Report

Direction-Specific Interactions Control Crystal Growth by Oriented Attachment

See allHide authors and affiliations

Science  25 May 2012:
Vol. 336, Issue 6084, pp. 1014-1018
DOI: 10.1126/science.1219643

Abstract

The oriented attachment of molecular clusters and nanoparticles in solution is now recognized as an important mechanism of crystal growth in many materials, yet the alignment process and attachment mechanism have not been established. We performed high-resolution transmission electron microscopy using a fluid cell to directly observe oriented attachment of iron oxyhydroxide nanoparticles. The particles undergo continuous rotation and interaction until they find a perfect lattice match. A sudden jump to contact then occurs over less than 1 nanometer, followed by lateral atom-by-atom addition initiated at the contact point. Interface elimination proceeds at a rate consistent with the curvature dependence of the Gibbs free energy. Measured translational and rotational accelerations show that strong, highly direction-specific interactions drive crystal growth via oriented attachment.

The growth of crystals through the aggregation and coalescence of nanoparticles is now recognized as a widespread phenomenon in biomineral (1, 2), biomimetic (3, 4), and natural systems (5) and during the synthetic production of nanoparticles and nanowires (68). The resulting crystals often exhibit complex forms ranging from quasi–one-dimensional (1D) chains to 3D hierarchical and self-similar superstructures. Yet the final structure typically diffracts as a single crystal, implying that the primary particles aligned during growth (4, 9). When coalignment is accompanied by coalescence, this growth process is often referred to as oriented attachment (OA) (10, 11). However, the pathway by which OA occurs has not been established. Although the preservation of primary particle morphology and the formation of twins and stacking faults at particle-particle boundaries strongly suggest a sequence of whole-particle alignment followed by interface elimination (10, 12), atom-by-atom reorientation via dislocation and grain-boundary migration after attachment are another potential mechanism (13, 14). If indeed the primary particles align before attachment, the dynamics of that process and the forces that drive it have yet to be revealed.

We used a liquid cell mounted within a high-resolution transmission electron microscope (TEM) for the direct in situ observation of iron oxide nanoparticle growth (fig. S1) (15). Previous studies have inferred that OA is important in the early crystal growth of iron oxyhydroxide nanoparticles (8, 1618), which are abundant and play important roles in biogeochemical processes that shape Earth’s near-surface environments (1921). Images were recorded with lattice fringe resolution, enabling us to track particle orientations throughout the experiments. Selected area electron diffraction pattern analysis indicates that the nanoparticles formed in the fluid cell through re-precipitation from akaganeite nanoparticle precursors are closely related to six-line ferrihydrite (5Fe2O3·9H2O) (fig. S2).

Low-resolution TEM images showed that after nucleation, the ferrihydrite nanoparticles grew both through monomer addition from solution and particle attachment events (fig. S3 and movie S1), as observed previously with Pt nanoparticles (22). We recorded several high-resolution movies that revealed the dynamics of the attachment process (movies S2 to S5): The particles continuously diffused and rotated, repeatedly contacting one another until attachment finally occurred. The surfaces of two adjoining particles made transient contact at many points and orientations before finally attaching and growing together (Fig. 1). Similar behavior was observed for smaller, 5- to 10-nm particles typical of six-line ferrihydrite (23) (movie S5).

Fig. 1

(A to G) Sequence of images from movie S2 showing typical dynamics of the attachment process (see fig. S4 for high-resolution images). The surfaces of particles I and II made transient contact at many points and orientations (points 1-1, 1-2, 2-3, and 3-4) before finally attaching and growing together (points 3-5). (H) High-resolution image of interface in (G) showing twin structure (an inclined twin plane). The yellow dashed line in (G) shows the original boundary of the attached particle. (I and J) High-resolution in situ TEM image (I) and fast Fourier transform (FFT) (J) of an interface from another OA event demonstrating formation of a (101) twin interface after attachment. The grain boundary is delineated by a dashed line in (I). Scale bars are 5 nm for (A) to (G).

