques, a study was undertaken to analyze the binding media and pigments of a selection of his paintings from the period of 1960-1977 (Lake et al., 1999). Table 5 provides a complete list of the paintings that were sampled.

Chemistry of Oil Paints The chemistry of oil paint is very complex, and even with modern analytical equipment, it is difficult to understand the precise details of the interactions between the polymerized oil media and pigments. Nonetheless, a substantial body of knowledge has been developed that sheds some light on the drying and subsequent aging of oil paints (van den Berg, 2002). Essentially, seed oils differ in terms of their fatty acid distribution on the triglyceride molecules. The so-called drying oils that are favored by artists (e.g., linseed, walnut, poppy seed) have a high proportion of multiple unsaturated fatty acids, whereas the semidrying and non- drying oils (such as castor, safflower, sunflower) do not. All of the aforemen-

192 SCIENTIFIC EXAMINATION OF ART

TABLE 3 Gum Sugar Compositions of Jacob Lawrence Paint Samples and Plant Gum Standards, with Correlation Coefficient Data

Normalized Weight % of Gum Sugars Weight % Sample Gum Sugars Rhamnose Fucose Arabinose

African Gold Miners: blue paint 10.4 13.0 0.1 36.3 African Gold Miners: black paint 6.3 10.2 0.1 29.1 Blind Beggars: blue paint 3.8 17.0 0.0 35.5 New Jersey: red paint 11.8 12.9 0.1 32.8 Painting the Bilges: blue paint 1.9 15.4 0.0 37.7 Gum Arabic standard 57.0 16.0 0.0 36.0 Cherry gum standard 95.7 0.6 0.0 46.6 Gum tragacanth standard 46.4 3.3 8.8 48.4

tioned seed oils contain approximately 10 percent by weight of glycerol, plus small amounts of saturated fatty acids (the two most important being hexade- canoic acid and octadecanoic acid, more commonly known as palmitic acid and stearic acid, respectively) (Mills, 1966). As oil paints dry, unsaturated fatty acids react with oxygen to form a polymerized oil matrix; triglycerides and diglycerides provide additional cross-links to the polymerized oil matrix via their glycerol backbones. The saturated fatty acids, being less reactive, neither oxidize nor cross-link, and so remain as marker compounds in oil paint. During the drying and aging processes, chain scission products such as dicarboxylic fatty acids are formed in oil paint. The most important dicarboxylic fatty acid marker compound is nonanedioic acid (azelaic acid), but other straight-chain dicarboxylic fatty acids (that contain from two to ten carbon atoms) also form (Mills, 1966). Other reactions also occur during aging that further alter the fatty acid distribution of oil paints. For example, free fatty acids are produced by hydrolysis of glycerides (e.g., glycerol esters of fatty acids). As hydrolysis progresses there is a reduction in residual triglycerides and diglycerides, and formation of free fatty acids. Complete hydrolysis of an oil paint would eventually yield three moles of free fatty acids per every mole of glycerol (van den Berg, 2002). Reaction of the carboxylate groups of free fatty acids with pigments that contain coordinating metal cations (e.g., lead, copper, cobalt) produces metal soaps. Evidence has shown that aged oil paints are essentially ionomeric polymers of fatty acids coordinated with pigments (van den Berg et al., 1999). Figure 1 illustrates the results from a Monte Carlo simulation of the hydrolysis of a model triglyceride, in which the extent of hydrolysis is estimated by the

PAINT MEDIA ANALYSIS 193

Correlation Coefficient to Plant Gum Standards

Xylose Mannose Galactose Gum Arabic Cherry Gum Gum Tragacanth

0.7 0.4 49.4 1.00 0.90 0.39 0.7 0.5 59.3 0.97 0.84 0.24 0.5 0.5 46.5 1.00 0.88 0.37 0.5 0.2 53.6 0.99 0.87 0.31 0.6 0.0 46.3 1.00 0.90 0.42 0.0 0.0 48.0 10.4 2.2 40.2 27.5 0.0 12.1

percent of free fatty acids. This figure shows that nontrivial amounts of triglycerides and diglycerides remain even after significant hydrolysis of the model compound. Undoubtedly this model is extremely simple, and does not account for differing rates of hydrolysis of the middle ester group on a glyceride compared to the end positions. Nonetheless, it does suggest that residual triglycerides and diglycerides may remain even in extremely hydrolyzed oil paints, which can additionally stabilize oil paint ionomeric polymers. Additional alteration of the fatty acid composition of oil paints occurs by evaporation of free fatty acids. So-called ”ghost images” that develop on the glass next to framed and glazed oil paintings provide clear evidence of fatty acid evaporation (Williams, 1989). Thermogravimetric analysis indicated that palmitic acid evaporated approximately twice as rapidly as stearic acid or azelaic acid (Schilling et al., 1999). Oil paints made with lead white pigment produced no visible ghost image, which is consistent with the fact that lead pigments readily coordinate free fatty acids.

Analysis of Oil Paints Two GC-MS procedures were employed to analyze the de Kooning paint samples. A quantitative procedure for fatty acid and glycerol analysis of food oils (Mason et al., 1964) was modified to work on oil paint samples (Schilling and Khanjian, 1996d). In this procedure FAMEs and isopropylidene glycerol (IPG, a volatile glycerol derivative) were produced quantitatively by overnight treatment with sodium methoxide in 2,2-dimethoxypropane, followed by addition of methanolic hydrochloric acid. The FAMEs and IPG were separated in a single

194

(1 part glue, 2 parts casein)

(yolk & ultramarine, light aged) (yolk, aged at 80 deg C) (Rowney cadmium red)

(yolk, light aged) (yolk & lead white, light aged)

(1 part glue, 3 parts casein)

(yolk, light aged) (Rowney cadmium red)

0.99

0.99 0.96 1.00

0.96 0.96

0.95

0.96 0.99

0.39

0.36

0.47 0.36 0.26

0.27 0.28 0.19

0.25 0.28

0.91

1.00

– – –

– –

0.13

0.02

0.06 0.07

0.03

0.98

1.00 0.95 0.95

– –

–

0.82 0.29 0.33

0.61 0.55 0.56

0.13 0.90 0.19

0.06 0.79 0.06

0.40 0.91 0.32 0.10 0.90 0.12 0.03 0.17 0.89 0.14 0.19 0.90 0.21

0.09 0.28 0.08 0.20 0.92 0.25 0.01 0.90 0.00

0.21 0.55

0.99 0.99 0.96

0.98 0.99

Collagen Egg Yolk Casein Other

Correlation Coefficient

–

– – – –

– – –

–

Red Paint & ground Yellow Ochre Brown Light black Blue & ground Brown Blue Blue & paper Red & fibers Black & paper Green & paper Yellow & blue Blue Brown Yellow Brown & ground Yellow Blue

Ground

Paint Sample

Bari

TABLE 4 Correlation Between Jacob Lawrence Paint Sample Data (Table 2) and Reference Material Data (Table 8)

Painting

Blind Beggars Checker Players Daybreak Library

Magic Man Men Exist Ordeal of Alice Parade Playing Card (Joker) Playing Card (King)

Sedation Street to M’

Struggle Series #11 Students and Books

Vaudeville

PAINT MEDIA ANALYSIS 195

TABLE 5 Willem de Kooning Paintings Analyzed in This Study

Door to the River, 1960, Whitney Museum of American Art, New York Spike’s Folly II, 1960, Robert and Jane Meyerhoff, Phoenix, MD Rosy-Fingered Dawn at Louse Point, 1963, Stedelijk Museum, Amsterdam Pastorale, 1963, private collection, New Orleans Woman, Sag Harbor, 1964, Hirshhorn Museum and Sculpture Garden, Washington, D.C. Woman, 1965, Hirshhorn Museum and Sculpture Garden, Washington, D.C. The Visit, 1966-67, Tate Gallery, London Two Figures in a Landscape, 1967, Stedelijk Museum, Amsterdam Amityville, 1971, private collection . . . Whose Name Was Writ in Water, 1975, Solomon R. Guggenheim Museum, New York Untitled I, 1977, Adriana and Robert Mnuchin Untitled V, 1977, Albright-Knox Art Gallery, Buffalo, New York

chromatogram using an HP-INNOWAX capillary column. This procedure derivatizes all three forms of fatty acids in oil paints (free fatty acids, pigment soaps, and glycerides). The second quantitative procedure used hexamethylene disilazane with trichloromethyl silane catalyst (HMDS/TMCS) in pyridine to form trimethyl silyl esters of the free fatty acids and soaps, and separation of the derivatives on a DB-5MS column (Pierce, 1968). In evaluating the quantitative test data several key parameters are calculated for identification and diagnostic purposes. For instance, the minimum content of oil in the paint sample can be approximated from the glycerol content, although the accuracy of this estimation is limited by loss of glycerol due to aging and solvent extraction. Second, the molar ratio of palmitic acid to stearic acid (P/S) is useful for identifying the type of oil present (Mills, 1966). Third, the extent of hydrolysis may be estimated by comparing the content of free azelaic acid and its soap to the total azelaic acid content. To probe the extent of alteration of the dried oil matrix, the two most diagnostic ratios are palmitic to glycerol (P/G), and the ratio of dicarboxylic fatty acids to glycerol (D/G, where D is the sum of all dicar-

boxylic fatty acids from C3 to C8 plus C10). Reduction in P/G from its original value in the fresh oil is caused by loss of palmitic acid due to evaporation or migration into the canvas and ground layer. Photo-oxidative reactions are responsible for increases in D/G (Schilling and Khanjian, 1996d; Schilling et al., 1997).

