Polyelemental nanoparticle libraries

See allHide authors and affiliations

Science  24 Jun 2016:
Vol. 352, Issue 6293, pp. 1565-1569
DOI: 10.1126/science.aaf8402

Multimetal nanoparticle synthesis

Multicomponent nanoparticles can be difficult to synthesize. Rather than mixing in one type of particle, the compounds often separate and form distinct particles. Using dip-pen lithography, Chen et al. show how adding reactants to very small volumes forces the reactants to form single particles containing various combinations of five different transition metal ions. Scanning transmission electron microscopy and energy-dispersive x-ray spectroscopy revealed the shapes of the nanoparticles and how metallic composition varied within them. For example, the quinary particle containing gold, silver, cobalt, copper, and nickel consisted of three domains of binary alloys.

Science, this issue p. 1565


Multimetallic nanoparticles are useful in many fields, yet there are no effective strategies for synthesizing libraries of such structures, in which architectures can be explored in a systematic and site-specific manner. The absence of these capabilities precludes the possibility of comprehensively exploring such systems. We present systematic studies of individual polyelemental particle systems, in which composition and size can be independently controlled and structure formation (alloy versus phase-separated state) can be understood. We made libraries consisting of every combination of five metallic elements (Au, Ag, Co, Cu, and Ni) through polymer nanoreactor–mediated synthesis. Important insight into the factors that lead to alloy formation and phase segregation at the nanoscale were obtained, and routes to libraries of nanostructures that cannot be made by conventional methods were developed.

Unary, binary, and even ternary nanostructures can be synthesized by incorporating up to three metals into a single nanoparticle (NP) in an alloyed or phase-segregated state with a variety of synthetic techniques (19). The controlled combination of multiple metals at the nanoscale offers an effective way to tune the chemical and physical properties of NPs, either (i) by facilitating hybrid chemical, electronic, and magnetic interactions between metal components (2, 3, 10, 11); or (ii) by combining different properties associated with each pure component (12, 13). These properties make multimetallic NPs promising materials for diverse applications, spanning catalysis (2, 3, 10, 1416), plasmonics (11, 17), therapy (12), and biological imaging (13, 18). Moreover, in the context of both catalysis and plasmonics, a multimetallic NP often has properties that exceed the capabilities of unary particles consisting of pure forms of each element (2, 3, 1417).

To date, several tools have been developed for screening the composition space of macroscopic multicomponent materials. Notably, Mallouk et al. used ink-jet printing to prepare a combinatorial library of multimetallic catalysts that resulted in the discovery of a quaternary electrochemical catalyst (19). However, at the nanoscale, controlled synthesis and positioning of multimetallic nanofeatures become increasingly important for systematic study. In NP synthesis, important advances have been made in bulk solution–based (either one-pot or stepwise) (57), microfluidic (8), and microemulsion methods (9), but none of these offers the ability to systematically examine compositional diversity in a high-throughput manner or beyond three-element particles. In contrast, scanning-probe block copolymer lithography (SPBCL) (2023), which can deliver attoliter-scale volumes of polymers with metal ion precursors to a desired location on a surface, can be used to generate nanoreactors, which upon thermal treatment can transform the precursors into individual site-isolated NPs. This method has proven useful for making bimetallic NPs (22, 23) and even a single trimetallic alloy NP (23). Although this is an impressive demonstration of synthetic capabilities, such structures could also be made by other methods.

Because SPBCL can now be used with millions of tips over areas as large as square centimeters (20, 23, 24) to generate nanoscale features consisting of, in principle, a limitless number of polymer types, the technique could become a powerful approach to nanocombinatorics, in which new particle compositions, including those not accessible by conventional techniques, can be generated en masse and characterized and screened in a systematic and site-isolated fashion. To test this hypothesis, we investigated the combinatorial synthesis of particles consisting of five different elements, Au, Ag, Cu, Co, and Ni, via SPBCL. We showed that all combinations of the binary, ternary, quaternary, and quinary NPs could be independently synthesized by SPBCL and characterized by scanning transmission electron microscopy (STEM) and energy-dispersive x-ray spectroscopy (EDS).

