FINAL REPORT RAMAN SPECTROSCOPIC ANALYSIS OF ROCK ART PIGMENT

MEDIEACKREDITERING ELITSERIEFINALEN I KONSTÅKNING – HALMSTAD ARENA IS HALMSTAD
EMAIL DISCLAIMER GUIDELINE QGEA EMAIL DISCLAIMER GUIDELINE FINAL DECEMBER
IMPLEMENTING INFORMATION GOVERNANCE QGEA IMPLEMENTING INFORMATION GOVERNANCE FINAL JUNE

LYNLEY SHIMAT LYS MIAP INTRO FINAL PROJECT HADASSAH FILM
PROJECT NAME NJDOT SCOPE STATEMENT TSM LIMITED SCOPE FINAL
TEXT LEGEND DELETE THIS LEGEND BEFORE FINALISING THE DOCUMENT

Preliminary Results of a Raman Spectroscopic Analysis of Rock Art Pigment from the Great Gallery, Maze District, Canyonlands N

Final Report: Raman Spectroscopic Analysis of Rock Art Pigment from the Great Gallery, Maze District, Canyonlands National Park, Utah


Dr. Pete E. Poston*and Gary Cox**


*Professor of Chemistry, Western Oregon University, Monmouth, OR 97361, Phone: (503) 838-8218, email: [email protected]


**Archeological Technician, Maze District, Canyonlands National Park, UT, Phone: (435) 259-2652. email: [email protected]



Abstract: Raman Spectroscopy and Scanning Electron Microscopy – Energy Dispersive X-ray spectrometry (SEM-EDX) were used to analyze a fragment of rock art from the Great Gallery site, Canyonlands National Park, Utah. Comparisons were made to previous Fourier Transform Infrared (FT-IR), X-ray Diffraction (XRD), Gas Chromatography-Mass Spectrometric (GC-MS), and Near Infrared (NIR) studies where the principle inorganic ingredients in the pigment were hematite, quartz, orthoclase, gypsum, kaolinite, and calcite. The Raman results confirmed the presence of hematite as the primary component, and also found quartz and gypsum in the pigment. Kaolinite was not found spectroscopically, but it was observed in the SEM images. Calcite is possibly in the spectrum, but close to the background noise in magnitude. Orthoclase was found to be confined exclusively to the rock underneath the pigment and bulk Navajo Sandstone samples collected at the Great Gallery and near the Hans Flat Ranger Station. A mineral not detected in the other reports except for the NIR study was fined-grained muscovite, or illite, which together with color centers is thought to be responsible for the background fluorescence of the sample, No organic binders were found in the Raman spectrum, in agreement with the GC-MS results. Instead, the presence of a Fe-O-Si bending frequency in the Raman spectrum indicates the pigment bonded inorganically to the rock substrate by linking the silica in the substrate to the hematite in the pigment. The mineral whewellite, CaC2O4H2O, was detected, indicating the presence of lichens in the past. This suggests that at the very least the time of burial of the detached slab could be carbon-14 dated. Until the nature of deposition of the whewellite is further understood, it’s not recommended to pursue this approach just yet. Finally, a comprehensive power study was undertaken to confirm no damage was being done to the pigment through laser heating at 785 nm, a serious concern with high-powered diode lasers commonly used in Raman systems.












I: Introduction


The goal of this study was to evaluate the use of Raman Spectroscopy as a nondestructive technique to help determine the composition of Barrier Canyon-style (BCS) rock art found in and near Canyonlands National Park, Utah.


The major advantage of Raman Spectroscopy is that it is nondestructive of the sample as compared to standard analytical techniques such as Gas Chromatography-Mass Spectrometry (GC-MS) and Fourier Transform-Infrared (FT-IR) Spectrometry, which rely on a sample being taken from the field and investigated in a laboratory. Raman Spectrometers are also available in portable, lightweight forms, making it possible to take them into the field, providing access to many previously inaccessible rock art sites in remote, hard-to-reach locations.


Barrier Canyon Style rock art is found extensively in the Canyonlands area (examples shown in Figure 1), and is thought to have originated from hunter-gatherer tribes during the Archaic Period. Joel Pederson of Utah State University, through the use of Optically Stimulated Luminescence (OSL) dating of alluvial terraces, has been able to bracket the age of the Great Gallery site in Horseshoe Canyon between ~15-6 ka (1).


The composition of BCS pigments in the Maze District of Canyonlands National Park has not been extensively researched as compared to European and African rock art, however (2-6), although there have been important studies of Pecos-style art in Texas (7,8). An analysis of the pigments used, and the detection of any organic binders that may be present, might allow radiocarbon dating as well as cultural comparisons to other similar rock art styles and materials. This could potentially reveal unknown cultural affiliations with, for example, the Pecos style in Texas as first suggested by rock art expert Polly Schaafsma in her book “The Rock Art of Utah” (9).


