Hydrophobic Metal Organic Framework Enhanced Acoustic Wave Formaldehyde Sensor Based on Polyethyleneimine and Bacterial Cellulose Nanolms

A surface acoustic wave (SAW) formaldehyde gas sensor was fabricated on a 42°75' ST-cut quartz substrate, with a composite sensing layer of zeolitic imidazolate framework (ZIF)-8 on polyethyleneimine (PEI)/ bacterial cellulose (BC) nanolms. The addition of snowake-like ZIF-8 structure on the PEI/BC sensitive lm signicantly improves the hydrophobicity of the SAW sensor and increases the sensor's sensitivity to formaldehyde gas. It also signicantly increases the surface roughness of the sensitive lm. Its hydrophobic nature prevents water molecules from entering into the internal pores of the BC lm, thereby avoiding signicant mass loading caused by the humidity change when the sensor is used to detect low-concentration formaldehyde gas. The Zn 2+ sites at the surface of ZIF-8 improves the sensor's response to formaldehyde gas through enhancing the physical adsorptions. Experimental results show that the ZIF-8@PEI/BC SAW sensor has a response (e.g., frequency shift) of 40.3 kHz to 10 ppm formaldehyde gas at 25 ℃ and 30% RH. When the relative humidity was increased from 30% to 93%, the response (frequency shift) of the sensor drifts only ~5%, and there is negligible drift at a medium humidity level (~56% RH).


Introduction
With the rapid growth of economy and increasing demand for comfortable living environment, decoration materials are extensively used in the built environment, which brings complex and diverse pollutants including formaldehyde into the indoor environment [1,2]. Among various types of air pollutants, formaldehyde has received signi cant attention because of its wide usage and toxicity. Being a reactive compound, formaldehyde can damage proteins, cause genetic mutations, DNA single-strand internal cross-linking and DNA-protein cross-linking, and inhibit DNA damage repair, etc. [3]. Long-term exposure to low-dose formaldehyde can cause chronic respiratory diseases, nasopharyngeal cancer, brain tumors and other diseases [4,5]. Therefore, a timely and precise detection of formaldehyde concentration is particularly important.
Many types of formaldehyde detection methods, such as spectrophotometry, chromatography, and uorescence spectroscopy, have been proposed. [6][7][8]. However, they are generally expensive, and often require a long cycle of sampling and analysis operated by professionals. The measurement results using these methods are usually the mean values over a period of time, which does not re ect the formaldehyde concentration in real time [9].
Up to now, the research of formaldehyde gas sensor is mostly focused on the development of detection techniques, which require both high sensitivity and short detection period, in the air in real time [10]. For example, Bouchikhi et al. developed a metal oxide thin lm-based sensor by vapor deposition of tungsten trioxide (WO 3 ) nanowires (NWs) and metal nanoparticles modi ed WO 3 NWs gas sensing layer on interdigital platinum electrode, and the sensor showed a high sensitivity for formaldehyde gas under both dark and ultraviolet light irradiation conditions [11]. Yin et al developed a polymer thin lm sensor by applying a ower-like compound with a heterostructure based on Sn 3 O 4 and reduced graphene oxide (rGO) to achieve a wide detection range of formaldehyde gas [12]. These types of sensors have the advantages of high sensitivity and low detection limit. However, they often suffer from the severe interferences of temperature and humidity changes. Up to now, there are not many studies to minimize the interferences of temperature and humidity. For example, Wang et al. proposed a formaldehyde gas sensor based on Cu-doped Sn 3 O 4 nano owers [13]. Its response to 100 ppm formaldehyde is 53 (the ratio of the resistances of the sensor in dry air and the gaseous environment), with a detection limit of 1 ppm, but the changes of humidity have a signi cant impact on the detection results. An offset of about 10% in response is reported between 25% and 75% RH environments [13]. Zeng et al. prepared a La 2 O 3 -In 2 O 3 and nanotube sensor using an electrospinning method, and the sensor has a response value of 101.9 to 50 ppm formaldehyde gas. They reported that when the RH value was lower than 60%, the sensor's response to formaldehyde gas was relatively stable, but when the RH value was higher than 60%, the sensor's response to formaldehyde gas was decreased sharply with the increase of RH values [14].
