Data Quality Control – the Sum of Major Components June 6, 2021 Christian Scheibe Note: Due to the use of data tables in this article it may look odd on smaller screens, e.g.; mobile/cell phones. For a quick data quality control, use the sum of the major components One of the first things to do for estimating the quality of geochemical data is to have a look at the sum of the major components, i.e., the sum of oxides.In theory, the sum of all components from a geochemical analysis should be 100%, which is seldom the case (unless the data are normalized to 100% ). Why is that? Geochemical analyses are reported in weight percent (wt%.). There are chemical components that cannot be determined by the analytical instruments. For instance, with Inductively Coupled Plasma (ICP) and X-Ray Fluorescence (XRF) methods a range of elements are not detected, which include, besides others, the important constituents H, C, O. Therefore, chemical components like H2O, CO2, and organics (H-C-O combinations) are not measured, and thus missed out from the sum. How to check the data quality with that knowledge? The first step is to calculate the sum of all major components, e.g., the major elements in their oxide format. If the sum is between 80 and 100%, there is probably nothing to worry about. Values slightly higher than 100% can be due to calibrations being slightly out of range for some elements.Sum = >80%-100% ⇒ the analysis is probably good and nothing to worry about.Issue 1: Sum < 80% [→ go to explanation]High CaO (and MgO) carbonate minerals are likely to be presentCheck for SO3, if available, and high CaO (and/or Ba concentrations)Issue 2: Sum > 100%this could be a calibration issue, e.g. an oxide that does not have a calibration for high concentrations, and has thus a high error when abundant in high amounts.One or more oxides are not present as oxides, e.g. Fe and S from pyrite. This however does usually not bring the sum to very high values. Remember ...… that the major elements are commonly reported in oxide format, e.g., SiO2, Al2O3, CaO, etc., but these are only calculated oxides; the instruments measure only the elements (Si, Al, Ca, etc.). What if the sum < 80% A sum (or totals) of the major components, i.e., major oxides, that is less than ca. 80% indicates components that cannot be detected by the analytical instrument. In the case that carbonate minerals (e.g., calcite and/or dolomite) are present, a quick mineralogy model can help to understand the data quality. Table 1: Geochemical data from two rock samples. Note that the sums for both samples are very low. Sample ID SiO2 [wt%] TiO2 [wt%] Al2O3 [wt%] Fe2O3 [wt%] Na2O [wt%] K2O [wt%] CaO [wt%] MgO [wt%] MnO [wt%] P2O5 [wt%] SO3 [wt%] SUM [wt%] A 2.79 0.06 0.39 0.13 2.74 0.17 49.91 1.06 0.006 0.005 1.83 59.09 B 17.47 0.18 4.78 1.12 1.26 1.34 22.03 14.62 0.027 0.021 1.11 63.96 On a second, more detailed view on the data, high CaO (sample A) and high CaO and MgO (sample B) concentrations are evident. This indicates that the samples most likely have high carbonate contents. Below are the chemical formulas and compositions listed for dolomite and calcite. Dolomite: CaMg(CO3)2 CaO = 30.41% MgO = 21.86% CO2 = 47.73% (MgO/CaO ∼ 2/3 ) Calcite: CaCO3 CaO = 56.03% CO2 = 43.97% As you can see, both minerals contain ca. 45% CO2, a component that is not determined by e.g., ICP or XRF, and thus missing from the sum of oxides.This may be enough to estimate the amount of (missing) CO2 and add it to the sum, which now should become much closer to 100%.However, we can have a closer look and do some basic mineral modeling. Some basic mineral modeling With this approach, we use molecular weights instead of weight percent.This means we need to know the molecular weights (in g/mol) of the oxides and minerals. [Note, there will be an article about the use of molecular proportions for basic geochemical calculations soon.]CaO = 56.08, MgO = 40.30, CO2 = 44.01, calcite = 100.09, dolomite = 184.40 Using molecular weights has several advantages:It is easier to calculate dolomite and calcite in case both are present.We can also calculate anhydrite/gypsum and subtract the required amount of CaO before modeling for carbonate minerals.We are more flexible in case we know the true Ca:Mg ratio for dolomite in the analyzed strata (which can vary significantly; e.g. Sperber et al. 1984).We are able to calculate our own conversion factors for future use. a) Let's start with dolomite: We will start with modeling dolomite first because we need to know how much CaO is in the dolomite of the sample before we can model for calcite.Dolomite = MgO/mol.wt.