Thin Film Gas Sensors Based on Planetary Ball-Milled Zinc Oxide Nanoinks: Effect of Milling Parameters on Sensing Performance

Planetary ball-milled zinc oxide (ZnO) nanoparticle suspensions (nanoinks) were used to produce thin film chemiresistive gas sensors that operate at room temperature. By varying milling or grinding parameters (speed, time, and solvent) different thin film gas sensors with tunable particle sizes and porosity were fabricated and tested with dry air/oxygen against hydrogen, argon, and methane target species, in addition to relative humidity, under ambient light conditions. Grinding speeds of up to 1000 rpm produced particle sizes and RMS thin film roughness below 100 nm, as measured by atomic force and scanning electron microscopy. Raman spectroscopy, photoluminescence, and X-ray analysis confirmed the purity and structure of the resulting ZnO nanoparticles. Gas sensor response at room temperature was found to peak for nanoinks milled at 400 rpm and for 30 min in ethylene glycol and deionized water, which could be correlated to an increased film porosity and enhanced variation in electron concentration resulting from adsorption/desorption of oxygen ions on the surfaces of ZnO nanoparticles. Sensor response and dynamic behavior was found to improve as the temperature was increased, peaking between 100 and 150 °C. This work demonstrates the use of low-cost PBM nanoinks as the active materials for solution-processed thin film gas/humidity sensors for use in environmental, medical, food packaging, laboratory, and industrial applications.


Introduction
Thin films and nanostructured coatings find broad application due to their unique properties, which is a result of both size and interface effects [1][2][3][4][5][6][7][8]. In particular, thin film gas sensing devices based on nanostructures are relevant for monitoring environmental gas concentration, controlling chemical processes, and agricultural and medical applications [9][10][11][12]. For efficient gas sensing, the film properties and structure play a vital role in sensor operation depending on particle size, porosity, and the diffusion time of the gas species [13]. Nanoparticles have a large relative surface area, thereby accelerating exchange with the surrounding atmosphere and leading to a larger detection signal relative to larger or bulk particles. A high porosity nanoparticle sensing film also facilitates the rapid diffusion of the gas species and improves sensing performance. Similarly, gas homogenization time for nanoparticles can be several orders of magnitude smaller than for micron-sized particles.
Semiconducting metal oxide nanostructures have shown exceptional chemical, optical, mechanical, and electrical properties for sensing [14][15][16]. Metal oxides show a large change in their electrical conductivity in the presence of various gas environments [15,17], and thus, by monitoring the variation of current (or resistance) through these films, they can function as sensitive gas detectors. Lowering the operating temperature needed for metal oxide gas sensors is an important goal, and room temperature operation, in particular, has attracted wide interest due to low-power consumption, portability, and cost-effectiveness [18].
Planetary ball milling (PBM) is known for its ability to reliably produce large amounts of nanoscale particles in appropriate solvents by grinding high-purity bulk powders [51,[55][56][57][58][59][60] without requiring complex physical or chemical processing. Planetary mills are very energy efficient by utilizing the high-impact forces during rotary motion of a grinding jar containing the sample, grinding beads and a liquid medium, arranged eccentrically on a so-called sun wheel, which facilitates the rapid production of nanostructured thin films in an inexpensive manner. PBM has been used to produce nanoscale suspensions, or nanoinks, of ZnO for various applications, including antibacterial materials [51], varistors [49], catalysts [61], antifouling [62] and anode materials [63], luminescence [64], composites and alloys [65][66][67], gas sensors [68][69][70][71], UV sensors, and photodetectors [72][73][74]. The PBM process depends on various configurable parameters such as speed of revolution, milling time, and the ratio of beads to feed material. The grinding parameters and solvent used influence the properties and size distribution of the resulting nanoparticle inks and thin films [60], which can be optimized for different applications, including gas sensing.
In this paper, we combine PBM and doctor blading to produce ZnO nanoparticle thin film gas sensors that operate at room temperature via changes in film resistance upon exposure to different gas species. By varying grinding parameters and examining the effect on nanoparticle structure and electrical characteristics of the resultant films, we are able to tune the response signal magnitude and response/recovery times of the ZnO gas sensor devices. Tests conducted in dry/humid air and different target gas environments allowed us to study the ZnO film fabrication conditions required for optimal gas sensing and validate the feasibility of using PBM nanoinks as the active material for thin film gas sensors.

Materials
Zinc oxide powder (ZnO; Anachemia), ethylene glycol (EG; 99% purity, BDH), deionized water (DI water), high-purity gases, and extra dry air (Praxair) were used as received. Glass slides were used as flat substrates for thin film coatings. Stainless steel single edge razor blades (GEM) were used for doctor blading. Conductive silver paint (Ted Pella) and copper tape (3M) were used to make electrical contacts to ZnO films.

