MULTICOMPONENT ANALYSIS OF FTIR SPECTRA: QUANTIFICATION OF AMORPHOUS AND

CRYSTALLIZED MINERAL PHASES IN SYNTHETIC AND NATURAL SEDIMENTS

JACQUES BERTAUX', FRANçOIS FRöHLICH2, AND PHILIPPE ILDEFONSE3
I ORSTOM, Laboratoire des Forriintioiis Siiper$cielles, Progranime PVC, Départenierit RED, 72, route d'Aiilriay, 93143 Bondy cedex, France
Labowtoire de Géologie du Mirshrrii Natiorial d'Histoire Nntiirelle, 43, tue Buffon, 75005 Paris, France
Loborntoire de Miriérnlogie-Crista~lograpliie,UA CNRS 09, Universités Paris 6 et 7 arid IPGP, 4 place Jussieu, 75252 Paris cedex OS

ABSTRACT: The study of weathering and erosion processes requires the matter, either transported to the sedimentary basin or produced in situ. The
establishment of a mass balance of minerals from the soils towards the relative proportions of these constituents can vary to form mixtures ranging
sedimentation,area. This calls for the development of quantitative tools in composition from slightly organic sediments to almost pure organic de-
that can be applied to mixtures of finely divided and sometimes poorly posits like peats. Observation of these variations with time helps in de-
crystallized minerals found in soils and sediments. We present here the scribing the climate variability during the late Quatemary. Such a com-
use of Fourier transform infrared (FTIR).absorbance spectroscopy in positional variability is also found in marine sediments, in soils that rep-
the mid-infrared as a method to calculate mineral modes for such ma- resent the main source of materials available for erosion and sedimentation,
terials. and in eolian dusts (Drees et al. 1993; Rognon et al. 1996).
Samples were prepared using the KBr disc method. This ensures Various techniques have been proposed to monitor the compositional
that Lambert-Beer's law is valid. A quantitative determination of the changes of soils or sediments. Elemental chemical analysis yields very
mineral content from various blends was performed by making a mul- accurate results, but a given element may be present in more than one
ticomponent analysis of the experimental spectrum using the spectra phase and the mineral composition cannot be deduced from the elemental
of each component in the mixture. To check the validity of the pro- composition. Data from selective chemical dissolution, usually used by soil
cedure, the multicomponent analysis was performed on spectra from scientists (Jones 1969; Teissier et al. 1979), are difficult to interpret in the
synthetic mixtures of standard minerals (quartz, kaolinse, gibbsite, case of complex mixtures in which several readily soluble phases are pres-
amorphous silica). Good agreement between actual and computed wt ent (Gehlen and Van Raaphorst 1993). X-ray diffraction, although used
% was obtained in the range 1315-315cm-l. largely for semiquantitative and quantitative determination in rocks and
This application of FTIR spectroscopy was used to quantify the min- sediments, suffers from strong limitations due to the difficulty of analyzing
eral components of a lacustrine sediment cored a t Salitre, Minas Ger- amorphous compounds and disordered clay minerals. Moreover, organic
ais, Brazil. The major phases are organic matter, kaolinite, gibbsite, matter can lead to great difficulties when preparing samples for X-ray dif-
quartz, anatase, and amorphous silica (Phytoliths and sponge spicules). fraction. For example, attempts to eliminate it with strong oxidizers can
Fifty-three samples were analyzed. Quantitative results from FTIR promote the formation of new compounds. Petrographic techniques such
spectroscopy were compared to chemical analyses to assess the validity as the optical microscope are useful for a qualitative description of the
of the method for the natural sediments. Phases with no sharp diag- samples, but their use for quantifying finely divided materials is impossible.
26 nostic spectral features, such as anatase or amorphous silica, were
quantified using the multicomponent analysis.
The aim of this paper is to investigate the use of Fourier transform
infrared (FTIR) transmittance/absorbance spectroscopy, in the mid-infrared
The variations of the main mineral phases along the core reflect the region, for determining the relative abundances of the mineral components
mechanisms of transfer from the surrounding soils to the sediment area in sediments. First, the proposed technique was tested on synthetic mixtures
and the role of the vegetal cover over the drainage area. of crystallized and amorphous mineral components. Then, natural sedi-
ments, originating from Salitre, west Minas Gerais state (Brazil), were an-
alyzed by FTIR spectroscopy, and the resulting modal analyses were com-
pared to chemical determinations of major elements on the same samples.
The abundances of major mineral phases in the sediment are discussed in
~
light of environmental changes over the drainage basin as revealed by the
U) .e ' formation of various form. They are an integration of components from pollen analysis.
CJ two distinct environments (Dean 1993): (1) the extemal terrestrial environ-
8)
t c r
Q 0 ment of the drainage basin and beyond (allochthonous components, mainly MATERIALS AND METHODS
U, 0 from soils), and (2) the intemal lake environment (autochthonous compo-
nents). The allochthonous components are tracers of weathering and erosion Materials
processes. Their behavior during their transport from soils to the lake re-
sults from several factors, including the type of soils, the vegetation cover, Reference Compounds.-The reference compounds consist of natural

