Cloete Huibrecht Wtms

COPYRIGHT AND CITATION CONSIDERATIONS FOR THIS THESIS/ DISSERTATION
o Attribution — You must give appropriate credit, provide a link to the license, and indicate if
changes were made. You may do so in any reasonable manner, but not in any way that
suggests the licensor endorses you or your use.
o NonCommercial — You may not use the material for commercial purposes.
o ShareAlike — If you remix, transform, or build upon the material, you must distribute your
contributions under the same license as the original.
How to cite this thesis
Surname, Initial(s). (2012). Title of the thesis or dissertation (Doctoral Thesis / Master’s
Dissertation). Johannesburg: University of Johannesburg. Available from:
http://hdl.handle.net/102000/0002 (Accessed: 22 August 2017).
THE ESTIMATION OF UNCERTAINTY OF MEASUREMENT BY X-RAY
FLUORESCENCE SPECTROMETRY
BY
HUIBRECHT CORNELIA CHRISTINA CLOETE
Thesis in fulfilment of the requirement for the degree
MASTER OF SCIENCE
in
CHEMISTRY
in the
FACULTY OF SCIENCE
of the
UNIVERSITY OF JOHANNESBURG
SUPERVISOR: PROF P. N. NOMNGONGO

CO-SUPERVISOR : DR A. BOTHA
Affidavit
Dedication
Dedicated to my father, Prof. Daan Pienaar, who passed away in 2020. As one of the
translators of the new Afrikaans Bible, he had a deep respect for a body of knowledge and,
along with my mother, Mrs Betta Pienaar, instilled into me an unquenchable thirst for
knowledge.
It is also dedicated to the scientists who, in the intricate age of COVID-19, fought tirelessly
to understand this menace and find reliable treatment as fast as possible.
Acknowledgements
- God, for all His grace for which I am grateful.

- Dr Angelique Botha of the National Metrology Institute of South Africa (NMISA), who
gave me the opportunity to pursue one of my professional goals at an advanced
stage of my career.
- Prof Philiswa Nomngongo of the University of Johannesburg (UJ) for her patient
assistance.
- The Council for Geoscience (CGS) for their financial support.
- My colleagues in the Analytical Services Unit for their support. A big ‘Thank you!’ to
Ms Prudence Mokoena, Mr Themba Mnguni and Mr Esau Tsaagane for their
assistance in the study experiments.
- Dr Bruno Vrebos and Mr Frik Burger from Malvern for their technical support.
- Ms Thelma Kotze for her sign language interpreting assistance during online
meetings.
- My sister, Mrs Elzabé Heyns, for her proofreading assistance.
- My family for their interest and support via social media in this social distancing
times.
- My beloved husband, Thinus, and children, Bard and Jana, for their belief in me and
chats about a dry scientific topic.
Abstract
The availability of a result’s uncertainty of measurement in an analytical procedure is an
integral part of a laboratory and is one of ISO17025:2017’s requirements. Uncertainty of
measurement is one of three indispensable parameters to ensure the quality of
measurement results – the others are metrological traceability and method validation.
Although there is numerous information available to estimate the uncertainty of
measurement, a step-by-step procedure covering a test method from sample preparation
to analysis by WD-XRF could not be located. The Guide to the Expression of Uncertainty
in Measurement, also known as the GUM, is the main referring document but cumbersome.
Furthermore, most publications in WD-XRF use the top-down approach to quantify a test
method’s uncertainty of measurement in contrast with the bottom-up approach, where a
test sample’s results’ uncertainty of measurement are estimated.
The study test method was set up to analyse major elements (SiO2, TiO2, Al2O3, total Fe
as Fe2O3, MnO, MgO, CaO, Na2O, K2O, P2O5 and Cr2O3) on glass disks by WD-XRF and
LOI by gravimetry. The test method analysed a wide range of sample types to cover typical
geological samples from South Africa. Method validation data was used to estimate
uncertainty, while metrological traceability was established by evaluating the test method’s
trueness using certified reference materials.
Bias and precision data of a test method was used by both the top-down and bottom-up
approaches. The top-down approach uses the quality control sample’s variance to quantify
precision and either certified reference materials or proficiency tests for bias quantification.
The bottom-up approach in the study required four functional models combining the
uncertainties of the input quantities such as the flux components, regression slope and
intercept on the Y-axis to quantify the measurand’s uncertainty of measurement.
EXCEL workbooks were developed for the GUM framework, Kragten and Monte Carlo
methods, and the top-down approach. Uncertainties for the different approaches were
compared, and significant differences, especially for SiO2, CaO and LOI at high
concentration values, exist between the bottom-up and top-down approaches.
Table of Contents
Affidavit ........................................................................................................................... i
Dedication ...................................................................................................................... ii
Acknowledgements ...................................................................................................... iii
Abstract ......................................................................................................................... iv
Table of Contents .......................................................................................................... v
List of Figures ............................................................................................................. viii
List of Tables .................................................................................................................. i
List of Abbreviations .................................................................................................... xi
Chapter 1: Introduction and problem statement ........................................................ 1

1.1 Introduction ......................................................................................................... 1
1.2 Motivations and objectives of the study .............................................................. 2
1.3 Problem statement ............................................................................................. 4
1.4 Key objectives of the study .................................................................................5
Chapter 2: Theoretical background and literature review ......................................... 6

2.1 Introduction .........................................................................................................6
2.2 Concepts in the uncertainty of measurement .....................................................6
2.2.1 Descriptive and inferential statistics....................................................................6
2.2.2 Metrological terms ..............................................................................................9
2.2.3 Method validation, metrological traceability and uncertainty of measurement .. 15
2.3 Established approaches to estimate the uncertainty of measurement .............. 20
2.3.1 The Bureau International des Poids et Mesures (BIPM) and the GUM ............ 20
2.3.2 Top-down and bottom-up approaches .............................................................. 28
2.4 Previous work in the estimation of uncertainty in WD-XRF measurements ...... 33
2.4.1 Previous studies of effects (errors) in WD-XRF ................................................ 34
2.5 Documentary standards of WD-XRF test methods ........................................... 37
2.6 Summary and conclusion ................................................................................. 37
Chapter 3: Methodology .............................................................................................. 39

3.1 Introduction ....................................................................................................... 39
3.2 Theory of wavelength-dispersive XRF .............................................................. 39
3.3 Scope of the test method .................................................................................. 43
3.4 Primary sample preparation.............................................................................. 43
3.5 Secondary sample preparation ......................................................................... 44
3.5.1 Instrumentation in the secondary sample preparation stage ............................ 45
3.5.2 Steps in the secondary sample preparation stage ............................................ 50
3.5.3 Flux used in the test method............................................................................. 51
3.6 Analysis by WD-XRF ........................................................................................ 52
3.6.1 The XRF spectrometer ..................................................................................... 52
3.6.2 Fundamental Parameters (FP) ......................................................................... 56
3.6.3 CRMs and RMs used for the test method calibration ....................................... 60
3.6.4 Other test materials and their portions .............................................................. 61
3.6.5 The quantification process ................................................................................ 62
3.7 Method performance characteristics ................................................................. 64
3.7.1 Selectivity ......................................................................................................... 64
3.7.2 Working range and linearity .............................................................................. 65
3.7.3 Analytical sensitivity .......................................................................................... 65
3.7.4 Trueness........................................................................................................... 66
3.7.5 Precision ........................................................................................................... 67
3.7.6 Limit of detection and limit of quantification ...................................................... 69
3.8 Metrological traceability .................................................................................... 69
3.9 Uncertainty of measurement............................................................................. 70
3.9.1 Sample preparation .......................................................................................... 70
3.9.2 Evaluation of precision ..................................................................................... 72
3.9.3 The regression line ........................................................................................... 72
3.9.4 The top-down approach .................................................................................... 75
3.9.5 The bottom-up approach .................................................................................. 78
3.9.6 The uncertainty budget ..................................................................................... 82
3.9.7 Software tools used in the study ....................................................................... 82
3.10 Summary .......................................................................................................... 85
Chapter 4: Results and Discussion............................................................................ 87

4.1 Introduction ....................................................................................................... 87
4.2 Method validation ............................................................................................. 87
4.2.1 Selectivity ......................................................................................................... 88
4.2.2 Linearity, working ranges, limit of detection (LOD) and limit of quantification (LOQ)
......................................................................................................................... 88
4.2.3 Trueness........................................................................................................... 90
4.2.4 Precision ........................................................................................................... 93
4.3 Uncertainty of measurement........................................................................... 103
4.3.1 The bottom-up approach ................................................................................ 103
4.3.2 The top-down approach .................................................................................. 134
4.3.2.1 CRM results................................................................................................. 135
4.3.2.2 PT results .................................................................................................... 136
4.3.2.3 EXCEL template for the top-down approach ............................................... 137
4.3.3 Software tools ................................................................................................. 139
4.3.4 Comparison of uncertainties from the bottom-up and top-down approaches . 141
4.4 Performance report for the study test method ................................................ 150
Chapter 5: Conclusions ............................................................................................ 151

5.1 Introduction ..................................................................................................... 151
5.2 Summary ........................................................................................................ 151
5.3 Conclusions .................................................................................................... 153
References ................................................................................................................. 155

Appendix ............................................................................ Error! Bookmark not defined.
List of Figures
Figure 2.1: The normal distribution plot of a population (the Gauss distribution) ....... 7
Figure 2.2: Precision, bias, and accuracy. The blue markers represent replicate
measurement results, and the red marker the assigned value of the
measurement. (a) reflects the highest accuracy and (d) the least accurate
result and test method ........................................................................... 13
Figure 2.3: The effect framework in terms of type, performance and expression and
their influence on the uncertainty of measurement. It is based on the error
framework in Eurachem Guide: The Fitness for Purpose of Analytical
Methods [15]. Systematic effects contribute to the trueness of a test method
and are quantified by the test method's bias, while the precision of a test
method is influenced by random effects and quantified by the standard
deviation of the random effects ............................................................. 14
Figure 2.4: Metrological traceability from each result (y1, y2 and y3) to the SI unit. The
SI unit is linked to the certified reference material (CRM) and the mass
piece used in the sample preparation. The CRM value is used in both the
test method calibration and the evaluation of the test method. Results (y1,
y2, y3) are compared by their traceability to the CRMs and the mass pieces
used in the test procedures. (Adapted from a figure[19]) ........................ 16
Figure 2.5: A simplified illustration of the three metrology cornerstones to evaluate a
test method and individual measurement results. The red diamond
represents a CRM's certified value. The term 'Fit for purpose' of a test
method indicates that the precision and accuracy of the test method is
sufficient so that the end-user can make an informed decision on the
intended use of the produced results or analysed samples[23][24]........... 18
Figure 2.6: The Monte Carlo method (GUM-2)[26] ..................................................... 27
Figure 2.7: The schematic differences between the bottom-up approach on the left and
the top-down approach on the right. The bottom-up approach evaluates all
(or at least the main) sources of uncertainty, while the top-down approach
uses results from QC, method verification/validation or PT Studies[50] .. 29
Figure 2.8: The five different uncertainty of measurement evaluation methods grouped
into the bottom-up and top-down approaches (Based on published
figures[21][49]). The data types obtained from the replicate measurements,
CRMs and PT samples are discussed in the study ............................... 31
Figure 2.9: Sources of effects in XRF[53] ................................................................... 34
Figure 2.10: Indirect determination of uncertainty due to sample preparation. The
symbols are as used in the publication[52].............................................. 35
Figure 2.11: Evaluation of measurement uncertainty in XRF.[5] .................................. 36
Figure 3.1: A simplified illustration of the WD-XRF basics. The element's wavelength
energy is identified by Bragg's law using the diffracting crystal's known
lattice space (2d) and the ϴ-value as measured by the goniometer while
the detector quantifies the identified elemental peak. The figure is based
on a figure drawn by Willis et al.[5] ......................................................... 42
Figure 3.2: An X-ray spectra of an andesite sample. The X-axis depicts the
goniometer's 2ϴ-theta positions from ~5 to 147 degrees, while the Y-axis
expresses the count rate in kCPS. Crystal LiF200 and duplex detector were
used for this scan made by the PANalytical Zetium XRF spectrometer 42
Figure 3.3: The furnace used to roast the dried sample material in porcelain crucibles
at 1000 ⁰C. After the roasting step, the crucibles are cooled inside
desiccators to prevent re-absorption of any moisture ............................ 46
Figure 3.4: Claisse TheOx fusion instrument is a front door electric furnace with a
position for six platinum crucibles and flat bottom moulds. The melting,
stirring and pouring steps are automated by a selected programme called
the "CGS Oxides" in this study. The steps are listed in
Table 3.3 ............................................................................................... 47
Figure 3.5: The platinum ware used by the fusion instruments. The Pt-crucible and flat
bottom mould on the left are used by the Claisse TheOx, while the
mouldible in the top right is used by the Beadmaster. The resulting glass
disk in the bottom right is used to present the sample into the WD-XRF
spectrometer ......................................................................................... 48
Figure 3.6: The Beadmaster is a robust fusion instrument which is an electric furnace
with a top door (a) and positions for four mouldibles (b). After the melting
stage, they are removed with a titanium tong and placed into cooling
positions in front of the furnace. There the glass disks are cooled with
controlled air.......................................................................................... 49
Figure 3.7: PANalytical Zetium XRF-spectrometer. It is both a wavelength and energy
dispersive XRF spectrometer and has a 4kW Rh-tube ......................... 53
Figure 3.8: Left bottom displays the measuring cups on a tray where one cup is
selected to handle the glass disk inside the XRF-spectrometer. The loading
position right of the measuring cups is where the cup is moved into the
measuring position inside the spectrometer. The combination of each cup
and sample position has slightly different distances between the tube and
glass disk surface .................................................................................. 56
Figure 3.9: The enhancement of Cr by Fe radiation in a sample[53]. Cr is the analyte
element i, Fe the matrix element j. IiA(λi) is the intensity from the primary
excitation, while IiE(λi) is the intensity from enhancement. The reverse is also
applicable, i.e. absorption of Fe by Cr. The effect of both enhancement and
absorption on the calibration line is illustrated in
Fig. 3.10 ................................................................................................ 57
Figure 3.10: Inter-element matrix effects on the intensity of the analyte in a calibration
plot for Fe in a high alloy steel sample[53]. In the left figure, enhancement
leads to higher than expected intensities, while lower intensities are the
product of absorption. In the right figure, matrix corrected intensities are
plotted linearly with their associated concentration values .................... 58
Figure 3.11: The FP equation calculating the primary fluorescence P i using the used
setup of the spectrometer, and the known elemental values of JK, the K
jump ratio, ωK , the fluorescent yield and PKα, the probability factor for the
Kα-line to be emitted instead of the Kβ-line. (From two figures in Willis et
al.[5]). Pi(λ0) is the primary fluorescence excited by the monochromatic
incident radiation λ0 from the Rh-tube in the study. I0(λ0) is the intensity of
the incident X-rays; Ci is the concentration weight fraction; ψ’ and ψ” are
respectively the incident and emergent angles of the X-rays with the test
µ
sample surface; is the mass attenuation coefficients for the X-rays
𝝆
(cm2/g). The blue dotted boxes (a) to (e) are explained in the text ........ 59
Figure 3.12: The calculation of the CRM and RM concentration for entering into the
SuperQ calibration application and the quantification process of the test
sample's major elements. Ci = first recalculated measurand concentration
(%); Ci,cert = original measurand concentration (%); Total (CRM and RM) =
the sum of the measurand concentrations and the measured LOI (%);
LOImeas = measured LOI (%); Ci,final = the concentration values for the
SuperQ application; MassSample and Massflux = mass (g) of the roasted
sample and flux respectively; Ci,unk = the result concentration for reporting
(%); Total (Measured sample) = the sum of all concentrations (%); Flux =
(B2O3 + Li2O) (%). The flux values are entered into the software with at
least five decimals, and the calculation process to obtain B 2O3 and Li2O is
illustrated in Figure 3.13 ........................................................................ 63
Figure 3.13: The 'flux calculator' used to obtain the Li2O and B2O3 values based on the
selected flux and the masses of the flux and roasted material .............. 64
Figure 3.14: The relationship between the measurand concentration and the absolute
(i) and relative (ii) uncertainties, respectively[6]. The dashed line divides the
measurement range into concentration areas where the absolute
uncertainty is constant and where it becomes larger with higher
concentration (i). In the case of the relative uncertainty (ii), the
measurement range is divided into concentration ranges where the
uncertainty is large at low concentrations and is approximately constant at
higher concentrations ............................................................................ 76
Figure 3.15: The measurement model and the cause-and-effect diagram to determine
the main sources of uncertainty. Ci is the concentration of the measurand,
Ei is the slope of the calibration's regression line, Di is the intercept value,
and MIi is the matrix corrected instrumental response. P is to account for
uncertainty due to precision and has a value of 1 ................................. 80
Figure 3.16: The connection between the sample preparation and analysis processes
with the determination of the measurand concentrations and associated
uncertainties .......................................................................................... 81
Figure 3.17: The main window of the NIST uncertainty machine web application. The
variables, probability distribution type, value and associated standard
uncertainty are entered into the middle area. If required, the correlation
values are entered too. The measurement model is defined in the bottom
area, and the green bar at the bottom is pressed to process the entered
information ............................................................................................ 83
Figure 3.18: The results from the GUM (bottom) and MCM (top) for the entered
information in Fig. 3.17. The calculated sensitivity coefficients, for example,
are compared with the sensitivity values calculated by the EXCEL template
to verify their accuracy .......................................................................... 84
Figure 4.1: The calibration application in SuperQ listing the compound, unit, regression
model, slope, correlation, and the goodness of fit value for the regression
slopes .................................................................................................... 88
Figure 4.2: The application setup with selected spectrometer parameters to analyse the
measurands by WD-XRF........................................................................ 88
Figures 4.3 (a) to (l):Graphic evaluation of the test method – the blue circles represent the
results of a CRM; the orange circles the proposed value of a CRM, and the
yellow boxes the proficiency test materials. The measurand results
compare favourably with their certified or assigned values ................... 90
Figure 4.4 Analysis of SiO2 in a fusion disk of MzBP-1 and arrangement of results in
EXCEL. The data is arranged per column for the cups and per row for the
measuring position. The alpha level (α) for the ANOVA is 0.05 ............ 99
Figure 4.5: Measurement model 1 at the top and fishbone diagram at the bottom to
identify contributors to the uncertainty of the output quantity,
LOI (%) ................................................................................................ 104
Figure 4.6: Measurement model 2 for the determination of Li2O (%) in the fusion disk.
The fishbone diagram at the bottom includes the contributions to
uncertainty by the purity of the flux, the balance and the atomic masses of
Li, O and B .......................................................................................... 105
Figure 4.7: Measurement model 2 for the determination of B2O3 (%) in the fusion
disk. The fishbone diagram at the bottom and the symbols are similar to
those in Fig. 4.6................................................................................... 106
Figure 4.8: Measurement model 3 for the results (Cmeas,i) as calculated by SuperQ and
is expressed as % ............................................................................... 107
Figure 4.9: Measurement model 1 in worksheet LOI of the EXCEL workbook, 'GUM
framework.xlsx'. The Cr, Cr100 and Cr1000 are the input variables, u(Cr),
u(Cr100) and u(Cr1000) the standard uncertainties while c(Cr), c(Cr100)
and c(Cr1000) are the sensitivity coefficients. The correlation coefficients
are determined using the Correlation function in the Data Toolpak in
EXCEL for Microsoft 365 and the results listed in cells J28 to L30 ..... 116
Figure 4.10: The calculation of Li2O(a), Li2O(b) (top left boxes), B2O3(a) and B2O3(b)
(bottom left boxes) values and uncertainties for the B flux. The calculated
result and uncertainty for Li2O are in cells O54 and P54, respectively, while
those for B2O3 are in cells O66 and P66. The template is applied with
applicable Li2B4O7 and LiBO2 values for the A and C fluxes ............... 117
Figure 4.11: The Measurement model 2 in the 'Flux in sample disk' worksheet. The top
three rows contain the values calculated in the 'Flux calculator’-worksheet
for specific flux types. The sample (T-column) and flux (P-column) masses
and the flux components based on the selected flux determine the Li 2O-
values in the Y-column and the B2O3-values in the AA-column. The input
values' calculated uncertainties are listed next to the input values' columns
............................................................................................................ 118
Figure 4.12: Measurement model 3 for the concentrations from SuperQ in the 'Conc
from SuperQ’-worksheet. Information from SuperQ such as the slope, its
uncertainty calculated from the slope's K-factor and the number of used
standards are listed in the top part of the worksheet (cells E10 to Y20). At
the bottom, the columns for the matrix corrected counts (MIi), precision (P)
and their uncertainty and sensitivity coefficients are organized for SiO 2. The
same setup for the other oxides is right in the worksheet and is detailed in
Appendix L .......................................................................................... 120
Figure 4.13: The last worksheet, 'Final conc', in the EXCEL workbook, 'GUM
framework.xlsx', is for measurement model 4. In the top from cells F27 to
K32, the correlation between CSiO2,meas, %Li2O, %B2O3, LOI and Total from
Measurement model 3 are calculated and used for the sensitivity
coefficients. The values, uncertainties and sensitivity coefficients for the
five variables in measurement model 4 for SiO2 are listed in cells B37 to
S41. The same setup for the rest of the oxides is from columns V in the
worksheet and is detailed in Appendix H. The input values such as C i,meas
and u(Ci,meas) refer to corresponding cells in the 'Conc from SuperQ’-
worksheet to prevent errors from copying and pasting of cell values. The
final results and associated uncertainty are reported in rows 46 to 50.…
............................................................................................................ 121
Figure 4.14: The measurement model 1 in the MCM's EXCEL spreadsheet template for
the test sample, MzBP-1 (workbook 'MCM MzBP-1.xlsx'). The input
quantities and associated standard uncertainties are in cells C4 to G5. In
rows 13 to 100 013 and columns C to G, permutations of the measurement
model 1 are calculated with the corresponding LOI-values in column I. The
standard deviation of the 100 000 LOI values in cell I5 (0.010 %) is then
the uncertainty of the LOI-value of 0.396% ......................................... 124
Figure 4.15: A plot of the frequency of MzBP-1's LOI results from the MCM's
permutations vs concentration (%). The black line is the LOI value of
0.396%, with the green lines being the uncertainty of 0.0100% and the red
lines the expanded uncertainty of 0.0200% at a confidence level of 95.45%
............................................................................................................ 125
Figure 4.16: Calculation of Li2O's uncertainty in sample MzBP-1 with the MCM
approach. The half-width interval, a, is calculated by multiplying the
standard uncertainty with the square root of 3. The standard deviation of
the results (0.019%) from the permutations in column U is in
cell U5 ................................................................................................. 126
Figure 4.17: The frequency plot of the 100 000 permutations for Li2O in sample MzBP-
1. The black line is the Li2O value of 21.19% (21.187%), with the green lines
being the uncertainty of 0.019% and the red lines the expanded uncertainty
of 0.038% at a confidence level of 95.45% ......................................... 127
Figure 4.18: B2O3's uncertainty in sample MzBP-1 as calculated by the MCM method in
the workbook, MCM MzBP-1.xlsx. The standard deviation of the results
(0.020 %) from the permutations is in cell AF5.................................... 128
Figure 4.19: The frequency plot of the 100 000 permutations for B2O3 in sample MzBP-
1 reflecting the combination of the two frequency distribution types of the
input variables. The black line represents the B2O3 value of 69.71%, with
the green lines being the uncertainty of 0.020% and the red lines the
expanded uncertainty of 0.040% at a confidence level of 95.45% ...... 129
Figure 4.20: Frequency plots of the permutations for SiO2 as determined by
measurement models 3 (in (a)) and 4 (in (b)) for test sample MzBP-1. The
final SiO2 result is 57.68\%, with an expanded uncertainty of 0.292% at the
confidence level of 95.45% ................................................................. 130
Figure 4.21: The Kragten method to calculate the LOI-value and uncertainty in
measurement model 1 for test sample MzBP-1. It is from rows 4 to 23 in a
worksheet allocated to this test sample in an EXCEL workbook in this study
(Kragten method.xlsx). The uncertainty budget for this model is in row 23
while the uncertainty value of LOI is in cell E20 .................................. 132
Figure 4.22: The Kragten method to calculate the Li2O-value and uncertainty in
measurement model 2 for test sample MzBP-1. It is from rows 32 to 54 in
the same worksheet in Fig. 4.21. The uncertainty budget for this model is
in row 54, while the uncertainty value of Li2O is in cell E51................. 132
Figure 4.23: The B2O3-value and uncertainty in measurement model 2 for test sample
MzBP-1 determined by the Kragten method. It is from rows 57 to 79 in the
same worksheet in Fig. 4.21. The uncertainty budget for this model is in
row 79, while the uncertainty value of B2O3 is in cell E76 ................... 133
Figure 4.24: The SiO2-value and uncertainty in measurement model 3 for test sample
MzBP-1 determined by the Kragten method. It is from rows 85 to 113 in the
same worksheet in Fig. 4.21. The uncertainty budget for this model is in
row 113, while the uncertainty value of SiO2 is in cell C110. The
contributions to the uncertainty budget are plotted in a bar graph which
illustrates that the calibration slope is the main contributor to SiO 2's
uncertainty for test sample MzBP-1 .................................................... 133
Figure 4.25: The concentration values and uncertainties for test sample MzBP-1
determined by the Kragten method. It is from rows 743 to 757 in the same
worksheet in Fig. 4.21 ......................................................................... 134
Figure 4.26: .... Quantification of CRM data for the top-down approach in the EXCEL
worksheet. .......................................................................................... 136
Figure 4.27: SiO2 results of eighteen GeoPT rounds to obtain the bias data for the test
method. This table is a screenshot of the EXCEL template for the top-down
approach. ............................................................................................ 136
Figure 4.28: The EXCEL worksheet for the top-down approach using PT data. The
uncertainty values for the test sample, SARM-36, is calculated in rows from
43 to 59 (Fig. 4.29) .............................................................................. 137
Figure 4.29: Uncertainty values for the test sample, SARM-36, as calculated in the
EXCEL worksheet in Fig. 4.28 ............................................................ 138
Figure 4.30: The EXCEL worksheet for the top-down approach using CRM data. The
uncertainty values for the test sample, SARM-36, is calculated in rows from
43 to 59. The bias data from CRM are in rows 16 to 18 ...................... 138
Figure C.1: Calibration certificate of balance used in the preparation of the glass disks.
Figure G1: ANOVA calculations to determine the effects of sample and flux masses
on the SiO2 results.
on the TiO2 results.
on the Al2O3 results.
on the Fe2O3 results.
on the MnO results.
on the MgO results.
on the CaO results.
on the Na2O results.
on the K2O results.
on the P2O5 results.
on the Cr2O3 results.
Figure H1: Two factor ANOVA by EXCEL to evaluate the precision of LOI-analysis in a
limestone test sample.
Figure I1: Two factor ANOVA analysis to determine fusion disk precision for SiO 2.
Figure I2: Two factor ANOVA analysis to determine fusion disk precision for TiO 2.
Figure I3: Two factor ANOVA analysis to determine fusion disk precision for Al2O3.
Figure I4: Two factor ANOVA analysis to determine fusion disk precision for Fe 2O3.
Figure I5: Two factor ANOVA analysis to determine fusion disk precision for MnO.
Figure I6: Two factor ANOVA analysis to determine fusion disk precision for MgO.
Figure I7: Two factor ANOVA analysis to determine fusion disk precision for CaO.
Figure I8: Two factor ANOVA analysis to determine fusion disk precision for Na 2O.
Figure I9: Two factor ANOVA analysis to determine fusion disk precision for K2O.
Figure I10: Two factor ANOVA analysis to determine fusion disk precision for P 2O5.
Figure I11: Two factor ANOVA analysis to determine fusion disk precision for Cr2O3.
Figure J1: SiO2 results to evaluate the precision of the sample cup and measuring
position in the Zetium XRF spectrometer.
Figure J2: TiO2 results to evaluate the precision of the sample cup and measuring
Figure J3: Al2O3 results to evaluate the precision of the sample cup and measuring
Figure J4: Fe2O3 results to evaluate the precision of the sample cup and measuring
Figure J5: MnO results to evaluate the precision of the sample cup and measuring
Figure J6: MgO results to evaluate the precision of the sample cup and measuring
Figure J7: CaO results to evaluate the precision of the sample cup and measuring
Figure J8: Na2O results to evaluate the precision of the sample cup and measuring
Figure J9: K2O results to evaluate the precision of the sample cup and measuring
Figure J10: P2O5 results to evaluate the precision of the sample cup and measuring
Figure J11: Cr2O3 results to evaluate the precision of the sample cup and measuring
Figure L1: The calibration and precision values with their associated uncertainties for
Measurement model 3.
Figure L2: Matrix corrected counts, associated uncertainties and sensitivity factors for
SiO2 analysed in the fusion disks of the four test samples. (Columns E to
Q in 'Conc from SuperQ’-worksheet).
Figure L3: Matrix corrected counts, associated uncertainties, and sensitivity factors for
TiO2 analysed in the fusion disks of the four test samples (Columns S to
AE in 'Conc from SuperQ’-worksheet).
Al2O3 analysed the fusion disks of the four test samples (Columns AG to
AS in 'Conc from SuperQ’-worksheet).
Fe2O3 analysed the fusion disks of the four test samples (Columns AU to
BG in 'Conc from SuperQ’-worksheet).
MnO analysed the fusion disks of the four test samples (Columns BI to BU
in 'Conc from SuperQ’-worksheet).
MgO analysed the fusion disks of the four test samples (Columns BW to CI
CaO analysed the fusion disks of the four test samples (Columns CK to
CW in 'Conc from SuperQ’-worksheet).
Na2O analysed the fusion disks of the four test samples (Columns CY to
DK in 'Conc from SuperQ’-worksheet).
K2O analysed the fusion disks of the four test samples (Columns DM to DY
P2O5 analysed the fusion disks of the four test samples (Columns EA to
EM in 'Conc from SuperQ’-worksheet).
Cr2O3 analysed the fusion disks of the four test samples (Columns EO to
FA in 'Conc from SuperQ’-worksheet).
Figure M1: The final concentrations (Column Q) and uncertainties (Column S) of SiO 2
analysed in the four test samples (Columns A to S in the 'Final conc’-
worksheet).
Figure M2: The final concentrations (Column AK) and uncertainties (Column AM) of
TiO2 analysed in the four test samples (Columns V to AM in the 'Final conc’-
worksheet).
Figure M3: The final concentrations (Column BD) and uncertainties (Column BF) of
Al2O3 analysed in the four test samples (Columns AO to BF in the 'Final
conc’-worksheet).
Figure M4: The final concentrations (Column BW) and uncertainties (Column BY) of
Fe2O3 analysed in the four test samples (Columns BH to BY in the 'Final
conc’-worksheet).
Figure M5: The final concentrations (Column CP) and uncertainties (Column CR) of
MnO analysed in the four test samples (Columns CA to CR in the 'Final
conc’-worksheet).
Figure M6: The final concentrations (Column DI) and uncertainties (Column DK) of
MgO analysed in the four test samples (Columns CT to DK in the 'Final
conc’-worksheet).
Figure M7: The final concentrations (Column EB) and uncertainties (Column ED) of
CaO analysed in the four test samples (Columns DM to ED in the 'Final
conc’-worksheet).
Figure M8: The final concentrations (Column EU) and uncertainties (Column EW) of
Na2O analysed in the four test samples (Columns EF to EW in the 'Final
conc’-worksheet).
Figure M9: The final concentrations (Column FN) and uncertainties (Column FP) of
K2O analysed in the four test samples (Columns EY to FP in the 'Final
conc’-worksheet).
Figure M10: The final concentrations (Column GG) and uncertainties (Column GI) of
P2O5 analysed in the four test samples (Columns FQ to GI in the 'Final
conc’-worksheet).
Figure M11: The final concentrations (Column GZ) and uncertainties (Column HB) of
Cr2O3 analysed in the four test samples (Columns GK to HB in the 'Final
conc’-worksheet).
Figure N1: Matrix corrected counts and associated uncertainties for the calculation of
SiO2 (cell AS4) and its uncertainty (cell AS5) in the measurement model 3
using the MCM method (Columns AJ to AS in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
TiO2 (cell BF4) and its uncertainty (cell BF5) in the measurement model 3
using the MCM method (Columns AW to BF in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
Al2O3 (cell BQ4) and its uncertainty (cell BQ5) in the measurement model
3 using the MCM method (Columns BH to BQ in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
Fe2O3 (cell CB4) and its uncertainty (cell CB5) in the measurement model
3 using the MCM method (Columns BS to CB in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
MnO (cell CM4) and its uncertainty (cell CM5) in the measurement model
3 using the MCM method (Columns CD to CM in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
MgO (cell CX4) and its uncertainty (cell CX5) in the measurement model 3
using the MCM method (Columns CO to CX in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
CaO (cell DI4) and its uncertainty (cell DI5) in the measurement model 3
using the MCM method (Columns CZ to DI in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
Na2O (cell DT4) and its uncertainty (cell DT5) in the measurement model 3
using the MCM method (Columns DK to DT in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
K2O (cell EE4) and its uncertainty (cell EE5) in the measurement model 3
using the MCM method (Columns DV to EE in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
P2O5 (cell EP4) and its uncertainty (cell EP5) in the measurement model 3
using the MCM method (Columns EG to EP in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
Cr2O5 (cell FA4) and its uncertainty (cell FA5) in the measurement model
3 using the MCM method (Columns ER to FA in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
Figure O1: Measurement model 4 processing the major elements' results from
Measurement model 3 (Appendix N), the LOI result from Measurement
model 1 and the flux compounds to estimate SiO 2 in cell FV4 and
associated uncertainty in cell FV5 (Columns FE to FV in the 'MCM MzBP-
1.xlsx' EXCEL workbook).
model 1 and the flux compounds to estimate TiO2 in cell GO4 and
associated uncertainty in cell GO5 (Columns FX to GO in the 'MCM MzBP-
model 1, the flux compounds to estimate Al2O3 in cell HH4 and associated
uncertainty in cell HH5 (Columns GQ to HH in the 'MCM MzBP-1.xlsx'
EXCEL workbook).
model 1, the flux compounds to estimate Fe2O3 in cell IA4 and associated
uncertainty in cell IA5 (Columns HJ to IA in the 'MCM MzBP-1.xlsx' EXCEL
workbook).
model 1 and the flux compounds to estimate MnO in cell IT4 and
associated uncertainty in cell IT5 (Columns IC to IT in the 'MCM MzBP-
model 1 and the flux compounds to estimate MgO in cell JM4 and
associated uncertainty in cell JM5 (Columns IV to JM in the 'MCM MzBP-
model 1 and the flux compounds to estimate CaO in cell KF4 and
associated uncertainty in cell KF5 (Columns JO to KF in the 'MCM MzBP-
model 1 and the flux compounds to estimate Na 2O in cell KY4 and
associated uncertainty in cell KY5 (Columns KH to KY in the 'MCM MzBP-
model 1 and the flux compounds to estimate K 2O in cell LR4 and
associated uncertainty in cell LR5 (Columns LA to LR in the 'MCM MzBP-
model 1 and the flux compounds to estimate P 2O5 in cell MK4 and
associated uncertainty in cell MK5 (Columns LT to MK in the 'MCM MzBP-
model 1 and the flux compounds to estimate Cr2O3 in cell ND4 and
associated uncertainty in cell ND5. (Columns MM to ND in the 'MCM MzBP-
1.xlsx' EXCEL workbook). The sum of the final major element results are
in cell NG4 with the uncertainty of the sum in cell NG5).
Figure P1: Frequency plot of TiO2 (1.07%) with an uncertainty of 0.014% and
expanded uncertainty of 0.028% at k= 2.
Figure P2: Frequency plot of Al2O3 (19.54%) with an uncertainty of 0.077% and
Figure P3: Frequency plot of Fe2O3 (4.78%) with an uncertainty of 0.035% and
Figure P4: Frequency plot of MnO (0.141%) with an uncertainty of 0.0033% and
Figure P5: Frequency plot of MgO (1.10%) with an uncertainty of 0.011% and
Figure P6: Frequency plot of CaO (3.73%) with an uncertainty of 0.067% and
Figure P7: Frequency plot of Na2O (6.82%) with an uncertainty of 0.034% and
Figure P8: Frequency plot of K2O (3.84%) with an uncertainty of 0.057% and
Figure P9: Frequency plot of P2O5 (0.61%) with an uncertainty of 0.008% and
Figure P10: Frequency plot of Cr2O3 (0.018%) with an uncertainty of 0.0017% and
Figure P11: Frequency plot of the sum of the major oxides and LOI (99.74%) with an
uncertainty of 0.198% and expanded uncertainty of 0.396% at k= 2.
Figure Q1: The TiO2-value and uncertainty in measurement model 3 for test sample
MzBP-1 determined by the Kragten method. It is from rows 116 to 144 in
in row 144, while the uncertainty value of TiO2 is in cell C141. The
illustrates that the CSE and calibration slope are the main contributors to
TiO2's uncertainty for test sample MzBP-1.
Figure Q2: The Al2O3-value and uncertainty in measurement model 3 for test sample
in row 174, while the uncertainty value of Al2O3 is in cell C171. The
illustrates that the calibration slope and the CSE are the main contributors
to Al2O3's uncertainty for test sample MzBP-1.
Figure Q3: The Fe2O3-value and uncertainty in measurement model 3 for test sample
in row 204, while the uncertainty value of Fe2O3 is in cell C201. The
to Fe2O3's uncertainty for test sample MzBP-1.
Figure Q4: The MnO-value and uncertainty in measurement model 3 for test sample
in row 235, while the uncertainty value of MnO is in cell C232. The
illustrates that the CSE, the intercept on the Y-axis and the calibration slope
are the main contributors to MnO's uncertainty for test sample MzBP-1.
Figure Q5: The MgO-value and uncertainty in measurement model 3 for test sample
in row 266, while the uncertainty value of MgO is in cell C263. The
illustrates that the calibration slope, the CSE and the intercept on the Y-
axis are the main contributors to MgO's uncertainty for test sample MzBP-
1.
Figure Q6: The CaO-value and uncertainty in measurement model 3 for test sample
in row 297, while the uncertainty value of CaO is in cell C294. The
illustrates that the calibration slope is the main contributor to CaO's
uncertainty for test sample MzBP-1.
Figure Q7: The Na2O-value and uncertainty in measurement model 3 for test sample
in row 327, while the uncertainty value of Na2O is in cell C324. The
illustrates that the calibration slope and the precision of dilution are the
main contributors to Na2O's uncertainty for test sample MzBP-1.
Figure Q8: The K2O-value and uncertainty in measurement model 3 for test sample
in row 357, while the uncertainty value of K2O is in cell C354. The
illustrates that the calibration slope is the main contributor to K 2O's
Figure Q9: The P2O5-value and uncertainty in measurement model 3 for test sample
in row 387, while the uncertainty value of P2O5 is in cell C384. The
to P2O5's uncertainty for test sample MzBP-1.
Figure Q10: The Cr2O3-value and uncertainty in measurement model 3 for test sample
in row 417, while the uncertainty value of Cr2O3 is in cell C414. The
illustrates that the CSE, calibration slope and the intercept on the Y-axis
are the main contributors to Cr2O3's uncertainty for test sample MzBP-1.
Figure Q11(a): The results and associated uncertainties from measurement model 3 are
arranged from columns C to AG in rows 424 and 425 for use in the
measurement model 4 for test sample MzBP-1 determined by the Kragten
method. In this figure, SiO2 to Na2O are in columns C to R.
Figure Q11(b):K2O to Cr2O3-values from measurement model 3 are in columns S to X,
while the flux components as calculated in model 2 are in columns Z to AC.
The sum of results in model 3 and its uncertainty are in AD and AE, with
the LOI-value and uncertainty as calculated previously (Fig. 4.17) are in
columns AF and AG. All the values in (a) and (b) refer to relevant cells to
prevent errors when copying and pasting values.
Figure Q12: The final SiO2-value and uncertainty calculated in measurement model 4 for
test sample MzBP-1 by the Kragten method. It is from rows 428 to 452 in
in row 452, while the uncertainty value of SiO2 is in cell C448.
Figure Q13: The final TiO2-value and uncertainty calculated in measurement model 4
for test sample MzBP-1 by the Kragten method. It is from rows 455 to 479
in the same worksheet in Fig. 4.17. The uncertainty budget for this model
is in row 479, while the uncertainty value of TiO2 is in cell C475.
Figure Q14: The final Al2O3-value and uncertainty calculated in measurement model 4
is in row 506, while the uncertainty value of Al2O3is in cell C502.
Figure Q15: The final Fe2O3-value and uncertainty calculated in measurement model 4
is in row 534, while the uncertainty value of Fe2O3is in cell C530.
Figure Q16: The final MnO-value and uncertainty calculated in measurement model 4
is in row 562, while the uncertainty value of MnO is in cell C558.
Figure Q17: The final MgO-value and uncertainty calculated in measurement model 4
is in row 590, while the uncertainty value of MgO is in cell C586.
Figure Q18: The final CaO-value and uncertainty calculated in measurement model 4 for
in row 619, while the uncertainty value of CaO is in cell C615.
Figure Q19: The final Na2O-value and uncertainty calculated in measurement model 4
is in row 648, while the uncertainty value of Na2O is in cell C644.
Figure Q20: The final K2O-value and uncertainty calculated in measurement model 4 for
in row 677, while the uncertainty value of K2O is in cell C673.
Figure Q21: The final P2O5-value and uncertainty calculated in measurement model 4
is in row 706, while the uncertainty value of P2O5 is in cell C702.
Figure Q22: The final Cr2O3-value and uncertainty calculated in measurement model 4
is in row 735, while the uncertainty value of Cr2O3 is in cell C731.
List of Tables
Table 1.1: Uncertainty of measurement requirements as listed in

ISO/IEC 17025:2017 ............................................................................... 2
Table 2.1: Main statistics equations used in this study. (Note: For a set of replicate
measurements, n = amount of replicates, for i = 1, 2, .. , n, xi = measured
values, N = amount of measurements in a population) ........................... 8
Table 2.2: Metrological terms used in this study as defined in VIM[10] .................... 10
Table 2.3: Common and specific actions during method validation, specifying the
metrological traceability and estimation of uncertainty of measurement
............................................................................................................ ..19
Table 2.4: Evaluation types for determining a standard uncertainty and usage in the
GUM framework and MCM. GUM regards all systematic errors have been
corrected for and that the only association with systematic error in the
measurement model is the uncertainty of this correction itself [31]. This
table is based on the summary in Theodorou's thesis[32]. (𝑥𝑖 = input
quantity, PDF = probability distribution frequency) ................................ 22
Table 2.5: The evaluation and calculation steps for uncertainty of measurement
according to the GUM framework[4] ....................................................... 25
Table 2.6: The three top-down approaches and the equations to calculate standard
uncertainty of measurement[6] ............................................................... 33
Table 3.1: Measurand’s spectral line properties[53]. The spectral line for all the listed
elements is K-L3 (also known as Kα1-line) ........................................... 41
Table 3.2: Sample types for major element analysis by the study test method ...... 42
Table 3.3: Instruments used in the sample preparation stage of the test method .. 45
Table 3.4: The sample preparation stage for WD-XRF analysis ............................ 50
Table 3.5: The three Lithium borate fluxes were used in the study and are produced
by Claisse. The terms Flux A, Flux B and Flux C are for operational
purposes ............................................................................................... 51
Table 3.6: The instrumental parameters and available settings of the PANalytical
Zetium ................................................................................................... 54
Table 3.7: The main instrumental parameters for the spectral lines. Due to the non-
wetting agent, bromide is present in each glass disk, and Br Lα1,2 is
measured to correct for peak overlapping with the Al Kα1,2-peak. The full
detail of the instrumental setup is detailed in Figure 4.1 ....................... 55
Table 3.8: Sample types in SuperQ ....................................................................... 62
Table 3.9: Line-overlap corrections applied by SuperQ software ........................... 65
Table 3.10: Effects under different precision conditions ........................................... 68
Table 3.11: Evaluation of a concentration value's uncertainty based on the available
information and the calculation of the associated standard uncertainty…
.............................................................................................................. 71
Table 3.12: The calibration regression line's goodness of fit as calculated per selected
EWF by SuperQ[76]. 𝐶𝑠𝑡𝑑 is the assigned concentration of the standard
glass disk, 𝐶𝑐𝑎𝑙𝑐 the calculated concentration, 𝐶0 the concentration at
(0,0), the constant is a value the user enters into SuperQ and has the
same unit as the concentration, k is the number of coefficients in the
regression[76] ......................................................................................... 73
Table 3.13: The top-down approach to establish a test sample's uncertainty of
measurement using QC, replicates and PT or CRM results .................. 77
Table 3.14: EXCEL formulas[47] to select a random value in a value range. RAND() is
an EXCEL formula to calculate random values between 0 and 1. a, b, and
c refer to cell addresses in the EXCEL spreadsheet for the input value (a),
standard uncertainty (b) and the half-width interval (c) ......................... 79
Table 4.1: The main method validation parameters of the measurands and their
values .................................................................................................... 89
Table 4.2: Evaluation of each measurand's trueness. The En and recovery values
are listed in Appendix E......................................................................... 94
Table 4.3: Mass of the samples and flux for the six split samples of a SARM-36
aliquot to investigate the dilution effect ................................................. 94
Table 4.4: Summary of the split samples of SARM-36 to evaluate the mass effects
.............................................................................................................. 92
Table 4.5: LOI-analysis of a limestone test sample under repeatability and
intermediate precision conditions .......................................................... 95
Table 4.6: Statistics (average, standard deviation and % relative standard deviation
of six results) of six fusion disks prepared from four aliquots of SARM-36
using different fluxers and moulds. A column of results represents an
aliquot’s split statistics, while the different columns represent the effects
of the fluxer and moulds. The full set of results is presented in
Appendix I ............................................................................................. 96
Table 4.7: Summary of the 189 major element results on a fusion disk to evaluate
the instrumental drift for two months ..................................................... 98
Table 4.8: Summary of two-factor ANOVA at alpha level (α) of 0.05 to evaluate eight
sample cups in two measuring positions of the Zetium XRF spectrometer
............................................................................................................ 100
Table 4.9: LOI’s concentration range, average standard deviation and % relative
standard deviation based on duplicate results from samples analysed at
the CGS during the past five years using the study test method. The
values are used for the quantification of LOI’s precision in the EXCEL
templates described in 4.3 .................................................................. 102
Table 4.10: MzBP-1's input quantity and associated standard uncertainty values for
measurement model 1......................................................................... 108
measurement model 2......................................................................... 109
measurement model 3. Ei is the calibration slope, Di the intercept on the
Y-axis and P the precision of measuring cup, dilution or fluxer ........... 110
Table 4.13: Correlation between the three variables in measurement model 1. Cr is
the mass of the crucible, Cr100 is the mass of crucible and sample after
drying at 100 °C, while Cr1000 is the mass of crucible and sample after
roasting at 1000 °C.............................................................................. 111
Table 4.14: The correlation between the variables in measurement model 4. The
largest correlation is between Li2O and B2O3 ...................................... 111
Table 4.15: The formulas to calculate the sensitivity coefficients for an input variable
in its measurement model. The corresponding formula calculates the
value of each sensitivity coefficient in an EXCEL template. The calculated
sensitivity coefficients are for SiO2 as calculated by the EXCEL template
and the NIST uncertainty machine for test sample
MzBP-1 ............................................................................................... 112
Table 4.16: Setup of the four measurement models in the EXCEL template for the
MCM approach. The columns containing the specific measurement model
are illustrated in the referred figures. Figures for Model 3 are in Appendix
N, while Model 4’s figures are in Appendix O ...................................... 122
Table 4.17: Arrangement of the four measurement models in the EXCEL template for
the Kragten method ............................................................................. 131
Table 4.18: Statistical data used in the Nordtest based EXCEL template developed
for the study ........................................................................................ 135
Table 4.19: Regression line properties for the major elements by XGENLINE v8.1.
‘1/slope’ is the inverse of XGENLINE’s slope and is compared with the
regression slopes from SuperQ ........................................................... 140
Table 4.20: Results (%) of test material SARM-36 C1 and uncertainties as evaluated
by the different approaches ................................................................. 141
Table 4.21: Results (%) of test material MzBP-1 and uncertainties as evaluated by the
different approaches ........................................................................... 142
Table 4.22: Results (%) of test material Li3 6c and uncertainties as evaluated by the
different approaches ........................................................................... 142
Table 4.23: Results (%) of test material G00006619001 and uncertainties as
estimated by the different approaches ................................................ 143
Table 4.24: Indirect determination of the uncertainty budget for SiO 2 in the GUM
framework EXCEL template for the four test samples. The column
headings are explained in the text ....................................................... 145
Table 4.25: Indirect determination of the uncertainty budget for TiO 2 in the GUM
headings are explained in the text....................................................... 145
Table 4.26: Indirect determination of the uncertainty budget for Al2O3 in the GUM
Table 4.27: Indirect determination of the uncertainty budget for Fe2O3 in the GUM
Table 4.28: Indirect determination of the uncertainty budget for MnO in the GUM
Table 4.29: Indirect determination of the uncertainty budget for MgO in the GUM
headings are explained in the text....................................................... 146
Table 4.30: Indirect determination of the uncertainty budget for CaO in the GUM
Table 4.31: Indirect determination of the uncertainty budget for Na2O in the GUM
Table 4.32: Indirect determination of the uncertainty budget for K 2O in the GUM
Table 4.33: Indirect determination of the uncertainty budget for P2O5 in the GUM
Table 4.34: Indirect determination of the uncertainty budget for Cr2O3 in the GUM
Table A1 Reported information for uncertainty evaluation for the CRMs and RMs
used in the test method calibration.
Table B1: The used CRM and RMs in the calibrations of the test method, their
description and the flux and fluxer used to prepare the fusion disk.
Table D1: The assigned concentration values (%), reported or calculated standard
deviations (s) and calculated standard uncertainties (u) of the CRMs and
RMs used in the calibration application in SuperQ. In cases of no reported
standard deviation or available information for its calculation, standard
uncertainty is calculated as described in Table 3.11.
Table D2: The calculated concentrations (%) of the fusion disks after excluding LOI
values in Table D1 and including the flux components (Li2O and B2O3).
These values are entered into the calibration application in SuperQ for
regression plots ag-ainst the matrix corrected intensities of the
measurands.
Table E1: Trueness of the test method for SiO2.
Table E2: Trueness of the test method for TiO2.
Table E3: Trueness of the test method for Al2O3.
Table E4: Trueness of the test method for Fe2O3.
Table E5: Trueness of the test method for MnO.
Table E6: Trueness of the test method for MgO.
Table E7: Trueness of the test method for CaO.
Table E8: Trueness of the test method for Na2O.
Table E9: Trueness of the test method for K2O.
Table E10: Trueness of the test method for P2O5.
Table E11: Trueness of the test method for Cr2O3.
Table E12: Trueness of the test method for LOI.
Table F1: The analysis of eleven test samples of a limestone on five pressed powder
pellets by WD-XRF.
Table K1: The input variables (g), standard uncertainties, and sensitivity coefficients
propagated to provide uncertainty of measured LOI (%) in measurement
model 1. The correlation coefficients are listed in Table 4.6. The LOI
results and calculated u(LOI) values are used in Measurement model 4.
Table K2: Measurement model 2 results for the calculation of Li2O and B2O3 in the
fusion disks of four test samples. The Li2O and B2O results and their
uncertainties are used in Measurement model 4.
Table K3: Concentration of SiO2 as calculated by Measurement model 3 and its
associated uncertainty for the four test samples. The sensitivity coefficient
value for precision is the same for the cup, fluxer and dilution and is
reported as c(P).
Table K4: Concentration of TiO2 as calculated by Measurement model 3 and its
reported as c(P).
Table K5: Concentration of Al2O3 as calculated by Measurement model 3 and its
reported as c(P).
Table K6: Concentration of Fe2O3 as calculated by Measurement model 3 and its
reported as c(P).
Table K7: Concentration of MnO as calculated by Measurement model 3 and its
reported as c(P).
Table K8: Concentration of MgO as calculated by Measurement model 3 and
its associated uncertainty for the four test samples. The sensitivity
coefficient value for precision is the same for the cup, fluxer and dilution
and is reported as c(P).
Table K9: Concentration of CaO as calculated by Measurement model 3 and its
associated uncertainty for four test samples. The sensitivity coefficient
reported as c(P).
Table K10: Concentration of Na2O as calculated by Measurement model 3 and its
reported as c(P). Although the concentration for sample Li3 6c is below
the lower limit of detection, it and its associated uncertainty are still used
in Measurement model 4.
Table K11: Concentration of K2O as calculated by Measurement model 3 and its
Table K12: Concentration of P2O5 as calculated by Measurement model 3 and its
reported as c(P).
Table K13: Concentration of Cr2O3 as calculated by Measurement model 3 and its
reported as c(P).
Table K14: The input, uncertainty and sensitivity variables and their values in
Measurement model 4 for SiO2. The concentration and uncertainty values
for the four test materials are reported in the last two rows.
Measurement model 4 for TiO2. The concentration and uncertainty values
Measurement model 4 for Al2O3. The concentration and uncertainty
values for the four test materials are reported in the last two rows.
Measurement model 4 for Fe2O3. The concentration and uncertainty
Measurement model 4 for MnO. The concentration and uncertainty values
Measurement model 4 for MgO. The concentration and uncertainty values
Measurement model 4 for CaO. The concentration and uncertainty values
Measurement model 4 for Na2O. The concentration and uncertainty
Measurement model 4 for K2O. The concentration and uncertainty values
Measurement model 4 for P2O5. The concentration and uncertainty values
Measurement model 4 for Cr2O3. The concentration and uncertainty
Table R1: Concentration and uncertainty values calculated by XGENLINE.
Table S1: Uncertainty budget of measurement model 1. All values are in %.

