Galvanic Replacement Reactions in Metal Oxide Nanocrystals

See allHide authors and affiliations

Science  24 May 2013:
Vol. 340, Issue 6135, pp. 964-968
DOI: 10.1126/science.1234751

Hollowing Out Metal Oxide Nanoparticles

Corrosion is normally a problem, but it can be useful, for example, when you wish to create hollow metal nanoparticles, whereby the reduction of one metal species in solution drives the dissolution of the core of the particle. Oh et al. (p. 964; see the Perspective by Ibáñez and Cabot) adapted this approach to metal oxide nanoparticles by placing Mn3O4 nanocrystals in solution with Fe2+ ions, which replaces the nanocrystal exterior with γ-Fe2O3. At sufficiently high Fe2+ concentrations, hollow γ-Fe2O3 nanocages formed. These hollow structures could be used as anode materials for lithium ion batteries.


Galvanic replacement reactions provide a simple and versatile route for producing hollow nanostructures with controllable pore structures and compositions. However, these reactions have previously been limited to the chemical transformation of metallic nanostructures. We demonstrated galvanic replacement reactions in metal oxide nanocrystals as well. When manganese oxide (Mn3O4) nanocrystals were reacted with iron(II) perchlorate, hollow box-shaped nanocrystals of Mn3O4/γ-Fe2O3 (“nanoboxes”) were produced. These nanoboxes ultimately transformed into hollow cagelike nanocrystals of γ-Fe2O3 (“nanocages”). Because of their nonequilibrium compositions and hollow structures, these nanoboxes and nanocages exhibited good performance as anode materials for lithium ion batteries. The generality of this approach was demonstrated with other metal pairs, including Co3O4/SnO2 and Mn3O4/SnO2.

The galvanic replacement reaction is the most versatile method of preparing hollow metallic nanostructures with controllable pore structures and compositions (17). These reactions involve a corrosion process that is driven by the difference in the electrochemical potentials of two metallic species. The hollow interior is generated from the oxidative dissolution of the metal nanocrystals (NCs) that are used as reactive templates. This strategy has also been used for the production of hollow semiconductor nanostructures (8). However, the chemical transformation of ionic systems via galvanic reactions has remained elusive. We demonstrated that a galvanic replacement reaction can occur in oxide NCs as well and can produce hollow oxide nanostructures.

Hollow oxide NCs have attracted much interest because of their potential for application in energy storage, catalysis, and medicine (9, 10). Considerable advances have been made in the synthesis of hollow oxide and semiconductor NCs (1113). The Kirkendall effect has been exploited to produce complex hollow nanostructures of metal oxides and chalcogenides (11, 1416). However, synthesizing hollow NCs of multimetallic oxides still remains a substantial challenge (17, 18). We showed that by using a nanoscale galvanic replacement reaction, monometallic oxide NCs could be completely transformed into hollow multimetallic oxide nanostructures. Contrary to what occurs in metallic systems, a redox-couple reaction between multivalent metallic ions took place, replacing the higher–oxidation-state ions in the NCs with lower–oxidation-state metal ions from solution. We focus here on one example, hollow heterostructured NCs of manganese oxide/iron oxide (Mn3O4/γ-Fe2O3) with a boxlike shape (“nanoboxes”), which we transformed into cage-shaped iron oxide (γ-Fe2O3) NCs (“nanocages”). We also showed that these hollow-structured multimetallic oxide nanostructures exhibit synergistic properties that make them attractive for use as anode materials in lithium ion batteries (LIBs).

Mn3O4 NCs were prepared by slightly modifying a previously reported method (19, 20). A transmission elecron microscope (TEM) image of the as-prepared Mn3O4 NCs showed that they were square prism–shaped, with sides of ~20 nm and a height of ~10 nm (Fig. 1A and fig. S1) (20). A high-resolution TEM (HRTEM) image and the corresponding Fourier transform (FT) pattern of the Mn3O4 NCs revealed that their top and side surfaces were enclosed by {001} and {100} facets, respectively, and their corners and edges were slightly truncated (Fig. 1B).

