Visible Light-Driven p-Type Semiconductor Gas Sensors Based on CaFe2O4 Nanoparticles

In this work, we present conductometric gas sensors based on p-type calcium iron oxide (CaFe2O4) nanoparticles. CaFe2O4 is a metal oxide (MOx) with a bandgap around 1.9 eV making it a suitable candidate for visible light-activated gas sensors. Our gas sensors were tested under a reducing gas (i.e., ethanol) by illuminating them with different light-emitting diode (LED) wavelengths (i.e., 465–640 nm). Regardless of their inferior response compared to the thermally activated counterparts, the developed sensors have shown their ability to detect ethanol down to 100 ppm in a reversible way and solely with the energy provided by an LED. The highest response was reached using a blue LED (465 nm) activation. Despite some responses found even in dark conditions, it was demonstrated that upon illumination the recovery after the ethanol exposure was improved, showing that the energy provided by the LEDs is sufficient to activate the desorption process between the ethanol and the CaFe2O4 surface.


Introduction
Metal oxide semiconductors have shown the best characteristics in term of sensitivity, selectivity, and stability in gas sensor technology. The development of new materials based on different methods, techniques and working principles has been carried out to achieve the best performance. However, the operating temperatures of metal oxide gas sensors are usually above 150 • C to activate the absorption and desorption processes between the targeted gases and the surfaces of the materials [1][2][3][4][5][6][7][8]. The high operating temperature is one of the main drawbacks of metal oxide-based gas sensors because it ultimately results in high power consumption and undesirable long-term drift problems caused by sintering effects in the metal oxide grain boundaries, yielding poor selectivity and stability [9,10]. Another disadvantage of the metal oxide-based gas sensors with the high operating temperature is

Sensor Preparation
Gold-interdigitated electrodes (Au-IDE) on glass (MicruX Technology, Asturias, Spain) were used as an electronic platform for measuring the electrical characteristics of the CaFe 2 O 4 . The IDE size is (10 × 6 × 0.75 mm), with 90 pairs of electrodes having a pitch of 10 µm and a line width 10 µm. CaFe 2 O 4 nanoparticles were deposited on the surface of the Au-IDEs by drop casting, followed by a spin-coating process. After setting the substrate on the sample holder of the spin coater, 5 µL of CaFe 2 O 4 nanoparticle 10 mg/l suspensions in ethylene glycol were deposited onto Au-IDE by a single layer spin coater and spin-coated at 2000 rpm for 40 s in the air and dried on the hot plate at 90 • C for a few minutes in order to evaporate the solvent. To attain the desired thickness of CaFe 2 O 4 nanoparticles film, the above procedure was repeated 3 times. Afterwards, an annealing process at 450 • C for 1 h with a ramping level 5 • C/min was applied to fix the material onto the substrate and achieve good electrical contact with the Au-IDEs.

Sensor Measurement
Gas-sensing experiments were conducted in a customized chamber of 200 mL in volume. The gas flow was maintained stably at 200 mL/min during all the measurements. Reference gaseous atmospheres were provided by independent mass flow controllers blending synthetic air (SA) and ethanol (100 ppm in SA). To investigate CaFe 2 O 4 optoelectronic properties by activating visible light, resistance measurements were conducted under synthetic air flow with different LED wavelengths (i.e., blue (465 nm), green (520 nm), yellow (590 nm), and red (640 nm)) and in dark condition (without illumination). To determine the sensitivity of the CaFe 2 O 4 sensors towards reducing gases, different concentrations of ethanol vapors from the low to the high concentration (i.e., 10,20,30,50, and 100 ppm) were then applied to the chamber under LED irradiation. The response was defined as [(R g − R a )/R a ] × 100%, where R a and R g are the electrical resistances of the sensor in the air and when exposed with ethanol, respectively. The response and recovery times were defined as the times needed by a sensor to achieve 90% of the total resistance change during the adsorption and desorption process, respectively. All experiments were performed at room temperature.

