Facet-Dependent Gas Adsorption Selectivity on ZnO: A DFT Study

Abstract

Semiconductor-based gas sensors are of great interest in both industrial and research settings, but poor selectivity has hindered their further development. Current efforts including doping, surface modifications and facet controlling have been proved effective. However, the “methods-selectivity” correlation is ambiguous because of uncontrollable defects and surface states during the experiments. Here, as a case study, using a DFT method, we studied the adsorption features of commonly tested gases—CH₂O, H₂, C₂H₅OH, CH₃COCH₃, and NH_3—on facets of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$. The adsorption energies and charge transfers were calculated, and adsorption selectivity was analyzed. The results show $\mathrm{ZnO}\left(000\overline{1}\right)$ has obvious CH₂O adsorption selectivity; $\mathrm{ZnO}\left(10\overline{1}0\right)$ has a slight selectivity to C₂H₅OH and NH₃; and $\mathrm{ZnO}\left(10\overline{1}1\right)$ has a slight selectivity to H₂, which agrees with the experimental results. The mechanism of the selective adsorption features was studied in terms of polarity, geometric matching and electronic structure matching. The results show the adsorption selectivity is attributed to a joint effort of electronic structure matching and geometric matching: the former allows for specific gas/slab interactions, the latter decides the strength of the interactions. As the sensing mechanism is probably dominated by gas–lattice interactions, this work is envisioned to be helpful in designing new sensing material with high selectivity.

1. Introduction

Due to high sensitivity, fast response, low cost and ease of fabrication and integrations, semiconductor-based gas sensors have been extensively used in leak alarms, pollution monitoring, building air quality control and heritage conservation airborne pollution monitors, etc. [1,2,3]. However, one of the biggest challenges of these sensors is poor selectivity, which hinders their further application [4,5]. To deal with this well-known challenge, researchers have developed a variety of methods, which can be divided into two types, i.e., indirect and direct. The indirect methods include using specific gas filters [6] and introducing the other output parameters apart from resistance, namely “multi-variable outputs”, such as pulse heating-induced parameters [7] and impedance-related parameters [8,9]. The indirect methods usually solve the challenge of selectivity perfectly at the initial stage. However, they might suffer from uncontrollable selectivity output along with the fading processes of the sensing materials and devices because complex parameters and algorithms make it hard to correct the selectivity. Worse still, the added complexity of the sensing system could greatly weaken the existing merits of the semiconductor-based gas sensors. The other type of method is the direct design of the sensing materials with high selectivity, which is much simpler and more efficient. The direct methods include heteroatom doping [10,11], facet controlling [12,13] and surface functioning [1,2,14], etc. Although great progress has been made in the past decade, the unambiguously correlated “methods-selectivity” relation is hard to achieve because of uncontrollable defect states, surface status and material stacking morphologies during the experiments [5,15], which severely hinders the development of new selective sensing materials.

Currently, some researchers have used density function theory (DFT) to analyze the semiconductor-based gas sensing properties and successfully explained the experimental results [1,16]. However, most calculations mainly focus on specific new materials in terms of doping or surface functioning; few contribute to the facet-dependent selectivity, even though it has been extensively studied by experiments due to ease of control and apparent structure selectivity correlations [15,17,18].

Here, by using the DFT method, we studied the most commonly seen ZnO facets of $\left(000\overline{1}\right)$, $\left(10\overline{1}0\right)$ and $\left(10\overline{1}1\right)$ [13,15,19,20] and the adsorption features of the most commonly tested gases (CH₂O, H₂, C₂H₅OH, CH₃COCH₃, NH₃) [2,13,15,20,21,22,23,24] on these facets. The adsorption selectivity and possible mechanisms were systematically investigated and analyzed. The choice of ZnO is because of it is ease of being synthesized in the form of single crystals with specific facet exposure. The choice of the $\left(000\overline{1}\right)$ facet of ZnO is because of its higher stability in air [25], which is representative in the air background sensing environment. Similarly, O-terminated ZnO $\left(10\overline{1}1\right)$ was studied instead of the Zn-terminated one. O₂ is not considered here, since this work mainly focuses on the “adsorption” process instead of the “adsorption and transduction” process of gas sensing [26]. Additionally, it has been pointed out that the sensing mechanism might be that the gases interact with lattice O on the surfaces, which causes a change of surface conductivity, and the O₂ in the air helps to restore the surface conductivity [27]. Out previous work also clearly showed this possibility, since the ethanol response in an N₂ background is much higher than in the air background [14].

