Highly responsive diabetes and asthma sensors with WO₃ nanoneedle films for the detection of biogases with low concentrations

Abstract

A diabetes sensor was developed to detect low concentrations of acetone gas, which is a diabetes biomarker. A WO₃ nanoneedle film was synthesized via an aqueous process for use as a sensitive sensing membrane. Acetone was adsorbed and oxidized on the WO₃ nanoneedle film, which changed the sensor resistance. The sensor exhibited a high response of R_a/R_g = 19.72, where R_a is the sensor resistance in air, and R_g is the sensor resistance in air containing 10 ppmv acetone gas. The sensor also exhibited a high response (25.36) to 1 ppmv NO₂, which is related to asthma. Furthermore, the sensor responded to various biogases associated with diseases. The sensor responses to 10 ppmv of the lung cancer marker gases acetaldehyde and toluene were 13.54 and 9.49, respectively. The sensor responses to 10 ppmv isoprene, ethanol, para-xylene, hydrogen, and NH₃ were 7.93, 6.33, 4.51, 2.08, and 0.90, respectively. Trace amounts of acetone and NO₂ gases (25 and 250 ppbv, respectively) were detected. The limits of detection for acetone and NO₂ gases were estimated to be 2.4 and 1.5 ppbv, respectively. The sensor exhibited superior ability to detect low concentrations of biomarker gases. The unique characteristics of the WO₃ nanoneedle film contributed to its high response rates.

Introduction

Interest in early detection of diseases and chain physical condition are increasing. Diseases induces changes in the components of the body and blood, and biogas components are emitted from the body, such as exhaled breath and skin gases. The correlations between various gases and diseases have been investigated, and disease marker gases have been identified^1,2,3,4. As such, there is strong demand for early detection of diseases by sensing low concentrations of biogases emitted from the living body. Acetone is a marker for diabetes^1,2,3,4, and patients with diabetes exhibit acetone skin gas concentrations of 188 ppbv, which is higher than that of a healthy person (87 ppbv)⁵. The acetone concentration in skin gas is correlated with the β-hydroxybutyrate, blood glucose, and HbA1c levels in the blood of diabetic patients. The skin gases from patients with diabetic ketoacidosis have high acetone concentrations (940 ppbv), which decrease to 80 ppbv after insulin therapy. It is expected that early screening will be possible by sensing the acetone concentration in exhaled breath and skin gas with small portable devices such as smartwatches. Nitrogen oxides (NO, NO₂, and HONO) are associated with asthma. NO gas is a marker for asthma, and sensing NO and its oxidation product, NO₂, is useful in patients with asthma^1,2,3,4. Acetaldehyde and toluene⁶ are lung cancer markers, and expectations are high for early detection of lung cancer^1,2,3,4. Isoprene is a biomarker produced from intermediates of cholesterol synthesis^1,2,3,4. Ethanol (C₂H₅OH) is a biomarker for alcohol consumption and is detectable in biogas^1,2,3,4. The H₂ concentrations in biogas increase due to overgrowth of intestinal anaerobic bacteria in patients with dyspepsia syndrome^1,2,3,4. Hepatic encephalopathy, congenital enzymatic abnormalities in the urinary tract, Helicobacter pylori infection, and other infectious diseases increase the concentrations of ammonia (NH₃) in the body^1,2,3,4. Additionally, various gases, such as para-xylene, are present in biogas and are useful in determining the state of the body^1,2,3,4.

Semiconductor gas sensors exhibit long-term stability, rapid recovery, and heat, ultraviolet, vibration, and chemical resistance. Therefore, they have been used in gas leak detectors. The need for biogas detection has increased in recent years, and semiconductor gas sensors capable of detecting biogas have been developed. In particular, metal oxide nanostructures have attracted considerable attention^7,8 as they are able to detect low-concentrations of gases^{9,10,11,12,13,14,15,16,17}. For example, SnO₂ nanosheets were developed via an aqueous process¹⁸ for use as lung cancer sensors^19,20. The SnO₂ nanosheets exhibited strong responses to various biogases^21,22,23. Psychological stress sensors have been developed from NiO/MnO₂@NiO²⁴ and SnO₂ nanosheets²⁵. ZnO nanobelts²⁶ and porous CuO²⁷ with metal-organic frameworks have been developed to detect low concentrations of biogases. However, new materials and gas sensors are needed to detect low biogas concentrations.

Many semiconductor gas sensors have been developed from SnO₂ or ZnO. However, the sensing properties of WO₃ are expected to differ from those of SnO₂ and ZnO. A WO₃ particulate film-type sensor was synthesized via the sol-gel method, followed by annealing at 500 °C for 2 h in a N₂ atmosphere²⁸. The average particle size was 76.77 nm, and the particle size distribution was 20–220 nm. The WO₃ particulate film sensor exhibited responses of 1.1, 1.6, and 20.36 to 5, 10, and 100 ppmv NO₂, respectively, at 200 °C. The detection limit was 5 ppmv NO₂ with a reasonable sensing response of 1.1. The sensor also exhibited a response of 18.51 to 100 ppmv acetone at 200 °C.

A microbrick sensor comprising WO₃ nanosheets was synthesized from a peroxy-tungstic acid (PTA) sol with a hydrothermal method at 180 °C²⁹. The widths of the WO₃ microbricks increased from 250–300 to 400–700 nm, and the spacings between them decreased with an increasing quantity of the PTA sol. The thicknesses of the WO₃ films increased from 3.0 to 8.8 µm with an increasing amounts of the PTA sol. The microbrick sensor with WO₃ nanosheets exhibited responses of 1.5, 2.5, and 11.5 to 5, 10, and 100 ppmv NO₂, respectively, at 300 °C. The sensor also exhibited a response of approximately 1 for 100 ppmv acetone.

A WO₃ polycrystalline film sensor was synthesized via direct thermal evaporation of metal tungsten foils in an oxygen atmosphere³⁰. The WO₃ film consisted of 800 nm crystals and contained both large and small pores. The WO₃ polycrystalline film-type sensor exhibited responses of 1, 4, and 30 to 1, 10, and 100 ppmv NO₂, respectively, at 250 °C. The sensor was less sensitive to NO and NH₃.

