Journal of Applied Geophysics 68 (2009) 423436

Contents lists available at ScienceDirect

Journal of Applied Geophysics

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a p p g e o

Induced polarization with in-loop transient electromagnetic soundings: A case study

of mineral discrimination at El Arco porphyry copper, Mexico
Carlos Flores , Sergio A. Peralta-Ortega 1
Departamento de Geofsica Aplicada. Centro de Investigacin Cientca y Educacin Superior de Ensenada (CICESE), km. 107 Carretera Tijuana-Ensenada, Ensenada, Baja California, Mxico

a r t i c l e

i n f o

Article history:
Received 27 May 2008
Accepted 25 March 2009
Keywords:
Induced polarization
Transient electromagnetic soundings
Mineral discrimination
Porphyry copper

a b s t r a c t
With the inversion of in-loop transient electromagnetic (TEM) soundings over the El Arco porphyry copper
deposit in terms of one-dimensional ColeCole dispersive models we found that the relationship between
the chargeabilities and the time constants shows a fair agreement with the actual mineral concentrations
measured in drillhole cores. This is the rst reported case of mineral discrimination with this technique.
Through a sensitivity analysis we show that these two parameters are the least well resolved.
2009 Elsevier B.V. All rights reserved.

1. Introduction
The induced polarization (IP) method has been shown to be an
effective tool in mineral exploration (Fink et al., 1990). In most IP
studies the data are acquired with ground electrodes, most commonly
with the dipoledipole conguration. A current is introduced into the
ground through two current electrodes, and a potential difference is
measured between two potential electrodes located at some distance
from the current electrodes. The polarization of the subsurface can be
detected in two domains. In the time domain, the decaying voltage is
measured at several times after a dc current is interrupted. In the
frequency domain, the voltage and its phase shift are measured for
different frequencies of a sinusoidal applied current.
An alternative to the current-injection or galvanic IP approach is
the use of a purely inductive technique, such as the transient electromagnetic (TEM) sounding in the in-loop or coincident-loop congurations (Spies and Frischknecht, 1991; Nabighian and Macnae, 1991). The
presence of polarizable material usually manifests in these data as a
change of sign in the measured voltages followed by the decay of
the negative voltages. Gubatyenko and Tikshayev (1979) and Weidelt
(1982) theoretically showed that for the coincident-loop conguration
these sign reversals cannot occur in a frequency-independent linear
medium, attributing them to polarization effects. In the in-loop array
a similar effect is expected when the ground is quasi-layered. In the
TEM in-loop conguration, a sounding is carried out with a large rec-

Corresponding author. Tel.: +646 1750500x26326; fax: +646 1750567.

E-mail addresses: cores@cicese.mx (C. Flores),
sortega2@villahermosa.oileld.slb.com (S.A. Peralta-Ortega).
1
Now at: Schlumberger Oileld Services. Km. 7 Carretera Villahermosa-Crdenas,
Ranchera Lzaro Crdenas, Villahermosa, Tabasco, Mxico.
0926-9851/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jappgeo.2009.03.009

tangular or square transmitting loop and a horizontal coil located at the

center of the loop. The injected dc current in the loop is periodically
interrupted in the form of a linear ramp. An induced current system,
owing in closed paths below the loop, is created each time the transmitter current is interrupted, producing a secondary magnetic eld. The
time variation of the vertical component of this magnetic eld induces
a voltage at the receiver coil. In the presence of polarizable rocks a
polarization current with opposite direction to that of the fundamental
induction current is also created (Smith and West, 1988). These current
systems usually decay at different rates. The decay rate of the polarization current is usually slower, such that in late-times the negative
polarization current may be greater than the positive fundamental
induced current, producing the sign reversal in the measured voltages.
As the spatial and temporal distribution of the subsurface current system depends upon the ground resistivity, the measured transient voltage
gives information about the subsurface resistivity. The locus of the
maximum amplitude of the induced currents diffuses downward and
outward with time, thereby giving information about deeper regions as
time increases (Nabighian, 1979; Hoversten and Morrison, 1982).
Pelton et al. (1978) and Vanhala and Peltoniemi (1992) carried
out galvanic spectral IP measurements over many mineralized outcrops with known mineral concentrations and textures. By inverting
their measurements to homogeneous ColeCole dispersive models,
they found that the relationship between two ColeCole parameters
(chargeability and time constant) gives information on the mineral
concentration and textures, an ability known as mineral discrimination.
Despite the fact that in-loop TEM soundings are also sensitive to
polarizable material, still there has not been any published eld case
showing that it is possible to do mineral discrimination with the TEM
method. In this paper we present a eld case of mineral discrimination
where the inversion of in-loop TEM responses to 1D dispersive models

424

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

over a porphyry copper deposit shows a reasonable agreement with the

drilling information on sulde concentrations (Peralta-Ortega, 2001).
1.1. The IP effect on TEM soundings with a ColeCole model
The dispersive character of the IP phenomena is usually described
with the ColeCole model (Cole and Cole, 1941; Pelton et al., 1978; Lee,
1981; Vanhala and Peltoniemi, 1992; Ghorbani et al., 2007)



= 0 1 m 1

1
1 + ic


1

where 0 is the zero frequency or direct current resistivity (m), m is

the chargeability (dimensionless), is the time constant (seconds),
and c is the frequency exponent (dimensionless). The four parameters
have different ranges of variation. The chargeability and the frequency
exponent can vary from 0 to 1. The dc resistivity and the time constant have wide ranges, from 10 2 to 105 m and from 10 5 to 104 s,
respectively.
Fig. 1 shows the amplitude and phase spectra of the ColeCole
model. Each graph displays the behavior of the dispersive resistivity
(expression (1)) when only one parameter of the reference model
[100, 0.5, 0.01, 0.5] is varied, keeping the other three xed. We will
be using the notation [0,m,,c] for dening the four ColeCole
parameters of a homogeneous medium. Solid lines are used for
the amplitudes and dashed lines for the phases, which are all negative.
The effects of varying the parameters are different. Varying the dc
resistivity affects the amplitudes but not the phases (Fig. 1a). The
variation of the chargeability from 0.1 to 0.9 has a profound effect; the
amplitudes show a sharp decrease for high frequencies and the phases
increase (Fig. 1b). Increasing the time constant from 10 4 to 1 shifts
the amplitudes and phases toward low frequencies (Fig. 1c), while

