Report

Tunable intraparticle frameworks for creating complex heterostructured nanoparticle libraries

See allHide authors and affiliations

Science  04 May 2018:
Vol. 360, Issue 6388, pp. 513-517
DOI: 10.1126/science.aar5597

Retrosynthesizing complex nanostructures

The solution synthesis of complex and asymmetric nanostructures is still challenging. For many applications, it will be important to gain simultaneous control over particle size and morphology, constituent materials, and internal interfaces. Fenton et al. have developed a strategy that mimics chemical retrosynthesis, starting with simple nanoparticle synthons—in this case, Cu1.8S nanoparticles, nanorods, and nanosheets. Various types of interfaces and junctions can be introduced, for example, by cation substitution. This intervention breaks the symmetry of the synthons and assembles them into higher-order structures. The nanostructures can thus be formed with asymmetric, patchy, porous, or sculpted regions.

Science, this issue p. 513

Abstract

Complex heterostructured nanoparticles with precisely defined materials and interfaces are important for many applications. However, rationally incorporating such features into nanoparticles with rigorous morphology control remains a synthetic bottleneck. We define a modular divergent synthesis strategy that progressively transforms simple nanoparticle synthons into increasingly sophisticated products. We introduce a series of tunable interfaces into zero-, one-, and two-dimensional copper sulfide nanoparticles using cation exchange reactions. Subsequent manipulation of these intraparticle frameworks yielded a library of 47 distinct heterostructured metal sulfide derivatives, including particles that contain asymmetric, patchy, porous, and sculpted nanoarchitectures. This generalizable mix-and-match strategy provides predictable retrosynthetic pathways to complex nanoparticle features that are otherwise inaccessible.

Nanoparticles (NPs) with compositional asymmetry and various types of morphologies and interfaces can be produced by top-down templating, surface patterning, and localized deposition (16). Because these approaches use planar surfaces to achieve the targeted complex features, only a relatively small number of NPs are produced compared with solution routes, which also can achieve size uniformity, morphology control, and tunable composition (7, 8). However, for multicomponent nanostructures, each distinct combination of materials generally requires a “tour de force” synthetic effort. Reaction conditions must be rigorously fine-tuned, and unavoidable incompatibilities among the multiple components and reagents must be sidestepped, typically many times during the synthesis of a single nanoarchitecture (9, 10). The robust modularity that characterizes molecular synthesis and top-down nanofabrication is largely absent from the approaches that are available to produce complex NP systems. Accordingly, the demand for complex nanostructures outpaces the synthetic methods available for obtaining them, and many high-value targets remain out of reach.

By analogy to the retrosynthetic framework through which complex molecules are synthesized, we deconstruct a target structure into smaller, synthetically tractable pieces and define a divergent synthesis strategy to access multicomponent NPs with arbitrary complexity. A key first step in this approach is to identify appropriate synthons that are simple, readily accessible, and can be rationally modified; we refer to these as first-generation (G-1) nanostructures. A structurally complex intraparticle framework of interfaces and junctions is then introduced in a predictable and generalizable manner to produce G-2 nanostructures, followed by systematic incorporation of a diverse library of materials at desired locations to produce higher-generation nanostructures. We selected Cu1.8S as a G-1 synthon because Cu1.8S NPs in a range of sizes and shapes can be routinely synthesized (11, 12) and Cu+ cations have high mobilities that facilitate partial or complete exchange with other cations (1319). These characteristics allow a rich collection of materials and interfaces to be introduced in subsequent steps across a diverse range of morphologies. Transmission electron microscopy (TEM) images are shown in Fig. 1 for three distinct classes of Cu1.8S NPs—zero-dimensional (0D) spheres (Fig. 1A), 1D rods (Fig. 1B), and 2D plates (Fig. 1C); additional characterization data are included in fig. S1 of the supplementary materials.

Fig. 1 Formation of G-2 nanostructures.

