Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Reversible microscale assembly of nanoparticles
driven by the phase transition of a thermotropic liquid
crystal
Niamh Mac Fhionnlaoich1, Stephen Schrettl2†, Nicholas B. Tito3, Ye Yang1,2, Malavika Nair2‡,
Luis A. Serrano1, Kellen Harkness2, Paulo Jacob Silva2, Holger Frauenrath2, W. Craig Carter4,
Francesco Stellacci2, Stefan Guldin1,2!
The arrangement of nanoscale building blocks into patterns with microscale periodicity is challenging to achieve via
self-assembly processes. Here, we report on the phase transition-driven collective assembly of gold nanoparticles
in a thermotropic liquid crystal. A temperature-induced transition from the isotropic to the nematic phase leads
to the assembly of individual nanometre-sized particles into arrays of micrometre-sized aggregates, whose size
and characteristic spacing can be tuned by varying the cooling rate. This fully reversible process offers hierarchical
control over structural order on the molecular, nanoscopic, and microscopic level and is an interesting model
system for the programmable patterning of nanocomposites with access to micrometre-sized periodicities.
Apart from their widespread use in display technolo-
gies, liquid crystals (LCs) have attracted considerable in-
terest in various fields of soft matter research.1–3 LCs
offer intriguing opportunities for colloid science.4–6 The
elastic distortion of the mesogen director field around
colloidal inclusions, i.e., topological defects, enables col-
loidal assembly into organized patterns by long-range in-
teractions.7Examples include nematic dipoles that can
lead to the formation of colloidal chains8, 9 and nematic
quadrupoles that enable the growth of 2D colloidal crys-
tals.9Furthermore, the mesogen director field may be
tailored through the particles’ shape.10 The interplay of
anisotropic elastic and weakly screened electrostatic in-
teractions opens up pathways to combined orientational
and positional order.11, 12 Colloid-induced topological de-
fects can act as templates for self-assembly processes on
the molecular scale.13 Such LC-mediated colloidal inter-
actions are size-dependent.14–16 Accordingly, the elastic
interaction with the director field should vanish when the
diameter of the colloidal particles is below the surface ex-
trapolation length, λ=K/W , where Kis the elastic con-
stant of the nematic LC and Wis the surface anchoring
strength, a measure of the interaction between the par-
ticle surface and the mesogen.16 Depending on the type
of mesogens and the nature of the colloidal surface, this
typically results in a minimum size of around 10 −100 nm
below which no elastic distortion of the director field is
expected.
Composites of LCs with nanoparticles (NPs) that are
below this length scale are nevertheless promising candi-
dates for novel optoelectronic applications, such as opti-
cal filters, metamaterials, polarizers or switches.17–24 The
large birefringence combined with versatile driving meth-
ods for switching make LC-NP composites an attractive
material system for reconfigurable active plasmonic de-
vices.25 The switching of such composites was recently
experimentally demonstrated for Ag and Au NPs where
structural changes of the LC-NP composite were induced
by a thermal stimulus and led to a tunable plasmonic re-
sponse.26, 27 A number of studies report on the successful
co-assembly of NPs by LCs into geometries that mimic the
mesogen arrangement.28–32
Solubilising such nanometer-sized objects in a LC is,
however, a known and widespread challenge.33 The dense
periodic packing of LC mesogens frustrates mixing and
free diffusion of additives.34 To-date, the most success-
ful strategies to solubilize NPs rely on creating a ligand
shell that resembles the mesogen environment, which is
typically achieved by a mixture of mesogen-like ligands
and shorter ligands that serve as spacers and thus cre-
ate supramolecular pockets for mesogen penetration.35, 36
This approach has made it possible to produce LC-NP
composites with NP loadings exceeding 50 wt%.37 At
significant NP loadings, the presence of non-nematogenic
species was found to have profound effects on the phase
behavior of nematic LCs, and several theoretical and ex-
perimental studies have established phase diagrams of LC-
NP composites.37–39 Exploitation of liquid crystal phase
transition processes offers access to structure formation
beyond the nanometer length scale.40–44 The presence of
an isotropic-nematic co-existence region above a critical
NP threshold enabled the formation of macroscopic NP-
rich networks.45 Recently, NP-based hollow microstruc-
tures were formed through a two-stage nematic nucleation
process.46 In a separate study, cooling of LC-NP compos-
ites allowed the spatial separation of quantum dot clusters;
however, control over the positional order was limited and
only achieved by the introduction of macroscopic beads as
nucleation points.47
Significance Statement: Hierarchical self-assembly
across multiple length scales is of common interest across
many scientific domains. Herein, we demonstrate how a
phase transition-driven process enables the collective and
reversible assembly of building blocks into a hierarchical
arrangements with a characteristic periodicity that is 4-5
orders of magnitude larger than the individual units.
1Department of Chemical Engineering, University College London, Torrington Place, London, WC1E 7JE, UK. 2Institute of
Materials, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland. 3Department of Applied Physics, Eindhoven
University of Technology, Eindhoven, The Netherlands. 4Department of Materials Science and Engineering, Massachusetts Institute
of Technology Cambridge, Massachusetts 02139, USA. †Now at: Adolphe Merkle Institute, Chemin des Verdiers, 1700 Fribourg, CH.
!e-mail: S.Guldin@ucl.ac.uk
In this work, we present how the temperature-induced
isotropic-to-nematic phase transition of the thermotropic
LC 4′-pentyl-biphenyl-4-carbonitrile (5CB) can drive a hi-
erarchical assembly of nanometer-sized gold particles into
micrometer-sized aggregates. This process is not only re-
versible but offers control over the characteristic size and
spacing of the resulting structures. We report on the dy-
namics of this process and correlate our findings to pa-
rameters such as the cooling rate, nanoparticle solubility,
phase separation kinetics, and the properties of the ne-
matic director field.
Results & discussion
Au NPs were made in a modular approach, which al-
lowed independent control over both the NP size and
ligand composition on the surface.48 First, oleylamine-
protected NPs were synthesized and these were sub-
sequently functionalized by ligand exchange with a
60:40 mol% mixture of 1-hexanethiol (HT) and 4′-
(12-mercaptododecyloxy)biphenyl-4-carbonitrile (MDD-
CBO), which led to an effective ligand composition on
the NP surface of 61:39 ±3 mol% according to analysis
by NMR spectroscopy. The Au NP size distribution was
4.7±0.7 nm according to TEM analysis. Another batch
of Au NPs with a size distribution of 6.0±1.0 nm and
comparable ligand composition was prepared that yielded
results similar to those reported here. The NPs were then
transferred to 5CB in the isotropic phase at a concen-
tration of 5 wt% following an established protocol using
dichloromethane as a volatile solvent.49 Subsequently, the
LC-NP composite was infiltrated into a LC cell with de-
fined gap thickness of 20 µm and homogeneous surface
alignment. Further details on the NP synthesis, cleaning
procedure, solubility, and material characterisation can
be found in the Supplementary Information (Supplemen-
tary Figs. S3-S9). Interestingly, the NPs showed supe-
rior solubility in isotropic 5CB compared to various com-
mon solvents, including chloroform, chlorobenzene, and
a 50/50 vol% mixture of acetonitrile and tetrahydrofuran
(Supplementary Fig. S10).