increasing the frequency exponent sharpens the amplitude and phase

spectra (Fig. 1d).
The effect of varying the ColeCole parameters on the in-loop
voltages of a TEM sounding is shown in Fig. 2. The transmitter and
receiver parameters used in this numerical simulation are representative of those employed in El Arco eld survey. The four drawings
of Fig. 2 result from varying one parameter at a time of the same
reference model [100,0.5,0.01,0.5] of Fig. 1. The most diagnostic feature of the IP cases is the sign reversal, the peaking of the negative
voltages (dashed lines), followed by a decay. We include the noise level
found in El Arco, which is about 1 10 8 V. Increasing the dc resistivity
to 1000 m (Fig. 2a) shifts the sign reversal to earlier times and
increases the negative peak, while decreasing it to 10 m causes the
opposite effect. For comparison, Fig. 2a also shows the voltages of a
non-polarizable homogeneous medium of 100 m, characterized by
all-positive voltages. Varying the chargeability (Fig. 2b) has a similar
effect to the variation of the dc resistivity. Notice, however, that the
sign reversal for m = 0.1 occurs well below the noise level, such that for
this case the IP effect might not be recognized by the interpreter. The
increase of the time constant (Fig. 2c) and the frequency exponent
(Fig. 2d) also produce shifts of the sign reversal toward earlier times.
The case for = 10 4 s does not produce the sign reversal within the
calculated time window, and for c = 0.25 the sign reversal is slightly
below the noise level. These are examples of the difculty of recognizing the IP effect for several combinations of parameters.
The importance of the ColeCole model stems on its ability to do
mineral discrimination, i.e., to perform semi quantitative inferences on
the concentration and size of the metallic minerals in a mineral deposit,
which is a very useful interpretation tool in mining prospecting. Pelton
et al. (1978), in a cornerstone geophysical work, carried out variablefrequency voltage measurements with the dipoledipole array over
several mineral outcrops in the United States and Canada. By inverting

Fig. 1. Amplitude (solid lines) and phase (dashed lines) spectra of the ColeCole relaxation model. a) Varying the dc resistivity, b) the chargeability, c) the time constant, and d) the
frequency exponent of the reference model [0 = 100, m = 0.5, = 0.01, c = 0.5].

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

425

Fig. 2. Transient in-loop voltages of polarizable homogeneous media. Each drawing corresponds to varying one ColeCole parameter of the reference model [100, 0.5, 0.01, 0.5], while
keeping the other three xed. The response of a non-polarizable 100 m medium is included in a).

the phase spectra to ColeCole models they studied the correlation

between the estimated parameters and the known concentrations and
mineral textures. Fig. 3 summarizes their results associated with the
ability of the IP spectral method to do mineral discrimination, where
their estimated values in 47 outcrops of 9 different porphyry copper
deposits are plotted in a graph of chargeability against the logarithm of
the time constant. Sites with high sulde concentrations are associated
with high chargeabilities and time constants (Fig. 3a).
The relationship of mineral grain texture with chargeability and
time constant is shown in Fig. 3b. Texture was divided by these
authors into two types: veinlet and disseminated. Veinlet is the
occurrence of minerals in small veins, where the connection between
the different sulde particles is high, while disseminated implies the
presence of isolated mineral grains. Sites with disseminated texture
(solid symbols) tend to have low chargeability and time constant
values. On the contrary, veinlets (open symbols) tend to present higher
values. It is important to note that these associations are not univocal,
i.e., a given concentration and texture does not correspond to a single
point in the m versus graphs.
2. Site description
2.1. Geology and geophysics of the El Arco deposit
The El Arco deposit is located in the central part of the Baja
California peninsula, in the western part of the Peninsular Batholith,
near the village of El Arco. In 1968 Industrial Minera Mexico started
geologic, geochemical and geophysical exploration campaigns of
the area, followed by an intensive drilling program (250 diamond
drill holes averaging 300 m in depth), leading to the discovery of the
porphyry deposit containing at least 600 million tons with 0. 6%
copper and 0. 2 g per ton of gold (Coolbaugh et al., 1995).
The regional and local geology have been described by several workers
(Echavarri-Prez and Rangin, 1978; Barthelmy, 1979; Coolbaugh et al.,
1995). The local geology is depicted in Fig. 4. The deposit is associated with
a mid-Jurassic (Valencia et al., 2006; Weber and Lpez-Martnez, 2006)

Fig. 3. Chargeability versus time constant plot estimated with galvanic spectral IP
measurements on 47 outcrops of 9 porphyry copper deposits of North America,
a) Association with sulde volume concentrations, b) Association with mineral grain
textures (after Pelton et al., 1978). Panel b has the same axes labels as a.

426

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

Fig. 4. a) Surface geology of the El Arco deposit. The layer of conglomerates covering the western portion of the area has been removed (after Echavarri-Prez and Rangin, 1978),
b) Geologic section from drillholes along the AA' prole (after Coolbaugh et al., 1995). The inset shows the El Arco location.

granodiorite porphyry stock intruding metavolcanic rocks (Fig. 4a).

