Ac a Geodyn. Geoma e ., Vol. 9, No. 4 (168), 511–520, 2012
DETERMINATION OF CHLORITE, MUSCOVITE, ALBITE AND QUARTZ IN
CLAYSTONES AND CLAY SHALES BY INFRARED SPECTROSCOPY AND PARTIAL
LEAST-SQUARES REGRESSION
Michal RITZ 1) *, Lenka VACULÍKOVÁ 2), E a PLEVOVÁ 2),
Dalibo MATÝSEK1) and Jiří MALIŠ 1)
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 Ap il 2012, accep ed Sep embe 2012)
ABSTRACT
The objec i e o his wo k is he chemome ic quan i ica ion o mine als in ocks. A chemome ic me hod was de eloped o
he de e mina ion o chlo i e, musco i e, albi e and qua z in clays ones and clay shales using in a ed spec oscopy. B omide
pelle s and di use e lec ance we e used o measu e he in a ed spec a; p incipal componen analysis and pa ial leas -
squa es eg ession we e used as chemome ic me hods. Spec al egions (4000-3000 cm-1 and 1300-400 cm-1) con aining
impo an spec al in o ma ion we e chosen by p incipal componen analysis. The calib a ion models we e c ea ed by a pa ial
leas -squa es eg ession. The mean ela i e e o and ela i e s anda d de ia ion we e calcula ed o he assessmen o
accu acy and ep oducibili y. The alue o he mean ela i e e o was abou 10 % o mos o he calib a ion models. The
alue o he ela i e s anda d de ia ion anged om 1.1 o 3.0 % o mos calib a ion models based on di use e lec ance
spec a and om 4.0 o 9.2 % o mos calib a ion models based on spec a ob ained wi h b omide pelle s.
KEYWORDS: clays one and clay shale, in a ed spec oscopy, chemome ics, chlo i e, musco i e, albi e, qua z
me hods can also be used o he pu pose o
quan i a i e analysis. The heo y o chemome ic
me hods has been desc ibed in many pape s and
books, such as F ede icks e al., 1985; Geladi and
Kowalski, 1986; Lo be and Kowalski, 1988; Ma ens
and Naes, 1989. Nume ical me hods, such as mul iple
linea eg ession (MLR), p incipal componen
eg ession (PCR) o pa ial leas -squa es (PLS)
eg ession, a e inc easingly being used o ci cum en
he p oblems posed by he p esence o in e e ences,
spec al o e lap o majo ma ix e ec s (Luis e al.,
2004). The e a e many applica ions o he
chemome ic me hod o IR spec oscopic analysis
(Fulle e a ., 1988; Haaland and Thomas, 1988; Iñón
e al., 2003; A men a e al., 2007; B een e al., 2008).
In ecen yea s, chemome ic me hods ha e also been
applied o o he analy ical me hods, such as
ol amme y o ch oma og aphy (Moneeb, 2006; Al-
Degs e al., 2008; Wagieh e al., 2010; Zapa a-U zua
e al., 2010).
The spec oscopic applica ions o chemome ic
me hods end o use he ull spec um. This app oach
p
o ides a mo e accu a e desc ip ion o he model
han a single measu emen a a speci ic wa eleng h o
wa enumbe , espec i ely. Howe e , he ull-
spec um applica ions pose some p oblems: 1) pa o
he in o ma ion ha is ga he ed can be edundan ; and
2) he measu ed signal o some wa eleng hs may be
noisy o nonlinea (Luis e al., 2004). The mos use ul
INTRODUCTION
The ype and con en o mine als p esen in ocks
ha e a signi ican in luence on he beha io and
p
ope ies o he ocks as well as on he whole oc
k
massi . A de ailed quali a i e and quan i a i e mine al
analysis is he e o e a necessa y s ep o he
cha ac e iza ion o he p ope ies o he ocks in
geological, geochemical and geomechanical s udies.
The expe imen al esul s can be applied subsequen ly
in geomechanics and mining ac i i ies.
Se e al exis ing con en ional analy ical me hods
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), in a ed (IR) spec oscopy, Raman
spec oscopy, he mal g a ime ic/di e en ial he mal
analysis (TG/DTA) and bulk chemis y analysis
(Kodama e al., 1989; Chipe a and Bish, 2001;
S odon, 2002; Vog e al., 2002). Un o una ely, he
exac de e mina ion o mine als (especially clay
mine als) in he ocks by 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 he a iable chemical
composi ion and common s uc u al anomalies o clay
mine als. The indi idual clay mine als occu in he
o m o mix u es wi h a ious a ios o he pa icula
clay mine als.
Cu en IR spec oscopy ep esen s a as ,
eliable and e icien ool o phase analysis. This
me hod especially combined wi h chemome ic
M. Ri z e al.
512
G aphically, equa ions (3) and (4) can
b
e shown as in
Figu e 1 (Geladi and Kowalski, 1986). In Figu e 1,
n is numbe o samples, m is numbe o independen
a iables (e.g. abso bance alues), a is numbe o
ac o s and pis numbe o dependen a iables
(concen a ion alues). PLS comp ises ela i ely
complica ed calcula ion p ocedu es. Thei de ailed
desc ip ion is beyond he in en ion o his
p
ape .
