 Mae'r gerdydd o leisir o blaesir ymlaesion, a'r FFMS, ysgronch ym Mastaf analysus, fally o'r strawnch i'n tufyn am. Wydd wedi bod yn fwy oherwydd, ychydig. Felly mae'r bryhau? Felly, bryhau. Bryhau. Felly, ychydig. Felly, yn meddwl. Felly, yn meddwl. Felly, dwi'n meddwl. Llywodraeth bryhau. Rwy'n meddwl. Efallai, roeddai. Rwy'n meddwl. I'm a PhD student at Durham University ond ar y brifysgol Llywodraeth, ac rwyf amser am yr emeysydd ynglyn â'r cyfnod hynny'r cyfnod yn phirpwyr. Mae yna sy'n gwybod bod y cyfnod erbyn o'r llwyddoedd yn cael ei defnyddio'r ffordd ac yn cael ei dyfanyddio'r cyfnod sy'n ffordd. Rwy'r cyfnod yw'r cyfnod yn ddod y cyfnod yn fwy o'r cyfnod yw'r cyfnod yw'r cyfnod for strontium isotope analysis and here we would be interested in the ratio of 87 to 86. There is micro drilling with Tim's but also you can use solution plasma spectrometry or laser ablation plasma spectrometry. So micro drilling and then either Tim's or solution plasma is the current gold standard and produces reproducibly accurate and precise results. However it's considered to be quite expensive and time consuming and can take several weeks to run 15 to 20 samples. It also requires clean of oratory facilities which not all places will have. Conversely laser ablation is being perceived as more rapid as it takes less than two minutes per sample and that includes the wash out time and moving from sample to sample within the enamel but post analysis data processing takes a considerable amount of time and can take several weeks or months afterwards to actually process the data before you can interpret it. It doesn't require column chemistry so it doesn't require clean laboratory or technicians in order to do this and it's perceived as less destructive and while this is true in relation to bulk sampling where you take a large sample and look at the average of the two the micro drilling samples are just as small as the laser ablation samples. Isobaric interferences and micro drilling are removed through this column chemistry which means that in the laser because we don't do it we don't remove isobaric interferences and isobaric interference is an interference which sits on the same mass as the element that you're interested in so particularly the problem with laser ablation is the interferences on mass 87 which artificially raise the 87 86 value and alter your interpretation of the data. So interference on mass 87 most papers who have published laser ablation data report seeing are higher than expected strontium 87 86 values and this is either compared to what they were expecting or the published Tim's values for samples of what you've done or the Tim's repeats that they did of their own data. Some papers consider this to be negligible with an uncertainty and don't affect their interpretation while others attribute it consider it to be significant and attribute it to either calcium phosphate or argon phosphate both of which have a mass 87 and both of which are potentially produced during the ablation of carbon 8 or well high carbon not high carbon sorry calcium high calcium substances such as tooth numb and you can't remove it because you're not doing any chemistry and everything that is ablated goes straight into the mass spec. Some authors don't detect it at all and interferences where they do appear to be machine specific with some machines producing it all the time to varying degrees some machines producing it on some days but not others and some machines appearing to not produce any interference at all and different groups use a variety of different methods to negate this interference from different tuning conditions so how you set up your mass spec to how you process the data afterwards or apply calibration equations. So what I did here was I used the two mass spectrometers at the NIGL or two of the mass spectrometers at the NIGL suite the thermonatune plus and the new plasma HR which is a plasma new plasma one but has a new plasma two interface and we coupled it with two of the three laser systems they have there and the selection of known concentration but varying strontium concentration varying isotope ratio geo and bio appetites including a variety of different enamel from different species ranging from 103 ppm strontium concentration up to 656 ppm and geo appetites ranging from 400 ppm strontium concentration to around 4300. So in an ideal world all of our data would sit along the slide so this is comparing the voltage of 88 so the most abundance strontium isotope to the accuracy or the bias of the ratio so the measured ratio that we got in our experiments to the Tim's value which I've either published or the ones that we calculated ourselves using Tim's. All of the all of the same samples have the same same people so all of the ones that a derango have which is a geoapposite have a little circle and all of the samples from the same run have the same colour but it doesn't matter too much about the data we just these are our low strontium enamel samples the squares which are quite far away from the line and these are our higher strontium appetites which are a bit close to the line but even these ones which are our sort of high strontium enamel samples aren't sitting exactly on the line one and I don't have any error bars on this graph because it would make the next run really difficult to read. So this is the Neptune and the previous slide was the new so here the values do sit a little bit closer to the line but they also have a higher concentrate no a higher strontium 88 voltage and potentially the reason they may sit closer to the line is because we don't have any of these higher they appear to be closer we don't have any of these higher signal samples however this sample is the same sample as I go back inside this one over here so we're getting a huge difference between our beam size of the same sample just in different runs at different settings. We're also seeing the new Neptune appears to be behaving well this Neptune appears to be having slightly better than this new because we get we don't get any of the really really far away from I mean this is still not good values but they're not quite as far away from the other so we also tested some of the different tuning conditions we tested 15 different tuning setups and this is the two these two previous slides of a combination of both and one of them that's been published and suggested to be a good method for reducing this interference is low oxide tuning so we tuned for low oxides and maximum sensitivity on both the new and the Neptune so this is the new data um these are are really rare samples so normally low oxides