WO₃ thin-film sensors were synthesized with a spray pyrolysis technique, and the film thicknesses were varied³¹. The response (%) was calculated as 100 × |R_g-R_a | /R_a. The sprayed 745 nm-thick WO₃ thin-film-type sensor exhibited responses of 20% (R_a/R_g = 1.2) and 97% (R_a/R_g = 1.97) to 10 and 100 ppmv NO₂, respectively, at 200 °C. The sprayed WO₃ thin film exhibited a dense dual-scale wire network. The crystallite sizes were estimated to be 69 nm. The detection limit for NO₂ was 10 ppmv. The responses to 100 ppmv H₂S, liquefied petroleum gas, NH₃, CO, and SO₂ gases were 7 (R_a/R_g = 1.07), 8 (R_a/R_g = 1.08), 9 (R_a/R_g = 1.09), 10 (R_a/R_g = 1.10), and 12% (R_a/R_g = 1.12), respectively.

WO₃ nanoflowers were prepared via a hydrothermal method³². The WO₃ nanoflower-type sensor exhibited responses of 10 (R_a/R_g = 1.1), 20 (R_a/R_g = 1.2), 45 (R_a/R_g = 1.45), 105 (R_a/R_g = 2.05), 150 (R_a/R_g = 2.15), 200 (R_a/R_g = 3.00), 210 (R_a/R_g = 3.10), and 230% (R_a/R_g = 3.23) to 1, 5, 10, 20, 40, 60, 80, and 100 ppmv NO₂, respectively, at 200 °C. The sensor exhibited negligible responses to 100 ppmv CO, H₂S, SO₂, and NH₃.

WO₃ and yttrium-doped WO₃ nanoplates (tungsten:oxygen:yttrium = 22.59:77.20:0.21 at. %) were prepared with a hydrothermal method³³. The WO₃ nanoplates exhibited responses of 1.36 (36.19%), 1.54 (54.45%), 2.10 (109.86%), and 12.10 (109.8%) to 5, 10, 15, or 20 ppmv NO₂, respectively, at 150 °C. The yttrium-doped WO₃ nanoplates exhibited responses of 3.7 (270%), 21 (1950%), 88 (8722%), and 101 (10022%) to 5, 10, 15, and 20 ppmv NO₂, respectively, at 150 °C.

Bilayer polyaniline-WO₃ (PANI–WO₃) thin films were prepared with a hydrothermal synthesis and in situ chemical oxidative polymerization³⁴. The p-type PANI exhibited one-dimensional nanofibers with diameters of approximately 100 nm. The nanofibers were randomly distributed on the surfaces of the n-type flower-like WO₃. p-n junctions were formed between the PANI and WO₃. The bilayer PANI-WO₃ thin-film sensor exhibited responses of 1.18, 1.23, 1.45, 1.66, 2.39, 14.47, and 60.9 to 2, 4, 6, 8, 10, 20, and 30 ppmv NO₂, respectively, at 50 °C.

WO₃ nanorods were synthesized via thermal oxidation³⁵. A tungsten film was deposited on an oxidized silicon substrate via sputtering and subsequently subjected to thermal oxidation at 500 °C in air. The tungsten film thickness was 100 nm. The WO₃ nanorods were approximately 80 nm in diameter and 400 nm in length. The WO₃ nanorod-type sensor exhibited responses of 1.50 (50%), 1.80 (80%), 2.115 (115%), 2.65 (165%), and 3.55 (255%) to 2, 5, 10, 20, and 50 ppmv NO₂, respectively, at 225 °C. The WO₃ nanorod-type sensor also exhibited responses of 5, 12, 22, 18, and 19 to 10 ppmv acetone, methanol (CH₃OH), ethanol, H₂S, and NH₃, respectively, at 225 °C.

Acetone gas sensors have also been developed with WO₃. Sea urchin-like WO₃ particles that measured 0.86 μm were prepared with a hydrothermal method at 180 °C³⁶. The deep-level emission in the photoluminescence spectra indicated that the particles contained many oxygen vacancies. The sea urchin-like WO₃ sensor exhibited responses of 2.8, 4.7, 6.4, 8.7, 16.9, 27.2, 41.1, 94.1, 144.1, 214.9, and 337.6 to 2, 5, 10, 20, 50, 100, 200, 500, 1000, 2000, and 5000 ppmv acetone, respectively, at 200 °C. The responses to 100 ppmv NO₂, H₂, N,N-dimethylformamide, ethylene glycol (C₂H₆O₂), CO, NH₃, CH₃OH, H₂S, and C₂H₅OH were −12, 9, 7, 4, 4, 3, 3, 2, and 2, respectively.

WO₃ nanosheets were prepared with a hydrothermal method at 110 °C and annealed at 350–500 °C³⁷. g-C₃N₄ nanosheets were prepared by ultrasonically treating bulk g-C₃N₄. The g-C₃N₄ and WO₃ nanosheets were mixed to form g-C₃N₄-modified WO₃ nanosheets. A 1 wt. % g-C₃N₄-modified WO₃ nanosheet sensor exhibited responses of 3.5, 4.8, 11, 25, 35, and 65 to 0.5, 1, 10, 50, 100, and 1000 ppmv acetone, respectively, at 340 °C. In contrast, the WO₃ nanosheet-type sensor exhibited a response of 12 to 100 ppmv acetone at 340 °C. The addition of g-C₃N₄ to the WO₃ nanosheet sensor enhanced the response of the sensor.

A Cr₂O₃-doped (5 mol%) WO₃ thin-film sensor exhibited responses of 4.19, 6.13, and 8.91 to 10, 15, and 20 ppmv acetone, respectively, at 320 °C³⁸.

Copper-doped (3 mol%) WO₃ fibers were synthesized via electrospinning³⁹. The samples were annealed at 500 °C and quenched to form 3 mol% copper-doped WO₃ hollow fibers. The hollow fibers had wall thicknesses of approximately 300–500 nm. The walls comprised numerous 20–30 nm nanocrystals. The 3 mol% copper-doped WO₃ hollow fiber sensor exhibited a response of 6.43 to 20 ppmv acetone at 300 °C.

