Clusters of charged Janus spheres.

Abstract

We study the assembly of spherical particles with opposite electric charge on both hemispheres, in the case that the particle diameter exceeds the electrostatic screening length. Clusters result, not strings. The cluster shapes are analyzed by combined epifluorescence microscopy and Monte Carlo computer simulations with excellent agreement, indicating that the particles assemble in aqueous suspension to form equilibrated aggregates. The simulations show that charge asymmetry of individual Janus particles is preserved in the clusters.

Full text

Page 1

Clusters of Charged Janus Spheres
Liang Hong,†Angelo Cacciuto,†Erik Luijten,* and Steve Granick*
Department of Materials Science and Engineering, UniVersity of Illinois,
Urbana, Illinois 61801
Received August 8, 2006; Revised Manuscript Received September 20, 2006
ABSTRACT
We study the assembly of spherical particles with opposite electric charge on both hemispheres, in the case that the particle diameter exceeds
the electrostatic screening length. Clusters result, not strings. The cluster shapes are analyzed by combined epifluorescence microscopy and
Monte Carlo computer simulations with excellent agreement, indicating that the particles assemble in aqueous suspension to form equilibrated
aggregates. The simulations show that charge asymmetry of individual Janus particles is preserved in the clusters.
Historically, colloid science has focused on understanding
particle interactions when the surface composition is so
uniform that the interaction potential depends only on
separation.1Though magnetic effects comprise an obvious
counter-example to the assumption of isotropic particle-
particle interactions, studies of magnetic colloids have
proceeded essentially independently.2Indeed, the assumption
that the relevant interactions are nondirectional3is the
premise for analyzing a vast number of ubiquitous techno-
logical and environmental problems involving colloids. Many
applications have ensued, including their use as proxies for
atoms in studying the dynamical structure of crystals and
glasses,4,5and their self-assembly into colloidal crystals is
useful in the important application of photonic materials.6
However, surface chemistry is commonly spotty, patchy, and
heterogeneous. Rather than dismiss this as imperfection, here
we show that qualitatively new behavior follows when
spherical symmetry is broken by anisotropic chemical
composition.
Synthetic routes to create colloidal-sized particles with
patchy surface chemical composition are known.7,8Also,
shape asymmetry can be created by causing block copoly-
mers to self-assemble in certain organic solvents.8In this
communication, we ignore the complexity of such shape
asymmetry and consider the “Janus” motif in which the
surface of a spherical particle is divided evenly into two areas
of different chemical composition. The effect of hemispheri-
cal interactions on the aggregation behavior of colloids
appears hitherto unstudied. Bearing in mind that electrostatic
potentials are better understood than those involving hydro-
phobic interactions, we consider the simpler case of hemi-
spheres having opposite surface charge density. Although
their net charge is zero, zwitterionic particles fail to behave
as classical (ideal) dipoles because their size greatly exceeds
the electrostatic screening length and the electrostatic interac-
tions thus have much shorter range than the particle size.
We study the structures into which zwitterionic Janus
particles self-assemble and show that colloids with directional
interactions can form clusters with definite shape and
anisotropic distribution of electric charge.
Experimental and Computational Details. We prepared
zwitterionic colloids using known methods for “Janus”
particles.9-11A suspension of (fluorescent) colloids was
spread onto a cleaned glass substrate such that a mono-
layer of colloid remained after the suspension liquid evapo-
rated. A thin (15 nm) gold film was deposited using elec-
tron-beam deposition. Monolayers of N,N,N-trimethyl-
(11-mercaptoundecyl)ammonium chloride were deposited
from ethanol solution to produce positive charge and
washed multiple times with 1% HCl ethanol solution to
remove the electrostatic adsorption. On the other hemi-
sphere, negative charge resulted from carboxylic acid groups
on the untreated side of carboxylate-modified polysty-
rene colloids. When this procedure was performed with
care, measurements of zeta potential showed that it
yielded particles whose charge on two hemispheres was
indeed nearly balanced. For example, the zeta potential of
bipolar particles is 0.6 mV, whereas it is -41.6 mV for bare
particles at the same salt concentration. The particles were
then suspended in aqueous solution, and epifluorescence
micrsocopy was used to image regions of relatively high
concentration formed when the particles settled to the bottom
of the glass sample container. In this thin volume ∼5 µm
from the surface, the maximum volume fraction of ∼10-3
was still dilute.
The particles in the experiments (F8819 from Invitrogen,
Inc.) had 1 µm diameter, but the size is not believed to be
fundamental, provided that it much exceeds the screening
length. The experiments were conducted in PBS buffer (pH
* Corresponding
uiuc.edu.
†These authors contributed equally.
authors.E-mail: luijten@uiuc.edu; sgranick@
NANO
LETTERS
2006
Vol. 6, No. 11
2510-2514
10.1021/nl061857i CCC: $33.50
Published on Web 10/12/2006
© 2006 American Chemical Society

