Investigation on the Transformation of Absorbed ... - ACS Publications

Jul 7, 2017 - In this work, the transformation of absorbed oxygen on ZnO (101̅0) and its ... Density functional theory (DFT) simulation has also been...
0 downloads 0 Views 6MB Size
Article pubs.acs.org/acssensors

Investigation on the Transformation of Absorbed Oxygen at ZnO {101̅0} Surface Based on a Novel Thermal Pulse Method and Density Functional Theory Simulation Tingqiang Yang,†,§ Yueli Liu,†,§ Wei Jin,† Yiyang Han,† Shuang Yang,† and Wen Chen*,‡ †

State Key Laboratory of Silicate Materials for Architectures, School of Materials Science and Engineering, Wuhan University of Technology, Wuhan 430070, P. R. China ‡ State Key Laboratory of Advanced Technology for Materials Synthesis and Processing, School of Materials Science and Engineering, Wuhan University of Technology, Wuhan 430070, P. R. China

ABSTRACT: Absorbed oxygen plays a key role in gas sensing process of ZnO nanomaterials. In this work, the transformation of absorbed oxygen on ZnO (101̅0) and its effects on gas sensing properties to ethanol are studied by a novel thermal pulse method and density functional theory (DFT) simulation. Thermal pulse results reveal that the absorbed O2 molecule dissociates into two individual oxygen adatoms by extracting electrons from ZnO surface layers when temperature is above 443 K. The temperature at which absorbed O2 molecule begins to dissociate is the lowest working temperature for gas sensing. DFT simulation demonstrates the dissociation process of O2 at ZnO (101̅0) surface, and the activation energy (Ea) of dissociation is calculated to be 351.71 kJ/mol, which suggests that the absorbed O2 molecule is not likely to dissociate at room temperature. The reactions between ethanol and absorbed O2 molecule, as well as reactions between ethanol and O adatom, are also simulated. The results indicate that ethanol cannot react with absorbed O2 molecule, while it can be oxidized by O adatom to acetaldehyde and then to acetic acid spontaneously. Mulliken charge analysis suggests electrons extracted by O adatom return to ZnO after the oxidation of ethanol. KEYWORDS: absorbed oxygen, ZnO, gas sensing, thermal pulse method, DFT

Z

temperature.4 The selective response depends on how many absorbed oxygen species each reducing gas molecule consumes5,6 or the barrier of reaction between absorbed oxygen and reducing gases.7 Therefore, it is absolutely necessary to study the transformation of absorbed oxygen on the ZnO surface. Lin and co-workers demonstrated that absorbed oxygen leads to the formation of build-in field at the surface by surface photovoltage spectroscopy (SPS) and field-induced surface photovoltage spectroscopy (FISPS).8,9 Regrettably, the tests were implemented at room temperature and the transformation of absorbed oxygen at different temperature was not studied. Iwamoto et al. demonstrated oxygen absorption and transformation on various metal oxides by temperature-programmed desorption (TPD) combined with

nO nanomaterials have drawn extensive attention in gas sensor applications due to wide band gap (3.37 eV), high electron mobility, and high response to volatile organic compounds (VOCs).1−3 However, the development of ZnO gas sensing materials has been severely hindered by high working temperature and poor selectivity. To overcome these shortcomings, the key factors for working temperature and selectivity as well as gas sensing mechanism must be revealed. It has been generally accepted that absorbed oxygen plays a key role in gas sensing process. When ZnO nanomaterials are exposed to air, oxygen molecules will absorb on the ZnO surface and extract electrons from surface layers, leading to the formation of depletion layer and enhancement of resistance. Reducing gas molecules can react with absorbed oxygen after being injected; afterward, electrons will return to the ZnO surface and resistance will decrease. The working temperature is restricted by the reactivity of absorbed oxygen, which is believed to be O2−, O−, and O2− in different ranges of © 2017 American Chemical Society