2. Computation Methods

All calculations were performed using the Vienna ab initio simulation package (VASP) based on the DFT framework [28]. The projector-augmented wave (PAW) with cut-off energy of 450 eV was adopted to describe the ion–electron interaction [29]. The gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) exchange correlation function was adopted to describe the electron exchange correlation [30]. During calculations, the Zn (3d¹⁰, 4p²), O (2s², 2p⁴), H (1s¹) and C (2s²2p²) were treated as valence electrons.

Slabs of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$ surfaces with 96 atoms (Zn48O48) were built by cleaving the optimized wurtzite primary cell along corresponding directions, with a thickness of four bilayers and supercells of 3 × 2. A vacuum layer of 15 Å was added to the top along the z direction to get rid of interactions between periodic image structures. During surface calculations, the bottom two bilayers were fixed, and the two top bilayers were allowed to relax. For the Brillouin zone integration, gamma-centered k-point mesh densities of 3 × 3 × 1 and 4 × 4 × 1 were adopted for geometry optimization and electron structure calculation, respectively.

Gases of hydrogen (H₂), ammonia (NH₃), formaldehyde (CH₂O), ethanol (C₂H₅OH) and acetone (CH₃COCH₃) were pre-optimized from available conformations [31] using gamma point in a box of 20 × 20.01 × 20.02 ų before adsorption calculations. The convergence criterion for geometry optimization and electron structure calculation were set as 1 × 10⁻⁵ eV and 1 × 10⁻⁶ eV, respectively, with force convergence set to 0.01 eV/Å. During calculations, the DFT-D3 correction proposed by Grimme [32] was adopted to correct the dispersion force.

The gas adsorption selectivity of various gases on specific surface slabs was evaluated via adsorption energy, which can be calculated as:

$$E_{\mathrm{ads}}=E_{\mathrm{slab}\text{-}\mathrm{gas}}-E_{\mathrm{gas}}$$

(1)

where E_slab-gas and E_gas are the energies of gas adsorption systems and isolated gases, respectively.

To clarify the charge transfer during gas adsorption, Bader charge was calculated using the code developed by the Henkelman group [33]. Additionally, the charge differences during gas adsorption were visualized via calculating the charge density difference (CDD) and visualizing in VESTA [34], as explained in our last work [35].

3. Results

3.1. Surface Properties

To understand the differences among the interested ZnO facets, slabs of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$ were built and relaxed. The relaxed slabs are shown in Figure 1. Figure 1a is the bulk wurtzite ZnO model projected along $\left[1\overline{2}10\right]$, which clearly illustrates $\left\{10\overline{1}1\right\}$, $\left\{0001\right\}$ and $\left\{10\overline{1}0\right\}$ facets.

Figure 1b portrays relaxed slabs of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$, respectively. For $\mathrm{ZnO}\left(000\overline{1}\right)$, the surface is terminated only by O ions, which results in polarity. The O ions in the first layer relaxed inward, which agrees well with previous results [36]. For $\mathrm{ZnO}\left(10\overline{1}0\right)$, terrace structures with O and Zn ions alternatively arranged on the surface. In the uppermost surface layers (the “freed bilayers”), the O ions remain almost in bulk positions with small relaxations away from the surface; the Zn ions relax inward (0.33 Å) and a movement (0.20 Å) parallel to the surface (b axis), which agree well with the measured values of 0.40 Å [37] and theoretical studies of 0.16 Å [38], respectively. For $\mathrm{ZnO}\left(10\overline{1}1\right)$, the surface is terminated by O ions only. Thus, the polarity is expected. The upmost O ions relax strongly due to fewer coordination features.