Page 2

) 6) at 1 mM ionic strength, such that the Debye screening
length was approximately 10 nm.
Janus particles were modeled as spherical shells composed
of 12 002 isotropically distributed spherical particles12(cf.
Figure 1E). The large number of surface particles ensures a
smooth representation of the charge distribution on each
hemisphere. Each pair of surface particles interacts via an
attractive or repulsive square potential (representing their
screened electrostatic interaction), according to the sign of
both hemispheres. We find that our results are insensitive to
the precise range (and shape) of the nanoparticle potential
(and hence to precise salt concentration), as long as this range
is less than 30% of the colloid diameter. Because the
equilibration time increases rapidly with decreasing interac-
tion range, we have adopted a range of 10% of the size of
a Janus particle. By systematically exploring all possible
relative orientations of a pair of Janus particles, we obtained
the colloid pair potential, which was then fitted to a
functional form (cf. Figure 1A-D). We explicitly verified
that variation of the number of surface particles does not
change the shape of this potential, but merely rescales the
contact energy of a pair of Janus particles. The magnitude
of the colloid potential quickly decays to zero with increasing
surface-to-surface distance.
Subsequently, we carried out Monte Carlo (MC) simula-
tions in which the Janus particles were represented as
monodisperse hard spheres with a pairwise interaction given
by the functional form obtained. Use of this functional
form allowed us to study systems containing relatively
large numbers of particles for a large number of steps,
which is essential to ensure that stable configurations are
reached. The simulations employed a standard Monte Carlo
algorithm and were all performed in the NVT ensemble
with periodic boundary conditions. We explored poten-
tials with contact interactions between 5 and 10 kBT; reduced
units were chosen such that T is equal to unity. Up to N )
400 particles were used in each simulation. The minimum-
energy states were reached fairly quickly, but to ensure
equilibration we continued all runs for a minimum of 106
MC steps.
We proceeded as follows. First, a dilute system of Janus
particles with random orientation and position was allowed
to evolve via local MC moves. Because of the strong contact
interactions, clusters assembled relatively quickly. These
clusters were identified, and their structure was analyzed
visually and compared to the experimental images as well
asto the known Lennard-Jones clusters of the same size.
Next, to check the stability of each cluster of size n, we
also performed simulations where only n particles were
placed within the simulation cell and allowed to self-
assemble. As the number of metastable states increases
quickly with increasing cluster size n, we started with a high
temperature, to permit continuous rearrangements within the
cluster. All different structural arrangements (i.e., aggregates
of n Janus particles) were enumerated and each was quenched
to a minimum-energy state using a standard simulated
annealing technique. Finally, the energies of all resulting
clusters were compared to find the absolute minimum energy.
Although such an approach becomes inefficient for large n,
our analysis only included rather small clusters, giving us
confidence that the clusters presented below are indeed the
most stable ones.
Last, we note that the rotational diffusion of the experi-
mental clusters facilitated the careful and thorough visual
comparison with the clusters obtained from the simulations.
Clusters of size up to n ) 7 were observed at least 20-30
times. The clusters of largest n were observed less frequently,
but at least 2-3 times, validating the conclusion that the
agreement between simulation and experiment was not
coincidence.
Results and Discussion. The complexity of the interaction
potential between Janus particles is obvious upon inspection.
Consider a polar axis pointing north-south through the
hemispheres. Only if the axes of the two particles lie in the
same plane (Figure 1A) the interaction switches smoothly
from repulsion to attraction depending on how the particles
face one another. This is illustrated in Figure 1B; a switch
between repulsion and attraction over a narrow range of
rotation angle. Generally, the polar axes passing through the
hemispheres do not lie in the same plane (Figure 1C) and
the interaction potential depends on relative orientation as
well as on separation. This multidimensional parameter space
illustrates the rapid increase of complexity as one goes
beyond the classical concept of isotropic potentials. Figure
1D presents the calculated dependence of the interaction
potential between a pair of particles on the two polar angles
describing the orientation of one particle for a fixed orienta-
tion of the second particle and for fixed particle separation.
The key conclusion is the existence of multiple energetic
minima, an energy “landscape” in the true sense of the word.
It suggests that in nature particles of this kind might be prone
to slide over one another by Brownian motion, toward
configurations of lower electrostatic energy. Indeed, we
observe this in our experiments.
The parameters in the modeling are selected to be
consistent with the experimental conditions; the main point
is that the particle size of 1 µm greatly exceeds the elec-
trostatic screening length of ∼10 nm. On the basis of
manufacturer’s specification and measurements of the zeta
potential of unfunctionalized particles, we estimate their
surface charge on the order of 1 elementary charge per square
nanometer. In our modeling, we set the net particle charge
equal to zero.
The epifluorescence images show that the particles cluster
into definite geometrical shapes depending on the number
of particles, as summarized in column A of Figure 2. Clusters
larger than about 12 zwitterionic particles are not observed
owing to the dilute concentration. Having validated the
simulations (column B of Figure 2) by comparing to the
shapes observed experimentally, we exploit the modeling to
determine the mutual orientation of the charged hemispheres
on neighboring particles, because this information is so far
inaccessible experimentally. For the case of n ) 2 particles,
one observes that the particles find a low-energy state even
if their polar axes are not strictly in the same plane. This
broad energy minimum is expected, because the area of
Nano Lett., Vol. 6, No. 11, 20062511