The dynamics of the LC phase transition and subse-
quent diffusion of NPs into the nematic phase are shown
in Fig. 1. The LC-NP composite was initially heated at
120s
270s
180s
1800s
150s
300s
210s
3600s
240s
7200s
0s
30s
10s 60s 90s
a
b c d
0 mins
120 mins
Figure 1 |Liquid crystal phase transition and dynamics of nanoparticle diffusion into the nematic liquid crystal. a, Kinetic
series of bright field microscopy images of liquid crystal - nanoparticle composite upon cooling to 34.3 ◦C (1 ◦C/min). Images were
acquired with a 20x objective. The scale bar represents 50 µm. b-d, Absorbance microspectroscopy was carried out during this
assembly process on a 25 µm-sized collection spot, represented by the dotted circle in a. In-situ absorbance spectra are plotted during
cooling at 1 ◦C/min from 38.5 to 34.3 ◦C (b) and then at a constant 34.3 ◦C with 5 min intervals (c). d,Absmax vs. time from 0 to
120 min, corresponding to the image series shown in a.
2
50 ◦C for two hours and subsequently cooled at a rate of
1◦C/min until a phase transition was observed, which oc-
curred at 34.3 ◦C. A series of images was then taken at
this temperature alongside microspectroscopic absorbance
measurements by partial out-coupling of light. By analy-
sis of the micrographs, the behavior observed upon cooling
may be divided into two stages. In the initial stage within
the first 60 s, a compartmentalisation occurred into NP-
depleted and NP-rich isotropic that resembled a spinodal-
type decomposition; over time, the characteristic continu-
ous pattern evolved to yield isolated NP aggregates at the
end of the process. Cross-polarized optical microscopy
revealed that the regions which appeared brighter under
bright field illumination in the initial decomposition phase
corresponded to nematic domains, while the dark regions
were isotropic (see Supplementary Figs. S11-S12).
The findings of in-situ absorbance microspectroscopy
are shown in Fig. 1b-d. Note that the absorbance of the
5CB mesogens within the 20 µm optical path is negligible
in both the nematic and isotropic state and absorbance
microspectroscopy hence offers a viable route to study the
concentration of NPs in this assembly process with a µm-
sized spatial resolution. Above the isotropic-to-nematic
phase transition temperature, gradual cooling did not re-
sult in significant changes in the maximum absorbance of
the LC-NP composite. However, once the phase transi-
tion was triggered at 34.3 ◦C, the maximum absorbance
(Absmax) decreased drastically from 1.21 to 0.86 in the
newly formed nematic domains (Fig. 1b). Absorbance
spectra were subsequently acquired at this temperature
in 5 min intervals. As shown in Fig. 1c and related mi-
croscopy images, a rapid increase in absorption to Absmax
= 1.05 within the first 5 min and a further increase to
Absmax = 1.15 was observed over time, which can be fitted
by a logistic growth function (Fig. 1d). Based on experi-
ments at concentrations below 1.5 wt%, where no residual
aggregates formed, a similar molar attenuation coefficient
was obtained for the LC-NP composite in the isotropic
and nematic phase, respectively (Supplementary Fig. 13).
The absorbance results are thus in line with a short-lived
depletion of NPs in the newly formed nematic domains
followed by subsequent diffusion of NPs from the enriched
receding isotropic into the nematic phase. After reaching
equilibrium conditions, the large majority of the NPs were
dispersed in the nematic phase with the residual fraction
of NPs embedded as aggregates in the LC-NP nematic
matrix.
To examine underlying effects, we developed a qualita-
tive coarse-grained molecular dynamics model of a single
NP complex with mixed ligand shell immersed in a host of
nematogenic mesogens, which is discoursed in detail in the
Supplementary Information (Supplementary Figs. S1-S2).
We find that while interdigitation of the ligand shell gener-
ally facilitates solvation of the NP complex, the solvation
free energy depends on whether the medium is isotropic
a
b
c
Figure 2 |Long-range nanoparticle aggregate formation. a,
Bright field transmission microscopy image of nematic liquid crys-
tal - nanoparticle composite at 28 ◦C that was cooled from the
isotropic phase at a rate of 5 ◦C/min. b, Size distribution of the
aggregates and c, radial distribution function of the aggregate
positions (center point) fitted with a damped sine function.
or nematic. In the nematic phase, solvation of an object
requires distorting the local order parameter, which intro-
duces a free energy penalty to the system. A solvation bar-
rier as driving force for a phase transition-driven collective
assembly may be conceptualized as follows. Cooling of the
LC-NP composite to the isotropic-to-nematic phase tran-
sition temperature Tiso−nem induces demixing and a short-
term phase co-existence of NP-enriched isotropic and NP-
depleted nematic regions. The build-up in NP concentra-
tion in the receding isotropic phase then forces additional
NPs to slowly diffuse into the nematic phase until, even-
tually, micrometer-sized aggregates of residual NPs are re-
tained.
Our findings are consistent with earlier studies on the
behavior of larger hard sphere colloids in thermotropic liq-
uid crystals. Petrov and Terentjev proposed a mechanism
by which particles smaller than the surface extrapolation
length, λ=K/W , form cellular networks upon cooling
from the isotropic to the nematic.50, 51 In the vicinity of
the weak first order phase transition, Kand Wapproach
0 at different rates. This would result in a decrease in
the ratio K/W and a corresponding increased elastic dis-
tortion due to the colloidal inclusions and expulsion of
the particles from the nematic domains. Poon, Terentjev
and coworkers showed that 100-1000nm PMMA particles
were expelled from nematic nuclei during cooling from the
isotropic phase, ultimately forming network-type struc-
3
tures with characteristic length scales.40, 41, 50, 51 Faster
cooling rates resulted in a greater number of nematic do-
mains which produced a more tightly packed network. Re-
cently, Abbott and coworkers demonstrated that such net-
works can be formed via a two step process involving an
initial spinodal decomposition of the colloidal dispersion in
the isotropic phase followed by the nucleation of nematic
domains.42
In Fig. 2a, a long-range bright transmission microscopy
image of a nematic composite after cooling at a rate of
5◦C/min to 28 ◦C is shown. In order to determine the
characteristic spacing between the aggregates and evalu-
ate ordering, the radial distribution function (RDF) was
calculated. Rastering acquisition with a 5X objective al-
lowed analysis of the entire active area of the LC cell
(10 ×10 mm), which contained between 1,500 and 60,000
aggregates depending on the cooling rate.52 We refer to
the Supplementary Information for a comprehensive com-
parison of the RDF with the nearest neighbor distance
(NND) approach, in particular on their robustness towards
small positional deviations (Supplementary Fig. S14). As
shown in Fig. 2c, the nanoscale building blocks formed a
pattern of µm-scale aggregates with characteristic inter-
aggregate distances of 68 ±2 µm, evidenced by a perio dic
fluctuation of the RDF that can be fitted with a modified
damped sine function.
We want to emphasize both the stability of the aggre-
gates upon formation and the reversibility of the process.
Once formed, the aggregates in the nematic phase were
found to be stable for months only to be redissolved by
heating to temperatures above the isotropic phase transi-
tion. To quantify the reversibility, i.e. whether the assem-
bly would fully dissolve upon heating to the isotropic state,
each image of the composite, comprising an area of approx-
imately 3000 µm2, was converted to a grayscale matrix.
The information entropy was then calculated and normal-
ized against the image area (see Figure 3b).53, 54 For a
perfectly homogenous system, each pixel value should be
equivalent, resulting in a maximum entropy per µm2of 1.
As variation increases, the entropy is reduced. Due to the
presence of the glass spacer beads and other minor aberra-
tions in the LC cell and optical path, the maximum cannot
be achieved but the reversibility over time can be tracked.
Prior to the phase transition, the entropy per µm2re-
mained relatively constant at an average of 0.979 µm−2.
At the onset of the nematic to isotropic transition, this
dropped drastically to 0.828 µm−2. With diffusion of the
AuNPs into the isotropic, the entropy per µm2increased
towards 1 following a logistic trend, demonstrating homo-
geneity.
In order to capture the phenomenology presented in Fig.