The calculated surface properties can be calculated as:

$$E_{\mathrm{surf}}=\frac{1}{2A}\left(E_{\mathrm{slab}}-N\times E_{\mathrm{bulk}}\right)$$

(2)

where A, E_slab, N, E_bulk are the surface area, total energy of a surface slab, number of Zn-O units in the slab, energy per Zn-O unit in the bulk material, respectively.

The calculated surface energies along with system energies and coordination numbers of Zn/O ions in the first bilayer are tabulated in

Table 1. Surface properties of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$ after relaxation.

Surface	Atoms	System Energy (eV)	Coordination Number of O on the 1st Bilayer	Coordination Number of Zn on the 1st Bilayer	Surface Energy (J/m2)
					This Work	Ref.
$\mathrm{ZnO}\left(000\overline{1}\right)$	Zn48O48	−406.12	3	4	1.39	1.01 [39]; 0.96 [25]
$\mathrm{ZnO}\left(10\overline{1}0\right)$	Zn48O48	−415.41	4, 3	4, 3	0.87	0.91 [40]; 0.82 [39]; 1.12 [41]
$\mathrm{ZnO}\left(10\overline{1}1\right)$	Zn48O48	−395.93	3, 2	4	2.00	1.74 [40]

. Due to the asymmetric feature when cutting a solid into two polar surfaces, the surface energies of an O-terminated polar surface might be overestimated, as the Zn-terminated surfaces usually have higher surface energies [39]. Nevertheless, the obtained trend agrees with the published results [25,39,40,41], i.e., the fewer coordinated ions the surface has, the higher the surface energy. In addition, it has been found that the surface energy is correlated to the surface catalytic activity [42]. Therefore, the gas-slab interaction of $\mathrm{ZnO}\left(10\overline{1}1\right)$ is foreseen to be the strongest.

3.2. Adsorption Configuration

The calculated adsorption conformations with the lowest energy are shown in Figure 2. For $\mathrm{ZnO}\left(000\overline{1}\right)$, CH₂O adsorbs vertically and decomposes H atoms; H₂ and CH₃COCH₃ adsorb horizontally over the surface, which indicates weak interactions with the slab; C₂H₅OH adsorbs vertically with a distance of 1.635 Å via interactions between its hydroxyl group with lattice O (O_L); NH₃ absorbs on lattice Zn via its N atom and all H atoms are aligned to specific O_L ions. For $\mathrm{ZnO}\left(10\overline{1}0\right)$, due to terrace structures alternatively arranged with Zn and O ions, all adsorbents adsorb on the slab in a structure of bridging the O_L and lattice Zn, except the case of H₂. For $\mathrm{ZnO}\left(10\overline{1}1\right)$, all gases adsorb and dehydrogenate on the less coordinated O_L, except CH₃COCH₃. Nevertheless, there is a strong surface reconstruction during CH₃COCH₃ adsorption. In addition, it is noticed that when NH₃ adsorb, a destructive adsorption of forming aminoxide happens.

3.3. Adsorption Energy and Charge Transfer

To understand the interactions between gases and slabs, the adsorption energy and Bader charge were calculated, and the results are shown in Figure 3 and Table S1. For $\mathrm{ZnO}\left(000\overline{1}\right)$, apart from the decomposed CH₂O, which has a high adsorption energy of −4.13 eV, the other four gases show much less adsorption energy, and in an order of NH₃ (−0.66 eV) > C₂H₅OH (−0.64 eV) > CH₃COCH₃ (−0.34 eV) > H₂ (−0.11 eV). A similar trend is also found in charge transfer. Therefore, the results show $\mathrm{ZnO}\left(000\overline{1}\right)$ has adsorption selectivity to CH₂O.