(MgO) *mol.wt.(dolomite) MgO/40.30*184.40CO2(dolo.) = MgO/40.30*44.01*2or dolomite/184.4*88.02In our example,sample A: dolomite = 1.06/40.30*184.4 = 4.84 [wt%]; CO2(dolo.) = 1.06/40.30*88.02 = 2.32 [wt%]sample B: dolomite = 14.62/40.30*184.4 = 66.90 [wt%]; CO2(dolo.) = 14.62/40.30*88.02 = 31.93 [wt%](see Table 2) b) We continue with calcite: First, we determine how much CaO is left available after modeling dolomite. CaO(calcite) = (CaO/56.08 – MgO/40.30)*56.08 Secondly, similar to our dolomite calculations, we calculate calcite and CO2 concentrations. Calcite = CaO(calcite)/mol.wt.(CaO)*mol.wt.(calcite) = CaO(calc.)/56.08*100.09 or (CaO/mol.wt.(CaO) – MgO/mol.wt.(MgO))*mol.wt.(calcite) CO2(calc.) = CaO/56.08*100.09 In our example, sample A: CaO(calc.) = (49.91/56.08 – 1.06/40.03)*56.08 = 48.43 [wt%] calcite = 48.43/56.08*100.09 = 86.45 [wt%]; CO2(calc.) = 48.43/56.08*44.01 = 38.01 [wt%] or (49.91/56.08 – 1.06/40.30)*44.01 = 38.01 [wt%] sample B: CaO(calc.) = 1.69 [wt%] Calcite = 3.01 [wt%]; CO2(calc.) = 1.32 [wt%] (see Table 2) Table 2: The calculated concentrations for dolomite, calcite, and CO2 from dolomite and calcite, respectively. Sample ID Dolomite [wt%] Calcite [wt%] CO2 (dolo.) [wt%] CO2 (calc.) [wt%] A 4.85 86.45 2.32 38.01 B 66.90 3.01 31.93 1.32 c) Final calculation of the sum again Finally, we add the calculated CO2 concentrations from dolomite and calcite to the sum of major oxides from the initial analysis of the two samples. As you can see from Table 3, the sum is now very close to 100 wt%. This enhances our confidence in the data quality. Table 3: Adding up the sum of measured oxides (Table 1) and CO2 from dolomite and calcite, reveals that the geochemical analyses are close to 100 wt%. Sample ID SUM [wt%] CO2 [wt%] SUM+CO2 [wt%]] A 59.09 40.33 99.42 B 63.96 33.25 97.21 Some fine-tuning Like mentioned before, there may be a case where also anhydrite (CaSO4) or gypsum (CaSO4•2H2O) may be present. CaO required for CaSO4 (CaSO4 = CaO+SO3) can then be calculated and subtracted from the total CaO before modeling for dolomite and/or calcite.Let’s assume there are 3.5 wt% SO3 in the analysis. With a molecular weight of 80.0632 g/mol for SO3: 3.5/80.06*100.09=4.38 (wt% CaO), which needs to be subtracted from the initial CaO concentration. One more thing ...… not only that we can be more confident in the quality of our data and analyses, we although have some information on the likely mineralogy, i.e., additional data to work with. Summary Calculating the sum (or totals) of the major oxides gives a first idea about the data quality.The sum should be >80% to 100%.If the sum is less than 80%, an estimation of the carbonate content of the sample might give a further indication of why the sum could be low. The carbonates calcite and dolomite contain ca. 45% CO2 (≅ 43.97%, and ≅ 47.73%, respectively).Anhydrite/gypsum may be present and may require a CaO correction to achieve a better result.Besides higher confidence in the data and sample quality, we have additional information about (parts) of the likely mineralogy. Reference in this Article Sperber, C.M., Wilkinson, B.H., and Peacor, D.R. (1984): Rock Composition, Dolomite Stoichiometry, and Rock/Water Reactions in Dolomitic Carbonate Rocks. The Journal of Geology, Vol. 92, Nr. 6. [Link]
Oh, I’m really sorry to bother you in replying. Thank you very much for your patience in explaining and narrating. I benefited a lot. Peace be upon you on your travels Reply
Thank you for you good wishes. You do not bother me at all. I am happy if I can help. As I am travelling, I do not have my references and notes with me, thus I am a bit restricted with information I can provide. Greetings from Amsterdam, C. Reply
I would thank you for this useful article, but I have some questions. If I have a revers case means I have total sum of Oxides but SO3 not find, how can I calculate it??? Can you help me please, q2/ how can I find percentage minerals abundance from chemical analysis? especially clay minerals Reply