Page 3

contact is so much smaller than the particle size. For clusters
larger than n ) 2, the polar axes of the hemispheres twist in
space as neighboring particles circle the structure. One
profound consequence is that if one considers the entire
Figure 1. Two particles with bipolar charge, analyzed using simulations. The right-hand object is fixed; the angles (θ,?) indicate the
orientation of the left-hand object. Red denotes positive charge, and yellow denotes negative charge; these charges are equal in magnitude.
(A) Charged hemispheres whose polar axes lie in the same plane. (B) Relative interaction energy for case A, plotted against angular
displacement for two particles at contact. The interaction potential switches from repulsive to attractive as the left-hand object rotates in the
θ direction even when the particle-particle separation does not change. (C) General case where the polar axes do not lie in the same plane.
(D) For every orientation of the right-hand particle, a potential energy landscape must be determined. Thus, the parameter space become
four-dimensional even at fixed surface separation, illustrating the rapid increase in complexity once one goes beyond the concept of isotropic
potentials. The potentials shown are obtained numerically using a pair of composite particles having 6001 fixed and uniformly distributed
“nanoparticles” per hemisphere, representing the surface charges (panel E). Each pair of nanoparticles interacts via an attractive or repulsive
square potential (representing their screened electrostatic interaction), according to the sign of both hemispheres. We find that the shapes
of the resulting clusters are insensitive to the precise range of the nanoparticle potential (and hence to precise salt concentration), as long
as this range is less than 30% of the colloid diameter. Because the equilibration time increases rapidly with decreasing interaction range,
we have adopted a range of 10% of the particle size. The magnitude of the resulting colloid potential quickly decays to zero with increasing
surface-to-surface distance.
2512
Nano Lett., Vol. 6, No. 11, 2006

