Improving gas sensing performance by oxygen vacancies in sub-stoichiometric WO_3−x

Abstract

Sub-stoichiometric WO_3−x has provided an alternative platform to investigate oxygen vacancies in gas sensors based on metal-oxides. We present an experimental study on the influence of sub-stoichiometric WO_3−x phase upon gas sensing performance. High-quality WO_3−x nanostructures with several x values (WO₃, W₁₉O₅₅, W₅O₁₄, W₁₈O₄₉) were synthesized and used to fabricate H₂S gas sensors. Temperature programmed desorption of oxygen (O₂-TPD) shows that oxygen absorption behaviors of the as-prepared WO_3−x nanostructures are affected by oxygen vacancies, which played a critical role in the detection of H₂S at varying temperature range. We find that oxygen vacancies in sub-stoichiometric WO_3−x facilitate the ionosorption process and in turn enhance the performance of the gas sensor.

---

Introduction

According to the “ionosorption model”, ionosorbed oxygen species (i.e. O₂⁻, O⁻, O²⁻) on semiconductor surfaces determine the gas sensing behavior.^1–3 So far, considerable efforts have been applied to facilitate the oxygen ionosorption for enhancing the gas sensing performance and various strategies are adopted to manipulate oxygen absorption, such as changing morphologies,^4–6 changing sizes and dimensions of micro or nano materials,^7–9 and using catalysts^10,11 or UV-light irradiation.^12,13 Recently, oxygen vacancies have been found to play a critical role in influencing the gas sensing process and determining the performance of gas sensors. For example, density functional theory (DFT) has been successfully employed to study the effect of oxygen deficiency on the electronic and structural properties of sub-stoichiometric tungsten oxides (WO_3−x).^14–17 F. Wang, et al., reported the computational results that revealed the strong dependence of WO_3−x electronic properties on oxygen vacancies.¹⁵ J. Song, et al., simulated the kinetic and thermodynamic processes of H₂ adsorption, activation, and dissociation on WO_2.72 surface.¹⁶ High sensitive NO₂ gas sensors based on W₁₈O₄₉ nanowires have been fabricated and the improvement of the sensing performance are due to oxygen vacancies in W₁₈O₄₉.¹⁸ Monoclinic W₁₈O₄₉ is a well-studied sub-stoichiometric phase and can easily be isolated in a pure form.^19–21 In fact, sub-stoichiometric WO_3−x materials have provided a powerful platform to investigate the gas sensing behaviour associated with oxygen vacancies in metal-oxides based gas sensors. However, it is a challenging task to control oxygen vacancies in sub-stoichiometric WO_3−x accurately.

In this work, we synthesised several sub-stoichiometric WO_3−x, including W₁₉O₅₅, W₅O₁₄ and W₁₈O₄₉, and fabricated gas sensors using the materials. We systematically studied the role of oxygen vacancies in sensing properties of the gas sensors. It was found that oxygen vacancies affected oxygen absorption on the surface of the WO_3−x nanostructures, and facilitated the ionosorption process, favourable for improving device performance and lowering operating temperature. Manipulation of oxygen vacancies in metal oxides brings us an alternative way to improve the performance of gas sensing devices.

Experimental

Synthesis of tungsten oxide samples

WO₃ powder was synthesized in a typical hydrothermal method as we had reported before.²² The sub-stoichiometric tungsten oxides (e.g. W₁₉O₅₅, W₅O₁₄, W₁₈O₄₉) were obtained by the reduction of WO₃ powders with S powders. The chemical compositions in the sub-stoichiometric tungsten oxides were controlled by the growth temperature. In the growth system, the as-prepared WO₃ powder was located in the middle of a vacuum tube furnace, while S powder was deposited at its upstream position. After evacuating pre-treatment and repeating flush with pure Ar gas, the tube furnace was heated up to the target temperature and dwelled for one hour.²³ The raw WO₃ powders were then annealed at 500 °C in atmospheric environment. W₁₉O₅₅ (dark-blue), W₅O₁₄ (purple-blue), and W₁₈O₄₉ (purple-red) were synthesized at 950, 1050, 1150 °C in the vacuum tube furnace, respectively. The corresponding colours of the resulting sub-stoichiometric tungsten oxides depend on their chemical compositions, as reported in literatures.^19,24–26