A mo e de ailed desc ip ion o a ma hema ical
algo i hm o PLS modeling can be ound in many
books o pape (e.g.: Geladi and Kowalski, 1986;
Wold e al., 2001; Hasegawa, 2002).
The PLS eg ession can be ca ego ized in o wo
p
ocedu es. PLS1 (some imes called s anda d PLS)
and PLS2 (some imes called global PLS). PLS1
employs in o ma ion om only one chemical
cons i uen o make he calib a ion; PLS1me hod
wo ks wi h he one-column Yma ix. PLS2 uses wo
and mo e chemical cons i uen simul aneously.
The aim o his s udy is he de e mina ion o
majo i y mine als (chlo i e, musco i e, albi e and
qua z) in clays ones and clay shales by I
R
spec oscopy combined wi h pa ial leas -squa es
eg ession.
EXPERIMENTAL
S
AMPLES
Eigh y-six samples (B1-B78 and CS1-CS8) o
clays ones and clay shales we e used o his esea ch.
The se en y-eigh samples (B1-B78) we e used as
a calib a ion se and eigh samples (CS1-CS8) we e
used as con ol samples. In calib a ion se we e hi y-
wo samples o clays ones (B1-B32) and o y-six
samples o clay shales (B33-B78). Fou con ol
samples (CS1-CS4) we e clays ones; ou con ol
samples (CS5-CS8) we e clay shales. The selec ion
o con ol samples was pe o med wi h ega d o
co e he en i e concen a ion ange o he analyzed
mine als. All he samples we e ob ained om he
collec ion o VŠB-Technical Uni e si y, Os a a.
echniques o chemome ic app oach in in a ed
spec oscopy a e p incipal componen eg ession
(PCR) and pa ial leas -squa es (PLS) eg ession. An
impo an ea u e o PCR is ac ha only spec al
in o ma ion (e.g. abso bance, Kubelka-Munk uni ,
e c…) is used o he gene a ion o basis ac o s. This
was no a p oblem when he concen a ion
in o ma ion is absolu ely accu a e. Howe e , in
p ac ice he concen a ion ma ix con ains e o o
noise. Ano he p oblem o PCR is ha collinea i y o
abso bance da a will make he calib a ion uns able
(Hasegawa, 2002). Some po en ial p oblems can be
sol ed by using a mo e s able chemome ic me hod
which akes bo h (abso bance and concen a ion)
ma ices in o accoun simul aneously. Fo his pu pose
PLS eg ession can be used as sui able me hod. In
PLS eg ession abso bance and concen a ion ma ices
a e used complemen a ily in a s able calib a ion.
PLS eg ession wo ks wi h wo ma ices, Xand
Y. The X ma ix con ains independen a iables – he
spec al da a (e.g. in abso bance uni ). The Y ma ix
consis s o he dependen a iables – quan i a i e
(concen a ion) da a. The NIPALS (Nonlinea
I e a i e Pa ial Leas Squa es) algo i hm is mos
o en used o c ea ion o PLS models. The PLS
models can be conside ed as consis ing o ou e
ela ions (X ma ix and Y ma ix indi idually) and an
inne ela ion (linking bo h ma ices) (Geladi and
Kowalski, 1986). The ou e ela ions o he X and
Y
ma ices can be w i en as ollows:
ETPEp X xh
h
h  (3)
FUQEquY Yh
hh  (4)
whe e h and uh a e sco e ec o s, ph and qh a e loading
ec o s; T and U a e sco e ma ices, P and Qa e
loading ma ices, E and F a e ma ices o esiduals.
Fig. 1 G aphical ep esen a ion o PLS (Geladi and Kowalski, 1986).
DETERMINATION OF CHLORITE, MUSCOVITE, ALBITE AND QUARTZ IN …
513
App oxima ely 5-10 mg o sample was g ound
wi h app oxima ely 400 mg o d ied KB . This
mix u e was used o collec IR spec a by he DRIFT
echnique. The IR spec a we e collec ed using he
FTIR spec ome e Nexus 470 (The moScien i ic,
USA) wi h a deu e a ed T iGlycine sul a e (DTGS)
de ec o . The measu emen pa ame e s we e he
ollowing: spec al egion 4000-400 cm-1; spec al
esolu ion 8 cm-1; 128 scans; and Happ-Genzel
apodiza ion. F eshly d ied KB was used o he
b
ackg ound measu emen . E e y sample was p epa e
d
and measu ed 3-5 imes. The mean IR spec um o
e e y sample was calcula ed. The mean IR spec a
we e subsequen ly used o he c ea ion o
chemome ic models.
