Report

Synthesis of ultrasmall, homogeneously alloyed, bimetallic nanoparticles on silica supports

See allHide authors and affiliations

Science  15 Dec 2017:
Vol. 358, Issue 6369, pp. 1427-1430
DOI: 10.1126/science.aao6538

Dispersing small, bimetallic nanoparticles

For applications of nanoparticles in sensing and catalysis, smaller nanoparticles are often more effective because they expose more active surface sites. The properties of metallic nanoparticles can also be improved by creating bimetallic alloys, but typical synthetic methods yield larger nanoparticles where the metals are poorly mixed. Wong et al. show that well-mixed bimetallic ∼1-nm-diameter nanoparticles can be made on silica supports. To do this, they exploited strong electrostatic adsorption, in which the metal precursors are strongly adsorbed onto the surface by controlling pH relative to the surface point of zero charge. Their method was successful for a wide range of metal alloys.

Science, this issue p. 1427

Abstract

Supported nanoparticles containing more than one metal have a variety of applications in sensing, catalysis, and biomedicine. Common synthesis techniques for this type of material often result in large, unalloyed nanoparticles that lack the interactions between the two metals that give the particles their desired characteristics. We demonstrate a relatively simple, effective, generalizable method to produce highly dispersed, well-alloyed bimetallic nanoparticles. Ten permutations of noble and base metals (platinum, palladium, copper, nickel, and cobalt) were synthesized with average particle sizes from 0.9 to 1.4 nanometers, with tight size distributions. High-resolution imaging and x-ray analysis confirmed the homogeneity of alloying in these ultrasmall nanoparticles.

Bimetallic nanoparticles (NPs), anchored on porous carbon or oxide supports with high surface area to prevent agglomeration, have many applications in chemical sensing, biomedicine, and especially catalysis (1, 2). The most common method to prepare supported metal NPs is impregnation (IMP), in which just enough metal precursor solution is used to fill the pore volume of the support. Although this method is simple and quick, IMP generally yields NPs with large average size, broad size distributions, and, in the case of multiple metals, inhomogeneous alloying. The highest possible metal utilization occurs when all atoms of a NP are accessible to the gas or liquid fluid phase, and this occurs when the particle size is about 1 nm, or clusters of several dozen atoms (3, 4). Optimal bifunctionality and synergy often occur when the atoms at the NP surface are well mixed (5, 6).

We demonstrate a simple method that can be applied to noble and base metals alike to synthesize ultrasmall (~1 nm) NPs with homogeneous alloying. The method is based on strong electrostatic adsorption (SEA), whereby charged metal precursors are strongly adsorbed onto oppositely charged oxide or carbon surfaces by controlling the pH relative to the surface point of zero charge (PZC) (79). This interaction holds the precursors in place during drying, as opposed to IMP, in which the metal precursors remain and aggregate in solution as drying occurs. The required surface charge is provided by the native OH groups on the surfaces; no explicit surface functionalization is needed. This method involves fewer steps than colloidal methods of NP synthesis and eliminates the problematic removal of the capping agents of those methods, because no capping agent is used.

Electrostatic adsorption has been demonstrated for a host of single metals on many different supports (717). Over composite oxide surfaces, pH-controlled adsorption has been used to achieve selective metal adsorption onto one surface component or the other (1820). Most recently, there have been preliminary attempts at simultaneous electrostatic adsorption of two metal precursors for the synthesis of bimetallic NPs, involving Pt and Pd (21) or Pt and Co (22); this has been termed simultaneous or co-SEA. This method is schematized in Fig. 1A, which shows an electrostatically adsorbed layer of a mixture of hydrated metal precursors, followed by a thermal reduction in H2 needed to remove the metal ligands, reduce the metals to their zero-valent state, and nucleate the atoms into NPs.

Fig. 1 SEA bimetallic adsorption schematic, adsorption density, and temperature-programmed reduction profiles.

(A) Metal precursors electrostatically adsorb as a mixed monolayer. Clusters of alloyed NPs form after H2 reduction. (B) Adsorption surveys of noble and base metal pairs depict relative affinities of metals as a function of pH on the support. (C) TPR profiles (downward peaks indicate H2 consumption) of the monometallic SEA NPs are at a higher temperature, indicating a stronger support interaction relative to IMP nanoparticles. Bimetallic SEA NPs show improved reducibility from H2 spillover over the monometallic analogs.