For $\mathrm{ZnO}\left(10\overline{1}0\right)$, adsorption energies of CH₂O, H₂, C₂H₅OH, CH₃COCH₃ and NH₃ are −1.21, −0.12 −1.43, −1.17 and −1.44 eV, respectively. Only a slight selectivity to C₂H₅OH and NH₃ is observed. Interestingly, the charge transfer results show negative values when CH₂O and H₂ adsorb on $\mathrm{ZnO}\left(10\overline{1}0\right)$, and a relatively small value in the case of C₂H₅OH.

For $\mathrm{ZnO}\left(10\overline{1}1\right)$, adsorption energies of CH₂O, H₂, C₂H₅OH, CH₃COCH₃ and NH₃ are −4.08, −4.62, −4.16 eV, −0.77 and −2.78 eV, respectively. The gas adsorption energies are much higher than on the other slabs due to dehydrogenation. A slight selectivity to H₂ of the slab is shown. The charge transfer shows a similar trend as the adsorption energy, but H₂ shows a relatively lower charge transfer.

Additionally, to understand how charge transfer happens during gas adsorption, CDD of the systems were calculated, and the results are shown in Figure 4.

For $\mathrm{ZnO}\left(000\overline{1}\right)$, it is noticed that the decomposed H atoms of CH₂O show strong interactions with O_L, as intense isosurfaces of charge overlap, while the remaining C-O has no obvious contribution. H₂ interacts weakly with both O_L and Zn, of which the saturated isosurface value is set as 4% of the default value (0.05 e/bohr³). It accepts electrons from Zn ions and donates electrons to O_L ions simultaneously, which further lowers the charge transfer to a negligible value. C₂H₅OH donates electrons via H–O_L interactions of the hydroxyl group. CH₃COCH₃ donates electrons via H atoms of the methyl group and C atoms of the ketone group. NH₃ donates electrons via H–O_L interactions, and accepts electrons via N–Zn interactions, resulting in little charge transfer.

For $\mathrm{ZnO}\left(10\overline{1}0\right)$, CH₂O donates and accepts electrons via C–O_L and O–Zn interactions, respectively. Due to stronger electronegativity differences of O–Zn compared with C–O_L, the net charge transfer is negative. Similarly, C₂H₅OH, CH₃COCH₃ and NH₃ adsorb with a bridge structure that donates and accepts electrons simultaneously. The difference is that those molecules donate electrons via H–O_L interactions, and therefore, the charge transfers are positive. The small values of the charge transfer can be attributed to weak interactions due to large H–O_L distances. The relatively higher charge transfer value of NH₃ can be explained by weaker electronegativity of N in NH₃ compared with O, which limits its ability in obtaining electrons from the Zn ions on the slab. The absorption conformation of H₂ is far away from the slab (2.37 Å), and the H atom weakly interacts with O_L, which results in little charge transfer.

For $\mathrm{ZnO}\left(10\overline{1}1\right)$, due to adsorption conformations with H decomposition, CH₂O, H₂, C₂H₅OH and NH₃ donate lots of electrons via dehydrogenation. CH₃COCH₃ donates electrons via H–O_L interactions and accepts electrons via O–Zn interactions. Therefore, a comparable charge transfer is obtained as on the slab of $\mathrm{ZnO}\left(10\overline{1}0\right)$, even though $\mathrm{ZnO}\left(10\overline{1}1\right)$ is more reactive due to less coordinated O_L.

4. Selectivity Analysis and Mechanism Discussion

As discussed above, $\mathrm{ZnO}\left(000\overline{1}\right)$ shows good selectivity for CH₂O adsorption; $\mathrm{ZnO}\left(10\overline{1}0\right)$ has a slight selectivity to C₂H₅OH and NH₃, and all gases except H₂ show comparable adsorption energies. $\mathrm{ZnO}\left(10\overline{1}1\right)$ has a slight selectivity to H₂, and all gases except CH₃COCH₃ can adsorb and decompose on the surface. For comparison, the gas adsorption selectivity is listed along with the published experimental results in gas sensing, as shown in

Table 2. Comparison between the calculated selectivity and the published experimental results.

