Anda di halaman 1dari 12

Geophys. J. Int.

(2000) 142, 83–94

Statistical analysis and origin of the magnetic susceptibility of drill


cuttings from the 9.1-km-deep KTB drill hole

Armin Rauen, Heinrich C. Soffel and Helmuth Winter


Geophysical Institute, University of Munich, T heresienstrasse 41, D-80333 Munich, Germany. E-mail: rauen@geophysik.uni-muenchen.de

Accepted 2000 February 2. Received 2000 February 2; in original form 1999 July 31

SU M MA RY
Magnetic susceptibility was measured at 2 m depth intervals on drill cuttings from
the main drill hole of the German Deep Drilling Project KTB. Metamorphic rocks
(metabasites and gneisses) were the rock types most frequently found down to a depth
of 9101 m. Petrophysical (susceptibility, density), geochemical (element concentrations),
lithological and petrological data (ore mineral concentrations, lithological components,
alteration index) were used for a statistical analysis. The histograms of magnetic
susceptibility show nearly log-normal distributions with two distinct peaks depending
on the lithology. The most frequent susceptibility values are 0.266×10−3 SI for gneissic
rocks and 0.847×10−3 SI for metabasic rocks (mainly amphibolites). The higher level
of metabasite susceptibility is caused by higher contents of paramagnetic silicates
such as hornblende. A theoretical paramagnetic susceptibility was calculated from the
iron and manganese contents derived from X-ray fluorescence (XRF) measurements.
The ferrimagnetic susceptibility was determined by subtracting the theoretical para-
magnetic susceptibility from the measured susceptibility. Cross-plots of the ferrimagnetic
susceptibility versus density are used to discriminate between samples with predominantly
magnetite or pyrrhotite as the main ferrimagnetic mineral. Samples containing mostly
pyrrhotite show susceptibilities not exceeding 6×10−3 SI, whereas the highest measured
susceptibilities of 66.5×10−3 SI correspond to zones exclusively with magnetite. A
factor analysis was applied to investigate the background factors representing the data
variabilities. The factor analysis reduces 13 original variables from the complete depth
section to five independent initial factors. These explain in total 66.2 per cent of the
total data variance. The most significant factor, 1, correlates with metabasite content,
density and paramagnetic susceptibility and it anticorrelates with gneiss content. The
next significant factor, 2, correlates with ferrimagnetic susceptibility and magnetite
content. Factor 3 correlates with the amount of cataclastic rocks, factor 4 with
hornblende gneiss and factor 5 with pyrrhotite
Key words: continental crust, density, magnetic anomalies, magnetic susceptibility,
magnetite, statistical methods.

continental sedimentary and crystalline rocks (e.g. Soffel 1991;


1 I NT R O DU C TI O N
Henkel 1994; Dunlop & Özdemir 1997). Pyrrhotite occurs in
Anomalies of the static geomagnetic field are commonly some metamorphic rocks and sediments (Thompson & Oldfield
observed features in geophysical prospecting. These anomalies 1986; Soffel 1991; Clark & Tonkin 1994; Henkel 1994; Pucher
are in general caused by increased amounts of ferrimagnetic 1994; Schwarz & Broome 1994). Due to weathering and
minerals in the rocks, mainly iron oxides of the solid-solution oxidation, pyrrhotite is often altered to magnetite and is
series between magnetite and ulvöspinel (titanomagnetites) and, therefore rare or even absent in surface outcrops, while it can
less frequently, by members from the solid-solution series still be preserved in lower parts of geological bodies. Magnetic
between haematite and ilmenite (titanohaematites). In addition, anomalies of low amplitudes (<10 nT) can also be caused
monoclinic pyrrhotite as an iron sulphide contributes signifi- exclusively by paramagnetic minerals.
cantly to surface magnetic anomalies. Other ferrimagnetic or The German Continental Deep Drilling Project KTB
ferromagnetic minerals play a minor role. Magnetite and titano- (Kontinentales Tief bohrprogramm der Bundesrepublik
magnetite are observed in ocean floor basalts and in many Deutschland) offered the unique chance to investigate the

© 2000 RAS 83

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
84 A. Rauen, H. C. SoVel and H. W inter