Previously, we developed a surface acoustic wave (SAW) formaldehyde gas sensor based on a bi-layer nano lm of bacterial cellulose (BC) and polyethyleneimine (PEI) [15]. The BC nano layer signi cantly improves the sensitivity of the PEI lm and reduces the response and recovery time for the low concentrations of formaldehyde. The sensor has a frequency shift of 35.6 kHz to 10 ppm formaldehyde gas at room temperature and 30% relative humidity (RH), with both good selectivity and stability. The sensor uses a ST-cut quartz substrate which has a low temperature coe cient of frequency (TCF).
However, it shows a poor performance with the change of humidity, because the amine groups of PEI and the hydroxyl groups of BC have strong adsorptions of H 2 O molecules [16,17], thus causing a signi cant mass loading effect of the SAW sensor. One way to solve this problem is to install a humidity sensor next to this formaldehyde SAW sensor which can quantitatively detect humidity to correct the SAW sensor output through o ine data analysis. However, this is an indirect compensation increasing the complexity, size and production cost of the formaldehyde sensor. Therefore, under the premise of maintaining the PEI/BC sensing lm's sensing performance for formaldehyde gas, solving the problem of its high sensitivity to RH values becomes our key research topic to improve the performance of the formaldehyde gas sensor.
Changes of hydrophobicity is currently a research hotspot in the eld of functional materials [18]. It is also important in the eld of gas sensing, and a high hydrophobic layer on the top of the sensor can prevent the sensor from interfering with the changes of environmental humidity [19]. The hydrophobic layer can prevent water molecules from easily entering the sensitive lm, thereby reducing the frequency shift of the mass loading caused by the water molecules in the formaldehyde gas detection process. Recently, researchers have applied various methods to improve the hydrophobicity of sensors. For example, Lee et al. used an ultra ltration method to successfully exchange water-based poly-(3,4ethylenedioxythiophene) polystyrene sulfonate (PEDOT: PSS) solution to organic solvent-based PEDOT: PSS solution, which was applied as a coating on a pressure sensor to increase the contact angle of water droplets and avoid the in uence of humidity changes to sensors [20]. Chen et al. used a hydrophobic cysteine-sensitized Cu 2 O(Cu(I)-Cys) nanocomposite as a sensing layer to fabricate a quartz microbalance gas sensor, which could determine hexanal and 1-octen-3-ol at room temperature, with a good hydrophobicity, sensitivity and selectivity [21].
Zeolite-based imidazole salt frameworks (ZIFs) are a type of metal organic frameworks (MOFs), which are porous crystalline materials formed by continuous and periodic connection of transition metal ions and imidazole-based organic linkers [22,23]. Due to its high porosity, thermal/chemical stability, surface functionality and diverse synthesis methods, ZIFs have been widely used in various elds, including gas storage, catalysis, and preparation of various nanostructures [24][25][26]. Li et al [27] and Yogapriya et al. [28] reported that the ZIFs have good hydrophobic properties (high water contact angle), and its hydrophobic properties can be further improved by compounding with other materials such as polyvinylidene uoride or porous uorinated graphene. The ZIF-8 is regarded as having the best hydrophobic properties among all the ZIF materials [29]. In this study, we proposed to use MOF ZIF-8 structure to improve the humidity insensitivity of the PEI/BC nano lms SAW formaldehyde gas sensor. We found that the contact angle of a water droplet on the sensing layer can reach ~ 135°, showing its high hydrophobicity. We also found that the creation of many metal ion (Zn 2+ ) sites on the porous surface of the sensor improves the sensor's response to formaldehyde gas through effective physical adsorptions [30].

SAW resonator and composite nano sensing layer
Surface acoustic wave resonator (SAWR) was made on a quartz substrate (12 mm × 3 mm × 0.5 mm, STcut type of 42°75' y-axis rotary cutting) with a TCF value of 0.24 ppm/°C [31]. Al electrode interdigital transducers (IDTs, 30 pairs) and re ective gratings (100 pairs) were deposited on the quartz substrate. IDTs have a nger width and spacing of 4 µm and an aperture of 3 mm. The re ection grating has the same structure and dimension as the IDTs, both of which are shown in Fig. 1. The resonant frequency of the designed SAWR is ~ 200 MHz. In order to prevent the growth of sensitive materials in the IDTs and their re ectors during the subsequent processes, we used a polyimide tape to cover most of SAWR area except the sensing region.
