week ending

PRL 116, 208001 (2016) PHYSICAL REVIEW LETTERS 20 MAY 2016

Hiding in Plain View: Colloidal Self-Assembly from Polydisperse Populations

Bernard Cabane,1,* Joaquim Li,2 Franck Artzner,3 Robert Botet,4 Christophe Labbez,5 Guillaume Bareigts,5
Michael Sztucki,6 and Lucas Goehring2,†
1
LCMD, CNRS UMR 8231, ESPCI, 10 rue Vauquelin, 75231 Paris Cedex 05, France
2
Max Planck Institute for Dynamics and Self-Organization (MPIDS), 37077 Göttingen, Germany
3
Institut de Physique de Rennes, CNRS UMR 6251, Université Rennes 1, 35042 Rennes Cedex, France
4
Physique des Solides, CNRS UMR 8502, Université Paris–Sud, F-91405 Orsay, France
5
ICB, CNRS UMR 6303, Université de Bourgogne Franche-Comté, F-21078 Dijon, France
6
ESRF—The European Synchrotron, CS40220, 38043 Grenoble Cedex 9, France
(Received 8 November 2015; revised manuscript received 10 March 2016; published 20 May 2016)
We report small-angle x-ray scattering experiments on aqueous dispersions of colloidal silica with
a broad monomodal size distribution (polydispersity, 14%; size, 8 nm). Over a range of volume fractions,
the silica particles segregate to build first one, then two distinct sets of colloidal crystals. These dispersions
thus demonstrate fractional crystallization and multiple-phase (bcc, Laves AB2 , liquid) coexistence. Their
remarkable ability to build complex crystal structures from a polydisperse population originates from the
intermediate-range nature of interparticle forces, and it suggests routes for designing self-assembling
colloidal crystals from the bottom up.

DOI: 10.1103/PhysRevLett.116.208001

What is the preferred structure for a population of different types of crystals, which cooperate to make the best
colloidal particles, dispersed in liquid? This simple ques- use of the whole population.
tion has been satisfactorily answered only in the case of The colloids that we have used are industrially produced.
spherical particles that are effectively monodisperse in size They consist of nanometric silica particles, dispersed in
[1–6]. As the volume fraction of particles increases, there is water (Ludox HS40). The particles are roughly spherical,
a well-defined transition from a liquid to a crystal state. with an average radius of 8 nm and a size polydispersity
Two types of structures can be found: close-packed and of 0.14 [13,14]. We used near-equilibrium dialysis to
body-centered-cubic crystals; the preferred form depends equilibrate them against NaCl solutions (5 mM, pH 9.5).
on the range of the interparticle forces [5–7]. They were then slowly concentrated by the addition
Polydisperse populations present a tougher problem. of poly(ethylene glycol) to the solution outside the dialysis
In one limit, for particles that interact as hard spheres, membranes, as in Refs. [14,15]. Under these conditions,
crystalline order is destroyed by even small amounts of the particles repel each other via a screened electrostatic
polydispersity [3,4,8–10]. Charged particles interact interaction, with an effective Debye length of 2.5–4.5 nm,
instead via soft potentials and are more tolerant of poly- depending on their volume fraction ϕ. Further details of
dispersity, especially where they have an effectively narrow our methods and the dispersion properties (e.g., charge,
size distribution, due to long-range interactions. In this equation of state, density) are given in the Supplemental
other limit, a crystal state can be retained at low volume Material [16].
fractions, regardless of significant size polydispersity, if the Samples were characterized through small-angle x-ray
interaction polydispersity remains low [6,11,12]. Between scattering (SAXS), using ID02 at the ESRF. The strength of
these two limits is a vast region of phase space where ordering in a colloidal dispersion can be evaluated by the
we do not know whether homogeneous crystallization or height, Smax , of the main peak of its effective structure
fractionated crystallization is possible. factor SðqÞ, for scattering vector q [30–32]. SðqÞ was found
Here, we address the self-organization of polydisperse by dividing the radially averaged scattering intensity IðqÞ
populations of particles that interact through forces with an by the form factor of a dilute (ϕ ¼ 10−3 ) dispersion, and
intermediate range, comparable to the variations in particle normalizing at high q, as in Refs. [14,32–34]. For low ϕ,
size. Using high-resolution scattering methods, we find that these SðqÞ had a broad main peak, indicative of disordered
such populations can evolve through fractionated crystal- liquid arrangements of particles [Fig. 1(a)]. Indeed, all of
lization to yield coexisting crystals with different struc- these samples also behaved rheologically as fluids. The
tures. These crystals can have large, complex unit cells with value of Smax (Table I) rose slowly with increasing ϕ, from
specific sites for particles of different sizes. To explain this 1.2 at ϕ ¼ 0.04, to 2.6  0.1 at ϕ ¼ 0.16. Despite our
result, we use numerical simulations to demonstrate how a polydispersity, which should lower Smax slightly [32,33]
broad distribution of particles can split spontaneously into and add a low-q incoherent scattering [33,34], these values