sources of magnetic anomalies in unweathered samples down


2 DATA B AS I S
to a depth of 9.1 km. The KTB drill site is located at the
western margin of the Bohemian Massif in an area with a The petrophysical, mineralogical and geochemical data from drill
prominent magnetic anomaly (Soffel et al. 1989; Bosum et al. cuttings used in this analysis were determined in the KTB field
1997). The KTB pilot drill hole (KTB-VB) was completed in laboratory (Emmermann et al. 1995). The data are stored at
1989 and reached a depth of 4.0 km. It was nearly completely the GeoForschungsZentrum (GFZ) in Potsdam (Germany)
cored. The KTB main hole (KTB-HB) was completed in 1994 and are available from http://icdp.gfz-potsdam.de/html/ktb/.
at a depth of 9101 m. The material extracted from the main The cuttings were sampled every 2 m (on average) in the
hole consists mostly of drill cuttings (sampled every 2 m) and depth interval 8–9101 m. Each sample was carefully washed
a few drill cores (80 m of core recovery in total ). Emmermann to remove drill mud and subsequently dried before the measure-
& Lauterjung (1997) give an overview and a summary of the ments were started (Emmermann & Lauterjung 1997). In some
major results of the KTB project. samples, metallic iron particles originating from the drill bits
Drill cores are the most valuable kind of drill samples, occurred. These artefacts were removed by hand magnets.
because they can be attributed to an accurately determined Table 1 shows the petrophysical, geochemical and lithological
depth. Furthermore, they can be used to obtain structural parameters, the sampling intervals, the range of the values and
information on the rocks at depth. Drill cuttings, on the other the amount of data obtained.
hand, are a mixture of all rock types from the section in the Fig. 1 shows a plot of these data as depth logs together with
vicinity of the drilling tool. They can be contaminated to some the simplified lithology. Some correlations are obvious. The high
extent by rock fragments that have fallen down from upper densities correlate with metabasites, whereas the low densities
parts of the open drill hole below the casing. There is an are mainly linked with gneisses. The magnetic susceptibility is
uncertainty in the depth assignment of drill cuttings. The higher in metabasic regions than gneissic regions, but pyrrhotite
original depth of cutting samples is calculated using the lag and magnetite increase the susceptibility more significantly.
time of the drill mud from the bottom of the hole to the Pyrrhotite is abundant in all depth intervals and in all litho-
surface. Borehole depth and diameter, composition of the drill logies. It occurs as a monoclinic, ferrimagnetic phase and also
mud and pump rate increase the uncertainty compared to as a hexagonal, antiferromagnetic phase. The relative abundance
the true depth (Dietrich et al. 1992). A calibration made by of the hexagonal phase increases with depth; however, the
comparing gamma-ray measurements of drill cuttings and two pyrrhotite phases are not distinguished in the database.
by borehole logging shows a difference of both depths in the Magnetite occurs rarely and is concentrated only in distinct
range 0–10 m at a depth of 6000 m (Umsonst et al. 1995). In regions (Kontny et al. 1997). Increased magnetite concentrations
addition, only poor structural information can be derived from correlate clearly with high susceptibility peaks.
drill cuttings. However, drill cuttings are available in large For the univariate and bivariate statistical analyses (Sections 3
quantities and were sampled more or less continuously at 1 m and 5) and the calculations in Section 4, the original data con-
depth intervals. sisting of 13 variables as given in Table 1 were used. Due to
In addition to measurements on drill hole samples, various different sampling intervals, the amount of data available varies
downhole logging measurements were performed. The results from 572 to 5921. For cross-plots and factor analysis (Sections 5
of these investigations are given in Bosum et al. (1997), and 6), only data measured on identical samples from the
same depth were chosen in order to avoid distortions of the
ELEKTB Group (1997), Pechnig et al. (1997) and Bram &
statistical analysis by interpolated data. This reduces the data
Draxler (1994).
set to 446 cases (cross-plots) and 444 cases (factor analysis).
All samples from the two KTB drill holes were investigated
Each case of the data matrix consists of a complete set of
in the KTB field laboratory immediately after recovery of the
variables measured at the same depth. We used the complete
material (Emmermann et al. 1995). They were also studied
depth section (0–9101 m) for the analysis and did not separate
later in several laboratories in order to obtain a complete
depth intervals.
set of petrophysical, geochemical and lithological properties
(Berckhemer et al. 1997). The processes leading to the formation
of the ore minerals at the KTB location were reported by
Kontny et al. (1997). The petrophysical analyses of the cores 3 UN IVA R IATE S TAT IS TI C S :
from the 4 km deep pilot hole are given in Soffel et al. (1992) H IS TO G R A M S
and Huenges et al. (1997). The frequency distributions of the measured magnetic
In this paper we present the results obtained exclusively susceptibilities of metabasites, gneisses and all data together
from drill cuttings sampled from the KTB main hole down to are plotted in Fig. 2. The histogram of all rock types displays
the maximum depth of 9101 m. We show that drill cuttings, two maxima and is not symmetrical, with a tail towards
despite the loss of data quality compared with drill cores, increased values. The two maxima can be assigned to the two
are suitable for statistical investigations due to the high data main lithologies. Separation into data from samples containing
density. We will progress from univariate ( histograms) to exclusively gneiss and those containing exclusively metabasite
multivariate (factor analysis) statistics to describe the observed yields two approximate log-normal distributions. The number
data. Huenges (1997) applied factor analysis on non-magnetic of samples containing only gneisses (503) or only metabasites
KTB data previously. Leonardi & Kümpel (1998) investigated (67) is low compared to the total data set (4621 samples). The
magnetic susceptibility and other data by spectral analysis, tale with increased values in the gneiss distribution is most
quantified fractal terms and interpreted the results with respect probably due to the enhanced contents of ferrimagnetic phases
to crustal inhomogeneities. such as pyrrhotite. Latham et al. (1989) explained the log-normal

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
Magnetic susceptibility of drill cuttings 85

Table 1. Variables used for statistical analysis. The amount of cataclastic rocks is the sum of
cataclasites and cataclastically deformed rocks (cataclastic gneisses and cataclastic metabasites). The
ore mineral contents (pyrrhotite, magnetite, pyrite, ilmenite) were determined by optical analysis and
measured in an ordinal scale in five classes: 0 (0 volume per cent=‘none’), 1 (%1 volume per cent),
2 (<1 volume per cent), 3 (1–2 volume per cent) and 4 (>2 volume per cent=‘much’). The alteration
parameter used is defined as chlorite/(chlorite+biotite+hornblende+garnet+muscovite). A total of
49 000 data points were used (see Fig. 1).

