Identification of Ivory Using a Handheld XRF Spectrometer

Ivory and Look-Alike Specimens

Ivory items from eight different species frequently encountered in the ivory trade were selected for analysis [6, 7]. Tooth and tusk specimens were selected for African and Asian elephants (Loxodonta africana, Elephas maximus), mammoth (Mammuthus sp.), walrus (Odobenus rosmarus), sperm whale (Physeter macrocephalus), narwhal (Monodon monoceros), warthog (Phacocherus sp.) and hippopotamus (Hippopotamus amphibius). African and Asian elephant specimens were pooled in one group. Analyses was performed on the cementum (n=20) and dentin (n=20) for each taxa on specimens that were either intact teeth or tusks or on carved items. Our assumption is that each specimen originated from a different individual. Additionally, 20 bone samples from 15 different species were selected for XRF analysis including tiger (Panthera tigris), lion (Pantera leo), gorilla (Gorilla gorilla), snow leopard (Panthera unica), hyena (Hyaena hyaena), leopard (Panthera pardus), wolf (Canis lupus), New Guinea singing dog (Canis hallstromi), California sea lion (Zalophus californianus), lynx (Lynx rufus), Stellar sea lion (Eumetopias jubatus), black colobus (Colobus satanas), American black bear (Ursus americanus), Asian black bear (Ursus tibetanus), and North American river otter (Lutrcanadenis). All the specimens used in this study are listed in the Microsoft Excel file labeled “Supplemental Data”.

To determine intra-specimen variability, 20 scans were collected from a single elephant and mammoth tusk, at various dentin locations.

Different biological and manufactured materials that could be mistaken for ivory (i.e. ivory look-alikes) were included in this study. Natural materials consisted of vegetable ivory from tagua nuts (Phytelephus macrocarpa) and alabaster (CaSO4 · 2H2O). Synthetic materials included Resin-Ivory+STM, Celluloid, Casein based Ivorina from 3 different manufacturers, Vigopas, Dekorit, Ivorite, Epoxy resin composite (Epoxy + ivory dust), and Alabrite from 2 different sources. For each material, 20 spectra were measured, except for Vigopas, Dekorit, Ivorite, Epoxy composite, and the Alabrite casts, where 3 replicas were collected for each material, due to the small quantity of specimens available. (Table 1) shows a summary of the composition and main distributor of the synthetic ivory substitutes used in this study.

All items selected for this study were part of the Reference Material Collection of the National Fish and Wildlife Service Forensic Laboratory.

XRF Device

XRF analyses of the ivory and ivory look-alike items was conducted using a SciAps X- 50 handheld XRF Analyzer (SciAps Inc., Boston, MA). The spectrometer measured the elemental composition of each specimen for 10 seconds. The percentage of light elements (LE) corresponded to the sum of the elements lighter than calcium (Ca, atomic number 20). Prior to each scanning session, the device was calibrated using the included Geo-mining protocol. Measurements were taken on the flattest parts of each sample to optimize the contact between the surface of the item and the scanning window.

Data Analysis

Elemental composition values were transcribed to a Microsoft Excel data spreadsheet. The raw data are shown in Microsoft Excel file labeled “Supplemental Data”.

To evaluate the quantitative variation of the XRF relative percentage, the mean (µ) and standard deviations (σ) were calculated from 20 measurements obtained from a single elephant and mammoth tusk.

To evaluate if dentin and cementum had the same elemental composition, the relative percentage for 20 unique specimens per species (i.e., each specimen was from a different individual) were measured and the mean (µ) and standard deviation (σ) were calculated. To evaluate if the percentage of LE and Ca was significantly different between cementum and dentin, t-tests were performed, using the values obtained from elephant and mammoth samples. Statistical analysis was conducted, and results were plotted using the software JASP.

A Linear Discriminant analysis (LDA) was firstly conducted to show the separation between ivory items and non-ivory items based on elemental composition. Since LE and Ca relative percentage are much higher than other elements, a Z-score normalization was applied. A second LDA was performed in the same way, to investigate whether ivory species and bone could be separated, based on their elemental composition. The software PAST (Paleontological statistics, Version 4.08) was used for this purpose.

Blind Test Validation Samples

Six items not used in the database development were selected to be used as validation samples. Three replicate XRF spectra were collected for each item and the mean (µ) and standard deviations (σ) were calculated. Their relative percentage results were then analyzed to determine whether they were ivory or a substitute.