Preparation of Nanostructured Thin Films and Sensor Fabrication
PBM was performed using a Fritsch Pulverisette 7 planetary micro mill (premium line) (silicon nitride grinding jars, 80 mL capacity; zirconia grinding beads, 2 mm in diameter). In a typical experiment, ZnO powder (~7 g) was ground in approximately 10 mL of either EG or DI water solvent (a.k.a. colloidal grinding) with approximately 100 g of beads. The grinding speed and time were varied between 200 and 1000 rpm and 10 min and 60 min (in cycles of 5 min grinding, 5 min resting), respectively. After the completion of each grinding trial, a syringe and stainless-steel mesh were used to extract and separate the resulting PBM nanoink suspension from the grinding beads. Several films of each of the ground ZnO trials were prepared for characterization via doctor blading on substrates as follows: A few µL of ZnO nanoink was used to coat the surface by sweeping the blade across the substrate (masked with a single layer of Scotch tape) along with the suspension (blade angle 20-30 • ). The film thickness for all prepared sensors was approximately 60 µm (based on nominal scotch tape thickness). The coatings were dried at~100 • C on a hotplate for approximately 5 min.
For thin film surface imaging, a Nanonics MultiView 1000 atomic force microscope (AFM) with Olympus BXFM optical microscope was employed. Additionally, Raman scattering measurements of thin film samples were performed using a Renishaw inVia Raman microscope. Raman spectra were collected using 632.8 nm HeNe laser line under ambient conditions. Photoluminescence (PL) spectra were acquired using an Olympus FV1000 confocal laser scanning microscope with solid state laser line of wavelength 405 nm. Scanning electron microscopy (SEM) and energy dispersive X-ray analysis (EDX) were performed using a Hitachi S-2600. The thin film porosity and particle size analysis was performed using ImageJ (version 1.46r) [75].
To create the gas sensor devices, two electrical contacts to the resulting films were made using silver paint and copper tape (see Figure 1a), followed by drying in a mechanical oven for 10-15 min at 75 • C.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 4 of 18 Figure 1. (a) Schematic diagram of thin film gas sensor devices: copper tape on glass substrates containing doctor bladed ZnO film served as the electrode contacts and silver paint were used for connecting sensor film to electrodes. (b) Illustration of the gas sensing apparatus used. The sensor was placed inside a quartz test chamber connected to gas cylinders (carrier and/or target gas) through MFCs. Cu wires were connected to the sample via a hermetic feedthrough for electrical measurements. The electrical response of the sensor was monitored using a precision source-measure system.

Sensor Characterization
The gas sensing studies were conducted using the setup shown in Figure 1b, under ambient pressure and temperature (approximately 22 °C, unless indicated otherwise), which consisted of a 3-inch diameter quartz tube chamber connected to a precision electrical source-measure system (Keithley 4200-SCS) and compressed gas source(s) (target/carrier species) through mass flow controllers (MFCs). Two-terminal current-voltage (I-V) characteristics and current response versus time for the ZnO film sensors were measured in different gas environments. Before injecting target gases inside the chamber, a baseline sensor behavior in dry air carrier gas was determined. Relative humidity (RH) inside the test Figure 1. (a) Schematic diagram of thin film gas sensor devices: copper tape on glass substrates containing doctor bladed ZnO film served as the electrode contacts and silver paint were used for connecting sensor film to electrodes. (b) Illustration of the gas sensing apparatus used. The sensor was placed inside a quartz test chamber connected to gas cylinders (carrier and/or target gas) through MFCs. Cu wires were connected to the sample via a hermetic feedthrough for electrical measurements. The electrical response of the sensor was monitored using a precision source-measure system.

Sensor Characterization
The gas sensing studies were conducted using the setup shown in Figure 1b, under ambient pressure and temperature (approximately 22 • C, unless indicated otherwise), which consisted of a 3-inch diameter quartz tube chamber connected to a precision electrical source-measure system (Keithley 4200-SCS) and compressed gas source(s) (target/carrier species) through mass flow controllers (MFCs). Two-terminal current-voltage (I-V) characteristics and current response versus time for the ZnO film sensors were measured in different gas environments. Before injecting target gases inside the chamber, a baseline sensor behavior in dry air carrier gas was determined. Relative humidity (RH) inside the test chamber was measured using a Digi-Sense 20250-11 pre-calibrated thermo-hygrometer. All experiments were performed under ambient room light.