ZnO Facets	Calculated Adsorption Selectivity	Gas Sensing Selectivity by Experiments
$\mathrm{ZnO}\left(000\overline{1}\right)$	CH2O 1 > NH3, C2H5OH > CH3COCH3 > H2	C2H5OH (s) 2 [13,15,20], C2H5OH, CH2O (s) [21]
$\mathrm{ZnO}\left(10\overline{1}1\right)$	NH3, C2H5OH > CH2O > CH3COCH3 > H2	CH2O (p) [2]; C2H5OH (s) [13,15,20]; C2H5OH, CH3COCH3 (s) [22]; C2H5OH, CH2O (s) [21]; NH3 (s) [10]
$\mathrm{ZnO}\left(10\overline{1}1\right)$	H2 > C2H5OH > CH2O > NH3 > CH3COCH3	CH2O (p) [2]; C2H5OH (s) [13,24]; NH3 (s) [10]; H2 (p) [23]

¹ The selectivity to CH₂O is obvious. ² The “s” in the bracket denotes the material is a ZnO single crystal; “p” in the bracket denotes ZnO polycrystal nanoparticles or grains.

. It is shown that the calculated adsorption selectivity roughly agrees with the experimental results, i.e., $\mathrm{ZnO}\left(000\overline{1}\right)$ has adsorption selectivity to CH₂O and C₂H₅OH; $\mathrm{ZnO}\left(10\overline{1}0\right)$ has adsorption selectivity to NH₃ and C₂H₅OH; $\mathrm{ZnO}\left(10\overline{1}1\right)$ has adsorption selectivity to H₂, C₂H₅OH and CH₂O, and H₂ > C₂H₅OH, CH₂O. The unmatched part can be attributed to: (1) there might be various surface states, such as defects and contaminations or coatings, that are not considered in the calculations; (2) the differences between “adsorption selectivity” and “sensing selectivity”, where the latter possesses the other process of “transduction” apart from “adsorption”.

To further understand the mechanism of the adsorption selectivity, polarity, electrostatic potential and electronic structure of both gases and slabs were further studied.

4.1. Polarity

It is reported that polarity plays an important role in gas sensing selectivity, such as selectivity to CH₃COCH₃ over C₂H₅OH [4]. Here, the experimental polarities of the gases are listed in

Table 3. Experimental polarity of the studied gases [43].

Gases	CH3COCH3	CH2O	C2H5OH	NH3	H2
Dipole moment (Debye)	2.88	2.33	1.52	1.48	0

. The polarity of gases can be listed in an order of CH₃COCH₃ > CH₂O > C₂H₅OH > NH₃, while H₂ is nonpolar. When compared to gas adsorption results shown in Figure 3a, it can be found that polarity seems to be true in the case of H₂/$\mathrm{ZnO}\left(000\overline{1}\right)$, where other polar gases show higher adsorption energies. However, it might be hard to explain the highest adsorption energy of H₂ on the polar surface of $\mathrm{ZnO}\left(10\overline{1}1\right)$. In addition, the order of adsorption energies on the typical polar surface of $\mathrm{ZnO}\left(000\overline{1}\right)$ does not match with the polarity of the gases. Therefore, our calculated results do not show evidence of polarity influences on selectivity.

4.2. Geometric Matching and Electrostatic Interactions

The electrostatic potential is another important factor that has been considered for interface adsorption [44,45]. The electrostatic potential mapping onto electron isosurfaces of the gas molecules according to ref. [45] are shown in Figure 5. The CH₂O has a symmetrical electrostatic potential along its vertical conformation, where the O atom is in negative potential and C/H atoms are in positive potential, as shown in Figure 5a. Figure 5b shows an unsymmetrical potential around C₂H₅OH, where the potential near the O atom is −171.17 kJ/mol, and potentials near the other atoms are positive, with the highest value of 230.96 kJ/mol showing up near the H atom of the hydroxyl group. Figure 5c shows an unsymmetrical electrostatic potential around the CH₃COCH₃ molecule, with a bigger negative potential near the O atom and positive potential near the other atoms. Figure 5d shows an unsymmetrical electrostatic potential around NH₃, where the bigger negative potential of −226.60 kJ/mol is near N and the smaller positive value of 145.71 kJ/mol near the other three H atoms. Figure 5f shows the electrostatic potential around H₂, where the positive potential is around two ends and the negative potential in the middle.

