Ac a Geodyn. Geoma e ., Vol. 8, No. 1 (161), 47–58, 2011
1.
APPLICATION OF INFRARED SPECTROSCOPY AND CHEMOMETRIC METHODS TO
IDENTIFICATION OF SELECTED MINERALS
Michal RITZ 1)*, Lenka VACULÍKOVÁ 2) and E a PLEVOVÁ 2)
1) VŠB-Technical Uni e si y Os a a, 17. lis opadu 15, 708 33 Os a a-Po uba, Czech Republic
2) Ins i u e o Geonics o he AS CR, S uden ská 1768, 708 00 Os a a-Po uba, Czech Republic
*Co esponding au ho ‘s e-mail: [email p o ec ed]
(Recei ed No embe 2010, accep ed Ma ch 2011)
ABSTRACT
A new way o iden i ica ion o mine als was sugges ed. The iden i ica ion was based on chemome ic analysis o measu ed I
R
spec a o selec ed mine als. IR spec a we e collec ed using di use e lec ance echnique. The disc iminan analysis an
d
p
incipal componen analysis we e used as chemome ic me hods. Fi e s a is ical models we e c ea ed o sepa a ion an
d
iden i ica ion o clay mine als. Up o 60 samples o a ious mine al s anda ds (clay mine als, eldspa s, ca bona es, sulpha es
and qua z) om di e en locali ies we e selec ed o he c ea ion o s a is ical models. The esul s o his s udy con i m ha
he disc iminan analysis o IR spec a o mine als could
p
o ide a powe ul ool o mine al iden i ica ion. E e
n
di e en ia ion o musco i e om illi e and iden i ica ion o mixed s uc u es o illi e-smec i e we e achie ed.
KEYWORDS: mine als, iden i ica ion, IR spec oscopy, disc iminan analysis, p incipal componen analysis
Ve y use ul o p ocessing o IR spec a a e
mul i a ia e s a is ical me hods. The big ad an age o
mul i a ia e s a is ical me hods is hei capabili y o
ex ac equi ed spec al in o ma ion om IR spec
a
and explo e his spec al in o ma ion o quali a i e o
quan i a i e applica ions. The mos equen ly use
d
mul i a ia e s a is ical me hods (o en called
chemome ic me hods) a e ac o analysis (FA),
p
incipal componen analysis (PCA), disc iminan
analysis (DA), p incipal componen eg ession (PCR),
mul iple linea eg ession (MLR) and pa ial leas -
squa es eg ession (PLS). DA, FA and PCA a e e y
use ul ools o quali a i e analysis.
The disc iminan analysis allows quan i ica ion
o he p obabili y wi h which an objec can be
assigned o an indi idual class. DA gene a es
a disc iminan unc ion as a linea combina ion o
measu ed a iables, which gi es maximu
m
sepa abili y o objec s om di e en classes. The
a io o be ween-class o pooled wi hin-class sample
a iance is commonly used as a c i e ion J o
measu ing he sepa abili y o di e en classes:
T
T
aBa
JaWa
= (1)
whe e ais a ec o maximizing he disc iminan
c i e ion and B and W a e he be ween-class and he
wi hin-class co a iance ma ixes, espec i ely (Jiang
e al., 2002).
1. INTRODUCTION
Knowledge o mine al composi ion is essen ial
o cha ac e ize he geochemical and physico-
mechanical p ope ies o ocks. Na u e and con en o
mine als (especially clay mine als) p esen in ocks
ha e a signi ican in luence on he beha iou an
d
p
ope ies o ocks as well as on he whole oc
k
massi .
In gene al, se e al con en ional analy ical
me hods exis ha can be used o examine he mine al
composi ion o ocks: op ical mic oscopy, elec on
mic oscopy, X- ay di ac ion (XRD), Fou ie
ans o m in a ed spec oscopy (FTIR spec oscopy),
Raman spec oscopy, he mal analysis (TG/DTA) and
b
ulk chemis y analysis (Kodama e a., 1989; Chipe a
and Bish, 2001; S odon, 2002; Vog e al., 2002;
Vaculíko á, 2006). Un o una ely, he exac
iden i ica ion o clay mine als by means o hese
me hods is a he complica ed and o en inaccu a e.