Powerful, highly sensitive, and accurate instrumental techniques such as X-ray Diffraction (XRD), Fourier Transform Infrared (FT-IR) Spectroscopy, and Gas Chromatography-Mass Spectrometry (GC-MS) have been used to study the composition of rock art pigments, including samples from Canyonlands National Park. Alan Watchman has studied the pigment from a detached fragment of rock art found buried beneath the Great Gallery (10), and David Hartman at the Australian National University performed a GC-MS analysis of the same pigment Watchman analyzed (11). Constance Silver, Principal of Preservar, Inc, also performed an analysis of rock art from a large variety of sites within Canyonlands, including the same detached slab that Watchman analyzed (12). More recently in 2008, Gregg Swayze of the United States Geological Study (USGS) used a portable Near Infrared (NIR) Spectrometer to study figures from the Shelter Site and the Great Gallery in Horseshoe Canyon (13).


Watchman was given permission to use XRD and GC-MS to identify the principle components found in this detached sample, and if enough carbon was identified, perform Accelerator Mass Spectrometric (AMS) dating as well. In his XRD and AMS report (10) Watchman identified the following components in the pigment (Table 1): quartz (silica), potassium feldspar (orthoclase), hematite (iron oxide), gypsum, kaolinite, and calcite.


Watchman was also able to isolate trace amounts of carbonaceous material and a brush fiber used to paint the figure, and the subsequent Accelerator Mass Spectrometric (AMS) dating placed the age of the fiber at 3000-2800 ybp. Subsequent analysis by David Harman to make sure that no recent carbon was contaminating the sample (11) led to the conclusion that the Great Gallery pigment had little to no organic content above 2 nL/gram. Extrapolating from these results, Watchman’s original date of 3000-2800 ybp must have suffered from contamination. From Pedeson’s results, it’s known the real date is much older, and more in line with Watchman’s radiocarbon date of the nearby Barrier Canyon-style Black Dragon Panel, which is around 8000 ybp.


Constance Silver, of Preservar, Inc, in collaboration with Prof. Richard Wolbers at the University of Delaware (12), was able to detect the presence of protein and lipids (oils) in the Great Gallery pigment, a finding at odds with Harman’s work. Using FT-IR, Wolbers detected iron oxide, carbonate, and natural resins in the pigment. Even though Wolbers used a scalpel to separate the pigment from the underlying rock, contamination by organics cannot be ruled out, however.


Gregg Swayze of the USGS studied the figures from the Shelter site and Great Gallery sites in Horseshoe Canyon during the Winter of 2008 (13). Using NIR Spectroscopy, Swayze was able to detect hematite as the primary pigment, with trace amounts of kaolinite, and the clay mineral illite. There were also possible indications of organic residue in the pigment as will be discussed later. An extremely interesting interpretation of the spectra suggested the presence of the iron oxide maghemite, with the possibility that it originated from blood in the pigment.

II: Description of Raman Spectroscopy


Briefly, Raman Spectroscopy is a technique that uses laser light to cause “vibrational transitions” within a molecule via nonelastic scattering of incident electromagnetic radiation. Simply put, the atoms in a bond within a molecule vibrate back and forth as “phonons”. These vibrations occur at frequencies that are fixed or “quantized”, which means that not all vibrational energies are “allowed”. Inputting or losing energy from these bonds places the molecule in different “quantized” vibrational energy levels.


A molecular vibration can be described using a simple spring analogy where the spring holds the atoms in a bond together. Laser light scatters off the vibrating “spring”, with energies that are either lower or higher in energy than the incoming laser photons, causing the molecule to change vibrational energy levels. The lower energy photons are known as the Stokes Shift, and have higher intensities that the other photons (the Anti-Stokes Shift). These scattered Stokes-shifted photons are detected using a spectrograph and a sensitive CCD detector, much like the detector commonly used today in digital cameras.


These characteristic photon frequencies are displayed as peaks in the resulting Raman spectrum, and the frequencies are converted to what spectroscopists refer to as “wavenumbers”. Wavenumbers are directly related to vibrational frequency and its energy, and have units of “inverse centimeters” or cm-1. The wavenumber of a given peak is highly diagnostic of the presence of that material in the sample, and all of the peaks together in the spectrum produce a “fingerprint” pattern to the exclusion of other molecules.


It’s important to point out that infrared spectra obtained by Raman spectroscopy and standard FT-IR are complementary in nature. This means that the same functional group within a molecule (e.g. carbonate or CO3) may appear at the same wavenumbers in both kinds of spectra, or sometimes only in one spectrum and not in the other. Taken together, a tremendous amount of useful information can be obtained about the structure of the molecule under study.