Due to the larger electronegativity of O compared with Zn [46], O ions accumulate electrons, while Zn ions deplete electrons, resulting in the negative potential of O and the positive potential of Zn. The calculated Bader charge states of O/Zn ions in specific slabs are shown in Figure S1 and Table S2. Combining with the adsorption conformations in Figure 2, it is found the gases prefer to adsorb on the slab with the atom that has the biggest electrostatic potential, and the adsorption sites are usually the opposite charge to the atom.

As a result, CH₂O prefers to adsorb on the ZnO slabs with H/C–O_L and O–Zn interactions, respectively. C₂H₅OH prefers to adsorb with H–O_L interactions in the hydroxyl group. CH₃COCH₃ prefers to adsorb with O–Zn interactions. NH₃ prefers to adsorb with N–Zn interactions or N–O_L interactions due to long pairs. H₂ prefers to adsorb with one side towards O_L or the central part close to Zn. Due to the surfaces of $\mathrm{ZnO}\left(000\overline{1}\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$ being terminated with O, CH₂O adorbs on the surfaces with two H atoms. Since lattice parameters of the slab surface match the molecule binding sites, the adsorbed CH₂O decomposed into adsorbed H atoms and isolated CO, as schematically shown in Figure 6. Since the preferred adsorption sites of the other gases cannot match the surface lattice of $\mathrm{ZnO}\left(000\overline{1}\right)$, only CH₂O exhibits obviously selective adsorption.

As for $\mathrm{ZnO}\left(10\overline{1}1\right)$, due to high chemical activity of the less coordinated O_L ions (as shown in Figure S1 and Table S2), the H bond in the gas molecules is vulnerable. As a result, H₂, CH₂O, C₂H₅OH and NH₃ all show high adsorption energies during adsorption on the surface. The geometric matching between the adsorption sites and the surface lattice allows for a slightly higher adsorption selectivity for H₂. As for $\mathrm{ZnO}\left(10\overline{1}0\right)$, it is because of its Zn/O_L alternative terrace structure that CH₂O, C₂H₅OH, CH₃COCH₃ and NH₃ can adsorb with a “bridge” conformation. In addition, all these molecules show high adsorption energies.

Therefore, geometry matching between the surface lattice and the adsorption sites indeed plays an important role in gas adsorption selectivity based on our calculations.

4.3. Electronic Structure Matching

In addition to geometry matching, the electronic structures of the slabs and the gases were also considered. The calculated highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies of the gases are tabulated in Table S4, which agrees well with the data in the Benchmark DataBase of NIST [43]. The density of states (DOS) of the adsorption systems (gas molecule and slabs) before adsorption interactions are shown in Figure 7. The DOSs of gas molecules are the DOSs of isolated gases shifting to energy alignment with the specific slabs, and the DOSs of slabs are the projected DOSs of isolated slabs at O 2p/Zn 3d orbitals of the first bilayer. Additionally, since the Fermi energies calculated by VASP is set as the valence band maximum [47], here the Fermi energies of specific slabs were shifted according to their experimental work function [48,49,50], as illustrated in part 3 of the Supplementary Materials.

It is found that the HOMO energies of four gases (CH₂O, C₂H₅OH, CH₃COCH₃ and NH₃) are higher than the Fermi energies (E_f) of $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$, but lower than the E_f of $\mathrm{ZnO}\left(000\overline{1}\right)$. The HOMO energy of H₂ is lower than the E_f of all slabs. All LUMO energies of the five gases are far higher than the E_f. Therefore, the possible gas–slab interactions are the interactions between the HOMO states of gas molecules and Zn/O states of the slabs in the valence band. It is beneficial when HOMO energy is higher than E_f, since the energies of antibonding states would be higher than E_f during gas–slab interactions. The higher energy of antibonding states dump electrons at the Fermi level, which keeps the adsorption system stable [51].