1:
the climate, and the drainage pattem over the drainage basin. The autoch- quartz; amorphous silica, or opal A according to the nomenclature of Jones
thonous components are indicators of processes acting within the lake and and Segnit (1971), constituted with Hexactinellid sponge spicules of Ho-
within the postdepositional (diagenetic) environment. In order to assess locene age, collected in the central Indian Basin (Fröhlich 1989); kaolinite
environmental conditions prevailing during sedimentation it is useful to from Macon, Georgia (USA), supplied by Wards; and synthetic gibbsite
perform mass-balance calculations among these various sources. However, supplied by Aluminium Pechiney. Punty of all samples was checked by
quantitative determination of major mineral phases in soils or sediments X-ray diffraction.
remains a difficult task, because the complexity of these natural mixtures Synthetic Mixtures.-Synthetic mixtures of these reference compounds
leads to the difficulty of using conventional methods such as X-ray dif- (Tables 1, 2) were prepared by weighing.They consist either of binary
fraction or optical microscopy. The methodology presented in this paper mixtures (quartz-kaolinite, quartz-gibbsite), or four-component mixtures
was developed to study tropical lacustrine sediments which contain com- (quartz5 kaolinite, gibbsite, amorphous silica). T ese mineral components
mon detrital minerals such as quartz, kaolinite, gibbsite, and amorphous P .
Were chosen because they are very widespread in soils and sediments.
mineral compounds such as biogenic silica in the form of Phytoliths, dia- Soils and Sediments.-The Salitre ultramafic-klkaline complex (west of
toms, and sponge spicules; these mineral phases are associated with organic Minas Gerais State, Brazil)
--.-.__ -- _I- - - has a central,.dolina-+aped
-
depression partially
Fonds Documentaire ORSTOM
JOURNAL
OF SEDIMENTARY VOL.68, NO. 3, MAY,1998, P. 4 4 0 4 7
RESEARCH,
Copyright O 1998, SEPM (Society for Sedimentary Geology) 1073-130X/98/068-0440/$03.00 (',
1I11l11I 010014669
I
h
 QUANTIFICATION OF MINERAL PHASES IN SEDIMENTS BY FTIR SPECTROSCOPY 441

TABLE
l.-Resulrs of the quantitative aiialyses of Ovo coniporierlts synt/ietic 2.-Resdts of the qrrantirative ailalyses of syiithetic mixtures with four
TABLE
rnixtlires. comnporierits.

Quartz Quartz Kaol. Kaol. Gibb. Gibb. Synthetic Quartz Kaolinite Gibbsite Amorphous
Synthetic Actual Computed Actual Computed Actual Computed Mixtures wt % wt % wt % Silica wt %
Mixture wt % wt % wt % wt % wt % Wt 7%
~~
MI 1 25 50 20 5
q80-k20 80 73.3 20 20.7 2 25.1 49.4 21.1 5.1
q60-k40 60 58.8 40 36.0
q40-kG0 40 38.2 60 57.8 M2 I 20 50 20 10
q20-k80 20 19.6 80 75.9 2 21.0 48.6 20.4 9.2
q80-g20 80 77.1 20 18.9 M3 1 10 60 15 15
q60-g40 60 56.4 40 38.8 2 13.1 57.8 17.8 14.9
q40-g60 40 40.3 60 60.3 M4 1 15 40 25 20
qZO-g80 20 19.1 80 81.3 2 19.5 42.0 25.5 21.0
Kaol. = Kaolinite: Gibb. = Gibbsite. 3 39.2 43.2 24.0 n.d.
I = actual wt % 2 = computed wt % 3 = computed wt % after multicomponent analysis using only
+
three end-members spectra (quartz kaolinite t gibbsite): n.d. = not determined.
filled by sediments. A thick, loose latosolic cover (goethite, gibbsite, ka-
olinite, quartz, anatase, crandallite-family phosphates) surrounds the de-
1989; Sunshine and Pieters 1993). Determination of hematite and goethite
pression (Sondag et al. 1996). Fifty-three samples were collected from a 6
concentrations in geological materials was performed by visible reflectance
m-long core (LC3) retrieved from the lake. A high organic-matter content
spectroscopy (Deaton and Balsam 1991; Barranco et al. 1989). Balsam and
(especially plant fibers) characterizes most of the core, as indicated by high Deaton (1996) used UVVNIR reflectance spectroscopy to estimate organic
loss on ignition and total organic carbon contents (ranging from 10 to 50%).
carbon, opal, and carbonate in marine sediments. Bishop et al. (1996) dem-
X-ray diffraction and optical microscopy data showed that quartz, kaolinite, onstrated a correlation between band depths of organic and carbonate fea-
gibbsite, anatase, and amorphous silica are the main inorganic components
tures measured on IR reflectance spectra of lacustrine sediments from Ant-
of the Salitre sediments (Sondag et al. 1993). Because reference spectra arctica and weight percent carbon (as carbonate or as organic) of the same
are required for quantitative analysis of the natural sediments by our meth-
sediments.
od (see below), the following mineral and organic phases were used for
Outline of IR Quantitative Analysis.-We used absorbance mid-infra-
this purpose: red spectroscopy in transmission mode as a quantitative method because it
quartz, gibbsite, and amorphous silica already used in the synthetic
has been demonstrated (see below) that, in given conditions (described in
mixture; the section on sample preparation), the amount of absorbed radiation is
-a kaolinite separated from a soil near Macaus (Brazil); it was chosen
proportional to the quantity of absorbing matter in the sample. This means ,
because of its spectral resemblance to the kaolinite from the Salitre sedi- that IR absorption obeys Lambert-Beer’s law. Lambert-Beer’s law has
ment; both are poorly ordered, as demonstrated by the relative amplitude
been established for dilute liquid solutions, in which the analyzed com-
of the structural OH bands between 3600 and 3700 cm-’ (Cases et al. ponent is evenly dispersed at a low concentration in a non-absorbing me-
1982), with three broad bands at 3620, 3650, and 3698 cm-l. Kaolinite
dium. The following relation exists when monochromatic radiation passes
was separated from the bulk sample after suspension in deionized water through an inert solution in which a substance is dissolved
and centrifugation (10 min at 3000 rpm);
-an anatase reference sample, hand-picked from a soil around the Salitre I = I, exp( -acl)
sedimentary deposit; where I = intensity of transmitted light, I, = intensity of incident light,
-two different references were used for organic matter. For the bottom
c = concentration (molell), 1 = length of sample cell (cm), and a =
part (160-600 cm), the sediment between 160 and 260 cm is purely organic absorption coefficient.
(peat), and the spectrum of a sample at 164 cm was used as the reference
The absorbance A is defined as
spectrum for organic matter. In the upper part (0-160 cm), the sediment
between O and 70 cm is composed of organic matter mixed with amorphous A = log (I,/I)= EC~
silica (Phytoliths and sponge spicules). A reference spectrum for organic where E = molar extinction (or absorbance) coefficient (cmz mole-’) =
matter was obtained by mathematically subtracting the amorphous silica
(l/cZ)log(I,/I)
component from the spectrum of a sample in this part of the core. For a system in which solid particles of the substances are pressed with
KBr to a transparent pellet (the KBr disc method), the following relation
Methods exists (Duyckaerts 1955, 1959):
Background of W I R Quantitative Analysis.-Spectroscopy in the ni k
ultraviolet (UV, 0.2-0.4 pm), visible (V, 0.4-0.7 pm), near-infrared (NIR, A = lOg(I,/I) = - x -
Sp 2.303
0.7-5 pm), and mid-infrared (MIR,2.5-40 pm) regions has been used
extensively for mineralogical studies. The UVVNIR region (0.2-5 pm) is where ni = mass of the sample dispersed in the KBr disc, p = density of
favorable to assess the mineralogy of a mixture because various mecha- the particles, S = surface area of the disc normal to the beam, and k =
nisms can produce photon absorption (Gaffey et al. 1993): electronic tran- absorption coefficient. This relation is the same as Lambert-Beer’s law
sitions in molecular orbitals, vibrational transitions in molecules and crys- when d p = cl and e = 2.303(k/S) are written.
tals, and electronic transition within and between atoms. On the other hand, According to Van der Mare1 and Beutelspacher (1976), Lambert-Beer’s
MIR is the region where the major rock-forming minerals have their fun- law is restricted to monochromatic light, non-interacting absorbers, very
damental molecular vibration bands (Farmer 1974; Salisbury 1993). Spec- dilute solutions (no appreciable change in the refractive index), absence of
troscopic data are measured either in transmission or in reflection mode, a Christiansen effect, particle size smaller than the wavelength used, ab-
and spectra record absorption or emission features. Reflectance spectros- sence of scattering, linearity of the detector, and the assumption that the
copy has been much investigated as a method for mineralogical analyses slits are infinitely small. Despite all these restrictions, it has been empiri-
of planetary surfaces via remote sensing (Bishop et al. 1996), and as a cally demonstrated by numerous authors that mid-IR absorbance of prop-
method for quantitative analysis of mineral mixture (Mustard and Pieters erly ground (< 2 pm), evenly dispersed minerals at low concentration
442 J. BERTAUX ET AL.