Received: May 31, 2017 Accepted: July 7, 2017 Published: July 7, 2017 1051

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors electron spin resonance (ESR).10 Nevertheless, the research did not reveal the effect of oxygen transformation on the resistance of ZnO materials. Furthermore, both of these studies were implemented in some vacuum, but gas sensing properties are tested in ambient atmosphere. It will have more practical significance to study oxygen transformation on ZnO surface under ambient environment. Temperature-dependent conductivity measurement is a common method to study defects and surface reactions under normal environment.11 However, surface reaction is usually slow, especially at relatively low temperature; as a result, the effects of surface reaction on conductivity might be concealed by the incessantly increasing temperature. Density functional theory (DFT) simulation has also been widely applied to study the gas sensing mechanism of ZnO surface.1,12−15 Yuan et al. simulated the absorptions of H2, NH3, CO, and ethanol on ZnO. Band structures of ZnO surface slab as well as charge transfers between ZnO surface and absorbates have also been calculated. The outcomes suggested that surface reconstruction and charge transfer result in gas sensing response.12 Hadipour et al. calculated highest occupied molecular orbitals (HOMOs) and lowest unoccupied molecular orbitals (LUMOs) of Al doped ZnO nanoclusters and CO molecules, and proposed that the relative locations of HOMOs and LOMOs play key roles in CO gas sensing.13 However, these works did not take absorbed oxygen into consideration. Korir et al. calculated absorption energy of O2 and ethanol, and proposed that gas sensing response should be attributed to stronger absorption ability of ethanol than that of O2 molecules.14 Whereas transformation of absorbed oxygen and reaction between absorbed oxygen and ethanol were not considered. In this work, a novel thermal pulse method is initially proposed to study the resistance change of ZnO nanorod arrays in ambient atmosphere. By analyzing the variation of stable resistance under thermal pulse at different temperatures, the transformation of absorbed oxygen can be revealed. The factors for the lowest and optimum working temperature for gas sensing can also be understood. To analyze the dissociation process of the absorbed O2 molecule more specifically, DFT simulation is executed. Activation energy of O2 dissociation, reaction energy of ethanol oxidation, and charge transfers to different absorbates are quantitatively calculated. Based on experimental and computational results, the gas sensing mechanism is revealed.



Scheme 1. Schematic Diagrams of (a) ZnO Nanorod Arrays Gas Sensor Device and (b) Test System of WS-30A

measurement system (WS-30A, Zhengzhou Winsen Technology Corp., China). A certain load resistance is in series circuit of the gas sensor under an overall testing voltage in the system. Scheme 1b shows the schematic diagram of WS-30A. During the test we can only observe the variation of output voltage, and the resistance of ZnO can be calculated by following eq 1: Uoverall − Uoutput R ZnO = R load Uoutput

(1)

The response is defined as Rair/Rgas, where Rair and Rgas are resistances of ZnO nanorod arrays in air and in reducing gas, respectively. As shown in Scheme 1a, a Ni−Cr alloy coil inside the ceramic tube can heat the device to a particular temperature under a particular voltage immediately. Thermal pulse is applying a voltage instantaneously on the alloy coil to heat the ZnO to a temperature promptly. The variation of output voltage under the thermal pulse can be obtained. This is called thermal pulse method. The heating voltage was maintained until the output voltage became steady, and it was 60 min before the next thermal pulse. We performed periodic density functional calculations using the Dmol3 4.4 program.17,18 The geometry optimization and transition state (TS) search were performed using spin unrestricted DFT methods with the general gradient approximation (GGA) in the form of a Perdew−Burke−Ernzerhof (PBE) functional.19 The doublenumeric quality basis set with polarization functions (DNP) was used. The inner electrons of Zn atoms were kept frozen and replaced by an effective core potential (ECP), and other atoms in this study were treated with an all-electron calculation. Brillouin-zone integrations were performed using a 2 × 2 × 1 Monkhorst−Pack grid. A Fermi smearing of 0.005 hartree was used to accelerate convergence, and a real-space cutoff was 4.5 Å to improve the computational performance. The tolerances of energy, force, and displacement convergence were 1 × 10−5 hartree, 2 × 10−3 hartree/Å, and 5 × 10−3 Å, respectively. Several possible configurations of different absorbates on ZnO(101̅0) were established for optimization, and all of the optimized atom structures presented in this paper are the most stable configurations. TS search was performed by complete linear synchronous transit/ quadratic synchronous transit (LST/QST) method.20 The root-meansquare (RMS) convergence was 0.005 hartree/Å.

