Precise control of surface oxygen vacancies in ZnO nanoparticles for extremely high acetone sensing response

ZnO has been studied intensely for chemical sensors due to its high sensitivity and fast response. Here, we present a simple approach to precisely control oxygen vacancy contents to provide significantly enhanced acetone sensing performance of commercial ZnO nanopowders. A combination of H2O2 treatment and thermal annealing produces optimal surface defects with oxygen vacancies on the ZnO nanoparticles (NPs). The highest response of ∼27,562 was achieved for 10 ppm acetone in 0.125 M H2O2 treated/annealed ZnO NPs at the optimal working temperature of 400 °C, which is significantly higher than that of reported so far in various acetone sensors based on metal oxide semiconductors (MOSs). Furthermore, first-principles calculations indicate that pre-adsorbed O formed on the surface of H2O2 treated ZnO NPs can provide favorable adsorption energy, especially for acetone detection, due to strong bidentate bonding between carbonyl C atom of acetone molecules and pre-adsorbed O on the ZnO surface. Our study demonstrates that controlling surface oxygen vacancies by H2O2 treatment and re-annealing at optimal temperature is an effective method to improve the sensing properties of commercial MOS materials.


Introduction 
Monitoring environmental acetone indoor and outdoor † Jihyun Lee and Youngmoon Choi contributed equally to this work. has been set to 250 ppm considering an 8 h timeweighted average [3]. Additionally, acetone is highly flammable owing to its high volatility even at room temperature [4]. Therefore, monitoring the acetone concentration in the environment is a critical issue. Furthermore, acetone is produced in the human body through metabolic processes and released through exhaled breath. A high concentration of breath acetone can be released due to diabetes [5], fat metabolism [6], and heart failure [7,8]. The breath acetone concentration ranges from 0.3 to 0.9 ppm for healthy people [5,6], but it is over 1.8 ppm for diabetes patients [5,6] and over 40 ppm for a person on a ketogenic diet [6,9]. Therefore, breath acetone can be used as a biomarker to diagnosis-related diseases such as diabetes, obesity, and systolic dysfunction.
The concentration of acetone in air or breath has been commonly measured by several techniques based on gas chromatography and mass spectrometry. These systems precisely measure the concentration; however, they are bulky, complex, time-consuming, and expensive [1]. Additionally, skilled operators are required, and gas samples must be collected before the measurement for the analysis. Resistive gas sensors based on metal oxide semiconductors (MOSs) are considered as an alternative. Among various MOS sensing materials, ZnO has been studied intensively due to its several advantages. It is chemically and thermally stable, and its various forms of nanostructure can be fabricated at a low cost. Moreover, it has favorable sensing properties such as high sensitivity and fast response.
To improve the sensing properties of ZnO, various approaches have been reported including nanostructuring [10][11][12], transition metal doping [13,14], and noble metal loading [15,16]. Additionally, the control of the surface defect [17] has been alternatively proposed because the MOS sensor response is highly dependent on the surface chemistry. The surface defect can be introduced during material synthesis [18,19] or subsequent annealing treatment [20,21]. Especially, oxygen vacancies, among various defects, of nanostructured ZnO play an important role in improving the sensing properties because they act as the electron donors and promote the ionization and chemisorption of oxygen molecules on the ZnO surface [17,18]. Hence, adsorbed oxygen ion species are easily generated on the surface of more oxygen-vacant ZnO, enhancing the sensing response significantly. It has been suggested that the number of oxygen vacancies of ZnO can be increased using a simple method of H 2 O 2 surface treatment and annealing [22,23]. The as-grown ZnO film and ZnO nanorods exhibited slightly enhanced sensing response compared to their untreated counterparts to detect ethanol due to the application of the surface treatment [24]. Therefore, more precisely controlled oxygen vacancy formation in ZnO is expected to be an effective strategy to improve the sensing properties of ZnO as a chemical sensor.