The hydrophobic composite sensing layer uses BC lm as the structural substrate, PEI particles to provide reaction sites, and ZIF-8 shell as the hydrophobic layer. The fermentation process of BC solution and the preparation process of PEI/BC bi-layer lms have been introduced in detail in our previous work [15]. The BC hydrosol (1 wt‰) and PEI hydrosol (1.25 wt‰) were sequentially coated onto the ST-cut quartz substrate with the spinning speeds of 6000 and 7000 rpm, respectively, then dried in an oven at 60℃ for 10 minutes to make a PEI/BC bi-layer on the SAWR. For synthesis of ZIF-8, 2-methylimidazole (MIN) was used as a linking agent. The molar ratio of the synthesis solution was Zn 2+ :MIN:H 2 O = 1:8:1000. Firstly, 3.28 g MIN was dissolved in 60 mL deionized water. Then, 1.96 g Zn(NO 3 ) 2 ·6H 2 O was dissolved in 30 mL deionized water. Finally, the zinc nitrate solution was mixed with the MIN solution under a constant stirring process. Zinc nitrate and MIN were obtained from Shanghai Aladdin Biochemical Technology Co., Ltd, China. All the preparation steps were carried out at room temperature (25°C). The top side of SAWR coated with PEI/BC bi-layer nano lms was gently put into the synthesis solution, which was placed in a constant temperature reactor at 60°C for 8 hours. After this, the SAWR with the composite sensing layer was thoroughly cleaned with methanol and deionized water to remove any physically adsorbed ZIF-8 crystals and unreacted precursors. After cleaning, the SAWR was placed in an oven at 60℃ for 10 minutes to obtain the SAWR with the ZIF-8@PEI/BC composite nano lm. The PEI/BC bi-layer nano lms without applied with the MIM solution, and a PEI lm which was soaked in the MIM solution were used as the control groups for comparisons. Finally, the SAWRs coated with the sensing lms were connected to oscillator-based readout circuits (including ampli cation and phase shifting) by spot welding using gold wires to form a SAW based formaldehyde gas sensor.

Sensing and characterization system
The experimental sensing system is illustrated in Fig. 2. The temperature and humidity of the laboratory environment were controlled at 25°C and 30% RH, respectively. The sensing system includes two enclosed chambers of 2 L inside and 20 L outside, a workbench with the constantly controlled temperature, evaporation station, saturated salt solution bottle, thermometer, hygrometer, frequency counter, and digital source meter. The 2 L testing chamber was placed inside the 20 L testing chamber, and the SAW device was kept inside the 2 L testing chamber. Formaldehyde solution was dropped on the evaporation station using a micropipette in the 20 L test chamber. The relative humidity control was implemented using a humidi er and the owing dried N 2 gas. After the formaldehyde solution evaporates, the power supply of the evaporator was disconnected and test chamber was returned back to room temperature of 25°C. A digital source meter and frequency counter were then switched on so that the output frequency of the SAW device is stabilized. Then the 2 L test chamber was slowly opened through a mechanical pulley to introduce the formaldehyde gas to the surface of the SAW gas sensor. The output frequency deviation of SAWR gas sensor was continuously monitored using a frequency counter (Agilent 53210A). Based on our previous work, the volume V liquid of the formaldehyde solution required to set the concentration of formaldehyde gas can be obtained using Eq. 1 [15], where C (ppm) is the required concentration of formaldehyde gas, and T (K) is the temperature of the test environment. A eld emission scanning electron microscope (FE-SEM, FEI-INSPECT F50) was used to characterize the surface morphology of the sensing lm. A Kruss G10 Drop Shape Analyzer Goniometer was employed to measure the water contact angle on the ZIF-8@PEI/BC composite nano lm. The contact angle values reported were an average of three separate measurements carried out at three different locations on the lm surface. Figure 3 shows the SEM images of PEI lm, PEI/BC bi-layer nano lms and ZIF-8@PEI/BC