XRF Reproducibility

The intra-variation measurements of dentin in an elephant and a mammoth tusk were determined by collecting 20 spectra from a single of each taxa.

Table 2 shows that the elemental composition varied little within a single specimen; summary values are shown and individual measurement for each data collection is shown in supplemental (Table 1). The standard deviations for LE and Ca did not exceed 1.1% for elephant and 1.5% for mammoth which demonstrates the accurate semi-quantitative reproducibility of the handheld XRF.

Tissue Type

On unmodified teeth and tusks, XRF measurements are mostly conducted on the exterior cementum tissue, since it forms the outer layer of the tusk. In contrast, most carved items consist of dentin, which is the tissue beneath the cementum layer. In museums, art galleries or in trade, both raw and carved items can be encountered. Therefore, it is of importance to determine whether the elemental composition of both tissues is significantly different. Table 3 shows the summary of mean and standard deviation of the elemental composition measured in cementum (Cem) and dentin (Den) from all the species used in this study. The t-tests analysis conducted on the relative percentage from elephant showed that the percentage of LE was significantly higher in dentin, whereas cementum contained significantly more Ca (p-value < 0,001) (Figure 1).

Click to enlarge

Relative Percentage Results of Ivory and Look- Alike Specimens

Figures 2 & 3 show the percentage of LE and Ca of all ivory species and ivory substitutes. The raw data is shown in the supplemental Microsoft Excel spreadsheet labeled “Ivory and Look-alikes Supplement File” The XRF measurement of ivory and bone samples showed that these two tissue types had a similar elemental composition. All ivory samples had a percentage of LE ranging from 50.3% to 75.5 %, and Ca from 24.5 % to 49.5% regardless of the species or tissue. As expected, Ca was the most abundant element detected while other elements detected at trace levels were Fe, Ni, Zn, Sr, and occasionally Cu. However, Cu, when detected, was only present in small proportions and with a large variability within species. Fe, Zn, and Sr were generally more abundant in cementum than in dentin and presented a relatively large intra- and inter- variability. The proportions of Fe varied greatly between samples, especially in cementum, giving sometimes relatively high values. These results are possibly due to the presence of blood contamination on the surface of the tusks or teeth. The relative percentage of Ni showed relatively little variability between species. Mn was detected in cementum for some hippopotamus specimens, which could also be due to surface contaminations. The percentage of Ca detected was overall higher in comparison to values reported by other studies [1, 5]. Non-ivory items presented a wider variability in elemental composition and those results are shown in Table 4.

Click to enlarge

Click to enlarge

Analysis of the look-alike materials revealed differences in the semiquantitative levels of LE% and Ca%. For example, Celluloid, Dekorit, Ivorina, Ivorite, Resin-Ivory+S, vegetable ivory from tagua nut, and Vigopas all had a high concentration of LE (> 78.9 %), and a low percentage of Ca (< 18.8 %), which was not detected at all in Dekorit, Resin-Ivory+S, vegetable ivory, and Vigopas. On the other hand, the average percentage of Ca measured in Alabaster, Alabrite, and Epoxy resin was on average higher than in Ivory. Additionally, all synthetic materials except Alabrite showed relatively high percentages of Ti. Therefore, the semi quantitative concentration of LE and Ca in the described range by XRF analyses confirms that a particular material is either bone or ivory regardless if the analysis was of the dentin or cementum. XRF detection of other elements (Table 3) or concentrations outside of the biological range (Figures 2 & 3) for ivory infers that material is an ivory substitute.

The elemental composition of bones was similar to ivory and could not be differentiated, regardless of the species and tissue. A t-test showed no significant statistical difference (p- value = 0.813).

The LDA conducted on all the dataset is shown in Figure 4. The model shows good separation between ivory and ivory look-alike items with a performance index of 92.45 %. All the ivory and bone specimen were included in one group, since the objective is to discriminate between genuine ivory and look-alikes.

A second LDA analysis was performed on the ivory and bones to determine if handheld XRF data can distinguish among species and the graphical result is shown in Figure 5. The variables selected here were LE, Ca, Ni, Zn, and Sr. Fe was eliminated due to the significant variability observed within species. The performance index in this case is 45.67 %. The model shows that the species of origin of an ivory item cannot be identified based on its elemental composition using a SciAps X-50 handheld XRF spectrometer.