1, we modeled the morphological evolution of the phase
transition and NP segregation with coupled Allen-Cahn
and Cahn-Hilliard equations. The Allen-Cahn dynamics
were chosen for non-conserved order parameters; in this
34.0 oC 35.0 oC 36.0 oC
38.0 oC
40.0 oC 10 mins 20 mins
40 mins 60 mins
37.0 oC
39.0 oC
a
b
30 mins
Figure 3 |Reversibility of aggregate array formation. a
A series of kinetic bright field microscopy images of the com-
posite heated at 1 ◦/min through the phase transition to 40 ◦
and held for 60 mins. bThe information entropy was used
to evaluate the homogeneity of the composite as it is heated
and maintained in the isotropic phase.
case, the magnitude of the LC director field, η(#x), was
taken to be zero in the isotropic region and one in the com-
pletely ordered LC phase. Boundaries that divide one LC
orientation from another were not modeled. The Cahn-
Hilliard dynamics were used for conserved order parame-
ters; in this case the composition, c(#x) which was taken
to be one when the system was pure mesogen and zero
when the system was pure NP. Cahn-Hilliard dynamics
allowed c(#x) to vary while preserving the system’s aver-
age composition The coupling of c(#x) and η(#x) occured
via a phenomenological free-energy function, f(c, η) such
as that illustrated in Fig. 4g where composition and or-
der are plotted along the horizontal- and vertical-axes and
4
Order
Composition
1
0
01
Figure 4 |Simulation of NP segregation and aggregate order-
ing. a-f Sequential snapshots of the evolving microstructure. The
black-and-white contrast corresponds to the composition, where
black/white indicates high/low NP concentration. The pink color
indicates the degree of order and is scaled to accentuate contrast
for more disordered states. gMorphology evolution superimposed
on a free-energy landscape, where the horizontal and vertical axes
are NP composition and degree of order, respectively. The free-
energy landscape was empirically chosen so that minima are lo-
cated at (i) low-NP composition and low-order and (ii) high NP
composition and high-order. hRadial distribution functions ob-
tained by the simulation and an example experiment.
the free-energy density is represented by isocontours. The
locations of the maxima, minima, and saddle points of
f(c, η) were chosen to be consistent with the observations
presented above as described in Fig. 4’s caption.
The equilibrium compositions are given by the com-
mon tangent construction on the projection of the free
energy surface f(c, η) onto a constant η-plane–or, equiv-
alently the common tangent construction on the curve
f(c)≡minηf(c, η)).55 The system will spontaneously
decompose into NP-rich and NP-depleted regions at any
composition and order parameter where at least one of the
principle curvatures of the surface f(c, η) is negative–this
is spinodal decomposition. The “direction” of decompo-
sition in the c-ηplane (i.e., the orientation of the grow-
ing curve in Fig. 4) is restricted to those directions that
have a negative curvature; the specific direction depends
on that curvature and the ratio of the two time constants
(i.e., if the time constant for LC ordering is fast, then the
developing curve in Fig. 4 will tend towards being ver-
tical but moderated by the curvature in that direction).
If sufficiently large order and composition fluctuations are
present, then the system will continue to decompose at any
composition that lies above the common tangent–that is,
some initial spinodal decomposition can produce a surface
that acts as a supercritical nucleus.
The simulations illustrated in Fig. 4 can be interpreted
as follows. At t= 0, the disordered (η≈0) system be-
gins to decompose into slightly NP-enriched (towards the
right) and depleted compositions. The regions that have
depleted compositions experience a larger driving force to
order and therefore the decomposition produces a con-
comitant order-disorder transition with the more isotropic
regions being slightly enriched. This initial decomposi-
tion can be observed in Fig. 4(a-b) and has the morphol-
ogy of a spinodal decomposition. The system’s equilib-
rium compositions and phase fractions are determined by
the common tangent construction described above; there-
fore the final state of the system will be mostly the NP-
depleted LC phase and a small amount of isotropic NP-
rich phase. Thus, the final state cannot have the both-
phase-percolating (e.g., “brain-like”) morphology that is
commonly associated with a XA=XB= 1/2 compo-
sition for the regular solution model. Instead, the ini-
tial fluctuations must break the minor phase into isolated
regions–this produces a structure that is more commonly
associated with nucleation and growth; but in this case,
the structure is initiated by a decomposition reaction. The
shape of the developing curve gives co-variation of compo-
sition and order across the interface. Most of the material
is located at the ends of the developing curve. The curve
only represents the interface and so most of the visual data
in Fig. 4 represent a small part of the microstructure.
The developing curve should evolve toward a fixed curve
that minimizes the surface tension in the system. Presum-
ably this minimizing curve should pass near the saddle
point of f(c, η) but is also influenced by the square gra-
dient coefficients. The full analysis of the direction of ini-
tial composition and the equilibrium interface is beyond
the scope of this paper. The initial decomposition pro-
duces a large amount of interface. At small times, points
in Fig. 4 come from a multitude of interfaces. At larger
times, the system coarsens and removes many of the in-
terfaces. It is possible to construct a free-energy function
for which the final microstructure is LC with solubilized
NPs–and which undergoes an initial decomposition reac-
tion. In other words, if the free energy function has regions
of negative curvature for small values of η, it is possible for
the system to undergo a transient spinodal composition.
As the system continues to order, the NP-rich regions will
dissolve and the system will evolve towards its common-
tangent compositions.
The outlined characteristics of structure formation by
phase transition are significantly different to conventional
spinodal decomposition-based processes that exhibit a
conserved order parameter, such as the demixing of poly-
5
a
b
c
d
ef
gh
Figure 5 |Effect of cooling rate on order parameters. a-b,
Zoom-in of aggregates resulting from a cooling rate of 0.1 ◦C/min
and 20 ◦C/min, respectively. c-f, Radial distribution function and
corresponding damped sine fit for liquid crystal - nanoparticle com-
posites exposed to different cooling rates. f, Overview of charac-
teristic spacing as a function of cooling rate. g, Overview of mean
aggregate size as a function of cooling rate.
mer blends56, 57, sol-gel-polymer58 and fullerene-polymer
composites59 , polyelectrolyte multilayers or the electro-
chemical or liquid metal dealloying.60,61 The approach
presented here starts with a homogeneous mixture that
is fully isotropic. Cooling induces demixing with NP-
enriched isotropic and NP-depleted nematic regions. The
phase co-existence is only transient and thus, at the end of
the process, the composite displays the nematic phase with
a characteristic spacing and size of aggregates depending
on the cooling rate.
The effect of the cooling protocol on the aggregate or-
dering parameters is shown in Fig. 5. Six different cooling
rates, ranging from 0.1 ◦C/min to 20 ◦C/min were stud-
ied. Samples were cooled to 28 ◦C and left to equilibrate
for two hours. Between cycles, the LC-NP composite was
heated to 50 ◦C for two hours to erase the sample history,
and the process was repeated multiple times to evidence
full reversibility. The corresponding RDFs demonstrate
a non-linear decrease in inter-aggregate spacing with in-
creasing cooling rate. Below 1 ◦C/min, a small increase
in the cooling rate produced a significant reduction in the
average inter-aggregate distance. Higher cooling rates, in
contrast, exhibited a much diminished dependence. Like-
wise, the mean aggregate size was found to depend on
the cooling rate as shown in Fig. 5h and in the Supple-
mentary Fig. 15. Note that the increased noise for slower
cooling rates is related to the significantly reduced popula-
tion size considered in the RDF, because fewer aggregates
were formed and these were spaced further apart. In this
series of experiments, an absorbance of 1.36 was found for
the LC-NP composite in the isotropic state, which was re-
duced to values of 1.32 to 1.24 in the nematic state for
cooling rates ranging from 0.1 ◦C/min to 20 ◦C/min (Sup-
plementary Fig. 16). The higher Absmax value indicates
that lower cooling rates also led to an overall improved
transfer rate of Au NPs from the isotropic to the nematic
phase.