The main analy ical di icul ies a e ela ed o a iable
chemical composi ion and common s uc u al
diso de s o clay mine als. Ano he eason consis s in
he ac ha he indi idual clay mine als occu in he
o m o mix u es (illi e – mon mo illoni e, chlo i e
–
mon mo illoni e, e c.) wi h a ious a ios o pa icula
clay mine als.
The cu en FTIR spec oscopy makes i possible
o analyze indi idual mine als, nonc ys alline
admix u es and, simul aneously, o de ec he p esence
o o ganic ma e .
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
M. Ri z e al.
48
b
ands ep esen spec al in o ma ion posi i ely
co ela ed h ough pa icula p incipal componen ; he
nega i e ones ep esen spec al in o ma ion
nega i ely co ela ed h ough pa icula p incipal
componen .
Bo h o he abo e-men ioned chemome ic
me hods (DA and PCA) a e desc ibed in a lo o
s a is ical ex books (Meloun and Mili ký, 2004;
Hendl, 2006) in ull de ails. Likewise, DA and/o
PCA o IR spec a a e used in many s udies (Tesch
and O o, 1995; Dahlbe g e al., 1997; Khe e al.,
2001; Tsuchikawa e al., 2003; Ri z, 2007).
This pape desc ibes a new app oach o clay
mine als and some accesso ies mine als iden i ica ion
b
ased on chemome ic analysis o IR spec a o clay
mine als. All IR spec a we e collec ed using di use
e lec ance echnique (DRIFT). The aim o he s ud
y
was he c ea ion o s a is ical models o mine als
iden i ica ion by disc iminan analysis o measu ed I
R
spec a.
2. MATERIALS AND METHODS
2.1. SAMPLES OF MINERALS
The esea ch has been ocused on he basic clay
mine al specimens (kaolini e, mon mo illoni e,
non oni e, illi e, chlo i e and e miculi e) including
micas (musco i e, bio i e) and he associa ed mine als
(qua z, eldspa , ca bona es and sulpha es). The
s anda ds o abo e-men ioned mine als we e selec ed
acco ding o hei au hen ici y and a deg ee o pu i y.
These mine al s anda ds we e ob ained om he
collec ion o mine als a ailable a he Ins i u e o
Geonics, Nano echnology Cen e o VŠB-Technical
Uni e si y Os a a and om Sou ce Clays Reposi o y
o he Clay Mine als Socie y, USA.
60 samples o mine al s anda ds we e collec ed
o he p epa a ion o s a is ical models. 51 samples o
clay mine als (including micas) and 9 samples o
associa ed mine als (qua z, eldspa s, ca bona es an
d
sulpha es) we e used. Some o mine al s anda ds we e
ea ed by je mill; o he ones we e p epa ed by
sedimen a ion o ob ain ac ion g ain size less han
5 μm. Some s anda ds we e used as unmodi ied
(“ aw” s anda ds). A lis o all used samples o
mine al s anda ds is shown in Table 1.
2.2. FTIR MEASUREMENTS
App oxima ely 5-10 mg o sample o mine al
s anda ds was g ound wi h app ox. 400 mg d ied KB .
This mix u e was used o collec IR spec a.
The IR spec a we e collec ed using FTI
R
spec ome e Nexus 470 (The moScien i ic, USA).
Di use e lec ance measu emen echnique (DRIFT)
was used. This echnique was selec ed o i s as ness
and simplici y. The measu emen pa ame e s we e as
ollows: spec al egion 4000-400 cm-1, spec al
esolu ion 8 cm-1; 128 scans; Happ-Genzel
apodiza ion.
