CMOS AlN and ScAlN Pyroelectric Detectors with Optical Enhancement for Detection of CO2 and CH4 Gases

Gas sensors are useful for monitoring of greenhouse gases. As the move toward complementary metal‐oxide‐semiconductor (CMOS) compatible pyroelectric room temperature detectors is gaining traction due to its scalability to 8‐in./12‐in. wafer area and integrable with CMOS electronics, CMOS compatible aluminum nitride (AlN)‐ and scandium aluminum nitride (ScAlN)‐based pyroelectric detectors are developed for sensing of CO2and CH4gases, which are two of the greenhouse gases that contribute significantly to global warming. Leveraging gas absorption at respective mid‐infrared (IR) wavelengths, CO2 and CH4 gases are detected at various concentrations with fast response time ≈1 s. A compound parabolic collector (CPC) is designed and integrated into the gas sensor to enhance the optical flux received by the detector, which demonstrates ≈10× signal improvement in its presence. Further factors such as the effect of sensing area reduction and response to random gas concentrations are also tested on AlN‐ and ScAlN‐based pyroelectric detectors respectively to observe the gas sensing behaviors of both detectors. The results obtained provide further understanding of the behavior of CMOS AlN‐ and ScAlN‐based pyroelectric detectors as IR gas sensors, which can potentially inspire new design and design selection for various gas sensing applications.


Introduction
Greenhouse gases have been of concern over the years. The ever-increasing emission of greenhouse gases signifies more heat trapped in the atmosphere, and hence increasing global DOI: 10.1002/aelm.202300256 temperatures, causing climate change, resulting in effects such as hotter temperatures, more severe storms, and increased droughts. Carbon dioxide (CO 2 ) and methane (CH 4 ) are both major contributors of global warming. In Year 2020, the United States (U.S.) Environmental Protection Agency reports 90% U.S. Greenhouse gas emissions coming from CO 2 (79%) and CH 4 (11%). [1] As CH 4 can trap heat 25 times more than that of CO 2 , [2] its presence cannot be ignored though it only accounts for 11% of greenhouse gas emissions. These two gases are both colorless and odorless and usually go undetected without sensors in place. When exposed to them, these gases could pose health threats and death. CO 2 gas at concentrations >1000 ppm in indoor areas have been reported to cause symptoms such as dizziness, headaches, drowsiness, and lack of concentration resulting in degraded performance known as sick building syndrome. [3,4] CH 4 gas, on the other hand, has an explosive range from 5% to 17%, [5] which can cause harm to lives and properties if gone undetected. There are a variety of different approaches for sensing CO 2 and CH 4 over the recent years. Some of these include semiconducting metal oxide based sensors, [6,7] optical and acoustics sensors, [8][9][10] and electrochemical-based sensors. [11,12] Among the optical sensors, pyroelectric-based nondispersive infrared (NDIR) sensors have proven promising, partly contributed by the advantages of pyroelectric detectors. Pyroelectric detectors are uncooled thermal detectors, functional at room temperature, and give an output signal when they sense sudden temperature change in their surroundings. They are operational over wide spectral wavelengths, from visible [13] and infrared (IR) [14,15] to millimeter wave [16] and at room temperature without the need to connect to thermoelectric cooler or liquid nitrogen (N 2 ) supply. This subsequently leads to cost and energy reduction as one detector can be used for a wide wavelength range instead of many detectors for specific operational wavelengths and no extra cost is required to set up the infrastructure for measurement, that is, no dewar, liquid N 2 or thermoelectric cooler needs to be purchased for operation of this detector. www.advancedsciencenews.com www.advelectronicmat.de In addition, pyroelectric detectors have a vast variety of applications. By itself, it is a thermal detector, which can be used in applications such as thermal detection and imaging, [17,18] gesture recognition, [19] fingerprint sensor, [20] and contactless button. [21] The effect of pyroelectricity