Analytical Chemistry News & Features
September 1, 1999; pp. 614A-620A
Copyright 1999 American Chemical Society

X-ray Photoelectron Spectroscopy and Archaeology

XPS offers unique advantages for analyzing a wide range of artifacts

Today, the workhorses of bulk inorganic analysis are neutron activation analysis and inductively coupled plasma MS, but X-ray fluorescence and, earlier, atomic absorption have also contributed much. These methods detect a wide variety of elements on small samples with great sensitivity and with minimal harm to the artifact. Nonetheless, they cannot be used to examine elements with very low atomic weights—most notably boron through fluorine. Furthermore, these methods nonselectively detect all types of a given element, that is, they fail to distinguish chemical variations based on oxidation state. Thus, bulk analysis for iron does not distinguish Fe(0) from Fe(II) or Fe(III). Moreover, because they are techniques for bulk analysis, these methods generally cannot examine surface effects on artifacts.

X-ray photoelectron spectroscopy (XPS) and the related Auger electron spectroscopy (AES) can provide elemental analysis for essentially the entire periodic table. Because the electrons whose energies are analyzed arise from a depth of no greater than about 2-5 nm, the techniques are surface-sensitive. Perhaps most importantly, chemical forms often give distinct results in the X-ray photoelectron (XP) spectrum, which is analogous to the chemical shift in NMR, but it is now used, for example, to identify copper or iron oxidation states in pottery or glaze. XPS occupies a unique niche in archaeological chemistry because it permits examination of an extensive array of elements from the surface of artifacts and offers sensitivity to elemental oxidation states. After a description of the analytical method, this Analytical Approach describes how XPS and AES have contributed to understanding archaeological problems.

Photoelectron spectroscopy

The irradiation of a material with soft X-rays can result in the expulsion of the inner shell, or lowest energy, electrons (Figure 1a). The amount of energy required to dislodge the electron (the binding energy, BE) is diagnostic for the type of atom. If the energy of the X-rays (h) corresponds precisely to the energy required to eject the electron, the electron leaves the atom with no extra kinetic energy. In practice, the energy of the X-rays exceeds that required to remove the electron, so that it is emitted with a specific amount of kinetic energy (KE), which may be measured by the XP spectrometer. The diagnostic BE is calculated from KE and the known value of h by equation 1,

Figure 1. Phenomena responsible for XPS and Auger.

in which , the work function, is a known function of a specific spectrometer and sample.

Sometimes, an electron of slightly higher energy drops down nonradiatively into the vacancy that is created after the emission of the first electron (Figure 1b). The energy released by this process can be transferred to an outer electron and cause it to be ejected. Such a process involves three orbitals (e.g., KLL) and produces a doubly ionized atom through the ejection of a secondary, or Auger, electron. The emission of both regular and Auger electrons may be found in the XP spectrum as the result of the X-ray irradiation of the sample. In the Auger experiment, the sample is irradiated with an electron beam rather than with X-rays, and only Auger electrons are examined.

The value of electron energies, which are measured for photoelectrons in XPS or for Auger electrons in AES, is used to identify the element from which the electron was ejected, just as X-ray energies in X-ray fluorescence and -ray energies in neutron activation analysis identify the elements. Because electron energies are the focus of interest in XPS, this method is a form of electron or photoelectron spectroscopy. Because there are numerous other types of electron spectroscopies (2-10), the alternative term ESCA (electron spectroscopy for chemical analysis) is not entirely accurate.

An XP spectrum can contain peaks from all the elements present, with the exception of hydrogen and helium, making this method one of the most general in all of analytical chemistry. The core electron binding energies fall in the 50- to 1500-eV range. For some elements, different types of core electrons may be emitted, so that multiple peaks may be observed (e.g., for aluminum, the 2s electron is at 120 eV, and the 2p electron is at 75 eV). Further multiplicity may be observed when the spins of p or d electrons interact with the orbital energies. Thus, for iron, spin-orbit coupling gives rise to two 2p peaks—the 2p_3/2 at 710 eV and the 2p_l/2 at 723 eV. Typical binding energies for elements from lithium to uranium have been listed (2-7). Occasionally, peaks occur in the spectrum from different processes, including shake-up satellites (for definitions, see the Appendix in [7]).