Lattice fringe images recorded in situ revealed that, irrespective of how many times particles made contact, at the time of attachment they either shared the same crystallographic orientation or their orientations were twin-related (fig. S4). Either way, after attachment, atoms filled the interface region on a time scale on the order of 10 to 100 s, beginning at the point of contact. Interface elimination resulted either in larger defect-free crystals (Fig. 2A and D to H, B and I to K, and movie S3) or crystals that contained a twin plane (Fig. 1, H and I). In all cases analyzed, twinning was on (101) ferrihydrite (Fig. 1J). This interface structure is consistent with studies demonstrating that periodic matching of the Bravais lattice geometry is associated with reduced boundary energy (24, 25), as well as the coincident-site lattice model, which relates the boundary energy to the number of coincident lattice sites (26).

Fig. 2

Sequences of in situ TEM images show the details of the attachment process. (A) and (D to H) Sequence from movie S3 showing attachment at a lattice-matched interface. (A) shows the arrangement of particles before attachment. The asymmetric particle in front of the smaller spherical particle is not involved in the attachment process. (D) to (G) show formation of the interface. Two edge dislocations denoted in (E) to (G) by red dashed lines translate to the right, leaving a defect-free structure in (H). (B) and (I to K) Sequence of images showing relative rotations of particles during the attachment process, leading to a lattice-matched interface. Particles I and II have their [–121] zone axes perpendicular to the viewing plane. FFT patterns are given in fig. S5. (C) and (L to P) Sequence from movie S4 showing how the interface expands laterally after attachment. All scale bars are 2 nm.

Sequences of lattice fringe images also provide direct insight into the dynamics of processes leading to growth and microstructure development, confirming some inferences from ex situ studies (10, 27, 28) and providing information about the forces involved. As the particles approached, they typically rotated until the lattice planes were close to a perfect match (Fig. 2, E and K). For example, particles initially misaligned by 54° (particles I and II in Fig. 2B), as shown by the orientations of their (101) planes, rotated in opposite directions by +45° and –9°, respectively (Fig. 2, I to K), leading to attachment with perfect crystallographic alignment (Fig. 2K). In some cases, a slight misalignment at the time of attachment led to defect formation at the interface, as in Fig. 2E where the lattice planes near the boundary are slightly bent because of the formation of two edge dislocations. Nonetheless, within 2 s, these dislocations translated laterally across the interface (Fig. 2, E to G), leaving behind perfectly flat planes and eliminating any trace of the interparticle boundary (Fig. 2H).

Analysis of particle motion leading to attachment shows that the translational and rotational speeds increased by factors of 2 to 4 as the particles approached one another. For example, the smaller particle in Fig. 2C diffused with a relatively constant translational speed of approximately 0.6 to 0.8 nm s−1 and a rotational speed of approximately 0.09 rad s−1 (Fig. 3A). Within the last 0.5 s before attachment, the translational component jumped to 1.8 to 2.5 nm s−1 and the rotational component increased to 0.21 to 0.28 rad s−1.

Fig. 3

(A) Relative translational and angular speeds leading up to the attachment process in sequence (L) to (P) of Fig. 2. (B) Dependence of the interface growth rate on curvature derived from the dependence of interface length on time (inset) for sequence (L) to (P) of Fig. 2 (grain boundary 1) and the sequence in movie S2 (grain boundary 2). The maximum error from delocalization is estimated to be less than 1 nm, and the error bars for each data point are shown in fig. S5.

In order to investigate the post-attachment kinetics, we measured the lateral growth rate at the grain boundary. Image sequences (Fig. 2C and L to P, from movie S4) show that after attachment, the boundary advanced by monomer addition. (Also see fig. S4, which shows the attachment event in Fig. 1 at higher resolution.) The lateral growth rate decreased from an initial value of 13.1 nm s−1 to 0.14 nm s−1 within the first 2 s, and to a constant value of 0.034 nm s−1 after approximately 20 s (Fig. 3B). The dependence of the speed of the growth front on the curvature of the interface is consistent with the well-known exponential dependence of chemical potential on interface curvature (29). However, a detailed form of the dependence cannot be determined, because the 3D shape of the interface is unknown, as is the variation in attachment and detachment kinetics with interface orientation.