Because of continuing developments in technology, Raman instrumentation has become highly portable and sensitive. There are a variety of models available for use in the field, with a range of prices associated with them. For example, the DeltaNu Rockhound is a basic unit that costs around $22,500, while a significantly more advanced instrument manufactured by B&W Tek cost around $100,000 apiece.


In this study, access to a laboratory-based, fiber-optically coupled Raman Spectrometer was kindly provided by Professor Joel M. Harris at the University of Utah in Salt Lake City, Utah. All data in this report were taken using this system, and specifications of this Raman system, including laser wavelength, power, and beam diameter (spot size) is given in the Instrumental Procedure section below. In addition, two portable units were provided by Ahura Scientific and DeltaNu for evaluation, and the assessment of these units is on-going and not described here.


III: Nature of the Samples


As described by Watchman in his report (10), a partially buried, fallen slab measuring 40 x 20 cm was discovered by Ranger Gary Cox below the Great Gallery. This fragment of rock contains thin red lines, red paint, as well as unpainted rock, and is part of a Great Gallery BCS anthromorph.


An approximately 1.0 x 0.5 x 0.5 cm (L x H x W) sample from this broken slab was furnished for analysis by rock art researchers Nancy Simon and Richard Reed with permission of the NPS. The sample arrived packaged in aluminum foil, and was physically handled using cotton gloves to prevent contamination with organic materials found on the fingers and hands, although this can’t exclude previous handling. The sandstone is rather friable, however, as are many of the spalled blocks found below the Great Gallery. The pigment is uniformly distributed on one side, and the backside of the fragment consists of freshly fractured sandstone with no adhered pigment.


Navajo Sandstone samples were collected for background studies, two from loose alluvium at the Great Gallery, and one from an outcrop near the Hans Flat Ranger Station, approximately twenty miles away aerially. These samples varied in texture, friability, and coloration in order to establish any ranges in composition, especially regarding the amount of hematite contained in them, and the calcite cement holding the sand grains together.


IV: Instrumental Procedure


A Process Instruments PI 200 fiber-optically coupled Raman system was used in the laboratory phase of this study (Figure 2). At the focus, the excitation beam has a diameter (spot size) of 400 nm. This instrument consists of a frequency-stabilized, narrow-linewidth diode laser, a temperature-controlled, optically fast (f/2.0) fiber-coupled spectrograph, a very low noise Andor CCD camera thermoelectrically-cooled to -60 oC, and PROspectTM data acquisition and analysis software. This Raman instrument design offers very high signal levels while retaining high resolution (~ 4 to 5 cm -1 depending upon excitation wavelength). The system includes a fiber-coupled probe with laser band-pass and elastic light rejection (~10 -8) filtering. With the 1200 grooves/mm grating, the spectral coverage is ~ 2100 cm-1 at 785 nm excitation.


A FEI Quanta 600 FEG was used for SEM-EDX work at 10.0 kV and a typical working distance of 10 mm. The company was not forthcoming about the energy density of the beam.


The freshly broken surface on the back of the sample, that held no pigment, was used as the background. As mentioned previously, the Raman spectrum of this surface should be almost pure quartz or silica (SiO2), with additional amounts of cementing material such as calcite and hematite. This background was then used to compare to the pigmented surface in order to identify those components that are unique to the pigment.


As will be shown in the Power Study section later, since no evidence of pigment damage was detected, the beam was focused at a laser power of 300 mW. The illumination or integration time was one second, and four accumulations were averaged together to increase the signal-to-noise ratio, resulting in a total of one hundred separate spectra.


V: Results and Discussion


Scanning Electron Microsopy (SEM) and Energy Dispersive Xray Spectra (EDX)


Figure 3 is an illustration of the surface of the pigment at a magnification of 1000X. This low-resolution scan was taken to search for any evidence of laser damage to the pigment in the form of pitting or “cratering”, and no obvious damage can be observed at this resolution. It was noted however, that at an electron beam energy of 15.0 kV, flakes of pigment were destroyed. The beam energy was therefore reduced to 10.0 kV and no flaking was observed to occur at this lower energy, and all subsequent scans were taken at this energy.


Figure 4 is a low-resolution SEM image and EDX spectrum obtained near the edge of the pigment, where the underlying sand grains intersect the overlying pigment. The EDX spectrum indicates the presence of the elements expected for the minerals reported by Watchman’s XRD study in the form of Fe, O, Si, C, Al, Mg, Na, and Ca (Table 1). At an electron beam energy of 10.0 kV, significant penetration into the subsurface of the pigment is expected, and explains why silicon is observed in later pigment spectra - that don’t overlap the sand grains - in such large quantities.