0031-9007=16=116(20)=208001(5) 208001-1 © 2016 American Physical Society

 week ending
PRL 116, 208001 (2016) PHYSICAL REVIEW LETTERS 20 MAY 2016

2.85 agree well with the Hayter-Penfold mean spherical approxi-

0
mation (MSA) model [35] of monodisperse Yukawa
spheres (Table I, using 8 nm particles with 5 mM salt
S(q)
10
and a surface charge of 170e).
1
(a) At ϕ ¼ 0.19 and 0.21, we found that the 2D interference
10
0 0.2 0.4 0.6 0.8 patterns of our dispersions also contained sharp diffraction
(b) spots, superimposed on the liquidlike scattering ring.
1 The spots are the powder-diffraction pattern of small
10
crystallites. Here, any fractionation between the liquid
I(q)q 2

and crystals would invalidate the decomposition of IðqÞ

10
0
0.22 into a form factor and effective structure factor. Instead, we
calculated the complex structure factor F2 ∼ IðqÞq2, which
0.24 does not require knowledge of the form factors of each
10
1
phase. The positions of the peaks of FðqÞ, as well as
0 0.2 0.4 1 ) 0.6 0.8
q (nm systematic extinctions (h þ k þ l odd), indicated that they
originated from colloidal crystals with a body-centered-
cubic (bcc) structure [see Fig. 1]. This is in empirical
agreement with liquid state theory, where, according to
Verlet and Hansen [30,31], the liquid state with short-range
order is unstable with respect to a crystalline structure
when Smax > 2.85. However, our dispersions were quite
(c) (d)
polydisperse, while the Verlet-Hansen criterion is strictly
true only for monodisperse populations. Our observations
FIG. 1. (a) Effective structure factors. At low volume fractions, suggest a possible reason why this agreement may still
ϕ, the dispersion has a liquid structure, with broad peaks. When hold. It involves growing the bcc crystals from a narrow
the intensity of the liquid peak would exceeded 2.85, bcc
subset (i.e., an effectively monodisperse set) of the original
colloidal crystals appear alongside the liquid phase. (b) At higher
ϕ, the scattering spectra show many sharp peaks in addition to
population, and leaving the remaining particles in a liquid
the (indexed) bcc peaks. Their positions and relative intensities phase that coexists with these crystals.
correspond to crystals of a Laves MgZn2 phase, in coexistence As the dispersions were compressed to a higher ϕ,
with the bcc and liquid phases. (c) These diffraction patterns between 0.22 and 0.24, their scattering spectra became
consist of spots arranged in rings and imply the existence of many more complex. The interference patterns of these disper-
micron-sized crystallites, which (d) can be seen directly by sions revealed a large number of spotty rings [Fig. 1(c)].
microscopy. Typically, hundreds of spots were seen, whose diameters,
δ ≃ 0.003 nm−1 , imply the presence of many crystallites
with a size of at least π=δ ¼ 1 μm. Microscope images
[Fig. 1(d)] of such dispersions confirm the presence of
TABLE I. Sample summary, showing the volume fraction ϕ stable, free-floating crystals.
(0.005), the intensity of the liquid peak Smax and its predicted In these spectra we detected, after radial averaging, a
value (MSA) using Ref. [35], and the observed phases. broad liquid peak, peaks from the bcc phase, and up to 14
additional well-resolved peaks, including a triplet at low q,
ϕ Smax (liq.) Smax (MSA) Phases implying the presence of a crystal phase with a large unit cell.
0.038 1.2 1.33 liquid The new peaks can all be indexed (see Table II and the
0.046 1.4 1.40 liquid Supplemental Material [16]) to the powder spectrum of a
0.057 1.5 1.50 liquid crystalline phase of compact hexagonal (P63 =mmc) sym-
pffiffiffiffiffiffiffi
ffi
0.067 1.6 1.58 liquid metry, with lattice constants a ¼ 43.58 nm and c ¼ 8=3a,
0.079 2.2 1.69 liquid pffiffiffi
0.085 1.8 1.74 liquid and a unit-cell volume of 2a3 ¼1.17×105 nm3 , in the
0.128 2.2 2.12 liquid ϕ ¼ 0.235 sample. In the same sample, the bcc peaks were
0.131 2.1 2.15 liquid indexed to a unit cell with lattice constant abcc ¼ 27.11 nm
0.159 2.7 2.42 liquid and volume a3bcc ¼ 1.99 × 104 nm3 . The unit-cell volume
0.161 2.5 2.44 liquid of the new phase is therefore 5.9 times larger than that of
0.188  2.72 liquid, bcc the coexisting bcc phase, which contains two particle sites
0.207  2.94 liquid, bcc per cell. Assuming that the number density of the sites is
0.219  3.08 liquid, bcc, Laves comparable in both phases—which, in conditions of close
0.235  3.28 liquid, bcc, Laves
equilibrium and not too large fractionation, is reasonable—
0.240  3.35 liquid, bcc, Laves
one finds that the new phase has 12 particles per unit cell.