196 SCIENTIFIC EXAMINATION OF ART

FIGURE 1 Monte Carlo simulation of the hydrolysis of a model triglyceride. NOTE: 10,000 triglyceride molecules and 100,000 iterations were used in this simulation, which was run using an Excel macro program.

Test Results for de Kooning Paintings Table 6 lists the analytical results for the de Kooning paint samples. From an examination of the test results a few generalizations can be made. The medium in the paints with the highest P/S ratios was identified either as poppy oil, or poppy oil mixed with linseed or castor oil. These paints tended to have the lowest values for oil content, extent of hydrolysis, and content of dicarboxylic fatty acid degradation products (as measured by D/G). In contrast, the medium in the paints with the lowest P/S ratios was identified either as linseed oil, or linseed oil mixed with castor oil; it is likely that de Kooning used unmixed tube colors in these paints. They had the highest values for oil content, extent of hydrolysis, and D/G ratios. Paints with intermediate P/S ratios contained linseed oil and safflower oil mixtures, and differed little from the high P/S paints in terms of hydrolysis and D/G. Void spaces inside paint samples indicate that de Kooning mixed water into them, and the test results show that this procedure had no deleterious effect on the extent of hydrolysis. Another finding was that the sticky paints tended to contain cadmium pigments or synthetic organic dyestuffs, whereas no clear relationship to the drying rate of the oil medium was apparent. The sticky paints exhibited the highest degree of hydrolysis, as measured by the free fatty acid

197

67 Linseed 40 Linseed

50 Castor/linseed 18 Poppy/linseed

24 Linseed/safflower 31 Linseed/safflower

33 Safflower

29 Linseed/safflower

23 Poppy/castor 48 Poppy

24 Poppy/castor

25 Poppy/linseed 29 Poppy 52 Castor

59 Safflower 23 Poppy/linseed

59 Linseed

67 Safflower

48 Castor

52 Linseed/safflower

31 Linseed/safflower

38 Poppy/linseed

P/S P/G D/G Hydrolyzed Oil Oil(s)

3.5 0.26 0.32 32 4.1 0.26 0.36 19 0.9 0.04 0.66 40 3.7 0.24 0.34 12 4.8 0.2 0.36 32 1.9 0.07 0.57 48

2.6 0.15 0.23 34

2.7 0.16 0.3 32 2.8 0.09 0.35 22 3.9 0.21 0.28 28

2.6 0.09 0.28 25

3.8 0.13 0.34 26

1.4 0.04 0.34 37 1.8 0.07 0.31 17

2.0 0.07 0.48 37 59 Linseed

0.91 0.03 0.72 66

2.6 0.12 0.5 33

2.7 0.14 0.26 26

1.3 0.05 0.36 23

3.4 0.1 0.23 29 2.6 0.14 0.65 17

3.7 0.24 0.25 15

2.1 0.17 0.32 29

White, voids, matte, brittle

Blue, brittle, matte Light blue, voids, brittle

Wrinkled gray-green, soft

Pink, wrinkled, soft Amber gel, wrinkled flesh

White, soft

Red, soft, sticky Maroon, soft, stringy

Pink-white, wrinkled

Black, wrinkled

Dark blue, wrinkled, soft

Yellow, soft, matte

Amber drip

Yellow, wrinkled, soft

Amber drip, sticky

White, wrinkled

Sample Description

Pink, voids, matte

Bright yellow, soft Pink, soft

Orange, soft

Orange, soft, soft

s..., 1960

TABLE 6 Test Results for Willem de Kooning Paint Samples

Painting, Date

Door..., 1960

Spike’

Pastorale, 1963 Rosy..., 1963 Woman..., 1964 Woman, 1965 Visit, 1966

Two..., 1967

Amityville, 1971

Whose..., 1975

Untitled I, 1977

Untitled V, 1977 Gray, wrinkled, voids

198 SCIENTIFIC EXAMINATION OF ART

content. In conclusion, the results supported the anecdotal evidence that de Kooning did occasionally add semidrying oils to his paints, which would have retarded their drying rate.

CONCLUSIONS Quantitative GC-MS is an important analytical technique for characterizing natural products that have been used by artists as organic binding media. In the study of modern paintings the technique provides valuable information that enhances our understanding of artists’ materials and techniques, permits changes in material composition to be monitored, and contributes to the development of appropriate conservation strategies.

ACKNOWLEDGEMENTS The following colleagues from the Getty Conservation Institute made invaluable contributions to the analytical research presented in this paper: Herant Khanjian, Joy Keeney, David Carson, Narayan Khandekar, Andrew Parker, and Luiz Souza. Jim Druzik was the creative influence behind the Monte Carlo simulation study. I am especially grateful to Dusan Stulik, who developed the organic binding medium research project and who was its director for many years. Our principal collaborator in the Jacob Lawrence study was Elizabeth Steele, paintings conservator at the Phillips Collection, whose extensive knowledge of Lawrence’s technique and materials greatly enhanced the interpretation of our research. I am also thankful for the enthusiastic support of Peter Nesbett, director of the Jacob Lawrence Catalogue Raisonné project, and Michelle DuBois, associate director. In the Willem de Kooning study we collaborated with Susan Lake, chief conservator at the Hirshhorn Museum and Sculpture Garden, who contributed in count- less ways to the success of the research. Suzanne Quillen-Lomax, organic chemist at the National Gallery of Art, Washington, D.C., studied a number of de Kooning paintings and was an important partner in the collaboration.

REFERENCES

Anderson, R. L. 1987. In Practical Statistics for Analytical Chemists. New York: Van Nostrand Reinhold. Columbini, M. P., R. Fuoco, A. Giacomelli, and B. Muscatello. 1998. Characterization of proteinaceous binders in wall painting samples by microwave-assisted acid hydrolysis and GC-MS determination of amino acids. Studies in Conservation 43:33-41. Halpine, S. 1992. Amino acid analysis of proteinaceous media from Cosimo Tura’s The Annunciation with Saint Francis and Saint Louis of Toulouse. Studies in Conservation 37:22-38. Kühn, H. 1986. In Conservation and Restoration of Works of Art and Antiquities, vol. 1, pp. 157-167. London: Butterworths.