Parameter Range of values Unit Sampling interval Number of data values

Petrophysics
magnetic susceptibility 0.1–66.5 10−3 SI 2m 4621
density 2.46–3.05 103 kg m−3 2m 4621
Lithology
gneiss 0–100 vol. % 1–4 m 5921
hornblende gneiss 0–97 vol. % 1–4 m 5921
metabasite 0–100 vol. % 1–4 m 5921
cataclastic rocks 0–100 vol. % 1–4 m 5921
Ore petrology
pyrrhotite 0–4 20 m 572
magnetite 0–4 20 m 572
pyrite 0–4 20 m 572
ilmenite 0–4 20 m 572
Geochemistry
iron (Fe O ) 2.4–15.2 weight % 2–4 m 4581
2 3
manganese (MnO) 0.05–0.31 weight % 2–4 m 4581
alteration parameter 0–1 2–4 m 4547

distribution of magnetic susceptibility by the similarly log- paramagnetic minerals such as hornblende in the metabasites,
normally distributed ore mineral concentration. We assume that which consist mostly of amphibolites (Bleil & Petersen 1982;
the distributions of paramagnetic and ferrimagnetic minerals Friedrich 1995). This effect will be explained in the next section.
both contribute to the distribution function.
The maximum values of the histograms (the modes)
represent the most frequent or characteristic values, which 4 EVA LUATI O N O F TH E PA R A M A G NET IC
are 0.266×10−3 SI for the gneisses and 0.847×10−3 SI for A N D FE R R IM A GN ET IC C O N TR I BU TIO N S
the metabasites. The most frequent gneiss value differs from the TO T HE M A G NET IC S U S CEP TI BI LIT Y
geometric mean (0.381×10−3 SI) because of the asymmetrical
distribution (attempts of arithmetic averaging led to inadequate The effective paramagnetic moment of an ion depends on the
results). The high values not belonging to a symmetric log- number of unpaired electrons in the 3-d atomic shell (Nagata
normal distribution are regarded as susceptibility anomalies. 1961; Bleil & Petersen 1982). The expected maximum number
The highest measured susceptibilities, >10×10−3 SI, are not of unpaired 3-d electrons is five. Fe3+ and Mn2+ represent
displayed in the gneiss or in the metabasite data. These this condition. Fe2+ ions contain four unpaired 3-d electrons.
susceptibility anomalies originate from mixed samples. These ions with an effective paramagnetic moment are most
The susceptibility distribution of the metabasites is shifted frequent in rocks of the Earth’s crust. Other ions contain-
to increased values compared to the values for the gneisses. ing four (Mn3+, Co3+, Cr2+) or three (Mn4+, Co2+, V2+, Cr3+)
The reason for this is the enhanced abundance of strongly or fewer unpaired 3-d electrons are less important (see Table 2).

Table 2. Element analysis from XRF measurements performed in the KTB field laboratory on
drill cuttings (0–9101 m depth, Emmermann & Lauterjung 1997) and the number of unpaired
3-d electrons (Bleil & Petersen 1982).

Element Naturally Number of unpaired Contents (weight %) analysed as


occurring ions 3-d electrons min. av max.

Fe Fe3+, Fe2+ 5,4 2.4 8.1 15.2 FeO+Fe O


2 3
Mn Mn2+ 5 0.05 0.13 0.31 MnO
Cr Cr2+, Cr3+ 4,3 0.002 0.013 0.074 Cr
V V2+–V5+ 3–0 0.005 0.016 0.047 V
Ni Ni2+ 2 0.001 0.006 0.031 Ni
Ti Ti4+ 0 0.21 1.22 3.39 TiO
2

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
86
A. Rauen, H. C. SoVel and H. W inter
© 2000 RAS, GJI 142, 83–94

Figure 1. Depth logs of measured properties of drill cuttings from the KTB main hole (0–9101 m) used in this paper. Simplified lithology (G=gneiss, G/M=alternating layer, M=metabasite);
petrophysical investigations: magnetic susceptibility, density; components from binocular observations: gneiss, hornblende gneiss, metabasite, cataclastic rocks; ore microscopy observations:
pyrrhotite, magnetite, pyrite, ilmenite, data are given in an ordinal scale of 5 classes: 0 (0 per cent=‘none’), 1 (%1 per cent), 2 (<1 per cent), 3 (1 per cent to 2 per cent) and
4 (>2 per cent=‘much’), all contents in volume per cent; monoclinic and hexagonal pyrrhotite are not distinguished; XRF measurements: Fe O , MnO; alteration parameter: defined as
2 3
chlorite/(chlorite+biotite+hornblende+garnet+muscovite) from XRD measurements (for sampling intervals, see Table 1).

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
Magnetic susceptibility of drill cuttings 87