Cu1.8S (A) spheres, (B) rods, and (C) hexagonal plates (shown in TEM images) are used as G-1 synthons for introducing diverse intraparticle frameworks through partial cation exchange with Cd2+ and Zn2+. The resulting library of segmented G-2 nanostructures includes (D) hemispherical CdS–Cu1.8S Janus particles, (E) sandwich-like ZnS–Cu1.8S–ZnS particles, (F) CdS–Cu1.8S capped rods, (G) ZnS–Cu1.8S striped rods, (H) CdS–Cu1.8S patchy plates, and (I) ZnS–Cu1.8S marbled plates. For each type of NP, the extent of partial cation exchange can be tuned by adjusting the reaction time. Three examples are shown for each type of nanostructure in (D) through (I). Each panel includes a TEM image and STEM-EDS element map. Cu, Cd, and Zn are shown in red, blue, and green, respectively. Crystal structure projections of wurtzite CdS, roxbyite Cu1.8S, and wurtzite ZnS are shown in (J) to highlight the crystallographic relationships between the adjacent phases within the intraparticle frameworks. Selected regions of the EDS spectra for all samples are shown in (K). For the spheres, rods, and plates, the Cd and Zn signals increase as the reaction time increases, whereas the Cu signals decrease (fig. S4 and table S2). The Ni signal is from the Ni TEM grid.

To transform the G-1 nanostructures into more complex G-2 systems, various types of interfaces and junctions must be introduced to break the symmetry of the synthons and produce structurally complex intraparticle frameworks. We replaced a fraction of the Cu+ cations in each of the G-1 Cu1.8S structures with Cd2+ or Zn2+ by applying the principles of nanocrystal (NC) cation exchange (20), intentionally arresting each reaction at various stages before it went to completion. TEM images and element maps generated from scanning TEM with energy dispersive spectroscopy (STEM-EDS) corresponding to the resulting G-2 NP library (Fig. 1) show that each type of G-2 NP has a distinct segmentation pattern and newly engineered internal interfaces between material components. The extent of cation exchange could be controlled by reaction time, which led to broad tunability of the intraparticle framework features. Spherical Cu1.8S NPs (Fig. 1A) reacted with Cd2+ to produce hemispherical CdS–Cu1.8S Janus particles (Fig. 1D), and reaction with Zn2+ formed two discrete ZnS domains in a layered, sandwich-like ZnS–Cu1.8S–ZnS structure (Fig. 1E) (15). Cu1.8S nanorods (Fig. 1B) reacted with Cd2+ preferentially at the tips to generate CdS–Cu1.8S nanorods (Fig. 1F), whereas striped ZnS–Cu1.8S nanorods were the dominant products formed upon reacting with Zn2+ (Fig. 1G). Hexagonal Cu1.8S nanoplates (Fig. 1C) reacted with Cd2+ to produce large patches of CdS in a Cu1.8S matrix (Fig. 1H), whereas reaction with Zn2+ yielded a highly interdigitated network of ZnS and Cu1.8S with small, marbled filaments throughout (Fig. 1I). Despite the different interfacial structures, the overall sizes and shapes of the G-1 Cu1.8S synthons were retained in the G-2 structures (table S1).

Across the library of G-2 NPs in Fig. 1, exchange with Zn2+ generally resulted in the formation of a larger number of smaller ZnS regions within the intraparticle frameworks, whereas exchange with Cd2+ resulted in a smaller number of larger CdS regions. This behavior can be rationalized by considering differences in cation radius, lattice matching, and interfacial strain (15, 21). Exchanging Cu+ for the larger Cd2+ cation to form CdS requires a volume expansion of 14% relative to the comparable subunit of the Cu1.8S crystal structure (fig. S2), whereas exchange with the smaller Zn2+ cation requires a smaller (10%) volume contraction (22, 23). Intraparticle phase segregation between CdS and Cu1.8S is expected to be greater than between ZnS and Cu1.8S, resulting in fewer interfaces in the CdS–Cu1.8S systems relative to ZnS–Cu1.8S, as observed. Additionally, the a-axis lattice parameters for comparable subunits of the hexagonal close-packed S lattices of Cu1.8S (3.87 Å) and ZnS (3.81 Å) are similar but differ substantially from CdS (4.13 Å), whereas the c-axis lattice parameters are closer for Cu1.8S (6.71 Å) and CdS (6.72 Å) than for ZnS (6.23 Å) (Fig. 1J) (17, 23).