Sensor Characterization Results
The sample as-prepared powder was tested using XRD as shown in Figure 1a. The mixture of various compounds (e.g., CaCO 3 (ICDD 04-007-4989), Fe 2 O 3 (ICDD 00-002-1047), γ-Fe 2 O 3 (ICDD 00-004-0755), and CaFe 2 O 4 (ICDD 04-007-4989)) was formed during the auto-combustion reaction. Nevertheless, the admixture of different compounds crystallizes to pure CaFe 2 O 4 compound (Figure 1b A scanning electron microscope (SEM) image of an as-prepared powder cross-section of gas sensor pellet annealed at 850 °C for 3 h shown in Figure 2a. The as-prepared sample powders are composed of amorphous-like anisotropically shaped and closely packed grains. Figure 2b shows a surface sensor on the Au-IDE after being prepared by the spin-coating process and annealed at 450 °C for 1 h. The porous structures of interconnected grains were observed, while the grains keep their anisotropic shape. The size of individual grains of smaller dimensions varies from 70 to 300 nm, while the length of anisotropic nanoparticles is up to 650 nm. Particles are very well interconnected and fused together, at the same time maintaining open structures for gas diffusion. Gas-accessible microstructures are preferred for the high gas response. Due to the particle size and form, the coating result was non-uniform on the IDEs. That was not the best result to make a uniform layer by a spin-coating process. However, knowing the response to the ethanol vapors will make a good opportunity to introduce this material as a potential candidate for p-type semiconductor which has a suitable bandgap for visible light-driven gas sensors at room temperature. Another possible method to make the uniform layer on the IDE is by the screen-printing process which is a widespread method in the industry of metal oxide gas sensors.  Figure 3a shows the optical adsorption UV-vis diffuse reflectance spectra (DRS) of CaFe2O4 nanoparticles and Tauc's plot approach to determine the bandgap. It has confirmed the adsorption spectra of CaFe2O4 in the visible light range (400-700 nm). The absorption peak at 425-455 nm (dashed green line/pattern) is due to the maximum adsorption spectra of CaFe2O4 nanoparticles. The dashed red line is the linear fit absorption spectra which are leading to the optical bandgap value. Its optical bandgap was determined to be ~1.9 eV according to the energy dependence relation of A scanning electron microscope (SEM) image of an as-prepared powder cross-section of gas sensor pellet annealed at 850 • C for 3 h shown in Figure 2a. The as-prepared sample powders are composed of amorphous-like anisotropically shaped and closely packed grains. Figure 2b shows a surface sensor on the Au-IDE after being prepared by the spin-coating process and annealed at 450 • C for 1 h. The porous structures of interconnected grains were observed, while the grains keep their anisotropic shape. The size of individual grains of smaller dimensions varies from 70 to 300 nm, while the length of anisotropic nanoparticles is up to 650 nm. Particles are very well interconnected and fused together, at the same time maintaining open structures for gas diffusion. Gas-accessible microstructures are preferred for the high gas response. Due to the particle size and form, the coating result was non-uniform on the IDEs. That was not the best result to make a uniform layer by a spin-coating process. However, knowing the response to the ethanol vapors will make a good opportunity to introduce this material as a potential candidate for p-type semiconductor which has a suitable bandgap for visible light-driven gas sensors at room temperature. Another possible method to make the uniform layer on the IDE is by the screen-printing process which is a widespread method in the industry of metal oxide gas sensors. A scanning electron microscope (SEM) image of an as-prepared powder cross-section of gas sensor pellet annealed at 850 °C for 3 h shown in Figure 2a. The as-prepared sample powders are composed of amorphous-like anisotropically shaped and closely packed grains. Figure 2b shows a surface sensor on the Au-IDE after being prepared by the spin-coating process and annealed at 450 °C for 1 h. The porous structures of interconnected grains were observed, while the grains keep their anisotropic shape. The size of individual grains of smaller dimensions varies from 70 to 300 nm, while the length of anisotropic nanoparticles is up to 650 nm. Particles are very well interconnected and fused together, at