The disc iminan analysis uses esul s o ano he
chemome ic me hod – p incipal componen analysis
(PCA). PCA is one o he mos common mul i a ia e
me hods wide-sp ead also in in a ed spec oscop
y
(e.g. Tesch and O o, 1995; Dahlbe g e al., 1997;
Rusnak e al., 2003). IR spec um is composed o
many poin s connec ed by he line; each poin in he
IR spec um is de ined by wa enumbe and in ensi y
( o example abso bance). In PCA, each spec um is
ep esen ed by a poin in mul idimensional space,
whe eas he numbe o dimensions co esponds o he
numbe o poin s in he spec um (in o he wo ds he
numbe o a iables). Because his mul ispace is e y
di icul o comp ehend, PCA in oduces a new
coo dina e sys em as a linea combina ion o he ol
d
coo dina es. Mos o he a iances (i means he mos
o in o ma ion om he IR spec um) a e concen a ed
in only a ew o he new coo dina es – called p incipa
l
componen s. This can be exp essed as he
decomposi ion o he da a ma ix (D) in o wo o he
ma ixes - he sco e ma ix (P) and he loading ma ix
(T). Fo be e unde s anding o he decomposi ion o
he da a ma ix we can use he ollowing g aphical
ep esen a ion (Geladi and Kowalski, 1986):
D
ma ix has he dimensions o numbe o samples (n)
b
y in ensi y (a pa icula wa enumbe s) o IR spec a
(m). The sco e ma ix (P) has he dimensions o
numbe o samples (n) by p incipal componen s (a)
and his ma ix ep esen s he loca ion o he samples
in he new p incipal componen space. The loading
ma ix (T) has dimensions o in ensi y (m) by
p incipal componen s (a); loading ma ix ep esen s
he ecipes by which each o he p incipal componen s
was cons uc ed om he old a iable - in ensi y a
each wa enumbe . The mos impo an in o ma ion is
included in he i s wo o h ee p incipal
componen s. The es o he p incipal componen s
con ain only mino cha ac e is ics o samples an
d
noise, so ha can be igno ed. The main ad an age o
PCA consis s in educing he in luence o noise and
exploi ing he sub le di e ences in he IR spec a o
samples o ob ain ele an and signi ican spec al
in o ma ion.
In PCA o spec al da a, he loading ma ix is
e y impo an . This ma ix shows us which spec al
in o ma ion ( egions o bands) a e impo an in he
p
incipal componen space. The mos signi ican wa
y
o p ojec ion o loading ma ix in spec al analysis is
so-called loading spec a. They o en look like
“ ypical spec a” and posi i e and/o nega i e ban
d
can be p esen a he loading spec a. The posi i e
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
APPLICATION OF INFRARED SPECTROSCOPY AND CHEMOMETRIC METHODS …
49
Table 1 Lis o samples.
Sample Mine al Locali y Desc ip ion
Bio1 Bio i e K íš, Czech Republic ac ion ≤ 5 μm
Bio2 Bio i e K íš, Czech Republic “ aw” s anda d
Bio3 Bio i e Dolní Bo y, Czech Republic ac ion ≤ 5 μm
Mus1 Musco i e India “ aw” s anda d
Mus2 Musco i e Měděnec, Czech Republic ac ion ≤ 5 μm
Mus3 Musco i e O o , Czech Republic ac ion ≤ 5 μm
Mus4 Musco i e Měděnec, Czech Republic “ aw” s anda d
Mus5 Musco i e Rožná, Czech Republic ac ion ≤ 5 μm
Mus6 Musco i e-Chlo i e Měděnec, Czech Republic “ aw” s anda d
Ill1 Illi e Hunga y ac ion ≤ 5 μm
Ill2 Illi e (IM -1) Mon ana, USA ac ion ≤ 5 μm
Ill3 Illi e Ka lo , Czech Republic “ aw” s anda d
Ill4 Illi e (IM -2) Mon ana, USA “ aw” s anda d
Ill5 Illi e (IM -1) Mon ana, USA je mill
Ill6 Illi e (IM -2) Mon ana, USA “ aw” s anda d