Figure 2 shows typical survey XP spectra, which may be used for the qualitative analysis of a sample. The elements are identified by the correspondence of the peak position with those of the standards. Quantitative analysis is a more complex process. The intensity of an XP peak depends on several factors other than the concentration of the element, such as sample area, sensitivity (cross section) for the absorption of X-ray photons, surface contamination, and probability of photoelectron escape. Experimental design can eliminate some of these factors, and tables of sensitivity factors are available for most elements (8, 9). The division of the observed peak intensities by these literature sensitivities provides a number that should depend directly on the amount of the element present. In the archaeological context, some authors have reported quantitative results in terms of absolute percentages (10), whereas others have reported relative amounts (compared with another element, such as oxygen [11]). Early workers claimed precisions of 10-15%, but they reported accuracies of only 50% or more. Relative amounts of two elements or two oxidation states of one element may be obtained with higher accuracy because of the cancellation of some sources of error. Recent studies have claimed accuracies of up to 5%.

An important characteristic of the XP experiment is its surface dependence. Although X-rays penetrate to a depth of several micrometers, ejected photoelectrons generally come from only the first several nanometers of material. Thus, XPS is very much a surface technique, much more so than X-ray fluorescence. This aspect of XPS necessitates great care in experimental design, as the surface may be contaminated, nonuniform, or unrepresentative. At the same time, surface phenomena may be addressed explicitly. Composition also may be studied as a function of distance from the surface through the use of ion sputtering or etching, whereby a stream of ions, usually Ar⁺, is used to remove a defined surface layer. The advantages and disadvantages of the surface sensitivity of XPS are illustrated by the applications that follow.

Pottery

Pottery is composed of clay mixed with a temper to improve firing properties (1). Clay is an aluminosilicate mineral containing relatively large amounts of iron. Tempers include sand, limestone, shells, straw, and dung. The first study of pottery by XPS (Figure 2b) was carried out on Kherbet Kerak shards (7), which were excavated during the operations of the University of Chicago's Oriental Institute in the 'Amuq on the Plains of Antioch in Turkey (12). Semiquantitative analysis of 12 shards gave a data set that could easily distinguish the two sources of the shards: Tell Ta 'Yinat for samples 1-4 and Tell a-Judaidah for samples 5-12. A plot of the intensity of the aluminum 2s peak versus that of the magnesium KLL (X-ray-excited) Auger peak clearly separated the two groups.

Figure 2. Survey X-ray photoelectron spectra.

Lambert et al. (13) carried out a more comprehensive study on 71 Mycenaean-style and related shards excavated at the site of Megiddo (ca. 13th century B.C.). This famous site, the biblical Armageddon, is located 30 km southwest of modern Haifa, Israel. Its strategic location along ancient trade routes led to interactions with many of the cultures of the era. During the Late Bronze Age, mainland Greece enjoyed a period of high culture associated with the site of Mycenae. In particular, attractive and well-executed pottery from Mycenae was widely traded throughout the Mediterranean. Some 70 fragments of stylistically Mycenaean pottery have been excavated at Megiddo in both occupational and funerary contexts. Many of the fragments were clearly from containers suitable for transport. The question arose as to whether these pottery fragments were manufactured locally or imported. Chemical analysis of the pottery could answer this question, because there are numerous samples of local ware to compare with the stylistically Mycenaean material.

Mycenaen pottery shards found at the Megiddo site. (Reprinted with permission from Ref. 13.)

With data from 10 XPS peaks, principal component analysis classified the 71 shards into 3 main clusters, 5 minor clusters, and a residue. The Mycenaean-style shards fell primarily into the three main clusters (see photographs above). The minor clusters mainly contained the local fabric vessels included in the study as controls. These distinctions indicated that the Mycenaean-style materials were not made of the same clays as the local ware and, therefore, were imported from at least three different sources, corresponding to the three main clusters. The compositional profile of the main clusters did not parallel that of Mycenaean pottery from the Peloponnisos in mainland Greece but had some resemblance to that of Cretan materials.

This study relied on bulk analyses, which required powdering a representative portion of the sample and collecting spectra from samples mounted on tape or foil. Thus, bulk analysis by XPS is destructive of the sample. However, whole samples may be attached mechanically to a holder and examined nondestructively. Such analyses are characteristic of only the surface.