The dynamics of the assembly process for different cool-
ing rates was further studied by time-lapse microscopy.
Though the principles of aggregate formation were found
to be similar to the observations reported in Fig. 1, the
process differed not only in the characteristic length scale
but also in the characteristic time scale. While the for-
mation of aggregates was completed in 1.5 min for a cool-
ing rate of 20 ◦C/min, the whole process required around
100 min at 0.1 ◦C/min. A comparison of the spinodal-
type patterns before break-up of the percolation for the
receding NP-enriched isotropic phase is shown in the Sup-
plementary Fig. 17 for different cooling rates. The ob-
served characteristic length scale is very closely in line
with the characteristic aggregate spacing after completion
of the isotropic-to-nematic phase transition, demonstrat-
ing that the NP aggregates are indeed a remnant of the
initial phase separation and can therefore be controlled
through parameters that govern this process.62
The concentration of NPs in the isotropic phase played
an important role in the overall aggregate formation. As
shown in the Supplementary Fig. 18, reducing the NP con-
centration down to 2.5wt% led to a decrease in the mean
aggregate size and a less pronounced peak in the RDF.
At concentrations below 1.5 wt%, no µm-sized aggregates
were observed in the final nematic composite, most likely
due to the fact that the transient local increase in NP
concentration in the receding isotropic phase was even-
tually consumed by NP diffusion into the nematic phase.
Furthermore, the extent of the diffusion into the bulk de-
pended on the initial concentration. For 5 wt% and a cool-
ing rate of 0.5 ◦C/min, the bulk concentration was found
to be 4.7 wt%. At an initial concentration of 4.5 wt%, this
dropped to 4.3 wt%. This relationship predicted that all
aggregates should vanish at an AuNP content of 2.2wt%
which agreed well with the experimental results. Note that
6
a
b
c
d
Director, n
Director, n
Figure 6 |2D orientation of nanoparticle aggregates in the
liquid crystal - nanoparticle composite. 2D histogram of the
aggregate spatial distribution for a cooling rate of 0.5 ◦C/min in
x-y coordinates (a) and polar coordinates (b), respectively. c-d,
2D histogram of the aggregate spatial distribution for a cooling
rate of 5 ◦C/min.
doping of the LC with Au NPs also led to a small decrease
in the observed transition temperature from 34.8 ◦C for
0 wt% to 34.3 ◦C for 5 wt%, which is in line with previous
studies that reported on a NP-induced dilution effect.37, 39
Further details on the effect of NP concentration and ap-
plied cooling rate on the observed phase transition tem-
perature can be found in Supplementary Figs. S19-S20.
The most important consequence of this process is the
hierarchical organisation of NPs into larger assemblies
within a nematic composite. While the individual 5 nm-
sized NPs are too small to be recognized by the LC director
field, the larger NP aggregates should indeed be subject
to elastic distortion.16 The RDF demonstrates positional
order of aggregates over the full radial profile and does,
therefore, not account for a possible anisotropy due the
nematic phase and alignment along a common director.
In Fig. 6 2D histograms of the spatial distribution are
shown to represent the aggregate spacing in x-y and po-
lar coordinates for 0.5 ◦C/min and 5 ◦C/min, respectively.
Details of the applied methodology,63 the experimental
results for other cooling rates as well as a simulation of
the effect of disorder on the resulting 2D histograms can
be found in Supplementary Figs. S21-S22. At 0.5◦C/min
(Fig. 6a), i.e. for larger aggregates that are further spaced
from each other, the aggregate depletion zone (the region
in which the likelihood of finding another aggregate is be-
low the mean) exhibited an ellipsoidal shape and thus a
clear anisotropy. In order to measure the eccentricity of
this behavior, an angular window was chosen that allowed
sufficient statistical evidence for the determination of the
characteristic spacing by a segmented radial profile, here
10 ◦. The segmented radial profile along the director field
was determined and compared with the segmented radial
profile of the respective angular window perpendicular to
the director field. Based on the characteristic spacing ob-
tained in both directions, an eccentricity of 0.71±0.02 was
found. The spatial modulation as a function of the rela-
tive orientation to the mesogen director field is directly
evident in the polar coordinate plot shown in Fig. 6b.
At higher cooling rates and thus smaller aggregates with
smaller spacing (5 ◦C/min to 20 ◦C/min), an anisotropy
in the depletion zone was not observed. Instead, an in-
creased probability of finding aggregates along the director
field became evident, observable in Fig. 6c as a yellow line
along the Y= 0 direction and in Fig. 6d as an increased
probability around the 180 ◦window. A segmented radial
profile over an angular window of 1 ◦in the direction of
director field was compared to that around 45 ◦and 90 ◦
with respect to the direction of the director field. Based on
the results obtained herein for a cooling rate of 5 ◦C/min,
the probability of finding another aggregate along the di-
rector field was 9 % higher than in any other direction. A
similar trend was observed for cooling rates of 10 ◦C/min
and 20 ◦C/min.
A common form of colloidal assembly in nematic LCs is
chain formation, typically associated to topological defects
that originate from the surface anchoring conditions of
mesogens.7The resulting elastic dipoles and quadrupoles
decay with distance as 1/l3and 1/l5, respectively.64,65
The herein described assembly process separates the ag-
gregates evenly over distances that are significantly larger
than the mean aggregate sizes. We determined for spac-
ing land diameter dthe dimensionless ratio to be in the
range of 8 < l/d < 18. At such separation distances, chain
formation through elastic multipoles is not expected. In-
deed, the formation of dimers was only a very rare event
and no evidence for trimers or longer chains was found.
Finally, we must emphasize the role of the homoge-
neous substrate, which imposes an in-plane alignment of
the mesogen director field. As shown in Supplementary
Fig. S23, we did not observe the formation of aggregates
with pronounced characteristic spacings when using LC
cells with homeotropic, i.e. vertical, alignment. This is
in line with a study by Reven and coworkers, who in-
vestigated NP ordering by 4′-Octyl-biphenyl-4-carbonitrile
(8CB).66 At the nematic-smectic A phase transition, NPs
were found to decorate the edge dislocation defect lines,
while no characteristic nearest neighbor distance was iden-
tified at the isotropic-nematic transition. Furthermore,
when we used cells without surface alignment, cooling of
the LC-NP composite led to the formation of NP-rich and
NP-depleted regions but these remained in co-existence
(see Supplementary Fig. S24), which is in line with earlier
findings.39, 45
7
Conclusions
In summary, we report on the reversible self-organized
collective assembly of gold NPs in a thermotropic LC.