the same time maintaining open structures for gas diffusion. Gas-accessible microstructures are preferred for the high gas response. Due to the particle size and form, the coating result was non-uniform on the IDEs. That was not the best result to make a uniform layer by a spin-coating process. However, knowing the response to the ethanol vapors will make a good opportunity to introduce this material as a potential candidate for p-type semiconductor which has a suitable bandgap for visible light-driven gas sensors at room temperature. Another possible method to make the uniform layer on the IDE is by the screen-printing process which is a widespread method in the industry of metal oxide gas sensors.  Figure 3a shows the optical adsorption UV-vis diffuse reflectance spectra (DRS) of CaFe2O4 nanoparticles and Tauc's plot approach to determine the bandgap. It has confirmed the adsorption spectra of CaFe2O4 in the visible light range (400-700 nm). The absorption peak at 425-455 nm (dashed green line/pattern) is due to the maximum adsorption spectra of CaFe2O4 nanoparticles. The dashed red line is the linear fit absorption spectra which are leading to the optical bandgap value. Its optical bandgap was determined to be ~1.9 eV according to the energy dependence relation of  The dashed red line is the linear fit absorption spectra which are leading to the optical bandgap value. Its optical bandgap was determined to be~1.9 eV according to the energy dependence relation of (αhν) 2 = A hν − E g , where α and E g are the absorption coefficient and the bandgap of CaFe 2 O 4 , respectively. In addition, it can be seen that the material shows efficient visible light absorption spectra and the bandgap of CaFe 2 O 4 . Figure 3b shows the optoelectronic properties of CaFe 2 O 4 under LED illumination with different wavelengths and intensities of visible light without gases. The responses indicated that the visible light is suitable for this material because its energy is equal or larger than the bandgap of CaFe 2 O 4 . It confirms that the response and recovery times depend on the energy and intensity of LEDs. When the light is ON (photo-activated), electron-hole (e-h) pairs are generated in CaFe 2 O 4 and will interact with an oxygen molecule and pre-chemisorb oxygen ion in the surface, thus facilitating their chemisorption and increasing the majority charge in CaFe 2 O 4 . This reaction will form a hole-accumulation layer, leading to decrease in electrical resistance. On the contrary, when the light is OFF the recombination process leads to an increase of the resistance to the initial value. According to the results, visible light activation works properly in this material. Moreover, the density of majority charge carriers is not only related to the intensity of visible light, but also to the absorption at this particular energy. In thin nanoparticle films, higher absorption leads to a larger generation rate of e-h pairs, and hence to a stronger impact on desorption.
Sensors 2020, 20, x 5 of 12 ( ℎ ) = ℎ − , where α and Eg are the absorption coefficient and the bandgap of CaFe2O4, respectively. In addition, it can be seen that the material shows efficient visible light absorption spectra and the bandgap of CaFe2O4. Figure 3b shows the optoelectronic properties of CaFe2O4 under LED illumination with different wavelengths and intensities of visible light without gases. The responses indicated that the visible light is suitable for this material because its energy is equal or larger than the bandgap of CaFe2O4. It confirms that the response and recovery times depend on the energy and intensity of LEDs. When the light is ON (photo-activated), electron-hole (e-h) pairs are generated in CaFe2O4 and will interact with an oxygen molecule and pre-chemisorb oxygen ion in the surface, thus facilitating their chemisorption and increasing the majority charge in CaFe2O4. This reaction will form a hole-accumulation layer, leading to decrease in electrical resistance. On the contrary, when the light is OFF the recombination process leads to an increase of the resistance to the initial value. According to the results, visible light activation works properly in this material. Moreover, the density of majority charge carriers is not only related to the intensity of visible light, but also to the absorption at this particular energy. In thin nanoparticle films, higher absorption leads to a larger generation rate of e-h pairs, and hence to a stronger impact on desorption. To