Mixtures of gold nanospheres and titanium dioxide nanorods formed different confined three-dimensional assemblies as the ratio between the two particle types changed. At a sphere-to-rod ratio of 5, microscopy showed colloidal membranes. Lowering the ratio to 0.6 produced binary smectic ordering, a layered arrangement confirmed by electron microscopy.
The work asked whether changing the sphere-to-rod number ratio could tune the internal phase and whether simulations would reproduce it. Experimental supraparticles were made in oil-in-water emulsion droplets, while molecular-dynamics simulations modelled sphere-rod mixtures.
How the particles were packed
The researchers placed the nanoparticle mixtures in oil-in-water droplets and used evaporation to drive the particles into close packing, producing micron-scale supraparticles. The prepared sphere-to-rod ratios were 0.05, 0.6 and 5.
The measured gold content followed that recipe. Energy-dispersive X-ray analysis found gold loadings of 1.1 ± 0.2 wt·% at a ratio of 0.05, 10.1 ± 1.3 wt·% at 0.6 and 40.6 ± 3.2 wt·% at 5. A ratio-5 supraparticle was about 2.4 μm across, with estimated contents of 3.8 million nanorods and 15 million nanospheres. Those particle counts were estimates based on interparticle spacing.
The experimental building blocks were ligand-capped brookite titanium dioxide nanorods. Their bare length was 49 ± 7 nm and their bare diameter was 3.5 ± 0.6 nm, with ligand-inclusive spacing of 2.7 ± 0.5 nm and an effective aspect ratio of 8.3. The gold spheres had an average diameter of 3.5 ± 0.4 nm.
The recipe changed the layers
The lowest gold content did not produce the cleanest version of the layered pattern. At a sphere-to-rod ratio of 0.05, the spaces between rod layers contained disordered titanium dioxide rods mixed with gold spheres rather than gaps filled only by gold spheres. The authors suggested that too few spheres or kinetic arrest, a structure becoming stuck before it fully reorganises, could explain the result.
As gold content increased, multiple layers of gold spheres appeared between titanium dioxide rod layers. Ratios between 5 and 10 disrupted and isolated the rod layers, and at a ratio of 10 the rods and spheres no longer mixed.
The researchers tracked the assembly process with polarised optical microscopy, or POM. Disordered supraparticles were optically isotropic, meaning their optical response did not depend on direction, while membrane and binary smectic structures showed birefringence. Brightness between crossed polarising filters served as a signal of internal ordering.
For the 0.05-ratio mixture, measurements of 300 droplets showed the mean droplet size falling from 14 ± 6 μm to 3 ± 2 μm over 90 minutes. The largest decrease came in the first 10 minutes, and the mean size stayed constant from 70 to 90 minutes as a late-stage ordering signal appeared. Initial size estimates were affected by droplet motion and sedimentation.
In a sample of 242 finished supraparticles, about 87% were bright under crossed polarisers and 13% were dark. The authors used that split to estimate anisotropic and isotropic behaviour, although optical thickness and particle orientation can affect whether birefringence is detected. High gold loading can also reduce the light reaching the detector through absorption.
The first birefringence response appeared at an estimated particle volume fraction of 24% ± 2% for the gold and titanium dioxide supraparticles, compared with 28% ± 2% for titanium dioxide-only supraparticles. The authors interpreted this difference as consistent with ordering in the mixed particles at a lower volume fraction, without substantial attraction effects.
A simulation that followed part of the picture
In a separate computer model, rigid rods were represented with a length 10 times the sphere diameter. The model used a repulsive interaction strength set to 40 times the usual thermal-energy scale and began with 17,500 rods.
The simulated binary smectic phase remained stable when the sphere-to-rod number ratio was greater than zero and no more than one, matching the reported experiments. The model used simplified repulsive interactions and finite systems, so the result describes the simulated range rather than every possible preparation.
The model also showed a possible route to smectic order. Single rod layers emerged from the disordered, or isotropic, phase, then joined into larger clusters as successive layers formed and grew. Whether the same nucleation pathway occurs in the experiments remains unresolved.
At high sphere packing fractions, the simulated nematic and smectic phases were not observed. Instead, the system showed a direct transition from an isotropic state to a membrane state, with membrane formation above a sphere-to-rod ratio of 1. Finite-size effects may help explain differences between the high-ratio simulations and experiments.
What remains unresolved
Several measurements carry built-in uncertainty. POM brightness depended on sample thickness and orientation, so birefringent particles could be missed optically. High gold loading could reduce detected light through absorption, and the initial droplet size used for volume-fraction estimates had to be inferred because droplets moved and sedimented early.
The lowest-ratio sample also leaves open whether the mixed interlayers reflect insufficient sphere concentration or kinetic arrest. Taken together, the experiments and simulations map a composition-dependent change in structure, while the experimental nucleation mechanism remains to be established.
Paper data and sources
Original title: Self-Assembly of Au/TiO2 Sphere–Rod Mixtures into Binary Supraparticles
Authors: Bijl M, Matias A, Kolácný L et al.
Journal/Repository: ACS nano
Status: Peer-reviewed
First online: 2026-10-19
DOI: Not available
Original paper · Full text