CASE STUDY
APPLICATION OF FOURIER TRANSFORM INFRA-RED
SPECTROSCOPY (FTIR) FOR MINERAL QUANTIFICATION
ISSUED – APRIL 2019
ABSTRACT
Fourier transform infra-red spectroscopy
is a technique which is used to obtain
an infra-red spectrum of absorption or
emission of solids, liquids or gases.
The absorption excites molecular
vibrations and rotations, which have
frequencies that are the same as those
within the infra-red part of the spectrum.
The frequency of the response relates
to specific bonds and the greater the
response the higher the concentration.
It is therefore possible in a similar way
to how X-ray diffraction (XRD) works to
both identify and quantify a wide range
of compounds.
Minerals contain a wide variety of bonds
and the more open their structure the
better the infra-red response. This
means that minerals which contain
covalent bonds such as clay minerals
can be both identified and quantified
with considerable accuracy. Further
to this the development of new
detectors and sampling plates based
on attenuated total reflection (ATR)
technology means that samples can be
presented to the instrument without
the need for complex preparation, they
simply need to be dry and ground to less
than 75 microns.
Instrument run time is typically two
minutes and SGS has set up fully
quantified methods for clay minerals
including illite, chlorite, kaolin and
smectite, micas, carbonates calcite,
dolomite and siderite, feldspars
plagioclase and k-feldspar, quartz,
spodumene and apatite. The
quantification ranges vary between
minerals but is usually between 5-100%
with error generally less than 15% i.e. R
2 greater than 90 even for clay minerals.
Although the base methodologies are
tolerant of a wide range in matrixes for
optimal results it is always preferable to
set up matrix matched methods.
INTRODUCTION using artificial intelligence (AI). The
frequency region and preprocessing
Fourier-transform infrared spectroscopy
method as well as the resulting
(FTIR) [1] is a technique used to obtain an
RMSECV value were displayed in order
infrared spectrum of absorption or
of relationship to the analyte calibration
emission of a solid, liquid or gas. An FTIR
values. The selection frequency
spectrometer simultaneously collects
parameters were then checked against
high-spectral-resolution data over a wide
the known spectral response to the pure
spectral range. In this study we utilised
minerals to ensure that the parameters
the range 500 – 4000 cm-1 with a
selected were valid. Further to this
resolution of 4 cm-1.
samples which gave a result more than
FTIR spectroscopy is an established 20% from the calibration line were
experimental technique for determining removed as outliers. The method
qualitative mineral identification. In this was then recalibrated with the outliers
paper we examine the use of the Bruker removed.
Alpha II unit (Figure 1) in the quantitative
The samples selected for validation
determination of common rock forming
encompassed the full concentration
minerals.
range for each analyte with the same
For this paper we used attenuated total variability in sample characteristics as
reflection (ATR) sample analysis unit. contained in the calibration set. The
ATR is a sampling technique used in validation set contained about 50%
conjunction with infrared spectroscopy as many samples as the calibration
which enables samples to be examined set and 5 samples were selected as
directly in the solid state without further independent testing.
preparation. [2] ATR uses a property of
total internal reflection resulting in an
evanescent wave. A beam of infrared
light is passed through the ATR crystal in
such a way that it reflects at least once
off the internal surface in contact with
the sample. The ATR system used in
this study was based on a diamond
system which made it ideal for minerals
which have a high hardness index.
EXPERIMENTAL
Rock samples of between 200-300g with
a wide range of mineralogy were dried
and pulverised until 100% passing a 75-
micron sieve. A split was then taken for
analysis by high resolution
X-ray diffraction (XRD) to act as
calibration standards. The remaining
portion was then analysed with the FTIR
using the ATR attachment.
Method optimisation was determined
RESULTS AND CONCLUSIONS MINERAL/ANALYTE R SQUARED MAHALANOBIS DISTANCE
R-squared is a statistical measure of how Apatite 99.59 0.22
close the data are to the fitted regression Calcite 98.66 0.24
line. It is also known as the coefficient Chlorite 99.67 0.44
of determination, or the coefficient
Dolomite 99.05 0.21
of multiple determination for multiple
regression. 0.00% indicates that the illite 98.39 0.55
model explains none of the variability, a Kaolinite 99.86 0.52
result of 100% would mean all the data
K-Feldspar 96.07 0.21
fits the model. So, R-squared can take
value between 0 and 1 where values Muscovite 95.36 0.17
closer to 0 represent a poor fit while Petalite 89.18 0.8
values closer to 1 represent a perfect Plagioclase 95.29 0.31
fit. If R-squared=0.93 then it means
Quartz 98.63 0.17
93% variations in dependent variable
Y are explained by the independent Siderite 98.23 0.26
variables present in our model. The Smecite 98.79 0.18
R-squared results for some of the mineral
Spodumene 97.65 1.2
methods are displayed in Table 1. It is
noted that only one of the methods has
an R-squared value of less than 0.90 As the FTIR only requires that the samples are dry and ground to less than 75 microns
therefore for nearly all the methods this offers the opportunity for low cost, rapid mineralogical analysis without the need
covered in this report have a 90% fit. for complex preparation. This type of instrumentation is portable and robust and as
The presence of outliers is determined the range of minerals which can be quantified expands may supplant more traditional
based on the Mahalanobis distance [3] technologies for a wide range of industrial applications.
and displayed in Table 1.
There are however limitations to what the technology can provide. Firstly, the
The calibration graph for illite is displayed technique relies on there being bonds present within a mineral which have a vibrational
in Figure 2 and the validation graph for mode. Halite and galena do not possess this type of bonding and so will always remain
illite is displayed in Figure 3. Outliers beyond the capability of this type of instrumentation to identify and quantify.
are displayed in red. RMSEE is the root
Secondly although there have been significant improvements in detector technology
mean square error of estimation RPD
limits of quantification for most minerals in a complex rock matrix remain at about 5%
is the residual predication deviation,
w/w.
RMSECV is the root mean square error of
cross validation and bias is the value for
the rank displayed in the graph. R2 Fit vs True / Illite [%] / Calibration
99.15
From this data FTIR can provide a highly 100
accurate method for the determination 90
RMSEE:
of a wide range of minerals. The degree 2.24 80
of accuracy is in general better than
70
that achieved by X-ray diffraction (XRD) RPD:
namely that it has error of less than 10.8 60