In this study, we found that the acetone sensing properties of commercial ZnO nanopowders were increased unprecedently via two consecutive steps: optimal H 2 O 2 solution treatment and stabilization by thermal annealing. ZnO powder-based sensors which are treated by the optimal H 2 O 2 concentration and high-temperature annealing showed a tremendously high sensing response from more than ~27,000 to 10 ppm acetone, which is significantly higher than that reported so far for the MOS-based various acetone sensors. The sensing mechanism of the greatly enhanced response is described by the effect of surface oxygen vacancies through the analysis of surface chemistry and simulation.

2 Characterization of sensing materials
The crystal structures and morphology of the prepared www.springer.com/journal/40145 powders were characterized by the X-ray diffractometer (Ultima IV, Rigaku) with Cu Kα radiation (1.5406 Å) and the high-resolution transmission electron microscope (JEM-ARM 200F, JEOL), respectively. The chemical states of the sample surface were investigated using the X-ray photoelectron spectroscope (K-alpha, Thermo Scientific Incorp.) with Al Kα source (1486.6 eV). Moreover, the X-ray photoelectron spectroscopy (XPS) results were calibrated by adventitious carbon C 1s peak at 285 eV. Raman spectra were measured by Horiba Jobin Yvon HR800 UV with Ar + laser excitation of 488 nm to understand chemical states further.

3 Fabrication of sensor devices
Interdigitated Pt electrodes were patterned on a SiO 2 /Si substrate (8.5 mm × 8.5 mm) using photolithographic technique after depositing Pt layer (100 nm) using a direct current (DC) magnetic sputtering system to fabricate the sensor devices. The Cr layer (20 nm) was deposited to provide good adhesion Pt layer and the substrate. The as-prepared nanopowders were mixed with an α-terpineol binder; and the mixtures were spread on the interdigitated Pt electrodes. The sensor device was annealed at 600 in a nitrogen environment ℃ for 1 h to remove the binding agent and enhance the stability of the sensor.

4 Gas sensing measurements
The sensing performance of the H 2 O 2 treated and annealed ZnO NPs for acetone was tested with a customized sensing measurement system. The sensor device was mounted on a chamber of tube furnace equipped with a gas flow system connected with mass flow controllers (MFCs) and gas cylinders. Furthermore, the acetone concentration was controlled by adjusting the flow rate between the high concentration of acetone and synthetic air of a balance gas. The adjusted amount of acetone and synthetic air was injected into the senor chamber at 100 sccm of a constant flow rate. The operating temperature of the sensor was controlled using a temperature controller of the tube furnace. The sensing performance of the samples was examined in the temperature range of 300-500 . Moreover, the ℃ electric sensing response of the samples was measured using a nanovoltmeter (Keithley 2182) at 10 nA of a constant current with a time interval of 1 s using a current source (Keithley 6220). The sensing response was defined as ΔR/R g , where ΔR = R a − R g , R a and R g represent the resistance in air and acetone gas, respectively.

5 Computational details
We performed first-principles calculations considering spin-polarized density functional theory using the Vienna ab initio Simulation Package [25,26] and projected augmented wave (PAW) potentials [27,28]. Perdew-Burke-Ernzerhof (PBE) functional using the generalized gradient approximation (GGA) was considered for electronic exchange and correlation [29]. All calculations used a plane-wave basis set with a cutoff of 500 eV. The convergence threshold for total energy calculations was set at 10 −4 eV. All geometry optimizations were conducted using a conjugate gradient algorithm considering the forces acting on each atom and terminated when they converged within 0.03 eV/Å for electronic relaxation. The k-point sampling of the Brillouin zone was used with a 3 × 3 × 1 Monkhorst-Pack grid [30]. We obtained the bulk ZnO parameters: a = b = 3.289 Å and c = 5.307 Å, which agree very well with the previous experimental data (a = b = 3.250 Å and c = 5.207 Å) [31]. The slab models representing the ZnO(1010 ) surface consisted of three ZnO double-layers separated by approximately 18 Å vacuum regions. In all cases, the atoms in the bottom ZnO double layer were fixed for simulating the underlying bulk structure. The calculations for gasphase molecules were performed in a 20 Å × 21 Å × 22 Å supercell.