Sample characterization and gas sensors measurements

The morphologies of the obtained sub-stoichiometric nanomaterials were examined by a field emission scanning electron microscope (SEM, FEI Sirion 200). Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM, JEOL2100F transmission electron microscope) was adopted to study the crystallographic structures. The phase and composition were investigated by X-ray diffractometry (XRD, Bruker D8 Avance) with a Cu Kα radiation source (40 kV and 40 mA). X-ray photoelectron spectrometer (XPS, ESCALAB 250Xi, Thermo Scientific) was employed to identify the elements and chemical states by using sub-monochromatized Al KR X-ray as the excitation source. The spectrometer was calibrated with the binding energy of the C 1s line (284.5 eV). Gas sensing test was performed on the HW-30A system (Hanwei Electronics Co. Ltd., PR China).

Results and discussions

Morphology, structure and composition characterization

The XRD results of the as-prepared WO₃, W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ are present in Fig. 1. Fig. 2 illustrates the SEM and HRTEM images of the samples. Grown at different temperatures, the pristine WO₃ crystal exhibits irregular rod-like shape as shown in Fig. 2(a) and the W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ samples show the similar micro and nano rod shapes with flat and smooth surface. The crystal nucleus grows promptly along the preferential growth direction, leading to such symmetric structures. The sub-stoichiometric W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ are well crystalline with clear grain boundaries and ordered crystal planes as shown in Fig. 2(b–d). The diffraction curve recorded in Fig. 1(a) (black line) can be well indexed to the phase of monoclinic WO₃ (JCPDS no. 43-1035), with three main diffraction peaks located at 2θ = 23.12, 23.59, and 24.38°, corresponding to the preferential growth directions of [002], [020], [200], respectively. The (002) plane with d-spacing of 3.84 Å is further demonstrated by the HRTEM image in Fig. 2(e).²⁷ The blue curve in Fig. 1(b) matches well with W₁₉O₅₅ (JCPDS no. 45-0167) and the strongest diffraction peak located at 2θ = 23.47° corresponds to the (010) facet with an interlayer distance of 3.7 Å as shown in Fig. 2(f).²⁸ The diffraction peaks of the red curve in Fig. 1(c) are in good agreement with JCPDS no. 71-0292 file for W₅O₁₄. The main peak located at 2θ = 23.41° suggests the (001) planes with interlayer distance of 3.797 Å, which is a close match to the HRTEM result of 3.79 Å in Fig. 2(g). The W₅O₁₄ sample grown in a longitudinal direction along the (001) crystallographic axis has the high aspect ratio of nanorods as shown in the SEM image.²⁴ At last, Fig. 1(d) depicts the XRD curve of monoclinic W₁₈O₄₉ (JCPDS no. 05-0392) with the main peak located at 23.45° corresponding to (010) planes, and the interlayer distance of 3.8 Å is consistent with the HRTEM value, 3.76 Å, as depicted in Fig. 2(h).^18,19,29

Fig. 1 XRD curves of the as-prepared WO₃ and sub-stoichiometric WO_3−x, with standard JCPDS card marked below the corresponding curves.

Fig. 2 SEM morphologies and HRTEM images of WO₃ (a and c), W₁₉O₅₅ (b and d), W₅O₁₄ (e and g), W₁₈O₄₉ (f and h).