I I I I I I I I

F:
:A .
8.
: \
.
I,

1300 1200 1100 1000 900 800 .700 600 500 400 300
Wavenumber (cm-')

Wavelength (pm)
7.7 8.3 9.1 10 11.1 12.5 14.3 16.7 20 25 33.3

u
o
B
P
E
2

FIG.l.--FTIR spectra of synthetic mixtures

(two components ). A) quartz-kaolinite mixtures;
B) quartz-gibbsite mixtures; dashed curves =
end member spectra; open circles = measured
1300 1200 Il00 I000 900 800 700 600 mixture spectra; solid curves = computed
'O0 400 300 mixture spectra. ~8O-k20means a mixture of 80
Wavenumber (cm-') wt % of quartz and 20 wt 9% of kaolinite.

(on the order of 0.2-0.3%) in a KBr pellet shows a linear correlation with than 2 pm is required to avoid excessive scattering of IR radiation. Then
concentration;see for instance the studies of Duyckaerts (1959) for calcite, the powder was carefully mixed by hand with KBr in an agate mortar. A
Chester and Elderfield (1967) for calcite, dolomite, and aragonite, Chester dilution of 0.25% was used for all samples studied. Weighing was per-
and Elderfield (1968) for amorphous silica and quartz, Van der Mare1 and formed within a dry atmosphere, with an accuracy of g. A 300 mg
Beutelspacher (1976) for clay minerals, Pichard and Fröhlich (1986) for pellet, 13 mm in diameter, was prepared by pressing the mixture in a
quartz and calcite, and Fröhlich (1989) for silica polymorphs. This point vacuum die, with up to 8 tonscm-2 of compression. The pellets were oven-
is very important because it provides an experimental confirmation that in dried two days at 110°C before data acquisition.
given conditions Lambert-Beer's law can be applied to a suspension of Data Acquisition.-IR spectra were recorded on a Perkin-Elmer FT 16
solid particles, as ground particles dispersed in KBr. PC spectrometer in the 4000-250 cm-' energy range with a 2 cm-' res-
Sample Preparation.-All samples (reference compounds, synthetic olution. For each spectrum, 50 scans were cumulated. Absorbance was
mixtures, and natural sediments) were analyzed using KBr pellets prepared computed relative to a background of sample holder containing a pure KBr
as follows: the samples were mechanically ground with small agate balls pellet.
in an agate vial under acetone and in a refrigerated area (4°C) in order to Data Reduction,-The spectral-energy range between 1315 and 315
prevent heating and structural changes of particles. A particle size of less cm-l was chosen for calculation since this region yields many absorption
3

.i
QUANTIFICATION OF MINERAL PHASES IN SEDIMENTS BY FTIR SPECTROSCOPY 443

Wavelength (pm)
7.7 8.3 9.1 10 11.1 12.5 14.3 16.7 20 25 33.3
, l I l I I J 1 ~ 1 ~ 1 t l

FIG.2.--FTIR spectra of synthetic mixtures

(four components). dashed curves = end-
member spectra; open circles = measured
mixture spectra; solid curves = computed
mixture spectra; M4*indicates the superposition
of the spectrum from mixture M4 and the
computed spectrum using only three components
+ +
(quartz kaolinite gibbsite); vertical arrows
show region where the computed spectrum using
1300 1200 1100 1000 900 800
700 600 500 400 300 three components differs significantly from the
measured M4 spectrum, and related to the
Wavenumber (cm-’) amorphous silica.