could further be extended toward ultraviolet (UV) [22,23] and mid-IR [24] detection, for energy harvesting and nanogenerators, [25][26][27][28][29] and self-powered sensing. [30][31][32] In recent years, complementary metal-oxide-semiconductor (CMOS) compatible pyroelectric materials have emerged as materials of interest due to the advantage of monolithic integration, thus allowing devices to be integrable with CMOS electronics. Fabricating these CMOS compatible pyroelectric films over 8-or 12-in. wafers using low thermal budget or reduced high temperature processing steps help with cost and energy savings in device fabrication. CMOS pyroelectric materials [33] include hafnium oxide (HfO 2 )-based, [34,35] gallium nitride (GaN)based, [36,37] aluminum nitride (AlN)-based, [38][39][40] and zinc oxide (ZnO)-based. [41] Among these materials, AlN and its doped counterpart scandium aluminum nitride (ScAlN) are attractive due to its high Curie temperatures above 1000°C [38,42] and have not only piezoelectric, pyroelectric, and ferroelectric properties, [43,44] but also suitable for photonics from UV to mid-IR wavelengths [45,46] and power electronics. [47] In particular, CMOS compatible AlN-and ScAlN-based pyroelectric detectors can be prepared by physical vapor deposition with low thermal budget at temperatures ≈200°C [38,39] and scalable up to 8-in. wafer area thus far. Doping of AlN with Sc has shown to increase its pyroelectric coefficient, [39,48] hence resulting in increase in specific detectivity, [39,49] bringing promise to increase performance of CMOS compatible AlN-based pyroelectric detectors to match that of commercially available non-CMOS compatible pyroelectric detectors based on lithium tantalate (LiTaO 3 ) [50] and lead zirconate titanate [51] pyroelectric materials. Although AlN and ScAlN pyroelectric and electrical properties have been studied in recent years, the gas sensing properties using these pyroelectric detectors have not been rigorously studied. To the best of our knowledge, we have not found any reports on the gas sensing properties from AlN-based pyroelectric detectors, though many have reported on electrical properties of AlN pyroelectric detectors [40,[52][53][54] and limited reports on gas sensing responses using ScAlN-based pyroelectric detectors. [14,15,55] Herein, we would like to examine AlN and ScAlN CMOS compatible pyroelectric sensing materials in CO 2 and CH 4 gas sensing based on NDIR technique at room temperature operation. These pyroelectric detectors have two terminals connected to the top and bottom electrodes respectively and no bias (no extra power) is required to be applied during operation, which helps in further saving energy and cost. Both AlN-and ScAlN-based pyroelectric detectors are integrated into our in-house NDIR gas sensing configuration, which includes a source emitter and gas channel. Gas sensing responses of both detectors at different CO 2 and CH 4 gas concentrations are measured. This includes concentration range and fast response times (t 90 ), which overall measure to be ≈1 s. Specifically, we are able to detect CH 4 down to 400 ppm using these pyroelectric detectors. One of the critical components to allow sensing ≤100 ppm CO 2 and down to 400 ppm CH 4 is the compound parabolic collector (CPC) that we have integrated into the gas sensing configuration. The CPC is designed and simulated using Zemax software, giving ≈10× im-provement in gas sensing signal, when tested with CO 2 . Other factors which may affect gas sensing response are also tested to give a flavor on how the pyroelectric detector reacts to each factor change. AlN-based pyroelectric detectors with two different sensing areas are tested for CO 2 gas response and ≈60% signal drop is recorded when AlN sensing area reduces from 500 μm × 500 μm to 300 μm × 300 μm. Random CO 2 gas concentrations are introduced when measuring the gas sensing response of the 12% ScAlN-based pyroelectric detector. The detector, when subjected to various random CO 2 gas concentrations, responds in ≈1 s. The results presented cover characteristics of AlN-and 12% ScAlN-based pyroelectric detectors in NDIR CO 2 and CH 4 gas sensing, which could further provide inspirations for improved gas sensing performance.