Film Morphology and Material Characterization
An optical microscope image of a typical ZnO thin film coating formed using PBM nanoink and doctor blading is shown in the inset of Figure 2a. The coatings displayed good uniformity, stability, and adhesion under ambient conditions and up to 200 • C. The SEM image in Figure 2a shows that the milled ZnO films consist of fine nanostructured particles with distributed pores. The ZnO film surface topography was further characterized using AFM ( Figure 2b). As expected, particles were ground into finer sizes as grinding speed was increased, and we observed a reduction in root mean square (RMS) film roughness, as measured with AFM (Figure 2c), which drops below 80 nm for grinding at 1000 rpm for 10 min. Higher grinding speed also concurrently reduced particle size below 100 nm (Figure 2d). A similar decreasing trend in particle size/film roughness was observed as grinding time was increased (at constant rpm) (see Figure 2e) [60].
The photoluminescence spectrum of a typical film is shown in Figure 2f. The milled ZnO thin films showed five different peaks of various intensity levels at different wavelength ranges, which can be correlated with the electronic and structural properties of the milled particles. Consistent with previous studies, the 465 nm peak (blue emission band) is attributed to deep level emission originating from oxygen vacancies or interstitial zinc ions of ZnO [76]. The green emission (525 nm) is due to singly ionized oxygen vacancy or oxygen antisite defect O Zn [55]. Similarly, the peak at 577 nm is most likely due to disorder at the surface of nanoparticles after milling [55]. Lastly, the other peaks near 590 nm and 655 nm emission are usually attributed to oxygen vacancies [77,78].
Raman spectra for both PBM nanoink thin films and bulk starting powder are shown in Figure 2g and h. ZnO most often has a wurtzite structure, and there are two A 1 , two E 1 , two E 2 , and two B 1 modes in the Raman spectra of its crystal structure [79].
The most common Raman intensive E 2 (low) mode at~99 cm −1 is just beyond the range of our detection; however, the other Raman mode, E 2 (high), at~437 cm −1 is visible, which is assigned to oxygen vibrational modes [80]. E 2 (high) mode is most prominent in the starting material; after milling, the intensity of the peak decreases and becomes broadened. Lowered intensity and peak broadening of the 437 cm −1 peak indicate a change in band structure and crystallinity of nanostructures after milling. The Raman spectra of both ground and bulk powder display three different peaks at about 206, 329, 379, and 412 cm −1 , where these peaks are attributed to 2TA:2E 2 , E 2 (high)-E 2 (low), A 1 (TO), and E 1 (TO) symmetry, respectively [80]. However, broader Raman peaks and lower intensity are attributed to size effects, lattice strain, and lower crystallinity [55,81]. The decrease in crystallinity is attributed to defects produced during ball milling. For the milled nanoparticles, we observed a broad peak around 580 cm −1 that became more prominent for longer grinding time or higher speeds. It is predicated that peaks with high intensity from 560 to 580 cm −1 indicate the presence of defects related to oxygen vacancy (V o ) and/or zinc interstitial (Zn i ) [80,82], which could be advantageous for gas sensing [19]. The milling media (grinding beads, solvents) can also affect the particles, but this is likely only significant for longer grinding times and/or higher speeds (see red curve in Figure 2h-the peak at 189 cm −1 is assumed to be related to the zirconia grinding beads [83]).
The EDX results for our ZnO gas sensor films (Figure 2i) confirmed the presence Zn and O elements, with average atomic percentages of 48.05 and 51.95, respectively. The oxygen content percentage in the milled powder is likely affected by milling time and medium (e.g., solvents) [55,84,85]. In general, ball milling can also cause contamination from the grinding jar and beads [55,85]; however, our EDX results did not indicate the presence of appreciable elemental impurities for the PBM nanoinks studied.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 18 ized using AFM ( Figure 2b). As expected, particles were ground into finer sizes as grinding speed was increased, and we observed a reduction in root mean square (RMS) film roughness, as measured with AFM ( Figure 2c), which drops below 80 nm for grinding at 1000 rpm for 10 min. Higher grinding speed also concurrently reduced particle size below 100 nm (Figure 2d). A similar decreasing trend in particle size/film roughness was observed as grinding time was increased (at constant rpm) (see Figure 2e) [60].  The photoluminescence spectrum of a typical film is shown in Figure 2f. The milled ZnO thin films showed five different peaks of various intensity levels at different wavelength ranges, which can be correlated with the electronic and structural properties of the milled particles. Consistent with previous studies, the 465 nm peak (blue emission band) is attributed to deep level emission originating from oxygen vacancies or interstitial zinc ions of ZnO [76]. The green emission (525 nm) is due to singly ionized oxygen vacancy or oxygen antisite defect OZn [55]. Similarly, the peak at 577 nm is most likely due to disorder at the surface of nanoparticles after milling [55]. Lastly, the other peaks near 590 nm and 655 nm emission are usually attributed to oxygen vacancies [77,78].
Raman spectra for both PBM nanoink thin films and bulk starting powder are shown in Figure 2g and h. ZnO most often has a wurtzite structure, and there are two A1, two E1, two E2, and two B1 modes in the Raman spectra of its crystal structure [79].