where k is the volume susceptibility, r is the density, L is


Avogadro’s number, R is the gas constant, T is the absolute
temperature, n is the amount of component i (in mol per
i
gram), and p is the magnetic moment of component i (Curie’s
i
law in cgs units, see Bleil & Petersen 1982; Multiplication by
4p gives k in SI units). The paramagnetic volume susceptibility
can be derived from Curie’s law and can be calculated from
the known contents of Fe2+, Fe3+ and Mn2+ ions (FeO, Fe O
2 3
and MnO, respectively) using the simplified formula in SI units
k =r(2.096 FeO+2.258 Fe O +2.543 MnO) ,
paramagnetic 2 3
(2)
where k is the volume susceptibility (in SI units), r is the
density (in kg m−3) and FeO, Fe O and MnO contents are
2 3
in weight per cent (from Friedrich 1995). The XRF analyses
used measure only the total iron content, which is the sum of
FeO+Fe O . For further calculations, three assumptions are
2 3
made:
(1) an FeO/Fe O ratio of 4 : 1 (according to e.g.
2 3
Wedepohl 1969);
(2) the measured manganese content consists exclusively of
Mn2+ ions (MnO);
(3) none of the paramagnetic ions is present in the form of
ferrimagnetic minerals.
Therefore, the calculated paramagnetic susceptibility represents
the maximum value for the contribution of paramagnetic
minerals. All measured susceptibility values exceeding this
maximum represent a ferrimagnetic contribution. The ferri-
Figure 2. Frequency distributions of magnetic susceptibility. Data magnetic portion of the volume susceptibility was calculated from
from drill cutting measurements from the main drill hole (KTB-HB). the difference between measured and calculated paramagnetic
(a) All rock types—all available data including gneisses, metabasites, volume susceptibilities:
other rock types and mixed samples (4621 values). (b) Gneisses—
data from samples containing 100 per cent gneiss (503 values). k =k −k . (3)
ferrimagnetic measured paramagnetic
(c) Metabasites—data from samples containing 100 per cent metabasites
(67 values). The logarithms of the measured volume susceptibilities The results of this calculation of the paramagnetic susceptibility
(in 10−3 SI units) are divided into 100 classes (metabasites: 50) between were verified by hysteresis measurements. Hysteresis curves of
−1 and +2 (33 classes per decade, corresponding to susceptibilities magnetization as a function of applied magnetic field give
from 0.1×10−3 to 100×10−3 SI). The modes of histograms (b) and information about the paramagnetic portion from the slope at
(c) are marked and quantified. high fields. The comparison was performed on seven samples
(Table 3); they are plotted together with calculated susceptibilities
in Fig. 3. The calculated paramagnetic susceptibilities are
Curie’s law gives the relationship between the effective
slightly higher (9–47 per cent). This is due to the simplifying
paramagnetic moments of the components of a mixture and
assumption that all paramagnetic ions contribute exclusively
the resulting magnetic volume susceptibility k:
to the paramagnetic susceptibility (eq. 2). In reality, some of
k=r[L 2/(3RT )] ∑ n p2 , (1) these ions are incorporated in ferrimagnetic minerals as well.
i i

Table 3. Comparison of calculated and measured paramagnetic susceptibilities. Calculated:


derived from low-field measurements of total susceptibility and element concentrations
(eq. 2); measured: from the high-field slope in hysteresis measurements (gns=gneisses;
amph=amphibolites; mob=mobilisates, hbl-gns=hornblende gneisses).

Depth (m) Main rock types Susceptibility (10−3 SI)


total calc. para. measured para.

648 gns. 95% 0.408 0.384 0.261


3302 amph. 90%, mob. 8% 0.578 0.565 0.476
4552 amph. 81%, hbl-gns. 12% 0.898 0.578 0.491
6000 amph. 95% 0.908 0.620 0.512
7160 amph. 91%, mob. 5% 1.518 0.571 0.395
7480 amph. 84%, mob. 10% 51.018 0.665 0.572
8300 gns. 80%, amph. 8% 1.179 0.401 0.367

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
88 A. Rauen, H. C. SoVel and H. W inter

Figure 3. Depth logs of measured volume susceptibility (same data as in Fig. 1), paramagnetic volume susceptibility as calculated from eq. (2)
(open squares) and derived from hysteresis data (stars; Table 3) and ferrimagnetic volume susceptibility as calculated from the difference between
measured and paramagnetic susceptibilities (eq. 3). Right: simplified lithology (for explanation see Fig. 1). Note the different susceptibility scales.

Fig. 3 presents the results of this separation of the para- versus volume susceptibility k of all available data (4243 values).
magnetic and ferrimagnetic contributions to the magnetic These data are separated into paramagnetic (Fig. 4b) and
susceptibility. The paramagnetic susceptibility reaches a maxi- ferrimagnetic (Fig. 4c) contributions of the susceptibility using
mum level of 1×10−3 SI and is clearly influenced by the eqs (2) and (3). The two lines in Fig. 4(b) represent the upper
lithology. High paramagnetic susceptibilities correspond to and lower limits of the paramagnetic susceptibility as a function
regions where metabasic rocks prevail, whereas low para- of density r according to the formulae
magnetic susceptibilities occur in gneissic sections. In the low-
susceptibility parts, all of the measured susceptibility can k =4p×10−6
upper
be explained exclusively by the paramagnetic contribution.
+[(r−2.55×103 kg m−3)×2×10−4]/0.75 , (4)
Susceptibility anomalies and increased values >1×10−3 SI
are caused by the ferrimagnetic portion in general.
k =4p×10−6
lower
+[(r−2.67×103 kg m−3)×6×10−5]/0.63 , (5)
5 B IVA RI AT E S TATI S TI C S : CR O SS - P LO T S
Bivariate distributions (cross-plots) of density and susceptibility where k is the volume susceptibility in SI units and r is the
are useful in analysing relationships between petrophysical and density in 103 kg m−3. Eqs (4) and (5) describing the boundary
lithological properties. Fig. 4(a) shows a plot of the density lines (Fig. 4b) have been calculated from the following equations,

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
Magnetic susceptibility of drill cuttings 89

Figure 4. Cross-plots of volume susceptibility as a function of density (4243 cases). (a) Measured susceptibility versus density (same data as in
Fig. 1). (b) Paramagnetic susceptibility (calculated with eq. 2) versus density. Lines of upper and lower paramagnetic susceptibility limits (calculated
with eqs 4 and 5) have been added. (c) Ferrimagnetic susceptibility (calculated with eq. 3) versus density.