Based on our observations, the phase-dependent solubil-
ity and the cooling temperature play a decisive role in
the formation of hierarchically structured LC-NP com-
posites comprising NP aggregates of controllable size and
characteristic spacing. While reduced solubility in the ne-
matic phase results in residual NPs to be expelled in ag-
gregates, their spacing depends on the cooling rate that
defines a characteristic length scale in the compartmen-
talisation through the imposed transient decomposition
process during phase transition. Once formed, the aggre-
gates were found to be stable for months and the pro-
cess remained reversible by heating above the nematic-
to-isotropic phase transition temperature. This observa-
tion of controlled and phase transition-driven reversible
assembly of gold NPs takes inspiration from other forms
of stimuli-directed self-assembly,67 where the stimuli may
be a redox reaction, solvent addition, pH change, or light
exposure.68, 69 The presented findings of such a controlled
assembly process may yield interesting opportunities to-
wards programmable composites for metamaterials,70 in
sensing and display applications. Furthermore, we antici-
pate this to be an interesting synthetic model system for
the further exploration of active soft matter.71–74
Methods
Ligand synthesis. The ligand synthesis was carried out following
an adapted literature procedure from Milette and coworkers.36
General Procedures: Unless otherwise noted, all reactions were
carried out in dried Schlenk glassware in an inert argon atmo-
sphere. Chromatography solvents were purchased as reagent grade
and distilled once prior to use. For reactions, dichloromethane
(DCM), methanol (MeOH), tetrahydrofuran (THF), and N,N-
dimethylformamide (DMF) were purchased dry over molecular
sieves from Acros Organics, and acetone was purchased dry from
Sigma-Aldrich. All reagents were commercially obtained and used
without further purification. 4’-Hydroxy-4-biphenylcarbonitrile
(99%) and hexamethyldisilthiane (97%) were purchased from
ABCR, 1,12-dibromododecane (98%) was purchased from TCI,
potassium thioacetate (98%) was purchased from Alfa Aesar,
and acetyl chloride (ACS reagent grade), anhydrous potassium
carbonate (99%), and tetrabutylammonium fluoride (1 M solution
in THF) were purchased from Acros Organics. TLC analyzes were
performed on TLC plates from Merck (Silica gel 60 F254). UV-light
(254 nm) or anisaldehyde staining was used for detection. Column
chromatography was conducted on Geduran silica gel Si 60 from
Merck (40-60 µm).
4’-(12-Bromododecyloxy)-4-biphenylcarbonitrile 1. 4’-
Hydroxy-4-biphenylcarbonitrile (4.38 g, 22.4 mmol) and 1,12-
dibromododecane (36.82 g, 112.2 mmol) were dissolved in dry
acetone (750 mL). Potassium carbonate (6.11g, 44.9 mmol) was
added and the mixture was heated to reflux for 10 h, after which
4’-hydroxy-4-biphenylcarbonitrile was consumed according to TLC
(DCM). After dilution with DCM (300 mL) the mixture was
washed twice with 1 mHCl and once with saturated NaCl solution.
The organic phase was dried over MgSO4, and concentrated in
vacuo. Column chromatography (silica gel; DCM/n-heptane 1:2)
afforded 4’-(12-bromododecyloxy)-4- biphenylcarbonitrile 1(8.69 g,
19.6 mmol, 88%) as a colorless powder. 1H NMR (400.13 MHz,
CDCl3): δ= 7.70-7.63 (m, 4H, PhH), 7.54-7.51 (m, 2H, PhH),
7.01-6.97 (m, 2H, PhH), 4.01 (t, J= 6.5 Hz, 2H, CH2OPh), 3.41 (t,
J= 6.9 Hz, 2H, CH2Br), 1.89-1.77 (m, 4H, 2 CH2), 1.51-1.29 (m,
16H, 8 CH2). 13C NMR (100.61 MHz, CDCl3): δ= 159.9, 145.4,
132.7, 131.4, 128.4, 127.2, 119.3, 115.2, 110.2 (8 PhC, 1 CN), 68.3
(CH2OR), 34.2 (CH2Br), 33.0, 29.7, 29.7, 29.6, 29.6, 29.5, 29.4,
28.9, 28.3, 26.2 (10 CH2).
S-(12-(4’-(4-Biphenylcarbonitrile)oxy)dodecyl)
ethanethioate 2. 4’-(12-Bromododecyloxy)-4-biphenylcarbonitrile
1(0.98 g, 2.2 mmol) was added to a dispersion of S-potassium
thioacetate (1.25 g, 11.0 mmol) in DMF (30 mL) and the mix-
ture was heated to 70 ◦C for 12 h. After dilution with DCM
(150 mL) the mixture was washed six times with 1 mHCl
and once with saturated NaCl solution. The organic phase
was dried over Na2SO4, and concentrated in vacuo. Column
chromatography (silica gel; DCM/n-heptane 5:1) afforded S-(12-
(4’-(4-biphenylcarbonitrile)oxy)dodecyl) ethanethioate 2(0.91 g,
2.1 mmol, 94%) as an off-white solid. 1H NMR (400.13 MHz,
CDCl3): δ= 7.70-7.63 (m, 2H, PhH), 7.54-7.51 (m, 2H, PhH),
7.00-6.98 (m, 2H, PhH), 4.00 (t, J= 6.6 Hz, 2H, CH2OPh), 2.86
(t, J= 7.4 Hz, 2H, CH2SAc), 2.32 (s, 3H, CH3), 1.81 (dt, J=
14.6, 6.7 Hz, 2H, CH2CH2OPh), 1.60-1.52 (m, 2H, CH2CH2SAc),
1.50-1.43 (m, 2H, 1 CH2), 1.35-1.28 (m, 14H, 7 CH2). 13 C NMR
(100.61 MHz, CDCl3): δ= 196.2 (CH3COS), 160.0, 145.4, 132.7,
131.4, 128.5, 127.2, 119.3, 115.2, 110.2 (8 PhC, 1 CN), 68.3
(CH2OR), 30.8 (CH3COS), 29.7, 29.7, 29.6, 29.5, 29.4, 29.3, 29.3,
29.0, 26.2 (11 CH2).
4’-(12-Mercaptododecyloxy)-4-biphenylcarbonitrile 3.
From S-(12-(4’-(4-Biphenylcarbonitrile)oxy) dodecyl) ethanethioate
2:S-(12-(4’-(4-Biphenylcarbo-nitrile)oxy) dodecyl) ethanethioate
2(440 mg, 1.0 mmol) was dissolved in dry DCM (7 mL) and
dry methanol (10 mL) was added. The mixture was stirred at
room temperature, acetyl chloride (0.5 mL, 7.0 mmol) was added
dropwise, and stirring was continued for 6 h. After dilution
with DCM (30 mL) the mixture was washed once with saturated
NH4Cl solution and once with saturated NaCl solution. The
organic phase was dried over Na2SO4, and concentrated in
vacuo. Column chromatography (silica gel; DCM) afforded 4’-(12-
mercaptododecyloxy)-4-biphenylcarbonitrile 3(183 g, 0.46 mmol,
46%) as an off-white solid.
From 4’-(12-Bromododecyloxy)-4-biphenylcarbonitrile 1: 4’-(12-
Bromododecyloxy)-4-bi-phenylcarbo-nitrile 1(5.00 g, 11.3 mmol)
was dissolved in dry THF (100mL) and the solution was cooled
to 0 ◦C. Hexamethyldisilathiane (2.85 mL, 13.56 mmol) was added
dropwise to the solution, followed by dropwise addition of tetra-
butylammonium fluoride (12.4 mL, 1.0 min THF, 12.43 mmol).
The reaction was allowed to warm up to room temperature and
stirred for 2 h. After dilution with DCM (200 mL) the mixture
was washed twice with saturated NH4Cl solution and once with
saturated NaCl solution. The organic phase was dried over Na2SO4,
and concentrated in vacuo. Column chromatography (silica gel;
DCM/n-heptane 1:1) afforded 4’-(12-mercaptododecyloxy)-4-
biphenylcarbonitrile 3(3.30 g, 8.3 mmol, 74%) as colorless solid.
1H NMR (400.13 MHz, CDCl3): δ= 7.70-7.63 (m, 4H, PhH),
7.54-7.51 (m, 2H, PhH), 7.01-6.97 (m, 2H, PhH), 4.00 (t, J=
6.5 Hz, 2H, CH2OR), 2.52 (q, J= 7.4 Hz, 2H, CH2SH), 1.81 (dt, J
= 14.6, 6 .7 Hz, 2H, C H2CH2OR), 1.64-1.57 (m, 2H, CH2), 1.51-1.43
(m, 2H, CH2) 1.37-1.28 (m, 14H, 7 CH2). 13C NMR (100.61 MHz,
CDCl3): δ= 160.0, 145.4, 132.7, 131.4, 128.5, 127.2, 119.3, 115.2,
110.2 (8 PhC, 1 CN), 68.3 (CH2OR), 34.2, 29.7, 29.6, 29.5, 29.4,
29.2, 28.5, 26.2, 24.8 (11 CH2).
