1-Butyl-3-Methylimidazolium Tetrafluoroborate Film as a Highly Selective Sensing Material for Non-Invasive Detection of Acetone Using a Quartz Crystal Microbalance

Breath acetone serves as a biomarker for diabetes. This article reports 1-butyl-3-methylimidazolium tetrafluoroborate ([bmim][BF4]), a type of room temperature ionic liquid (RTIL), as a selective sensing material for acetone. The RTIL sensing layer was coated on a quartz crystal microbalance (QCM) for detection. The sensing mechanism is based on a decrease in viscosity and density of the [bmim][BF4] film due to the solubilization of acetone leading to a positive frequency shift in the QCM. Acetone was detected with a linear range from 7.05 to 750 ppmv. Sensitivity and limit of detection were found to be 3.49 Hz/ppmv and 5.0 ppmv, respectively. The [bmim][BF4]-modified QCM sensor demonstrated anti-interference ability to commonly found volatile organic compounds in breath, e.g., isoprene, 1,2-pentadiene, d-limonene, and dl-limonene. This technology is useful for applications in non-invasive early diabetic diagnosis.


Introduction
Human breath contains a large number of volatile organic compounds (VOCs) derived from a metabolic origin [1,2]. Concentration of VOCs is related to the degree of oxidative damage inside the body. VOCs commonly found in exhaled breath include alkanes, alkenes, aldehydes, ketones, sulfides, nitrides, etc. [3]. There are about 340 types of VOCs from a normal individual's breath. However, VOC composition in breath varies largely from person to person [3]. Breath testing for VOCs is intrinsically safe and noninvasive. Hence, breath VOCs find lots of applications in early disease diagnosis. It has been reported that alveolar-arteriole C 4 -C 20 alkane gradients were more abundant in patients with lung and breast cancer than those in a normal person [4]. The mean concentrations of pentane and carbon disulfide are significantly higher in the breath of patients with schizophrenia than in a normal person [5]. The nitric oxide level in the exhaled breath of a patient with asthma and community-acquired pneumonia is higher than that in healthy subjects [6,7]. 9 MHz AT-cut quartz crystals (Chen Jing Electronics, Beijing, China) were used as the QCM sensor in this work. The diameter of the quartz crystal was 12.5 mm, and the gold electrode diameter was 6.0 mm. The quartz crystal was fixed to a glass tube using silicone rubber, and only one side of the crystal was coated with film for sensing. Then, 10 µL of [bmim] [BF 4 ] in 10% acetonitrile was dropped onto the center of the QCM device and it was then loaded on a spin coater (Laurell Technology, North Wales, PA, USA). It was rotated at a speed of 500 rpm for 30 s, and 2000 rpm for 2 min. The film was then blown and dried with N 2 . Excess RTILs on the edge of the quartz wafer were wiped off by acetonitrile soaked cotton. The acetonitrile impurity introduced in this process was removed by purging with dry N 2 . Quartz crystal could be recovered by rinsing with acetonitrile.

Sample Handling Setup and Quantification of Acetone Vapor Concentration
Static mode was adopted for the sample handling system. The modified QCM was placed in a custom-made sealed glass chamber of 273 mL. A fixed amount of pure acetone solution was injected into the chamber, followed by its natural evaporation. The acetone sample is measured when the chamber is full of acetone vapor. The acetone concentration (C a ) in the cell was calculated by: where ρ a and V a are the relative density and sample volume of acetone. V cell is the chamber volume, which is 35 mL as designed in this work. The relative density of acetone is 0.79 g/mL. Without loss of generality, the acetone vapor concentration is converted into parts per million (ppm) in this work. The conversion formula is: where M a is the molar mass of acetone, T is the chamber temperature, and P is pressure. Under room temperature (22 • C) and normal atmospheric pressure conditions, Equation (2) can be simplified to:

Viscoelasticity Effect of [bmim][BF4] Film
Most films behave neither as a rigid layer nor as an ideally viscous (Newtonian) membrane and therefore have to be treated as viscoelastic films [30]. The viscoelasticity of a film is evaluated by measuring the conductance using the PQCI technique [31]. Therefore, we investigated in detail if there was any difference in half bandwidth of the conductance spectra between the bare and modified electrodes. The conductance (G) at half peak height and half bandwidth of the conductance spectrum ∆ / can be calculated as [32]: where 1 2 HG f and 1 2 LG f are the higher and lower frequency at half peak height (G = Gmax/2) in the conductance spectrum, respectively; and R1 and L1 are the motional resistance and the motional inductance, respectively. f0 is the resonant frequency and Q is the quality factor. Figure 1 shows the conductance spectrum for the bare and modified electrodes. The calculated Gmax and