investigate the sensitivity of the CaFe2O4 sensors towards reducing gases (i.e., ethanol), different concentrations of ethanol vapors (i.e., from 10 to 100 ppm) were then applied to the chamber under LED illumination. The first phenomenon to be noticed was that the resistance increased in the presence of ethanol (reducing gas), confirming that CaFe2O4 is a p-type material. For comparison, the experiments were also performed by introducing another reducing gas (i.e. NH3), as shown in Figure S1 and the oxidizing gas (i.e., NO2) with different concentrations ( Figure S2). Figure 4a-e shows the dynamic response to different concentrations of ethanol under blue, green, yellow and red LED illumination and also in dark condition (without illumination). The response comparison toward different conditions is shown in Figure 4f. The blue LED had better sensitivity than the others because of its energy. The higher energy and intensity are illuminated as the material increases the energetic state and density of charge carrier on the surface, which influences the sensitivity while being exposed to the target gas. The maximum response corresponding to adsorption spectra is 3.6% at 100 ppm of ethanol for blue LED. In this case, the response and recovery times were ~18 min and ~41 min, respectively. The results obtained provide evidence that CaFe2O4 is a good candidate for visible light-driven gas sensor because of its suitable bandgap (energy of visible light spectra is 1.9-2.7 eV). The sensor responses from other LEDs are too slow compared to that from blue LED, which can be due to their insufficient energy to detect gas.
Some responses were, however, also found in dark conditions. The sensitivity in dark conditions (Figure 4f) indicates that p-type material has the ability to chemisorb the higher To investigate the sensitivity of the CaFe 2 O 4 sensors towards reducing gases (i.e., ethanol), different concentrations of ethanol vapors (i.e., from 10 to 100 ppm) were then applied to the chamber under LED illumination. The first phenomenon to be noticed was that the resistance increased in the presence of ethanol (reducing gas), confirming that CaFe 2 O 4 is a p-type material. For comparison, the experiments were also performed by introducing another reducing gas (i.e., NH 3 ), as shown in Figure S1 and the oxidizing gas (i.e., NO 2 ) with different concentrations ( Figure S2). Figure 4a-e shows the dynamic response to different concentrations of ethanol under blue, green, yellow and red LED illumination and also in dark condition (without illumination). The response comparison toward different conditions is shown in Figure 4f. The blue LED had better sensitivity than the others because of its energy. The higher energy and intensity are illuminated as the material increases the energetic state and density of charge carrier on the surface, which influences the sensitivity while being exposed to the target gas. The maximum response corresponding to adsorption spectra is 3.6% at 100 ppm of ethanol for blue LED. In this case, the response and recovery times were~18 min and~41 min, Sensors 2020, 20, 850 6 of 12 respectively. The results obtained provide evidence that CaFe 2 O 4 is a good candidate for visible light-driven gas sensor because of its suitable bandgap (energy of visible light spectra is 1.9-2.7 eV). The sensor responses from other LEDs are too slow compared to that from blue LED, which can be due to their insufficient energy to detect gas. Some responses were, however, also found in dark conditions. The sensitivity in dark conditions (Figure 4f) indicates that p-type material has the ability to chemisorb the higher concentrations of oxygen which react with gases since the formation of a hole-accumulation space charge layer is not limited by concentrations of free charge carriers [50] despite no illumination of light. The light irradiation has not only provided visible light activation on CaFe 2 O 4 , but also contributed to the desorption process when the ethanol was removed from the chamber. The energy provided by LEDs is sufficient to activate the desorption process between ethanol and the surface of CaFe 2 O 4 . In a dark condition, there was no external energy to break the bonding of target gases on the surface sensing. Thus, the sensor signal was not able to be well recovered. The maximum recovery ability in the dark condition is 35% at 100 ppm of ethanol and the average recovery ability is less than 25% for all concentrations of ethanol.
