Particle analogs of electrons in colloidal crystals

See allHide authors and affiliations

Science  21 Jun 2019:
Vol. 364, Issue 6446, pp. 1174-1178
DOI: 10.1126/science.aaw8237

Mobile particles in colloidal crystals

The crystallization of nanoparticles can be controlled by functionalizing them with DNA strands that direct assembly through hybridization. The design rules for interactions between pairs of particles resemble those for ionic compounds. Inspired by molecular dynamics simulations, Girard et al. show that larger particles (∼10 nanometers in diameter) that have mutual repulsive interactions can form a stable lattice only if much smaller conjugate particles (∼1.5 nanometers in diameter) are present. These smaller particles are mobile and diffuse through the lattice, so the bonding interaction resembles the classical picture of electrons in metals.

Science, this issue p. 1174


A versatile method for the design of colloidal crystals involves the use of DNA as a particle-directing ligand. With such systems, DNA-nanoparticle conjugates are considered programmable atom equivalents (PAEs), and design rules have been devised to engineer crystallization outcomes. This work shows that when reduced in size and DNA grafting density, PAEs behave as electron equivalents (EEs), roaming through and stabilizing the lattices defined by larger PAEs, as electrons do in metals in the classical picture. This discovery defines a new property of colloidal crystals—metallicity—that is characterized by the extent of EE delocalization and diffusion. As the number of strands increases or the temperature decreases, the EEs localize, which is structurally reminiscent of a metal-insulator transition. Colloidal crystal metallicity, therefore, provides new routes to metallic, intermetallic, and compound phases.

The interactions among electrons and atoms to form molecules and materials are foundational in physics and chemistry. However, in the science of colloidal crystals, in which particles are often analogized with atoms, a particle analog to electrons has not been invoked, despite the synthesis of hundreds of colloidal crystals (124) and the development of certain approaches into elaborate forms of crystal engineering (921). In particular, colloidal crystal engineering with DNA has led to the design of structures with diverse symmetries, lattice parameters, and crystal habits (1222). However, to date, the particles modified with DNA that define such structures behave as programmable atom equivalents (PAEs) and have fixed particle positions at set stoichiometric ratios. These systems are governed by a set of design rules and the complementary contact model (CCM) (14), the premise that they organize themselves to maximize contacts that lead to hybridization and structures such as ionic compounds.

We report on an electron-atom duality analog in colloidal crystal engineering with DNA, in which the resulting colloidal assemblies are better classified as “metallic” structures. In such structures, small DNA-functionalized NPs become mobile and “electron-like” [or electron equivalents (EEs)] and are essential for maintaining the positions of the larger PAE “atom” components. Mixtures of complementary DNA-functionalized nanoparticles (NPs) that vary in size and DNA surface density were assembled and characterized by means of electron microscopy, synchrotron small-angle x-ray scattering (SAXS), and scale-accurate molecular dynamics (MD) simulations with explicit hybridization (17, 25, 26). Through a combination of theory, simulations, and experiments, we show that small particles grafted with low numbers of DNA strands (for example, <6), when mixed with complementary functionalized NPs (Fig. 1A), form crystals but do not occupy specific lattice sites and diffuse through the crystal in a manner reminiscent of classical electrons in metals, as described by the original Drude model. The PAEs alone will not form crystals because they are almost solely repulsive. The delocalized EEs that move freely through the lattice are responsible for stabilizing it, a type of bonding more reminiscent of metals than ionic compounds (Fig. 1B).

Fig. 1 Transition from PAE-PAE systems to PAE-EE systems.

(A) Illustrations of DNA-functionalized Au NPs behaving as programmable atom equivalents (PAEs) or electron equivalents (EEs) used in the MD simulation. (B) Snapshots from the MD simulation depicting “ionic” bonding behavior shown by PAE + PAE assemblies, and “metallic” bonding behavior shown by PAE + EE assemblies, where roaming EEs hold the crystal of repulsive PAEs together. (C and D) Four crystalline lattices assembled from 10-nm PAEs and complementary DNA–functionalized Au NPs (nominal core diameters of 10, 5, 2, and 1.4 nm, respectively). Shown are (C) SAXS spectra, (D) models, and cross-sectional LAADF images of silica-encapsulated samples. Scale bar, 25 nm. The 1.4-nm Au NPs in (D) (yellow arrows indicate visually identified ones) are dispersed randomly in the lattice and do not occupy specific lattice sites. (E) SAXS-determined distance between bonding 10-nm PAE pairs (same DNA type, defined in the inset according to CCM assumptions). (F) Quantification of linker DNA strands duplexed on 1.4-nm Au NPs (EEs) as a function of the input number of linkers per EE in the solution.