Ill7 Illi e-Smec i e (ISCz-1) Czechoslo akia “ aw” s anda d
Ch1 Chlo i e (CCa-2) Cali o nia, USA ac ion ≤ 5 μm
Ch2 Chlo i e Le o ice, Czech Republic ac ion ≤ 5 μm
Ch3 Chlo i e O lické ho y, Czech Rep. ac ion ≤ 5 μm
Ch4 Chlo i e (CCa-2) Cali o nia, USA “ aw” s anda d
Ch5 Chlo i e (CCa-2) Cali o nia, USA je mill
Kao1 Kaolini e Božíčany, Czech Republic “ aw” s anda d
Kao2 Kaolini e Ho ní Bříza, Czech Republic “ aw” s anda d
Kao3 Kaolini e (KGa-1b) Geo gia, USA ac ion ≤ 5 μm
Kao4 Kaolini e Kadaň, Czech Republic “ aw” s anda d
Kao5 Kaolini e Kaznějo , Czech Republic “ aw” s anda d
Kao6 Kaolini e Sedlec, Czech Republic “ aw” s anda d
Kao7 Kaolini e Úna o , Czech Republic “ aw” s anda d
Kao8 Kaolini e (KGa-1b) Geo gia, USA “ aw” s anda d
Kao9 Kaolini e (KGa-2) Geo gia, USA “ aw” s anda d
Kao10 Kaolini e Sedlec, Czech Republic “ aw” s anda d
Kao11 Kaolini e Sedlec, Czech Republic ac ion ≤ 5 μm
Dic1 Dicki e K ásno, Czech Republic ac ion ≤ 5 μm
Hal1 Halloysi e Bílá Ho a, Slo akia “ aw” s anda d
Mon1 Mon mo illoni e I ančice, Czech Republic ac ion ≤ 5 μm
Mon2 Mon mo illoni e Jelšo ý po ok, Slo akia ac ion ≤ 5 μm
Mon3 Mon mo illoni e (SWy-1) Wyoming, USA “ aw” s anda d
Mon4 Mon mo illoni e (SWy-1) Wyoming, USA ac ion ≤ 5 μm
Mon5 Mon mo illoni e (SWy-2) Wyoming, USA “ aw” s anda d
Mon6 Mon mo illoni e (SAz-2) A izona, USA ac ion ≤ 5 μm
Mon7 Mon mo illoni e (SAz-2) A izona, USA “ aw” s anda d
Mon8 Mon mo illoni e (STx-1b) Texas, USA “ aw” s anda d
Non1 Non oni e (Nau-1) Uley Mine, Sou h Aus alia “ aw” s anda d
Non2 Non oni e (Nau-2) Uley Mine, Sou h Aus alia “ aw” s anda d
Hec1 Hec o i e (SHCa-1) Cali o nia, USA “ aw” s anda d
Ve 1 Ve miculi e China expanded
Ve 2 Ve miculi e Ald ich s anda d expanded
Ve 3 Ve miculi e Ald ich s anda d expanded, je mill
Ve 4 Ve miculi e Le o ice, Czech Republic no expanded
Ve 5 Ve miculi e B asil no expanded
Ve 6 Ve miculi e B asil no expanded, je mill
Acc1 Qua z O o , Czech Republic “ aw” s anda d
Acc2 Side i e Dúb a a, Slo akia “ aw” s anda d
Acc3 Magnesi e Nižná Slaná, Slo akia “ aw” s anda d
Acc4 Calci e Š ambe k, Czech Republic “ aw” s anda d
Acc5 Gypsum No he n Bohemia, Czech Rep. “ aw” s anda d
Fel1 Albi e Wes . Bohemia, Czech Rep. “ aw” s anda d
Fel2 Feldspa (Na, K) Wes . Bohemia, Czech Rep. “ aw” s anda d
Fel3 Feldspa (Na, K) Wes . Bohemia, Czech Rep. “ aw” s anda d
Fel4 Feldspa (Na, Ca) Wes . Bohemia, Czech Rep. “ aw” s anda d
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
M. Ri z e al.
50
Model 1 was c ea ed o he sepa a ion o clay
mine als om accesso y mine als like eldspa s,
qua z, ca bona es (calci e, side i e and magnesi e)
and sulpha es (gypsum). Gene al disc imina ion o
clay mine als was pe o med in Model 2. The clay
mine als we e classi ied in o pa icula g oups o cla
y
mine als in his model: bio i e, musco i e-illi e,
chlo i es, kaolin-se pen ines, smec i es an
d
e miculi es. The emaining models we e c ea ed o
iden i ica ion o indi idual membe s o selec ed clay
mine al g oups. Model 3 has allowed iden i ica ion o
mine als o he kaolin-se pen ine g oup (kaolini e,
dicki e and halloysi e), Model 4 has allowe
d
iden i ica ion o mine als o he smec i e g oup (Na-
mon mo illoni e, Ca-mon mo illoni e, non oni e and
hec o i e) and Model 5 has allowed iden i ica ion o
micas (bio i e, musco i e) and illi e.