The most common Raman intensive E2 (low) mode at ~99 cm −1 is just beyond the range of our detection; however, the other Raman mode, E2 (high), at ~437 cm −1 is visible, which is assigned to oxygen vibrational modes [80]. E2 (high) mode is most prominent in the starting material; after milling, the intensity of the peak decreases and becomes broadened. Lowered intensity and peak broadening of the 437 cm −1 peak indicate a change in band structure and crystallinity of nanostructures after milling. The Raman spectra of both ground and bulk powder display three different peaks at about 206, 329, 379, and (c) Plot of RMS roughness from AFM of ZnO thin films using PBM nanoink ground in EG for 10 min at different speeds. Average particle size of ZnO particles obtained from AFM and SEM data of thin films using PBM nanoink ground in (d) EG for 10 min at different speeds and (e) DI water at 200 rpm for different grinding time (data at 0 rpm/min denote nominal starting bulk powder particle size). (f) Photoluminescence spectra for approximately 450-700 nm of ZnO thin film ground at 200 rpm for 60 min using DI water solvent (inset (left) optical map image of total PL intensity obtained for 415-715 nm and (right) overall range averaged full spectrum). (g) Raman spectra of ZnO ground film prepared using DI water ground at 200 rpm for 60 min. (h) Raman spectra of different ZnO films ground (EG solvent) and unground: black, bulk sample; blue, ground at 200 rpm for 10 min; red, ground at 750 rpm for 90 min. (i) EDX spectrum of ZnO film using PBM nanoink ground at 600 rpm for 10 min in EG solvent. Figure 3 shows current (at +2 V applied bias) vs. time and I-V curves for ZnO thin film sensors formed using the PBM nanoinks. In Figure 3a, the current levels vs. time for a flow sequence of dry air/pure argon/dry air are shown (similar behavior was observed for nitrogen target gas environments). The time-dependent sensor behavior was observed upon exposure to a target gas atmosphere after reaching a minimum in pure dry air flow, whereby output current rises upon injection of the target gas source(s) and falls upon release of the gas (and re-exposure to dry air/oxygen). Accordingly, the resistance of the sensor underwent a decrease as it was exposed to a target and an increase as dry air was reintroduced. Initially, the sensor sample current (in ambient atmosphere) was observed to increase, and started to decrease as soon as dry air was introduced, indicating that the response of the ZnO gas thin film sensors are also strongly dependent on operating atmospheric humidity or RH. The humidity or moisture sensing ability of our ZnO films was confirmed by the resistance data shown in Figure 3b, which shows a large sensor response vs. RH. Inset shows sensor current vs. time for a similar sequence of on/off dry air/argon gas pulses for ZnO thin film sensor formed using PBM nanoinks ground for 10 min in EG (600 rpm). (d) Sensor current vs. time for 500 sccm dry air followed by 400 sccm dry air/100 sccm hydrogen (sensor formed using PBM nanoinks ground for 10 min in EG (600 rpm)). Inset shows current vs. time for ZnO sensor (prepared using nanoinks ground at 400 rpm for 10 min in EG solvent) exposed to 450 sccm dry air/50 sccm hydrogen (red curve) and 475 sccm dry air/25 sccm hydrogen (black curve). (e) Current vs. time for ZnO thin film sensor (prepared using PBM nanoinks ground at 400 rpm for 10 min in EG solvent) at different H2 gas concentrations indicated. Inset shows response as a function of gas concentration. (f) Current vs. time for ZnO thin film sensor (sample as in (e)) when exposed to different gases (hydrogen, argon, and methane). Inset shows relative response or selectivity to different target gas species.

Gas Sensing Results
The target gas response of our sensors was repeatable over multiple flow sequences; however, due to the size of the test quartz chamber and maximum flow rates used (in addition to the film thickness), it took some time to reach a stable baseline as displayed in Figure 3. (a) Time dependence of sensor current upon exposure to dry air (~2000 s mark) followed by pure argon gas (~5500 s mark) and then dry air again (~8000 s mark) (all flows 500 sccm) for ZnO thin film sensors formed using PBM nanoinks ground for 10 min in EG (400 rpm). Inset shows I-V plots as current increases near start of argon flow. (b) Resistance vs. RH for a ZnO thin film sensor formed using PBM nanoinks ground at 200 rpm for 30 min in DI water. Inset shows individual I-V plots for different humidity values. (c) Gas sensor (prepared using nanoinks ground at 400 rpm for 10 min in EG solvent) showing approach to stable baseline vs. time during repeated exposure to 250 sccm of H 2 pulses. Inset shows sensor current vs. time for a similar sequence of on/off dry air/argon gas pulses for ZnO thin film sensor formed using PBM nanoinks ground for 10 min in EG (600 rpm). (d) Sensor current vs. time for 500 sccm dry air followed by 400 sccm dry air/100 sccm hydrogen (sensor formed using PBM nanoinks ground for 10 min in EG (600 rpm)). Inset shows current vs. time for ZnO sensor (prepared using nanoinks ground at 400 rpm for 10 min in EG solvent) exposed to 450 sccm dry air/50 sccm hydrogen (red curve) and 475 sccm dry air/25 sccm hydrogen (black curve). (e) Current vs. time for ZnO thin film sensor (prepared using PBM nanoinks ground at 400 rpm for 10 min in EG solvent) at different H 2 gas concentrations indicated. Inset shows response as a function of gas concentration. (f) Current vs. time for ZnO thin film sensor (sample as in (e)) when exposed to different gases (hydrogen, argon, and methane). Inset shows relative response or selectivity to different target gas species.
