Methods
AgCl-KCl eutectic synthesis
As-received AgCl (99.999%, Sigma Aldrich) and KCl (99.99%, Sigma
Aldrich) were mixed as per the binary eutectic composition^34 of
81.77 wt% AgCl and 18.23 wt% KCl, and melted in a glass vial (covered
with aluminium foil to exclude light) on a hotplate at 470 °C for 1 h, and
subsequently cooled to room temperature.
Fabrication of pillar templates
Under dark-room conditions, SU-8 (MicroChem, SU-8 2000
series) photoresist was mixed with 0.5 wt% photo-initiator,
cyclopentadienyl(fluorene)iron(ii) hexafluorophosphate (Aldrich).
Indium tin oxide coated glass substrates (with a sheet resistance of
70–100 Ω; Delta Technologies Ltd) were cleaned with acetone and
isopropanol, followed by O 2 reactive ion etching (RIE) at 100 W for 5 min
(with an O 2 flow of 19.6 cm^3 min−1 at STP at 10 mtorr chamber pressure;
Plasma-Therm 790 MF RIE Plasma System). Immediately after RIE, the
SU-8 solution was spin-coated (2,000 r.p.m. for 30 s) on the substrates.
These SU8 layers were soft baked on hotplates first at 65 °C for 10 min
and then at 90 °C for 20 min, followed by slow cooling. A laser interfer-
ence lithography setup^35 (laser output of 5.0 W, frequency-doubled
Nd:YVO 4 laser, 532 nm) was used to create a hexagonal pattern in the
SU8 film, by focusing three identical beams (arranged at 120° to each
other) onto the sample plane for 0.8 s. The parameters of this pattern
(with a lattice constant of 760–810 nm, and diameter of 500–620 nm)
were adjusted by controlling the exposure time and the incident angles
of each beam. All the beams were linearly polarized in their incident
planes. After exposure, the samples were oven-baked at 85 °C for 20 min
in dry air followed by slow cooling. The post-baked samples were then
developed in propylene glycol monomethyl ether acetate for 4 h, fol-
lowed by rinsing in an isopropanol bath for 10 min. The samples were
gently blow dried with N 2 and then baked at 90 °C for 15 min. To cre-
ate the pillar template, Ni electrodeposition was carried out within the
SU8 scaffold. The electrodeposition was performed potentiostatically
at −1.8 V in a commercial sulfamate nickel plating solution (Transene,
SN-10). A drop of isopropanol was put on the scaffold before dipping it
in the electroplating solution to improve the infiltration of the solution
into the hydrophobic scaffold. By controlling the electrodeposition
time, we were able to control the pillar height. The SU8 scaffold was then
etched using O 2 plasma RIE (Nordson March RIE Plasma System) with
500 mtorr pressure and 200 W power for 15 min, thus exposing the Ni
pillars. Subsequently, the exposed pillars were surface-protected with
a conformal coating of ~25 nm of alumina using atomic layer deposition
(300 cycles of the standard recipe; Savannah S100 Cambridge Nanotech).
The usable spot size of the pillar template was about 2 mm in diameter.
Eutectic solidified in pillar templates
Directional solidification experiments were performed using two tech-
niques, a Linkam THMS 600 hot-stage (with a TMS 94 controller) that
sets a cooling rate, and a tube-furnace setup^23 where a syringe pump
draws the samples out of the heated furnace at fixed speeds. The rate
of solidification (controlled by either a set cooling rate or the draw
rate) was varied to obtain a certain λ of the eutectic (see Extended Data
Fig. 1a). With the hot-stage setup, the stage was first heated to 470 °C
and then the sample was placed on it. For the tube-furnace setup, the
temperature was first set to 550 °C, and then the samples were placed
at the location where the temperature reached 470 °C. A small piece
of the prepared eutectic (around 150 mg) was placed on top of the
pillar templates (see schematic in Extended Data Fig. 1b), and heated
to 470 °C for 8–10 min to enable complete infiltration of the molten
eutectic. Partial infilling was obtained with the dwell time less than
5 min. Subsequently, the infilled samples were directionally solidified at
various rates, such that the nominal direction of solidification was along
the axis of the pillars, until the sample cooled to room temperature.
