Detection of acetone vapours using solution-processed tin oxide thin-film transistors

Abnormal concentrations of volatile organic compounds (VOCs) in human breathe can be used as disease-specific biomarkers for the non-invasive diagnosis of medical conditions, such as acetone for diabetes. Solution-processed bottom gate top contact metal oxide thin-film transistors (TFTs) are used to detect acetone vapours, as part of a proof-of-concept study. The effect of increasing annealing temperature (T) and channel length (L) on electrical and sensing performance are explored. Drain current (Ids) increases following exposure as acetone undergoes a redox reaction with the adsorbed oxygen species on the semiconductor surface, which results in free electrons being released back into the conduction band. Responsivity (R) is maximized at negative bias (Vgs < 0). For L = 50 μm, the peak R of the TFT annealed at 450 °C is three times greater than that of the TFT annealed at 350 °C, with Vgs = − 37.5 V and − 33 V, respectively.


Introduction
The remote detection of chemical vapours, specifically volatile organic compounds (VOCs), is paramount to public health and environmental monitoring. The detection of acetone in human breathe is useful for the non-invasive and real-time diagnosis of diabetes. The concentration of acetone from a healthy individual varies from 0.3 to 0.9 ppm and exceeds 1.8 ppm for a diabetic patient [1,2]. Common sensing technologies such as gas chromatography, mass spectrometry and optical spectroscopy are costly, bulky and require a skilled technician [3,4]. Therefore, investigating new devices that are compact, simple to operate, and can detect sub-ppm traces of VOCs with high selectivity and specificity, under a highly humid environment (relative humidity (RH) > 80% [5]), is critical. Chemiresistive gas sensors based on highly abundant metal oxides, such as SnO 2 , are well publicised due to their miniaturized size, ease of fabrication and long-term stability [6,7]. The sensing mechanism is reversible and involves monitoring changes in surface resistance following exposure to analytes of interest [8]. Several groups have developed highly sensitive exhaled breath sensors that incorporate 1D nanostructures, owing to their high surface-to-volume ratio and open porosity. For example, Shin et al., obtained short response (< 11 s) and recovery times (< 6 s) upon exposure to acetone and humid air by combining Pt catalytic nanoparticles with SnO 2 fibres [5]. However, accurate cross-sensitivity toward exhaled breath containing multiple gases such as acetone, toluene and H 2 S remains a major challenge.
Metal oxide semiconductor field effect transistors (MOS-FETs) have been widely reported to detect acetone vapours and other harmful gases (e.g., NO 2 ) via two distinct sensing mechanisms [9,10]. In contrast with chemiresistive gas sensors, current modulation by the gate electrode yields an amplified response and cross-sensitivity can be improved by operating the MOSFET in depletion or enhancement modes. Changes in other electrical parameters such charge carrier mobility (μ) and threshold voltage (V th ) can also be exploited. For example, Andringa et al., deposited ZnO thin films by spray pyrolysis and source-drain electrodes were patterned using photolithography, resulting in finger transistors. I ds decreased upon exposure to NO 2 and an applied positive bias (V gs > 0), which caused a shift in V th for concentrations as low as 10 ppb [9].
Considering the aforementioned points, solution-processed bottom gate top contact (BGTC) single layer SnO 2 TFTs were used to detect acetone vapours under ambient conditions. Source-drain electrodes were deposited via thermal evaporation, which is simple, ideal for large batch processing and can achieve high deposition rates. The effect of increasing T and L on electrical and sensing performance were explored. I ds increased following exposure as acetone underwent a redox reaction with the adsorbed oxygen species on the semiconductor surface, which resulted in free electrons being released back into the conduction band.
To the best of our knowledge, this is the first study examining the use of SnO 2 TFTs to detect acetone vapours and will be pivotal in developing novel exhaled breath sensors.

Materials and methods
Experimental and characterisation techniques were conducted under ambient conditions, unless otherwise stated. The thermal behaviour of the sol-gels was examined using thermal gravimetric analysis (TGA) (Perkin Elmer Pyris 1 TGA) at a ramp rate of 10 °C min −1 from 30 to 500 °C. The crystalline properties of the films were determined using X-ray diffraction (XRD) (Bruker AXS D8 Advance GX003410) and surface morphology examined by atomic force microscopy (AFM) (Asylum Research MFP-3D Infinity). AFM analysis was performed by open-source software, Gwyddion. The electrical performance of the TFTs was characterised using Agilent B2912A source measurement units (SMUs) and probe station. A bespoke chamber ( Fig. 3d) was manufactured to record changes in transfer characteristics upon exposure to analytes.