EXPERIMENTAL AND COMPUTATIONAL DETAILS

ZnO nanorod arrays were synthesized by a low temperature aqueous solution growth method followed by annealing the substrates with deposited nanostructures at 673 K for 30 min.16 ZnO nanorod arrays film was separated from the substrate after annealing. The exfoliated ZnO nanorod arrays film was put into Zn(Ac)2·2H2O solution with equal moles of Zn(Ac)2 ·2H2O and ethylamine dissolved in methoxyethonol. The ZnO nanorod arrays film was flexible and had a good tendency to be attached on the Al2O3 ceramic tube with Au interdigital electrodes when the ceramic tube was dipped in the solution. Finally, the Al2O3 tube was annealed at 673 K for 3 h to remove the organics and stabilize the gas sensor device. The schematic diagram of the device is shown in Scheme 1a. The crystal structures and microstructures of the as-prepared samples were characterized by X-ray diffraction (XRD, Pert-Pro, PANalytical, Netherlands) and field-emission scanning electron microscopy (FESEM, S-4800, Hitachi, Japan). The gas sensing properties were measured by using a commercial gas sensing 1052

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

Figure 1. Characterization of ZnO nanorod arrays: (a) XRD patterns and FESEM images of (b) top view and (c) section view.



RESULTS AND DISCUSSION Characterization of ZnO Nanorod Arrays. The XRD pattern of as-prepared sample is shown in Figure 1a, and the diffraction peaks can be indexed to hexagonal wurtzite ZnO. The peak intensity of (002) plane is far higher than those of (100) and (101) planes, revealing the preferentially oriented growth along the c-axis. It is also verified by FESEM images shown in Figure 1b, c. The FESEM images indicate the nanorod arrays with diameter and length of ∼100 nm and ∼1.5 μm, respectively. The nanorods are well vertically aligned and uniformly distributed. Particularly, the cross sections of nanorods are hexagons. Accordingly, the side sections of nanorods belong to {101̅0} crystal planes which take a dominant part of the ZnO surface of nanorod arrays. Thermal Pulse Method. It is widely accepted that O2− will transform to O− by extracting one electron from ZnO surface layers when the temperature surpasses a particular value. The process can be expressed as the following equation. k1, k −1

O2− + e ←⎯⎯→ 2O−

According to the value of output voltage, the resistance of ZnO nanorod arrays can be obtained by eq 1. Figure 2b shows the variation of stable resistances after thermal pulses of different temperatures. It shows that the stable resistance decreases to a minimum value when temperature is below 443 K. Afterward, the resistance increases to a peak at 553 K followed by a second decline. Due to the depletion layer resulting from absorbed oxygen at ZnO {101̅0} surfaces, the conductivity model is equivalent to double-Schottky barrier model.21 The resistance is mainly dependent on the density of electrons [e] in depletion layers, and [e] can be expressed as eq 3 derived from Poisson’s equation. ⎛ qV ⎞ [e] = [e]0 exp⎜ − S ⎟ ⎝ kBT ⎠

(3)

[e]0 is the initial density of electrons at the depletion layer before oxygen absorption, q is elemental charge of electron, and VS is the electric potential at surface. kB and T are Boltzmann constant and temperature, respectively. As the conductivity is appropriately proportional to electron concentration due to the minute change of electron mobility, the correlation between resistance (R) and temperature (T) can be expressed as the following:

(2)

Here k1 and k−1 are the rate constants of forward and reverse reactions. If a thermal pulse is imposed on ZnO nanorod arrays at relatively low temperature which cannot stimulate the dissociation of O2−, the resistance will definitely decrease due to the thermal excitation of charge carriers. In contrast, the resistance will vary by a more complex curve if the thermal pulse is at higher temperature which can cause O2− to dissociate. Figure 2a shows the variation of output voltages upon thermal pulses. When the temperature is below 443 K, the output voltage keeps increasing to a stable value after a thermal pulse. In contrast, when temperature is above 443 K, the output voltage decreases after a dramatic enhancement, and the decrease rate speeds up with improvement of temperature.

⎛ qV ⎞ 1/R ∝ [e]0 exp⎜ − S ⎟ ⎝ kBT ⎠

(4)

Then we calculate their natural logarithms to obtain the following equation: ln(1/R ) ∝ − 1053

qVS 1 + ln[e]0 kB T

(5) DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

Figure 2. (a) Variation of output voltage under different thermal pulse from 303 to 683 K. (b) Curve of stable resistance after thermal pulse. (c, d) Experimental data (dots) and fitting function (line) between stable resistance and temperature in the range of 303−443 K and 463−519 K, respectively.