3.2. MODEL 1 – SEPARATION OF CLAY MINERALS
FROM ACCESSORY MINERALS
This model was c ea ed o he selec ion o clay
mine als om accesso y mine als. The e we e h ee
classes used in Model 1 (Clay mine als, Feldspa s and
Accesso y mine als). The i s class included all
used samples o clay mine als (bio i e, musco i e,
illi e, chlo i e, kaolini e, dicki e, halloysi e,
mon mo illoni e, no oni e, hec o i e and e miculi e).
The i s class (Clay mine als) in ol ed 51 samples as
membe s o his class. The second class (Feldspa s)
had 4 membe s; h ee samples o mixed eldspa s an
d
one sample o albi e. The hi d class (Accesso
y
mine als) included 5 membe s (qua z, side i e,
calci e, magnesi e and gypsum). The Mahalanobis
dis ance plo is shown in Figu e 1. All mine als we e
clea ly dis ibu ed in o h ee clus e s whe eas he
clus e o Feldspa s is ela i ely close o he clus e o
Clay mine als. The clus e o Accesso y mine als is
e iden ly isola ed om he o he wo clus e s.
The p incipal componen analysis was used as
in eg al pa o he disc iminan analysis. Ten
p
incipal componen s (PCs) we e calcula ed in
Model 1. These en PCs desc ibed 99.3 % o a iance
o spec al in o ma ion, whe eas almos 95 % o
a iance was in ol ed in he i s ou PCs. Loadings
spec um o he i s p incipal componen (PC1)
includes spec al bands o kaolini e only: s e ching
bands o inne -su ace hyd oxyl g oups (3695 cm-1,
3668 cm-1, 3652 cm-1 and 3620 cm-1), Si-O s e ching
band (1100 cm-1), in-plane Si-O s e ching bands
(1030 cm-1 and 1010 cm-1 ), de o ma ion band o
inne -su ace hyd oxyl g oups (915 cm-1), Si-O
s e ching bands (795 cm-1, 755 cm-1 and 695 cm-1)
and de o ma ion bands Al-O-Si, Si-O-Si and Si-O,
espec i ely (540 cm-1, 470 cm-1 and 430 cm-1).
Assignmen o IR spec al bands o mine als
acco ding o li e a u e (Russel e al., 1994; Madejo
á
and Komandel, 2001; Vaculíko á and Ple o á, 2005)
was used in his pape . PC1 desc ibed 76.3 % o
a iance. The second p incipal componen (PC2)
E e y sample was p epa ed and consequen ly
measu ed se e al imes (3 – 5 imes). The mean I
R
spec um o e e y sample was calcula ed o
subsequen s a is ical p ocessing.
2.3. DISCRIMINANT ANALYSIS
The disc iminan analysis (DA) was pe o med
using TQ Analys so wa e (The moScien i ic,
USA). Two spec al egions 4000-3000 cm-1 and
1300-400 cm-1 o each IR spec um we e used o
DA. A one-
p
oin baseline was used in e e y spec al
egion. A mul iplica i e signal co ec ion o
p
a hleng h was used o calcula ion. The Mahalanobis
dis ance was used o o mula e a dis ance be ween
clus e s.
Be o e DA,
p
incipal componen analysis
(PCA) ca ied ou . The ask o PCA was educ ion
o edundan spec al in o ma ion and inding o
impo an spec al ea u es which had a signi ican
in luence on spec al a iance. The numbe o
p incipal componen s used o he p epa a ion o
model was 10.