To minimize interfacial lattice distortion, exchange with Cd2+ propagated along the a direction of Cu1.8S, which interfaced Cu1.8S and CdS along the closely lattice-matched c direction. In contrast, exchange with Zn2+ propagated along the c direction, which interfaced Cu1.8S and ZnS along the closely lattice-matched a direction. The high-resolution TEM (HRTEM) images of CdS–Cu1.8S Janus NPs and ZnS–Cu1.8S–ZnS sandwich NPs, shown in fig. S3, confirmed the different crystallographic orientations within the NPs and helped to rationalize the distinct intraparticle frameworks that emerged from the various cation exchange reactions. Thus, ionic radii, lattice matching, and reaction time can be used to predict the number of segments. The extent of cation exchange, the corresponding sizes of the regions each material occupies within the intraparticle framework, and the interfacial area between the material components all can be rationally tuned while maintaining the key morphological features of the G-1 synthons.

The G-2 nanostructures contained a residual Cu1.8S region, as well as either CdS or ZnS (Fig. 1K). All three of these metal sulfides can be targeted by NC cation exchange reactions, but their different chemical driving forces in principle allow each region to be addressed selectively and independently. Cu+ will diffuse faster than either Cd2+ or Zn2+ because of its lower charge density and higher mobility (15). Cu+ is also a softer cation than either Cd2+ or Zn2+, making its coordination with the soft phosphorus atom in trioctylphosphine (TOP), which is included in the reaction solution, more favorable than either of the divalent cations (20, 24). Accordingly, the Cu1.8S regions of the G-2 CdS–Cu1.8S and ZnS–Cu1.8S nanostructures in Fig. 1 could be exchanged selectively, keeping the CdS or ZnS regions intact while introducing other materials into the remainder of the intraparticle framework. After reaction with Zn2+ in the presence of TOP, the G-2 CdS–Cu1.8S nanostructures (Fig. 2, A, C, and E) transformed into G-3 CdS–ZnS (Fig. 2, G, I, and K). Likewise, after reaction with Cd2+ in the presence of TOP, the G-2 ZnS–Cu1.8S nanostructures (Fig. 2, B, D, and F) transformed to G-3 ZnS–CdS (Fig. 2, H, J, and L). The G-3 products retained the overall morphologies and segmentation patterns defined by their G-1 and G-2 precursors, respectively, while containing barely detectable amounts (<3%) of residual Cu (figs. S7 and S8), based on analysis by energy-dispersive x-ray spectroscopy (EDS). Powder x-ray diffraction (XRD) confirmed the presence of each of the expected phases across all three generations of spherical nanostructures (figs. S9 and S10). The observed color changes were also consistent with those expected for the transformations of G-1 NPs into G-2 and G-3 nanostructures (fig. S11).

Fig. 2 Formation of G-3 nanostructures.

G-2 (A) hemispherical CdS–Cu1.8S Janus particles, (B) sandwich-like ZnS–Cu1.8S–ZnS particles, (C) CdS–Cu1.8S capped rods, (D) ZnS–Cu1.8S striped rods, (E) CdS–Cu1.8S patchy plates, and (F) ZnS–Cu1.8S marbled plates are transformed into (G to L) six distinct classes of heterostructured G-3 CdS–ZnS isomers via selective cation exchange of Cu+ with Cd2+ or Zn2+. Selective exchange of Cu+ in the Cu1.8S regions of G-2 CdS–Cu1.8S Janus and ZnS–Cu1.8S–ZnS sandwich particles with Co2+, Mn2+, and Ni2+ results in (M) CdS–CoS, (N) CdS–MnS, and (O) CdS–Ni9S8 Janus NPs and (P) ZnS–CoS–ZnS, (Q) ZnS–MnS–ZnS, and (R) ZnS–Ni9S8–ZnS sandwich particles. The G-3 NPs preserve the morphology of the G-1 Cu1.8S NPs and the intraparticle frameworks of the G-2 NPs while containing none of the original Cu1.8S. STEM-EDS maps and TEM images are shown for each population of particles. Cu, Cd, Zn, Co, Mn, and Ni are shown in red, blue, green, purple, cyan, and orange, respectively. For additional characterization data, see figs. S5 to S8 and S12 to S15.