Page 4

cluster as a particle in its own right then the envelope of
electrostatic charge is not distributed evenly. To quantify this,
Column C of Figure 2 displays the envelope of electrostatic
charge presented to the environment (the caption of Figure
2 describes the color coding). Charge is distributed asym-
metrically; one half of each (non-spherical) cluster tends to
be predominantly positive in charge, the other half predomi-
nantly negative. Thus, charge asymmetry on the elemental
spherical particles is preserved in the surface charge asym-
metry of the clusters into which they assemble.
In principle, there appears to be no limit to the size of
clusters that can form from these electrostatic interactions;
our modeling shows that each cluster of larger size is
energetically more stable than all of the smaller ones. This
implies that the clusters depicted in Figure 2 might merge
into larger ones, if only their concentration were higher. This
has not yet been possible to investigate experimentally owing
to the limited quantity of sample available. However, because
the charge anisotropy of individual particles is preserved in
the clusters analyzed here, it is logical to conclude that
energetically the positive side of one cluster will attract the
negative side of another cluster, resulting in facile assembly
into even larger aggregates, based on short-range, orientation-
dependent interactions of the kind presented by our calcula-
tions summarized in Figure 1.
The compact shapes in Figure 2 differ fundamentally from
the lines and rings formed by magnetic particles2and electric
dipoles, which are much studied in recent literature.13Our
particles would behave as ideal dipoles only at distances that
much exceed the actual interaction range. It is appropriate
to compare the shapes we observe to those of colloidal-sized
spheres without electrostatic interaction. A pioneering study
by Pine and co-workers of uncharged spheres that self-
assemble owing to van der Waals attraction14revealed
structures that are identical to those predicted mathematically
for most favorable packing.15For n ) 2-5 and n ) 7, we
observe the same overall shapes albeit with the internal
structure and charge asymmetry depicted in Figure 2. An
intriguing technical point is that packing considerations14,15
lead, for n > 7, to structures different from those predicted
Figure 2. Comparison of experimental epifluorescence images and Monte Carlo computer simulations of the self-assembled structures of
particles with near-equal positive and negative charges on the two hemispheres (denoted by red and yellow colors). Clusters form with
monotonically decreasing energy as the number of colloidal particles increases. Their computed structures (column B) agree quantitatively
with the observed structures (column A). The charge distribution in these clusters is also computed; the color goes smoothly from red to
yellow depending on the inner product of the vector from the center of mass of the cluster to each colloid and the axial vector (pointing
to the positive side) of this colloid (column C). In the charge distributions, a wrapping has been added to emphasize the outer surface. As
described in the text, experimental measurements of zeta potential showed it to be <1 mV, confirming that charge on two hemispheres was
nearly balanced.
Nano Lett., Vol. 6, No. 11, 20062513