Levanthal and Thompson (14) exploited the surface sensitivity of XPS in a study of the composition of an Egyptian pottery slip. A slip is a thin layer applied to the surface of a pottery vessel, usually by dipping it into a dilute suspension of clay in water (1). The material is so thin that bulk analysis is not possible by traditional methods. Qualitative analysis of the clay slip by XPS showed the presence of calcium, chlorine, silicon, iron, and fluorine, which suggested that the material was a fluorspar. Few alternative analytical methods could have detected the key element fluorine.

The second major strength of XPS is its ability to distinguish oxidation states. Lambert and co-workers (15) were able to distinguish iron oxidation states in pottery samples from the Bahamas (ca. 1250-1360 A.D.) and Puerto Rico (ca. 120-1200 A.D.). Amerindian populations migrated from the Greater Antilles, which includes Puerto Rico, south to the Bahamas archipelago around 800 A.D. Because the Amerindians in both locations usually used a relatively uncontrolled open pit rather than a kiln for firing, their pottery exhibits a range of colors from light brown to red to black, depending on the oxidation state of iron. XPS can determine the relative contributions of the various forms of the iron.

A significant difference between the binding energies of the black constituent FeO and the reddish-brown constituent Fe₂O₃ can be seen in Figure 3. The proportions of iron oxides were measured semiquantitatively and found to range from 89% Fe₂O₃ (the remainder expressed as FeO) in a light brown Puerto Rican sample to 33% Fe₂O₃ in a black Bahamian sample. The analysis of different layers, which was performed using ion etching, was also enlightening. The brown outermost material of a Bahamian sample contained 74% Fe₂O₃, and the darker inner material contained 49% Fe₂O₃. Thus, the surface consisted of more highly oxidized material. Error was estimated at ±5%.

Figure 3. X-ray photoelectron spectra of iron in a pottery shard from Crooked Island, Bahamas.

Pigments and glazes

Colored inorganic materials, or pigments, often were used to decorate the surface of ancient objects (1). Decorations were placed directly on the surface of the pottery, on a slip, or on a glaze (a thicker, nonporous, glassy layer baked onto the surface). Because XPS is primarily a surface technique, it has been used extensively in the analysis of surface pigments. Moreover, colors often are associated with specific oxidation states of an element, which XPS is particularly well suited to identify.

Bruno et al. (16, 17) examined several colored medieval pottery fragments from southern Italy. A longstanding manufacturing tradition in Apulia (the heel of Italy's boot) produced glazed pottery with red, green, and brown decorations. Red in pottery almost always is associated with iron, specifically with Fe(III) as Fe₂O₃. The XPS analysis revealed that the red colors are also associated with the presence of lead. This element can serve, depending on its oxidation state, either as a pigment or as a flux, which enables silica (sand) in the glaze to melt at a reasonable temperature. Thus, a white, transparent glaze might contain PbO as the fluxing agent, but an opaque, red glaze might contain PbO as a flux and either Pb₃O₄ (minium) or Fe₂O₃ (hematite or red ocher) as the pigment.

The XPS spectra confirmed that the white glaze contained PbO, and the red glazes from several sites contained a mixture of PbO and Pb₃O₄. Unglazed, painted pottery from Castel Fiorentina in Foggia, on the other hand, contained no lead but had iron, suggesting that Fe₂O₃ was the pigment. A glazed sample from Gallana in Brindisi had both iron and lead, so the situation is complex.

McNeil (18, 19) has used XPS and scanning Auger microscopy to examine the age of ink on cellulose substrates, such as paper, or on animal fibers, such as parchment. He argued that iron in iron gallotannate ink migrates over the substrate surface, with the area of migration proportional to age. This method closely replicated the known dates for 37 substrates from the period 1350-1950 A.D. He also used the Auger technique to scan the substrate for ink erasures, thereby determining if a document was written at a single time or over time.