Viscoelasticity Effect of [bmim][BF 4 ] Film
Most films behave neither as a rigid layer nor as an ideally viscous (Newtonian) membrane and therefore have to be treated as viscoelastic films [30]. The viscoelasticity of a film is evaluated by measuring the conductance using the PQCI technique [31]. Therefore, we investigated in detail if there was any difference in half bandwidth of the conductance spectra between the bare and modified electrodes. The conductance (G) at half peak height and half bandwidth of the conductance spectrum ∆ f G 1/2 can be calculated as [32]: where f HG 1/2 and f LG 1/2 are the higher and lower frequency at half peak height (G = G max /2) in the conductance spectrum, respectively; and R 1 and L 1 are the motional resistance and the motional inductance, respectively. f 0 is the resonant frequency and Q is the quality factor. Figure 1 shows the conductance spectrum for the bare and modified electrodes. The calculated G max and ∆ f G 1/2 values are given in Table 1. In Figure 1, plots a and b show the conductance spectrum corresponding to the bare quartz crystal and the [bmim] [BF 4 ] modified piezoelectric quartz crystal in air, respectively. Appreciable change in both peak height and half bandwidth of the conductance  4 ] film is nearly viscoelastic, not rigid, which agrees with the report [26].
conductance spectrum can be found. It can be concluded that the [bmim][BF4] film is nearly viscoelastic, not rigid, which agrees with the report [26].  The absolute value of Δf0/ΔR1 is conventionally used to reflect whether frequency shift is dominated by mass loading or by the viscosity-density effect of the solution near the interface. If the value is more than 10 Hz/Ω for 9 MHz quartz crystal, the frequency shift is dominated by mass loading [33]. Otherwise, the frequency shift is dominated by the viscosity-density effect of the solution near the interface. From Table 1, the absolute value of Δf0/ΔR1 is 123 Hz/Ω, much greater than 10 Hz/Ω. This indicates that the frequency shift due to the deposition of the [bmim][BF4] film on the QCM device is dominated by mass loading, and the viscosity-density effect of the solution near the interface can be negligible.

Response Curve on the [bmim][BF4]-Modified QCM Sensor
Commonly, there are two types of sample handling systems: dynamic and static. In this work, the static system was adopted. The modified-QCM sensor was placed in a sealed glass chamber with a specific volume. A typical [bmim][BF4]-modified QCM sensor response curve under exposure to different acetone concentrations is shown in Figure 2. Noticeably, the resonance frequency of the [bmim][BF4]-modified QCM device increased sharply when in contact with acetone vapor. The response time was measured at less than 10 s. The positive response, whose value is 1355 Hz for 329 ppmv acetone, is opposite to that of conventional QCM devices based on the Sauerbrey equation. If all 329 ppmv acetone were adsorbed on a QCM sensor coated with a rigid film, the corresponding negative frequency shift would be less than 13.3 Hz based on the Sauerbrey equation. The principle is based on the detection of mass loading, where the adsorption of gas to a rigid (or solid) polymeric or metallic film decreased the resonance frequency of the quartz crystal. Since the [bmim][BF4] film has been shown to be nearly viscoelastic, not rigid, we cannot use the Sauerbrey equation to explain our results.  The absolute value of ∆f 0 /∆R 1 is conventionally used to reflect whether frequency shift is dominated by mass loading or by the viscosity-density effect of the solution near the interface. If the value is more than 10 Hz/Ω for 9 MHz quartz crystal, the frequency shift is dominated by mass loading [33]. Otherwise, the frequency shift is dominated by the viscosity-density effect of the solution near the interface. From Table 1, the absolute value of ∆f 0 /∆R 1 is 123 Hz/Ω, much greater than 10 Hz/Ω. This indicates that the frequency shift due to the deposition of the [bmim][BF 4 ] film on the QCM device is dominated by mass loading, and the viscosity-density effect of the solution near the interface can be negligible.