Table S2: Uncertainty budget of measurement model 2 to determine Li2O. All values
are in %.
Table S3: Uncertainty budget of measurement model 2 to determine B 2O3. All
values are in %.
Table S4: Uncertainty budget of measurement model 3 to determine the major
elements by SuperQ. All values are in %.
Table S5: Uncertainty budget of measurement model 4 to determine SiO2 in
measurement model 4 (shaded rows). All values are in %. The
contributors to SiO2’s uncertainty are the major elements from SuperQ
(rows 3 to 13), flux components and LOI.
Table S6: Uncertainty budget of measurement model 4 to determine TiO2 in
contributors to TiO2’s uncertainty are the major elements from SuperQ
Table S7: Uncertainty budget of measurement model 4 to determine Al2O3 in
contributors to Al2O3’s uncertainty are the major elements from SuperQ
Table S8: Uncertainty budget of measurement model 4 to determine Fe2O3 in
contributors to Fe2O3’s uncertainty are the major elements from SuperQ
Table S9: Uncertainty budget of measurement model 4 to determine MnO in
contributors to MnO’s uncertainty are the major elements from SuperQ
Table S10: Uncertainty budget of measurement model 4 to determine MgO in
contributors to MgO’s uncertainty are the major elements from SuperQ
Table S11: Uncertainty budget of measurement model 4 to determine CaO in
contributors to CaO’s uncertainty are the major elements from SuperQ
Table S12: Uncertainty budget of measurement model 4 to determine Na 2O in
contributors to Na2O’s uncertainty are the major elements from SuperQ
Table S13: Uncertainty budget of measurement model 4 to determine K 2O in
contributors to K2O’s uncertainty are the major elements from SuperQ
Table S14: Uncertainty budget of measurement model 4 to determine P 2O5 in
contributors to P2O5’s uncertainty are the major elements from SuperQ
Table S15: Uncertainty budget of measurement model 4 to determine Cr2O3 in
contributors to Cr2O3’s uncertainty are the major elements from SuperQ
List of Abbreviations
Abbreviation Description
ANOVA Analysis of variance
CGS Council for Geoscience
CRM Certified reference material
CRMs Certified reference materials
CSE Counting statistical error
EWF Error weighting function
FP Fundamental Parameters
GUM Guide to the expression of uncertainty in measurement
GLP Good Laboratory Practices
IAG International Association of Geoanalysts
IUPAC The International Union of Pure and Applied Chemistry
ISO International Organization for Standardization
LOD Limit of detection
LOI Loss on ignition (at 1000 °C)
LOQ Limit of quantification
MCM Monte Carlo method
MU Uncertainty of measurement
PDF Probability distribution frequency
PT Proficiency testing programme
RM Reference material
SANAS South African National Accreditation System
SI International System of Units
UPS Uninterrupted power supply system
VIM International Vocabulary of Metrology
WD-XRF Wavelength dispersive X-ray fluorescence spectrometry
XRF X-ray fluorescence spectrometry
Chapter 1
Chapter 1:
Introduction and problem statement
1.1 Introduction
In today's analytical environment, calculation of the uncertainty of measurement is not

a luxury but an integral part of any measured result in an analytical laboratory.
Uncertainty of measurement is one of three indispensable parameters to ensure the
quality of measurement results – the others are metrological traceability and method
validation[1]. Without any stated uncertainty, technical personnel cannot evaluate the
quality of the result, and the customer will be unable to make informed decisions, such
as whether to use a product according to a set of specifications. The customer can
also use the uncertainty of measurement to evaluate the accuracy of the reported
result[2]. Furthermore, ISO/IEC 17025:2017[3] requires that all sources of uncertainty
of measurement are identified and that all significant contributing factors be evaluated
(Table 1.1).
The Council for Geoscience (CGS) Laboratory aims to obtain ISO/IEC 17025
accreditation for major and trace element analysis by wavelength X-ray fluorescence
spectrometry (WD-XRF). During the literature study to obtain at least a template or
step-by-step procedure for the estimation of uncertainty of measurement in WD-XRF
test methods, numerous detailed publications were found for most of the analytical
techniques except for WD-XRF. The main reference document for estimating the
uncertainty of measurement is the Guide to the Expression of Uncertainty in
Measurement[4], also known as the GUM. However, the GUM is not an easy document
to digest and is not accessible for many technical personnel in an XRF laboratory.
1
Chapter 1
1.2 Motivations and objectives of the study
It is an ISO/IEC 17025:2017[3] requirement for an accredited test method to have an

established process for the estimation of uncertainty of measurement. Table 1.1 lists
the items in ISO/IEC 17025:2017 requiring this estimation process for testing and
calibration laboratories. In ISO/IEC 17025:2017, the uncertainty of measurement is
not only part of the metrological traceability and validation process for test methods
but is also part of conformity statements should the customer require it (Clauses 7.1.3
and 7.8.6). Without reported uncertainty of measurement, conformity assessment is
not viable.
Table 1.1: Uncertainty of measurement requirements as listed in ISO/IEC

17025:2017[3]
Clause number
in ISO/IEC Statement
17025:2017
The equipment used for measurement shall be capable of achieving the …
6.4.5
measurement uncertainty…..
Measuring equipment shall be calibrated when:
6.4.6
… measurement uncertainty affects the validity of the reported results….
The laboratory shall establish and maintain metrological traceability of its
6.5.1 measurement results by means of a documented unbroken chain of calibrations,
each contributing to the measurement uncertainty …
7.1.3 When the customer requests a statement of conformity to a specification …
The laboratory shall use appropriate methods … for evaluation of the
7.2.1
measurement uncertainty …
The performance characteristics of validated methods … can include . .
7.2.2.3
measurement range, accuracy, measurement uncertainty of the results.
7.6.1 Laboratories shall identify the contributions to measurement uncertainty. . .
A laboratory performing calibrations … shall evaluate the measurement
7.6.2
uncertainty for all calibrations.
7.6.3 A laboratory performing testing shall evaluate measurement uncertainty. …
… calibration certificates shall include the following:
a) the measurement uncertainty of the measurement result . . ;
7.8.4.1
b) the conditions … under which the calibrations were made that have an
influence on the measurement results;
7.8.6 Reporting statements of conformity
Googling "uncertainty of measurement for XRF" produced about 317 000 results (as
of 15 November 2019), but no document detailing the evaluation of uncertainty of
2
Chapter 1
measurement for all XRF analytical processes from sample preparation to the
production of results was found. It could be that some laboratories already did such
work but did not make it available to the scientific public. Publications for the estimation
of uncertainty for XRF are available for either a part of the XRF analytical process or
for the final uncertainty of measurement as deduced from the calibration graphs or
quality control graphs. These available publications will be discussed in Chapter 2.
This study aims to present an approach based on GUM[4] to estimate the uncertainty
of measurement of major elements on fused beads analysed by WD-XRF. All sources
of error, both systematic and random, are to be identified in the whole analytical
process, from sample preparation to the measuring stage of major elements by WD-
XRF. Systematic errors are to be eliminated and, if not, be minimized and then
quantified with the associated random errors. This process of error identification and
quantification will contribute to:
i. the validation of the major element analysis on fused beads by the WD-XRF
test method at the CGS,
ii. the verification of the robustness of the test method,
iii. establishing the metrological traceability of the results,
iv. the estimation of uncertainty of measurement at specific confidence levels, and
v. the development of a template to estimate the uncertainty of measurement in
WD-XRF test methods.
Various approaches to obtain a practical process to estimate the uncertainty of

measurement by WD-XRF will be investigated.
3
Chapter 1
1.3 Problem statement
The GUM[4], which is internationally the main reference document for the evaluation of
uncertainty of measurement, prescribes a specific approach to evaluate the
uncertainty of measurement. This process is based on frequentist statistics, and each
significant contributing uncertainty component is identified and quantified according to
the evaluation type in terms of obtained statistical information. Furthermore, the
combined uncertainty and confidence levels utilize formulas depending on whether
there is any covariance between the contributing components. This detailed and
complex approach is also called the bottom-up approach as each uncertainty
component in the whole analytical process is identified and quantified. The statistical
principles and steps as prescribed by the GUM will be discussed during the literature
review in Chapter 2.
This bottom-up approach in XRF is labelled as a first-principles approach[5], where all

sources of variation are identified and summarised to calculate the uncertainty of
measurement (Equation 1.1) and is "inconvenient and time-consuming" (p.11-10) [5].
This approach is a simplified GUM approach, and it does not give direction on how to
treat the separate uncertainty components in Equation 1.1 according to the evaluation
types, which are discussed in Chapter 2 during the literature review. The bottom-up
4
Chapter 1
approach both assists in identifying sources of systematic errors and minimizing such
errors.
On the other hand, the top-down approach is much less complex than the bottom-up
approach as it utilizes quality control (QC) and validation data to estimate the total
uncertainty of each test quantity. It is described in a Nordtest technical report [6].
Fundamentally this approach is also a simplified GUM approach aiming to be more
practical and user-friendly in the sometimes complex process of estimating the
uncertainty of measurement. The differences between the top-down and bottom-up
approaches are also outlined in Chapter 2.
Although the bottom-up approach is time-consuming and requires a sound

understanding of classic (or frequentist) statistics, the separate steps to identify and
quantify errors in the study test method should contribute to a reliable quantification of
uncertainty of measurement in WD-XRF.
1.4 Key objectives of the study
Main sources of errors on the measurement results are to be identified in the test
method from sample preparation to the measuring process. This process of error
identification and calculations involves:
i) the validation of the test method,
ii) the traceability of the test results, and
iii) the estimation of uncertainty of measurement for test samples analysed by
WD-XRF.
The study's main aims are to compare bottom-up and top-down approaches for the
estimation of uncertainty in WD-XRF analysis of major elements in a test sample and
to set up a template to estimate the uncertainty of measurement by WD-XRF.
5
Chapter 2
Chapter 2:
Theoretical background and literature review
2.1 Introduction
This chapter outlines the terminology used in the theory of uncertainty of

measurement, which includes metrological terms, descriptive and inferential statistics,
method validation, metrological traceability, and uncertainty of measurement.
Furthermore, the description of internationally accepted processes to estimate the
uncertainty of measurement and salient aspects of existing studies in WD-XRF are
discussed.
2.2 Concepts in the uncertainty of measurement
2.2.1 Descriptive and inferential statistics
Descriptive statistics provides tools to describe and interpret a set of data, while
inferential statistics assist one to use a set of data to make assumptions about an
unknown population leading to estimate population parameters[7]. A population can
contain some possible but not measured values[8]. In this study, the actual measured
values are regarded as representative of a population as it is impossible to make all
measurements for this population. Parameters such as the average, variance and
standard deviation based on the sample data are then used to estimate the
corresponding population parameters (Figure 2.1). Tools in inferential statistics
include the population plot and significance testing (Table 2.1).
6
Chapter 2
Figure 2.1: Parameters of a normal distribution plot of a population (the Gauss

distribution), population average (µ) and mean (σ).
If all the possible measured values around its mean are plotted, and the spread of
those values is symmetrical, a bell-shaped figure is formed depending on the
population's mean (µ) and standard deviation (σ). Figure 2.1 depicts the normal or
Gaussian distribution with its equation in Table 2.1. The curve area between ±3 σ of
the mean contains 99.73% of the population values while the areas between ±2 σ and
±1 σ of the mean contain 95.45% and 68.27% of the population values, respectively[4].
The normal distribution is the basis of the main statistics in the evaluation of
uncertainty in measurement[4]. In the normal distribution, the expectation of a
measurement value being close to the population mean is high, and the interval half
widths are expected to be equal, i.e. (µ-1σ) = (µ+1σ), (µ-2σ) = (µ+2σ) and (µ-3σ) =
(µ+3σ).
If the sample size is too limited to assume a normal distribution, a t-distribution, also
termed the Student's t-distribution, is used instead. The t-distribution also has a bell-
shaped figure but has a lower probability density close to the mean and higher in the
wings than the normal probability distribution. Its shape depends on its associated
degrees of freedom which is the sample amount minus one. Probability values based
7
Chapter 2
on the degrees of freedom are deduced from a table listing the critical values of
t-distribution tables available in scientific literature such as the NIST Engineering and
Statistics Handbook[9]. In cases of infinite degrees of freedom, the t-distribution has
the same shape as the normal probability distribution.
This study's other probability distribution types are the rectangle and triangular
probability distributions. In the rectangle probability distribution, a sample can be
expected anywhere in a range. On the other hand, in the triangular probability
distribution, a sample is expected more to be in the middle of a range than at the range
ends.
Table 2.1: Main statistics equations used in this study. (Note: For a set of replicate
measurements, n = amount of replicates, for i = 1, 2, .. , n, x i = measured
values, N = amount of measurements in a population)
Descriptive statistics
Parameter Equation/description
Mean of a sample set
Variance of a sample set
∑𝑛𝑖=1(𝑥𝑖 − 𝑥)2
Standard deviation of a sample set 𝑠(𝑥𝑖 ) = √
(𝑛 − 1)
Coefficients of variation (CV) or 𝑠(𝑥𝑖 )

𝐶𝑉 =
relative standard deviation (R.S.D.) 𝑥
∑𝑁
𝑖=1 𝑥𝑖
Mean of population 𝜇=
𝑁
8
Chapter 2
Table 2.1: Continued
Descriptive statistics
Parameter/Test Equation/description
∑𝑛
𝑖=1(𝑥𝑖 −𝑥)
2
𝜎= √ for a sample set,
𝑛
Standard deviation of a population ∑𝑁
𝑖=1(𝑥𝑖 −𝜇)
2
𝜎= √ for a population containing all
𝑁
sample sets
Degrees of freedom ν=n–1
Inferential statistics
Parameter/Test Equation/description
−(𝑥−𝜇)2
𝑒 2𝜎 2
Normal distribution of population 𝑦=
𝜎√2𝜋
To evaluate the difference between
- means of two sample sets
t-test - pairs of different measurements
- mean of a sample set and a value of a certified
reference material
To evaluate the difference between standard

F-test
deviations of two sample sets
2.2.2 Metrological terms

Metrology is termed from the Greek metron and logos, which can be translated to
measure and reason. It evolved from the Egyptian pharaoh's forearm length as a
standard for a cubit to build the pyramids to establishing the International System of
Units (SI), which is intrinsically part of the modern world. International Vocabulary of
Metrology (VIM)[10] defines metrology as the science of measurement, which also
includes the application of this science (VIM 2.2), i.e. from the acts of measurements
to the use of measurement results in decision making. Metrological terms are listed in
Table 2.2 with their VIM definitions and discussed in this section.
9
Chapter 2
Table 2.2: Metrological terms used in this study as defined in VIM[10].

VIM item Metrological
Definition
no term
process of experimentally obtaining one or more
2.1 measurement quantity values that can reasonably be attributed to a
quantity
2.3 measurand quantity intended to be measured
set of quantity values being attributed to a measurand
together with any other available relevant information
measurement
2.9 NOTE 2 A measurement result is generally expressed
result
as a single measured quantity value and a
measurement uncertainty.
2.11 true value quantity value consistent with the definition of a quantity
closeness of agreement between a measured quantity
2.13 accuracy
value and a true quantity value of a measurand
closeness of agreement between the average of an
2.14 trueness infinite number of replicate measured quantity values
and a reference quantity value
closeness of agreement between indications or
measured quantity values obtained by replicate
2.15 precision
measurements on the same or similar objects under
specified conditions
measured quantity value minus a reference quantity
2.16 error
value
component of measurement error that in replicate
systematic
2.17 measurements remains constant or varies in a
error
predictable manner
2.18 bias estimate of a systematic measurement error
component of measurement error that in replicate
2.19 random error
measurements varies in an unpredictable manner
non-negative parameter characterizing the dispersion of
measurement
2.26 the quantity values being attributed to a measurand,
uncertainty
based on the information used
property of a measurement result whereby the result
metrological can be related to a reference through a documented
2.41
traceability unbroken chain of calibrations, each contributing to the
measurement uncertainty
provision of objective evidence that a given item fulfils
2.44 verification
specified requirements
verification, where the specified requirements are
2.45 validation
adequate for an intended use
10
Chapter 2
As noted in the VIM's definition of measurement in Table 2.2, it is a process of

various action steps to estimate a measurand's value. In this study, the different
action steps include weighing and ashing the sample to determine loss of ignition
(L.O.I.) at 1000 ⁰C and then the analysis by WD-XRF. The study measurands are
principal oxides present in powdered sample form (< 75 µm), usually received from
the Sample Preparation Unit in the CGS Laboratory after milling. These include
SiO2, TiO2, Al2O3, Fe2O3(t), MnO, MgO, CaO, Na2O, K2O, P2O5, Cr2O3 and L.O.I.
The measurement result consists of the following information:

i) the estimated quantity of the measurand,
ii) a unit, which should be an SI unit, and
iii) associated uncertainty of measurement.
The term 'true value' is misleading as it can never be known, and with this value
unknowable, error calculations cannot be established[11]. The GUM[4] notes that the
true value results from a perfect measurement, i.e. this term is 'an idealized concept'
(GUM item D.3.5), and its use should be discouraged. However, the measurement
result is regarded as the best estimate of the measurand's true value (GUM item
D.3.1). It also means that this measurement can have a set of true values in a stated
range of uncertainty. In metrology, the assigned value of a certified reference material
(CRM) is viewed as 'a surrogate for a true quantity value'[12]. In this study, values with
established metrological traceability are regarded as 'true values' and can be found in
the certificates of CRMs and accredited proficiency test (PT) samples.
The true value concept is needed in the determination and quantification of errors
which in chemistry can be classified as either systematic or random errors[12]. An error
can be quantified as the difference between the measured and assigned value. It is
either positive or negative depending on the position of the measured to its assigned
value. It should be noted that the error concept in the evaluation of uncertainty
measurement is different from errors in everyday usage, where it could be the
consequence of a mistake. Therefore, the term 'error' is replaced by 'effect' in this
study to avoid the inherent conflict of the quantification of errors[13].
11
Chapter 2
Various authors[14][15][16][4] describe effects impacting the result and methods to quantify
them. Random effects can be observed mathematically as a dispersion of values
around a mean value and originate mainly from instrumental behaviour. They cannot
be eliminated but can be quantified by standard deviation. In contrast, systematic
effects can be seen in the difference between their mean and assigned values. Both
instrumental and sample behaviour contribute to systematic effects, which can be
minimized by applying background, line-overlap and matrix corrections in WD-XRF.
Precision of a test method quantifies the closeness between the various measurement
results to each other and is influenced by random effects. On the other hand, bias is
caused by systematic effects such as incorrect calibration of the instrument or
instrumental drift. Both precision and bias contribute to the accuracy of a test
method[17]. In Figures 2.2 (a) and (c), the small spread of replicate values reflects high
precision, while Figures 2.2 (b) and (d) depicts a flatter curve due to a broader spread
of values around its mean reflecting lower precision. Bias in Figures 2.2 (c) and (d)
indicates lower trueness in comparison with Figures 2.2 (a) and (b). An accurate result
has both high trueness and precision (Fig 2.2 (a)). Although systematic effects cannot
be eliminated, they can be corrected using appropriate adjustments to the test method.
Blunders such as operational mistakes and improper correction factors are to be
avoided and are not part of the uncertainty of measurement estimation process[14].
12
Chapter 2
(a) (b)
(d)
(c)
Figure 2.2: Precision, bias and accuracy. The blue markers represent replicate
measurement results, and the red marker the assigned value of the
measurement. (a) reflects the highest accuracy and (d) the least
accurate result and test method.
The effect framework in Figure 2.3 illustrates the contribution of effects to the
uncertainty of measurement, and it shows that trueness and precision are both to be
quantified as they contribute to optimal accuracy of a measurement. Therefore, both
trueness and precision are parameters to evaluate the quality of a measurement.
13
Chapter 2
Figure 2.3: The effect framework in terms of type, performance and expression and
their influence on the uncertainty of measurement. It is based on the error
framework in Eurachem Guide: The Fitness for Purpose of Analytical
Methods [15]. Systematic effects contribute to the trueness of a test
method and are quantified by the test method's bias, while the precision
of a test method is influenced by random effects and quantified by the
standard deviation of the random effects.
The effect contributions mean that the reverse can be applied so that the uncertainty
of measurement can be used to assess the quality of a measurement result[18].
14
Chapter 2
2.2.3 Method validation, metrological traceability and uncertainty of

measurement
Informed decisions are based on reliable measurements, and this need for reliable
measurements led to the concept of "Tested once, accepted everywhere" [18]. Reliable
measurements depend on:
i) competent staff,
ii) implemented quality management systems,
iii) validated test methods and
iv) traceability of measurements to measurement references.
Accreditation bodies such as the South African National Accreditation System

(SANAS) and International Organization for Standardization (ISO) ensure that the
accredited laboratory upholds the abovementioned dependencies and can
demonstrate its capability to undertake reliable measurements. Method validation,
metrological traceability and uncertainty of measurement are the three cornerstones
of reliable measurements.
As per VIM's definition of method validation, it is the verification that the test method
is fit for purpose. Method validation by a test laboratory is mainly in the form of a report
outlining the scope of the measurement process, the test conditions for the
measurement process and a list of method validation parameters to state that this
measurement process is fit for purpose in this laboratory. For a standard test method,
the method validation process is replaced by a method verification process to
demonstrate the test laboratory's ability to perform such a test[15].
Metrological traceability links the produced measurement result to one or more

reference standards in a documented process. Primary examples of reference
standards are mass pieces to verify a weighing balance and reference materials to
calibrate an instrument. This metrological traceability assists a laboratory in evaluating
a test method's performance (Figure 2.4).
15
Chapter 2
Figure 2.4: Metrological traceability from each result (y1, y2 and y3) to the SI unit.
The SI unit is linked to the certified reference material (CRM) and the
mass piece used in the sample preparation. The CRM value is used in
both the test method calibration and the evaluation of the test method.
Results (y1, y2, y3) are compared by their traceability to the CRMs and
the mass pieces used in the test procedures. (Adapted from a figure[19])
Uncertainty of measurement characterizes the dispersion of possible measurand

values and is expressed by a non-negative parameter to describe this range of
dispersion. The phrase 'based on the information used'[10] in VIM's definition of
uncertainty of measurement is significant because it indicates that for the same
measurement result, different values of uncertainty of measurement could be reported
depending on the usage and evaluation of available information. The main goal of
uncertainty of measurement evaluation should not be to obtain the smallest or best
possible uncertainty as this strategy could be too costly in terms of time and monetary
costs. Instead, the most optimal uncertainty which is fit for purpose should be the basis
of the evaluation strategy[20][21].
The tight relationship between metrological traceability, method validation and

uncertainty of measurement starts with the unambiguous identification of the
16
Chapter 2
measurand[21]. For example, total iron is analysed and reported as Fe2O3 and not FeO
in this study. Furthermore, the parameter for the measurand is to be specified, i.e.
whether the measurand results are for the submitted sample only or for the entity from
which the sample originated.
Results of the method validation are also used in the evaluation of the uncertainty of
measurement. All associated measurements and uncertainties of measurement
throughout the measurement process are recorded to establish the measurement
result's metrological traceability[19]. CRMs are also used to establish metrological
traceability and for trueness evaluation during the method validation.
The validation process is used to gather evidence to confirm that the analytical method
is fit for purpose and to evaluate the uncertainty of measurement. In the review of
metrology principles in the analysis of clinical samples by LA-ICP-MS, the authors[22]
noted that the use of calibrated instrumentation and the standard operating procedures
without establishing the three metrological cornerstones would not guarantee correct
results. In Chapter 3, the most important validation parameters such as working range,
linearity, detection limit, limits of quantification, trueness, precision, sensitivity and
robustness are defined and discussed for this study. Figure 2.5 and Table 2.3
illustrates the tight relationships between the three metrology cornerstones.
17
Chapter 2
Figure 2.5: A simplified illustration of the three metrology cornerstones to evaluate

a test method and individual measurement results. The red diamond
represents a CRM's certified value. The term 'Fit for purpose' of a test
method indicates that the precision and accuracy of the test method is
sufficient so that the end-user can make an informed decision on the
intended use of the produced results or analysed samples[23][24].
The three metrology cornerstones start with the specification of the measurand. The
second common step is the description of the measurement procedure, the
calculations used in the test method leading to the definition of the measurement
model relating the measurand to the input coefficients. Other steps are demonstrating
the test method's fitness for purpose and estimating the uncertainty of measurement.
18
Chapter 2
Table 2.3: Common and specific actions during method validation, specifying the
metrological traceability and estimation of uncertainty of measurement.
Specific actions
Common action
Metrological Uncertainty of
Method validation
traceability measurement
Set the scope of
Specify the measurand measurement and
required uncertainty
 Describe the
measurement
procedure,
 List the main method
calculations and
equations,
 Define the
measurement
conditions and
 Relate the output with
the input coefficients in
a measurement model
f(y)= (x1,…,xn)
Evaluate results of List the uncertainty
Demonstrate the Evaluate the accuracy
CRM or PT samples by sources, and evaluate
measurement procedure of the test method in
comparing the result each source according to
and test each significant terms of method
with the assigned evaluation type (Type A
influence coefficient validation parameters.
values. or B)*
Set up the measurement
Identify the relative
uncertainty budget and
importance of each
evaluate each error
influence coefficient
contribution.
List evidence of used
chemical and
Choose and apply Check the calibration
calibration certificates
measurement standards standards' uncertainties
(mandatory for ISO
17025)
Estimate the uncertainty Calculate the total/
of measurement relative MU
State test method's Establish test method's Report on test method's
Report
fitness for purposes metrological traceability MU
19
Chapter 2
2.3 Established approaches to estimate the uncertainty of

measurement
2.3.1 The Bureau International des Poids et Mesures (BIPM) and the GUM
In 1977, the BIPM was requested to investigate a methodology to evaluate and

express uncertainty in measurement, which led to GUM[4][25][26]. At that stage, there
was no accepted process internationally, similar to the International System of Units
(SI) to evaluate uncertainty in measurement. The BIPM established two working
groups in the Joint Committee for Guides in Metrology (JCGM).
The JCGM Working Group 1 published the GUM[4], while JCGM Working Group 2
produced the VIM[10]. The first version of GUM was published in 1993, and the most
recent version in 2008. As the GUM is mainly based on frequentist statistics with the
Central Limit Theorem as its foundation, the following supplements were published to
include uncertainty evaluation not covered in the GUM:
i) Evaluation of measurement data – Supplement to the "Guide to the expression
of uncertainty in measurement" – Propagation of distributions using a Monte
Carlo method[26]. It will be referred to as GUM-1 in this study.
ii) Evaluation of measurement data – Supplement 2 to the "Guide to the
expression of uncertainty in measurement" – Extension to any number of output
quantities[27] (GUM-2).
iii) Evaluation of measurement data – An introduction to the "Guide to the
expression of uncertainty in measurement" – and related documents[28] (GUM-
4).
iv) Evaluation of measurement data – The role of measurement uncertainty in
conformity assessment[29](GUM-6).
The BIPM states on its website[30] that two more supplements are still in preparation
stages:
v) Evaluation of measurement data – Supplement 3 to the "Guide to the
expression of uncertainty in measurement" – Modelling and
vi) Evaluation of measurement data – Applications of the least-squares method.
20
Chapter 2
GUM-1 and GUM2 use different methods to propagate uncertainties from the different
uncertainty sources, while GUM-4 outlines the content and use of each GUM
supplement document. GUM-6 describes how to assess the reported measurement
uncertainty to evaluate reported results for conformity to a set of standards or
specifications. It is mainly applicable in cases where decisions are based on product
specifications.
As noted previously, the main difference between GUM and GUM-1 is the method to
propagate. The propagation method in GUM is referred to as the GUM framework,
while the Monte Carlo method (MCM) is applied in GUM-1. According to GUM-4[28],
both methods share the main stages of formulation and calculation. In the formulation
stage, the following are defined:
i) the output quantity, Y, i.e., the measurand,
ii) identification of the input quantities (xi) upon which the output quantity
depends,
iii) the definition of the functional relationship between the output and input
quantities, and
iv) The input quantity's standard deviation is evaluated based on the used
statistical analysis to determine its standard uncertainty. Type A evaluation
of uncertainty uses statistical analysis of a series of values for an input
quantity, while Type B evaluation uses information obtained by non-
statistical means. Table 2.4 lists the equations of both evaluations, which are
then used in the GUM framework. This process also provides information
regarding the probability distributions required for the MCM.
v) The calculation stage involves propagating the different input quantities
described in sections 2.3.1.1 and 2.3.1.2.
21
Chapter 2
Table 2.4: Evaluation types for determining a standard uncertainty and usage in the
GUM framework and MCM. GUM regards all systematic effects have
been corrected for and that the only association with systematic effect in
the measurement model is the uncertainty of this correction itself[31]. This
table is based on the summary in Theodorou's thesis[32]. (𝑥𝑖 = input
quantity, PDF = probability distribution frequency)
Evaluation type and
Monte Carlo method (PDF
description of input GUM-framework
-type)
quantity
Type A
For a series of observations in a
Gaussian
A series of observations - measurement:
with assumed Gaussian 𝑢(𝑥𝑖 ) = 𝑠(𝑥𝑖 )
distribution For the average from m replicates:
(𝑥𝑖 = mean) 𝑠(𝑥𝑖 )
𝑢(𝑥𝑖 ) =
√𝑚
t-distribution
A series of observations -
with assumed non-
√𝑛 − 1 𝑠(𝑥𝑖 )
Gaussian distribution (𝑥𝑖 = 𝑢(𝑥𝑖 ) =
mean) √𝑛 − 3 √𝑚
Type B
Calibration certificate Gaussian
reporting expanded 𝑈
𝑢(𝑥𝑖 ) =
uncertainty U and coverage 𝑘
factor k (𝑥𝑖 = calibration
value)
Calibration certificate t-distribution
reporting expanded
uncertainty Up, coverage 𝑈𝑝
𝑢(𝑥𝑖 ) =
factor kp and stated 𝑘𝑝
effective degrees of
freedom νeff
(𝑥𝑖 = calibration value)
Manufacturer's 𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒 Gaussian
𝑢(𝑥𝑖 ) =
specification certificate 𝑘
reporting a tolerance range
at a confidence level k = 2 for confidence level of 95%
(𝑥𝑖 = specification value) k = 2.58 for confidence level of 99%
Rectangular
𝑥𝑖 = any value between the 𝑎−𝑏
𝑢(𝑥𝑖 ) =
upper limit of a and lower 2√3
limit of b.
Triangular
𝑥𝑖 = a value between the
𝑎−𝑏
upper limit of a and lower 𝑢(𝑥𝑖 ) =
2√6
limit of b, but is more likely
in the middle of the range.
22
Chapter 2
During the past ten years, numerous publications about the GUM and its supplements
indicate that a standardised approach for uncertainty measurement evaluation is not
as established as the SI. One of the GUM's scopes (item 0.3) [4] is to establish such a
standard, and after more than twenty years, a standardised approach to evaluate the
uncertainty of measurement is still evolving. The various inconsistencies[25] in the
GUM, GUM-1 and GUM-2 are contradictory to one of GUM's main goals for the
measurement uncertainty to be "internally consistent" as outlined in its scope[4]. The
main topics in the debates about the GUM and its supplements for the past six years
are:
- the terminology in measurement uncertainty as described in the GUM framework
documents[12][33][34].
- the case for or against the suitability of frequentist and Bayesian statistics in the
evaluation of measurement uncertainty[34][35][36][37], and
- the different approaches to the uncertainty of measurement evaluation [38][39]
There was even a proposal to withdraw the supplements GUM-1 and GUM-2 and the
current GUM be revised to accommodate new and accepted understandings of
statistics[18]. Updates to the GUM are ongoing using the presently accepted
interpretations of statistics[40]. Areas for updates are the determination of coverage
intervals and in vector measurements. However, the current GUM is still the main
reference document for measurement uncertainty internationally. The 20th-
anniversary issue for GUM in Metrology declared: "the GUM principles still apply" [42].
The simplified main GUM principles[42][43] are:
i) The concept of uncertainty of measurement.
ii) An unambiguous definition of the measurand.
iii) The measurement process where the input and output items are to be defined
clearly to account for all primary sources of measurement effects.
iv) Both systematic and random effects contribute to the uncertainty of
measurement, and their variance is processed identically.
v) Standard deviation is used to express uncertainty for each identified effect
source.
vi) Each uncertainty is then estimated into standard uncertainty based on the data
evaluation types A or B.
23
Chapter 2
vii) The standard uncertainties are combined using the propagation of uncertainties
which takes the form of the variance sum law.
viii) Uncertainty of measurement is expressed as standard deviation (standard
uncertainty) or as multiples of standard deviation with a specified coverage
factor.
2.3.1.1 The GUM framework
Table 2.5 outlines the GUM framework steps to estimate the uncertainty of
measurement. The establishment of the measurement model, i.e. the unambiguous
definition of the measurand, Y and the identification of all input quantities X i on which
Y depends, will be the first step in this evaluation process. A functional relationship
model Y and Xi is then arranged, with each input value Xi being estimated by xi, which
is the uncertainty source. Tools such as the Ishikawa diagram, also called the fishbone
diagram, is usually used to assist in this process and will be employed in this study.
For the quantification of all the significant uncertainty contributions, the available
information for each considered contribution is to be converted to a standard deviation
(standard uncertainty). Each standard uncertainty is calculated based on the
evaluation type (Type A or Type B) listed in Table 2.4. Correlation between any two
input quantities is to be determined as it affects the calculations for the combination of
all the uncertainty contributions (equation in step six). A correlation value of 1 indicates
strong interdependence between the two input quantities while a value of zero shows
total independence between the two input quantities, and a simpler step six (a)
equation is then used. The partial derivatives in step six (a) equation also serve as
sensitivity coefficients as each uncertainty contribution does not contribute equally to
the measurand's combined standard uncertainty but depends on the measurement
model.
The expanded uncertainty is calculated to report the measurement result with a level
of confidence (usually 95%). The estimated combined standard uncertainty is
multiplied with a coverage factor as detailed in step seven of Table 2.5. Reporting the
result with the uncertainty concludes the evaluation process of the GUM framework.
24
Chapter 2
Table 2.5: The evaluation and calculation steps for uncertainty of measurement
according to the GUM framework[4]
GUM
Description of evaluation and calculation process
step no
- Specify the measurand Y
- Identify all input qualities Xi on which Y depends
1
- Model functional relationship between Y and Xi
Y = f(X1, X2, …., XN)
Determine each input estimate of xi based on whether this value was obtained statistically
2
(Type A) or by other means such as calibration certificates (Type B).
Convert each input quantity, xi, into standard uncertainty u(xi) based on their evaluation
3
in the previous step.
Evaluate the covariance 𝑟(𝑥𝑖 , 𝑥𝑗 )between any two input estimates:

4 𝑠(𝑥𝑖 , 𝑥𝑗 )
𝑟(𝑥𝑖 , 𝑥𝑗 ) = 𝑟(𝑥𝑗 , 𝑥𝑖 ) =
𝑠(𝑥𝑖 ) 𝑠(𝑥𝑗 )
5 Calculate y, the estimate of the measurand Y from the function f in Step 2.

a) Calculate the combined standard uncertainty uc(y) from the standard uncertainties
and covariances associated with the input estimates:
𝑁 𝑁−1 𝑁
𝑢𝑐2 (𝑦) = ∑ 𝑐𝑖2 𝑢2 (𝑥𝑖 ) + 2 ∑ ∑ 𝑐𝑖 𝑐𝑗 𝑢(𝑥𝑖 )(𝑥𝑗 )𝑟(𝑥𝑖 , 𝑥𝑗 )

𝑖=1 𝑖−1 𝑗=𝑖+1
6 Only if co-variance is present
Where 𝑐𝑖 is the partial derivative of the function in terms of the input quantity, i.e.
𝜕𝑓 𝜕𝑓
𝑐𝑖 = and 𝑐𝑗 =
𝜕𝑥𝑖 𝜕𝑥𝑗
b) 𝑢𝑐 (𝑦) = √𝑢𝑐2 (𝑦)

For the determination of expanded uncertainty, U:
a) If there is no domination of a specific type of uncertainty: 𝑈𝑝 = 𝑘𝑝 𝑢𝑐 (𝑦) where k
is a confidence factor obtained from the normal distribution, and p is the
confidence level.
b) If a normal distribution cannot be assumed for 𝑢𝑐 (𝑦), the effective degrees of
freedom (νeff) is to be calculated. For normal distribution Type A evaluation of an
7 input coefficient, it is one less the number of observations. Otherwise, it is
calculated as follows:
𝑢4 (𝑦)
𝑣𝑒𝑓𝑓 = 4
𝑢 (𝑦)
where νi is the degrees of freedom of the input quantity
∑𝑖 𝑖
𝑣𝑖
The expanded uncertainty, in this case, is then: U = 𝑡𝑝 (𝑢)uc(y)

where 𝑡𝑝 (𝑢) is deduced from a table listing this value[4] based on the degrees of
freedom and associated fraction p of the distribution.
Report the measured report and its associated uncertainty or expanded uncertainty with
8
its associated confidence level.
25
Chapter 2
2.3.1.2 The Monte Carlo method (MCM)
GUM-4[28] states that the GUM framework is not applicable where:

i) the functional relationship is not linear,
ii) the uncertainty contributions are not approximately equal, i.e. the presence of
one or two uncertainties being large in comparison with the others,
iii) the probability distribution of the input quantities are not symmetric, i.e. is not
normal, and
iv) the probability distribution of the output quantity is asymmetric.
It is not always clear in advance whether the GUM framework is applicable for a
measurement model, and this is where the MCM is versatile. Furthermore, the
challenging partial derivatives in the GUM framework is avoided in the MCM. The
MCM uses pseudo-random values from each input quantity's probability distribution
(p(xj) in Figure 2.6) to be processed in the mathematical function for M times. The
coverage interval for the measurand is then the shortest interval length to cover a
specific coverage probability (p(y) in Figure 2.6) – even if the frequency distribution is
not symmetric. The standard uncertainty is determined by the simulations' standard
deviation using the M model values for yM in Equation 2.1[32].
1
𝑢(𝑦) = √𝑀−1 ∑𝑀
𝑟=1(𝑦𝑟 − 𝑦) Equation 2.1
26
Chapter 2
Figure 2.6: The Monte Carlo method (GUM-2)[26].
The MCM is also used to validate the uncertainty values estimated by the GUM
framework[26][44]. This evaluation involves comparing the coverage intervals from both
the GUM framework and MCM and determining whether it agrees within a specific
numeric tolerance. The comparison steps are as follows:
i) Determine the absolute difference between the right-hand endpoints of both the
confidence intervals as established by the GUM framework and MCM. This
value is then assigned to dhigh.
ii) The same step is done for the left-hand endpoints and the value assigned to
dlow.
iii) A numerical tolerance, δ, is compared with both dhigh and dlow.
iv) For the calculation of δ:
- Determine ndig, where ndig is the number of significant digits in the output
value.
- Express the output value in c x 10l, where c is an integer and l the number
of digits.
1
- δ= 10𝑙
2
27
Chapter 2
v) The GUM framework evaluation is validated if both d high and dlow are not larger
than δ. If dhigh or/and dlow is larger than δ, the MCM is the preferred evaluation
approach, and its resulting uncertainty evaluation is used instead.
Although the GUM framework and MCM create similar standard uncertainty values,
the differences become apparent when the measurement function is not linear. In such
cases, the MCM produces better uncertainty estimates than the GUM framework[44].
2.3.1.3 Software tools for GUM based calculations
A spreadsheet is a popular tool to create templates for calculating uncertainty of

measurement. The Kragten approach[45], for example, simplified the partial derivation
calculations in step six of Table 2.5 and broke down the GUM framework calculations
into a spreadsheet while co-variation coefficients are included[39] in the calculations if
present. The Kragten approach provides acceptable accuracy in the approximation of
uncertainty, even in cases of non-linearity of the measurement model[46]. Microsoft
EXCEL 2010 is suitable for the MCM as more than 106 trials were successfully
tested[47]. A practical example[44] is the basis of the MCM of this study.
The NIST Uncertainty Machine is a web application[48] that uses both the GUM
framework and the MCM to evaluate the uncertainty of measurement. Up to fifteen
input coefficients with their distribution type and associated interval are entered into
the assigned fields. The associated function expressed in R then process the values
for the input coefficients. Both approaches' results are compared with each other.
2.3.2 Top-down and bottom-up approaches
The detailed identification, evaluation and propagation of uncertainty components by

the GUM framework and the MCM is regarded as the bottom-up approach, also called
the modelling approach. On the other hand, the top-down approach, also termed the
empirical approach, is not as exhaustive as the bottom-up approach. Variance data
from quality control (QC), method validation data or proficiency test results are used
instead (Figure 2.7). Another main difference between the two approaches is that the
bottom-up approach evaluates the uncertainty of a measurement result. The top-down
approach, in contrast, evaluates the uncertainty of a test method[49]. The bottom-up
28
Chapter 2
approach has been discussed in the description of the GUM framework and the MCM
in section 2.3.1.
Figure 2.7: The schematic differences between the bottom-up approach on the left
and the top-down approach on the right. The bottom-up approach
evaluates all (or at least the main) sources of uncertainty, while the top-
down approach uses results from QC, method verification/validation or
PT studies[50].
The top-down approaches are also based on GUM principles such as identifying of
the measurand, equal treatment of random and systematic effects, and the variances'
summation according to the variance sum law[49]. Furthermore, the estimation process
in equations 2.2(a) and (b) is also model-based, which is a GUM principle[15].
29
Chapter 2
Based on the used data used for the evaluation, there are three different top-down
approaches, and the organisation of data is outlined in Figure 2.8. Data from a single
laboratory is used in the intralaboratory approach, while more than one laboratory is
involved in the interlaboratory approach.
30
Chapter 2
Figure 2.8: The five different uncertainty of measurement evaluation methods

grouped into the bottom-up and top-down approaches (Based on
published figures[21][49]). The data types obtained from the replicate
measurements, CRMs and PT samples are discussed in the study.
ISO5725-1:1994[17] defines repeatability as 'precision obtained under repeatability

conditions', which means that the test samples were measured under identical
laboratory conditions in terms of site, environmental conditions, test method,
equipment and chemicals. Precision under reproducibility conditions are obtained from
the analysis of identical test samples, but in different laboratories, under different
operators, using different test methods and instruments. Intermediate precision is from
the same test samples but with some constant variables. Examples are:
(i) the same laboratory and test method, but different equipment or operators,
(ii) the same equipment is used but during different periods.
31
Chapter 2
The accuracy of a measurement method can be estimated by a basic model of a test

result[17], i.e. the test result obtained under repeatability conditions as expressed in
Equation 2.3.
𝑦 =𝑚+𝐵+𝑒 Eq. 2.3

where
𝑦 = test result
𝑚 = general expected value for result
𝐵 = bias (laboratory component).
𝑒 = random error
Replicate measurement data obtained under different laboratory conditions leads to

different standard deviation values. The resulting standard deviation terms used in the
top-down equations identifies the precision conditions. Table 2.6 lists the equations
for the uncertainty of measurement in the three top-down approaches and the data
types used in the evaluations.
There are views[51] stating no significant differences between the top-down and
bottom-up methods, although detailed evaluation studies are difficult to obtain.
Whether there are significant differences between the uncertainty values obtained
from the different bottom-up and top-down approaches in this study remains to be
seen.
32
Chapter 2
Table 2.6: The three top-down approaches and the equations to calculate standard
uncertainty of measurement[6].
Approach Data used Equations
Single laboratory Routine sample replicates 𝑢(𝑅𝑤 ) = √𝑠𝑅2𝑤 + 𝑠𝑟2

validation approach QC samples
CRMs 2
𝑢(𝑏𝑖𝑎𝑠) = √𝑅𝑀𝑆𝑏𝑖𝑎𝑠 + 𝑢2 (𝐶𝑟𝑒𝑓 )
Interlaboratory validation CRMs 2
approach QC samples
PT sample to be analysed at
Proficiency test approach 2
least six times
Symbols and equations used in this table:
𝑢(𝑅𝑤 ) = Uncertainty of within-laboratory reproducibility component

𝑠𝑅𝑤 = standard deviation of within-laboratory reproducibility component
𝑠𝑟 = standard deviation of repeatability component
𝑏𝑖𝑎𝑠 = difference between the result of PT sample or CRM and its assigned value
𝐶𝑟𝑒𝑓 = assigned value of CRM or PT sample
∑𝑁
𝑖=1 𝑢(𝐶𝑟𝑒𝑓𝑖 )
𝑢 (𝐶𝑟𝑒𝑓 ) = 𝑁
where i is the different laboratories used in the evaluation, and N is
the amount of analysis.
∑(𝑏𝑖𝑎𝑠𝑖 )2
𝑅𝑀𝑆𝑏𝑖𝑎𝑠 = √ 𝑤ℎ𝑒𝑟𝑒 𝑛 = 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑠𝑎𝑚𝑝𝑙𝑒𝑠/𝐶𝑅𝑀𝑠
𝑛
2.4 Previous work in the estimation of uncertainty in WD-XRF

measurements
The WD-XRF technique used in this study is described in Chapter 3. A glass disk
prepared with lithium tetra- and metaborate flux and the roasted sample is irradiated
with a 4 kW Rh-tube and the detected secondary X-rays quantified using Bragg's Law,
the basis of WD-XRF[5] and is stated in Equation 2.4.
nλ = 2d sinϴ Eq. 2.4

where
n = order of diffracted beam
d = spacing between diffracting crystal planes
λ = wavelength of secondary (diffracted) X-rays
ϴ = Bragg angle - the angle between primary, incident X-rays and the sample
surface
33
Chapter 2
The ISO/IEC 17025 standard[3] in 2005 created a need for accredited laboratories to
state specifically uncertainty of measurements in their XRF test methods [5]. Before
ISO/IEC 17025[3], standard quality control processes were sufficient as long as there
was confidence in the resulting precision and accuracy of the XRF test methods.
2.4.1 Previous studies of effects (errors) in WD-XRF
The main body of previous studies in estimating uncertainty in WD-XRF results

focused on describing specific sources of effect and practical methods to quantify
them. Rousseau[52] addressed the process to estimate the uncertainty of XRF results
with the detection limit concept, which the author regarded as being misunderstood
and the evaluation process being questionable. The global uncertainty of XRF
analytical results depends on a variety of sources, with the main being as follows:
- secondary sample preparation, i.e. preparation for sample presentation into the
XRF instrument,
- peak and background intensity measurements,
- the calibration line's slope, and
- sample matrix effects' corrections.
The effects are also grouped into random and systematic categories (Figure 2.9).
Figure 2.9: Sources of effects in XRF[53].
The addition law of variance propagates the various effects in terms of relative
standard deviation[52]. This publication describes two tests set up by analysing ten
glass disks once each and one glass disk ten times. The value of (𝐸%) 𝑇𝑜𝑡𝑎𝑙 is
determined from the first analytical process using the mean counts per second (cps)
and the standard deviation of the first test, while the statistics of the second test is due
to the counting statistics and instrument (Figure 2.10). With those values known, the
(𝐸%) 𝑆𝑎𝑚𝑝𝑙𝑒 can be indirectly determined.
𝑃𝑟𝑒𝑝𝑎𝑟𝑎𝑡𝑖𝑜𝑛
34
Chapter 2
Figure 2.10: Indirect determination of uncertainty due to sample preparation. The

symbols are as used in the publication[52].
Similarly, the uncertainty due to the instrument,(𝐸%)𝐼𝑛𝑠𝑡𝑟𝑢𝑚𝑒𝑛𝑡 , can be indirectly

100
determined as (𝜀%)𝐶𝑜𝑢𝑛𝑡𝑖𝑛𝑔 = where T is the measuring time.
√𝑐𝑝𝑠 𝑥 𝑇
𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐𝑠
The early papers using the GUM approach in WD-XRF estimate the uncertainty of Fe-
measurement in silicon[54] and describe the influence of the sample matrix on the
uncertainty of measurement budget[55] for improvement of the test method. As the test
methods in both papers use a simple relationship between concentration and intensity,
a measurement model could be set up, and the rest of the GUM approach be followed.
Effects of the quality of glass disks on the quality of the instrumental response are
discussed[56], where repeated glass disks of cement are created to evaluate the
repeatability of the Ca intensity in the resulting glass disks and each effect investigated
and quantified. The different sample preparation effects on the glass disks are
evaluated and quantified in this study using a similar approach in this paper.
Publications since 2015 describing test methods by WD-XRF similar to the study test
method describe the test method and report measurement uncertainty using a top-
down approach, but the evaluation process of the uncertainty of measurement is not
detailed[57][58][59][60].
35
Chapter 2
The easiest approach[5] to calculate measurement uncertainty in WD-XRF is to

determine the precision of the XRF test method and analyse a certified reference
material (CRM). The precision of both the test method and CRM is combined with the
CRM's uncertainty data reported by its certificate (Figure 2.11). For expanded
uncertainty, the coverage factor k is typically 2 for an approximate level of confidence
of 95%.
𝑈𝑐(𝑡𝑒𝑠𝑡 𝑟𝑒𝑠𝑢𝑙𝑡) = √{𝑆𝑟2 (𝑡𝑒𝑠𝑡) + 𝑆𝑟(𝑣𝑎𝑙)

2 2
+ 𝑈(𝑣𝑎𝑙) }
where
𝑈𝑐(𝑡𝑒𝑠𝑡 𝑟𝑒𝑠𝑢𝑙𝑡) = uncertainty of test result
𝑆𝑟 (𝑡𝑒𝑠𝑡) = repeatability of test result
𝑆𝑟(𝑣𝑎𝑙) = repeatability of CRM
𝑈(𝑣𝑎𝑙) = uncertainty of CRM as stated on certificate
Figure 2.11: Evaluation of measurement uncertainty in XRF.[5]
The GUM bottom-up approach is, however, used in the energy-dispersive XRF test
methods[61][62]. The authors simplified the complex algorithm into a relationship
between the measured XRF intensity and concentration of the measurand for the
measurement model. The uncertainty sources were categorised according to their
origin, which is mainly the analytical stages such as weighing for drying, preparation
of powder pellet and setup of the calibration curve. The two papers also illustrated
the tight relationship between method validation, metrological traceability, and
uncertainty of measurement, which applies to this study.
36
Chapter 2
2.5 Documentary standards of WD-XRF test methods
Some international standards[63][64][65][66] describe specific sample types for analysis by