Results and discussion
To characterize the crystal structure modification of the commercial ZnO NPs by the H 2 O 2 treatment and the subsequent thermal annealing, the X-ray diffraction (XRD) measurements were conducted as shown in Fig. 1 (Fig. 1) [24]. The diffraction peaks of the annealed samples after 1 M H 2 O 2 treatment were well-indexed to the crystal structure of the ZnO phase without any impurities and secondary phases. However, it releases oxygen by thermal decomposition at a temperature above 200 , ℃ and easily converts to a more stable phase, ZnO, at high temperatures (Reaction (2)) [24]. Particularly, oxygen vacancies were formed on the surface of ZnO during its transformation from ZnO 2 to ZnO [24].
The XRD results indicated that the crystal structure of the ZnO NPs changed to cubic ZnO 2 phase during H 2 O 2 treatment, and returned to the hexagonal ZnO phase after annealing at 400 ℃. Hence, the surface of ZnO particles was damaged presumably during the H 2 O 2 treatment and re-annealing step, forming surface defects [24].
We prepared various ZnO NPs that were annealed at 400 after H ℃ 2 O 2 treatments with different molar concentrations (x M) to investigate the effect of H 2 O 2 on the sensing performance of the final ZnO product.
The crystalline phase of these NPs was compared to that of pure ZnO. Figure 2(a) shows the XRD spectra   c)). The intensity of the main peak (*) was compared to that of the pure ZnO NPs. As shown in Fig. 2(d), the peak intensity slightly decreases at 0.125 M compared to that of the pure ZnO; however, it is abruptly decreased at 0.250 M. Additionally, the sharp diffraction peaks broadened from 0.250 M ( Fig. 2(c)), which means a significant decrease in crystallinity at the concentrations of H 2 O 2 over 0.250 M. We found that the particle size decreases for H 2 O 2 concentrations above 0.250 M from the broadening of the diffraction peaks.
We calculated the crystallite size of the as-prepared nanopowders using the Halder-Wagner method as follows: 2 2 cos 16 sin tan sin where θ is the Bragg angle, D represents the crystallite size, β denotes the peak width at half-maximum intensity, K is the crystallite configuration constant, λ is the wavelength of Cu Kα radiation, and ε denotes the weighted average strain [32]. The Halder-Wagner method gives more weight to the diffraction peaks at low and intermediated angle ranges [33]. The estimation results were more accurate than various other crystallite size estimation methods for the spherical metal-oxide NPs, such as CeO 2 and TiO 2 [33,34]. The crystallite size can be determined from the slope of (β/tanθ) 2 vs. β/(tanθ·sinθ) plot [33]. The Halder-Wagner plots for all the prepared samples after adopting the K = 4/3 for the volume-weighted average size of spherical crystallites [33] are shown in Figs. 3(a)-3(e). The calculated crystallite sizes are summarized in Fig. 3   decomposition during annealing at 400 ℃ (Reaction (2)). Therefore, the particle size can be reduced during this transformation from ZnO 2 to ZnO. Highly . It was seen that the small-sized particles of ~10 nm were aggregated into particles with an average size of ~55 nm. The size obtained from the SEM image agreed well with the results obtained from the XRD (Fig. 2) and TEM (Fig. 4) analyses. Therefore, the significant size reduction at high concentrations of H 2 O 2 can be explained by the decomposition of a large amount of 2 2 O  from the excessively oxidized ZnO 2 that was converted to ZnO during the annealing process.
