Full Paper

www.mts-journal.de

A Study of a Novel Synergy Definition for Ternary

CB/CNT Composites Suggesting a Representative
Model for CB and CNT
Guanda Yang,* Dirk W. Schubert, Fritjof Nilsson, Muchao Qu, and Michael Redel

areas that require antistatic performance,

For polymer composites containing two types of conductive fillers, the electrostatic dissipation, or electromag-
electrical conductivity can be predicted with a ternary mixing equation. netic shielding. Intrinsically conductive
polymers are usually expensive and age
However, a slight deviation from the foreseen trend is recently observed for
easily.[1–3]
ternary polymethyl methacrylate (PMMA) nanocomposites containing carbon Due to their high conductivity, low
nanotubes (CNT) and carbon black (CB). In this study, the reason for the weight and simple processing, conductive
deviation is examined. Representative composite geometries are generated polymer composites (CPCs) have a prom-
with Monte-Carlo technique and the composite conductivity is calculated ising property for many electrical applica-
with Finite Element Modeling (FEM) Traditional FEM strategies would give tions. And as two of the most commonly
used conductive fillers in CPCs, Carbon
results resembling the ternary mixing equation, but when the CB particles are
nanotubes (CNTs) and carbon black (CB)
modeled as solid spheres whereas the CNT are modeled as solid cylinders are studied in lots of ternary systems with
surrounded by soft, conductive shells, the composite conductivity gained the CPCs (i.e., three-component CPCs). The
same shape as observed experimentally. One possible interpretation is that conductivity of binary CPCs can be mod-
a small geometrical rearrangement of the CNT fibers has a larger impact on eled, as a function of filler fraction, ori-
entation of the fillers as well as the ratio
the composite conductivity than a corresponding movement of the spherical
between the fillers. The conductivity of ter-
CB particles. As a consequence, the CNT/CB/PMMA nanocomposite exhibits nary composites can be modeled with an
a special kind of synergy (“synergism”), caused by the grape-shaped mor- analytical model reported in Qu et al.[3–6]
phology. These findings can be utilized in the future development and optimi- Carbon nanotubes (CNTs) have been
zation of conductive polymer nanocomposites. identified as an ideal reinforcement phase
in composite materials due to their excel-
lent mechanical and physical proper-
ties.[3–6] Their high aspect ratio (over 1000) and high conduc-
1. Introduction
tivity (up to 104 S cm−1) enable them to significantly increase
1.1. Background the conductivity of polymer matrix composites (PMCs), by
adding a very small volume fraction of the CNT to the bulk
Polymers are widely used in the society due to their excellent polymer.[7–9] A large number of experiments have been carried
plasticity, light weight and low cost. However, most polymers out to examine the electrical percolation threshold of CNT/
are galvanically isolated, which severely limits their use in polymer composites.[10–14] Several theoretically and numerical
studies have also been performed to estimate the percolation
threshold of CNT/polymer composites,[15] often utilizing the
G. Yang, Prof. D. W. Schubert, Dr. M. Qu, M. Redel
Institute of Polymer Materials excluded volume method[16–24] and Monte Carlo simulations.
Friedrich-Alexander-University Erlangen-Nuremberg All these studies conclude that a better dispersion, a larger
Martensstr. 7, Erlangen 91058, Germany aspect ratio and stable properties of the CNTs lead to a lower
E-mail: [email protected] percolation threshold. However, large aspect ratio CNTs can
G. Yang, Prof. D. W. Schubert, Dr. M. Qu, M. Redel easily be entangled and agglomerated when dispersed in
Bavarian Polymer Institute
KeyLab Advanced Fiber Technology
polymer matrices.[25–27] Carbon black (CB) is one of the most
Dr.-Mack-Straße 77, Fürth 90762, Germany commonly used conductive nanoparticles in industrial applica-
Prof. F. Nilsson tions today. It has low cost, good dispersibility, low percolation
KTH Royal Institute of Technology threshold, high conductivity. Additionally, experimental studies
School of Chemical Science and Engineering have revealed a synergistic effect between the electrical conduc-
Fibre and Polymer Technology tivities of CNT and CB.[27–31] The combination of CB and CNT
Stockholm, SE 10044, Sweden
is therefore a promising candidate when designing conductive
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/mats.202000035.
polymer composites.
For example, Socher et al.[32] observed the synergistic influ-
DOI: 10.1002/mats.202000035 ence of CB in hybrid nanocomposites containing multi-walled

Macromol. Theory Simul. 2020, 2000035 2000035 (1 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 1. Morphology of the anisotropic ternary CPCs filament with 5.44 vol. % CNTs and 6 vol. % CB: a) the cross section of the specimen; b) the
middle section of the specimen.