Nanoparticle synthesis. Oleylamine AuNPs were synthesized
as follows.48 A precursor solution containing 10 ml n-o ctane
(Sigma Aldrich, puriss.), 10ml oleylamine (Acros, C18 80-90%)
and 0.25mmol HAuCl4.3H2O (Sigma Aldrich, 99.9+% metals ba-
sis) was prepared and stirred under inert atmosphere at a tempera-
ture of 15◦C, which was controlled with 0.1K precision. Separately,
0.25mmol of the reducing agent t-butylamine borane (Strem, 97%+)
was dissolved in a solvent mixture of 1 ml n-octane and 1 ml oley-
lamine. Subsequently, the reducing solution was injected quickly into
8
the precursor solution and left stirring at for 1 h. The oleylamine-
protected AuNPs were subsequently 2x washed in ethanol (Fluka,
HPLC grade) with a minimal amount of DCM (Carlo Erba, ACS
grade) and subsequently redispersed in DCM (Carlo Erba, ACS
grade). Ligand exchange was carried out by preparing a thiol so-
lution containing 15 ml DCM with 0.127 mmol of 1-hexanethiol
(Alfa Aesar, 97%+) and 0.085 mmol of MDD-CBO, i.e. a 60/40
mol% mixture. Subsequently, a solution with 25 mg oleylamine-
protected AuNPs and 5 ml DCM was added and left stirring for 24 h
at room temperature. AuNPs were cleaned by dispersing them in a
10/90 vol% mixture of DCM and acetone (Sigma Aldrich, puriss.)
and subsequent precipitation by ultracentrifugation (32,000rpm,
1 h). This step was repeated 3 times. A mixture of 10/10/80 vol%
tetrahydrofuran, acetonitrile (Carlo Erba, HPLC grade), and ace-
tone (Sigma Aldrich, puriss.) was used for two subsequent cleaning
cycles. We want to note that usual protocols such as the repeated
precipitation in a poor solvent (here: acetone) and subsequent cen-
trifugation (at 5,000 rpm for 10 min) as well as cleaning by a Soxhlet
extractor did not result in an NMR signal that is dominated by
surface-bound ligands.
Sample fabrication. AuNP-LC composites were fabricated ac-
cording to a protocol published by Qi et al..49 In short, thiol-
protected AuNPs were dissolved in DCM and mixed in the targeted
w% with the LC (5CB; 4-Cyano-4′-pentylbiphenyl, Synthon, 99,8 %).
The mixture was stirred, then sonicated for 1 min before the volatile
components were evaporated overnight at 60 ◦Cunder a stream of ni-
trogen (Eppendorf ThermoMixer C). Subsequently, the solution was
transferred to a vacuum oven (Heraeus Vacutherm), which was set
to 50 ◦Cfor 3 hours to remove any remaining traces of solvent. Once
complete, the sample was sealed and stored in the thermomixer at
40 ◦Cuntil use.
Subsequently, the AuNP-LC composite was infiltrated in a glass
sandwich with defined gap thickness and surface functionalisation.
Cells with a homogeneous or homotropic surface alignment and 4-
20 µmthickness were supplied by Instec Inc. Reference cells with-
out surface alignment were built with pre-cleaned microscope slides
where an initially 25 µmthick thermoplastic sealing film (DuPont
Surlyn, Meltonix 1170-25) served as spacer. Prior to infiltration, the
cell was placed on the temperature controlled stage and warmed to
40 ◦C. Once the cell had been heated, 10 µlof the AuNPLC compos-
ite was taken from the vial and slowly deposited on the cell near the
opening. Capillary action drew the composite into the cell. The cell
was left at 40 ◦Cfor at least 15 minutes to allow a uniform film to
form inside.
Optical microscopy. Optical microscopy was carried out in trans-
mission on an Olympus BX61. The following Objectives were used:
5x (UMPlanFI, NA 0.15), 10x (UMPlanFI, NA 0.30), 20x (UM-
PlanFI, NA 0.46). The samples were-temperature controlled by a
Peltier-driven hot stage (Linkam, PE120). The effective tempera-
ture within the LC cell was determined by a calibration run.
Microspectroscopy. Spectroscopic absorption measurements were
carried out on the BX61 with a 20x magnifying objective and the
in-built halogen lamp (100 W). The signal was collected through a
microspectroscopy port with a 200 µmfibre and an Ocean Optics QE
65000 spectrometer. This resulted in a collection spot size of around
25 µm.
Nuclear magnetic resonance (NMR) spectroscopy. NMR ex-
periments were carried out at 297.2 K on a Bruker Avance III
400 spectrometer at frequencies of 400.13 MHz for 1H nuclei and
100.62 MHz for 13 C nuclei or on a Bruker Avance 400 spectrometer
with a BBIz 5 mm probe at a frequency of 400.13 MHz for 1H nu-
clei. Spectra were calibrated to the residual solvent peak of CDCl3
(7.26 ppm 1H NMR; 77.16 ppm 13C NMR).75
Transmission electron microscopy (TEM). TEM was carried
out on a Philips/FEI CM12 with a LaB6source that was operated
at 120 kV accelerating voltage. Size analysis was carried out using
imageJ.
Free energy modeling with coupled Allen-Cahn and Cahn-
Hilliard equations The relevant equations are given by:76
∂c
∂t=∇·Dc∇(∂f(c, η)
∂c−%c∇2c)
∂η
∂t=Dη(%η∇2η−∂f(c, η)
∂η )
(1)
where the composition (c) and order (η) are understood to be time-
dependent fields: c(&x, t), η(&x, t). Dcand Dηset the time scales for
the diffusion of NP and the rate of ordering of the mesogens and
are assumed to be constant in the simulations. Interfaces separat-
ing two regions of differing compositions are associated with large
|∇c|and the width of that interface scales with !%c/fmax (%cis the
square-gradient coefficient) where fmax is maximum value of the free
energy f(c, η) in the interfacial region. The interfacial tension scales
as !%cfmax. Scaling for the order-disorder interface is the same
but with subscripts changed. In the simulation, an initial uniform
composition and order were chosen as initial conditions to which a
small amount of noise was added. The two equations 1 were then up-
dated in a leap-frog process: one field is updated with its own value
at −∆tand the other’s value at −∆t/2. A semi-implicit spectral
method was employed.77
References
1. Lagerwall, J. P. F. & Scalia, G. A new era for liquid crystal
research: Applications of liquid crystals in soft matter nano-
, bio- and microtechnology. Curr. Appl. Phys. 12, 1387–1412
(2012).
2. Wang, L. & Li, Q. Stimuli-Directing Self-Organized 3D Liquid-
Crystalline Nanostructures: From Materials Design to Photonic
Applications. Adv. Funct. Mater. 26, 10–28 (2016).
3. Kato, T., Uchida, J., Ichikawa, T. & Sakamoto, T. Func-
tional Liquid Crystals towards the Next Generation of Mate-
rials. Angewandte Chemie-International Edition 57, 4355–4371
(2018).
4. Blanc, C., Coursault, D. & Lacaze, E. Ordering nano- and
microparticles assemblies with liquid crystals. Liq. Cryst. Rev.
1, 83–109 (2013).
5. Smalyukh, I. I. Liquid Crystal Colloids. Annu. Rev. Condens.
Matter Phys. 9, 207–226 (2018).
6. Dierking, I. From colloids in liquid crystals to colloidal liquid
crystals 46, 2057–2074 (2019).
7. Tkalec, U. & Musevic, I. Topology of nematic liquid crystal
colloids confined to two dimensions. Soft Matter 9, 8140–8150
(2013).