Furthermore, when the interactions are tailored by increasing the number of potential DNA bonds or lowering the temperature, these EEs condense into specific locations, yielding a transition akin to a metal-insulator transition. Last, by taking advantage of this duality and the structural features of the DNA-modified particles that govern it, we realized three polymorphic crystal phases—body-centered cubic (bcc), face-centered cubic (fcc), and Frank-Kasper A15—and analyzed the distribution and diffusion of the particles (EEs and PAEs) within them as a function of temperature and number of linkers per EE.

In a typical set of experiments, 10-nm-diameter Au NPs were densely modified with single-stranded propylthiolated DNA to yield conjugates with ~160 strands per NP. These modified NPs were hybridized with a complementary strand to form a rigid duplex region (18 bases) with a six-base single-stranded overhang (fig. S1). NPs with average diameters of 10, 5, 2, and 1.4 nm, respectively, were modified in a similar manner but with a second type of complementary DNA overhang (Fig. 1A). The average number of DNA overhanging strands available for bonding (number of linkers per EE) is a function of input linker concentrations in the solution (Fig. 1F). For the first three combinations of NPs (10 + 10, 10 + 5, and 10 + 2 nm), all formed the expected CsCl lattice (space group Pm3¯m) (27) based on the conventional CCM model and the description of them as ionic compound analogs (14).

However, with the 10 + 1.4 nm combination, the 10-nm NPs assumed a bcc lattice (space group Im3¯m), but the 1.4-nm NPs were invisible to SAXS (Fig. 1C). The lattice assignments were all verified by means of electron microscopy with low-angle annular dark field (LAADF) imaging after the structures were encased in silica (28), whereas the 1.4-nm NPs in the 10 + 1.4 nm combination did not appear at specific lattice sites (Fig. 1D). For the first three combinations, the NP positions were determined by the length of the DNA bonding elements that define them (Fig. 1E). However, for the 10 + 1.4 nm combination, there was a marked decrease in the interparticle distance compared with the expected value based on CCM prediction (14).

To visualize the positions of the 1.4-nm NPs, we performed cryogenic transmission electron microscopy (cryo-TEM) on the as-synthesized bcc lattice formed from the 10 + 1.4 nm NPs and then stacked the repeating “unit cell” images and EE locations along the [111] zone axis (Fig. 2A). Cryo-TEM showed that the large NPs (PAEs) assumed a bcc lattice, and the small NPs (EEs) were randomized throughout that lattice (Fig. 2B), which is in agreement with the MD simulations (Fig. 2D). In the simulations, crystalline structures were obtained for mixtures of complementary DNA–functionalized Au NPs at a fixed size ratio (10- to 2-nm diameter) but with variable EE:PAE ratios from 4:1 to 12:1 and number of linkers per EE from four to eight (Fig. 2, D and E). Because the MD simulations were performed in the isobaric-isothermal ensemble (NPT) with pressure near zero (supplementary materials), these complementary EEs, which are delocalized from specific lattice sites (Fig. 2D), were responsible for the attraction that holds the large PAEs in crystalline positions in these metal-like assemblies.

Fig. 2 Spatial probability distribution of EEs in PAE-EE assemblies in the bcc structure.

(A) Workflow for obtaining EE location-labeled “unit cell” images from cryo-TEM by using image segmentation. ACF, auto-correlation function. (B and C) Overlay of averaged-intensity TEM images and identified EE locations in “unit cells” along the [111] direction. The input parameters refer to the ratio of each substance added to the solution, not within a crystal. (D and E) MD simulation snapshots of PAE-EE assemblies. (F) Measure of clustering tendency (Scl) of EEs in MD simulations. (G) Temperature-dependent trapping time (τ) of EEs in MD simulations. kB, Boltzmann constant. (H to J) Simulated Boltzmann volumes of EEs viewed along the [111] direction with EE:PAE = 4:1 and (H) 4 or (I) 8 linkers per EE, or with (J) EE:PAE = 9:1 and 8 linkers per EE. Orange dashes approximate a repeating “unit cell” used in cryo-TEM image analysis.