originally given by Henkel (1994) in cgs units: data available for these correlations is less because of the
increased sampling interval (up to 20 m) for the ore microscopy
upper curve:
data (446 samples). Nevertheless, some interesting trends are
r=(1−v)×2.55 g cm−3+v×3.3 g cm−3 , (6a) apparent. All samples containing pyrrhotite show susceptibilities
not exceeding 6×10−3 SI, whereas magnetite-bearing samples
k=10−6+v×2×10−4 ; (6b)
reach the highest measured values, of up to 66.5×10−3 SI.
lower curve: The number of pyrrhotite samples exceeds that of magnetite
samples, although magnetite is responsible for the highest
r=(1−v)×2.67 g cm−3+v×3.3 g cm−3 , (7a)
susceptibility values.
k=10−6+v×6×10−5 , (7b)
where v is the volume fraction of the paramagnetic minerals, 6 M ULTI VA R I AT E S TATI S TIC S: FA C T O R
r is the density and k is the volume susceptibility (all parameters A N A LY S I S
in cgs units). Henkel (1994) made the following assumptions:
Correlations between susceptibility, different lithologies, ore
(1) the rocks are composed of non-magnetic and paramagnetic minerals and rock alteration indicators are investigated by
minerals (volume fraction v); multivariate factor analysis of the data. In general, factor
(2) the densities of non-magnetic minerals vary from 2.55 analysis attempts to identify underlying factors that explain
to 2.67 g cm−3; the pattern of correlations within a data set of observed
(3) the density of paramagnetic minerals is 3.3 g cm−3 variables (see e.g. Backhaus et al. 1996 or Cooley & Lohnes
(1 g cm−3=103 kg m−3); 1971). A set of data contains variables that are, in general, not
(4) the susceptibilities of paramagnetic minerals vary from independent and may correlate with other variables. Each of
6×10−5 to 2×10−4 cgs (7.5×10−4−2.5×10−3 SI). the extracted factors accounts for a certain portion of the
original variance and is not correlated with the other factors.
Henkel (1994) derived these values from metasedimentary
Thus, factor analysis reduces a set of input data containing i
rocks, amphibolites and eclogites containing biotite, amphibole,
variables in j cases to a set of f factors in the same j cases,
pyroxene and garnet as the main paramagnetic minerals.
where f <i. In our analysis the j cases correspond to j
The most frequent paramagnetic minerals from X-ray
depth values.
diffraction (XRD) analyses of KTB samples are amphibole,
The fundamental theorem of factor analysis is given by the
biotite, garnet, clinopyroxene, orthopyroxene and olivine,
matrix equation
mainly in amphibolites and metagabbros. The calculated para-
magnetic susceptibilities using KTB data fit nearly completely Z=AP , (8)
in the range of naturally occurring paramagnetic susceptibilities
written in coefficients
originally given by Henkel (1994). No values occur below the
lower line in Fig. 4( b), which confirms the natural boundary z = ∑ a p . (8a)
ij ix xj
character of this lower susceptibility line (Henkel 1994). (x=1 to f)
Susceptibility anomalies are explained solely by the ferri- Z is the matrix with coefficients z containing standardized
ij
magnetic contribution. The ferrimagnetic data of the KTB original values x of i variables in j cases:
ij
range from low values typical of paramagnetic rocks to very
z =(x −X )/s , (9)
high values, which contribute to susceptibility anomalies (Fig. 4c). ij ij i i
A discrimination was also possible for pyrrhotite- and where z is the standardized value of variable i at depth j, x
ij ij
magnetite-bearing rocks, as shown in Fig. 5. The amount of is the original value of variable i at depth j, X is the mean
i
© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
90 A. Rauen, H. C. SoVel and H. W inter

Figure 5. Cross-plots of the ferrimagnetic portion of susceptibility (eqs 2 and 3; see Fig. 4c) as functions of density. (a) All available data
(446 cases). (b) All samples containing only magnetite (i.e. pyrrhotite=0 and magnetite>0; 21 cases). (c) All samples containing only pyrrhotite
(i.e. pyrrhotite>0 and magnetite=0; 317 cases).

value (for all j) of variable i and s is the standard deviation the respective factors in Fig. 7. Fig. 7 also displays the linear
i
of variable i. The mean of the standardized values is zero. correlation coefficients between one variable and the respective
Each case is a set of variables measured at the same depth. A factor, which equals the respective factor loading.
is the matrix with coefficients a containing factor loadings of Factor 1 explains 27.21 per cent of the variance of the
if
i variables and f factors. The factor loadings range from −1 data set and is therefore the most significant factor. The
to 1. Factor loadings near −1 or +1 mean a high loading of contents of metabasites and the density have factor loadings
the respective variable. P is the matrix with coefficients p of the of 0.92 and 0.90, respectively. This is equal to a correlation
fj
factor values (also called factor scores) containing f factors in coefficient of 0.92 between factor 1 and metabasite content
j cases (depth points). The arithmetic mean of all j cases of and of 0.90 between factor 1 and density (Fig. 7). The gneiss
each factor p is zero and the standard deviation is 1. content anticorrelates strongly with a factor loading of −0.91.
fj
The calculation steps of factor analysis are standardization The logarithm of the paramagnetic susceptibility also correlates
of the given data, calculation of the correlation matrix, esti- with factor 1 with a loading of 0.83. As the rock-forming
mation of the explained part of the variance of each variable minerals create the paramagnetic susceptibility, this property
(communalities) and extraction of factors. We used principal is closely connected to the lithology and contributes to the
axis factoring and the so-called Kaiser criterion for determining factor 1 too. The correlation with lithology is obvious when
the number of factors. For an easier interpretation, the initial comparing factor 1 as a function of depth with the lithological
factor loading matrix A is rotated into a suitable coordinate column of Fig. 6. The plots of factor 1 as a function of
system to obtain maximum loadings of as few variables as metabasites, density, gneisses and logarithm of paramagnetic
possible. We applied the so-called Varimax rotation, which is susceptibility in Fig. 7 therefore display two clusters corre-
an orthogonal rotation of the factor loading matrix A (e.g. sponding to the two main lithological units. In general, factor 1
Backhaus et al. 1996). is loaded by variables that represent the lithology, so factor 1
Huenges (1997) applied factor analysis to selected petro- can be called the lithology factor.
physical (not magnetic) and geochemical data from measure- Factor 2 contains variables controlled by the content of
ments on cores of the 4-km-deep KTB pilot drill hole. Our ferrimagnetic minerals and explains 19.92 per cent of the total
analysis is based on the abundant data from measurements on variance. The most important variables are the measured
drill cutting samples from the 9.1-km-deep KTB main drill susceptibility (loading 0.93), the extracted ferrimagnetic suscep-
hole. This data set consists of 13 petrophysical, geochemical, tibility ( loading 0.92) and the magnetite content ( loading 0.81).
ore petrological and lithological variables (i=13) measured at The most pronounced anomaly in the depth range 7300–8000 m
444 depth points ( j=444, see Table 4). We used the total correlates with increased magnetite contents in that depth
depth range of the KTB main hole (0–9101 m) for a common range (see Fig. 1). Factors 1 and 2 together explain 47.13 per
analysis. Several distinct depth intervals were investigated with cent of the observed data.
respect to electrical properties by Winter et al. (1998). Factors 3, 4 and 5 contain only one variable each with
For the calculations we used the  software package loadings >0.5 or <−0.5. Factor 3 includes only the content of
( for Windows, version 6.1.3 and above). Factor analysis cataclastic rocks with a loading of 0.93 and explains 8.05 per
of the data set described extracts five factors, explaining a total cent of the total variance. The content of hornblende gneiss
of 66.23 per cent of the original variance. The resulting factor only loads factor 4 ( loading 0.88, variance 6.64 per cent) and
values p (containing f =5 factors for j=444 cases or depth the pyrrhotite content only loads factor 5 ( loading 0.51) and
fj
points) are shown in Fig. 6 as depth plots of the five factors. explains the smallest part (4.41 per cent) of the total variance.
Variables with high factor loadings (>0.5 or <−0.5) are These exclusive loadings also show the uniqueness of cataclastic
shown in Fig. 6. The high loading variables are plotted versus rocks, hornblende gneiss and pyrrhotite. Each of these variables