Flower-like WO₃ particles composed of intersecting nanoflakes with diameters of 1.5–2 μm were synthesized via a hydrothermal method⁴⁰. The flower-like WO₃ sensor exhibited responses of 12.5 and 10.9 to 100 ppmv C₂H₅OH and acetone, respectively, at 320 °C. Flower-like MoO₃/WO₃ particles with molybdenum:tungsten molar ratios of 4 were also prepared. These particles exhibited responses of 28.5 and 18.2 to 100 ppmv C₂H₅OH and acetone, respectively, at 320 °C. These responses were approximately 2.3- and 1.7-fold higher than those of the flower-like WO₃.

WO₃ particles were prepared with a surfactant-mediated method and annealed at 500 °C^41,42. Single-crystalline WO₃ plates were prepared via a hydrothermal method with cetyltrimethyl ammonium bromide⁴². The particulate WO₃ gas sensor exhibited responses of 11, 27 to 100, and 1000 ppmv acetone, respectively, at 307 °C⁴². The single-crystalline WO₃ plate-type gas sensor exhibited responses of 15, 105 to 100, and 1000 ppmv acetone, respectively, at 307 °C. The single-crystalline WO₃ plate-type gas sensor also exhibited responses of 10, 19, 5, and 19 to 100 ppmv NH₃, 1000 ppmv NH₃, 100 ppmv alcohol, and 1000 ppmv alcohol, respectively, at 307 °C.

WO₃ hollow spheres with diameters of 400 nm were synthesized in solution via hydrolysis of WCl₆ with carbon microspheres⁴³. The WO₃ hollow spheres were annealed at 450 °C for 2 h in air. The shell thicknesses were approximately 30 nm, and the spheres contained holes and wrinkles and were constructed from nanoparticles with diameters of approximately 12 nm. The WO₃ hollow sphere gas sensor exhibited responses of 3.53, 4.56, 6.04, 13.5, 16, 18.6, and 23.1 to 50, 100, 200, 500, 1000, 2000, and 5000 ppmv acetone, respectively, at 400 °C.

WO₃ nanosheets were prepared by deposition from solution and annealed at 300 °C⁴⁴. The surfaces of the nanosheets were modified with silver nanoparticles to form silver quantum dot-sensitized WO₃ nanosheets. The WO₃ nanosheet gas sensor exhibited responses of 2.5, 4.5, 7.5, 9.5, 13, and 14 to 10, 20, 50, 100, 300, and 500 ppmv acetone, respectively, at 340 °C. The silver quantum dot-sensitized WO₃ nanosheet gas sensor exhibited higher responses of 2.5, 5, 8, 12, 17, and 19.5 to 10, 20, 50, 100, 300, and 500 ppmv acetone, respectively, at 340 °C.

WO₃ nanostructures were used as sensors for NO₂ and acetone. However, the responses to low concentrations of gases must be improved for biogas detection. Therefore, WO₃ nanoneedle-derived films were fabricated in this study using an aqueous process. A diabetes sensor was developed with the WO₃ nanoneedle films. The sensor exhibited a high response to 10 ppmv acetone gas, a diabetes marker. The gas sensor responded to acetone and other gases, such as NO₂, acetaldehyde, toluene, isoprene, C₂H₅OH, para-xylene, H₂, and NH₃. The developed semiconductor-type gas sensor was small and lightweight with no size restrictions; therefore, it can be used in small portable devices such as smartwatches.

Experimental

Syntheses of WO₃ and WO₃(H₂O)_0.33 nanoneedles

Tungstic acid (CAS registration number: 7783-03-1; molecular formula: H₂WO₄; molecular weight: 249.85, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan), acetonitrile (CAS registration number: 75-05-8; molecular formula: CH₃CN; molecular weight: 41.05, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan), and hydrochloric acid (CAS registration number: 7647-01-0; molecular formula: HCl; molecular weight: 36.46, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) were used as received. Tungstic acid (0.25 mol/L, 120 mL), CH₃CN (10 mL), and HCl (6 mol/L, 5 mL) were dissolved in distilled water (25 mL) and maintained at 70 °C for 1 week. The synthesized particles were then removed from the solution by centrifugation and subsequently heated at 300 °C for 2 h in air.

Synthesis of a seed layer for a WO₃ nanoneedle film

Comb-shaped platinum electrodes were formed on an Al₂O₃ substrate (3 mm × 3 mm) via screen printing (Fig. 1a). A platinum heater electrode was deposited on the back surface of the Al₂O₃ substrate. The comb-shaped platinum electrodes were irradiated with vacuum ultraviolet light for 20 min to remove adsorbed organic molecules. Subsequently, the surface became superhydrophilic.

Fig. 1: Schematic illustration showing the synthesis of the WO₃ nanoneedle film for acetone gas sensing.

a Al₂O₃ substrate with Pt electrodes, b WO₃ nanoneedles, and c acetone gas sensing.

Hydrogen peroxide (30%) (CAS registration number: 722-84-1; molecular formula: H₂O₂; molecular weight: 34.01) with a density of approximately 1.11 g/mL (at a mass fraction of 30%), H₂WO₄ (CAS registration number: 7783-03-1; molecular formula: H₂WO₄; molecular weight: 249.85), and polyvinyl alcohol (PVA; CAS registration number: 9002-89-5; molecular formula: (-CH₂CHOH-)_n; average degree of polymerization: 1500 or more (1500–1800) were used as received.

Tungstic acid (1.25 g) and PVA (0.5 g) were dissolved in H₂O₂/water (30%) (10 mL). The Al₂O₃ substrate was coated with a tungstic acid solution via dip-coating. The substrate was heated at 500 °C for 2 h in air to form a seed layer.

Synthesis of a WO₃ nanoneedle film gas sensor

Tungstic acid (0.25 mol/L, 120 mL), CH₃CN (10 mL) and HCl (6 mol/L, 5 mL) were dissolved in distilled water (25 mL). The substrate with the seed layer was immersed in the solution and maintained at 70 °C for 7 d to form a WO₃ nanoneedle film (Fig. 1b). The substrate was immersed face-down to suppress precipitation of the particles generated in the solution onto the substrate.