X-ray photoelectron spectroscopy (XPS) provides an appropriate solution to investigate the chemical composition and element states. As we know, tungsten has a complex distribution of W 4f binding energies consisted of three pairs of doublets (six peaks). As exhibited in Fig. 3(a–d), the thin black line is raw data and the red one is the result of smooth fitting. The first doublet peaks (blue line) locate at 37.85 eV and 35.75 eV corresponding to W 4f_5/2 and W 4f_7/2, which confirms that the W ions are dominated by W⁶⁺ in WO₃. No W⁵⁺ and W⁴⁺ peaks are observed in Fig. 3(a). However, on the other three samples synthesized at 950, 1050, 1150 °C, we observe the mixture of W⁶⁺, W⁵⁺ and W⁴⁺ oxidation states. As revealed in Fig. 3(b), the second doublet peaks (green dot-curve) appearing at 36.85 and 33.85 eV are attributed to the W⁵⁺ state. The third doublet peaks occur at 35.25 and 32.5 eV (see the navy dot line), implying the presence of the W⁴⁺ oxidation state.²⁶ Thus, as shown by Fig. 3, the contents of W⁵⁺ and W⁴⁺ tend to increase gradually with increasing growth temperature. Moreover, we have concluded the O 1s-level XPS spectra of the four WO_3−x samples in Fig. 3(e–h), the thin black line is raw data and the red one is the sum of four deconvolution spectra. The first peak (blue line) locate at 530.9 eV corresponding to oxygen atoms O²⁻ in the lattice, and the second peak (green line) locate at 532 eV corresponding to adsorbed oxygen species and oxygen state O¹⁻.^30,31 Metal oxide surfaces that are exposed to the atmosphere are always hydrated, so that hydroxyl groups (–OH) and water molecules (H₂O) signals are also found at 531.4, 532.8 eV, respectively.³¹ Our XPS results are in perfect agreement with the report by A. P. Shpak in 2007.³¹

Fig. 3 (a–d) The W 4f-level and (e–h) O 1s-level XPS deconvolution spectra of the as-prepared sub-stoichiometric tungsten oxides, WO₃, W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉.

As we can see in Fig. 3(e), oxygen states of O²⁻ are dominated in WO₃ sample, the content of O¹⁻ or remnants of adsorbed oxygen species are much smaller than that of O²⁻ state. However, the intensity ratio of O²⁻/O¹⁻ is increasing continuously from WO₃ to W₁₈O₄₉, which further confirms the enhanced oxygen adsorption performance in the sub-stoichiometric tungsten oxide. This is the main reason why sensing performance is improved in WO_3−x samples. However, the adsorption of H₂O molecules in atmosphere and the formation of –OH groups are also enhanced in the non-stoichiometric WO_3−x, which is supposed to suppress the sensing process. This explains why our sensing response of WO_3−x samples are improved but with a limited improvement.

From the Raman curves in Fig. 4, Raman characteristic peaks of WO₃ are very clearly exhibited at 806 cm⁻¹, 716 cm⁻¹, 326 cm⁻¹ and 272 cm⁻¹.^28,31 However, much wider peaks are observed at 300 and 800 cm⁻¹ on the sub-stoichiometric tungsten oxide samples. Due to the distortions in the crystalline structure, it is difficult to assign each of the bands to certain vibrations, particularly for the bands below 300 cm⁻¹ which might be deconvoluted into several deformation and lattice vibrations. As exhibited in Fig. 4(b–d), the main bands of W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ are located at 816 cm⁻¹, 805 cm⁻¹ and 801 cm⁻¹, respectively. The main band shifts to lower wave number (lower energy) with increasing x value, indicating that more content of reduced tungsten state (W⁴⁺, W⁵⁺) were formed in the sub-stoichiometric phases.^28,29,32

Fig. 4 The Raman spectra of the prepared sub-stoichiometric tungsten oxides, WO₃ (a), W₁₉O₅₅ (b), W₅O₁₄ (c), and W₁₈O₄₉ (d).