Fig. 1

(A and B) TEM image of Mn3O4 NCs. (A) Low-magnification image of Mn3O4 NCs. The inset shows the corresponding HRTEM image of a single NC recorded along the [111] axis. (B) HRTEM image of a single Mn3O4 NC recorded along the [001] axis. The inset shows the corresponding FT pattern. (C and D) (C) Low-magnification TEM and (D) HRTEM image of the γ-Fe2O3 nanocages synthesized by reacting the Mn3O4 NCs with 1 ml of 2.0 M aqueous iron(II) perchlorate solution. The inset shows the corresponding FT pattern. (E) ICP-AES data showing the molar fraction of Fe in the reaction product as a function of the amount of iron(II) perchlorate added during the synthesis (solid circles) and their bulk counterpart (open circles). The molar fractions expected in the cases of the complete replacement of Mn by Fe (dotted line) and the addition of Fe to Mn (dashed line) are also shown. (F) Powder XRD patterns of the original Mn3O4 NCs and for the samples synthesized by the reaction with Mn3O4 NCs with 1 ml of aqueous solutions of iron(II) perchlorate having different concentrations. For comparison, the standard XRD patterns of Mn3O4 and γ-Fe2O3 are also shown.

The galvanic replacement reaction was initiated by adding an aqueous iron(II) perchlorate solution into a suspension of the Mn3O4 NCs in xylene and subsequently heating the reaction mixture at 90°C for 1.5 hours (20). The reaction seemed to take place at the organic/aqueous interface of reverse vesicles (fig. S2) (20). Figure 1, C and D, shows the TEM and HRTEM images of the product derived from the galvanic replacement reaction between the Mn3O4 NCs and 1 ml of 2.0 M iron(II) perchlorate solution. The original NCs transformed completely into γ-Fe2O3 nanocages with a hollow interior and holes on their shell. The exterior shape of the nanocages was nearly the same as that of the original Mn3O4 NCs. The average particle size increased from 19 to 23 nm after the galvanic reaction, because of the formation of the γ-Fe2O3 shell (fig. S3) (20). The shell of these nanocages had a single-crystalline structure with highly ordered continuous lattice fringes (Fig. 1D). Inductively coupled plasma–atomic emission spectrometry (ICP-AES) showed that the molar ratio of Fe to Mn in the γ-Fe2O3 nanocages was 91:9 (Fig. 1E) and that almost complete metal replacement took place. We could also synthesize hollow nanocages of Co3O4/SnO2 and Mn3O4/SnO2 from the reactions of Co3O4 and Mn3O4 NCs with aqueous tin(II) chloride solutions (fig. S4) (20), demonstrating the generality of the transformation process. These nanocages exhibited polycrystalline structures, because the crystal structure of SnO2 (rutile) is different from that of Co3O4 and Mn3O4 (spinel).

Surface precipitation of γ-Fe2O3 reduces the dissolution rate of Mn3O4 by preventing the diffusion of ions and electrons to its surface (21). However, if the volume of Mn3O4 was below a certain critical limit, its complete dissolution and thus the near-total replacement of Mn by Fe could be achieved. Indeed, the ICP-AES results showed that the Mn molar ratio decreased to less than 0.1% when the Mn3O4 NCs were treated with more concentrated 3.0 M iron(II) perchlorate. The volume of Mn3O4 affected this transformation; for the same reaction performed with bulk Mn3O4 powder, the molar fraction of Fe in the thus-formed product was substantially lower (less than 25%) (Fig. 1E). Powder x-ray diffraction (XRD) patterns of the reactant and the products (Fig. 1F) confirmed that the tetragonally distorted Mn3O4 spinel transformed into a cubic γ-Fe2O3 spinel as the concentration of the iron(II) perchlorate solution was increased. Superconducting quantum interface device measurements showed that magnetization of the NCs increased steadily with an increase in the Fe content (fig. S5) (20).