Carbon is another element that is poorly analyzed by standard techniques. Gillies and Urch (10) used XPS to study black-surfaced ware to distinguish between carbon and iron sources of the blackening. The black coloration of shards from Little Waltham and Orsett in the U.K. (Iron Age), and from Servia, Macedonia, (Early Bronze Age) was found to be caused by carbon, whereas the black coloration on black polished ware from northern India (Iron Age, 6th-2nd century B.C.) was attributed to biotitic iron. Argon ion etching was used to remove the dark metallic surface to examine the underlying slip.

Wilson-Yang and Burns (11) have used XPS to study the degradation of mural paintings at the Beni Hasan tombs (ca. 2100 B.C.) in Egypt. A gray deposit has been forming over the murals, obscuring the coloration. Various mechanisms were excluded by the XPS experiments. For example, the lack of extra chlorine in the XP spectra of samples cleaned with HCl eliminated CaCl₂, and the absence of abnormal levels of sulfur meant that sulfate from the underlying plaster was not diffusing through the mural. However, XPS analysis for carbon, silicon, and aluminum found some surfaces with carbonates and others with aluminosilicates. Thus, the favored mechanism involves the adsorbance of water, enhanced by the deposition of dust or sand on the surface. The diffusion of surface water through the mural to the rock on which the paintings had been applied then could result in the dissolution of CaCO₃ from the limestone walls. Fluctuations of temperature and humidity can cause ions to diffuse to the surface and deposit CaCO₃. This knowledge may lead to the development of procedures for stabilizing the mural coloration. This study exemplifies the strengths of XPS in examining surface phenomena and analyzing nonmetals.

Glass

Raw materials were heated at high temperature in the production of glass and pottery (1). Glass is composed of silicon dioxide (sand) and various metals, which provide properties such as color, opacity, and a lower melting point during manufacture. Egypt was the first great producer of glass, starting around 1500 B.C. at the beginning of the New Kingdom. These early materials often exhibited bright colors through selective addition of metal oxides. XPS is a useful method for examining the oxidation states of these colorants, since color often varies with oxidation state.

Figure 2a shows the survey spectrum of a fragment of an Egyptian glass from the 18th dynasty (7). A detailed XPS study of the metals' oxidation states was carried out on the glass to determine the chemical species responsible for the blues, greens, and reds (20). Although blue colors often result from the presence of cobalt, elemental analysis found no significant amount of this element. If copper is responsible for both red and blue-green colors, it must be the result of differences in oxidation states. Lambert and McLaughlin used XPS to establish these differences. Relative binding energies, peak intensities, and the presence of shake-up satellites (7) served to distinguish the Cu(II) oxidation state from Cu(I) and Cu(0). Red glasses contained the lower oxidation states of copper and blue-green glasses the Cu(II) state. Even 3500 years ago, Egyptian artisans were able to control color by manipulating oxidation states.

Metals

Ancient civilizations have exploited numerous metals, including gold and silver for decorative purposes, iron and copper as utilitarian bulk metals, tin and lead to enhance the properties of the bulk metals, and a variety of metals in their oxidized forms to add color to glass, pottery, and glaze. Although other methods are superior for bulk analysis, XPS is particularly effective for identifying metal oxidation states and for examining surface effects, such as corrosion and patinas.

The earliest XPS study of an archaeological metal was of the Luristan bronze (7) in Figure 2c. Figure 4 shows the Cu 2p region. An unetched sample (upper spectrum) shows peaks from both the Cu(0) and Cu(II) oxidation states (the peak multiplicities are from spin-orbit coupling and shake-up processes). Etching the surface with Ar⁺ produced the lower spectrum, which corresponds to a Cu(0) species. Thus, the thin oxidation layer was easily removed. Although etching has been used widely to remove surface features sequentially in layers, the process has several drawbacks. Damage caused by ion bombardment probably is not significant, so there is no aesthetic problem. Etching, however, not only removes surface constituents but also can cause elemental segregation. Both the depletion and augmentation of elements at the surface can make the interpretation of etching experiments difficult. An alternative method for depth profiling is to examine the angle dependence of the emitted electrons by tilting the sample with respect to the electron analysis lens (21). When perpendicular to the sample, the analyzer looks more deeply into the sample than with a shallower angle. In this way, composition can be studied as a function of depth, although only with a nicely flat surface.

Figure 4. Copper 2p region in the X-ray photoelectron spectrum of a Luristan bronze.