The horizontal dimensions of the mineralized body are approximately
1.5 by 1.5 km, with a maximum depth not exceeding 300 m. The deposit
can be divided into oxide, transitional, and sulde zones (Fig. 4b). The
oxidized zone extends down to an average depth of 50 m and contains
several copper oxides, with chrysocolla being the most abundant.
Underlying and parallel to the oxidized zone is a transition zone which
varies in thickness from 0 to 18 m, with an average thickness of 9 m,
containing secondary and primary suldes and a mixture of iron and
copper oxides. The primary sulde zone underlies the transition zone and
represents the major portion of the mineralized ore body. Its top coincides
with the water table while its lower limit is not well dened. Exploration
drilling to a depth of 500 m has bottomed in plus 0. 4% copper. Pyrite
and chalcopyrite are the most abundant suldes in this zone. Lesser
amounts of bornite, magnetite, molybdenite, galena, and sphalerite are

also present. Copper mineralization occurs as fracture lling (about 60%)

and disseminated (about 40%) (Coolbaugh et al., 1995). Sixty percent of
the copper values are in the granodiorite porphyry and the remaining
40% in the andesite wallrock. The topographic relief is smooth.
Fig. 5a shows the total suldes concentration from drillhole cores,
expressed in percentage by volume. It varies from 1 to 2. 5% in the
richest part of the deposit to lower values toward the periphery. The
pyritechalcopyrite weight ratio (Fig. 5b) is less than 1 in the central
part of the deposit, increasing to about 14 toward the periphery
(Echavarri-Prez and Rangin, 1978).
Faras (1978) reported the results of IP, resistivity, gravity, and ground
magnetic studies carried out before the exploratory drilling. Non-spectral
galvanic IP measurements in the time domain were done every 100 m
along NS proles covering the area of study, where current was
injected into the ground with a three-electrode Wenner array, using a

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

427

Fig. 5. a) Total suldes content in percentage by volume (after Faras, 1978), b) pyritechalcopyrite weight ratio (after Echavarri-Prez and Rangin, 1978).

xed 100 m electrode separation. Fig. 6 shows the chargeability map

with contours every 5 ms. Although the deposit is partially covered by the
10, 15, and 20 ms contours, the maximum values of 22.5 ms do not occur
over the center of the deposit, but displaced to the southeast. This feature
was interpreted as being associated with the pyrite halo surrounding
the mineral deposit. A common problem in the electrical exploration of
mineral deposits is the usually strong response of pyritized zones, a
common mineral with no economic use. Simultaneous to the IP survey
apparent resistivity measurements were also taken. Fig. 6 shows the map
of apparent resistivities. Even though there is not a perfect agreement
between the ore boundary and the area enclosed by the 200 m contour,
the deposit is clearly associated with low apparent resistivities.
Fig. 7 shows the measured responses obtained from four geophysical methods on a SWNE line crossing the deposit (location indicated
in Fig. 6). The IP anomaly (apparent chargeability), with a maximum of
18 ms, is the one which better reects the presence of the mineralized
body. The apparent resistivity response also shows low values over the
deposit, although in the SW end of the line the increase of values is not
so clear as at the NE boundary of the deposit. The gravity low occurring
over the central part of the ore body was interpreted by Faras (1978) as
the result of a lower density of the granodiorite intrusive (2690 kg/m3)
with respect to that of the host andesite (2720 kg/m3). The total eld
magnetic anomaly also shows low values over the mineralized body,
interpreted by Faras (1978) as the result of the hydrothermal obliteration of the primary magnetite in the mineralized area and the relatively
high magnetite content in the surrounding andesites.
3. The TEM survey
A total of 14 TEM soundings were acquired with the in-loop array,
where the time derivative of the vertical magnetic eld at the center of
each loop was measured with a 100 m2 coil. Fig. 8 shows the location
of 13 soundings; sounding 14 is located 1 km southwest of the mineral
deposit. A contiguous 300 by 150 m transmitting loop was employed
in 11 soundings along Line 19, a northeastern drillhole prole crossing
the mineral deposit. Sounding 8 was acquired with a smaller loop (150
by 150 m) and is located 170 m off Line 19. The purpose of sounding
13, located south of the southwestern ore boundary, was to test the
response of the pyrite halo.

We used the Geonics TEM57 system, employing in most of the

soundings the three available repetition frequencies (30, 7.5, and 3 Hz),
and injecting currents to the loops of approximately 5.5 A. The stacking times for the three frequencies were 128, 256, and 256 s, respectively, resulting in stacked voltages representing the average of
about 7700, 3800, and 1500 individual voltage decays. At least three
realizations for each repetition frequency were carried out, from which
the nal average voltage and standard error were estimated. Figs. 9
and 10 display the voltage data as a function of time. The data quality is
generally good. All soundings show the voltage sign reversal, characteristic of the presence of IP. In early and intermediate times the voltage
standard deviations are usually very small. A 5% error was assigned to
these data to avoid overweighting them in the inversion process. In all
soundings the late-time voltages were eliminated because they atten
out, indicating that the ambient noise level (1 10 8 V) was reached.
4. Inversion approach
The in-loop voltages in each sounding were inverted to homogeneous or two-layered dispersive models using an iterative leastsquares, linearized algorithm based on the Singular Value Decomposition (SVD) of the Jacobian matrix (Jupp and Vozoff, 1975). In these
models the relationship between the N measured voltages di at times
ti with the M unknown parameters pj is a nonlinear functional y,
represented in matrix notation by
d = yp; t
This relationship is linearized by expanding the functional with a
Taylor's series and neglecting second and higher order terms, resulting
in a set of N linear equations in M unknowns
y = A p

where


yi = di y p0 ; ti
Aij =

Ayp; ti
Apj

p0

i = 1; N ; N

428

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

Fig. 8. Location of TEM soundings. The location of prole AA' of Figs. 6 and 7 is also
shown.

and
0

pj = pj pj

j = 1; N ; M

The vector y is the residual error or difference between the

measured and calculated voltage from the initial model p0, the
matrix A is the derivative matrix, also known as the sensitivity or
jacobian matrix, and p is the difference between the unknown
parameters and the parameters of the initial model. The goal is to
nd the vector of unknown parameters that minimizes yyT, the
squared sum of residual errors. The least-squares solution of expression (2) is

1
T
T
A y
p = A A
Fig. 6. Maps of apparent chargeability and apparent resistivity over the El Arco deposit
(after Faras, 1978). The location of prole AA' of Fig. 7 is indicated.

whereT stands for the matrix transpose. This solution is ill-posed

due to the presence of unimportant parameters and redundant

Fig. 7. Apparent chargeability (ms-scale), apparent resistivity (m-scale), gravity (mGal-scale) and total eld magnetics (nT-scale) responses along the AA' line of Fig. 6 (after
Faras, 1978).