We determined the degree of delocalization of EEs in the MD simulations by discretizing the unit cell volumes Vcell into (128)3 voxels of equal volume a3 so that Vcell = (128a)3 and then counting the EE visitation frequency in each voxel. This frequency gives a probability distribution fk in each voxel, k. A quantitative measure of clustering tendency, Scl, is defined asScl=kfkln(fk)+ln(Vcell/a03)(1)where a0 is the average of a over all simulations and used as a constant value to normalize the volume. The quantity Scl can be associated with an information entropy (29). This quantity is minimized if the EEs are localized and maximized if they are delocalized (when they have a uniform distribution). The resulting Scl shows a strong correlation with both EE:PAE ratio and number of linkers per EE (Fig. 2F): Either an increase in the number of EEs in the lattice or a decrease in the number of duplexed DNA linkers on the EEs resulted in a more randomized density distribution of EEs in the lattice.

The EEs in the PAE-EE assemblies are classical particles and as such follow a Boltzmann distribution. Thus, the movement of EEs in time is related to free-energy barriers that allow for an exponential relationship between their trapping time and temperature (Fig. 2G), which is directly related to the degree of EE delocalization. To visualize the spatial distribution of EEs from MD simulations, we calculated the cumulative density of EEs by integrating fk from its maxima (where the density of EEs is the highest) and drew isosurfaces that separate the unit cell into equally probable accessible volumes for the EEs, termed Boltzmann volumes (supplementary materials). The Boltzmann volumes for an assembly with a low number of DNA linkers per EE were widely dispersed across the volume of the crystal (Fig. 2H). Increasing the number of linkers per EE resulted in the localization of EEs into a group of locations near the B sites in AB6-type binary compounds (Fig. 2I and Table 1). This localized state is similar to the electron charge-density distributions in semiconductors and insulators (30). Furthermore, the Boltzmann volumes became more dispersed as the EE:PAE ratio increased (Fig. 2J and fig. S33). Such a response was also experimentally observed by comparing the projected EE locations determined with cryo-TEM on samples of bcc assemblies with varying input EE:PAE ratios and EE linker concentrations in the solution. The EE local density in the crystal with a low EE:PAE ratio and high linker concentration (Fig. 2C) around the predicted localization sites was substantially higher than the uniform distribution baseline and than the local densities in the assemblies with higher EE:PAE ratios and lower linker concentrations (Fig. 2B and fig. S17).

Table 1 Symmetry of PAE-EE assemblies in the fully localized state.

Wyckoff positions in square brackets have higher energies than the ground-state configurations.

View this table:

Compared with PAEs that reside on relatively fixed lattice sites, EEs in a PAE-EE assembly are macroscopically mobile beyond local vibrations. Time-series MD simulation snapshots showed that the EEs could diffuse between unit cells (Fig. 3A, fig. S39, and movie S4). To further probe the diffusion of EEs in experimentally realized assemblies, 10-nm PAEs and Cy5-DNA–labeled 1.4-nm EEs were assembled in the bcc structure. Subsequently, these crystals were incubated with a solution of Cy3-DNA–labeled EEs. To track any exchange of EEs between the crystalline assembly and EEs in solution, the ultraviolet-visible (UV-vis) extinction spectra of the supernatant were measured over time (Fig. 3B, top). A simultaneous increase of Cy5 signal and reduction of Cy3 signal suggested that the EEs were highly mobile and could diffuse macroscopically between the colloidal assemblies and solution. In comparison, no appreciable exchange events were observed for the bcc assemblies formed by 10 + 10 nm PAEs modified with identical dye-labeled DNA (Fig. 3B, bottom), suggesting a much weaker diffusion of PAEs as compared with that of EEs. In both cases, the crystallinity of the assemblies was preserved after the exchange (fig. S18).

Fig. 3 Diffusion of EEs in PAE-EE assemblies in the bcc structure.