Gas sensing properties and discussion

The as-grown sub-stoichiometric tungsten oxides have stable chemical properties and have become an alternative platform to investigate the behaviour of oxygen vacancies in metal-oxides based gas sensors.²³ Fig. 5 shows temperature-dependent sensitivity curves of WO₃ (a), W₁₉O₅₅ (b), W₅O₁₄ (c), and W₁₈O₄₉ (d) when the sensors are exposed to different concentrations of H₂S (1, 2, 3, 4, 10, 50 ppm), marked in different shades of colours. Firstly, it is found that the low-temperature performance of the sub-stoichiometric WO_3−x sensors has a tendency to improve as the x value increases. The WO₃ and W₁₉O₅₅ sensors show high response at temperature above 380 °C, while the W₅O₁₄ and W₁₈O₄₉ sensors hold the best performance approximately at 280 °C. The W₁₈O₄₉ sensor presents a broad band sensitivity curve with a broad shoulder extending to the low temperature side ranging from 150 to 250 °C approximately. W₁₈O₄₉ has the highest density of oxygen vacancies among the tested sensors, suggesting that low-temperature sensing performance could be improved as a result of increasing concentration of oxygen vacancies. Thus, oxygen deficiency facilitates chemisorption of oxygen molecules on the WO_3−x surface. The W₅O₁₄ sensor has a narrow temperature-dependent response curve with its highest response at around 280 °C. As a general trend, it is found that increasing oxygen vacancies give rise to enhanced device performance at low temperature as shown in Fig. 5(a–d), which is beneficial for applications.

Fig. 5 (a–d) Temperature-dependent response curves of the gas sensors based on WO₃, sub-stoichiometric W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ when exposed to different concentrations of H₂S (1, 2, 3, 4, 10, 50 ppm, marked in different shades of colour). (e–h) The O₂-TPD profiles for the four sensors and their corresponding photographs presented next to the curves.

To understand the oxygen absorption behaviours of the sub-stoichiometric tungsten oxide sensors, Fig. 5(e–h) display the temperature programmed desorption (O₂-TPD) profiles of the WO_3−x sensors, together with their photographs presented next to their corresponding curves. As the temperature increases, ambient oxygen molecules that are adsorbed on the tungsten-oxide surfaces undergo changes following the sequence of O₂ → O₂⁻→ O⁻→ O²⁻.² Physically adsorbed oxygen molecules (O₂) make ignorable contribution to gas sensing. The peaks of chemisorbed oxygen species (O₂⁻, O⁻, O²⁻) are reported to centre around 400, 600 and >700 °C on the TPD curves.³³ Thus, Fig. 5(e) shows that the pristine WO₃ exhibits a broad desorption bands around 360 and 500 °C, attributed to the chemisorption of oxygen species O₂⁻ and O⁻. For the W₁₉O₅₅ sensor, the two peaks are shifted to lower temperature positions at ∼350 and ∼490, as shown in Fig. 5(f). In the case of W₅O₁₄, strong peaks are observed at 310–350 °C (see Fig. 5(g)), while the peak at ∼490 °C is rather weak. This indicates that a large amount of oxygen ions are chemisorbed in 310–350 °C, which might explain why there is a relatively strong response peak at ∼280 °C, close to 310–350 °C where there are a lot of chemisorbed oxygen ions. As x increases to 0.28 (W₁₈O₄₉), a strong peak occurs at the low temperature of 140 °C shown by Fig. 5(h). The peak at 350 or 360 °C remains essentially unchanged. These results indicate that oxygen deficiency enhances chemisorption of oxygen ions on WO_3−x surface at low temperature by offering a large number of chemically active absorption sites, which in turn facilitates the ionosorption process and improves the low-temperature gas sensing sensitivity.^34–36 Thus, both the temperature-dependent response curves and the O₂-TPD profiles measured on the sub-stoichiometric WO_3−x sensors evidence that the working temperature of the gas sensors can be lowered down by using sub-stoichiometric nano WO_3−x.

The W₅O₁₄ sensor presents the best performance over the narrow temperature range around 280 °C. However, W₁₈O₄₉ possesses the highest concentration of oxygen vacancies among all the sub-stoichiometric WO_3−x ones studied here, and the W₁₈O₄₉ sensor presents considerable response at low temperature. Obviously, the x value and the sensing sensitivity of the WO_3−x sensor do not have a simple relationship. Some other factors might affect device performance, such as specific surface area, crystallographic planes on the surface, and density of interfacial states, according to the published works.^18,37 Qin's work suggests that agglomeration of W₁₈O₄₉ at a comparatively high operating temperature weakens the gas diffusion, which in turn gives rise to low response.³⁴ Overall speaking, W₁₈O₄₉ is a metal-like material with abnormally high carrier concentration.^19,38 It is expected that the carrier concentration in W₁₈O₄₉ is comparable to the change of electron population induced by the target gas molecules. This leads to a relatively small resistance change when the sensor is exposed to the target gas, compared to the result measured on the W₅O₁₄ sensor. The W₁₈O₄₉ sensor, with the highest concentration of oxygen vacancies, thus possesses relatively low sensing response compared to the W₅O₁₄ sensor, though its low-temperature performance is quite good.