Hello. Thank you for your comment and questions. Q1: This is a tricky one. First, what type of rock do you have, siliciclastic, carbonate, or igneous? What mineral do you expect to host S, e.g. pyrite, anhydrite, or barite? Do you have other volatiles such as CO2 or H2O? The next step would be to do some mineral modelling, at least with the elements that are involved in the likely sulphides or sulphates. But this can be difficult already. In the easiest case, only one mineral has S, e.g. barite. Then you can do a normative calculation of how much SO3 is required to make all Ba into barite. It is more difficult with anhydrite when you also have carbonates. If you have CO2 then you may be able to model calcite/dolomite and use the remaining Ca for calculating how much SO3 you need to make anhydrite. If pyrite is the S host, you need to know how much Fe is in other minerals. As you can see, it is getting complicated. Q2:Mineral modelling from element data is challenging, particularly clay minerals. There are a lot of papers and algorithms published. Unfortunately, most of them, particularly for clay minerals, require to know how much Fe2+ (FeO) and Fe3+ (Fe2O3) and often how much volatiles are in the analysis. Algorithms that only use a minimum on elements are those in petrophysical software, such as Minlog, but you would also need petrophysical data such as GR, RHOB, etc. (and obviously access to one of those software packages. I plan to write an article about mineral modelling from chemical data, but that may take some time. If you have some mineral data you can try cross-plots of minerals vs. elements, e.g. K vs. illite and/or K-feldspar. If you get positive correlation lines, you can use the regression formula to calculate/estimate the mineral abundance for that element-mineral pair for the other data (assuming they have the same mineral assemblage). For siliciclastic rocks I found MinLith (Rosen, 2004) often appropriate. You may have to recalculate gibbsite in (?) kaolinite, if you don’t have gibbsite. If you know about the actual chemical composition of your minerals, you may want to have a look at a program called MinSolv (or MinSolver?). It uses the Solver Add-on in Excel. (If you understand how it works, you can easily alter the chemical composition to the minerals according to your requirements.) For igneous rocks, the good old CIPW norm (there are a few alterations published) may do the job, at least give you an estimate. – I am currently on vacation and do not have my references list with me, thus cannot provide further details in the moment. I hope this gives you some ideas. Mineral modelling from chemical data is difficult, but possible. However, whatever approach/algorithm you want to use, make sure it is fit for your purpose/data. Don’t apply them blind, check if the minerals make sense for the type of rock you have. Reply
Hello, Thank you very much for your reply and your help. I am very grateful, In fact, the type of rocks are carbonate rocks. I have one of the samples, for example, SiO2, Al2O3, Fe2O3 ,CaO, MgO, Na2O, K2O, Cr2O3, TiO2 ,MnO, P2O5, SrO, BaO ,LOI, Total 0.16, 0.06, 0.04, 56.0, 0.33 ,0.03, 0.008 ,0.0018 ,0.008, 0.09, 0.008, 0.01 ,0.008 ,43.6 ,100.52 respectively, how do I find SO3 and H2O from these results and the percentages range of clay minerals Thanks in advance and sorry for the inconvenience Reply
Hello again, Thank you for the additional information. Referring to the data you have a very clean limestone. The maximum CaO concentration possible is theoretically 56.03%. Your analysis has 56.0%. For this you need 43.57% CO3 to make it into calcite, which is almost exactly your LOI. The Total is 100.52%, something is slightly overestimated, but with such a clean lithology this is not surprising, due to calibrations. I do not think that you have any noteworthy amounts of SO3 or H2O (Totals are over 100% already). With these small amounts of other oxides it is difficult to estimate what clay minerals may be present. Fe and Mg could be either impurities in calcite or maybe chlorite. Na and K could be from feldspars or illite. There could also be tiny grains of quartz and feldspar, and/or stylolites. The non-carbonate fraction could be simply wind-blown dust. You see, a lot of uncertainties. If you have mineralogy, e.g. from XRD, for a few samples that would be helpful in restricting the mineral assemblage in an attempt to model mineralogy. That is all I can do in the moment from the airport lounge in Dammam. Reply