composite nano lms. Due to the high viscosity and small particle size of PEI colloids, the surface of the PEI lm shows some uneven agglomerates without obvious large pores. The PEI/BC bi-layer nano lms inherit the porous and brous network structure of the BC lm, and the PEI particles are evenly distributed on the surface of the BC lm. For the ZIF-8@PEI/BC composite nano lms, a snow ake-like structure of zeolite imidazole metal salt framework can be seen formed in the sensing area of the resonator. This snow akelike structure increases the roughness of the sensitive lm and increases the contact angle of water droplets on the lm [32]. The hydrophobic ZIF-8 structure prevents water molecules from entering the BC pores, thereby reducing the frequency shift of the mass loading caused by the water molecules in the formaldehyde gas detection process. Figure 4 shows the response frequency of SAW sensors coated with ZIF-8@PEI nano lms, PEI/BC bi-layer nano lms, and ZIF-8@PEI/BC composite nano lms when exposed to 10 ppm formaldehyde gas in 30% RH environment. All three types of sensors respond well, which is due to the reversible nucleophilic reactions between formaldehyde molecules and PEI at room temperature [33]. The response mechanism of PEI adsorbing formaldehyde molecules is shown in Fig. 5. The electron pairs of the polar double bonds for the formaldehyde molecules are transferred to the oxygen atoms, thus the oxygen atom is negatively charged, and the carbon atom is positively charged. This makes the carbon atoms becoming electrophilic. However, there is a lone pair of electrons in the nitrogen atom of the amine group, which is easy to be used as a nucleophile to attack the electrophile. This causes a nucleophilic reaction. The two bonds in the formaldehyde molecules are broken, thus forming two new covalent bonds. The formaldehyde molecules are then adsorbed by the sensing layer, thus causing a mass loading of the SAW sensor.

Gas sensing detection and sensing mechanism
The frequency responses of SAW sensors coated with ZIF-8@PEI/BC lms and PEI/BC lms to 10 ppm formaldehyde gas are much larger than that coated with ZIF-8@PEI. This can be attributed to the porous network structure of the BC nano ber lm facilitating more sites for the absorption of formaldehyde gas molecules. The hydrogen bonds formed between the hydroxyl group on the surface of BC and the amine group of PEI prevent the aggregation of PEI particles, which ensure the uniform distribution of PEI on the BC lm and enhance the adsorption of formaldehyde gas molecules. This has been reported in our previous work [15]. In addition, for 10 ppm formaldehyde gas, the frequency shift of the SAW sensor coated with ZIF-8@PEI/BC lms is 13% higher than that of PEI/BC lms. This is because the Zn 2+ sites provide extra physical adsorption positions for the formaldehyde molecules. Each Zn 2+ ion on the surface of ZIF-8 can adsorb one formaldehyde molecule, and all the Zn ions on the surface can adsorb formaldehyde molecules simultaneously. When there are many formaldehyde molecules in the chamber, the formaldehyde molecules are preferentially adsorbed on the surface of Zn 2+ ions. After the formaldehyde molecules are covered the Zn 2+ surfaces, the remaining formaldehyde molecules can be weakly connected with the those already adsorbed on the Zn 2+ surface. For more details of the adsorption mechanism of Zn 2+ sites on formaldehyde molecules, the readers can refer to the work reported by Chen et al using a density functional theory [30]. Our study reveals that the presence of the ZIF-8 metal framework enhances the response of the sensitive lm to formaldehyde gas. Figure 6 showed the response dynamic curves of SAW sensors based on ZIF-8@PEI/BC composite nano lm, PEI/BC lm and ZIF-8@PEI lm, when exposed to different concentrations of formaldehyde gas from 100 ppb to 10 ppm at 25℃ and 30% RH. The response curves clearly show that the presence of ZIF-8 enhances the sensor's response to formaldehyde gas. The sensor also shows good stability to formaldehyde gas, as there is no apparent baseline drift.