The target gas response of our sensors was repeatable over multiple flow sequences; however, due to the size of the test quartz chamber and maximum flow rates used (in addition to the film thickness), it took some time to reach a stable baseline as displayed in Figure 3c, which shows the sensor stabilizing after approximately 2.5 h. Sensing with dry air as the carrier gas to balance the target species was also possible using our PBM nanoink sensors, as evidenced by the detection of hydrogen in Figure 3d, which shows that sensor current is proportional to target gas concentration. Further study of sensor response was performed for a sequence of hydrogen gas concentrations ranging from 5000 to 20,000 ppm (Figure 3e), which showed a linear response region (Figure 3e, inset) with a sensitivity of approximately 2.4 × 10 −2 ppm −1 . In addition, sensor selectivity was tested by exposure to the same concentration for different gas species (Figure 3f), indicating a strong response to methane, followed by hydrogen and argon, which shows potential for selective gas sensing (Figure 3f, inset). Preferential adsorption of different gas molecules [86] likely leads to the sensor selectivity we observed at room temperature. Upon cycling back to the first gas pulse (H 2 ), the original sensor signal returned. In general, the gas sensor devices studied gave stable and reproducible responses that stayed within 25% of the original signal after several months.
The gas sensing response is usually defined as the ratio of resistances in two different gas atmospheres. In order to compare the performance of the gas sensors prepared using PBM nanoinks from different grinding conditions, the sensor response [87,88] was calculated as where R a and R g are the sensor resistance in dry air and in the presence of target gas, respectively. Figure 4 compares the effects of different grinding parameters (speed/time/solvent) on gas sensor response in the presence of argon target gas. It can be seen in Figure 4a that sensor response decreases with grinding speed beyond 400 rpm, possibly due to particle agglomeration and porosity variations (see Section 3.3). Response can also be seen to increase with grinding time, peaking near 30 min for a constant grinding speed of 200 rpm (Figure 4b). Grinding of materials using high energy ball milling can produce both surface and bulk defects. It is possible that surface defects would enhance gas sensing performance by exposing adsorption sites, whereas bulk defects could limit the adsorption surface area [55]. The presence of surface defects such as Zn i -and V o -related defects is indicated in the Raman data of Figure 2g and h as grinding was increased. This could lead to a significant increase in gas sensor response because more oxygen molecules can be easily absorbed and ionized on the surface of ZnO [89]. However, grinding at a higher speed or for a longer time can also give rise to bulk defects; therefore, it is anticipated that there will be a point at which bulk defects arise during grinding, resulting in lower response, which is discussed further below.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 9 of 18 Figure 3c, which shows the sensor stabilizing after approximately 2.5 h. Sensing with dry air as the carrier gas to balance the target species was also possible using our PBM nanoink sensors, as evidenced by the detection of hydrogen in Figure 3d, which shows that sensor current is proportional to target gas concentration. Further study of sensor response was performed for a sequence of hydrogen gas concentrations ranging from 5000 to 20,000 ppm (Figure 3e), which showed a linear response region (Figure 3e, inset) with a sensitivity of approximately 2.4 × 10 −2 ppm −1 . In addition, sensor selectivity was tested by exposure to the same concentration for different gas species (Figure 3f), indicating a strong response to methane, followed by hydrogen and argon, which shows potential for selective gas sensing (Figure 3f, inset). Preferential adsorption of different gas molecules [86] likely leads to the sensor selectivity we observed at room temperature. Upon cycling back to the first gas pulse (H2), the original sensor signal returned. In general, the gas sensor devices studied gave stable and reproducible responses that stayed within 25% of the original signal after several months. The gas sensing response is usually defined as the ratio of resistances in two different gas atmospheres. In order to compare the performance of the gas sensors prepared using PBM nanoinks from different grinding conditions, the sensor response [87,88] was calculated as where Ra and Rg are the sensor resistance in dry air and in the presence of target gas, respectively. Figure 4 compares the effects of different grinding parameters (speed/time/solvent) on gas sensor response in the presence of argon target gas. It can be seen in Figure 4a that sensor response decreases with grinding speed beyond 400 rpm, possibly due to particle agglomeration and porosity variations (see Section 3.3). Response can also be seen to increase with grinding time, peaking near 30 min for a constant grinding speed of 200 rpm (Figure 4b). Grinding of materials using high energy ball milling can produce both surface and bulk defects. It is possible that surface defects would enhance gas sensing performance by exposing adsorption sites, whereas bulk defects could limit the adsorption surface area [55]. The presence of surface defects such as Zniand Vo-related defects is indicated in the Raman data of Figure 2g and h as grinding was increased. This could lead to a significant increase in gas sensor response because more oxygen molecules can be easily absorbed and ionized on the surface of ZnO [89]. However, grinding at a higher speed or for a longer time can also give rise to bulk defects; therefore, it is anticipated that there will be a point at which bulk defects arise during grinding, resulting in lower response, which is discussed further below.