Response Curve on the [bmim][BF 4 ]-Modified QCM Sensor
Commonly, there are two types of sample handling systems: dynamic and static. In this work, the static system was adopted. The modified-QCM sensor was placed in a sealed glass chamber with a specific volume. A typical [bmim][BF 4 ]-modified QCM sensor response curve under exposure to different acetone concentrations is shown in Figure 2. Noticeably, the resonance frequency of the [bmim][BF 4 ]-modified QCM device increased sharply when in contact with acetone vapor. The response time was measured at less than 10 s. The positive response, whose value is 1355 Hz for 329 ppmv acetone, is opposite to that of conventional QCM devices based on the Sauerbrey equation. If all 329 ppmv acetone were adsorbed on a QCM sensor coated with a rigid film, the corresponding negative frequency shift would be less than 13 The frequency shift of the viscoelastic film-modified QCM sensor results from the change in the physicochemical properties of the surrounding medium primarily is attributed to the film density and viscosity. The dependence of frequency shift (Δf) on medium density and viscosity can be derived from Kanazawa and Gordon's work [34] and is shown below: where f0 is the resonance frequency of the bare QCM sensor, and ρL and ηL are the absolute density and viscosity of the medium (or the film), respectively. µQ and ρQ are the elastic modulus and density of quartz, respectively. The solubility of organic vapor in imidazolium-based RTILs are known to be higher as compared with those of polymers [26]. The dissolution of organic vapor in RTILs can result in changes to its physicochemical properties [34][35][36][37]. In this work, adsorption of acetone vapor onto [bmim][BF4] film caused a reduction in viscosity and density of the film. The viscosity for pure acetone and [bmim][BF4] are 0.31 and 136.90 mPas, respectively. For the RTILs system, the reduction of viscosity is more pronounced than that of density after the addition of low-viscosity organic compounds [37,38]. The reason may be that entrapped low-viscosity organic compounds decrease ion-pairing or ion aggregation of RTILs by solvating interacting anions and cations. Therefore, the observed large and positive frequency shift (1355 Hz) must be related to a decrease in viscosity and density of the [bmim][BF4] film upon adsorption of acetone vapor. The limit of detection is 5.0 ppmv, which is much lower than 77.6 ppmv and a little higher than 1.2 ppmv reported in previous works [39,40]. We define LOD as the lowest acetone concentration in air that can be measured with a signal to noise ratio of 3:1. For diabetic patients under different conditions, mean acetone concentration is found from 2.05 to 5.58 ppmv, while in healthy individuals, the value is in the range of 0.48 to 0.51 ppmv. Therefore in a real situation, a 10-fold pre-concentration value is needed. Meanwhile, the large response sensitivity is higher than previously reported values, such as 8.58 × 10 −3 Hz/ppmv [26] and 0.52 Hz/ppmv [27]. The frequency shift of the viscoelastic film-modified QCM sensor results from the change in the physicochemical properties of the surrounding medium primarily is attributed to the film density and viscosity. The dependence of frequency shift (∆f ) on medium density and viscosity can be derived from Kanazawa and Gordon's work [34] and is shown below:

Calibration Curve
where f 0 is the resonance frequency of the bare QCM sensor, and ρ L and η L are the absolute density and viscosity of the medium (or the film), respectively. µ Q and ρ Q are the elastic modulus and density of quartz, respectively. The solubility of organic vapor in imidazolium-based RTILs are known to be higher as compared with those of polymers [26]. The dissolution of organic vapor in RTILs can result in changes to its physicochemical properties [34][35][36][37]. In this work, adsorption of acetone vapor onto [bmim][BF 4 ] film caused a reduction in viscosity and density of the film. The viscosity for pure acetone and [bmim][BF 4 ] are 0.31 and 136.90 mPa·s, respectively. For the RTILs system, the reduction of viscosity is more pronounced than that of density after the addition of low-viscosity organic compounds [37,38]. The reason may be that entrapped low-viscosity organic compounds decrease ion-pairing or ion aggregation of RTILs by solvating interacting anions and cations. Therefore, the observed large and positive frequency shift (1355 Hz) must be related to a decrease in viscosity and density of the [bmim][BF 4 ] film upon adsorption of acetone vapor. Figure 3 shows measurements of different acetone concentrations ranging from 0 to 1410 ppmv. A linear response range from 7.05 ppmv to 705 ppmv was found. The regression equation is: ∆f (Hz) = 3.49C acetone (ppmv) + 5.29, with a correlation coefficient of 0.990. The slope of this linear zone is 3.49 Hz/ppmv, which could be viewed as the sensitivity of the [bmim][BF 4 ] modified-QCM sensor. The limit of detection is 5.0 ppmv, which is much lower than 77.6 ppmv and a little higher than 1.2 ppmv reported in previous works [39,40]. We define LOD as the lowest acetone concentration in air that can be measured with a signal to noise ratio of 3:1. For diabetic patients under different conditions, mean acetone concentration is found from 2.05 to 5.58 ppmv, while in healthy individuals, the value is in the range of 0.48 to 0.51 ppmv. Therefore in a real situation, a 10-fold pre-concentration value is needed. Meanwhile, the large response sensitivity is higher than previously reported values, such as 8.58 × 10 −3 Hz/ppmv [26] and 0.52 Hz/ppmv [27].