Calculation of saturation and effective mobility
Differences in I ds following exposure to analytes are caused by effects on μ and V th . Assuming that μ is independent of V gs and neglecting short channel effects, I ds in the saturation regime is given by Eq. (1) [11]: where I ds is the drain current, W L is the channel width-tolength ratio, C i is the capacitance per unit area of the dielectric layer, is the charge carrier mobility, V gs is the gate voltage and V th is the threshold voltage. is obtained by taking the partial derivative with respect to V gs . Rearranging the resulting equation yields Eq. (2): resulted in a several erroneous reported mobilities. The reliability factor (r) given in Eq. (3) accounts for nonlinearities in transfer characteristics [11]. It is defined as the ratio (expressed in %) of the maximum channel conductivity experimentally achieved in a FET to that calculated in an equivalent ideal FET: accounts for the claimed device parameters and mobility in the saturation regime. | | I ds | | max is the experimental maximum drain current reached at the maximum gate voltage, Effective mobility, eff , is thus given by: The values quoted in this report are eff and calculated using C i = 15 nFcm −2 and W = 1 mm. L = 50, 80, 100 and 120 μm unless otherwise stated.

Transistor fabrication
The SnO 2 sol-gel was prepared from anhydrous SnCl 2 that was dissolved in 2-methoxyethanol (10 mL, 0.15 M). The solution was sonicated for 15 min, stirred for 24 h to promote hydrolysis and passed through a syringe fitted with a 22 μm filter prior to spin-coating. The SnO 2 sol-gel was spin-coated at 5000 RPM for 30 s onto pre-cleaned substrates consisting of a 400 nm SiN x dielectric on a highly doped Si wafer, which acted as the back gate. The samples were annealed at 350 °C or 450 °C for 1 h and finally, 50 nm thick Al source-drain electrodes were deposited via thermal evaporation using a shadow mask under high vacuum (≈ 10 -6 mbar) in a system integrated in a nitrogen glovebox. The mask comprised four transistors with variable channel lengths, L = 50, 80, 100 and 120 μm.
The phase transitions of a sol-gel during high temperature annealing can be observed by monitoring the weight change that occurs when a sample is heated at a constant rate. Initially, metal-hydroxide (M-OH) or metal-oxygen (M-O) bonds are created, which is referred to as a hydrolysis reaction. As temperature is increased, condensation and dehydroxylation remove the remaining impurities to generate metal-oxide-metal (M-O-M) bonds. The lattice becomes denser, which can lead to a phase transition from amorphous to polycrystalline by means of an exothermic reaction [12,13]. A TGA curve from the SnO 2 sol-gel is shown in Fig. 1. The derivative of the curve is plotted to highlight the

Acetone sensing in air
An open 3 mL vial was filled with acetone and placed in a sealed custom-built chamber (Fig. 3d), which was then allowed to form a saturated atmosphere overnight. When I ds plateaued, the taps on the chamber were opened in a fume cupboard. The TFT was removed and heated at 100 °C for 1 h under ambient conditions to recover I ds . R was calculated using Eq. (5) for each TFT: where ΔI ds represents the difference in I ds before ( I ds [0] ) and after ( I ds analyte ) exposure to acetone vapours.

Optimisation of TFT charge transport layer
Single layer SnO 2 TFTs were fabricated and characterised to investigate the effect of increasing T and L on electrical performance. The device structure and the corresponding transfer characteristics for the TFTs annealed 350 °C and 450 °C are shown in Fig. 2a and b, respectively. XRD analysis of the film annealed at 450 °C showed peaks at 2θ = 26.6° and 33.9° (Fig. 2c) (Fig. 2d) suggested that the film was polycrystalline, which would have allowed the M-O-M backbones in the metal oxide to serve as electron conductance pathways between atoms. The maximum grain size and root mean square (rms) roughness were calculated from the AFM image as 32.2 nm and 1.28 nm, respectively, which indicated a high-quality morphology. Conversely, XRD and AFM analyses of the film annealed at 350 °C (Fig. S2) suggested that the film was amorphous, as the maximum grain size and rms roughness were calculated as 7.4 nm and 0.340 nm, respectively.
Maximum drain current (I ds [max] ) and μ eff increased with T for all L (Table 1 (L = 50 μm) and S1 (L = 50, 80, 100 and 120 μm). The output characteristics (Fig. S1) show that high I gs was present, as the curves shifted right with increasing V gs .