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

429

Fig. 9. Observed data (symbols) with error bars. Calculated responses (solid and dashed lines) from the inverted homogeneous models.

information. To alleviate this problem a process of SVD is applied to

the sensitivity matrix,

meter matrix. With this approach, expression (3) now has the form
(Hohmann and Raiche, 1991)
0

A = USV

p = VS U y

where U is the N M eigendata matrix, S is an M M diagonal matrix

containing the ordered eigenvalues sj, and V is the M M eigenpara-

with s'j = 1 / sj. To stabilize the solution, we truncate the elements of

the matrix S associated with small singular values.

Fig. 10. Observed data (symbols) with error bars. Calculated responses (solid and dashed lines) from the inverted two-layer models.

430

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

In the optimization process we weight y and the elements of

the sensitivity matrix by the data errors and use the logarithms of the
parameters to avoid obtaining negative values. The elements of the
sensitivity matrix which contain the derivatives of each voltage data
with respect to each ColeCole parameter were calculated analytically
following a procedure similar to that proposed by Constable et al.
(1987). Since expression (3) was established by a linearization of a
nonlinear system, the solution is reached only after several iterations.
The selection of the best inverted model was based on obtaining the
minimum root mean squared (rms) mist error with the minimum
number of layers. The denition of the rms error used in this work is
v
u N

u 1 X d y 2
i
i
;
e=t
i
N i=1
where i are the standard errors of the observed data.
The solution of the forward problem required by the inversion is
calculated following the procedures outlined by Stoyer (1990), Poddar
(1982), and Newman et al. (1986). The vertical component of the
magnetic eld in the frequency domain produced by a rectangular
loop requires integrating the eld of a horizontal electric dipole along
the wire forming the loop (Poddar, 1982). The eld at the receiver coil
located at P(0,0,0) produced by one side of a rectangular loop of
dimensions 2a by 2b over a layered earth has the form,
Hz =

bI
4

Za
a

1
R

Z




0
K ; ; j ; dj J1 Rd dx

with R = [x'2 + b2]1/2. In this expression is the angular frequency, I

is the current in the loop, K is the kernel of the Hankel transform and
J1() is the Bessel function of order 1. The kernel corresponds to the
vertical component of the magnetic eld produced by a horizontal
electric dipole on a layered earth (Ward and Hohmann, 1987). It
depends on the layer thicknesses dj and on the dispersive resistivities
of each layer j(), dened by the ColeCole model. Similar
expressions hold for the remaining sides of the loop. This Hankel
transform is evaluated with a convolution with the lter proposed
by Anderson (1975). To circumvent the time-consuming integration
along the loop, each loop side is divided into N wire segments of equal
length, approximating each segment by an equivalent electric dipole
(Stoyer, 1990). We considered at least three equivalent dipoles for
each loop side. Compared to a GaussLegendre numerical integration
with 20 points per side, the voltages calculated with this approximation are better than 2%, except close to the sign reversal, where they
are better than 15%.
The transient voltage due to a current step turn-off is,
v

off

= 0 ARx

Ahz
At

where ARx is the effective area of the receiver coil and the time
derivative of the vertical magnetic eld is obtained from the Fourier
sine transform (Newman et al., 1986)
Ahz
2
=
At

ImHz  sint d

The lter weights proposed by Anderson (1979) are used to

calculate this transformation from the frequency domain to the time
domain. Before applying the Fourier transform, the effect of the
receiver coil nite bandwidth is incorporated by multiplying the
transfer function of the coil by Hz(). Finally, the effect of the actual
current waveform (linear turn-off periodic ramps) is accounted for by
using the procedure described by Fitterman and Anderson (1987).
This approach requires extrapolating the voltage response beyond the

last late-time gate. We tted an exponential function to the last ve

voltages to perform this extrapolation, calculating additional voltages
points if necessary.
5. Results
5.1. Data inversion
During the rst inversion runs we realized that the time constant
is one of the least resolved parameters. Given the importance of this
parameter for mineral discrimination and its wide range of possible
values, in a rst stage we focused our attention on a gross estimation
of this parameter. Using the same initial model, we inverted each
sounding to a homogeneous half-space dispersive model, using in each
inversion the same ColeCole parameters except the time constant,
which was varied over 5 orders of magnitude, from 10 5 to 1.
In the second inversion stage, all soundings were inverted to a
homogeneous dispersive earth, keeping all parameters free to vary.
The best model was determined with a Monte Carlo search (e.g., Jones,
1982) around the model found in the rst stage. In this approach all
model parameters were randomly perturbed by up to +/25%
around the preliminary model to dene initial guess models in new
data inversions. The inverted model replaced the previous one if it
produced a signicant decrease in the mist error. The process was
repeated about 50 times. This procedure was also used to estimate the
uncertainties in the model parameters. To dene the upper and lower
parameter uncertainties we considered their corresponding upper
and lower variation bounds in all models with mist errors falling
within 10% of the error of the best model.
Only in six soundings a homogeneous earth reproduced well the
observed data. For the remaining eight soundings we inverted the
data to two-layered earths, using again a Monte Carlo search. Figs. 9
and 10 show the calculated voltages for the homogeneous and
the two-layer inverted models, respectively, comparing them with
the measured voltages. At the top of each sounding we include the
inverted ColeCole parameters, together with the mist error (). The
matches between measured and model responses are adequate,
the rms errors vary from 1.73 to 3.85, with an average of 2.70. The
residual errors do not show any particular bias. Fig. 11 shows the
superposition of the residual errors for the models of minimum, closeto-average, and maximum mist errors. As mentioned above, these
residuals are the difference between calculated and observed voltage,
normalized by the standard deviation of the observed voltage. Except
for a couple of high values close to 12 for the model with maximum
mist, the remaining points cluster around zero, without any apparent
bias in early or late times. Most of the estimated uncertainties are not
symmetric with respect to the best ColeCole parameters. This could
be due to the limited number of inversions performed in the Monte
Carlo search, although for such non-linear problems symmetry of
standard errors is not a required result (Ghorbani et al., 2007).
From the 22 media estimates, the mean parameter values and
standard deviations are: dc resistivity 110 b 230 b 470, chargeability
0.32 +/ 0.15, time constant 2.2 10 4 b 2.4 10 3 b 2.6 10 2, frequency exponent 0.66 +/ 0.16. The standard deviations for the
dc resistivity and time constant are not symmetric because the
mean is evaluated in the logarithmic domain. Of the four ColeCole
parameters, the chargeability and time constant have the highest
percent uncertainties and the dc resistivity has the lowest. This is
relevant because m and are the two parameters used for mineral
discrimination.
5.2. Resolution analysis
In the inversion of geophysical data is not only important to nd a
model that ts the data, but also to evaluate which parameters of the
model are well resolved, which are not, and which part of the data is