XRF, and it was hoped to obtain instructions to evaluate the uncertainty of
measurements from these standards.
Analysis of cement for SiO2, Al2O3, Fe2O3, CaO, MgO, SO3, K2O, Na2O, TiO2, P2O5,
Mn2O3, SrO, Cl and Br by XRF is described in ISO 29581-2:2010[63], which is mainly
performance-based, i.e. only specification limits are tabled. It is, however, a good
source of Good Laboratory Practices (GLP) in the preparation of fused beads. ISO
12677:2011[64] is a similar standard for the analysis of ceramic materials.
ISO/TR 18336:1996[66] provides guidelines in quality control practices for XRF

laboratories in the iron ore industry. Handling of fusion flux, calibration reagents and
all equipment are prescribed. The preparation of glass disks is detailed with an EXCEL
workbook with macros to evaluate the precision of produced glass disks. A process to
calculate the precision of the XRF instrument is also included. A glass disc is placed
into a height-adjustable cup to investigate the effect of distance between the
measuring surface of the glass disc and iron concentration. Other sources of effects,
such as loss of accuracy caused by instrumental drift and change in backgrounds, are
briefly described but not quantified. The standards include applicable information for
the experimental design, validation steps and general GLP, but no evaluation
processes for the uncertainty of measurement.
2.6 Summary and conclusion
- Uncertainty of measurement estimation requires descriptive and inferential

statistics.
- The normal distribution is the basis of the Central Limit Theorem in the GUM
framework which means that the expectation is high for a measurement result
to be close to the population mean. The standard deviation and its multiples
determine the probability area under the distribution area.
37
Chapter 2
- VIM is the reference document of metrological terms, while GUM, GUM-1 and
GUM-3 are the reference documents for estimating the uncertainty of
measurement.
- All measurements are affected by systematic and random effects, and
uncertainty of measurement quantifies identified measurement effects.
- The uncertainty of measurement and bias expresses the performance of a test
method. It is therefore required to identify the main effects in the test method.
- The GUM approach remains a challenge for scientific personnel in an XRF
laboratory. It could be the reason for the lack of published work in the estimation
of uncertainty in WD-XRF analysis.
- The GUM framework and the Monte Carlo method are both bottom-up
approaches as the main sources of uncertainty are identified, quantified, and
propagated into a combined uncertainty for a specific test result.
- The top-down approach uses primarily precision and bias (trueness)
information to calculate a combined uncertainty for a test method.
38
Chapter 3
Chapter 3:
Methodology
3.1 Introduction
This chapter outlines the basic theory of WD-XRF, with the view to use the principle
of operation of the analytical technique for the quantification of major elements to
establish metrological traceability to the International System of Units (SI). The details
of the test method, the main validation parameters, and the estimation of the
uncertainty of measurement are described in this chapter, with the method validation
results being reported in Chapter 4 using both CRMs and proficiency test (PT) items.
The methodology of the study test method to estimate the uncertainty of measurement
is also outlined in this chapter.
3.2 Theory of wavelength-dispersive XRF
WD-XRF is a versatile analytical tool to qualify and quantify elements in a wide range
of sample types. Photoelectric effects are the basis of WD-XRF analysis and are the
product of matter absorption of primary X-rays. During the bombardment of matter by
primary X-rays, electrons are ejected from the sample, resulting in the emission of
characteristic X-ray photons. Two fundamental equations in physics describe the
wavelength (Equation 3.1) and energy (Equation 3.2) of a photon. Combining these
equations leads to Equation 3.3a, illustrating the direct relationship between an X-ray
photon's energy (E) and wavelength (λ)[5][53][67].
𝑐
𝑣= (Eq. 3.1)
𝜆
𝐸 = ℎ𝑣 (Eq. 3.2)
where 𝑣 = frequency of wave motions (vibrations per second)

h = Planck’s constant (4.135 x 10-15 eV sec)
c = velocity of light (3 x 108 m/sec)
λ = wavelength (m) (1 Å = 10-10 m)
E = energy of the X-ray photon (keV)
39
Chapter 3
In Equation 3.3a, the product of two fundamental physical constants, Planck's constant
(h) and the speed of light (c), can be substituted into either equation 3.3b or equation
3.3c depending on the unit used for wavelength.
After the ejection of an electron from a sample atom, a transition process occurs in the
unstable ion to fill all the orbital vacancies created by both the original ejection and the
movement of electrons from outer to inner orbital shells to stabilise the ionised atom.
Each electron movement from the outer into the inner orbital subshells (K, L, M and
N-shells) results in a formation of a characteristic photon. The International Union of
Pure and Applied Chemistry (IUPAC) notation describes the resulting spectral line
regarding the subshells from where the electron moved into a created vacancy [53]. The
characteristic photon energy can be determined by the difference between the binding
energies of the ejected electron in a shell and the stabilising electron from an outer
subshell[67]. The binding energy of an electron in, for example, Si's K-shell is 1.837
keV, while the binding energy of the stabilising electron in its L-3 subshell is 0.098
keV. The energy of the emitted Si's K-L2 photon will then be 1.740 keV (Equation 3.4),
and in terms of wavelength, it can be calculated to be 7.126 Å. The measurands'
photon properties are listed in Table 3.1.
40
Chapter 3
𝐸𝐾−𝐿3 = 𝐸𝐾 − 𝐸𝐿3 Equation 3.4
Where
𝐸𝐾−𝐿3 = photon energy of spectral line
𝐸𝐾 = binding energy of an electron in the K-shell
𝐸𝐿3 = binding energy of an electron in the L3-energy level of the L-shell
Table 3.1: The test measurand’s spectral line properties[53]. The spectral line for all
the listed elements is K-L3 (also known as Kα1-line).
Binding energy of an electron Photon Wavelength
in a shell (keV) energy of of spectral
Element
spectral line line
K L3
(keV) (Å)
Na 1.080 0.034 1.041 11.909
Mg 1.303 0.049 1.254 9.889
Al 1.559 0.072 1.487 8.339
Si 1.838 0.098 1.740 7.126
P 2.142 0.128 2.015 6.155
K 3.607 0.294 3.313 3.744
Ca 4.038 0.349 3.691 3.360
Ti 4.964 0.454 4.510 2.750
Cr 5.988 0.574 5.414 2.291
Mn 6.537 0.639 5.898 2.103
Fe 7.111 0.708 6.403 1.937
Although the binding energy values of some photons differ slightly in some
sources[53][68], the reported differences are not significant. Malvern (previously
Panalytical), a manufacturer of XRF spectrometers, uses mainly the values in the
database set up by Elam et al.[68] in the spectrometer software for the qualitative and
quantitative XRF processes[69]. The WD-XRF spectrometer employs Bragg's Law (Eq.
2.4) to identify the wavelength of the spectral line where the lattice spacing of the
detector crystal (2d) and the angular position of the goniometer (ϴ) are known. Figures
3.1 and 3.2 illustrate Bragg's Law in the WD-XRF spectrometer.
41
Chapter 3
Figure 3.1: A simplified illustration of the WD-XRF basics. Bragg's law identifies
the element's wavelength energy using the diffracting crystal's known
lattice space (2d) and the ϴ-value as measured by the goniometer
while the detector quantifies the identified elemental peak. The figure
is based on a figure drawn by Willis et al.[5]
Figure 3.2: An X-ray spectrum of an andesite sample. The X-axis depicts the
goniometer's 2ϴ-theta positions from ~5 to 147 degrees, while the Y-
axis expresses the count rate in kCPS. Crystal LiF200 and duplex
detector were used for this scan made by the PANalytical Zetium XRF
spectrometer.
42
Chapter 3
3.3 Scope of the test method
The test method was developed to analyse a wide range of sample types for a set of
major elements and loss on ignition (LOI) using WD-XRF and gravimetry, respectively.
The major elements are SiO2, TiO2, Al2O3, total Fe as Fe2O3, MnO, MgO, CaO, Na2O,
K2O, P2O5 and Cr2O3. The sample types include igneous, sedimentary, metamorphic
rocks, ores and industrial materials (Table 3.2). The test method is not applicable for
samples containing a major or minor element not listed in the abovementioned set of
major elements.
Table 3.2: Sample types for major element analysis by the study test method.
Sample type Samples in South Africa
Granite, gabbro, dolerite, basalt, granodiorite,
Igneous rocks
carbonatite
Limestone, shale, sandstone, calcrete, silcrete,
Sedimentary rocks
mudstone, conglomerate
Amphibolite, diabase, schist, granulite, gneiss,
Metamorphic rocks
marble
Industrial materials Kaolinite, bauxite, wollastonite
Ores Iron, manganese and chrome ores
3.4 Primary sample preparation
Although the XRF analyst is usually not part of the initial sampling process, information
regarding the origin and sample type is useful in analysing the submitted sample. Such
information assists the XRF analyst:
- to ensure the expected set of major and minor elements is the same as the set of
major elements in the study test method and
- to expect any possible adverse refractory compounds in the submitted sample.
Powders from quarries, for example, could contain silicon carbide (SiC) or tungsten
carbide (WC), e.g. drill bit fragments which are detrimental to the platinum (Pt) ware
used to prepare the glass disks.
43
Chapter 3
Before sample preparation for WD-XRF, the submitted samples are crushed, split and
milled to produce a powder with particle size less than 75 µm. It is also assumed that
the produced test sample is homogenous.
3.5 Secondary sample preparation
Glass disks are used to analyse the sample material by WD-XRF. The main advantage
of glass disks is their homogeneity as the particle size effects are overcome. The test
method to analyse major elements on fused beads at the Council for Geoscience
(CGS) evolved from Norrish and Hutton’s method[70] using one type of lithium borate
flux to the current test method involving various lithium borate fluxes to dissolve the
sample material. This evolution can be ascribed to the wide range of sample types to
be analysed and the rapidly growing variety of chemicals available as a flux.
The main requirement of the current test method is the consistent total mass of the
fused bead. It must also be the same for both the CRMs and the sample material.
The first stage of the test method is to dry about 5 g sample powder at 100 ⁰C for at
least three hours in an oven to remove moisture before roasting it at 1000 ⁰C in a
furnace. The roasting step has three main purposes:
i) to remove crystalline water and other volatiles. This process ensures a

consistent mass amount of the produced glass disk. Pure limestone, for
example, consists of approximately 40 % carbon dioxide (CO2), and without the
roasting step, the resulting count rate for an analyte will be less than that of the
glass disk produced from the roasted sample,
ii) to pre-oxidise the dried sample material. Iron(II) oxide (FeO), for example, is
converted to iron(III) oxide (Fe2O3), i.e. the total iron (Fe) in the sample material
is analysed and reported as Fe2O3. In the case of sulphur (S), which is
detrimental to the Pt ware used in the fusing process, this oxidation step assists
in the conversion of sulphur (S) into sulphur dioxide (SO 2), which is then
partially removed from the sample as a gas, and
iii) to determine loss on ignition (LOI) gravimetrically. The user utilises the
sample’s LOI value to gauge the sample nature, especially its mineralogy.
44
Chapter 3
Furthermore, the sum of the major elements and LOI is to be as close as

possible to 100%, which is indicative of the analytical accuracy of the major
element results[71].
3.5.1 Instrumentation in the secondary sample preparation stage
The main instruments in the second sample preparation stage are the balance, drying
oven, furnace and fusion instruments. The balance is verified each day when in use,
while the other instruments are serviced and calibrated annually.
Table 3.3: Instruments used in the sample preparation stage of the test method.
Instrument used in the study Purpose of instrument and comments
Mettler AE 240 four decimal - To determine LOI gravimetrically, and
balance - to weigh the sample and flux mixture accurately.
To dry the samples during the determination of the
Labotec drying oven
moisture content.
To roast the sample and reference materials at 1000
Lenton furnace (Fig. 3.3)
⁰C.
To prepare glass disks. TheOx is an automated
fusion instrument that fuses the sample and flux
mixture in a Pt crucible (Fig. 3.5) and pours the melt
into a hot mould according to a predefined
programme using a set of parameters in the “CGS
Oxides”- programme as follows:
Steps 1 to 2: After starting the programme, the steps
Claisse TheOx fusion instrument are used to stabilise the temperature to 980 ⁰C.
(Fig. 3.4) Steps 3 to 4: Heat for 10 min at 980 ⁰C in still
position, i.e. no rotating or shaking positions.
Steps 5 to 10: Heat for 2 to 3 min each at 980 ⁰C and
different combinations of rotating speeds, angles
and shaking styles.
Step 11: Pour into moulds
Step 12: Cool for 1.15 min at a low fan speed
Step 13: Cool for 4 min at high fan speed
To prepare glass disks. In the Beadmaster,
mouldibles (Fig. 3.5) are used to melt and cool the
sample mixture to produce a glass disk. Parameters
Beadmaster FM-4 (Fig. 3.6) are set in terms of time and are as follows:
- no rocking for 10 minutes
- rocking for 30 minutes
- cool for 4 minutes
45
Chapter 3
Figure 3.3: The furnace used to roast the dried sample material in porcelain
crucibles at 1000 ⁰C. After the roasting step, the crucibles are cooled
inside desiccators to prevent re-absorption of any moisture.
46
Chapter 3
(a)
(b)
Figure 3.4: Claisse TheOx fusion instrument is a front door electric furnace with a
position for six platinum crucibles and flat bottom moulds. The melting,
stirring and pouring steps are automated by a selected programme
called the “CGS Oxides” in this study. The steps are listed in Table 3.3.
47
Chapter 3
Figure 3.5: The platinum ware used by the fusion instruments. The Pt-crucible and
flat bottom mould on the left are used by the Claisse TheOx, while the
mouldible in the top right is used by the Beadmaster. The resulting
glass disk in the bottom right is the sample presented to the WD-XRF
spectrometer.
48
Chapter 3
(a) (a)
(b)
Figure 3.6: The Beadmaster is a robust fusion instrument which consists of an

electric furnace with a top door (a) and positions for four mouldibles (b).
After the melting stage, they are removed with a titanium tong and placed
into cooling positions in front of the furnace. There the glass disks are
cooled with controlled air.
49
Chapter 3
3.5.2 Steps in the secondary sample preparation stage
Table 3.4. details the preparation steps, recorded information, and the formulas to
calculate the resulting LOI.
Table 3.4: The sample preparation stage for WD-XRF analysis.

Recorded information
Step Description with the associated
symbol in parentheses
Weigh 10 g, 5 g, 2 g, 1 g, 0.5 g and 0.1 g mass pieces to check The masses of the
1
the balance. different mass pieces
2 Weigh a marked and pre-dried porcelain crucible. Crucible mass (Cr)
Weigh approximately 5 g of sample powder into the crucible Crucible + sample mass
3
from Step 2. (CrS)
Dry the sample for at least three hours at 100 ⁰C in a drying
4
oven.
Cool the crucible inside a desiccator, and after cooling, record Crucible + sample mass
5
its mass. after drying (Cr100)
Place the sample from Step 4 into a cool muffle furnace. Close
the furnace, set the furnace’s temperature to 1000 ⁰C, and after
stabilising the furnace at 1000 ⁰C, roast the sample for at least Crucible + sample mass
6
two hours. Samples with expected high LOI is to be roasted after roasting (Cr1000)
longer (or overnight). After cooling inside a desiccator, record
the crucible’s mass.
Record the following:
Weigh into a glass vial 10 g flux and 1 g roasted material from - selected flux,
7
Step 5 - mass of flux, and
- mass of sample
In the case of the Claisse TheOx fusion instrument:
- Pour the flux and sample mixture into the Pt-crucible.
- Ensure that each Pt-crucible is correctly placed with the
supporting rod being secured. Close the door and start the
“CGS Oxide”-program.
- Remove the cooled glass disk from the instrument after the
conclusion of the selected program.
8 In the case of the Beadmaster FM-4 fusion instrument:

- Pour the flux and sample mixture into the mouldible.
- Initiate the process.
- After thirty minutes, open the top and check whether the
sample melt has fully dissolved its contents.
- If the melt contains no solids, stop the rocking movement
process, remove the mouldible using titanium tongs and put
the mouldible into a cooling rack.
- Start the fan and cool for about five minutes.
9 Place an identification sticker label on the back of the glass disk.
Calculate the moisture and LOI as follows:
(𝐶𝑟𝑆 − 𝐶𝑟100)
% 𝑀𝑜𝑖𝑠𝑡𝑢𝑟𝑒 = × 100
10 (𝐶𝑟𝑆 − 𝐶𝑟)
(𝐶𝑟100 − 𝐶𝑟1000)
% 𝐿𝑂𝐼 = × 100
(𝐶𝑟100 − 𝐶𝑟)
50
Chapter 3
3.5.3 Flux used in the test method

The two main purposes of flux in the preparation of glass disks for WD-XRF are to
dissolve the sample powder and form a stable, homogenous glass disk for
presentation into the WD-XRF spectrometer. The flux and sample acidity combination
should be close to neutral to dissolve the sample by the flux. With the target sample
types varying from acidic to basic, three different fluxes are used in the study method
and are listed in Table 3.5. Furthermore, the flux should also contain a non-wetting
agent to assist the glass disk to release from the mould after cooling. The three fluxes
are granular eutectic mixtures of lithium metaborate and lithium tetraborate. The fluxes
contains lithium bromide as a non-wetting agent for the release of glass disks from the
moulds.
Table 3.5: The three lithium borate fluxes were used in the study and are produced
by Claisse. The terms Flux A, Flux B and Flux C are for operational
purposes.
Flux and
catalogue Composition Target sample type
number
66.67% Lithium tetraborate ‘Basic’ samples, such as
Flux A
32.83% Lithium metaborate carbonatite, calcrete, marble
0.50% Lithium bromide and limestone
Most alumino-silicates
49.50% Lithium metaborate including amphibolite, diabase,
Flux B 49.50% Lithium tetraborate schist, granite, gabbro,
1.00% Lithium bromide dolerite, basalt and
granodiorite.
Iron and chrome ores.
64.67% Lithium metaborate Alumino-silicates containing
Flux C 34.83% Lithium tetraborate high amounts of acidic oxides
0.50% Lithium bromide such as kaolinite, bauxite,
sandstone and silcrete.
51
Chapter 3
3.6 Analysis by WD-XRF
Analysis by WD-XRF involves both the instrument’s spectral components and the
software used to operate the spectrometer and set up the calibration test method. The
CRMs and RMs used in the calibration are also discussed in this section.
3.6.1 The XRF spectrometer
PANalytical Zetium equipped with a 4kW Rh-tube (Figure 3.7) was used for the study
test method. SuperQ-software assists the analyst to operate the XRF spectrometer,
setting up calibrations for specific sample types, and managing the results for reports.
The setup of the test method calibration for the WD-XRF instrument involves mainly
the following processes:
(i) the selection of CRMs and RMs with similar sample types as the expected
sample materials for future analysis and the preparation of the standard glass
disks;
(ii) the selection of instrumental parameters (Table 3.6) optimal for the analysis of
the measurands,
(iii) the calibration of the angular positions,
(iv) the checking of the detector conditions for the elemental peaks,
(v) the verification of line overlap positions,
(vi) the selection of sufficient measurement time for each background, interfering and
elemental peaks; and
(vii) the set up of a calibration application to process the resulting count rates into
concentrations.
52
Chapter 3
Figure 3.7: PANalytical Zetium XRF-spectrometer. It is both a wavelength and

energy dispersive XRF spectrometer and has a 4kW Rh-tube.
53
Chapter 3
Table 3.6: The instrumental parameters and available settings of the PANalytical
Zetium.
Parameter Available setting
Tube voltage Up to 60 kV with a maximum power of 4 kW
Tube current Up to 160 mA with a maximum power of 4 kW
Diffracting crystal LiF200, LiF220, Ge, PET, PX1
No filter, 100 µm brass, 400 µm brass, 200 µm Al and
Tube filter
750 µm Al
Detector Flow, Xe sealed, Scintillator and Duplex
Angular position Depending on the selected detector, from 5 to 145 °2ϴ
Measurement time From 2 to 10000 seconds
Collimator 150 µm, 300 µm and 700 µm
Mask diameter 6 mm, 10 mm, 20 mm, 27 mm, 30 mm and 37 mm
Spinner On, off
Analysis medium Vacuum
The main instrumental parameters, as used in the study, are listed in Table 3.7. Major
corrections applied on the elemental peaks are subtractions of background and, in the
case of Al Kα1,2, line overlapping by Br Lα1,2. Instrumental drift corrections are applied
by analysing a stable drift monitor disk, AUSMON, each day with the standard and test
glass disks. The initial monitor values are created using a measurement application
with the same channel setup used for the test method but with at least twice the
measurement times to obtain the correction factor based on the difference between
the obtained and initial counts. The drift correction factor is calculated by dividing the
measured counts with the original counts.
54
Chapter 3
Table 3.7: The main instrumental parameters for the spectral lines. Due to the non-
wetting agent, bromide is present in each glass disk, and Br Lα1,2 is
measured to correct for peak overlapping with the Al Kα1,2-peak. The full
detail of the instrumental setup is detailed in Figure 4.1.
Spectral Tube setting Angular

Crystal Detector
line (kV, mA) position (°2ϴ)
Si Kα1,2 40, 100 PE 108.9632 Gas flow
Ti Kα1,2 60, 66 LiF200 86.1456 Gas flow
Al Kα1,2 40, 100 PE 144.7858 Gas flow
Fe Kα1,2 60, 66 LiF220 85.7470 Duplex
Mn Kα1,2 60, 66 LiF220 95.2100 Duplex
Mg Kα1,2 30, 130 PX1 22.8748 Gas flow
Ca Kα1,2 50, 80 LiF200 113.1294 Gas flow
Na Kα1,2 30, 130 PX1 27.6316 Gas flow
K Kα1,2 50, 80 LiF200 136.6908 Gas flow
P Kα1,2 30, 130 Ge 141.0030 Gas flow
Cr Kα1,2 40, 100 LiF200 69.3308 Duplex
Br Lα1,2 40, 100 PE 146.4174 Gas flow
The Zetium XRF spectrometer has two sample positions, the loading and measuring
positions for the sample cup containing a glass disk for analysis (Fig. 3.8). The
spectrometer moves a glass disk from a sample tray into a sample cup, which is then
placed into the loading position and then into the measuring position to conserve the
vacuum of the goniometer chamber. Various sample cups are available for the fusion
disks of the test samples and standards, while the same sample cup is used for the
drift monitor glass disk.
55
Chapter 3
Figure 3.8: Left bottom displays the measuring cups on a tray where one cup is
selected to handle the glass disk inside the XRF-spectrometer. The
loading position right of the measuring cups is where the cup is moved
into the measuring position inside the spectrometer. The combination
of each cup and sample position has slightly different distances
between the tube and glass disk surface.
3.6.2 Fundamental Parameters (FP)

Matrix effects are inherent in each test material and can be both chemical and physical.
Physical matrix effects are caused by particle size and mineralogical effects[5]. As this
study employs glass disks for the analysis, both described effects are eliminated. In
the case of chemical effects, some elements in different chemical states can lead to
wavelength shifts of the elemental peak[53]. Other chemical effects are the elements'
interaction with each other (Fig. 3.9) and must be corrected for in this study.
Fundamental parameters are used in the study to apply such corrections.
56
Chapter 3
Figure 3.9: The enhancement of Cr by Fe radiation in a sample[53]. Cr is the analyte

element i, Fe the matrix element j. IiA(λi) is the intensity from the primary
excitation, while IiE(λi) is the intensity from enhancement. The reverse is
also applicable, i.e. absorption of Fe by Cr. The effect of both
enhancement and absorption on the calibration line is illustrated in
Fig. 3.10.
57
Chapter 3
Figure 3.10: Inter-element matrix effects on the intensity of the analyte in a

calibration plot for Fe in a high alloy steel sample[53]. In the left figure,
enhancement leads to higher than expected intensities, while lower
intensities are the product of absorption. In the right figure, matrix
corrected intensities are plotted linearly with their associated
concentration values.
The FP matrix model as available in SuperQ is used in this study to correct for the
inter-element effects. It employs the various instrumental parameters used in the
measurement and fundamental parameters such as the K-jump ratio, the K-
fluorescent yield and the probability for the emission of a Kα-peak instead of Kβ-peak
to calculate theoretical intensity for the test sample[5]. Figure 3.11 demonstrates the
different components contributing to the calculations of the theoretical intensity. The
incident radiation is calculated in the dotted box (a) in Figure 3.11, while box (b)
quantifies the generation of the characteristic or elemental photon. The geometry
factor is calculated in box (c) with boxes (d) and (e) calculating the absorption of the
incident and characteristic photons, respectively.
58
Chapter 3
Figure 3.11: The FP equation calculating the primary fluorescence Pi using the
used setup of the spectrometer, and the known elemental values of JK,
the K jump ratio, ωK , the fluorescent yield and PKα, the probability
factor for the Kα-line to be emitted instead of the Kβ-line. (From two
figures in Willis et al.[5]). Pi(λ0) is the primary fluorescence excited by
the monochromatic incident radiation λ0 from the Rh-tube in the study.
I0(λ0) is the intensity of the incident X-rays; Ci is the concentration
weight fraction; ψ’ and ψ” are respectively the incident and emergent
µ
angles of the X-rays with the test sample surface; is the mass
𝝆
attenuation coefficients for the X-rays (cm2/g). The blue dotted boxes
(a) to (e) are explained in the text.
SuperQ uses the original intensities corrected for background, line interferences and
instrumental drift in the test sample analysis to determine all apparent measurand
concentrations used with the described parameters to determine theoretical
intensities. This process is iterated with the adjusted concentrations until the
calculated theoretical intensities match the measured intensities. The matrix factor is
determined from this process and is multiplied with the drift, background and line
overlap corrected counts for use in the regression model (Eq. 3.5), which is essentially
59
Chapter 3
the final plot between the CRM concentration and corrected instrumental
responses[72].
The regression model in SuperQ used in the study is outlined in Equation 3.5.
𝐶 = 𝐷 + 𝐸 × 𝑀𝐼 Eq. 3.5
where C = analyte concentration in %

D = intercept of the regression line with the Y-axis (%)
E = slope of regression line (%/kCPS)
MI = Background, drift, line overlap and matrix corrected countrate
(kCPS)
The plot of concentration instead of the instrumental responses on the Y-axis in

SuperQ is contrary to the practice in analytical chemistry where the instrumental
response is plotted on the Y-axis. It is, however, outside the scope of this study to
evaluate this approach.
3.6.3 CRMs and RMs used for the test method calibration
CRMs and RMs are used to establish the relationship between the corrected count
rate produced by WD-XRF and the measurand concentration. The RMs and CRMs’
certificates express uncertainty information in different ways, and where there was no
information available, publications were used instead to obtain the best estimate of the
standard uncertainty of the reported values. For example, South African MINTEK’s
NIMROC series’ CRM certificates only report certified and provisional values for the
major and trace elements. As submitted samples for major element analysis by XRF
at the CGS are mainly from Southern Africa, those CRMs are therefore being used in
this study. In such cases, the publication containing statistics of the CRM preparation
was perused to provide the necessary uncertainty information. PT materials are also
used to evaluate and validate the calibration. Appendix A lists the used CRMs, and
RMs grouped per producer and the provided information to evaluate the uncertainty
of concentration, while Appendix B details the sample preparation to prepare the glass
60
Chapter 3
disks. In Appendix D, their certified or assigned values are listed with their associated
uncertainty and standard uncertainty.
3.6.4 Other test materials and their portions
Other test materials used in the study are MzBP-1, SARM-36, and a limestone sample.
MzBP-1 is a monzonite sample from GeoPT, a PT programme managed by the
International Association of Geoanalysts (IAG), and was employed to validate the test
method. SARM-36 is an isotope reference material prepared by MINTEK, and as it is
assumed to be sufficiently homogenous, the primary sample preparation effects are
excluded in this sample. Limestone was selected as a study material due to the
availability of sufficient material to be developed as an in-house standard with an
estimated uncertainty of measurement.
The various test sample portions for the quantification of uncertainty in this study are
defined as follows:
- A submitted sample is the original sample submitted by the customer to the CGS
Laboratory for testing.
- A test sample is the portion of the original sample that has been pulverised to less
than 75 µm. A test sample can also be a submitted sample if the customer prepared
it.
- An aliquot is the 5 g portion of the test sample which is dried and roasted. LOI
results are from the aliquots of a test sample.
- Split samples are 1 g subsamples of the aliquot for glass disk preparation to
determine the major elements by XRF.
61
Chapter 3
3.6.5 The quantification process
SuperQ has various options of sample types to set up a measure and calibration
application. The sample type list provides the analyst with sample preparation
information listed in Table 3.8.
Table 3.8: Sample types in SuperQ.

Sample type Details for information capture
Weight (g), diameter (mm), area (cm2), thickness (mm) and
Solid
density (g/cm3)
Diameter (mm), area (cm2), thickness (mm) and density
Bead
(g/cm3), additive(s), sample weight before and after fusion (g)
Pre-treated bead
(g/cm3), additive(s), sample weight (g)
Pressed powder
(g/cm3), additive(s), sample weight (g)
Weight (g), diameter (mm), area (cm2), thickness (mm) and
Loose powder
density (g/cm3)
Liquid
(g/cm3), solvent(s), sample weight (g), volume (cm3)
The ‘Bead’ and ‘Pre-treated bead’ options were investigated but not used as the
resulting totals varied between 97 and 101%, which is unsuitable. It is outside the
scope of this study to discuss the various reasons for it. The ‘Solid’ sample type is
used in the study while the flux compounds are entered into the software along with
the adapted CRM and RM-concentrations. The quantification process to obtain the
concentration for each fusion disk prepared from CRMs and test samples are outlined
in Figure 3.12. The CRM and RM concentrations are entered into SuperQ during the
setup of a calibration application for the test method.
62
Chapter 3
Figure 3.12: The calculation of the CRM and RM concentration for entering into the
SuperQ calibration application and the quantification process of the test
sample’s major elements. Ci = first recalculated measurand
concentration (%); Ci,cert = original measurand concentration (%); Total
(CRM and RM) = the sum of the measurand concentrations and the
measured LOI (%); LOImeas = measured LOI (%); Ci,final = the
concentration values for the SuperQ application; MassSample and Massflux
= mass (g) of the roasted sample and flux respectively; Ci,unk = the result
concentration for reporting (%); Total (Measured sample) = the sum of
all concentrations (%) obtained from SuperQ; Flux = (B2O3 + Li2O) (%).
The flux values are entered into the software with at least five decimals,
and the calculation process to obtain B2O3 and Li2O is illustrated in
Figure 3.13.
63
Chapter 3
Figure 3.13: The ‘flux calculator’ used to obtain the Li2O and B2O3 values based on
the selected flux and the masses of the flux and roasted material.
3.7 Method performance characteristics
As noted in Chapter 2, the method validation starts with the specification of the
measurands, which are SiO2, TiO2, Al2O3, total iron as Fe2O3, MnO, MgO, CaO, Na2O,
K2O, P2O5, Cr2O3 and LOI. Furthermore, the method validation data is used to confirm
the test method is fit for purpose and to estimate the uncertainty of measurement. The
method performance characteristics follow Eurachem’s guide, ‘The Fitness for
Purpose of Analytical Methods’[15].
3.7.1 Selectivity
The selectivity parameter refers to the test method’s ability to analyse the measurand
in the presence of other possible interfering elements and in different sample matrixes.
The measurement application was set up to optimally measure each measurand’s
spectral line in terms of power settings (kV and mA), crystal, detector, collimator, and
pulse height settings. Appropriate line corrections are also applied on some elements
and are listed in Table 3.9. Furthermore, as the major elements are analysed in fusion
64
Chapter 3
disks, the particle size effects are removed. However, the matrix effects are still
present and are enhanced, especially in ore-type samples. Most of the effects are
corrected by the Fundamental Parameters algorithm used by SuperQ as described
previously in 3.6.2.
Table 3.9: Line-overlap corrections applied by SuperQ software.

Measurand peak Interfering compound or peak
Mn Kα Cr2O3 concentration
Mg Kα Al2O3 concentration
Ca Kα K2O concentration
Al Kα Br Lα-peak
3.7.2 Working range and linearity
In this study, linearity is the relationship between the measured responses (kCPS) and
the CRM concentration (%) and is expressed by the calibration application in the
SuperQ software as R2. Calibration curves with correlation coefficients larger than
0.999 indicate a good relationship between the instrumental response and certified
concentrations. After evaluating the results of the CRMs, RMs and PT samples, the
working range for each measurand is determined to be from the limit of quantification
(LOQ) to the highest value of the CRM or RM.
3.7.3 Analytical sensitivity
The slope of a calibration line where the instrument responses are plotted against the
concentration indicates the test method’s sensitivity. A calibration with high sensitivity
has a slope with more than 45 ° angle between the slope and the X-axis[13]. As
described in eq. 3.6, SuperQ plots the concentration on the Y-axis, and this parameter
check is not applicable in this study.
This validation parameter is not an important test method performance

characteristic[73], but is useful to check the method performance at low concentration
levels. The dilution factor of 1:10 is high but had to be used due to practical reasons.
65
Chapter 3
The high dilution factor means that at least five decimals are used for the measured
concentrations (Ci,meas in Figure 3.12).
3.7.4 Trueness
As the evaluation of a test method requires relevant CRM or RM’s certificate

information and it is not always available, or the certificate does not follow the modern
reporting format of stating the assigned value along with the associated uncertainty,
PT results were also used to evaluate the trueness of the test method. During the initial
stages of the calibration process, the result concentrations of the used CRMs, RMs
and PT are plotted against their associated certified concentrations to gauge the
trueness of the test method. Two approaches are used to establish the bias of the test
method and are Recovery (in 3.7.4.1)and the En score (in 3.7.4.2).
3.7.4.1 Recovery
Relative recovery of a CRM or RM is outlined in Equation 3.6. Acceptable recovery

should be between 95 and 105% for a level of confidence of 95%.
𝐶𝑅𝑒𝑠𝑢𝑙𝑡
𝑅𝑒𝑐𝑜𝑣𝑒𝑟𝑦 = × 100 Eq. 3.6
𝐶𝐶𝑒𝑟𝑡𝑖𝑓𝑖𝑒𝑑/𝐴𝑠𝑠𝑖𝑔𝑛𝑒𝑑
3.7.4.2 The En-score
The bias of the test method is also evaluated using the En-scores of the used CRMs,
RMs and PT samples. The En-score uses the absolute difference between the CRM’s
(𝑥𝑙𝑎𝑏 −𝑋𝐶𝑅𝑀 )
𝐸𝑛 = | | Eq. 3.7
2
√(𝑈𝑙𝑎𝑏 2
+𝑈𝐶𝑅𝑀 )
Where 𝑥𝑙𝑎𝑏 = CRM’s result

𝑋𝐶𝑅𝑀 = CRM’s certified value
𝑈𝑙𝑎𝑏 = expanded uncertainty of CRM’s result
𝑈𝐶𝑅𝑀 = expanded uncertainty of CRM’s certified value
66
Chapter 3
certified and result values and their expanded uncertainties (Equation 3.7)[74]. An
acceptable score is to be less than one.
3.7.5 Precision
As noted in the previous chapter, the closeness of agreement between the test results
obtained during the same and different analytical conditions contributes to the
concepts of repeatability, reproducibility and intermediate precision[17]. However, with
the reproducibility applicable to different laboratories[74], only repeatability and
intermediate precision types are investigated for this study. The main areas
contributing to the precision of the test method are in sample preparation and the XRF-
spectrometer (Table 3.10). Various contributors to the precision of the test method are
identified and quantified using analysis of variance (ANOVA) in EXCEL. The results
from the experiments and their organisation for ANOVA is described in Chapter 4.
As part of quality control at the CGS, ten percent of a batch of samples for analysis by
WD-XRF is done in duplicate. In the past five years, major element results have been
captured in Starlims, a laboratory information management system (LIMS). For this
study, such quality control data is extracted to evaluate the repeatability of duplicate
results. This information is used for the top-down approach described in 3.9.5, where
absolute or relative uncertainty is used depending on the concentration result. It is also
used in this study to quantify LOI’s precision in the bottom-up approaches (Duplicates
in Table 3.10).
A test material investigated for use as an in-house standard was prepared by

homogenizing about 100 kg limestone and packed into 64 split bags. Eleven test
samples were evaluated for primary sample preparation by preparing five pressed
powder pellets per test sample. The powder pellets were analysed for TiO 2, Fe2O3,
MnO and CaO using the trace element test method to evaluate the homogeneity of
the milled limestone. The four major elements are part of the CGS trace element test
method for QC purposes. The trace instead of the major element test method was
used due to its minimal sample preparation and large sample volume (12 g sample
per pellet). Two test samples were used in the precision experiments to evaluate the
sampling effect in aliquot and split samples.
67
Chapter 3
Table 3.10: Effects under different precision conditions.

Intermediate precision
Effect Repeatability conditions
conditions
Sampling: primary
Milled limestone powder split
sample
into test samples
preparation
Different aliquots and split
One aliquot prepared into
Aliquot and split samples prepared into fusion
various fusion disks using the
sampling disks using different moulds
same mould and fluxer
and fluxers
Variation in a split Six SARM-36 aliquots, same
sample and flux mould and fluxer, but
masses (Dilution variations in split sample and
effect) flux masses
One aliquot sample in the Different aliquots in different
LOI
same position in the furnace positions in the furnace
One aliquot prepared into

various fusion disks using the
Fusion disk
same mould and the same
fluxer
Same fusion disk analysed on

Instrumental drift
different days for two months
XRF spectrometer:
A glass disk in the same A glass disk in the same
sample position
position and cup position and different cups
and sample cup
Evaluation of repeatability
between duplicate results in a
Duplicates
dataset extracted from
Starlims.
68
Chapter 3
3.7.6 Limit of detection and limit of quantification
The lower limit of detection (LOD) and limit of quantification (LOQ) are used to indicate
a concentration value more than zero and to determine the lowest concentration value
for reporting at a specified confidence level[73]. In WD-XRF, the first limit parameter,
LOD, in Equation 3.8 is three times the measurand background, and the LOQ is six
times the measurand background (Equation 3.9). The LOD is at the confidence limit
of 95 %, while the LOQ is the LOD at the 99 % confidence limit[53]. The LOQ can also
be evaluated graphically from the working range plots. In the case of LOI, the LOD is
not determined as it depends on the sample type. Iron ores rich in FeO, for example,
becomes heavier due to the oxidation of FeO to Fe2O3, leading to negative LOI values.
3 𝐼
𝐿𝐿𝐷 = × √ 𝑖,𝑏 Equation 3.8
𝑚 𝑇𝑏
6 𝐼𝑖,𝑏
𝐿𝑂𝑄 = × √𝑇 Equation 3.9
𝑚 𝑡𝑜𝑡𝑎𝑙
Where m = sensitivity (cps/wt%)

𝐼𝑖,𝑏 = total background countrate after corrections (cps)
𝑇𝑏 = total background measuring time (s)
𝑇𝑡𝑜𝑡𝑎𝑙 = peak and background measuring time (s)
3.8 Metrological traceability
In Chapter 2, metrological traceability is described to link the produced measurement

result with the SI units using CRMs and mass pieces in a documented process (Fig.
2.4). The following were documented for the test method validation:
i) the used CRMs’ calibration certificates,
ii) records of the mass pieces results, and
iii) the calibration records of the instruments, for example, the Zetium’s service
record, the balance, drying oven and the furnace.
69
Chapter 3
3.9 Uncertainty of measurement
As noted in Chapter 2, precision and bias are the most important contributors to the
evaluation of each uncertainty component or influence quantity in the estimation of
uncertainty of a measurement result. Three main areas for evaluating the uncertainty
components are from the sample preparation, the XRF instrument and the regression
line of the calibration. The standard uncertainty of each evaluated uncertainty
component is estimated according to the evaluation type as listed in Table 2.4. in
Chapter 2. Most precision effects measured using repeated observations in the study
are evaluated as Type A uncertainties, i.e., the standard deviation of the number of
repeated observations divided by the square root of the number of observations.
3.9.1 Sample preparation
The uncertainty contributors in sample preparation for the glass disks are from
weighing and CRM or RM’s certified concentration. In the case of weighing, the
uncertainty is Type A and is stated by the used equipment’s calibration certificate. In
Appendix C, the balance’s calibration certificate, for example, states the balance’s
uncertainty of measurement to be ±0.0004g at a confidence level of approximately
95 %.
The standard uncertainty of each CRM and RM concentration is based on the

available information on the certificate. Table 3.11 lists the approaches used to
determine standard uncertainty of the CRM and RM values depending on the available
information.
70
Chapter 3
Table 3.11: Evaluation of a concentration value (C)’s uncertainty based on the

available information and the calculation of the associated standard
uncertainty.
Information of
Standard uncertainty Evaluation approach
uncertainty as reported
calculation notes
on calibration certificate
- t-Student probability
distribution
𝑠 - At N data points, the
𝑈 = 𝑡𝛼 × degrees of freedom (df) is
√𝑁
N-1
𝑈
Where 𝑢(𝑥𝑖 ) = - The k factor is obtained
𝑘 from the Student’s t table
at df and stated
α = 0.01, or 0.05
confidence level.
s = standard deviation,
- At confidence level of 99
N = number of data points
%, α = 0.01, at 95 % α =
0.05
- Rectangular probability
Standard deviation, s distribution is assumed
due to lack of other
or information
𝑠𝑒𝑚𝑖 − 𝑟𝑎𝑛𝑔𝑒
𝑢(𝐶) = - In the case of reported s, s
√3 is the semi-range
%COV
- In the case of reported
No sample number is %COV, the s is first
reported determined by
%𝐶𝑂𝑉
𝑠= ×𝐶
100
- Normal probability
distribution
- Tolerance is the difference
Upper (UL) and lower (LL) between the UL and LL
limits at a specific 𝑈𝐿 − 𝐿𝐿 - At a specific confidence
𝑢(𝐶) =
confidence level for the 𝑘 level, CL, which is also the
certified value, C coverage probability
distribution, the coverage
factor k = 2 for 95 % CL
and k = 2.58 for 99 % CL
Expanded uncertainty, U 𝑈 - Normal probability
at coverage factor k 𝑢 (𝐶 ) = distribution
𝑘
71
Chapter 3

Information of
Standard uncertainty - Evaluation approach
uncertainty as reported
calculation notes
on calibration certificate
A list of concentration
- Normal probability
values with the average to
distribution
be used as the certified 𝑠(𝐶)
𝑢(𝐶) = - Calculate standard
value, C √𝑛 deviation, s(C) from the
n = number of
set of concentrations
concentration values
Robust standard deviation
s(C) is reported with the - Normal probability
𝑠(𝐶)
number (n) of 𝑢(𝐶) = distribution
participating laboratories √𝑛
or analysis
Robust standard deviation - Rectangular distribution is
𝑠(𝐶)
s(C) is reported without 𝑢(𝐶) = assumed due to lack of
other information √3 other information
CRM certificate states,
“value is not expected to - Rectangular probability
deviate from "true" value 𝑠𝑒𝑚𝑖 − 𝑟𝑎𝑛𝑔𝑒 distribution is assumed
𝑢(𝐶) =
by more than ± 1 in the √3 - Semi range is determined
last significant figure from the value ± 1 or 5 the
reported. For a subscript significant figure.
figure - not more than ± 5
3.9.2 Evaluation of precision
Information from two-factor ANOVA in the repeatability and intermediate precision

tests, as listed in Table 3.10 and described in 3.7.5, is used to quantify the standard
uncertainty.
3.9.3 The regression line
The following factors are used for the estimation of the uncertainty of the regression
line (Eq. 3.5) in SuperQ:
- the counting statistical error (CSE) of the matrix, line and background corrected
instrumental response (MIi) as calculated by the operating software, i.e. u(MIi) =
CSE with its unit being kCPS,
- the value of the CRMs and RMs used in the regression (%), \
72
Chapter 3
- the amount of the used CRMs and RMs in the regression,

- the slope (E) (%/kCPS), and
- the intercept on the X-axis (D).
SuperQ provides three options of ‘error weighting function’ (EWF) for fitting the
regression line through the data points with the reported goodness of fit, which is then
used to determine the uncertainty of the slope (Table 3.12). Square root error
weighting was selected for the study, and the standard uncertainty of the regression
line is K divided by the number of used standards.
Table 3.12: The calibration regression line’s goodness of fit as calculated per
selected EWF by SuperQ[76]. 𝐶𝑠𝑡𝑑 is the assigned concentration of the
standard glass disk, 𝐶𝑐𝑎𝑙𝑐 the calculated concentration, 𝐶0 the
concentration at (0,0), the constant is a value entered by the user into
SuperQ and has the same unit as the concentration (typical value is
0.1 %), k is the number of coefficients in the regression[76].
Error Goodness of
EWF Formula
weighting fit
Root Mean 1
None 1 𝑅𝑀𝑆 = √ ∑(𝐶𝑐ℎ𝑒𝑚 − 𝐶𝑐𝑎𝑙𝑐 )2
Square (RMS) 𝑛−𝑘
RE (Relative 1 𝐶𝑐ℎ𝑒𝑚 − 𝐶𝑐𝑎𝑙𝑐 2

Linear 𝐶𝑠𝑡𝑑 + 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑅𝐸 = √ ∑[ ]
Error) 𝑛−𝑘 𝐶𝑐ℎ𝑒𝑚 + 𝐶0
Square 1 (𝐶𝑐ℎ𝑒𝑚 − 𝐶𝑐𝑎𝑙𝑐 )2

√𝐶𝑠𝑡𝑑 + 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 K (K-factor) 𝐾= √ ∑
root 𝑛−𝑘 𝐶𝑐ℎ𝑒𝑚 + 𝐶0
Although the intercept of the regression line in this study is locked at (0,0), the
uncertainty of D is still to be determined, and SuperQ does not provide information for
it. EXCEL can be used to plot the CRMs’instrumental response against their
concentrations using the linest function to determine the standard deviation of the
slope and intercept using Equations 3.10 to 3.12[77]. Dividing the relevant standard
deviation by the number of used CRMs in the regression provides the standard
73
Chapter 3
uncertainty of the slope and intercept. However, this approach was not used for this
study as XGENLINE 8.1[78] was available and used instead.
Random errors in the Y-direction:

∑𝑖(𝐶𝑠𝑡𝑑 −𝐶𝑐𝑎𝑙𝑐 )2
𝑠𝑦/𝑥 = √ Equation 3.10
𝑛−2
Standard deviation of the slope (E):

𝑠𝑦/𝑥
𝑠𝐸 = Equation 3.11
√∑𝑖(𝑥𝑖 −𝑥̅ )2
Standard deviation of the intercept (D):

∑ 𝑥2
𝑠𝐷 = 𝑠𝑦/𝑥 √𝑛 ∑ (𝑥𝑖 −𝑥̅
𝑖
)2
Equation 3.12
𝑖 𝑖
Where
𝐶𝑠𝑡𝑑 = assigned concentration value of the CRM or RM
𝐶𝑐𝑎𝑙𝑐 = calculated concentration of the CRM or RM
n = amount of CRMs and RMs used in the regression
𝑥𝑖 = instrumental response (background, line overlap and matrix corrected)
𝑥𝑖 = average instrumental response
The SuperQ calibration data and the uncertainty of intercept on the Y-axis as
calculated by XGENLINE 8.1 are used to evaluate the effect of the intercept’s
uncertainty on the final uncertainty of measurement.
74
Chapter 3
3.9.4 The top-down approach
Two of the three top-down approaches are used in the study: the Single Laboratory
validation and Proficiency Testing approaches. Both approaches use replicate and QC
results to quantify the precision uncertainty while the bias uncertainty is quantified
differently. CRM results of at least five different analytical series such different runs
are used to quantify the uncertainty of the test method bias. For this study, five CRMs
used in the calibration were evaluated to quantify the bias’ uncertainty. In the PT
approach, the PT results from at least six rounds are used. The philosophy of this
approach is that the control sample’s results and replicates of the test sample should
include the effects of the preparation and analytical instruments, while the PT results
and CRMs quantify the test method’s bias.
This approach also uses either relative or absolute uncertainty of measurement

depending on whether the values are close to the LOD[6]. Uncertainty of concentration
values close to LOD is expressed as absolute uncertainty, mainly based on replicate
results and is illustrated in Figure 3.14.
75
Chapter 3
Figure 3.14: The relationship between the measurand concentration and the absolute
(i) and relative (ii) uncertainties, respectively[6]. The dashed line divides
the measurement range into concentration areas where the absolute
uncertainty is constant and where it becomes larger with higher
concentration (i). In the case of the relative uncertainty (ii), the
measurement range is divided into concentration ranges where the
uncertainty is large at low concentrations and is approximately constant
at higher concentrations.
Table 3.13 lists the top-down approach’s steps to calculate the measurands’
uncertainty of measurement and largely follows the Nordtest method[6].
76
Chapter 3
Table 3.13: The top-down approach to establish a test sample’s uncertainty of

measurement using QC, replicates and PT or CRM results.
Step Action Process
Specify the measurands’
Measurands and working range have been
working range.
established during method validation. At low
Decide on either relative or
1 concentration levels, absolute uncertainty is
absolute uncertainty based
used, while relative uncertainty is selected for
on the measurand
medium and higher concentration levels.
concentration.
Quantify the control sample’s
uncertainty from its control
limits. Calculate the standard
2 2
𝑢(𝑅𝑤 ) = √𝑠𝑅𝑤 + 𝑠𝑟2
deviation (sr) if the test
sample is analysed in
replicates.
Quantify the test method’s
bias:
- RMS of the different - Using PT results:
rounds’ bias in the case
2
of PT results. 𝑅𝑀𝑆𝑏𝑖𝑎𝑠 = √∑ 𝐵𝑖𝑎𝑠𝑟𝑜𝑢𝑛𝑑
3
- Recovery in the case of - Using CRM results:
CRM results. 𝑅𝑒𝑠𝑢𝑙𝑡 − 𝐶𝑟𝑒𝑓
𝐵𝑖𝑎𝑠 = %𝑅𝑒𝑐𝑜𝑣𝑒𝑟𝑦 =
- For relative uncertainty, 100
convert the bias to
relative bias.
- Using PT results:
Convert the uncertainties in 𝑢(𝑏𝑖𝑎𝑠) = √𝑅𝑀𝑆𝑏𝑖𝑎𝑠 2 + 𝑢(𝐶𝑟𝑒𝑓 )
2
steps 2 and 3 to standard

4 uncertainties.
- Using CRM results:
n = amount of CRM analysis 𝑠𝑏𝑖𝑎𝑠 2 2

𝑢(𝑏𝑖𝑎𝑠) = √𝐵𝑖𝑎𝑠 2 + ( ) + 𝑢(𝐶𝑟𝑒𝑓 )
√𝑛
77
Chapter 3
Step Action Process

Combine the standard
5 uncertainties of the QC, 𝑢𝑐 = √𝑢(𝑅𝑊 )2 + 𝑢(𝑏𝑖𝑎𝑠)2
replicates, and bias.
Calculate the expanded
6 uncertainty at the confidence 𝑈 = 2 × 𝑢𝑐
level of approximately 95 %.
Apply the relative uncertainty 𝑅𝑒𝑠𝑢𝑙𝑡 ∗ 𝑈𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒
7 𝑈=
on the test sample’s results. 100
3.9.5 The bottom-up approach
The main steps in the evaluation and calculation steps as per the GUM framework
(Table 2.5) are:
i) The specification of the measurand, identification of all the input qualities on which
the output quantity depends and the setup of a functional relationship between the
input and output quantities. A fishbone diagram is used to identify the main
sources of uncertainty affecting the measurand result (Figure 3.15). More than one
measurement model and fishbone diagrams are created to cover the test method.
ii) Evaluation of each input estimate to establish the standard uncertainty process as
listed in Table 3.11.
iii) Evaluation of the correlation between any input estimates. The correlation function
in EXCEL is used for this process.
iv) Determination of the sensitivity coefficients for each input quantity using partial
derivatives, which describes the change in the output estimate with any change in
the input estimates.
v) Calculation of the combined standard uncertainty.
For the MCM, the abovementioned steps (i) to (iii) are the same, and other steps are:
vi) The identification of the probability distribution type as it defines the creation of
random numbers in the specified range based on the standard uncertainty.
78
Chapter 3
vii) Using the same measurement model, different permutations of the measurement
model are calculated using pseudo-random values of the input quantities. The
random values of the input quantities are restricted to the range as determined by
the input quantity’s uncertainty. The selection of random values in a range is based
on the uncertainty’s probability distribution type (Table 3.14). The permutation
process is repeated at least 100000 times.
viii) Calculation of the standard deviation of the set of results in step vii.
Table 3.14: EXCEL formulas[47] to select a random value in a value range. RAND()
is an EXCEL formula to calculate random values between 0 and 1. a,
b, and c refer to cell addresses in the EXCEL spreadsheet for the input
value (a), standard uncertainty (b) and the half-width interval (c).
Probability distribution type EXCEL formula used in study
Normal probability distribution =NORM.INV(RAND(),a,b)
Rectangle probability distribution =(a-c)+(2*c*(RAND()))
79
Chapter 3
Figure 3.15: The measurement model and the cause-and-effect diagram to

determine the main sources of uncertainty. Ci is the concentration of
the measurand, Ei is the slope of the calibration’s regression line, Di is
the intercept value, and MIi is the matrix corrected instrumental
response. P is to account for uncertainty due to precision and has a
value of 1.
As the test method involves more than one process for determining the test sample
concentrations, more than one measurement model with associated uncertainty
estimation is created. Figure 3.16 illustrates the four different processes in the test
method. On the right side of this figure, the four main processes have their
measurement model, contributing to the final estimation process.
80
Chapter 3
Figure 3.16: The connection between the sample preparation and analysis
processes with the determination of the measurand concentrations
and associated uncertainties.
EXCEL for Microsoft 365 is used to set up a template to calculate the measurand’s
uncertainties for each of the two GUM and Kragten methods. A dataset of samples’
preparation information and matrix corrected counts along with the calibration data
such as the slope and intercept from SuperQ and EXCEL is processed by the different
templates to compare the results.
81
Chapter 3
3.9.6 The uncertainty budget

The uncertainty budget assists in identifying the main contributors to uncertainty in a
measurement model, and the setup of such a budget is part of the Kragten method. In
the GUM framework, the identification of main contributors is done indirectly by
removing an uncertainty value and observing its effect on the rest of the contributors’
uncertainty values.
3.9.7 Software tools used in the study

Besides Microsoft EXCEL, the NIST Uncertainty Machine[48] and XGENLINE (v8.1)[78]
are used in the study to compare the results with those in the EXCEL templates
developed for the GUM approaches. The NIST Uncertainty Machine[48] uses both the
GUM framework and the MCM to evaluate measurement uncertainty. The input
quantities for the four measurement models used in the EXCEL templates are
processed by the web application to obtain the results. Figure 3.17 illustrate the input
process for the LOI measurement model, and figure 3.18 shows the results of this
process.
82
Chapter 3
Figure 3.17: The main window of the NIST uncertainty machine web application. The
variables, probability distribution type, value and associated standard
uncertainty are entered into the middle area. If required, the correlation
values are entered too. The measurement model is defined in the
bottom area, and the green bar at the bottom is pressed to process the
entered information.
83
Chapter 3
Figure 3.18: The results from the GUM (bottom) and MCM (top) for the entered
information in Fig. 3.17. The calculated sensitivity coefficients, for
example, are compared with the sensitivity values calculated by the
EXCEL template to verify their accuracy.
84
Chapter 3
XGENLINE (v8.1)[78] is a software package developed by the National Physical

Laboratory of the UK for data regression and is used to evaluate the uncertainty of the
calibration’s regression line and the intercept on the Y-axis. Uncertainty of both the
instrumental responses and the CRM values are involved in calculating a best fit
function which is then used to obtain estimates of both concentration and uncertainty
for the unknown samples. The results are compared with the concentrations obtained
from SuperQ and the uncertainty values as calculated by the developed EXCEL
templates.
3.10 Summary
- The basic theory of WD-XRF is based on the relationship between the energy
and wavelength of an X-ray photon (Eqs. 3.3a, b and c). The WD-XRF
spectrometer uses Bragg’s law (Fig. 3.1) to identify and quantify present
elements in a sample.
- The study test method is set up to analyse major elements (SiO 2, TiO2, Al2O3,
total Fe as Fe2O3, MnO, MgO, CaO, Na2O, K2O, P2O5 and Cr2O3) on glass disks
by WD-XRF and LOI by gravimetry. A wide range of sample types (Table 3.2)
to cover typical samples from South Africa is to be analysed by the test method.
- One of three types of flux, a eutectic mix of lithium metaborate and lithium
tetraborate, is selected to dissolve the roasted sample material and fuse it into
a stable glass disk for analysis by WD-XRF. The flux components Li2O and
B2O3 are to be quantified based on the selected flux, and its mass is entered
into SuperQ before analysis of the glass disk.
- SuperQ applies corrections for instrumental drift, background and line-overlaps
while FP in SuperQ corrects for matrix effects. The corrected instrumental
responses are plotted on the Y-axis against the glass disk concentration on the
X-axis.
- Test materials for the study are MzBP-1, SARM-36, and a limestone sample
and an in-house QC sample.
- The test method’s measurement processes can be divided into four main
stages:
(i) Roast of sample and determination of LOI by gravimetry.
85
Chapter 3
(ii) Weigh 10 g flux and calculate % Li2O and % B2O3 based on selected flux
and its mass.
(iii) Prepare glass disk with 1 g roasted material and 10 g flux. Enter flux
components into SuperQ before analysis of glass disk by WD-XRF.
(iv) Calculate final results by removing flux concentrations from SuperQ
results and adding LOI result.
- The test method performance parameters are selectivity, working range and
linearity, trueness (Recovery and the En-score), precision, LOD and LOQ.
- Metrological traceability links the produced measurement result with SI as the
calibration uses certified CRMs to be linked to SI.
- Precision and bias strongly affect the accuracy of measurement, and both are
to be quantified.
- The uncertainties of the regression line’s slope and intercept on the Y-axis is
calculated by XGENLINE in this study, although they can be calculated using
Eqs. 3.11 and 3.12.
- Two top-down approaches, the Single Laboratory validation and Proficiency
Testing approaches, use replicate and QC results to quantify the precision
uncertainty while the bias uncertainty is quantified differently. The first approach
employs CRM results and the other results from PT-rounds to quantify the test
method’s bias.
- The bottom-up approach’s main steps are the definition of the measurand, the
measurement model, the identification and quantification of the effect sources,
determination of the standard uncertainty based on the effect source’s
evaluation type and finally, the propagation of the standard uncertainties. The
GUM framework, MCM and the Kragten method is used in the study.
- Software tools in the study are EXCEL used for ANOVA and the development
of templates for the GUM framework, MCM and the Kragten method. Other
tools are XGENLINE and the NIST uncertainty machine. In XGENLINE, the
uncertainties of both the CRMs’ instrumental responses and concentration
values are used. The NIST uncertainty machine is used in the study to verify
the values of the sensitivity coefficients in the GUM framework template.
- The uncertainty budget identifies the main contributors of uncertainty.
86
Chapter 4
Chapter 4:
Results and Discussion
4.1 Introduction
The method validation results leading to the quantification of the uncertainty of

measurement of major elements (SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaO, Na2O,
K2O, P2O5 and LOI) are described in this chapter. The different quantification
processes and results of the top-down and bottom-up approaches for the evaluation
of uncertainty of measurement are presented and discussed here. The EXCEL
templates for the bottom-up approaches for the evaluation of the uncertaint of
measurement are also described in Chapter 4.
4.2 Method validation
In the method validation process, the test method characteristics such as the working
range, linearity, trueness and precision are used to evaluate the test method and to
confirm that the test method is fit for purpose. Appendix D lists the certified or assigned
values for the concentrations in the CRMs, and RMs used to set up the calibration for
each measurand (Table D1 in Appendix D). The certified values are then calculated
to include the flux (Li2O and B2O3) and exclude LOI (Table D2). The calculated values
(Table D2 in Appendix D) are entered into the calibration application in SuperQ for the
regression plot with the matrix corrected intensities (Figure 4.1). The method validation
parameters are based on the results of the calibrations in SuperQ.
87
Chapter 4
Figure 4.1: The calibration application in SuperQ listing the compound, unit,
regression model, slope, correlation, and the goodness of fit value for
the regression slopes.
4.2.1 Selectivity
Figure 4.2 depicts the measurement application to analyse measurand peaks,

background, and interfering peaks. As LOI is determined gravimetrically, there is no
selectivity parameter value for it.
Figure 4.2: The application setup with selected spectrometer parameters to

analyse the measurands by WD-XRF.
4.2.2 Linearity, working ranges, limit of detection (LOD) and limit of

quantification (LOQ)
Linearity of the calibrations for each measurand is expressed as the correlation

between the instrumental response (kCPS) and the CRMs’ concentration which should
be close to one (Table 4.1). There is no linearity evaluation in the case of LOI as it is
88
Chapter 4
analysed by gravimetry. The working ranges are also listed in Table 4.1. They are
established by the minimum and maximum values of the RMs and CRMs in the
calibration. The LOD replaces the minimum value if the minimum value is less than
the LOD. The test method’s LOD and LOQ values are the averages of the LOD and
LOQ of the used CRM and RMs (Table 4.1). Except for Al2O3 and K2O, the LOD and
LOQ are the same for all the measurands.
Table 4.1: The main method validation parameters of the measurands and their
values.
Linearity Working range