Characterization of the WO₃ nanoneedle film

The crystal structures of the synthesized WO₃ nanoneedles and WO₃ nanoneedle film were evaluated with X-ray diffractometry (XRD; SmartLab, Rigaku Co. Ltd., Japan) operating at 40 kV and 30 mA with Cu Kα radiation (λ = 1.5418 Å). The morphologies of the WO₃ nanoneedles and sensor structures were observed with field emission scanning electron microscopy (FE-SEM; JSM-6335FM, JEOL, Japan).

The samples were also examined with transmission electron microscopy (TEM; Tecnai Osiris, 200 kV, FEI), high-angle annular dark-field scanning TEM (HAADF-STEM), energy-dispersive X-ray spectroscopy (EDS; JED-2300T, DRY SD100GV, JEOL Ltd.), and electron energy-loss spectroscopy (Quantum ER, Gatan, Inc.). Cross-sectional samples were prepared with a focused ion beam (10–40 kV). The WO₃ nanoneedle film was fixed with a resin for cross-sectional observations. Some of the WO₃ nanoneedle films embedded in the resin were peeled off the Al₂O₃ substrate during cross-sectioning via ion milling.

Evaluation of the gas sensing performance

An Al₂O₃ substrate with a WO₃ nanoneedle film was evaluated as a gas sensor (Fig. 1c). The sensor was heated to 300 °C wit a platinum heater on the back side of the sensor. The resistance of the sensor in air (R_a) was evaluated with a small resistance-measuring instrument under a dry air flow (N₂:O₂ = 80:20). Because of the measurement range of the resistance meter, the noise in the measurement data was high at resistance values ≥ 1 MΩ. N₂ and O₂ gases in gas cylinders (G1 grade, > 99.99995 vol. % and G1 grade, > 99.99995 vol.%, respectively) were used for the evaluations. The resistance of the sensor to target gases (R_g) was evaluated by measuring the resistance under flowing dry air (N₂:O₂ = 80:20) containing 10 ppmv of the target gas. Acetone, H₂, NH₃, NO₂, C₂H₅OH, toluene, acetaldehyde, para-xylene, and isoprene were used as the target gases. The total flow rate was 100 cm³ min⁻¹. The sensor response (R_a/R_g) was calculated from the sensor resistance in air (R_a) and sensor resistance in air containing the target gas (R_g).

The 50% (T50 (res)) and 90% response times (T90 (res)) are the times needed for the sensor resistance to achieve 50 and 90%, respectively, of the total resistance change resulting from exposure to the target gas, and there were calculated using:

$$T50({\rm{res}}):{R}_{a}-\Delta R\times 0.5$$

(1)

$$T90({\rm{res}}):{R}_{a}-\Delta R\times 0.9$$

(2)

$$\Delta R={R}_{a}-{R}_{g}$$

(3)

where the 50% (T50 (rec)) and 90% recovery times (T90 (rec)) are the times needed for the sensor resistance to reach 50 and 90%, respectively, of the total resistance change upon exposure to the target gas and were calculated as:

$$T50({\rm{rec}}):{R}_{g}+\Delta R\times 0.5$$

(4)

$$T90({\rm{rec}}):{R}_{g}+\Delta R\times 0.9$$

(5)

$$\Delta R={R}_{a}-{R}_{g}$$

(6)

The resistance in NO₂ gas was higher than that in air because NO₂ is an oxidizing gas. The sensor response was calculated from R_g/R_a. The 50% (T50 (res)) and 90% response times (T90 (res)) were defined as:

$$T50({\rm{res}}):{R}_{a}+\Delta R\times 0.5$$

(7)

$$T90({\rm{res}}):{R}_{a}+\Delta R\times 0.9$$

(8)

$$\Delta R={R}_{g}-{R}_{a}$$

(9)

The 50% (T50 (rec)) and 90% recovery times (T90 (rec)) were defined as:

$$T50({\rm{rec}}):{R}_{g}-\Delta R\times 0.5$$

(10)

$$T90({rec}):{R}_{g}-\Delta R\times 0.9$$

(11)

$$\Delta R={R}_{g}-{R}_{a}$$

(12)

The concentrations of the gases produced in this study are displayed in parts per million by volume (ppmv), parts per billion by volume (ppbv), and parts per trillion by volume (pptv). The concentrations of the gases produced in this study, namely, 1 ppmv, 1 ppbv, and 1 pptv, were equivalent to 1 µmol/mol, 1 nmol/mol, and 1 pmol/mol, respectively. Parts per million by volume (ppmv, µmol/mol) was used for the volume fractions, and parts per million weight (ppmw) were used for the mass fractions. In several gas sensor studies, 1 ppmv, 1 ppbv, and 1 pptv were referred to as 1 ppm, 1 ppb, and 1 ppt, respectively.

The sensing evaluations presented in Section “Low-concentration gas sensing and limit of detection (LOD) for the WO₃ nanoneedle filmsensor” for low gas concentrations were conducted precisely and compared with the simple evaluations described in Section “Gas sensing performance and gas selectivity of the WO₃ nanoneedle film sensor”. Trace amounts of the reducing gas were adsorbed by the evaluation system, which decreased the responses to NO₂ in the evaluations described in Section “Gas sensing performance and gas selectivity of the WO₃ nanoneedle film sensor”.

Results and discussion

Syntheses of WO₃ and WO₃(H₂O)_0.33 nanoneedles

Nanoneedles were crystallized from an aqueous solution containing tungstic acid (0.25 mol/L, 120 mL), CH₃CN (10 mL), HCl (6 mol/L, 5 mL), and water (25 mL) and were maintained at 70 °C for 1 week (Fig. 2a). The nanoneedles were approximately 100 nm long and 10 nm wide. The aspect ratio was approximately 10 (100 nm/10 nm). Their structures and sizes were maintained after annealing at 300 °C for 2 h in air (Fig. 2b). The X-ray diffraction (XRD) pattern for the as-synthesized nanoneedles was assigned to WO₃(H₂O)_0.33 (Powder Diffraction Data File (PDF) No. 01-087-1203) (Fig. 3a). The nanoneedles underwent a phase transformation into WO₃ (PDF No. 01-085-2459) while annealing at 300 °C for 2 h in air (Fig. 3b). The XRD pattern for WO₃ was similar to that of WO₃(H₂O)_0.33. The diffraction peak at 18°, which was observed for WO₃(H₂O)_0.33 but not for WO₃, was characteristic of WO₃(H₂O)_0.33 and was attributed to the {111} plane of WO₃(H₂O)_0.33. This peak was useful for distinguishing WO₃(H₂O)_0.33 and WO₃. The slight residual WO₃(H₂O)_0.33 was completely transformed into WO₃ by extending the annealing period.