The evolution of the internal hollow core was examined by reacting the Mn3O4 NCs with aqueous iron(II) perchlorate solutions of various concentrations (Fig. 2 and fig. S6) (20). When a less concentrated solution was used, the cores of the Mn3O4 NCs were dissolved partially, and nanoboxes with relatively thick walls were formed (Fig. 2, B, C, F, and G). The images in Fig. 2, B and F, and fig. S7 (20) show that pinholes were formed on the surface of the nanoboxes. These observations indicate that pores develop inside the NCs by a mechanism analogous to pinhole corrosion, in which pinholes serve as transport paths during the dissolution of the NC core. As the concentration of iron(II) perchlorate was increased to 1.0 M, the pores increased further in size, and the nanoboxes evolved into nanocages with clear openings (Fig. 2H). During the intermediate stages of the replacement reaction, both the nanoboxes and the nanocages exhibited continuous fringe patterns. Figure 2, B to E, shows a slight increase in the interplanar distance between (100) of Mn3O4 NCs and (110) of γ-Fe2O3 NCs. This change, however, did not alter the alignments of the lattices along the square basal planes of the NCs, demonstrating topotactic transformation. The color-mapped energy-filtered TEM (EFTEM) images in the inset of Fig. 2G and fig. S8 (20) show the accumulation of Fe species in the shell of the nanoboxes, whereas the image in Fig. 2I shows the deposition of Fe species over the entire surface of the nanocages. We did not observe any noticeable amount of Mn species in the γ-Fe2O3 shell of the nanoboxes (Fig. 2C). These results suggest that the Kirkendall effect does not play a major role in the current hollowing process (figs. S9 and S10), probably because of the low reaction temperature of 90°C, where the interdiffusion rates of Mn and Fe ions are low (table S1) (20).

Fig. 2 (A) Schematic illustration of the transformation of Mn3O4 NCs, showing the evolution of their morphology via the localized dissolution of Mn3O4 and the surface precipitation of γ-Fe2O3.

(B to E) HRTEM images of the hollow nanostructures synthesized by the reaction of Mn3O4 NCs with 1 ml of aqueous solutions of iron(II) perchlorate having different concentrations: (B) 0.4 M, (C) 0.6 M, (D) 1.0 M, and (E) 1.6 M. Insets show the corresponding FT patterns. (F) High-angle annular dark-field scanning TEM (HAADF-STEM) image of the nanoboxes shown in (B). (G) TEM image and a corresponding EFTEM image of the nanoboxes shown in (C). (H) HAADF-STEM image of the nanocages shown in (D). (I) TEM image and a corresponding EFTEM image of the nanocages shown in (E).

The galvanic reaction observed in this study could be driven by the difference in the standard reduction potentials of Fe3+/Fe2+ (0.77 V) and Mn3O4/Mn2+ (1.82 V) pairs, in which Fe2+ ions reduce Mn3+ ions in the Mn3O4 NCs (fig. S11) (20). Indeed, the Mn3O4 NCs were not transformed into hollow nanocages when reacted with cobalt(II) perchlorate (fig. S12) (20), which can be attributed to the higher standard reduction potential of the Co3+/Co2+ pair (1.87 V) than that of the Mn3O4/Mn2+ pair.

The mechanism of redox replacement was collaborated by measurement of the valence state with x-ray absorption spectroscopy (XAS) and x-ray magnetic circular dichroism (XMCD) measurements (fig. S13) (20). The Mn L2,3-edge x-ray absorption spectra of the nanoboxes were almost identical to those of the original Mn3O4 NCs. The Mn3+ peaks gradually disappeared with increasing Fe2+ concentration ([Fe2+]), and only Mn2+ peaks remained after the complete replacement reaction (22). However, the decrease in [Mn2+] to less than 9% suggested that most of the Mn2+ were also removed from the tetrahedral sites. Both XAS and XMCD spectra at the Fe L2,3-edges of the nanocages were similar to those of γ-Fe2O3 that contained only Fe3+, but not to those of Fe3O4 that contained both Fe3+ and Fe2+.