208001-2
 week ending
PRL 116, 208001 (2016) PHYSICAL REVIEW LETTERS 20 MAY 2016

TABLE II. Positions and relative scattering intensities of the

observed and fitted diffraction peaks of the Laves phase, for
ϕ ¼ 0.235. F is a complex structure factor corrected for the
multiplicity of the peaks, m, and the averaging of the powder-
diffraction pattern; zero indicates a systematic extinction, while
not obs. indicates that a line was not observed experimentally.

h k l m qexp (nm−1 ) qfit (nm−1 ) Fexp Ffit

FIG. 2. (a) Unit cell of the Laves MgZn2 phase. (b) Larger-than-
0 0 1 2 not obs. 0.0883 not obs. 0 average particles (yellow) occupy central sites and are surrounded
1 0 0 6 0.1667 0.1665 12.9 12.9 by rings of smaller-than-average particles (blue).
0 0 2 2 0.1769 0.1766 26.9 21.6
1 0 1 12 0.1885 0.1884 9.0 12.0 have shown, however, is that similar phases also naturally
1 0 2 12 0.2431 0.2427 18.1 21.9 arise in the solidification of broad and continuous pop-
0 0 3 2 not obs. 0.2648 not obs. 0
ulations of nanoparticles.
1 1 0 6 0.2891 0.2883 84.3 84.3
1 1 1 12 not obs. 0.3015 not obs. 0 An explanation for the coexistence of different crystal
1 0 3 12 0.3132 0.3128 98.2 79.4 types, each composed of a subset of particle radii, can be
2 0 0 6 0.3329 0.3329 52.4 42.3 made by seeking the equilibrium phases of the particle
1 1 2 12 0.3378 0.3381 73.7 87.6 population. To this end, we investigated the fractionation of
2 0 1 12 0.3441 0.3444 57.5 76.2 polydisperse charged particles through Gibbs-ensemble
0 0 4 2 0.3530 0.3531 86.2 84.8 Monte Carlo numerical simulations [44] of a combination
2 0 2 12 0.3767 0.3768 29.1 32.1
of a Laves MgZn2 phase and a bcc phase, with a fcc phase
1 0 4 12 0.3903 0.3904 25.8 25.8
1 1 3 12 not obs. 0.3915 not obs. 0 added as a control. The model is similar to that used in
2 0 3 12 0.4256 0.4254 19.6 18.2 Ref. [45]. Each phase was treated as an isolated volume
2 1 0 12 0.4402 0.4404 8.4 10.0 (avoiding grain boundaries), but particles could move
randomly between sites within each phase, and between
One can reasonably expect that this phase is constituted phases, according to a Monte Carlo Metropolis algorithm at
by a mixture of nanoparticles with distinct mean diameters. room temperature [46]. Although, for simplicity of dem-
Among the varied options [36], only one is of the compact onstration, no colloidal liquid was modeled, we would
hexagonal space group and contains 12 atoms per unit cell: expect such a phase to act as the medium of particle
the MgZn2 Laves phase. Here, four Mg atoms sit on the exchange and an acceptor of misfit particles. The propor-
four equivalent f Wyckoff positions, while eight Zn atoms tions of particles and the lattice constants of the three