Fig. 2: SEM images of the WO₃ (H₂O)_0.33 nanoneedles.

a The WO₃ (H₂O)_0.33 nanoneedles before annealing at 300 °C for 2 h in air. b The WO₃ (H₂O)_0.33 nanoneedles after annealing at 300 °C for 2 h in air.

Fig. 3: XRD patterns for the WO₃ (H₂O)_0.33 nanoneedles.

a The WO₃(H₂O)_0.33 nanoneedles before annealing at 300 °C for 2 h in air. b The WO₃ nanoneedles after annealing at 300 °C for 2 h in air.

Synthesis of a WO₃ nanoneedle film for use in a gas sensor

A WO₃ nanoneedle film was formed on the Al₂O₃ substrate via a seed layer in an aqueous solution containing tungstic acid (0.25 mol/L, 120 mL), CH₃CN (10 mL), HCl (6 mol/L, 5 mL), and distilled water (25 mL) at 70 °C for 7 d (Fig. 4a). The substrate was uniformly covered with WO₃ nanoneedles, which were elliptical with tips (Fig. 4c1–d2). The nanoneedles were approximately 500 nm long and 5–50 nm wide (Fig. 4c2, d2). The aspect ratios were approximately 10 (500 nm/50 nm). Smaller nanobranches grew from the nanoneedles via dendritic crystal growth. The angles between the nanoneedles and nanobranches were approximately 90°, as shown in Fig. 4d2. The WO₃ crystals were hexagonal. Considering the shapes and branching angles, the longitudinal direction of the nanoneedles may be along the c-axis of WO_3, and the longitudinal direction of the nanobranches may be perpendicular to the c-axis of WO₃. The nanoneedle structures were maintained during annealing at 300 °C for 2 h in air (Fig. 4b).

Fig. 4: SEM images of the WO₃ nanoneedle film.

a The WO₃ nanoneedle film before annealing at 300 °C for 2 h in air. b The WO₃ nanoneedle film after annealing at 300 °C for 2 h in air. (c1, d1) Magnified images of a. c2 Image in (c1) with a WO₃ nanoneedle illustrated to show its size and morphology. d2 Image in (d1) with a WO₃ nanoneedle illustrated to show its size and morphology.

Cross-sectional images of the WO₃ nanoneedle film on the Al₂O₃ substrate show that the Al₂O₃ substrate was covered with a dense particle layer and that WO₃ nanoneedles were formed on the dense particle layer (Fig. 5a–b). Because the Al₂O₃ substrate was a sintered body of Al₂O₃ particles, the surface exhibited irregularities measuring several micrometers. In general, it is difficult to form a uniform film on a substrate with an unevenness of several micrometers. However, in this study, a dense particle layer and WO₃ nanoneedle film were formed along an uneven substrate.

Fig. 5: Cross-sectional image of WO₃ nanoneedle films.

Cross-sectional image of a WO₃ nanoneedle film on the Al₂O₃ substrate. b A low magnification image of (a).

The XRD pattern of the WO₃ nanoneedle film was assigned to WO₃ (PDF No. 01-085-2459) (Fig. 6a). A small WO₃(H₂O)_0.33 peak (PDF No. 01-087-1203) was observed at 18°. The diffraction peaks of Al₂O₃ and platinum were also detected because the Al₂O₃ substrate contained comb-shaped platinum electrodes on its surface. The WO₃ nanoneedle film was composed of a single phase of WO₃ (PDF No. 01-085-2459) after the annealing treatment at 300 °C for 2 h in air (Fig. 6b). Annealing completely transformed the small amount of the WO₃(H₂O)_0.33 phase into WO₃.

Fig. 6: XRD patterns of the WO₃ nanoneedle films.

a The WO₃ nanoneedle film before annealing at 300 °C for 2 h in air. b The WO₃ nanoneedle film after annealing at 300 °C for 2 h in air.

The morphologies of the dense particle layers and WO₃ nanoneedle films were observed with TEM (Fig. 7a). The dense particle layer was composed of spherical particles. The WO₃ nanoneedles were elliptical with tips similar to those observed with FE-SEM. The nanoneedles were approximately 500 nm long and 5–50 nm wide. Moreover, smaller nanoneedles were observed with approximate 10–100 nm lengths and 5–10 nm widths. The lattice spaces were calculated from the selected-area electron diffraction (SAED) patterns of the nanoneedles (Fig. 7b). The lattice spacings of 0.316, 0.317, 0.381, 0.364, and 0.628 nm were assigned to the {020}, {020}, {001}, {001}, and {010} planes of WO₃, respectively. Additionally, a weak diffraction pattern was observed. The calculated lattice spacing was 0.480 nm. This peak was assigned to the {111} plane of WO₃(H₂O)_0.33. The main phase of the nanoneedles was WO₃ and included a small amount of WO₃(H₂O)_0.33, which was consistent with the XRD pattern of the WO₃ nanoneedle film.

Fig. 7: TEM analysis of the WO₃ nanoneedles.

a TEM image of the WO₃ nanoneedle film. b SAED pattern for the WO₃ nanoneedles.

The magnified cross-sectional TEM image in Fig. 8a shows that a dense particle layer was formed along the uneven surfaces of the Al₂O₃ substrate, indicating that the WO₃ nanoneedle film was formed on the dense particle layer.

Fig. 8: TEM analysis of the WO₃ nanoneedles.

a Magnified cross-sectional TEM image of the WO₃ nanoneedle. b HAADF-STEM image of the WO₃-nanoneedle film. c EDS maps for tungsten and aluminum on the WO₃ nanoneedle film.

HAADF-STEM images provide contrast based on factors that affect electron scattering, such as the sample density, sample thickness, and composition (atomic number) (Fig. 8b). Detailed morphologies of the dense particle layer and WO₃ nanoneedle film were observed. Tungsten was detected in both the dense particle layer and the WO₃ nanoneedle film in the EDS maps (Fig. 8c). The dense particle layer was composed of WO₃ particles. Aluminum was uniformly distributed on the Al₂O₃ substrate. A clear interface was observed between the dense particle layer and the Al₂O₃ substrate. No reactions occurred between the WO₃ and Al₂O₃.