PAINT MEDIA ANALYSIS 199

Lake, S., S. Lomax, and M. Schilling. 1999. A technical investigation of Willem de Kooning’s paintings from the 1960s and 1970s. In ICOM Committee for Conservation Preprints, 12th Triennial Meeting, Lyon, France 29 August-3 September 1999, ed. J. Bridgland, pp. 381-385. London: James and James. Mason, M. E., M. E. Eager, and G. R. Waller. 1964. A procedure for the simultaneous quantitative determination of glycerol and fatty acid contents of fats and oils. Analytical Chemistry 36(3):597- 590. Matteini, M. 1998. Different integrated analytical methods for the study of the pictorial techniques in the Vasari and Zuccari wall paintings of Florence Cathedral: Comparison and discussion. Sci- ence and Technology for Cultural Heritage 7(1):83-94. Mayer, R. 1940. In The Artists Handbook of Materials and Techniques, p. 223. New York: Viking Press. Mills, J. S. 1966. The gas chromatographic examination of paint media. Part I. Fatty acid composition and identification of dried oil films. Studies in Conservation 11:92-107. Murphy, D., and C. A. Pennock. 1972. Gas chromatographic measurement of blood and urine glucose and other monosaccharides. Clinica Chimica Acta 42:67-75. Neeser, J. R., and T. F. Schweizer. 1983. A quantitative determination by capillary gas-liquid chromatography of neutral and amino sugar (as O-methyloxime acetates). Analytical Biochemistry 142:58-67. Pierce, A. E. 1968. In Silylation of Organic Compounds, pp. 160-162. Rockford, Ill.: Pierce Chemical Co. Ronca, F. 1994. Protein determination in polychromed stone sculptures, stuccoes and gesso grounds. Studies in Conservation 39:107-120. Schilling, M., and H. Khanjian. 1996a. Gas chromatographic investigations of organic materials in art objects: New insights from a traditional technique. In Innovation et Technologie au Service du Patrimoine de l’Humanite, pp. 137-143. Paris: UNESCO/Admitech. Schilling, M., and H. Khanjian. 1996b. Gas chromatographic analysis of amino acids as ethyl chloroformate derivatives. II. Effects of pigments and accelerated aging on the identification of proteinaceous binding media. Journal of the American Institute of Conservation 35:123-144. Schilling, M., and H. Khanjian. 1996c. Gas chromatographic analysis of amino acids as ethyl chloroformate derivatives. III. Identification of proteinaceous binding media by interpretation of amino acid composition data. In ICOM Committee for Conservation Preprints, 11th Triennial Meeting, Edinburgh, Scotland 1-6 September 1996, ed. J. Bridgland, pp. 220-227. London: James and James. Schilling, M., and H. Khanjian. 1996d. Gas chromatographic determination of the fatty acid and glycerol content of lipids. I. The effects of pigments and aging on the composition of oil paints. In ICOM Committee for Conservation Preprints, 11th Triennial Meeting, Edinburgh, Scotland 1-6 September 1996, ed. J. Bridgland, pp. 220-227. London: James and James. Schilling, M., D. Carson, and H. Khanjian. 1999. Gas chromatographic determination of the fatty acid and glycerol content of lipids. IV. Evaporation of fatty acids and the formation of ghost images by framed oil paintings. In ICOM Committee for Conservation Preprints, 12th Triennial Meeting, Lyon, France 29 August-3 September 1999, ed. J. Bridgland, pp. 242-247. London: James and James. Schilling, M., H. Khanjian, and D. Carson. 1997. Fatty acid and glycerol content of lipids; effects of ageing and solvent extraction on the composition of oil paints. Techné 5:71-78. Schilling, M., H. Khanjian, and L. Souza. 1996. Gas chromatographic analysis of amino acids as ethyl chloroformate derivatives. I. Composition of proteins associated with objects of art and monuments. Journal of the American Institute of Conservation 35:45-59. Schilling, M., N. Khandekar, J. Keeney, and H. Khanjian. 2000. Identification of binding media and pigments in the paintings of Jacob Lawrence. In Over the Line: The Art and Life of Jacob Lawrence, eds. P. T. Nesbett and M. DuBois, pp. 266-269. Seattle: University of Washington Press.

200 SCIENTIFIC EXAMINATION OF ART

Simek, P., A. Heydová, and A. Jegorov. 1994. High resolution capillary gas chromatography and gas chromatography-mass spectrometry of protein and non-protein amino acids, amino alcohols, and hydroxycarboxylic acids as their tert-butyldimethylsilyl derivatives. Journal of High Resolu- tion Chromatography 17:145-152. Steele, E. 2000. The materials and techniques of Jacob Lawrence. In Over the Line: The Art and Life of Jacob Lawrence, eds. P. T. Nesbett and M. DuBois, pp. 247-265. Seattle: University of Washing- ton Press. Steele, E., and S. M. Halpine. 1993. Precision and spontaneity: Jacob Lawrence’s materials and techniques. In Jacob Lawrence: The Migration Series. Washington, D.C.: Rappahannock Press in association with the Phillips Collection. Van den Berg, J. D. J. 2002. In Analytical Chemical Studies on Traditional Linseed Oil Paints. Ph.D. thesis, pp. 45-52. University of Amsterdam. Van den Berg, J. D. J., K. J. van den Berg, and J. Boon. 1999. Chemical change in curing and aging oil paints. In ICOM Committee for Conservation Preprints, 12th Triennial Meeting, Lyon, France 29 August-3 September 1999, ed. J. Bridgland, pp. 248-253. London: James and James. Williams, S. R. 1989. Blooms, blushes, transferred images and mouldy surfaces: What are these dis- tracting accretions on art works? In Proceedings of the Fourteenth Annual IIC-CG Conference, pp. 65-84. Ottawa: IIC-Canadian Group.

APPENDIX A PROCEDURE FOR QUANTITATIVE GC-MS ANALYSIS OF AMINO ACIDS, FATTY ACIDS, AND GLYCEROL AS (TERT -BUTYL-DIMETHYLSILYL) DERIVATIVES

All analytical standards were obtained from Aldrich Chemical Company. Weigh paint sample on an ultra-microbalance and then transfer to a 0.1 ml conical vial. Add norleucine internal standard to give a final concentration of 20 ppm in the final injection volume. Add 100 µl of 6.0N hydrochloric acid (Pierce, sequanal- grade) to the sample vial and close the vial with a screw-top lid and PTFE septum. Heat the vial at 105°C for 24 hours in an oven. Remove vial from oven, let stand until cool, and centrifuge. Evaporate the contents to dryness under a stream of nitrogen gas while warming the vial to 60°C. Add 40 µl of HPLC-grade water (VWR Scientific), replace lid, stir on a vortex mixer, centrifuge, and evaporate the contents to dryness. Add 40 µl of absolute ethanol (Spectrum Chemical), replace lid, stir with a vortex mixer, centrifuge, and evaporate the contents to dryness. Prepare a solution of 40 mg of pyridine hydrochloride (Aldrich) to 1 ml of silylation-grade pyridine (Pierce Chemical). The silylating reagent consists of 300 µl of 99 percent MTBSTFA/1 percent TBDMCS mixture (Pierce Chemical) in 700 µl of pyridine hydrochloride solution. Add silylating reagent to the vial and replace lid (Note: Use 1 µl of reagent per 2 µg of sample for typical paint samples; use 300 µl per 50 to 100 µg of pure

PAINT MEDIA ANALYSIS 201

proteinaceous reference materials. Use a minimum of 20 µl of reagent if an oven is used for heating and 50 µl if a heating block is used). Stir with a vortex mixer, warm the vial at 60°C for 30 minutes on a hotplate, and heat in an oven at 105°C for five hours. Remove vial from heat, let stand until cool, centrifuge, and transfer solution to an injection vial. GC-MS Conditions for 30 M × 0.25 mm × 1 µm DB-5MS Column: Helium carrier set to a linear velocity of 45 cm/sec; splitless injector at 300°C with a 60 sec purge off time; MS transfer line set to 300°C. GC oven temperature program: 80°C for one minute; 75°C/m to 180°C; 10°C/m to 320°C; isothermal for three minutes; solvent delay of seven minutes. MSD source temperature is approximately 200°C. Figure 2 shows the GC-MS result for a standard mixture of amino acids, fatty acids, and glycerol. Calibration parameters: See Table 7 for the list of quantitation ions for the TBDMS derivatives. Using a quadratic curve fit forcing through the origin gives correlation coefficients of 0.995 or better for most analytes over the calibration range of 2 to 50 ppm. Stable amino acid compositions for various reference materials are listed in Table 8.

FIGURE 2 GC-MS analysis of (tert-butyl dimethylsilyl) derivatives of amino acid, fatty acid, and glycerol standards.