The response time of the sensor is de ned as the time required to reduce the resonant frequency to 90% of the maximum frequency shift, and the recovery time is the time required to restore the resonant frequency to 10% of the maximum frequency shift. Figure 7 shows the typical response and recovery characteristic curves of SAW sensors with ZIF-8@PEI/BC composite nano lm and PEI/BC lm, when exposed to 10 ppm formaldehyde gas in the test chamber environment at 25℃ and 30% RH. Even though there is an added ZIF-8 layer, the response time and recovery time of the SAW sensor are only slightly increased, which does not show signi cant effect on the overall sensing performance. This may be due to the fact that the ZIF metal framework can signi cantly increase the speci c surface area of the sensitive membrane and does not reduce the effective transport and adsorption of gas molecules. Figure 8(a) shows the repeated response recovery curves of SAW sensors, coated with PEI/BC lm, ZIF-8@PEI lm and ZIF-8@PEI/BC composite nano lm, when exposed to 10 ppm formaldehyde gas in different chamber environments with relative humidity values of 30%, 56% and 84%, respectively. The responses of PEI/BC lms SAW sensor to 10 ppm formaldehyde gas are increased with the increase of relative humidity levels. This is because the amine group of PEI and the hydroxyl group of BC have strong attractions to H 2 O molecules, which increases the mass loading of the SAW sensor. The frequency shifts of SAW sensors with ZIF-8@PEI lm and ZIF-8@PEI/BC composite nano lm to 10 ppm formaldehyde gas show little changes with the humidity levels, and the sensors show good responses in different RH environments, with response drifts less than 3%. Figure 8(b) shows the frequency shift of three SAW sensors exposed to 10 ppm formaldehyde gas under different relative humidity levels between 43%~93%.

Detection and mechanism of hydrophobicity
Compared with the low humidity (10% RH) environment, the PEI/BC bi-layer nano lms SAW sensor without growth the ZIF-8 has a nearly 50% drift in the frequency response to 10 ppm formaldehyde gas under a high humidity (93% RH) environment. Humidity changes have considerable impact on the sensor's response to formaldehyde. The frequency response of ZIF-8@PEI and ZIF-8@PEI/BC composite nano lms SAW sensors to 10 ppm formaldehyde gas has a response drift of less than 5% under these two types of humidity (10% RH and 93% RH). This shows that the in uence of humidity on the concentration of formaldehyde has been greatly reduced. Therefore, it is con rmed that the presence of ZIF-8 makes the SAW sensor much less sensitive to humidity.
In order to verify that the growth of ZIF-8 improves the hydrophobic performance of the sensing lm, we measured contact angles of water droplets placed on the lm surface. As shown in Fig. 9(a), the PEI/BC lms are hydrophilic with a contact angle is ~ 20° due to the strong hydrogen bonds formed between the amine group of PEI and the hydroxyl group of BC. The addition of ZIF-8 increases the contact angle of the ZIF-8@PEI/BC composite nano lms to 135°, as shown in Fig. 9(b). This is attributed to the fact that the growth of ZIF-8 crystals with a snow ake ake structure signi cantly increases the surface roughness (R f ) of the sensing layer. As is well reported, the hydrophobicity and surface roughness of the lm are the two main factors which affect the water contact angle value of the surface [32,34]. When water droplets are placed on the rough surface of the lm, the following equation is normally used to predict the water contact angle [35]: where θ ω and θ 0 are the contact angles of water droplets on the rough lm surface and the smooth lm surface, respectively. R f is a dimensionless factor, equal to the ratio of the surface area to its at projected area. Accordingly, the contact angle of water droplets on the rough hydrophobic lm surface is much larger than that on the smooth surface. This means that the hydrophobic properties of the lm will increase with the increase of R f . Therefore, ZIF-8 with snow ake-like structure greatly improves the hydrophobicity of the sensitive lm. The hydrophobic ZIF-8 structure prevents water molecules from entering the internal pores of the BC membrane, and avoids the frequency shift of the sensor's mass loading caused by the change of environmental humidity when the sensor detects low-concentration formaldehyde gas.