Gas Sensing Mechanisms and Data Analysis
An important parameter for gas sensing performance is the response/recovery time of the sensor. Response time is usually defined as the time required for a sensor to reach 90% of the total response signal, such as resistance or current upon exposure to the target gas [88]. On the other hand, recovery time is typically defined as the time required for a sensor to return to within 10% of the original signal upon removal of the target gas [88]. The response and recovery times depend upon diffusion paths (distance travelled by gas molecules through the nanoparticle sensing surface), film porosity, and particle agglomeration [13,[90][91][92]. Based on the results shown in Figure 5a, gas sensors prepared from PBM nanoinks with lower grinding speeds have a faster response compared to sensors prepared from higher grinding speeds, with response time reaching a maximum near 600 rpm before decreasing again at higher speeds. Similarly, increased grinding time also appears to increase response time, with a leveling off near 30 min observed (Figure 5c). Gas sensor recovery time generally increased with grinding speed up to 600 rpm before dipping at 800 rpm (Figure 5b), while a grinding time of approximately 30 min also displayed a dip (Figure 5d), indicating a faster sensor recovery. In general, to have an effective sensor that responds/recovers quickly, one should choose the optimal combination of nanoink grinding conditions, and based on Figure 5, this appears to occur for low to moderate grinding speeds (200-400 rpm) and moderate grinding times (~30 min) for the devices studied, or potentially higher grinding speeds (~800 rpm) with short grinding times (~10 min).

Gas Sensing Mechanisms and Data Analysis
An important parameter for gas sensing performance is the response/recovery time of the sensor. Response time is usually defined as the time required for a sensor to reach 90% of the total response signal, such as resistance or current upon exposure to the target gas [88]. On the other hand, recovery time is typically defined as the time required for a sensor to return to within 10% of the original signal upon removal of the target gas [88]. The response and recovery times depend upon diffusion paths (distance travelled by gas molecules through the nanoparticle sensing surface), film porosity, and particle agglomeration [13,[90][91][92]. Based on the results shown in Figure 5a, gas sensors prepared from PBM nanoinks with lower grinding speeds have a faster response compared to sensors prepared from higher grinding speeds, with response time reaching a maximum near 600 rpm before decreasing again at higher speeds. Similarly, increased grinding time also appears to increase response time, with a leveling off near 30 min observed (Figure 5c). Gas sensor recovery time generally increased with grinding speed up to 600 rpm before dipping at 800 rpm (Figure 5b), while a grinding time of approximately 30 min also displayed a dip (Figure 5d), indicating a faster sensor recovery. In general, to have an effective sensor that responds/recovers quickly, one should choose the optimal combination of nanoink grinding conditions, and based on Figure 5, this appears to occur for low to moderate grinding speeds (200-400 rpm) and moderate grinding times (~30 min) for the devices studied, or potentially higher grinding speeds (~800 rpm) with short grinding times (~10 min). Note: Response/recovery time evaluation is constrained by gas flow rates (maximum 500-700 sccm), test chamber volume, and gas sensor film thickness used in the present study.
As mentioned earlier, the response of nanostructured thin film gas sensors is related to particle size and porosity. Figure 2c-e shows that grinding at a higher speed or for a Note: Response/recovery time evaluation is constrained by gas flow rates (maximum 500-700 sccm), test chamber volume, and gas sensor film thickness used in the present study.