202 SCIENTIFIC EXAMINATION OF ART

TABLE 7 Quantitation Ions for TBDMS Derivatives

Analyte m/z Analyte m/z

Norleucine 200.1 Glycerol 189.1 Alanine 260.1 Decanoic acid 229.1 Glycine 246.1 Lauric acid 257.1 Valine 186.1 Myristic acid 285.1 Leucine 200.1 Pentadecanoic acid 299.2 Isoleucine 200.1 Palmitic acid 313.2 Proline 184.1 Heptadecanoic acid 327.2 Methionine 218.1 Oleic acid 339.2 Serine 390.2 Stearic acid 341.2 Threonine 303.2 Nonadecanoic acid 355.2 Phenylalanine 234.1 Eicosanoic acid 369.2 Aspartic acid 302.2 Oxalic acid 261.1 Hydroxyproline 416.2 Malonic acid 115.1 Glutamic acid 432.2 Succinic acid 289.1 Lysine 488.2 Glutaric acid 303.2 Arginine 460.2 Adipic acid 317.2 Histidine 196.1 Pimelic acid 331.2 Tyrosine 302.2 Subaric acid 345.2 Azelaic acid 359.2 Sebacic acid 373.2

APPENDIX B PROCEDURE FOR QUANTITATIVE GC-MS ANALYSIS OF SUGARS IN PLANT GUMS AS O-METHYLOXIME ACETATE DERIVATIVES All analytical standards were obtained from Aldrich Chemical Company. Weigh sample on an ultra-microbalance, transfer to a conical reaction vial, and add allose internal standard to give a final concentration of 20 ppm in the injection volume. Add 100 µl of 1.2N trifluoroacetic acid (Pierce Chemical), purge vial with nitrogen for 30 seconds, and cap. Heat the vial for one hour at 125°C, remove from heat, let stand until cool, and centrifuge. Transfer contents to a 2 ml autosampler vial, rinse conical vial with 40 µl of water, and combine in the 2 ml vial. Evaporate the contents to dryness using a nitrogen stream while warming the vial to 50°C. Rinse with 40 µl of absolute ethanol (Spectrum Chemical), and evaporate to dryness. Add 80 µl of a 1 percent solution of methoxyamine hydrochloride (Sigma) in pyridine (Pierce Chemical), replace cap, and heat for 10 minutes at 70°C. Remove from heat, let stand until cool, and add 40 µl of acetic anhydride (Supelco). Replace the cap, and heat the vial for 10 minutes at 70°C. Remove vial from heat, let stand until cool, and centrifuge.

203

e hypothetical mixtures

sted for this study).

Concentration (mole percent)

ALA VAL ILE LEU GLY PRO HYP

15.7 3.0 1.8 3.7 46.7 16.7 12.4

23.5 15.7 13.1 22.3 15.8 9.5 0.0

23.6 18.7 14.4 24.5 16.1 2.7 0.0

23.3 16.0 14.0 21.0 18.5 7.1 0.0 13.3 10.1 7.3 13.0 27.3 22.9 6.1 12.7 11.6 8.5 15.0 23.2 24.2 4.8

C 21.0 19.2 14.6 21.9 13.6 9.8 0.0

TABLE 8 Stable Amino Acid Compositions of Various Reference Materials

Reference Material

Casein (mean) 10.9 16.9 12.8 22.0 8.6 28.8 0.0 Collagen & Gelatine (mean) Egg Yolk (mean) 20.9 16.6 13.3 23.2 14.1 11.9 0.0 Yolk & ultramarine, light aged Yolk, aged at 80 Rowney egg tempera with cadmium red Yolk, light aged 23.5 16.0 13.4 22.7 15.5 8.9 0.0 Yolk & lead white, light aged 1 part glue, 2 parts casein 1 part glue, 3 parts casein

NOTE: All data taken from (M. Schilling, H. Khanjian, and L. Souza, 1996), except for the mixtures of glue and casein (which ar calculated from the equations in M. Schilling and H. Khanjian, 1996), and the Rowney egg tempera with cadmium red (which was te

204 SCIENTIFIC EXAMINATION OF ART

Evaporate the contents using a nitrogen stream while warming the vial to 50°C. Reconstitute the contents in 100 µl chloroform (Spectrum Chemical), and transfer into a clean 2 ml vial. Rinse the first vial with 100 µl chloroform, and combine. Evaporate chloroform to about 50 µl under a nitrogen stream while warming the vial to 50°C. Transfer the chloroform to a conical glass insert. Evapo- rate to dryness and reconstitute to desired final volume. Reconstitute the contents in chloroform (use an amount equivalent to 1 µg of gum per 5 µl), and inject into GC-MS. GC-MS Conditions for a 15 M × 0.25 mm × 0.25 µm DB-WAX Capillary Column: Helium carrier set to a linear velocity of 60 cm/sec; splitless injector at 240°C with a 60 sec purge off time; MS transfer line set to 240°C. GC oven temperature program: 105°C for one minute; 30°C/m to 180°C; 5°C/m to 240°C; isothermal for two minutes; solvent delay of seven minutes. MSD source temperature is approximately 200°C. Figure 3 shows the GC-MS result for a standard mixture of carbohydrates.

FIGURE 3 GC-MS analysis of O-methyloxime acetate derivatives of carbohydrate standards.

PAINT MEDIA ANALYSIS 205

Calibration parameters: See Table 9 for the list of quantitation ions for the MOA derivatives. Using a linear curve fit forcing through the origin gives correlation coefficients of 0.995 or better for most analytes over the calibration range of 2 to 50 ppm.

TABLE 9 Quantitation Ions for MOA Derivatives

Analyte m/z Analyte m/z

Allose 115.1 Xylose 115.1 Rhamnose 129.2 Mannose 131.2 Fucose 129.2 Fructose 203.2 Ribose 115.1 Glucose 89.2 Arabinose 115.1 Galactose 131.2

A Review of Some Recent Research on Early Chinese Jades

Janet G. Douglas Department of Conservation and Scientific Research Freer Gallery of Art/Arthur M. Sackler Gallery Washington, D.C.

ABSTRACT Chinese jades produced in the earliest periods of China, during the Neolithic period (5000 to 1700 BCE) to the Han dynasty (206 BCE to 220 CE), were typically fashioned by abrasive techniques using fine mineral powders without the advantage of metal tools. Most of these jades are composed of nephrite, a fine-grained variety of the tremolite-actinolite series of amphiboles, although other stone materials were used as well. The study of early Chinese jades using scientific techniques is a relatively narrow field aimed at developing the cultural and archaeological contexts of these materials. The primary areas of investigation include mineralogical identification, geological source of jade, early jade working methods, detection of heating in jade, burial alteration ,and surface accretions. Research in this field is particularly exciting given the large number of excavations in China during the past few decades.

INTRODUCTION In early China most jade manufacturing involved abrasion, using fine mineral powders, without the advantage of metal tools. The material of choice was nephrite, a fine-grained variety of the tremolite-actinolite series of amphiboles, although other stone materials were used as well. Nephrite is a calcium magnesium hydroxyl silicate that occurs in a massive form consisting of interlocking fibrous crystals (Hurlbut and Switzer, 1979). Another jade material, jadeite, was not

206

A REVIEW OF SOME RECENT RESEARCH ON EARLY CHINESE JADES 207

known in China until the eighteenth century, when it was imported from Myanmar (Burma) for working by Chinese artisans. Scientific study of Chinese jades produced in the earliest periods of China, during the Neolithic period (5000 to 1700 BCE) to the Han dynasty (206 BCE to 220 CE), is leading to a richer understanding of these early cultures and their use of jade. Study of well-documented, preferably excavated Chinese jades is helping to place into context those jades of uncertain origin and address issues of authentication.

MINERALOGICAL IDENTIFICATION During the last few decades, analysis of early Chinese jades has focused on the identification of the mineral content of jade materials. In China over 500 excavated jades from a wide variety of sites dating from the Neolithic period to the Han dynasty have been analyzed for their mineral content at the Chinese Acad- emy of Geological Sciences (Wen, 1996, 1997, 1998; Wen and Jing, 1992). Many of the over 800 early jades at the Freer and Sackler galleries have been analyzed for mineral content, thus making this collection one of the most extensively studied in the West. In the last decade minimally invasive analytical methods such as X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FTIR) have been routinely used for identification. Most of these early Chinese jades have been found to be composed of nephrite, a fine grained, massive variety of tremolite-actinolite. Other materials identified include serpentine, marble, olivine, and corundum. Examples of FTIR spectra from three of these jades are given in Figure 1. Similar findings appear throughout a wide range of Chinese archaeological reports. In addition to the study of individual jades, some composite works have been studied in detail, such as the Freer Gallery’s jade and gold pectoral from the Jincun site in Henan province, dating to about the third century BCE (Douglas and Chase, 2001). The pectoral consists of 10 jades attached to a gold chain, and was examined to determine whether its configuration was correct. The jades were found to be similar in material and workmanship and consistent with other jades from the site. The pectoral, however, was found to be a pastiche where the jades were attached to the gold chain with modern gold wires and cut links from the chain.

