2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 300-4
Presentation Time: 9:00 AM-6:30 PM

DETERMINATION OF DIAMOND PROVENANCE IS POSSIBLE WITH MULTIVARIATE ANALYSIS OF LIBS SPECTRA


MCMANUS, Catherine E., Materialytics, LLC, P.O. Box 10988, Killeen, TX 76547, DOWE, James, Materialytics, PO Box 10126, Killeen, TX 76547 and MCMILLAN, Nancy J., Geological Sciences, New Mexico State University, Box 30001 MSC 3AB, Las Cruces, NM 88003, c.mcmanus@materialytics.com

The ability to accurately determine the provenance of gem diamonds impacts economic, political, and national security arenas. Currently, provenance determinations rely on: 1) gemological and mineralogical features of stones, such as spectroscopic measurements, geochemistry, and inclusions, and 2) certification and tracking of individual stones through the Kimberly Process Certificate Scheme. Unfortunately, during cutting and polishing, many gemological features are obliterated and tracking individual stones through the chain of custody can be difficult. This study resulted in a highly successful method for determining provenance of cut diamonds from information in the stone itself.

A set of 30 cut diamonds from each of ten controlled localities and one set of 30 synthetic diamonds were analyzed by Laser-Induced Breakdown Spectroscopy (LIBS). The sample set (330 total diamonds) includes both kimberlite and placer diamonds from five countries and five different cratons. LIBS acquires the atomic emission spectra released from a material during laser ablation. The spectra contain information from nearly every element in the periodic table, and thus are unique chemical, or quantagenetic, signatures of the material. Spectra were analyzed using a Bayesian statistical method that compares groups of samples defined by the reported locations of the stones to clusters of samples defined by spectral similarity. Ideally, each spectral cluster coincides with a group of stones. The spectrum of each sample is compared to a set of reference spectra from each group to determine the probable provenance of the sample. The correlation between groups and clusters was excellent, with average accuracy of 98%, suggesting that diamonds from each location are spectrally similar to each other and distinct from those from other locations. This is true even for diamonds from kimberlites in close proximity to each other. Synthetic diamonds are easily distinguished from natural diamonds (100% success). Some groups of diamonds in the study are more heterogeneous than others. For instance, a placer group has five recognizable spectrally-defined sub-clusters. This work demonstrates that diamond provenance can be determined at a high level of confidence on individual cut gemstones.