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
Magnetic susceptibility of drill cuttings 91

Table 4. Input data set for the factor analysis. The ore mineral contents (pyrrhotite, magnetite, pyrite,
ilmenite) were determined by optical analysis and measured in an ordinal scale in five classes:
0 (0 volume per cent=‘none’), 1 (%1 volume per cent), 2 (<1 volume per cent), 3 (1–2 volume per
cent) and 4 (>2 volume per cent=‘much’). The amount of cataclastic rocks is the sum of cataclasites
and cataclastically deformed rocks (cataclastic gneisses and cataclastic metabasites). The alteration
parameter is defined as chlorite/(chlorite+biotite+hornblende+garnet+muscovite). The data
set contains 444 cases, i.e. depth points.

Variable Range of values Unit Sampling interval

Petrophysics
log. of measured susceptibility in 10−3 SI −1 to 1.82 2m
log. of paramagnetic susceptibility in 10−3 SI −0.9 to −0.02 2m
log. of ferrimagnetic susceptibility in 10−3 SI −1 to 1.82 2m
density 2.46 to 3.05 103 kg m−3 2m
Ore Petrology
pyrrhotite 0–4 20 m
magnetite 0–4 20 m
pyrite 0–4 20 m
ilmenite 0–4 20 m
Lithology
gneiss (all gneisses except hornblende-gneisses) 0–100 vol.% 2m
metabasite (amphibolites and metagabbros) 0–100 vol.% 2m
hornblende-gneiss 0–97 vol.% 2m
cataclastic rocks 0–100 vol.% 2m
Geochemistry
alteration parameter 0–1 2m

is necessary to explain the total variance of the data set. This (1) Two ferrimagnetic ore minerals (magnetite and mono-
result documents also that pyrrhotite is abundant in all rock clinic pyrrhotite) and some strongly paramagnetic minerals
types. Factor 5 describing mainly pyrrhotite tends to have ( hornblende, biotite) contribute to the magnetic susceptibility.
increased values with depth. This is also seen in the depth plot The theoretically expected paramagnetic susceptibility, calcu-
of pyrrhotite (Fig. 1). Although magnetite and pyrrhotite lated from iron and manganese contents and the density,
are the most important ferrimagnetic minerals, they are not correlates clearly with lithology. High values of the measured
represented by a common factor. This is documented by their volume susceptibility k are typical of metabasic rocks such as
low linear correlation coefficient of 0.32. metagabbros and amphibolites due to their contents of para-
There are some distinct zones with increased contents of magnetic minerals such as hornblende, while much lower values
cataclastic components in the drill cuttings, especially at 6900 have been found in gneissic rocks. The characteristic values for
and 7200 m, where the drill hole intersected prominent cata- metabasites and gneisses are derived from the modes of the
clastic fault zones (Fig. 1). Factor 3 represents these cataclastic respective histograms as k=0.847×10−3 SI (metabasites) and
fault zones well, but no correlation with magnetic properties k=0.266×10–3 SI (gneisses).
was found, as suggested by Bosum et al. (1997). The most (2) Pyrrhotite is abundant in all rock types of the KTB drill
probable explanation for the lack of correlation is that no holes. Magnetite is concentrated only in distinct regions in limited
cataclasis is observed in the depth interval 7300 to 7800 m, depth intervals. Susceptibility anomalies up to k=6×10–3 SI
are caused by pyrrhotite, whereas the highest measured values
where the highest measured susceptibilities occurred. Future
of up to k=67×10−3 SI are caused by magnetite.
investigations of several distinct depth intervals are planned
(3) Factor analysis applied to KTB drill cutting data in
to verify these hypotheses.
the total depth interval 0–9101 m reduces 13 input variables
All five factors together explain most (66.23 per cent) of the
to five independent factors. In order of their significance, these
total variance of the input data set.
five factors are controlled by (1) lithology and paramagnetic
susceptibility, (2) ferrimagnetic susceptibility and magnetite
content, (3) the amount of cataclastic rocks, (4) the amount of
7 CO N CLU S IO N S hornblende gneiss and (5) the pyrrhotite content. 66 per cent
of the variance of the data set used is explained by these five
Drill cuttings are often regarded as low-quality drill samples due factors. The reported sequence of factors also demonstrates
to possible mixing during sample recovery and due to reduced the sequence of importance of geological and petrophysical
depth information. In spite of this apparent disadvantage, properties for the total variance of the observed data. The type
important results were obtained using data from drill cuttings. of lithology controls the greatest proportion of data variance
Histograms, cross-plots and factor analysis lead to the following (27 per cent). The rock-forming minerals also create the para-
conclusions. magnetic susceptibility, which is therefore closely connected to