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

431

eigenparameter matrix. The size of the circles is proportional to the

coefcient values according to the displayed scale, using dashed circles
for negative coefcients. The coefcients can vary from 1 to 1. Each
row corresponds to a linear combination of differentials of the logarithmic parameters arranged in order of increasing standard error
(Edwards et al., 1981). The rst row in Fig. 12b is
0:25dln 0 + 0:50dln m 0:03dln + 0:86dln c

Fig. 11. Normalized residual errors as a function of time of the inverted models of
soundings 9, 11, and 5.

contributing to their resolution. For this purpose Fig. 12a shows the
elements of the sensitivity matrix for the homogeneous model of
sounding 11. Each curve is the normalized derivative of the calculated
response with respect to the logarithmic parameter, i.e. yi / (i)ln(pj)
where i = 1,,N data and pj is one of the four ColeCole parameters.
These derivatives are a measure of how much the model response
changes for a small change in one of the parameters. Fig. 12b graphically
displays the coefcients of the 4 by 4 matrix VT, the transpose of the

Neglecting the small coefcients of 0 and , this combination is

dominated by m and c. When two coefcients of approximately the
same magnitude have the same sign, their product is the relevant
combination; if they have different signs, it is their ratio. Then, this
linear combination indicates that the product mc is the best resolved
parameter combination because its standard error is 0.06%. The fourth
row is also dominated by m and c, but here their coefcients have
different signs, indicating that the ratio m/c is the least resolved
combination. If the standard error of this row were large, a problem of
equivalence, similar to that frequently found in resistivity soundings
when there is a thin resistive layer, would be affecting the model.
Nonetheless, as the standard error of this latter combination is only 2.
7%, such equivalence problem is not present in this model. The second
and third rows are dominated by 0 and , respectively, resulting in
good resolutions for both parameters. In the six homogeneous models
0 and c are the best resolved parameters, while and m are the least
resolved, but in all of them the standard error of the least resolved
combination is less than 20%.

Fig. 12. Resolution analysis of the homogeneous model of sounding 11, a) time variation of the normalized data derivatives with respect to the four ColeCole parameters, b) elements
of the eigenparameter matrix VT. Dashed circles denote negative values.

432

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

From the analysis of the sensitivity curves we can learn what part
of the data is contributing to the resolution of each parameter. The
behavior of the curves of m, , and c (Fig. 12a) are similar in the
sense that they are close to zero over the entire time range except in
the vicinity of 3 ms, which is the time where the voltages change
sign in this model (see Fig. 9). Then, obtaining a good t in the times
located in the vicinity of the time reversal is critical, especially for m
and , the two parameters used in mineral discrimination. The
sensitivity curve for 0 also peaks at the time of sign reversal, but
differs from the behavior of the other three parameters in that it has
values different from zero in all previous times. This is the reason
why 0 is the best resolved parameter in the six homogeneous
models.
Compared to the simple situation of the homogeneous models, in
the two-layer models almost all parameters are correlated, resulting
in a signicant decrease of their resolutions. Fig. 13 shows the nine
sensitivity curves, separated into rst and second layer, and the
coefcients of the VT matrix of the model of sounding 3. Most of the
maxima of the sensitivity curves (Fig. 13a) still occur in the vicinity

of the voltage sign reversal (approximately at 2 ms; Fig. 10) but their
peak intensities are considerably less than those of the homogeneous case (Fig. 12a). Regarding the parameter correlation, let us
consider the rst row of the eigenparameter matrix (Fig. 13b),
where the linear combination of parameters is dominated by m1 and
0,2, such that the product m10,2 is the combination best resolved.
However, the same two parameters also dominate the ninth row,
but here the parameter combination is 0,2/m1 (coefcients with
different sign), corresponding to the least resolved combination
as its standard error is 1800%. Therefore, 0,2 and m1 are not individually resolved. In contrast, c1 is well resolved because this
parameter dominates the second row of the matrix. Besides 0,2/m1,
the other two worst resolved parameter combinations in this model
are d/1 and 2. Although the order of resolution is not the same in
the eight two-layer models, the best resolved parameters tend to be
c1, 0,1, c2, and d, while 0,2, m1, and 2 are among the least resolved.
Notice that one or more of the parameters used for mineral discrimination (chargeability and time constant) usually are among the
least resolved parameters.

Fig. 13. Resolution analysis of the two-layer model of sounding 3, a) time variation of the normalized data derivatives with respect to the eight ColeCole parameters and thickness d,
b) elements of the eigenparameter matrix VT. Dashed circles denote negative values.

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

5.3. Correlation with mineral concentrations and textures

The space variation of the estimated ColeCole parameters and
their uncertainties are displayed in Fig. 14a. We use different symbols
to indicate if the parameter belongs to a homogeneous model, rst
or second layer of a two-layered model and employ a line to join
parameters of the same medium. Fig. 14b shows the volume concentration prole for total suldes, interpolated from Fig. 5a. Fig. 14c
shows the section of total copper suldes and oxides (pyrite not
included) measured in core samples recovered from drillholes, ex-

433

pressing these values with contours of percent volume. This gure

also includes the thicknesses and uncertainties of the rst layer for the
two-layer models.
In the northeastern portion of the prole (Fig. 14c) ve contiguous
two-layered models (soundings 1 to 5) are very useful for the comparison with drillhole concentrations. In these models the average
rst-layer thickness is 62 +/ 19 m, a value that agrees well with the
depth of 60 m to the primary sulde zone (Coolbaugh et al., 1995).
Furthermore, the chargeability and time constants of the second layer
in these models are higher than the corresponding values of the rst

Fig. 14. a) Spatial variation of the inverted ColeCole parameters along Line 19. In the dc resistivity plot we include the VLF and Wenner apparent resistivity proles (dashed lines),
b) Total suldes volume concentration prole, c) Copper oxide and sulde concentration section (excluding pyrite). Contours every 0.2%. The thicknesses of the rst layer in the twolayered inverted models are also included.