The sensing performance of the pure ZnO NPs was measured after exposing them to acetone at different concentrations (0.01-10 ppm) or at different operating temperatures (300-500 ). Figure 5  (see ℃ the inset of Fig. 5(a)). The response time was estimated using a definition-the time required to reach 90% of the total change in response. Figure 5 According to the sensing results, the 0.125 M H 2 O 2 treated/annealed ZnO NPs exhibited extremely high sensing response for the detection of 10 ppm acetone among the prepared samples. Because the sensing performances are strongly influenced by the chemical states of the metal oxide surface, we performed surface elemental composition analysis by XPS. The chemical states of the sample surface were analyzed by probing the Zn 2p and O 1s states. Figure 6(a) shows the XPS spectra of the Zn 2p state for all samples. All samples show two main peaks corresponding to the electronic states of Zn 2p 1/2 (~1044 eV) and Zn 2p 3/2 (~1021 eV), indicating the existence of a divalent oxidation state of Zn atoms in ZnO crystals. Furthermore, we observed that the Zn 2p 3/2 spectrum peak was shifted to the lower binding energy for 0.075 and 0.125 M treated/annealed ZnO NPs, as seen in Fig. 6(b). Additionally, more shift was observed for the 0.125 M samples. The shift of Zn 2p 3/2 peak to lower binding energy is closely related to the existence of oxygen vacancies in ZnO. The binding energies of Zn 2p 3/2 in ZnO were decreased with an increasing concentration of oxygen vacancies [35].
The existence of oxygen vacancies in the as-prepared ZnO NPs was analyzed by the detailed characterization of the O 1s state in the XPS spectra.   [36]. Among the three different types of oxygen states, the O-deficient state is considered to be directly related to the sensing performance [37]. The number of oxygen vacancies was estimated by calculating the relative percentage of the area fraction of the O-deficient peak. The percentage of oxygen vacancies of the commercial ZnO NPs was ~36% (Fig. 6(c)). Furthermore, oxygen vacancies of the H 2 O 2 treated/annealed ZnO NPs for various concentrations are as follows: ~41% for 0.075 M, ~46% for 0.125 M, ~38 for 0.250 M, and ~37% for 1.000 M. According to the XRD analysis, 0.250 and 1.000 M samples have smaller particles than the rest. Therefore, a larger number of oxygen vacancies is expected due to their higher surface-to-volume ratio.
However, that was not the case (Figs. 6(f) and 6(g)) due to the reduced ZnO phase formation, as revealed in Fig. 2. Particularly, the 0.075 and 0.125 M samples have higher contents of oxygen vacancies than the other two. Specifically, the 0.125 M sample has more vacancies compared with the 0.075 M sample. As described above, the ZnO 2 phase, transformed by H 2 O 2 treatment, is easily converted to the more stable ZnO phase at high annealing temperature; however, the converted ZnO NPs have oxygen vacancies on the surface [24]. According to the XRD analysis (Fig. 2) The analysis of Raman spectra was performed to confirm the existence of oxygen vacancies further. Figure 7 shows the Raman spectra of the commercial ZnO NPs and those of 0.075, 0.125, and 0.250 M H 2 O 2 treated/annealed ZnO NPs. Four Raman modes were observed at 331, 381, 440, and 584 cm −1 , associated with the ZnO wurtzite structure [38,39]. The sharp and strong peak at 440 cm −1 can be assigned to the nonpolar optical phonon E 2 (high) mode [40]. The small peaks at ~331 and 381 cm −1 can be attributed to E 2 (high) − E 2 (low) multiphonon scattering phonon mode and A 1 (TO) transverse optical phonon mode, respectively. The polar phonon peaks centered at 584 cm −1 represent A 1 (LO) longitudinal optical phonon mode, which is attributed to the defects such as oxygen vacancies [41]. The Raman spectra were normalized with respect to the E 2 (high) − E 2 (low) peak for the ease of interpretation (Fig. 8). The ratio between A 1 (LO) and E 2 (high) peak intensities (I(A 1 (LO))/I(E 2 (high)) indicates the number of oxygen vacancies in the sample [38,39]. The contents of oxygen vacancies increase in the 0.125 M  Fig. 7). This trend is consistent with the results obtained from the XPS analysis, as shown in Figs. 6(c)-6(g).