Exac ly 0.5 mg o sample was g ound wi h
200 mg o d ied KB . This mix u e was used o
p epa e he b omide pelle . In his s udy, 13 m
m
diame e pelle s we e used. The pelle s we e p esse
d
b
y 10 ons o 30 seconds unde acuum. The I
R
spec a we e collec ed using he FTIR spec ome e
A a a 320 (The moScien i ic, USA) wi h DTGS
de ec o . The measu emen pa ame e s we e he
ollowing: spec al egion 4000-400 cm-1; spec al
esolu ion 8 cm-1; 64 scans; and Happ-Genzel
apodiza ion. An emp y sample compa men was use
d
o backg ound measu emen . E e y sample was
p
epa ed and measu ed only once. The IR spec
a
we e subsequen ly used o he c ea ion o
chemome ic models.
CHEMOMETRIC ANALYSIS
The chemome ic analysis was pe o med using
The Unsc amble 9.7 so wa e package (CAMO
So wa e AS, No way). PCA and PLS eg essions
we e used as ep esen a i e o chemome ic me hods.
PCA was used o p elimina y da a analysis: de ec ing
ou lie spec a and he speci ica ion o impo an
spec al egions. The PLS1 echnique was employed
o c ea e chemome ic models o he de e mina ion o
mine als (chlo i e, musco i e, albi e and qua z) in
clays ones and clay shales. Mul iplica i e sca e
co ec ion ( ull MSC) was pe o med o ans-
o ma ion o DRIFT spec a. The numbe o op imal
PLS pa ame e s we e de e mined by s a is ical
compa ison o PRESS alues as a unc ion o numbe s
o ac o . The model alida ion was pe o med by he
c oss- alida ion (CV). The segmen ed c oss-
alida ion was pe o med as he alida ion me hod o
calib a ion models. The size o he c oss- alida ion
segmen was wo samples.
The da a ma ices o spec al and concen a ion
in o ma ion o all samples in calib a ion se we e
p
epa ed. One da a ma ix was p epa ed om I
R
spec a o b omide pelle s; he nex da a ma ix was
p
epa ed om DRIFT spec a. P incipal componen
analysis (PCA) was pe o med o de ec ing he
ou lie spec a in he calib a ion se and o selec ion
o he impo an spec al egions. The ollowing plo s
we e p epa ed: he sco e plo o he i s wo p incipal
componen s, he in luence plo o he i s h ee
Clays one is a compac e y ine-g ained
sedimen a y ock consis ing p ima ily o clay-size
mine al pa icles ha a e commonly ep esen ed by
clay mine als. Clays ones a e only pa ially
decomposed in wa e and hei po osi y gene ally
a ies om 25 o 5 %. Clay shales a e simila in
composi ion bu do no decompose in he wa e , and
hei o al po osi y is less han 5 %. Mo eo e , shale is
lamina ed ( he ock is made up o many hin laye s).
Shales ha a e subjec o hea and p essu e o
me amo phism al e in o a ha d, issile ( he oc
k
eadily spli s in o hin pieces along he lamina ions),
me amo phic ock known as sla e. The samples o
clays ones we e aken om se e al lowe - o-middle
C e aceous s a a belonging o he Silesian uni o he
Mo a ian-Silesian Beskydy Moun ains. The samples
o clay shales a e aken om he Kyjo ice laye s ha
a e s a ig aphically adhe en o he Lowe
Ca boni e ous pe iod o he Mo a ian-Silesian a ea.
E e y sample was pul e ized in an aga e mill.
Pul e ized samples we e homogenized by ca e ul
shu ling and by epea edly being spilled.
A lis o calib a ion se including he con en o
he mine als is shown in Table 1.
P
OWDER XRD ANALYSI
S
The con en o mine als in samples o clays ones
and clay shales was de e mined by he quan i a i e
e e ence me hod - powde XRD analysis. The
Rie eld echnique, used as he quan i ica ion
echnique o di ac ion da a (Rie eld, 1969), is he
quan i a i e echnique o he c ys al s uc u e analysis
om powde di ac ion da a. The heo e ical
di ac og am is calcula ed on he basis o s uc u al
da a (e.g., c ys al symme y, uni cell pa ame e s,
a omic coo dina es and occupancy) o he mine als
ha a e p esen . The heo e ical di ac og am is
subsequen ly compa ed wi h he measu ed di ac ion
p
a e n using mul idimensional eg ession. The
Rie eld echnique is conside ed one o he bes
echniques o he quan i ica ion o powde di ac ion
da a, and i has been used in many s udies (Chipe a
and Bish, 2001; B ingley, 1980; Bish and Howa d,
1988; Hillie , 2000).
The pul e ized and homogenized samples we e
measu ed in he cu e es. Powde di ac ion
measu emen s we e ca ied ou on a ully-au oma ed
di ac ome e ID3003 (Rich Sei e -FPM, Ge many)
unde he ollowing condi ions: CoK adia ion/Fe
il e , 2 goniome e geome y, s ep mode wi h
0.05o 2s eps, 3s measu emen ime pe s ep and
wi h he digi al p ocessing o esul an da a. Fo
measu emen and semi-quali a i e e alua ion, he
so wa e packages Ray leX and Ray leX Au oquan
(GE Sensing & Inspec ion Technologies, USA) we e
used.