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
92 A. Rauen, H. C. SoVel and H. W inter

Figure 6. Depth logs of factor values (=factor scores) p with factor number f ( f =1–5) and case number j ( j=1–444). This is the result of the
fj
factor analysis (eqs 8 and 9) of the input data set (Table 4). Below the depth logs, the percentage factor variances and the sum of all factor
variances are given. The factor loadings of all variables with loadings >0.5 or <−0.5 are also given. The maximum theoretical loading is ±1.
Left: simplified lithology (see Fig. 1).

the lithology. The next significant properties are ferrimagnetic without any participation of magnetic properties or magnetic
ores, most importantly the magnetite content (20 per cent). minerals. This is also seen in 2-D linear regression: the corre-
These two factors explain in total 47 per cent of the total lations between the susceptibility and the amount of cataclastic
variance. Of less importance are the amount of cataclastic rocks and the alteration index are very poor (Fig. 8, correlation
rocks (8 per cent), the amount of hornblende gneiss (7 per cent) coefficients C=0.17 and C=−0.15). These results are derived
and the amount of pyrrhotite (4 per cent). Magnetite and from statistical analyses of the total depth interval. Here, the
pyrrhotite are loaded by different factors. This demonstrates increased magnetite concentrations in the amphibolite section
the different characters of these two ferrimagnetic minerals. between 7300 and 7800 m, where almost no cataclasis was
The amount of cataclastic rocks, hornblende gneiss and observed, dominate the variance of the complete depth section.
pyrrhotite control one factor each. This shows that these Future detailed investigations of distinct depth intervals are
three properties are necessary to explain the variance of the planned in order to study the influence of cataclasis zones on
data set. magnetic properties.
(4) Bosum et al. (1997) suggested a correlation between (5) Drilling through cataclastic fault zones in general results
the observed anomalies in magnetic borehole log data and the in a lack of core material due to poor recovery of compact
occurrence of cataclastic zones and faults in the deeper part of cores in these mechanically unstable zones (Berckhemer et al.
the KTB borehole. This suggestion is not supported by the 1997). Nevertheless, in such difficult zones drill cuttings are
statistical investigations on drill cutting data presented here. sampled. Therefore, the analysis of drill cutting data has the
The amount of cataclastic rocks loads factor 3 exclusively, advantage that the complete lithology is investigated.

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
Magnetic susceptibility of drill cuttings 93

Figure 7. Cross-plots of factors 1–5 versus all variables with loadings >0.5 or <−0.5. Linear regression lines and correlation coefficients C are
also given.

Figure 8. Cross-plots of (a) the amount of cataclastic rocks (3965 values) and (b) the alteration index (4180 values) versus the logarithm of
susceptibility. Cataclastic rocks and alteration index are defined in Table 4. Linear regression lines and correlation coefficients C are given.

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018
94 A. Rauen, H. C. SoVel and H. W inter