are distributed on the six h and two a positions. This phases were allowed to vary with volume exchange
suggests that the new phase is composed of particles with between them, keeping the total volume constant.
two or three separate sizes organized into a Laves phase We considered a model of 22 466 particles with a
[37]. Within this hypothesis, the intensities of the Bragg Gaussian distribution of size r, an average radius of
peaks were fit with three free parameters corresponding to 8 nm, and a polydispersity of 0.14 [13], with a global
the radii ra , rf , and rh of particles at the a, f, and h sites ϕ ¼ 0.22. Interactions between particles were modeled as
(see the Supplemental Material [16]). The fit, the results of hard core plus Yukawa pair potentials, with an effective
which are shown in Table II, converges when rf ¼ 9.1  Debye length of κ−1 ¼ 2.8 nm and an effective surface
0.3 nm and ra ¼ rh ¼ 7.3  0.3 nm. The stoichiometry is charge density of 0.2e= nm2 (i.e., the charge on particle i
consequently AB2, with four large particles and eight small scales as r2i ). These parameters are estimated as in
particles per unit cell. The larger particles occupy relatively Ref. [25–27], accounting for modest charge renormaliza-
spacious truncated tetrahedron environments, where they tion, and they agree with the dispersion’s experimentally
are comfortably surrounded by rings of smaller particles determined equation of state [15,28].
in octahedral sites (Fig. 2). In contrast to repulsive Over time, the system evolved to find a configuration of
monodisperse crystals [2], the density of this Laves phase minimal Madelung energy, and the proportion of each
thus appears to be slightly lower (0.22) than that of the phase stabilized; Fig. 3 shows the final distribution of
coexisting liquid (0.235); this situation could relate to the particle sizes, according to phases and sites. It shows how
size selection of the individual sites. the coexistence of a Laves phase with the bcc phase is
Various AB2 phases are well known in binary mixtures of possible: the bcc phase uses the most populated part of
hard spheres [38–43]. For example, the AlB2 structure is a the distribution of particle sizes, near the center of the
preferred crystal phase for binary mixtures with a size ratio distribution. In this example, a small minority of particles
of the smaller to larger particles between approximately 0.4 was also taken into the fcc phase, although this phase
and 0.6 [39] and occurs in gem opals [40,41], while the disappears if a longer screening length (3 nm) is used. In
MgCu2 phase can be templated by walls [43]. What we either case, the remaining particles have a bimodal size