Microstructure characterization
Directionally solidified solid eutectic samples were peeled off from the
substrate and their bottom surface was characterized (see schematic
in Extended Data Fig. 1b) by scanning electron microscopy (SEM) using
a Hitachi S4800-SEM. SEM images from all the samples are shown in
Extended Data Figs. 3 and 4. Note that we did not coat the samples dur-
ing the SEM visualization: to visualize the cross-section, the samples
were mechanically cleaved. After SEM imaging, the samples were stored
in an argon glove-box to minimize the corrosive reaction between the
chloride salts and the exposed Ni pillars.Image analysis
The average lamellar spacing values (from samples without pillar tem-
plates) were determined by performing the fast Fourier transform
function of ImageJ on the SEM images. For templated eutectic samples,
λ was determined outside the template region (see Extended Data
Fig. 1c), whereas g and the emergent patterns were determined from
SEM images taken around the centre of the pillar template samples.
Patterns were analysed using a custom MATLAB code based on ref.^36.
The SEM images were first false-coloured to match the colour scheme
of the simulated patterns, and the false-coloured SEM images were used
in further analysis. These images were divided into unit cells and these
unit cells were stacked together, followed by averaging the intensity
at each pixel (shown in Extended Data Fig. 5). Moreover, the deviation
between the simulated pattern and the SEM image was calculated for
each unit cell and mapped over the entire SEM image. The unit cells
having the lowest deviation are denoted in blue and those with the
greatest deviation in red. Cells where the template pillars appeared
damaged were not included in the analysis.Eutectic solidified in colloid template
Silica microspheres (Fibre Optic Center Inc.) were dispersed in etha-
nol (1% by weight). Using a vertical deposition method^37 , the 560 nm
diameter silica colloids were deposited as a monolayer (2D colloidal
crystal) and as multilayers (3D colloidal crystal) on (piranha cleaned)
silicon substrates in an incubator (Isotemp 125D, Fisher) set at 50 °C.
Multilayer regions were scratched off before the directional solidifica-
tion experiments. The eutectic was side-infilled (as demonstrated in
ref.^31 ) into the template, using a hotplate set at 470 °C, and subsequently
directionally solidified using the hot-stage setup. The patterns were
imaged with top-view SEM.Thermal profile simulations
The temperature profile in the eutectic during solidification was calcu-
lated using COMSOL in order to evaluate the assumptions regarding the
orientation of the solidification front. For the pillar template case, the
heat equation was solved in a two-dimensional axisymmetric domain
consisting of (from bottom to top) an aluminium plate (1 cm thick, 8 cm
in diameter, thermal conductivity (κ) = 238 W m−1 K−1), a glass substrate
(0.7 mm thick, 20 mm in diameter, κ = 1.38 W m−1 K−1), a pillar-eutectic
composite layer (6 μm thick, 2 mm in diameter, κ = 2 3.4 W m−1 K−1), and
a eutectic overlayer (a truncated hemisphere with a base diameter
of 5 mm and a height of 2 mm). Unless otherwise noted, the thermal
conductivity values were assumed from COMSOL’s default material
library. The thermal conductivity of the pillar-eutectic composite
(κ = 23.4 W m−1 K−1, as previously stated) was determined by weight-
ing the thermal conductivity of each material by its volume fraction—
namely, the Ni pillar (κ = 90.7 W m−1 K−1, taken from a pure Ni value) and
the eutectic (κ = 3.25 W m−1 K−1, based on the solid value^22 )—along with
an assumption of a volume fraction of 23% and 77%, respectively, which
is for the case when the pillar diameter and the edge gap are equal.
The thermal properties of the ~25 nm alumina layer on the surface of
the Ni pillars were not considered. Since the overlayer would solidify
before the composite layer, the overlayer is assumed to have the average