As mentioned earlier, the response of nanostructured thin film gas sensors is related to particle size and porosity. Figure 2c-e shows that grinding at a higher speed or for a longer time tends to result in smaller particle sizes. The smaller particle size is ac-companied by an increase in the total sensing surface area; thus, sensor performance is proportional to the decrease in grain size [13]. This explains the increase in sensor response ( Figure 4) as grinding speed approaches 400 rpm and grinding time reaches 30 min. As particle size decreases further (<200 nm), it is possible that the nanoparticles suffer from increased agglomeration, which reduces the effective surface area and thus the free surface energy [93][94][95] (in addition, a large particle size dispersion as in the 600 rpm sample of Figure 2d may exacerbate this effect). The influence of agglomeration is consistent with decreased porosity observed for the samples made with nanoinks milled at a higher rpm (see Figure 6a) [96]. Lower porosity could also explain the faster response/recovery times observed for some samples in Figure 5. Further to the material characterization data in Section 3.1, grinding of the ZnO material eventually leads to an increase in surface and bulk defects. In general, the defect density on the surface is considered higher than in the bulk [97]. Surface defects function as (i) charge carrier traps and (ii) adsorption sites improving the electron-hole separation [98]. Bulk defects, on the other hand, serve as recombination sites [99]. Thus, the influence of grinding on the gas sensing behavior is dependent on the ratio of bulk-to-surface defects. Initially, as the grinding time/speed increases, the ratio of bulk-to-surface defects decreases, which can be attributed to an increase in the total surface area leading to comparatively higher surface defects. However, based on our results at higher grinding speeds and for longer times, this ratio seems to increase due to the formation of excess bulk defects, which may be attributed to bulk crystal distortion caused by excess mechanical attrition. longer time tends to result in smaller particle sizes. The smaller particle size is accompanied by an increase in the total sensing surface area; thus, sensor performance is proportional to the decrease in grain size [13]. This explains the increase in sensor response (Figure 4) as grinding speed approaches 400 rpm and grinding time reaches 30 min. As particle size decreases further (<200 nm), it is possible that the nanoparticles suffer from increased agglomeration, which reduces the effective surface area and thus the free surface energy [93][94][95] (in addition, a large particle size dispersion as in the 600 rpm sample of Figure 2d may exacerbate this effect). The influence of agglomeration is consistent with decreased porosity observed for the samples made with nanoinks milled at a higher rpm (see Figure 6a) [96]. Lower porosity could also explain the faster response/recovery times observed for some samples in Figure 5. Further to the material characterization data in Section 3.1, grinding of the ZnO material eventually leads to an increase in surface and bulk defects. In general, the defect density on the surface is considered higher than in the bulk [97]. Surface defects function as (i) charge carrier traps and (ii) adsorption sites improving the electron-hole separation [98]. Bulk defects, on the other hand, serve as recombination sites [99]. Thus, the influence of grinding on the gas sensing behavior is dependent on the ratio of bulk-to-surface defects. Initially, as the grinding time/speed increases, the ratio of bulk-to-surface defects decreases, which can be attributed to an increase in the total surface area leading to comparatively higher surface defects. However, based on our results at higher grinding speeds and for longer times, this ratio seems to increase due to the formation of excess bulk defects, which may be attributed to bulk crystal distortion caused by excess mechanical attrition. Since ZnO is a chemiresistive sensing material [29], the sensor response mechanism is influenced by the gas-surface absorption-desorption process, surface diffusion-reaction, and redox reactions between active species on the sensor surfaces [100]. In addition, defects of nanostructured ZnO such as oxygen vacancies (see Raman data, Figure 2) can act as electron donors, which enhances conductivity and promotes gas sensing performance [19,90]. Grinding in the presence of solvents during PBM can also result in surface modification [101], since ball milling is an energy intensive process. The energy from the high speed impact of the beads can result in the formation of reactive radicals [101]. The grinding media, therefore, can influence the baseline resistivity of the ZnO nanostructured films [29,102,103]. Since ZnO is a chemiresistive sensing material [29], the sensor response mechanism is influenced by the gas-surface absorption-desorption process, surface diffusion-reaction, and redox reactions between active species on the sensor surfaces [100]. In addition, defects of nanostructured ZnO such as oxygen vacancies (see Raman data, Figure 2) can act as electron donors, which enhances conductivity and promotes gas sensing performance [19,90]. Grinding in the presence of solvents during PBM can also result in surface modification [101], since ball milling is an energy intensive process. The energy from the high speed impact of the beads can result in the formation of reactive radicals [101]. The grinding media, therefore, can influence the baseline resistivity of the ZnO nanostructured films [29,102,103].