208001-3
 week ending
PRL 116, 208001 (2016) PHYSICAL REVIEW LETTERS 20 MAY 2016
1
Init. distribution colloidal crystals is performed with particles that interact as
F requency (norm.)
0.8 bcc hard spheres, and which crystallize when they are in close
0.6
Tetragonal to direct contact, at ϕ ∼ 0.5. When such particles have a
Octahedral
fcc
broad distribution of sizes, the unavoidable overlaps of any
0.4 large adjacent particles inhibit the formation of a structure
0.2 with long-range order [3,4,8,53], and dynamic arrest turns
the dispersion into a glass [2,8]. Our particles interact instead
0
4 6 8 10 12 through soft potentials. Assuming an effective Yukawa
Radius (nm) potential [25–27], the pair potential of two average-sized
particles reaches about 3 kT at a volume fraction of 20%,
FIG. 3. Monte Carlo simulations of the fractionation of a corresponding to a surface separation (for bcc) of 8 nm.
polydisperse colloidal dispersion into three preset crystalline
In this state, overlap of the particles themselves is still a
structures (bcc, fcc, Laves MgZn2 ). Shown are the final equi-
librium particle-size distributions in each phase. The vertical lines rare occurrence, determined by the frequency of very large
show the average radii extracted from the SAXS data for particles particles. These few “outliers” can easily be rejected away
in the bcc phase (black), and in the tetragonal (blue) and from the surfaces of growing crystals, as the soft potentials
octahedral (red) sites of the Laves phase. also keep the mobility of such particles high.
The width of the particle-size distribution and the range
of particle interactions together control the frequency of
distribution and thus fit efficiently into the differently such outliers, which are then available to build more diverse
shaped sites of the Laves phase. Exploring various param- structures. We consider three cases. If the interactions are
eter values, we found that these results were robust to long range (with an effective diameter ≫ a), then variations
doubling the charge density of the particles, or letting their in the particle size will be screened, and simple fcc or bcc
charge scale with r (as occurs for strong charge condensa- crystals are both expected and seen [1,6,11,12]. If the
tion [29]), but were sensitive to changes to the effective interaction range is intermediate—for example, κa ∼ 1—
screening length (between 2.2 and 3.0 nm). The model’s but the polydispersity σ is too high, then there will be too
average radii of 7.0, 8.2, and 9.6 nm, for particles at many overlaps to nucleate the first bcc crystals, and the
equilibrium in the Laves tetragonal, bcc, and Laves dispersion may remain in a liquid or glass phase. Inverting
octahedral sites, respectively, correspond well to the cor- Pusey’s criterion [8] suggests that this will be the case when
responding experimental values of 7.3, 8.3, and 9.1 nm.
ϕ ≥ c(1=ð1 þ σÞ)3 , where the order-1 constant c depends
We have thus described how a polydisperse population
on how tolerant a crystal is to overlaps. If, however, the
can split into coexisting phases of a colloidal liquid, a bcc
effects of the soft potential and the number of overlaps
crystal that preferentially selects the most abundant particle
are balanced against each other, as in this Letter,
sizes, and a Laves phase that accommodates the remaining
then fractionation is encouraged, and the phase space of
binary distribution of particles. This segregation by particle
polydisperse colloidal dispersions is opened.
size is known as fractionated crystallization; similar proc-
The behavior of such polydisperse nanometric dispersions
esses are known in molecular systems [47], including
points to directions that have not been explored thus far,
geochemistry [48]. For hard-sphere colloids fractionation
despite theoretical predictions [49,51–54,60]. We demon-
has been predicted beyond a terminal polydispersity of
strate here fractionated crystallization, with the coexistence
about 6% [49–53]. For medium-range Yukawa interactions
of at least three very different phases (liquid, bcc, and Laves),
(where κa is between 2.5 and 10), recent simulations [54]
and the formation of complex crystals that efficiently utilize
have suggested that a size polydispersity of 10%–15%,
the full size distribution. The link between the particle-size
comparable to ours, is required to hinder crystallization,
distribution and the structures also gives us a scheme for gen-
and thus to potentially trigger fractionation.
erating even more complex phases through the crystallization
Experimentally, the best prior evidence of colloidal
of populations of particles with broader size distributions,
fractionation is the work of van Megen and collaborators
provided that they interact through soft medium-range
[10,55,56], who invoke it to explain the nucleation proc-
potentials. The variety of structures waiting to be discovered
esses of colloidal crystals near a terminal polydispersity.
could be enormous, given that, within the limits defined
The coexistence of multiple solid phases is also known in
above, there exists a huge phase space of different size
cases of low-dimensional systems such as platelets [57] or
distributions and interaction potentials to explore.
particles confined to a plane [58]. Further evidence may
also be hiding in old data such as Fig. 13 of Ref. [59],
which appears to imply the presence of large-unit-cell
crystals in dispersions similar to ours (10.2 nm silica with *
bernard.cabane@espci.fr
9% size polydispersity). †
lucas.goehring@ds.mpg.de
The fractionation of particles in our experiments depends [1] A. Kose, M. Ozaki, K. Takano, Y. Kobayashi, and
on their intermediate range of interactions. Much work on S. Hachisu, J. Colloid Interface Sci. 44, 330 (1973).