CNTs (MWCNTs) and CB. A relationship between maximum

conductivity and MWCNT dispersion was also identified. Sum-
fleth et al.[33] found synergistic effects in network formation and
in charge transportation for ternary CB/CNTs epoxy systems.
They also related electrical conductivity with microstructure.
Ma et al.[34–35] also examined CB/CNTs-epoxy composites and
found that, in addition to the synergistic effects on the conduc-
tivity, the addition of both CB and CNT improved the composite
ductility and fracture toughness more than corresponding frac-
tions of pure CB or CNT.

1.2. Preceding Experimental Results

Qu et al.[3] studied ternary composites systems consisting of

CB/CNT and polymethyl methacrylate (PMMA). A synergistic
effect between CB and CNT was revealed in that work. Figure 1
shows a) the cross section b) the middle section of a ternary
composite filament with 5.44 vol. % CNTs and 6 vol. % CB. Figure 2. Logarithm of the composite conductivity versus volume filler frac-
Figure 2 shows the logarithm of the electrical conductivity tion of CNTs and CB in a contour plot diagram, defining a quantifiable “syn-
ergasm” between CNTs and CB. The inset synergasm map illustrates the
plotted versus the CB (black squares) and CNT (red circles)
synergasm effect of the different compositions of CNTs and CB from the
volume fraction, respectively. Each data point represents the previous study.[12,36]
average of 10 measurements, with standard deviations shown
with vertical error bars. Horizontal error bars indicate the uncer-
tainty in filler fraction. From Figure 2 it is evident that a syner-
gistic effect of CB and CNT is present in this ternary composite,
otherwise all the field-lines would have been linear or convex.
Figure 3 shows the binary conductivity curve of CB and CNT
including the region beyond percolation. The curves for CB and
CNT were surprisingly similar, considering that the electrical
percolation threshold typically decreases when the aspect ratio
(i.e., length/width) of the particles increases.[23] The reason is
that the spherical CB particles agglomerate in a way such that
their effective aspect ratio becomes much larger.

2. FEM Simulation Procedure

Finite element modeling (FEM) was utilized to predict the
electrical conductivity of three-component polymer com-
posites. Representative, periodic composite structures were Figure 3. The logarithm of the conductivity versus filler volume fraction
generated with Monte-Carlo technique, the input model of CNTs and CB, respectively.[12,36]

Macromol. Theory Simul. 2020, 2000035 2000035 (2 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 4. The spring-formed CNT are modeled as cylinders, whereas aggregates of small, spherical primary CB particles are modeled as large spheres.