10% from true rather than being in the 50

order of 20%. The degree of accuracy 40
is, in general, better than that achieved 30
by X-ray diffraction (XRD) in that it has 20
error of less than 10% from true, rather
10
than being in the order of 20%. This is Line
especially notable for clay minerals which 0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
present considerable difficulties for XRD Color

analysis where preparation is a limiting R2 Prediction vs True / Illite [%] / Cross Validation
factor in accuracy. 98.39
100
90
RMSECV:
2.89 80
70
RPD: 60
7.9
50

Bias: 40
0.173 30
20
10
0
Line
-10
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Color
 FURTHER WORK
As mentioned in the previous section
the list of minerals which can be
analysed and quantified by FTIR is ever
expanding. New methods are currently
under development which will provide
accurate determinations of minerals such
as biotite mica, potash, ferron dolomite,
haematite, goethite, topaz and sellaite.
Recently, we have developed methods
for tantalite and heavy minerals. Other
minerals will soon follow as commercial
requirements dictate.

ACKNOWLEDGMENTS
My co-worker Katie Donnelly and
reviewers Hugh De Souza, Peta Hughes

REFERENCES

1. Griffiths, P.; de Hasseth, J. A. (18

May 2007). Fourier Transform Infrared
Spectrometry (2nd ed.). Wiley-
Blackwell. ISBN 978-0-471-19404-0.
2. FT-IR Spectroscopy—Attenuated
Total Reflectance (ATR)» (PDF). Perkin
Elmer Life and Analytical Sciences.
2005. Archived from the original (PDF)
on 2007-02-16. Retrieved 2007-01-26.
3. Mahalanobis, Prasanta Chandra
(1936). «On the generalised distance
in statistics» (PDF). Proceedings of the
National Institute of Sciences of India.
2 (1): 49–55. Retrieved 2016-09-27
© SGS Group Management SA – 2019 – All rights reserved – SGS is a registered trademark of SGS Group Management SA

ABOUT SGS
SGS is the world’s leading inspection,
verification, testing and certification
company. SGS is recognized as the
global benchmark for quality and
integrity. With more than 97,000
employees, SGS operates a network
of over 2,600 offices and laboratories
around the world.

CONTACTS
SGS AUSTRALIA PTY LTD
John Woods
28 Reid Road, Perth Airport
Western Australia
minerals@sgs.com

SGS IS THE WORLD’S LEADING INSPECTION, VERIFICATION, TESTING AND CERTIFICATION COMPANY