We chose Au, Ag, Cu, Co, and Ni as example elements because they are often components of nanomaterials used in a wide variety of fields, spanning catalysis (10, 14), plasmonics (11, 17), magnetics (25), electronics (26), and biology (12, 13). In addition, there are readily available salt precursors that can be easily incorporated into block copolymers such as poly(ethylene oxide)-b-poly(2-vinyl pyridine) (PEO-b-P2VP). To synthesize NPs composed of these elements, a polymer preloaded with the appropriate metal salt(s) was used as an ink and deposited onto a substrate as hemispherical domes via dip-pen nanolithography (DPN) (Fig. 1A) (27). After deposition, the polymer domes were thermally annealed in a two-step procedure. The first step was carried out in Ar at 120°C for 48 hours and induced aggregation of the metal salts within the polymer dome (22). The second step was carried out under flowing H2 at 500°C for 12 hours, which results in decomposition of the polymer as well as reduction and coalescence of the aggregate metal ion precursors into a single-metal NP (22). Moreover, the long-term thermal treatment at relatively high temperature allows the constituent metals to adopt a stable configuration (alloy or phase-separated) based on the compatibility of the elements. This approach allows one to make and study all combinations of the unary, binary, ternary, quaternary, and quinary NPs in a systematic and controlled manner [Fig. 1, B to F, and figs. S1 to S4 (28)].

Fig. 1 The SPBCL-mediated synthesis of multimetallic NPs and a five-element library of unary and multimetallic NPs made via this technique.

(A) Scheme depicting the process: A polymer loaded with metal ion precursor is deposited onto a substrate in the shape of a hemispherical dome via dip-pen nanolithography. After two-step thermal annealing, the metal precursors are aggregated and reduced, the polymer is decomposed, and individual NPs result from each dome feature. The interwoven patterns in the NP schemes are only meant to indicate the alloying of the elements and not the actual atomic structure. (B) Unary NPs (top row is a color-coded diagram of the anticipated result; bottom row is each particle, as characterized by EDS mapping). (C) Binary NPs consisting of every two-element combination of the five metals; the alloy versus phase-segregated state was consistent with the bulk phase diagrams for each two-element combination. (D) Ternary NPs consisting of every three-element combination of the five metals; the prediction of alloy versus phase-segregated state was based on the miscibility of the three components, extracted from the binary data. (E) Quaternary NPs consisting of every four-element combination of the five metals; the prediction of alloy versus phase-segregated state was based on the miscibility of the four components, extracted from the binary and ternary data. (F) A quinary NP consisting of a combination of Au, Ag, Cu, Co, and Ni; the prediction of alloy versus phase-segregated state was based on all of the previous data for the unary through quaternary systems.

To evaluate the capability of SPBCL to develop a combinatorial library of NPs, we first prepared the simplest structures, monometallic NPs, on TEM grids using ink solutions containing only one type of metal ion precursor. The resulting NPs were characterized by high-angle annular dark-field (HAADF) STEM and EDS. Fig. S1 (28) shows HAADF-STEM images of SPBCL-prepared Au, Ag, Cu, Co, and Ni NPs with diameters deliberately tailored to ~40 nm. The sizes of NPs were controlled by adjusting the volume and metal loading of the polymer nanoreactors, which dictates the number of metal ions used for forming each NP (2022). EDS analysis provides an effective way to reveal the chemical composition and elemental distribution of each NP product (Fig. 1B and fig. S1).