are thought to make better produce better data than tuning for maximum sensitivity but in this instance these are our low oxide samples and these are our maximum sensitivity samples so actually tuning for low oxides made the values worse probably because they reduced the beam size and at lower beam size there is more potential for interfering but if we zoom into this bit over here we also see that the symbols of the same one so the closed symbols are the low oxide and the open symbols are the maximum sensitivity that they're not pushing them any further up sort of away from the line one with the bias but they are just pulling them back in the beat in terms of the beam intensity and the sensitivity and we did the same for the Neptune but in this case it appears like tuning for low oxides did improve um the results and create a sort of less bias more accurate value with this one the open symbol is being the low oxide and the closed symbol is being the maximum sensitivity the maximum sensitivity one here is further up and away from the line one which would be what we would want and we do have the error bars but if you also notice the error bars on here are really quite large but so what so we have we have our sort of 12 of the tuning conditions that worked because we have three tuning conditions that produce no data at all and all of the same symbols are the same sample so it kind of shows that we have a big range of the values that we got from this sample which is exactly the same as the sample down here which is our 103 ppm so our low strontium sample but actually 103 ppm isn't that low so here it appears that run three and run seven on the Neptune the different tuning conditions gave the best results but interestingly session number two and session number seven were the same tuning conditions six months apart but they still give different data but what does this mean for archaeological samples so these are all of my samples that we ran um the just but we've split them by bio appetites and geo appetites so the reds are the geo appetites and the blue are the bio appetites and if we take the evans 2010 map and apply the same colour scheme and assume one is 709 they're all different this is what it would look like so you can see that the interpret any interpretations you took from these numbers would be very different from the sample from their other values which sit down here in the sort of 709 range as opposed to the 713 range so we clearly can't use data that like this one up here but we wouldn't necessarily know unless we tested all of the tuning conditions what was going to create good data or what was going to create bad data or less good data so we can also apply correction regressions and this is another thing that has been proposed by a variety of authors as to be a good way to create correct for this interference so i tried to create regressions for all of my data sets and these were the best four which all happened to be on the next tune but you can still see that the r square values of these regression line not always very good for applying a regression to correct the data so this one actually 13 we have a nice r square value of 0.9 but this one up here is an r square value of 0.5 which really doesn't sort of doesn't instill confidence and these are all based on the exception of this one because i have to take out this which is a derango which is a geo appetite which has very high rare earths which cause interference but these are all based on the average of apart from that one which is based on all the spots apart from derango um but we also have these runs which runs for which doesn't produce any regression at all so even the 0.5 r squared was now looking quite good um so when i apply the regression it doesn't always make the data much better so this was the run three data i didn't apply the one with the lowest r square value um because it just means that instead of being too high we still have the values but they're now seeming too low and the error is still huge now these regressions don't include any of the propagated error from the regression equations um but just include the machine error so these error bars would be even bigger if we applied the propagated error from the correction equations as well but these were our two best correction regressions where they do sit along the line of one but even here we have the really really large error bars on our low strontium 103 ppm in ammo samples which really aren't that low in the ground scheme thanks so while the best tuning conditions appear to be run three and run seven when you look at the graph of just the tuning conditions when you try and apply the regressions actually the best data that we get out at the end is from run six and run seven but you wouldn't necessarily know by looking at your data when it comes off the machine but run six was going to produce sensible values or what appear to be sensible values so in summary we can see this interference on all and interference on both machines at bgs but we don't know exactly what what it is we're only correcting for it because we think it looks right or we think that the corrections we're applying look right we've been invited to repeat this experiment at a further three two machines at a further three machines at two different institutions one of which has a collision cell and one of which is a prototype at Bristol which I think has a different kind of collision cell so we might be able to improve our data this way or we might see different problems with different machines um the bgs Neptune appears to produce better results from the bgs new but that doesn't mean that all net tunes produce better data than all news or even that this Neptune produces better data all the time because some days we had runs where it started off looking sensible and then it would just drop halfway through the same tuning conditions do not always produce the same degree of bias so as we showed showed with the two runs that had much bigger error on one day than on the other when the tuning conditions were pretty much the same and the published method of tuning for lower oxides does not always produce less interference it did on the Neptune here but it didn't on the new we can create a regression for some but not all runs with varying degrees of successful correction as we don't really know what we're correcting it only appears correct we're not sure if we're actually dealing with all of the problems that we might potentially be encountering or just the ones that we think we can see so i'd like to thank my supervisors and knurc for funding me and thank you all for listening