features that are relevant for distinguishing mineral components in mixture. ing after heating to 1000°C. The chemical composition was determined
We used the method described by Saarinen and Kauppinen (1991) for mul- after melting with LiBO, and dilution in 2N HNO,. The content of SiO,,
ticomponent analysis of FTIR spectra of .mixtures of gases. Starting from Alzo,,Fe,O,, Tio,, MnO, and P,O, were analyzed using ICP-ES, while
a set of spectra of pure components, they demonstrated that the spectrum Na,O, K,O, MgO, and Ca0 were analyzed using AAS. Carbon was deter-
of the mixture can be mathematically processed to quantify the contribution mined with a CHN elemental analyzer.
of each component. The measured absorbance spectrum is recorded with a
constant wave-vector step and thus corresponds to a S,(i) function [i = RESULTS AND DISCUSSION
1, . . . ,N I , where N is the number of measured points. According to Lam- Calibration Procedure with Synthetic Mixtures
bert-Beer’s law, the measured absorbance spectrum S,(i) of a mixture can
be approximated by a computed spectrum S,( i) that is a linear combination Two-Component Mixtures in KBr.-Spectra of these mixtures
of M reference spectra &?(i) [i = l,,th. . . , M, i= 1,. . . , NI recorded in (quartz-kaolinite and quartz-gibbsite) were measured, and these spectra
the same conditions (i.e., quartz, kaolinite,. . .): were then compared to spectra of the mineral components (Fig. lA, B).
M
Some characteristicspectral features can be recognized in the mixture spec-
[ i = 1, . . . , NI tra after comparison with the end-member spectra; for instance, the doublet
S,(i) = xjKj(i)
j= 1 at 914-936 cm-I corresponding to the Al-OH bending vibrations of ka-
olinite (Van der Mare1 and Krohmer 1969), the doublet at 780-798 cm-l
Determination of the xj coefficients to obtain a least-squares adjustment due to the Si-O-Si intertetrahedral bridging bonds in quartz (Moenke
between the computed and measured spectra is done by minimizing the 1974), and the OH deformation band of gibbsite at 1000 cm-’ are easily
quantity recognized. Our model was used to generate a computed mixture spectrum
from the end-member spectra. The computed mixture spectra are super-
[s‘(ij - s(i)]z imposed on the measured mixture spectra (Fig. 1). The high quality of the
i= 1
fits shows that an IR spectrum of a two-component mixture can be calcu-
Calculations were performed using the “GRAPH” software (Bonnin et lated from the individual IR spectra of the components in the mixture. Our
al. 1989). The scale factors xj associated with each component spectrum model was then applied to the mixture spectra in order to determine the
reflect the proportion of the component in the mixture, because in our study wt % of the end members (referred to as computed wt %) in these known
reference spectra and mixture (artificial or natural) spectra correspond to mixtures, which are shown in Table 1. Computed wt % of quartz, kaolinite,
0.75 mg of a given constituent or blend of constituents in a 300 mg KBr and gibbsite are in good agreement with the actual (weighed) wt %. The
pellet. maximum relative error is 10%.
Chemical Analyses.-After drying at 60”C, the samples were crushed Four-Component Mixtures in KBr.-The measured spectra of the
and sieved (200 pm). The loss on ignition (LOI) was determined by weigh- four-component mixtures (quartz, kaolinite, gibbsite, and amorphous silica)
 b'
444 J.BERTAUX ET AL.

Wavelength (pm)
5 2.9 3.3 4 5 6.7 10 20 coreLC3
I I I I I I f I I I c t I I I I I I I I I I I I I I I I Sampledepth
(cm)
17-18
38-39
50-51
68-69
!-

80-81
105-107
129-131

n JL-/ 158-159
164-165
286-290

K d!
302-303

387-391

1.50-452

J .
wS.sil.+K+G+d
- 342-546 FIG.3.--FTIR spectra from selected sediment
samples from the LC3 core. O.M. = organic
4000 3500 3000 2500 , 2000 1500 1000 500 matter; am. sil. = amorphous silica; K =
kaolinite; G = gibbsite; Q = quartz; A =
Wavenumber (cm-') anatase.

are shown in Figure 2 (open circles) together with the four end-member amount of amorphous silica (mixture M4, with 20% amorphous silica),
spectra (dashed curves). Some characteristic spectral features can be rec- + +
using only three end-member spectra, namely quartz kaolinite gibbs-
ognized in the mixture spectra after comparison with the end-member spec- ite. We observed only slight changes in the kaolinite (43.2% instead of
tra; for instance, the doublet at 914-936 cm-l corresponding to the Al- 42%) and gibbsite (24% instead of 25.5%) calculated wt % (Table 2). On
OH bending vibrations of kaolinite (Van der Marel and Krohmer 1969) the other hand, the calculated quartz wt % changed from 19.5% to 39.2%.
and the doublet at 780-798 cm-I due to the S i - O 4 intertetrahedral bridg- The resulting computed spectrum (solid curve) differs significantly from
ing bonds in quartz (Moenke 1974) are easily recognized. The computed the measured spectrum, as indicated by the arrows in Figure 2, Sample
wt % of each end member determined after application of our model are M4*. This example demonstrates that our model is sensitive to the presence
compared with the actual wt % in Table 2, and the computed mixture of amorphous silica in a complex sample, and that it is important to check
spectra generated from the end-member spectra by application of our model the fit between the computed and the measured spectra. Important missing
are shown (solid curves) in Figure 2. The high quality of the fit between end-member spectra can eventually be determined. In mixtures where
the computed spectra (solid curves) and the measured spectra (open circles) quartz and amorphous silica are both present in similar proportions, the
for the fow synthetic mixtures prepared is again a demonstration that our calculated quartz wt % can be significantly overestimated when the amor-
measurements obey Lambert-Beer's law. As a consequence, the spectrum phous silica content is greater than 10%. This will not usually cause lim-
of a mixture can be accurately calculated starting from the spectra of the itation for the mineral quantitative analysis of soils and sediments, where
components, as described in the section on data reduction. There is good the quartz content is generally larger than the amorphous silica content.
agreement between the computed and the actual wt % for kaolinite, gibbs-
ite, and amorphous silica (Table 2), as shown by the relative error, which Natural Sediments
is generally below 5%. For quartz, the relative error increases increasing
amorphous silica content in the mixture: the relative error is 0.4%, 5%, Selected mid-infrared absorbance spectra of the Salitre sediments are
31%, and 30% for 5, 10, 15, and 20 wt % amorphous silica, respectively, shown in Figure 3. Several components of the sediments are easily rec-
in the mixture. The incidence of the amorphous silica content upon the ognized by their spectral features. Organic matter gives a set of absorption
accuracy of the computed quartz content is attributed to the similarity of bands between 1000 and 1800 cm-l (C=O, C=C, 6 CH aliphatic) and a
the amorphous silica and quartz spectra: the mid-IR absorbance spectrum doublet between 2800 and 3000 cm-1 (6 CH aliphatic) (Rochdi et al. 1991;
of amorphous silica is composed of the three fundamental vibration modes Landais et al. 1993). Bands at 3698, 3650, and 3620 cm-I are due to the
for silica, namely Si-O stretching mode at 1100 cm-I, Si-O-Si stretching structural OH of kaolinite (Van der Marel and Krohmer 1969). This clay
mode at 800 cm-I, and Si-O bending mode at 470 cm-i (Moenke 1974). can also be recognized by the doublet at 914 and 936 cm-l. The bands at
These fundamental modes are also present in the quartz spectrum. To check 3420 and 3526 cm-' indicate the presence of gibbsite (Van der Marel and
this hypothesis, we applied our model to the mixture with the greater Beutelspacher 1976). Small quantities of quartz are detected by the doublet
I 'u
QUANTIFICATION OF MINERAL PHASES IN SEDIMENTS BY FTIR SPECTROSCOPY 445