Figure 5(a) is a high-resolution image and EDX spectrum of the pigment at a magnification of 9000X. The same elemental composition is present as observed in the low-resolution image shown in Figure 4, but the abundance of iron oxide is much higher. Figure 5(b) is a “digital zoom” into the region where the EDX spectrum was obtained, and illustrates the size and morphology of the hematite grains found in the pigment. A reference SEM image of hematite was obtained as corroboration that hematite has been correctly identified (Figure 5(c)). As can be seen from the scale, the hematite particles in the pigment are less than 10 m in size. This could result in peak-shifting of hematite in the Raman spectrum as will be discussed later.


Finally, Figure 6 is of a sandstone grain magnified 2000X. As expected, the principle elements are Si and O, but there are also trace amounts of C, Mg, Ca, K, Fe, and even a smaller amount of S, perhaps due to small amounts of anhydrite which as will be seen, was detected in one of the Navajo Sandstone samples. The large carbon peak is from the CaCO3 cement that holds the Navajo Sandstone grains together.


Rock and Pigment Composition


In order to identify the presence of a particular mineral in the pigment or Navajo Sandstone, reference spectra of known materials are used to match-up peaks. In addition to reference spectra obtained in the lab, there are several excellent Internet databases available free of charge for this purpose. The reference spectra used in this study were obtained from the Internet databases RRUFF.INFO (14) and RASMIN (15). The peaks that were observed in this study are summarized in Table 2.


Figure 7(a) is an illustration of what the raw Raman spectra look like without data processing. The rock spectrum without pigment (bulk rock substrate under the pigment) is displayed in red, and the pigment spectrum is in blue. It’s clear that when the pigment is applied to the rock surface, the background signal from the rock is diminished by a factor of about five. The pigment doesn’t completely block the laser from penetrating into and exciting the background rock, as indicated by peaks and background features that are common to both (e.g. the quartz peak at 464 cm-1 on the left. The technique used to identify these peaks will be explained shortly). This does make it difficult to unequivocally differentiate whether a particular peak is due to either the rock or the pigment, but two data analysis methods have been devised that will be discussed shortly.


Figure 7(b) contains the “normalized” spectra of the data. Since the two spectra significantly differed in magnitude, each spectrum was divided by the largest value in the data, producing a maximum value of 1.0. When plotted together, this has the advantage of allowing easier comparison of the finer spectral details. This is important because the other mineral constituents are producing very small Raman signals relative to the quartz peak, and relative to the hematite peak in the center of the spectrum at around 1300 cm-1.


As can be seen in Figure 7(b), both spectra are superimposed on a large background “hump” centered at approximately 1300 cm-1, with “wavy” ripples on either side. The appearance of this background is surprising because it is very similar to the fluorescent background produced by laser excitation at visible wavelengths, rather than the 785 nm used in this study. It is known that silica (the mineral quartz or SiO2) fluoresces in the visible, but not at the laser excitation wavelength of 785 nm. In this study, it’s believed that “color centers” and the clay mineral illite are the source of fluorescence. Color centers are produced by charge-transfer bands between trace amounts of ions in the sample. Illite is a very fine-grained variety of the mineral muscovite which, as will be shown, was detected in the pigment and rock. Since the pigment spectrum has a diminished background fluorescence, then a large portion of the fluorescence background must originate in the sandstone, with a smaller amount fluorescing in the pigment. From the EDX spectra of the sand grain shown in Figure 6, the color center trace ions could be Fe, Ca, O, or Mg among others.


Figure 8(a) shows the result of removing the slanting offset from each spectrum, followed by scaling the magnitude of the quartz peak at 464 cm-1 in the pigment spectrum to match its magnitude in the rock spectrum. It’s important to note that all spectra were treated this way, including the bulk samples of Navajo Sandstone collected that had no pigment on them. This scaling procedure has the effect of removing the influence of the pigment on the rock background, in effect making it “invisible” to the probing lasing excitation. This is just one way the data were analyzed; it allows comparison of minerals that may be independently present in the rock and the pigment, as well as in both.


The interpretation that will be used here is if a peak is present in both background and pigment, and decreases in the pigment, then it is not a component of the pigment mixture. Similarly, if the peak is larger in the pigment spectrum, then it is present in both. The unfortunate aspect of the operation, however, is to make it impossible to determine if quartz is also in the pigment. But it seems likely that quartz would be in the pigment anyway, if the original artists ground up the pigments in a sandstone metate. And of course, Watchman detected it anyway. So a second data analysis procedure was followed to see if quartz was present in the pigment, and to corroborate the quartz scaling procedure.