434

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

layer, agreeing with higher mineral concentrations in the primary

sulde zone (Coolbaugh et al., 1995). This feature is also seen in the
other two-layered models (soundings 9 and 13). The exceptions to this
pattern occur in sounding 5, where 1 N 2, and in sounding 9, where
m1 N m2. Then, although the highly irregular concentration contours of
Fig. 14c indicate that the distribution of the electrical properties is very
likely three-dimensional (3-D), the general agreement of some
features of the layered models with the drillhole information suggests
that interpreting the data with 1D models might be a good approximation to the actual 3-D distribution.
In all two-layered models the zero frequency resistivity of the
second layer is higher than that of the rst layer (Fig. 14a). This result
indicates that, despite the higher mineral concentrations in the
primary sulde zone, the sulde grains in the second layer are not
sufciently well connected as to decrease its zero frequency resistivity
with respect to that of the rst layer. This increase of the zero frequency
resistivity with depth is supported by the behavior of both apparent
resistivity proles (dashed lines) included in this plot. One of them
was acquired by us with the VLF (Very Low Frequency) technique
measuring the apparent resistivity every 100 m. We used the NAA
antenna located at Cutler, Maine, USA, which transmits at the
frequency of 24.0 kHz. Considering the ColeCole dispersive resistivity
of expression (1) and the average values for the parameters, the skin
depth of the VLF method is of only 34 m. This shallow skin depth
explains why the VLF apparent resistivities are less than the zerofrequency resistivities of the homogeneous models and those of the
rst layer of the two-layered models, both derived from the TEM data.

The other apparent resistivity prole was estimated from the map
of Fig. 6. As mentioned above, this map was congured by Faras
(1978) from proles measured with a three-electrode Wenner array
with a xed 100 m electrode separation. With this electrode separation, the theoretical depth of exploration must be about 75 m, which
explains why in the central and northern portions of the section,
these apparent resistivities agree with the zero frequency resistivities
of the homogeneous models and are between the zero frequency
resistivities of the rst and second layers of the two-layered models.
Regarding the frequency exponent, the estimated values do not
show any systematic behavior in the two-layered models (Fig. 14a).
From studies in the galvanic IP frequency band, Seigel et al. (1997)
and Luo and Zhang (1998) suggested that values of about 0.1 of this
ColeCole parameter are associated with a wide distribution of the size
of the mineral particles, while high values (about 0.6) indicate a
sharply peaked distribution. Our estimated frequency exponents are
generally high (average 0.66 +/ 0.16). If results derived from the
galvanic IP frequency band can be extrapolated to the much higher
TEM frequency band, these values would suggest a relatively uniform
distribution of grain sizes.
Fig. 15 shows our estimated points plotted over the chargeability
versus logarithmic time constant graphs of Pelton et al. (1978). Fig. 15a
includes the uncertainties plotted as error bars. For clarity, the error
bars are omitted in Fig. 15b and c, where we use different symbols to
indicate if the parameters belong to a homogeneous model, or to the
rst or second layer of a two-layered model. Twenty of the twenty two
points fall within the area delimited by Pelton et al. It is worth

Fig. 15. El Arco chargeability versus time constant points superimposed on the Pelton et al. (1978) plots. a) Parameter estimates with uncertainties shown with error bars. Numbers in
brackets denote the rst or second layer of the two-layered models, b) Association with suldes volume concentration, c) Association with mineral texture. Panels b and c have
the same axes labels as a.

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

mentioning that this experimental boundary is somewhat arbitrary.

Vanhala and Peltoniemi (1992) carried out a similar study with
galvanic spectral measurements over 81 mineralized outcrops of six
different ore deposits in Finland. Their points in the m versus plot
show a general agreement with Fig. 15, but 21 points fall close but
outside the Pelton et al. boundary.
All our estimates from the contiguous loops (soundings 1 through
12) fall or are close to the region of disseminated textures (Fig. 15c).
In terms of the total suldes concentrations, there is a reasonable
agreement between the El Arco drillhole values, our estimates, and the
data by Pelton et al. (Fig. 15b). The drillhole total suldes under these
soundings vary from 0.8 to 2.3% (Fig. 14b), while our points from the
contiguous loops plot close to four of the low-concentration ranges
(.15.3,.3.7,.71.5, and 1.53%) of Pelton et al. (Fig. 15b). On the other
hand, Coolbaugh et al. (1995) estimated that 40% of the mineralization
occurs as disseminated and 60% as veinlets. As the agreement of the
concentration estimates is fair, we believe the strong discrepancy
between the textures interpreted by Coolbaugh et al. (based on visual
inspection of core samples), and our interpretation can be explained by
the use of different threshold criteria in the denition of mineral
textures.
Only two points out of the 22 estimates fall within the veinlet
texture zone (Fig. 15c), corresponding to the second layers of soundings 13 and 14. Sounding 13 is located in the pyrite halo, where the
total sulde concentration is higher than 3% (Echavarri-Prez and
Rangin, 1978). The location of the corresponding point in the
chargeability versus time constant graph of Fig. 15b agrees with the
concentrations predicted by Pelton et al., where two nearby points of
Pelton et al have concentrations that range from 3 to 7%. Although no
drillhole information exists in the neighborhood of sounding 14,
located about 1 km southwest of the ore body, its location in the
chargeability versus time constant plot suggests that the pyrite halo
might extend to this sounding.
Data obtained with galvanic methods, either in the frequency or
in the time domain, have practical limitations for the detection of the
IP source. Among them are electromagnetic coupling, telluric noise,
and incomplete frequency coverage in the time-domain voltage
decay (e.g., Bertin and Loeb, 1976; Soininen, 1984). The TEM sounding
method is not immune from their own limitations. Smith and West
(1989) proposed three favorable conditions to generate the sign
reversal in coincident-loop or in-loop TEM soundings:
a) The transmitter must couple well to the polarizable body so that
the fundamental inductive current is large, and the resulting
polarization current is strong.
b) The voltage response associated with the fundamental inductive
current must decay quickly.
c) The polarization current must couple well to the receiver to enable
its small negative response to be greater than the positive response
of the fundamental induced current.
In terms of the measured TEM response, these conditions mean
that if the sign reversal is below the ambient noise level or falls
beyond the range of recording times the presence of a frequencydependent conductivity will be hard to be noticed. In these cases
mineral discrimination will be unsuccessful because the sensitivity of
the data from the chargeability and time constant are concentrated
close to the sign reversal.
Fig. 16 illustrates some cases where the above conditions do not
hold for the homogeneous half-space representing the El Arco average
model [0 = 227 m, m = 0.321, = 2.35 10 3 s, c = 0.66], represented with a star in this gure. The time of sign reversal is plotted
against one variable parameter, keeping the other three parameters
xed. We use El Arco survey parameters (300 by 150 m loops, 5.5 A
currents, a 100 m2 receiver coil, and a recording time window from
0.086 to 70 ms) to determine these polarity change times. The size of
the loop also has an effect on the sign reversal (El-Kaliouby et al.,1997),