Figure 2c is the experimental data (dots) and fitting function (line) between the natural logarithms of stable resistances and reciprocals of temperatures. It is clear that the values of ln(1/R) and 1/T have negatively linear correlation, and according to the slope we obtain that Schottky barrier qVS is 0.17 eV. The reduction of stable resistance after thermal pulse below 443 K can be attributed to the temperature-induced enhancement of election concentration in the depletion layer. The resistances at temperatures above 443 K are the interplay of thermal excitation and oxygen transformation; thus, the reaction of eq 2 has to be taken into consideration. The dynamic process of the reaction is largely dependent on an activation energy (Ea) and temperature. The correlation between reaction rates (k1 and k−1) and temperature obeys the rules of eq 6 which successfully explains the dynamic process of absorbed oxygen on TiO2 (110) surface.22 ⎛ E ⎞ k1 = k10 exp⎜ − a1 ⎟ ⎝ kBT ⎠ ⎛ E ⎞ k −1 = k −10 exp⎜ − a − 1 ⎟ ⎝ kBT ⎠

ref 10, the equilibrium constant K of eq 2 can be expressed by following rule: K = [O−]2 /[O2−][e] = k1/k −1

(7)

The change of electron concentration is defined as Δ[e]. Hence, (2Δ[e])2 ([O2 ]initial − Δ[e])([e]initial − Δ[e]) ⎛ E − Ea − 1 ⎞ k = 10 exp⎜ − a1 ⎟ k −10 kBT ⎝ ⎠ −

(8)

Here [O2−]initial and [e]initial are initial concentrations of O2− at surface and electron in depletion layer, respectively. Actually, [e]initial is same with [e] in eq 3. (Ea1 − Ea−1) is the reaction energy Ereact of eq 2, and the value of (Ea1 − Ea−1)/kBT is large. Hence, eq 8 can be simplified to

(6a)

⎛ E ⎞ (2Δ[e])2 ∝ [O2−]initial [e]initial exp⎜ − react ⎟ ⎝ kBT ⎠

(6b)

Here k10 and k−10 are prefactors. Ea1 and Ea−1 are activation energies of forward and reverse reactions of eq 2. According to

(9)

According to eqs 3 and 9, the change of electron concentration after the oxygen transformation must obey the relation: 1054

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors 1/2 ⎛ ⎛ Ereact ⎞⎞ ⎜ ⎟ Δ[e] ∝ ⎜[e]exp⎜ − ⎟⎟ ⎝ kBT ⎠⎠ ⎝

⎛ qV + Ereact ⎞ ∝ [e]01/2 exp⎜ − S ⎟ 2kBT ⎠ ⎝

improvement of output voltage takes place, demonstrating that absorbed oxygen has been largely consumed. In the next step, nitrogen gas is injected for 15 min to sweep the remaining ethanol molecule, and the thermal pulse is applied again. It can be shown that there is little decrease of output voltage compared with the first thermal pulse. Therefore, the resistance enhancement of ZnO nanorod arrays above 443 K is definitely related to the oxygen transformation. To probe the relationship between the form of absorbed oxygen and the working temperature, the response variation in the range of 423−693 K has been measured as shown in Figure 4a. It can be seen that the lowest working temperature is 443 K at which absorbed O2 molecule starts to transform to O adatoms. Hence, the lowest working temperature is the temperature when O adatoms begin to form. As for the optimum temperature which is 673 K, it is determined by variation of Rair and Rgas. Figure 4b is variation of Rair and Rgas in the range of 423−693 K. Even though Rair in Figure 4b is somehow different from that in Figure 2b mainly due to the intrinsic instability of ZnO nanamaterials, the overall trend of resistance variation is nearly identical. Rair reaches the maximum value at 643 K due to interplay of thermal excitation, oxygen transformation, and oxygen desorption. In comparison, Rgas decreases continuously due to the enhanced consumption of absorbed oxygen by improvement of temperature. The response is the ratio of Rair and Rgas; thus, adverse change of Rair and Rgas from 643 to 693 K leads to the optimum temperature being 673 K. DFT Simulation. According to the results of XRD and FESEM, ZnO nanorod arrays are dominantly exposed with {101̅0} crystal planes. Therefore, we established ZnO (101̅0) surface to carry out the DFT simulation. The ZnO (101̅0) surface is modeled using a six-slab p(5 × 3) supercell with a 20 Å vacuum region. In all calculations, two layers of the slab in the bottom were fixed at their bulklike position, whereas the remaining atoms in the top four layers, as well as the adsorbed molecules or atoms, were allowed to relax. The adsorption energy Eads, reaction energy Ereact and activation energy (reaction barrier) Ea are defined as

(10)