materials were given realistic material parameters, the geom- boundaries, they are cut in two sections. Numerical troubles
etry was meshed and an applied voltage was assigned to oppo- can arise when one of the two sections is too small, and there-
site faces of the geometry. Finally, the Poisons equation was fore the smallest objects were removed. All the fillers were
solved in order to compute the effective electrical conductivity modeled using a hard-core and soft-schell strategy where no
of the composite as function of filler content of the two fillers. overlap of the hard core between particles was allowed, nei-
This procedure resembled the strategy presented in Yang,[2] ther between CBs, between CNTs or between CBs and CNT.
but now with a combination of cylinders and spheres instead The geometry was meshed using a fine tetrahedral mesh, as
of only cylinders. shown in Figure 5. One side of the domain was grounded, the
The material parameters in the FEM simulations corre- opposite side was given a fixed potential electrical potential and
spond to the material data reported by Qu et al.[3] The polymer the remaining sides were given periodic boundary conditions.
matrix was PMMA Plexiglas 7 N from Evonik Röhm GmbH Finally, the electrical conductivity of each composite section
(Germany) with Mw = 99 kg mol−1, polydispersity index was computed by solving Poisons equation.
1.52, density 1.19 g cm−3 and electrical conductivity 3.64 ×
10−9 S cm−1. The CNTs (Bayertubes) had outer diameter 13 nm,
inner diameter 4 nm, average length 1 µm, bulk density of
130–150 kg m−3 and electrical conductivity 134 S cm−1. The
CB (“Printex XE2” from Evonik Degussa GmbH) had a spe-
cific surface area of 900 m2/g measured according to the BET
method, according to the data sheet. The mean diameter of
the primary CB particles was ≈35 nm, the CB density at room
temperature was 2.13 g cm−3 and the electrical conductivity of
the CB was 100 S cm−1.
As seen in the SEM images of Figure 1, most CNTs in this
study were shaped as springs. Using the excluded volume
theory presented by Balberg,[22] the CNTs were modeled as
springs contained in surrounding cylinders with diameter
78 nm and length 1800 nm (Figure 4 (left)). The CBs were mod-
eled as spherical 350 nm diameter aggregates, each containing
multiple primary CB particles. (Figure 4 (right)).[4] The conduc-
tive volumes are shown in Figure 4 (transparent).
In all the simulations, the nanofillers (CB and CNT) were
randomly inserted, using Monte-Carlo technique, into cubic
composite boxes with 300 µm side length, An isotropic orienta-
tion of the CNT was assigned and the geometry had periodic Figure 5. Mesh strategy of the whole domain for simulation with CNT as
boundary conditions. When particles are bridging the periodic capped cylinders and CB as spheres.

Macromol. Theory Simul. 2020, 2000035 2000035 (3 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 6. FEM simulation result with streamlines of current density with carbon black as spheres and CNT as capped cylinders.

3. Calculation and Results

3.2. Rescaling Using Binary Composite Systems
3.1. FEM Simulation Results
Preliminary rescaling parameters were determined by com-
Typical solutions of the FEM modeling are displayed in paring simulated and experimental conductivity versus particle
Figure 6, showing electrical potential with colors red to blue, fraction for binary CB/PMMA and CNT/PMMA composites.
electrical field lines with red lines and current flow with red The data points were fitted with a power function (Equation 1),
arrows. As expected, the electrical current tends to flow through where σ is the composite conductivity, σM is the polymer matrix
the nanoparticles, since they have a higher intrinsic electrical conductivity, n = 3 and k is an adjustable parameter:
conductivity than the surrounding insulating polymer matrix.
Simulations like those presented in Figure 7 were sys- σ
= (1 + k × ϕ )n (1)
tematically performed for ternary composites with CNT σM
filler fractions between 0 and 8 vol. % and CB filler fractions
between 0 and 10 vol. %. The effective electrical conductivity For CB/PMMA composites the fitted parameter k became 208
was computed for those composite compositions and the and 246 for simulations and experiments, respectively, whereas
results were displayed in a contour plot (Figure 9), where for CNT/PMMA composites the corresponding k values became
each conductivity data point correspond to the average of 83 and 175 (Figure 8). The simulations thus underestimated k
40 FEM simulations. with a factor 1.2 and 2.1, respectively for CB and CNT. These
For the binary and ternary CNT/CB/PMMA composites, numbers can be used to rescale the x–y axis in the simulation
the thickness of the soft shells surrounding the solid particles contour plot of Figure 9 to make the results more readily com-
is 25% to the diameter of hard core. With these settings, the parable with their experimental counterparts in Figure 2. It was
contour plot in Figure 9 became qualitatively similar to shape however noted that an even more efficient rescaling algorithm
of the corresponding experimental contour plot (Figure 2). A could be obtained by using data for ternary composites.
locally enlarged contour plot, with adjusted X and Y axis, with
0–7 vol% of CB and 0–5. vol% of CNT is shown in Figure 11.
The simulated percolation thresholds of CB and CNT did how- 3.3. Rescaling Using Ternary Composite Systems
ever somewhat higher than their experimental counterparts.
In order to enable also quantitative comparisons between sim- A rescaling of the FEM conductivity data can also be performed
ulated and experimental conductivities for ternary compos- with an algorithm that uses information from the experimental
ites, a rescaling of the simulation data was therefore needed. measurements, with the aim to facilitate comparisons between