435

Fig. 16. Variation of the time of sign reversal as each of the four ColeCole parameters of
the model [227,0.321,2.3510 3,0.66 is varied. This model is shown with a star. Dashed
lines denote the sign reversal is below the noise level. The recording time window of the
Geonics TEM57 system is indicated.

but we do not analyze it here. Dashed lines indicate that the peak of
negative voltages is below the ambient noise level (1 10 8 V) at
El Arco. For the dc resistivity, 30 m is the lowest value with a
sign reversal above the noise level. With a value of 28 m the sign
reversal is below the noise level, while for values less than 28 m
there is no sign reversal, indicating that the polarization current is
never less than the fundamental inductive because of its slow decay.
For the other three ColeCole parameters, the voltage sign reversal
occurs below the noise level for chargeabilities lower than 0.2, time
constants higher than 0.05 s, and frequency exponents lower than
0.36. Cooper and Swift (1994) carried out coincident-loop and in-loop
soundings over three mineralized bodies of the Troodos massive,
Cyprus, over which galvanic IP methods had shown positive responses.
The TEM soundings failed to show a voltage sign reversal from these
bodies, possibly due to slow decaying fundamental inductive currents
produced by the low dc resistivities of the host rocks. The large
dimensions of the polarizable medium compared to the loop dimensions, the moderate resistivity of the subsurface, and adequate
chargeabilities, time constants, and frequency exponents, are all factors
which contributed to the observed sign reversal in the El Arco data.
6. Conclusions
The three conditions to observe the presence of a polarizable target
with in-loop TEM soundings are fullled by our data: An intense
fundamental inductive current with a fast decay and good EM
coupling between polarization current and receiving coil.
Two model features suggest our 1-D interpretation of the data
might be an adequate approximation to the actual 3-D resistivity
distribution; in the northeastern portion of the prole the thickness of
the rst layer does not vary rapidly and approximately coincides with
the thickness of the oxidized zone of about 60 m, and the chargeability
and time constant values in the second layer generally agree with the
higher concentrations in the primary sulde zone. When our
estimates are plotted on the chargeability versus time constant plots
of Pelton et al., most of the points fall in the zone of disseminated
texture and there is a reasonable agreement between the actual
sulde concentrations and those predicted by Pelton et al., agreeing
with the actual total suldes volume concentrations. The point
corresponding to the sounding located in the pyrite halo also agrees
with a higher sulde concentration and a likely veinlet texture. We
obtained high values for the frequency exponent, suggesting a
homogeneous distribution of particle sizes.
The parameter resolution analysis for homogeneous half-spaces
showed that the chargeabilities and the time constants are the least

436

C. Flores, S.A. Peralta-Ortega / Journal of Applied Geophysics 68 (2009) 423436

resolved of the four ColeCole parameters but their correlations with

other parameters are low, resulting in good resolutions for all the
parameters. A similar result was obtained for the two-layered models,
but here the resolution of these two parameters decreases due to
increased parameter correlations. Except for the dc resistivity, the
sensitivities of the ColeCole parameters are generally concentrated
close to where the voltage sign reversal occurs.
Acknowledgements
We thank Jaime Caldern for his invaluable help during the eld
work, and Hctor Romero for his fruitful discussions on the mineral
deposit. We are particularly grateful to Ings. Remigio Martnez Muller
and Benjamn Rascn for their logistic support. The permit granted by
Grupo Mxico, Mexicana de Cobre to publish this study is acknowledged. This work was funded by CICESE and CONACYT projects.
References
Anderson, W.L., 1975. Improved digital lters for evaluating Fourier and Hankel
transform integrals. United States Geological Survey Report GD-75-012.
Anderson, W.L., 1979. Numerical integration of related Hankel transforms of order 0 and
1 by adaptive digital ltering. Geophysics 44, 12871305.
Barthelmy, D.A., 1979. Regional geology of El Arco porphyry copper deposit, Baja
California. In: Abbot, P.L., Gastil, R.G. (Eds.), Baja California Geology. San Diego State
University, pp. 127138.
Bertin, J., Loeb, J., 1976. Experimental and theoretical aspects of induced polarization,
vol. II. Gebruder Borntraeger.
Cole, K.S., Cole, R.H., 1941. Dispersion and absorption in dielectrics. Journal of Chemical
Physics 9, 341.
Constable, S.C., Parker, R.L., Constable, C.G., 1987. Occam's inversion: a practical algorithm for generating models from electromagnetic sounding data. Geophysics 52,
289300.
Coolbaugh, D.F., Osoria-Hernndez, A., Echvarri-Prez, A., Martnez-Mller, R., 1995. El
Arco porphyry copper deposit, Baja California, Mexico. In: Pierce, F.W., Bolm, J.G.
(Eds.), Porphyry copper deposits of the American Cordillera. Arizona Geological
Society Digest, vol. 20, pp. 524538.
Cooper, N.J., Swift, R., 1994. The application of TEM to Cyprus-type massive sulde
exploration in Cyprus. Geophysics 59, 202214.
Echavarri-Prez, A., Rangin, C., 1978. El yacimiento cuprfero de El Arco, B.C.: Su
ambiente geolgico y sus caractersticas de alteracin y mineralizacin. Boletn del
Departamento de Geologa 1, 118 Universidad de Sonora.
Edwards, R.N., Bailey, R.C., Garland, G.D., 1981. Conductivity anomalies: lower crust or
asthenosphere? Physics of the Earth and Planetary Interiors 25, 263272.
El-Kaliouby, H.M., El-Diwany, E.A., Hussain, S.A., Hashish, E.A., Bayoumi, A.R., 1997.
Optimum negative response of a coincident-loop electromagnetic system above a
polarizable half-space. Geophysics 62, 7579.
Faras, R., 1978. Geophysical exploration of the El Arco-Calmalli mining district, Baja
California, Mexico, M.Sc. Thesis, Dept. of Geosciences, The University of Arizona.
Fink, J.B., McAlister, E.O., Sternberg, B.K., Wieduwilt, W.G., Ward, S.H., 1990. Induced
polarization, applications and case histories. Investigations in Geophysics. Society
of Exploration Geophysicists, vol. 4.
Fitterman, D.V., Anderson, W.L., 1987. Effect of transmitter turn-off time on transient
soundings. Geoexploration 24, 131146.
Ghorbani, A., Camerlynck, C., Florsch, N., Cosenza, P., Revil, A., 2007. Bayesian inference
of the ColeCole parameters from time- and frequency-domain induced polarization. Geophysical Prospecting 55, 589605.