GEOLOGICAL SOURCE OF JADE Both nephrite and jadeite are known to occur in geological environments through metasomatic processes in a variety of worldwide localities (Harlow and Sorensen, 2001). Two major types of geologic occurrences of Chinese nephrite are known: nephrite associated with metamorphosed dolomitic marbles and nephrite associated with serpentinites.

208 SCIENTIFIC EXAMINATION OF ART

FIGURE 1 FTIR spectra of some early Chinese jades.

The geological sources of nephrite used by early cultures in China are not currently known. Such sources may have been depleted in antiquity, as nephrite can occur in small localized deposits. Research involving scientific methods on early Chinese jades has been addressing issues related to the geological origin of nephrite in early China, as well as jade production and use (Douglas, 2003). Analysis of early Chinese jades at the Freer and Sackler galleries using X-ray fluorescence spectroscopy (XRF) indicates that the geological sources of the material used for these jades are most likely associated with dolomitic marbles. The

145 jades analyzed by XRF were found to be consistently low in Cr2O3 (< 0.08 percent by wt.) and NiO (< 0.01 percent by wt.), characteristic of nephrites associated with dolomitic marbles. Future work on geological sourcing of nephrite should concentrate on these types of deposits in China. Of particular interest are the FeO and MnO contents, which have been determined by XRF to point to simple source patterns for the nephrite used by the Neolithic cultures of Hongshan, Liangzhu, and Longshan, possibly involving one or more related geologic sources for each culture. Longshan jades were found to have unusually high FeO (0.35-17.95 percent by wt.) and MnO (0.02-0.89 percent by wt.) contents, suggesting a source particularly rich in iron and manganese. Jades of the Shang and Western Zhou dynasties show a wide range of compositions, suggesting multiple nephrite sources for these objects. XRF is a simple noninvasive tool for determining minor elemental oxide concentrations, but further work on jade sources will need to involve an expanded suite of analytical methods on a wider range of jade objects and geological samples from China.

A REVIEW OF SOME RECENT RESEARCH ON EARLY CHINESE JADES 209

EARLY JADE WORKING METHODS Jade working methods have been investigated by a variety of researchers, and we are beginning to understand how early jades were worked. This type of study can include examination of tool marks on finished and unfinished jades as well as remains from jade working (Wu, 1994). It is particularly important to understand the working methods used on jades of the Neolithic Hongshan culture because of the large numbers of forgeries that have been produced (So and Douglas, 1998; Forsythe, 1990). The remains of one jade workshop were discovered in 1997 north of Dingshadi near Nanjing in the proximity of the remains of the Neolithic cultures of Majiaban, Songze, and Liangzhu (Lu and Tao, 2001). The Nanjing Museum Institute of Archaeology and the Institute of Geological Research of Huadong are currently excavating this area. To date, the workshop has yielded a variety of stone tools that may have been used to work jade through cutting, drilling, surface abrading, polishing, and incising. Raw jade pebbles can still be found along a nearby river’s bank that may have been a source of jade for craftsmen during the Neolithic period. The Lingjiatan site in Anhui province was discovered in 1987, and jades yielded from the site are being studied with the aid of stereomicroscopy (Zhang et al., 2002). The Lingjiatan site is thought to be the location of the earliest agriculture-based city in China, dating to 4000 BCE or earlier. A proficient jade- producing culture inhabited the area as evidenced by the approximately 1,200 jades that have been unearthed there. This work is showing the presence of highly developed working methods, and evidence of the earliest use of the “tuo,” a small rotary disk tool to create fine incised decoration. A cutting edge of the tuo would be similar to the flat head of a nail, although other shapes could have been used for different purposes. In most cases the technique for drilling holes through jade was typically done from both sides of the object. The high level of craftsmanship is exemplified by the glossy polish on these jades, which has left little or no surface striation visible under the stereomicroscope. Tool marks preserved on jades from the collection at the British Museum have begun to be studied using detailed impressions with silicone dental resin (Michaelson and Sax, 2003), which follows from previous work on Mesopotamian seals composed of several quartz varieties. Impressions of small tool marks from jades are imaged using scanning electron microscopy (SEM), which greatly facilitates examination and documentation of these marks for comparison among jades. One Eastern Zhou jade plaque mentioned in this work was worked with several different handheld tools. Polishing techniques used in early China has been a largely unexplored area of research, but quartz sand and related fine-grained materials are generally accepted as the abrasives employed. After a Liangzhu corundum-rich axe was studied at the Freer Gallery of Art in 1998, a fragment from a similar axe was studied

210 SCIENTIFIC EXAMINATION OF ART

in a polishing replication experiment using commercially available abrasives that approximate natural diamond and corundum (Lu et al., 2005). The resulting polished surfaces were compared with the original surface polish produced by Liangzhu jade workers in antiquity. The polished surfaces were examined using optical and electron microscopes and characterized using atomic force microscopy. Three abrasives were used to polish the corundum-rich stone, and the diamond-polished surface most closely matched the surface polished in antiquity. The data suggest that extremely hard mineral abrasives composed of diamond may have been used to polish jades to the high gloss observed on these jades today. Likewise, corundum would have been another likely hard abrasive used. Ornamental jade rings from the Spring and Autumn period (771 to 475 BCE) are decorated with spiral grooves that were created through a mechanical method involving the use of a precision compound machine (Lu, 2004). Such rings must have had their spiral design drafted or directly carved through precisely linked rotational and linear motion of the type that has been demonstrated in recent experiments. These findings imply greater mechanical sophistication than has previously been assumed for this period in ancient China.

DETECTION OF HEATING Some physical and chemical changes that occur with the heating of nephrite are known from studies of the amphibole group minerals, tremolite-actinolite (Whittels, 1951; Vermaas, 1952). The dehydration of actinolite occurs in three stages, including the loss of adsorbed water, the loss of structural water, and a very small quantity of absorbed water. Studies using differential thermal analysis (DTA) show that an exothermic reaction takes place between 815oC and 824oC, and is associated with the oxidation of the small amounts of ferrous iron present in the mineral. This oxidation is not associated with any structural change in the crystal structure. Structural water is liberated at temperatures between 930oC and 988oC, and at lower temperatures with increasing iron in the mineral structure. This change occurs through a solid-state reaction:

→ . Ca2Mg5Si8O22(OH)2 + heat 2 CaSiO3 5MgSiO3 + SiO2 + H2O[1] nephrite (tremolite) pyroxene (diopside) cristobalite water

Detection of heating in jades using minimally invasive analytical methods is of interest because some jades may have been heated in antiquity prior to working or during burial rituals involving burning. Heat treatment may also be used in the production of modern-day forgeries to make jade appear older due to natural weathering or alteration. At the Freer and Sacker galleries, XRD and FTIR have been used to detect heating in jade, but these techniques have been found to be successful only if the object has been heated to at least 900oC (Douglas, 2001). In this study a nephrite pebble was sliced and heated in 100oC increments

A REVIEW OF SOME RECENT RESEARCH ON EARLY CHINESE JADES 211

FIGURE 2 (a) Heating series of low-iron (tremolitic) nephrite heating from Hetian (Kho- tan), Xinjiang province (nephrite slice length approximately 3 cm). (b) Heated bracelet (F1917.43) dating to the Neolithic period or Shang dynasty (bracelet diameter 6.0 cm).

from 500oC to 1100oC to observe visual changes and to investigate XRD and FTIR as methods to identify heating in jade. This heating series is shown in Figure 2, along with an example of a heated jade bracelet dating to the Neolithic or Shang dynasty. The heating series samples became more white and opaque with heating. Vickers hardness measurements on the heated samples showed that nephrite becomes slightly harder, rather than softer up to 800oC. After this temperature the material becomes brittle and tends to fracture more easily. In addition, black areas developed in the nephrite upon heating, which then became brown at 900oC. This black coloration may be due either to carbonization of small amounts of organic material trapped in crevices or oxidation of iron in the nephrite. The applicability of noninvasive Raman spectroscopy to the detection of heating in jade is also being investigated on the same nephrite heating series (unpublished information from P. P. Knops-Gerrits). Some of the XRD, FTIR, and microRaman data that can be used to identify heating in jade composed of nephrite are summarized in Table 1.