This second procedure involved simply ratioing the pigment spectrum by the rock spectrum, removing any portions common to both (Figure 8(b)). As can be seen, the fluorescent background is essentially totally removed, although traces of it can be discerned in places such as at around 2100 cm-1. The final course of action was to removing the curving baseline from the ratioed spectrum via cubic spline interpolation, and the result is illustrated in Figure 8(c).


In Figures 8(d) and (e), the ratioed and baseline-subtracted Raman spectrum in Figure 8(c) is broken up into two wavenumber intervals, 300 cm-1 to 1400 cm-1, and then from 1400 cm-1 to 2400 cm-1. The Raman spectrum is colored brown, and below it is a reference spectrum of hematite taken with the Ahura Scientific Raman spectrometer (blue) since it has a greater range than the RRUFF database spectrum. The identity of various peaks is also given as identified in the following discussion.


This ratioing process illustrates that quartz is indeed present in the pigment, as indicated by the quartz peaks at 365 cm-1 and 465 cm-1, corroborating Watchman’s work. Additionally, the negative-going peak due to orthoclase at 512 cm-1 proves that orthoclase is not present in the pigment. So at least from the perspective of this study, perhaps Watchman wasn’t able to totally remove Navajo Sandstone grains from his pigment sample prior to his XRD analysis.

The problem with cubic-spline baseline subtraction is the selection of zero points is arbitrary, and this can produce false positive as well as false negative peaks. A cubic spline can only be squeezed into so much area before artificial curvature is produced between zero points. Regions where this is thought to occur include 596 cm-1 and 1155 cm-1, based on analysis of the data in Figure 8(c).


The ratioed spectra in Figures 8(d) and (e), in addition to the non-ratioed data in Figure 8(a), were used to identify peaks in the pigment. Data in Figure 8(a) were “zoomed into”, and the results are givevn in Figures 9 through Figure 17. In each of these figures, the spectrum in (a) is for the pigment (shown in blue), while the unpigmented or underlying rock substrate on the opposite side to the pigment is displayed in red. It’s important to differentiate this from bulk samples of Navajo Sandstone that have no pigment associated with them that are shown in (b). For purposes of this discussion, the underlying rock substrate is simply referred to as “rock” or “rock substrate”; the rest will be called Navajo Sandstone or “bulk”. The green spectrum in all figures is the reference spectrum obtained from references 14 and 15 unless otherwise noted, and is an example of the peak-matching technique used to identify unknowns in a Raman spectrum.


The sandstone labeled “Nav SS Red” is for a distinctly red-colored outcrop found near the Hans Flat Ranger Station. The remaining two Navajo Sandstone spectra are from alluvium lying free at the Great Gallery, and are labeled Nav SS GG” and “Nav SS Friable” respectively. The former Great Gallery sandstone sample was firmly cemented together, While the latter was loosely bound and friable. This was done to see if in the Raman spectra the amount of calcite cement was less in the more friable samples.


Figure 9(a) “zooms in” between 350 cm-1 and 550 cm-1, where as discussed previously, the peaks at 345 cm-1 and 465 cm-1 occurring in both spectra are due to quartz. In Figure 9(b) the presence of quartz in all the bulk Navajo Sandstone samples was also verified as expected. The ratioed spectrum in Figure 8(d) proves that quartz is also in the pigment.


Initially it was observed that there was an offset of approximately 10 cm-1 between the quartz reference spectrum and the laboratory spectrum, indicating a wavenumber calibration difference between the two laser systems. This error was only observed for the data taken with the Process Instruments Raman spectrometer, even though the system was calbrated prior to use. Therefore all laboratory spectra, not internet reference spectra, were corrected for this offset by matching up the quartz peaks. This assumes a linear calibration error throughout the scanned spectral range.


The presence of orthoclase at 514 cm-1 in the underlying rock substrate and the bulk samples of Navajo Sandstone is verified in Figure 10. Again, it is not present in the pigment because the magnitude of the Raman peak is less after the quartz scaling procedure outline previously, and because the peak is negative-going in the ratioed spectrum.


In Figure 8(d) and Figure 11(a), there is a characteristic peak at 1008 cm-1 that is due to gypsum, CaSO42H2O, and is unique to the pigment. The bulk Navajo Sandstone samples in Figure 11(b) are free of gypsum, although the friable sample taken at the Great Gallery has a peak at 1019 cm-1 that reference spectra indicate is due to anhydrite instead (violet spectrum). Anhydrite is related to gypsum in that it is anhydrous calcium sulfate.