208001-4
 week ending
PRL 116, 208001 (2016) PHYSICAL REVIEW LETTERS 20 MAY 2016

[2] P. N. Pusey and W. van Megen, Nature (London) 320, 340 [27] E. Trizac, L. Bocquet, M. Aubouy, and H. H. von Grünberg,
(1986). Langmuir 19, 4027 (2003).
[3] P. N. Pusey, E. Zaccarelli, C. Valeriani, E. Sanz, W. C. K. [28] J. Li, M. Turesson, C. A. Haglund, B. Cabane, and
Poon, and M. E. Cates, Phil. Trans. R. Soc. A 367, 4993 M. Skepö, Polymer 80, 205 (2015).
(2009). [29] N. Garbow, M. Evers, T. Palberg, and T. Okubo, J. Phys.
[4] W. C. K. Poon, in The Oxford Handbook of Soft Condensed Condens. Matter 16, 3835 (2004).
Matter, edited by E. M. Terentjev and D. A. Weitz (Oxford [30] L. Verlet, Phys. Rev. 165, 201 (1968).
University Press, New York, 2015), Chap. 1. [31] J.-P. Hansen and L. Verlet, Phys. Rev. 184, 151 (1969).
[5] Y. Monovoukas and A. P. Gast, J. Colloid Interface Sci. 128, [32] P. van Beurten and A. Vrij, J. Chem. Phys. 74, 2744 (1981).
533 (1989). [33] B. D’Aguanno and R. Klein, J. Chem. Soc., Faraday Trans.
[6] W. B. Russel, D. A. Saville, and W. R. Schowalter, Colloidal 87, 379 (1991).
Dispersions (Cambridge University Press, Cambridge, [34] J. S. Pedersen, J. Appl. Crystallogr. 27, 595 (1994).
England, 1989), p. 525. [35] J. B. Hayter and J. Penfold, Mol. Phys. 42, 109 (1981).
[7] A. Yethiraj and A. van Blaaderen, Nature (London) 421, 513 [36] L. Filion and M. Dijkstra, Phys. Rev. E 79, 046714 (2009).
(2003). [37] R. L. Berry and G. B. Raynor, Acta Crystallogr. 6, 178
[8] P. N. Pusey, J. Phys. (Paris) 48, 709 (1987). (1953).
[9] S. R. Williams, I. K. Snook, and W. van Megen, Phys. Rev. [38] P. Bartlett, R. H. Ottewill, and P. N. Pusey, Phys. Rev. Lett.
E 64, 021506 (2001). 68, 3801 (1992).
[10] H. J. Schöpe, G. Bryant, and W. van Megen, Phys. Rev. E [39] A. B. Schofield, P. N. Pusey, and P. Radcliffe, Phys. Rev. E
74, 060401 (2006). 72, 031407 (2005).
[11] M. E. Leunissen, A. van Blaaderen, A. D. Hollingsworth, [40] J. V. Sanders, Philos. Mag. A 42, 705 (1980).
M. T. Sullivan, and P. M. Chaikin, Proc. Natl. Acad. Sci. [41] M. J. Murray and J. V. Sanders, Philos. Mag. A 42, 721
U.S.A. 104, 2585 (2007). (1980).
[12] N. J. Lorenz, H. J. Schöpe, H. Reiber, T. Palberg, P. Wette, [42] E. V. Shevchenko, D. V. Talapin, C. B. Murray, and S.
I. Klassen, D. Holland-Moritz, D. Herlach, and T. Okubo, O’Brien, J. Am. Chem. Soc. 128, 3620 (2006).
J. Phys. Condens. Matter 21, 464116 (2009). [43] A.-P. Hynninen, J. H. J. Thijssen, E. C. M. Vermolen, M.
[13] V. Goertz, N. Dingenouts, and H. Nirschl, Part. Part. Syst. Dijkstra, and A. van Blaaderen, Nat. Mater. 6, 202 (2007).
Charact. 26, 17 (2009). [44] A. Z. Panagiotopoulos, N. Quirke, M. Stapleton, and D. J.
[14] J. Li, B. Cabane, M. Sztucki, J. Gummel, and L. Goehring, Tidesley, Mol. Phys. 63, 527 (1988).