Valida ion o s a is ical models has been ca ied
ou by means o a alida ion spec a se (TQ Analys
so wa e does no allow c oss- alida ion o
disc iminan analysis). The IR spec a o mine als
we e spli in o wo g oups in each class used in DA
–
calib a ion g oup o spec a and alida ion g oup o
spec a. The g oup o calib a ion spec a was used o
c ea e a disc imina ion model; he g oup o alida ion
spec a was used o e i y his model. Mos o he I
R
spec a in each class we e used as calib a ion spec a;
up o h ee IR spec a in each class we e used as
alida ion spec a. The e we e no alida ion spec a
used in one-membe classes only. The selec ion o
calib a ion and alida ion spec a has been pe o med
b
y TQ Analys so wa e. The ollowing samples we e
used o collec ion o alida ion spec a: Bio2, Mus1,
Mus4, Ill2, Ill4, Ch3, Kao2, Kao7, Kao8, Mon3,
Mon6, Non2, Ve 4, Fel3 and Acc3. The models in his
s udy we e conside ed co ec when all alida ion
spec a occu ed inside hei class clus e s. All
disc imina ion models men ioned in his pape ul ill
his condi ion.
3. RESULTS AND DISCUSSION
3.1. STRATEGY OF MODEL CREATION
The c ea ion o a unique s a is ical model o all
used samples was a p ima y idea o he au ho s.
Du ing he p ocess o c ea ion o he unique model
se ious p oblems occu ed. A lo o mine als we e
classi ied in o inco ec classes because o
conside able a iabili y o spec al bands. The
solu ion o his p oblem was a c ea ion o pa ial
models; i e pa ial models we e c ea ed in his s udy.
These pa ial models ha e allowed sepa a ion an
d
subsequen iden i ica ion o clay mine als in a ew
s eps. A b ie summa y o he c ea ed models is in
Table 2.
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
APPLICATION OF INFRARED SPECTROSCOPY AND CHEMOMETRIC METHODS …
51
Table 2 B ie summa y o s a is ical models.
Name o model Used mine als Classes
Model 1 All mine al samples Clay mine als
Feldspa s
Accesso y mine als
Model 2 Clay mine als Bio i e
Musco i e-illi e
Chlo i e
Kaoline-se pen ine
Smec i e
Ve miculi e
Model 3 Mine als o kaoline-se pe ine g oup Kaolini e
Dicki e
Halloysi e
Model 4 Mine als o smec i e g oup Na-Mon mo illoni e
Ca-Mon mo illoni e
Non oni e
Hec o i e
Mixed illi e-smec i e s uc u e
Model 5 Micas Bio i e
Musco i e
Illi e
Mixed illi e-smec i e s uc u e
Fig. 1 Mahalanobis dis ance plo o h ee clus e s o mine als (Model 1).
he hyd oxyl g oup o gypsum (3550 cm-1 and
3400 cm-1). The bands o qua z and gypsum had
no mal o ien a ion (e.g. hey did no ha e nega i e
in ensi ies o bands). The ou h p incipal componen
(PC4) desc ibed 2.8 % o a iance and PC4 loadings
spec um included he same bands as PC 3 loadings
spec um bu en i e bands had nega i e in ensi ies.
The loadings spec a o he abo e-men ioned PCs a e
shown in Figu e 2.
desc ibed 9.3 % o a iance and i s loadings spec u
m
included again he kaolini e bands only. Con a y o
he PC1, he en i e spec al band in PC2 loadings
spec um had nega i e in ensi ies o bands. The hi
d
p
incipal componen (PC3) desc ibed 5.9 % o
a iance. The PC3 loadings spec um included also
nega i e bands o kaolini e and mo eo e a
double o Si-O s e ching bands o qua z (800 cm-1
and 780 cm-1) and a double o s e ching bands o
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012

M. Ri z e al.
52
Fig. 2 Loadings spec a o he i s ou p incipal componen s (Model 1).