The WO₃ nanoneedles were examined with high-resolution images, which showed that they had one-dimensional shapes (Fig. 9a). The nanoneedles exhibited lattice fringes perpendicular to their long directions (Fig. 9b). The magnified image clearly shows the lattice spaces of the {001} plane of WO₃ (Fig. 9c), which were perpendicular to the long directions of the nanoneedles. The distance between the yellow lines, which were the lattice spaces for {001} WO₃, was 0.38 nm. The fast Fourier transform (FFT) image in Fig. 9d shows clusters of white dots at the 1 and 7 o’clock positions. A lattice spacing of 0.380 nm was calculated from the FFT image, which was assigned to the {001} plane of WO₃. This indicated that the long direction of the WO₃ nanoneedles was arrayed in the stacking direction of the {001} planes of WO₃. That is, the long direction of the WO₃ nanoneedles was along the <001> direction, which was the c-axis of WO₃.

Fig. 9: High-resolution TEM analysis of the WO₃ nanoneedles.

High-resolution TEM image of a the WO₃ nanoneedles, b Magnified image of the demarcated area in (a). c Magnified image of the demarcated area in (b) showing the lattice spaces of the {001} plane of WO₃. d FFT image of (b) showing the lattice spaces of the {001} plane of WO₃.

The lengths of the nanobranches were calculated as follows. The long direction of the WO₃ nanoneedles in the TEM image shown in Fig. 10a was the 4 o’clock direction. The magnified image showed a clear lattice fringe (Fig. 10b). The FFT image showed white spots at the 1 o’clock position, perpendicular to the lengths of the WO₃ nanoneedles (Fig. 10c). A lattice spacing of 0.637 nm was calculated from the FFT image, which was assigned to the {010} plane of WO₃. This indicated that the direction perpendicular to the length of the WO₃ nanoneedles was <010>. The long direction of the WO₃ nanoneedle was <001>, which was the c-axis, whereas the long direction of the nanoneedle branches was <010>. The angle between the long direction of the WO₃ nanoneedles and the long direction of the nanobranches was 90°. This is consistent with the FE-SEM observations (Fig. 4d2) and the hexagonal crystal structure of WO₃.

Fig. 10: High-resolution TEM analysis of the WO₃ nanoneedles.

a High-resolution TEM image of the WO₃ nanoneedles. b Magnified image of the demarcated area in (a). c FFT image of (b) showing the lattice spaces of the {010} plane of WO₃.

The growth direction of the WO₃ nanoneedles was also evaluated. The WO₃ nanoneedles exhibited clear lattice fringes (Fig. 11a). The lattice spaces were calculated from the strong rings in the FFT images (Fig. 11b). The lattice spacings of 0.370, 0.376, 0.650, and 0.321 nm were assigned to the {001}, {001}, {010}, and {020} planes of WO₃, respectively. The <001> and <010> directions represented the long directions of the WO₃ nanoneedles and nanobranches, respectively. This was also consistent with our hypothesis regarding the growth directions and connection angles of the WO₃ nanoneedles and nanoneedle branches.

Fig. 11: High-resolution TEM image and FFT image of the WO₃ nanoneedles.

a High-resolution TEM image of the WO₃ nanoneedles. b FFT image of (a) showing the lattice spaces of WO₃.

Gas sensing performance and gas selectivity of the WO₃ nanoneedle film sensor

A WO₃ nanoneedle film was formed on an Al₂O₃ substrate for use as a gas sensor. The gas sensor was heated to 300 °C with a platinum heater (Fig. 12). Heating of the gas sensor improved its response and response and recovery times.

Fig. 12: Thermal image ofthe WO₃ nanoneedle.

Thermal image of the WO₃ nanoneedle film on the Al₂O₃ substrate with a platinum heater.

The gas sensor exhibited a resistance in air of R_a = 2.02 MΩ. The resistance decreased significantly to R_g = 0.10 MΩ in 10 ppmv acetone gas (Fig. 13a). The gas sensor exhibited a high response of R_a/R_g = 19.72 (Fig. 14). The gas sensor exhibited a fast response to acetone with T50 (res) <5 s and T90 (res) <5 s. The gas sensor also exhibited fast recoveries of T50 (rec) = 595 s and T90 (rec) = 2355 s. This gas sensor could be used to detect acetone, a marker gas for diabetes.

Fig. 13: Electrical resistances at 300 °C for 10 ppmv gases.

a Acetone, b H₂, c NH₃, d NO₂, e C₂H₅OH, f Toluene, g Acetaldehyde, h para-xylene, and i isoprene.

Fig. 14: Sensor signal responses.

Sensor signal responses to 10 ppmv NO₂, acetone, acetaldehyde, toluene, isoprene, C₂H₅OH, para-xylene, H₂, and NH₃.

The gas sensor exhibited high responses to various biogases. The sensor responses to acetaldehyde, toluene, isoprene, C₂H₅OH, para-xylene, and H₂ were 13.54, 9.49, 7.93, 6.33, 4.51, and 2.08, respectively (Fig. 14). Additionally, the gas sensor exhibited a high response to NO₂, which is an oxidizing gas; R_g was within the detection range. The response rates for NO₂ were 40–10 ppmv. Considering the width of the noise, the response may be several times greater than that at 40. In contrast, a low sensor response of 0.90 was obtained for NH₃. This indicated that acetaldehyde, toluene, isoprene, C₂H₅OH, para-xylene, H₂, and NO₂ was detected without the influence of coexisting NH₃ gas.

The gas sensor exhibited fast responses (T50 (res) < 5 s and T90 (res) < 5 s) to C₂H₅OH, H₂, and NO₂. The gas sensor exhibited fast responses (T50 (res) < 5 s and T90 (res) < 10 s) to acetaldehyde, toluene, isoprene, and paraxylene. The gas sensor exhibited fast recoveries (T50 (rec) < 10 s and T90 (rec) < 10 s) for H₂ and NO₂. The T50 (rec)/T90 (rec) recovery times for toluene, isoprene, para-xylene, C₂H₅OH, and acetaldehyde were 10 s/50 s, 35 s/145 s, 50 s/1990 s, 115 s/2210 s, and 410 s/2390 s, respectively.