Measurand LOD (%) LOQ (%)
R2 (%)
SiO2 0.99927 0.37 to 99.78 0.01 0.01
TiO2 0.99958 0.005 to 3.22 0.002 0.002
Al2O3 0.99984 0.036 - 88.2 0.02 0.03
Fe2O3 0.99995 0.012 to 95.26 0.002 0.002
MnO 0.99982 0.00013 to 61.39 0.002 0.002
MgO 0.99994 0.0013 to 44.11 0.02 0.02
CaO 0.99993 0.006 to 55.4 0.002 0.002
Na2O 0.99972 0.003 to 10.36 0.02 0.02
K2O 0.99955 0.004 to 11.23 0.002 0.003
P2O5 0.99986 0.008 to 30.2 0.01 0.01
Cr2O3 0.99984 0.0001 to 48.97 0.001 0.001
LOI - 0.08 to 43.4 Not determined
89
Chapter 4
4.2.3 Trueness
Appendix E (Tables E1 to E10) lists the evaluation of each calibration in terms of
recovery and the En-scores, while Figures 4.3 (a) to (l) demonstrate the test method’s
trueness and concentration ranges graphically. In Table 4.2, the trueness of each
measurand is summarized.
Figures 4.3 (a) to (d): Graphic evaluation of the test method – the blue circles
represent the results of a CRM; the orange circles the
proposed value of a CRM, and the yellow boxes the
proficiency test materials. The measurand results compare
favourably with their certified or assigned values.
90
Chapter 4
Figures 4.3 (e) to (l): Graphic evaluation of the test method .(Continued)
91
Chapter 4
Table 4.2: Evaluation of each measurand’s trueness. The En and recovery values are
listed in Appendix E.
Measurand Notes on measurand’s trueness
SiO2 From a total of 59 fusion disks, En-values for ten disks are more than one.
The average recovery is 100.6 % which is satisfactory.
From a total of 56 fusion disks, En-values for eight disks are more than
TiO2 one. The average recovery is 103 %. Most high En values are for the
standards close to the LOD. Overall, the TiO2-calibration is acceptable.
From a total of 59 fusion disks, En-values for eight disks are more than
Al2O3 one. The average recovery is 102. 8%. Overall, this calibration is
acceptable.
From 59 fusion disks, En-values for eight disks are more than one. Two
disks, 1298 and 1299, are both from the same CRM (BCS-313/1, high
Fe2O3 purity silica), and both disks display similar results indicating possible
contamination of the CRM with Fe2O3. Excluding the two fusions, the
average recovery is 101.8 % which is satisfactory.
From 45 fusion disks, En-values for twelve disks are more than one,
indicating insufficient bias correction for some sample types such as Cr-
ore (1280, 1334), while fusions such as 1260, 1261,1284, 1285, 1286 and
MnO 1338 are RMs and were not part of the calibration for MnO. Excluding
those RMs, the average recovery is 109.2 % which is at this stage
acceptable as the high recovery values are mainly from disks from MnO
values close to LOD.
From 45 fusion disks, En-values for fifteen disks are more than one.
Fusions such as 1282 and 1284 are RMs and are not part of the MgO
MgO
calibration. Excluding those RMs, the average recovery is 104.1 %, which
is acceptable at this stage.
From 59 fusion disks, En-values for thirteen disks are more than one,
CaO mainly due to the proximity to LOD. The average recovery is 96.5 % which
is acceptable.
From 46 fusion disks, En-values for thirteen disks exceeds one. The main
reasons are the proximity to LOD (1298, 1278, 1300), the fusions being
Na2O prepared from an RM (1256, 1271, 1278) and contamination of the fusion
disks by handling (1266). Excluding those fusions, the average recovery
is 96.9 % which is acceptable.
From 53 fusion disks, En-values for seven disks exceeds one. Low
recovery values for 1270, 1271, 1278, 1371, 1372 and 1375 indicate
K2O insufficient bias correction for some sample types such as limestones and
Fe-ores. Excluding the two sample types, the average recovery is
100.2 % which is satisfactory.
92
Chapter 4

Measurand Notes on measurand’s trueness
From 43 fusion disks, En-values for six disks exceeds one. Most of those
disks are from RMs (1262, 1267, 1276) and excluding them in the recovery
P2O5
calculations produce an average recovery value of 99.7 %, which is
satisfactory.
From 34 fusion disks, En-values for nine disks exceeds one. Most of those
disks are from RMs (1255, 1267, 1270, 1271, 1278, 1284, 1285, 1299).
Cr2O3
Excluding all the RMs in the recovery calculations produce an average
recovery value of 109.2 %, which is acceptable at this stage.
LOI values are not available for most CRMs and RMs used in the test
study. From thirty available LOI values, eighteen results are satisfactory.
However, if all evaluation values for the RM standards are excluded from
LOI
the evaluation, the average recovery is 99.3 % which is acceptable.
Furthermore, the most recent PT evaluation for LOI (GeoPT-49, z= 2.12)
indicates the fitness for purpose of the LOI analysis.
4.2.4 Precision
The effects under different precision conditions as listed in Table 3.10 are described
and, where applicable, quantified for estimating the uncertainty of measurement.
4.2.4.1 Sampling: primary sample preparation
In Appendix F, the results of eleven test sample portions of limestone are listed. Except
for two pellets (fourth pellet in 1-13abba and second pellet in 1-24bbab), the results of
the test samples in each portion are similar, indicating that they are sufficiently
homogeneous.
4.2.4.2 Variation in the split sample and flux masses
Six fusion disks of a SARM-36 aliquot were prepared with variations in the split sample
and flux mass to evaluate the mass effects on the dilution factor. Table 4.3 lists the
different masses for the flux and roasted test material, while Table 4.4 summarises the
mass effects. Appendix G lists the split (a to f) results with the ANOVA calculations.
Due to the low number of replicate analyses, the P- and F-critical values in the ANOVA
calculations are not available.
93
Chapter 4
Table 4.3: Mass of the samples and flux for the six split samples of a SARM-36
aliquot to investigate the dilution effect.
Sample mass Flux mass
Sample name
(g) (g)
SARM-36 (a) 1.0512 9.9505
SARM-36 (b) 1.0208 9.9802
SARM-36 (c) 1.0006 10.0008
SARM-36 (d) 0.9903 10.0102
SARM-36 (e) 0.9802 10.0203
SARM-36 (f) 0.9701 10.0311
Table 4.4: Summary of the six split samples of SARM-36 to evaluate the mass
effects.
Average Standard Relative standard
Measurand
(%) deviation (%) deviation (%)
SiO2 63.87 0.095 0.15
TiO2 0.042 0.0011 2.77
Al2O3 17.59 0.028 0.16
Fe2O3 1.41 0.007 0.005
MnO 0.010 0.0007 0.08
MgO 0.40 0.011 2.62
CaO 0.75 0.003 0.40
Na2O 0.42 0.018 4.24
K2O 15.43 0.050 0.32
P2O5 0.12 0.002 1.45
Cr2O3 0.002 0.0005 24.1
The relative standard deviation of each measurand except for Cr2O3 is less than 5 %,
which is acceptable and reflects the test method's robustness. The high relative
standard deviation of Cr2O3 is due to its concentration being close to the LOD, but the
standard deviation of 0.0005 % is, however, acceptable. Each measurand's relative
uncertainty values are used to determine the standard uncertainty of the precision of
dilution during the bottom-up approach calculations and are listed in the last row in
Figures G1 to G11.
4.2.4.3 LOI
The LOI results of aliquots of a limestone test sample, 2-13baba, are listed in Table
4.5. The data is arranged according to the aliquot position in the furnace, enabling two-
94
Chapter 4
way ANOVA analysis (Appendix H) to evaluate the precision according to a specific

positional area in the furnace.
Table 4.5: LOI-analysis of a limestone test sample under repeatability precision

conditions. The average of the LOI concentrations is 44.74 %, with a
standard deviation of 0.146 %
Left Middle Right
44.78 44.65 44.58
44.94 44.69 44.59
44.74 44.65 44.56
44.76 44.69 44.70
Back
44.97 44.74 44.76
44.95 45.28 44.92
44.96 44.83 44.86
44.91 44.92 44.84
44.76 44.59 44.56
44.72 44.66 44.52
44.68 44.63 44.49
44.97 44.76 44.74
Middle
44.72 44.79 44.74
44.88 44.88 44.88
44.77 44.77 44.81
44.75 44.78 44.72
44.65 44.62 44.50
44.78 44.50 44.45
44.57 44.46 44.51
Front 44.86 44.72 44.66
44.98 44.73 44.67
44.68 44.80 44.78
44.71 44.70 44.85
44.78 44.67 44.72
The LOI values for aliquots roasted in the furnace's left area are higher than those
from the middle and right areas. The precision under repeatability conditions is
0.146 %. A set of 18 aliquots analysed during a different day averaged 44.38 %, with
a standard deviation of 0.175 %. Combining the two datasets leads to an intermediate
precision for the different conditions of 0.210 %.
95
Chapter 4
4.2.4.4 The fusion disk
Four aliquots of SARM-36 were used for six fusion disks per aliquot, with each fusion
disk for each aliquot prepared using the same mould and fluxer. Both the new and
used (with scratches) moulds were used to investigate the effect of mould quality on
the results. Table 4.6 lists the summary of the moulds and fluxer effects of the results
with the full set of results in Appendix I.
Table 4.6: Statistics (average, standard deviation and % relative standard deviation
of six results) of six fusion disks prepared from four aliquots of SARM-36
using different fluxers and moulds. A column of results represents an
aliquot’s split statistics, while the different columns represent the effects
of the fluxer and moulds. The full set of results is presented in
Appendix I.
FM fluxer TheOx fluxer
Old New Old New
mouldible mouldible mould mould
Average 64.04 64.08 63.66 63.77
SiO2 Standard deviation 0.132 0.133 0.037 0.039
% Relative. std. dev. 0.21 0.21 0.06 0.06
Average 0.048 0.048 0.047 0.047
TiO2 Standard deviation 0.002 0.002 0.001 0.001
% Relative. std. dev. 4.20 3.57 2.08 2.30
Average 17.37 17.37 17.29 17.30
Al2O3 Standard deviation 0.068 0.052 0.041 0.039
% Relative. std. dev. 0.39 0.30 0.24 0.22
Average 1.41 1.40 1.41 1.41
Fe2O3 Standard deviation 0.005 0.008 0.003 0.004
% Relative. std. dev. 0.35 0.56 0.20 0.30
Average 0.010 0.010 0.010 0.010
MnO Standard deviation 0.0007 0.0008 0.0007 0.0010
% Relative. std. dev. 7.02 7.70 7.27 9.61
Average 0.331 0.334 0.338 0.336
MgO Standard deviation 0.0103 0.0089 0.0098 0.0081
% Relative. std. dev. 3.12 2.68 2.90 2.42
Average 0.706 0.706 0.716 0.715
CaO Standard deviation 0.0037 0.0035 0.0035 0.0036
% Relative. std. dev. 0.53 0.49 0.49 0.51
Average 0.403 0.389 0.402 0.401
Na2O Standard deviation 0.0150 0.0125 0.0132 0.0125
% Relative. std. dev. 3.73 3.21 3.29 3.11
Average 15.37 15.33 15.56 15.55
K2O Standard deviation 0.066 0.077 0.029 0.014
% Relative. std. dev. 0.43 0.50 0.19 0.09
96
Chapter 4

FM fluxer TheOx fluxer
Old New Old New
mouldible mouldible mouldible mouldible
Average 0.12 0.12 0.12 0.12
P2O5 Standard deviation 0.001 0.002 0.003 0.002
% Relative. std. dev. 0.98 1.63 2.20 1.67
Average 0.003 0.002 0.005 0.004
Cr2O3 Standard deviation 0.0005 0.0004 0.0028 0.0016
% Relative. std. dev. 19.58 17.64 55.27 40.14
The average concentration value for the old and new moulds prepared for any fluxer
is mostly similar, however, the average value per fluxer differs for SiO2, CaO, K2O and
Cr2O3. For SiO2, TiO2, Al2O3, Fe2O3 and Cr2O3, the % relative standard deviation
differs between the two fluxers.
Both the average and standard deviation of the SiO 2-values are higher for the
Beadmaster than the TheOx fluxer. Such differences are not significant for TiO 2, Al2O3,
MnO, MgO, Na2O and P2O5. For Fe2O3, the average values are similar for both fluxers,
but the standard deviation is higher for the Beadmaster fluxer. In the case of CaO, the
average values are higher for the TheOx fluxer, while the standard deviation values
are similar for both fluxers. For K2O, the average values produced by the TheOx fluxer
are higher than the Beadmaster fluxer, while the standard deviation is larger in the
Beadmaster. In the case of Cr2O3, both the average and the standard deviation is
higher for the TheOx fluxer.
The precision values per fluxer (bottom rows in each figure of Appendix I) are used to
quantify the standard uncertainty of the fluxer precision in the bottom-up approaches.
4.2.4.5 Sampling: aliquot and split samples
The experiments to evaluate the effects of the split sample and flux masses (in 4.2.4.2)
and the fusion process (in 4.2.4.5) on the results also involves aliquot and split
sampling effects. Except for Cr2O3, the precision for split sampling effects are
acceptable. The high relative standard deviation of Cr2O3 can be ascribed to its
concentration value being close to its LOD.
97
Chapter 4
4.2.4.6 Instrumental drift
A fusion disk of MzBP-1 was analysed 189 times from 27 June 2021 to 31 August
2021 to evaluate the instrumental drift correction applied by SuperQ. This period was
a period that included load shedding, which is not ideal for this evaluation. Although
the Zetium instrument is connected to an uninterrupted power supply system (UPS),
the generator did not operate regularly for various reasons leading to the shutdown of
the UPS and the Zetium. The results of the 189 analyses are summarised in Table
4.7. reflecting the robustness of the test method and the Zetium XRF spectrometer.
Table 4.7: Summary of the 189 major element results on a fusion disk to evaluate
the instrumental drift for two months.
Standard
Average Minimum Maximum
deviation
SiO2 57.73 0.046 57.61 57.84
TiO2 1.08 0.004 1.06 1.09
Al2O3 19.70 0.047 19.51 19.81
Fe2O3 4.80 0.010 4.76 4.82
MnO 0.147 0.0014 0.141 0.150
MgO 1.01 0.024 0.98 1.13
CaO 3.77 0.009 3.73 3.79
Na2O 6.81 0.029 6.75 6.91
K2O 3.86 0.007 3.83 3.87
P2O5 0.62 0.004 0.61 0.63
Cr2O3 0.018 0.0005 0.017 0.020
4.2.4.7 The XRF spectrometer: sample position and cup
A fusion disk of MzBP-1 was analysed five times in a measuring cup in both sample
positions 1 and 2. Eight cups were evaluated, leading to 80 results per major element.
The results were arranged with the columns arranged for the cups and rows for the
measuring positions (Figure 4.4), and two-factor ANOVA was applied to the data.
98
Chapter 4
SiO2
Cup 1 Cup 2 Cup 3 Cup 4 Cup 5 Cup 6 Cup 7 Cup 8
Pos 1 57.87 57.77 57.71 57.81 57.84 57.92 57.82 57.78
57.88 57.82 57.75 57.79 57.85 57.90 57.84 57.78
57.82 57.80 57.82 57.78 57.84 57.79 57.77 57.81
57.90 57.80 57.83 57.86 57.81 57.83 57.84 57.79
57.90 57.90 57.79 57.85 57.86 57.83 57.87 57.79
Pos 2 57.83 57.79 57.80 57.79 57.77 57.80 57.86 57.75
57.91 57.78 57.82 57.81 57.76 57.84 57.88 57.80
57.90 57.77 57.81 57.87 57.88 57.82 57.90 57.78
57.88 57.80 57.88 57.80 57.86 57.82 57.80 57.76
57.85 57.79 57.82 57.82 57.85 57.86 57.87 57.86
Figure 4.4 Analysis of SiO2 in a fusion disk of MzBP-1 and arrangement of results
in EXCEL. The data is arranged per column for the cups and per row for
the measuring position. The alpha level (α) for the ANOVA is 0.05.
Figures J1 to J11 in Appendix J details the evaluation data per major element, and the
ANOVA results are summarised and evaluated in Table 4.8. The values of the relative
standard uncertainty for measuring cup 5 used in the bottom-up approach for the
evaluation of the uncertainty of measurement are also listed in the last rows in Figs.
J1 to J11.
99
Chapter 4
Table 4.8: Summary of two-factor ANOVA at alpha level (α) of 0.05 to evaluate eight sample cups in two measuring positions of
the Zetium XRF spectrometer.
Major Sample P- Sample Columns Columns Interaction
Evaluation notes
element value F<Fcrit P-value F<Fcrit P-value
The measuring position has no significant impact on SiO2 results,
while there are significant differences between the sample cups.
SiO2 0.96 Yes 0.00005 No 0.23
There is no significant interaction between the measuring cups
and sample positions.
The measuring position has no significant impact on TiO2 results,
while there are significant differences between the sample cups.
TiO2 0.44 Yes 0.02 No 0.001
There is significant interaction between the measuring cups and
sample positions.
The measuring position has no significant impact on the Al2O3
results, while there are significant differences between the sample
Al2O3 0.19 Yes 0.02 No 0.16 cups.
The measuring position has no significant impact on the Fe2O3
results, while there are significant differences between the sample
Fe2O3 0.63 Yes 0.008 No 0.008 cups.
sample positions.
The measuring position and sample cups have no significant
impact on the MnO results.
MnO 0.51 Yes 0.50 Yes 0.32
100
Chapter 4

Major Sample P- Sample Columns Columns Interaction
Evaluation notes
element value F<Fcrit P-value F<Fcrit P-value
impact on the MgO results.
MgO 0.66 Yes 0.56 Yes 0.93
impact on the CaO results.
CaO 0.70 Yes 0.06 Yes 0.01
There is some interaction between the measuring cups and
sample positions.
The measuring position and sample cups impact the Na2O results.
Na2O 0.04 No 0.0000003 No 0.17 There is interaction between the measuring cups and sample
positions
The measuring position impacts the K2O results, while there are
no significant differences between the sample cups.
K2O 0.02 No 0.73 Yes 0.003
sample positions.
impact on the P2O5 results.
P2O5 0.90 Yes 0.59 Yes 0.85
impact on the Cr2O3 results.
Cr2O3 0.52 Yes 0.10 Yes 0.79
101
Chapter 4
4.2.4.8 Duplicates
About 1600 duplicate results were extracted from a database of results by WD-XRF
to evaluate the repeatability of the analysis. Table 4.9 lists the LOI concentration range
and the average standard deviation per concentration range. The other major
elements were also evaluated but are not included as the information is not used in
this study.
Table 4.9: LOI’s concentration range, average standard deviation and % relative
standard deviation based on duplicate results from samples analysed at
the CGS during the past five years using the study test method. The
values are used for the quantification of LOI’s precision in the EXCEL
templates described in 4.3.
Average Absolute
Number Average % standard Relative
Concentration of Average standard relative uncertain standard
Range (%) duplicate (%) deviation standard ty uncertain
results (%) deviation ty (%)
(%) (%)
Less than zero 66 0.039 7.04 0.00483
0 to 5 658 2.23 0.046 5.27 0.00181 0.000812
5 to 10 522 6.70 0.060 0.90 0.00262 0.000391
10 to 20 139 12.80 0.101 0.83 0.00860 0.000672
20 to 30 24 24.69 0.049 0.20 0.01010 0.000409
30 to 40 38 35.21 0.087 0.25 0.01407 0.000399
40 to 50 54 43.41 0.043 0.10 0.00592 0.000136
50 to 60 8 56.15 0.030 0.07 0.01071 0.000191
Greater than
24 74.70 0.095 0.13 0.01943 0.000260
60
The average standard deviation values for the LOI duplicates are the largest, close to
0, reflecting the relationship between the measurand concentration and the relative
uncertainty described in Fig. 3.14. The absolute standard uncertainties are used for
LOI concentration values between 0 and 5 %, while the relative standard uncertainties
are used for LOI concentration values larger than 5 %.
102
Chapter 4
4.3 Uncertainty of measurement
Both the top-down and bottom-up approaches for the evaluation of the uncertainty of
measurement for the results are described in this section, and the evaluated
uncertainties are compared with each other.
4.3.1 The bottom-up approach
The first two phases of the bottom-up approach, the modelling and the determination
of the standard uncertainties for the input quantities, are identical for both the classic
GUM and MCM approaches, while the propagation of the standard uncertainty values
differs for the two approaches. Three EXCEL templates were developed for this study:
i) the GUM framework,
ii) the MCM, and
iii) the Kragten method.
4.3.1.1 The four measurement models in the test method
The four analytical processes contributing to the final uncertainty of measurement are
outlined in Figure 3.16 and are as follows:
i) Measurement model 1: the determination of LOI (Fig. 4.5),
ii) Measurement model 2: the ‘flux calculator’ to determine the flux components,
Li2O (Fig. 4.6) and B2O3 (Fig. 4.7), in the fusion disk,
iii) Measurement model 3: the results from SuperQ (Fig. 4.8), and
iv) Measurement model 4: the major oxide results after including LOI and excluding
Li2O and B2O3 (Eq. 4.1).
4.3.1.1.1 Measurement model 1: Determination of LOI
The test method’s first process is to determine the test sample’s LOI, and its
measurement model is defined in Figure 4.5, while the fishbone diagram is used to
identify the contributors to the LOI’s uncertainty of measurement.
103
Chapter 4
Figure 4.5: Measurement model 1 at the top and fishbone diagram at the bottom
to identify contributors to the uncertainty of the output quantity, LOI (%).
Cr = mass of the empty porcelain crucible (g)
Cr100 = mass of crucible and test material after drying at 100 ⁰C (g)
C1000 = mass of crucible and test material after roasting at 1000 °C (g)
P = a value of 1 allocated to precision.
4.3.1.1.2 Measurement model 2: Li2O and B2O3 in the fusion disk
As noted in Chapter 3, the selected flux, the masses of flux and the test sample
determine the amount of Li2O and B2O3 in the fusion disk. Before analysing the fusion
disk by WD-XRF, those values are entered into SuperQ. The measurement models
for the determination of Li2O and B2O3 are identical (Figs. 4.6 and 4.7).
104
Chapter 4
Figure 4.6: Measurement model 2 for the determination of Li2O (%) in the fusion
disk. The fishbone diagram at the bottom includes the contributions to
uncertainty by the purity of the flux, the balance and the atomic masses
of Li, O and B.
Li2O(a) = Li2O from the lithium tetraborate portion in the flux (%)
Li2O(b) = Li2O from the lithium metaborate portion in the flux (%)
massflux = mass of the flux (g)
masssample = mass of the roasted test sample (g)
MmLi2O = molecular mass of Li2O (g)
MmB2O3 = molecular mass of B2O3 (g)
%Li2B4O7 = % amount of lithium tetraborate in the flux
%LiBO2 = % amount of lithium metaborate in the flux
105
Chapter 4
Figure 4.7: Measurement model 2 for the determination of B2O3 (%) in the fusion
disk. The fishbone diagram at the bottom and the symbols are similar
to those in Fig. 4.6.
4.3.1.1.3 Measurement model 3: SuperQ results

In measurement model 3, the instrumental responses corrected for drift, background
and line-overlaps as well as matrix effects are converted into concentrations by
regression analysis in SuperQ (Fig. 4.8).
106
Chapter 4
Figure 4.8: Measurement model 3 for the results (Cmeas,i) as calculated by SuperQ
and is expressed as %, Ei = the slope of the regression line, MIi = the
instrumental response corrected for background, line overlap and
matrix effects (kCPS), Di = intercept on the Y-axis, P = a value of 1 and
is used for the propagation of uncertainty due to dilution, the fluxer and
sample cup.
4.3.1.1.4 Measurement model 4: Major oxide concentrations
The final stage of the test method involves multiplication of the major oxide values
(Ci,meas) in the previous stage with a factor based on the Li 2O, B2O3-values entered
into SuperQ and the measured LOI (Equation 4.1). The Total value is the sum of the
measured major oxides from SuperQ, Li2O and B2O3.
𝑇𝑜𝑡𝑎𝑙−𝐿𝑂𝐼𝑚𝑒𝑎𝑠
𝐶𝑖, 𝑢𝑛𝑘 = 𝐶𝑖,𝑚𝑒𝑎𝑠 × 𝑇𝑜𝑡𝑎𝑙−(𝐿𝑖2 𝑂+𝐵2 𝑂3 )
(Eq. 4.1)
4.3.1.2 Standard uncertainty of the input quantities
MzBP-1, a PT sample, is used to illustrate the four measurement models, and Tables
4.10 to 4.5 list the input quantities' values and associated standard uncertainties in the
first three measurement models for the test method.
107
Chapter 4
Table 4.10: MzBP-1’s input quantity and associated standard uncertainty values for
measurement model 1.
Input Standard
Input value Comments
variable uncertainty
5
Temperature 5⁰C has no significant effect on the
1000⁰C √6
(T) result – it is therefore not included.
= 2.0412 °𝐶
0.0003
2 0.0003 g uncertainty as stated on
Cr 19.9448 g
calibration certificate at k = 2
= 0.00015𝑔
0.0003
Cr100 22.0653 g 2 It is used twice in the model.
= 0.00015𝑔
0.0003
Cr1000 22.0569 g 2
= 0.00015𝑔
Absolute value for LOI values
Precision 1 0.0018098
between 0 to 5 % (See Table 4.8)
In the case of the determination of the flux compounds (Li2O(a), Li2O(b), B2O3(a),
B2O3(b)), for a specific flux, the values and uncertainties are similar. The three fluxes
in the test method have similar purity specifications of 99.98 %, and the purity value is
used to determine the concentration interval for the Li-tetra and -metaborate in the
flux. The atomic masses of B, Li and O and associated uncertainties are used to
determine the molecular masses and the associated standard uncertainties of Li2O
and B2O3, respectively (Table 4.11).
108
Chapter 4
Table 4.11: MzBP-1’s input quantity and associated standard uncertainty values
for measurement model 2.
Input Input
Standard uncertainty Comments
variable value
Mass of flux 10.0018 g 0.00015 g
Mass of
1.0014 g
sample 0.00015 g
50 − 50 × 0.9998
Li2B4O7 in
50 % √3 Purity = 99.98 %
flux
= 0.00577
LiBO2 in flux 50 % 0.00577 Purity = 99.98 %
10.821 − 10.806
Atomic mass
10.81g 2√3
of B
= 0.00433𝑔
6.997 − 6.938
Atomic mass Interval range and conventional
6.94g 2√3
of Li atomic weight from IUPAC[77]
= 0.00214𝑔
15.99977 − 15.99903
Atomic mass
15.999g 2√3
of O
= 0.00214𝑔
Molecular
mass of Li2O 28.9774g 0.03406g
(MLi2O)
Molecular
mass of B2O3 69.6202g 0.008683g
(MB2O3)
For measurement model 3, the uncertainty of the slope was calculated by dividing the
goodness of fit factor, K, with the number of used standards in the regression line,
while the uncertainty of the intercept is obtained from XGENLINE8.1. The precision of
the cup is obtained from experimental data using MzBP-1, while the precision ascribed
to the fluxer (Appendix I), and dilution (Appendix G) is from experimental data using
SARM-36. The standard uncertainty of the matrix corrected instrumental response is
obtained from SuperQ’s CSE reports. The final measurement model’s values and
associated uncertainties are obtained from the first three models and are described in
4.3.1.5.
109
Chapter 4
Table 4.12: MzBP-1’s input quantity and associated standard uncertainty values for measurement model 3. 𝐸𝑖 is the calibration
slope, Di is the intercept on the Y-axis, and P is the precision of measuring cup, dilution or fluxer.
u(Pcup) u(PFluxer) u(PFluxer)
Oxide MIi u(MI)i 𝐸𝑖 u(𝐸𝑖 ) Di u(Di) P u(PDilution)
Cup 5 TheOx FM
SiO2 28.6 0.032 0.1793 0.00084 0 0.00126 1 0.0001 0.0029 0.0058 0.0034
TiO2 3.01 0.017 0.0316 0.00013 0 0.00008 1 0.0003 0.0598 0.107 0.151
Al2O3 11.9 0.021 0.1463 0.00041 0 0.00193 1 0.0002 0.0063 0.0095 0.0087
Fe2O3 18.8 0.058 0.0226 0.00015 0 0.00009 1 0.0001 0.0074 0.0141 0.0261
MnO 0.459 0.010 0.0275 0.00010 0 0.00008 1 0.0009 0.233 0.205 0.419
MgO 1.66 0.008 0.0593 0.00048 0 0.00023 1 0.0010 0.0746 0.0808 0.146
CaO 16.1 0.040 0.0206 0.00038 0 0.00007 1 0.0002 0.0140 0.0140 0.0220
Na2O 6.54 0.010 0.0929 0.00035 0 0.00046 1 0.0004 0.0905 0.106 0.236
K2O 17.4 0.037 0.0196 0.00030 0 0.00010 1 0.0001 0.0042 0.0133 0.0179
P2O5 1.71 0.009 0.0318 0.00038 0 0.00012 1 0.0006 0.0530 0.0432 0.0807
Cr2O3 0.123 0.008 0.0135 0.00084 0 0.00005 1 0.0025 2.61 0.509 1.32
110
Chapter 4
4.3.1.3 The GUM framework: Correlation between the input quantities
Correlation close to one indicates high interdependence between the input variables
and must then be part of the propagation of the standard uncertainties (GUM step 6 in
Table 2.5). The correlation between the input variables in the four measurement
models was determined using Microsoft EXCEL. Correlation between some variables
is high in measurement models 1 and 4 (Tables 4.13 and 4.14), and the GUM step 6
is fully applied in the two models.
Table 4.13: Correlation between the three variables in measurement model 1. Cr is

the mass of the crucible, Cr100 is the mass of the crucible and sample
after drying at 100 °C, while Cr1000 is the mass of the crucible and
sample after roasting at 1000 °C.
Cr Cr100 Cr1000
Cr 1
Cr100 0.90556 1
Cr1000 0.94013 0.96958 1
Table 4.14: The correlation between the variables in measurement model 4. The
largest correlation is between Li2O and B2O3.
CSiO2, CTiO2, CAl2O3, CFe2O3, CMnO, CMgO, CCaO, CNa2O,

meas meas meas meas meas meas meas meas
Ci,meas 1 1 1 1 1 1 1 1
%Li2O 0.685 -0.005 0.000 0.060 0.223 0.358 0.497 0.013
%B2O3 0.686 0.004 -0.001 -0.060 -0.223 -0.357 -0.496 -0.013
LOI -0.537 0.037 0.102 -0.088 0.025 0.402 0.515 -0.325
Total 0.246 0.005 0.062 -0.005 -0.316 -0.085 -0.171 -0.122
CK2O, CP2O5, CCr2O3,

%Li2O %B2O3 LOI Total
meas meas meas
Ci,meas 1 1 1
%Li2O -0.117 0.082 0.044 1
%B2O3 0.117 -0.082 -0.043 -1.000 1
LOI -0.268 -0.020 -0.134 0.630 -0.630 1
Total 0.122 -0.179 0.215 -0.417 0.417 -0.154 1
111
Chapter 4
4.3.1.4 The GUM framework: Sensitivity coefficients of the input quantities
Sensitivity coefficients of the input quantities are determined by the partial

differentiation of the output quantity with respect to the input quantity. Table 4.15 lists
the formula for each sensitivity coefficient in its measurement model and the calculated
values for MzBP-1 using an EXCEL template described in 4.3.1.6. The calculated
values are compared with those calculated by the NIST uncertainty machine for
verification. Both values are similar, indicating the correctness of the developed
EXCEL template.
Table 4.15: The formulas to calculate the sensitivity coefficients for an input variable
in its measurement model. The corresponding formula calculates the
value of each sensitivity coefficient in an EXCEL template. The
calculated sensitivity coefficients are for SiO2 as calculated by the
EXCEL template and the NIST uncertainty machine for MzBP-1.
Sensitivity
Measur
Sensitivity coefficient
ement
coefficient calculated
model Input
Formula for a partial derivative calculated by NIST
and variable
with uncer-
output
EXCEL tainty
variable
machine
𝜕𝐿𝑂𝐼 (𝐶𝑟100 − 𝐶𝑟1000)
Cr = × 𝑃 × 100 0.187 0.19
𝜕𝐶𝑟 (𝐶𝑟100 − 𝐶𝑟)2
𝜕𝐿𝑂𝐼 (𝐶𝑟1000 − 𝐶𝑟)
Cr100 = × 𝑃 × 100 46.97 47.00
1 𝜕𝐶𝑟100 (𝐶𝑟100 − 𝐶𝑟)2
LOI 𝜕𝐿𝑂𝐼 −1
Cr1000 = × 𝑃 × 100 -47.159 -47.00
𝜕𝐶𝑟1000 (𝐶𝑟100 − 𝐶𝑟)
𝜕𝐿𝑂𝐼 (𝐶𝑟100 − 𝐶𝑟1000)
P =( × 100) 0.396 0.40
𝜕𝑃 (𝐶𝑟100 − 𝐶𝑟)
𝜕(𝐿𝑖2 𝑂(𝑎)) 2 × 𝑀𝐵2 𝑂3
Li2O = 2 × %𝐿𝑖2 𝐵4 𝑂7 0.246 0.25
𝜕(𝑀𝐿𝑖2 𝑂 ) (𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 ))
2 𝜕(𝐿𝑖2 𝑂(𝑎)) −2 × 𝑀𝐿𝑖2 𝑂
= × %𝐿𝑖2 𝐵4 𝑂7
Li2O(a) B2O3 𝜕(𝑀𝐵2 𝑂3 ) 2 -0.102 -0.10
(𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 ))
𝜕(𝐿𝑖2 𝑂(𝑎)) 𝑀𝐿𝑖2 𝑂
Li2B4O7 = 0.172 0.17
𝜕(%𝐿𝑖2 𝐵4 𝑂7 ) 𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 )
112
Chapter 4

Measur Sensitivity
ement Sensitivity coefficient
model Input coefficient calculated
and variable
with uncer-
output EXCEL tainty
variable machine
𝜕(𝐿𝑖2 𝑂(𝑏)) (0.25 × 𝑀𝐵2 𝑂3 )
=
Li2O 𝜕(𝑀𝐿𝑖2 𝑂 ) [(0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )]
2 0.358 0.36
× %𝐿𝑖𝐵𝑂2
2 𝜕(𝐿𝑖2 𝑂(𝑏)) −0.25 × 𝑀𝐿𝑖2 𝑂
=
Li2O(b) B2O3 𝜕(𝑀𝐵2 𝑂3 ) [(0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )]
2
-0.149 -0.15
𝜕(𝐿𝑖2 𝑂(𝑏)) 0.5 × 𝑀𝐿𝑖2 𝑂
LiBO2 = 0.294 0.29
𝜕(%𝐿𝑖𝐵𝑂2 ) (0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )
𝜕(𝐵2 𝑂3 (𝑎)) −2 × 𝑀𝐵2 𝑂3
Li2O = 2 × %𝐿𝑖2 𝐵4 𝑂7 -0.246 -0.25
𝜕(𝑀𝐿𝑖2 𝑂 ) [𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 )]
2
𝜕(𝐵2 𝑂3 (𝑎)) 2 × 𝑀𝐿𝑖2 𝑂
B2O3(a) B2O3 = 2 × %𝐿𝑖2 𝐵4 𝑂7 0.102 0.10
𝜕(𝑀𝐵2 𝑂3 ) [𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 )]
𝜕(𝐵2 𝑂3 ) 2 × 𝑀𝐵2 𝑂3
Li2B4O7 = 0.828 0.83
𝜕(%𝐿𝑖2 𝐵4 𝑂7 ) 𝑀𝐿𝑖2 𝑂 + (2 × 𝑀𝐵2 𝑂3 )
𝜕(𝐵2 𝑂3 (𝑏)) −0.25 × 𝑀𝐵2 𝑂3
=
Li2O 𝜕(𝑀𝐿𝑖2 𝑂 ) [(0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )]
2
-0.358 -0.36
2 𝜕(𝐵2 𝑂3 (𝑏)) 0.25 × 𝑀𝐿𝑖2 𝑂
=
B2O3(b) B2O3 𝜕(𝑀𝐵2 𝑂3 ) [(0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )]
2
0.149 0.15
𝜕(𝐵2 𝑂3 (𝑏)) 0.5 × 𝑀𝐵2 𝑂3
LiBO2 = 0.706 0.71
𝜕(%𝐿𝑖𝐵𝑂2 ) (0.5 × 𝑀𝐿𝑖2 𝑂 ) + (0.5 × 𝑀𝐵2 𝑂3 )
2
𝜕%𝐿𝑖2 𝑂 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥
Li2O(a) = 0.909 0.91
𝜕𝐿𝑖2 𝑂(𝑎) 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒
𝜕%𝐿𝑖2 𝑂 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥
Li2O(b) = 0.909 0.91
2 𝜕𝐿𝑖2 𝑂(𝑏) 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒
Li2O 𝜕%𝐿𝑖2 𝑂 𝑚𝑠𝑎𝑚𝑝𝑙𝑒 ∙ (𝐿𝑖2 𝑂(𝑎) + 𝐿𝑖2 𝑂 (𝑏))
Massflux = 0.193 0.19
𝜕𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 (𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 )2
masssam 𝜕%𝐿𝑖2 𝑂 −𝑚𝑓𝑙𝑢𝑥 ∙ (𝐿𝑖2 𝑂(𝑎) + 𝐿𝑖2 𝑂 (𝑏))
= -1.925 -1.90
ple 𝜕𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 (𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 )2
113
Chapter 4

Measur Sensitivity
ement Sensitivity coefficient
model Input coefficient calculated
and variable
with uncer-
output EXCEL tainty
variable machine
𝜕%𝐵2 𝑂3 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥
B2O3(a) = 0.909 0.91
𝜕𝐵2 𝑂3 (𝑎) 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒
𝜕%𝐵2 𝑂3 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥
B2O3(b) = 0.909 0.91
2 𝜕𝐵2 𝑂3 (𝑏) 𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒
B2O3 𝜕%𝐵2 𝑂3 𝑚𝑠𝑎𝑚𝑝𝑙𝑒 ∙ (𝐵2 𝑂3 (𝑎) + 𝐵2 𝑂3 (𝑏))
Massflux = 0.634 0.63
𝜕𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 (𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 )2
masssam 𝜕%𝐵2 𝑂3 −𝑚𝑓𝑙𝑢𝑥 ∙ (𝐵2 𝑂3 (𝑎) + 𝐵2 𝑂3 (𝑏))
= -6.336 -6.30
ple 𝜕𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 (𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥 + 𝑚𝑎𝑠𝑠𝑠𝑎𝑚𝑝𝑙𝑒 )2
𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 𝑃 ∙ (𝐷𝑖 −𝑀𝐼𝑖 )
Ei = 28.644 29.00
𝜕𝐸𝑖 (𝐸𝑖 ∙ 𝑃)2
𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 1
MIi = 0.179 0.18
𝜕𝑀𝐼𝑖 𝐸𝑖 ∙ 𝑃
3 𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 𝐸𝑖 ∙ (𝐷𝑖 −𝑀𝐼𝑖 )
PCup = 5.135 5.10
Ci,meas 𝜕𝑃 (𝐸𝑖 ∙ 𝑃)2
(i = 𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 𝐸𝑖 ∙ (𝐷𝑖 −𝑀𝐼𝑖 )
PFluxer = 5.135 5.10
SiO2) 𝜕𝑃 (𝐸𝑖 ∙ 𝑃)2
𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 𝐸𝑖 ∙ (𝐷𝑖 −𝑀𝐼𝑖 )
PDilution = 5.135 5.10
𝜕𝑃 (𝐸𝑖 ∙ 𝑃)2
𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 −1
Di = 1 1.00
𝜕𝐷𝑖 𝐸𝑖 ∙ 𝑃
𝜕𝐶𝑖,𝑢𝑛𝑘 𝑇𝑜𝑡𝑎𝑙 − 𝐿𝑂𝐼𝑚𝑒𝑎𝑠
Ci,meas = 11.23 11.00
𝜕𝐶𝑖,𝑚𝑒𝑎𝑠 𝑇𝑜𝑡𝑎𝑙 − (𝐿𝑖2 𝑂 + 𝐵2 𝑂3 )
𝜕𝐶𝑖,𝑢𝑛𝑘 𝐶𝑖,𝑚𝑒𝑎𝑠 × (𝐿𝑂𝐼𝑚𝑒𝑎𝑠 − (𝐿𝑖2 𝑂 + 𝐵2 𝑂3 ))
Total = 2 -5.94 -5.90
4 𝜕𝑇𝑜𝑡𝑎𝑙 (𝑇𝑜𝑡𝑎𝑙 − (𝐿𝑖 𝑂 + 𝐵 𝑂 ))
2 2 3
Ci,unk 𝜕𝐶𝑖,𝑢𝑛𝑘 −𝐶𝑖,𝑚𝑒𝑎𝑠
LOI = -0.58 -0.58
(i = 𝜕𝐿𝑂𝐼 (𝑇𝑜𝑡𝑎𝑙 − (𝐿𝑖2 𝑂 + 𝐵2 𝑂3 ))
SiO2) 𝜕𝐶𝑖,𝑢𝑛𝑘 𝐶𝑖,𝑚𝑒𝑎𝑠 (𝑇𝑜𝑡𝑎𝑙 − 𝐿𝑂𝐼𝑚𝑒𝑎𝑠 )
Li2O = 2 6.52 6.50
𝜕𝐿𝑖2 𝑂 (𝑇𝑜𝑡𝑎𝑙 − (𝐿𝑖 𝑂 + 𝐵 𝑂 ))
2 2 3
𝜕𝐶𝑖,𝑢𝑛𝑘 𝐶𝑖,𝑚𝑒𝑎𝑠 (𝑇𝑜𝑡𝑎𝑙 − 𝐿𝑂𝐼𝑚𝑒𝑎𝑠 )
B2O3 = 6.52 6.50
𝜕𝐵2 𝑂3 (𝑇𝑜𝑡𝑎𝑙 − (𝐿𝑖 𝑂 + 𝐵 𝑂 ))2
2 2 3
4.3.1.5 The GUM framework: Propagation of the standard uncertainty values
As noted in Chapter 2, the propagation of the standard uncertainties, sensitivity and

correlation coefficients are obtained with Equation 4.2, and the last term in this
114
Chapter 4
equation is used in cases of covariance between variables in a measurement model.

In the cases of Measurement models 1 and 4, it is applied.
𝑢𝑐2 (𝑦) = ∑𝑁 2 2 𝑁−1 𝑁

𝑖=1 𝑐𝑖 𝑢 (𝑥𝑖 ) + 2 ∑𝑖−1 ∑𝑗=𝑖+1 𝑐𝑖 𝑐𝑗 𝑢(𝑥𝑖 )(𝑥𝑗 )𝑟(𝑥𝑖 , 𝑥𝑗 ) (Eq 4.2)
Appendix F lists the input and calculated values and associated uncertainties in the
GUM framework for the four test samples SARM-36, MzBP-1, Li3 6c and
G00006619001 as calculated by the GUM framework in an EXCEL template described
in 4.3.1.6. Values for measurement model 1 are listed in Table K1 and contribute to
the used LOI concentration and uncertainty values in measurement model 4.
Measurement model 2 values for the flux components are detailed in Table K2, while
Tables K3 to K13 list the uncertainty propagation results for Measurement model 3,
where the eleven major oxides (Ci,meas) are calculated from the calibration regression
plots in SuperQ. The flux components are removed in the fourth and last measurement
model, and the measured LOI is included, and their values are reported in
Appendix K (Tables K14 to K24). The total uncertainty of the reported results’ sum is
calculated by propagating the uncertainties of the major oxides and LOI using
Equation 4.3.
2
𝑢(𝑇𝑜𝑡𝑎𝑙) = √∑ 𝑢(𝐶𝑖,𝑢𝑛𝑘 ) + 𝑢(𝐿𝑂𝐼)2 Eq. 4.3
4.3.1.6 The GUM framework: EXCEL template
An EXCEL workbook, GUM framework.xlsx, was developed to propagate the

uncertainty values of the input quantities in the four different measurement models
described previously in 4.3.1.1.1 to 4.3.1.1.4. This template workbook evaluates more
than one sample with a worksheet set up for each measurement model. Measurement
model 1 for the determination of LOI in worksheet LOI is depicted in Figure 4.9.
115
Chapter 4
Figure 4.9: Measurement model 1 in worksheet LOI of the EXCEL workbook,

‘GUM framework.xlsx’. The Cr, Cr100 and Cr1000 are the input
variables, u(Cr), u(Cr100) and u(Cr1000) the standard uncertainties
while c(Cr), c(Cr100) and c(Cr1000) are the sensitivity coefficients.
The correlation coefficients are determined using the Correlation
function in the Data Toolpak in EXCEL for Microsoft 365 and the
results listed in cells J28 to L30.
For measurement model 2, two worksheets, ‘Flux calculator’ and ‘Flux in sample disk’,
are used. In the first worksheet (Fig. 4.10), the standard uncertainties for Li2O(a),
Li2O(b), B2O3(a) and B2O3(b) per flux type are calculated. The second worksheet (Fig.
4.11) uses the sample and flux masses to determine the Li 2O and B2O3 values and
associated uncertainties in the fusion disks for analysis by WD-XRF.
116
Chapter 4
Figure 4.10: The calculation of Li2O(a), Li2O(b) (top left boxes), B2O3(a) and B2O3(b) (bottom left boxes) values and uncertainties
for the B flux. The calculated result and associated uncertainty for Li2O are in cells O54 and P54, respectively, while
those for B2O3 are in cells O66 and P66. The template is applied with applicable Li 2B4O7 and LiBO2 values for the
A and C fluxes.
117
Chapter 4
Figure 4.11: The Measurement model 2 in the ‘Flux in sample disk’ worksheet. The top three rows contain the values calculated
in the ‘Flux calculator’-worksheet for specific flux types. The sample (T-column) and flux (P-column) masses and
the flux components based on the selected flux determine the Li2O-values in the Y-column and the B2O3-values in
the AA-column. The input values’ calculated uncertainties are listed next to the input values’ columns.
118
Chapter 4
The ‘Conc from SuperQ’-worksheet processes the background and line-overlap

corrected counts with associated uncertainties from previously described precision
into concentration values (Fig 4.12).
The last worksheet, ‘Final conc’, is for Measurement model 4, where the major oxide
results and uncertainties are calculated after including LOI and excluding Li 2O and
B2O3 (Fig. 4.13 for SiO2). The rest of the processing of the major elements’ results in
this worksheet is in Appendix M (Figs. M1 to M11).
119
Chapter 4
Figure 4.12: Measurement model 3 for the concentrations from SuperQ in the ‘Conc from SuperQ’-worksheet. Information
from SuperQ such as the slope, its uncertainty calculated from the slope’s K-factor and the number of used
standards are listed in the top part of the worksheet (cells E10 to Y20). At the bottom, the columns for the
matrix corrected counts (MIi), precision (P) and their uncertainty and sensitivity coefficients are organized
for SiO2. The same setup for the other oxides is right in the worksheet and is detailed in Appendix L.
120
Chapter 4
Figure 4.13: The last worksheet, ‘Final conc’, in the EXCEL workbook, ‘GUM framework.xlsx’, is for measurement model 4. In
the top from cells F27 to K32, the correlation between C SiO2,meas, %Li2O, %B2O3, LOI and Total from Measurement
model 3 are calculated and used for the sensitivity coefficients. The values, uncertainties and sensitivity coefficients
for the five variables in measurement model 4 for SiO2 are listed in cells B37 to S41. The same setup for the rest of
the oxides is from columns V in the worksheet and is detailed in Appendix M. The input values such as Ci,meas and
u(Ci,meas) refer to corresponding cells in the ‘Conc from SuperQ’-worksheet to prevent errors from copying and
pasting of cell values. The final results and associated uncertainties are reported in rows 46 to 50.
121
Chapter 4
4.3.1.7 The Monte Carlo Method
A spreadsheet per test sample is used to estimate the uncertainties of the results using
MCM. The four measurement models’ processes are placed into columns with the
100 000 permutations of the measurement model in rows from 13 to 100 013. Table
4.16 lists the four measurement models’ positions in the spreadsheet with associated
figures for illustration.
Table 4.16: Setup of the four measurement models in the EXCEL template for the
MCM approach. The columns containing the specific measurement
model are illustrated in the referred figures. Figures for Model 3 are in
Appendix N, while Model 4’s figures are in Appendix O.
Measurement model Columns in MCM EXCEL template
Model 1 (Determination of LOI) A to I (Fig. 4.14)
L to U for Li2O (Fig. 4.16)
Model 2 (Determination of flux components)
W to AF for B2O3 (Fig. 4.18)
AJ to AS for SiO2 (Fig. N1)
AW to BF for TiO2 (Fig. N2)
BH to BQ for Al2O3 (Fig. N3)
BS to CB for Fe2O3 (Fig. N4)
CD to CM for MnO (Fig. N5)
Model 3 (Major elements from SuperQ) CO to CX for MgO (Fig. N6)
CZ to DI for CaO (Fig. N7)
DK to DT for Na2O (Fig. N8)
DV to EE for K2O (Fig. N9)
EG to EP for P2O5 (Fig. N10)
ER to FA for Cr2O3 (Fig. N11)
122
Chapter 4