AlN-and 12% ScAlN-Based Pyroelectric Detectors Gas Sensing Responses to CO 2 and CH 4
A gas sensing configuration is set up to test the gas sensing responses of AlN-based and 12% ScAlN-based pyroelectric detectors. We adopt the NDIR gas sensing approach where detection is based on CO 2 and CH 4 absorptions at specific optical wavelengths. Figure 1a shows a schematic of our NDIR gas sensing configuration where the pyroelectric detector is put at one end of the gas channel. This gas channel is of length 10 cm and diameter 0.5 cm. The source emitter is attached to one end of the gas channel, while the other end consists of the pyroelectric detector. Optical bandpass filters are put between the gas channel and the emitter and detector, respectively. This prevents the gas from being directly in contact with the detector and emitter, thereby preventing performance deterioration due to poisoning or high humidity, which could occur when devices come in direct contact with some gases or high humidity. Figure 1b shows a photo image of this 10 cm long gas channel. The gas channel is made of Al with the source emitter and detector at each end of the gas channel. The source emitter is connected to a printed circuit board with a microcontroller unit, which is programmed for the emitter to emit at a frequency of ≈17.4 Hz and powered via a universal serial bus connector plugged into a laptop. On the detector end is a CPC placed in front of the AlN-based pyroelectric detector. The detector is connected to a current amplifier, followed by a lock-in amplifier for signal acquisition.
The pyroelectric detector is attached onto the detector end of the gas channel configuration. It sits on a transistor outline (TO)-39 header with its metal contacts wire-bonded on the TO-header leads. Figure 1c,d show the schematics of AlN-based and 12% ScAlN-based pyroelectric detector, respectively. The detector device is of die size ≈2.1 mm by 3 mm with the sensing layer in a membrane stack sandwiched between the top and bottom electrode. The sensing area is ≈500 μm by 500 μm. The pyroelectric coefficients of both detectors are calculated from their measured output current and presented in Figure S2 of the Supporting Information.
The pyroelectric coefficients obtained from our in-house AlN and 12%-doped ScAlN are comparable with those reported in literature. [15,39,42,48,56,57] It is also worthwhile to note that the pyroelectric coefficients calculated for AlN and 12% ScAlN are in general independent of background temperatures, agreeing with what Kurz et al. [48] have reported. This is because the AlN and 12%-doped ScAlN films deposited in this work are columnar and do not contain abnormally oriented grains (AOGs) that usually resulted in bigger grains in the film. Ng et al. [15] have previously reported increasing pyroelectric coefficient as background temperature increases due to bigger grains in the ScAlN films caused by AOGs. Figure 2a,b depict the CO 2 gas responses obtained when AlNbased and 12% ScAlN-based pyroelectric detectors are used in the NDIR gas sensing configuration. The gases cycle between 99.9999% purified N 2 and different concentrations of CO 2 at 2min intervals. In this case, N 2 acts as a reference gas and presents a voltage output of ≈0.81 mV. It can be observed that for AlNbased pyroelectric detector, at the presence of 5000 ppm CO 2 concentration, the voltage output from the detector drops ≈39% from ≈0.81 to ≈0.50 mV. Subsequently, when CO 2 gas concentration decreases from 5000 down to 100 ppm, we note that the output voltage measured due to presence of CO 2 increases from ≈0.50 to ≈0.80 mV. This is in agreement with NDIR approach where CO 2 absorbs at ≈4.26 μm wavelength. The wavelength of the light entering the gas channel is limited by the optical bandpass filter which only allows light of ≈4.26 μm into the gas chan-nel. When N 2 flows into the gas channel, there is no gas absorption and the AlN-based pyroelectric detector reads directly the filtered light from the emitter at a distance 10 cm away. When CO 2 gas is introduced, it absorbs some of the light at ≈4.26 μm wavelength depending on its concentration. The higher the CO 2 gas concentration, the more the absorption will be. With CO 2 absorption at 4.26 μm, the detector detects less light, resulting in output voltage drop. Compared to AlN-based pyroelectric detector, the output voltage from 12% ScAlN-based pyroelectric detector is higher (≈1.13 mV) when reference gas N 2 flows in the gas channel.