In contrast to freely diffusing nanoparticles, those that were flanked by two adjacent particles (such as particle II in Fig. 4A) became trapped and then fused across a mismatched interface (Fig. 4C and movie S6). At the time of attachment, particles aligned along different zone axes (Fig. 4, G and H). Grain boundaries then migrated toward the adjacent particle, leading to the growth of one orientation at the expense of the other, which was quickly consumed (Fig. 4, A to F). This post-attachment behavior of misaligned particles is consistent with strain-induced grain boundary migration, through which the energy stored in dislocations and point defects at the grain boundary of a highly strained nanoparticle drives recrystallization (13, 14). In close proximity to large particles, some <5-nm particles undergo dissolution even though these particles are stable before their encounter with the larger one (movie S6). As expected from the curvature dependence of the free energy, the dissolution rate increases rapidly with decreasing size. These two phenomena—attachment and recrystallization of a misaligned particle and dissolution of small particles in the vicinity of larger ones—show that, even in systems where OA is dominant, Ostwald ripening can play a role.

Fig. 4

(A to F) Sequence of in situ TEM images from movie S6 showing a rare example of attachment across a mismatched interface due to trapping between two adjacent particles. (G to I) FFT diffraction patterns of particle I [in (C)], particle II [in (C)], and the fused product particle IV [in (F)], respectively. The 5-nm scale bar in (F) applies to all high-resolution TEM images.

In all examples, nanoparticles explored multiple configurations before attachment occurred. This was possible because the primary particles tended to aggregate into clusters where they interacted in close proximity for sufficient periods of time to become nearly aligned through Brownian motion (fig. S3 and movie S1). This tendency implies the existence of a long-range attractive force that is necessary to keep them in proximity. Osmotic forces, which largely drive colloidal stabilization (30), are most likely responsible. Previous ex situ studies of OA also proposed this as a source of initial aggregation (5); however, this colloidal stabilization must be sufficiently weak to allow the nanoparticles to approach each other within the primary minimum of the interaction potential, where van der Waals interactions could lead to further attraction. In addition, the diffusional dynamics must be sufficiently rapid to allow rearrangement into low-energy configurations. Because the OA event is accompanied by a jump to contact only after orientation is established (Fig. 1), OA is ultimately driven by a short-range force that acts over <1 nm and is highly sensitive to orientation. The orientation dependence would seem to argue for Coulomb interactions as the source of this force, although one cannot rule out van der Waals interactions with anisotropic polarizability. The observation that the acceleration phase occurs over the last 0.5 to 1 nm is consistent with the expected Debye length for a 50 mM 3:1 electrolyte solution (30). In addition, the fact that the particles tend to remain separated by nanometer-scale distances until a correct configuration is reached implies the existence of a slight energetic barrier. This provides further support for the importance of electrostatic interactions, which, in the case of molecular crystals such as iron oxides, would produce cation-cation and anion-anion repulsion when the lattices are mismatched.

Irrespective of its source, to estimate the strength of this short-range attraction, we determined the translational and angular acceleration by measuring the frame-to-frame velocity and rotation rate (Fig. 3A). Using the density of ferrihydrite and the measured particle size to determine the mass and moment of inertia for spherical particles, we calculated that the forces and torques acting on the approaching particle arising from particle-attraction are almost completely offset by the resistance of the solution (supplementary text). Even so, the potential energy for the particle-particle interaction, which equals 1.6 × 10−19 J, far exceeds the initial kinetic energy of 7.5 × 10−40 J. In addition, if we assume that this interaction energy is due to the electrostatic force, we find that the particles interact with an effective number of fundamental charges (1.6 × 10−19 C) on the order of unity (supplementary text).

Supplementary Materials

www.sciencemag.org/cgi/content/full/336/6084/1014/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S5

Table S1

References (3133)

Movies S1 to S6

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank V. Altoe and S. Aloni for their assistance with electron microscopy and L. Zhang for his help with analysis of the data. This research was supported by the U.S. Department of Energy, Office of Basic Energy Sciences (OBES), by LBNL under contract no. DE-AC02-05CH11231 and Lawrence Livermore National Laboratory (LLNL) under contract DE-AC52-07NA27344. Development of the TEM fluid cell was supported by the OBES, Division of Chemical, Biological and Geological Sciences; analysis of iron oxide formation was supported by the OBES, Division of Materials Science and Engineering; and cell fabrication and TEM analysis were performed at the Molecular Foundry, LLNL, which is supported by the OBES, Scientific User Facilities Division. M.H.N. acknowledges government support under and awarded by the Department of Defense, the Air Force Office of Scientific Research, and a National Defense Science and Engineering Graduate Fellowship, 32 CFR 168a. C.F. acknowledges support from The Danish Council for Independent Research. Data are available in the supplementary materials.
View Abstract

Navigate This Article