Ten p incipal componen s (PCs) we e calcula ed
in Model 2. These en PCs desc ibed 99.9 % o
a iance o spec al in o ma ion, whe eas almos 97 %
o a iance was in ol ed in he i s ou PCs. PC1
loadings spec um included spec al bands o kaolini e
only as well as PC1 in Model 1. PC1 desc ibed 79.4
% o a iance. PC2 desc ibed 10.8 % o a iance an
d
i s loadings spec um included nega i e bands o
kaolini e and “posi i e” b oad and e y weak band a
3400 cm-1 (s e ching ib a ion o wa e ). PC3
desc ibed 4.0 % o a iance and i s loadings spec u
m
con ained nega i e bands o kaolini e and “posi i e”
b
ands o s e ching ib a ion o wa e (highe in ensi y
han in PC2). PC4 desc ibed 2.3 % o a iance an
d
PC4 loadings spec um included also nega i e bands
o kaolini e and “posi i e” band o wa e , whe eas he
in ensi y o he las -men ioned band was highe han
in PC3 loadings spec um.
The band a 1645 cm-1 (in PC2, PC3 and PC4
loadings spec um) belongs o de o ma ion ib a ion
o wa e . This band e y p obably p esen s mois u e
o samples. The band a 1645 cm-1 has no been used
o he c ea ion o Model 2.
The loadings spec a o he abo e-men ioned
PCs a e shown in Figu e 4.
3.3. MODEL 2 – IDENTIFICATION OF CLAY
MINERALS GROUPS
This model was c ea ed o classi ica ion o clay
mine als in o pa icula g oups o clay mine als.
Classi ica ion o clay mine als acco ding o Weiss and
Kuž a (Weiss and Kuž a , 2005) was used in his
s udy. Model 2 in ol ed six classes (Bio i e,
Musco i e-illi e, Chlo i e, Kaoline-se pen ine,
Smec i e and Ve miculi e). In his model he i s wo
classes ep esen ed he micas g oup o clay mine als;
he o he classes ep esen ed he emaining g oups o
clay mine als. The micas g oup had o be spli in o
wo classes (Bio i e and Musco i e-illi e) by eason o
b
ad c ea ion o clus e s. In he Bio i e class we e
samples o bio i e only; in he Musco i e-illi e class
we e samples o musco i e and illi e. In he Chlo i e
class we e chlo i e samples only. In he Kaoline-
se pen ine class we e samples o kaolini e, dicki e and
halloysi e. In he Smec i e class we e samples o
mon mo illoni e, non oni e and hec o i e. In he
Ve miculi e class we e samples o e miculi e only.
Wi h excep ion o sample Ill7 all mine als we e
clea ly classi ied in o app op ia e classes a e
pe o mance o disc iminan analysis. The abo e-
men ioned sample (mixed illi e-smec i e s uc u e)
was classi ied be ween he class Musco i e-illi e and
class Smec i e (Fig. 3).
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
APPLICATION OF INFRARED SPECTROSCOPY AND CHEMOMETRIC METHODS …
53
Fig. 3 Mahalanobis dis ance plo o classes Musco i e-illi e and Smec i e (Model 2).
Fig. 4 Loadings spec a o he i s ou p incipal componen s (Model 2).
Halloysi e). All mine als we e clea ly dis ibu ed in o
used classes by disc iminan analysis.
Ten p incipal componen s (PCs) we e calcula ed
in Model 3. These en PCs desc ibed 99.9 % o
a iance o spec al in o ma ion, whe eas almos
97.5 % o a iance was in ol ed in he i s h ee PCs.
PC1 desc ibed 82.5 % o a iance, PC2 desc ibe
d
11.3 % o a iance and PC3 desc ibed 3.6 % o
3.4. MODEL 3 – IDENTIFICATION OF MINERALS
OF KAOLINE-SERPENTINE GROUP
This model was c ea ed o iden i ica ion o
indi idual membe s o he Kaoline-se pen ine g oup
o clay mine als. Kaolini e, dicki e and halloysi e we e
used in his s udy as ep esen a i es o he Kaoline-
se pen ine g oup o clay mine als. Thus he Model 3
included h ee classes (Kaolini e, Dicki e an
d
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
M. Ri z e al.
54
Fig. 5 Loadings spec a o he i s h ee p incipal componen s (Model 3).
Fig. 6 Loadings spec a o he i s h ee p incipal componen s (Model 4).