Auger peaks (7) were examined for a tri-wing metallic arrowhead (see opening art on p 612 A) found in the northern Sinai and attributed to the Saitean period (6th century B.C.) (22). Qualitative and quantitative XPS analysis found that the arrowhead has an unusual Cu/Sn/Mg alloy, with minor peaks from calcium and chlorine. Moreover, the magnesium peaks were characteristic of MgO (Mg(II)) rather than Mg(0), whereas the copper peaks correspond to Cu(0). Elemental maps were obtained over a cross section. The edges appeared to be richer in magnesium and oxygen, whereas the core was richer in copper. Using an electron beam in this Auger experiment provided better spatial resolution (3 m) than standard XP peaks, which are limited by the size of the X-ray beam (2-3 mm).

Gillies et al. (23) studied intentional surface enrichment of silver on a predominantly copper Roman coin by sequential Ar⁺ sputtering. They found that surface copper and silver were present only in their oxidized forms, but the free metals appeared after sputtering. Chlorine was present from the surface to the depths achieved by three rounds of sputtering, suggesting that the surface silver might have been applied by dipping the copper object into molten AgCl. The original artisans were probably cosmetically improving a predominantly copper object and presumably increasing its value by giving it a silver appearance.

In a search for hard evidence for iron smelting in early Iron Age Syria, Ingo and co-workers (24) examined slag-like material found at the site of Tell Afis. Bulk analysis indicated that the slag contained about 25% Si and 40% Fe. Small-area XPS examined iron-rich inclusions in the slag. This technique permits the analysis of areas of only 0.1-0.2 mm. They identified oxidation states of iron, calcium, magnesium, silicon, and aluminum consistent with the expected slag by-products of iron smelting.

Paparazzo (25) and Moretto (26) examined the joining process of Roman lead pipes (fistulae) from the 3rd century A.D. by XPS. The presence of tin at the joint indicated that the Romans preferred soldering with a Pb/Sn solder over welding (direct Pb/Pb joining). The presence of tin in the solder means that a relatively low temperature could be used for joining (1/1 Sn/Pb is a paste at 183-252 C and a fluid at >252C). Moreover, Paparazzo found that the more easily oxidized iron at the joint protected the lead from oxidation, as demonstrated in Figure 5. Metallic Pb(0) is found at 137 eV and lead oxides (PbO and PbO₂) at 139 eV. The nonjoint portion of the pipe contains appreciable levels of oxidized Pb (see Figure 5a and 5b, the latter of which was obtained after Ar⁺ etching). In contrast, the material is primarily metallic lead at the joint (see Figure 5c and particularly 5d after etching). Paparazzo also found high concentrations of carbon at the joint, indicating that an organic material such as olive oil was used during soldering to minimize oxidation. These experiments confirmed Pliny's assertion that "lead is not able to be joined to itself without tin, nor is the former to the latter without oil" (26).

Figure 5. X-ray photoelectron spectra of a Roman lead pipe joint.

Summary

XPS complements existing analytical techniques for the study of materials of interest to the archaeologist. It has a number of particular strengths: Oxidation states are identified, all elements are measured simultaneously, and nonmetals that are poorly studied by standard methods may be readily observed by XPS. The technique is applicable to both organic and inorganic materials. Because it is a surface method, XPS can provide considerable information about vertical structure, and Ar⁺ etching provides a means of sequential sampling as a function of depth. Alternatively, angle-dependent studies can provide similar information. The technique requires very little sample (milligrams to micrograms) and can be nondestructive.

Offsetting these advantages are poor absolute sensitivity (elements must be present at the level of 0.1-0.5% to be detected, albeit in a very small sample), difficulties in quantitation associated with matrix matching, surface contamination (a particular part of the surface may not be representative of the rest of the surface or of the bulk), and difficulties with nondestructive analysis of large objects. With due consideration for these problems, XPS provides an important method for the analysis of archaeological materials.

References

 (1) Lambert, J. B. Traces of the Past; Addison-Wesley: Reading, MA, 1997.

 (2) Carlson, T. A. Photoelectron and Auger Spectroscopy; Plenum Press: New York, 1975.