The sensing mechanisms in our fabricated sensors likely depend on surface interaction and charge transfer between adsorbed gas species, which results in variation of sensor resistance (or conductance) [104]; it has been reported that O 2 − species will be present at ZnO film surfaces at room temperature [105,106], which results in an electron-depletion layer (barrier layer) on the outer surface of the ZnO in air. Between grains, the merging of the two depletion regions creates a Schottky barrier, the magnitude of which determines the conductivity of the materials [96]. The reaction between the adsorbed oxygen species with adsorbed gas molecules can modify the height of the Schottky barrier, resulting in the variation of sensor resistivity; in dark conditions, when the ZnO surface is exposed to an air environment, electrons from the ZnO conduction band ionize atmospheric oxygen to produce negative oxygen ions at the surface of the nanostructured film (Equation (2)), thereby creating a low conductivity depletion layer near the surface [107].
When light with higher energy than the bandgap of ZnO is incident, electron-hole pairs will be generated. Holes produced by the light absorption migrate to the surface along the potential slope created by band bending and neutralize the negatively charged adsorbed oxygen ions (Equation (3)), which are then desorbed, while the photogenerated electrons reduce the width of the depletion layer, thereby increasing conductivity [107].
Thus, conduction in dark and photoconduction of the ZnO films are strongly dependent on the ambient gas conditions and oxidizing atmosphere. In addition, RH, defined as the amount of water vapor in air, expressed as a percentage of the maximum amount that the air could hold at a given temperature, can also strongly impact a metal oxide surface [108] and thus alter the sensor response [109][110][111]. Water molecules adsorbed on the surface tend to dope ZnO with electrons and/or displace previously adsorbed ionized oxygen, releasing electrons back to the conduction band (i.e., the reverse of Equation (2)) [110]. In both cases, sensor conductivity increases as consistent with Figure 3b.
For the ZnO thin film gas sensors based on PBM nanoinks reported here, the ultimate performance likely depends on a combination of grain size, porosity, and surface-to-bulk defect ratio within the films. The results from porosity ( Figure 6) indicate that there is high surface area in sensors based on PBM nanoinks, peaking around a grinding speed of 400 rpm and grinding time of 30 min, which ultimately creates more active sites and gas diffusion channels and hence improves sensing signal magnitude.
To probe the gas sensor performance further, we examined the effect of operating temperature on sensor response; thermally activated processes can impact reaction kinetics, carrier concentration, and mobility on/near the sensing surface, all of which affect gas sensor detection response and dynamic behavior [112]. Figure 7a shows temperaturedependent sensor response data indicating optimal operation near 100 • C, followed by a decline at higher temperatures, consistent with prior work using metal oxides [113]. Similarly, a shortening of response and recovery times (Figure 7b) at elevated temperatures is due to reduction in activation energy required for surface reactions, i.e., faster absorption and desorption occurs on the surface of the ZnO, leading to shorter response/recovery time [114]. Table 1 summarizes data for ZnO gas sensors fabricated using various synthesis techniques, including ball milling, and their response to different gas species.

Conclusions
In summary, chemiresistive ZnO thin film gas sensors operating at room temperature were fabricated using PBM nanoinks and doctor blading for the effective detection of gas species, including dry air/oxygen, argon, nitrogen, hydrogen, and methane, in addition to atmospheric humidity. By varying grinding parameters, nanoparticle structure and electrical characteristics of the resultant films could be optimized for efficient gas sensing. The response of different fabricated gas sensors versus milling speed and time revealed that nanostructured films made using ZnO nanoinks milled at 400 rpm for 30 min produced the best combination of sensor signal magnitude and dynamic behavior (time response), and further performance enhancement was observed up to temperatures of 100 °C. Future studies could examine the effect of different materials and grinding solvents on ZnO PBM nanoink-based sensors along with different grinding times/speeds to refine the optimal gas sensor structure as a function of film thickness. In addition, sensing different solvent vapors (acetone, IPA, ethanol, etc.) and humidity along with multiple gases for multi-  a S = R a /R g = I g /I a ; b S = (R a −R g )/R a × 100%.

Conclusions
In summary, chemiresistive ZnO thin film gas sensors operating at room temperature were fabricated using PBM nanoinks and doctor blading for the effective detection of gas species, including dry air/oxygen, argon, nitrogen, hydrogen, and methane, in addition to atmospheric humidity. By varying grinding parameters, nanoparticle structure and electrical characteristics of the resultant films could be optimized for efficient gas sensing. The response of different fabricated gas sensors versus milling speed and time revealed that nanostructured films made using ZnO nanoinks milled at 400 rpm for 30 min produced the best combination of sensor signal magnitude and dynamic behavior (time response), and further performance enhancement was observed up to temperatures of 100 • C. Future studies could examine the effect of different materials and grinding solvents on ZnO PBM nanoink-based sensors along with different grinding times/speeds to refine the optimal gas sensor structure as a function of film thickness. In addition, sensing different solvent vapors (acetone, IPA, ethanol, etc.) and humidity along with multiple gases for multiplexed E-nose applications [120,121] should all be feasible using the low-cost PBM nanoink thin film gas sensor approach presented.