We now present a systematic application of this method using a common silica support with a variety of noble and base metal ammine precursor pairs (Pt, Pd, Co, Ni, and Cu ammines) as shown in Table 1. The silica support (Evonik Aerosil 300) has a PZC of 3.6, a value similar to that of other fumed silicas (10, 11, 21). Adsorption of metal ammine precursors was measured as a function of pH for all pairs of metal complexes to determine their relative affinities for the support (23). Uptake surveys for representative noble-noble (Pt-Pd), noble-base (Pd-Cu), and base-base (Ni-Co) metal pairs are shown in Fig. 1B. For the Pt-Pd adsorption surveys, no adsorption of metals occurred near the PZC.

Table 1 Compositions (weight percent) of bimetallic NPs supported on silica.
View this table:

As the pH was increased, the OH groups of the silica became deprotonated, which allowed the cationic Pt/Pd ammine precursors to adsorb. At lower pH values, Pd had a greater affinity than Pt to the support; the opposite was true at higher pH values. Maximum adsorption of 1.3 μmol/m2 occurred at an optimum pH (pH ≈ 11.5) followed by a decrease in uptake as more base was added. This decrease in uptake is attributed to the strong ionic strength in solution that creates a double-layer screening effect (9); the metal complexes were prevented from experiencing the surface charge. A similar trend was observed in the Pd-Cu adsorption survey, but the maximum adsorption occurred at a higher pH value. Adsorption uptake surveys of Ni- and Cu-containing solutions were kept above a pH of 11 to prevent precipitation. The base-base metal pair of Ni-Co (Fig. 1B) was consistent with the other uptake surveys.

In all cases, the maximum adsorption of complexes for complete monolayer coverage for the bimetallic systems was ~1.0 to 1.3 μmol/m2. This result indicates that all of the metal complexes adsorbed with the same number of hydration sheaths and were of similar size during adsorption (12). The Co surveys with Pd, Ni, and Cu (Fig. 1B and fig. S1) revealed competitive adsorption, in that the Co ammine favored the silica surface. The high affinity of Co ammines for silica is known and ultimately results in the formation of a new phyllosilicate phase (24, 25). Thus, the Co adsorption was independent of pH in these experiments. The relative concentration of Co was lessened to allow the adsorption of the other metal to achieve an atomic ratio of 1:1. All other metal pairs were adsorbed onto the support in a 1:1 atomic ratio at the monolayer capacity of the precursors; this was achieved by controlling the metal precursor concentrations in solution (table S1).

Temperature-programmed reduction (TPR) was performed to characterize the intimacy of alloying of the metals in the nascent NPs. We compared monometallic and bimetallic NPs prepared by SEA and IMP (see tables S1 and S2 for synthesis parameters). TPR profiles for the Pd-Cu series (Fig. 1C) showed that the monometallic Pd in the IMP sample exhibited a low reduction temperature of ~100°C but that the monometallic Pd in the SEA sample required 150°C for reduction, indicating a much stronger interaction with the silica support. The TPR for Cu IMP reduction peaks began with a shoulder at ~190°C and had a sharp peak at 230°C. The broad shoulder of the IMP profile suggests a wide particle size distribution in which some of the larger particles were more easily reduced. The TPR for the Cu SEA profile contains a sharp peak reduction peak at 220°C and a broader peak around 500°C.

Among the bimetallic NPs, the Pd-Cu co-SEA NPs (Fig. 1C, center) showed two distinct peaks. The first peak at ~180°C was from the reduction of a Pd-Cu alloy, and the second peak at 400°C was the monometallic Cu peak at 500°C shifted to a lower temperature. We attribute the shift to lower temperature to hydrogen spillover from Pd to unreduced Cu complexes, indicating that there must be close proximity of Pd-Cu atoms in the co-SEA–derived NPs. The bimetal Pd-Cu IMP NPs contained multiple reduction peaks between 80° and 250°C with distinct peaks that corresponded to the nonshifted monometallic Pd and Cu. Thus, unalloyed Pd or Cu existed in the IMP NPs.