As part of the library, bimetallic NPs were accessed by equally loading two metal ion precursors into the polymer ink (Fig. 1C and fig. S2). Among the 10 binary combinations consisting of Au, Ag, Cu, Co, and Ni, four of them (AuAg, AuCu, CuNi, and CoNi) resulted in alloy NPs, as validated by the even contrast in HAADF-STEM images of NPs and the overlap of EDS element maps (Fig. 1C and fig. S2). In contrast, the other six combinations resulted in heterodimers, indicative of phase segregation in each binary mixture. The difference in contrast in the HAADF-STEM images of these heterostructured particles reflects the difference in atomic number between two metal domains, whereas EDS analysis confirms the separation of the majority of the elements that make up the NPs. Slight mixing between incompatible metals may still occur, but if so, it is below the detection limit of EDS. The orientation of phase boundaries in heterostructured NPs is random with respect to the substrate (i.e., the image plane). Here, particles with phase boundaries perpendicular to the substrate are used to clearly show the separation of metals. Although there is no reason to believe that nanostructures will necessarily follow the mixing behavior associated with bulk phase diagrams, the binary NPs we explored and prepared do (23, 29). Because no one knows how three-, four-, and five-component particle systems will behave, it is important to build a library of nanostructures in successively more complex fashion, so that each layer of complexity can be used to understand the next series of entries (e.g., ternary structures).

The syntheses of ternary NPs via conventional methods are usually constrained by the difficulties in balancing the reduction kinetics of the different metal ion precursors (1, 30), controlling the site-selective nucleation of each metal as opposed to forming multicomponent structures (5, 6, 31), and controlling phase separation at the nanoscale (32, 33). However, trimetallic particles are easily accessible via SPBCL because this method preconfines precursor ions in a single nanoreactor and subsequently transforms them into a single particle, thus bypassing the issues involving precursor reduction potentials and element-specific nucleation. To evaluate the scope of this approach, we synthesized all of the ternary NPs consisting of every combination of Au, Ag, Cu, Co, and Ni. Although the study of binary particles shows the possibility of controlling particle structure based on metal compatibility, the combination of metal compatibility in ternary particles is more complicated, which includes four possible combination types. The first combination type (I) occurs when all three components are miscible with each other at any composition, which results in a single alloy particle (23). This combination type has been made before but was not observed in the library we generated. For the 10 ternary combinations of Au, Ag, Cu, Co, and Ni, the other combination types that occur are ones in which (II) two metals are miscible with each other but the third is not with either, (III) all three metals are immiscible with each other, and (IV) two metals are immiscible and the other is miscible with the other two.

We first explored the combination of Au, Ag, and Ni (Fig. 2, A to E and figs. S5 to S9), which resulted in type II particles, in which the first two metals are miscible (Au and Ag) and the third one (Ni) is immiscible with the other two (29). We synthesized AuAgNi NPs on silicon substrates or on TEM grids using an ink solution with a metal loading ratio of 25% Au, 25% Ag, and 50% Ni. As shown by scanning electron microscopy (SEM) and HAADF-STEM images (figs. S5 and S6), arrays of AuAgNi NPs could be synthesized, and only one NP was observed in each position within the array. Increasing the magnification revealed that each AuAgNi NP was heterogeneously structured (Fig. 2, A and B, and fig. S6). Each AuAgNi NP was a dimeric heterostructure consisting of an AuAg alloy domain and a Ni domain. The contrast in the HAADF-STEM image arises from the difference in atomic number between the AuAg alloy and Ni. High-resolution TEM shows that the two domains in the AuAgNi NP are crystalline (Fig. 2C and fig. S5B). More than 10 million NPs were generated simultaneously over the centimeter scale by combining SPBCL with a large-area cantilever-free scanning probe technique such as polymer pen lithography (24) (fig. S8). Finally, in addition to AuAgNi, the other five ternary combinations that lead to type II particles are AuAgCo, AuCuCo, AuCoNi, AgCuNi, and AgCoNi (Fig. 1D and fig. S3).

Fig. 2 AuAgNi and AgCuCo heterostructured NPs generated by SPBCL.