The resulting library of nanostructures in Fig. 2, G to L, represents six distinct classes of heterostructured CdS–ZnS isomers that have the same material components in different spatial and interfacial arrangements within colloidal NPs that maintain the uniform shapes and sizes defined by the G-1 Cu1.8S synthons. Although selected members of the G-2 CdS–Cu1.8S and ZnS–Cu1.8S NP library were chosen as representative examples for transformation into G-3 products, any of the other G-2 members from Fig. 1 could also be used to further expand the scope of accessible isomeric CdS–ZnS heterostructures.

The process that transformed G-1 Cu1.8S NPs into G-2 CdS–Cu1.8S and ZnS–Cu1.8S NPs and G-3 CdS–ZnS NPs was also extended to other materials systems through cation exchange with different metals; this step could be implemented at any stage of the reaction sequence. For example, the G-1 Cu1.8S nanorods underwent partial cation exchange with Co2+ and Ni2+ to produce G-2 CoS–Cu1.8S and Ni9S8–Cu1.8S striped nanorods (figs. S12 and S13). G-3 CoS–ZnS striped nanorods, formed by exchange of the Cu1.8S region of G-2 ZnS–Cu1.8S striped nanorods, are shown in figs. S12 and S14. Similarly, the residual Cu+ in the G-2 CdS–Cu1.8S hexagonal plates exchanged with Co2+ to produce hexagonal plates of CoS with embedded CdS islands (figs. S12 and S14). Figure 2, M to R, shows the products formed by reacting the G-2 spherical CdS–Cu1.8S Janus NPs and ZnS–Cu1.8S–ZnS sandwich structures with Co2+, Mn2+, and Ni2+ in the presence of TOP. The resulting library of G-3 nanostructures included CdS–CoS, CdS–MnS, and CdS–Ni9S8 Janus NPs and ZnS–CoS–ZnS, ZnS–MnS–ZnS, and ZnS–Ni9S8–ZnS sandwich structures.

The formation of diverse intraparticle frameworks can be decoupled from the integration of targeted materials in a predictive manner, but the modularity and scope of this approach to complex NP synthesis extends far beyond the G-2 and G-3 systems highlighted in Fig. 2. Any point in the reaction sequence can serve as a springboard to a more complex target through a rational and predictive mix-and-match process. In addition to cation exchange, several other classes of nanochemical reactions can be integrated, including seeded growth (25, 26) and selective etching (27). Figure 3 highlights several examples of highly sophisticated nanostructures that would otherwise not be accessible. For example, G-2 ZnS–Cu1.8S striped nanorods (Fig. 3A) were transformed to G-3 ZnS–CdS striped nanorods (Fig. 3B), followed by reaction with Ag+ to produce complex segmented nanorods that formed through regioselective cation exchange (Fig. 3C). Under these mild reaction conditions, Ag+ should exchange with both ZnS and CdS (28). However, with a substoichiometric amount of Ag+, cation exchange only occurred with Cd2+, due to the more closely matched sizes of the Ag+ and Cd2+ cations (22), and the ZnS regions remained intact. Alternatively, the Cu1.8S regions of the G-2 ZnS–Cu1.8S striped nanorods in Fig. 3A were partially etched in the presence of a trialkylphosphine and air (27). The nanorod products maintained the morphologies of the ZnS regions while adopting concave Cu1.8S regions, resulting in intricately sculpted ZnS–Cu1.8S NPs (Fig. 3D). Exchanging the remaining Cu+ in the sculpted segments with Zn2+ transformed them to ZnS, resulting in a 1D NP composed entirely of ZnS but maintaining the complex sculpted morphology (Fig. 3E).