(A) Trajectory of one EE over time in a lattice of PAEs (yellow) from the MD simulation. The color of the EE positions (red to green to blue) represents diferent time points. (B) The exchange of dye-labeled particles between crystalline lattices and solution was monitored by the change of light extinction in solution by means of UV-vis spectroscopy. Cy5-DNA–labeled EEs were exchanged from “metallic” PAE-EE assemblies (10 + 1.4 nm, bcc) by Cy3-DNA–labeled EEs in the supernatant over 24 hours (top), whereas no appreciable exchange was observed for PAE-PAE assemblies (10 + 10 nm, bcc) (bottom). (C and D) Predicted colloidal crystal metallicity (Mcc) in bcc assemblies with different number of linkers per EE. The Mcc value has a minimum against the EE:PAE ratio at 6:1 (C) but is monotonic against temperature (D).

The tunable spatial density distribution of EEs (based on number, temperature, and linker density) and their macroscale diffusion inside a crystalline PAE framework establish the concept of colloidal crystal metallicity, Mcc, asMcc = Scl – ln(Ncell)(2)where Ncell is the number of EEs per unit cell and is used to make the metallicity independent of crystal unit cell and to reduce to the ideal entropy of a gas in the limit of noninteracting particles. In Eq. 1, the EEs are considered as a group (Fig. 2F), but in Eq. 2, Mcc is a quantitative measure of the average degree of delocalization per EE, which can decrease when the EE:PAE ratio increases (Fig. 3C) even if Scl increases or remains nearly the same (Fig. 2F). A metallicity minimum was attained at EE:PAE = 6:1, suggesting that the metal-like bonds in the colloidal system become saturated. Increasing the number of duplexed strands on EEs led to crystals with lower Mcc values (Fig. 3C) because the EEs were more localized or trapped as the frequency of DNA binding events increased. When the system temperature increased, Mcc also increased (Fig. 3D) because of the decrease in the number of hybridized sticky ends. These results show that the EEs can undergo a classical (nonquantized) process analogous to a metal-insulator transition observed in atomic solids (31) because the degree of EE delocalization and diffusion drastically changed when either the temperature, the number of linkers per EE, or the EE:PAE ratio was varied.

Colloidal crystal metallicity depends on the number of particles, the number of DNA strands that can engage in bonding, and the strength of the bonds formed. Because these parameters are difficult if not impossible to control in purely electrostatic systems owing to electroneutrality requirements, metallicity has been neither observed nor defined in conventional ionic colloidal crystals (crystals formed from oppositely charged particles) (4, 9). The complementary DNA design on NPs ensures that a crystal will not form from PAEs alone (and vice versa from EEs alone), which distinguishes this system from a description of small particles doped in a crystalline solid formed from larger particles. Moreover, although this concept of metallicity superficially resembles “sublattice melting” in superionic conductors (32), in which one sublattice of an ionic compound loses long-range order while the other is fixed (for example, AgI), ion diffusion in such systems is mediated by Frenkel defects because of the constraint of charge balance. Thus, ionic systems have limited tunability and more strict stoichiometry requirements compared with those of the colloidal crystals formed through DNA-directed assembly events.

New phases were accessed by adjusting the input EE:PAE ratio in solution and the total DNA coverage on EEs. For example, as the input EE:PAE ratio was progressively increased, an fcc lattice (space group Fm3¯m) emerged (Fig. 4, A and C). The increase in EE:PAE ratio resulted in stronger cumulative bonding interactions (there are more DNA bonding connections under such circumstances), which was reflected by the increase in the crystal melting temperature, Tm, from 31° (bcc) to 41°C (fcc) (Fig. 4B). In addition, if both the total DNA coverage on EEs (characterized by the total number of duplexed and nonduplexed strands) and the input EE:PAE ratio in solution increase, the Frank-Kasper A15 phase (space group Pm3¯n) emerged (Fig. 4, D and E). The formation of the A15 phase may be associated with its tendency to minimize the contact area between repulsive PAEs, which is similar to the trend observed in dendrimer assemblies (fig. S27) (33) but different from the Cr3Si structure in binary superlattices (14, 34). These phases were also observed in the simulations (table S9). The three phases realized in this system mimic the metal tungsten, in which bcc, fcc, and A15 structures have all been realized either in the bulk or in thin films (35) (table S6).

Fig. 4 Equilibrium phases realized by PAE-EE assemblies.