I
R MEASUREMENT
S
Two echniques o IR spec osco
p
y we e used in
his s udy;
b
omide pelle s and di use e lec ance I
R
Fou ie T ans o m (DRIFT) spec oscopy.
M. Ri z e al.
514
Table 1 Lis o calib a ion se o mine als.
Sample Chlo i e (w/w %) Musco i e (w/w %) Albi e (w/w %) Qua z (w/w %)
B1 8.2 40.3 3.0 44.9
B2 8.2 42.1 3.3 39.5
B3 8.9 39.0 2.6 46.2
B4 3.5 38.2 3.1 39.1
B5 5.4 8.0 3.5 56.4
B6 9.7 23.5 2.5 64.3
B7 10.5 35.1 1.8 52.6
B8 7.9 48.8 9.2 34.1
B9 6.2 65.5 8.4 20.0
B10 7.8 53.3 4.7 29.4
B11 8.8 56.0 5.6 36.1
B12 5.1 52.1 4.0 33.2
B13 < 1.0 2.2 < 1.0 48.7
B14 3.8 9.7 < 1.0 57.0
B15 < 1.0 16.5 < 1.0 30.6
B16 3.4 17.7 1.5 19.1
B17 < 1.0 1.6 < 1.0 31.4
B18 5.0 12.4 < 1.0 15.7
B19 2.8 17.5 < 1.0 37.2
B20 3.8 10.5 < 1.0 13.5
B21 5.5 14.4 < 1.0 27.3
B22 5.8 14.9 2.3 41.6
B23 < 1.0 10.9 < 1.0 23.1
B24 < 1.0 9.1 < 1.0 23.5
B25 6.0 10.1 1.8 22.1
B26 < 1.0 19.0 2.6 26.6
B27 < 1.0 10.7 1.1 25.2
B28 < 1.0 28.4 < 1.0 28.4
B29 < 1.0 13.4 < 1.0 34.5
B30 1.6 22.6 < 1.0 61.4
B31 < 1.0 9.3 1.7 34.0
B32 5.4 15.4 < 1.0 23.9
B33 22.2 39.5 18.9 19.5
B34 17.6 32.4 19.8 30.1
B35 15.6 27.8 16.8 39.8
B36 9.1 21.6 15.6 53.8
B37 < 1.0 < 1.0 1.8 69.6
B38 < 1.0 34.1 < 1.0 29.2
B39 < 1.0 24.1 < 1.0 6.5
B40 10.4 23.2 18.0 48.4
B41 28.1 41.6 9.1 18.9
B42 17.2 31.1 17.9 26.3
B43 17.2 53.8 3.7 25.2
B44 22.7 25.1 21.4 30.3
B45 17.6 37.2 8.9 22.0
B46 < 1.0 71.9 2.9 5.0
B47 17.0 29.2 14.6 25.8
B48 11.6 47.6 3.8 36.9
B49 29.6 43.3 7.0 20.2
B50 19.2 33.7 19.0 28.1
B51 54.6 12.5 2.5 27.4
B52 3.3 38.1 11.2 43.0
B53 9.6 18.6 41.5 30.4
B54 12.3 25.0 2.3 49.5
B55 20.2 33.3 13.4 32.5
B56 18.0 25.4 14.7 36.9
B57 22.8 43.9 10.6 21.1
B58 25.5 42.5 2.7 27.5
B59 19.8 50.4 1.4 29.4
B60 19.6 30.4 13.2 33.1
B61 20.9 31.1 12.9 35.2
B62 23.4 36.5 12.3 27.8
B63 22.8 34.0 10.9 32.4
B64 13.9 33.8 16.4 35.9
B65 10.8 9.9 27.0 52.3
B66 9.9 42.0 13.5 34.3
B67 17.3 27.3 17.3 32.7
B68 17.6 34.7 13.2 34.5
B69 21.0 44.9 3.0 27.7
B70 22.5 45.9 3.0 25.9
B71 18.9 32.5 15.3 33.4
B72 16.9 23.6 13.0 40.6
B73 21.6 31.9 12.6 34.0
B74 19.5 35.0 16.2 25.0
B75 14.7 31.8 18.3 31.0
B76 16.6 31.3 16.7 33.7
B77 20.5 30.2 17.0 31.3
B78 19.0 29.7 15.8 35.5
DETERMINATION OF CHLORITE, MUSCOVITE, ALBITE AND QUARTZ IN …
515
Fig. 2 Sco e plo (a) and in luence plo (b) o DRIFT spec a.
Fig. 3 Line loading plo o b omide pelle s spec a.
loading plo s we e used o he selec ion o he
impo an spec al egions. This ype o loading plo
looks like spec um. Thus he impo an spec al
egions ha e cha ac e o spec al bands. The
impo an spec al egions de e mined by loading plo s
o bo h da a ma ices we e 4000-3000 cm-1 and 1300-
400 cm-1. The spec al bands p esen a he spec al
egion 4000-3000 cm-1 belonged o he s e ching
ib a ion o s uc u al hyd oxyl g oups (3630 cm-1)
p incipal componen s and he loading plo o he i s
p
incipal componen . The sco e plo s and he
in luence plo s we e used o de ec he ou lie spec a.