Effect of Humidity on the [bmim][BF4]-Modified QCM Sensor
Water vapor is a common interference to most sensors, and is especially important for gas sensors. Since the saturated water vapor concentration is 19.  In this work, all measurements were performed with a relative humidity of 50% to 60% at a stable room temperature of 22 ± 1 °C. It means that the highest water vapor concentration did not exceed 11.58 ppmv. Therefore, the reported results do not need to be corrected. The effect of humidity will be considered for future iterations under highly humid environments, such as exhaled

Effect of Humidity on the [bmim][BF 4 ]-Modified QCM Sensor
Water vapor is a common interference to most sensors, and is especially important for gas sensors. Since the saturated water vapor concentration is 19.

Effect of Humidity on the [bmim][BF4]-Modified QCM Sensor
Water vapor is a common interference to most sensors, and is especially important for gas sensors. Since the saturated water vapor concentration is 19.  In this work, all measurements were performed with a relative humidity of 50% to 60% at a stable room temperature of 22 ± 1 °C. It means that the highest water vapor concentration did not exceed 11.58 ppmv. Therefore, the reported results do not need to be corrected. The effect of humidity will be considered for future iterations under highly humid environments, such as exhaled gas in human breath. In that case, the background response will be measured and subtracted for  In this work, all measurements were performed with a relative humidity of 50% to 60% at a stable room temperature of 22 ± 1 • C. It means that the highest water vapor concentration did not exceed 11.58 ppmv. Therefore, the reported results do not need to be corrected. The effect of humidity will be considered for future iterations under highly humid environments, such as exhaled gas in human breath. In that case, the background response will be measured and subtracted for more accurate sensing results.
In real sample testing, there are at least two efficient routes to decrease the water vapor concentration. One is to pre-concentrate sample through the drying unit. Another route to address this issue is to entrap a useful additive into the RTIL sensing layer, such as chromium (III) based MIL-101, which is reported in Ref. [41].

Selectivity
The most abundant VOCs with positive alveolar gradients in human breath are as follows: isoprene (48.60%), 1,2-pentadiene (15.00%), acetone (14.59%), d-limonene (8.43%), and dl-limonene (2.31%) [42]. In Table 2, a small response was observed in the three VOCs: 1,2-pentadiene, d-limonene and dl-limonene. The response values in Table 2 were comparable to background noise (around 13.0 Hz). However, the response to isoprene (18.1 Hz) is a little bit higher than the background noise, which is not comparable to that of 9.25 ppmv acetone (38.1 Hz). We also determined 9.25 ppmv of acetone in the presence of mixed VOC gases with comparable concentrations: isoprene (29.98 ppmv), 1,2-pentadiene (9.25 ppmv), d-limonene (5.2 ppmv), and dl-limonene (1.43 ppmv). The response of the [bmim]BF 4 -modified QCM sensor to mixed VOC gases in the absence of acetone were also tested as a control. Results are shown in Figure 5. There is a 104.0 Hz response for acetone in the presence of mixed VOC gases and 46.0 Hz for only mixed VOC gases. Therefore, we can determine acetone in the mixed VOC gases. This demonstrates that the [bmim][BF 4 ] film presents anti-interference ability to the four VOCs to some extent. The reason may be that: the viscosity of isoprene, 1,2-pentadiene, d-limonene, and dl-limonene are 0.23, 0.25, 0.92, and 0.88 mPa·s, respectively. Only limonene has a higher viscosity value than that of acetone (0.31 mPa·s). Meanwhile, there is a carbonyl group on acetone molecules (CH 3 COCH 3 ), while there is no such functional group or oxygen atom in the other VOCs (C 5 H 8 , C 10 H 20  In real sample testing, there are at least two efficient routes to decrease the water vapor concentration. One is to pre-concentrate sample through the drying unit. Another route to address this issue is to entrap a useful additive into the RTIL sensing layer, such as chromium (III) based MIL-101, which is reported in Ref. [41].