Defining RT is the practical resistance at high temperature and R0 is the resistance without taking eq 2 into consideration. Then, we can attain the following relation: ⎛ qV + Ereact ⎞ 1 1 − ∝ Δ[e] ∝ [e]01/2 exp⎜ − S ⎟ R0 RT 2kBT ⎠ ⎝

(11)

Calculating their natural logarithms: ⎛ 1 qV + Ereact 1 1 ⎞ 1 ln⎜ − + ln[e]0 ⎟∝− S RT ⎠ 2kB T 2 ⎝ R0

(12)

The R0 value can be obtained by extending the line of Figure 2c to higher temperature. Figure 2d shows that the values of

(

ln

1 R0



1 RT

) and 1/T have negatively linear correlation up to

519 K. It can be seen from Figure 2d that the slope at higher temperature tends to be less steep, which is attributed to obvious oxygen species desorption peak starting from around 519 K.10 The electrons extracted by absorbed oxygen is less than the expected at higher temperature. According to the slope and qVS obtained before, Ereact is estimated to be 50.80 kJ/mol. The increase of stable resistance after thermal pulse above 443 K mainly results from the transformation of absorbed oxygen. When the temperature increases continuously, resistance variation is much more complicated. The absorbed oxygen has a higher tendency to desorb,10 and even lattice oxygen may transform to oxygen molecule. It is also restricted by WS-30A, the gas sensing test system, which cannot impose a high voltage on the alloy coil for a long period. Therefore, the resistance variation above 553 K is not deeply discussed. To make sure that the decrease of output voltage after a rapid increase is attributed to oxygen transformation, thermal pulse is also applied in nitrogen atmosphere. Figure 3 reveals that the output voltage increases dramatically and then decreases under a thermal pulse at 513 K in air. Whereafter, 104 ppm ethanol is injected into the sealed chamber to consume the absorbed oxygen, resulting in a huge rise of output voltage. Then, another 104 ppm ethanol is poured into the chamber and little

Eads = EZnO(10 1̅ 0) + gas − EZnO(10 1̅ 0) − Egas

(13a)

Ereact = Eproduct − Ereactant

(13b)

Ea = E TS − Ereactant

(13c)

Accordingly, a negative value of Eads or Ereact corresponds to an exothermic adsorption or reaction. First, the dissociation of O2 on ZnO surface was simulated. The ZnO (101̅0) is optimized to stable configuration. In sequence, O2 molecule is added ∼3.5 Å apart from surface to establish the model of O2 absorbing on ZnO (101̅0), i.e., ZnO (1010̅ )−O2. Figure 5 shows top views (a, b) and front views (c, d) of ZnO (101̅0)−O2 surface structure before and after optimization. The optimized structure is consistent with other works.23,24 The Eads of O2 is 9.86 kJ/mol, suggesting that the process of O2 absorption is slightly endothermic. The O−O bonds elongate from 1.233 to 1.306 Å. Table 1 lists the charge transfer to absorbates of different optimized atom structures by Mulliken charge analysis; it is shown that 0.335q (q is elemental charge of electron) transfers from ZnO (1010̅ ) slab to O2 molecule. The O−O bond extension and charge transfer suggest that O2 has a tendency to dissociate. Figure 6a shows the dissociation processes of O2 on ZnO (101̅0). At first step,

Figure 3. Variation of output voltage under thermal pulse in air and nitrogen. 1055

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

Figure 4. (a) Responses of ZnO nanorod arrays to 100 ppm ethanol at different temperatures. (b) Rair and Rgas of ZnO nanorod arrays at different temperatures.

Figure 5. (a, b) Top views and (c, d) front views of ZnO (1010̅ ) with O2 before and after optimization, respectively. The atoms and bonds of top two layers and absorbates are in the style of ball and stick, whereas the remaining are in the style of line.

pulse method. It is ascribed to three factors. First, the temperature of DFT simulation is 0 K at which little electron jumps to the conduction band, making the highly electrophilic O adatom less stable. Second, not taking oxygen vacancies in the bulk of ZnO nanorods into consideration also leads to underestimation of electron concentration in surface slab. Third, the configuration of 2O(a) is not stable. It is more likely to transform to 2O(b) with the O adatom migrating along the [121̅ 0] direction, which eliminates the probability of adverse reaction in eq 2.23 The Ea of O adatom migration is 63.53 kJ/ mol which is much lower than that of the first step. The Ereact is −22.94 kJ/mol, suggesting 2O(b) is more stable. Due to the scope of ZnO (101̅0) supercell, we cannot simulate further separation of O adatoms. Alternatively, we calculate the energy of ZnO (101̅0) with only one O adatom (ZnO (101̅0)-O). Figure 6b and c shows the top view and front view of ZnO (101̅0)−O before and after optimization. As for ZnO (101̅0)−