The response and recovery times of the sensors are also investigated. Fig. 6(a–d) presents the dynamic response curves of the WO₃, W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ sensors measured by the HW-30A gas sensing test system at given temperatures. The measured response (recovery) times are 67 s (65 s), 53 s (37 s), 21 s (11 s), and 32 s (15 s) for the WO₃, W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉ sensors, respectively. This overall results show the tendency that the use of sub-stoichiometric tungsten oxides accelerates response–recovery processes.

Fig. 6 Dynamic response curves of the sub-stoichiometric WO_3−x sensors: (a) WO₃ at 380 °C, (b) W₁₉O₅₅ at 325 °C, (c) W₅O₁₄ and (d) W₁₈O₄₉ at 280 °C after exposed to 1, 2, 3, 4, 10, 50 ppm H₂S.

Conclusions

We have demonstrated that oxygen deficiencies facilitate the low-temperature oxygen ionosorption process, leading to the improvement of sensing performance of the WO_3−x-based gas sensors. High-quality sub-stoichiometric WO_3−x with several x values (WO₃, W₁₉O₅₅, W₅O₁₄, and W₁₈O₄₉) was synthesized and H₂S gas sensors were fabricated using the materials. The W₁₈O₄₉ sensor shows good device performance in a wide temperature range, particularly at the temperature as low as 150 °C. The W₅O₁₄ gas sensor presents very high response value in a narrow temperature range centred at 280 °C. Our work shows that the sub-stoichiometric WO_3−x can be used to enhance the low temperature sensing performance through controlling the concentration of oxygen vacancies.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work is financially supported by National Key R&D Program of China (2016YFA0202200), National Natural Science Foundation of China (Grant no. 11574335, 61471345, 51772213), the Frontier Science Research Project (Key Programs) of Chinese Academy of Sciences (Grant no. QYZDJ-SSW-SLH018).