Figure 12 shows the spectra for calcite, which is known to be the cement for the Navajo Sandstone, and was reported in the pigment by Watchman. Calcite is clearly present in all rock samples, i.e. underlying the pigment and in the Navajo Sandstone. But its presence in the pigment is not certain since the magnitude of its peak is less than or equal to the rock peak. In the ratioed spectrum a small peak at 1085 cm-1 does show up, but it’s not clear if this is due to noise fluctuations, or caused by cubic-spline artifacts. If quartz is showing up in the spectrum because the pigment was ground up in a stone metate, then there should be some calcite cement present also. In the Navajo sandstone samples given in Figure 12(b), the variability in the amount of calcite is probably due to weathering, although microbial effects can not be discounted (7). The smaller quantities observed in the Great Gallery samples reflects the friable nature of the rock, as well.


Several studies have demonstrated the formation of calcium oxalates as the result of lichen growth (16, 17), and this appears to be observed here as well as demonstrated in Figure 8(e) and Figure 13. Whewellite is the monohydrate variety of calcium oxalate, CaC2O4H2O, and the characteristic doublets at 1465 cm-1 and 1490 cm-1 shown in Figure 13(a) are clearly present on the pigment, but not on the underlying rock. In Figure 13(b) whewellite is probably present on the surface of the Navajo Sandstone samples, but because of the poor signal-to-noise ratio, it’s not definite.


The presence of whewellite suggest a possible use in carbon-14 dating, resulting in the at least the time of burial of the detached slab. But until the nature of whewellite deposition is further understood, it’s not recommended to pursue this approach just yet. So far no traces of lichens have been located in SEM images to date, but this will be an ongoing line of research.


In reference 16, J. Russ et al studied oxalate crusts on Peco-style rock art in Texas, and described them in SEM images as botryoidal in appearance, and composed mainly of whewellite with minor amounts of gypsum. There was also an elevated phosphorous concentration, but no phosphorous-containing minerals were detected by XRD. The authors suggest that the whewellite was created by the lichen Aspicilia calcarea.


Gypsum was contained within the crust, which they attributed to an efflorescent origin as the gypsum migrated up from within the rock substrate. This mechanism is not thought to have taken place here, since no gypsum was found in the rock substrate under the pigment, or in bulk samples of Navajo sandstone collected at the Great Gallery. Kaolinite was common at the surface of the oxalate, which they attributed to an aeolian origin.


The presence of whewellite, of course, indicates the pigment was a source of food energy, and has obvious implications regarding the degradation of Barrier Canyon-style rock art over time. However, lichens are not commonly observed on the Navajo Sandstone today, especially in the overhang protecting the Great Gallery figures from precipitation. It is possible that thousands of years ago when the climate was wetter, more lichens grew, leaving behind a thin crust of oxalate.


The clay minerals kaolinite, montmorillonite, jarosite, afwillite, halloysite, and illite (fine-grained muscovite) were searched for in the pigment (14, 15, 18, 19), and the only one conclusively identified was illite (using a standard muscovite sample as reference). There are two characteristic peaks for illite (muscovite) at 411 cm-1 and 700 cm-1 as illustrated in Figures 14(a-c). Because these peaks are so small, and are larger in the Navajo Sandstone spectra, they don’t show up in the ratioed spectrum above the baseline noise. It also needs to be pointed out that correcting the spectrum for the slanting baseline resulted in significant diminishment of the tiny illite shoulder seen at 700 cm-1 in Figure 14(b), so the data for this spectrum only is from Figure 7(b).


As previously mentioned, it is thought that in combination with “color centers”, illite contributes to the background fluorescence of the samples. Standard Raman spectra of illite are unobtainable due to this background fluorescence (15). The Navajo sandstone samples illustrated in Figure 14(c) show a larger amount of illite in it, accounting for its larger background fluorescence.


Since kaolinite was specifically identified in Watchman and Swayze’s work, it was searched for in detail (Figure 15). Unlike their results, however, there is no trace of kaolinite in the pigment or underlying unpigmented rock in the Raman spectrum. In the bulk Navajo Sandstone samples, a small amount appears to be in the Hans Flat sample, but not in the others.


The presence of kaolinite crystals was observed in the SEM results, however (Figure 16). Returning to the study by J. Russ et al, perhaps these crystals weren’t part of the pigment, but also Aeolian in origin. Kaolinite might be missing in the Raman spectra due to several reasons. The first is since the laser spot size was less than 1 m, it’s possible that it simply missed any existing larger crystals. The other reason may be due to the small change in polarizability of the mineral lattice phonons. The lack of the change in polarizability of the vibrating electron cloud between lattice atoms, would greatly diminish the Raman signal. Reference spectra obtained on a laboratory sample seem to bear this conclusion out, with typically poor signal-to-noise ratios encountered relative to other standard reference materials given the same instrumental operating parameters.