In he sco e plo , he ou lie spec a a e loca ed ou side
o main clus e . In he in luence plo , he ou lie
spec a do no show dec easing endency. No ou lie
spec a we e ound in ei he o he da a ma ices.
Examples o bo h plo s o spec a measu ed by
DRIFT echnique a e shown in Figu e 2. The line

M. Ri z e al.
516
Table 2 Pa ame e s o PLS models.
Mine al Me hod RMSEC
(% w/w)
RMSECV
(% w/w)
No. o ac o s Explained a iance
(%)
Chlo i e DRIFT 2.67 3.34 8 98.3
KB pelle 3.71 4.51 9 96.3
Musco i e DRIFT 5.58 6.85 8 97.8
KB pelle 11.39 11.74 9 95.9
Albi e DRIFT 2.38 2.71 8 98.5
KB pelle 2.22 3.44 9 96.5
Qua z DRIFT 4.91 6.50 11 90.1
KB pelle 4.79 6.64 11 87.7
alida ion. The alida ion e o o he model was
exp essed by RMSECV, analogous o RMSEC:
n
cc
RMSECV
n
i e e ip ed ali



1
2
,,, )(
(2)
whe e ci, al,p ed is he alue o he mine al con en o
he i h alida ion sample p edic ed by he PLS model,
and ci, e e is he alue o he mine al con en o he i h
alida ion sample ob ained by he e e ence me hod
and nis he numbe o samples in he calib a ion se .
The numbe o ac o s is he op imal numbe o “la en
a iables” necessa y o e ec i ely desc ibe he PLS
model. The numbe o ac o s was ob ained by so
called PRESS plo (i.e., he plo o PRESS s. he
numbe o ac o s). PRESS mean p edic ed esidual
e o sum o squa es and his pa ame e was
calcula ed as:


 n
i e e ip edi ccPRESS
1
2
,, )( (3)
whe e ci,p ed is he alue o he mine al con en o he
i h sample p edic ed om he PLS model, and ci, e e is
he alue o he mine al con en o he i h sample
ob ained by he e e ence me hod and n is he numbe
o samples in he calib a ion se . The explained
a iance is he pe cen age o he a iance o he
sys em desc ibed by he exp essed numbe o ac o s.
RESULTS AND DISCUSSION
A
NALYSIS OF CONTROL SAMPLES (ACCURACY
A
ND PRECISION)
The p edic i e abili y o he PLS models was
es ed by analysis o he eigh con ol samples (CS1-
CS8). The KB pelle s we e p epa ed o each con ol
sample and he IR spec a o hese con ol samples
we e subsequen ly measu ed. The IR spec a o he
con ol samples we e also ob ained by he DRIFT
echnique (each spec um was p epa ed like he mean
spec um om h ee independen DRIFT measu e-
and o he s e ching ib a ion o wa e (3350 cm-1).
The mos signi ican spec al bands in he egion
1300-400 cm-1 could be assigned o ollowing
ib a ions: Si-O s e ching ib a ion (1030 cm-1),
he de o ma ion ib a ion o Al-Al-OH (930 cm-1),
he Si-O s e ching ib a ions o qua z (800 cm-1 and
780 cm-1) and he de o ma ion ib a ions o Al-O-Si
and Si-O-Si (530 cm-1 and 480 cm-1, espec i ely).
The assignmen o spec al bands was pe o med
acco ding o (Ri z e al., 2010). Example o line
loading plo o spec a o b omide pelle s (including
signi ican wa enumbe s) a e shown in Figu e 3. The
“nega i e” ea u e in Figu e 3 belonged o he
ib a ion o ca bona es; his band was no used o
c ea ion o calib a ion models.
The da a ma ices men ioned p e iously we e
used o he c ea ion o he PLS models. The PLS1
echnique was used: a sepa a e calib a ion model was
c ea ed o de e mina ion o each mine al. The
impo an pa ame e s o he PLS models ha we e
c ea ed a e shown in Table 2. The example o
eg ession be ween p edic ed and measu ed alues o
PLS model o musco i e (DRIFT spec a) is shown in
Figu e 4.
The ollowing pa ame e s a e shown in Table 2:
RMSEC ( oo mean squa ed e o o calib a ion),
RMSECV ( oo mean squa ed e o o c oss-
alida ion), numbe o PLS ac o s and pe cen age o
explained a iance. The calib a ion e o o he PLS
model was exp essed by RMSEC:
n
cc
RMSEC
n
i e e ip edcali



1
2
,,, )(
(1)
whe e ci,cal,p ed is he alue o he mine al con en o
he i h calib a ion sample p edic ed om he PLS
model, and ci, e e is he alue o he mine al con en o
he i h calib a ion sample ob ained by he e e ence
me hod (XRD analysis) and nis he numbe o
samples in he calib a ion se . The alida ion o PLS
models was pe o med by he segmen ed c oss-
DETERMINATION OF CHLORITE, MUSCOVITE, ALBITE AND QUARTZ IN …
517
Fig. 4 P edic ed s. measu ed plo (Musco i e PLS model; DRIFT spec a).