We compared the sensing responses of all the prepared samples upon exposure to 10 ppm acetone at the optimal working temperature of 400 (Fig. 8).  oxygen species on the ZnO surface. We also confirmed that the ZnO surface covered with oxygen species is more preferred for the adsorption of acetone molecules thermodynamically. This result was obtained from computational calculations, and we discuss more about it in the following. Therefore, a higher sensing response can be explained by the increased number of oxygen vacancies. The H 2 O 2 treated/annealed ZnO NPs show a highly improved sensing response than the commercial ZnO NPs, as shown in Fig. 8. In particular, the 0.125 M sample displays the best sensing response. The results reveal that the H 2 O 2 treatment and annealing process can easily produce oxygen vacancies on the surface of ZnO NPs. The optimal concentration of H 2 O 2 for the best acetone sensing response of ZnO NPs was found to be approximately 0.125 M. Additionally, the highest sensing response value of ~27,562 for acetone 10 ppm is remarkably higher than that of previously reported various acetone sensors based on MOSs [42][43][44][45][46][47][48][49][50][51][52][53][54], as shown in Fig. 9. This indicates that controlling the number of oxygen vacancies by H 2 O 2 treatment is a very effective way to improve the sensing performance using simple commercial NPs.
Furthermore, we tested other sensing properties, such as sensitivity for other target gases and repeatability for the detection of acetone, of the 0.125 M H 2 O 2 treated/ annealed ZnO NPs at its optimal working temperature of 400 . Figure 10(a) shows the response of the ℃ sample to various gases, such as acetone, dimethyl methylphosphonate (DMMP), hydrogen sulfide, ethanol, toluene, isoprene, benzene, formaldehyde, methane, hydrogen, CO, and CO 2 , at the same concentration of 10 ppm at 400 . The response to acetone was ℃ remarkably higher than that of the other gases, indicating that the sensor based on the 0.125 M H 2 O 2 treated/annealed ZnO NPs provides a clear distinction during acetone gas detection. Figure 10(b) shows the repeatability of the 0.125 M H 2 O 2 treated/annealed ZnO NPs during sensing measurement for the detection of 10 ppm acetone over 60 cycles at 400 . Furthermore, ℃ Fig. 10(c) shows the magnified plot of the real-time resistance curves marked by the dotted box in Fig. 10(b). Figures 10(b) and 10(c) clearly show that the sensing resistance is approximately constant throughout 60 cycles of continuous detection of 10 ppm acetone. This implies a good repeatability in the response of the 0.125 M H 2 O 2 treated/annealed ZnO NPs for sensing 10 ppm acetone.
We also tested the humidity influence on the sensing response of the 0.125 M H 2 O 2 treated/annealed ZnO NPs to detect 10 ppm acetone under different humidity conditions (0-90 RH%). The response decreased due to the contamination of the sensing active sites by the adsorption of water molecules (Fig. 10(d)). We found that the response of the sensor at a dry condition was significantly degraded by two orders of magnitude at 90 RH%. This indicates that the surface defects of the H 2 O 2 treated/annealed samples provide a higher sensing response under dry environmental conditions; however, they readily lead to higher adsorption of water molecules under high humidity conditions, resulting in greater contamination of the sensing active sites. Consequently, the response is significantly reduced. However, this disadvantage of humidity interference can be overcome using a moisture filter when this material is applied to a breath analyzer [55].