(A) HAADF-STEM image and EDS elemental mapping, and (B) schematic illustration of a representative AuAgNi NP (24% Au, 24% Ag, 52% Ni). Au and Ag are miscible and both elements are immiscible with Ni, resulting in a heterodimer. (C) HR-TEM image of an AuAgNi NP. The observed lattice spacings near the interface are 2.36 and 2.02 Å (extracted from fast Fourier transform, fig. S5B), which closely match the AuAg alloy (111) and Ni (111) planes, respectively. (D) A comparison of the NP composition (20 to 60 nm) with the precursor ink composition. The dashed black line is the ideal 1:1 correlation case. (E) HAADF-STEM image and corresponding EDS mapping of AuAgNi NPs as a function of Ni content. Scale bars, 10 nm. (F) EDS elemental mapping, (G) schematic illustration, and (H) HAADF-STEM image of a representative AgCuCo NP (35% Ag, 38% Cu, 27% Co). The Ag, Cu, and Co are immiscible with each other, resulting in a heterotrimer. (I) Dark-field STEM images and EDS elemental mapping of AgCuCo NPs consisting of different compositions. From left to right, the content of Cu is approximately constant at 45% and the content of Ag is 39, 25, and 14%, respectively. Scale bars, 15 nm.

One advantage of using SPBCL for generating NPs is the homogeneity of the patterned nanoreactor, which yields particles that have metallic constituent ratios reflecting those of the ink. This enables straightforward tuning of the particle composition (23). We used this ability to study the effect of altering metal ratios on the particle structure by examining five ink solutions (Fig. 2, D and E). The five solutions contained increasing Ni content from 20 to 80% while a 1:1 ratio of Au and Ag was maintained; the NPs were then investigated by EDS to determine elemental distribution and composition. As shown in Fig. 2E, the varying Ni content in the ink solution effectively changed the relative size of the Ni domains and AuAg alloy domains, respectively, while the dimeric heterostructure of AuAgNi NPs was maintained for all compositions studied. Experimentally, the majority of the AuAgNi NPs (yield, 80%; sample size, 50) matched the composition of the ink solutions, with a compositional variation of less than 10% (Fig. 2D and fig. S9). The remaining 20% of the NPs exhibited large composition fluctuations and even included bimetallic NPs (i.e., AuAg, AuNi, or AgNi), which is probably caused by the imperfect distribution of precursors in the ink solution. The formation of one NP per polymer reactor is crucial for this composition control. Indeed, attempts to synthesize NPs in a bulk polymer solution through a similar thermal annealing procedure resulted in heterogeneous mixtures of particles of varying size and composition (figs. S10 and S11), as is commonly observed in one-pot solution syntheses of multimetallic NPs, because different metal precursors exhibit different reduction kinetics and element-specific nucleation (5, 30). In some polymer reactors (ones consisting of a features diameter >1 μm), failure of the atomic precursors to coalesce into a single NP was also observed, which prevented control over NP size and composition because of the formation of multiple particles (fig. S12).

In combination type III, the three constituent metals, such as Ag, Cu, and Co, are not miscible with each other, and trimeric heterostructures were expected. As a proof-of-concept experiment, we synthesized AgCuCo NPs using an ink solution with a metal loading of 33% for each metal. HAADF-STEM images along with EDS characterization confirm that Ag, Cu, and Co segregate into three segments in a AgCuCo NP (Fig. 2, F to H, and figs. S13 and S14), which is consistent with the compatibility of these metals (29). If the Cu content was kept constant while the relative loading ratio of Ag and Co in the ink solution was changed, the sizes of the Ag and Co domains, respectively, varied according to the ink formulation, but no apparent alloying of either two metals was observed for all compositions examined (Fig. 2I). In the case of all ink compositions studied, the three metal domains in the NPs shared a common structural feature: A central Cu domain is capped by an Ag or Co domain on each end. The phase boundaries between Ag-Cu and between Cu-Co are not always parallel, as evidenced by STEM and EDS (figs. S15 and S16). This structural motif for AgCuCo NPs suggests a higher interfacial energy between Ag and Co than the interfacial energy between Ag and Cu or between Cu and Co. Nevertheless, AgCuCo NPs with adjacent Ag-Co domains can be realized when the Cu content is <20%. Indeed, under these circumstances, the amount of Cu is not sufficient to completely segregate the Ag and Co domains (fig. S17).