Table 1. Charge Transfer to Absorbates of Different Optimized Atom Structures by Milliken Charge Analysis atom structure ZnO ZnO ZnO ZnO ZnO ZnO ZnO

(101̅0)−O2 (101̅0)−2O(a) (101̅0)−2O(b) (101̅0)−O (1010̅ )−O2−C2H5OH (101̅0)−O−C2H5OH (101̅0)−O−CH3CHO

charge transfer to absorbates 0.335q 1.052q 1.070q 0.547q 0.385q 0.006q 0.171q

the absorbed O2 molecule dissociates into two O adatoms connecting with three adjacent Zn atoms. In this process, the Ea is 351.71 kJ/mol, which is so high that dissociation of O2 is not likely to take place at room temperature. The Ereact is 258.84 kJ/ mol, which is much higher than that obtained from thermal 1056

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

Figure 6. (a) Dissociation process of O2 molecule on ZnO (1010̅ ). (b) Top view and (c) front view of ZnO(1010̅ ) with one O adatom after optimization. The atom structures in panel (a) is in the region of the black rectangle in panel (b), and the O adatom is in the blue circle in panel (b).

Figure 7. (a, c) Top views and (b, d) side views of ethanol and O2 on ZnO (1010̅ ) before and after optimization, respectively.

O, Eads of O2 is −233.84 kJ/mol, which suggests that ZnO (1010̅ )−O is stable. Mulliken charge analysis shows that 1.052q and 1.070q transfer to O adatoms in 2O(a) and 2O(b) configuration, respectively. 0.547q transfers to the only O adatom in ZnO (101̅0)−O, while lattice oxygen possesses negative charge about 0.8q. Hence, the O adatoms can extract much more electrons from ZnO surface than absorbed O2 molecules. According to the computational results, transformation of absorbed oxygen on ZnO (101̅0) is verified. O2 molecule absorbs on ZnO (101̅0) and extracts a few electrons without dissociation. If a thermal pulse above 443 K is imposed, absorbed O2 molecules begin to dissociate with extraction of much more electrons from surface layers. Under the interaction between thermal excitation and oxygen dissociation, the density of electrons in the depletion layers decreases and the resistance of ZnO increases. Second, reactions between ethanol molecule and two types of absorbed oxygen were simulated. Figure 7 shows configurations of ethanol on ZnO (101̅ 0 )−O 2 (ZnO (101̅0)−O2−C2H5OH) before and after optimization. It can be seen that H−O bond of ethanol elongates from 0.972 to 0.984 Å and O−O bond elongates from 1.233 to 1.325 Å.

Mulliken charge analysis suggests that O2 molecule obtains 0.385q from ZnO surface layers and 0.073q from ethanol molecule. As is mentioned above, it is 0.335q that transfers from ZnO to sole O2 molecule. Hence, ethanol absorption cannot enhance conductivity of ZnO nanorod arrays, and O−O bond elongates more severely. Comparatively, the reaction between ethanol and O adatom is completely different. Figure 8 shows configurations of ethanol on ZnO (1010̅ )−O (ZnO (101̅0)−O−C2H5OH) before and after optimization. Ethanol dehydrogenates into acetaldehyde with one hydrogen bonding to the O adatom and the other bonding to three-coordinated lattice oxygen at surface. The reaction is spontaneous with −454.98 kJ/mol reaction energy. Mulliken charge analysis indicates that only 0.006q transfers from ZnO to absorbates, which means that electrons extracted by O adatom mostly return to ZnO depletion layer. Therefore, this process can reduce the resistance of ZnO nanorod arrays. With existence of hydrogen and hydroxyl, the Eads of acetaldehyde molecule is −100.14 kJ/mol, which suggests that acetaldehyde molecule still has high tendency to absorb on ZnO (101̅0) surface and can be continuously oxidized. Figure 9 shows configurations of acetaldehyde on ZnO (101̅ 0 )−O (ZnO (101̅ 0 )−O− 1057

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

Figure 8. (a, c) Top views and (b, d) side views of ethanol and O adatom on ZnO (101̅0) before and after optimization, respectively.