A possible mechanism for the pristine ZnO and oxygen-vacant ZnO NPs is illustrated in Fig. 11 for the detection of acetone based on the characterization results of surface chemistry. The mechanism is based on the change in the electrical resistance of ZnO during adsorption and desorption of gas molecules on the ZnO surface. When the ZnO NPs are under ambient air, the O 2 molecules in the air are adsorbed on the surface of ZnO NPs, and the adsorbed O 2 molecules take the electrons from the conduction band of the ZnO (Fig. 11(a), left). Subsequently, O 2− ions are formed on the ZnO surface at the optimal operating temperature of 400 ℃ in the air ( Fig. 11(a) These reactions show that an electron depletion layer is formed on the surface of ZnO, resulting in high resistance of the sample (Fig. 11(a), left). When the sample is exposed to acetone (CH 3 COCH 3 ) gas, the gas molecules react with the adsorbed oxygen ion species (O 2− ). Many electrons are released back to the ZnO conduction band (Fig. 11(a), right), as shown by Reaction (7) [19]. Subsequently, the thickness of the depletion layer decreases, resulting in a decrease in the resistance (Fig. 11(a), right). These resistance change, before and after the chemical reaction of the target gas molecules with oxygen ion species on the sample surface, corresponds to the sensing response.
The oxygen vacancies act as acceptors of oxygen ion species in the air at high temperatures in the case of oxygen-vacant ZnO NPs through H 2 O 2 treatment and annealing [29]. Many oxygen ion species are adsorbed, and the thicker depletion layer is generated (Fig. 11(b), left). Therefore, the resistance of oxygen-vacant ZnO NPs in the air becomes higher than that of the untreated ones. Then the adsorbed acetone molecules interact with a greater number of oxygen ion species during the sensing reaction, leading to a significant decrease in resistance. Accordingly, the oxygen-vacant sample exhibited a significantly sensitive and improved acetone gas sensing response.
We further performed density functional theory calculations to elucidate the high sensing response of H 2 O 2 treated/annealed ZnO NPs for acetone. The non-polar ZnO(1010 ) surface was used as the slab model because it is not only the most stable surface of ZnO [57,58], but also the dominating surface as verified from our XRD results. It is expected that O 2 can be easily adsorbed on the oxygen-vacant ZnO surface due to O 2 molecules those exist in the air under gas-sensing conditions. Therefore, we calculated the dissociative adsorption energy in which O 2 molecules are dissociated on the oxygen-vacant ZnO surface such that one O atom fills the oxygen vacancy site, and the other O atom is adsorbed at the Zn bridge site (Fig. S1 in the ESM). Hence, the O 2 dissociative adsorption on the oxygen-vacant ZnO surface (−1.49 eV) is significantly more favorable than on the clean ZnO surface (+2.96 eV). Therefore, we considered the oxygen-adsorbed ZnO surface model, reflecting the oxygen adsorption under gas-sensing conditions. We constructed the oxygen full-covered ZnO( 1010 ) surface and calculated the adsorption energy of acetone, ethanol, and methane on the oxygen full-covered ZnO surface to describe the oxygen-adsorbed ZnO surface in gas molecule sensing conditions. The adsorption energies of acetone, ethanol, and methane were calculated using Eq. (8): where ΔE ads denotes the adsorption energy, E molecule+slab represents the total energy of surface and adsorbed molecule, E slab is the total energy of the surface, and E molecule denotes the total energy of an isolated molecule. The optimized adsorption configurations of acetone, ethanol, and methane on the oxygen full-covered ZnO(1010 ) surface are depicted in Fig. 12(a), and their adsorption energies are plotted in Fig. 12(b). The adsorption energies reveal that ethanol (−0.35 eV) and methane (−0.03 eV) can be considered as weak physisorption; however, acetone is most strongly adsorbed on the ZnO surface (−1.47 eV). In the adsorption configuration of acetone, pre-adsorbed oxygen (darkred) of the ZnO surface binds to carbonyl C atom of acetone when