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012
APPLICATION OF INFRARED SPECTROSCOPY AND CHEMOMETRIC METHODS …
55
ib a ion (840 cm-1), Si-O s e ching band o silica
(800 cm-1), coupled ou -o -plane ib a ion band
o Al-O and Si-O (625 cm-1) and de o ma ion
bands Al-O-Si, Si-O-Si, espec i ely (525 cm-1
and 470 cm-1). PC1 desc ibed 86.0 % o a iance.
PC2 desc ibed 7.2 % o a iance and i s loadings
spec um included nega i e bands a 3640 cm-1,
3440 cm-1 and 1650 cm-1 and posi i e bands a
800 cm-1, 625 cm-1, 525 cm-1 and 470 cm-1. PC3
desc ibed 4.3 % o a iance and he loadings spec u
m
con ained nega i e bands a 3640 cm-1, 3440 cm-1,
1650 cm-1, 525 cm-1 and 470 cm-1. All bands in he
loadings spec a o PC2 and PC3 belonged o
mon mo illoni e oo. The band a 1650 cm-1 has no
b
een used o he c ea ion o Model 4 by eason o
p
o en ial in e e ence in ensi y o his band by
samples mois u e.
The loadings spec a o he abo e-men ioned
PCs a e shown in Figu e 6.
3.6. MODEL 5 – IDENTIFICATION OF MINERALS IN
GROUP OF MICAS
This model was c ea ed o iden i ica ion o
indi idual membe s o micas. Bio i e, musco i e an
d
illi e we e used in his s udy as ep esen a i es o his
g oup o clay mine als. Model 5 included ou classes
(Bio i e, Musco i e, Illi e and Mixed illi e-smec i e
s uc u e). All mine als we e clea ly dis ibu ed in o
used classes by disc iminan analysis. The
Mahalanobis dis ance plo o Musco i e and Illi e
classes is shown in Figu e7.
The clus e s o musco i e and illi e we e
ela i ely close o each o he bu hey we e clea ly
sepa a ed. IR spec a o musco i es and illi es can be
e y simila (on accoun o simila chemical
composi ion and s uc u e) and common iden i ica ion
a iance. All h ee PC loadings spec a included
nega i e bands o kaolini e. The pa icula loadings
spec a di e ed only in in ensi ies o bands. Bands o
he highes in ensi ies we e in PC1 loadings spec um.
The band a 1440 cm-1 (nega i e in ensi y in PC2
loadings spec um and posi i e in ensi y in u eC3
loadings spec um) belongs o s e ching ib a ion o
ca bona es p esen as impu i y in he sample o
dicki e. This band has no been used o he c ea ion
o Model 3. The loadings spec a o he abo e-
men ioned PCs a e shown in Figu e 5.
3.5. MODEL 4 – IDENTIFICATION OF MINERALS
OF SMECTITE GROUP
This model was c ea ed o iden i ica ion o
indi idual membe s o he smec i e g oup o cla
y
mine als. Na-mon mo illoni e, Ca-mon mo illoni e,
non oni e and hec o i e we e used in his s udy as
ep esen a i es o his g oup o clay mine als. Model
4 included i e classes: Na-mon mo illoni e (Mon1 -
Mon5), Ca-mon mo illoni e (Mon6 - Mon8),
N
on oni e, Hec o i e and Mixed illi e-smec i e
s uc u e. All mine als we e clea ly dis ibu ed in o
used classes by disc iminan analysis.
Ten p incipal componen s (PCs) we e calcula ed
in Model 4. These en PCs desc ibed 99.9 % o
a iance o spec al in o ma ion, whe eas almos
97.5 % o a iance was in ol ed in he i s
h ee PCs. The loadings spec um o he i s
p incipal componen (PC1) included spec al bands
o mon mo illoni e only: s e ching bands o
s uc u al hyd oxyl g oups (3640 cm-1), b oad
s e ching band o wa e (3440 cm-1), de o ma ion
band o wa e (1650 cm-1), band o Si-O s e ching
(1040 cm-1), de o ma ion bands o Al-Al-OH
ib a ion (915 cm-1), de o ma ion bands o Al-Mg-OH
Fig. 7 Mahalanobis dis ance plo o Musco i e and Illi e classes (Model 5).
DSpace VŠB-TUO
h p://hdl.handle.ne /10084/84629
21/06/2012