High and fast sensor responses to various biogases were obtained due to the nano/microstructures of the WO₃ nanoneedle films. Numerous spaces were formed between the WO₃ nanoneedles, as shown in the TEM image in Fig. 8a. Sensors for detecting low gas concentrations require a large surface areas and a gas flow paths for the diffusion of gas molecules. The spaces in the WO₃ nanoneedle film led to large surface areas and provided gas flow paths for sensing gases with low concentrations. Additionally, the WO₃ nanoneedles contained nanobranches, which increased the surface area and number of adsorption sites for gas molecules. The WO₃ nanoneedles exhibited high electrical conductivity in the longitudinal direction owing to the absence of grain boundaries. Shape anisotropy and high electrical conductivity are beneficial for gas sensors that detect changes in resistance.

The nano/microstructure of the sensitive membrane in a gas sensor requires a high surface area, porosity, gas flow paths, and a suitable thickness and conductivity. In particular, gas molecules should flow to reach the interface between the sensitive membrane and substrate to provide high responses to low gas concentrations. The sensitive membrane should be porous to form flow paths for the gas molecules because they cannot flow through a dense membrane. Additionally, the membrane should be thin to ensure that the gas molecules reach the interface between the sensitive membrane and the substrate to obtain a strong sensor response. Thick and porous films are effective in detecting high gas concentrations because many molecules react with them. Thin and porous films are effective for detecting gases with low concentrations. Low-concentration gases react only on the surface of the thick porous film because of the small number of molecules. Only the surface of a thick, porous film contributes to the change in resistance. The internal areas of thick, porous films do not react with the gas and therefore do not contribute to the resistance change. Because the resistance changes only on the surface of the thick, porous film, the resistance change in the sensor element is small. Therefore, thick, porous films are unsuitable for detecting gases with low concentrations. However, almost all of a thin porous film contributes to the resistance change. Therefore, thin, porous films exhibit high sensor responses to low gas concentrations. The optimum film thickness depends on the gas concentration and porosity of the sensitive film. The molecular weight, adsorption capacity, oxidation reactions, and material properties of the sensitive film also determine the optimal film thickness.

Furthermore, the resistance of the sensor must be within the measurable ranges of small resistance meters. However, thin, porous films generally exhibit high resistances that cannot be measured with small resistance meters. In this study, the resistance of a thin, porous membrane was controlled to within the measurement range with the high electrical conductivity in the longitudinal directions of the WO₃ nanoneedles. The WO₃ nanoneedle film was effective in measuring low-concentration gases.

Low-concentration gas sensing and limit of detection (LOD) for the WO₃ nanoneedle film sensor

The sensor responses increased in the order 10 ppmv NO₂, acetone, acetaldehyde, toluene, isoprene, C₂H₅OH, para-xylene, H₂, and NH₃ (Fig. 14). In particular, the sensor exhibited strong responses to NO₂ and acetone. The performance of the sensor was evaluated with low concentrations of NO₂ and acetone.

The sensor exhibited responses of 1.049, 1.218, 1.491, 2.043, 2.392, 3.591, 5.713, 6.777, 7.418, and 7.656 to 25, 50, 75, 100, 125, 250, 500, 750, 1000, and 1250 ppbv acetone gas, respectively (Fig. 15a, b). The approximate straight line calculated over the range 25–125 ppbv is represented in Eq. 13 (Fig. 15c).

$${\rm{y}}=14.044{\rm{x}}-0.5855$$

(13)

Fig. 15: Sensing performance for acetone gas.

a Electrical resistances to 25–1.25 ppmv acetone at 300 °C. b Responses to 25–1.25 ppmv acetone. c Responses to 25–125 ppbv acetone.

The LOD is a key figure of merit in trace analyses. The analytical signal is clearly differentiated from noise at the LOD concentration. According to the IUPAC definition, the LOD of a sensor is the concentration at which the noise level is less than one-third of the signal level⁴⁵. The theoretical LOD can be calculated with Eq. (14)⁴⁵. The LOD is the analyte concentration that produces a response that is threefold higher than the noise level of the device⁴⁵.

$${\rm{LOD}}=3\times {\rm{RMS}}/{\rm{S}}$$

(14)

where RMS is the standard root-mean-square deviation of the device noise and S is the analytical sensitivity. The RMS value of the sensor resistance in air was calculated from 100 data points.

An LOD of 2.4 ppbv was estimated for acetone gas from an RMS of 0.01106 and S of 14.044. Acetone is a well-known diabetes marker. There is a strong demand in the health care field for gas sensors that can monitor trace amounts of acetone in the exhaled breath and skin gases. The WO₃ nanoneedle sensor film can be used to detect low concentrations of acetone in skin gas and to distinguish the acetone levels of patients with diabetes (188 ppbv) from that of healthy individuals (87 ppb)v⁵. This sensor exhibited detection of trace amounts of acetone in health care applications.

The sensing capabilities were evaluated with trace amounts of NO₂. The sensor responses were 3.93, 12.06, 17.83, 25.36, and 29.24 to 0.25, 0.5, 0.75, 1, and 1.25 ppmv NO₂ gas (Fig. 16a). The approximate straight line calculated over the range 0.25–1.25 ppmv is shown in Eq. 15 (Fig. 16b).

$${\rm{y}}=25.568{\rm{x}}-1.492$$

(15)

Fig. 16: Sensing performance for NO₂.

a Electrical resistances for 250–1.25 ppmv NO₂ at 300 °C. b Response to 250–1.25 ppmv NO₂.

An LOD of 1.5 ppbv was estimated from an RMS of 0.01291 and S of 25.568 for NO₂ gas. This indicated that the sensor detected low concentrations of NO₂ gas. The sensor can be used to sense low concentrations of NO₂ gas in biogases, such as in exhaled breath and skin gases, as well as gases in living and mobile spaces. NO₂ is dangerous, and there is a need to detect minute amounts. Additionally, the detection of NO₂ gas is useful in monitoring asthma because NO is a marker for asthma and is easily oxidized to NO₂ gas.

