Robert F. B. Weigel and
Michael Schmiedeberg
*
Theoretical Physics: Lab for Emergent Phenomena, Friedrich-Alexander-Universität, Erlangen-Nürnberg, 91058 Erlangen, Germany. E-mail: michael.schmiedeberg@fau.de
First published on 18th June 2025
Using a modified phase field crystal model that we have recently introduced [Weigel et al., Modelling and Simulation in Materials Science and Engineering, 2022, 30, 074003], we study grain boundaries that occur in two-dimensional structures composed of particles with preferred binding angles like patchy colloids. In the case of structures with a triangular order, we show how particles with a 5-fold rotational symmetry that differs from the usual 6-fold coordination of a particle in bulk affect the energy of the dislocations in the grain boundaries. Furthermore, for quasicrystals we find that the dislocation pairs recombine easily and the grain boundaries disappear. However, the resulting structure usually possesses a lot of phasonic strain. Our results demonstrate that the preferred symmetry of a particle is important for grain boundaries, and that periodic and aperiodic structures may differ in how stable their domain boundaries are.
Design, System, ApplicationWe perform calculations with a phase field crystal model of particles with preferred binding angles. The system is described by a density-like field that gives the relative deviation from a homogeneous density and a complex orientational field that contains both the strength of the orientation and a phase to denote the angle with respect to some reference direction. The phase field crystal approach uses a functional that corresponds to a power law expansion of the free energy that we minimize in our calculations. We seed grains with different orientations and study the free energy and the number of dislocations as a function of the angle between the seeded grains. For periodic crystals, we are especially interested in those composed of particles that possess a 5-fold rotational symmetry and find that they do not affect the dependence of the free energy on the angle between the grains. However, if quasicrystalline grains are considered, the outcome is totally different from that of periodic crystals as dislocations recombine. |
Recently, we have developed a phase field crystal model to describe the phase behavior of particles with preferred binding angles.4 Phase field crystal theories5,6 are based on the expansion of the free energy similar to that in the Swift–Hohenberg approach,7,8 can be derived from microscopic theories,9,10 and have been proposed to study the phase behavior in two dimensions.5,6 A similar approach had been introduced to explain the stability of quasicrystals.11,12
Quasicrystals possess a long range order but no translational symmetry. They were first reported in a physical system by Dan Shechtman.13 In the meantime, there are also many soft matter systems where quasicrystals have been observed.14–18 There are special collective rearrangements called phasons that occur in quasicrystals, which – similar to phonons in a crystal – do not increase the free energy;19,20 quasicrystals differ from periodic crystals, for e.g., in their growth behaviour21,22 and in how grain boundaries are formed or might disappear.23,24 Quasicrystals are also expected to occur for systems composed of patchy colloids.25–30
In this paper, we study grain boundaries in both periodic and aperiodic crystals composed of particles with preferred binding angles.
The system is described by a density-like field ψ that gives the relative deviation from a homogeneous density; thus its mean is 0. Furthermore, there is a complex orientational field U that contains both the strength |U| of the orientation and a phase to give the angle with respect to some reference direction. A homogeneous state is represented by ψ = 0 and a constant U. A state with isotropic orientation has a vanishing U = 0.
The free energy functional of our phase field crystal approach was derived in ref. 4 and is given by
![]() | (1) |
The phase behavior has been discussed in ref. 4. Note that as usual for the phase field crystal or similar models the resulting structures correspond to periodic or aperiodic cluster crystals.4,34
As we consider grain boundaries in the triangular phase for n = 2, 5, and ∞ (i.e., isotropic interactions), we choose parameter values for which the triangular phase is stable:4 Bl = 3, Bx = 3.5, D = 0.5, and E = 0.5. Furthermore, f(∇2) = ∇6 in order to suppress instabilities.4 Choosing F = 0 results in a vanishing orientation field U = 0. In this sense, F = 0 is equivalent to considering isotropic particles with n = ∞. When considering other values of n, we set F = 1, such that the orientation field plays a role.
The dodecagonal quasicrystalline phase is stable for n = 6, Bl = 3, Bx = 3.5, D = 2, E = 1, and F = 14. In this case, to enhance modulations in the orientation, we have to choose a non-monotonic function for f(∇2), which in reciprocal space has an extended flat minimum at small wave numbers k and for large k, it grows like k8. For further details, please see the discussion in ref. 4.