Fig. 3 Higher-generation nanostructures formed by rationally applying multiple exchange and/or etching steps.

(A) G-2 ZnS–Cu1.8S striped rods react with Cd2+ to form (B) G-3 ZnS–CdS striped rods, which further react with Ag+ to selectively target the CdS regions, forming (C) G-4 Ag2S–ZnS striped rods. Alternatively, the Cu1.8S regions of the ZnS–Cu1.8S striped rods in (A) can be partially etched with TOP to form the sculpted ZnS–Cu1.8S rods shown in (D); the Cu+ can then exchange with Zn2+ to replace the remaining Cu1.8S with ZnS, forming the complex sculpted rods in (E) that are composed entirely of ZnS. Partial exchange of Cu+ in the Cu1.8S region of (F) G-2 CdS–Cu1.8S Janus NPs with Zn2+ results in the new type of particle shown in (G), which merges the features of partial Zn2+ and Cd2+ cation exchanges by retaining the CdS hemisphere while introducing sandwich-like ZnS caps on the Cu1.8S hemisphere. The remaining Cu1.8S in these complex (ZnS–Cu1.8S–ZnS)–CdS Janus particles can be (H) etched to produce particles containing two noncontiguous ZnS domains on a CdS hemisphere or (I) replaced with CoS to form (ZnS–CoS–ZnS)–CdS Janus NPs. Alternatively, the Cu+ of the Cu1.8S region of the (ZnS–Cu1.8S–ZnS)–CdS Janus NPs can be exchanged with Cd2+ to form (J) (ZnS–CdS–ZnS)–CdS Janus NPs, which can also be described as a spherical CdS particle containing two distinct embedded regions of ZnS. STEM-EDS maps, TEM images, and drawings are shown for each population of NPs. Cu, Cd, Zn, Ag, and Co are shown in red, blue, green, pink, and purple, respectively. For additional characterization data, see figs. S16 to S18.

To further demonstrate the mix-and-match modularity of this synthetic framework, partial exchange of Cu+ with Zn2+ was carried out on the G-2 CdS–Cu1.8S Janus NPs shown in Fig. 3F. The product, shown in Fig. 3G, combined both Janus and sandwich features by replacing the top and bottom regions of the Cu1.8S hemisphere with ZnS while maintaining the hemisphere of CdS. The resulting NP contained one hemisphere of CdS and one hemisphere of a ZnS–Cu1.8S–ZnS sandwich-type particle, and was further modified in multiple ways to produce complex derivative products. Selectively etching the Cu1.8S region produced a NP with a CdS hemisphere having two attached, noncontiguous ZnS domains (Fig. 3H). Alternatively, exchanging the Cu+ with Co2+ transformed the central region to CoS, producing a (ZnS–CoS–ZnS)–CdS Janus NP (Fig. 3I). As yet another alternative, exchanging the Cu+ with Cd2+ transformed the Cu1.8S region to CdS, resulting in a spherical CdS NP with two embedded regions of ZnS (Fig. 3J).

Other classes of nanostructured NPs could also be intentionally designed with this modular framework. For example, nanoscale spherical colloids lack compositional asymmetry, or valency, on their surfaces and hence are generally unable to facilitate directional or anisotropic interactions with other particles and/or surface-based chemical processes (29). Likewise, many targeted nanoscale architectures, including those for applications in photocatalytic water splitting, require deposition of NPs selectively on different regions of a central particle (30). Figure 4 shows multiple design strategies for achieving such features. The G-2 CdS–Cu1.8S Janus NPs had two distinct surfaces exposed, and Pt deposited selectively on the Cu1.8S region (Fig. 4A). The remaining Cu+ in the CdS–Cu1.8S–Pt NP was then exchanged with Cd2+ to transform the spherical particle entirely to CdS, forming asymmetrically functionalized CdS–Pt (Fig. 4B). The CdS–Cu1.8S intermediate introduced asymmetry that led to regioselective surface deposition of Pt. Au was then deposited on the exposed CdS surface to form asymmetric (Au)x–CdS–Pt NPs (Fig. 4C). For comparison, direct deposition of Pt and/or Au onto spherical CdS particles resulted in Pt and Au NPs decorating the entire surface (fig. S20). To further expand the scope of regioselective functionalization, Au was deposited exclusively on the CdS patches of CdS–ZnS hexagonal plates (Fig. 4D).