Comparison of the sensing performance for NO₂ or acetone

Various WO₃ nanostructures have been developed for sensing NO₂ (Table 1)^{28,29,30,31,32,33,34,35} or acetone (Table 2)^{36,37,38,39,40,42,43,44}. The WO₃ nanostructures exhibited high performance in the detection of low-concentration gases. Moreover, the WO₃ nanoneedle film exhibited responses of 25.36 and 7.418 to 1 ppmv NO₂ and acetone, respectively. The WO₃ nanoneedle film exhibited a higher response than previously reported films.

Table 1 Comparison of the NO₂ sensing capability of the WO₃ nanoneedle film with those of previously reported WO₃-based NO₂ gas sensors.

Table 2 Comparison of the acetone sensing capability of the WO₃ nanoneedle film with those of previously reported WO₃-based acetone gas sensors.

The concentrations of NO₂ (Table 1) or acetone (Table 2) differed in each study. The responses (R_a/R_g) for each sensor to 0 ppmv NO₂ and acetone were 1, indicating no response. The responses to 1 ppmv NO₂ and acetone gas were calculated for simple comparison. For a more accurate comparison, the responses were compared for the same gas concentration. Additionally, it was necessary to perform evaluations with the same measuring device. Measurement method standardization is also required for gas sensors used in detecting low gas concentrations.

Sensing mechanism for the WO₃ nanoneedle film sensor

In air, oxygen is adsorbed on the surfaces of n-type semiconductor particles such as WO₃, SnO₂, or ZnO. Electrons in the n-type semiconductors are transferred to the adsorbed oxygen atoms, and an electron depletion layer is formed on the surfaces of the particles. The electron depletion layer has a low electron density and electrical conductivity. The electrical resistance of the particulate film increases due to formation of the electron-depletion layer at the grain boundaries. For combustible gases such as volatile organic compounds, the combustible gas molecules and adsorbed oxygen on the particle surface undergo oxidation reactions. The adsorbed oxygen is consumed by the oxidation reaction. Electrons trapped in the adsorbed oxygen return to the particles, reducing the thickness of the electron depletion layer. The electrical resistance of the gas sensor decreases with increasing electrical conductivity between the particles. The sensing mechanism for the semiconductor gas sensor is the change in the electrical resistance of the particulate film owing to the oxidation reaction between the combustible gas and the adsorbed oxygen. The sensing mechanisms were reviewed in a previously published article⁷.

The WO₃ nanoneedle film is very effective as a gas sensor because the nanoneedles were approximately 100 nm long and 10 nm wide and the electron depletion layer was several nanometers thick. Therefore, the entire nanoneedle became an electron-depletion layer. The electrical resistance of the electron-depletion layer was changed after oxidation of the combustible gas by the adsorbed oxygen; however, the electrical resistance inside the particles did not change, which is problematic for the detection of low gas concentrations. However, the electrical resistance of the nanoneedles changed drastically because almost all the nanoneedles became electron-depletion layers.

Furthermore, it is necessary to diffuse the gas over the entire sensitive film to detect low concentrations. Therefore, porous thin films are effective in the detection of low gas concentrations. However, low-conductivity semiconductor particles increase the electrical resistance of the porous thin films. A large measuring instrument is needed to measure the high electrical resistance, which is problematic for miniature gas sensors. In contrast, nanoneedles have one-dimensional structures and exhibit high conductivities in the longitudinal direction of the particles. The highly conductive nanoneedles contribute to both the high porosity and conductivity. Therefore, nanoneedles are effective for detecting gases at low concentrations.

Notably, the WO₃ nanoneedle film sensor exhibited high responses to both NO₂ and acetone. NO₂ is an oxidizing and combustion-supporting gas. Combustion-supporting gases, such as NO₂, oxygen, air, chlorine, fluorine, and nitrogen monoxide, provide oxygen for combustion of combustible substances. Acetone is a reducing and combustible gas. Reducing gases such as acetone, H₂, carbon monoxide, hydrogen sulfide, and sulfur dioxide deplete oxygen. Combustible gases, such as acetone, H₂, methane, propane, and isobutane, burn in air or oxygen. The high responses to NO₂ and acetone, which have very different properties, is a significant characteristic of the WO₃ nanoneedles.

Moreover, the gas selectivities of metal-oxide gas sensors differ depending on the metal ions and their valences. Therefore, various metal oxide-based gas sensors have been developed. It is also known that different nanostructures present different gas selectivities because they exhibit different specific surface areas, crystal planes, and crystal defects. These features affect the gas selectivities and responses substantially, which has led to the development of various nanostructures for use in gas sensors. The high responses to both NO₂ and acetone is a major feature of the WO₃ nanoneedle film sensor, which distinguishes it from other materials. The metal type, metal valence, high specific surface area, crystal planes, and crystal defects contribute to the high responses to both NO₂ and acetone.

Conclusions

A highly responsive diabetes sensor was developed in which a WO₃ nanoneedle film was used to detect low concentrations of acetone gas. The WO₃ nanoneedle film was formed by heating an aqueous solution containing tungstic acid, CH₃CN, and HCl at 70 °C for 7 d. The WO₃ nanoneedle film was obtained by precisely controlling the crystallization process. The sensor exhibited a high sensor response of R_a/R_g = 19.72 to 10 ppmv acetone gas, demonstrating the ability to detect low concentrations of acetone. Furthermore, the sensor exhibited a high sensor response of R_g/R_a = 25.36 to 1 ppmv NO₂, which is a marker for asthma. The sensor responses to the lung cancer marker gases acetaldehyde and toluene were 13.54 and 9.49, respectively. The sensor responses to 10 ppmv isoprene, C₂H₅OH, para-xylene, H₂, and NH₃ were 7.93, 6.33, 4.51, 2.08, and 0.90, respectively. The WO₃ nanoneedle film sensor successfully detected 25 ppbv acetone and 250 ppbv NO₂. LODs of 2.4 and 1.5 ppbv were estimated for acetone and NO₂, respectively. This sensor detects low gas concentrations and is selective for acetone and NO₂. Biogases such as those in exhaled breath and skin gas are detected by the sensor, which will contribute to early screening of diabetes, asthma, and lung cancer.