Therefore, for $\mathrm{ZnO}\left(000\overline{1}\right)$, no gas is expected to interact with the slab strongly. Still, because the HOMO level of CH₂O, C₂H₅OH, NH₃ and CH₃COCH₃ is close to O 2p states and E_f at the same time, it has the possibility to interact with O 2p of the slab and form antibonding states that are higher than E_f. Due to the closer energy levels and the stronger interactions, weak adsorptions of these gases with smaller adsorption energy of CH₃COCH₃ are expected, which agrees with the calculated adsorption energies, except for the adsorbed and decomposed CH₂O.

For $\mathrm{ZnO}\left(10\overline{1}0\right)$, due to HOMO of CH₂O, C₂H₅OH, NH₃ and CH₃COCH₃ being higher than the E_f of the slab, these four gases can form stable bonds with the slab and are expected to have high adsorption energies. The bigger energy difference between Zn 3d/O 2p and HOMO energy of CH₃COCH₃ limits the adsorption energy. Therefore, strong adsorptions of these gases with smaller adsorption energy of CH₃COCH₃ are expected, which agrees with the calculated adsorption energies, expect for minor differences among CH₂O, C₂H₅OH and NH₃.

Similarly, for $\mathrm{ZnO}\left(10\overline{1}1\right)$, strong adsorptions of CH₂O, C₂H₅OH and NH₃ with smaller adsorption energy of CH₃COCH₃ are expected. The interaction strength should be CH₂O > C₂H₅OH > NH₃, which is slightly different from the adsorption energy of C₂H₅OH > CH₂O > NH₃, and it can be attributed to geometric matching-induced molecule decomposition and conformation relaxation. It is also noted that the calculated results show the H₂ can adsorb and decompose on $\mathrm{ZnO}\left(10\overline{1}1\right)$ and has the highest adsorption energy among all gases. This indicates the HOMO states of H₂ might be able to interact with the O 2p states of the slab and form antibonding states over E_f.

Therefore, the matching analysis of the electronic structure between gases and slabs before interaction is helpful in understanding the adsorption preference, and the geometric matching-induced electrostatic interactions decide the molecule decomposition and conformation relaxation, which finally decides the gas adsorption selectivity.

5. Conclusions

This work studied gas adsorption selectivity on three typical ZnO facets of $\mathrm{ZnO}\left(000\overline{1}\right)$, $\mathrm{ZnO}\left(10\overline{1}0\right)$ and $\mathrm{ZnO}\left(10\overline{1}1\right)$ using the DFT method. The adsorption conformations, adsorption energies and charge transfer features of the gases (CH₂O, H₂, C₂H₅OH, CH₃COCH₃ and NH₃) during adsorption were calculated and systematically analyzed. Based on these, the adsorption selectivity was evaluated and the mechanism beneath was analyzed in terms of polarity, geometric matching and electronic structure matching. It was found that: (1) the $\mathrm{ZnO}\left(000\overline{1}\right)$ has adsorption selectivity to CH₂O; (2) the $\mathrm{ZnO}\left(10\overline{1}0\right)$ has high adsorption energies for C₂H₅OH, NH₃, CH₂O and CH₃COCH₃, with a slight selectivity to C₂H₅OH and NH₃; (3) the $\mathrm{ZnO}\left(10\overline{1}1\right)$ has much better adsorption energies for CH₂O, C₂H₅OH, H₂ and NH₃, with a slight selectivity to H₂. We conclude the selectivity is attributed to the geometric matching of the adsorption sites and surface lattices caused by electrostatic interactions and the electronic structure matching, which allows for the interactions.

However, since the work principle of semiconductor-based gas sensing is “adsorption-transduction”, to better understand the experimental results and guide experimental works, future work will focus on “transduction” and the combination of “adsorption” and “transduction”.