Macromol. Theory Simul. 2020, 2000035 2000035 (4 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 7. The current flow built up by both CNT and carbon black, which explains the synergism of the both particles.

simulations and experimental. The average of 40 FEM simula- determined by the difference in volume fraction of CB and
tions was used to compute the conductivity of each simulation CNT, and the three points with the smallest square of the dif-
data point. For each of these averaged simulation data points, ference between the two components will be selected.
the following algorithm was applied: 3. Determine the normal vector N = [a b c], which is perpendicu-
lar to the plane through these three experimental points, by
1. Fetch the coordinates (x0, y0, and z0) of the FEM data point,
computing the determinant:
where x0 is CB filler fraction, y0 is CNT filler fraction and z0
is simulated conductivity of a composite with these CB and
CNT filler fractions. a b c
2. Search the experimental data for the three data points (x1, y1, x 2 − x 1 y 2 − y 1 z 2 − z1 (2)
z1), (x2, y2, z2), (x3, y3, z3) which have x and y values closest x 3 − x 1 y 3 − y 1 z3 − z1
to the simulation data point (x0, y0, z0). The closest point is

Figure 8. Normalized electrical conductivity versus vol.% filler faction in binary composites. The lines are fitted with Equation 1. a) CB b) CNT.

Macromol. Theory Simul. 2020, 2000035 2000035 (5 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 9. Simulation data points with both CB and CNT. Each point is
calculated out of 40 FEM simulation. All the data are converted to the
same filler fraction of CB and CNT by Equation 7.

which gives: Figure 10. Calculating the interpolated point from the simulational data.
S(x0, y0, z0) is a simulation result, but it is not possible to make a direct
comparison if the CB and CNT value are not same as the experimental
a = ( y 2 − y1)( z3 − z1) − ( z2 − z1)( y 3 − y1)(3)
results (point A, B, C). By calculating the plane normal vector composed
of ABC three points, the projection Sr on the plane formed by the A,B,C
b = ( z 2 − z1)( x 3 − x 1) − ( x 2 − x 1)( z3 − z1)(4) points can be calculated. Sr represents the value that the experimental
data can draw. By comparing the S point with Sr, comparison between
c = ( x 2 − x 1)( y 3 − y1) − ( y 2 − y1)( x 3 − x 1)(5) the simulation and experimental results can be made.

The rescaled simulation point (x0, y0, z0r) must fulfil:

S value shows the deviation between the simulation and the
a ∗ ( x 0 − x 1) + b ∗ ( y 0 − y1) + c ∗ ( z0r − z1) = 0 (6) experiment. It is obvious that the smaller the S value, the closer
the two results are. The A and B values obtained at the smallest
Accordingly, the rescaled conductivity z0r can be calculated as: S value will represent the best tuning parameters.

a ∗ ( x 1 − x 0) + b( y 1 − y 0)
σ (φCB,φCNT ) = z0r = z1 + (7)
c

Thus, all the simulated data points with known CB- and CNT
filler fractions (x0,y0) could be interpolated to the experimental
data at position (x0,y0). The process of calculating the interpo-
lated points are shown in Figure 10.
It is obvious that in the simulation, the synergism point
could be obtained by a higher filler fraction around 4% filler
fraction of CB, and 2% filler fraction of CNT, where in the
experimental data it is 5% of CB and 3% of CNT. A possible
reason is that during the simulation, all the particles are set
as “hardcore” without soft shell. But in the reality, due to the
tunnel effect and the break-down voltage, particles could be
connected to each other even they have a distance to each
other.
In order to compare the experimental data and the simu-
lation results, two correction factors A and B are used in fol-
lowing functions:
Figure 11. Simulation data points with zoomed out 0–7 vol% of CB and
0–5. vol% of CNT. The synergism kurve pointed out is very obvious to
∑
n
(ϕ ABi − ϕ expi )2
identify at the volume fraction at 5 vol% of CB and 2 vol% of CNT. In
S= i =1
(8) order to more prominently express the synergism curve, the ratio of the
∑
n
i =1
ϕ expi
2
horizontal and vertical coordinates was adjusted accordingly.