BURIAL ALTERATION AND SURFACE ACCRETIONS Burial alteration is a particular type of alteration known to occur on early Chinese jades composed of nephrite. Such alteration usually appears as opaque, white, chalky areas on otherwise translucent, polished jades. In many cases these patchy

212

–

1 1

– –

1 1

–

– –

Peaks

393, 368, 348, 1058, 1029, 930 cm

393, 369, 349, 1058, 1028, 930 cm

391, 365, 349, 1058, 1027, 930 cm

391, 365, 1058, 1026, 929 cm

391, 367, 1058, 1025, 927 cm

391, 323, 1340, 1010 cm

392, 327, 1341, 1011 cm

392, 327, 1341, 1013 cm

) MicroRaman

–

Rel % T

(475 and 512 cm

475 cm

512 cm

–

1075 cm

–

Peaks

FTIR

+ +

_ _

d-spacings d-spacing 3670 cm

XRD

_ _

_ +

+ +

C C

o o

C C

Philips RG-2600 X-ray diffractometer with Gandofi camera. Mattson Alpha Centauri Fourier transform infrared spectrometer.

o o o

Renishaw R1000 microRaman spectrometer, 785 nm excitation.

a b c

TABLE 1 Some XRD, FTIR, and MicroRaman Data That Can Be Used to Identify Heating in Jade Based on a Nephrite Heating Experiment

Heating Series Diffuse 8.38

Sample

Unheated _ + + _ 475 cm

500

600 700

800 900 1000

1100

NOTE: _ absent; + present.

A REVIEW OF SOME RECENT RESEARCH ON EARLY CHINESE JADES 213

areas of alteration are softer than the unaltered areas of the jade. This type of alteration consists of a selective dissolution or leaching on a microscopic scale along mineral grain boundaries by solutions of high pH (pH > 9) rather than a mineralogical change (Gaines and Handy, 1975). This type of high-pH environment can occur during decay of the corpse(s) with which the jades were buried. Experiments to produce burial alteration on jade indicate that it is likely that this type of alteration occurs during the months immediately after the burial when a corpse decomposes (Aerts et al., 1995). Optical coherence tomography (OCT) is a noninvasive technique that is being used to study the subsurface morphologies of jade objects to determine whether surface whitening is due to burning or natural alteration (Yang et al., 2004). Tomography images are used to show the refractive index or dielectric constant variations in jades, which reflect their internal structures. To date, OTC has been applied to a relatively small number of early jades but may prove to be useful in the future to answer questions relating to the authenticity of jade objects. Surface accretions remain an unexplored area of focused research, probably because it can be difficult to determine the significance and relative age of such deposits. Many jades are heavily cleaned and waxed, which often obliterates any accretions on their surfaces. Other accretions may unintentionally find their way to the surface of a jade but are typically not related to its early history. Earthy encrustations typically include calcareous deposits and soil. Occasionally lacquer and other organic remains can be seen.

CONCLUSIONS AND FUTURE DIRECTIONS Research using scientific techniques is helping us to understand the mineral composition and early history of well-documented and excavated jades. Similar work on unknown jades is helping to solve questions of authenticity (Douglas, 2000). Such study is most fruitful if it can include thorough visual examination using a stereomicroscope along with comparison to similar, preferably excavated jades. No direct methods of dating jade materials exist. Some future areas for research include

1. identification and distribution of surface accretions, weathering, and alteration; 2. continued study of jade working methods, with particular emphasis on the study of large groups of related jades from individual sites and cultural areas; 3. study of jade working remains, including tools and jade debris; 4. analytical and technical methods of dating jade workmanship; and 5. study of early jades from areas neighboring China, such as Korea, Taiwan, Siberia, and Southeast Asia, as all of these areas had jade-producing cultures.

214 SCIENTIFIC EXAMINATION OF ART

REFERENCES

Aerts, A., K. Janssens, and F. Adams. 1995. Orientations Nov.:79-80. Douglas, J. G. 2000. Orientations Feb.:86. Douglas, J. G. 2001. Proceedings of the Conference on Archaic Jades across the Taiwan Straits. Taipei: Guo li Taiwan da xue li xue yuan di zhi ke xue xi yin xing and Guo li Taiwan da xue chu ban wei yuan hui. pp. 543-554. Douglas, J. G. 2003. In Scientific Research in the Field of Asian Art, Proceedings of the first Forbes Symposium at the Freer Gallery of Art, ed. P. Jett, with J. G. Douglas, B. McCarthy, and J. Winter, pp. 192-199. London: Archetype Publications in association with the Freer Gallery of Art, Smithsonian Institution, Washington, D.C. Douglas, J. G., and W. T. Chase. 2001. Studies in Conservation 46:35-48. Forsythe, A. 1990. Orientations May:54-63. Gaines, A. M., and J. L. Handy. 1975. Nature 253:433-434. Harlow, G., and S. Sorensen. 2001. Australian Gemmologist 21:7-10. Hurlbut, C. S., and G. S. Switzer. 1979. Gemology. 243 pp., Canada: John Wiley. Lu, J., and H.Tao. 2001. In Enduring Art of Jade Age China, vol. 2, ed. E. Childs-Johnson, pp. 31-42. New York: Throckmorton Fine Art. Lu, P. 2004. Science 304:38. Lu, P. J., N. Yao, J. F. So, G. E. Harlow, J. F. Lu, G. F. Wang, and P. M. Chaikin. February 2005. Archaeometry 47:1-12. Michaelson, C., and M. Sax. 2003. APOLLO Nov.:3-8. So, J. F., and J. G. Douglas. 1998. In East Asian Jades: Symbol of Excellence, vol. 1, ed. C. Tang, pp.148- 163: Hong Kong: Chinese University of Hong Kong. Vermaas, F. H. S. 1952. Transactions of the Geological Society of South Africa 55:1. Wen, G. 1996. Acta Geological Taiwanica 32:55-83. Wen, G. 1997. Chinese Jades-Colloquies on Art & Archaeology in Asia 18:105-122. Wen, G. 1998. In East Asian Jades: Symbol of Excellence, vol. 2, ed. C. Tang, pp. 217-221. Hong Kong: Chinese University of Hong Kong. Wen, G., and Z. Jing. 1992. Geoarchaeology 7:251-255. Whittels, M. 1951. American Mineralogist 36:851. Wu, T. 1994. Renshi Guyu: Gudai yuqi zhizuo yu xingzhi. Taiwan: Zhonghua ziran wenhua xuehui. Yang, M. L., C. W. Lu, I. J. Hsu, and C. C. Yang. 2004. Archaeometry 46(2):171-182. Zhang J., Z. Yang, and Q. Cheng. 2002. Dong nan wen hua 5:17-27.

APPENDIXES

APPENDIX A

Contributors

Roy S. Berns is the Richard S. Hunter Professor in Color Science, Appearance, and Technology at the Munsell Color Science Laboratory and Graduate Coordi- nator of the Color Science master’s degree program within the Center for Imaging Science at Rochester Institute of Technology. He received B.S. and M.S. degrees in textile science from the University of California at Davis and a Ph.D. degree in chemistry with an emphasis in color science from Rensselaer Polytechnic Insti- tute. His research includes spectral-based imaging, archiving, and reproduction of cultural heritage; algorithm development for multi-ink printing; the use of color and imaging sciences for art conservation science; and colorimetry. He is active in the International Commission on Illumination, the Council for Optical Radiation Measurements, the Inter-Society Color Council, and the Society for Imaging Science and Technology. He has authored over 150 publications including the third edition of Billmeyer and Saltzman’s Principles of Color Technology. During the 1999-2000 academic year, he was on sabbatical at the National Gallery of Art, Washington, D.C., as a Senior Fellow in Conservation Science. During 2000, Dr. Berns was invited to participate in the Technical Advisory Group of the Star-Spangled Banner Preservation Project. He is currently involved in a joint research program in museum imaging with the National Gallery of Art, Washing- ton, D.C., and the Museum of Modern Art, New York. He is also collaborating with the Art Institute of Chicago and the Van Gogh Museum in digitally rejuve- nating paintings that have undergone undesirable color changes.

