Visualizing defect dynamics by assembling the colloidal graphene lattice

Two-dimensional materials combine extraordinary mechanical, optical and electronic properties. Graphene, a monolayer of carbon atoms in a honeycomb lattice, exhibits strong in-plane covalent bonding and a honeycomb geometry that yields photonic and phononic band gaps. Scaling these concepts to micrometer-sized particles to form colloidal graphene would enable 2D multifunctional materials with analogous band-gap features. However, producing large, defect-free single-crystalline graphene (atomic or colloidal) remains challenging. Defects critically influence graphene’s properties and can be introduced during growth or deliberately for tuning functionalities, yet their formation pathways and healing kinetics are not fully understood. The trivalent coordination allows many polygonal arrangements so coherent lattices can exist with defects and multiple rearrangement pathways. Direct visualization at the atomic scale during high-temperature deposition is difficult. Colloidal model systems, governed by thermal forces similarly to atoms, allow real-time, single-particle-resolution observation of crystallization and defect dynamics. Advances in anisotropic “patchy” particles with programmed valency and angles provide a route to assemble molecule analogues and covalently bonded crystals. This study assembles a colloidal analogue of graphene using pseudo-trivalent patchy particles to directly follow crystallization, defect formation, and healing, and to quantify the associated energy landscape.

Prior work established colloids as model systems for atomic crystallization due to shared thermodynamic principles and accessible real-time microscopy. Defect formation and dynamics have been visualized in colloidal crystals, while atomic graphene studies using electron microscopy face limitations during high-temperature growth. Advances in synthesizing anisotropic and patchy colloids with directional bonding have enabled assembly of complex structures (e.g., kagome lattices, clathrates, diamond), though kinetic traps (e.g., 5-membered motifs) can hinder equilibrium structures. Simulations of surface-confined trivalent particles predict honeycomb lattices at intermediate densities. Atomic graphene studies report grain boundaries comprised of alternating pentagon–heptagon pairs, vacancy reconfigurations, and strain effects, but the initial stages of defect nucleation are less accessible. This work leverages near-equilibrium colloidal assembly with tunable critical Casimir interactions to bridge these gaps and probe defect origin and dynamics at the particle level.

Patchy particle synthesis and design: Patchy colloids were fabricated by colloidal fusion of polystyrene (PS) and 3-(trimethoxysilyl)propyl methacrylate (TPM) spheres, yielding tetrahedrally coordinated particles with a PS core and fluorescent TPM patches. Particle diameter d ≈ 2.0 µm, patch diameter d1 ≈ 0.2 µm (sufficiently small to enforce single-patch bonding). When adsorbed to a substrate via one patch, the particles become pseudo-trivalent, with three patches available for bonding and bond angles near 120°, mimicking sp2-like coordination with slight out-of-plane tilt.

Solvent and interaction control: Particles were dispersed in a near-critical binary solvent of water and 2,6-lutidine (volume fraction c1 = 0.25, near critical c1^c ≈ 0.27), with demixing temperature T_cx ≈ 33.95 °C. Added 1 mM MgSO4 screened electrostatics and enhanced hydrophobic patch adsorption. Samples were loaded into silanized, hydrophobic glass capillaries; at ΔT = T − T_c ≤ 0.6 °C particles adsorbed via one patch to the wall and diffused freely in-plane. Patch–patch critical Casimir attractions were tuned by temperature: at ΔT ≤ 0.25 °C free patches began to attract; assembly was driven near-equilibrium by approaching T_c in 0.05 °C steps from ΔT = 0.25 °C, equilibrating for 4 h at each step, mimicking slow cooling in atomic deposition. In this range, pair binding energies reached 12–15 k_BT (from a benchmarked critical Casimir model and measurements).

Bending potential measurement: The bond-bending potential was obtained from angle fluctuations in three-particle clusters (two neighbors bonded to a central particle). Angle distributions were converted to bending potentials via Boltzmann inversion, exhibiting approximately harmonic behavior. This enabled estimation of a bending force constant (harmonic fit to U_bend(θ)).

Imaging and tracking: Confocal microscopy (100× oil immersion) and bright-field imaging were used to monitor growth, structure, and defects with single-particle and single-bond resolution. Temperature control combined a stage and objective heater to achieve ~0.01 °C relative accuracy with minimal gradients. Particles were allowed to sediment prior to measurements; during heating to ΔT ≈ 0.6 °C one patch anchored to the glass. 3D image stacks and time series were analyzed with particle tracking software (Trackpy) to determine particle centers and bonding states (bonded patches appear as merged fluorescent blobs), enabling identification of rings, defects (pentagons, heptagons, vacancies), and bond angles.

Energy analysis: Configurational energy per particle was computed as the sum of contributions from (i) dangling (unsaturated) bonds and (ii) bond-bending distortions (from measured angles relative to 120° and the bending potential). Time-resolved energy traces were obtained for selected regions to correlate structural reconfiguration with energy changes.

Protocols for growth studies: Early-stage assembly was probed at lower attraction (higher ΔT), revealing small clusters dominated by pentagons; as ΔT decreased in steps, clusters grew, hexagons increased, and extended flakes formed. Defect formation was tracked in situ, including grain boundary nucleation from pre-existing pentagons and grain-boundary creation upon grain merging. Long-term defect evolution was followed over 9 h at fixed temperature to study vacancy coarsening/reconfiguration and energy landscape navigation.