Measurement model Columns in MCM EXCEL template
FE to FV for SiO2 (Fig. O1)
FX to GO for TiO2 (Fig. O2)
GQ to HH for Al2O3 (Fig. O3)
HJ to IA for Fe2O3 (Fig. O4)
IC to IT for MnO (Fig. O5)
Model 4 (Major elements without flux IV to JM for MgO (Fig. O6)
components and with LOI) JO to KF for CaO (Fig. O7)
KH to KY for Na2O (Fig. O8)
LA to LR for K2O (Fig. O9)
LT to MK for P2O5 (Fig. O10)
MM to ND for Cr2O3 (Fig. O11)
NG for Total (Fig. O11)
123
Chapter 4
Figure 4.14: The measurement model 1 in the MCM’s EXCEL spreadsheet

template for the test sample, MzBP-1 (workbook ‘MCM MzBP-1.xlsx’).
The input quantities and associated standard uncertainties are in cells
C4 to G5. In rows 13 to 100 013 and columns C to G, permutations of
the measurement model 1 are calculated with the corresponding LOI-
values in column I. The standard deviation of the 100 000 LOI values
in cell I5 (0.010 %) is then the uncertainty of the LOI-value of 0.396%.
The results of the 100 000 permutations are plotted to illustrate the normal distribution
of the calculated LOI-uncertainty using the Histogram function in EXCEL’s Data
Analysis Toolpak, where the results are grouped into 317 bins (Figure 4.15).
124
Chapter 4
Figure 4.15: A plot of the frequency of MzBP-1’s LOI results from the MCM’s
permutations vs concentration (%). The black line is the LOI value of
0.396%, with the green lines being the uncertainty of 0.010 % and
the red lines the expanded uncertainty of 0.020 % at a confidence
level of 95.45 %.
The flux components’ uncertainties determined by the MCM approach (Figs. 4.16 and
4.18) involve four rectangular and two normal distribution types leading to somewhat
a rectangular frequency distribution plots (Figs. 4.17 and 4.19).
125
Chapter 4
Figure 4.16: Calculation of Li2O’s uncertainty in sample MzBP-1 with the MCM
approach. The half-width interval, a, is calculated by multiplying the
standard uncertainty with the square root of 3. The standard deviation
of the results (0.019%) from the permutations in column U is in cell U5.
126
Chapter 4
Figure 4.17: The frequency plot of the 100 000 permutations for Li 2O in sample
MzBP-1. The black line is the Li2O value of 21.19 % (21.187%), with the
green lines being the uncertainty of 0.01 9% and the red lines the
expanded uncertainty of 0.038% at a confidence level of 95.45%.
127
Chapter 4
Figure 4.18: B2O3’s uncertainty in sample MzBP-1 as calculated by the MCM

method in the workbook, MCM MzBP-1.xlsx. The standard deviation of
the results (0.020 %) from the permutations is in cell AF5.
128
Chapter 4
Figure 4.19: The frequency plot of the 100 000 permutations for B 2O3 in sample
MzBP-1 reflecting the combination of the two frequency distribution
types of the input variables. The black line represents the B2O3 value
of 69.71%, with the green lines being the uncertainty of 0.02% and the
red lines the expanded uncertainty of 0.04% at a confidence level of
95.45%.
As stated in Table 4.15, the results and associated uncertainties from the
measurement models 3 and 4 are depicted in Appendixes N and O. The frequency
distribution of the major elements in both Measurement models 3 and 4 reflect their
normal probability distribution types (Figs. 4.20 (a) and (b) for SiO2). The distribution
plots of the other final major element results for test material MzBP-1 are illustrated in
Appendix P. The sum of the major elements and LOI is also permutated to obtain its
uncertainty (Fig. P11).
129
Chapter 4
(a)
(b)
Figure 4.20: Frequency plots of the permutations for SiO2 as determined by

measurement models 3 (in (a)) and 4 (in (b)) for test sample MzBP-1.
The final SiO2 result is 57.687%, with an expanded uncertainty of
0.292 % at the confidence level of 95.45%.
130
Chapter 4
4.3.1.8 The Kragten method
For the Kragten method, a single spreadsheet per test sample is used with the four
measurement models grouped into rows (Table 4.17). The EXCEL template is based
on the Kragten example in the Eurachem CITAC Guide CG4 [46]. The standard
uncertainty values for each input values are similar to those used in the GUM
framework and MCM EXCEL templates. Figures 4.19 to 4.23 illustrate the Kragten
method per measurement model for test sample MzBP-1 in an EXCEL workbook
(Kragten method.xlsx) developed for this study, while all the major elements’
processes for measurement models 3 and 4 in the same worksheet are illustrated in
Appendix Q.
Table 4.17: Arrangement of the four measurement models in the EXCEL template
for the Kragten method.
Measurement model Rows
Model 1 4 to 23
Model 2 32 to 79
Model 3 85 to 417
Model 4 428 to 735
The uncertainty budget is included with each uncertainty calculation, while the final
results are reported from rows 743 to 757 (Fig. 4.25). The four test samples’
uncertainty budgets are described in 4.3.4.1.
131
Chapter 4
Figure 4.21: The Kragten method to calculate the LOI-value and uncertainty in
measurement model 1 for test sample MzBP-1. It is from rows 4 to 23
in a worksheet allocated to this test sample in an EXCEL workbook in
this study (Kragten method.xlsx). The uncertainty budget for this model
is in row 23, while the uncertainty value of LOI is in cell E20.
Figure 4.22: The Kragten method to calculate the Li2O-value and uncertainty in
measurement model 2 for test sample MzBP-1. It is from rows 32 to 54
in the same worksheet in Fig. 4.21. The uncertainty budget for this
model is in row 54, while the uncertainty value of Li2O is in cell E51.
132
Chapter 4
Figure 4.23: The B2O3-value and uncertainty in measurement model 2 for test
sample MzBP-1 determined by the Kragten method. It is from rows 57
to 79 in the same worksheet in Fig. 4.21. The uncertainty budget for
this model is in row 79, while the uncertainty value of B 2O3 is in cell
E76.
Figure 4.24: The SiO2-value and uncertainty in measurement model 3 for test
sample MzBP-1 determined by the Kragten method. It is from rows 85
to 113 in the same worksheet in Fig. 4.21. The uncertainty budget for
this model is in row 113, while the uncertainty value of SiO 2 is in cell
C110. The contributions to the uncertainty budget are plotted in a bar
graph which illustrates that the calibration slope is the main contributor
to SiO2’s uncertainty for test sample MzBP-1.
133
Chapter 4
Figure 4.25: The concentration values and uncertainties for test sample MzBP-1
determined by the Kragten method. It is from rows 743 to 757 in the
same worksheet in Fig. 4.21.
4.3.2 The top-down approach
An EXCEL workbook based on the Nordtest approach[6] was developed for the study
(Top down method.xlsx). The method varies depending on whether CRM or PT results
are used with QC results obtained from the in-house standard analysed with the test
samples. The QC data is listed in Table 4.18 and is used to quantify the QC part of
the test results.
134
Chapter 4
Table 4.18: Statistical data used in the Nordtest based EXCEL template
developed for the study.
% Relative
Standard
Average standard
deviation
deviation
SiO2 45.624 0.417 0.915
TiO2 1.530 0.018 1.195
Al2O3 16.640 0.246 1.479
Fe2O3 9.833 0.108 1.098
MnO 0.179 0.003 1.598
MgO 7.989 0.097 1.215
CaO 10.716 0.153 1.429
Na2O 3.591 0.099 2.766
K2O 0.702 0.017 2.360
P2O5 0.266 0.005 1.885
Cr2O3 0.077 0.004 4.869
LOI 2.696 0.099 3.662
4.3.2.1 CRM results
Five CRMs prepared for fusion disk standards for the study (1252, 1253, 1254, 1255,
1268, 1269, 1270, 1271, 1318, 1319 and 1356) were used to determine the test
method’s bias. Their uncertainty values are available in Appendix E and contribute to
the determination of the Relative u(Cref), Bias and sbias illustrated in Figure 4.26.
Relative data is used for all results as the absolute value of the bias’ contribution is too
high for the low test result value as suggested by the Nordtest method[6], leading to
unrealistic uncertainty values.
135
Chapter 4
Figure 4.26: Quantification of CRM data for the top-down approach in the EXCEL
template for this approach.
4.3.2.2 PT results
Eighteen GeoPT rounds were used to obtain the PT data for the top-down approach.
In Figure 4.27, the SiO2 results are organised and summarised.
Figure 4.27: SiO2 results of eighteen GeoPT rounds to obtain the bias data for the
test method. This table is a screenshot of the EXCEL template for the
top-down approach.
136
Chapter 4
4.3.2.3 EXCEL template for the top-down approach
In the EXCEL workbook for the top-down approach, the u(bias) from previous PT
rounds are summarised in a spreadsheet (Nordtest u(bias) from PT rounds) while the
u(bias) from CRM data is organised in the ‘Bias from CRMs’-worksheet. In the QC
Results-worksheet, the in-house standard’s QC data is referenced from the worksheet
templates using PT data (Fig. 4.28) and CRM data (Fig. 4.30).
Figure 4.28: The EXCEL worksheet for the top-down approach using PT data. The
uncertainty values for the test sample, SARM-36, is calculated in rows
from 43 to 59 (Fig. 4.29).
137
Chapter 4
Figure 4.29: Uncertainty values for the test sample, SARM-36, as calculated in the
EXCEL worksheet in Fig. 4.28
Figure 4.30: The EXCEL worksheet for the top-down approach using CRM data.
The uncertainty values for the test sample, SARM-36, is calculated in
rows from 43 to 59. The bias data from CRM are in rows 16 to 18.
138
Chapter 4
4.3.3 Software tools
The results from two software applications, XGENLINE v 8.1[78] and the NIST
uncertainty machine[48], are summarised as both were used in the study to evaluate
the uncertainty of measurement.
4.3.3.1 XGENLINE
In XGENLINE v8.1[78], the uncertainties of both the instrumental responses and the
CRMs were entered into the software. The properties of each major element’s
regression line are summarised in Table 4.19. As noted previously, SuperQ plots the
instrumental response on the X-axis in contrast with XGENLINE and the inverse of
XGENLINE’s regression slopes were calculated and are listed in Table 4.19. As both
slope values compare favourably, the uncertainty values of the intercept on the X-axis
are used in the estimation of uncertainty for the bottom-up approach. Furthermore, the
values of the co-variance between the slope and intercept are also reported by
XGENLINE, enabling the decision not to use the second part of the GUM equation
(Eq. 4.2) for measurement model 3. The study test materials’ concentrations from
XGENLINE are similar to those from SuperQ, and the results are listed in Appendix R.
139
Chapter 4
Table 4.19: Regression line properties for the major elements by XGENLINE v8.1. ‘1/slope’ is the inverse of XGENLINE’s slope
and is compared with the regression slopes from SuperQ.
Co-variance
Intercept X- u(Intercept between SuperQ
Slope u(slope) Intercept u(Intercept) 1/slope
axis X-axis) slope and slope
intercept
SiO2 5.58 0.00168 -0.287 0.00471 0.00515 0.000842 -0.000004 0.179 0.179
TiO2 31.6 0.0592 0.0111 0.00106 -0.00035 0.000034 -0.000015 0.0316 0.0316
Al2O3 6.83 0.00360 0.0121 0.00161 -0.00177 0.000237 -0.000002 0.146 0.146
Fe2O3 44.1 0.0381 0.0603 0.00122 -0.00137 0.000028 -0.000007 0.0227 0.0226
MnO 38.3 0.166 0.00447 0.00051 -0.00012 0.000001 -0.000014 0.0261 0.0275
MgO 17.3 0.0230 -0.00824 0.00189 0.00477 0.000108 -0.000011 0.0579 0.0593
CaO 48.1 0.0377 -0.00327 0.00150 0.00007 0.000031 -0.000006 0.0208 0.0206
Na2O 10.7 0.0170 0.0105 0.00119 -0.00099 0.000112 -0.000005 0.0938 0.0929
K2O 51.0 0.0402 0.0665 0.00207 -0.00130 0.000041 -0.000014 0.0196 0.0196
P2O5 31.3 0.0893 -0.00554 0.00201 0.00018 0.000064 -0.000054 0.0319 0.0318
Cr2O3 73.7 0.0670 -0.0814 0.00214 0.00110 0.000029 -0.000018 0.0136 0.0135
140
Chapter 4
4.3.3.2 NIST uncertainty machine
The MCM and GUM concentration and uncertainty values produced by the NIST
uncertainty machine web application[48] are similar to those calculated by the EXCEL
templates for the GUM framework described in 4.3.1. The uncertainty values are listed
in Table 8 with the uncertainty values from other approaches.
4.3.4 Comparison of uncertainties from the bottom-up and top-down

approaches
Tables 4.20 to 4.23 compare the uncertainties calculated by the bottom-up and top-
down approaches, the NIST uncertainty machine, and XGENLINE. As the
measurement results from the different EXCEL templates and the NIST uncertainty
machine are similar, only the measurement results calculated by the GUM framework
are listed in the first column of the four tables.
Table 4.20: Results (%) of test material SARM-36 C1 and uncertainties as evaluated
by the different approaches.
Top-down Software used in
Bottom-up approach
approach study
Major
Result GUM NIST
element Kragten XGENLI
frame- MCM PT CRM uncertainty
method NE
work machine
SiO2 64.084 0.353 0.193 0.195 1.00 1.92 0.354 0.192
TiO2 0.0471 0.0016 0.0016 0.0017 0.0009 0.0044 0.0016 0.0015
Al2O3 17.213 0.088 0.080 0.080 0.503 0.418 0.088 0.057
Fe2O3 1.418 0.013 0.013 0.013 0.031 0.076 0.013 0.008
MnO 0.0087 0.0020 0.0019 0.0020 0.0005 0.0008 0.0020 0.0018
MgO 0.404 0.006 0.006 0.006 0.013 0.009 0.006 0.004
CaO 0.6785 0.0137 0.0132 0.0132 0.0272 0.0256 0.0137 0.0043
Na2O 0.418 0.006 0.006 0.006 0.017 0.050 0.006 0.003
K2O 15.664 0.238 0.208 0.210 0.599 1.52 0.238 0.049
P2O5 0.119 0.003 0.003 0.003 0.011 0.010 0.003 0.002
Cr2O3 0.0098 0.0013 0.0013 0.0026 4.78E-04 0.0013 0.0013 0.0011
LOI 0.192 0.0005 0.002 0.002 0.007 0.013 0.0005
Total 100.256 0.436 0.296 0.297 0.436
141
Chapter 4
Table 4.21: Results (%) of test material MzBP-1 and uncertainties as evaluated by
the different approaches.
Software used in
Bottom-up approach Top-down approach
study
Major
Result GUM NIST
element Kragten XGENLIN
method E
work machine
SiO2 57.687 0.303 0.147 0.148 0.90 1.73 0.304 0.190
TiO2 1.069 0.008 0.008 0.014 0.021 0.099 0.008 0.007
Al2O3 19.543 0.092 0.077 0.078 0.572 0.474 0.092 0.070
Fe2O3 4.780 0.038 0.035 0.035 0.106 0.256 0.038 0.021
MnO 0.1416 0.0034 0.0033 0.0032 0.0078 0.0129 0.0034 0.0031
MgO 1.104 0.012 0.011 0.011 0.036 0.024 0.012 0.007
CaO 3.731 0.072 0.068 0.067 0.150 0.141 0.072 0.015
Na2O 6.819 0.041 0.033 0.034 0.283 0.815 0.041 0.025
K2O 3.842 0.058 0.058 0.058 0.147 0.372 0.058 0.015
P2O5 0.611 0.009 0.008 0.008 0.056 0.050 0.009 0.004
Cr2O3 0.0186 0.0018 0.0018 0.0018 9.1E-04 0.0025 0.0018 0.0013
LOI 0.396 0.002 0.010 0.010 0.015 0.027 0.002
Total 99.743 0.338 0.196 0.198 0.335
Table 4.22: Results (%) of test material Li3 6c and uncertainties as evaluated by the
different approaches.
Software used in
study
Major
Result GUM NIST
method E
work machine
SiO2 0.197 0.009 0.009 0.009 0.003 0.006 0.009 0.006
TiO2 0.0475 0.0011 0.0011 0.0014 0.0010 0.0044 0.0011 0.0008
Al2O3 0.043 0.012 0.012 0.012 0.001 0.001 0.012 0.003
Fe2O3 0.4654 0.0078 0.0075 0.0076 0.0104 0.0249 0.0078 0.0032
MnO 0.1326 0.0029 0.0025 0.0025 0.0073 0.0121 0.0029 0.0017
MgO 5.248 0.083 0.082 0.084 0.173 0.114 0.083 0.021
CaO 49.482 1.14 0.253 0.243 1.99 1.87 1.14 0.192
Na2O <0.02
K2O <0.002
P2O5 <0.01
Cr2O3 0.0136 0.0010 0.0010 0.0010 0.0008 0.0018 0.0010 0.0006
LOI 44.556 0.271 0.090 0.036 1.70 3.02 0.271
Total 100.174 1.255 0.281 0.261 1.17
142
Chapter 4
Table 4.23: Results (%) of test material G00006619001 and uncertainties as

estimated by the different approaches.
Software used in
study
Major
Result GUM NIST
method E
work machine
SiO2 46.228 0.244 0.160 0.160 0.720 1.39 0.245 0.137
TiO2 1.548 0.010 0.010 0.020 0.031 0.144 0.010 0.009
Al2O3 16.653 0.078 0.068 0.068 0.487 0.404 0.078 0.053
Fe2O3 9.839 0.074 0.067 0.066 0.219 0.527 0.074 0.033
MnO 0.1818 0.0035 0.0033 0.0033 0.0100 0.0166 0.0035 0.0031
MgO 7.895 0.073 0.065 0.065 0.261 0.172 0.073 0.026
CaO 10.503 0.202 0.176 0.177 0.422 0.396 0.205 0.035
Na2O 3.614 0.022 0.019 0.019 0.150 0.432 0.022 0.013
K2O 0.6937 0.0106 0.0111 0.0111 0.0266 0.0672 0.0106 0.0038
P2O5 0.264 0.004 0.004 0.004 0.024 0.022 0.004 0.002
Cr2O3 0.0749 0.0049 0.0050 0.0040 3.65E-03 0.0101 0.0049 0.0016
LOI 2.827 0.005 0.005 0.007 0.108 0.192 0.005
Total 100.320 0.351 0.266 0.267 0.344
The uncertainty values from the bottom-up approaches are mostly similar for TiO2,
Fe2O3, MnO, MgO, Na2O, K2O, P2O5 and Cr2O3. The GUM framework and NIST
uncertainty values are higher for SiO2 and CaO, while they are lower for LOI. The top-
down approach’s uncertainty values are the highest for SiO 2, CaO and LOI at high
concentration values.
4.3.4.1 The uncertainty budget
One of the Kragten method’s main advantages is its ability to determine the
contribution of each input value to the output value’s uncertainty. Tables S1 to 4.26
list the uncertainty budgets for the four measurement models per test sample and
identify the main contributors of uncertainty to the result. For measurement model 1
(Table S1), except for Li3 6c, the masses of the crucible with dried and roasted
samples (Cr100 and Cr1000) are the main contributors to LOI’s uncertainty. In the
case of Li3 6c, limestone with a high LOI value.
In the case of measurement model 2 (Tables S2 and S3), the molecular mass of Li 2O
is the main contributor to B2O3’s uncertainty for all four test samples. The uncertainty
values of the flux and sample masses are less than 0.5%. For measurement model 3,
143
Chapter 4
Table S4 lists the main contributors to uncertainty in the results from SuperQ: the CSE,
calibration slope, and intercept of the slope on the Y-axis. The influence of dilution’s
precision on the uncertainty of most elements is minimal, except for Na2O in MzBP-1.
The precision of the fluxer and cup does not contribute significantly to any uncertainty
in this measurement model.
Tables S5 to S15 list the uncertainty budgets for the final results from measurement
model 4. The main contributors to each final result are its corresponding result from
SuperQ and other present oxides with high concentrations (SiO 2 from SuperQ in all
the final results, and CaO in sample Li3 6c, K2O from SuperQ on all final results for
SARM-36). The effect of input quantities from other measurement models on the final
result was also evaluated using the GUM framework’s EXCEL template to determine
such contributions indirectly. The uncertainty of the main contributors, LOI, the flux
components, the corrected instrumental responses, the calibration’s slope and
intercept and precision are set to 0 at a time. Tables 4.24 to 4.34 list the uncertainty
budgets for SiO2 to Cr2O3 with the concentration in the second column. The samples
listed in the first column are: Sample 1 is SARM-36, Sample 2 is MzPP-1, Sample 3
Li3 6c and Sample 4 G00006619001. The third to eight columns list the following
uncertainty values of the result:
(a) The original calculated uncertainty.
(b) The LOI’s uncertainty is removed.
(c) Both the flux components’ uncertainty is removed (Flux).
(d) All the instrumental responses’ uncertainties are removed (MI).
(e) The uncertainty of the calibration slope is set to 0.
(f) The uncertainty of the intercept on the Y-axis is removed (Int.).
(g) All the precision components’ uncertainty values are removed (P).
The last five columns in the tables list the % contribution of the six contributors to the
concentration’s uncertainty of measurement.
144
Chapter 4
Table 4.24: Indirect determination of the uncertainty budget for SiO2 in the GUM framework EXCEL template for the four test
samples. The column headings are explained in the text.
u(SiO2) u(SiO2) u(SiO2) u(SiO2) u(SiO2) u(SiO2) u(SiO2) LOI Flux MI Slope Int. P
Sample SiO2
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 64.084 0.353 0.353 0.336 0.345 0.130 0.353 0.353 0.1 6.9 3.1 89.7 0.2 0.2
2 57.687 0.303 0.303 0.286 0.295 0.128 0.303 0.303 0.2 8.7 3.9 86.7 0.2 0.2
3 0.197 0.009 0.008 0.009 0.008 0.009 0.004 0.009 9.3 0.5 11.1 1.6 77.5 0.0
4 46.228 0.244 0.243 0.232 0.238 0.094 0.244 0.244 0.6 7.1 3.6 88.2 0.3 0.2
Table 4.25: Indirect determination of the uncertainty budget for TiO2 in the GUM framework EXCEL template for the four test
u(TiO2) u(TiO2) u(TiO2) u(TiO2) u(TiO2) u(TiO2) u(TiO2) LOI Flux MI Slope Int. P
Sample TiO2
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.0471 0.0016 0.0016 0.0016 0.0009 0.0016 0.0014 0.0016 0.0 0.2 71.4 1.8 26.6 0.0
2 1.0694 0.0083 0.0083 0.0080 0.0056 0.0065 0.0082 0.0083 0.0 5.8 56.0 36.9 1.2 0.1
3 0.0475 0.0011 0.0011 0.0011 0.0008 0.0009 0.0010 0.0011 0.5 0.5 47.4 32.4 18.9 0.4
4 1.5477 0.0104 0.0104 0.0101 0.0080 0.0072 0.0104 0.0104 0.1 5.7 39.7 53.6 0.7 0.2
Table 4.26: Indirect determination of the uncertainty budget for Al2O3 in the GUM framework EXCEL template for the four test
u(Al2O3) u(Al2O3) u(Al2O3) u(Al2O3) u(Al2O3) u(Al2O3) u(Al2O3) LOI Flux MI Slope Int. P
Sample Al2O3
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 17.213 0.088 0.088 0.084 0.082 0.049 0.086 0.088 0.0 8.8 11.6 75.0 4.4 0.2
2 19.543 0.092 0.092 0.084 0.085 0.057 0.090 0.092 0.1 14.7 13.8 66.4 4.8 0.2
3 0.043 0.012 0.012 0.012 0.012 0.012 0.003 0.012 0.2 0.0 3.1 0.2 96.5 0.0
4 16.653 0.078 0.078 0.073 0.072 0.045 0.075 0.078 0.3 10.7 11.6 71.7 5.5 0.3
145
Chapter 4
Table 4.27: Indirect determination of the uncertainty budget for Fe2O3 in the GUM framework EXCEL template for the four test
u(Fe2O3) u(Fe2O3) u(Fe2O3) u(Fe2O3) u(Fe2O3) u(Fe2O3) u(Fe2O3) LOI Flux MI Slope Int. P
Sample Fe2O3
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 1.418 0.013 0.013 0.013 0.011 0.008 0.013 0.013 0.0 2.8 31.0 65.7 0.5 0.0
2 4.780 0.038 0.038 0.037 0.035 0.018 0.038 0.038 0.0 4.9 13.0 81.9 0.1 0.0
3 0.4654 0.0078 0.0076 0.0078 0.0072 0.0038 0.0078 0.0078 3.4 0.9 11.7 82.9 0.4 0.7
4 9.839 0.074 0.074 0.072 0.071 0.027 0.074 0.074 0.0 3.6 5.5 90.7 0.1 0.1
Table 4.28: Indirect determination of the uncertainty budget for MnO in the GUM framework EXCEL template for the four test
u(MnO) u(MnO) u(MnO) u(MnO) u(MnO) u(MnO) u(MnO) LOI Flux MI Slope Int. P
Sample MnO
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.0087 0.0020 0.0020 0.0020 0.0008 0.0020 0.0018 0.0020 0.0 0.1 85.3 0.5 14.1 0.0
2 0.1416 0.0034 0.0034 0.0034 0.0012 0.0033 0.0033 0.0034 0.0 1.5 88.9 4.8 4.8 0.0
3 0.1326 0.0029 0.0029 0.0029 0.0021 0.0019 0.0029 0.0029 0.1 0.5 42.4 53.8 2.6 0.5
4 0.1818 0.0035 0.0035 0.0035 0.0014 0.0033 0.0034 0.0035 0.0 1.4 87.0 7.6 3.9 0.0
Table 4.29: Indirect determination of the uncertainty budget for MgO in the GUM framework EXCEL template for the four test
u(MgO) u(MgO) u(MgO) u(MgO) u(MgO) u(MgO) u(MgO) LOI Flux MI Slope Int. P
Sample MgO
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.404 0.006 0.006 0.006 0.005 0.005 0.005 0.006 0.0 1.0 43.0 39.0 17.0 0.0
2 1.104 0.012 0.012 0.011 0.010 0.007 0.011 0.012 0.1 3.2 19.3 73.0 4.4 0.0
3 5.248 0.083 0.088 0.083 0.082 0.026 0.083 0.082 7.2 0.6 1.6 89.4 0.0 1.2
4 7.895 0.073 0.073 0.072 0.071 0.021 0.073 0.072 0.3 2.1 3.1 93.0 0.1 1.3
146
Chapter 4
Table 4.30: Indirect determination of the uncertainty budget for CaO in the GUM framework EXCEL template for the four test
u(CaO) u(CaO) u(CaO) u(CaO) u(CaO) u(CaO) u(CaO) LOI Flux MI Slope Int. P
Sample CaO
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.6785 0.0137 0.0137 0.0137 0.0132 0.0042 0.0137 0.0137 0.0 0.5 5.4 93.8 0.2 0.0
2 3.731 0.072 0.072 0.072 0.072 0.013 0.072 0.072 0.1 1.0 1.2 97.8 0.0 0.0
3 49.482 1.139 1.231 1.139 1.130 0.238 1.139 1.131 9.1 0.0 0.9 89.2 0.0 0.8
4 10.502 0.202 0.202 0.201 0.201 0.026 0.202 0.202 0.2 0.5 0.4 98.8 0.0 0.0
Table 4.31: Indirect determination of the uncertainty budget for Na2O in the GUM framework EXCEL template for the four test
u(Na2O) u(Na2O) u(Na2O) u(Na2O) u(Na2O) u(Na2O) u(Na2O) LOI Flux MI Slope Int. P
Sample Na2O
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.418 0.006 0.006 0.006 0.006 0.006 0.004 0.006 0.0 1.6 20.9 12.9 64.5 0.0
2 6.819 0.041 0.041 0.038 0.037 0.023 0.040 0.039 0.1 10.7 13.3 68.6 1.5 5.9
3 <0.02 0.003 0.003 0.003 0.003 0.003 0.001 0.003 1.0 0.0 10.4 0.6 87.9 0.0
4 3.614 0.022 0.022 0.021 0.020 0.012 0.021 0.021 0.3 8.0 12.0 73.3 4.4 2.1
Table 4.32: Indirect determination of the uncertainty budget for K2O in the GUM framework EXCEL template for the four test
u(K2O) u(K2O) u(K2O) u(K2O) u(K2O) u(K2O) u(K2O) LOI Flux MI Slope Int. P
Sample K2O
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 15.664 0.238 0.238 0.237 0.237 0.033 0.238 0.238 0.0 0.5 0.3 99.2 0.0 0.0
2 3.842 0.058 0.058 0.058 0.058 0.009 0.058 0.058 0.0 0.5 0.3 99.2 0.0 0.0
3 <0.002 0.001 0.001 0.001 0.001 0.001 0.000 0.001 0.6 0.0 3.0 1.0 95.4 0.0
4 0.694 0.011 0.011 0.011 0.011 0.002 0.011 0.011 0.1 0.4 1.0 97.9 0.5 0.0
147
Chapter 4
Table 4.33: Indirect determination of the uncertainty budget for P2O5 in the GUM framework EXCEL template for the four test
u(P2O5) u(P2O5) u(P2O5) u(P2O5) u(P2O5) u(P2O5) u(P2O5) LOI Flux MI Slope Int. P
Sample P2O5
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.1186 0.0027 0.0027 0.0027 0.0021 0.0022 0.0024 0.0027 0.0 1.0 41.6 33.5 23.9 0.0
2 0.6109 0.0088 0.0088 0.0086 0.0081 0.0040 0.0087 0.0088 0.0 2.6 11.4 83.9 2.1 0.0
3 <0.01 0.0009 0.0009 0.0009 0.0008 0.0009 0.0006 0.0009 0.1 0.0 34.2 1.3 64.4 0.0
4 0.2635 0.0042 0.0042 0.0042 0.0037 0.0025 0.0040 0.0042 0.0 1.8 20.8 69.1 8.4 0.0
Table 4.34: Indirect determination of the uncertainty budget for Cr2O3 in the GUM framework EXCEL template for the four test
u(Cr2O3) u(Cr2O3) u(Cr2O3) u(Cr2O3) u(Cr2O3) u(Cr2O3) u(Cr2O3) LOI Flux MI Slope Int. P
Sample Cr2O3
(a) (b) (c) (d) (e) (f) (g) (%) (%) (%) (%) (%) (%)
1 0.0098 0.0013 0.0013 0.0013 0.0008 0.0012 0.0012 0.0013 0.0 0.1 62.4 19.6 17.9 0.0
2 0.0186 0.0018 0.0018 0.0018 0.0013 0.0013 0.0017 0.0018 0.0 0.2 46.9 42.2 10.6 0.0
3 0.0136 0.0010 0.0010 0.0010 0.0009 0.0006 0.0010 0.0010 1.5 0.0 19.7 70.1 8.6 0.0
4 0.0749 0.0049 0.0049 0.0049 0.0047 0.0016 0.0049 0.0049 0.0 0.2 6.8 92.1 0.9 0.0
148
Chapter 4
Except for sample 3, the calibration slope has the highest impact on the uncertainty of
measurement for SiO2, while the flux components and the instrumental responses
have a minor impact. In the limestone sample with high LOI and low SiO2-value, the
LOI and the intercept on Y-axis are major contributors to SiO2’s uncertainty.
In the case of TiO2, Al2O3 and MnO, the instrumental responses is a major contributor
to the uncertainty of their measurements, while the calibration slope and intercept on
Y-axis are major or minor contributors depending on their concentration values. The
flux components are minor contributors.
The calibration slope is a major contributor to the uncertainty of Fe 2O3, while the
instrumental responses and flux components are minor contributors. LOI is a minor
contributor in Sample 3.
The instrumental responses, calibration slope, and intercept on Y-axis are major or
minor contributors to the uncertainty of MgO, Na2O, P2O5 and Cr2O3, depending on
the concentration value. The flux components are minor contributors, except for
sample 3, where the concentration value is below LOD. The precision is a minor
contributor in two test samples with the highest concentration values. Furthermore, in
sample Li3 6c with high LOI-concentration, the LOI is a minor contributor.
The calibration slope and intercept on Y-axis are major or minor contributors to the
uncertainty of K2O depending on their concentration value, while the instrumental
responses and the flux components are minor contributors. The precision’
contributions for K2O are insignificant.
149
Chapter 4
4.4 Performance report for the study test method
The main method performance parameters, selectivity, linearity, working ranges, LOD,
LOQ, trueness and precision, to validate the study test method are described in 4.2.
The test method’s metrological traceability is established by the CRM’s used to
calibrate each measurand and evaluate the test method’s trueness. The uncertainty
of measurement for a specific sample analysed by the study test method can be
calculated using the GUM framework template developed for the study.
Therefore, the study test method is fit for purpose.
150
Chapter 5
Chapter 5:
Conclusions
5.1 Introduction
The study's main work is summarised, and its outcomes are evaluated in Chapter 5 to
compare them with the study’s objectives and aims.
5.2 Summary
The uncertainty of measurement and bias (trueness) are two important performance
characteristics of a test method that are evaluated during method validation.
Metrological traceability links the test results to internationally recognised references
(usually SI units) by using CRMs to calibrate and/or evaluate the test method’s
trueness. This process assents to the concepts of method validation, metrological
traceability and uncertainty of measurement being the three cornerstones to evaluate
a test method to ensure the accuracy and reliability of the measurement results
(Fig. 2.5). Without a documented method validation process, the uncertainty of
measurement cannot be estimated.
The GUM[4] is the main reference document, although there is no consensus on a

standardised approach to evaluate the uncertainty of measurement. The bottom-up
and top-down approaches use method validation data differently to quantify the
uncertainty of measurement. In the top-down approach, results from QCs, method
validation studies (CRMs) or PT studies[50] are used to quantify the uncertainty of a
test method[49]. The bottom-up approach is a rigorous process where experimental
data in the test method validation is used to quantify the test sample’s uncertainty of
measurement. The GUM[4] and the GUM-1’s Monte Carlo method[26] use the same
evaluation process for the experimental data but differ in the approach to calculate the
uncertainty of measurement. The main steps for the GUM are:
i) Specification of the measurands, identification of the input quantities on
which the output quantity depends and setup of a functional model between
the input and output quantities. Four measurement models were identified
in the study test method.
151
Chapter 5
ii) Evaluation of input quantities based on associated information to establish

the standard uncertainty process as listed in Table 3.11. Type A evaluation
is applied for statistically obtained information, and Type B for information
obtained by other means.
iii) Evaluation of the correlation between any input estimates. If present, the
sensitivity coefficients for each input quantity are calculated, and the GUM
formula (Eq. 4.2) is used.
For the MCM, different permutations of a measurement model are calculated using
pseudo-random values of the input quantities. The standard deviation of the result
from the permutations defines the uncertainty of measurement.
The scope of the test method is the analysis of typical samples from South Africa for
major elements (SiO2, TiO2, Al2O3, total Fe as Fe2O3, MnO, MgO, CaO, Na2O, K2O,
P2O5 and Cr2O3) on glass disks by WD-XRF and LOI by gravimetry. Glass disks are
prepared with 10 g flux and 1 g roasted material and analysed by WD-XRF. The main
corrections applied by the SuperQ software are for instrumental drift, background, line-
overlaps, and matrix effects. The study test method performance parameters are
selectivity, working range and linearity, trueness (Recovery and the En-score),
precision, LOD and LOQ (in 4.2), and with the calculated uncertainty of measurement
for four test samples, the test method is reported to be fit for purpose.
Repeatability and Intermediate precision were determined to evaluate the following

effects on the results: (i) sample and flux masses; (ii) the different cups and measuring
positions; (iii) the quality of glass disks, indicating the test method to be robust.
Although results from fusion disks from the two fluxers are different for SiO 2 and K2O,
they are acceptable. Values obtained from the precision evaluations were used for the
bottom-up approaches.
Four EXCEL templates were developed during the study for:

1) the GUM framework. (‘GUM framework.xlsx’)
2) the Kragten method. (‘Kragten method.xlsx’)
3) the Monte Carlo method. (‘MCM.xlsx’)
4) the top-down method. (‘Top down method.xlsx’)
152
Chapter 5
XGENLINE v8.1 was used for the uncertainty of intercept on the Y-axis, while the NIST
uncertainty machine was used to verify the values in the GUM framework EXCEL
workbook – especially the sensitivity coefficients determined by the partial derivatives.
The uncertainty values of the different methods were compared, and there are
significant differences between the uncertainty values produced by the bottom-up and
top-down approaches. For SiO2, CaO and LOI at high concentration values, the GUM
framework approach produces uncertainty values higher than those from the Kragten
method and MCM.
The Kragten method determines the main contributors to the uncertainty budget per
measurement model, while the GUM framework EXCEL workbook indirectly
determines the test method's main contributors. The matrix corrected counts
(instrumental responses), calibration slopes, and intercepts on the Y-axis are major
contributors to the uncertainty budget for all the major elements analysed by WD-XRF.
The flux components are minor contributors, while precision could be excluded for
most elements.
5.3 Conclusions
Evaluating the test method, documenting the processes in the test method, and
calculating the uncertainty of measurement is an iterative process. It illustrates that
the three cornerstones of the test method’s performance are intrinsically part of each
other.
Although the GUM framework is not applicable when the uncertainty contributions are
not approximately equal [28], it should still be applicable when the functional
relationships are linear, and probability distributions of the input and output quantities
are symmetric. The uncertainty budget is an important tool to evaluate the effect of
input quantities on the final result’s uncertainty of measurement. The intercept of the
regression slope on the Y-axis in SuperQ, for example, is set to be at (0,0), but as the
uncertainty of the intercept is large for a result’s concentration close to the LOD, it still
must be calculated.
153
Chapter 5
The study’s main objectives of test method validation, metrological traceability set up,
and estimation of uncertainty of measurement for test samples analysed by WD-XRF
have been accomplished. Furthermore, the study aims to compare the bottom-up and
top-down approaches and set up a template to estimate the uncertainty of
measurement by WD-XRF, resulted in the development of four EXCEL templates that
will be useful in future. The EXCEL templates can also be developed into user-friendly
applications.
.
154
References
References
[1] Milde, D (2018) Traceability, validation and measurement uncertainty - 3 pillars
for quality of measurement results. Eurachem Scientific Workshop. Data –
Quality, Analysis and Integrity. Downloaded from www.eurachem2018.com
[2] ILAC-G17:2002 Introducing the Concept of Uncertainty of Measurement in
Testing in Association with the Application of the Standard ISO/IEC 17025
[3] ISO/IEC 17025:2017 General requirements for the competence of testing and
calibration laboratories.
[4] JCGM 100:2008 GUM 1995 with minor corrections. Evaluation of measurement
data – Guide to the expression of uncertainty in measurement.
[5] Willis J.P., Turner K., and Pritchard G. (2011) XRF in the workplace PANalytical
Australia
[6] Magnusson B., Nȁykki T., Hovind H., Krysell M. and Sahlin E. (2017) Handbook
for calculation of measurement uncertainty in environmental laboratories.
Nordtest Report NT TR 537 (ed. 4) Available from www.nordtest.info
[7] Illowsky, B. and Dean, S. (2017) Introductory Statistics. Samurai Media Limited,
906 pp, ISBN 9888407309, (downloaded from
https://openstax.org/details/books/introductory-statistics on 12 July 2019)
[8] Kirkup L. and Frenkel R.B. (2006) An introduction to uncertainty in
measurement using the GUM (Guide to the expression of uncertainty in
measurement). Cambridge University Press, 233 pp
[9] NIST/SEMATECH e-Handbook of Statistical Methods,
https://www.itl.nist.gov/div898/handbook/eda/section3/eda3672.htm, 25
August 2020
[10] JCGM 200:2012 International vocabulary of metrology – Basic and general
concepts and associated terms (VIM) 3rd edition 2008 version with minor
corrections.
[11] Kacker, R. (2018) True value and uncertainty in the GUM. J. Phys.: Conf. Ser.
1065 212003
[12] Kacker, R.N. (2018) Measurement uncertainty and its connection with true
value in the GUM versus JCGM documents. Measurement, 127, 525-532
[13] Botha, A. (2021) Personal communication.
155
References
[14] Birch, K (2004) Estimating Uncertainties in Testing. Measurement Good

Practice Guide No. 36, British Measurement and Testing Association, ISSN
1368-6550
[15] Magnusson B. and Örnemark U. (eds.) (2014) Eurachem Guide: The Fitness
for Purpose of Analytical Methods – A Laboratory Guide to Method Validation
and Related Topics, (2nd ed.) ISBN 978-91-87461-59-0. Available from
www.eurachem.org
[16] Rozet E., Marini R.D., Ziemons E., Dewé W., Rudaz S., Boulanger B. and
Hubert Ph. (2011a) Total error and uncertainty: Friends or foes? Trends in
Analytical Chemistry, 30 (5), 797-806
[17] ISO 5725-1:1994 Accuracy (trueness and precision) of measurement methods
and results Part 1: General principles and definitions.
[18] White, D.R. (2016) In pursuit of a fit-for-purpose uncertainty guide. Metrologia,
53, S107-S124
[19] Ellison S.L.R and Williams A. (editors) (2019) Eurachem/CITAC Guide:
Metrological Traceability in Chemical measurement (2nd ed.) ISBN: 978-0-
948926-34-1. Available from www.eurachem.org
[20] Thompson, M. (2004) What is uncertainty from sampling, and why is it
important? Analytical Methods Committee, AMC TB 16A, ISSN 1757- 5958
[21] Majcen N and Gegevičius V (editors) (2012) Analytical measurement:
measurement uncertainty and statistics Available on-line from
http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/29537/1/la
na2207enn-web.pdf Downloaded on 15 May 2020
[22] Sajnóg A., Hanć A. and Baralkiewicz D. (2018) Metrological approach to
quantitative analysis of clinical samples by LA-ICP-MS: A critical review of
recent studies. Talanta, 182, 92-110.
[23] Bullock, DG and Thompson M (2015) Fitness for purpose: the key feature in
analytical proficiency testing. Analytical Methods Committee, AMCTB No 68,
Anal. Methods, 2015, 7, 7404-7405
[24] Thompson, M. ed. (2005) Analytical and sample strategy, fitness for purpose,
and computer games. Analytical Methods Committee, AMCTB No 20
[25] Stant L.T., Aaen P.H., Ridler N.M. (2016) Comparing methods for evaluating
measurement uncertainty given in the JCGM' Evaluation of Measurement Data'
documents. Measurement, 94, 841-851
156
References
[26] JCGM 101:2008 Evaluation of measurement data – Supplement to the "Guide

to the expression of uncertainty in measurement" – Propagation of distributions
using a Monte Carlo method
[27] JCGM 102:2011 Evaluation of measurement data – Supplement 2 to the "Guide
to the expression of uncertainty in measurement" – Extension to any number of
output quantities
[28] JCGM 104:2009 Evaluation of measurement data – An introduction to the
"Guide to the expression of uncertain uncertainty in measurement" – and
related documents
[29] JCGM 106:2012 Evaluation of measurement data – The role of measurement
uncertainty in conformity assessment
[30] BIPM website: https://www.bipm.org/en/publications/guides/ (accessed 13 May
2020)
[31] Farrance I. and Frenkel R. (2012) Uncertainty of Measurement: A Review of the
Rules for Calculating Uncertainty Components through Functional
Relationships Clin Biochem Rev, vol 33, 49-75
[32] Theodorou, D.G. (2015) Measurement data analysis in quality management
systems. Application to test fuel test methods. PhD thesis, School of Chemical
Engineering, National Technical University of Athens.
[33] Ehrlich, C. (2014) Terminological aspects of the Guide to the Expression of
Uncertainty in Measurement (GUM) Metrologia, 51, S145-S154.
[34] Kacker, R.N. (2015) Probability distributions and coverage probability in GUM,
JCGM documents, and statistical inference. Measurement, 65, 61-70
[35] O'Hagan, A (2014) Eliciting and using expert knowledge in metrology.
Metrologia, 51, S237-S244
[36] Lira, I. (2016) Dealing with prior knowledge about the measurand.
Measurement, 78, 344-347
[37] van der Veen, A.M.H. (2018) Bayesian methods for type A evaluation of
standard uncertainty. Metrologia, 55, 670-684
[38] Possolo, A. (2014) Statistical models and computation to evaluate
measurement uncertainty. Metrologia, 51, S228-S236
[39] Separovic L., Bettencourt da Silva R.J.N. and Lourenço F.R. (2019) Improved
spreadsheet method for determination of between components metrological
157
References
correlation due to the sharing of analytical steps. Chemometrics and Intelligent

Laboratory Sytems 189, 161-168.
[40] Bich W., Cox M. and Michotte C. (2016) Towards a new GUM – an update.
Metrologia, 53, S149-S159
[41] Cox, M. and Harris, P. (2014) GUM anniversary issue. Metrologia, 51, S141-
S143
[42] Thompson, M. and Ellison, S.L.R. (2012) Dark uncertainty. Analytical Methods
Committee, AMCT No 53. Anal. Methods, 4, 2609-2612
[43] Dèsenfant M. and Priel M. (2017) Reference and additional methods for
measurement uncertainty evaluation. Measurement, 95, 339-344
[44] Farrance I, Badrick T and Frenkel R (2018) Uncertainty in measurement: A
review of the procedures for determining uncertainty in measurement and its
use in deriving the biological variation of the estimated glomerular filtration rate.
Practical Laboratory Medicine, 12, 1-13
[45] Kragten J. (1994) Calculating Standard Deviations and Confidence Intervals
with a Universally Applicable Spreadsheet Technique. Analyst, vol 119, 2161-
2165
[46] Ellison S.L.R and Williams A. (editors) (2012) Eurachem/CITAC Guide CG4
Quantifying Uncertainty in Analytical Measurement (3rd edition) ISBN 978-0-
948926-30-3. Available from www.eurachem.org
[47] Farrance I, Frenkel R (2014) Uncertainty in Measurement: A Review of Monte
Carlo Simulation Using Microsoft EXCEL for the Calculation of Uncertainties
Through Functional Relationships, Including Uncertainties in Empirically
Derived Constants Clin Biochem Rev, 35(1), 37-61
[48] National Institute of Standards and Technology (NIST)
https:/uncertainty.nist.gov (accessed during October and November 2021)
[49] EUROLAB Technical Secretariat (2007) Measurement uncertainty revisited:
Alternative approaches to uncertainty evaluation. EUROLAB Technical Report
1/2007. Available from www.eurolab.org
[50] Magnusson B., Ossowicki H., Rienitz O, and Theodorsson E. (2011) Routine
internal- and external-quality control data in clinical laboratories for estimating
measurement and diagnostic uncertainty using GUM principles. Scandinavian
Journal of Clinical & Laboratory Investigation, 2011; 00: 1–9
158
References
[51] Martinello F., Snoj N., Skitek M. and Jerin A. (2020) The top-down approach to
measurement uncertainty: which formula should we use in laboratory
medicine? Biochem Med (Zagreb) 30(2), 1-9
[52] Rousseau, R.M. (2001) Detection limit and estimate of uncertainty of analytical
XRF results. The Rigaku Journal, vol 18, no.2, 33-47
[53] Willis J.P. and Duncan A.R. (2008) Understanding XRF Spectrometry. Volume
1 Basic concepts and instrumentation. PANalytical B.V. Almelo, Netherlands.
ISBN 978-90-809086-4-2
[54] Rashmi, Dhawan U. and Suri D.K. (1999) Evaluation of Uncertainty of
Measurement in Quantitative XRF Analysis: I. Scattered Radiation Method. X-
Ray Spectrom., 28, 157-162.
[55] Morgenstern P., Brȕggemann L. and Wennrich R. (2005) Effect of the sample
matrix on measurement uncertainty in X-ray fluorescence analysis.
Spectrochmica Acta Part B 610, 1373-1379
[56] Blanchette J, Dessureault M and Slim F (2001) ASTM C-114 Qualification Test
Method to Improve Analytical Results Using the Fused Beads Technique. In
Proceedings of the Twenty-Third Annual International Conference on Cement
Microscopy, April 8-11, 2002, San Diego, California USA (Publication copy
received from authors)
[57] Ichikawa S, Onuma H. and Nakamura T. (2016) Development of undersized
(12.5 mm diameter) low-dilution glass beads for X-ray fluorescence
determination of 34 components in 200 mg of igneous rock for applications with
geochemical and archaeological silicic samples. X-Ray Spectrom. 45, 34-47
[58] Younis A., Ahmadi Z, Adams M.G. and Iqbal A. (2017) A simple method for
quantitative analysis of elements by WD-XRF using variable dilution factors in
fusion bead technique for geologic specimens. X-ray Spectrom. 46, 69-76
[59] Gazulla M.F., Rodrigo M., Vincente S. and Orduňa M (2010) Methodology for
the determination of minor and trace elements in petroleum cokes by
wavelength-dispersive X-ray fluorescence (WD-XRF) X-ray Spectrom, 39, 321-
327
[60] Gazulla M.F., Ventura M.J., Orduňa M., Rodrigo M. and Gómez M.P. (2017)
Analysis of corrosion residues by WDXRF X-ray Spectrom, 46, 271-276
[61] Wegrzynek D., Markowicz A., Chinea-Cano E. and Bamford S. (2003)
Evaluation of the uncertainty of element determination using the energy-
159
References
dispersive x-ray fluorescence technique and the emission-transmission

method. X-Ray Spectrom., 32, 317-335
[62] Álvarez R.P., Markowicz A., Wegrzynek D., Cano E.C., Bamford S.A. and
Torres D.H. (2007) Quality management and method Validation in EDXRF
analysis. X-Ray Spectrom., 36, 27-34
[63] ISO 29581-2:2010(E) (2010) Cement – Test methods. Part 2: Chemical
analysis by X-ray fluorescence
[64] ISO 12677:2011(E) (2011) Chemical analysis of refractory products by X-ray
fluorescence (XRF) – Fused cast-bead method.
[65] ISO 18227:2014(E) (2014) Soil quality – Determination of elemental
composition by X-ray fluorescence
[66] ISO 18336:2016 Guidelines for good XRF laboratory practice for the iron ore
industry
[67] Lachance G.R. and Claisse F. (1995) Quantitative X-ray fluorescence analysis.
Theory and application. John Wiley & Sons, West Sussex, England
[68] Elam W.T., Ravel B.D. and Sieber J.R. (2002) A new atomic database for X-
ray spectroscopic calculations. Radiation Physics and Chemistry, 63, 121-128.
[69] Vrebos Bruno (2021) Personal communication via email in January 2021
[70] Norrish, K and Hutton, J.T. (1969) An accurate X-ray spectrographic method
for the analysis of a wide range of geological samples. Geochimica et
Cosmochimica Acta, 33(4), 431-453
[71] Taggart J.E. and Siems D.F. (2002) Major element analysis by wavelength
dispersive X-ray fluorescence spectrometry. U.S. Geological Survey Open File
Report 02-223-T.
[72] Vrebos B. (2021) Personal communication via Microsoft Teams on 5 November
2021.
[73] Barwick V. (editor) (2016) Eurachem/CITAC Guide: Guide to Quality in
Analytical Chemistry: An Aid to Accreditation (3rd ed.) ISBN 978-0-948926-32-
7. Available from www.eurachem.org.
[74] National Institute of Standards and Technology (NIST)
https://www.nist.gov/document/pt-ilc-report-terminology-guidance-8-4-2011-
2doc (accessed on 21 September 2021)
160
References
[75] Gonzáles A.G., Ángeles M.H. and Asuero A.G. (2010) Intra-laboratory
assessment of method accuracy (trueness and precision) by using validation
standards. Talanta, 82, 1995-1998
[76] SuperQ’s help application. (SuperQ v6)
[77] Miller J.N. and Miller J.C. (2010) Statistics and Chemometrics for Analytical
Chemistry ISBN: 978-0-273-73042-2
[78] National Physics Laboratory, UK (2020)
https://www.npl.co.uk/resources/software/xgenline (downloaded on 20 October
2020)
[79] Steele, T.W. and Hansen R.G. (1979) Major element data (1966 – 1978) for the
six “Nimroc” reference samples. Geostandards Newsletter, vol 3 (2), 135-172
161
Appendix A
Appendix A
Available information of the used CRMs and for the determination

of standard uncertainty
This appendix lists the CRMs and RMs used in the study and their reported information
to estimate associated uncertainty (Table A1). The method to estimate their standard
uncertainty is also detailed here.
Table A1: Reported information for uncertainty evaluation for the CRMs and RMs
used in the test method calibration.
BCS is Bureau of Analysed Samples Ltd. In the United Kingdom
(Originally British Chemical Standards).
CCRM is a Canadian Certified Reference Materials Project
CNAC is China National Analysis Center for Iron and Steel
CTM is Centro de Tecnologia Mineral in Brazil.
ECRM is the European Committee for Iron and Steel Standardization
GSJ is the Geological Survey of Japan
MINTEK is from South Africa
NIST is the National Institute of Standards and Technology, USA.
(Originally National Bureau of Standards)
Reported information on Information for uncertainty

Producer CRM
result certificate estimation
Standard deviation reported Standard deviation and amount
BCS BCS-203a
with the results. of results.
Standard deviation calculated
BCS BCS-176/2 A set of results.
from set of results
BCS BCS No. 267 A set of results.
from set of results
A set of results are reported with
standard deviation. Certified Confidence limits at k = 2 for
BCS BCS-CRM 301/1
values with half width confidence level of 95%
confidence interval at 95%.
Both intralaboratory mean and
BCS-CRM No. Standard deviation and amount
BCS standard deviation are reported
313/1 of results.
along with a set of 9 results
162
Appendix A
Table A1: Continued

Producer CRM
Standard deviation reported Standard deviation and amount
BCS BCS-314
with set of results. of results.
BCS-CRM No. Half-width confidence interval at Confidence limits at k = 2 for
BCS
348 95% confidence level of 95%
from set of results
from set of results
Half-width confidence interval at Confidence limits at k = 2 for
BCS BCS-381
95% confidence level of 95%
A set of results with half with 𝑡 × 𝑠𝑚
BCS BCS-393 𝐶(95%) =
confidence interval at 95% √𝑛
BCS BCS-394 A set of results.
from set of results
Both intra and inter-laboratory Inter-laboratory standard
CTM BXGO-1 standard deviation is reported deviation with amount of
with certified values. certified values.
CTM GXMG-4 standard deviation is reported deviation with amount of
CTM BXPA-4 standard deviation is reported deviation with amount of
Semi-range determined from
CCRM FER-2 Results only
last significant figure in result
CCRM FER-4 Results only
Expanded uncertainty are
reported as ± values and is
Results with expanded 𝑠
CNAC GSR-1 expressed as 𝑈 = 𝑡𝛼 ∙ where
uncertainty √𝑁
α = 0.01, s = standard deviation
and N = number of data (N>6)
𝑠
Results with expanded expressed as 𝑈 = 𝑡𝛼 ∙
CNAC GSR-2 √𝑁
uncertainty where α = 0.01, s = standard
deviation and N = number of
data (N>6)
𝑠
CNAC GSR-3 √𝑁
data (N>6)
Table A1: Continued
163
Appendix A

Producer CRM
𝑠
CNAC GSR-4 √𝑁
data (N>6)
𝑠
CNAC GSR-5 √𝑁
data (N>6)
𝑠
CNAC GSR-6 √𝑁
data (N>6)
A set of results, standard
Standard deviation and number
ECRM ECRM-610-1 deviation, and number of used
of results
results
Results, standard deviation and
amount of analysis obtained Standard deviation and number
GSJ JA-2
from Geological Survey of of results
Japan’s website.
GSJ JA-2
Japan’s website.
GSJ JA-2
Japan’s website.
Robust standard deviation was
MINTEK SARM-3 Results only
obtained from a report[79].
Low and high limit values at
MINTEK SARM-8 Confidence limits
95% confidence levels.
164
Appendix A
Table A1: Continued

Producer CRM
Low and high limit values at Robust standard deviation was
MINTEK SARM-11
95% confidence levels. obtained from a report[79].
Low and high limit values at Robust standard deviation was
MINTEK SARM-12
95% confidence levels. obtained from a report[79].
Estimated uncertainty is
expressed in significant digits,
Results with estimated
NIST NBS-76a for example, ± 1 in the last
uncertainty
significant figure of
concentration.
uncertainty
concentration.
uncertainty
concentration.
NIST NBS-91 None
NIST NBS-98 A set of results.
from set of results
± values are reported along with ± values are used to determine
NIST NBS-98b
results confidence interval
Results with estimated Significant digits, i.e., ± 1 in the
NIST NBS-99a uncertainty as expressed in last significant figure is used to
significant digits determine confidence interval
165
Appendix A
Table A1: Continued

Producer CRM
NIST NBS-103 A set of results.
from set of results
± values are reported along with ± values is used to determine
NIST NBS-278
results confidence interval
Significant digits, i.e., ± 1 in the
uncertainty which is reported as
NIST NBS-600 last significant figure is used to
based on judgement. No
determine confidence interval
confidence level is reported.
Significant digits, i.e. ± 1 in the
uncertainty which is reported as
NIST NBS-679 last significant figure is used to
based on judgement. No
determine confidence interval
Results with uncertainty being
NIST NBS-694 Two standard deviations
two standard deviations
uncertainty which is reported as ± values is used to determine
NIST NBS-1413
based on judgement. No confidence interval
166
Appendix B
Appendix B
CRM/RM’s used in the calibration, their description and flux used to prepare
glass disk standards
The standard glass disks’ sample preparation is detailed in Table B.1. Flux used to
prepare the glass disks is either A or B or C. (Flux A = 66.67:32.83:0.05, Flux B =
49.75:49.75:0.5, Flux C = 66.67:32.83:0.50 Lithium metaborate, Lithium tetraborate
and Lithium bromide) FM is the Beadmaster fluxer (Fig. 3.6) while TO is the Claisse
TheOx fluxer (Fig. 3.4).
Table B1: The used CRM and RMs in the calibrations of the test method, their
description and the flux and fluxer used to prepare the fusion disk.
Glass Roasted
Fluxer Flux mass
disk CRM/RM Description Flux sample
used (g)
label mass (g)
1252 SARM-39 Kimberlite B FM 1.0004 10.0001
1253 SARM-39 Kimberlite B TO 1.0004 10.0034
1254 SARM-43 Magnesite C FM 1.0007 10.0009
1255 SARM-43 Magnesite C TO 1.0002 10.0002
1256 SARM-11 Fe ore B FM 1.0008 10.0006
1260 GSR-1 Biotite granite A FM 1.0001 10.0007
1261 GSR-1 Biotite granite A TO 1.0014 10.0001
Quartz hornblende
1262 GSR-2 B FM 1.0002 10.0031
andesite
Arkosic quartz
1266 GSR-4 A FM 1.0004 10.0044
sandstone
Arkosic quartz
1267 GSR-4 A TO 1.0007 10.0031
sandstone
1268 GSR-5 Shale B FM 1.0001 10.0021
1269 GSR-5 Shale B TO 1.0015 10.0036
1270 GSR-6 Limestone C FM 1.0004 10.0026
167
Appendix B
Table B1: Continued.