For the final combination type, IV, two metals (Ag and Cu) are not miscible but the third one (Au) is miscible with the other two. Compared with the previous two types, the particle structure is more dependent on composition because of the increased complexity of interactions between the metals. To further show the advantage of SPBCL in studying composition-dependent phase separation in multimetallic NPs, we synthesized AuAgCu NPs using four ink formulations. The four solutions contained increasing Au content from 10 to 70% while the Ag and Cu contents were kept equal in each solution. The NPs obtained from these inks are shown in Fig. 3 and figs. S18 to S21. When the Au content in the NPs was only 10%, Ag and Cu phase-segregated into two domains (Fig. 3A and fig. S18). Experimentally, Au showed a higher affinity for Cu, as indicated by the enrichment of Au in the Cu domain. When the Au content was increased to 30%, Au diffused into the Ag domain, resulting in a Janus NP in which one part is an AuAg alloy and the other part is an AuCu alloy (Fig. 3B). Further increasing the Au content to 50% resulted in NPs without a distinct boundary between Ag and Cu (Fig. 3C). Instead, the Ag and Cu started to diffuse into each other’s domains, indicating the formation of an intermediate region in the center that consists of an AuAgCu alloy. Finally, when the Au content was increased to 70%, Au was found throughout the entire NP structure and formed an alloy (Fig. 3D). In addition to AuAgCu, the other two ternary combinations that satisfy type IV are AuCuNi and CuCoNi. Similar to the AuAgCu system where Au shows preference for Cu over Ag, Cu had a higher affinity for Au than Ni in the AuCuNi system, and Ni had a higher affinity for Co than Cu in the CuCoNi system (Fig. 1D and fig. S3). Less noble metals such as Co, Ni, and Cu were readily reduced to a metallic state under an H2 atmosphere under annealing conditions, as evidenced by XPS measurements (figs. S7, S14, and S21). Particles with these elements slowly oxidized upon exposure to air.

Fig. 3 AuAgCu heterostructured NPs.

(A to D) Schematic illustration, HAADF-STEM images, and EDS elemental mapping of four representative AuAgCu NPs with different Au content. From (A) to (D), the Ag content and Cu content are equal in each NP and the Au content is 10, 30, 50, and 70%, respectively. The dashed white line in (A) and (B) indicates the position of the phase boundary. Scale bars, 20 nm.

We further extended the library by synthesizing all quaternary combinations of Au, Ag, Cu, Co, and Ni. The structures of the quaternary NPs were highly dependent on particle composition. Here, we examined one specific ink composition for each combination of metals (Fig. 1E and fig. S4). The aforementioned studies on bimetallic and trimetallic systems are instructive for elucidating and assigning the structure of each quaternary NP. For example, the structure of the AuAgCoNi NP was assigned as an AuAgCo heterodimer primary particle and Ni as the fourth component. The location of the fourth component depends on its miscibility with the three components in the primary particle. Ni is miscible with Co but immiscible with Au and Ag. Thus, AuAgCoNi NPs still adopt a dimeric structure, with one part being an AuAg alloy and the other part a CoNi alloy. For the other quaternary particles synthesized, the structure can be similarly understood by the miscibility of components (fig. S4), which we previously described in the binary and ternary analysis. The incorporation of four components in one NP via SPBCL enables the creation of nanostructures with unprecedented compositional and structural complexity.