• Colloidal graphene assembly: Pseudo-trivalent patchy particles assembled into large honeycomb flakes with three bonds per particle at ~120°, forming repeating hexagonal rings characteristic of graphene.
• Early-stage kinetics favor pentagons: At low interaction strength, small clusters are dominated by 5-membered rings; as attraction increases, hexagon motifs grow at the expense of pentagons, eventually forming extended honeycomb flakes. Energy-time traces of isolated rings show pentagons form by rapid closure after the 5th particle attachment, whereas hexagons remain open after five particles to incorporate a sixth before closure, yielding lower energy. This demonstrates a kinetically favored pathway to pentagons.
• Grain boundaries of alternating 5–7 rings: Fully grown flakes commonly contain grain boundaries comprising alternating pentagons and heptagons that separate grains rotated by ~6–17°; a typical rotation ≈17° was observed. Geometrically, a 5–6–6 combination leaves ~12° mismatch, while 5–6–7 leaves only ~3–4°, stabilizing 5–7 scars.
• Quantified angular strain energetics: At final ΔT = 0.05 °C, adding two adjacent hexagons to a pentagon imposes a bending energy cost ~0.43 k_BT (angle deviation ~12°), while adding a heptagon and hexagon costs ~−0.14 k_BT (deviation ~3.4°), rationalizing stabilization of alternating 5–7 chains.
• Strain localization near grain boundaries: Bond-bending energy increases near the 5–7 boundary (variance comparable to thermal energy) and decays to near-bulk levels within about one hexagon layer from the interface.
• Vacancies: Monovacancies, divacancies, and larger polyvacancies were directly imaged. In contrast to atomic graphene (where vacancies often reconfigure to lower-dangling-bond motifs), colloidal vacancies remained stable and did not readily reconstruct, attributed to short-range, relatively rigid critical Casimir bonds and higher barriers to rearrangement.
• Defect evolution and energy landscape: Over 9 h, regions with multiple-grain junctions showed significant reconfiguration: polyvacancies merged/split while the total energy per particle decreased via reductions in both dangling-bond and bending contributions (dangling-bond reduction dominated). The number of hexagons increased over time in dynamic regions, consistent with approach to lower-energy configurations.
• Interaction energetics: Critical Casimir patch–patch bond strengths reached ~12–15 k_BT under near-equilibrium assembly, similar in scale to effective thermalized energies in high-temperature atomic graphene processing, supporting the colloidal system as a model for defect dynamics.
• Grain boundary formation pathways: Boundaries nucleate from early pentagons (which then promote adjacent heptagons) and also form during merging of misoriented grains as the system stitches to minimize dangling bonds, yielding 5–7 scars.

The study demonstrates a colloidal analogue of graphene in which trivalent coordination and near-equilibrium, tunable critical Casimir interactions enable direct, time-resolved observation of crystallization, defect nucleation, and healing. A key insight is that pentagons are kinetically favored during the earliest stages of growth, serving as seeds that stabilize grain boundaries via adjacent heptagons, thereby limiting single-crystal growth. Subsequent reconfiguration is governed by a balance between saturating dangling bonds and minimizing lattice distortions; dynamic regions evolve towards lower-energy states with increasing hexagon count, while some grain boundaries remain immobile. The observed 5–7 grain boundary motifs, misorientations, and stabilized scars parallel those in atomic graphene, although differences in interaction potentials and reconstruction kinetics exist. Energy scales in the colloidal system (~12–15 k_BT bonds) are comparable to those relevant under high-temperature atomic growth, lending support to the colloidal model for capturing defect pathways. Strategies used in atomic systems (single-seed growth, aligning islands before merging) and proposed colloidal tactics (e.g., A/B selective binding to enforce even-membered rings) could mitigate pentagon formation and promote defect-free growth, although implementation in critical Casimir systems is non-trivial. The results underscore the central role of early-stage kinetics in defect genesis and highlight how controlled interaction tuning enables mapping of the energy landscape underlying defect dynamics.

This work assembles colloidal graphene from pseudo-trivalent patchy particles and directly visualizes the formation, stabilization, and healing of common defects at single-particle resolution. It identifies kinetically favored pentagons in early growth as seeds for 5–7 grain boundaries and demonstrates that defect evolution proceeds through an interplay of dangling-bond saturation and bond-angle strain relaxation, increasing the number of hexagons and lowering the lattice energy. The colloidal model provides access to growth pathways and defect energetics not readily observable in atomic systems and offers quantitative insight into strain localization and vacancy dynamics. Future directions include engineering interaction specificity (e.g., binary A/B patch chemistries) to disfavor pentagon formation, implementing single-seed nucleation protocols within the narrow temperature window of critical Casimir systems, and extending assembly to more complex 2D metamaterials with targeted photonic/phononic band structures and topological properties.

• Model system vs atomic graphene: While energy scales are comparable at elevated temperatures, the colloidal critical Casimir potential and out-of-plane constraints differ from covalent bonding, affecting reconstruction kinetics and defect stability.
• Interaction encoding: Specific, selective patch–patch chemistries (e.g., A/B selectivity) are difficult to implement with critical Casimir interactions, limiting direct tests of strategies to suppress odd-membered rings.
• Temperature window and control: The accessible ΔT range is narrow; achieving true single-seed growth and precise synchronization of island merging may be challenging in practice.
• Energy model simplifications: The configurational energy considers dangling bonds and bond-bending distortions; other contributions (e.g., bond-stretching, multi-body or longer-range effects) are not explicitly included.
• Reconstruction barriers: The short-range, relatively rigid interactions may over-stabilize certain defects compared with atomic graphene, potentially limiting generalizability of reconstruction pathways.