Further evidence of hydrogen spillover in co-SEA bimetallic NPs is observed in the Pd-Co TPR profiles (Fig. 1C, right). Although Co complexes are very difficult to reduce over silica and require temperatures of 700°C (20), alloying Pd-Co enabled us to observe hydrogen spillover–assisted reduction of Co cations by Pd at a temperature of 400°C. TPR profiles of additional metal combinations are shown in figs. S2 and S3. These additional metal combinations exhibited similar results: The SEA NPs showed greater support interaction and commonly required a higher reduction temperature than the IMP NPs. The co-SEA NPs often had peaks shifted to lower temperatures than the monometallic SEA analogs, indicative of hydrogen spillover–assisted reduction between the two metals. The IMP NPs contained broader peaks and more reducible species, which are consonant with a larger particle size distribution and various NP compositions.

After reduction in 10% H2 balanced in He for 1 hour at 400°C, x-ray diffraction (XRD) profiles were taken of each material. The XRD patterns of the co-SEA materials (Fig. 2A) showed no distinct features between 30° and 45° 2θ, the range containing the most intense diffraction peaks. The very broad peaks observable in some patterns, with respect to the metal-free supports, were the result of highly dispersed NPs. This interpretation was later confirmed by electron microscopy. XRD patterns of the co-IMP NPs, reduced at the same conditions (Fig. 2B), showed strong XRD peaks for all of the IMP metal combinations. The intense XRD peaks between 30° and 45° 2θ were indicative of large NPs (i.e., x-ray diameters dXRD > 5 nm). Moreover, the non-Gaussian shape of the peaks, containing shoulders and sharp centers, suggests variable phase compositions and a wide particle size distribution, respectively.

Fig. 2 Bimetallic NP size characterization after reduction in 10% H2 balanced in He for 1 hour at 400°C.

(A and C) XRD patterns and STEM images of co-SEA bimetallic NPs reveal extremely high dispersion. (B and D) XRD patterns and STEM images of co-IMP bimetallic samples show larger particles.

Electron microscopy Z-contrast images of the co-SEA NPs (Fig. 2C) show that they existed as clusters ~1.0 to 1.3 nm in diameter with narrow particle size distributions (standard deviations were about 20 to 30% of average particle size). This finding is consistent with the absence of peaks in the high-sensitivity x-ray diffractometer, which has a size limit of ~1 nm (26). In comparison, the microscopy images in Fig. 2D showed the co-IMP NPs to have broader particle size distributions and overall larger particle sizes (~5 to 25 nm). A difference in NP shape was also observed between the two synthesis methods. The co-SEA NPs produced more regular spherical particles, whereas more frequent irregular shapes and particle agglomeration were observed in the co-IMP NPs (Table 2; see fig. S4 for particle size distributions). SEA with single metals typically yields particles in the size range 1 to 2 nm (10); with co-SEA, the NPs were about the same size or smaller.

Table 2 Average particle sizes of bimetallic NPs.

STEM sizes are number-average diameters; XRD sizes are from the Scherrer equation.

View this table:

Although the 1-nm NPs were too small for individual NP elemental mapping, we were able to use energy-dispersive x-ray spectroscopy (EDXS) in spot mode on multiple individual bimetallic NPs produced by co-SEA. These measurements can help to determine composition consistency by comparing the nominal weight ratios to the actual weight ratios determined through x-ray analysis. The NPs were prepared near a 1:1 atomic ratio; thus, the weight percent (wt %) and weight ratio of the two metals will vary depending on their molecular weights. Table S3 shows the spot analysis of the Pt-Pd NPs, which have an expected weight ratio of 0.65 Pt:0.35 Pd. The results showed that the Pt-Pd NPs were consistent with the expected ratio. In all cases (tables S4 to S12), the co-SEA NPs were bimetallic, with most individual NPs being near the predicted wt % for each metal (23).

EDXS mapping was performed over the larger IMP NPs to determine the degree of alloying (figs. S5 and S6). Incomplete alloying occurred in most of the co-IMP bimetallic NPs, and the extent of alloying varied from particle to particle. The segregation of the metals was independent of the predicted structure obtained from phase diagrams, where most metal combinations do form alloys. This incomplete alloying agrees with XRD data of the IMP NPs, which showed multiple phase compositions.