Creating combinatorial libraries with SPBCL even allowed for the synthesis of higher-order nanostructures, such as pentametallic NPs consisting of Au, Ag, Cu, Co and Ni, using an ink solution with an equal loading of the five ion precursors (Figs. 1F and 4 and fig. S4). Composition control was more difficult than with the lower-order NPs. Experimentally, we successfully formed quinary particles from half of the polymer reactors patterned by SPBCL (yield, 50%; sample size, 50; figs. S22 to S24). The remaining polymer reactors generated tetrametallic NPs. The AuAgCuCoNi NPs had three distinct structural domains (Fig. 4B and figs. S25 and S26). The overlay of EDS element maps revealed the distribution of metals and their spatial relationship. AgCuCo segregated into three domains, which formed the primary NP (Fig. 4E). Au was present in the Ag and Cu domains because it is compatible with Ag and Cu but immiscible with Co (Fig. 4D). Ni was only present in the Co domain because it phase-segregates from Au and has a higher affinity for Co than Cu (Fig. 4F). Therefore, AuAgCuCoNi NPs composed of three segments (i.e., an AuAg alloy, an AuCu alloy, and a CoNi alloy) were obtained via SPBCL (Fig. 4G). The structure of the AuAgCuCoNi NP was highly composition-dependent. Other substrates or annealing conditions may result in different particle structures, providing the potential for additional structure tuning. As such, this result validates SPBCL as a viable platform for synthesizing and studying previously undiscovered multimetallic NPs that are defined by fundamentally interesting and complex heterostructures.

Fig. 4 Quinary heterostructured NPs.

(A) HAADF-STEM image of a typical AuAgCuCoNi NP (19% Au, 24% Ag, 28% Cu, 14% Co, 15% Ni). (B) Schematic illustration of the structure of the AuAgCuCoNi NP in (A). The NP phase segregates into three domains: AuAg alloy, AuCu alloy, and CoNi alloy. (C to G) EDS elemental mapping of the AuAgCuCoNi NP in (A). (C) Distribution of each metal inside the NP. (D) Overlay of the element maps of Au, Ag, and Cu. (E) Overlay of the element maps of Ag, Cu, and Co. (F) Overlay of the element maps of Co and Ni. (G) Overlay of the element maps of all five metals.

The ability to synthesize and characterize such structures provides an experimental platform to study alloy formation and phase segregation at the nanoscale, which is important for understanding both structure and function. Given the enormous library of nanostructures that can be tailored based on particle composition and metal compatibility, this work advances the field of multimetallic NPs toward higher compositional diversity and structural complexity, which has the potential to affect a broad range of fields, such as catalysis (10, 14, 19), plasmonics (11, 17), magnetics (25), electronics (26), biology (13, 18), and medicine (12).


Materials and Methods

Figs. S1 to S26

Reference (34)


  1. See the supplementary materials on Science Online.
Acknowledgments: This material is based on work supported by GlaxoSmithKline, the Air Force Office of Scientific Research (award FA9550-12-1-0141), and the Asian Office of Aerospace R&D (award FA2386-13-1-4124). The authors thank B. R. Meckes for helpful discussion. P.-C.C., S.W., and Q.-Y.L. gratefully acknowledge support from the Ryan Fellowship and the Northwestern University International Institute for Nanotechnology. X.L. and M.C.H. acknowledge support from the Materials Research Science and Engineering Center of Northwestern University (grant NSF DMR-1121262). V.P.D. acknowledges support by the Air Force Research Laboratory and the National Science Foundation under the Ceramics program. J.L.H. acknowledges support from the U.S. Department of Defense through the National Defense Science & Engineering Graduate Fellowship Program. This work made use of the EPIC facility of the NUANCE Center at Northwestern University, which has received support from the Soft and Hybrid Nanotechnology Experimental (SHyNE) Resource (NSF NNCI-1542205). A patent has been issued on this work: C. A. Mirkin, G. Liu, D. J. Eichelsdoerfer, K. A. Brown, “Method for synthesizing nanoparticles on surfaces” World Intellectual Property Organization Patent 2014039821-A1.

Stay Connected to Science

Navigate This Article