at 780 and 798 cm-' (Moenke 1974). In the three samples of the top of Wavelength (pm)
Figure 3, the presence of amorphous silica can be recognized by the three 7.7 8.3, 9.1 10 11.1 12.5 14.3 16.7 20 25 33.3
1 , 1 , 1 1 1 , 1 , I I I I I I I I
broad bands at 470, 800, and 1100 cm-I (Fröhlich 1989), because it is the
only mineral phase associated with organic matter in this part of the core.
Because anatase, which was identified by XRD, is always mixed with other Sample depth (cm)
mineral phases, no specific spectral feature can be assigned to this mineral.
Mid-IR spectral data of Figure 3, together with information from X-ray
diffraction, optical microscopy, and chemical analyses (Sondag et al. 1993),
show that four types of sediments are present throughout the core. At the 23-24
top, between O and 70 cm, the sediment contains organic matter and amor- .......................... organic matter
...................................................................
phous silica represented by Phytoliths. Between 70 and 160 cm, organic R
I\
.i!. .
matter constitutes at least 80% of the sediment, but kaolinite, gibbsite,
quartz, amorphous silica (Phytoliths), and anatase are also present. Between * !

160 and approximately 300 cm, organic matter is the only constituent of
the sediment. The sediment below 300 cm contains the highest proportion \ :..
:
of kaolinite and gibbsite associated with quartz, anatase, amorphous silica,
and organic matter.
i, amorphous silica
..!, :,... ........
.....................................................
Between O and 70 cm, the multicomponent analysis was processed using
two components, organic matter + amorphous silica; the two reference
spectra are shown in Figure 4A (dashed curves), together with the spectrum
measured on Sample 23-24 (open circles). Below 70 cm, calculations were
+
performed assuming that the sediments are a mixture of organic matter
+ +
quartz kaolinite 4- gibbsite 4- anatase amorphous silica. In Figure
4B, the reference spectra of these phases are shown (dashed curves) with
a
E 129-131

spectra of two samples at 129-131 and 302-303 cm (open circles). Analysis a

2
4
of these natural mixtures produces wt % of various components, which are
then used to generate computed spectra. The computed spectra for the var-
ious samples analyzed are shown (solid curves) superimposed upon the
measured spectra (open circles) in Figure 4. The fit between the measured 302-303
IR spectra of the natural sediment and the computed spectra is not as good
as that observed when synthetic mixtures are processed (compare with Fig-
ures 1 and 2) but remains fairly acceptable. The resulting variations of
quartz, amorphous silica, gibbsite, anatase, and kaolinite along the Salitre
LC3 core are reported in Figure 5B. ..
Comparison with Chemical Analyses,-The three most abundant ele- h'':. . . . . amorphous
................
silica
....................... *.
'i
*.
.-
e
-.._.
ments, except C, are Si, Al, and Ti, with the maximum values of SO,, i: ,
i;,q
1; ;
Alzo3, and Tio, being 34.03%, 33.64%, and 10.39%, respectively. The
high Tio, content of the soils and sediment (Sondag et al. 1993) is due to ::!: D

/i. . .
p.
the particular mineralogy of the bedrock on which soils surrounding the .i:, , :
.
II
;

.....
i: :1 :i d ' fi * :, i\
lacustrine sediment have developed, with perovskite, calcizirtite, and apatite
often present as essential minerals. Unlike Tio,, Fe,O, in sediments has a
i, : .i i kaolinit:...
.......
\.-; 6. .,j
.............. :. ..................
:\ t.2 8.
._-.' $. : :< * *.
maximum value of 3.72%, and is generally lower than 3%. We have used ;\
: :
......... .p.. .......,i:: -i'. . >
..a
;**
the quantitative mineral analyses by application of our model to calculate ........ b.................... ...._..............'.,'.a, '*
:;
.........
a FTIR %O,, Alzo,, and Tio,. The FTIR SiO, is the sum of the weight
% of silica in quartz, amorphous silica, and kaolinite as would be stoichi-
................
anatase ............ .*.......
.......*-',