When the galvanic reaction is initiated, some of Mn3O4 at the surface is dissolved into the solution (fig. S14) (20). The precipitation of γ-Fe2O3 started with the outermost shell of the Mn3O4 NCs, preventing the outward Mn2+ diffusion. This transformation occurred preferentially around the edges of the shell (Fig. 2G, inset) (23). An increase in the number of empty octahedral sites caused by dissolution of the reduced Mn2+ could lead to the complete disruption of the remaining lattice that consists of tetrahedral Mn2+ and oxygen anions (24). This process formed pinholes during the initial stage of reaction in the areas not covered by the γ-Fe2O3 layer; i.e., in the middle of the basal plane. As the reaction progresses, the electrons released from the Fe2+ migrate inward and reduce the octahedral Mn3+ in the interior. The pinholes formed facilitated the dissolution of the residual core species. Because of the open structure of the nascent γ-Fe2O3 nanocages, precipitation of γ-Fe2O3 occurred in their hollow interior at later stages. We also observed the formation of isolated γ-Fe2O3 NCs during the galvanic reaction (fig. S8A) (20), indicating that some of oxidized Fe3+ ions are converted to γ-Fe2O3 NCs via homogeneous nucleation. Together with the initial dissolution of Mn3O4 at the surface, the formation of isolated γ-Fe2O3 NCs may account for the generation of relatively thin shells of γ-Fe2O3, although the overall amounts of the dissolved Mn species and the precipitated Fe species are similar (fig. S15) (20). The perchlorate moiety in the aqueous iron(II) perchlorate solution did not contribute to the formation of the hollow structure, because no morphological change was observed when the Mn3O4 NCs were reacted with perchloric acid (fig. S16) (20). In addition, we found that cagelike structures with a different morphology could also be produced via the same method by using different-shaped Mn3O4 NCs as the starting material (fig. S17) (20).

Transition-metal oxides such as Mn3O4 and Fe3O4 are attractive anode materials for LIBs because of their higher specific capacities as compared to typical graphitic anodes. However, their capacities decrease drastically with cycling because of their severe volume changes during Li insertion and extraction (25, 26). Despite its theoretically high capacity (936 mAh g−1), Mn3O4 usually exhibits a much lower capacity (≤ ~400 mAh g−1) because of its low electrical conductivity (~10−7 to 10−8 S cm−1) (25). The current galvanic replacement reactions in oxide NCs can be a promising technique for engineering their structures and tuning their chemical compositions to improve their electrochemical properties for their applications to LIB anodes. Our hollow nanoboxes and nanocages were coated with polypyrrole in situ in the reaction solution and carbonized at 500°C for 2 hours in an Ar atmosphere (20). Through the thermal treatment, the samples transformed into solid solutions of Mn3O4 and Fe3O4 (i.e., Mn3-xFexO4) with a ferrite structure and varying compositions ranging from Mn2.0Fe1.0O4 to Mn0.3Fe2.7O4 (Fig. 3A), while retaining their original hollow structures (Fig. 3B and fig. S18) (20).

Fig. 3

Characterization and LIB anode applications of carbon-coated hollow Mn3-xFexO4 NCs: (a) Mn2.0Fe1.0O4, (b) Mn1.5Fe1.5O4, (c) Mn1.1Fe1.9O4, (d) Mn0.6Fe2.4O4, and (e) Mn0.3Fe2.7O4. (A) XRD patterns of the samples. a.u., arbitrary units. (B) TEM image of the carbon-coated Mn1.1Fe1.9O4 NCs. (C and D) Electrochemical measurements of the samples: (C) first discharge curves and corresponding results of the dQ/dV analysis in the inset, and (D) cycle performances at a current density of 100 mA g−1 in the voltage range from 3.0 to 0.01 V. Open symbols indicate discharge (lithiation) cycles, and solid symbols are charge (de-lithiation) cycles.