We interpolate between the two grains with the functions (1 ± cos(0 ·
))/2 with
0 = dk(cos
γ, sin
γ) and dk = 2π/L. As L is the length of the square-shaped simulation box with periodic boundaries, these interpolation functions make just one oscillation across the length of the box. Since the functions are smooth, they do not create a sharp edge between the grains but introduce a Moiré pattern between them instead (see Fig. 1b). Upon minimizing the free energy, the Moiré pattern relaxes to a grain boundary, which may not follow a straight line; see Fig. 1c. The angle γ is the angle between the straight line, where the grain boundary most likely will emerge, and a side of the simulation box. Since the free energy of the system may depend on the orientation of the simulation box relative to the grain boundary or relative to the orientation of the grains themselves, we vary γ when necessary to average over this dependence, effectively tilting the simulation box against the two grains; see Fig. 2. The angle α is always defined relative to the straight line, around which the grains are initialized. Hence, when γ is varied at constant α, both grains rotate relative to the simulation box.
![]() | ||
Fig. 2 Schematic of the grains and grain boundaries tilted by an angle γ against the simulation box. |
To obtain statistics, we perform computations for 10 different random initial configurations for each α: both grains' lattices are displaced by independent random vectors drawn from their respective rotated primitive unit cells; see Fig. 1a. The orientation field is initialized as low-amplitude random noise.
For our study regarding the triangular phase, it suffices to not vary γ and keep γ = 0, as shown in section 3.4.
To study grain boundaries in dodecagonal quasicrystals, we initialize the seed of each grain as 12 modes in reciprocal space with equal amplitudes and random phases. Hence, the position of the symmetry center and other degrees of freedom (isomorphism class) are random. In the same range of α ∈ and for 11 values of γ in the inclusive range
, we perform 10 simulations each.
We freely minimize the free energy with respect to the fields ψ and U by following the negative of the functional derivatives,
![]() | (2) |
![]() | (3) |
With the equilibrium nearest-neighbor peak distance being roughly 2π, we perform simulations in a square box of length L = 24.39 · 2π with periodic boundary conditions, discretized to 512 × 512 points, i.e. with a spatial resolution of 0.0476 · 2π.
When the simulation box is resized, stress might be reduced. The terms proportional to Bx, E, and F are the only ones sensitive to the size, and their reciprocal-space representation is a polynomial in the resolutions dkx and dky. The free energy can be minimized with respect to the resolutions with, for e.g. the Nelder–Mead algorithm.35 We do this to reduce stress once during the simulation, at half the simulation time, after the fields have almost become stationary. Note the size difference between Fig. 1b and c. This step adapts the size of the simulation box or the spatial resolution, respectively. However it does not change the mean density.
Dislocations can be created and annihilated in pairs of opposing Burgers vector. However, as we initialize the system without special care for the defect composition, in some configurations an odd number of dislocations is present. Due to the topology, the number of defects must stay odd (or even) under any peak-conserving dynamics. Some dislocations are deformed: it is very hard to tell apart the 5- and 7-coordinated peaks from some of their 6-coordinated neighbors. The Burgers vector reveals that they are normal dislocations nevertheless; see Fig. 4.
![]() | ||
Fig. 4 A distorted dislocation. The density-like field and orientation field are visualized as shown in Fig. 3. The open Burgers path around the dislocation is shown in red. |
Rarely, we observe disclinations, i.e. peaks with 5 or 7 neighbors surrounded by 6-coordinated neighbors exclusively. They only occur in configurations with many dislocations. Often, they are located close to chains of alternating 5- and 7-coordinated peaks with an odd length. In these cases, it is fundamentally ambiguous to decide which of the ends of the chain belongs to the dislocations and which is from another disclination. For statistics, we count a disclination pair as one dislocation.
Without this knowledge, but based on the observation of strongly oriented 5-coordinated peaks in uncontrolled grain boundaries, we conjectured in 4 that an orientation between grains matching the 5-fold symmetry may be energetically favored for n = 5. In the next section, we show that our new calculations of the free energy as a function of the angle between grains do not support this conjecture.