Henkel, H., 1994. Standard diagrams of magnetic properties and


AC KN O WL ED GM E NTS density—a tool for understanding magnetic petrology, J. appl.
These investigations were supported by the Deutsche Geophys., 32, 43–53.
Forschungsgemeinschaft (DFG, reference number So72/59 Huenges, E., 1997. Factors controlling the variances of seismic velocity,
density, thermal conductivity and heat production of cores from the
and 60). We thank H. Küchenhoff for advice in statistical
KTB pilot hole, Geophys. Res. L ett., 24, 341–344.
details and J. Pohl for making hysteresis data available. We Huenges, E., Lauterjung, J., Bücker, Ch., Lippmann, E. & Kern, H.,
also thank F. Heider and the two reviewers, D. Eberle and 1997. Seismic velocity, density, thermal conductivity and heat
E. Huenges, for substantial improvements to the manuscript. production of cores from the KTB pilot hole, Geophys. Res. L ett.,
24, 345–348.
Kontny, A., Friedrich, G., Behr, H.J., de Wall, H., Horn, E.E., Möller, P.
& Zulauf, G., 1997. Ore mineral forming processes in the meta-
morphic rocks of the German continental deep drilling site (KTB),
RE FE R ENC ES J. geophys. Res., 102 (B8), 18 323–18 336.
Latham, A.G., Harding, K.L., Lapointe, P., Morris, W.A. & Balch, S.J.,
Backhaus, K., Erichson, B., Plinke, W. & Weiber, R., 1996. Multivariate 1989. On the lognormal distribution of oxides in igneous rocks, using
Analysemethoden, Springer, Berlin. magnetic susceptibility as a proxy for oxide mineral concentration,
Berckhemer, H. et al., 1997. Petrophysical properties of the Geophys. J., 96, 179–184.
9-km-deep crustal section at KTB, J. geophys. Res., 102 (B8), Leonardi, S. & Kümpel, H.J., 1998. Variability of geophysical log data
18 337–18 361. and the signature of crustal heterogeneities at the KTB, Geophys.
Bleil, U. & Petersen, N., 1982. Magnetic properties, in Landolt- J. Int., 135, 964–974.
Börnstein: Numerical Data and Functional Relationships in Science Nagata, T., 1961. Rock Magnetism, Maruzen Company Ltd., Tokyo.
and T echnology, New Series, Group V, Vol. 1b, Physical Properties Pechnig, R., Haverkamp, S., Wohlenberg, J., Zimmermann, G. &
of Rocks, pp. 308–365, ed. Hellwege, K.H., Springer, Heidelberg. Burkhardt, H., 1997. Integrated log interpretation in the German
Bosum, W., Casten, U., Fieberg, F.C., Heyde, I. & Soffel, H.C., 1997. Continental Deep Drilling Program: lithology, porosity and fracture
Three-dimensional interpretation of the KTB gravity and magnetic zones, J. geophys. Res., 102 (B8), 18 362–18 390.
anomalies, J. geophys. Res., 102 (B8), 18 307–18 321. Pucher, R., 1994. Pyrrhotite-induced aeromagnetic anomalies in
Bram, K. & Draxler, J.K., eds., 1994. KT B-Report 94–1, Basic western Germany, J. appl. Geophys., 32, 33–42.
Research and Borehole geophysics (Final Report), Niedersächsisches Schwarz, E.J. & Broome, J., 1994. Magnetic anomalies due to
Landesamt für Bodenforschung, Hannover. pyrrhotite in Paleozoic metasediments in Nova Scotia, eastern
Clark, D.A. & Tonkin, C., 1994. Magnetic anomalies due to pyrrhotite: Canada, J. appl. Geophys., 32, 1–10.
examples from the Cobar area, N.S.W., Australia, J. appl. Geophys., Soffel, H., Plaumann, S., Pucher, R., Bücker, C., Götze, H.J.,
32, 11–32. Wagener, M. & Haak, V., 1989. Gravity and magnetic investigations
Cooley, W.W. & Lohnes, P.R., 1971. Multivariate Data Analysis, John at the KTB locations Schwarzwald and Oberpfalz, in T he German
Wiley, New York. Continental Deep Drilling Program (KT B), pp. 409–431, eds
Dietrich, H.-G., Lauterjung, J. & Wöhrl, T., 1992. Introduction, in Emmermann, R. & Wohlenberg, J., Springer, Heidelberg.
KT B-Report 92–2, KT B Hauptbohrung, Results of Geoscientific Soffel, H.C., 1991. Paläomagnetismus und Archäomagnetismus, Springer,
Investigation in the KT B Field L aboratory, 0–6000 m, pp. A1–A26, Heidelberg.
eds Emmermann, R., Dietrich, H.-G. Lauterjung, J. & Wöhrl, T., Soffel, H.C. et al., 1992. Physical parameters measured on cores
Project Management of the Continental Deep Drilling Programme and cuttings from the pilot well (0 m–4000.1 m) of the German
of the Federal Republic of Germany in the Geological Survey of Continental Deep Drilling Program (KTB) in the Oberpfalz area,
Lower Saxonia, Hannover. Bavaria, Federal Republic of Germany, Surv. Geophys., 13, 1–34.
Dunlop, D.J. & Özdemir, Ö., 1997. Rock Magnetism, Fundamentals Thompson, R. & Oldfield, F., 1986. Environmental Magnetism, Allen
and Frontiers, Cambridge University Press, Cambridge. & Unwin, London.
ELEKTB Group, 1997. KTB and the electrical conductivity of the Umsonst, T., Lauterjung, J., Tran Viet, T. & Wöhrl, T., 1995. Technical
crust, J. geophys. Res., 102 (B8), 18 289–18 305. aspects and sampling procedures at the KTB -Hauptbohrung, in
Emmermann, R. & Lauterjung, J., 1997. The German Continental KT B-Report 95–2, KT B Hauptbohrung. Results of Geoscientific
Deep Drilling Program KTB: overview and major results, J. geophys. Investigation in the KT B Field L aboratory. Final Report: 0–9101 m,
Res., 102 (B8), 18 179–18 202. pp. A1–A11, eds Emmermann, R., Althaus, E., Giese, P. &
Emmermann, R., Althaus, E., Giese, P. & Stöckhert, B., eds, 1995. Stöckhert, B., Schweizerbart’sche Verlagsbuchhandlung, Stuttgart.
KT B-Report 95–2, KT B Hauptbohrung. Results of geoscientific Wedepohl, K.H., ed., 1969. Handbook of Geochemistry, Springer,
investigation in the KT B field laboratory, Final report: 0–9101 m, Heidelberg.
Schweizerbart’sche Verlagsbuchhandlung, Stuttgart. Winter, H., Rauen, A. & Soffel, H., 1998. Die Ursachen elektrischer
Friedrich, D., 1995. Gefügeuntersuchungen an Amphiboliten der Leitfähigkeitsanomalien in der KTB, eine Analyse mit multivariater
Böhmischen Masse unter besonderer Berücksichtigung der Anisotropie Statistik, in Protokoll über das Kolloquium Elektromagnetische
der magnetischen Suszeptibilität, Geotektonische Forschungen, 82, T iefenforschung, pp. 12–20, eds Junge, A. & Bahr, K., Deutsche
1–118, Stuttgart, Germany. Geophysikalische Gesellschaft, Frankfurt.

© 2000 RAS, GJI 142, 83–94

Downloaded from https://academic.oup.com/gji/article-abstract/142/1/83/593835


by guest
on 09 April 2018

Anda mungkin juga menyukai