Click to enlarge

Click to enlarge

Blind Test Validation Samples

XRF analyses was conducted on 6 blind validation samples. Validation samples 3 and 6 were easily excluded as ivory based on the proportion of LE, which was above 90%.

The proportion of Zn was also significantly higher. The other four validation samples had a similar elemental composition to ivory, and therefore, could not be excluded as ivory. Table 5 shows the elemental composition of the validation samples.

Among the validation samples, samples 3 and 6 were excluded as ivory due to the LE and Ca proportions which fall outside of the range observed for ivory and bone materials. These carvings were made out of celluloid. The values measured of these unknown samples are consistent with results obtained for celluloid analyzed in the first part of this study (paragraph 3.3), and allowed correct classification as look-alikes. The other samples couldn’t be excluded as ivory, as the proportions of LE and Ca fall into the range established for ivory and bone specimen. Samples 1 and 2 were African elephant dentin, specimen 4 was Asian elephant dentin and sample 5 was whale cementum. These objects had been identified via traditional morphological analysis and are part of the specimen collection at the National Fish and Wildlife Forensic Laboratory.

XRF Criteria and Tissue Type

Handheld XRF analysis is a quick in-situ screening tool. The experimental range determined for ivory aids the traditional morphological analysis in identifying materials of unknown origin. In order to obtain accurate results, the item scanned must be larger than the scanning window (diameter of 9.5 mm). Carved objects can have curves and intricate shapes that can make the analysis more difficult. When the XRF device was not in immediate proximity to the object occasional spurious data was reported; if an elemental concentration seems inconsistent, a second scanning is advised.

It must be noted that XRF instruments from different manufacturers have different instrument settings, and also have different calibration parameters based on the material being tested. Preset instrumental parameters are optimized for the elements of interest. Because of this, the presence and / or concentrations of certain elements vary depending on calibration or instrument manufacturer. Moreover, single spectrometer instruments, like the SciAps X-50 used in this study, can only detect a defined number of elements, which are determined by the manufacturer. This set of elements can be customized according to the desired application. As a result, different instruments, even from the same manufacturer and model, will not detect the same elements, if programmed differently. Therefore, the results presented in the present study cannot be compared to results obtained using a different instrument, or with different settings.

Scientists conducting XRF measurements must consider the relative percentage variation between dentin and cementum to avoid spurious conclusions. The distinction between both tissue types has not been addressed in previous studies of ivory identification using handheld XRF spectrometers. We have shown a significant statistical difference between the concentration of LE and Ca in the dentin and cementum of the elephant analysis. Nevertheless, even though LE and Ca percentages varied between dentin and cementum, the overall semi-quantitative results for each species is relatively similar, and different from ivory look- alikes. Therefore, in the absence of morphological characters, it is feasible to determine a material is from an ivory source or an ivory look-alike. Ivory objects have a range of LE 50.3% – 75.5% and Ca 24.5% and 49.5%, whereas ivory look-alike materials fall outside this range. This is useful to infer the presence of ivory in the analysis of small, carved ivory objects when morphological features are absent. The experimental range was corroborated by the analysis of validation blind samples, which were correctly identified as ivory or look- alike materials.

There are synthetic materials manufactured is such a manner to resemble the presence of the proboscidean diagnostic Schreger lines [7]. For example, Casein based Ivorina and Resin-Ivory+S (Resin-Ivory+STM) show lines that imitate Schreger lines. Based on our empirical studies, XRF analysis is a good technique to apply to delicate art objects that appear to show Schreger lines, especially when the suspected material cannot be removed from the art. As the data shown in the Figures 2 and Figures 3, handheld XRF analysis easily excludes all synthetic ivory look-alikes even in the presence of imitation Schreger lines.

In conclusion, the XRF device used in this study can be used to distinguish between ivory and non-ivory items. If the analysis of a suspected item falls within the diagnostic range (LE 50.3% – 75.5% and Ca 24.5% and 49.5%) then a robust inference can be made that the material is ivory or bone, regardless if the tissue sampled was cementum or dentin.

Additionally, XRF analysis allowed to quickly eliminate most non-ivory materials (except bone). The data produced by this research also indicates that the taxonomic species source cannot be determined using the criteria defined in the manuscript [8, 9, 10, 11, 12, 13, 14, 15].