Table 3 Lis o con ol samples and esul s o hei analysis.
Chlo i e (% w/w) Musco i e (% w/w) Albi e (% w/w) Qua z (% w/w)
PLS PLS PLS PLS
Sample
XRD
DRIFT Pelle s
XRD
DRIFT Pelle s
XRD
DRIFT Pelle s
XRD
DRIFT Pelle s
CS-1 10.7 8.1 7.5 12.1 16.2 10.9 4.5 5.6 3.9 31.6 23.9 26.2
CS-2 3.0 5.0 2.6 37.4 39.8 42.3 6.7 4.8 6.2 52.9 47.2 47.9
CS-3 5.0 5.2 5.0 40.8 43.2 37.7 8.2 4.6 7.0 48.9 46.1 44.9
CS-4 14.8 12.4 13.0 87.5 59.2 33.8 5.9 7.3 5.8 14.9 15.8 19.5
CS-5 19.1 19.9 18.7 36.5 37.0 36.3 14.3 13.6 15.1 30.0 30.4 32.9
CS-6 24.3 20.8 23.8 30.9 29.8 33.5 14.5 15.2 14.2 29.4 28.3 29.4
CS-7 11.7 15.5 10.2 30.0 27.1 29.0 41.5 31.8 47.0 30.4 32.5 26.8
CS-8 21.0 20.3 16.5 27.2 30.2 25.1 12.0 12.3 10.4 39.8 33.3 34.4
The p edic i e abili y o chemome ic models
can be desc ibed using se e al alida ion diagnos ics.
The ollowing pa ame e s we e used in his s udy:
b
ias, s anda d e o o p edic ion (SEP) and mean
ela i e e o (RE). Bias and SEP pa ame e s we e
used acco ding (Esbensen, 2006); RE pa ame e s was
c ea ed o he pu pose o his s udy:


n
cc
bias
n
i e e ip edi



1,,
(4)

n
cc
SEP
n
i e e ip edi



1
2
,,
(5)
men s). All o hese spec a we e used o p edic ion
o he con en o mine als by he PLS models ha
we e c ea ed. The esul s o he analysis o con ol
samples om PLS models a e shown in Table 3
oge he wi h he con en o mine als in he con ol
samples ob ained om he e e ence me hod (XRD
analysis).
Fi s , he esul s o p edic ion o he con en o
mine als in con ol samples we e es ed o s a is ical
compliance wi h he e e ence alues ( esul s o XRD
analysis) o he con ol samples. The es ing
echniques we e he ollowing: F- es , - es (S uden ’s
es ) and pai ed compa ison (Meloun and Mili ký,
2004). All o hese echniques showed s a is ical
compliance be ween he p edic ed and e e ence
alues o he con ol samples.
M. Ri z e al.
518
Table 4 Pa ame e s o p edic ed abili y o PLS models.
Chlo i e Musco i e Albi e Qua z
Pa ame e DRIFT Pele s DRIFT Pele s DRIFT Pele s DRIFT Pele s
bias (w/w %) -0.3 -1.5 -2.5 -6.7 -0.1 0.1 -1.9 -1.2
SEP (w/w %) 2.4 2.1 10.3 19.1 1.7 1.6 4.9 4.2
RE (%) 20.7 11.5 12.9 13.9 19.4 9.1 10.5 12.6
Table 5 Rep oducibili y - lis o esul s.
Chlo i e (w/w %) Musco i e (w/w %) Albi e (w/w %) Qua z (w/w %)
CS5 B25 CS5 B25 CS5 B25 CS5 B25
DRIFT Pelle s DRIFT Pelle s DRIFT Pelle s DRIFT Pelle s DRIFT
P
elle s DRIFT Pelle s DRIFT
Pelle s DRIFT Pelle s
18.8 17.3 2.5 6.2 38.0 28.4 12.7 11.5 11.9 11.4 1.4 2.6 32.7 32.2 27.6 23.9
19.0 21.5 2.2 6.1 37.6 30.9 12.0 12.8 12.3 14.2 1.7 1.0 31.9 30.6 27.6 23.1
19.3 17.1 2.3 5.2 38.1 27.1 11.7 10.8 12.1 14.3 1.9 1.5 31.6 33.5 27.3 25.3
18.5 19.4 2.1 6.3 36.2 32.3 11.9 11.8 11.8 12.4 1.5 1.3 31.7 33.3 27.7 25.7
18.8 18.7 2.2 4.5 37.3 27.4 11.8 11.0 11.9 14.0 1.6 1.6 31.9 33.2 27.2 22.1
18.8 20.7 2.3 6.9 37.0 28.6 12.0 11.5 11.8 12.7 1.3 3.2 31.2 34.0 27.6 24.5
18.5 18.7 2.1 6.7 37.1 25.3 10.9 10.1 11.4 11.7 1.6 2.5 32.4 30.8 28.2 25.7
18.7 17.2 2.7 5.6 38.0 27.9 9.7 9.4 11.2 13.6 1.7 3.2 32.7 34.4 28.1 25.5
18.5 18.0 3.0 5.5 37.2 26.8 10.3 12.2 11.4 14.1 1.4 1.1 32.7 32.0 27.6 22.3
18.9 17.5 2.1 5.6 37.9 28.5 11.9 10.7 11.6 13.2 1.5 2.2 32.8 32.6 27.8 23.3
RSD
(%) 1.4 8.3 12.8 12.1 1.6 7.1 7.9 8.8 3.0 9.2 10.9 40.5 1.7 4.0 1.1 5.8
exp essed by he RE showed alues simila o he RE
o he es o models. The eason is ob ious om he
di e en ma hema ical o mulas o bias and RE (see
equa ions 4 and 6). The bias equa ion (4) does no
ope a e wi h he absolu e alues o he di e ence
p edic ed and he e e ence con en o he mine al.