Macromol. Theory Simul. 2020, 2000035 2000035 (6 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

Figure 12. 3D Figure which shows how the rescaling factor A and B (from 0.5 to 1.5 with interval of 0.01) influent the S value (Z-axis). The minimium
S value is pointed out on the figure.

According to function 7, the filler fraction of CB and CNT are 4. Conclusion

corrected by A and B via:
In this study, a 3D Monte Carlo model was developed for sys-
ϕ AB = ϕ ( A × CB,B × CNT ) tematically studying the electrical conductivity in different ter-
a × ( A × x 1 − A × x 0) + b × (B × y1 − B × y 0) (9) nary CNT–CB/polymer composites, in which CB, modeled as a
= z1 + sphere, and CNT, modeled as a capped cylinder, were randomly
c
distributed within the simulating space. The aspect ratio of the
The A and B value are defined between 0,5 and 1,5, and the S
value are calculated by function 8 and 9 to determine the min-
imum value of S.
The minimum S value is 0.60614 which happens at the point
of:
A = 1.12 (to modify CB)
B = 0.79 (to modify CNT) (Figure 12)
The rescaling factors A and B shows how we could rescale
the simulation results to make it fit the experimental results
best (Figure 13). Since the rescaling factor A is larger than
1 and B is smaller than 1, we can draw a conclusion as fol-
lowing: Although the cylindrical structure corresponding
to the carbon nanotubes can make it better in conducting
conductivity in simulation experiments, the enhancement
effect of the conductivity of CNT must be discounted with
about 21% (corresponding to the B value). Correspondingly,
although carbon black is not as effective as CNT in enhancing
the conductivity in the simulation, in the actual experimental
process, it should be multiplied by an enhancement coeffi- Figure 13. Figure that shows the electrical conductivity in simulation(X-
cient of 112% (corresponding to A value). The finite element axis) and results measured in experiment(Y-axis). The ideal case would
analysis by computer simulation has a deviation of about 12% be the linear fit with function y = x (showed as the red line). The blue line
on the cylindrical structure. shows the actual linear fit with the points.

Macromol. Theory Simul. 2020, 2000035 2000035 (7 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.mts-journal.de