From the CO 2 gas sensing results in Figure 2a,b, using 12% ScAlN-based pyroelectric detector has shown an output voltage increase of ≈40% (Figure 2c) compared to using AlN-based pyroelectric detector while keeping the thermal emitter, optical filters, and gas channel unchanged. Figure 2d shows the output voltage difference (when N 2 is flowing and when CO 2 gas is flowing) plotted against respective CO 2 gas concentration for both AlN-and 12% ScAlN-based pyroelectric detectors. The results show that 12% ScAlN-based pyroelectric detector measured higher voltage difference compared to AlN-based pyroelectric detector, exhibiting the performance superiority when compared to undoped AlN-based pyroelectric detector. Inset shows a zoom-in on the voltage difference at lower CO 2 gas concentration. We note that at 100 ppm CO 2 gas concentration, the voltage difference for 12% ScAlN-based pyroelectric detector is ≈0.022 mV, higher than that of AlN-based pyroelectric detector at ≈0.012 mV. This indicates that for gas sensing, AlN-based pyroelectric detector will achieve a higher gas concentration for its lower limit of detection compared to 12% ScAlN-based pyroelectric detector. Hence using 12% ScAlN-based pyroelectric detector will allow us Figure 2. Output voltage signal measured from a) AlN-based and b) 12% ScAlN pyroelectric detector in the NDIR gas sensing configuration at CO 2 gas concentration from 100 to 5000 ppm. c) Comparison of output voltages measured from AlN-based and 12% ScAlN-based pyroelectric detectors across different CO 2 gas concentrations. The results show ≈40% increase in voltage output from 12% ScAlN-based pyroelectric detector compared to AlN-based pyroelectric detector. d) Comparison of voltage differences (between V N 2 and V CO 2 Gas Concentration ) against respective CO 2 gas concentration for AlN-based and 12% ScAlN-based pyroelectric detectors. Inset shows the voltage difference at low ppm CO 2 gas concentration for both detectors. CO 2 gas response across different CO 2 gas concentrations for e) AlN-based and f) 12% ScAlN-based pyroelectric detectors. The sensing area is ≈500 μm × 500 μm for both detectors.
to stretch the lower limit of detection in gas sensing compared to using AlN-based pyroelectric detector. With the change from AlN-based pyroelectric detector to 12% ScAlN-based pyroelectric detector, factors such as pyroelectric coefficients and absorptivity of the detector stacks are among those that will affect the detector's performance, reflected from the output voltage. Pyroelectric coefficients ( Figure S2, Supporting Information) have shown that 12% ScAlN demonstrates ≈3× increase in pyroelectric coefficient as compared to AlN. Fourier transform infrared spectra have also reported higher absorption from 12% ScAlN-based pyroelectric detector stack [14,55,58,59] compared to AlN-based pyroelectric detector stack. [38,58] Although the pyroelectric coefficient shows ≈3× improvement for 12% ScAlN-based pyroelectric detector compared to AlN-based pyroelectric detector, absorptivity of both detector stacks at specified gas absorption wavelength also need to be considered and both factors will affect the final output gas sensing signal. The effect of doping of Sc which helps to increase pyroelectric coefficient of AlN-based pyroelectric detector will impact the limit of detection of the current gas sensor configuration, allowing the gas sensor to sense lower concentration of the specified gas with increased pyroelectric coefficient, hence increase detectivity. We further examine the measured dark currents of both detectors in Figure S3 of the Supporting Information and noted that 12% ScAlN-based pyroelectric detector has www.advancedsciencenews.com www.advelectronicmat.de Figure 3. Output voltage signal measured from a) AlN-based and b) 12% ScAlN-based pyroelectric detector in the NDIR gas sensing configuration at CO 2 gas concentration above 5000 ppm from 10% to ≈100%.
≈4× lower dark current compared to AlN-based pyroelectric detector.