Finally, atomically resolved Z-contrast images were taken to individually characterize the co-SEA NPs (Fig. 3). Careful examination of these images revealed a speckling effect, where elements of higher atomic number are brighter and those of lower atomic number are dimmer. This phenomenon could be best observed in the samples in which the two metals have large differences in atomic number. Figure 3, A and B, shows the clearest speckling of the Pt-Co and Pt-Cu NPs; this is direct visual evidence of a well-mixed alloy. Other combinations of Pd-Ni, Pd-Cu, and Pt-Pd (Fig. 3, C to E) with lower differences in Z also showed speckling, although less pronounced. The speckling effect has been reported most clearly in the case of solution-derived and somewhat larger Pt/Pd and other bimetallic NPs (27). No speckling was observed in the Cu-Co NPs because of the very small Z difference and poor contrast; however, EDXS spot analysis of this sample (table S11) reveals that this as well as the other two base metal combinations, Ni-Cu and Ni-Co (tables S9 and S10), are as thoroughly alloyed as the other samples.

Fig. 3 Atomically resolved Z-contrast images showing NP speckling of alloys.

(A and B) Pt-Co and Pt-Cu, separated by two rows in the table of elements. (C to E) Pd-Ni, Pd-Cu, and Pt-Pd, separated by one row. (F) Cu-Co, in the same row. Scale bars, 5 nm.

Simultaneous SEA can be extended to other surfaces such as alumina, titania, and carbon, as long as soluble metal precursors are available; in the current case, cationic ammine complexes were adsorbed over a low-PZC silica surface. Over high-PZC surfaces such as alumina and carbon, anions such as chloride complexes can be used. The method is limited to the availability of suitably charged complexes. A potential drawback with the use of silica is the solubility of silica at the high pH at which the adsorption is performed. Silica dissolution during SEA was seen to be appreciable above pH 11 and with 24-hour contact times (28). Dissolution can be circumvented by working at a pH of 10.5 or below, where dissolution is negligible (28), or can be minimized by using short contact times, as metal adsorption reaches equilibrium within minutes (4).

Another limitation is the amount of metal that can be applied in one co-SEA application. The normal uptake of charged precursors (~1 μmol/m2, or ~1 complex per nm2) corresponds to a steric maximum of close-packed complexes retaining one or two hydration sheaths (4). Metal loading thus depends on the specific surface area of the support; for a support such as precipitated silica (200 m2/g), typical metal loadings are limited to a few wt %, whereas for materials with high surface area such as mesoporous silica (1000 m2/g), loadings of 10 to 15 wt % are possible in a single SEA application (10). Higher loadings can be achieved by repeated applications of SEA and reduction [this has been demonstrated for Pt on carbon (29)]; the reduction transforms the hydrated, ligated precursors into metal nanoparticles and uncovers the support surface, onto which more precursor can be subsequently adsorbed.

Other methods for synthesizing highly dispersed, well-alloyed bimetallic nanoparticles exist. It is suggested that bifunctional organics that chelate both a metal and a silica surface, demonstrated for single metals of ruthenium and iridium (30), can be applied to mixtures of metals (2). The method appears to be limited to silica and is also limited to low metal loadings because of the exotherm generated in removing the organic chelating agent by oxidation (2). Alternatively, in limited instances, supported nanoparticles may be derived from the decomposition of metastable bimetallic precursor species such as Keggin-type complexes continuing cobalt and molybdenum (31) or organometallic cluster complexes containing platinum and iron (32). Finally, well-alloyed, size- and composition-controlled bimetallic nanoparticles can be synthesized in the solution phase by colloidal methods (33) such as a recent study of 50 wt %, 4.5- to 6.5-nm PdPt nanoparticles on carbon (34). The higher achievable weight loading of this approach may offset the complexity of the synthesis and the difficulty in distributing the NPs into the support. If, on the other hand, ultrasmall, homogeneously alloyed bimetallic nanoparticles at moderate weight loadings are desired, simultaneous SEA provides a more generalizable and straightforward synthesis strategy.

Supplementary Materials

www.sciencemag.org/content/358/6369/1427/suppl/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S12

References and Notes

  1. See supplementary materials.
  2. Acknowledgments: Supported by NSF grants CBET 1160023 and IIP 1464630. All data are reported in the main text and supplement.
View Abstract

Navigate This Article