Selectivity
The most abundant VOCs with positive alveolar gradients in human breath are as follows: isoprene (48.60%), 1,2-pentadiene (15.00%), acetone (14.59%), d-limonene (8.43%), and dl-limonene (2.31%) [42]. In Table 2, a small response was observed in the three VOCs: 1,2-pentadiene, d-limonene and dl-limonene. The response values in Table 2 were comparable to background noise (around 13.0 Hz). However, the response to isoprene (18.1 Hz) is a little bit higher than the background noise, which is not comparable to that of 9.25 ppmv acetone (38.1 Hz). We also determined 9.25 ppmv of acetone in the presence of mixed VOC gases with comparable concentrations: isoprene (29.98 ppmv), 1,2-pentadiene (9.25 ppmv), d-limonene (5.2 ppmv), and dl-limonene (1.43 ppmv). The response of the [bmim]BF4-modified QCM sensor to mixed VOC gases in the absence of acetone were also tested as a control. Results are shown in Figure 5. There is a 104.0 Hz response for acetone in the presence of mixed VOC gases and 46.0 Hz for only mixed VOC gases. Therefore, we can determine acetone in the mixed VOC gases. This demonstrates that the [bmim][BF4] film presents anti-interference ability to the four VOCs to some extent. The reason may be that: the viscosity of isoprene, 1,2-pentadiene, d-limonene, and dl-limonene are 0.23, 0.25, 0.92, and 0.88 mPas, respectively. Only limonene has a higher viscosity value than that of acetone (0.31 mPas). Meanwhile, there is a carbonyl group on acetone molecules (CH3COCH3), while there is no such functional group or oxygen atom in the other VOCs (C5H8, C10H20). So the other organic interferents are non-polarizable molecules, while acetone is. The [bmim][BF4] film is a polarizable compound, which prefers binding with partially polarized acetone according to the principle that similarly polarized substances are more likely to be dissolved by each other. Therefore, the decrease of viscosity would be more pronounced for the acetone-     3.6. Comparison of QCM and Gas Chromatography Measurements Figure 6 shows the chromatograms of 470 ppmv acetone. An obvious peak was observed in the chromatograms at a retention time of 6.05 min. The same sample was also added into the gas chamber, and then tested by the [ Figure 6 shows the chromatograms of 470 ppmv acetone. An obvious peak was observed in the chromatograms at a retention time of 6.05 min. The same sample was also added into the gas chamber, and then tested by the [bmim][BF4]-modified QCM sensor. The response frequency shift was about 1425 Hz. From the regression equation, the acetone concentration obtained from the QCM method was about 407 ppmv. This result is comparable to the value determined by the GC technique. Therefore, the QCM method can be a complementary method to the GC technique. In addition, the presence of water vapor is an unfavorable factor in the detection of VOCs by the GC technique, while the [bmim][BF4]-modified QCM sensor is not subject to this restriction; it allows trace water to be present.

Conclusions
A 9 MHz QCM sensor has been successfully developed, evaluated, and tested for acetone vapor employing room temperature ionic liquids ([bmim][BF4]) as sensing materials. The [bmim][BF4] film has been proven to be viscoelastic, which relates the response of the modified QCM sensor to the viscosity and density change of the film. The [bmim][BF4] modified-QCM sensor demonstrated a linear range from 7.5 to 705 ppmv, with a fast response to acetone vapors because of the fast diffusion of analytes in RTILs. Our preliminary results will stimulate further research towards developing noninvasive, inexpensive, portable, and compact sensors for early diabetes diagnosis.