With due respect thus paid to events that have overtaken us, I return to my sort-of-scheduled programming here at A Corner of Tenth-Century Europe and also to the subject of Byzantine coinage, from which it seems I will probably never entirely escape. And why indeed would one want to? But I can bring one thing to a close with this post, which is my reports on the experiments that my collaborators and I did on the All That Glitters project analysing Byzantine gold coins by X-ray Fluorescence of which you have by now heard so much, at least until we actually publish properly on it. So here is the last post on the theme for the time being.1
And therefore last chance to re-use this photograph of the S8TIGER WD-XRF machine in the Department of Chemistry at the University of Birmingham!
I’ve done so many of these posts now that summarising the experiments in any detail would be quite difficult as well as tedious, so for the purposes of this post, let’s just say, we’d gone into the project hoping to say something about changes in the trace elements present in the coins that might tell us something about changing metal sources and minting practices, and we’d found that for two reasons that wasn’t really possible, although we did still find some stuff out doing it. The first and more historical of those reasons why not was simply the nature of the Byzantine tax system, which persistently called in coins from across the Empire, melted them down and redistributed their metal centrally to the mints for striking, thus ineluctably mixing all the different mints’ practices together; the only place we could really see compositional difference was provinces that were falling off the Empire, and even then interaction was usually sufficient to keep things mixed up. But the other reason was that the detection machinery, be it never so sophisticated, couldn’t really tell us what we wanted to know, and that also for two reasons, one being because of invisible surface deposits from the soil that we couldn’t properly see through with the X-rays (and couldn’t safely remove very effectively), and the other being simple and frustrating variation in results.
At what had been supposed to be the end of the project, therefore, because of the various constraints and inefficiencies of getting the coins safely to the analysis machinery and back, we had unspent money left in our grant budget, and so I thought—I think this was me, but if not, I’m sorry to either Rebecca Darley or Maria Vrij for stealing their credit here—that one useful thing we could do with half a day was get some kind of baseline figure for how bad that variability was. So on 16th February 2016 we did a very simple experiment. We took one coin, put it in a sample cup and then without touching it, moving it or changing it in any way beyond what the automatic handling in the spectrometer put it through, ran it through exactly the same test five times, and then turned it over in the cup and did that again for the reverse side, giving us ten runs on the same object in which there was literally nothing more that we could have done to reduce variation.
Apparently I wasn’t taking the security photographs that day, so I cannot show you that coin in its sample cup, but here it is in shiny catalogue image; it is a gold-ish tremissis of Emperors Leo III and Constantine V probably struck in Sicily or Italy, as we demonstrated in the last one of these posts, between 717 and 741, Barber Institute of Fine Arts B4542
The results were not encouraging. Admittedly, in terms of gold content they were not too far apart, ranging from 85.26% to 87.21%, an error margin of only about 3%, but even that is 2% raw difference in apparent metal content. People have founded theories about currency alteration on the basis of disparities like that, so if you’re getting them between measurements of a single coin that’s a problem. But the less present elements had similar amounts of variation: silver 7.63% to 8.84%, copper 1.78% to 2.39%, iron 0.15% to 1.04%, aluminium 0.38% to 0.67%, magnesium 0.13% to 0.67%, and then a host of other elements that one didn’t even see in some or most tests. Again, these margins may not seem like a lot, less than a per cent in some cases, but those less-than-per-cent margins are in some cases more than the total percentage of a metal in question, which meant that the error margins we were seeing were mathematically huge, in the order of 300% to 500%. Basically, no respectable scientist would trust such figures, because they could have no confidence that they would be able to reproduce them, and fair enough, because here we were trying to do so and more or less failing.
So what did all this tell us? One gloomy conclusion, that for all we’d hoped to find differently, XRF probably still isn’t a workable way of analysing trace elements in coins, we had more or less already reached, but this let us actually put numbers on why not, which is worth something. I’ve since looked at quite a lot of papers using XRF analyses on coins, and I’ve found only one that used an average of several experiments as their working figure, and that was from 1983 (and was by none other than the late lamented Michael Metcalf, and he was dealing with variations of over 20% depending on what he’d done to Philip Grierson’s coins to get those results, which we know because he actually said so in his write-up).2 He wasn’t even a metallurgist! And he presumably also wasn’t paying for machine time, which is the basic reason that I guess people don’t otherwise do this. But it is, one might say, a little embarrassing for the subdiscipline. Still, I’m not sure that even an average figure from our tests would be very safe to use. How many tests would one need to run on each object to make safe a 300% error margin? What if one of those tests increased that margin? In general, I think that even the best XRF machinery we can get just can’t give accurate figures for small-percentage composition elements, even if it probably still has some application for the big-ticket components. It’s not the conclusion we’d aimed for but when we can get anyone to publish such a negative finding it may not be without value.3 And thus endeth, for now, the sequence.
1. Of course, we have actually published on the project a tiny bit, in the form of Rebecca Darley, “All that glitters…: the Byzantine gold solidus, c. 300-1092″, in Maria Caccamo Caltabiano (ed.), XV International Numismatic Congress, Taormina 2015: Proceedings (Rome 2017), II, pp. 982-985, but that was actually written and given before we’d done these final experiments.
2. That being D. M. Metcalf, “Interpreting the Alloy of the Merovingian Silver Coinage” in C. N. L. Brooke, B. H. I. H. Stewart, J. G. Pollard and T. R. Volk (eds), Studies in Numismatic Method Presented to Philip Grierson (Cambridge, 1983), pp. 113–126, out of 16 studies I know of from 1966 to 2019; I’m sure there are more, though, and if you feel I’ve missed an important one it would be really useful to know!
3. The main reason that we haven’t yet done more on publication of these experiments, even three years down the line, I should explain, is professional mobility. Even in the course of the project, every single project member changed jobs and only two of them even stayed within the same company/institution. Since then several more of us have moved again. Of course, our new employers all hired us for our own individual qualities and while some of them might like us to do this kind of collaborative inter-disciplinary research, they would prefer to have been part of it, so that only those of us who remain in Birmingham have any immediate professional interest in making this part of our workload. We will publish something on it, because we spent money on the assurance that we would, but it will be when one of us needs it more than whatever else we’re supposed to be working on, and the path to that isn’t yet clear.