For Mn2.0Fe1.0O4, two plateaus were observed around 0.55 and 0.68 V in the first discharge curve, with a peak ratio of 2:1, as shown in the inset of Fig. 3C. Mn1.5Fe1.5O4 and Mn1.1Fe1.9O4 exhibited two plateaus at slightly higher voltages of around 0.59 and 0.73 V, respectively. The ratio of the plateau at the upper potential gradually increased with the Fe content of the sample. For Mn0.6Fe2.4O4, the plateau around 0.60 V eventually disappeared and the upper-potential plateau shifted to 0.83 V, a value similar to that of commercial Fe3O4 (<50 nm, no. 637106 from Aldrich) (Fig. 3C), indicating that almost all Mn ions were replaced by Fe ions. The changes in the ratios and potentials of the two plateaus seem to be related to the content of substituted Fe in Mn3O4. The potential of Mn3-xFexO4 as a function of the Fe content was investigated through density functional theory calculations, which indicated that a solid solution of Mn3-xFexO4 can occur between Mn3O4 and Fe3O4 (fig. S20) and that their average reaction voltages are proportional to the amount of the Fe3O4 component from 0.400 to 0.818 V at 0 ≤ x ≤ 3 (table S2) (20). Our results suggest that the potentials of multicomponent spinel electrode materials can be easily tuned by galvanic replacement reaction and subsequent thermal treatment.

Figure 3D shows cycle performances of hollow Mn3-xFexO4 NCs. Mn1.1Fe1.9O4 exhibited first discharge and charge capacities of 1339 and 984 mAh g−1, respectively. Mn1.1Fe1.9O4 exhibited a high reversible capacity of ~1000 mAh g−1 with almost no fading up to 50 cycles. The reversible capacities and cyclic stabilities of Mn2.0Fe1.0O4 and Mn1.5Fe1.5O4 were comparable to those of Mn1.1Fe1.9O4. Most of the samples showed higher specific capacities and better cyclic stabilities than those of carbon-coated solid Mn0.8Fe2.2O4 NCs (fig. S21) (20). These good cyclic stabilities can be attributed to their hollow structure, which provides extra free space for alleviating the structural strain caused by the large volume change (26, 27) and additional sites for lithium ion storage (28). Mn0.6Fe2.4O4 and Mn0.3Fe2.7O4, which are close to Fe3O4, showed less cyclic stability than the other samples, indicating that multimetallic composition contributes to the enhanced cyclic stabilities (29). Furthermore, Mn1.1Fe1.9O4 exhibited the highest rate capability (fig. S22) (20), which is due to the improvement in its electronic conductivity resulting from the mixed valency of the multicomponent Mn3-xFexO4 (30).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S22

Tables S1 and S2

References (31–43)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We acknowledge financial support by the Research Center Program of the IBS in Korea (T.H. and Y.-E.S.); the WCU (R31-10013) Program of the National Research Foundation (NRF) of Korea (T.H., N.P., and Y.-E.S.); Portuguese Fundação para a Ciência e a Tecnologia projects (PTDC/CTM/100468/2008 and REDE/1509/RME/2005) for TEM work (M.-G.W. and N.P.); a Human Resources Development of the Korea Institute of Energy Technology Evaluation and Planning grant funded by the Korea government Ministry of Trade, Industry and Energy (20124010203320) (K.K. and D.-H.S.); and National Creative Initiative (2009-0081576), WCU (R31-2009-000-10059), and Max Plank POSTECH/KOREA Research Initiative (2011-0031558) programs though the NRF of Korea (K.-T.K. and J.-H.P.). The Pohang Accelerator Laboratory is supported by POSTECH and the Ministry of Science, ICT and Future Planning.
View Abstract

Navigate This Article