Metal nanoparticles with the size of 1-10 nm have been widely applied to catalyze chemical reactions in the petrochemical and fine chemical industries, energy processes, and environmental protection. The general characteristics of these solid catalysts are that the coordinatively unsaturated atoms on their surface act as the active sites and the surface electronic properties and geometrical structure determine the catalytic activity, selectivity, and stability ^{[1, 2, 3, 4, 5, 6]}. The microstructure of metal nanoparticles is usually given in terms of size, morphology, and crystal phase. The size and morphology effect of metal nanoparticles have been extensively studied ^{[7, 8, 9, 10, 11, 12, 13, 14, 15, 16]}. However, the impact of the crystal phase was rarely considered, mainly because of the facile transformation of metastable crystal phases to thermodynamically stable ones during the synthesis process or under reaction conditions.

Metal nanoparticles usually crystallize in the forms of face-centered cubic (fcc), body-centered cubic (bcc), and hexagonal close-packed (hcp) structures as the space packing pattern of the metal atoms. The coordination environment and the spatial distribution of the metal atoms vary significantly in these crystal phases. The atomic coordination number of the fcc, bcc and hcp structures are 12, 8 and 12, respectively, and the corresponding spatial utilization degree of the atoms are 74%, 68% and 74%. For a specific metal nanoparticle, the occupation degree of the surface atoms is also largely determined by the crystal phase ^[17]. Therefore, changing its crystal phase would change the geometrical structure. This in turn modifies the adsorption and desorption properties of reactants and products. More importantly, the stacking mode of the atoms in different crystal phase also tunes the electronic properties of the metal nanoparticle because the atomic nuclei at the lattice sites provide Coulombic attraction for the valence electrons according to the free electron model ^[18]. The electronic structure of transition metals near the Fermi level is primarily determined by the d electrons, which significantly affect the adsorption and activation of the reactant molecules as well as their reaction pathway ^{[19, 20]}. A higher energy d state, compared to the Fermi energy, gives a stronger interaction with the adsorbed substrate. Cobalt, for example, has three polymorphs, namely, the hcp, fcc and primitive cubic phases (ε phase) (Fig. 1) ^{[21, 22]}. The hcp phase is more stable at room temperature, while the fcc and ε phases are metastable and are transformed into the hcp phase at elevated temperatures ^{[22, 23]}. The Fermi energy of the hcp, fcc and ε phases are 5.49, 5.29, and 5.15 eV, respectively ^[23]. The electron density around the Fermi level follows the order of fcc-Co < ε-Co < hcp-Co. The electron localization function of hcp-Co is a radial electron distribution, while fcc-Co electron distribution has an evident polarization. The density of states (DOS) for the d band in ε-Co is broader than in the others. Therefore, it is expected that tuning the crystal phase of cobalt nanoparticles would change their geometric environment and electronic property, and in turn the catalytic performance ^{[23, 24]}.

Fig. 1. Crystal structures of hcp-Co, fcc-Co and ε-Co.

Nanoalloys consist of two or more types of metals and present different electronic structures compared with the individual metals ^{[25, 26, 27, 28]}. This allows tuning their electronic structure and catalytic properties more efficiently ^{[29, 30, 31]}. For instance, the d state of the AuCu alloy varies considerably between Au and Cu (Fig. 2), depending on the molar ratio of Au and Cu ^[32]. Au₃Cu nanoparticles with an fcc structure showed a much higher activity for the electrochemical reduction of CO₂ than the other compositions. Nickel particles are traditionally used to catalyze CO₂ hydrogenation to methane, but alloyed Ni₅Ga₃ particles catalyzed CO₂ hydrogenation to methanol with a comparable activity to industrial Cu/Zn/Al₂O₃ catalysts ^[33]. The Al₁₃Fe₄ alloy exhibited a similar activity to commercial Pd catalysts for the semi-hydrogenation of acetylene ^[34]. The AgCo alloy gave a comparable activity to Pt nanoparticles for electrochemical oxygen reduction ^[35]. These examples are convincing evidence that tuning the electronic structure of metal nanoalloys can improve the catalytic activity and mediate the reaction route. Moreover, metal alloys crystallize in versatile crystal phases through the flexible hybrid pathways of d electrons from the component metals, providing more options for mediating the electronic properties ^{[26, 36]}. For example, Pt and Fe typically crystallize in the fcc and bcc phases, respectively, but alloyed FePt particles may be fcc-Fe₃Pt, fcc-FePt₃ or FePt with a face-centered tetragonal (fct, L1₀) structure ^[37]. As shown in Fig. 3, the disordered Fm3-m and ordered Pm3-m structures of Cu₆₀Pd₄₀ alloy had dramatically different DOS ^[38]. The ordered phase gave a lower DOS than the disordered one, showing a higher degree of covalent bonding. As a result, the ordered crystal phase enhanced ethylene selectivity up to 90% in the semi-hydrogenation of acetylene.

Fig. 2. Crystal lattices (a) and surface valence band photoemission spectra (b) of Au, Cu and their alloys. Reprinted with permission from Ref. [32]. Copyright 2014, Macmillan Publishers Limited.

Fig. 3. Crystal structures of disordered PdCu (Fm3–m) and ordered PdCu (Pm3–m) (a) and their DOS (b). Reprinted with permission from Ref. [38]. Copyright 2013, MDPI Publishing.

In this perspective, we summarize recent progress on the crystal phase effect of metal nanoparticles in nanocatalysis. We first briefly introduce the principle of crystal phase controlled synthesis of metal nanoparticles and then examine the effect of the crystal phase of the metal and alloy nanoparticles on their catalytic performance. Finally, we present our perspectives on the future development of the crystal phase effect of metal nanoparticles, which is a challenging topic in nanocatalysis.

Metal atoms usually pack in the structure of an appropriate space group to decrease the surface energy to be as low as possible and thus form the most stable crystal phase. This means that mediating the crystal phase of metal nanoparticles requires modifying their surface energy during the synthesis process. Both theoretical calculations and experimental studies have confirmed that the control of crystal phase of metal particles strongly depends on controlling the size and morphology ^{[39, 40]}. For example, Co nanoparticles favor an fcc phase as twinned icosahedrons at sizes of less than 20 nm, whereas they are present as a hcp structure in the form of polyhedrons at sizes of more than 110 nm ^[39]. This is because of the different atom density in the exposed crystal faces induced by the surface energy. The phase transition of PtCo nanoparticles was strongly correlated with their size and morphology ^[40]. Bulk PtCo alloy transforms from disordered fcc to ordered L1₀ at 1098 K, whereas the 2.4-3 nm sized particles complete the phase transition at 773-923 K. Moreover, this transformation did not require the size of the PtCo nanoparticles to be less than 3 nm for the three axes; only one axial size smaller than the critical value would facilitate the phase transition at lower temperature. This implies that the phase transformation is also linked to the morphology and the metastable crystal phase can be obtained by tuning the shape of the metal nanoparticles as well.

To date, the crystal phase control of metal nanoparticles has been achieved by two approaches: liquid phase reduction and solid phase conversion. The synthesis of metal nanoparticles in liquid solution is initialized by the chemical reduction or thermal decomposition of a metal precursor. The generated metal atom then aggregates into nuclei after reaching a degree of supersaturation ^{[41, 42, 43, 44, 45, 46, 47]}. The growth pattern of the metal particles is a balance of thermodynamics and kinetics. Accordingly, the systematic manipulation of synthesis parameters, especially the use of suitable surfactants, is the most effective strategy to tune the surface energy of the growing metal particle and its crystal phase. For example, reduction of RuCl₃ by ethylene glycol (EG) with the presence of poly(N-vinyl-2-pyrrolidone) at 473 K yielded hcp-Ru nanoparticles of 2-5 nm (Fig. 4). However, when the metal precursor and the solvent were replaced by Ru(acac)₃ and triethylene glycol, respectively, fcc-Ru nanoparticles with similar sizes were produced ^[48]. Reduction of RuCl₃ in ethanol by NaBH₄ at room temperature produced fcc-Ru particles of 2.2 nm in size ^[49]. The control of particle size and crystal structure was achieved by adjusting the reduction rate of the ruthenium ion in the liquid phase and simultaneously using a surfactant to stabilize the metastable phases. Thermal decomposition of Co₂(CO)₈ in 1,2-dichlorobenzene with the presence of oleic acid and trioctylphosphine oxides yielded ε-Co particles of 8.7 nm (Fig. 5), while fcc-Co nanoparticles with a similar size (8.0 nm) was produced in dodecane with the aid of oleic acid and dioctylamine ^[50]. This result demonstrated that the surfactant mediated the crystal phase of the metal nanoparticles through effectively lowering the surface energy upon adsorption.

Fig. 4. TEM images of hcp-Ru (a) and fcc-Ru nanoparticles (b), and their catalytic performance for CO oxidation (c). Reprinted with permission from Ref. [48]. Copyright 2013, American Chemical Society.

Fig. 5. TEM images of hcp-Co, fcc-Co and ε-Co nanoparticles and their oxides on heating at 493 K under oxygen. Reprinted with permission from Ref. [50]. Copyright 2015, WILEY-VCH Verlag GmbH & Co.

In addition, metal nanoparticles with a specific crystal phase were also used as the seed or crystallite nuclei, based on its inherent structural symmetry, to direct the growth of metal nanoparticles with a desired crystal phase ^{[51, 52]}. Applying fcc-AuCu nanoparticles of 10.2 nm as the seed, the thermal decomposition of Co₂(CO)₈ and Fe(CO)₅ at 503 K generated a AuCu@FeCo core-shell structure (Fig. 6), where FeCo crystallized as the bcc phase ^[53]. Annealing of the composite at 653 K caused the FeCo shell to transform into the body-centered tetragonal (bct) phase while the AuCu core converted into the L1₀ phase ^[54]. This transformation was driven by the lattice match of the core and shell layers, which induced a heteroepitaxial growth of the FeCo shell.

Fig. 6. TEM images of fcc-AuCu@bcc-FeCo nanoparticles (a) and L10-AuCu@bct-FeCo nanoparticles (b); Illustration of the heteroepitaxial interfacial structure of the AuCu@FeCo core-shell nanostructure (c). Reprinted with permission from Ref. [54]. Copyright 2014, American Chemical Society.

Solid phase conversion is another route to mediate the crystal phase of the metal nanoparticles. It typically involves the thermal treatment of metal nanoparticles that are in the conventional crystal phase at elevated temperature and under a proper atmosphere. It is worth noting that this method differs from the traditional metallurgy process that is usually performed at temperatures above the melting point of metals. A solid phase conversion is conducted at much milder conditions by taking advantage of the chemical interaction of the reactive gases, mainly H₂ or CO, with the metal nanoparticles ^{[55, 56, 57]}. For example, fcc-AuCu nanoparticles of 3.4 nm were transformed into the ordered L1₀ phase under a H₂ atmosphere above 433 K (Fig. 7) ^[58]. Molecular hydrogen induced the generation of lattice strain and possible charge transfer between Au and Cu, which facilitated the phase transformation under such a mild condition. Similarly, treating fcc-AuPdCo nanoparticles at 1073 K in H₂ produced a mixed system consisting of primitive trigonal, rhombohedral, and face-centered cubic phases ^[59]. Treatment of disordered fcc-Pt₃Co nanoparticles at 973 K in H₂ produced the ordered L1₂intermetallic structure ^[60]. More interestingly, CO exerted a promotional effect on the phase transformation of metal nanoparticles through a carbonization process. Fcc-Co nanoparticles of 7 nm in a 25 wt%Co-0.1%Pt/Al₂O₃ sample that experienced hydrogen at 623 K were converted into hcp-Co nanoparticles of 9 nm after a successive treatment in flowing CO and H₂ at 623 K (Fig. 8) ^{[61, 62]}. In situ XRD measurement of this phase conversion process confirmed that the interaction of fcc-Co nanoparticles with CO formed cobalt carbides, which were finally evolved into hcp-Co nanoparticles by H₂ reduction.

Fig. 7. Element arrangements in the ordered L10 lattice of the AuCu nanoalloy (a) and TEM images of fcc-AuCu and L10-AuCu nanoparticles (b). Reprinted with permission from Ref. [58]. Copyright 2014, The Royal Society of Chemistry.

Fig. 8. XRD patterns of fcc-Co and hcp-Co nanoparticles in CoPt/Al2O3 catalysts (a) and their activity in FTS (b). Reprinted with permission from Ref. [62]. Copyright 2011, Elsevier.

Cobalt nanoparticles are typically used to catalyze Fischer-Tropsch synthesis (FTS) to convert syngas into hydrocarbons. Cobalt particles undergo chemical and structural evolution during the pretreatment and reaction, and the overall catalytic performance relies on the size of the resulting cobalt particles ^{[63, 64, 65, 66]}. One interesting observation was that the reaction becomes structure insensitive when the cobalt particles are larger than 6 nm ^{[67, 68]}. This remarkable size effect was ascribed to the enhanced proportion of low coordinated surface sites on small Co particles ( < 6 nm), which favored a higher coverage of irreversibly bonded CO. Surface studies have indentified that CO dissociation on the zigzag grooved (112-0) facet proceeds faster than that on the close-packed (0001) facet. More importantly, the chain growth probability on the former (0.36 at 523 K) is larger than that on the latter (0.20 at 523 K) ^[69].

The crystal phase transition of cobalt nanoparticles in practical catalysts, mainly achieved through the initial carbonization by CO and the subsequent H₂ reduction, also played a vital role in determining the performance in FTS. For example, the reaction rate and product distribution of 23 wt%Co-0.05 wt%Pt/Al₂O₃ catalysts clearly showed a crystal phase dependent effect. Under the same reaction conditions, hcp-Co nanoparticles provided a space time yield of 4.29 × 10^-5 mol _CO/(g_Co · s) and a C₅₊ selectivity of 94.1%, whereas fcc-Co particles showed a lower reaction rate of 3.55 × 10^-5 mol _CO/(g_Co · s) with a C₅₊ selectivity of 89.3% ^[70]. Similarly, with 25 wt%Co-0.1 wt%Pt/Al₂O₃ catalysts, the time yield of hcp-Co nanoparticles (9 nm) was two times that of fcc-Co nanoparticles (7 nm) (Fig. 8) ^{[61, 62]}. The turnover frequency (TOF) of hcp-Co particles (12.0 nm) in a 10 wt% Co/ZrO₂ catalyst approached 0.054 s^-1, being much larger than that of fcc-Co nanoparticles of 12.4 nm (0.033 s^-1) ^[71]. The superior performance of hcp-Co particles, in terms of activity and C₅₊ selectivity, was further shown with 20 wt%Co/SiO₂ catalysts ^[72]. The presence of surface defects and stacking faults on hcp-Co nanoparticles facilitated the inner layer diffusion of carbide from the surface to bulk and thus promoted CO activation ^[73]. Recent theoretical calculations further verified the promotional mechanism of hcp-Co nanoparticles in FTS ^[74]. According to the principle of Wulff construction, hcp-Co nanoparticles preferentially expose the (101-1) and (101-0) facets (Fig. 9). The former surface is richened with the B5-type site, an ensemble of four atoms of the surface and a fifth atom in the subsurface, which is the active center for CO dissociation and carbon chain growth ^{[75, 76, 77]}. Fcc-Co nanoparticles, however, were mainly enclosed by the close-packed (111) facet that is much less active for CO dissociation.

Fig. 9. Wulff constructions of hcp-Co and fcc-Co (a) and their reaction rates for CO dissociation (b). Reprinted with permission from Ref. [74]. Copyright 2013, American Chemical Society.

However, a recent example showed the higher activity of fcc-Co nanoparticles in 20 wt%Co/SiO₂ catalysts for FTS ^[78]. Calcination at 823 K and successive H₂ reduction at 673 K of the catalyst precursor yielded mainly hcp-Co nanoparticles of 31.2 nm while direct reduction by H₂ at 673 K predominately formed fcc-Co nanoparticles of 16.4 nm. The TOF of fcc-Co was 48.2 × 10^-3 s^-1 but it was 18.0 × 10^-3 s^-1 for hcp-Co nanoparticles, clearly demonstrating the importance of the crystal phase of the cobalt nanoparticles. The large hcp-Co particles showed a more regular surface with a prevalence of planar sites, while the smaller fcc-Co particle gave a higher fraction of coordinatively unsaturated sites. Therefore, the crystal phase effect of cobalt nanoparticles in FTS may be related to the chemical state, particle size, crystal facets exposed, and more importantly the dynamic change of the catalyst particles under reaction conditions. Nevertheless, these experimental results have identified that tuning the crystal phase of the cobalt nanoparticles, in conjunction with the size and morphology effects, provides a potential way to develop highly efficient FTS catalysts.

The crystal phase of cobalt nanoparticles also affects their surface rearrangement under FTS conditions as well as their oxidation to cobalt oxides under O₂ atmosphere. During FTS, CO activation and dissociation induced the surface reconstruction of the Co particles ^{[66, 79]}. The mostly exposed (111) facet of fcc-Co was reconstructed to the (100) plane ^[79], which has more uncoordinated atoms that are highly active for FTS ^[80]. The close-packed (0001) facet on hcp-Co was kept almost unchanged and only the stepped (101-2) facet experienced surface rearrangement to push Co rows apart in the ^[010] direction ^[81]. The oxidation of cobalt nanoparticles generally follows the Kirkendall process, which is determined by the mutual diffusion rates of cobalt and oxygen anions ^[82]. Oxidation of hcp-Co, ε-Coand fcc-Co nanoparticles (with a similar size of 8 nm) at 473 K formed a core-shell Co@CoO nanostructure, CoO hollow nanoparticle, and a mixture of hollow CoO and yolk-shell Co@CoO, respectively (Fig. 5) ^{[50, 83]}. This was attributed to the different stacking patterns of the cobalt atoms in these crystal phases, which influenced the rates of Kirkendall diffusion and cobalt oxide construction.

Iron nanoparticles are the active component in catalysts for ammonia synthesis and FTS ^{[84, 85]}. The crystal phase of Fe particles determines the activation of N₂ and CO molecules. The activation of N₂ on the Fe surface includes the initial dissociative adsorption and the subsequent formation of the nitride phase. The total nitrogen uptake on fcc-Fe is higher than that on bcc-Fe ^[86]. In the case of CO activation, bcc-Fe dissociates CO at 300 K, whereas the fcc phase only provides weak adsorption of CO (Fig. 10) ^[87]. This is due to the different adsorption geometry of CO on these phases. The smaller distance of the iron atoms (0.29 nm) and the fourfold hollow site on bcc-Fe facilitates CO adsorption through its π orbital, but the relatively larger distance of the atoms (0.34 nm) on fcc-Fe is too large to provide the necessary active site. A similar observation was Fe(CO)₅ decomposition on iron surfaces ^[88]. Fe(CO)₅ was dissociative activated on bcc-Fe but only molecularly adsorbed on fcc-Fe. All these studies on model samples suggested the importance of the iron crystal phase in the activation of N₂ and CO molecules and are helpful in designing and preparing highly efficient catalysts for ammonia synthesis and FTS.

Fig. 10. CO adsorption geometry on fcc-Fe and bcc-Fe, where the distance between Fe (B) and Fe (C) is the decisive factor. Reprinted with permission from Ref. [87]. Copyright 1996, Elsevier.

Ruthenium nanoparticles effectively catalyze the low temperature synthesis of ammonia ^{[89, 90]}, hydrogenation of benzene to cyclohexene ^[91], FTS ^{[85, 92]} and CO oxidation ^[93]. Ru prefers to adopt a hexagonal close-packed mode because its Gibbs free energy is smaller than that of the face-centered cubic model ^[94]. With the progress of solution chemistry, metastable fcc-Ru nanoparticles have been obtained by precisely modifying the surface energy and the reduction kinetics. An obvious crystal phase effect of Ru nanoparticles in Ru/Al₂O₃ was observed for CO oxidation. For fcc-Ru nanoparticles (larger than 3 nm), the temperature for 50% conversion (T₅₀) of CO was 426 K (Fig. 4), while the T₅₀ for hcp-Ru particles was increased to 446 K ^[48]. CO oxidation over Ru is initialized with the oxidation of the Ru(001) facet and the formed RuO₂(110) layers act as the active sites ^{[95, 96]}. fcc-Ru nanoparticles are enclosed by the (111) facet and present twin boundaries with fivefold symmetry axes, which facilitate the formation of RuO_x active sites and therefore enhanced the catalytic activity. The impact of the crystal phase of Ru nanoparticles was recently demonstrated in the hydrolysis of ammonia borane. The activation energy over hcp-Ru nanoparticles (1.7-2.9 nm) was 33-92 kJ/mol, but it decreased to 27.5 kJ/mol over fcc-Ru nanoparticles (2.2 nm) ^[49]. A similar behavior of the crystal phase was observed over Ru nanoparticles supported on layered MgAl oxides for the same reaction. The activation energy over hcp and fcc-Ru nanoparticles was 17.7 and 50.2 kJ/mol, respectively, although both samples have a very similar size of 2.0 nm ^[97].

Pallidum nanoparticles are widely used in the hydrogenation of organic substrates. Pd commonly crystallizes as a face -centered cubic structure, but the dissociation of molecular hydrogen into hydrogen atoms isotropically expands its lattice under reaction conditions ^{[98, 99, 100, 101]}. The formed hydride species transform from the dilute phase of PdH_x to a hydrogen-rich and lattice expanded phase (β-PdH_x) that is characterized with a higher activity for deep hydrogenation. This transformation is governed by the size and morphology of the Pd nanoparticles ^{[102, 103]}, and further influences their catalytic performance ^{[104, 105]}. In the case of the hydrogen reduction of nitrite to N₂ over Pd/γAl₂O₃ catalysts, the activity slightly decreased from 21 mg_nitrite /₍g_{surface Pd}·min) to 16 mg_nitrite / (g_{surface Pd}·min) as the size of the Pd particles decreased from 10 nm to 2 nm, but the selectivity to N₂ dramatically increased from 85% to 99% ^[106]. The promotion in N₂ selectivity was attributed to the fact that the smaller Pd particles prohibited the formation of the β-PdH_x species and hence limited the over-hydrogenation of nitrite to NH₃. Meanwhile, the strong hydrogenation ability of β-PdH_x species has been reported to benefit the hydrodesulfurization of thiophene and benzothiophene over PtPd alloy nanoparticles ^[107]. Recent theoretical calculations have revealed the chemical mechanism of PdH_x species for ethyl hydrogenation to ethane over Pd nanoparticles (Fig. 11) ^[108]. When the Pd surface was saturated by H₂ (H^ad₇₈H^sub_oPd₇₉), the rate constant of ethyl hydrogenation to ethane at 298 K was 5 × 10⁷ s^-1 with the activation energy of 29 kJ/mol. On further increasing the adsorbed amount of H₂ with the formation of subsurface-saturated H^ad₇₈H^sub₂₄Pd₇₉ species, the hydrogenation rate constant was strikingly increased to 8 × 10¹¹ s^-1 and the activation energy drastically decreased to only 5 kJ/mol. This was because the subsurface H species destabilized adsorbed H on the surface by changing the electronic structure of the Pd nanoparticles with the occupation of the H^ad-Pd anti-bonding orbital. These results convincingly demonstrated that the phase transition of the Pd hydride species should be taken into account when preparing efficient hydrogenation catalysts.

Fig. 11. Influence of adsorbed hydrogen on the rate constant of ethyl hydrogenation at 298 K over Pd nanoparticles. Reprinted with permission from Ref. [108]. Copyright 2014, WILEY-VCH Verlag GmbH & Co.

Alloyed metal nanoparticles present another strategy to enhance catalytic performance because of their versatile crystal phases. Recent progress in nanomaterials, especially the simultaneous control of size, shape, and crystal phase, provide a solid base for exactly tuning their electronic properties and coordination environment ^{[41, 42, 43, 44, 45, 46]}. When the crystal phase of Pt₃Co nanoparticles was changed from disordered fcc to ordered L1₂ under a H₂ atmosphere, the mass activity for the oxygen reduction reaction (ORR) significantly increased from 0.16 to 0.52 mA^-1·μg^-1_Pt (Fig. 12) ^[60]. Moreover, the ordered crystal structure also prevented the leaching of Co during the electrochemical reaction. The detailed structural analysis indicated that the phase transformation induced the formation of a Pt rich shell and a stable intermetallic Pt₃Co core. AuPdCo nanoparticles with a mixture of the P3-m, R3-m, and Fm3-m phases and a twin lattice structure showed a much higher ORR performance than the pure Fm3-m phase ^[59]. Hydrogen treatment of FePt@Pt catalysts with a core-shell structure at 923 K converted the crystallite phase of FePt from fcc to fct. This transformation released the over-compressed surface Pt strain and subsequently enhanced the ORR activity ^[109]. In the case of the electro-oxidation of ethanol, the activity of PtRhSn nanoparticles in the Niggliite mineral phase was twice that on the fcc-PtRhSn alloy because of the better ability in breaking the C-C bond of ethanol ^[110]. For the electrooxidation of ethanol and formic acid on ZrPt₃ nanoparticles, the hexagonal structure showed a superior performance to the fcc-phase, even when the former has a large particle size of 200 nm (Fig. 13). This was assigned to the higher surface energy of hcp-ZrPt₃ (1.47 J/m²) than that of the fcc phase (1.34 J/m²) ^[111]. Similarly, the crystal phase transformation of Ag₃In nanoparticles from disordered Fm3-m to ordered Pm3-m remarkably enhanced the activity for p-nitrophenol reduction by NaBH₄ because the ordered structure isolated the active sites ^[112]. PdRu nanoalloys with a mixture of the fcc and hcp phases catalyzed CO oxidation more effectively than fcc-PdRu nanoalloys, simply because of the generation of significant amounts of defect sites that facilitated the adsorption and activation of CO and O₂ molecules ^[113].

Fig. 12. HAADF-STEM image (a) of a L12-Pt3Co particle and its catalytic performance for ORR at room temperature (b). Reprinted with permission from Ref. [60]. Copyright 2013, Macmillan Publishers Limited.

Fig. 13. TEM image of fcc-ZrPt3 (a) and hcp-ZrPt3 (b) nanoparticles and their specific activities for ethanol electro-oxidation (c). Reprinted with permission from Ref. [111]. Copyright 2014, American Chemical Society.

In the past three decades, we have experienced and benefited from the impressive impact of size and morphology controlled synthesis of metal nanoparticles. The concept of nanocatalysis not only advanced the development of more efficient catalysts but also deepened fundamental understanding in the structure-performance relationship at the atomic, molecular, and nano scales. The size and morphology effects of metal nanoparticles are generally associated with the variations in electronic properties and geometrical structures. These nanocatalysts are crystallized in the thermodynamically stable phase. For examples, Ru, Fe, and Pt₃Co nanoparticles are present as the hcp, bcc and fcc phases, respectively. However, the metastable phases of fcc-Ru, fcc-Fe and L1₂-Pt₃Co were recently found to be more active for several chemical reactions, primarily due to the changes in electronic properties and coordination environment. This is a newly emerging and growing topic in nanocatalysis. There are several critical issues that should be taken into account when elaborating the effect of the crystal phase of the metal nanoparticles on the catalytic properties. First, the synthetic method for tuning the crystal phase of metal nanoparticles with a specific size and morphology is still empirical. The in-depth study of the synthesis mechanism in liquid solution would enable the simultaneous control of microstructure in terms of chemical composition, size, morphology and crystal phase. More importantly, there would be advances in the fabrication of metal nanoparticles in metastable crystal phases. Second, the phase transition of a metal nanoparticle is usually achieved by thermal treatment at elevated temperatures and under a reactive atmosphere, but this is accompanied with the sintering of nanoparticles or leaching of chemical elements. For example, the phase transformation of Pt₃Co nanoparticles from disordered fcc to ordered L1₂ structure under a hydrogen atmosphere at 973 K increased the particle size from 4.8 to 7.2 nm and caused Pt leaching with the formation of a Pt-rich shell ^[60]. Accordingly, the changes in ORR activity cannot be simply assigned to the crystal phase of the Pt₃Co particles, and should take the additional size effect and composition change into consideration. Spatial confinement of metal nanoparticles, for example, the construction of core-shell or yolk-shell nanostructures ^{[114, 115, 116, 117, 118, 119]} and the confinement of metal nanoparticles inside carbon nanotubes ^{[120, 121, 122, 123, 124, 125]}, may be technically applicable for avoiding sintering and leaching of metal nanoparticles during the phase transformation. One promising example is the confinement of fcc-Re nanowires ~0.4 nm wide in the channels of carbon nanotubes that resulted in a stronger interaction to stabilize the metastable fcc phase ^[126]. Finally, it is necessary to address the structural stability of the metastable crystal phase of metal nanoparticles under reaction conditions. The reaction temperature and atmosphere often induce reconstruction of the catalyst particle and therefore convert the metastable phase into the thermodynamically stable one during the course of catalytic reaction. A detailed structural analysis of working or used catalysts, especially the lattice structure and exposed crystal face, combined with the use of cutting-edge physical and chemical techniques, should be applied for clarifying the crystal phase effect in metal nanoparticles.