Barbara H. Berrie is senior conservation scientist at the National Gallery of Art, Washington, D.C. She received her B.Sc.(Hons) in chemistry from St. Andrews

217

218 APPENDIX A

University, Scotland, and her Ph.D. on electron transfer reactions from Georgetown University. She was awarded a National Research Council Postdoctoral Fellowship at the Naval Research Laboratory where she investigated the reaction of carbon dioxide with low-valent palladium compounds. She has worked at the National Gallery since 1984. Dr. Berrie has always been interested in the alchemy of turning base materials into art; now she is involved in studying the materials and painting methods of artists and analysis of materials in works of art in order to understand the artist’s original intention, and address issues of authenticity and preservation. She has used chemical analysis in the study of over 300 works of art in all media, including works on paper, easel paintings, and sculpture. She has published on paintings by Dosso Dossi, Gerard David, and Orazio Gentileschi among others and on the watercolors of Winslow Homer. Berrie is a Fellow of the International Institute for Conservation. She is the editor of the forthcoming volume of Artists’ Pigments that will be published by the National Gallery of Art.

Robin J. H. Clark is the Sir William Ramsay Professor of Chemistry and former Dean of Science at University College London. His research on physical inorganic chemistry and spectroscopy is concerned with synthesis, characterisation, and structure, focusing mainly on the electronic and vibrational spectroscopy of inorganic compounds, on matrix isolation infrared spectroscopy of photochemically generated species, and on infrared-based spectroelectrochemistry of redox-active species. In particular, he has made seminal contributions to virtually all aspects of Raman spectroscopy, notably to the characterisation of deeply coloured materials n- (e.g., TiI4) and to metal-metal bonded (e.g., M2X8 ) and linear-chain species, to gas-phase Raman band contour analysis, to Raman band intensities and the nature of the chemical bond, to the theory and practice of resonance Raman spectroscopy (including its application to the determination of excited state geometries), to nanostructures and thin, photoactive oxide/sulfide films on glass and, at the Arts/Science interface, to the application of Raman microscopy to the characterisation of pigments on medieval manuscripts, paintings, icons, papyri, sherds, and other artefacts. His research is embodied in nearly 500 scientific papers, 3 books, and 36 edited books. He has acted as Visiting Professor in 13 universities and has lectured in over 350 universities and institutions in 33 countries throughout the world. He was elected Hon FRSNZ (1989), FRS (1990), FRSA (1992), FUCL (1992), Hon DSc(Cant 2001), HonFRI (2004), and a Com- panion of the New Zealand Order of Merit (CNZM, 2004).

John K. Delaney has been a scientific consultant to the Conservation Division of the National Gallery of Art, Washington, D.C., since 1990, and has consulted on infrared imaging applications for other major American museums. He has published over 20 peer-reviewed papers on spectroscopy and several papers on infrared imaging in art conservation. He received his Ph.D. in Biophysics from The

APPENDIX A 219

Rockefeller University. He completed post-doctoral studies in the spectroscopy of biomolecules at the University of Arizona’s Department of Chemistry and the Department of Biological Chemistry at John Hopkins University’s School of Medi- cine. He is currently Chief Scientist for the Business Unit of Surveillance and Reconnaissance Systems, which is a part of Optical and Space Systems of Goodrich Corporation.

Janet G. Douglas is a Conservation Scientist in the Arthur M. Sackler Gallery and Freer Gallery of Art’s Department of Conservation and Scientific Research at the Smithsonian Institution. Her area of research involves the analysis of a variety of inorganic materials relating to Asian art such as jade, stone, pigments, metals, and corrosion products. She holds an M.A. in Geology (Metamorphic petrology) from Bryn Mawr College, awarded in 1980. She was a mineralogist at the U.S. Bureau of Mines for 5 years, involved in asbestos research. At the Freer and Sackler Galler- ies, her work involves research on Asian art and archaeological materials to answer questions relating to their authenticity, cultural context, and method of manufacture. Recent projects involve the mineralogical study of early Chinese jades, characterization of glass and stone gokok beads from Korea, and petrographic analysis of stone sculpture from Cambodia.

Molly Faries received her Ph.D. from Bryn Mawr College in 1972. During her years at Indiana University/Bloomington from 1975 on, she directed two long- term infrared reflectography (IRR) research projects: one a National Endowment for the Humanities Basic Research Grant (1984-1987) and the second, a Samuel H. Kress Foundation Grant for Art Historical Study Using Infrared Reflectography (1990-1997). Since 1998, she has also held a chair in Technical Studies in Art History at the University of Groningen in the Netherlands. Currently, she is involved in research for the catalogue of the fifteenth- and sixteenth-century northern collection of the Centraal Museum, Utrecht (funded by the Mondrian Foun- dation), and she is CPI for a project entitled, “Infrared Reflectography: Evaluative Studies,” in the interdisciplinary De Mayerne Program funded by the Dutch Or- ganization for Scientific Research (NWO), linking the exact sciences, conservation, and art history. In 1995, for her many publications in the field of northern European painting, she was awarded the College Art Association/National Insti- tute for Conservation Joint Award for Distinction in Scholarship and Conserva- tion, and in 2001, she was awarded the American Institute for Conservation Caroline and Sheldon Keck Award for Excellence in Education. Recent publications include The Madonnas of Jan van Scorel, Serial Production of a Cherished Motif (2000) and Recent Developments in the Technical Examination of Early Neth- erlandish Painting: Methodology, Limitations & Perspectives (2003).

Colin F. Fletcher is a Program Manager of Mouse Genetics at the Genomics Institute of the Novartis Research Foundation (GNF). His area of research is the

220 APPENDIX A

genetic analysis of mouse models of human disease, specifically neurological mutants that display ataxia. While at GNF he was a scientific co-founder of Phenomix Corp. Prior to joining GNF, he was a staff scientist in the Mammalian Genetics Laboratory at the National Cancer Institute. In the course of his research Dr. Fletcher employs a variety of imaging modalities, including confocal microscopy, magnetic resonance, X-ray, and low-light luciferase imaging with cryogeni- cally cooled CCDs. A long-standing interest in the scientific examination of works of art has lead to the previous publication of two reports in Studies in Conserva- tion. Dr. Fletcher received his Ph.D. from The Rockefeller Institute in Biochemis- try and Molecular Biology and his A.B. from Dartmouth College in Biochemistry.

Joyce Hill Stoner has taught for the Winterthur/University of Delaware (UD) Program in Art Conservation for 29 years and served as its director for 15 years (1982-1997). She graduated Phi Beta Kappa summa cum laude from the College of William and Mary in 1968. She received her Master’s degree in Art History at the Institute of Fine Arts of New York University (1970), her diploma in conservation at the NYU Conservation Center (1973), and a Ph.D. in Art History (1995, UD). She has been a Visiting Scholar in Painting Conservation at the Metropoli- tan Museum and at the J. Paul Getty Museum. In 1976, she founded the oral history project for the Foundation of the American Institute for Conservation and has interviewed more than 45 major art conservation professionals interna- tionally. Both an art historian and a practicing paintings conservator, Stoner has treated paintings for many museums and private collectors and was senior conservator of the team for the five-year project of examination and treatment of Whistler’s Peacock Room at the Freer Gallery of Art in Washington, D.C. Stoner has authored more than 60 book chapters and articles, and has recently been studying the paintings of the Wyeth family. She is currently serving as a Vice President of the Board of Directors of the College Art Association and a Vice President of the Council of the International Institute for Conservation. In June 2003 she received the AIC “Lifetime Achievement Award” sponsored by Univer- sity Products.

Tom Learner is a Senior Conservation Scientist at Tate in London, the UK’s national collection of British and international 20th/21st century art. He received a Master’s degree in Chemistry from Oxford University in 1988 and a Diploma in the Conservation of Easel Paintings from the Courtauld Institute of Art, London in 1991. He spent a year as a Getty Intern in the Painting Conservation and Scientific Research Departments at the National Gallery of Art (NGA), Washing- ton, D.C., and then joined the Conservation Department of the Tate Gallery in 1992, where he established appropriate analytical protocols for the identification and characterisation of twentieth-century painting materials with Fourier Trans- form infrared spectroscopy (FTIR) and pyrolysis-gas chromatography-mass spectrometry (PyGCMS). During this time he received his Ph.D. in Chemistry on The

APPENDIX A 221

Characterisation of Acrylic Painting Materials and Implications for their Use, Con- servation and Stability from Birkbeck College, University of London in 1997. He was a guest scholar at the Getty Conservation Institute (GCI) in 2001, assessing analytical techniques to follow changes in artists’ acrylic emulsion paints with accelerated light aging and with water immersion. He has written two books: The Impact of Modern Paints, co-authored with Jo Crook and published in 2000, and The Analysis of Modern Paints, published in 2004. He is currently coordinating a collaborative research venture into modern paints between the GCI, the NGA, and Tate, in which three initial areas of focus are improving methods for chemical analysis, studying their physical properties and assessing cleaning treatments.