CNTs and the diameter ratios of the CB aggregates to CNTs [5] J. B. Klauda, B. R. Brooks, R. W. Pastor, J. Chem. Phys. 2006, 125,
were taken into account. Simulation results were obtained by 144710.
a simple and effective pathway-finder algorithm. The following [6] O. N. de Souza, R. L. Ornstein, Biophys. J. 1997, 72, 2395.
was concluded from this study: [7] F. Carmona, P. Prudhon, F. Barreau, Solid State Commun. 1984, 51,
255.
1. Compared with other theoretical and experimental results, [8] W. Niesel, Ann. Phys. 1952, 445, 336.
the pathway-finder algorithm, in which the connection pos- [9] G. Ondracek, Mater. Chem. Phys. 1986, 15, 281.
sibility was deduced from the reachability matrix, proved to [10] L. E. Nielsen, Ind. Eng. Chem. Fundam. 1974, 13, 17.
be effective. [11] D. M. Bigg, Polym. Eng. Sci. 1977, 17, 842.
2. A synergistic effect can be obtained in the ternary CPCs of [12] E. P. Mamunya, V. V. Davidenko, E. V. Lebedev, Compos. Interfaces
1996, 4, 169.
CNT and CB. It is obvious that in the simulation, the syner-
[13] M. Weber, M. R. Kamal, Polym. Compos. 1997, 18, 711.
gism can be obtained by a higher filler fraction around 8% [14] D. S. McLachlan, M. Blaszkiewicz, R. E. Newnham, J. Am. Ceram.
filler fraction of CB, and 1% filler fraction of CNT, where in Soc. 1990, 73, 2187.
the experimental data it is 5% of CB and 2% of CNT. The [15] F. Lux, J. Mater. Sci. 1993, 28, 285.
results fit the theory of novel theoretical self-consistent mean- [16] S. Kirkpatrick, Rev. Mod. Phys. 1973, 45, 574.
field approach published by Schubert in 2017[1]. [17] M. Qu, D. W. Schubert, Compos. Sci. Technol. 2016, 136, 111.
3. In the simulation, the effect of carbon fiber to enhance the [18] J. Y. Yi, G. M. Choi, J. Electroceram. 1999, 3, 361.
conductivity is larger than in the experiments with a factor of [19] N. A. M. Radzuan, A. B. Sulong, J. Sahari, Int. J. Hydrogen Energy
112%, while the relative carbon black is smaller with a factor 2017, 42, 9262.
of 79%. [20] Y. P. Mamunya, V. V. Davidenko, P. Pissis, E. V. Lebedev, Eur. Polym.
J. 2002, 38, 1887.
[21] F. Dalmas, R. Dendievel, L. Chazeau, J. -. Y. Cavaillé, C. Gauthier,
Acta Mater. 2006, 54, 2923.
Conflict of Interest [22] I. Balberg, C. H. Anderson, S. Alexander, N. Wagner, Phys. Rev. B
1984, 30, 3933.
The authors declare no conflict of interest. [23] M. L. Clingerman, J. A. King, K. H. Schulz, J. D. Meyers, J. Appl.
Polym. Sci. 2002, 83, 1341.
[24] Z. Starý, J. Krückel, C. Weck, D. W. Schubert, Compos. Sci. Technol.
Keywords 2013, 85, 58.
[25] J. Krückel, Z. Starý, C. Triebel, D. W. Schubert, H. Münstedt, Polymer
conductivity, CPCs, simulation, ternary systems 2012, 53, 395.
[26] D. S. McLachlan, G. Sauti, J. Nanomater. 2007, 2007, 15.
Received: May 5, 2020 [27] F. Nilsson, U. W. Gedde, M. S. Hedenqvist, Compos. Sci. Technol.
Revised: June 16, 2020 2011, 71, 216.
Published online: [28] F. Nilsson, J. Krueckel, D. W. Schubert, F. Chen, M. Unge,
U. W. Gedde, M. S. Hedenqvist, Compos. Sci. Technol. 2016, 132, 16.
[29] F. Nilsson, M. Unge, J. Phys. D: Appl. Phys. 2016, 49, 335303.
[30] I. Balberg, N. Binenbaum, N. Wagner, Phys. Rev. Lett. 1984, 52,
[1] D. W. Schubert, Macromolecular Theory and Simulations, Wiley, New 1465.
York 2018 [31] I. Balberg, N. Binenbaum, Phys. Rev. B 1983, 28, 3799.
[2] G. Yang, D. W. Schubert, M. Qu, F. Nilsson, Macromol. Theory [32] D. A. G. Bruggeman, Ann. Phys. 1935, 416, 636.
Simul. 2018, 27, 1700105. [33] D. A. G. Bruggeman, Ann. Phys. 1936, 25, 645.
[3] M. Qu, F. Nilsson, Y. Qin, G. Yang, Y. Pan, X. Liu, G. H. Rodriguez, [34] D. A. G. Bruggeman, Ann. Phys. 1935, 24, 636.
J. Chen, C. Zhang, D. W. Schubert, Compos. Sci. Technol. 2017, 150, 24. [35] H. Fricke, Phys. Rev. 1924, 24, 575.
[4] a) J. Krückel, Z. Starý, C. Triebel, D. W. Schubert, H. Münstedt, [36] M. Qu, Characterization and Modeling of Anisotropic Electrically Con-
Polymer 2012, 53, 395; b) C. D. Lorenz, R. M. Ziff, Phys. Rev. E 1998, ductive Composite Filaments comprising PMMA and Carbon Fillers
57, 230. 2019.

Macromol. Theory Simul. 2020, 2000035 2000035 (8 of 8) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim