Strain and support effects on phase transition and surface reactivity of ultrathin ZnO films: DFT insights

Oxide-based catalysts have wide applications in heterogeneous catalysis. In the past decades, many important progress have been made toward fundamental understanding and the practical development of oxide-catalyzed processes, for example, CeO₂ for partial hydrogenation of alkynes to olefins,¹ ZnCrO_x/zeolite composite catalysts for syngas conversion to light olefins,² ZnO–ZrO₂ solid solution for CO₂ hydrogenation to methanol,³ and NiFe layered double hydroxides for the oxygen evolution reaction.⁴ Furthermore, oxides can be combined with metals to form inverse oxide/metal catalysts for advanced catalysis, for instance, interface-confined FeO_1−x/Pt catalysts for CO oxidation⁵ as well as CeO_x/Cu⁶ and ZnO/Cu⁷ catalysts for CO₂ hydrogenation. These catalysts show excellent performance owing to the synergy between the oxide and metal. In the oxide-containing composite catalysts, oxides may be either nanoparticles or two-dimensional nanofilms.^8–12 Particularly, the oxide films present well defined surface structures and provide a good opportunity for fundamental studies of oxide catalysis. A few systems have been well-investigated, including monolayer (ML) FeO supported on Pt¹³ and monolayer MgO supported on Au¹⁴ for CO oxidation, monolayer nickel (hydroxy) oxide supported on Pt(111) for the hydrogen evolution reaction,¹⁵ sub-monolayer ceria supported on Au(111) for the CO₂ reduction reaction,¹⁶ and sub-monolayer ZnO supported on Cu(111) for CO₂ hydrogenation.⁷ 

Ultrathin oxide films may exhibit distinct structures from their bulk phases,^17–21 which are dependent on external strains, epitaxial substrates, and preparation conditions.^22–24 Taking ZnO as an example, ZnO(0001)-oriented films transform from the wurtzite-like geometry (W-ZnO) to the graphene-like structure (G-ZnO) when the thickness decreases down to a threshold, which has been extensively demonstrated by both experiments and calculations.^21,25–33 A body centered tetragonal (BCT) phase has been theoretically predicted,^30,34 which is observed in quantum-confined ZnO nanowires.²² Besides the three phases, there are also others reported so far.^19,30 It is well known that the catalyst structure strongly influences its catalytic reactivity.^35–37 For example, W-ZnO is more reactive than G-ZnO toward CO oxidation and H₂ activation.²⁹ Thus, it is crucial to understand the structural transition of the ZnO film and its dependence on the factors that are crucial in catalysis such as strain and support.

For the strain effect, Wu et al.²⁶ demonstrated that tensile strain can delay the G-ZnO to W-ZnO phase transition in the free-standing (FS) ZnO films. However, a strict correlation among strain, phase transition, and surface reactivity is not yet established. For the support effect, several groups prepared supported ultrathin ZnO films but did not systematically study the influence of the substrate on the critical thickness (Tc) at which phase transition from G-ZnO to W-ZnO occurs. In addition, both the strain and support effects on the surface reactivity have not been clearly studied for ZnO films and other oxide catalysts, although they have been extensively investigated in transition metal catalysts.^38–44 

Here, we study the strain and support effects on the phase transition and surface adsorption of the ZnO films by employing density functional theory (DFT) calculations. vdW-DF (optPBE) functionals^45,46 are particularly used to address the van der Waals interaction, which show excellent performance toward the highly accurate description of oxides.⁴⁷ CO and NH₃ are chosen as the indicators of the surface reactivity since they are often used for the surface titration analysis of solid catalysts.^48,49 Both graphene-like and wurtzite-like phases are considered here, while the BCT phase has not been explored²⁶ considering that this structure may not follow epitaxial growth on substrates of Au(111), Pt(111), Ru(0001), and graphene. As a result, we find that besides delaying the G-ZnO to W-ZnO phase transition, tensile strain enhances the surface reactivity of ZnO films as indicated by the enhanced CO and NH₃ adsorptions. The metal substrates (Au, Pt, and Ru) show inverse effects on the phase transition as compared to the tensile strain, exhibiting larger manipulation of the surface reactivity than the strain. The results provide insights toward a fundamental understanding of the strain and support effects on oxide-based catalysts.

DFT calculations were implemented using a plane wave basis set in the Vienna Ab initio Simulation Packages (VASP 5.4).^50,51 The exchange-correlation energy was evaluated using the Perdew–Burke–Ernzerhof (PBE)⁵² functional within the generalized gradient approximation (GGA).⁵³ The projected-augmented wave (PAW)⁵⁴ pseudopotentials were utilized to describe the core electrons, and a cutoff energy of 400 eV was used for the plane-wave expansion. The following valence electron configurations were included in the self-consistent field calculations: Zn (3d¹⁰ and 4s²), O (2s² and 2p⁴), Au (5d¹⁰ and 6s¹), Pt (5d⁹ and 6s¹), Ru (4d⁷ and 5s¹), C (2s² and 2p²), N (2s² and 2p³), and H (1s¹). The van der Waals (vdW) dispersion forces were corrected by adopting the vdW-DF (optPBE) functionals.^45,46 The energies and residual forces were converged to 10⁻⁵ eV and 0.02 eV Å⁻¹, respectively. The effective charge was calculated through the Bader charge analysis.⁵⁵ The lattice constants of wurtzite ZnO were calculated to be a, b = 3.288 Å and c = 5.347 Å, agreeing well with the experimental results.⁵⁶ For free-standing (FS) ZnO, we used a (1 × 1) unit cell for the structural study and a (2 × 2) supercell for the adsorption study along the [0001]-orientation of wurtzite ZnO. The 7 × 7 × 1 and 4 × 4 × 1 k-point grids were used for the Brillouin zone sampling, respectively. For supported ZnO, the following supercells were used: ZnO-(3 × 3)/graphene-(4 × 4),⁵⁷ ZnO-(7 × 7)/Au(111)-(8 × 8),^58,59 ZnO-(5 × 5)/Pt(111)-(6 × 6),³³ and ZnO-(5 × 5)/Ru(0001)-(6 × 6), with an interface Zn/M (M denotes the substrate) ratio of 0.563, 0.766, 0.694, and 0.694, respectively. These models are adopted based on the experimentally observed moiré patterns or minimum lattice mismatch. A 3 × 3 × 1 k-point grid was used for ZnO/graphene, and a 2 × 2 × 1 k-point grid was used for the others. Dense k-point grids were also examined, which showed negligible difference on the trends. Three-layer metal slabs were adopted to mimic the substrate, and only the top two were constrained. The slabs were separated by an ∼15 Å vacuum layer. More details can be seen from the supplementary material.

Before describing the supported ZnO films, we first overview the geometries and energetics of a set of free-standing (FS) ZnO films with the thickness from 1 ML to 12 ML [1 ML denotes one monolayer, equal to one Zn–O layer shown in Fig. 1(a) and Fig. SI1 of the supplementary material]. Structural features include the in-plane lattice constants (a) [Fig. 1(b)] and rumple factors (R) defined as the average vertical distance between O and Zn atoms [Fig. 1(c)]. The energetics are denoted by the relative chemical potential (Δμ), which is the energy difference between the films and the bulk [Fig. 1(d)]. All involved formulas are compiled in the supplementary material.

With increasing film thickness [Fig. 1(b) and Table SI1 of the supplementary material], the in-plane lattice constants of FS-ZnO films gradually increase from 3.295 Å (0.21% of intrinsic strain, relative to the bulk) at 1 ML to a peak value of 3.431 Å (4.35% of intrinsic strain) at 8 ML and then sharply decrease to 3.355 Å (2.04% of intrinsic strain) at 9 ML and 3.315 Å (0.82% of intrinsic strain) at 12 ML. The dramatic change of lattice constants at 9 ML is also reflected by the rumple factor. While R is less than 0.04 Å from 1 to 8 MLs, it sharply increases to 0.41 Å at 9 ML [the value of bulk wurtzite ZnO is 0.62 Å], as shown by the blue lines in Fig. 1(c). The changes in the lattice and the rumple indicate a phase transition from the graphene-like ZnO (G-ZnO) to wurtzite-like ZnO (W-ZnO) geometry. For convenience, we here define 9 ML (i.e., the thinnest film with the W-ZnO phase) as the Tc of the FS-ZnO film system. To clarify the phase transition from energetics, we compared the relative chemical potential (Δμ) of the G-ZnO and W-ZnO phases [Fig. 1(d)]. Clearly, G-ZnO is more stable than W-ZnO with the film thickness less than 9 ML since G-ZnO can lead to the release of the nominal surface energy up to 1.5 eV/ZnO, as shown in Fig. 1(e). With increasing the thickness of ZnO films, the bulk energetic becomes dominant, which leads to the phase transition from G-ZnO to W-ZnO at 9 ML. Such a phase transition is in good agreement with the previous reports.^25,26

The responses of ZnO films upon the biaxial strain are investigated, namely, by varying their lattices in the basal plane. Here, the largest in-plane lattice of 8 ML ZnO (a = 3.431 Å, 4.35% of intrinsic strain vs bulk lattice) was chosen as the zero-strain reference for convenience. Then, ZnO films with relative strain from −12% to +3% were constructed, which correspond to the intrinsic strains of −8.17% to +7.48% vs the bulk lattice. As shown in Figs. 1(f) and 1(g) and Figs. SI2 and SI3 of the supplementary material, Tc decreases (increases) with the compressive (tensile) strain. For example, Tc changes from 9 ML (FS-ZnO) to 14 ML and 5 ML with +2% and −5% of relative strain, respectively. Such a trend indicates that tensile strain enables a later phase transition, while the effect is opposite for the compressive strain.²⁶ In addition, the quantitative analysis indicates that the tensile strain has a larger impact on the shift of Tc, i.e., ∼2 ML/1% strain, than the compressive strain, i.e., ∼1 ML/1% strain [Fig. 1(g)]. Strain needs to be −11% (∼−6% of intrinsic strain) in order to keep the monolayer ZnO film with the W-ZnO phase.

The strain effect on the surface reactivity of ZnO films has been investigated by comparing adsorption energies of CO and NH₃ on the monolayer FS-ZnO films. Figure 2(a) shows the optimal adsorption configurations of CO and NH₃, and a set of strained monolayer G-ZnO films with 1%–10% of tensile strain is constructed with respect to the lattice (a = b = 3.295 Å) of monolayer FS-ZnO. Figure 2(b) illustrates that adsorption energies of CO and NH₃ increase linearly from −0.17 eV and −0.45 eV at zero strain to −0.25 eV and −0.72 eV at 10% strain, respectively, suggesting that the tensile strain enhances the molecular adsorptions. The calculations of CO adsorption on the thick (2 ML–5 ML) FS-ZnO films [Fig. 2(c)] confirm the trend established on monolayer ZnO, and furthermore, the thickness-dependent adsorption energy shows that the adsorption energy increases with the ZnO thickness. This should be from the intrinsic tensile strain effect⁴³ because the film lattices expand with the thickness increasing from 1 ML to 5 ML.

Density of states (DOS) analysis [Fig. 2(d)] indicates that despite the underestimated bandgap by PBE,⁶⁰ all ZnO films show the semiconducting feature, which explains the weak adsorptions of both CO and NH₃. However, the bandgap decreases and peaks of the unfilled states (blue arrow) shift toward the Fermi level^61–63 with increasing tensile strain. This indicates that the Zn–O bond is weakened, resulting in the enhanced binding to CO and NH₃. Analogous to the cases of transition metal catalysts,^39–41,43 electronic structures of the oxides can be tuned by the strain causing the change in metal–oxygen bonds, which then modulate the surface adsorption and catalytic activity.

Ultrathin ZnO films are supported on four substrates, including graphene (Gr), Au(111), Pt(111), and Ru(0001), which are abbreviated as ZnO/Gr, ZnO/Au, ZnO/Pt, and ZnO/Ru, respectively. Taking ZnO/Au and ZnO/Pt as examples, structures around the phase transition are shown in Fig. 3(a) (see Figs. SI4–SI7 of the supplementary material for more details). The rumple factor (R) and relative chemical potential (Δμ) of ZnO overlayers vs the film thicknesses are shown in Figs. 3(b) and 3(c). For ZnO/Gr, ZnO/Au, ZnO/Pt, and ZnO/Ru, Tc values are 6 ML, 6 ML, 5 ML, and 4 ML, respectively, which are 3 ML, 3 ML, 4 ML, and 5 ML thinner than 9 ML for FS-ZnO. Residual strain in the film due to the lattice mismatch coupled with the film–substrate interaction should contribute to the above differences in Tc. The disentanglement of the interfacial interaction from the strain is the key to understand the pure electronic effect of the support. We note that the relative strains of the ZnO overlayer are −3.9%, −1.3%, −1.4%, and −4.8%, respectively, for ZnO/Gr, ZnO/Au, ZnO/Pt, and ZnO/Ru, as compared to the lattice of 8 ML FS-ZnO (see details from Table SI2 of the supplementary material). Figure 3(d) shows that Tc values purely due to −1%, −4%, and −5% strain are 8 ML, 6 ML, and 5 ML, respectively. Thus, the decreased Tc values induced by the support effect are ∼2 ML, 3 ML, 1 ML, and 0 ML, respectively, for the Au, Pt, Ru, and Gr-supported ZnO films. Such a disentanglement indicates that the metal substrates have similar effects on the phase transition as the compressive strain, i.e., promoting an early phase transition.

To deepen understanding of the support effect, we calculated ZnO-M interface adhesion energy (E_adh) for the supported 1 ML–6 ML ZnO films [Fig. 3(e)]. Before the phase transition (≤3 ML), E_adh values decrease from graphene, Au, Pt, to Ru-supported ZnO. Taking monolayer ZnO as a case, E_adh values are −0.18 eV, −0.27 eV, −0.42 eV, and −0.45 eV/interface ZnO, where the corresponding ZnO–substrate charge transfer is 0.001, 0.022, 0.054, and −0.051 |e| per interfacial ZnO, respectively, for the ZnO/Gr, ZnO/Au, ZnO/Pt, and ZnO/Ru systems ( see Fig. SI8 and Table SI3 of the supplementary material). The strong ZnO–substrate interaction^64,65 leads to the enhanced rumple and polarization of the G-ZnO films, which can be regarded as the precursor of the W-ZnO phase. Accordingly, it is expected that the strong ZnO–support interaction induces an early phase transition of ZnO films. It should be noted that E_adh shows a sharp change at Tc due to the different interaction of the substrates with G-ZnO and W-ZnO phases.

Because of the weak interaction between ZnO and the graphene substrate, we here only consider Au, Pt, and Ru-supported ZnO and the support effect on CO and NH₃ adsorptions on the surfaces of ZnO overlayers. Figure 4(a) shows the optimal configurations of CO and NH₃ adsorption (the other adsorption sites and energies are given in Fig. SI9 and Table SI4 of the supplementary material). As shown in Fig. 4(b), CO adsorption energy increases from −0.17 eV on FS-ZnO to −0.19 eV, −0.39 eV, and −0.47 eV on the ZnO/Au, ZnO/Pt, and ZnO/Ru surfaces, respectively. The corresponding NH₃ adsorption energy increases from −0.45 eV to −0.75, −0.96, and −1.20 eV. It is inferred that the magnitude of the increased adsorption energy varies from −0.03 eV (ZnO/Au) to −0.22 eV (ZnO/Pt) to −0.30 eV (ZnO/Ru) for CO adsorption and from −0.30 eV to −0.51 eV to −0.75 eV for NH₃ adsorption.

To understand the origin of the enhanced adsorption, we first evaluate the contribution from the residual strain effect. On monolayer FS-ZnO, the increased adsorption energies are −0.01 eV for CO and −0.07 eV for NH₃ when applying a 3% tensile strain [Fig. 2(b)]. Relative strains (vs monolayer FS-ZnO) in supported ZnO monolayers on Au, Pt, and Ru substrates are 2.81%, 2.67%, and −0.87%, respectively, which are all less than 3%. Thus, the increase in CO adsorption energy from 0.03 eV to 0.30 eV and that in NH₃ adsorption energy from 0.30 eV to 0.75 eV indicate that the support effect plays a major role in the enhanced surface reactivity. Moreover, when applying 10% tensile strain, the increased amounts of adsorption strengths are 0.08 eV for CO and 0.27 eV for NH₃, which are lower than those induced by the substrates (except for CO adsorption on ZnO/Au). This suggests that the substrate exerts a larger effect on the surface reactivity than the strain. Such an enhancement can be correlated with the strong ZnO–substrate interaction, which weakens the Zn–O interaction and increases the activity of Zn sites. As shown in Fig. 4(c), the DOS close to the Fermi level marked by yellow increases in the order of FS-ZnO, ZnO/Au, ZnO/Pt, and ZnO/Ru, which is consistent with the trend of the ZnO–substrate interaction strength and the molecule adsorption.

We find that compressive strain promotes an early phase transition of ZnO ultrathin films from the inert graphene-like structure to the more reactive wurtzite-like one and tensile strain leads to stronger CO and NH₃ adsorption on the surface due to the elongated Zn–O bond. The metal substrates induce a similar effect as the compressive strain, i.e., causing an early phase transition toward the W-ZnO film phase. The stronger the ZnO–substrate interaction, the earlier the phase transition. In comparison with the strain, the substrate has a larger effect on the surface reactivity of the ultrathin ZnO films. The enhanced surface adsorption of the supported ZnO films can be attributed to the increased density of electronic states close to the Fermi level, which is due to the strong ZnO–substrate interaction and weakened Zn–O bonding.

See the supplementary material for additional details including figures, tables, and explanations.

This work was financially supported by the National Natural Science Foundation of China (Grant Nos. 21688102, 21825203, and 91945302), the National Key R&D Program of China (Grant Nos. 2017YFB0602205 and 2016YFA0200200), and the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB17020000).

The data that support the findings of this study are available from the corresponding author upon reasonable request.