The Victoria Cross (VC), the UK's highest military award for bravery, has been awarded 1358 times since 1856. Its rarity and historical significance make it a highly sought-after item among collectors, leading to concerns about authenticity. A long-standing, yet unsubstantiated, tradition claims that all VCs were cast from bronze recovered from Russian guns captured at Sebastopol during the Crimean War. This research aims to investigate the source material of VCs and assess the potential of XRF spectrometry, a non-destructive technique, to determine the authenticity of questionable medals. The study's importance lies in its ability to provide a scientific method for authenticating these historically significant artifacts, addressing the existing controversies surrounding their origin and provenance.

Previous studies have examined VCs using XRF, primarily focusing on smaller samples and often not publicly releasing their findings. The Royal Armouries, for instance, conducted XRF analysis in the 1980s and 90s, successfully identifying authentic medals from forgeries. Other analyses in Australia and New Zealand have confirmed that VCs predominantly consist of copper and zinc, although the precise ratios have varied between medals. However, a comprehensive study examining a large number of VCs to investigate compositional changes over time and to characterize medals of questionable provenance was lacking, which this research aimed to address. Existing literature predominantly relies on visual examination and subjective expert judgment for the authentication of earlier VCs, lacking the objective and non-destructive approach provided by XRF analysis.

This study utilized XRF data from two sources: (1) 50 VCs analyzed by Marriott between 2016 and 2018 using a Bruker Tracer III PXRF spectrometer, spanning medals from various conflicts; and (2) data from the Royal Armouries' XRF investigations of the 1980s and 1990s, comprising 71 VCs (excluding three copies/duplicates). The data encompassed medals from different eras, including pre- and post-1914 medals, source metal samples from the Ministry of Defence (MoD), Hancocks (the medal suppliers), and the 'VC Guns' displayed at the Royal Artillery Museum. The measurement consistency between the two datasets was assessed using Pearson's correlation coefficient. Gaussian finite mixture modeling and hierarchical clustering were employed to partition the medals based on their elemental composition (iron, copper, zinc, lead, tin, and arsenic). Principal component analysis (PCA) was performed to visualize the relationships between the medals and potential source materials. To assess the authenticity of questionable medals, a weighted squared Euclidean distance metric was calculated, comparing the composition of each medal to the median composition of medals from the same year, effectively mitigating batch effects and improving resolution.

The analysis of the combined dataset (110 VCs) revealed a clear compositional difference between VCs issued before and after 1914. Gaussian mixture modeling identified four distinct clusters, with two primarily containing pre-1914 VCs and one primarily containing post-1914 VCs. The remaining cluster had a mix of VCs from both eras. These findings strongly suggest a change in the source material around 1914. PCA indicated that the MoD's Donnington metal block is compositionally similar to post-1914 VCs, supporting its role as a source material. However, the 'VC Guns' from Woolwich showed dissimilar compositions, raising questions about their role as a source material. The composition of bronze cannon from Dutch shipwrecks surprisingly showed greater similarity to pre-1914 VCs than the Woolwich cannon. The study also examined the authenticity of three questionable VCs. The medal found in the River Thames in 2015 showed an exceptionally close compositional match to other 1854 VCs, suggesting authenticity. The medal awarded to Lieutenant John Chard was confirmed to closely match other early VCs, supporting its authenticity. Similarly, the medal awarded to Lt Harry Prendergast, despite its worn appearance, demonstrated high compositional similarity to other 1857 VCs. The VC awarded to Captain Herbert Clogstoun presented an unusual composition, prompting further investigation that revealed the existence of a replacement VC issued in 1938. Further analysis comparing the Clogstoun VC to other medals from both the 1859 and 1938 periods may shed further light on the provenance of the medal. The developed weighted squared Euclidean distance metric proved effective in evaluating the compositional similarity of questioned medals to other VCs of a given year, facilitating authenticity assessment.

This study's findings strongly support the hypothesis of a change in the source metal for VCs around 1914. The compositional differences between pre- and post-1914 medals, coupled with the compositional similarity between post-1914 VCs and the Donnington metal block, provide compelling evidence for this assertion. The unexpectedly close match between the Dutch shipwreck cannon and pre-1914 VCs challenges the long-held belief that the Woolwich 'VC Guns' were the primary source for all VCs. The analysis of questionable VCs demonstrates the power of XRF in verifying authenticity, providing a non-destructive and objective method that complements existing visual and historical approaches. This research highlights the potential of XRF analysis combined with robust statistical modeling for studying artifacts whose history and authenticity are debated. Future research could focus on expanding the dataset to further refine the clustering and distance metrics, potentially incorporating machine learning techniques to improve the accuracy of authenticity assessments. Examining the impact of casting techniques and other factors on compositional variability may also provide further insights into VC production.

This research successfully applied XRF spectrometry in conjunction with advanced statistical methods to investigate the source materials and authenticity of Victoria Cross medals. The study confirmed a change in source metal around 1914, and the developed distance metric proved effective in evaluating the authenticity of questionable medals. Future work should aim to expand the dataset and incorporate machine learning to improve the accuracy and precision of authenticity assessments, further exploring the influence of casting methods and other factors.

While the study analyzed a substantial number of VCs (110), it still represents a relatively small proportion of the total number awarded (1358). The analysis relies on the availability of XRF data; some medals may be inaccessible for analysis. The developed distance metric's effectiveness is influenced by the precision of compositional measurements and the size of the comparison groups. In this study, the compositional analysis has been used to suggest the authenticity of several medals. In several cases, the authenticity of these medals has been subsequently confirmed; however, the study did not include an analysis that could confirm if this was a result of the approach suggested or coincidence.