Fig. 4 Complex nanostructures formed by combining multiple cation exchange steps with regioselective deposition or etching.

(A) Selective deposition of Pt on the Cu1.8S surface of a G-2 CdS–Cu1.8S Janus particle, followed by (B) subsequent exchange of the Cu+ in the Cu1.8S hemisphere with Cd2+, formed a spherical NP that was fully CdS while being asymmetrically functionalized with Pt on one side. (C) Further deposition of small Au NPs on the exposed CdS surface forms an asymmetric Au-CdS-Pt NP. In (A) to (C), Cu, Cd, Pt, and Au are shown in red, blue, green, and yellow, respectively. (D) Au NPs were selectively deposited on the CdS patches of a G-3 CdS–ZnS patchy hexagonal plate, as shown in the high-angle annular dark-field STEM image. (E and F) The Cu1.8S regions of ZnS–Cu1.8S marbled plates were selectively etched, forming the nanoporous ZnS plates shown in the TEM images and STEM-EDS maps. By tuning the reaction time in the initial partial cation exchange step that produces the ZnS–Cu1.8S marbled plates, the size of the nanopores in the resultant plates can also be tuned. In (D) to (F), Cu, Cd, Zn, and Au are shown in red, blue, green, and yellow, respectively. For additional characterization data, see figs. S19, S21, and S22.

As an additional design goal, introducing porosity into nanostructures increases surface area and is desirable for applications in catalysis (31, 32), gas storage and separation (33), and batteries and supercapacitors (34), where chemical interactions with exposed surfaces must be maximized. By leveraging the diversity of accessible intraparticle frameworks, along with chemistry that can selectively etch Cu1.8S (27), a variety of porous NPs were produced. Two distinct pore sizes were incorporated into morphologically identical ZnS nanoplates by starting with two distinct intraparticle frameworks incorporated into ZnS–Cu1.8S precursors (Fig. 4, E and F).

This modular and divergent synthetic strategy allows us to reenvision the synthesis of multicomponent nanostructures by first carving a substructure of interfaces into a reactive nanoparticle synthon and then integrating desired materials, in mix-and-match fashion, at precise locations. Complex colloidal nanoarchitectures that would otherwise be inaccessible can now be designed and synthesized in a predictive manner. Combining these capabilities with a broader class of nanoparticle chemical transformation reactions will provide synthetic entryways into other materials systems, including oxides, metals, phosphides, and other chalcogenides (20, 3537). Such advances will enable the design and synthesis of the increasingly complex NP systems that are in demand across a wide range of application areas, including semiconductor-semiconductor interfaces for the controllable separation or confinement of excitons, anisotropic particles for nonlinear optics, and precision integration of semiconductors and catalysts for light-driven chemical transformations.

Supplementary Materials

www.sciencemag.org/content/360/6388/513/suppl/DC1

Materials and Methods

Figs. S1 to S22

Tables S1 to S5

References

References and Notes

Acknowledgments: TEM imaging was performed in the Penn State Microscopy and Cytometry facility. HRTEM imaging, STEM imaging, and EDS mapping were performed at the Materials Characterization Laboratory of the Penn State Materials Research Institute. The authors thank K. Wang and J. L. Gray for assistance with TEM experiments and helpful discussions. Funding: This work was supported by the U.S. National Science Foundation under grant DMR-1607135. Author contributions: J.L.F., B.C.S., and R.E.S. conceived of the concept, designed the experiments, and wrote the paper; B.C.S. and J.L.F. synthesized and characterized the nanoparticles; J.L.F. performed HRTEM and STEM-EDS characterization. Competing interests: The authors declare no competing financial interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.
View Abstract

Navigate This Article