N
ega i e and posi i e inc emen s could he e o e
cancel each o he . The RE equa ion (6) used absolu e
alues o he di e ence. RE is he e o e he mo e
obus pa ame e o accu acy, whe eas bias can
desc ibe he sys ema ic e o o he PLS model. The
accu acy pa ame e (RE) showed simila alues in
mos o he models (app oxima ely 10 %). Only
DRIFT models o he p edic ion o chlo i e and albi e
had alues o RE o app oxima ely 20 %. The alues
o he accu acy o he PLS models ha we e c ea ed
a e e y simila o he alues o accu acy o he
esul s o he XRD analysis ci ed in he li e a u e (e.g.
Moo e and Reynolds, 1997). Requi emen o p ac ice
o RE pa ame e o quan i a i e phase analysis esul s
o ock a e abou 20-25 %. The alues o RE
p
a ame e s o con ol samples achie ed by e e ence
quan i a i e me hod (XRD analysis) in his s udy
we e be ween 10 % and 20 %. Thus bo h me hods
(XRD analysis and chemome ic analysis o I
R
spec a) p o ided esul s accep able by po en ial
cus ome s.
The bes alues o he pa ame e o p ecision
(SEP) had PLS models o he p edic ion o albi e an
d
chlo i e ( alues o SEP abou 2 w/w %). The models
o p edic ion o qua z showed sligh ly wo se
p ecision. The model o he p edic ion o musco i e
had d ama ically wo se p ecision han o he PLS
models, p obably caused by he wo se pa ame e s o
he musco i e PLS models (see Table 2). Wi h he
100
1,
,,












n
c
cc
RE
n
i e e i
e e ip edi
(6)
whe e ci,p ed is he alue o he mine al con en o he
i h con ol sample p edic ed om he PLS model, ci, e e
is he alue o he mine al con en o he i h con ol
sample ob ained by he e e ence me hod (XRD) and
n is numbe o con ol samples.
Bias ep esen s he a e age di e ence be ween
he p edic ed alues and he e e ence alues o
con ol samples and is a commonly-used measu e o
he accu acy o a chemome ic model. Bias is also
used o check any sys ema ic di e ences obse ed
b
e ween he a e age alues o he con ol samples and
he alida ion samples (Esbensen, 2006). Ano he way
o exp ess he accu acy o a chemome ic model is he
mean ela i e e o . The s anda d e o o p edic ion
(SEP) exp esses he p ecision o he p edic ed esul s.
The alues o he pa ame e s o he alida ion
diagnos ics a e shown in Table 4.
The bias o almos all o he PLS models was
nega i e; only he bias o he PLS model o albi e o
KB pelle s had a posi i e alue. The absolu e alues
o bias we e usually e y low, anging om 0.1 o
1.9. Only he bias o he models o musco i e showed
highe alues; he model o DRIFT had a bias o -2.5
and he model o he KB pelle s had a bias alue o
-6.7. These high alues o bias we e e y p obably
caused by he wo se alues o he pa ame e s o he
musco i e PLS models (see Table 2).
The accu acy as exp essed by he bias showed
wo se alues o he musco i e PLS models, whe eas
he accu acy o he musco i e PLS models as
DETERMINATION OF CHLORITE, MUSCOVITE, ALBITE AND QUARTZ IN …
519
Howe e , he use o a hea ie sample could induce
o al abso bance o some spec al bands and cause
poo e u iliza ion o spec al in o ma ion.