Acetone sensing in air
Following exposure to acetone vapours, I ds of the SnO 2 TFTs were examined in both depletion and enhancement modes. I ds for both devices (L = 50 um) were monitored following overnight exposure, as shown in Fig. 3a and b. R was plotted against V gs (Fig. 3c) for L = 50, 80 and 100 μm.

Discussion
The substantial improvement in μ eff and I ds [max] with increasing T may have been owing to a greater proportion of M-O-M bonds, which are known to result in a highly ordered and dense lattice. It is likely that the TFT annealed at 350 °C would have contained a greater proportion of M-OH bonds due to incomplete dehydroxylation and condensation, which acted as traps and hence hindered electron transport [15]. Table S1 suggested that I ds [max] decreased with increasing L, as the resistance of the channel is inversely proportional to the width-to-length ratio. Therefore, I gs was dominant at larger L and less charges were able to accumulate at the semiconductor-dielectric interface.
Chen et al., found that acetone acts as a reducing agent that transfers electrons to the SnO 2 (110) surface [15]. The sensing mechanism is shown in Eqs. (6)- (8). Due to the large electronegativity of oxygen atoms, adsorbed oxygen depletes electrons from the SnO 2 film, which increases its resistance and forms reduced oxygen species (O 2− , O 2 − or O − , depending on temperature). The film resistance in air is therefore larger than in vacuum. Acetone undergoes a redox reaction with the reduced oxygen species, which results in free electrons being released back into the conduction band of SnO 2 [16] . I ds increased following exposure to acetone vapours, as depicted in Fig. 3a and b. This was consistent with the conductivity of the film increasing due to surplus electrons donated by acetone. The increase in V th (Table 1) suggested that excess electrons prevented normally off operation as the film was too conductive. R was maximized at negative bias (V gs < 0) (Fig. 3c). For L = 50 μm, the peak R of the TFT annealed at 450 °C was three times greater than that of the TFT annealed at 350 °C, with V gs = − 37.5 V and − 33 V, respectively. The shape of the peak broadened with increasing L irrespective of T, which suggested that a high R can be achieved at the V gs required for a battery-driven handheld sensor. Oxygen species readsorbed to the SnO 2 surface during heating under ambient conditions. I ds was recovered as the I on/off was comparable to that measured before exposure. V th was also noticeably lower, which could be owing to the detrapping of electrons and increased density of conducting states.
To conclude, the TFT annealed at 450 °C (L = 50 μm) was more responsive to the presence of acetone vapours across the whole bias range than the TFT annealed at 350 °C, due to increased electron transport facilitated by a denser M-O-M network. AFM and XRD analyses indicated clear evidence of higher crystallinity with T. At smaller L, a greater density of charges were able to accumulate at the semiconductordielectric interface and promoted electron transfer between acetone and SnO 2 . A new, improved vapour testing chamber is currently being designed, which will allow vapour concentration to be accurately determined. Testing and recovery under nitrogen will also be possible. This study could further benefit from Fourier transform infrared spectrometry (FT-IR) and X-ray photoelectron spectroscopy (XPS), which may confirm that the film annealed at 450 °C contained a greater fraction of M-O-M bonds. Differential scanning calorimetry (DSC) could also be performed to help identify the temperature range at which crystallization occurred. However, we believe that the results presented here are sufficient for this paper.
Going forward, SnO 2 TFTs annealed at T > 450 °C will be fabricated and characterised to investigate whether electrical and sensing performance can be further enhanced. These devices will be exposed to VOCs for the diagnosis of other conditions, such as toluene for lung cancer [17].

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1557/ s43580-023-00494-5. Fig. 3 a Transfer characteristics in the saturation regime at V ds = 40 V, V gs = − 40 to 40 V and L = 50 μm, obtained before and after the TFTs annealed at 350 °C and. b 450 °C were exposed to acetone overnight. I gs is shown as dashed lines. c R versus V gs of the TFTs annealed at 350 °C and 450 °C for L = 50, 80, and 100 μm. 450-50 corresponds to T = 450 °C and L = 50 μm. d Testing chamber