Figure 2e,f shows the experimental data of CO 2 gas response using AlN-and ScAlN-based pyroelectric detectors across CO 2 gas concentration fitted using modified Beer-Lambert equation [60,61] where x represents the gas concentration, span provides an indication on the amount of light that is absorbed by the gas, is the effective absorption coefficient of the gas, and l is the optical path length of the light in the gas channel, which is 0.1 m in this case. CO 2 gas response is calculated using the following equation [14] CO 2 gas response = V N 2 − V CO 2 Gas Concentration V N 2 (2) where V N 2 is the output voltage measured when only N 2 (as reference gas) is flowing in the gas channel and V CO 2 Gas Concentration is the output voltage measured when CO 2 of different gas concentration is in the gas channel. From the fitted data of the CO 2 gas response curves (Figure 2e,f), we note that both fitted equations from AlN-and ScAlN-based pyroelectrics detectors in CO 2 gas sensing are approximately the same. For CO 2 gas using the same gas channel, the fitted curves seem to be independent of the pyroelectric sensing layer. Rather, span and change when the gas tested changes [60] or when the gas channel change. [15] The CO 2 gas concentration is increased further to test CO 2 gas sensing at percentage concentrations. CO 2 gases from 10% to ≈100% (99.995%) are input into the gas channel. Measurements are made at 10%, 25%, 50%, and 100% CO 2 concentrations. Figure 3a,b shows the voltage output gas sensing response of the AlN-and ScAlN-based pyroelectric detector respectively at high CO 2 gas concentration. For AlN-based pyroelectric detector (Figure 3a), at 10% CO 2 concentration, the output voltage dropped to ≈0.176 mV, whereas at ≈100% CO 2 concentration, output voltage dropped to ≈0.122 mV. For 12% ScAlN-based pyroelectric detector (Figure 3b), output voltage ranges from ≈0.294 mV (10% CO 2 ) to ≈0.214 mV (≈100% CO 2 ). As CO 2 gas concentration increases, saturation seems to occur as we notice reduced voltage drop as the concentration differences increase.
In addition to CO 2 gas, we also test the pyroelectric detectors' gas responses to CH 4 gas. This is done by replacing the optical bandpass filter centered wavelength at 4.26 μm (meant for CO 2 gas sensing) to an optical bandpass filter centered wavelength at 7.3 μm as CH 4 gas has an absorption peak at ≈7 μm wavelength region. The output signals from different concentration of CH 4 gas are tested from 400 ppm CH 4 to 2% CH 4 . Figure 4 shows the performance of CH 4 gas sensing using AlN-based pyroelectric detector (Figure 4a,c,d) and 12% ScAlNbased pyroelectric detector (Figure 4b,c,e). The gases are cycled between N 2 and different concentrations of CH 4 at intervals of 2 min. When N 2 is in the gas channel, AlN-based pyroelectric detector registered an output voltage of ≈ 0.67 mV (Figure 4a) while 12% ScAlN-based pyroelectric detector registered an output voltage of ≈0.81 mV (Figure 4b). Figure 4c shows the output voltage difference (when N 2 is flowing, V N 2 and when CH 4 gas is flowing, V CH 4 Gas Concentration ) plotted against respective CH 4 gas concentration for both AlN-and 12% ScAlNbased pyroelectric detectors. Similar to the trend observed for CO 2 in Figure 2d, 12% ScAlN-based pyroelectric detector measured higher voltage difference compared to AlN-based pyroelectric detector. Inset shows a zoom-in on the voltage difference at lower CH 4 gas concentration. At 400 ppm CH 4 gas concentration, the voltage difference for 12% ScAlN-based pyroelectric detector is ≈0.0054 mV, higher than that of AlN-based pyroelectric detector at ≈0.0045 mV, further indicating reduction of lower limit of detection with higher pyroelectric coefficients. The output voltages of both detectors vary with different concentrations of CH 4 . Figure 4d,e then shows CH 4 gas response from respective detectors plotted against different CH 4 gas concentration and fitted using modified Beer-Lambert law, similar to Equation (1). Unlike CO 2 gas that absorbs well at wavelength of ≈4.26 μm, CH 4 gas has a lower absorbance at wavelength of ≈7.3 μm. This is why from the fitted data in Figure 4, span for CH 4 is lower than for CO 2 (Figure 2). According to National . Output voltage signal due to sensing response from CH 4 gas from 2% concentration down to 400 ppm CH 4 gas concentration measured from a) AlN-based pyroelectric detector and b) 12% ScAlN-based pyroelectric detector in the NDIR gas sensing configuration. c) Comparison of voltage differences (between V N 2 and V CH 4 Gas Concentration ) against respective CH 4 gas concentration for AlN-based and 12% ScAlN-based pyroelectric detectors. Inset shows the voltage difference at low ppm CH 4 gas concentration for both detectors. CH 4 gas responses from d) AlN-based pyroelectric detector and e) 12% ScAlN-based pyroelectric detector fitted using modified Beer-Lambert law based on the experimental data measured in (a) and (b), respectively. Response time (t 90 ) taken for f) AlN-based and g) 12% ScAlN pyroelectric detector when sensing 2% CH 4 in the gas channel. The time measured (t 90 ) is ≈1 s for both detectors.