CONCLUSIONS
P incipal componen analysis (PCA) and
especially PLS eg ession we e used as chemome ic
me hods in his pape . PCAs we e used o de ec ing
ou lie s and o selec ion o impo an spec al
egions. No ou lie s we e de ec ed. Regions 4000-
3000 cm-1 and 1300-400 cm-1 we e selec ed o he
c ea ion o calib a ion models by he PLS eg ession
echnique. The se ies o PLS models p oposed in his
wo k showed RMSEC and RMSECV alues up o 5-
6 w/w % (wi h he excep ion o he PLS model o
p
edic ion o musco i e based on he KB pelle
echnique). Fo a se o con ol samples, alues o
mean ela i e e o (RE) o abou 10 % we e achie ed
in mos o he PLS models ha we e c ea ed. The
sligh ly be e alues o RE we e achie ed o PLS
models based on he KB pelle echnique. The
ep oducibili y o he PLS models was e alua ed o
wo samples and he ep oducibili y was exp essed by
he ela i e s anda d de ia ion (RSD); he alues o
he RSD anged om 1.1 o 12.8 % (wi h he
excep ion o he PLS model o p edic ion o albi e
b
ased on he DRIFT echnique). The ela i ely be e
alues o he RSD we e achie ed o he PLS models
based on he DRIFT echnique.
The e a e some impo an ad an ages o
chemome ic analysis o IR spec a o e mos used
quan i a i e phase analysis me hod (Rie eld
echnique o XRD analysis): analysis ime,
accessibili y and simplici y. Fi s , he analysis ime o
chemome ic analysis o IR spec oscopy is much
sho e han analysis ime o XRD analysis. O cou se,
c ea ion o calib a ion models is a he ime-
consuming. Bu subsequen analysis is e y as .
Second, IR spec ome e s a e p esen in signi ican ly
la ge numbe s o labo a o ies han X- ay
di ac ome e . The main easons a e he
p
u chase
p ice o IR spec ome e s and ope a ing cos s, which
helped o sp ead he use o in a ed spec oscopy as
a common analy ical echnique. And inally,
chemome ic esul p ocessing o IR spec a is
conside ably simple han Rie eld ea men o
di ac og am da a.
The use o chemome ic da a ea men o he I
R
spec a is simple and eliable and can be used o
de e mine mine als in ocks. This s udy showed ha
IR spec oscopy in conjunc ion wi h he PLS
eg ession me hod p o ided an accep able al e na i e
o he mos commonly used me hods o quan i a i e
phase analysis – he Rie eld echnique o XRD
analysis.
ACKNOWLEDGMENTS
This pape was c ea ed by he p ojec No.
CZ.1.05/2.1.00/01.0040 "Regional Ma e ials Science
excep ion o he models o he p edic ion o
musco i e, he o he models had e y simila alues
o SEP o DRIFT and KB pelle spec a. The
p
ecision o analysis o con ol samples by e e ence
quan i a i e me hod (XRD analysis) in his s udy we e
be ween 10 % and 20 %.
R
EPRODUCIBILITY
One sample om he calib a ion se (B25) and
one con ol sample (CS5) we e used o es ima e he
ep oducibili y o he PLS models ha we e c ea ed.
Fo h ee weeks, en KB pelle s we e p epa ed om
each sample, and hei IR spec a we e measu ed. In
he same pe iod, en IR spec a we e also ob ained by
he DRIFT echnique; each spec um was p epa ed
like he mean spec um om h ee independen
DRIFT measu emen s (including homogeniza ion and
g inding wi h KB ). All o hese spec a we e used o
he p edic ion o he con en o mine als by he PLS
models ha we e c ea ed. The ep oducibili y was
exp essed by he ela i e s anda d de ia ion (RSD).
RSD was calcula ed om he esul s ha we e
ob ained:

100
1
1
2






P
n
iiP
x
n
xx
RSD
(11)
whe e xi is he p edic ed alue o he mine al con en
o he i h analysis o ep oducibili y, xPis he mean
alue o he p edic ed mine al con en and nis numbe
o analyses. The esul s and he calcula ed RSD a e
shown in Table 5.
Wo se alues o ep oducibili y we e ob ained
om he analysis o he IR spec a using KB pelle s.
Values o he RSD in he analyses o chlo i e,
musco i e and albi e we e app oxima ely 10 %; he
alues o he RSD in he analysis o qua z we e
app oxima ely 5 %. The ex emely high alue o he
RSD o sample BP5 (in he analysis o albi e) was
caused by he e y low con en o albi e in his
sample.
Signi ican ly be e alues o he RSD we e
ob ained om he analysis o he IR spec a measu ed
b
y he DRIFT echnique. Mos o alues o he RSD
we e wi hin he ange 1-3 %. The highe alues o he
RSD (abou 10 %) we e ob ained om he analysis o
samples wi h a low con en o mine als (chlo i e,
musco i e and albi e in sample BP5).
Di e en alues o he ep oducibili y om he
analysis o DRIFT spec a and he spec a om KB
p
elle s we e e y p obably caused by he di e en
weigh s o samples in p epa ing o IR measu emen .
Fo p epa a ion o DRIFT measu emen s, 5-10 mg o
he samples was used, whe eas o p epa ing KB
pelle s, only 0.5 mg o sample was used. This
ela i ely low weigh is a he limi s o accu a e
measu emen o common analy ical balances and
p
obably caused an ele a ed e o o weighing.