References

1. A. Gurlo and R. Riedel, Angew. Chem., Int. Ed., 2007, 46, 3826–3848.
2. A. Gurlo, ChemPhysChem, 2006, 7, 2041–2052.
3. N. H. Al-Hardan, M. J. Abdullah, A. Abdul Aziz, H. Ahmad and L. Y. Low, Vacuum, 2010, 85, 101–106.
4. S. W. Choi, J. Y. Park and S. S. Kim, Nanotechnology, 2009, 20, 465603.
5. X. Song, Z. Wang, Y. Liu, C. Wang and L. Li, Nanotechnology, 2009, 20, 075501.
6. H. G. Moon, Y. S. Shim, D. H. Kim, H. Y. Jeong, M. Jeong, J. Y. Jung, S. M. Han, J. K. Kim, J. S. Kim, H. H. Park, J. H. Lee, H. L. Tuller, S. J. Yoon and H. W. Jang, Sci. Rep., 2012, 2, 588.
7. D. Zhang, Z. Liu, C. Li, T. Tang, X. Liu, S. Han, B. Lei and C. Zhou, Nano Lett., 2004, 4, 1919–1924.
8. Y. Zhou, G. Xie, T. Xie, H. Yuan, H. Tai, Y. Jiang and Z. Chen, Appl. Phys. Lett., 2014, 105, 033502.
9. J. D. Fowler, M. J. Allen, V. C. Tung, Y. Yang, R. B. Kaner and B. H. Weiller, ACS Nano, 2009, 3, 301–306.
10. L. Xu, R. Xing, J. Song, W. Xu and H. Song, J. Mater. Chem. C, 2013, 1, 2174.
11. Z. Meng, H. Jian Xing and O. Chung Wo, Nanotechnology, 2012, 23, 315503.
12. D. Jung, M. Han and G. S. Lee, ACS Appl. Mater. Interfaces, 2015, 7, 3050–3057.
13. D. Jung, M. Han and G. S. Lee, Sens. Actuators, A, 2014, 211, 51–54.
14. D. B. Migas, V. L. Shaposhnikov and V. E. Borisenko, J. Appl. Phys., 2010, 108, 093714.
15. F. Wang, C. Di Valentin and G. Pacchioni, Phys. Rev. B, 2011, 84, 073103.
16. J. Song, Z.-F. Huang, L. Pan, J.-J. Zou, X. Zhang and L. Wang, ACS Catal., 2015, 5, 6594–6599.
17. Y. Qin, D. Hua and M. Liu, J. Alloys Compd., 2014, 587, 227–233.
18. Y. Qin, W. Xie, Y. Liu and Z. Ye, Sens. Actuators, B, 2016, 223, 487–495.
19. S. Cong, F. Geng and Z. Zhao, Adv. Mater., 2016, 28, 10518–10528.
20. G. Xi, S. Ouyang, P. Li, J. Ye, Q. Ma, N. Su, H. Bai and C. Wang, Angew. Chem., 2012, 124, 2445–2449.
21. Y. Li, Y. Bando and D. Golberg, Adv. Mater., 2003, 15, 1294–1296.
22. W. Yu, Y. Sun, T. Zhang, K. Zhang, S. Wang, X. Chen and N. Dai, Part. Part. Syst. Charact., 2016, 33, 15–20.
23. J. Qian, Z. Peng, Z. Shen, Z. Zhao, G. Zhang and X. Fu, Sci. Rep., 2016, 6, 25574.
24. M. Remškar, J. Kovac, M. Viršek, M. Mrak, A. Jesih and A. Seabaugh, Adv. Funct. Mater., 2007, 17, 1974–1978.
25. Z. Shen, Z. Zhao, J. Wen, J. Qian, Z. Peng and X. Fu, J. Nanomater., 2018, 2018, 1–12.
26. Z. Shen, Z. Zhao, J. Qian, Z. Peng and X. Fu, J. Mater. Res., 2016, 31, 1065–1076.
27. C.-Y. Su, H.-C. Lin, T.-K. Yang, C.-H. Chang and C.-K. Lin, Mater. Trans., 2009, 50, 2593–2597.
28. J. Qian, Z. Zhao, Z. Shen, G. Zhang, Z. Peng and X. Fu, RSC Adv., 2016, 6, 8061–8069.
29. K. Thummavichai, L. Trimby, N. Wang, C. D. Wright, Y. Xia and Y. Zhu, J. Phys. Chem. C, 2017, 121, 20498–20506.
30. J. C. Dupin, D. Gonbeau, P. Vinatier and A. Levasseur, Phys. Chem. Chem. Phys., 2000, 2(6), 1319–1324.
31. A. Shpak, M. Korduban, M. M. Medvedskij and V. O. Kandyba, J. Electron Spectrosc. Relat. Phenom., 2007, 156–158, 172–175.
32. C. Jian, L. Dongyu, Z. Weihong, X. Fangyan, Z. Jun, G. Li, L. Xiao, D. Shaozhi and X. Ningsheng, J. Phys. D: Appl. Phys., 2008, 41, 115305.
33. X. Huang, Y. Peng, X. Liu, K. Li, Y. Deng and J. Li, Catal. Commun., 2015, 69, 161–164.
34. Y. Qin, M. Hu and J. Zhang, Sens. Actuators, B, 2010, 150, 339–345.
35. J. Liu, Y. Gao, X. Wu, G. Jin, Z. Zhai and H. Liu, Sensors, 2017, 17, 1852.
36. J. Wu, Q. Huang, D. Zeng, S. Zhang, L. Yang, D. Xia, Z. Xiong and C. Xie, Sens. Actuators, B, 2014, 198, 62–69.
37. C. T. Pan, C. Y. Su and Y. C. Luo, Microsyst. Technol., 2016, 23, 2113–2123.
38. E. Salje and B. Güttler, Philos. Mag. B, 1984, 50, 607–620.

---