Institute of Standards and Technology (NIST) Chemistry Web-Book, SRD 69, CO 2 gas has absorbance of ≈0.615 [62] at wavelength of ≈4.26 μm while CH 4 gas has absorbance of ≈0.05 [63] at wavelength of ≈7.3 μm. This is ≈90% lower absorbance for CH 4 . If we compare (effective absorption coefficient of gas) for CH 4 and CO 2 , we note that of CH 4 is ≈83% lower than of CO 2 for both AlN-based and ScAlN-based pyroelectric detector. This is close to lower absorbance ≈90% extracted from data reported in NIST Chemistry WebBook.
The response time (t 90 ) when sensing from N 2 to CH 4 is then measured. This is the time taken for the output voltage to change 90% due to the introduction of another gas or gas concentration. Figure 4f,g shows t 90 when 2% CH 4 is introduced into the gas channel initially flowed with N 2 , measured for AlN-and 12% ScAlN-based pyroelectric detectors, respectively. For AlN-based pyroelectric detector, t 90 ≈ 1.06 s when sensing 2% CH 4 , while 12% ScAlN-based pyroelectric detector gives t 90 ≈ 0.87 s. It can be seen that for both detectors, the response times are very fast, ≈1 s.

CPC for Gas Sensing Signal Enhancement
In the above gas measurements presented, a CPC is placed in front of the pyroelectric detector device. This CPC is designed to maximize light collection from the thermal emitter to the pyroelectric detector. Figure 5a shows the schematics from Zemax simulations for 10 cm long, 5 mm diameter gas channels with and without CPC. The light rays travel from the thermal emitter at one end of the gas channel towards the detector surface at the other end of the gas channel. The simulation schematics show that for the gas channel with CPC, most of the light rays are concentrated in front of the detector's sensing surface. Whereas when without CPC, the light rays at the end of the channel seem scattered with some of the light rays impinging outside the detector's sensing surface area. Figure 5b shows 2D images of the pyroelectric detectors' sensing surfaces with color mapping of the power intensity experienced by the detectors' surfaces. We note that the detector with a CPC placed in front of it exhibits higher power intensity on its surface. Figure 5a,b shows the simulations at source emission angle of 12°. The simulated output power received on the detectors' surfaces when with and without CPC at different source emission angles are then plotted as shown in Figure 5c. We observe that when CPC is used, high output power is received by the detector at small source emission angle ≈5°, and gradually reduces as the source emission angle increases toward 50°. Beyond 50°, the simulated output power received by the detector seems comparable for both with and without CPC. Figure 5d plots the effect of CPC at different source emission angles from 5°to 50°based on simulation results. The percentage of power drop in the absence of CPC and enhancement in the presence of CPC are both tabulated across different source emission angles. At source emission angle of 12°, the simulated power drop is ≈90% in the absence of CPC. On the other hand, if CPC is used at this emission angle (12°), we note ≈10× enhancement. This agrees well with our experimental data shown in Figure 5e,f. The enhancement exhibits a parabolic trend, where it starts high (≈40×) at 5°source emission angle and reduces rapidly to ≈4× enhancement as the source emission angle increases to 20°. At wide 50°source emission angle, the enhancement becomes <1× (≈0.69×), negative effect when CPC is used. Figure 5e shows the CO 2 gas response cycled between different concentrations of CO 2 and N 2 when using AlN-based pyroelectric detector in the NDIR gas configuration without CPC. The device performs poorly with CO 2 gas detectable only down to 600 ppm concentration, compared to 100 ppm CO 2 gas concentration or lower detectable when CPC is used (Figure 2a). The output voltage measured from the AlN-based pyroelectric detector when N 2 is in the gas channel is ≈66 μV. This is ≈12× lower as compared to when CPC is used in the gas configuration (Figure 2a). Figure 5f shows a comparison on the output voltages measured at different CO 2 gas concentration with and without use of CPC during the gas sensing measurement. The results show ≈10× improvement in output voltage signal when CPC is used, which agrees well with simulation results in Figure 5d when source emission angle is at ≈12°. This ≈10× improvement in the presence of CPC makes a huge difference and will improve the lower detection limit of the gas. As seen from Figure 5e, the lowest CO 2 concentration detectable without the use of CPC is at only 600 ppm compared to concentrations of 100 ppm or lower when CPC is used.