............................................ organic matter

ometrically calculated. The FTIR Alzo3 is the sum of the weight % of ............................ .....I.......'.......I.......~........I.......'......I ' ...................
I ~ I ' I ' I ~ I '
Alzo, in kaolinite and gibbsite as would be stoichiometrically calculated. 1 IO 1200 1100 lD00 900 800 700 600 500 400 2 1
The FTIR Tio, is equal to the weight % of anatase. Figure 5C presents Wavenumber (cm-')
the comparison between the FTIR determined and the chemically deter-
mined %O,, Alzo3,and Tio,. One can see the very good agreement be- FIG.4.-Multicomponent analysis of the Salitre sediments. A) spectra of the 23-
tween the chemically and the R I R determined SiO, (SiO, quartz SiO, + 24 cm sample and the two end members used to run the multicomponent analysis;
B) spectra of the 129-131 and 302-303 samples and the six end members used to
amorphous silica -t SiO, kaolinite). For Tio, and Alzo3, the agreement is run the multicomponentanalysis; dashed curves = end-member spectra; open circles
also fairly good. The slight discrepancies observed for Tio, can be ex- = measured mixture spectra; solid curves = computed mixture spectra.
plained by the absence of spectral features clearly related to anatase in the
sediment spectra, and to minor ilmenite. For A1,0, the small discrepancies
could be attributed to minor phases (amorphous hydrated Al oxides pos- ues are underestimated below 8% and overestimated above 20%. The cor-
sibly associated with O.M., crandallite), contamination from the aluminum responding correlation coefficients ( r ) are 0.982,0.954, and 0.990 for SiO,,
coring tube, enhancing the chemical signal, and to difference of crystallinity Tioz, and Alzo3, respectively.
between the natural gibbsite and the synthetic one used for calibration. In Discussion.-The rather good correlation between the RIR and chem-
Figure 5D we determined the regression lines between chemical and FTIR ical SiO, determined on 53 samples support the quantitative results for
SiO,, Tio,, and Alzo,; for SiO, and Tio, the regression lines that fit the quartz, amorphous silica, and kaolinite. The good correlation between FTIR
data are very close to the 1/1 line. For A1,03 the regression line is slightly determined wt % of anatase and Tio, also support the quantitative results
different from the 1/1 line; compared to chemical values, Alzo, FTIR val- for anatase. Although the IR spectrum of this mineral is simple, with only
446 J.BERTAUX ET AL.

A B C
% Si02 % Ti02 % Al203
Pollen
O 50 30 60 I O ì 4 6 5 10 20 40 102030 10 20 3 6 9 10 20 30 40

io u O I I u 1 I I I I I I
O

3060
0440

100 100 100

10
b

16800
!a740 7s

200 200 200

5

i;I
300 300 300
O
O 25 5 7.5 10
x = 0.947~+ 0.215 r = 0.954
400 400 400 40

30

500 500 500 20

10

600 600 600 O

Depth (cm) Depth (cm) O 10 20 30 40
x = 1.276~- 3.855 r = 0.990

FIG.5.-Results of the mineralogical analyses of the Salitre core; comparison with chemical, pollen and total organic carbon (TOC) data. A) Pollen, total organic carbon
(TOC) and I4C data; for the pollen column: plain circles = arboreal pollen except Myrtacea (Ledru 1996); open squares = non-arboreal pollen; B) Variations of mineral
content along the LC3 core as determined by ITIR spectroscopy; C) comparison of the chemically (i-) and the ITIR (o) determined (see text) Sioz, Tioz, and Alzo3.
+
FTIR Sioz = SiO, quartz f Sioz amorphous silica f SiOz kaolinite; FTIR Tio, = Tio, anatase; FTIR Alzo, = Alzo, kaolinite Alzo, Gibbsite. D) linear regression
for the chemical and FTIR determined Sioz, Tio, and Alzo, values. The straight lines represent the best linear fit of the data; the resulting equations and correlation
coefficients ( r ) are given below each plot; for SiOz and Tio,, the lines that fit the data are very close to the 1/1 line; for A1,0,, the 111line is drawn as a dotted line.

two broad bands not observable when overlapped by the spectra of the pearance of the detrital phases in the sediments. This opposition between
other phases of our sediment, the R I R quantitative data fit reasonably well the forest development over the drainage basin and the transport of detrital
the chemical Tio, determination. This confirms the results from X-ray phases towards the lake results from the increased susceptibility of soils to
diffraction showing that anatase is indeed the major Tio, phase, and shows erosion after the degradation of the vegetation cover. Therefore mineral
how the multicomponent analysis of FTIR spectra can provide quantitative phases are tracers of regional environmental variations (Bertaux et al.
results on phases with no visible specific band. The quantitative determi- 1996).
nation of amorphous silica is also an advantage of FTIR spectroscopy when
compared to a method like X-ray diffraction commonly used in sediment CONCLUSION
composition studies (Mumme et al. 1996).
The mineralogical similarity between the soils surrounding the depres- The above example demonstrates that multicomponent analysis of IR
sion (Sondag et al. 1993) and the sediments support the origin of kaolinite, absorbance spectra is a powerful tool for quantitative mineralogical analysis
gibbsite, quartz, and anatase from these soils. The comparison between of sediments. The analysis is performed on the bulk sediment after grind-
mineralogy (Fig. 5B) and pollen contents (Ledru 1993; Ledru et al. 1996) ing. Treatments like sequential dissolution, which in many cases can affect
(Fig. 5A) of the Salitre lacustrine sediments reveals an opposite trend be- the mineral or organic components, are not needed. Knowledge of the qual-
tween the forest cover, as inferred by the arboreaunon-arboreal pollen con- itative composition of the mixture, using X-ray diffraction analysis and
tents, and the detrital content of sediments: below 300 cm, a high content microscopic observation on smear slides, is required. A second requirement
in detrital minerals corresponds to a low content in arboreal pollen; between is the availability of reference spectra of the various constituents. In soils
160 and 300 cm, an increase of the arboreal pollednon-arboreal pollen or sediments, many components are susceptible to exhibiting a range of
ratio characterizes a period with a pure organic sedimentation; between 70 structural ordeddisorder and compositional variations. Clays are character-
and 160 cm, a decrease in arboreal pollen and an increase in non-arboreal ized by large variations in layer stacking and chemical compositions, both
pollen coincides with the presence of detrital minerals in the sediments; resulting in modifications of their IR absorbance spectral features, as evi-
above 70 cm, an increase in the arboreal pollen content matches a disap- denced in kaolinite minerals (Cases et al. 1982; Brindley et al. 1986) or in
 QUANTIFICATION OF MINERAL PHASES IN SEDIMENTS BY FTIR SPECTROSCOPY 441