8. Poulin, P., Stark, H., Lubensky, T. & Weitz, D. Novel col-
loidal interactions in anisotropic fluids. Science 275, 1770–1773
(1997).
9. Musevic, I., Skarabot, M., Tkalec, U., Ravnik, M. & Zumer,
S. Two-dimensional nematic colloidal crystals self-assembled by
topological defects. Science 313, 954–958 (2006).
10. Lapointe, C. P., Mason, T. G. & Smalyukh, I. I. Shape-
Controlled Colloidal Interactions in Nematic Liquid Crystals.
Science 326, 1083–1086 (2009).
11. Mundoor, H., Senyuk, B. & Smalyukh, I. I. Triclinic nematic
colloidal crystals from competing elastic and electrostatic inter-
actions. Science 352, 69–73 (2016).
12. van der Asdonk, P. & Kouwer, P. H. J. Liquid crystal templating
as an approach to spatially and temporally organise soft matter.
Chem. Soc. Rev. 46, 5935–5949 (2017).
13. Wang, X., Miller, D. S., Bukusoglu, E., de Pablo, J. J. & Abbott,
N. L. Topological defects in liquid crystals as templates for
molecular self-assembly. Nat. Mater. 15, 106+ (2016).
14. Skarabot, M. & Musevic, I. Direct observation of interaction of
nanoparticles in a nematic liquid crystal. Soft Matter 6, 5476–
5481 (2010).
9
15. Tomar, V., Roberts, T. F., Abbott, N. L., Hernandez-Ortiz, J. P.
& de Pablo, J. J. Liquid Crystal Mediated Interactions Between
Nanoparticles in a Nematic Phase. Langmuir 28, 6124–6131
(2012).
16. Ryzhkova, A. V. & Musevic, I. Particle size effects on nanocol-
loidal interactions in nematic liquid crystals. Phys. Rev. E 87
(2013).
17. Draper, M. et al. Self-assembly and shape morphology of liquid-
crystalline gold metamaterials. Adv. Funct. Mater. 21, 1260–
1278 (2011).
18. Podoliak, N., Bartczak, D., Buchnev, O., Kanaras, A. G. &
Kaczmarek, M. High optical non linearity of nematic liquid
crystals doped with gold nanoparticles. J. Phys. Chem. C 116,
12934–12939 (2012).
19. Dintinger, J. et al. A Self-Organized Anisotropic Liquid-Crystal
Plasmonic Metamaterial. Adv. Mater. 25, 1999–2004 (2013).
20. Choudhary, A., Singh, G. & Biradar, A. M. Advances in gold
nanoparticle-liquid crystal composites. Nanoscale 6, 7743–7756
(2014).
21. Quint, M. T. et al. All-optical switching of nematic liquid crystal
films driven by localized surface plasmons. Opt. Express 23,
6888–6895 (2015).
22. Zhang, Y., Liu, Q., Mundoor, H., Yuan, Y. & Smalyukh, I. I.
Metal Nanoparticle Dispersion, Alignment, and Assembly in Ne-
matic Liquid Crystals for Applications in Switchable Plasmonic
Color Filters and E-Polarizers. ACS Nano 9, 3097–3108 (2015).
23. Atorf, B. et al. Liquid Crystals and Precious Metal: from
Nanoparticle Dispersions to Functional Plasmonic Nanostruc-
tures. Liq. Cryst. 44, 1929–1947 (2017).
24. Gabinet, U. R. & Osuji, C. O. Optical materials and metamate-
rials from nanostructured soft matter. Nano Res. 12, 2172–2183
(2019).
25. Stratford, K., Henrich, O., Lintuvuori, J. S., Cates, M. E. &
Marenduzzo, D. Self-assembly of colloid-cholesteric composites
provides a possible route to switchable optical materials. Nat.
Commun. 5, 3954 (2014).
26. Lewandowski, W., Lojewska, T., Szustakiewicz, P.,
Mieczkowski, J. & Pociecha, D. Reversible switching of
structural and plasmonic properties of liquid-crystalline gold
nanoparticle assemblies. Nanoscale 8, 2656–2663 (2016).
27. Baginski, M., Szmurlo, A., Andruszkiewicz, A., Wojcik, M. &
Lewandowski, W. Dynamic self-assembly of nanoparticles using
thermotropic liquid crystals. Liq. Cryst. 43, 2391–2409 (2016).
28. Mitov, M., Portet, C., Bourgerette, C., Snoeck, E. & Verelst,
M. Long-range structuring of nanoparticles by mimicry of a
cholesteric liquid crystal. Nat. Mater. 1, 229–231 (2002).
29. Zeng, X. et al. 3D Ordered Gold Strings by Coating Nanopar-
ticles with Mesogens. Adv. Mater. 21, 1746–1750 (2009).
30. Gardner, D. F., Evans, J. S. & Smalyukh, I. I. Towards Recon-
figurable Optical Metamaterials: Colloidal Nanoparticle Self-
Assembly and Self-Alignment in Liquid Crystals. Mol. Cryst.
Liq. 545, 1227–1245 (2011).
31. Gharbi, M. A. et al. Reversible Nanoparticle Cubic Lattices in
Blue Phase Liquid Crystals. ACS Nano 10, 3410–3415 (2016).
32. Zhang, S., Pelligra, C. I., Feng, X. & Osuji, C. O. Directed
Assembly of Hybrid Nanomaterials and Nanocomposites. Adv.
Mater. 30, 1705794 (2018).
33. Araya, K. & Iwasaki, K. Solubility parameters of liquid crystals.
Mol. Cryst. Liq. Cryst. 392, 49–57 (2002).
34. Goodby, J. W., Mandle, R. J., Davis, E. J., Zhong, T. & Cowl-
ing, S. J. What makes a liquid crystal? The effect of free volume
on soft matter. Liq. Cryst. 42, 593–622 (2015).
35. Qi, H., Kinkead, B., Marx, V. M., Zhang, H. R. & Heg-
mann, T. Miscibility and Alignment Effects of Mixed Monolayer
Cyanobiphenyl Liquid-Crystal-Capped Gold Nanoparticles in
Nematic Cyanobiphenyl Liquid Crystal Hosts. Chemphyschem
10, 1211–1218 (2009).
36. Milette, J., Toader, V., Reven, L. & Lennox, R. B. Tuning the
miscibility of gold nanoparticles dispersed in liquid crystals via
the thiol-for-DMAP reaction. J. Mater. Chem. 21, 9043–9050
(2011).
37. Milette, J. et al. A Molecular and Thermodynamic View of
the Assembly of Gold Nanoparticles in Nematic Liquid Crystal.
Langmuir 29, 1258–1263 (2013).
38. Soule, E. R., Milette, J., Reven, L. & Rey, A. D. Phase equi-
librium and structure formation in gold nanoparticles-nematic
liquid crystal composites: experiments and theory. Soft Matter
8, 2860–2866 (2012).
39. Gorkunov, M. V. et al. Phase separation effects and the nematic-
isotropic transition in polymer and low molecular weight liquid
crystals doped with nanoparticles. Soft Matter 9, 3578–3588
(2013).
40. Meeker, S., Poon, W., Crain, J. & Terentjev, E. Colloid-liquid-
crystal composites: An unusual soft solid. Physical Review E
61, R6083–R6086 (2000).
41. Vollmer, D. et al. Formation of self-supporting reversible cellular
networks in suspensions of colloids and liquid crystals. Langmuir
21, 4921–4930 (2005).
42. Bukusoglu, E., Pal, S. K., de Pablo, J. J. & Abbott, N. L.
Colloid-in-liquid crystal gels formed via spinodal decomposition.
Soft Matter 10, 1602–1610 (2014).