Langmuir 28, 200 (2012). [45] R. Botet, B. Cabane, L. Goehring, J. Li, and F. Artzner,
[15] B. Jönsson, J. Persello, J. Li, and B. Cabane, Langmuir 27, Faraday Discuss. 186, 229 (2016).
6606 (2011). [46] D. Landau and K. Binder, A Guide to Monte Carlo
[16] See Supplemental Material at http://link.aps.org/ Simulation in Statistical Physics (Cambridge University
supplemental/10.1103/PhysRevLett.116.208001, which in- Press, Cambridge, England, 2000).
cludes Refs. [13–15,17–29], for an extended materials and [47] R. E. Timms, Eur. J. Lipid Sci. Technol. 107, 48 (2005).
methods section, and for details of the powder spectra [48] M. S. Ghiorso, Contrib. Mineral. Petrol. 90, 107 (1985).
analyses. [49] P. Bartlett, J. Chem. Phys. 109, 10970 (1998).
[17] K. Wong, B. Cabane, and R. Duplessix, J. Colloid Interface [50] P. Bartlett, Prog. Colloid Polym. Sci. 115, 137 (2000).
Sci. 123, 466 (1988). [51] M. Fasolo and P. Sollich, Phys. Rev. Lett. 91, 068301
[18] Colloidal Silica: Fundamentals and Applications, edited by (2003).
H. E. Bergna and W. O. Roberts (Taylor & Francis, London, [52] P. Sollich and N. B. Wilding, Phys. Rev. Lett. 104, 118302
2005), p. 895. (2010).
[19] The Chemistry of Silica: Solubility, Polymerization, Colloid [53] P. G. Bolhuis and D. A. Kofke, Phys. Rev. E 54, 634 (1996).
and Surface Properties and Biochemistry of Silica, edited [54] M. N. van der Linden, A. van Blaaderen, and M. Dijkstra,
by R. K. Iler (Wiley, New York, 1979), p. 896. J. Chem. Phys. 138, 114903 (2013).
[20] A. van Blaaderen and A. Vrij, J. Colloid Interface Sci. 156, [55] S. Martin, G. Bryant, and W. van Megen, Phys. Rev. E 67,
1 (1993). 061405 (2003).
[21] V. M. Masalov, N. S. Sukhinina, E. A. Kudrenko, and G. A. [56] H. J. Schöpe, G. Bryant, and W. van Megen, Phys. Rev. Lett.
Emelchenko, Nanotechnology 22, 275718 (2011). 96, 175701 (2006).
[22] F. Boulogne, L. Pauchard, F. Giorgiutti-Dauphiné, R. Botet, [57] D. V. Byelov, M. C. D. Mourad, I. Snigireva, A. Snigirev,
R. Schweins, M. Sztucki, J. Li, B. Cabane, and L. Goehring, A. V. Petukhov, and H. N. W. Lekkerkerker, Langmuir 26,
Europhys. Lett. 105, 38005 (2014). 6898 (2010).
[23] G. H. Bolt, J. Phys. Chem. 61, 1166 (1957). [58] N. Geerts, S. Jahn, and E. Eiser, J. Phys. Condens. Matter
[24] J. Persello, in Adsorption on Silica Surfaces, edited by 22, 104111 (2010).
E. Papirer (Marcel Dekker, New York, 2000), Chap. 10. [59] J. Chang, P. Lesieur, M. Delsanti, L. Belloni, C. Bonnet-
[25] S. Alexander, P. M. Chaikin, P. Grant, G. J. Morales, Gonnet, and B. Cabane, J. Phys. Chem. 99, 15993 (1995).
P. Pincus, and D. Hone, J. Chem. Phys. 80, 5776 (1984). [60] L. A. Fernández, V. Martin-Mayor, and P. Verrocchio, Phys.
[26] L. Belloni, Colloids Surf. A 140, 227 (1998). Rev. Lett. 98, 085702 (2007).

208001-5