Gubatyenko, V.P., Tikshayev, V.V., 1979. On the variation of sign of the electromagnetic
force of induction in the transient electromagnetic eld method. Izvestiya, Physics
of the Solid Earth 15, 217219 (in Russian).
Hohmann, G.W., Raiche, A.P., 1991. Inversion of controlled-source electromagnetic data.
In: Nabighian, M.N. (Ed.), Electromagnetic methods in Applied GeophysicsTheory.
Society of Exploration Geophysicists, vol. 1, pp. 469503.
Hoversten, G.M., Morrison, H.F., 1982. Transient elds of a current loop source above a
layered earth. Geophysics 47, 10681077.
Jones, A.G., 1982. On the electrical crustmantle structure in Fennoscandia: no Moho,
and the asthenosphere revealed? Geophysical Journal of the Royal Astronomical
Society 68, 371388.
Jupp, D.L.B., Vozoff, K., 1975. Stable iterative methods for the inversion of geophysical
data. Geophysical Journal of the Royal Astronomical Society 42, 957976.
Lee, T., 1981. Transient electromagnetic response of a polarizable ground. Geophysics 46,
10371041.
Luo, V., Zhang, G., 1998. Theory and application of spectral induced polarization. Society
of Exploration Geophysicists. Geophysical Monograph Series, vol. 8.
Nabighian, M.N., 1979. Quasi-static transient response of a conductive half-space an
approximate representation. Geophysics 44, 17001705.
Nabighian, M.N., Macnae, J.C., 1991. Time domain electromagnetic prospecting methods.
In: Nabighian, M.N. (Ed.), Electromagnetic methods in Applied Geophysics. Society of
Exploration Geophysicists, vol. 2, pp. 427479. Applications, Part A.
Newman, G.A., Hohmann, G.W., Anderson, W.L., 1986. Transient electromagnetic
response of a three-dimensional body in a layered earth. Geophysics 51, 16081627.
Pelton, W.H., Ward, S.H., Hallof, P.G., Sill, W.R., Nelson, P.H., 1978. Mineral discrimination
and removal of inductive coupling with multifrequency IP. Geophysics 43, 588609.
Peralta-Ortega, S.A., 2001, Polarizacin inducida con sondeos electromagnticos
transitorios sobre el prdo cuprfero de El Arco, Baja California, M.Sc. Thesis,
Divisin de Ciencias de la Tierra, CICESE, Mexico.
Poddar, M., 1982. A rectangular loop source of current on a two-layered earth.
Geophysical Prospecting 30, 101114.
Seigel, H.O., Vanhala, H., Sheard, S.N., 1997. Some case histories of source discrimination
using time-domain spectral IP. Geophysics 62, 13941408.
Smith, R.S., West, G.F., 1988. Inductive interaction between polarizable conductors: an
explanation of a negative coincident-loop transient electromagnetic response.
Geophysics 53, 677690.
Smith, R.S., West, G.F., 1989. Field examples of negative coincident-loop transient
electromagnetic responses modeled with polarizable half-planes. Geophysics 54,
14911498.
Soininen, H., 1984. Inapplicability of pulse train time-domain measurements to spectral
induced polarization. Geophysics 49, 826827.
Spies, B.R., Frischknecht, F.C., 1991. Electromagnetic sounding. In: Nabighian, M.N. (Ed.),
Electromagnetic methods in Applied Geophysics. Society of Exploration Geophysicists, vol. 2, pp. 285386. Applications, Part A.
Stoyer, C.H., 1990. Efcient computation of transient sounding curves for wire segments
of nite length using an equivalent dipole approximation. Geophysical Prospecting
38, 87100.
Vanhala, H., Peltoniemi, M., 1992. Spectral IP studies of Finnish ore prospects.
Geophysics 57, 15451555.
Ward, S.H., Hohmann, G.W., 1987. Electromagnetic theory for geophysical applications.
In: Nabighian, M.N. (Ed.), Electromagnetic Methods in Applied GeophysicsTheory.
Society of Exploration Geophysicists, vol. 1, pp. 130311.
Weber, B., Lpez-Martnez, M., 2006. Pb, Sr, and Nd isotopic and chemical evidence for a
primitive island arc emplacement of the El Arco porphyry copper deposit (Baja
California, Mexico). Mineralium Deposita 40, 707725.
Weidelt, P., 1982. Response characteristics of coincident loop transient electromagnetic
systems. Geophysics 47, 13251330.
Valencia, V.A., Barra, F., Weber, B., Ruiz, J., Gehrels, G., Chesley, J., Lpez-Martnez, M.,
2006. ReOs and UPb geochronology of the El Arco porphyry copper deposit, Baja
California Mexico: implications for the Jurassic tectonic setting. Journal of South
American Earth Sciences 22, 3951.