AlN-Based Pyroelectric Detector: Response and Sensing Area Reduction
The time taken to detect presence of CO 2 is measured for AlNbased pyroelectric detector. Figure 6 shows the response times (t 90 ) taken for AlN-based pyroelectric detector to detect presence of 400 ppm CO 2 (Figure 6a), 1000 ppm CO 2 (Figure 6b), and 5000 ppm CO 2 (Figure 6c) from prior N 2 environment. The results show t 90 ≈ 1.18 s (Figure 6a), t 90 ≈ 1.69 s (Figure 6b), and t 90 ≈ 1.56 s (Figure 6c), all 3 data showing response time < 2 s. The CO 2 gas response time detected by AlN-based pyroelectric detector is in general comparable with CO 2 gas sensing response time when detecting using a ScAlN-based pyroelectric detector [14] in similar gas channel configuration. Figure 6d plots the output voltages from two AlN-based pyroelectric detectors of different AlN sensing areas when sensing different CO 2 gas concentrations from 0 to 5000 ppm. Both detectors have sensing areas ≈500 μm × 500 μm and ≈ 300 μm × 300 μm. As seen from Figure 6d, the voltage drops by ≈60% for the detector with smaller sensing area. The drop in voltage is consistent with the following equation [15,[38][39][40] where i is the output current, A is the sensing area, is the pyroelectric coefficient, and dT/dt is the temperature change with time. As seen from Equation (3), a reduction in sensing area, A will result in drop in i if and dT/dt remain unchanged. We note that as AlN sensing area reduces by ≈64%, the output signal drops by ≈60%. The reduction in sensing area and drop in output signal are comparable, indicating that in this case, the sensing area is a major determining factor for the magnitude of the output signal. Strategies and new designs will need to be implemented if we desire to reduce the detector size further while maintaining a high signal output. Figure 6e further shows the CO 2 gas sensing response across CO 2 gas concentrations from 100 to 5000 ppm where the experimental data are plotted, fitted using modified Beer-Lambert law from Equation (1) and compared with the CO 2 gas response results obtained from Figure 2d,e. Earlier in Figure 2d,e, we note that the fitted data equations are comparable and seem independent to the change in the pyroelectric sensing layer. Here, we observe that additionally, the fitted data equations are also independent to the change in the sensing area of the pyroelectric detector. Figure 6. Response time (t 90 ) taken for AlN-based pyroelectric detector to sense a) 400 ppm CO 2 , b) 1000 ppm CO 2 , and c) 5000 ppm CO 2 in the gas channel. The response time (t 90 ) is measured when the voltage dropped by 90% after switching from N 2 to respective CO 2 concentration. Response time measured for all three CO 2 gas concentration is <2 s. d) Voltage measured when sensing different CO 2 gas concentration from two AlN-based pyroelectric detectors with different AlN sensing areas. The results show ≈60% drop in voltage measured when AlN sensing area reduces by ≈64% from area ≈500 μm × 500 μm to area ≈300 μm × 300 μm. e) CO 2 gas response across different CO 2 gas concentrations from 100 to 5000 ppm for AlN-based pyroelectric detectors with a sensing area of ≈300 μm × 300 μm. The fitted data shows approximately similar CO 2 gas response curves as AlN-based and ScAlN-based pyroelectric detectors in Figure 2d,e.