Carol Mancusi-Ungaro serves as Associate Director for Conservation and Re- search at the Whitney Museum of American Art and Founding Director of the Center for the Technical Study of Modern Art at Harvard University Art Muse- ums. She graduated with a Bachelor of Arts degree from Connecticut College in 1968 and a Master of Arts degree from the Institute of Fine Arts, New York University in 1970. She trained and worked in conservation at the Yale University Art Gallery until she assumed a position of Conservator of Paintings at the British Art Center at Yale. Subsequent positions included Conservator of Paintings at the J. Paul Getty Museum in Malibu, California, and at the Intermuseum Conserva- tion Association in Oberlin, Ohio. For 19 years she served as Chief Conservator of The Menil Collection in Houston, Texas, and during that time she also consulted on the conservation of twentieth-century paintings at the National Gallery of Art in Washington, D.C. She has lectured widely on the conservation of modern art and written for retrospective catalogues on Mark Rothko and Jackson Pollock and most recently for the catalogue raisonné of Barnett Newman. In 2004 she received the College Art Association/Heritage Preservation Award for Distinction in Scholarship and Conservation. In her joint position, she teaches undergraduate and graduate students at Harvard University and continues to engage in research documenting the materials and techniques of living artists as well as other issues pertaining to the conservation of modern art.

Louisa C. Matthew received her Ph.D. in Italian Renaissance Art History from Princeton University with a thesis on Venetian painting. She received fellowships from the Delmas Foundation, the Harvard Center for Italian Renaissance Studies at Villa I Tatti in Florence, Italy, and recently a paired Kress fellowship together with Dr. Barbara Berrie at the Center for Advanced Studies in the Visual Arts at the National Gallery of Art in Washington, D.C. Currently, Dr. Matthew is an Associate Professor of Art History at Union College in Schenectady, N.Y.

Christopher J. McNamara received his Ph.D. in Aquatic Ecology in 2001 from the Department of Biological Sciences at Kent State University. Since receiving his doctorate, he has worked in the Division of Engineering and Applied Sciences at

222 APPENDIX A

Harvard University, first as a Postdoctoral Fellow and currently as a Research Associate. His research focuses on the ecology of biofilm bacteria and he has studied biofilms in diverse systems such as streams and aircraft fuel tanks. He has also studied the role of biofilms in deterioration of cultural heritage materials such as limestone from Maya ruins, protective coatings for bronze statues, synthetic cloth in Apollo spacesuits, and wax sculptures by Edward Degas.

Ralph Mitchell is the Gordon McKay Professor of Applied Biology in the Divi- sion of Engineering and Applied Sciences at Harvard. His Laboratory of Applied Microbiology has as its focus the microbiology of surfaces. The laboratory investigates the basic processes involved in the formation of biofilms on surfaces. His research group emphasizes the effects of biofilms on degradation of stone, metals, and artificial polymers. Current research in his laboratory involves the role of microorganisms in the biodeterioration of Maya sites in Mexico and in microbial processes resulting in corrosion of metals in the U.S.S. Arizona memorial.

Richard Newman is Head of Scientific Research at the Museum of Fine Arts, Boston, where he has worked since 1986. His lab oversees scientific research on the Museum’s collections carried out in collaboration with curatorial and conservation activities. One of his research interests is scientific methods for establishing the provenance of stone sculptures and the application of alteration layers in helping to resolve questions of authenticity. He is particularly interested in interdisciplinary research projects on works of art involving scientists, conservators, and art historians, and subjects he has studied range from qero cups produced in the Inka and colonial periods in Peru, to stone sculptures from the Indian sub- continent, to painting materials used in ancient Egypt. He collaborated with an art historian, conservator, and technical photographer in a 1988 book, Examining Velazquez, which received the 1991 Award for Distinction in Scholarship and Conservation from the College Art Association and National Institute for Con- servation.

Thomas D. Perry IV is the Sandia National Laboratories Campus Executive Graduate Fellow at Harvard University. His research involves understanding the processes of deterioration of materials, including stone, aluminum, and artificial polymers, by microorganisms living in biofilms. Biofilms are thin films of microorganisms living on surfaces that are capable of changing their surrounding environment through production of metabolites resulting in affected material deterioration. He is particularly interested in the specific mineral binding and mineralization caused by microbially produced polymers.

Michael R. Schilling earned his B.S. and M.S. degrees in chemistry from The California State Polytechnic University, Pomona. He has worked at The Getty Conservation Institute (GCI) since 1983 and presently holds the position of Se-

APPENDIX A 223

nior Scientist in charge of the Analytical Research Section. Michael oversees and coordinates a wide variety of projects in his section: applied research in materials analysis, scientific support to GCI’s field conservation projects, study of museum collections, evaluation of the air quality in museums, assessment of safe levels of lighting in museum galleries, and characterization of building materials. One research area in which Analytical Research scientists have developed considerable expertise is the characterization and analysis of organic materials. In this project, several gas chromatography-mass spectrometry procedures were developed for qualitative and quantitative analysis of natural organic binding media in paints. He and other scientists in the Analytical Research Section have conducted numerous workshops that were developed to inform conservation professionals about these GC-MS procedures. Since 1997, Michael and his staff have been studying the materials and techniques of modern and contemporary artists. Much of this work has involved the analysis of modern synthetic binding media and synthetic organic pigments. Michael has participated in collaborative projects to study and preserve wall paintings in the tomb of Nefertari, located in Luxor, Egypt, and also in the Mogao Grottoes, which is near the city of Dunhuang in the Gansu Province of China. He was also a member of a GCI research team that studied the Dead Sea Scrolls.

Elizabeth Walmsley is a painting conservator at the National Gallery of Art (NGA), Washington, D.C. She has also worked on the NGA’s systematic catalogue project from which has stemmed her interests in the technical examination of Old Master paintings using digital imaging, infrared reflectography, and x- radiography, and in the history of conservation. She graduated with an AB from Dartmouth College and received an M.A. in Art History with a Certificate in Art Conservation from Buffalo State College.

Paul M. Whitmore was trained as a chemist, getting a B.S. from Caltech and a Ph.D. from the University of California at Berkeley. He has worked in art conservation science for his entire professional career, starting at the Environmental Quality Laboratory at Caltech, working with Professor Glen Cass studying the effects of air pollution on works of art. From there, he went to the Fogg Art Museum at Harvard University, where he worked as a scientist in what is now the Straus Center for Conservation. Since 1988 he has been at Carnegie Mellon Uni- versity, directing the Research Center on the Materials of the Artist and Conserva- tor. His current research interests are in material degradation chemistries, intrinsic and environmental risk factors for those processes, and chemical sensors for material aging processes and risk factors. He has published on paper deterioration, its treatment, and damage induced by humidity changes; acrylic paint media stability and the physical damage to acrylic coatings from shrinkage stresses during drying; fading of colorants from air pollutant exposure; fading of transparent paint glazes from light exposure and the relationship between photochemical

224 APPENDIX A

degradation and color changes; and projects utilizing a new nondestructive probe of light stability for colored artifact materials. He has edited a book, Contributions to Conservation Science, a compilation of research papers published by the first director of the Center, Robert Feller. He is currently senior editor of the Journal of the American Institute for Conservation.

John Winter is a chemist (B.A., Cambridge University, Ph.D., University of Manchester) who, after a period of academic and industrial research, moved into the field of archaeological science and then into research on works of art using scientific methods. He holds the position of Conservation Scientist on the staff of the Department of Conservation and Scientific Research, Freer Gallery of Art/ Arthur M. Sackler Gallery, Smithsonian Institution. These museums hold the national collections of Asian art, which form the chief focus of research in the department. Dr. Winter’s own studies have mainly centered around East Asian paintings, their components, structural aspects, and the influence of microstructure and macrostructure on deterioration processes. He has published work on Chinese ink (the ubiquitous black design component across China, Japan, and Korea), lead-based white pigments, methods for the identification of organic colorants used as design components or as support dyes, painting techniques including those based on the use of precious metals, deterioration processes in East Asian paintings, and on other aspects of paintings as physical objects. Dr. Winter is a past President of the International Institute for Conservation of His- toric and Artistic Works, and has received research support from the Andrew W. Mellon Foundation as well as from the Smithsonian Institution itself.