smectites (Decarreau et al. 1992). In our study, we used kaolinite separated hematite and goethite concentration in geological materials: Journal of Sedimentary Petrol-
ogy, v. 61, p. 628-632.
from a Brazilian soil to prepare a reference spectrum. Because only one DECAIIREAU, A., GRAunY, O., ANO PETIT,S., 1992, The actual distribution of octahedral cations
reference spectrum per phase is required to run the calculation, this step is in 2:l minerals: Results from clay synthesis: Applied Clay Science, v. 7, p. 147-167.
relatively easy to achieve provided that a good separation is made. Our DIIEES,L.R., MANU,A., AND WiLniNG, L.P., 1993, Characteristics of aeolian dusts in Niger,
West Africa: Geoderma, v. 59, p. 213-233.
study on a natural sediment shows that although we used only a limited DUYCKAERTS, G., 1955, Contribution 1 l’analyse quantitative par-les spectres d’absorption in-
set of reference spectra, corresponding to the main phases of the sediments, h u g e des poudres. I. Examen théorique de la question: Spectrmhimica Acta, v. 7, p. 25-31.
and despite the fact that the reference compounds were slightly different DUYCKAERTS, G., 1959,The infrared analysis of solid substances: Analyst, v. 84, p. 201-214.
from the natural one, we obtained FTIR determined wt % of the mineral FARMER, V.C., ed., 1974, The infrared spectra of minerals: Mineralogical Society [London],
Monograph 4, 539 p.
phases consistent with available chemical data. This suggests that our meth- FRETIGNY, c., ANO DESBARRES, J., 1988, Graph: software for “Ecole 0“:structures fines
od can be applied to a large variety of natural mixtures. The multicom- d’absorption des rayons X. Des données expérimentales à leur analyse”.
ponent analysis of IR absorbance spectra succeeds in quantifying compo- FR~HLICH, F., 1982,Evolution minéralogique dans les dépots azoïques rouges de l’océan indien.
Relations avec la stratigraphie: SociEté Géologique de France, Bulletin, v. 3, p. 563-571.
nents identified by XRD (anatase) or optical microscopy (amorphous silica) FROHLICH, F., 1989, Deep-sea biogenic silica: new structural and analytical data from infrared
but with no visible spectral features in the mixture spectrum. This is a great analysis-geological implications: Terra Nova, v. I,p. 267-273.
advantage over previous use of IR absorbance spectroscopic data for quan- FROHLICH, F., AND SERVANT-VILDARY, S., 1989, Evaluation of diatom content by counting and
infrared analysis in quaternary tluvio-lacustrine deposits from Bolivia: Diatom Research, v.
tifying mineral phases by measuring specific band surface (Fröhlich 1982; 4, p. 241-248.
Sifeddine et al. 1994; Chester and Elderfield 1968; Fröhlich and Servant- GAFWY,S.J., MCFADDEN, L.A., NASH,R., AND PIETERS, C., 1993, Ultraviolet, visible, and near-
Vildary 1989). Correlation between pollen data and quantitative mineralogy infrared reflectance spectroscopy: Laboratory spectra of geologic materials, in Pieters, C.M.,
and Englert, P.A.J., eds., Remote Geochemical Analysis: Elemental and Mineralogical Com-
in the Salitre sediments reveals that degradation of the vegetal cover is the position, Cambridge University Press, p. 43-77.
main factor that controls the transport of the detrital minerals (quartz, ka- GEHLEN,M., AND VANRAAPHORST, W., 1993, Early diagenesis of silica in sandy North Sea
olinite, gibbsite, and anatase) from the surrounding soils towards the lake. sediments: quantification of the solid phase: Marine Chemistry, v. 42, p. 71-83.
JONES,J.B., AND SEGNIT, E.R., 1971, The nature of opal I. Nomenclature and constituent phases:
Because changes in the vegetal cover result from climatic variability, the Geological Society of Australia, Journal, v. 18, p. 57-68.
mineral content of the lacustrine sediments can be used as an indicator of JONES,R.L., 1969, Determination of opal in soil by alkali dissolution analysis: Soil Science
environmental changes. Society of America, Proceedings, v. 33, p. 976-978.
LANDAIS, P., ROCHOI, A., LARGEAU, C., ANO DERENNE, S., 1993, Chemical characterization of
torbanites by transmission micro-mIR spectroscopy: origin and extent of compositional
ACKNOWLEDGMENTS heterogeneities: Ceochimica et Cosmochimica Acta, v. 57, p. 2529-2539.
LEDRU,M.P., 1993, Late Quaternary Environmental and climatic changes in central Brazil:
Quaternary Research, v. 39, p. 90-98.
This work was supported b y t h e V R 12 of the “Terre-Océan-Atmosphère” De- LEDRU,M.P., SOARES BRAGA, P.I., SouniB, F., FOURNIER, M., MARTIN, L., SUGUIO, K., ANOTURCQ,
partment of Institut Français d e Recherche Scientifique pour le Développement en B., 1996, The last 50,000 years in the Neotropics (Southern Brazil): evolution of vegetation
Coopération (ORSTOM). I t is part of the AIMPACT-PVC (ORSTOM) and “ECOF- and climate: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 123, p. 239-257.
IT” (CNRS-ORSTOM) programs. Caroline Duong is greatly acknowledged for her MOENKE,H.H.W., 1974, Silica, the three dimensional silicates, borosilicates and beryllium
contribution in the samples preparation and analysis. The authors wish to thank F. silicates, ir1 Farmer, V.C., ed., The Infrared Spectra of Minerals: Mineralogical Society
Sondag, L. Martin, J.M. Flexor, G. Morin, and T. Corrège for their help during the [London], Monograph 4, p. 365-382.
MUMME, W.G., TSAMROURAKIS, G., MAOSEN, I.C., AND HILL,R.J., 1996, Improved petrologiCd
preparation of the manuscript. J.L. Bishop and a n anonymous reviewer are thanked