(A) Schematic representation of the equilibrium conditions of bcc, fcc, and A15 phases and (B) their corresponding thermal melting transitions. (C) SAXS spectra showing the equilibrium phase transition from bcc (red, input EE:PAE = 10:1) to a bcc/fcc mixture (purple, input EE:PAE = 20:1), and then to a majority fcc phase (blue, input EE:PAE = 40:1). (D) Experimental (green) and simulated (black) SAXS spectra of A15 assemblies. (E) Cryo-TEM image of an A15 assembly. (Inset) A15 lattice model along the [001] direction. (F and G) Simulated Boltzmann volumes of fcc (F) and A15 (G) phases as a function of the number of linkers per EE and EE:PAE ratio at a constant temperature (kBT = 1.30). A whole graph is reconstructed by 1/8 of the unit cells from each combination. (H and I) Predicted colloidal crystal metallicity (Mcc) in (H) fcc and (I) A15 assemblies as a function of the number of linkers per EE and EE:PAE ratio. Two low-metallicity configurations (gray shades) are present in (I).

Critically, the MD simulations allowed us to identify the positions that the EEs reside in at low Mcc values. For example, when the number of DNA linkers per EE was maximized and the EE:PAE ratio was decreased, the EEs settled into distinct locations in the fcc and A15 assemblies, as shown by the Boltzmann volumes (Fig. 4, F and G) and the corresponding Mcc values (Fig. 4, H and I). The localized lattice for the fcc structure resembled a fully filled high-temperature Cu2Se lattice (fig. S34, A4B40), whereas the A15 structure shows two possible configurations, either clathrate type I (fig. S35A, A8B46) (36) or an unreported lattice (fig. S35B, A8B82). The latter configuration, in which Mcc reached a local minimum, contained all sites in the clathrate structure that were fully occupied, and the additional EEs that could not occupy the lowest energy positions began to fill the higher-energy 12f and 24j positions (Table 1).

Taken together, this work makes the case for describing certain classes of colloidal crystals in a fundamentally new way, in which, in the case of mobile particles (EEs), the concept of metallicity becomes important. By understanding the factors that govern EE diffusion and delocalization, we have a better understanding of the structures and phases that can be accessed through colloidal crystals, potentially including metals, intermetallics, and complex metal alloys. It also challenges the colloidal science community to identify exotic new properties that arise from the PAE-to-EE transition and structures that exhibit high degrees of metallicity as well as to develop theoretical models that capture the effects that lead to metallicity.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S39

Tables S1 to S11

References (3752)

Movies S1 to S6

References and Notes

  1. When characterized by means of SAXS and electron microscopy, the 10 + 10 nm assembly shows a bcc lattice because the complementary DNA-functionalized Au NPs are indistinguishable.
Acknowledgments: The authors thank H. Lopez-Rios [Northwestern University (NU)] and M. G. Kanatzidis (NU) for helpful discussions, A. M. Geller (NU) for rendering the Boltzmann volume data, E. W. Roth (NU) for ultramicrotomy, and J. Remis (NU) for cryo-TEM tomography. Funding: This material is based on work supported by the Center for Bio-Inspired Energy Science (CBES), an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE) Office of Basic Energy Sciences (DE-SC0000989, for computational studies), the Air Force Office of Scientific Research (FA9550-17-1-0348, for synthesis, spectroscopy, and electron microscopy), the Vannevar Bush Faculty Fellowship program sponsored by the Basic Research Office of the Assistant Secretary of Defense for Research and Engineering and funded by the Office of Naval Research (N00014-15-1-0043), the Sherman Fairchild Foundation (for electron microscopy and computational support), and the Biotechnology Training Program of NU (for cryo-TEM). This work made use of facilities at the NUANCE Center at NU (NSF ECCS-1542205 and NSF DMR-1720139), the Structural Biology Facility at NU (NCI CCSG P30 CA060553), and the DuPont-Northwestern-Dow Collaborative Access Team (DND-CAT) of the Advanced Photon Source (APS) Sector 5 (DOE DE-AC02-06CH11357). Author contributions: C.A.M. and M.O.d.l.C. directed the research. M.G. performed simulations. S.W. and A.D. performed synthesis and x-ray scattering experiments. J.S.D., S.W., and Z.H. performed electron microscopy studies. All authors contributed to data analysis and manuscript preparation. Competing interests: The authors declare no competing interests. Data and materials availability: All data needed to evaluate the conclusions in this manuscript are present in the main text or the supplementary materials. Additional data or codes are available upon request to the corresponding authors.
View Abstract

Stay Connected to Science

Navigate This Article