Figure 9. (a, c) Top views and (b, d) side views of acetaldehyde and O adatom on ZnO (1010̅ ) before and after optimization, respectively.

According to computational results, at room temperature, the absorbed O2 molecule on ZnO (101̅0) cannot oxidize the ethanol molecule, and the charge transfer from ZnO to absorbates changes little after absorption of ethanol molecule. In contrast, when the temperature is high enough, the absorbed O2 molecule dissociates into two O adatoms which can oxidize ethanol into acetaldehyde and then into acetic acid. The

CH3CHO) before and after optimization. Acetaldehyde is oxidized into acetic acid, which is also spontaneous with −576.17 kJ/mol reaction energy larger than that of ethanol oxidation. Hence, ethanol is likely to be oxidized into acetic acid. Mulliken charge analysis indicates that 0.171q transfers to acetic acid, which also means that electrons return to ZnO surface in the oxidation process of acetaldehyde. 1058

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059

Article

ACS Sensors

(4) Hongsith, N.; Wongrat, E.; Kerdcharoen, T.; Choopun, S. Sensor Response Formula for Sensor Based on ZnO Nanostructures. Sens. Actuators, B 2010, 144, 67−72. (5) Kaneti, Y. V.; Yue, J.; Jiang, X.; Yu, A. Controllable Synthesis of ZnO Nanoflakes with Exposed (101̅0) for Enhanced Gas Sensing Performance. J. Phys. Chem. C 2013, 117, 13153−13162. (6) Liao, L.; Lu, H. B.; Li, J. C.; He, H.; Wang, D. F.; Fu, D. J.; Liu, C.; Zhang, W. F. Size Dependence of Gas Sensitivity of ZnO Nanorods. J. Phys. Chem. C 2007, 111, 1900−1903. (7) Arshak, K.; Gaidan, I. Development of a Novel Gas Sensor Based on Oxide Thick Films. Mater. Sci. Eng., B 2005, 118, 44−49. (8) Lin, Y.; Wang, D.; Zhao, Q.; Yang, M.; Zhang, Q. A Study of Quantum Confinement Properties of Photogenerated Charges in ZnO Nanoparticles by Surface Photovoltage Spectroscopy. J. Phys. Chem. B 2004, 108, 3202−3206. (9) Lin, Y.; Wang, D.; Zhao, Q.; Li, Z.; Ma, Y.; Yang, M. Influence of Adsorbed Oxygen on the Surface Photovoltage and Photoluminescence of ZnO Nanorods. Nanotechnology 2006, 17, 2110−2115. (10) Iwamoto, M.; Yoda, Y.; Yamazoe, N.; Seiyama, T. Study of Metal Oxide Catalysts by Temperature Programmed Desorption. 4. Oxygen Adsorption on Various Metal Oxides. J. Phys. Chem. 1978, 82, 2564−2570. (11) Zhao, J.; Xie, C.; Yang, L.; Zhang, S.; Zhang, G.; Cai, Z. Enhanced Gas Sensing Performance of Li-Doped ZnO Nanoparticle Film by the Synergistic Effect of Oxygen Interstitials and Oxygen Vacancies. Appl. Surf. Sci. 2015, 330, 126−133. (12) Yuan, Q.; Zhao, Y. P.; Li, L.; Wang, T. Ab Initio Study of ZnOBased Gas-Sensing Mechanisms: Surface Reconstruction and Charge Transfer. J. Phys. Chem. C 2009, 113, 6107−6113. (13) Hadipour, N. L.; Ahmadi, P. A.; Soleymanabadi, H. Theoretical Study on the Al-Doped ZnO Nanoclusters for CO Chemical Sensors. J. Phys. Chem. C 2015, 119, 6398−6404. (14) Korir, K. K.; Catellani, A.; Cicero, G. Ethanol Gas Sensing Mechanism in ZnO Nanowires: An ab Initio Study. J. Phys. Chem. C 2014, 118, 24533−24537. (15) Bai, S.; Guo, T.; Zhao, Y.; Luo, R.; Li, D.; Chen, A.; Liu, C. C. Mechanism Enhancing Gas Sensing and First-Principle Calculations of Al-Doped ZnO Nanostructures. J. Mater. Chem. A 2013, 1, 11335. (16) Ranjith, S.; Pandian, R.; McGlynn, E.; Rajendra Kumar, R. T. Alignment, Morphology and Defect Control of Vertically Aligned ZnO Nanorod Array: Competition between “Surfactant” and “Stabilizer” Roles of the Amine Species and Its Photocatalytic Properties. Cryst. Growth Des. 2014, 14, 2873−2879. (17) Delley, B. An All-Electron Numerical Method for Solving the Local Density Functional for Polyatomic Molecules. J. Chem. Phys. 1990, 92, 508−517. (18) Delley, B. From Molecules to Solids with the DMol3 Approach. J. Chem. Phys. 2000, 113, 7756−7764. (19) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868. (20) Halgren, T. A.; Lipscomb, W. N. The Synchronous-Transit Method for Determining Reaction Pathways and Locating Molecular Transition States. Chem. Phys. Lett. 1977, 49, 225−232. (21) Yamazoe, N.; Shimanoe, K. Theory of Power Laws for Semiconductor Gas Sensors. Sens. Actuators, B 2008, 128, 566−573. (22) Wahlström, E.; Vestergaard, E. K.; Schaub, R.; Rønnau, A.; Vestergaard, M.; Lægsgaard, E.; Stensgaard, I.; Besenbacher, F. Electron Transfer-Induced Dynamics of Oxygen Molecules on the TiO2 (110). Science 2004, 303, 511−513. (23) Yan, Y.; Al-Jassim, M.; Wei, S.-H. Oxygen-Vacancy Mediated Adsorption and Reactions of Molecular Oxygen on the ZnO (101̅0) Surface. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 72, 161307. (24) Ling, L.; Han, P.; Wang, B.; Zhang, R. DFT Study on the Regeneration Mechanism of ZnO Surface during the Desulfurization of H2S. Fuel Process. Technol. 2013, 109, 49−56.