43. Bartolo, M., Amaral, J. J., Hirst, L. S. & Ghosh, S. Directed
assembly of magnetic and semiconducting nanoparticles with
tunable and synergistic functionality. Sci. Rep. 9, 1–11 (2019).
44. Sudha, D. G., Ochoa, J. & Hirst, L. S. Colloidal aggregation
in anisotropic liquid crystal solvent. Soft Matter 17, 7532–7540
(2021).
45. Milette, J. et al. Reversible long range network formation in gold
nanoparticle - nematic liquid crystal composites. Soft Matter 8,
173–179 (2012).
46. Riahinasab, S. T. et al. Nanoparticle-based hollow microstruc-
tures formed by two-stage nematic nucleation and phase sepa-
ration. Nat. Commun. 10, 894 (2019).
47. Rodarte, A. L., Pandolfi, R. J., Ghosh, S. & Hirst, L. S. Quan-
tum dot/liquid crystal composite materials: self-assembly driven
by liquid crystal phase transition templating. J. Mater. Chem.
C1, 5527–5532 (2013).
48. Yang, Y., Serrano, L. A. & Guldin, S. A versatile aunp synthetic
platform for decoupled control of size and surface composition.
Langmuir 34, 6820–6826 (2018). PMID: 29768005.
49. Qi, H. & Hegmann, T. Formation of periodic stripe patterns in
nematic liquid crystals doped with functionalized gold nanopar-
ticles. J. Mater. Chem. 16, 4197–4205 (2006).
50. Petrov, P. & Terentjev, E. Formation of cellular solid in liquid
crystal colloids. Langmuir 17, 2942–2949 (2001).
51. Anderson, V., Terentjev, E., Meeker, S., Crain, J. & Poon, W.
Cellular solid behaviour of liquid crystal colloids - 1. Phase sep-
aration and morphology. Eur. Phys. J. E 4, 11–20 (2001).
52. Preibisch, S., Saalfeld, S. & Tomancak, P. Globally optimal
stitching of tiled 3D microscopic image acquisitions. Bioinfor-
matics (Oxford, England) 25, 1463–5 (2009).
53. Shannon, C. E. A Mathematical Theory of Communication.
Bell System Technical Journal 27, 379–423 (1948).
54. Mac Fhionnlaoich, N. & Guldin, S. Information entropy as a
reliable measure of nanoparticle dispersity. Chem. Mater. 32,
3701–3706 (2020).
55. Boettinger, W., Warren, J., Beckermann, C. & Karma, A.
Phase-field simulation of solidification. Annu. Rev. Mater. Res.
32, 163–194 (2002).
56. Bates, F. Polymer-Polymer Phase-Behavior. Science 251, 898–
905 (1991).
10
57. Walheim, S., Boltau, M., Mlynek, J., Krausch, G. & Steiner,
U. Structure formation via polymer demixing in spin-cast films.
Macromolecules 30, 4995–5003 (1997).
58. Nakanishi, K. Pore Structure Control of Silica Gels Based on
Phase Separation. J. Porous Mater. 4, 67–112 (1997).
59. Vaynzof, Y. et al. Surface-Directed Spinodal Decomposition
in Poly[3-hexylthiophene] and C-61-Butyric Acid Methyl Ester
Blends. ACS Nano 5, 329–336 (2011).
60. Erlebacher, J., Aziz, M., Karma, A., Dimitrov, N. & Sieradzki,
K. Evolution of nanoporosity in dealloying. Nature 410, 450–
453 (2001).
61. Geslin, P.-A., Mccue, I., Gaskey, B., Erlebacher, J. & Karma, A.
Topology-generating interfacial pattern formation during liquid
metal dealloying. Nat. Commun. 6, 8887 (2015).
62. Carter, W. C. & Johnson, W. C. The Selected Works of John
W. Cahn (John Wiley & Sons, Inc., 2013).
63. Mac Fhionnlaoich, N., Qi, R. & Guldin, S. Application of the
spatial distribution function to colloidal ordering. Langmuir 35,
16605–16611 (2019).
64. Lubensky, T., Pettey, D., Currier, N. & Stark, H. Topological
defects and interactions in nematic emulsions. Phys. Rev. E 57,
610–625 (1998).
65. Stark, H. Physics of colloidal dispersions in nematic liquid crys-
tals. Phys. Rep. 351, 387–474 (2001).
66. Milette, J. et al. Reversible long-range patterning of gold
nanoparticles by smectic liquid crystals. Soft Matter 8, 6593–
6598 (2012).
67. Grzelczak, M., Vermant, J., Furst, E. M. & Liz-Marzan, L. M.
Directed Self-Assembly of Nanoparticles. ACS Nano 4, 3591–
3605 (2010).
68. Klajn, R., Bishop, K. J. M. & Grzybowski, B. A. Light-
controlled self-assembly of reversible and irreversible nanoparti-
cle suprastructures. Proc. Natl. Acad. Sci. U.S.A. 104, 10305–
10309 (2007).
69. Montelongo, Y., Yetisen, A. K., Butt, H. & Yun, S.-H. Recon-
figurable optical assembly of nanostructures. Nat. Commun. 7,
12002 (2016).
70. Fan, J. A. et al. Self-Assembled Plasmonic Nanoparticle Clus-
ters. Science 328, 1135–1138 (2010).
71. Mann, S. Self-assembly and transformation of hybrid
nano-objects and nanostructures under equilibrium and non-
equilibrium conditions. Nat. Mater. 8, 781–792 (2009).
72. Marchetti, M. C. et al. Hydrodynamics of soft active matter.
Rev. Mod. Phys. 85, 1143–1189 (2013).
73. Menzel, A. M. Tuned, driven, and active soft matter. Phys.
Rep. 554, 1–45 (2015).
74. Lavrentovich, O. D. Active colloids in liquid crystals. Curr.
Opin. Colloid Interface Sci. 21, 97–109 (2016).
75. Fulmer, G. R. et al. NMR Chemical Shifts of Trace Impu-
rities: Common Laboratory Solvents, Organics, and Gases in
Deuterated Solvents Relevant to the Organometallic Chemist.
Organometallics 29, 2176–2179 (2010).
76. Balluffi, R., Allen, S. & Carter, W. Kinetics of Materials (Black-
well Science Publ, 2005).
77. Chen, L. Phase-field models for microstructure evolution. Annu.
Rev. Mater. Res. 32, 113–140 (2002).
Acknowledgements
NMF acknowledges funding by the EPSRC under a Doctoral Train-
ing Partnership (EP/M507970/1). YY is grateful to University
College London for support through an Overseas Research Scholar-
ship. LSG acknowledges funding from the European Union’s Horizon
2020 research and innovation programme under grant agreement No
633635 (DIACHEMO). NBT is grateful for financial support from
the 4TU.High-Tech Materials research programme ‘New Horizons in
designer materials’ (www.4tu.nl/htm), and for discussions on this
work with Paul van der Schoot, Andela Saric, Stefan Paquay, and
Wouter Ellenbroek. SG is grateful for support by the German Na-
tional Academy of Sciences Leopoldina, Fellowship LPDS2012-13
and by a start-up fund from University College London. The authors
thank Prof Ullrich Steiner for valuable feedback on the manuscript.
Author contributions
NMF and SG carried out the aggregate formation experiments and
analysed data. SS, LAS, and HF synthesised and characterised the
ligand and contributed to NP solvation aspects. YY and MN synthe-
sised the NPs. YY, KH, and PJS characterised the NPs. NBT car-
ried out molecular dynamics simulations. WCC modeled the phase
transition process. FS and SG designed the experiment. NMF, SS,
NBT, HF, FS, and SG wrote the manuscript. All authors discussed
the results and provided comments on the manuscript. Correspon-
dence and requests for materials should be addressed to SG.
11