modal analyses from X-ray powder diffraction data by use of the Rietveld method. Part II.
‘for their critical reading o f the manuscript and helpful suggestions. Selected sedimentary rocks: Journal of Sedimentary Research, v. 66, p. 132-138.
MUSTARD, J.F., ANO PIETERS,C.M., 1989, Photometric phase functions of common geologic
minerals and applications to quantitative analysis of mineral mixture reflectance spectra:
REFERENCES Joumal of Geophysical Research, v. 94, p. 13.619-13,634.
PICHARD, C., AND FROHLICH, F., 1986, Analyses IR quantitatives des sédiments. Exemple du
BALSAM, W.L., AND DEATON,B.C., 1996, Determining the composition of late Quaternary ma- dosage du quartz et de la calcite: Institut Français du Pétrole, Revue, v. 41, p. 809-819.
rine sediments from NUV, VIS and NIR diffuse reflectance spectra: Marine Geology, v. ROCHOI, A., LANDAIS, P., ANO BURNEAU, A., 1991. Analysis of coal by transmission FTIR mi-
134, p. 31-55. crospectroscopy: methodological aspect: Société GCologique de France, Bulletin, v. 162,p.
BARRANCO, F.T., BALSAM, W.L., AND DEATON,B.C., 1989, Quantitative reassessment of brick 155-162.
red lutites: evidence from reflectance spectrophotometry: Marine Geology, v. 89, p. 299- ROGNON, P., COUOE-GAUSSEN, G., REVEL,M., GROUSSET, F.E., AND PEOEMAY, P., 1996, Holocene
314. Saharan dust deposition on the Cape Verde Islands: sedimentological and Nd-Sr isotopic
BhllTAUx, J., LEDRU, M.P., ~OUSICS, F., AND SONDAG, F., 1996,The use ofquantitative mineralogy evidence: Sedimentology, v. 43, p. 359-366.
linked to palynological studies in palaeoenvironmental reconstruction: the case study of the SAARINEN, P., AND KAUIWNEN, J., 1991, Multicomponent analysis of FT-IR spectra: Applied ..
“Lagoa Campestre” lake, Salitre, Minas Gerais, Brazil: Académie des Sciences [Paris], Spectroscopy, v. 45, p. 953-963.
Comptes Rendus, v. 323,série Il a, p. 65-71. SALISBURY, J.W.,1993, Mid-infrared spectroscopy: Laboratory data, in Pieters, C.M., and En-
BISHOP, J.L., KOEHEIIL,C., KRALK,C., FROSCHL, H., ENGLEIIT,P.A.J., ANDERSEN, D.W., PETERS, glert, P.A.J., eds., Remote Geochemical Analysis: . Elemental and Mineraloaical Comoosi-
C.M., AND WHARTON, R.A.. JR., 1996, Reflectance spectroscopy and geochemical analyses Ïion, Cambridge University Press, p. 79-98.
of Lake Hoare sediments, Antarctica: Implications for remote sensing of the Earth and Mars: SIEDDINE,A., FR~HLICH, F., FOURNIER, M., MARTIN, L., SERVANT, M., SouniFS, F., TURCQ,B.,
Ceochimica et Cosmochimica Acta, v. 60, p. 765-785. ~ U G U I O ,ti., ANO VOLKMEII-RIBEIRO, c., 1994, La sédimentation lacostre indicateur de chan-
BONNIN, D., KAISER,P., FKETIGNY, C., ANO DEBAIIIIES, J., 1989, Logiciel d’analyse sur micro- gement des paléoenvironnements au cours des 30 O00 dernières années (Carajas, Amazonie,
ordinateur IBM-PC, in Dexpert, H., Michalowicz, A., and Verdaguer, M., eds, Structures Brésil): Académie des Sciences [Paris], Comptes Rendus, v. 318, p. 1645-1652
fines d’absorption des rayons X en chimie, Ecole du CNRS, Orsay-Garchy, 19-24 Septem- SONDAG, F., SouniÈs, F., FORTUNE, J.P., DUPIIÉ,B., MAGAT,P., ANO MELFI,A., 1996,Hydrogeo-
bre 1988, Orsay, v. 3-2,p. 1-32. chemistry in soils and sediments in the area of the Lagoa Campestre lake (Salitre, MG,
BRINDI.EY, G.W., KAO,C.C., HARRISON, J.L., LIPSICAS,
M., AND RAYTHATHA, R., 1986, Relation Brazil): chemical balances of major and trace elements and dynamics of rare earth elements:
between structural disorder and other characteristics of kaolinites and dickites: Clays and Applied Geochemistry, v. 12, p. 155-162.
Clay Minerals, v. 34, p. 239-249. SONOAG, F., Soue~ts,F., LEDIIU,M.P., AND DELAUNE, M.,1993,Geochemical markers of palaeo-
CASES,J.M., LI~TAIID, O., YVON,J., AND DELON,J.F., 1982, Etude des propriétés cristallochi- environments: relations between climatic changes, vegetation and geochemistry of lake sed-
miques, morphologiques et superficielles de kaolinites désordonnées: Bulletin de Minéral- iments, southern Brazil: Applied Geochemistry, Supplementary Issue no. 2, p. 165-170.
ogie, v. 105, p. 439357. SUNSHINE, J.M., AND PIETERS, C.M., 1993, Estimating modal abundances from the spectra of
CHESTER, H., 1967,The application of infra-red absorption spectroscopy to
R., AND EI.DERFI~I.D, natural and laboratory pyroxene mixtures using the modified gaussian model: Journal of
carbonate mineralogy: Sedimentology, v. 9, p. 5-21. Geophysical Research, v. 98, p. 9075-9087.
CHESTER, R., AND ELDERFIELD, H., 1968, The infrared determination of opal in siliceous deep- TEISSIEII, A., CAMPnELL, P.G.C., AND BISSON,M., 1979,Sequential extraction procedure for the
sea sediments: Ceochimica et Cosmochimica Acta, v. 32, p. 1128-1140. speciation of particulate trace metals: Analytical Chemistry, v. 51, p. 844-851.
DEAN,W.E., 1993, Physical properties, mineralogy, and geochemistry of Holocene varved VANDER MAREL,H.W., AND KROHMER, P., 1969, OH stretching vibrations in kaolinite. and
sediments from Elk lake, Minnesota, in Bradbury, J.P., and Dean, W.E., eds., Elk Lake. related minerals: Contributions to Mineralogy and Petrology, v. 22, p. 73-82.
Minnesota: Evidence for Rapid Climate Change in the North-Central United States: Geo- VANDER MAREL,H.W., AND BEUTELSPACHER, H., 1976, Atlas Of infrared spectroscopy of clay
minerals and their admixtures: Amsterdam, Elsevier, 396 p.
logical Society of America, Special Paper 276. p. 135-157.
DEATON, B.C., AND BAISAM,W.L., 1991,Visible spectroscopy-A rapid method fordetermining Received 27 September 1996; accepted 9 December 1997.