SAlN-Based Pyroelectric Detector: Response to Random CO 2 Concentrations
Finally, we test the ScAlN-based pyroelectric responses to CO 2 when random concentrations of CO 2 gas are flowed into the gas channel. Figure 7a shows the CO 2 gas responses change randomly at different concentrations over a total duration of >70 min. While the CO 2 gas concentrations are selected randomly, the gas is set to flow for ≈10 min at each specific gas concentration before switching to the next concentration. The response times (t 90 ) are then measured at four different random gas concentration changes. These changes are from 100 ppm CO 2 increase to 4000 ppm CO 2 (Figure 7b), 200 ppm CO 2 increase to 1000 ppm CO 2 (Figure 7c), 1500 ppm CO 2 decrease to 150 ppm CO 2 (Figure 7d), and 3000 ppm CO 2 decrease to 300 ppm CO 2 (Figure 7e). The response times obtained are as follows: t 90 ≈ 1.12 s (100 ppm CO 2 to 4000 ppm CO 2 ), t 90 ≈ 0.98 s (200 ppm CO 2 to 1000 ppm CO 2 ), t 90 ≈ 0.75 s (1500 ppm CO 2 to 150 ppm CO 2 ), and t 90 ≈ 0.94 s (3000 ppm CO 2 to 300 ppm CO 2 ). The results show ≈1 s response time when CO 2 gas concentration changes. For CO 2 sensor applications toward indoor air quality monitoring, a response time of ≈1 s from healthy CO 2 levels (≈400 ppm) [64] to unhealthy CO 2 levels (≥1000 ppm) (Figure 7b,c) will provide sufficient time to kick-start ventilation measures to lower the CO 2 concentration level. Similarly, a recovery time of ≈1 s (Figure 7d,e) will allow the ventilation systems to work more efficiently, hence energy saving.

Conclusion
Gas sensors are one of the essentials for monitoring of greenhouse gases where CO 2 and CH 4 are the two major contributors for global warming. We have demonstrated CO 2 and CH 4 gas sensing behaviors from CMOS-compatible AlN-and 12% ScAlN-based pyroelectric detectors fabricated over 8-in. wafer. With CMOS-compatible pyroelectric detectors gaining traction in recent years, [33,65] the adoption of CMOS-compatible pyroelectric thin films for gas sensing not only allow for wafer-level manufacturing, scalability toward 8-and 12-in. production, but also integrable with CMOS electronics. In summary, these sensors have exhibit response to CO 2 gas from 100 ppm-100% concentration and CH 4 gas from 400 ppm-2% concentration. The gas sensing response time (t 90 ) is ≈1 s. In addition, 12% ScAlN-based pyroelectric detector has shown to have higher sensitivity with its higher output signal (when in N 2 reference gas) and voltage difference (between N 2 reference gas and tested gas concentration) than AlN-based pyroelectric detector. The higher sensitivity from Sc-doped AlN allows it to further reduce the lower limit of detection in gas sensing. These gas sensing responses are measured with a designed CPC placed in front of the detector to concentrate light onto the detector's sensing surface for signal enhancement. At source emission angle ≈12°, we note ≈10× improvement in output signal when CPC is used. Other factors that could affect the gas sensing behaviors of these two detectors are also tested.
AlN-based pyroelectric detector shows response to CO 2 within 2 s. Its gas sensing signal is also observed to vary linearly with its sensing area. 12% ScAlN-based pyroelectric detectors have shown response time (t 90 ) ≈ 1 s when subjected to random CO 2 gas concentrations. We report here, one of the fastest gas response times compared to most NDIR CO 2 gas sensors that report gas response time is nearly tens of seconds. [66][67][68] So far, our experimental data have shown superiority of 12% ScAlN-based pyroelectric detector compared to its AlN counterpart under the same measurement conditions, indicating that doping with Sc is the direction to push up device performance. Despite that, when CO 2 gas responses are fitted with modified Beer-Lambert equation, we note comparable span and regardless of the pyroelectric material used and sensing area of the pyroelectric material, which signifies that with the same gas, the gas channel is oblivious of the change in detector. For CH 4 gas responses, the data fitted with modified Beer-Lambert equation shows lower span and compared to CO 2 gas, indicating that while we maintain the same gas channel, different gases will alter the gas response curve. To further improve robustness of the system, future work could include adding another pyroelectric detector next to the current detector as a reference channel to help monitor any changes in the source emitter caused by factors such as source aging, degradation, and power variations. The results obtained in this work provide further understanding on the behaviors of CMOS-compatible AlNand 12% ScAlN-based pyroelectric detectors in gas sensing, and could potentially inspire future design and tailoring of these devices for NDIR gas sensing.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.