carbonyl O atom (red) of acetone binds to Zn atom at the surface. This bidentate adsorption configuration is thermodynamically favorable for acetone on the oxygen full-covered ZnO(1010 ) surface, compared to other weakly physisorbed molecules. Additionally, we calculated the adsorption energies on the ZnO surface with different contents of preadsorbed oxygen-oxygen-vacant surface, clean surface, one oxygen atom-adsorbed surface, and oxygen fullcovered surface-to explain the effect of pre-adsorbed oxygen on the adsorption strength at the ZnO surface. The results of the calculation are presented in Fig. 12(c). The adsorption energies of ethanol on the oxygenvacant ZnO and clean ZnO surfaces (black and orange bars) were −0.73 and −1.00 eV, respectively, which were stronger than those of acetone (−0.45 and −0.58 eV). In this case, ethanol forms not only the bonds between hydroxyl O atom and surface Zn atom but also the hydrogen bonding between hydroxyl H atom and lattice O atom. However, acetone (−1.31 eV) binds more strongly on the ZnO surface in one oxygenadsorbed ZnO surface (blue bar) with bidentate configuration bonding between carbonyl C of acetone and pre-adsorbed O on the ZnO surface. This difference in adsorption configuration between the clean and oxygen pre-adsorbed surfaces of acetone makes it more favorable to be adsorbed on the ZnO surface with pre-adsorbed oxygen. In the case of methane adsorption, all ZnO surfaces show the weak physisorption of methane. Acetone is most favorable to be bound on the ZnO surface with pre-adsorbed oxygen among the three gas molecules due to the adsorption energies on different ZnO surfaces.
The surface of oxygen-vacant ZnO NPs can be covered with the large number of oxygen ion species, as mentioned in the sensing mechanism part of Fig. 11. Therefore, highly enhanced sensing response of 0.125 M H 2 O 2 treated/annealed ZnO NPs than that of bare ZnO NPs (Fig. 8) is attributed to the higher adsorption energy of acetone molecules, especially in the oxygen full-covered ZnO surface due to the bidentate configuration bonding between carbonyl C of acetone and pre-adsorbed O on the ZnO surface, as described by the simulation (Fig. 12(c)). Furthermore, the remarkably high response of the 0.125 M H 2 O 2 treated/annealed ZnO NPs to acetone than other gases, such as ethanol and methane ( Fig. 10(a)), is explained by significantly higher adsorption energies of acetone than those of ethanol and methane ( Fig. 12(b)).

Conclusions
We show that only a simple H 2 O 2 treatment and annealing process can significantly improve the acetone sensing performance of the commercial ZnO NPs by creating abundant oxygen vacancies on the surface of ZnO NPs. The ZnO NPs, annealed after surface treatment with an optimized H 2 O 2 content of 0.125 M, exhibited a significantly higher sensing response of ~27,562 to 10 ppm acetone at the optimal working temperature of 400 compared with the ℃ untreated commercial ZnO NPs. Moreover, this value is remarkably higher than that reported so far for various acetone sensors based on MOSs. The surface defect of oxygen vacancies was verified by analyzing surface chemistry using XPS and Raman spectroscopy. Therefore, the significantly high response of the 0.125 M sample can be attributed to the generation of many oxygen vacancies on the surface of ZnO NPs. Consequently, we developed a highly sensitive acetone sensor based on ZnO with a simple surface treatment. In addition, first-principles calculations indicate that pre-adsorbed O formed on the surface of H 2 O 2 treated ZnO NPs can provide a favorable adsorption energy, especially for acetone detection, due to strong bidentate bonding between carbonyl C atom of acetone molecules and pre-adsorbed O on the ZnO surface. Our study demonstrates that the control of surface oxygen vacancies by H 2 O 2 treatment is an effective method to improve the sensing response significantly. Furthermore, the surface treatment method can be adopted to other commercial MOS NPs without the complicated material synthesis process, reducing the manufacturing time of sensor devices and enabling large-scale fabrication.