processes return electrons extracted by O adatoms to the ZnO surface.



CONCLUSION Thermal pulse method is initially applied to study transformation of absorbed oxygen. Under thermal pulse below 443 K, the output voltage increases until it reaches a stable value. While a decline appears following a dramatic improvement under thermal pulse above 443 K. By analyzing the stable resistance after thermal pulse, it is demonstrated that the enhancement of resistance above 443 K is due to absorbed O2 molecules extracting electron from ZnO, then transforming to two O adatoms. The lowest working temperature is the initial temperature of O2 dissociation, and the optimum temperature is determined by variation of Rair and Rgas. The dissociation process of O2 and reaction of ethanol with absorbed oxygen are simulated by the DFT calculation. The O2 molecule absorbs on the ZnO (101̅0) surface with Eads 9.86 kJ/mol, and 0.335q of charge transfers from the ZnO surface to the molecule. The absorbed O2 molecule dissociates into two O adatoms when temperature is high enough. After O adatoms migrate along [12̅10] to a stable state, 0.547q transfers to the only O adatom; thus, the density of electrons in the depletion layer decreases severely. Although ethanol gases cannot react with absorbed O2 molecules, they can transform to acetaldehyde and then to acetic acid by consuming two O adatoms, which makes electrons return to the ZnO depletion layer and reduces the resistance of ZnO nanorod arrays.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Yueli Liu: 0000-0001-5960-8166 Author Contributions §

T.Y. and Y.L. contributed equally to this study and share first authorship. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work is supported by the National Nature Science Foundation of China (No. 11674258, 51506155), Science and Technology Support Program of Hubei Province (No. 2014BAA096), National Nature Science Foundation of Hubei Province (No. 2014CFB165, 2015CKC898), and Wuhan Science and Technology Project (No. 2016010101010020). Thanks to Center for Materials Research and Analysis at Wuhan University of Technology (WUT) for the measurements support.



REFERENCES

(1) Xu, J.; Xue, Z.; Qin, N.; Cheng, Z.; Xiang, Q. The Crystal FacetDependent Gas Sensing Properties of ZnO Nanosheets: Experimental and Computational Study. Sens. Actuators, B 2017, 242, 148−157. (2) Wen, Z.; Zhu, L.; Zhang, Z.; Ye, Z. Fabrication of Gas Sensor Based on Mesoporous Rhombus-Shaped ZnO Rod Arrays. Sens. Actuators, B 2015, 208, 112−121. (3) Wang, Z.; Xue, J.; Han, D.; Gu, F. Controllable Defect Redistribution of ZnO Nanopyramids with Exposed {101̅0} Facets for Enhanced Gas Sensing Performance. ACS Appl. Mater. Interfaces 2015, 7, 308−317. 1059

DOI: 10.1021/acssensors.7b00363 ACS Sens. 2017, 2, 1051−1059