![]() | ||
Fig. 5 The rescaled free energy ![]() ![]() ![]() ![]() ![]() |
We rescale the energy by the absolute of the free energy of the perfect triangular lattice for the respective n, which is given in Table 1. The rescaled /|
tri| of a perfect crystal is −1 and the rescaled free energy of a homogeneous isotropic configuration is still 0. Contrary to our conjecture,
or
is not energetically favored for n = 5. Instead, the maximum rescaled free energy occurs roughly at
, as the data in Fig. 5 show. The situation is almost identical for all considered n, up to the constant rescaling factors and a slightly smaller increase of the free energy in the case of n = 5 in comparison to the increase for isotropic interactions (n = ∞).
n | |
---|---|
2 | −0.644 |
5 | −0.755 |
∞ | −0.362 |
The increase of the free energy can be explained by a growing number of dislocations that are present in the respective configurations. Of course, this number depends only on the angle α and not on n as can be seen in the lower part of Fig. 5. It follows that the free energy gain of oriented 5-coordinated peaks for n = 5 is small in comparison to the typical free energy cost of adding a dislocation in the triangular lattice.
![]() | ||
Fig. 6 The distribution of the free energy per area ![]() ![]() |
Which free energy level is reached does not directly depend on angle α, which characterizes the initial grain boundaries. It rather depends on the relation between the simulation box and the dodecahedra that appear in the system. If the edges of the dodecahedron align with the edges of the rectangular box, lower free energies are typically reached, than in the case of vertices pointing to the box edges.
The different free energy levels correspond to periodic approximants of different sizes. Some of them do not fit well into the simulation box. In these cases, they are combined with filler structures, which likely can be viewed as smaller approximants in turn. The smaller the approximant's asymmetric unit, the higher is its free energy (see Fig. 7 for some examples).
![]() | ||
Fig. 7 The columns (a) through (d) show typical configurations corresponding to the free energies labeled in Fig. 6. The top row displays the full density-like fields, with the orientation fields omitted. The insets in the top row show the absolute magnitudes of the Fourier transform of the respective density fields. The bottom row shows zoomed-in sections of the respective configurations, where the density-like field and orientation field are visualized as seen in Fig. 3. |
A mode analysis36,37 reveals that the dislocations, which had initially been present, usually annihilate during the minimization. So, in contrast to periodic structures, free energy changes are not caused by dislocations. This holds for all the free energy levels. The approximants seem to contain phasonic flips as the structures differ from the initial quasicrystalline patterns. The phasonic flips cause phasonic strain that probably causes the different free energy levels. This is similar to the situation in ref. 23 where two quasicrystals merge by relaxing the possible defects by phasonic flips but at the cost of exciting phasonic strain.
Although our minimization scheme is not a dynamical simulation, the process of annihilation observed during the minimization steps looks like what one would expect from a dynamical simulation: dislocations of opposing Burgers vectors approach each other and then recombine. The approach is facilitated by the local phasonic rearrangements. In very rare cases, we observe that dislocation pairs are formed and do not annihilate or move anymore.
We have found that the increase in free energy due to grain boundaries is slightly less pronounced for particles with 5-fold symmetry than for isotropic particles. Furthermore, for particles with 5-fold rotational symmetry, we observe a very strong orientational field at the position of the particles that have only five nearest neighbors. The strong orientation might be the reason for the less pronounced increase of the free energy.
However, we have not observed any preferred angles between the crystals in contact due to a preferred binding angle. To be specific, except for a small overall decrease, the free energy as a function of the relative angle between the crystals does not significantly change if particles with n = 5 instead of particles with isotropic symmetry (n = ∞) are considered.
If the preferred binding angles are used to stabilize a quasicrystal, we find a completely different behavior at the grain boundaries. The phasonic degrees of freedom are used to move the dislocations such that they can recombine with opposite dislocations from another grain boundary. As a result, one finds patterns with motives of quasicrystals but no dislocation at all. We argue that the situation is similar to that in ref. 23 and 24 where the number of defects was reduced due to phasonic rearrangements. However, here we even observe a more extreme behavior, as usually all dislocations disappear and the resulting pattern is distorted a lot which corresponds to a large phasonic strain. It seems that even local phasonic rearrangements can be easily excited in our model allowing an efficient way to repair defects.
As already pointed out in ref. 4, an important advantage of our approach is that the orientation field is treated separately from the local density field. Therefore, the strength of the orientation field at a grain boundary might differ from the strength of the orientation close to a particle in bulk. Furthermore, the effect of phasonic rearrangements can be studied with our model. As a consequence, our mean field approach is suitable to study and understand the behavior of particles with preferred binding angles close to grain boundaries.
This journal is © The Royal Society of Chemistry 2025 |