 (3) Carlson, T. A. X-ray Photoelectron Spectroscopy; Dowden, Hutchinson & Ross: Stroudsburg, PA, 1978.

 (4) Practical Surface Analysis, Vol. 1: Auger and X-ray Photoelectron Spectroscopy; 2nd ed.; Briggs, D., Seah, M. P., Eds.; Wiley: New York, 1990.

 (5) Hüfner, S. Photoelectron Spectroscopy; Springer-Verlag: Berlin, 1995.

 (6) Spoto, G.; Ciliberto, E. In Modern Analytical Methods in Art and Archaeometry; Ciliberto, E., Spoto, G., Eds; Wiley: New York, in press; Chapter 13.

 (7) Lambert, J. B.; McLaughlin, C. D. Archaeometry 1976, 18, 169.

 (8) Scofield, J. H. J. Electron Spectrosc. Relat. Phenom. 1976, 8, 129.

 (9) Wagner, C. D.; Davis, L. E.; Zeller, M. V.; Taylor, J. A.; Raymond, R. H.; Gale, L. H. Surf. Interface Anal. 1981, 3, 211.

(10) Gillies, K. J. S.; Urch, D. S. Archaeometry 1983, 25, 29.

(11) Wilson-Yang, K. M.; Burns, G. Can. J. Chem. 1987, 65, 1058.

(12) Braidwood, R. J.; Braidwood, L. S. Excavations in the Plains of Antioch; Oriental Institute Publications, Vol. LXI; University of Chicago Press: Chicago, 1960.

(13) Lambert, J. B.; McLaughlin, C. D.; Leonard, A., Jr. Archaeometry 1978, 20, 107.

(14) Levanthal, E. T.; Thompson, M. J. Intern. Inst. Conserv. Can. Gr. 1978, 3, 16.

(15) Lambert, J. B.; Xue, L.; Weydert, J. M.; Winter, J. H. Archaeometry 1990, 32, 47.

(16) Bruno, P., et al. Fresenius' J. Anal. Chem. 1994, 350, 168.

(17) Bruno, P.; Caselli, M.; Curri, M. L.; Favia, P.; Laganara, C.; Traini, A. Ann. Chim. 1997, 87, 539.

(18) McNeil, R. Literary Research 1988, 13, 137.

(19) McNeil, R. J. In Archaeological Chemistry III; Lambert, J. B., Ed; Advances in Chemistry Series No. 205; American Chemical Society: Washington, DC, 1984; pp 255-69.

(20) Lambert, J. B.; McLaughlin, C. D. In Archaeological Chemistry II; Carter, G. F., Ed.; Advances in Chemistry Series No. 171; American Chemical Society: Washington, DC, 1978; pp 189-99.

(21) Baschenko, O. A.; Nefedov, V. I. J. Electron Spectrosc. Relat. Phenom. 1990, 53, 1.

(22) Polak, M.; Baram, J.; Pelleg, J. Archaeometry 1983, 25, 59.

(23) Gillies, K. J. S.; Moss, G. P.; Urch, D. S. Revue d'Archeometrie 1980, Vol. III (Suppl. 121), 121.

(24) Ingo, G. M.; Scoppio, L.; Bruno, R.; Bultrini, G. Mikrochim. Acta 1992, 109, 269.

(25) Paparazzo, E. Appl. Surf. Sci. 1994, 74, 61.

(26) Paparazzo E.; Moretto, L. J. Electron Spectrosc. Relat. Phenom. 1995, 76, 65.

Joseph B. Lambert is Clare Hamilton Hall Professor of Chemistry at Northwestern University, where he combines research interests in physical organic chemistry and archaeological chemistry. Charles D. McLaughlin received his doctoral degree in archaeological and organic chemistry at Northwestern, worked for several years at Chevron Research Co., and now pursues a career as a professional sculptor. Catherine E. Shawl is a graduate student in inorganic chemistry at Northwestern University. Her research interest is in the area of archaeological chemistry. Liang Xue, a native of Shanghai, China, received his doctoral degree in archaeological and physical chemistry at Northwestern and is now a nuclear magnetic resonance spectroscopist at Alcon Laboratories, Inc. Address correspondence to Lambert, Dept. of Chemistry, 2145 Sheridan Rd., Evanston, IL 60208.