The primary mineral observed in the Raman spectra for the Great Gallery pigments is hematite, as also reported by Watchman and Swayze (Figure 8(d-e), Figure 17). The presence of many hematite peaks makes it easy to identify. Hematite appears to a much smaller extent in the underlying rock and in the Navajo Sandstone spectra. However, the hematite peak at 1295 cm-1 observed in the Great Gallery pigments is shifted to a lower wavenumber relative to the reference spectra, and this may be due to the effect of pigment particle size as mentioned previously. Standard reference spectra show hematite absorbing at 1320 cm-1, but there is a downward shift to 1295 cm-1 in the Great Gallery spectrum. A recent paper by Owens and Orosz (20) established that as the particle size of the hematite decreased to the 5-10 nm range, so did the wavenumber the Raman peak. The Scanning Electron Microscopy (SEM) results given in Figure 5 indicate the size of the hematite crystals is in the 1-10 m in size, but not in the nanometer range. So it’s not certain the same effect is being observed here, but it does seem likely.


Additional iron oxides and oxyhydroxides such as maghemite, magnetite, limonite, and goethite were also compared to the pigment spectrum, but none of these minerals matched. As mentioned before, Swazey’s NIR spectra possibly showed maghemite, but this could not be confirmed here.


These iron-containing minerals undergo physical and chemical transformations to hematite as a function of temperature, however. Since it is known that lasers can heat up pigments, and destroy organics, the relationship of heating with laser power will be addressed in the Power Study section that follows.


Power Study


When using a Raman spectrometer, and not knowing conclusively a priori whether there are any organics in the sample, great care must be exercised in minimizing the energy flux onto the sample pigment, as any organic residue present (e.g. from the binder) can easily be photooxidized. Additionally, some inorganic constituents of the pigment may be transformed or degraded as well.


It’s known that high-power diode lasers operating at 785nm can heat inorganic pigments up to 125 oC or more (21). This temperature is not enough to alter the composition of hematite since it is stable at higher temperatures, nor would it convert any magnetite or goethite in the pigment since these phase transitions to hematite occur at 200 oC and 260-280 oC respectively (22, 23)


To avoid pigment damage to any organic binder that may be present, Raman spectra were initially obtained using a low power of 50mW, and a defocused excitation beam distance of 1 cm from the sample. This radiant power corresponds to a total energy delivered to the sample of 50 mW x 10 sec = 500 mJ. Additional spectra were obtained at 100 mW and 300 mW, with various numbers of accumulations per spectrum (signal averaging), and with given integration times. All of these data are summarized in Table 3, and the total energy is plotted versus peak height for the hematite peak at 1295 cm-1 in Figure 18.


As can be seen in Figure 18, the linear plot is almost a two-point fit, but the middle point is on the trend line. The Raman signal is responding proportionally to total laser energy incident on the sample, which wouldn’t happen if pigment was being destroyed. The point at a total energy of 500 mJ (red data point) is lower than the rest clustered at this energy, because the laser beam was defocused for this point. This resulted in a smaller signal, and therefore the point falls below the trend.


Figure 19 consists of the Raman spectra produced at the given total laser energies tabulated in Table 4, and plotted in Figure 18. As expected, there is an increase in signal strength, but the shape of the spectra remain the same, i.e. no peaks are disappearing or appearing with power, and there is no evidence of peak shifting. All told, these observations confirm that unless the laser beam destroys any organics instantly, there are no organics in the pigment, and laser heating does not damage the inorganic components of the pigment either. In addition, the baseline offset for the lowest energy spectra is approximately 9600 counts, corresponding to the y-intercept in Figure 18.


As a final illustration of the effect of heating on the sample, the data sets at total energies of 400 and 1200 mJ are plotted in Figure 20(a-b) versus time, so that the accumulated effect of continuously acquired spectra can be observed. Again, no peak shifting or alteration of the spectra is observed.


Organic Binders


The presence of organics in BCS rock art pigments is not to be expected, because it’s hard to imagine organic compounds surviving for thousands of years without degrading because of microbial activity. That’s not to say, however, that pollen or vegetation detritus couldn’t be found, as again, Watchman was able to extract brush fibers used to apply the original pigments from his sample of the Great Gallery pigment.


More generally, if organic binders are present, they might be best determined by Raman Spectroscopy compared to NIR techniques due to the greater number of “fingerprint” peaks present in Raman spectra, as opposed to single overtone bands commonly found in NIR spectra. Diagnostic Raman bands of commonly employed organic binders collated from the literature are summarized in Table 4, including the oxalates already talked about.