Page 5

by modeling spheres with Lennard-Jones interactions. By
contrast, our findings differ from either prediction already
for n ) 6, but reproduce the Lennard-Jones clusters for n )
7-10. For cluster sizes of n ) 6, 11, 12, and 13, the shapes
found for hemispherical interactions differ from those
obtained from isotropic interactions; the difference is that
the shapes we observe are less symmetric. We anticipate even
larger distinctions for particles of lesser symmetry, for
example, unmatched positive and negative surface charge,
and also for larger clusters.
A key aspect of these particles is that real-time imaging
reveals the process of self-assembly. The constituent zwit-
terionic spheres assemble into clusters, the clusters coalesce
with one another, and structural defects heal; all of this on
the convenient time scale of seconds to minutes. In earlier
work on colloid clusters, the conditions were such that in
the ultimate structures the elements were quenched in place.14
In the present system, the forces of interaction are relatively
weak. From simulation we estimate them as on the order of
5-10 kBT and this is consistent with the unambiguous
observation that equilibration events, dynamical particle
attachment to and detachment from clusters, and reconfigu-
ration of cluster shapes, can be observed in video images as
these clusters self-assemble. This experimental approach will
allow comparison with prior computer-based calculations of
how patchy particles associate.13,16-20
Looking to the future, the idea of directional self-assembly
between colloidal-sized particles suggests many possibilities.
It is obvious to generalize the situation to consider colloids
whose shape is not just spherical, as in this study, but also,
for example, rodlike or oblate, more complex than the
spherical shape considered in this study for simplicity. Also
obvious is that the bipolar functionality studies in this
communication can be generalized to ternary; a simple place
to start will be to divide spheres into three regions of different
chemical composition, as appears to be possible by vacuum
deposition of metal at oblique angles. Finally, it is intriguing
to note that asymmetric charge distribution has been pro-
posed, based on experiments, to explain the aggregation of
some proteins.21Thus, the model particle systems that we
study here may have relevance to proteins and other more
complex particles where the surface charge distribution is
similarly patchy.
However, a limitation of our present method to produce
Janus particles is that their yield is low. Methods to scale
up the production of Janus particles to larger quantities are
under development and will be reported presently.
Acknowledgment. We thank J. F. Douglas for discussion
and H. Tu and P.V. Braun for the samples of N,N,N-
trimethyl(11-mercaptoundecyl)ammonium chloride. This work
is supported (A.C. and E.L.) by the National Science
Foundation, CAREER Award DMR-0346914 and (L.H. and
S.G.) by the Donors of the Petroleum Research Fund,
Administered by the American Chemical Society, no. 45523-
AC7.
References
(1) Israelachvili, J. N. Intermolecular and Surface Forces, 2nd Ed.;
Academic: San Diego, CA, 1992.
(2) Klokkenburg, M.; Dullens, R. P. A.; Kegel, W. K.; Erne ´, B. H.;
Philipse, A. P. Phys. ReV. Lett. 2006, 96, 037203.
(3) Lyklema, J. Fundamental of Colloid and Interface Science; Academic:
San Diego, CA, 2005; Vol. 4.
(4) Alsayed, A. M.; Islam, M. F.; Zhang, J.; Collings, P. J.; Yodh, A.
G. Science 2005, 309, 1207.
(5) Weeks, E. R.; Crocker, J. C.; Levitt, A. C.; Schofield, A.; Weitz, D.
A. Science 2000, 287, 627.
(6) Qi, M. H.; Lidorikis, E.; Rakich, P. T.; Johnson, S. G.; Joannopoulos,
J. D.; Ippen, E. P.; Smith, H. I. Nature 2004, 429, 538.
(7) Perro, A.; Reculusa, S.; Ravaine, S.; Bourgeat-Lami, E.; Duguet, E.
J. Mater. Chem. 2005, 15, 3745.
(8) Erhardt, R.; Bo ¨ker, A.; Zettl, H.; Kaya, H.; Pyckhout-Hintzen, W.;
Krausch, G.; Albetz, V.; Mu ¨ller, A. H. E. Macromolecules 2001,
34, 1069.
(9) Lu, Y.; Xiong, H.; Jiang, X.; Xia, Y.; Prentiss, M.; Whitesides, G.
M. J. Am. Chem. Soc. 2003, 125, 12724.
(10) Anker, J. N.; Behrend, C.; Kopelman, R. J. Appl. Phys 2003, 93,
6698.
(11) Behrend, C. J.; Anker, J. N.; Kopelman, R. Appl. Phys. Lett. 2004,
84, 154.
(12) Hardin, R. H.; Sloane, N. J. A.; Smith, W. D. Tables of Spherical
Codes with Icosahedral Symmetry, published electronically at http://
www.research.att.com/∼njas/icosahedral.codes/.
(13) Van Workum, K.; Douglas, J. F. Phys. ReV. E 2006, 73, 031502.
(14) Manoharan, V. N.; Elsesser, M. T.; Pine, D. J. Science 2003, 301,
483.
(15) Sloane, N. J. A.; Hardin, R. H.; Duff, T. D. S.; Conway, J. H. Discrete
Comput. Geom. 1995, 14, 237.
(16) Doye, J. P. K.; Miller, M. A.; Mortenson, P. N.; Walsh, T. R. AdV.
Chem. Phys. 2000, 115, 1.
(17) Wales, D. J. Int. J. Mod. Phys. B 2005, 19, 2877.
(18) Zhang, Z.; Glotzer, S. C. Nano Lett. 2004, 4, 1407.
(19) De Michele, C.; Gabrielli, S.; Tartaglia, P.; Sciortino, F. J. Phys.
Chem. B 2006, 110, 8064.
(20) Zerrouki, D.; Rotenberg, B.; Abramson, S.; Baudry, J.; Goubault,
C.; Leal-Calderon, F.; Pine, D. J.; Bibette, M. Langmuir 2006, 22,
57.
(21) Stradner, A.; Sedgwick, H.; Cardinaux, F.; Poon, W. C. K.; Egelhaaf,
S. U.; Schurtenberger, P. Nature 2004, 432, 492.
NL061857I
2514
Nano Lett., Vol. 6, No. 11, 2006