Glass Roasted
Fluxer Flux mass
used (g)
label mass (g)
1271 GSR-6 Limestone C TO 1.0005 10.0025
1272 SARM-50 Dolerite B FM 1.0001 10.0021
1274 SARM-6 Dunite B FM 1.0001 10.0032
1276 SARM-3 Lujavrite B FM 1.0001 10.0015
1278 SARM-12 Fe ore B FM 1.0007 10.0011
1280 SARM-8 Cr ore B FM 1.0007 10.0041
1282 BXGO-1 Bauxite B FM 1.0012 10.0018
1284 BXMG-4 Bauxite B FM 1.0007 10.0017
1285 BXMG-4 Bauxite B TO 1.0011 10.0005
1286 BXPA-4 Bauxite B FM 1.0002 10.0027
1292 BCS-375 Soda feldspar B FM 1.0017 10.0025
1293 BCS-375 Soda feldspar B TO 1.0013 10.0041
1294 BCS-376 Potash feldspar B FM 1.0002 10.0039
1295 BCS-376 Potash feldspar B TO 0.9999 10.0009
1298 BCS-313/1 High purity silica A FM 1.0001 10.0001
1299 BCS-313/1 High purity silica A TO 1.0006 10.0032
1300 BCS-314 Silica brick A FM 1.0004 10.0001
1304 BCS-267 Silica brick A FM 1.0007 10.0014
1305 BCS-267 Silica brick A TO 1.0002 10.0002
1310 BCS-348 Ball clay A FM 1.0008 10.0017
1311 BCS-348 Ball clay A TO 1.0002 10.0013
1314 BCS-394 Bauxite B FM 1.0011 10.0011
1318 BCS-301/1 Iron ore B FM 1.0002 10.0022
1319 BCS-301/1 Iron ore B TO 1.0001 10.0021
1320 BCS-381 Basic slag B FM 1.0015 10.0029
1324 BCS-176/2 Manganese - ore C FM 1.0006 10.0013
Western Phosphate
1328 NBS-694 B FM 1.0003 10.0038
Rock
1330 SARM-41 Shale B FM 1.0003 10.0031
1334 NBS-103 Cr refractory B FM 0.9998 10.0016
1336 NBS-600 Bauxite B FM 1.0003 10.0034
1338 NBS-679 Brick clay A FM 1.0001 10.0023
1344 NBS-91 Opal glass powder B FM 1.0011 10.0041
1346 NBS-77a Burnt refractory B FM 1.0012 10.0012
1350 NBS-98b Plastic clay A FM 1.0008 10.0009
1352 NBS-98 Plastic clay A FM 1.0001 10.0005
1354 NBS-1413 High Al sand A FM 1.0005 10.0007
1356 NBS-99a Soda feldspar B FM 1.0005 10.0001
1358 FER-2 Iron-formation sample B FM 1.0001 10.0035
168
Appendix B
Table B1: Continued.

Glass Roasted
Fluxer Flux mass
used (g)
label mass (g)
1358 FER-2 Iron-formation sample B FM 1.0001 10.0035
1360 FER-4 Iron-formation sample A FM 1.0004 10.0022
1362 JF-1 Feldspar B FM 1.0001 10.0001
1366 JA-2 Andesite B FM 1.0011 10.0009
1368 NBS-78a Burnt refractory B FM 1.0001 10.0001
1371 SARM-40 Carbonatite C FM 1.0006 10.0000
1372 SARM-40 Carbonatite C TO 0.9996 10.0001
1375 BCS-393 Limestone C FM 1.0003 9.9993
1377 NBS-98 Plastic clay B FM 1.0004 10.0008
169
Appendix C
Appendix C
Calibration certificate of balance
Figure C.1: Calibration certificate of balance used in the preparation of the glass
disks
170
Appendix D
Appendix D
Table D1: The assigned concentration values (%), reported or calculated standard deviations (s) and calculated standard
uncertainties (u) of the CRMs and RMs used in the calibration application in SuperQ. In cases of no reported standard
deviation or available information for its calculation, standard uncertainty is calculated as described in Table 3.11.
Fusion ID CRM/RM SiO2 s(SiO2) u(SiO2) TiO2 s(TiO2) u(TiO2) Al2O3 s(Al2O3) u(Al2O3) Fe2O3 s(Fe2O3) u(Fe2O3) MnO s(MnO) u(MnO) MgO s(MgO) u(MgO)
1252 SARM-39 33.44 0.26 0.150 1.58 0.17 0.098 4.29 0.33 0.191 9.29 0.31 0.179 0.17 0.01 0.006 26.24 0.78 0.450
1253 SARM-39 33.44 0.26 0.150 1.58 0.17 0.098 4.29 0.33 0.191 9.29 0.31 0.179 0.17 0.01 0.006 26.24 0.78 0.450
1254 SARM-43 5.99 0.45 0.260 0.01 0.009 0.005 0.06 0.040 0.023 0.26 0.06 0.035 0.01 0.005 0.003 44.11 0.68 0.393
1255 SARM-43 5.99 0.45 0.260 0.01 0.009 0.005 0.06 0.040 0.023 0.26 0.06 0.035 0.01 0.005 0.003 44.11 0.68 0.393
1256 SARM-11 3.10 0.02 0.01 0.064 0.002 0.0009 1.38 0.02 0.009 94.6 0.03 0.014 0.0146 0.0004 0.00019 0.0206 0.0004 0.00019
1260 GSR-1 72.83 0.10 0.027 0.287 0.012 0.0031 13.40 0.07 0.019 2.14 0.06 0.016 0.047 0.002 0.0006 0.42 0.04 0.011
1261 GSR-1 72.83 0.10 0.027 0.287 0.012 0.0031 13.40 0.07 0.019 2.14 0.06 0.016 0.047 0.002 0.0006 0.42 0.04 0.011
1262 GSR-2 60.6 0.14 0.04 0.52 0.015 0.004 16.17 0.12 0.032 4.90 0.060 0.016 0.078 0.00 0.0006 1.72 0.060 0.016
1266 GSR-4 90.36 0.15 0.040 0.264 0.013 0.0036 3.52 0.09 0.024 3.22 0.07 0.019 0.020 0.0009 0.0002 0.082 0.020 0.0054
1267 GSR-4 90.36 0.15 0.040 0.264 0.013 0.0036 3.52 0.09 0.024 3.22 0.07 0.019 0.020 0.0009 0.0002 0.082 0.020 0.0054
1268 GSR-5 59.23 0.16 0.043 0.659 0.022 0.0058 18.82 0.14 0.038 7.60 0.09 0.024 0.022 0.0014 0.0004 2.01 0.05 0.013
1269 GSR-5 59.23 0.16 0.043 0.659 0.022 0.0058 18.82 0.14 0.038 7.60 0.09 0.024 0.022 0.0014 0.0004 2.01 0.05 0.013
1270 GSR-6 16.60 0.06 0.016 0.327 0.015 0.0040 5.03 0.08 0.022 2.52 0.07 0.019 0.056 0.0035 0.0009 5.19 0.12 0.032
1271 GSR-6 16.60 0.06 0.016 0.327 0.015 0.0040 5.03 0.08 0.022 2.52 0.07 0.019 0.056 0.0035 0.0009 5.19 0.12 0.032
1272 SARM-50 51.56 0.42 0.243 0.86 0.025 0.014 15.28 0.30 0.173 11.0 0.29 0.17 0.17 0.015 0.009 7.57 0.18 0.104
1274 SARM-6 38.96 0.31 0.180 0.0200 0.0001 6E-05 0.3 0.135 0.08 17.00 0.41 0.236 0.22 0.0011 0.001 43.51 0.7832 0.452
1276 SARM-3 52.40 0.26 0.151 0.48 0.0024 0.001 13.64 0.341 0.197 9.91 0.36 0.206 0.77 0.00385 0.002 0.28 0.0952 0.055
1278 SARM-12 0.34 0.01 0.005 0.717 0.005 0.0025 0.77 0.02 0.009 95.3 0.07 0.04 0.22 0.004 0.002 2.80 0.04 0.019
171
Appendix D
Table D1: Continued

Fusion ID CRM/RM SiO2 s(SiO2) u(SiO2) TiO2 s(TiO2) u(TiO2) Al2O3 s(Al2O3) u(Al2O3) Fe2O3 s(Fe2O3) u(Fe2O3) MnO s(MnO) u(MnO) MgO s(MgO) u(MgO)
1280 SARM-8 4.30 0.04 0.020 0.24 0.01 0.005 10.57 0.11 0.055 20.2 0.1 0.05 0.25 0.01 0.005 14.69 0.06 0.030
1282 BXGO-1 0.634 0.046 0.023 0.461 0.016 0.008 60.72 0.36 0.180 4.645 0.098 0.0490 0.0066 0.0011 0.00055 0.03 0.02 0.010
1284 BXMG-4 9.53 0.14 0.070 1.778 0.030 0.0150 50.44 0.50 0.250 9.88 0.13 0.065 0.4179 0.0172 0.00858 0.04 0.02 0.010
1285 BXMG-4 9.53 0.14 0.070 1.778 0.030 0.0150 50.44 0.50 0.250 9.88 0.13 0.065 0.4179 0.0172 0.00858 0.04 0.02 0.010
1286 BXPA-4 4.651 0.079 0.040 1.306 0.026 0.0130 57.26 0.37 0.185 6.95 0.11 0.055 0.0458 0.0027 0.00135 0.03 0.02 0.010
1292 BCS-375 67.1 0.10 0.03 0.38 0.020 0.007 19.8 0.06 0.02 0.12 0.011 0.004 0.05 0.02 0.007
1293 BCS-375 67.1 0.10 0.03 0.38 0.020 0.007 19.8 0.06 0.02 0.12 0.011 0.004 0.05 0.02 0.007
1294 BCS-376 67.1 0.14 0.05 <0.02 17.7 0.04 0.01 0.10 0.005 0.002 0.03 0.011 0.004
1295 BCS-376 67.1 0.14 0.05 <0.02 17.7 0.04 0.01 0.10 0.005 0.002 0.03 0.011 0.004
1.225E-
1298 BCS-313/1 99.78 0.07 0.025 0.017 0.003 0.0011 0.036 0.004 0.0013 0.012 0.001 0.0004 0.0001 0.00003 0.0013 0.002 0.00082
05
1.225E-
1299 BCS-313/1 99.78 0.07 0.025 0.017 0.003 0.0011 0.036 0.004 0.0013 0.012 0.001 0.0004 0.0001 0.00003 0.0013 0.002 0.00082
05
1300 BCS-314 96.2 0.09 0.03 0.19 0.016 0.005 0.77 0.01 0.004 0.53 0.046 0.015 <0.01 0.05 0.008 0.003
1304 BCS-267 95.9 0.16 0.05 0.17 0.010 0.003 0.85 0.017 0.006 0.79 0.024 0.008 0.15 0.019 0.006 0.06 0.024 0.009
1305 BCS-267 95.9 0.16 0.05 0.17 0.010 0.003 0.85 0.017 0.006 0.79 0.024 0.008 0.15 0.019 0.006 0.06 0.024 0.009
1310 BCS-348 51.1 0.2 0.09 1.08 0.02 0.009 31.6 0.2 0.09 1.04 0.02 0.009 0.30 0.01 0.004
1311 BCS-348 51.1 0.2 0.09 1.08 0.02 0.009 31.6 0.2 0.09 1.04 0.02 0.009 0.30 0.01 0.004
1314 BCS-394 4.98 0.077 0.029 3.11 0.034 0.013 88.8 0.11 0.04 1.90 0.028 0.011 0.12 0.016 0.006
1318 BCS-301/1 7.40 0.03 0.012 0.16 0.01 0.004 4.26 0.05 0.020 34.10 0.0572 0.023 1.25 0.04 0.016 1.73 0.04 0.016
1319 BCS-301/1 7.40 0.03 0.012 0.16 0.01 0.004 4.26 0.05 0.020 34.10 0.0572 0.023 1.25 0.04 0.016 1.73 0.04 0.016
1320 BCS-381 8.78 0.09 0.038 0.35 0.03 0.013 0.67 0.05 0.021 19.02 0.14 0.060 3.16 0.08 0.034 1.03 0.07 0.030
1324 BCS-176/2 2.53 0.052 0.016 0.30 0.017 0.006 5.16 0.176 0.059 9.81 0.162 0.051 61.4 0.12 0.04 0.04 0.0043 0.001
1328 NBS-694 11.2 0.4 0.23 1.8 0.1 0.06 0.79 0.06 0.035 0.0116 0.0012 0.00069 0.33 0.02 0.012
1330 SARM-41 56.67 0.91 0.525 0.55 0.035 0.020 13.50 0.50 0.289 4.23 0.13 0.075 0.06 0.013 0.008 8.10 0.21 0.121
1334 NBS-103 8.24 0.024 0.009 0.93 0.018 0.006 20.83 0.064 0.022 15.99 0.111 0.039 0.21 0.016 0.007 16.27 0.060 0.021
1336 NBS-600 20.3 0.4 0.23 1.31 0.0 0.023 40.0 0.4 0.23 17.0 0.3 0.17 0.013 0.004 0.0023 0.05 0.01 0.006
Table D1: Continued
172
Appendix D
Fusion u(Fe2O3
CRM/RM SiO2 s(SiO2) u(SiO2) TiO2 s(TiO2) u(TiO2) Al2O3 s(Al2O3) u(Al2O3) Fe2O3 s(Fe2O3) MnO s(MnO) u(MnO) MgO s(MgO) u(MgO)
ID )
1338 NBS-679 52.07 0.64 0.370 0.96 0.055 0.032 20.80 0.64 0.371 12.94 0.30 0.173 0.2234 1.252 0.0114 0.0066
1344 NBS-91 67.5 0.06 0.019 0.0006 6.0 0.06 0.079 0.0006
1346 NBS-77a 35.0 0.1 0.06 2.66 0.05 0.029 60.2 0.1 0.06 1.00 0.05 0.029 0.38 0.01 0.006
1350 NBS-98b 57.01 0.34 0.684 1.35 0.020 0.040 27.02 0.38 0.218 1.69 0.01 0.029 0.0150 0.0006 0.00129 0.594 0.015 0.0310
1352 NBS-98 59.11 0.006 1.43 0.006 25.5 0.06 2.05 0.006 0.005 0.0006 0.72 0.006
1354 NBS-1413 82.77 0.40 0.231 0.11 0.02 0.012 9.90 0.13 0.075 0.24 0.01 0.006 0.06 0.01 0.006
1356 NBS-99a 65.22 0.05 0.029 0.007 0.001 0.0006 20.5 0.1 0.06 0.065 0.005 0.02887 0.02 0.01 0.006
1358 FER-2 49.21 0.006 0.18 0.058 5.16 0.006 39.47 0.058 0.12 0.006 2.1 0.058
1360 FER-4 50.07 0.006 0.07 0.058 1.7 0.06 39.97 0.058 0.19 0.006 1.41 0.006
3.78E-
1362 JF-1 66.69 0.77 0.148 0.005 0.002 0.0007 18.08 0.32 0.062 0.08 0.01 0.003 0.001 0.0001 0.006 0.001 0.0003
05
1366 JA-2 56.42 0.63 0.124 0.66 0.04 0.008 15.41 0.36 0.069 6.21 0.18 0.050 0.108 0.007 0.0013 7.60 0.27 0.054
1368 NBS-78a 19.4 0.1 0.057 3.22 0.05 0.029 71.7 0.1 0.06 1.2 0.1 0.06 0.70 0.01 0.006
1371 SARM-40 3.08 0.18 0.104 0.045 0.01 0.0058 0.41 0.085 0.049 2.75 0.14 0.081 0.18 0.01 0.006 1.97 0.29 0.167
1372 SARM-40 3.08 0.18 0.104 0.045 0.01 0.0058 0.41 0.085 0.049 2.75 0.14 0.081 0.18 0.01 0.006 1.97 0.29 0.167
1375 BCS-393 0.70 0.02 0.009 0.009 0.002 0.0008 0.12 0.01 0.004 0.045 0.005 0.0022 0.010 0.001 0.0004 0.15 0.01 0.004
1377 NBS-98 59.11 0.006 1.43 0.006 25.5 0.058 2.05 0.006 0.005 0.0006 0.72 0.006
173
Appendix D
Table D1: Continued

Fusion
CRM/RM CaO s(CaO) u(CaO) Na2O s(Na2O) u(Na2O) K2O s(K2O) u(K2O) P2O5 s(P2O5) u(P2O5) Cr2O3 s(Cr2O3) s(Cr2O3) LOI s(LOI) u(LOI)
ID
1252 SARM-39 9.69 0.21 0.121 0.5 0.19 0.11 1.04 0.065 0.038 1.46 0.075 0.043 0.19 0.02 0.012
1253 SARM-39 9.69 0.21 0.121 0.5 0.19 0.11 1.04 0.065 0.038 1.46 0.075 0.043 0.19 0.02 0.012
1254 SARM-43 0.75 0.15 0.087 0.05 0.16 0.092 0.04 0.02 0.012 0.02 0.005 0.003 0.0285 0.00014
1255 SARM-43 0.75 0.15 0.087 0.05 0.16 0.092 0.04 0.02 0.012 0.02 0.005 0.003 0.0285 0.00014
1256 SARM-11 0.0452 0.0031 0.00154 0.0152 0.0015 0.0007 0.14 0.01 0.006 0.096 0.002 0.0009 0.0060 0.0003 0.00015
1260 GSR-1 1.55 0.05 0.013 3.13 0.06 0.016 5.01 0.07 0.019 0.0928 0.0046 0.00124 0.00053 0.00013 0.00004 0.70 0.70 0.014
1261 GSR-1 1.55 0.05 0.013 3.13 0.06 0.016 5.01 0.07 0.019 0.0928 0.0046 0.00124 0.00053 0.00013 0.00004 0.70 0.70 0.014
1262 GSR-2 5.20 0.070 0.019 3.86 0.070 0.019 1.89 0.050 0.013 0.236 0.0055 0.00148 0.0047 0.0004 0.00012 4.44 4.44 0.032
1266 GSR-4 0.30 0.04 0.011 0.061 0.014 0.0038 0.65 0.03 0.008 0.2223 0.0089 0.00241 0.0029 0.0004 0.00012 1.10 1.10 0.019
1267 GSR-4 0.30 0.04 0.011 0.061 0.014 0.0038 0.65 0.03 0.008 0.2223 0.0089 0.00241 0.0029 0.0004 0.00012 1.10 1.10 0.019
1268 GSR-5 0.60 0.04 0.011 0.35 0.02 0.005 4.16 0.10 0.027 0.1581 0.0078 0.00210 0.0145 0.0009 0.00024 5.95 5.95 0.014
1269 GSR-5 0.60 0.04 0.011 0.35 0.02 0.005 4.16 0.10 0.027 0.1581 0.0078 0.00210 0.0145 0.0009 0.00024 5.95 5.95 0.014
1270 GSR-6 35.67 0.25 0.067 0.08 0.014 0.78 0.04 0.011 0.0518 0.0071 0.00192 0.0047 0.0009 0.00024 34.1 34.1 0.05
1271 GSR-6 35.67 0.25 0.067 0.08 0.014 0.78 0.04 0.011 0.0518 0.0071 0.00192 0.0047 0.0009 0.00024 34.1 34.1 0.05
1272 SARM-50 10.80 0.21 0.121 2.30 0.21 0.121 0.61 0.015 0.009 0.15 0.027 0.016 0.0522 0.00014
1274 SARM-6 0.28 0.0896 0.052 0.04 0.0256 0.015 0.01 0.0058 0.003 0.009 5E-05 2.6E-05 0.42 0.0021 0.001
1276 SARM-3 3.22 0.2705 0.156 8.37 0.5273 0.304 5.51 0.1873 0.108 0.060 0.0003 0.0002
1278 SARM-12 1.09 0.03 0.014 0.0123 0.0016 0.0008 0.0130 0.0010 0.00048 0.109 0.002 0.0011 0.0031 0.0003 0.00015
1280 SARM-8 0.26 0.02 0.010 0.009 0.0014 48.97 0.05 0.025
1282 BXGO-1 0.02 0.01 0.005 0.006 0.0014 0.004 0.0014 0.0420 0.0029 0.00145 0.0096 0.0017 0.00085 33.26 33.26 0.205
1284 BXMG-4 0.03 0.03 0.015 0.04 0.014 0.773 0.029 0.0145 0.1440 0.0036 0.00180 0.002 0.001 0.0005 26.30 26.30 0.300
1285 BXMG-4 0.03 0.03 0.015 0.04 0.014 0.773 0.029 0.0145 0.1440 0.0036 0.00180 0.002 0.001 0.0005 26.30 26.30 0.300
174
Appendix D
Table D1: Continued

Fusion
ID
1286 BXPA-4 0.01 0.01 0.005 0.02 0.014 0.009 0.004 0.0020 0.0267 0.0036 0.00180 0.0074 0.0015 0.00075 29.72 29.72 0.130
1292 BCS-375 0.89 0.045 0.015 10.4 0.09 0.03 0.79 0.047 0.016 0.39 0.39 0.020
1293 BCS-375 0.89 0.045 0.015 10.4 0.09 0.03 0.79 0.047 0.016 0.39 0.39 0.020
1294 BCS-376 0.54 0.031 0.010 2.83 0.060 0.020 11.2 0.05 0.016
1295 BCS-376 0.54 0.031 0.010 2.83 0.060 0.020 11.2 0.05 0.016
1298 BCS-313/1 0.006 0.001 0.0004 0.003 0.001 0.0004 0.005 0.002 0.0008 0.0001 0.00014 0.08 0.08 0.014
1299 BCS-313/1 0.006 0.001 0.0004 0.003 0.001 0.0004 0.005 0.002 0.0008 0.0001 0.00014 0.08 0.08 0.014
1300 BCS-314 1.81 0.034 0.011 0.05 0.007 0.002 0.09 0.011 0.004 0.10 0.10 0.014
1304 BCS-267 1.75 0.061 0.020 0.06 0.005 0.002 0.14 0.025 0.009
1305 BCS-267 1.75 0.061 0.020 0.06 0.005 0.002 0.14 0.025 0.009
1310 BCS-348 0.17 0.01 0.004 0.34 0.02 0.008 2.23 0.03 0.013 0.071 0.009 0.0037 0.016 0.004 0.0014 11.8 11.8 0.04
1311 BCS-348 0.17 0.01 0.004 0.34 0.02 0.008 2.23 0.03 0.013 0.071 0.009 0.0037 0.016 0.004 0.0014 11.8 11.8 0.04
1314 BCS-394 0.08 0.022 0.008 0.02 0.0076 0.003 0.02 0.0063 0.002 0.22 0.022 0.008 0.07 0.014 0.40 0.40 0.034
1318 BCS-301/1 22.6 0.2 0.082 0.07 0.02 0.008 0.32 0.02 0.008 0.802 0.0458 0.01783 25.77 25.77 0.130
1319 BCS-301/1 22.6 0.2 0.082 0.07 0.02 0.008 0.32 0.02 0.008 0.802 0.0458 0.01783 25.77 25.77 0.130
1320 BCS-381 49.0 0.3 0.127 15.7 0.1 0.04 0.33 0.02 0.00846
1324 BCS-176/2 0.09 0.015 0.005 0.10 0.010 0.003 1.30 0.07 0.022 0.199 0.0077 0.0026
1328 NBS-694 43.6 0.4 0.231 0.86 0.04 0.023 0.51 0.02 0.012 30.2 0.1 0.06
1330 SARM-41 1.50 0.09 0.052 0.93 0.12 0.069 1.39 0.065 0.038 0.053 0.015 0.0087 0.0180 0.00014
1334 NBS-103 0.79 0.030 0.011 0.07 0.015 0.008 36.97 0.051 0.026
1336 NBS-600 0.22 0.02 0.012 0.022 0.007 0.0040 0.23 0.02 0.012 0.039 0.007 0.0040 0.024 0.004 0.0023 20.5 20.5 0.12
1338 NBS-679 0.2278 0.0018 0.00105 0.1758 0.0051 0.00296 2.931 0.057 0.0327
1344 NBS-91 10.49 0.006 8.47 0.006 3.24 0.006 0.023 0.0006
1346 NBS-77a 0.05 0.01 0.006 0.037 0.001 0.0006 0.090 0.005 0.003 0.092 0.001 0.0006 0.22 0.22 0.014
1350 NBS-98b 0.1062 0.0049 0.00979 0.2017 0.0089 0.01779 3.38 0.08 0.169 0.0174 0.0007 0.00146
175
Appendix D
Table D1: Continued

Fusion
ID
1352 NBS-98 0.21 0.006 0.28 0.006 3.17 0.006 0.08 0.00577 0.021 0.0006 7.28 7.28 0.006
1354 NBS-1413 0.74 0.04 0.023 1.75 0.06 0.035 3.94 0.09 0.052
1356 NBS-99a 2.14 0.01 0.006 6.2 0.1 0.06 5.2 0.1 0.058 0.02 0.01 0.00577 0.26 0.26 0.006
1358 FER-2 2.17 0.006 0.51 0.006 1.33 0.006 0.27 0.00577
1360 FER-4 2.23 0.006 0.05 0.006 0.29 0.006 0.13 0.00577
1362 JF-1 0.93 0.06 0.011 3.37 0.18 0.035 9.99 0.22 0.040 0.010 0.005 0.0016
1366 JA-2 6.29 0.19 0.037 3.11 0.12 0.022 1.81 0.06 0.010 0.146 0.014 0.0032
1368 NBS-78a 0.11 0.01 0.006 0.078 0.001 0.0006 1.22 0.01 0.006 1.3 0.1 0.05774 0.42 0.42 0.006
1371 SARM-40 49.77 1.65 0.953 0.05 0.21 0.121 0.03 0.026 0.015 2.05 0.08 0.0462
1372 SARM-40 49.77 1.65 0.953 0.05 0.21 0.121 0.03 0.026 0.015 2.05 0.08 0.0462
1375 BCS-393 55.4 0.2 0.088 0.02 0.01 0.004 43.4 43.4 0.04
1377 NBS-98 0.21 0.006 0.28 0.006 3.17 0.006 0.08 0.00577 0.021 0.0006 7.28 7.28 0.006
176
Appendix D
Table D2: The calculated concentrations (%) of the fusion disks after excluding LOI values in Table D1 and including the flux
components (Li2O and B2O3). These values are entered into the calibration application in SuperQ for regression plots
against the matrix corrected intensities of the measurands.
Fusion Id SiO2 TiO2 Al2O3 Fe2O3 MnO MgO CaO Na2O K2O P2O5 Cr2O3 Li2O B2O3
1252 3.43514 0.162306 0.44069 0.95432 0.017463 2.69552 0.99541 0.05136 0.10683 0.14998 0.019518 21.6794 69.2292
1253 3.43613 0.162353 0.44082 0.95459 0.017468 2.69629 0.99570 0.05138 0.10687 0.15002 0.019523 21.6787 69.2271
1254 1.05305 0.001758 0.01055 0.04571 0.001758 7.75456 0.13185 0.00879 0.00703 0.00352 0.005010 23.3317 67.5759
1255 1.05482 0.001761 0.01057 0.04579 0.001761 7.76763 0.13207 0.00880 0.00704 0.00352 0.005018 23.3308 67.5732
1256 0.28363 0.005827 0.12613 8.64950 0.001334 0.00188 0.00413 0.00139 0.01322 0.00878 0.000548 21.6790 69.2280
1260 6.67765 0.026306 1.22862 0.19621 0.004309 0.03851 0.14212 0.28698 0.45936 0.00851 0.000048 19.7660 71.1316
1261 6.66921 0.026273 1.22707 0.19596 0.004304 0.03846 0.14194 0.28662 0.45878 0.00850 0.000048 19.7684 71.1404
1262 5.76235 0.048995 1.53707 0.46578 0.007413 0.16350 0.49430 0.36692 0.17966 0.02243 0.000445 21.6787 69.2272
1266 8.30783 0.024231 0.32363 0.29605 0.001840 0.00754 0.02758 0.00561 0.05976 0.02044 0.000269 19.7678 71.1381
1267 8.30634 0.024227 0.32358 0.29600 0.001840 0.00754 0.02758 0.00561 0.05975 0.02043 0.000269 19.7686 71.1409
1268 5.70293 0.063440 1.81207 0.73176 0.002151 0.19353 0.05777 0.03370 0.40054 0.01522 0.001393 21.6772 69.2224
1269 5.69815 0.063387 1.81056 0.73115 0.002149 0.19337 0.05772 0.03367 0.40021 0.01521 0.001392 21.6797 69.2303
1270 2.28960 0.045094 0.69378 0.34758 0.007729 0.71585 4.91989 0.01103 0.10758 0.00714 0.000645 23.3316 67.5754
1271 2.28996 0.045101 0.69389 0.34763 0.007731 0.71596 4.92066 0.01104 0.10760 0.00714 0.000645 23.3318 67.5761
1272 4.67638 0.078000 1.38586 0.99768 0.015419 0.68658 0.97954 0.20861 0.05533 0.01360 0.004732 21.6808 69.2339
1274 3.51098 0.001804 0.02704 1.53200 0.019826 3.92102 0.02523 0.00360 0.00090 0.00083 0.037849 21.6788 69.2275
1276 4.89644 0.044853 1.27457 0.92602 0.071951 0.02616 0.30089 0.78212 0.51487 0.00557 21.6796 69.2300
1278 0.03057 0.064062 0.06919 8.50827 0.019605 0.25025 0.09748 0.00110 0.00116 0.00976 0.000274 21.6789 69.2276
1280 0.39299 0.021934 0.96602 1.84630 0.022848 1.34255 0.02376 0.00082 4.475487 21.6794 69.2293
1282 0.08642 0.062838 8.27659 0.63315 0.000902 0.00409 0.00273 0.00082 0.00055 0.00572 0.001309 21.6768 69.2209
177
Appendix D
Table D2: Continued

1284 1.17892 0.219950 6.23976 1.22222 0.051698 0.00495 0.00371 0.00495 0.09563 0.01781 0.000247 21.6774 69.2230
1285 1.17893 0.219951 6.23978 1.22222 0.051698 0.00495 0.00371 0.00495 0.09563 0.01781 0.000247 21.6784 69.2263
1286 0.60194 0.169026 7.41072 0.89949 0.005932 0.00388 0.00129 0.00259 0.00116 0.00346 0.000958 21.6786 69.2266
1292 6.13396 0.034601 1.80964 0.01066 0.00426 0.08141 0.94674 0.07217 21.6777 69.2240
1293 6.13709 0.034618 1.81056 0.01066 0.00427 0.08145 0.94722 0.07221 21.6766 69.2205
1294 6.12006 1.61493 0.00912 0.00308 0.04914 0.25787 1.02450 21.6799 69.2308
1295 6.11857 1.61454 0.00912 0.00308 0.04913 0.25781 1.02425 21.6799 69.2308
1298 9.07777 0.001547 0.00328 0.00109 0.000012 0.00012 0.00055 0.00027 0.00045 0.000009 19.7678 71.1381
1299 9.07874 0.001547 0.00328 0.00109 0.000012 0.00012 0.00055 0.00027 0.00045 0.000009 19.7678 71.1381
1300 8.74351 0.017578 0.07011 0.04778 0.00434 0.16497 0.00475 0.00788 19.7686 71.1409
1304 8.71385 0.015642 0.07757 0.07155 0.013625 0.00545 0.15855 0.00545 0.01298 19.7686 71.1409
1305 8.71431 0.015643 0.07758 0.07155 0.013626 0.00545 0.15856 0.00545 0.01298 19.7686 71.1409
1310 5.24777 0.110912 3.24520 0.10680 0.03081 0.01746 0.03492 0.22901 0.00729 0.001643 19.7684 71.1401
1311 5.25126 0.110985 3.24735 0.10687 0.03083 0.01747 0.03494 0.22916 0.00730 0.001644 19.7674 71.1365
1314 0.45451 0.283842 8.10457 0.17341 0.01095 0.00730 0.00183 0.00183 0.02008 0.006389 21.6794 69.2294
1318 0.91299 0.019740 0.52559 4.20696 0.154222 0.21344 2.78833 0.00864 0.03948 21.6797 69.2303
1319 0.91263 0.019733 0.52538 4.20529 0.154160 0.21336 2.78722 0.00863 0.03947 21.6795 69.2297
1320 0.81744 0.032586 0.06238 1.77035 0.294205 0.09590 4.56203 1.46171 0.030724 21.6775 69.2232
1324 0.26817 0.031799 0.54695 1.03961 6.506615 0.00424 0.00954 0.01060 0.13737 0.02113 23.3321 67.5769
1328 1.09821 0.17650 0.07746 0.001137 0.03236 4.27519 0.08433 0.05001 2.96126 21.6770 69.2218
1330 5.89817 0.057244 1.40507 0.44025 0.006245 0.84304 0.15612 0.09679 0.14467 0.00520 0.001871 21.6791 69.2284
1334 0.75120 0.084822 1.89994 1.45883 0.018850 1.48370 0.07240 0.00616 3.371810 21.6795 69.2297
1336 2.33249 0.150520 4.59603 1.95331 0.001494 0.00575 0.02528 0.00253 0.02643 0.00448 0.002758 21.6796 69.2298
1338 5.11969 0.094638 2.04560 1.27226 0.021965 0.12311 0.02240 0.01728 0.28818 19.7673 71.1363
178
Appendix D
Table D2: Continued

1344 6.22249 0.001752 0.55403 0.00728 0.96702 0.78081 0.29868 0.00212 21.6782 69.2256
1346 3.18943 0.242397 5.48583 0.09113 0.03463 0.00456 0.00337 0.00820 0.008384 21.6788 69.2273
1350 5.61280 0.132861 2.66029 0.16610 0.001475 0.05844 0.01046 0.01985 0.33326 0.001712 19.7691 71.1429
1352 5.78999 0.140073 2.50171 0.20080 0.000490 0.07053 0.02057 0.02743 0.31051 0.00784 0.002057 19.7686 71.1409
1354 7.61077 0.010115 0.91031 0.02207 0.00552 0.06804 0.16091 0.36229 19.7682 71.1395
1356 5.94731 0.000638 1.86936 0.00593 0.00182 0.19514 0.56537 0.47418 0.00182 21.6793 69.2290
1358 4.45143 0.016282 0.46676 3.57035 0.010855 0.18996 0.19629 0.04613 0.12031 0.02442 19.7692 71.1433
1360 4.72965 0.006612 0.16058 3.77563 0.017948 0.13319 0.21065 0.00472 0.02739 0.01228 21.6791 69.2285
1362 6.07692 0.000456 1.64748 0.00729 0.000091 0.00055 0.08474 0.30708 0.91031 0.00091 21.6793 69.2290
1366 5.18147 0.060613 1.41522 0.57031 0.009918 0.69796 0.57766 0.28561 0.16623 0.01341 21.6775 69.2232
1368 1.77073 0.293904 6.54439 0.10953 0.06389 0.01004 0.00712 0.11135 0.11866 21.6793 69.2290
1371 0.46184 0.007497 0.06148 0.41236 0.026991 0.29540 7.46289 0.00750 0.00450 0.30739 23.3308 67.5733
1372 0.46134 0.007489 0.06141 0.41191 0.026961 0.29508 7.45483 0.00749 0.00449 0.30706 23.3330 67.5795
1375 0.11040 0.001419 0.01893 0.00710 0.001577 0.02366 8.73707 0.00315 23.3313 67.5747
1377 5.79194 0.140120 2.50256 0.20087 0.000490 0.07055 0.02058 0.02744 0.31061 0.00784 0.002058 21.6789 69.2276
179
Appendix E
Appendix E
In Appendix E, each standard fusion disk’s result and calculated uncertainty is
compared with its CRM concentration value and associated standard to evaluate the
trueness of the test method using En and Recovery values. The ‘Included’-column
indicates whether the CRM was used in the calibration. If not, it is mainly due to its
value not being certified but assigned. (Ci,CRM = Certified or assigned concentration
value for CRM or RM)
Table E1: Trueness of the test method for SiO2.

Result
Fusion CSiO2,CRM
u(CSiO2,CRM) (%) for u(SiO2) En Recovery Included
ID (%)
SiO2
1252 33.44 0.150 33.87 0.232 0.8 101.3 Yes
1253 33.44 0.150 33.55 0.232 0.2 100.3 Yes
1254 5.99 0.260 6.11 0.088 0.2 101.9 Yes
1255 5.99 0.260 5.99 0.088 0.0 100.0 Yes
1256 3.10 0.01 3.09 0.074 0.1 99.8 No
1260 72.83 0.027 73.48 0.488 0.7 100.9 Yes
1261 72.83 0.03 73.18 0.491 0.4 100.5 Yes
1262 60.6 0.038 61.28 0.412 0.8 101.1 Yes
1266 90.36 0.040 90.87 0.643 0.4 100.6 Yes
1267 90.36 0.040 90.87 0.649 0.4 100.6 Yes
1268 59.23 0.043 59.56 0.417 0.4 100.6 Yes
1269 59.23 0.043 59.59 0.423 0.4 100.6 Yes
1270 16.60 0.016 15.60 0.235 2.1 94.0 Yes
1271 16.60 0.016 15.65 0.241 2.0 94.3 Yes
1272 51.56 0.242 50.66 0.405 1.0 98.2 Yes
1274 38.96 0.180 38.45 0.372 0.6 98.7 Yes
1276 52.40 0.151 54.15 0.431 1.9 103.3 Yes
1278 0.34 0.005 0.359 0.252 0.0 104.9 No
1280 4.30 0.020 4.51 0.313 0.3 104.8 Yes
1282 0.634 0.0230 0.674 0.186 0.1 106.2 Yes
1284 9.53 0.070 9.72 0.227 0.4 102.0 Yes
1285 9.53 0.070 9.60 0.236 0.1 100.7 Yes
1286 4.651 0.0395 4.63 0.227 0.0 99.6 Yes
1292 67.1 0.03 66.92 0.594 0.2 99.7 Yes
1293 67.1 0.03 66.98 0.605 0.1 99.8 Yes
1294 67.1 0.05 67.41 0.619 0.2 100.5 Yes
180
Appendix E
Table E1: Continued

Result
Fusion CSiO2,CRM
u(CSiO2,CRM) (%) for u(SiO2) En Recovery Included
ID (%)
SiO2
1295 67.1 0.05 67.09 0.640 0.0 100.0 Yes
1298 99.78 0.025 99.89 0.893 0.1 100.1 Yes
1299 99.78 0.025 99.81 0.909 0.0 100.0 Yes
1300 96.2 0.03 96.20 0.899 0.0 100.0 Yes
1304 95.9 0.05 96.52 0.886 0.3 100.6 Yes
1305 95.9 0.05 95.83 0.922 0.1 99.9 Yes
1310 51.1 0.09 51.70 0.570 0.5 101.2 Yes
1311 51.1 0.09 51.65 0.580 0.5 101.1 Yes
1314 4.98 0.029 4.99 0.507 0.0 100.2 Yes
1318 7.40 0.012 7.48 0.399 0.1 101.0 Yes
1319 7.40 0.012 7.44 0.409 0.1 100.6 Yes
1320 8.78 0.038 8.55 0.543 0.2 97.3 Yes
1324 2.53 0.016 2.68 0.517 0.1 106.1 Yes
1328 11.2 0.23 11.89 0.573 0.6 106.2 Yes
1330 56.67 0.525 56.79 0.715 0.1 100.2 Yes
1334 8.24 0.009 8.39 0.651 0.1 101.9 Yes
1336 20.3 0.23 19.98 0.529 0.3 98.4 Yes
1338 52.07 0.371 52.73 0.734 0.4 101.3 No
1344 67.5 0.06 69.05 0.931 0.8 102.3 Yes
1346 35.0 0.06 35.05 0.732 0.0 100.2 Yes
1350 57.01 0.685 57.72 0.802 0.3 101.3 No
1352 59.11 0.006 59.50 0.828 0.2 100.7 Yes
1354 82.77 0.231 82.38 1.028 0.2 99.5 Yes
1356 65.22 0.029 65.34 0.943 0.1 100.2 Yes
1358 49.21 0.006 48.20 0.884 0.6 97.9 Yes
1360 50.07 0.006 49.67 0.866 0.2 99.2 Yes
1362 66.69 0.148 67.11 0.996 0.2 100.6 Yes
1366 56.42 0.124 57.20 0.938 0.4 101.4 Yes
1368 19.4 0.06 19.40 0.826 0.0 100.0 Yes
1371 3.08 0.104 3.07 0.496 0.0 99.6 Yes
1372 3.08 0.104 3.05 0.504 0.0 99.0 Yes
1375 0.70 0.009 0.729 0.490 0.0 104.1 Yes
1377 59.11 0.006 59.11 1.00 0.0 100.0 Yes
181
Appendix E
Table E2: Trueness of the test method for TiO2.
Result
Fusion CTiO2,CRM
u(CTiO2,CRM) (%) for u(TiO2) En Recovery Included
ID (%)
TiO2
1252 1.58 0.098 1.66 0.012 0.4 105.2 Yes
1253 1.58 0.098 1.64 0.013 0.3 104.1 Yes
1254 0.01 0.005 0.005 0.001 0.5 51.3 No
1255 0.01 0.005 0.004 0.001 0.6 41.0 No
1256 0.064 0.0009 0.068 0.002 0.9 107.2 No
1260 0.287 0.0031 0.291 0.004 0.4 101.5 No
1261 0.287 0.0031 0.291 0.004 0.5 101.6 No
1262 0.52 0.004 0.512 0.005 0.2 99.4 No
1266 0.264 0.0036 0.263 0.004 0.1 99.7 No
1267 0.264 0.0036 0.267 0.004 0.3 101.2 No
1268 0.659 0.0058 0.660 0.006 0.1 100.2 No
1269 0.659 0.0058 0.663 0.006 0.3 100.7 No
1270 0.327 0.0040 0.324 0.005 0.3 99.0 No
1271 0.327 0.0040 0.328 0.005 0.1 100.5 No
1272 0.86 0.014 0.877 0.008 0.5 102.0 Yes
1274 0.0200 5.8E-05 0.020 0.002 0.1 97.9 No
1276 0.48 0.001 0.497 0.005 1.6 103.6 Yes
1278 0.717 0.0025 0.739 0.008 1.3 103.1 No
1280 0.24 0.005 0.229 0.010 0.5 95.6 Yes
1282 0.461 0.008 0.433 0.005 1.5 93.9 Yes
1284 1.778 0.0150 1.79 0.014 0.4 100.9 Yes
1285 1.778 0.0150 1.77 0.014 0.3 99.3 Yes
1286 1.306 0.0130 1.25 0.011 1.5 96.0 Yes
1292 0.38 0.007 0.372 0.005 0.4 98.3 Yes
1293 0.38 0.007 0.373 0.005 0.3 98.5 Yes
1294 <0.02 0.007 0.001 No
1295 <0.02 0.007 0.001 No
1298 0.017 0.00106 0.024 0.002 1.7 138.3 Yes
1299 0.017 0.00106 0.023 0.002 1.6 136.8 Yes
1300 0.19 0.005 0.201 0.003 0.6 103.8 Yes
1304 0.17 0.003 0.167 0.003 0.6 97.0 Yes
1305 0.17 0.003 0.170 0.003 0.2 98.9 Yes
1310 1.08 0.009 1.09 0.010 0.5 101.2 Yes
1311 1.08 0.009 1.10 0.010 0.6 101.6 Yes
1314 3.11 0.013 3.12 0.025 0.1 100.2 Yes
1318 0.16 0.004 0.162 0.003 0.2 101.3 Yes
182
Appendix E
Table E2: Continued

Result
Fusion CTiO2,CRM
u(CTiO2,CRM) (%) for u(TiO2) En Recovery Included
ID (%)
TiO2
1319 0.16 0.004 0.162 0.003 0.2 101.2 Yes
1320 0.35 0.013 0.334 0.005 0.6 95.4 Yes
1324 0.30 0.006 0.315 0.008 0.8 105.1 Yes
1328 0.127 0.003 No
1330 0.55 0.020 0.554 0.006 0.1 100.7 Yes
1334 0.93 0.006 0.942 0.026 0.2 101.3 Yes
1336 1.31 0.023 1.28 0.012 0.6 97.5 Yes
1338 0.96 0.032 0.988 0.010 0.4 102.7 No
1344 0.019 0.0006 0.026 0.002 1.9 137.5 Yes
1346 2.66 0.029 2.68 0.024 0.3 100.8 Yes
1350 1.35 0.040 1.38 0.013 0.3 102.2 No
1352 1.43 0.006 1.46 0.014 1.0 102.1 Yes
1354 0.11 0.012 0.114 0.002 0.2 103.6 Yes
1356 0.007 0.0006 0.008 0.001 0.4 116.6 Yes
1358 0.18 0.058 0.185 0.003 0.0 102.7 Yes
1360 0.07 0.058 0.071 0.002 0.0 101.4 Yes
1362 0.005 0.0007 0.010 0.001 1.5 190.6 Yes
1366 0.66 0.008 0.672 0.008 0.6 101.9 Yes
1368 3.22 0.029 3.27 0.032 0.6 101.6 Yes
1371 0.045 0.0058 0.045 0.001 0.0 99.2 Yes
1372 0.045 0.0058 0.043 0.001 0.1 96.2 Yes
1375 0.009 0.0008 0.011 0.001 0.9 124.0 Yes
1377 1.43 0.006 1.46 0.016 0.8 102.0 Yes
183
Appendix E
Table E3: Trueness of the test method for Al2O3.
Result
Fusion CAl2O3,CRM
u(CAl2O3,CRM) (%) for u(Al2O3) En Recovery Included
ID (%)
Al2O3
1252 4.29 0.191 4.26 0.035 0.1 99.4 Yes
1253 4.29 0.191 4.23 0.035 0.2 98.6 Yes
1254 0.06 0.023 0.066 0.011 0.1 110.0 No
1255 0.06 0.023 0.071 0.012 0.2 118.6 No
1256 1.38 0.009 1.38 0.026 0.1 100.3 No
1260 13.40 0.019 13.46 0.079 0.3 100.4 Yes
1261 13.40 0.019 13.40 0.079 0.0 100.0 Yes
1262 16.17 0.032 16.28 0.092 0.6 100.7 Yes
1266 3.52 0.024 3.50 0.032 0.2 99.4 Yes
1267 3.52 0.024 3.49 0.033 0.4 99.2 Yes
1268 18.82 0.038 19.16 0.108 1.5 101.8 Yes
1269 18.82 0.038 19.10 0.109 1.2 101.5 Yes
1270 5.03 0.022 4.90 0.064 1.0 97.4 Yes
1271 5.03 0.022 4.93 0.064 0.7 98.1 Yes
1272 15.28 0.173 15.29 0.093 0.0 100.0 Yes
1274 0.3 0.08 0.24 0.023 0.4 78.7 No
1276 13.64 0.197 13.98 0.084 0.8 102.5 Yes
1278 0.77 0.009 0.812 0.024 0.7 104.8 No
1280 10.57 0.055 10.58 0.422 0.0 100.1 Yes
1282 60.72 0.180 61.08 0.522 0.3 100.6 Yes
1284 50.44 0.250 50.93 0.336 0.6 101.0 Yes
1285 50.44 0.250 51.28 0.339 1.0 101.7 Yes
1286 57.26 0.185 57.47 0.435 0.2 100.4 Yes
1292 19.8 0.02 20.08 0.128 1.0 101.4 Yes
1293 19.8 0.02 20.05 0.129 0.9 101.2 Yes
1294 17.7 0.01 18.06 0.120 1.5 102.0 Yes
1295 17.7 0.01 17.85 0.122 0.6 100.8 Yes
1298 0.036 0.0013 0.042 0.022 0.1 116.9 Yes
1299 0.036 0.0013 0.063 0.022 0.6 174.7 Yes
1300 0.77 0.004 0.780 0.024 0.2 101.1 Yes
1304 0.85 0.006 0.881 0.024 0.6 103.1 Yes
1305 0.85 0.006 0.889 0.025 0.7 104.1 Yes
1310 31.6 0.09 31.62 0.210 0.0 100.1 Yes
1311 31.6 0.09 31.75 0.213 0.3 100.5 Yes
1314 88.8 0.04 88.77 0.605 0.0 100.0 Yes
1318 4.26 0.020 4.31 0.046 0.5 101.2 Yes
184
Appendix E
Table E3: Continued

Result
Fusion CAl2O3,CRM
u(CAl2O3,CRM) (%) for u(Al2O3) En Recovery Included
ID (%)
Al2O3
1319 4.26 0.020 4.27 0.046 0.1 100.3 Yes
1320 0.67 0.021 0.718 0.024 0.8 107.2 Yes
1324 5.16 0.059 5.47 0.124 1.1 106.0 Yes
1328 1.8 0.06 1.95 0.032 1.1 108.3 Yes
1330 13.50 0.289 13.49 0.107 0.0 99.9 Yes
1334 20.83 0.022 20.61 0.555 0.2 98.9 Yes
1336 40.0 0.23 39.77 0.308 0.3 99.4 Yes
1338 20.80 0.371 20.65 0.159 0.2 99.3 No
1344 6.0 0.06 6.14 0.060 0.8 102.2 Yes
1346 60.2 0.06 60.32 0.469 0.1 100.2 Yes
1350 27.02 0.218 27.13 0.217 0.2 100.4 No
1352 25.5 0.06 25.34 0.206 0.5 99.2 Yes
1354 9.90 0.075 9.77 0.088 0.5 98.7 Yes
1356 20.5 0.06 20.56 0.177 0.2 100.3 Yes
1358 5.16 0.006 5.15 0.053 0.1 99.9 Yes
1360 1.7 0.06 1.64 0.027 0.4 96.7 Yes
1362 18.08 0.062 18.45 0.166 1.0 102.0 Yes
1366 15.41 0.069 15.75 0.141 1.1 102.2 Yes
1368 71.7 0.06 71.93 0.635 0.2 100.3 Yes
1371 0.41 0.049 0.417 0.016 0.1 101.6 Yes
1372 0.41 0.049 0.408 0.016 0.0 99.5 Yes
1375 0.12 0.004 0.143 0.013 0.8 119.0 Yes
1377 25.5 0.058 25.52 0.241 0.0 99.9 Yes
185
Appendix E
Table E4: Trueness of the test method for Fe2O3.
Result (%)
Fusion CFe2O3,CRM
u(CFe2O3,CRM) for u(Fe2O3) En Recovery Included
ID (%)
Fe2O3
1252 9.29 0.179 9.40 0.078 0.3 101.2 Yes
1253 9.29 0.179 9.45 0.078 0.4 101.7 Yes
1254 0.26 0.035 0.246 0.005 0.2 94.6 Yes
1255 0.26 0.035 0.247 0.005 0.2 94.9 Yes
1256 94.6 0.014 94.60 0.917 0.0 100.0 No
1260 2.14 0.016 2.12 0.019 0.4 99.1 Yes
1261 2.14 0.016 2.15 0.020 0.2 100.5 Yes
1262 4.90 0.016 4.84 0.040 0.7 98.8 Yes
1266 3.22 0.019 3.16 0.028 0.8 98.2 Yes
1267 3.22 0.019 3.22 0.029 0.0 99.9 Yes
1268 7.60 0.024 7.63 0.062 0.2 100.4 Yes
1269 7.60 0.024 7.71 0.063 0.8 101.5 Yes
1270 2.52 0.019 2.51 0.035 0.1 99.8 Yes
1271 2.52 0.019 2.55 0.036 0.4 101.3 Yes
1272 11.0 0.17 11.24 0.093 0.6 102.2 Yes
1274 17.00 0.236 17.24 0.153 0.4 101.4 Yes
1276 9.91 0.206 10.13 0.083 0.5 102.3 Yes
1278 95.3 0.04 95.56 0.944 0.2 100.3 No
1280 20.2 0.05 19.84 0.797 0.2 98.2 No
1282 4.645 0.0490 4.54 0.048 0.7 97.8 Yes
1284 9.88 0.065 9.92 0.088 0.2 100.4 Yes
1285 9.88 0.065 9.80 0.087 0.4 99.2 Yes
1286 6.95 0.055 6.79 0.066 1.0 97.6 Yes
1292 0.12 0.004 0.118 0.003 0.1 100.9 Yes
1293 0.12 0.004 0.122 0.003 0.5 104.4 Yes
1294 0.10 0.002 0.105 0.003 0.7 104.9 Yes
1295 0.10 0.002 0.106 0.003 0.9 106.1 Yes
1298 0.012 0.0004 0.030 0.002 4.1 252.3 Yes
1299 0.012 0.0004 0.031 0.002 4.2 261.7 Yes
1300 0.53 0.015 0.515 0.007 0.3 98.1 Yes
1304 0.79 0.008 0.781 0.010 0.3 99.1 Yes
1305 0.79 0.008 0.814 0.010 1.0 103.3 Yes
1310 1.04 0.009 1.06 0.011 0.8 102.3 Yes
1311 1.04 0.009 1.06 0.011 0.7 101.9 Yes
1314 1.90 0.011 1.91 0.020 0.3 100.6 Yes
1318 34.10 0.023 35.00 0.385 1.2 102.6 No
186
Appendix E
Table E4: Continued

Result (%)
Fusion CFe2O3,CRM
u(CFe2O3,CRM) for u(Fe2O3) En Recovery Included
ID (%)
Fe2O3
1319 34.10 0.023 34.96 0.386 1.1 102.5 No
1320 19.02 0.060 19.15 0.236 0.3 100.7 No
1324 9.81 0.051 10.79 0.248 1.9 110.0 No
1328 0.79 0.035 0.790 0.011 0.0 100.0 Yes
1330 4.23 0.075 4.27 0.043 0.2 101.0 Yes
1334 15.99 0.039 15.87 0.436 0.1 99.2 Yes
1336 17.0 0.17 17.28 0.168 0.6 101.7 Yes
1338 12.94 0.173 13.25 0.127 0.7 102.4 No
1344 0.079 0.0006 0.084 0.003 0.9 106.6 Yes
1346 1.00 0.029 1.00 0.012 0.0 100.0 Yes
1350 1.69 0.029 1.71 0.019 0.4 101.4 No
1352 2.05 0.006 2.06 0.022 0.2 100.4 Yes
1354 0.24 0.006 0.266 0.005 1.8 111.0 Yes
1356 0.065 0.02887 0.069 0.003 0.1 106.7 Yes
1358 39.4697 0.058 40.17 0.423 0.8 101.8 Yes
1360 39.9704 0.058 40.61 0.427 0.7 101.6 Yes
1362 0.08 0.003 0.087 0.003 0.8 108.4 Yes
1366 6.21 0.050 6.33 0.068 0.7 101.9 Yes
1368 1.2 0.06 1.28 0.016 0.7 106.5 Yes
1371 2.75 0.081 2.76 0.054 0.0 100.3 Yes
1372 2.75 0.081 2.74 0.053 0.1 99.5 Yes
1375 0.045 0.0022 0.055 0.002 1.8 122.0 Yes
1377 2.05 0.006 2.06 0.025 0.1 100.4 Yes
187
Appendix E
Table E5: Trueness of the test method for MnO.
Result (%)
Fusion CMnO,CRM
u(CMnO,CRM) for u(MnO) En Recovery Included
ID (%)
MnO
1252 0.17 0.006 0.172 0.003 0.1 100.9 Yes
1253 0.17 0.006 0.175 0.004 0.4 103.0 Yes
1254 0.01 0.003 0.010 0.001 0.0 103.0 No
1255 0.01 0.003 0.011 0.001 0.1 108.1 No
1256 0.0146 0.00019 0.012 0.003 0.5 84.0 No
1260 0.047 0.0006 0.061 0.003 2.7 No
1261 0.047 0.0006 0.062 0.003 2.9 No
1262 0.078 0.0006 0.079 0.003 0.1 101.1 No
1266 0.020 0.0002 0.021 0.002 0.3 106.1 No
1267 0.020 0.0002 0.022 0.002 0.4 110.0 No
1268 0.022 0.0004 0.024 0.002 0.3 105.3 No
1269 0.022 0.0004 0.025 0.002 0.5 109.8 No
1270 0.056 0.0009 0.060 0.002 0.9 106.3 No
1271 0.056 0.0009 0.058 0.002 0.4 103.0 No
1272 0.17 0.009 0.182 0.004 0.6 106.9 Yes
1274 0.22 0.001 0.224 0.004 0.5 102.0 Yes
1276 0.77 0.002 0.762 0.008 0.5 99.0 Yes
1278 0.22 0.002 0.219 0.004 0.0 99.9 No
1280 0.25 0.005 0.379 0.016 3.9 151.7 Yes
1282 0.0066 0.00055 0.006 0.001 0.3 88.3 No
1284 0.4179 0.00858 0.301 0.004 6.1 No
1285 0.4179 0.00858 0.295 0.004 6.4 No
1286 0.0458 0.00135 0.033 0.002 2.7 No
1292 0.005 0.002 No
1293 0.006 0.002 No
1294 0.005 0.002 No
1295 0.004 0.002 No
1298 0.0001 1.225E-05 <0.002 Yes
1299 0.0001 1.225E-05 <0.002 Yes
1300 <0.01 0.006 0.002 No
1304 0.15 0.006 0.166 0.004 1.1 110.9 Yes
1305 0.15 0.006 0.176 0.004 1.7 117.1 Yes
1310 0.004 0.002 No
1311 0.006 0.002 No
1314 0.011 0.002 No
1318 1.25 0.016 1.29 0.014 0.8 102.8 Yes
188
Appendix E
Table E5: Continued

Result (%)
Fusion CMnO,CRM
u(CMnO,CRM) for u(MnO) En Recovery Included
ID (%)
MnO
1319 1.25 0.016 1.30 0.014 1.1 103.9 Yes
1320 3.16 0.034 3.15 0.037 0.1 99.8 No
1324 61.4 0.04 65.19 2.47 0.8 106.2 No
1328 0.0116 0.00069 0.013 0.002 0.3 111.0 Yes
1330 0.06 0.008 0.059 0.003 0.1 98.6 Yes
1334 0.21 0.007 0.332 0.010 5.1 Yes
1336 0.013 0.0023 0.014 0.002 0.1 105.5 Yes
1338 0.2234 0.0001 0.234 0.004 1.2 104.8 No
1344 0.008 0.002 No
1346 0.004 0.002 No
1350 0.0150 0.0012912 0.014 0.002 0.1 96.3 No
1352 0.005 0.00058 0.008 0.002 0.7 157.4 Yes
1354 0.005 0.002 No
1356 0.003 0.002 No
1358 0.12 0.006 0.125 0.003 0.4 104.3 Yes
1360 0.19 0.006 0.197 0.004 0.5 103.4 Yes
1362 0.001 3.78E-05 0.001 0.002 0.1 132.6 Yes
1366 0.108 0.0013472 0.112 0.003 0.5 103.3 Yes
1368 0.005 0.002 No
1371 0.18 0.006 0.188 0.004 0.6 104.6 Yes
1372 0.18 0.006 0.190 0.004 0.7 105.7 Yes
1375 0.010 0.0004087 0.012 0.001 1.1 124.2 Yes
1377 0.005 0.00058 0.009 0.002 0.9 170.5 Yes
189
Appendix E
Table E6: Trueness of the test method for MgO.
Result (%)
Fusion CMgO,CRM
u(CMgO,CRM) for u(MgO) En Recovery Included
ID (%)
MgO
1252 26.24 0.450 26.51 0.278 0.3 101.0 Yes
1253 26.24 0.450 26.62 0.278 0.4 101.5 Yes
1254 44.11 0.393 44.68 0.892 0.3 101.3 Yes
1255 44.11 0.393 44.92 0.889 0.4 101.8 Yes
1256 0.0206 0.00019 0.022 0.005 0.2 107.9 No
1260 0.42 0.011 0.351 0.006 2.9 83.5 Yes
1261 0.42 0.011 0.355 0.006 2.7 84.6 Yes
1262 1.72 0.016 1.61 0.016 2.4 93.5 Yes
1266 0.082 0.0054 <0.02 Yes
1267 0.082 0.0054 <0.02 Yes
1268 2.01 0.013 1.97 0.020 0.7 98.2 Yes
1269 2.01 0.013 1.97 0.020 0.9 97.8 Yes
1270 5.19 0.032 5.26 0.077 0.4 101.4 Yes
1271 5.19 0.032 5.18 0.076 0.1 99.8 Yes
1272 7.57 0.104 7.49 0.072 0.3 98.9 Yes
1274 43.51 0.452 43.88 0.525 0.3 100.9 Yes
1276 0.28 0.055 0.200 0.005 0.7 71.4 Yes
1278 2.80 0.019 2.848 0.032 0.6 101.7 No
1280 14.69 0.030 15.46 0.627 0.6 105.2 Yes
1282 0.03 0.010 0.067 0.003 1.8 No
1284 0.04 0.010 0.065 0.003 1.2 No
1285 0.04 0.010 0.055 0.003 0.7 137.1 No
1286 0.03 0.010 0.046 0.003 0.7 151.7 No
1292 0.05 0.007 <0.02 No
1293 0.05 0.007 <0.02 No
1294 0.03 0.004 <0.02 No
1295 0.03 0.004 <0.02 No
1298 0.0013 0.00082 <0.02 Yes
1299 0.0013 0.00082 <0.02 Yes
1300 0.05 0.003 <0.02 Yes
1304 0.06 0.009 <0.02 Yes
1305 0.06 0.009 <0.02 Yes
1310 0.30 0.004 0.298 0.005 0.2 99.2 Yes
1311 0.30 0.004 0.290 0.005 0.8 96.5 Yes
1314 0.12 0.006 0.204 0.005 5.3 169.9 Yes
1318 1.73 0.016 1.71 0.021 0.4 98.8 Yes
190
Appendix E
Table E6: Continued

Result (%)
Fusion CMgO,CRM
u(CMgO,CRM) for u(MgO) En Recovery Included
ID (%)
MgO
1319 1.73 0.016 1.72 0.021 0.3 99.2 Yes
1320 1.03 0.030 0.915 0.013 1.8 88.9 Yes
1324 0.04 0.001 0.093 0.005 5.3 232.5 Yes
1328 0.33 0.012 0.338 0.006 0.3 102.5 Yes
1330 8.10 0.121 8.13 0.089 0.1 100.4 Yes
1334 16.27 0.021 16.33 0.458 0.1 100.4 Yes
1336 0.05 0.006 0.071 0.004 1.5 142.7 Yes
1338 1.252 0.0066 1.26 0.015 0.3 100.9 No
1344 0.031 0.004 No
1346 0.38 0.006 0.416 0.007 2.0 No
1350 0.594 0.0310 0.602 0.008 0.1 101.4 No
1352 0.72 0.006 0.653 0.009 3.2 90.7 Yes
1354 0.06 0.006 0.013 0.004 3.3 21.5 Yes
1356 0.02 0.006 <0.02 Yes
1358 2.1 0.058 2.10 0.025 0.0 99.8 Yes
1360 1.41 0.006 1.33 0.017 2.1 94.7 Yes
1362 0.006 0.0003 <0.02 Yes
1366 7.60 0.054 7.76 0.091 0.7 102.0 Yes
1368 0.70 0.006 0.758 0.011 2.4 108.3 Yes
1371 1.97 0.167 1.86 0.037 0.3 94.4 Yes
1372 1.97 0.167 1.88 0.038 0.3 95.6 Yes
1375 0.15 0.004 0.152 0.004 0.2 101.3 Yes
1377 0.72 0.006 0.647 0.010 3.3 89.8 Yes
191
Appendix E
Table E7: Trueness of the test method for CaO.
Result (%)
Fusion CCaO,CRM
ID
u(CCaO,CRM) for u(CaO) En Recovery Included
(%)
CaO
1252 9.69 0.121 9.54 0.182 0.4 98.4 Yes
1253 9.69 0.121 9.58 0.183 0.2 98.9 Yes
1254 0.75 0.087 0.775 0.018 0.1 103.3 Yes
1255 0.75 0.087 0.777 0.018 0.2 103.6 Yes
1256 0.0452 0.00154 0.039 0.002 1.3 No
1260 1.55 0.013 1.54 0.030 0.2 99.1 Yes
1261 1.55 0.013 1.56 0.030 0.1 100.4 Yes
1262 5.20 0.019 5.17 0.098 0.1 99.4 Yes
1266 0.30 0.011 0.252 0.005 2.0 83.9 Yes
1267 0.30 0.011 0.261 0.006 1.6 87.1 Yes
1268 0.60 0.011 0.594 0.012 0.2 99.0 Yes
1269 0.60 0.011 0.597 0.012 0.1 99.4 Yes
1270 35.67 0.067 36.27 0.798 0.4 101.7 Yes
1271 35.67 0.067 36.20 0.797 0.3 101.5 Yes
1272 10.80 0.121 10.87 0.207 0.1 100.6 Yes
1274 0.28 0.052 0.280 0.006 0.0 100.2 Yes
1276 3.22 0.156 3.13 0.060 0.3 97.3 Yes
1278 1.09 0.014 1.12 0.023 0.6 103.1 No
1280 0.26 0.010 0.208 0.010 1.9 80.2 Yes
1282 0.02 0.005 0.004 0.001 1.6 No
1284 0.03 0.015 0.018 0.001 0.4 60.9 No
1285 0.03 0.015 0.016 0.001 0.5 54.8 No
1286 0.01 0.005 0.003 0.001 0.7 27.8 No
1292 0.89 0.015 0.804 0.016 2.0 90.3 Yes
1293 0.89 0.015 0.805 0.016 2.0 90.3 Yes
1294 0.54 0.010 0.459 0.009 2.8 85.1 Yes
1295 0.54 0.010 0.464 0.010 2.6 86.1 Yes
1298 0.006 0.0004 0.007 0.001 0.3 112.1 Yes
1299 0.006 0.0004 0.007 0.001 0.5 119.3 Yes
1300 1.81 0.011 1.86 0.037 0.5 102.3 Yes
1304 1.75 0.020 1.65 0.033 1.2 94.7 Yes
1305 1.75 0.020 1.70 0.034 0.5 97.6 Yes
1310 0.17 0.004 0.172 0.004 0.2 101.2 Yes
1311 0.17 0.004 0.173 0.004 0.2 101.7 Yes
1314 0.08 0.008 0.080 0.003 0.0 99.8 Yes
1318 22.6 0.082 22.93 0.468 0.3 101.5 Yes
192
Appendix E
Table E7: Continued

Result (%)
Fusion CCaO,CRM
u(CCaO,CRM) for u(CaO) En Recovery Included
ID (%)
CaO
1319 22.6 0.082 22.97 0.470 0.4 101.7 Yes
1320 49.0 0.127 48.67 1.030 0.2 99.3 Yes
1324 0.09 0.005 0.113 0.004 1.8 125.3 Yes
1328 43.6 0.231 44.61 0.954 0.5 102.3 Yes
1330 1.50 0.052 1.50 0.030 0.0 100.1 Yes
1334 0.79 0.011 0.757 0.025 0.7 95.4 Yes
1336 0.22 0.012 0.223 0.005 0.1 101.1 Yes
1338 0.2278 0.00105 0.231 0.005 0.3 101.3 No
1344 10.49 0.006 10.75 0.216 0.6 102.5 Yes
1346 0.05 0.006 0.050 0.002 0.0 100.7 Yes
1350 0.1062 0.00979 0.106 0.003 0.0 100.3 No
1352 0.21 0.006 0.183 0.004 1.8 87.3 Yes
1354 0.74 0.023 0.757 0.016 0.3 102.3 Yes
1356 2.14 0.006 2.08 0.042 0.7 97.0 Yes
1358 2.17 0.006 2.17 0.044 0.1 100.2 Yes
1360 2.23 0.006 2.22 0.045 0.1 99.4 Yes
1362 0.93 0.011 0.915 0.019 0.3 98.4 Yes
1366 6.29 0.037 6.24 0.126 0.2 99.2 Yes
1368 0.11 0.006 0.122 0.003 0.9 110.7 Yes
1371 49.77 0.953 50.44 1.309 0.2 101.4 Yes
1372 49.77 0.953 50.46 1.311 0.2 101.4 Yes
1375 55.4 0.088 57.35 1.471 0.7 103.5 Yes
1377 0.21 0.006 0.186 0.005 1.6 88.7 Yes
193
Appendix E
Table E8: Trueness of the test method for Na2O.
Result (%)
Fusion CNa2O,CRM
u(CNa2O,CRM) for u(Na2O) En Recovery Included
ID (%)
Na2O
1252 0.5 0.11 0.527 0.007 0.1 105.3 No
1253 0.5 0.11 0.511 0.007 0.1 102.3 No
1254 0.05 0.092 <0.02 No
1255 0.05 0.092 <0.02 No
1256 0.015 0.0007 0.035 0.006 1.7 230.4 No
1260 3.13 0.016 3.10 0.020 0.7 98.9 Yes
1261 3.13 0.016 3.08 0.020 1.0 98.3 Yes
1262 3.86 0.019 3.89 0.025 0.5 100.8 Yes
1266 0.061 0.0038 0.271 0.006 15.0 443.8 Yes
1267 0.061 0.0038 0.039 0.005 1.7 64.0 Yes
1268 0.35 0.005 0.401 0.007 3.0 114.6 Yes
1269 0.35 0.005 0.335 0.006 0.9 95.7 Yes
1270 0.08 0.014 0.067 0.004 0.5 83.3 No
1271 0.08 0.014 0.036 0.004 1.5 45.3 No
1272 2.30 0.121 2.50 0.018 0.8 108.8 Yes
1274 0.04 0.015 <0.02 No
1276 8.37 0.304 8.67 0.057 0.5 103.6 Yes
1278 0.012 0.0008 0.149 0.006 11.2 1217.3 No
1280 0.124 0.008 No
1282 0.006 0.0014 <0.02 No
1284 0.04 0.014 0.026 0.004 0.5 65.9 No
1285 0.04 0.014 0.036 0.004 0.2 88.8 No
1286 0.02 0.014 <0.02 No
1292 10.4 0.03 10.45 0.075 0.6 100.9 Yes
1293 10.4 0.03 10.38 0.074 0.1 100.2 Yes
1294 2.83 0.020 2.78 0.021 0.8 98.3 Yes
1295 2.83 0.020 2.82 0.022 0.1 99.7 Yes
1298 0.003 0.0004 0.103 0.006 8.5 3424.4 Yes
1299 0.003 0.0004 <0.02 Yes
1300 0.05 0.002 0.082 0.006 2.5 157.8 Yes
1304 0.06 0.002 0.052 0.005 0.7 86.1 Yes
1305 0.06 0.002 0.057 0.006 0.3 94.6 Yes
1310 0.34 0.008 0.327 0.006 0.6 96.1 Yes
1311 0.34 0.008 0.321 0.006 0.9 94.4 Yes
1314 0.02 0.003 <0.02 Yes
1318 0.07 0.008 0.077 0.005 0.4 109.4 Yes
194
Appendix E
Table E8: Continued

Result (%)
Fusion CNa2O,CRM
u(CNa2O,CRM) for u(Na2O) En Recovery Included
ID (%)
Na2O
1319 0.07 0.008 0.072 0.005 0.1 103.3 Yes
1320 0.193 0.006 No
1324 0.10 0.003 0.110 0.006 0.7 109.9 Yes
1328 0.86 0.023 0.867 0.012 0.1 100.8 Yes
1330 0.93 0.069 1.05 0.010 0.8 112.6 Yes
1334 <0.02 No
1336 0.022 0.0040 0.025 0.005 0.2 113.3 Yes
1338 0.1758 0.00296 0.180 0.006 0.3 102.4 No
1344 8.47 0.006 8.72 0.080 1.5 102.9 Yes
1346 0.037 0.0006 <0.02 Yes
1350 0.2017 0.01779 0.200 0.006 0.0 99.4 No
1352 0.28 0.006 0.279 0.006 0.1 99.5 Yes
1354 1.75 0.035 1.67 0.016 1.1 95.2 Yes
1356 6.2 0.06 6.23 0.057 0.2 100.6 Yes
1358 0.51 0.006 0.487 0.008 1.2 95.4 Yes
1360 0.05 0.006 <0.02 Yes
1362 3.37 0.035 3.29 0.031 0.9 97.6 Yes
1366 3.11 0.022 3.06 0.029 0.7 98.4 Yes
1368 0.078 0.0006 0.081 0.006 0.3 104.0 Yes
1371 0.05 0.121 0.035 0.004 0.1 70.7 Yes
1372 0.05 0.121 0.036 0.004 0.1 71.5 Yes
1375 <0.02 No
1377 0.28 0.006 0.253 0.006 1.6 90.3 Yes
195
Appendix E
Table E9: Trueness of the test method for K2O.
Result (%)
Fusion CK2O,CRM
u(CK2O,CRM) for u(K2O) En Recovery Included
ID (%)
K2O
1252 1.04 0.038 0.987 0.016 0.7 94.9 Yes
1253 1.04 0.038 0.997 0.016 0.5 95.9 Yes
1254 0.04 0.012 <0.002 No
1255 0.04 0.012 <0.002 No
1256 0.14 0.006 0.163 0.003 1.4 113.0 No
1260 5.01 0.019 5.06 0.080 0.3 101.1 Yes
1261 5.01 0.019 5.13 0.081 0.7 102.5 Yes
1262 1.89 0.013 1.85 0.029 0.6 98.0 Yes
1266 0.65 0.008 0.650 0.011 0.0 100.0 Yes
1267 0.65 0.008 0.654 0.011 0.1 100.6 Yes
1268 4.16 0.027 4.31 0.068 1.0 103.7 Yes
1269 4.16 0.027 4.31 0.068 1.0 103.6 Yes
1270 0.78 0.011 0.654 0.013 3.8 83.8 Yes
1271 0.78 0.011 0.630 0.012 4.6 80.7 Yes
1272 0.61 0.009 0.631 0.010 0.8 103.5 Yes
1274 0.01 0.003 0.008 0.001 0.3 81.5 No
1276 5.51 0.108 5.51 0.087 0.0 99.9 Yes
1278 0.0130 0.00048 0.004 0.001 3.6 No
1280 0.020 0.001 No
1282 0.004 0.0014 <0.002 No
1284 0.773 0.0145 0.784 0.013 0.3 101.4 Yes
1285 0.773 0.0145 0.773 0.013 0.0 100.0 Yes
1286 0.009 0.0020 0.008 0.001 0.2 89.3 No
1292 0.79 0.016 0.753 0.012 0.9 95.3 Yes
1293 0.79 0.016 0.753 0.012 0.9 95.3 Yes
1294 11.2 0.016 11.06 0.178 0.5 98.4 Yes
1295 11.2 0.016 11.21 0.181 0.1 99.7 Yes
1298 0.005 0.0008 0.007 0.001 0.8 145.0 Yes
1299 0.005 0.0008 0.006 0.001 0.3 114.3 Yes
1300 0.09 0.004 0.089 0.002 0.3 102.9 Yes
1304 0.14 0.009 0.125 0.002 0.9 87.3 Yes
1305 0.14 0.009 0.131 0.003 0.6 91.5 Yes
1310 2.23 0.013 2.27 0.037 0.5 101.8 Yes
1311 2.23 0.013 2.26 0.037 0.4 101.5 Yes
1314 0.02 0.002 0.021 0.001 0.3 106.9 Yes
1318 0.32 0.008 0.319 0.006 0.1 99.5 Yes
196
Appendix E
Table E9: Continued

Result (%)
Fusion CK2O,CRM
ID (%)
K2O
1319 0.32 0.008 0.317 0.006 0.2 99.0 Yes
1320 0.010 0.001 No
1324 1.30 0.022 1.15 0.031 1.9 89.0 Yes
1328 0.51 0.012 0.532 0.010 0.7 104.4 Yes
1330 1.39 0.038 1.41 0.024 0.3 101.8 Yes
1334 0.043 0.002 No
1336 0.23 0.012 0.234 0.004 0.2 101.7 Yes
1338 2.931 0.0327 2.98 0.050 0.4 101.6 No
1344 3.24 0.006 3.41 0.058 1.4 105.1 Yes
1346 0.090 0.003 0.093 0.002 0.5 103.8 Yes
1350 3.38 0.169 3.50 0.059 0.3 103.3 No
1352 3.17 0.006 3.26 0.055 0.8 102.9 Yes
1354 3.94 0.052 3.93 0.067 0.1 99.7 Yes
1356 5.2 0.058 5.31 0.091 0.5 102.1 Yes
1358 1.33 0.006 1.35 0.023 0.5 101.6 Yes
1360 0.29 0.006 0.287 0.005 0.2 98.9 Yes
1362 9.99 0.040 9.92 0.171 0.2 99.3 Yes
1366 1.81 0.010 1.78 0.031 0.5 98.1 Yes
1368 1.22 0.006 1.25 0.022 0.6 102.2 Yes
1371 0.03 0.015 0.015 0.001 0.5 Yes
1372 0.03 0.015 0.015 0.001 0.5 Yes
1375 0.02 0.004 0.005 0.001 1.7 Yes
1377 3.17 0.006 3.24 0.057 0.6 102.2 Yes
197
Appendix E
Table E10: Trueness of the test method for P2O5.
Result (%)
Fusion CP2O5,CRM
u(CP2O5,CRM) for u(P2O5) En Recovery Included
ID (%)
P2O5
1252 1.46 0.043 1.48 0.020 0.2 101.5 Yes
1253 1.46 0.043 1.46 0.020 0.0 100.1 Yes
1254 0.02 0.003 <0.01 No
1255 0.02 0.003 <0.01 No
1256 0.096 0.0009 0.093 0.003 0.6 96.5 No
1260 0.0928 0.00124 0.097 0.002 0.8 104.7 No
1261 0.0928 0.00124 0.095 0.003 0.4 102.6 No
1262 0.236011 0.00148 0.246 0.004 1.1 No
1266 0.2223 0.00241 0.229 0.004 0.7 102.8 No
1267 0.2223 0.00241 0.232 0.004 1.0 No
1268 0.1581 0.00210 0.163 0.003 0.7 103.3 No
1269 0.1581 0.00210 0.161 0.003 0.4 101.9 No
1270 0.0518 0.00192 0.051 0.002 0.1 99.4 No
1271 0.0518 0.00192 0.051 0.002 0.1 98.8 No
1272 0.15 0.016 0.152 0.003 0.1 101.4 Yes
1274 0.009 2.64585E-05 0.007 0.002 0.7 No
1276 0.060 0.0002 0.028 0.002 8.1 No
1278 0.109 0.0011 0.109 0.003 0.0 99.9 No
1280 0.009 0.0014 0.006 0.002 0.7 No
1282 0.0420 0.00145 0.041 0.001 0.2 97.7 Yes
1284 0.1440 0.00180 0.144 0.003 0.1 100.3 Yes
1285 0.1440 0.00180 0.146 0.003 0.3 101.3 Yes
1286 0.0267 0.00180 0.026 0.001 0.1 98.2 Yes
1292 0.023 0.002 No
1293 0.017 0.002 No
1294 <0.01 No
1295 0.012 0.002 No
1298 <0.01 No
1299 <0.01 No
1300 0.023 0.002 No
1304 0.025 0.002 No
1305 0.026 0.002 No
1310 0.071 0.0037 0.077 0.002 0.8 109.1 Yes
1311 0.071 0.0037 0.074 0.002 0.3 104.2 Yes
1314 0.22 0.008 0.235 0.004 0.8 106.7 Yes
1318 0.80198 0.017827511 0.806 0.013 0.1 100.5 No
198
Appendix E
Table E10: Continued

Result (%)
Fusion CP2O5,CRM
u(CP2O5,CRM) for u(P2O5) En Recovery Included
ID (%)
P2O5
1319 0.80198 0.017827511 0.801 0.013 0.0 99.9 No
1320 15.7 0.04 15.58 0.250 0.2 99.2 Yes
1324 0.199 0.0026 0.199 0.006 0.0 100.0 Yes
1328 30.2 0.06 31.62 0.521 1.4 104.7 Yes
1330 0.053 0.0087 0.047 0.002 0.3 89.4 Yes
1334 0.07 0.008 <0.01 Yes
1336 0.039 0.0040 0.037 0.002 0.3 93.7 Yes
1338 0.168 0.003 No
1344 0.023 0.0006 0.018 0.002 1.3 76.8 Yes
1346 0.092 0.0006 0.097 0.003 0.9 105.1 Yes
1350 0.064 0.002 No
1352 0.08 0.006 0.089 0.002 0.7 111.1 Yes
1354 0.042 0.002 No
1356 0.02 0.006 0.016 0.002 0.3 80.8 Yes
1358 0.27 0.006 0.272 0.005 0.1 100.8 Yes
1360 0.13 0.006 0.127 0.003 0.2 97.6 Yes
1362 0.010 0.0016 0.009 0.002 0.2 91.6 Yes
1366 0.146 0.0032 0.156 0.003 1.1 106.7 Yes
1368 1.3 0.06 1.24 0.019 0.5 95.4 Yes
1371 2.05 0.046 1.97 0.043 0.6 96.1 Yes
1372 2.05 0.046 1.98 0.044 0.6 96.4 Yes
1375 <0.01 No
1377 0.08 0.006 0.088 0.002 0.7 110.5 Yes
199
Appendix E
Table E11: Trueness of the test method for Cr2O3.
Result (%)
Fusion CCr2O3,CRM
u(C Cr2O3,CRM) for u(Cr2O3) En Recovery Included
ID (%)
Cr2O3
1252 0.19 0.012 0.187 0.012 0.1 98.2 Yes
1253 0.19 0.012 0.188 0.012 0.1 98.9 Yes
1254 0.0285 0.000141 0.025 0.002 0.9 No
1255 0.0285 0.000141 0.025 0.002 1.1 No
1256 0.0060 0.00015 0.007 0.002 0.3 No
1260 0.00053 0.00004 0.003 0.001 0.9 No
1261 0.00053 0.00004 0.003 0.001 1.0 No
1262 0.004677 0.00012 0.005 0.001 0.2 No
1266 0.0029 0.00012 0.005 0.001 0.8 No
1267 0.0029 0.00012 0.006 0.001 1.1 No
1268 0.0145 0.00024 0.016 0.002 0.5 No
1269 0.0145 0.00024 0.016 0.002 0.6 No
1270 0.0047 0.00024 0.007 0.001 1.1 No
1271 0.0047 0.00024 0.007 0.001 1.1 No
1272 0.0522 0.000141 0.050 0.004 0.3 No
1274 0.42 0.001 0.451 0.029 0.6 107.5 Yes
1276 0.003 0.001 No
1278 0.0031 0.00015 0.009 0.002 1.8 No
1280 48.97 0.025 48.41 4.74 0.1 98.9 Yes
1282 0.0096 0.00085 0.011 0.001 0.5 115.5 Yes
1284 0.002 0.0005 0.005 0.001 1.3 No
1285 0.002 0.0005 0.006 0.001 1.7 No
1286 0.0074 0.00075 0.009 0.001 0.5 117.9 Yes
1292 0.006 0.001 No
1293 0.045 0.003 No
1294 0.002 0.001 No
1295 0.001 0.001 No
1298 0.0001 0.000141 0.002 0.001 0.9 No
1299 0.0001 0.000141 0.003 0.001 1.3 No
1300 0.008 0.001 No
1304 0.018 0.002 No
1305 0.020 0.002 No
1310 0.016 0.0014 0.024 0.002 1.6 149.4 Yes
1311 0.016 0.0014 0.017 0.002 0.3 108.9 Yes
1314 0.07 0.014 0.077 0.005 0.2 110.3 Yes
1318 0.020 0.002 No
200
Appendix E

Result (%)
Fusion CK2O,CRM
ID (%)
K2O
1319 0.020 0.002 No
1320 0.33 0.008458 0.367 0.023 0.8 111.3 Yes
1324 0.025 0.002 No
1328 0.127 0.008 No
1330 0.0180 0.000141 0.019 0.002 0.3 106.6 Yes
1334 36.97 0.026 36.35 3.00 0.1 98.3 Yes
1336 0.024 0.0023 0.025 0.002 0.1 102.6 Yes
1338 0.018 0.002 No
1344 0.005 0.001 No
1346 0.031 0.002 Yes
1350 0.0174 0.00146 0.020 0.002 0.6 No
1352 0.021 0.0006 0.023 0.002 0.5 No
1354 0.005 0.001 No
1356 0.001 0.001 No
1358 0.008 0.001 No
1360 0.004 0.001 No
1362 0.002 0.001 No
1366 0.066 0.004 No
1368 0.047 0.003 No
1371 0.010 0.001 No
1372 0.008 0.001 Yes
1375 0.001 0.001 No
1377 0.021 0.0006 0.022 0.002 0.2 104.6 Yes
201
Appendix E
Table E12: Trueness of the test method for LOI.
Result (%)
Fusion CLOI,CRM
u(CLOI,CRM) for u(LOI) En Recovery
ID (%)
LOI
1252 11.42 0.088
1253 11.42 0.088
1254 48.04 0.315
1255 47.91 0.313
1256 0.57 0.007
1260 0.70 0.014 0.72 0.007 0.6 102.7
1261 0.70 0.014 0.72 0.006 0.7 103.1
1262 4.44 0.032 4.32 0.010 1.7 97.4
1266 1.10 0.019 1.11 0.007 0.2 100.6
1267 1.10 0.019 1.09 0.007 0.4 98.7
1268 5.95 0.014 5.46 0.013 12.5 91.8
1269 5.95 0.014 5.43 0.013 13.3 91.3
1270 34.1 0.05 34.30 0.470 0.2 100.6
1271 34.1 0.05 34.27 0.469 0.2 100.5
1272 -0.16 0.007
1274 -0.91 0.008
1276 2.79 0.008
1278 -1.87 0.011
1280 0.50 0.006
1282 33.26 0.205 33.13 0.439 0.1 99.6
1284 26.30 0.300 26.32 0.283 0.0 100.1
1285 26.30 0.300 26.27 0.282 0.0 99.9
1286 29.72 0.130 29.78 0.363 0.1 100.2
1292 0.39 0.020 0.41 0.007 0.3 103.3
1293 0.39 0.020 0.41 0.006 0.3 103.2
1294 0.30 0.007
1295 0.33 0.007
1298 0.08 0.014 0.05 0.006 0.9 63.5
1299 0.08 0.014 0.04 0.007 1.3 50.2
1300 0.10 0.014 0.03 0.007 2.2 29.6
1304 -0.07 0.006
1305 -0.08 0.007
1310 11.8 0.04 11.41 0.088 2.0 96.7
1311 11.8 0.04 11.40 0.088 2.1 96.6
1314 0.40 0.034 0.40 0.007 0.0 99.6
1318 25.77 0.130 25.64 0.269 0.2 99.5
202
Appendix E

Result (%)
Fusion CK2O,CRM
u(CK2O,CRM) for u(K2O) En Recovery
ID (%)
K2O
1319 25.77 0.130 25.68 0.270 0.1 99.7
1320 2.25 0.008
1324 13.42 0.121
1328 6.90 0.020
1330 12.55 0.106
1334 0.33 0.007
1336 20.5 0.12 20.93 0.179 1.0 102.1
1338 7.41 0.023
1344 1.27 0.007
1346 0.22 0.014 0.20 0.007 0.7 90.1
1350 7.62 0.024
1352 7.28 0.006 7.20 0.021 1.8 98.9
1354 1.15 0.007
1356 0.26 0.006 0.35 0.007 4.7 133.6
1358 -0.46 0.007
1360 3.74 0.010
1362 0.22 0.007
1366 0.91 0.007
1368 0.42 0.006 0.39 0.010 1.4 92.8
1371 39.12 0.611
1372 39.13 0.612
1375 43.4 0.04 41.45 0.234 4.1 95.5
1377 7.28 0.006 7.28 0.050 0.0 100.0
203
Appendix F
Appendix F
In this Appendix, eleven test sample portions of limestone are evaluated for primary
sampling effects by analysing five elements with the trace element test method at the
CGS. Although CaO is part of the trace element test program, it is not included as its
values far exceed its calibration range's upper limit. Sr and Y values are included to
illustrate the homogeneity of the milled limestone.
Table F1: The analysis of eleven test samples of a limestone on five pressed
powder pellets by WD-XRF.
Name of test sample portion of TiO2 MnO Fe2O3 Sr Y
limestone (%) (%) (%) (ppm) (ppm)
1-13aaab 0.020 0.100 0.489 69 8.6
1-13aaab 0.020 0.100 0.491 68 8.7
1-13aaab 0.019 0.100 0.489 70 8.6
1-13aaab 0.019 0.099 0.493 69 9.3
1-13aaab 0.019 0.100 0.494 69 8.9
Average 0.019 0.100 0.491 69 8.8
Standard deviation 0.0005 0.0004 0.0023 0.4 0.30
1-13abba 0.024 0.099 0.484 68 8.4

1-13abba 0.024 0.100 0.488 70 8.6
1-13abba 0.024 0.099 0.487 69 8.9
1-13abba 0.030 0.134 0.617 81 10.9
1-13abba 0.024 0.100 0.490 69 8.9
Average 0.025 0.106 0.513 71 9.1
1-13bbaa 0.024 0.097 0.480 68 8.8

1-13bbaa 0.023 0.099 0.480 69 9.1
1-13bbaa 0.023 0.097 0.481 68 8.7
1-13bbaa 0.024 0.100 0.483 69 8.7
1-13bbaa 0.023 0.099 0.482 68 8.2
Average 0.023 0.098 0.481 69 8.7
1-24aaab 0.024 0.097 0.482 68 8.5

1-24aaab 0.025 0.100 0.483 69 8.8
1-24aaab 0.025 0.098 0.483 68 8.9
1-24aaab 0.024 0.099 0.487 68 8.9
1-24aaab 0.024 0.099 0.485 69 8.4
Average 0.024 0.099 0.484 69 8.7
204
Appendix F
Table F1: Continued

1-24abbb 0.022 0.097 0.479 67 8.7
1-24abbb 0.022 0.098 0.483 68 8.2
1-24abbb 0.022 0.098 0.484 68 8.5
1-24abbb 0.023 0.098 0.483 69 8.9
1-24abbb 0.022 0.099 0.484 69 8.8
Average 0.022 0.098 0.483 68. 8.6
1-24bbab 0.023 0.098 0.478 69 8.9

1-24bbab 0.028 0.127 0.573 79 10.5
1-24bbab 0.022 0.099 0.489 69 8.8
1-24bbab 0.022 0.100 0.495 69 8.8
1-24bbab 0.022 0.100 0.489 69 8.8
Average 0.023 0.105 0.505 71 9.1
2-13aabb 0.023 0.099 0.487 69 9.1

2-13aabb 0.023 0.098 0.487 68 8.8
2-13aabb 0.023 0.100 0.491 70 9.2
2-13aabb 0.023 0.101 0.492 70 8.9
2-13aabb 0.023 0.101 0.491 69 9.3
Average 0.023 0.100 0.490 69 9.0
2-13baba 0.023 0.099 0.475 68 8.4

2-13baba 0.023 0.097 0.477 69 8.9
2-13baba 0.023 0.098 0.478 68 8.7
2-13baba 0.023 0.098 0.477 69 8.5
2-13baba 0.023 0.098 0.479 69 9.1
Average 0.023 0.098 0.477 69 8.7
2-24aaaa 0.023 0.082 0.347 68 8.4

2-24aaaa 0.023 0.081 0.349 69 8.7
2-24aaaa 0.024 0.082 0.349 69 9.1
2-24aaaa 0.024 0.082 0.356 68 8.7
2-24aaaa 0.024 0.100 0.471 69 8.7
Average 0.024 0.086 0.374 68 8.7
2-24abba 0.027 0.100 0.486 69 8.7

2-24abba 0.027 0.099 0.487 69 9.1
2-24abba 0.027 0.100 0.491 69 8.7
2-24abba 0.027 0.099 0.489 69 8.9
2-24abba 0.028 0.100 0.491 69 8.5
Average 0.027 0.099 0.489 69 8.8
205
Appendix F
Table F1: Continued

2-24bbaa 0.022 0.098 0.470 67 8.7
2-24bbaa 0.023 0.098 0.474 68 7.9
2-24bbaa 0.023 0.099 0.472 68 8.5
2-24bbaa 0.023 0.099 0.474 67 8.9
2-24bbaa 0.023 0.091 0.425 68 8.7
Average 0.023 0.097 0.463 68 8.6
Average of averages 0.023 0.099 0.477 69 8.8

Standard deviation of averages 0.0020 0.0052 0.0366 1.1 0.2
206
Appendix G
Appendix G
This Appendix presents the results of six fusion disks (split samples a, b, c, d, e, f) of
a SARM-36 test sample, while the sample flux masses are listed in Table 4.3. Each
disk was analysed in triplicate, and the results are listed in the first rows of Figures G1
to G11. At the bottom of each figure, the standard deviation (Stdev) value is calculated
from the total value divided by the degrees of freedom (df) in the ANOVA table’s last
row, and this value is used to determine the relative standard uncertainty of the
dilution’s precision.
on the SiO2 results.
207
Appendix G
Figure G2: ANOVA calculations to determine the effects of sample and flux
masses on the TiO2 results.
208
Appendix G
on the Al2O3 results.
209
Appendix G
on the Fe2O3 results.
210
Appendix G
masses on the MnO results.
211
Appendix G
masses on the MgO results.
212
Appendix G
masses on the CaO results.
213
Appendix G
on the Na2O results.
214
Appendix G
masses on the K2O results.
215
Appendix G
masses on the P2O5 results.
216
Appendix G
masses on the Cr2O3 results.
217
Appendix I
Appendix H
This appendix lists the two-way ANOVA analysis of the LOI results of 72 aliquots of a
limestone test sample to evaluate the effect of different areas in a furnace on the LOI-
values.
Figure H1: Two factor ANOVA by Excel to evaluate the precision of LOI-analysis in
a limestone test sample.
218
Appendix G
Appendix I
In this Appendix, four SARM-36 aliquots are used for six fusion disks per aliquot, and
each fusion disk is analysed three times. The top parts of each figure list the triplicate
results of six fusion disks prepared by the following:
i) Old mouldible and FM fluxer,

ii) new mouldible and FM fluxer,
iii) old mould and TheOx fluxer, and
iv) new mould and TheOx fluxer.
The values in the last two rows in each figure are used to quantify the relative standard
uncertainty of the two fluxers and are used in the bottom-up approaches in this study.
219
Appendix G
FM fluxer_old FM fluxer_new TheOx_old TheOx_new

Fusion 1 64.052 64.146 Fusion 1 63.728 63.747
64.205 64.134 63.701 63.806
64.094 64.201 63.662 63.819
Fusion 2 64.129 63.918 Fusion 2 63.695 63.795
64.144 63.862 63.711 63.852
64.160 63.900 63.693 63.731
Fusion 3 63.739 63.961 Fusion 3 63.646 63.796
63.870 63.928 63.639 63.788
63.901 63.939 63.673 63.727
Fusion 4 64.049 64.298 Fusion 4 63.628 63.707
64.056 64.229 63.638 63.818
64.030 64.258 63.636 63.757
Fusion 5 63.913 64.090 Fusion 5 63.615 63.783
63.929 64.177 63.648 63.766
63.899 64.091 63.660 63.735
Fusion 6 64.189 64.114 Fusion 6 63.714 63.746
64.157 64.141 63.664 63.733
64.138 64.066 63.592 63.770
Anova: Two-Factor With Replication Anova: Two-Factor With Replication

FM
FM fluxer_old Total SUMMARY TheOx_old TheOx_new Total
SUMMARY fluxer_new
Fusion 1 Fusion 1
Count 3 3 6 Count 3 3 6
Sum 192.3511 192.4810 384.8321 Sum 191.0918 191.3718 382.4636
Average 64.1170 64.1603 64.1387 Average 63.6973 63.7906 63.7439
Variance 0.0062 0.0013 0.0036 Variance 0.0011 0.0015 0.0037
Fusion 2 Fusion 2
Sum 192.4331 191.6801 384.1132 Sum 191.0989 191.3776 382.4765
Average 64.1444 63.8934 64.0189 Average 63.6996 63.7925 63.7461
Fusion 3 Fusion 3
Sum 191.5098 191.8281 383.3379 Sum 190.9581 191.3112 382.2692
Average 63.8366 63.9427 63.8896 Average 63.6527 63.7704 63.7115
Fusion 4 Fusion 4
Sum 192.1345 192.7851 384.9196 Sum 190.9015 191.2820 382.1835
Average 64.0448 64.2617 64.1533 Average 63.6338 63.7607 63.6973
Fusion 5 Fusion 5
Sum 191.7406 192.3577 384.0982 Sum 190.9237 191.2845 382.2082
Average 63.9135 64.1192 64.0164 Average 63.6412 63.7615 63.7014
Fusion 6 Fusion 6
Sum 192.4838 192.3210 384.8048 Sum 190.9702 191.2493 382.2195
Average 64.1613 64.1070 64.1341 Average 63.6567 63.7498 63.7033
Total Total
Count 18 18 Count 18 18
Sum 1152.6528 1153.4530 Sum 1145.9442 1147.8763
Average 64.0363 64.0807 Average 63.6636 63.7709
Variance 0.0174 0.0177 Variance 0.0014 0.0015
ANOVA ANOVA
Source of Variation SS df MS Source of Variation SS df MS
Sample 0.3179 5 0.0636 Sample 0.0145 5 0.0029
Columns 0.0178 1 0.0178 Columns 0.1037 1 0.1037
Interaction 0.2348 5 0.0470 Interaction 0.0019 5 0.0004
Within 0.0448 24 0.0019 Within 0.0329 24 0.0014
Total 0.6153 35 Total 0.1530 35

Repeatability:
Relative
Standard standard
uncertainty uncertaint
y
FM fluxer 0.1326 0.00368 0.00575
TheOx 0.0661 0.00184 0.00287
Figure I1: Two factor ANOVA analysis to determine fusion disk precision for SiO 2.
220
Appendix G
Figure I2: Two factor ANOVA analysis to determine fusion disk precision for TiO 2.
221
Appendix G
Figure I3: Two factor ANOVA analysis to determine fusion disk precision for Al2O3.
222
Appendix G
Figure I4: Two factor ANOVA analysis to determine fusion disk precision for Fe2O3.
223
Appendix G
Figure I5: Two factor ANOVA analysis to determine fusion disk precision for MnO.
224
Appendix G
Figure I6: Two factor ANOVA analysis to determine fusion disk precision for MgO.
225
Appendix G
Figure I7: Two factor ANOVA analysis to determine fusion disk precision for CaO.
226
Appendix G
Figure I8: Two factor ANOVA analysis to determine fusion disk precision for Na2O.
227
Appendix G
Figure I9: Two factor ANOVA analysis to determine fusion disk precision for K 2O.
228
Appendix G
Figure I10: Two factor ANOVA analysis to determine fusion disk precision for P 2O5.
229
Appendix G
Figure I11: Two factor ANOVA analysis to determine fusion disk precision for Cr2O3.
230
Appendix J
Appendix J
Appendix J lists the results per major element to determine the precision of the sample
cup and measuring position. The data of cup 5 is used to quantify the relative standard
uncertainty of the cup in the bottom-up approaches (last row in Figs J1 to J11).
SiO2
Pos 1 57.87 57.77 57.71 57.81 57.84 57.92 57.82 57.78
57.88 57.82 57.75 57.79 57.85 57.90 57.84 57.78
57.82 57.80 57.82 57.78 57.84 57.79 57.77 57.81
57.90 57.80 57.83 57.86 57.81 57.83 57.84 57.79
57.90 57.90 57.79 57.85 57.86 57.83 57.87 57.79
Pos 2 57.83 57.79 57.80 57.79 57.77 57.80 57.86 57.75
57.91 57.78 57.82 57.81 57.76 57.84 57.88 57.80
57.90 57.77 57.81 57.87 57.88 57.82 57.90 57.78
57.88 57.80 57.88 57.80 57.86 57.82 57.80 57.76
57.85 57.79 57.82 57.82 57.85 57.86 57.87 57.86
Anova: Two-Factor With Replication
SUMMARY Cup 1 Cup 2 Cup 3 Cup 4 Cup 5 Cup 6 Cup 7 Cup 8 Total
Pos 1
Count 5 5 5 5 5 5 5 5 40
Sum 289.3679538 289.09385 288.8976 289.09431 289.1901 289.26175 289.14868 288.95393 2313.008
Average 57.87359076 57.81877 57.779521 57.818862 57.838021 57.85235 57.829735 57.790786 57.8252
Variance 0.001108207 0.002452 0.0023959 0.0013716 0.0003972 0.0028899 0.001217 0.0001672 0.002079
Pos 2
Count 5 5 5 5 5 5 5 5 40
Sum 289.3591453 288.92684 289.13017 289.09422 289.11859 289.13355 289.31286 288.94966 2313.025
Average 57.87182906 57.785368 57.826034 57.818845 57.823719 57.82671 57.862572 57.789932 57.82563
Variance 0.001182281 0.0001307 0.0011168 0.0008935 0.0027528 0.000437 0.0014244 0.0016662 0.001811
Total
Count 10 10 10 10 10 10 10 10
Sum 578.7270991 578.02069 578.02778 578.18853 578.3087 578.3953 578.46154 577.90359
Average 57.87270991 57.802069 57.802778 57.818853 57.83087 57.83953 57.846154 57.790359
Variance 0.001018857 0.0014578 0.0021622 0.0010067 0.0014568 0.0016612 0.0014735 0.000815
ANOVA
Source of Variation SS df MS F P-value F crit
Sample 3.55665E-06 1 3.557E-06 0.0026342 0.9592267 3.9909238
Columns 0.052253377 7 0.0074648 5.5287469 5.266E-05 2.156424
Interaction 0.0130546 7 0.0018649 1.3812616 0.2286112 2.156424
Within 0.08641111 64 0.0013502
Total 0.151722644 79

Average 57.873 57.802 57.803 57.819 57.831 57.840 57.846 57.790
Stdev 0.0319 0.0382 0.0465 0.0317 0.0382 0.0408 0.0384 0.0285
Standard u 0.00319 0.00382 0.00465 0.00317 0.00382 0.00408 0.00384 0.00285
Relative u 0.00006 0.00007 0.00008 0.00005 0.00007 0.00007 0.00007 0.00005
Figure J1: SiO2 results to evaluate the precision of the sample cup and measuring
231
Appendix J
Figure J2: TiO2 results to evaluate the precision of the sample cup and measuring
232
Appendix J
Figure J3: Al2O3 results to evaluate the precision of the sample cup and measuring
233
Appendix J
Figure J4: Fe2O3 results to evaluate the precision of the sample cup and
measuring position in the Zetium XRF spectrometer.
234
Appendix J
Figure J5: MnO results to evaluate the precision of the sample cup and measuring
235
Appendix J
Figure J6: MgO results to evaluate the precision of the sample cup and measuring
236
Appendix J
Figure J7: CaO results to evaluate the precision of the sample cup and measuring
237
Appendix J
Figure J8: Na2O results to evaluate the precision of the sample cup and
238
Appendix J
Figure J9: K2O results to evaluate the precision of the sample cup and measuring
239
Appendix J
Figure J10: P2O5 results to evaluate the precision of the sample cup and measuring
240
Appendix J
Figure J11: Cr2O3 results to evaluate the precision of the sample cup and
241
Appendix K
Appendix K
This appendix details the propagation of the standard uncertainty values in the GUM
framework for the four test samples, SARM-36, MzBP-1, Li3 6c and G00006619001.
Table K1: The input variables (g), standard uncertainties, and sensitivity coefficients
propagated to provide uncertainty of measured LOI (%) in measurement
model 1. The correlation coefficients are listed in Table 4.6. The LOI
results and calculated u(LOI) values are used in Measurement model 4.
Test sample Cr u(Cr) c(Cr) Cr100 u(Cr100) c(Cr100)

SARM-36 C1 28.1882 0.00015 0.0214 37.1861 0.00015 11.092
MzBP-1 19.9448 0.00015 0.1868 22.0653 0.00015 46.972
Li3 6c 27.6082 0.00015 3.3226 41.0182 0.00015 4.135
G00006619001 29.4098 0.00015 0.6661 33.6544 0.00015 22.893
Table K1: Continued

Test sample Cr1000 u(Cr1000) c(Cr1000) P u(P) c(P) LOI u(LOI)
SARM-36 C1 37.1688 0.00015 -11.114 1 0.00181 0.192 0.192 0.0005
MzBP-1 22.0569 0.00015 -47.159 1 0.00181 0.396 0.396 0.0019
Li3 6c 35.0432 0.00015 -7.457 1 0.00607 44.556 44.556 0.2706
G00006619001 33.5344 0.00015 -23.559 1 0.00181 2.827 2.827 0.0052
242
Appendix K
Table K2: Measurement model 2 results for the calculation of Li2O and B2O3 in the
fusion disks of four test samples. The Li2O and B2O results and their
uncertainties are used in Measurement model 4.
SARM-36 C1 MzBP-1 Li3 6c G00006619001
Selected flux A B C A
Li2O(a) (%) 11.54237 8.61306 6.07981 11.54237
u(Li2O(a)) 0.01137 0.00849 0.00599 0.01137
c(Li2O(a)) 0.90897 0.90899 0.90900 0.90885
Li2O(b) (%) 9.69708 14.69478 19.01677 9.69708
u(Li2O(b)) 0.00817 0.01238 0.01602 0.00817
c(Li2O(b)) 0.90897 0.90899 0.90900 0.90885
B2O3(a) (%) 55.46266 41.38694 29.21431 55.46266
u(B2O3(a)) 0.01298 0.00969 0.00929 0.01298
c(B2O3(a)) 0.90897 0.90899 0.90900 0.90885
B2O3(b) (%) 23.29790 35.30522 45.68911 23.29790
u(B2O3(b)) 0.00853 0.01292 0.01673 0.00853
c(B2O3(b)) 0.90897 0.90899 0.90900 0.90885
massflux (g) 10.00040 10.00180 10.00190 10.00430
u(massflux) 0.00015 0.00015 0.00015 0.00015
c(massflux) in calculation
0.17574 0.19278 0.20756 0.17588
for Li2O
c(massflux) in calculation
0.65166 0.63434 0.61948 0.65221
for B2O3
masssample (g) 1.00150 1.00140 1.00130 1.00340
u(masssample) 0.00015 0.00015 0.00015 0.00015
c(masssample) in
-1.75479 -1.92549 -2.07329 -1.75362
calculation for Li2O
c(masssample) in
-6.50715 -6.33565 -6.18794 -6.50283
calculation for B2O3
Li2O (%) 19.306 21.187 22.813 19.303
u(Li2O) 0.0127 0.0136 0.0155 0.0127
B2O3 (%) 71.591 69.712 68.087 71.581
u(B2O3) 0.0141 0.0147 0.0174 0.0141
243
Appendix K
Table K3: Concentration of SiO2 as calculated by Measurement model 3 and its

associated uncertainty for the four test samples. The sensitivity
SARM-36 C1 MzBP-1 Li3 6c G00006619001
Fluxer used TheOx TheOx FM FM
Cup 5 5 5 5
MIi (kCPS) 33.4335 28.6440 0.1831 24.9551
u(MIi) 0.035 0.032 0.003 0.028
c(MIi) 0.17928 0.17928 0.17928 0.17928
c(Ei) 33.434 28.644 0.183 24.955
u(Pcup) 3.96E-06 3.39E-06 2.17E-08 2.95E-06
u(PFluxer) TheOx 1.72E-04 1.47E-04
u(PFluxer) FM 1.89E-06 2.57E-04
u(PDilution) 2.05E-04 1.76E-04 1.12E-06 1.53E-04
c(P) 5.9940 5.1353 0.0328 4.4739
c(D) 1 1 1 1
CSiO2,meas (%) 5.9940 5.1353 0.0328 4.4739
u(CSiO2,meas) 0.0288 0.0248 0.0014 0.0216
Table K4: Concentration of TiO2 as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.1393 3.0118 0.2507 4.7385
u(MIi) 0.004 0.017 0.004 0.02
c(MIi) 0.03161 0.03161 0.03161 0.03161
c(Ei) 0.139 3.012 0.251 4.739
u(Pcup) 1.18E-08 2.55E-07 2.12E-08 4.01E-07
u(PFluxer) FM 8.51E-06 1.61E-04
u(PDilution) 6.63E-06 1.43E-04 1.19E-05 2.25E-04
c(P) 0.00440 0.09520 0.00792 0.14979
244
Appendix K
Table K4: Continued

SARM-36 C1 MzBP-1 Li3 6c G00006619001
c(D) 1 1 1 1
CTiO2,meas (%) 0.00440 0.09520 0.00792 0.14979
u(CTiO2,meas) 0.000153 0.000662 0.000155 0.000872
Table K5: Concentration of Al2O3 as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 11.0079 11.8949 0.0493 11.0189
u(MIi) 0.020 0.021 0.003 0.018
c(MIi) 0.14626 0.14626 0.14626 0.14626
c(Ei) 11.008 11.895 0.049 11.019
u(Pcup) 2.43E-06 2.63E-06 1.09E-08 2.43E-06
u(PFluxer) FM 6.87E-07 1.53E-04
u(PDilution) 1.40E-04 1.52E-04 6.29E-07 1.40E-04
c(P) 1.6100 1.7397 0.0072 1.6116
c(D) 1 1 1 1
CAl2O3,meas (%) 1.6100 1.7397 0.0072 1.6116
u(CAl2O3,meas) 0.00568 0.00605 0.00198 0.00555
245
Appendix K
Table K6: Concentration of Fe2O3 as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 5.8692 18.8349 3.4349 42.1529
u(MIi) 0.031 0.058 0.020 0.083
c(MIi) 0.02259 0.02259 0.02259 0.02259
c(Ei) 5.869 18.835 3.435 42.153
u(Pcup) 1.73E-07 5.55E-07 1.01E-07 1.24E-06
u(PFluxer) FM 1.10E-05 1.35E-04
u(PDilution) 3.46E-05 1.11E-04 2.02E-05 2.48E-04
c(P) 0.1326 0.4255 0.0776 0.9522
c(D) 1 1 1 1
CFe2O3,meas (%) 0.1326 0.4255 0.0776 0.9522
u(CFe2O3,meas) 0.00111 0.00304 0.00068 0.00642
Table K7: Concentration of MnO as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.0295 0.4588 0.8047 0.6402
u(MIi) 0.006 0.010 0.010 0.011
c(MIi) 0.02748 0.02748 0.02748 0.02748
c(Ei) 0.030 0.459 0.805 0.640
u(Pcup) 7.07E-09 1.10E-07 1.93E-07 1.53E-07
u(PFluxer) FM 4.53E-05 3.60E-05
246
Appendix K
Table K7: Continued

SARM-36 C1 MzBP-1 Li3 6c G00006619001
u(PDilution) 3.40E-06 5.28E-05 9.26E-05 7.37E-05
c(P) 0.00081 0.01261 0.02211 0.01759
c(D) 1 1 1 1
CMnO,meas (%) 0.00081 0.01261 0.02211 0.01759
u(CMnO,meas) 0.000182 0.000289 0.000297 0.000319
Table K8: Concentration of MgO as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.6380 1.6576 14.7652 12.8907
u(MIi) 0.006 0.008 0.019 0.017
c(MIi) 0.05927 0.05927 0.05927 0.05927
c(Ei) 0.638 1.658 14.765 12.891
u(Pcup) 0.00000036 0.00000095 0.00000844 0.00000737
u(PFluxer) TheOx 0.000028 0.000073
u(PFluxer) FM 0.000707 0.000617
u(PDilution) 0.000055 0.000143 0.001273 0.001112
c(P) 0.0378 0.0982 0.8751 0.7640
c(D) 1 1 1 1
CMgO,meas (%) 0.0378 0.0982 0.8751 0.7640
u(CMgO,meas) 0.000182 0.000289 0.000297 0.000319
247
Appendix K
Table K9: Concentration of CaO as calculated by Measurement model 3 and its

reported as c(P).
SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 3.0838 16.1400 400.9126 49.3889
u(MIi) 0.017 0.040 0.245 0.070
c(MIi) 0.02058 0.02058 0.02058 0.02058
c(Ei) 3.084 16.140 400.913 49.389
u(Pcup) 1.23E-07 6.46E-07 1.60E-05 1.98E-06
u(PFluxer) FM 1.16E-03 1.42E-04
u(PDilution) 1.39E-05 7.29E-05 1.81E-03 2.23E-04
c(P) 0.06346 0.33216 8.25078 1.01642
c(D) 1 1 1 1
CCaO,meas (%) 0.06346 0.33216 8.25078 1.01642
u(CCaO,meas) 0.001220 0.006164 0.152834 0.018748
Table K10: Concentration of Na2O as calculated by Measurement model 3 and its

and is reported as c(P). Although the concentration for sample Li3 6c is
below the lower limit of detection, it and its associated uncertainty are
still used in Measurement model 4.
SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.4208 6.5363 -0.0237 3.7664
u(MIi) 0.003 0.010 0.002 0.007
c(MIi) 0.09287 0.09287 0.09287 0.09287
c(Ei) 0.421 6.536 -0.024 3.766
u(Pcup) 1.40E-07 2.18E-06 -7.92E-09 1.26E-06
u(PFluxer) FM -2.34E-06 3.72E-04
u(PDilution) 9.20E-05 1.43E-03 -5.19E-06 8.24E-04
248
Appendix K
Table K10: Continued

SARM-36 C1 MzBP-1 Li3 6c G00006619001
c(P) 0.0391 0.6070 -0.0022 0.3498
c(D) 1 1 1 1
CNa2O,meas (%) 0.0391 0.6070 -0.0022 0.3498
u(CNa2O,meas) 0.000555 0.002701 0.000493 0.001585
Table K11: Concentration of K2O as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 74.6346 17.4244 -0.0189 3.4202
u(MIi) 0.077 0.037 0.003 0.016
c(MIi) 0.01963 0.008213192 0.007357324 0.008199451
c(Ei) 74.635 17.424 -0.019 3.420
u(Pcup) 1.26E-06 2.95E-07 -3.20E-10 5.80E-08
u(PFluxer) FM -4.95E-08 8.95E-06
u(PDilution) 2.62E-04 6.12E-05 -6.63E-08 1.20E-05
c(P) 1.4651 0.3420 -0.0004 0.0671
c(D) 1 1 1 1
CK2O,meas (%) 1.4651 0.3420 -0.0004 0.0671
u(CK2O,meas) 0.0222 0.00518 0.00010 0.00103
249
Appendix K
Table K12: Concentration of P2O5 as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.3493 1.7128 0.0160 0.8031
u(MIi) 0.005 0.009 0.003 0.006
c(MIi) 0.03175 0.03175 0.03175 0.03175
c(Ei) 0.349 1.713 0.016 0.803
u(Pcup) 6.46E-08 3.17E-07 2.96E-09 1.49E-07
u(PFluxer) FM 2.20E-07 1.10E-05
u(PDilution) 8.95E-06 4.39E-05 4.10E-07 2.06E-05
c(P) 0.0111 0.0544 0.0005 0.0255
c(D) 1 1 1 1
CP2O5,meas (%) 0.0111 0.0544 0.0005 0.0255
u(CP2O5,meas) 0.000242 0.000727 0.000156 0.000383
Table K13: Concentration of Cr2O3 as calculated by Measurement model 3 and its

SARM-36 C1 MzBP-1 Li3 6c G00006619001
Cup 5 5 5 5
MIi (kCPS) 0.0682 0.1231 0.1677 0.5379
u(MIi) 0.007 0.008 0.006 0.011
c(MIi) 0.01348 0.01348 0.01348 0.01348
c(Ei) 0.068 0.123 0.168 0.538
u(Pcup) 2.33E-08 4.21E-08 5.74E-08 1.84E-07
u(PFluxer) FM 1.15E-05 3.69E-05
250
Appendix K

SARM-36 C1 MzBP-1 Li3 6c G00006619001
u(PDilution) 1.21E-05 2.19E-05 2.99E-05 9.58E-05
c(P) 0.00092 0.00166 0.00226 0.00725
c(D) 1 1 1 1
CCr2O3,meas (%) 0.00092 0.00166 0.00226 0.00725
u(CCr2O3,meas) 0.000123 0.000159 0.000171 0.000479
Measurement model 4 for SiO2. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CSiO2,meas 5.9940 5.1353 0.0328 4.4739
u(CSiO2,meas) 0.0288 0.0248 0.0014 0.0216
c(CSiO2,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.0127 0.014 0.016 0.013
c(Li2O) 6.847 6.523 0.021 4.899
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.014 0.015 0.017 0.014
c(B2O3) 6.8472 6.5228 0.0212 4.8995
LOI 0.1923 0.396 44.556 2.827
u(LOI) 0.0005 0.0019 0.2706 0.0052
c(LOI) -0.6404 -0.5807 -0.0035 -0.4742
Total 100.256 99.743 100.174 100.320
u(Total) 0.0415 0.0337 0.1548 0.0360
c(Total) -6.207 -5.942 -0.018 -4.425
SiO2 (%) 64.08 57.69 0.20 46.23
u(SiO2) 0.353 0.303 0.009 0.244
251
Appendix K
Measurement model 4 for TiO2. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CTiO2,meas 0.00440 0.09520 0.00792 0.14979
u(CTiO2,meas) 0.00015 0.00066 0.00015 0.00087
c(CTiO2,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.00503 0.12093 0.00512 0.16403
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.00503 0.12093 0.00512 0.16403
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00047 -0.01076 -0.00085 -0.01587
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.00456 -0.11016 -0.00427 -0.14816
TiO2 (%) 0.047 1.07 0.047 1.55
u(TiO2) 0.0016 0.008 0.0011 0.010
Measurement model 4 for Al2O3. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CAl2O3,meas 1.61001 1.73975 0.00722 1.61163
u(CAl2O3,meas) 0.00568 0.00605 0.00198 0.00555
c(CAl2O3,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 1.83919 2.20983 0.00467 1.76492
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 1.83919 2.20983 0.00467 1.76492
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
252
Appendix K

SARM-36 C1 MzBP-1 Li3 6c G00006619001
c(LOI) -0.17202 -0.19672 -0.00078 -0.17081
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -1.66717 -2.01311 -0.00389 -1.59411
Al2O3 (%) 17.21 19.54 0.04 16.65
u(Al2O3) 0.088 0.092 0.012 0.078
Measurement model 4 for Fe2O3. The concentration and uncertainty values
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CFe2O3,meas 0.13259 0.42548 0.07759 0.95223
u(CFe2O3,meas) 0.00111 0.00304 0.00068 0.00642
c(CFe2O3,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.15146 0.54044 0.05018 1.04280
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.15146 0.54044 0.05018 1.04280
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.01417 -0.04811 -0.00837 -0.10092
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.13729 -0.49233 -0.04181 -0.94188
Fe2O3 (%) 1.42 4.78 0.465 9.84
u(Fe2O3) 0.013 0.038 0.0078 0.074
253
Appendix K
Measurement model 4 for MnO. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CMnO,meas 0.00081 0.01261 0.02211 0.01759
u(CMnO,meas) 0.00018 0.00029 0.00030 0.00032
c(CMnO,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.00093 0.01602 0.01430 0.01927
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.00093 0.01602 0.01430 0.01927
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00009 -0.00143 -0.00238 -0.00186
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.00084 -0.01459 -0.01192 -0.01740
MnO (%) 0.009 0.142 0.133 0.182
u(MnO) 0.0020 0.0034 0.0029 0.0035
Measurement model 4 for MgO. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CMgO,meas 0.03781 0.09825 0.87513 0.76403
u(CMgO,meas) 0.00053 0.00096 0.00734 0.00639
c(CMgO,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.04320 0.12479 0.56594 0.83670
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.04320 0.12479 0.56594 0.83670
LOI 0.19227 0.39613 44.55630 2.82712
254
Appendix K
Table K19: Continue

SARM-36 C1 MzBP-1 Li3 6c G00006619001
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00404 -0.01111 -0.09437 -0.08098
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.03916 -0.11368 -0.47157 -0.75573
MgO (%) 0.40 1.10 5.25 7.89
u(MgO) 0.006 0.012 0.083 0.073
Measurement model 4 for CaO. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CCaO,meas 0.06346 0.33216 8.25078 1.01642
u(CCaO,meas) 0.00122 0.00616 0.15283 0.01875
c(CCaO,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.07250 0.42191 5.33570 1.11310
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.07250 0.42191 5.33570 1.11310
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00678 -0.03756 -0.88969 -0.10773
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.06572 -0.38435 -4.44601 -1.00537
CaO (%) 0.679 3.73 49.48 10.50
u(CaO) 0.0137 0.072 1.139 0.202
Measurement model 4 for Na2O. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CNa2O,meas 0.03908 0.60702 -0.00220 0.34978
255
Appendix K
SARM-36 C1 MzBP-1 Li3 6c G00006619001

u(CNa2O,meas) 0.00056 0.00270 0.00049 0.00158
c(CNa2O,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.04464 0.77104 -0.00143 0.38305
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.04464 0.77104 -0.00143 0.38305
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00418 -0.06864 0.00024 -0.03707
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.04047 -0.70240 0.00119 -0.34598
Na2O (%) 0.42 6.82 <0.02 3.61
u(Na2O) 0.01 0.041 0.022
Measurement model 4 for K2O. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CK2O,meas 1.46508 0.34204 -0.00037 0.06714
u(CK2O,meas) 0.02221 0.00518 0.00010 0.00103
c(CK2O,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 1.67363 0.43446 -0.00024 0.07352
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 1.67363 0.43446 -0.00024 0.07352
256
Appendix K

SARM-36 C1 MzBP-1 Li3 6c G00006619001
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.15654 -0.03868 0.00004 -0.00712
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -1.51709 -0.39578 0.00020 -0.06641
K2O (%) 15.66 3.84 <0.003 0.694
u(K2O) 0.238 0.058 0.0106
Measurement model 4 for P2O5. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CP2O5,meas 0.01109 0.05438 0.00051 0.02550
u(CP2O5,meas) 0.00024 0.00073 0.00016 0.00038
c(CP2O5,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.01267 0.06907 0.00033 0.02792
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.01267 0.06907 0.00033 0.02792
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00119 -0.00615 -0.00005 -0.00270
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.01148 -0.06293 -0.00027 -0.02522
P2O5 (%) 0.12 0.61 <0.01 0.26
u(P2O5) 0.003 0.009 0.004
Measurement model 4 for Cr2O3. The concentration and uncertainty
SARM-36 C1 MzBP-1 Li3 6c G00006619001
CCr2O3,meas 0.00092 0.00166 0.00226 0.00725
257
Appendix K
SARM-36 C1 MzBP-1 Li3 6c G00006619001

u(CCr2O3,meas) 0.00012 0.00016 0.00017 0.00048
c(CCr2O3,meas) 10.691 11.233 5.997 10.333
Li2O 19.306 21.187 22.813 19.303
u(Li2O) 0.01273 0.01365 0.01555 0.01273
c(Li2O) 0.00105 0.00211 0.00146 0.00794
B2O3 71.591 69.712 68.087 71.581
u(B2O3) 0.01415 0.01471 0.01742 0.01415
c(B2O3) 0.00105 0.00211 0.00146 0.00794
LOI 0.19227 0.39613 44.55630 2.82712
u(LOI) 0.00054 0.00188 0.27057 0.00519
c(LOI) -0.00010 -0.00019 -0.00024 -0.00077
Total 100.256 99.743 100.174 100.320
u(Total) 0.04150 0.03371 0.15480 0.03604
c(Total) -0.00095 -0.00192 -0.00122 -0.00717
Cr2O3 (%) 0.010 0.019 0.014 0.075
u(Cr2O3) 0.0013 0.0018 0.0010 0.0049
258
Appendix L
Appendix L
In this Appendix, the calculation of major elements and associated uncertainties in the
Measurement model 3 as applied in the Excel workbook template (‘GUM
framework.xlsx’), is illustrated. This process is in the worksheet, ‘Conc from SuperQ’.
Symbols in the worksheet are as follows:
Ei = Calibration slope
u(Ei) = Uncertainty of the calibration slope
c(Ei) = Sensitivity factor of the calibration slope
Di = Intercept of the slope on the Y-axis
c(Di) = Sensitivity factor of the intercept
P = Precision value
u(Pcup) = Uncertainty of the measuring cup’s precision
u(PFluxer) = Uncertainty of the fluxer’s precision (TheOx or FM)
u(PDilution) = Uncertainty of the dilution’s precision
c(P) = Sensitivity factor of the precision
RMS = Root Mean Square
K = The K-factor evaluating the calibration slope
Ii = Background and line overlap corrected counts (kCPS)
Mi = Matrix factor
MIi = Matrix corrected counts (kCPS)
u(MIi) = Uncertainty of the matrix corrected counts
c(MIi) = Sensitivity factor of the matrix corrected counts
Ci = Measured concentration of the major element in the fusion disk (%)
u(Ci) = Uncertainty of the measured concentration in the fusion disk
259
Appendix L
Figure L1: The calibration and precision values with their associated uncertainties
for Measurement model 3.
Figure L2: Matrix corrected counts, associated uncertainties and sensitivity factors
for SiO2 analysed the fusion disks of the four test samples (Columns E
to Q in ‘Conc from SuperQ’-worksheet).
for TiO2 analysed in the fusion disks of the four test samples (Columns
S to AE in ‘Conc from SuperQ’-worksheet).
260
Appendix L
for Al2O3 analysed the fusion disks of the four test samples (Columns
AG to AS in ‘Conc from SuperQ’-worksheet).
for Fe2O3 analysed the fusion disks of the four test samples (Columns
AU to BG in ‘Conc from SuperQ’-worksheet).
for MnO analysed the fusion disks of the four test samples (Columns BI
to BU in ‘Conc from SuperQ’-worksheet).
261
Appendix L
for MgO analysed the fusion disks of the four test samples (Columns
BW to CI in ‘Conc from SuperQ’-worksheet).
for CaO analysed the fusion disks of the four test samples (Columns
CK to CW in ‘Conc from SuperQ’-worksheet).
for Na2O analysed the fusion disks of the four test samples (Columns
CY to DK in ‘Conc from SuperQ’-worksheet).
262
Appendix L
for K2O analysed the fusion disks of the four test samples (Columns
DM to DY in ‘Conc from SuperQ’-worksheet).
for P2O5 analysed the fusion disks of the four test samples (Columns
EA to EM in ‘Conc from SuperQ’-worksheet).
for Cr2O3 analysed the fusion disks of the four test samples (Columns
EO to FA in ‘Conc from SuperQ’-worksheet).
263
Appendix M
Appendix M
Appendix M depicts the calculation of major elements and associated uncertainties in
the Measurement model 4, in the ‘Final conc’-worksheet of the Excel workbook
template (‘GUM framework.xlsx’). The correlation between the input variables are
calculated in rows 28 to 32. Symbols in the worksheet are as follows:
Ci,meas = Measured concentration of the major element, i, as calculated in the
Measurement model 3 in the ‘Conc from SuperQ’-worksheet (%)
u(Ci,meas) = Uncertainty of the concentration calculated by the Measurement model
3 in the ‘Conc from SuperQ’-worksheet. Two uncertainties are calculated
– one using the sensitivity factors and one without the sensitivity factors
for comparison with the MCM and Kragten approaches
c(Ci,meas) = Sensitivity factor of the measured concentration
%Li2O = Concentration of Li2O calculated in Measurement model 2
u(%Li2O) = Uncertainty of Li2O calculated in Measurement model 2
c(%Li2O) = Sensitivity factor of Li2O
%B2O3 = Concentration of B2O3calculated in Measurement model 2 (%)
u(%B2O3) = Uncertainty of B2O3calculated in Measurement model 2
c(%B2O3) = Sensitivity factor of B2O3
Total = Sum of the major oxides calculated in Measurement model 3 and the
flux compounds
264
Appendix M
Figure M1: The final concentrations (Column Q) and uncertainties (Column S) of

SiO2 analysed in the four test samples (Columns A to S in the ‘Final
conc’-worksheet).
Figure M2: The final concentrations (Column AK) and uncertainties (Column AM)
of TiO2 analysed in the four test samples (Columns V to AM in the ‘Final
conc’-worksheet).
265
Appendix M
Figure M3: The final concentrations (Column BD) and uncertainties (Column AM)
of Al2O3 analysed in the four test samples (Columns AO to BF in the
‘Final conc’-worksheet).
Figure M4: The final concentrations (Column BW) and uncertainties (Column BY)
of Fe2O3 analysed in the four test samples (Columns BH to BY in the
266
Appendix M
Figure M5: The final concentrations (Column CP) and uncertainties (Column CR)
of MnO analysed in the four test samples (Columns CA to CR in the
Figure M6: The final concentrations (Column DI) and uncertainties (Column DK) of
MgO analysed in the four test samples (Columns CT to DK in the ‘Final
conc’-worksheet).
267
Appendix M
Figure M7: The final concentrations (Column DM) and uncertainties (Column ED)
of CaO analysed in the four test samples (Columns EB to ED in the
Figure M8: The final concentrations (Column EU) and uncertainties (Column EW)
of Na2O analysed in the four test samples (Columns EF to EW in the
268
Appendix M
Figure M9: The final concentrations (Column FN) and uncertainties (Column FP)
of K2O analysed in the four test samples (Columns EY to FP in the
Figure M10: The final concentrations (Column GG) and uncertainties (Column GI)
of P2O5 analysed in the four test samples (Columns FQ to GI in the
269
Appendix M
Figure M11: The final concentrations (Column GZ) and uncertainties (Column HB)
of Cr2O3 analysed in the four test samples (Columns GK to HB in the
270
Appendix N
Appendix N
Appendix N portrays the calculation of major elements and associated uncertainties in
the Measurement model 3 for test sample MzBP-1 in the Excel workbook template for
the Monte Carlo Method (‘MCM MzBP-1.xlsx’). The symbols are as follows:

P = Precision value of the measuring cup, fluxer and Dilution
271
Appendix N
Figure N1: Matrix corrected counts and associated uncertainties for the calculation
of SiO2 (cell AS4) and its uncertainty (cell AS5) in the measurement
model 3 using the MCM method (Columns AJ to AS in the ‘MCM MzBP-
1.xlsx’ Excel workbook).
272
Appendix N
of TiO2 (cell BF4) and its uncertainty (cell BF5) in the measurement
model 3 using the MCM method (Columns AW to BF in the ‘MCM MzBP-
273
Appendix N
of Al2O3 (cell BQ4) and its uncertainty (cell BQ5) in the measurement
model 3 using the MCM method (Columns BH to BQ in the ‘MCM MzBP-
274
Appendix N
of Fe2O3 (cell CB4) and its uncertainty (cell CB5) in the measurement
model 3 using the MCM method (Columns BS to CB in the ‘MCM MzBP-
275
Appendix N
of MnO (cell CM4) and its uncertainty (cell CM5) in the measurement
model 3 using the MCM method (Columns CD to CM in the ‘MCM MzBP-
276
Appendix N
of MgO (cell CX4) and its uncertainty (cell CX5) in the measurement
model 3 using the MCM method (Columns CO to CX in the ‘MCM MzBP-
277
Appendix N
of CaO (cell DI4) and its uncertainty (cell DI5) in the measurement model
3 using the MCM method (Columns CZ to DI in the ‘MCM MzBP-1.xlsx’
Excel workbook).
278
Appendix N
of Na2O (cell DT4) and its uncertainty (cell DT5) in the measurement
model 3 using the MCM method (Columns DK to DT in the ‘MCM MzBP-
279
Appendix N
of K2O (cell EE4) and its uncertainty (cell EE5) in the measurement
model 3 using the MCM method (Columns DV to EE in the ‘MCM MzBP-
280
Appendix N
of P2O5 (cell EP4) and its uncertainty (cell EP5) in the measurement
model 3 using the MCM method (Columns EG to EP in the ‘MCM MzBP-
281
Appendix N
of Cr2O5 (cell FA4) and its uncertainty (cell FA5) in the measurement
model 3 using the MCM method (Columns ER to FA in the ‘MCM MzBP-
282
Appendix O
Appendix O
In Appendix J, the Measurement model 4’s calculations per major element by the MCM
in the Excel workbook (MCM MzBP-1.xlsx) are depicted for the test sample, MzBP-1.
Values of the flux compounds and associated uncertainties (%Li 2O and %B2O3) are
obtained using measurement model 2 in cells U4 and AF4, respectively, while the LOI-
value is from cell I4. Their associated uncertainty values are referred similarly.
Figure O1: Measurement model 4 processing the major elements’ results from
model 1 and the flux compounds to estimate SiO 2 in cell FV4 and
associated uncertainty in cell FV5 (Columns FE to FV in the ‘MCM
MzBP-1.xlsx’ Excel workbook).
283
Appendix O
model 1 and the flux compounds to estimate TiO 2 in cell GO4 and
associated uncertainty in cell GO5 (Columns FX to GO in the ‘MCM MzBP-
model 1, the flux compounds to estimate Al2O3 in cell HH4 and associated
uncertainty in cell HH5 (Columns GQ to HH in the ‘MCM MzBP-1.xlsx’
Excel workbook).
284
Appendix O
model 1, the flux compounds to estimate Fe2O3 in cell IA4 and associated
uncertainty in cell IA5 (Columns HJ to IA in the ‘MCM MzBP-1.xlsx’ Excel
workbook).
model 1 and the flux compounds to estimate MnO in cell IT4 and
associated uncertainty in cell IT5 (Columns IC to IT in the ‘MCM MzBP-
285
Appendix O
model 1 and the flux compounds to estimate MgO in cell JM4 and
associated uncertainty in cell JM5 (Columns IV to JM in the ‘MCM MzBP-
model 1 and the flux compounds to estimate CaO in cell KF4 and
associated uncertainty in cell KF5 (Columns JO to KF in the ‘MCM MzBP-
286
Appendix O
model 1 and the flux compounds to estimate Na 2O in cell KY4 and
associated uncertainty in cell KY5 (Columns KH to KY in the ‘MCM MzBP-
model 1 and the flux compounds to estimate K 2O in cell LR4 and
associated uncertainty in cell LR5 (Columns LA to LR in the ‘MCM MzBP-
287
Appendix O
model 1 and the flux compounds to estimate P2O5 in cell MK4 and
associated uncertainty in cell MK5 (Columns LT to MK in the ‘MCM
MzBP-1.xlsx’ Excel workbook).
model 1 and the flux compounds to estimate Cr2O3 in cell ND4 and
associated uncertainty in cell ND5. (Columns MM to ND in the ‘MCM
MzBP-1.xlsx’ Excel workbook). The sum of the final major element
results are in cell NG4 with the uncertainty of the sum in cell NG5
288
Appendix P
Appendix P
Measurement model 4’s frequency plots per major element as created by the MCM
approach to determine uncertainty of measurement in the Excel template are depicted
in this Appendix for test sample, MzBP-1. The black dotted line represents the
measured value, the green lines the uncertainty of the value with the red lines
respresenting the expanded uncertainty at confidence level of 95.45% (k=2).
Figure P1: Frequency plot of TiO2 (1.07%) with an uncertainty of 0.014% and
289
Appendix P
Figure P2: Frequency plot of Al2O3 (19.54%) with an uncertainty of 0.077% and
Figure P3: Frequency plot of Fe2O3 (4.78%) with an uncertainty of 0.035% and
290
Appendix P
Figure P4: Frequency plot of MnO (0.141%) with an uncertainty of 0.0033% and
Figure P5: Frequency plot of MgO (1.10%) with an uncertainty of 0.011% and
291
Appendix P
Figure P6: Frequency plot of CaO (3.73%) with an uncertainty of 0.067% and
Figure P7: Frequency plot of Na2O (6.82%) with an uncertainty of 0.034% and
292
Appendix P
Figure P8: Frequency plot of K2O (3.84%) with an uncertainty of 0.057% and
Figure P9: Frequency plot of P2O5 (0.61%) with an uncertainty of 0.008% and
293
Appendix P
Figure P10: Frequency plot of Cr2O3 (0.018%) with an uncertainty of 0.0017% and
Figure P11: Frequency plot of the sum of the major oxides and LOI (99.74%) with an
uncertainty of 0.198% and expanded uncertainty of 0.396% at k= 2.
294
Appendix Q
Appendix Q
The Kragten method as applied to the measurement models 3 and 4 for test sample MzBP-1 is illustrated in Appendix Q. The used
worksheet is in an Excel workbook, Kragten method.xlsx.
295
Appendix Q
Figure Q1: The TiO2-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
It is from rows 116 to 144 in the same worksheet in Fig. 4.17. The uncertainty budget for this model is in row 144, while
the uncertainty value of TiO2 is in cell C141. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the CSE and calibration slope are the main contributors to TiO 2’s uncertainty for test sample
MzBP-1.
296
Appendix Q
Figure Q2: The Al2O3-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of Al2O3 is in cell C171. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope and the CSE are the main contributors to Al2O3’s uncertainty for test sample
MzBP-1.
297
Appendix Q
Figure Q3: The Fe2O3-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of Fe2O3 is in cell C201. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope and the CSE are the main contributors to Fe 2O3’s uncertainty for test sample
MzBP-1.
298
Appendix Q
Figure Q4: The MnO-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of MnO is in cell C232. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the CSE, the intercept on the Y-axis and the calibration slope are the main contributors to MnO’s
299
Appendix Q
Figure Q5: The MgO-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of MgO is in cell C263. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope, the CSE and the intercept on the Y-axis are the main contributors to MgO’s
300
Appendix Q
Figure Q6: The CaO-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of CaO is in cell C294. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope is the main contributor to CaO’s uncertainty for test sample MzBP-1.
301
Appendix Q
Figure Q7: The Na2O-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of Na2O is in cell C324. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope and the precision of dilution are the main contributors to Na 2O’s uncertainty
for test sample MzBP-1.
302
Appendix Q
Figure Q8: The K2O-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method. It
is from rows 329 to 357 in the same worksheet in Fig. 4.17. The uncertainty budget for this model is in row 357, while
the uncertainty value of K2O is in cell C354. The contributions to the uncertainty budget are plotted in a bar graph which
illustrates that the calibration slope is the main contributor to K 2O’s uncertainty for test sample MzBP-1.
303
Appendix Q
Figure Q9: The P2O5-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of P2O5 is in cell C384. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the calibration slope and the CSE are the main contributors to P 2O5’s uncertainty for test sample
MzBP-1.
304
Appendix Q
Figure Q10: The Cr2O3-value and uncertainty in measurement model 3 for test sample MzBP-1 determined by the Kragten method.
the uncertainty value of Cr2O3 is in cell C414. The contributions to the uncertainty budget are plotted in a bar graph
which illustrates that the CSE, calibration slope and the intercept on the Y-axis are the main contributors to Cr2O3’s
305
Appendix Q
Figure Q11(a): The results and associated uncertainties from measurement model 3 are arranged from columns C to AG in
rows 424 and 425 for use in the measurement model 4 for test sample MzBP-1 determined by the Kragten
method. In this figure, SiO2 to Na2O are in columns C to R.
Figure Q11(b): K2O to Cr2O3-values from measurement model 3 are in columns S to X, while the flux components as calculated
in model 2 are in columns Z to AC. The sum of results in model 3 and its uncertainty are in AD and AE, with the
LOI-value and uncertainty as calculated previously (Fig. 4.17) are in columns AF and AG. All the values in (a)
and (b) refer to relevant cells to prevent errors when copying and pasting values.
306
Appendix Q
Figure Q12: The final SiO2-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
method. It is from rows 428 to 452 in the same worksheet in Fig. 4.17. The uncertainty budget for this model is in row
452, while the uncertainty value of SiO2 is in cell C448.
307
Appendix Q
Figure Q13: The final TiO2-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
479, while the uncertainty value of TiO2 is in cell C475.
308
Appendix Q
Figure Q14: The final Al2O3-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
506, while the uncertainty value of Al2O3is in cell C502.
309
Appendix Q
Figure Q15: The final Fe2O3-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
534, while the uncertainty value of Fe2O3is in cell C530.
310
Appendix Q
Figure Q16: The final MnO-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
562, while the uncertainty value of MnO is in cell C558.
311
Appendix Q
Figure Q17: The final MgO-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
590, while the uncertainty value of MgO is in cell C586.
312
Appendix Q
Figure Q18: The final CaO-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
619, while the uncertainty value of CaO is in cell C615.
313
Appendix Q
Figure Q19: The final Na2O-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
648, while the uncertainty value of Na2O is in cell C644.
314
Appendix Q
Figure Q20: The final K2O-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
677, while the uncertainty value of K2O is in cell C673.
315
Appendix Q
Figure Q21: The final P2O5-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
706, while the uncertainty value of P2O5 is in cell C702.
316
Appendix Q
Figure Q22: The final Cr2O3-value and uncertainty calculated in measurement model 4 for test sample MzBP-1 by the Kragten
735, while the uncertainty value of Cr2O3 is in cell C731.
317
Appendix R
Appendix R
The concentration and uncertainty values as calculated by Xgenline v8.1 for the four
test materials are listed in this Appendix.
Table R1: Concentration and uncertainty values calculated by Xgenline.
SARM-36 C1 MzPP-1 Li3 3c G00006619001
SiO2 63.99 57.52 0.192 46.15
u(SiO2) 0.192 0.190 0.0062 0.137
TiO2 0.043 1.06 0.045 1.54
u(TiO2) 0.0015 0.007 0.0008 0.009
Al2O3 17.19 19.50 0.03 16.63
u(Al2O3) 0.057 0.070 0.0028 0.053
Fe2O3 1.41 4.76 0.455 9.82
u(Fe2O3) 0.008 0.021 0.0032 0.033
MnO 0.004 0.134 0.126 0.174
u(MnO) 0.0018 0.0031 0.0017 0.0031
MgO 0.59 1.30 5.13 7.86
u(MgO) 0.004 0.007 0.021 0.026
CaO 0.693 3.79 49.80 10.67
u(CaO) 0.0043 0.015 0.192 0.035
Na2O 0.41 6.87 LLD 3.64
u(Na2O) 0.003 0.025 0.013
K2O 15.61 3.82 LLD 0.685
u(K2O) 0.049 0.015 0.0038
P2O5 0.12 0.61 0.01 0.27
u(P2O5) 0.002 0.004 0.001 0.002
Cr2O3 0.016 0.025 0.008 0.080
u(Cr2O3) 0.0011 0.0013 0.0006 0.0016
LOI 0.19 0.40 44.54 2.83
u(LOI) 0.002 0.010 0.090 0.005
Total 100.28 99.78 100.34 100.34
u(Total) 0.212 0.229 0.357 0.173
318
Appendix S
Appendix S
The uncertainty budgets of the four measurement models as determined by the
Kragten methods are listed in Appendix S.
Table S1: Uncertainty budget of measurement model 1. All values are in %.
Test sample LOI u(LOI) Cr Cr100 Cr1000 P
SARM-36 C1 0.19 0.002 0.0 50.0 50.0 0.0
MzBP-1 0.40 0.010 0.0 50.0 50.0 0.0
Li3 6c 44.56 0.090 0.0 0.0 0.0 100.0
G00006619001 2.83 0.005 0.0 49.8 49.8 0.4
Table S2: Uncertainty budget of measurement model 2 to determine Li2O. All values
are in %.
Test sample Li2O u(Li2O) MLi2O MB2O3 Li2B4O7 LiBO2 massFlux massSample
SARM-36 C1 19.31 0.018 98.1 1.1 0.5 0.3 0.0 0.0
MzBP-1 21.19 0.019 98.0 1.1 0.2 0.7 0.0 0.0
Li3 6c 22.81 0.020 97.8 1.1 0.1 1.0 0.0 0.0
G00006619001 19.31 0.018 98.1 1.1 0.5 0.3 0.0 0.0
Table S3: Uncertainty budget of measurement model 2 to determine B 2O3. All values
are in %.
Test sample Li2O u(Li2O) MLi2O MB2O3 Li2B4O7 LiBO2 massFlux massSample
SARM-36 C1 19.31 0.018 87.4 1.0 9.6 1.7 0.0 0.3
MzBP-1 21.19 0.019 90.3 1.0 4.9 3.5 0.0 0.2
Li3 6c 22.81 0.020 91.2 1.0 2.2 5.4 0.0 0.2
G00006619001 19.31 0.018 87.4 1.0 9.6 1.7 0.0 0.3
319
Appendix S
Table S4: Uncertainty budget of measurement model 3 to determine the major

elements by SuperQ. All values are in %.
P = Precision value of the measuring cup, fluxer and Dilution
Concen- Uncer-
Test sample MII Ei Di PCup PFluxer PDilution
tration tainty
SiO2
SARM-36 C1 5.9940 0.0290 4.7 93.9 0.2 0.0 0.1 1.0
MzBP-1 5.1353 0.0249 5.3 93.6 0.3 0.0 0.1 0.8
Li3 6c 0.0328 0.0014 15.2 1.2 83.5 0.0 0.0 0.0
G00006619001 4.4739 0.0217 5.4 93.4 0.3 0.0 0.3 0.6
TiO2
SARM-36 C1 0.0044 0.0002 68.7 1.3 30.0 0.0 0.0 0.0
MzBP-1 0.0952 0.0007 65.9 32.5 1.6 0.0 0.0 0.0
Li3 6c 0.0079 0.0002 66.7 4.1 29.2 0.0 0.0 0.0
G00006619001 0.1498 0.0009 52.6 46.3 0.9 0.0 0.0 0.2
Al2O3
SARM-36 C1 1.6100 0.0057 26.5 61.7 11.6 0.0 0.1 0.2
MzBP-1 1.7397 0.0060 25.8 63.7 10.2 0.0 0.1 0.2
Li3 6c 0.0072 0.0020 4.9 0.0 95.1 0.0 0.0 0.0
G00006619001 1.6116 0.0055 22.5 64.9 12.2 0.0 0.2 0.2
Fe2O3
SARM-36 C1 1.6100 0.0057 40.0 59.4 0.7 0.0 0.0 0.0
MzBP-1 1.7397 0.0060 18.6 81.3 0.1 0.0 0.0 0.0
Li3 6c 0.0072 0.0020 44.2 54.0 1.8 0.0 0.0 0.0
G00006619001 1.6116 0.0055 8.5 91.3 0.0 0.0 0.0 0.1
MnO
SARM-36 C1 0.0008 0.0002 81.9 0.0 18.1 0.0 0.0 0.0
MzBP-1 0.0126 0.0003 90.2 2.6 7.2 0.0 0.0 0.0
Li3 6c 0.0221 0.0003 85.6 7.6 6.8 0.0 0.0 0.0
G00006619001 0.0176 0.0003 89.9 4.2 5.9 0.0 0.0 0.0
320
Appendix S
Table S4: Continued

Concen- Uncer-
Test sample MII Ei Di PCup PFluxer PDilution
tration tainty
MgO
SARM-36 C1 0.0378 0.0005 45.8 34.4 19.8 0.0 0.0 0.0
MzBP-1 0.0982 0.0010 24.4 69.6 5.9 0.0 0.0 0.0
Li3 6c 0.8751 0.0073 2.4 94.5 0.1 0.0 0.7 2.3
G00006619001 0.7640 0.0064 2.5 95.1 0.1 0.0 0.5 1.8
CaO
SARM-36 C1 0.0635 0.0012 8.2 91.5 0.3 0.0 0.0 0.0
MzBP-1 0.3322 0.0062 1.8 98.2 0.0 0.0 0.0 0.0
Li3 6c 8.2508 0.1528 0.1 98.5 0.0 0.0 0.4 1.0
G00006619001 1.0164 0.0187 0.6 99.4 0.0 0.0 0.0 0.0
Na2O
SARM-36 C1 0.0391 0.0006 25.2 7.2 67.6 0.0 0.0 0.0
MzBP-1 0.6070 0.0025 0.1 83.2 3.2 0.0 1.7 11.7
Li3 6c <0.02 0.0005 14.2 0.0 85.8 0.0 0.0 0.0
G00006619001 0.3498 0.0016 16.8 70.9 8.3 0.0 0.7 3.3
K2O
SARM-36 C1 1.4651 0.0222 0.5 99.5 0.0 0.0 0.0 0.0
MzBP-1 0.3420 0.0052 1.9 98.0 0.0 0.0 0.0 0.0
Li3 6c <0.002 0.0001 26.5 0.2 73.2 0.0 0.0 0.0
G00006619001 0.0671 0.0011 8.7 90.5 0.8 0.0 0.0 0.0
P2O5
SARM-36 C1 0.0111 0.0002 43.1 30.7 26.2 0.0 0.0 0.0
MzBP-1 0.0544 0.0007 15.4 81.6 2.9 0.0 0.0 0.0
Li3 6c <0.01 0.0002 37.1 0.2 62.8 0.0 0.0 0.0
G00006619001 0.0255 0.0004 24.8 64.8 10.5 0.0 0.0 0.0
Cr2O3
SARM-36 C1 0.0009 0.0001 58.8 21.7 19.5 0.0 0.0 0.0
MzBP-1 0.0017 0.0002 46.0 42.4 11.6 0.0 0.0 0.0
Li3 6c 0.0023 0.0002 22.3 67.7 10.0 0.0 0.0 0.0
G00006619001 0.0073 0.0005 9.6 89.1 1.3 0.0 0.0 0.0
321
Appendix S
Table S5: Uncertainty budget of measurement model 4 to determine SiO2 in

contributors to SiO2’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 64.08 57.69 0.197 46.24
u(SiO2) 0.193 0.147 0.009 0.160
SiO2 44.9 79.3 90.4 63.7
TiO2 0.0 0.1 0.0 0.1
Al2O3 3.3 5.9 0.0 2.3
Fe2O3 0.1 1.5 0.0 3.1
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.1 0.0 3.1
CaO 0.2 6.2 9.4 26.7
Na2O 0.0 1.0 0.0 0.2
K2O 50.6 4.4 0.0 0.1
P2O5 0.0 0.1 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.3 0.6 0.0 0.3
B2O3 0.4 0.6 0.0 0.3
LOI 0.0 0.2 0.1 0.0
Table S6: Uncertainty budget of measurement model 4 to determine TiO 2 in

contributors to TiO2’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.047 1.069 0.048 1.55
u(SiO2) 0.0016 0.008 0.0011 0.010
SiO2 0.6 11.8 0.0 10.3
TiO2 98.9 85.8 67.2 79.2
Al2O3 0.0 0.7 0.0 0.7
Fe2O3 0.0 0.2 0.0 0.9
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.1 0.9
CaO 0.0 0.7 32.2 7.7
Na2O 0.0 0.1 0.0 0.1
K2O 0.4 0.5 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.1 0.0 0.1
B2O3 0.0 0.1 0.0 0.1
LOI 0.0 0.0 0.5 0.0
322
Appendix S
Table S7: Uncertainty budget of measurement model 4 to determine Al 2O3 in

contributors to Al2O3’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 17.21 19.54 0.04 16.66
u(SiO2) 0.080 0.077 0.012 0.068
SiO2 36.5 41.8 0.0 25.5
TiO2 0.0 0.0 0.0 0.0
Al2O3 41.5 52.1 99.8 50.3
Fe2O3 0.1 0.6 0.0 2.2
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.1 0.0 2.2
CaO 0.1 2.6 0.2 19.1
Na2O 0.0 0.4 0.0 0.1
K2O 21.5 1.9 0.0 0.1
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.1 0.2 0.0 0.2
B2O3 0.2 0.3 0.0 0.2
LOI 0.0 0.1 0.0 0.0
Table S8: Uncertainty budget of measurement model 4 to determine Fe 2O3 in

contributors to Fe2O3’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 1.42 4.78 0.465 9.84
u(SiO2) 0.013 0.035 0.0075 0.067
SiO2 9.7 12.0 0.0 9.4
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.4 0.7 0.0 0.6
Fe2O3 84.1 85.7 28.9 81.9
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.2 0.8
CaO 0.0 0.7 69.8 7.0
Na2O 0.0 0.1 0.0 0.1
K2O 5.7 0.5 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.1 0.0 0.1
B2O3 0.0 0.1 0.1 0.1
LOI 0.0 0.0 1.0 0.0
323
Appendix S
Table S9: Uncertainty budget of measurement model 4 to determine MnO in

contributors to MnO’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.009 0.142 0.133 0.009
u(SiO2) 0.0019 0.0033 0.0025 0.0019
SiO2 0.0 1.2 0.0 0.0
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.0 0.1 0.0 0.0
Fe2O3 0.0 0.0 0.0 0.0
MnO 100.0 98.5 49.2 100.0
MgO 0.0 0.0 0.1 0.0
CaO 0.0 0.1 49.9 0.0
Na2O 0.0 0.0 0.0 0.0
K2O 0.0 0.1 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.0 0.0 0.0
B2O3 0.0 0.0 0.0 0.0
LOI 0.0 0.0 0.7 0.0
Table S10: Uncertainty budget of measurement model 4 to determine MgO in

contributors to MgO’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.404 1.10 5.25 7.90
u(SiO2) 0.0058 0.011 0.082 0.065
SiO2 3.8 6.4 0.0 6.3
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.1 0.4 0.0 0.4
Fe2O3 0.0 0.1 0.0 0.6
MnO 0.0 0.0 0.0 0.0
MgO 93.7 92.2 24.3 87.9
CaO 0.0 0.4 74.5 4.7
Na2O 0.0 0.1 0.0 0.0
K2O 2.3 0.3 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.0 0.1 0.0
B2O3 0.0 0.0 0.1 0.1
LOI 0.0 0.0 1.1 0.0
324
Appendix S
Table S11: Uncertainty budget of measurement model 4 to determine CaO in

contributors to CaO’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.679 3.731311 49.48 10.50
u(SiO2) 0.0132 0.0676 0.253 0.176
SiO2 2.1 2.0 0.1 1.5
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.1 0.1 0.1 0.1
Fe2O3 0.0 0.0 0.0 0.1
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 1.7 0.1
CaO 96.6 97.7 87.0 98.1
Na2O 0.0 0.0 0.0 0.0
K2O 1.2 0.1 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.0 0.5 0.0
B2O3 0.0 0.0 0.5 0.0
LOI 0.0 0.0 10.1 0.0
Table S12: Uncertainty budget of measurement model 4 to determine Na 2O in

contributors to Na2O’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.418 6.818978 <0.02 3.61
u(SiO2) 0.0061 0.032818 0.003 0.019
SiO2 3.7 28.2 0.0 15.4
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.1 1.7 0.0 1.0
Fe2O3 0.0 0.4 0.0 1.4
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.0 1.3
CaO 0.0 1.7 0.4 11.5
Na2O 94.0 66.3 99.6 69.0
K2O 2.2 1.2 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.2 0.0 0.1
B2O3 0.0 0.2 0.0 0.1
LOI 0.0 0.0 0.0 0.0
325
Appendix S
Table S13: Uncertainty budget of measurement model 4 to determine K 2O in

contributors to K2O’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 15.66 3.84 <0.002 0.694
u(SiO2) 0.208 0.058 0.001 0.0111
SiO2 4.4 2.9 0.0 1.7
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.2 0.2 0.0 0.1
Fe2O3 0.0 0.0 0.0 0.1
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.0 0.1
CaO 0.0 0.2 0.2 1.2
Na2O 0.0 0.0 0.0 0.0
K2O 95.3 96.6 99.8 96.7
P2O5 0.0 0.0 0.0 0.0
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.0 0.0 0.0
B2O3 0.0 0.0 0.0 0.0
LOI 0.0 0.0 0.0 0.0
Table S14: Uncertainty budget of measurement model 4 to determine P 2O5 in

contributors to P2O5’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.12 0.61 <0.01 0.26
u(SiO2) 0.003 0.008 0.001 0.004
SiO2 1.6 3.5 0.0 1.8
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.1 0.2 0.0 0.1
Fe2O3 0.0 0.1 0.0 0.2
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.0 0.2
CaO 0.0 0.2 0.2 1.4
Na2O 0.0 0.0 0.0 0.0
K2O 0.9 0.2 0.0 0.0
P2O5 97.4 95.7 99.8 96.3
Cr2O3 0.0 0.0 0.0 0.0
Li2O 0.0 0.0 0.0 0.0
B2O3 0.0 0.0 0.0 0.0
LOI 0.0 0.0 0.0 0.0
326
Appendix S
Table S15: Uncertainty budget of measurement model 4 to determine Cr2O3 in

contributors to Cr2O3’s uncertainty are the major elements from SuperQ
SARM-36 C1 MzBP-1 Li36c G00006619001
SiO2 0.010 0.019 0.014 0.075
u(SiO2) 0.0013 0.0018 0.0010 0.0050
SiO2 0.0 0.1 0.0 0.1
TiO2 0.0 0.0 0.0 0.0
Al2O3 0.0 0.0 0.0 0.0
Fe2O3 0.0 0.0 0.0 0.0
MnO 0.0 0.0 0.0 0.0
MgO 0.0 0.0 0.0 0.0
CaO 0.0 0.0 3.1 0.1
Na2O 0.0 0.0 0.0 0.0
K2O 0.0 0.0 0.0 0.0
P2O5 0.0 0.0 0.0 0.0
Cr2O3 99.9 99.9 96.9 99.8
Li2O 0.0 0.0 0.0 0.0
B2O3 0.0 0.0 0.0 0.0
LOI 0.0 0.0 0.0 0.0
327

Cloete Huibrecht Wtms

Uploaded by

Copyright:

Available Formats

Cloete Huibrecht Wtms

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cloete Huibrecht Wtms

Uploaded by

Copyright:

Available Formats

COPYRIGHT AND CITATION CONSIDERATIONS FOR THIS THESIS/ DISSERTATION

How to cite this thesis

HUIBRECHT CORNELIA CHRISTINA CLOETE

Thesis in fulfilment of the requirement for the degree

SUPERVISOR: PROF P. N. NOMNGONGO

- God, for all His grace for which I am grateful.

Chapter 1: Introduction and problem statement ........................................................ 1

Chapter 2: Theoretical background and literature review ......................................... 6

Chapter 3: Methodology .............................................................................................. 39

Chapter 4: Results and Discussion............................................................................ 87

Chapter 5: Conclusions ............................................................................................ 151

References ................................................................................................................. 155

Table 1.1: Uncertainty of measurement requirements as listed in

Table R1: Concentration and uncertainty values calculated by XGENLINE.

Table S1: Uncertainty budget of measurement model 1. All values are in %.

In today's analytical environment, calculation of the uncertainty of measurement is not

1.2 Motivations and objectives of the study

It is an ISO/IEC 17025:2017[3] requirement for an accredited test method to have an

Table 1.1: Uncertainty of measurement requirements as listed in ISO/IEC

Various approaches to obtain a practical process to estimate the uncertainty of

1.3 Problem statement

This bottom-up approach in XRF is labelled as a first-principles approach[5], where all

Although the bottom-up approach is time-consuming and requires a sound

1.4 Key objectives of the study

This chapter outlines the terminology used in the theory of uncertainty of

2.2 Concepts in the uncertainty of measurement

2.2.1 Descriptive and inferential statistics

Figure 2.1: Parameters of a normal distribution plot of a population (the Gauss

Mean of a sample set

Variance of a sample set

Coefficients of variation (CV) or 𝑠(𝑥𝑖 )

Table 2.1: Continued

Degrees of freedom ν=n–1

To evaluate the difference between standard

2.2.2 Metrological terms

Table 2.2: Metrological terms used in this study as defined in VIM[10].

As noted in the VIM's definition of measurement in Table 2.2, it is a process of

The measurement result consists of the following information:

2.2.3 Method validation, metrological traceability and uncertainty of

Accreditation bodies such as the South African National Accreditation System

Metrological traceability links the produced measurement result to one or more

Uncertainty of measurement characterizes the dispersion of possible measurand

The tight relationship between metrological traceability, method validation and

Figure 2.5: A simplified illustration of the three metrology cornerstones to evaluate

2.3 Established approaches to estimate the uncertainty of

In 1977, the BIPM was requested to investigate a methodology to evaluate and

2.3.1.1 The GUM framework

Evaluate the covariance 𝑟(𝑥𝑖 , 𝑥𝑗 )between any two input estimates:

5 Calculate y, the estimate of the measurand Y from the function f in Step 2.

𝑢𝑐2 (𝑦) = ∑ 𝑐𝑖2 𝑢2 (𝑥𝑖 ) + 2 ∑ ∑ 𝑐𝑖 𝑐𝑗 𝑢(𝑥𝑖 )(𝑥𝑗 )𝑟(𝑥𝑖 , 𝑥𝑗 )

6 Only if co-variance is present

b) 𝑢𝑐 (𝑦) = √𝑢𝑐2 (𝑦)

The expanded uncertainty, in this case, is then: U = 𝑡𝑝 (𝑢)uc(y)

2.3.1.2 The Monte Carlo method (MCM)

GUM-4[28] states that the GUM framework is not applicable where:

Figure 2.6: The Monte Carlo method (GUM-2)[26].

2.3.1.3 Software tools for GUM based calculations

A spreadsheet is a popular tool to create templates for calculating uncertainty of

2.3.2 Top-down and bottom-up approaches