Several representative spectra obtained in this study (using the Raman spectrometer furnished by Ahura Scientific) are shown in Figure 21. In Figure 21(a) the Raman spectrum of olive oil is shown which contains the fatty acids such as palmitic, stearic, and oleic acid, common ingredients in animal fat. Figure 21(b) was obtained of pinyon pine resin collected in the field. Pinyon pine resin contains terpenoids, and is very sticky. As anticipated, none of the peaks illustrated in these spectra appear in the Great Gallery pigment spectrum.


In Swayze’s NIR study though, he reports weak signals (< 1% deep) from potential C-H stretch overtone absorptions, both in the pigment and in bare rock But since the surface was analyzed directly without removing surface residue, it’s hard to conclusively state the organics are entirely from the pigment, particularly in light of Harman’s careful analysis to the contrary. He does comment that this very weak band may be due to H2O in clay minerals instead of C-H from pigment additives. Water has no significant Raman signal, however, so this possibility cannot be further investigated.


In Silver and Wolbers’ investigation, they separated the pigment from the sandstone with a scalpel, and then obtained the infrared spectrum using a related technique to Raman called Attenuated Total Internal Reflection FTIR (ATR-FTIR). They report “natural resins” at 743, 789, 877, and 1421 cm-1. The presence of these organic bands in the pigment is highly doubtful, however. So even though Wolbers was careful to use a scalpel to separate the pigment from the rock, it would appear that contamination of the surface still occurred, possibly during sample collection in the field.


So if there was no organic binder applied, what physical-chemical process has held these pigments in place for thousands of years? Earlier it was mentioned that the cement bonding the Navajo Sandstone was calcium carbonate, but correspondence with University of Utah geologist Marjorie Chan indicated that hematite can act as cement in the Navajo sandstone, too. Since it’s known that brushes were used to apply the pigment (some BCS panels display the texture of the brush as well), and therefore present as a liquid slurry, then the principle mode of binding must be inorganic. This would require a basic pH, however, at least to partially dissolve the silica, which becomes soluble at a pH greater than 9. Covalent bonds between hematite in the pigment and in the underlying rock would be far more permanent than hydrogen-bonding anyway, in such a dry environment. Also, if hydrogen-bonding was responsible for binding, the pigments would suffer from severe flaking over the millenia due to freeze-thaw cycles and the associated volume expansion of ice relative to liquid water.


There is spectroscopic evidence for this hypothesis as shown in Figure 22. The reference spectrum used was for fayalite, a mineral in the olivine family that is composed of iron silicate, Fe2SiO4. This mineral has a distinguishing doublet at 814 cm-1 and 841 cm-1 in the reference spectrum. This doublet appears at 797 cm-1 and 810 cm-1 in both the pigment, underlying rock, and Navajo Sandstone samples, although it is stronger in the pigment. This doublet corresponds to the Fe-O-Si bending vibration, and is decreased in wavenumber (frequency) relative to the reference. It is hypothesized that this lowering of the vibrational energy is due to incomplete bonding between pigment and sandstone substrate, resulting in the weakening of the Fe-O-Si bond.


VI: Summary and Conclusions


Raman spectroscopy was successfully used to determine the pigment composition of a sample of rock art from the Great Gallery. The spectra were obtained in a noncontact, nondestructive manner, with no damage to the underlying pigment due to the laser power.


Hematite was confirmed as the primary pigment component, along with quartz and gypsum. Kaolinite was not found spectroscopically, but it was observed in the SEM images. Calcite was possibly present, but close to the background levels in the spectra. Orthoclase was only found in the underlying rock and bulk Navajo Sandstone samples. Trace amounts of the clay mineral illite were detected, and together with “color centers” is thought to be responsible for the background fluorescence of the sample.


No organic binders were found in the Raman spectrum, in agreement with the GC-MS results. Instead, the presence of a Fe-O-Si bending frequency in the Raman spectrum indicates the pigment bonded inorganically to the rock substrate by linking the silica in the substrate to the hematite in the pigment. The mineral whewellite, CaC2O4H2O, was detected, indicating the presence of lichens in the past. This suggests that at the very least the time of burial of the detached slab could be carbon-14 dated.



VII: Acknowledgements


Prof. Joel M. Harris at the University of Utah is acknowledged for donating the use of the Process Instruments PI 200 Raman Spectrometer system. The author also extends acknowledgements to Dr. Nancy Simon, whose literature research and rock art expertise were very helpful.




FINAL INSTRUCTIONS NO 1 THANK YOU FOR
CONCLUSIONES Y RECOMENDACIONES FINALES EN LOS HOSPITALES
EU FP6 PROJECT IMPROVE FINAL CONFERENCE DUBROVNIK CROATIA


Tags: analysis of, careful analysis, report, raman, pigment, spectroscopic, analysis, final