Stability and selectivity of SAW sensors
In order to characterize the sensor response with respect to temperature changes, we measured its resonance frequency in a temperature range of 25-60℃. Figure 10(a) shows that as the temperature increases, the resonant frequency of the sensor decreases slightly. The changes of resonant frequency show a good linear relationship with the temperature change. The TCF of the SAW sensor coated with ZIF-8@PEI/BC composite nano lms is -0.24 ppm/℃. This level shows the sensor response is not affected by temperature in practical gas sensing conditions. The SAW sensor coated with ZIF-8@PEI/BC composite nano lms were repeatedly tested with the formaldehyde for every 5 days by exposing to 100 ppb-10 ppm formaldehyde gas. The responses (or frequency shifts) were recorded for 40 days to test its long-term stability. The results are shown in Fig. 10(b). The uctuation of the frequency shift of the sensor is less than 5%, and there is almost no uctuation at low concentrations, indicating that the sensor has a good long-term stability.
We further tested the selectivity of the SAW sensor with the ZIF-8@PEI/BC composite nano lms to different types of commonly used gases, including 100 ppm reducing gases CO, H 2 , NH 3 , H 2 S, and oxidizing gases NO 2 , volatile gases ethanol, benzene, toluene, etc., and 10 ppm of target gas formaldehyde. The frequency response tests were carried out separately, and the results are shown in Fig. 10(c). Similar to the results of the PEI/BC bi-layer nano lms SAW sensor reported in Ref. [15], the frequency shifts of CO, H 2 , NH 3 , H 2 S, NO 2 , benzene, and toluene remain almost unchanged. The SAW sensor coated with ZIF-8@PEI/BC composite nano lms has almost no response to ethanol gas, whereas that coated with PEI/BC bi-layer nano lms has a slight deviation in response to ethanol gas. This may be due to the improved hydrophobic properties of the ZIF-8@PEI/BC composite nano lms, which prevents water molecules from entering the internal pores of the lm, thereby reducing the adsorption of ethanol gas by water molecules. Therefore, compared with the response of 40.3 kHz of the formaldehyde gas, the response of non-target gases at low concentrations is weak and negligible, which indicates that the sensor has a good selectivity to formaldehyde gas.

Conclusions
In summary, we report that snow ake-like structure of MOF ZIF-8 coated on SAW resonator can improve the hydrophobic properties of PEI/BC bi-layer nano lms, and enhance the sensor's sensitivity to formaldehyde gas. The snow ake ZIF-8 structure signi cantly increases the surface roughness of the sensitive lm and de ning a hydrophobic sensing layer. The hydrophobic ZIF-8 structure prevents water molecules from entering the internal pores of the BC lm, and avoids the mass loading change caused by the environmental humidity change when the sensor detects low-concentration formaldehyde gas. The experimental results show that the response of the SAW sensors with ZIF-8@PEI/BC composite nano lms exposed to 10 ppm formaldehyde gas has a drift of less than 5% in an environment where the RH changes between 10% and 93%. This is a signi cant improvement over SAW sensors with PEI/BC bi-layer nano lms that exhibit 50% frequency drift within the same RH variation. In addition, the Zn 2+ sites on the surface of ZIF-8 further improve the sensor's response to formaldehyde gas through physical adsorption. The SAW sensor coated with ZIF-8@PEI/BC composite nano lms also has good selectivity and long-term stability to the targeted formaldehyde gases.  The response of ZIF-8@PEI nano lms, PEI/BC bi-layer nano lms and ZIF-8@PEI/BC composite nano lms SAW sensor to 10 ppm formaldehyde gas under 25oC and 30% RH. The mechanism diagram of PEI particles adsorbing formaldehyde molecules and the schematic diagram of ZIF-8 hydrophobic structure blocking water molecules from entering the pores of the lm.  Water contact angle measurement of (a) PEI/BC nano lms and (b) ZIF-8@PEI/BC composite nano lms.

Figure 10
ZIF-8@PEI/BC composite nano lms SAW sensor: (a) the relationship between resonance frequency and temperature; (b) repetitive detection of exposure to different concentrations of formaldehyde for 40 days; (c) with PEI/BC bi-layer nano lms SAW sensor exposure to 100 ppm of various common industrial gases and 10 ppm of formaldehyde gas frequency shift, respectively.