Sensors 2020, 20, x 6 of 12 concentrations of oxygen which react with gases since the formation of a hole-accumulation space charge layer is not limited by concentrations of free charge carriers [50] despite no illumination of light. The light irradiation has not only provided visible light activation on CaFe2O4, but also contributed to the desorption process when the ethanol was removed from the chamber. The energy provided by LEDs is sufficient to activate the desorption process between ethanol and the surface of CaFe2O4. In a dark condition, there was no external energy to break the bonding of target gases on the surface sensing. Thus, the sensor signal was not able to be well recovered. The maximum recovery ability in the dark condition is 35% at 100 ppm of ethanol and the average recovery ability is less than 25% for all concentrations of ethanol.

Sensing Mechanism
The working principle of gas sensors based on metal oxide depends on chemisorbed oxygen molecules on the surface (i.e., adsorption and desorption), which ionize into species such as O , O and O by taking electrons near the surface of the metal oxides. Generally, ionosorption species of O , O and O are known to be dominant at <150 °C, between 150 and 400 °C, and at >400 °C, respectively [57]. In case of a p-type metal oxide formed by an aggregate of nanoparticles the conduction mechanism is governed by the grain boundaries. However, unlike in the case of n-type metal oxide materials where the outer shell of the nanoparticles is "insulating" because of the ionosorption of oxygen species, in p-type metal oxides the outer shell develops a hole accumulation layer (i.e., "conducting" layer). Figure 5a shows the condition of CaFe2O4 when it is exposed to air in the dark, in which the adsorbed oxygen molecules trap electrons from the valence band of CaFe2O4 and form pre-chemisorbed oxygen ion (O ) on the surface at room temperature [57,58]. Pre-chemisorbed O on the surface results in the presence of a high-conductivity hole-accumulation region in the surface layer of CaFe2O4. Consequently, the energy bands bend upward near the surface of CaFe2O4 (Figure 5d) in comparison with the flat band situation before any surface reaction (Figure 5c) [59]. In the dark condition, the pre-chemisorbed oxygen ion is thermally stable and difficult to remove from the surface of CaFe2O4 at room temperature due to the large absorption

Sensing Mechanism
The working principle of gas sensors based on metal oxide depends on chemisorbed oxygen molecules on the surface (i.e., adsorption and desorption), which ionize into species such as O − 2 , O − and O 2− by taking electrons near the surface of the metal oxides. Generally, ionosorption species of O − 2 , O − and O 2− are known to be dominant at <150 • C, between 150 and 400 • C, and at >400 • C, respectively [57]. In case of a p-type metal oxide formed by an aggregate of nanoparticles the conduction mechanism is governed by the grain boundaries. However, unlike in the case of n-type metal oxide materials where the outer shell of the nanoparticles is "insulating" because of the ionosorption of oxygen species, in p-type metal oxides the outer shell develops a hole accumulation layer (i.e., "conducting" layer). Figure 5a shows the condition of CaFe 2 O 4 when it is exposed to air in the dark, in which the adsorbed oxygen molecules trap electrons from the valence band of CaFe 2 O 4 and form pre-chemisorbed oxygen ion O − 2 on the surface at room temperature [57,58]. Pre-chemisorbed O − 2 on the surface results in the presence of a high-conductivity hole-accumulation region in the surface layer of CaFe 2 O 4 . Consequently, the energy bands bend upward near the surface of CaFe 2 O 4 (Figure 5d) in comparison with the flat band situation before any surface reaction (Figure 5c) [59]. In the dark condition, the pre-chemisorbed oxygen ion is thermally stable and difficult to remove from the surface of CaFe 2 O 4 at room temperature due to the large absorption energy [60]. The kinetic reaction can be explained as follows [6]: When the light illuminates the materials (Figure 5b), electrons are excited from the valence band to conduction band and electron-hole pairs are generated. The holes react with the pre-chemisorbed O − 2(ads) to form oxygen molecules which will be desorbed from the surface of CaFe 2 O 4 ( h + + O − 2(ads) ↔ O 2(g) ). At the same time, new oxygen molecules will be adsorbed and capture the photo-electrons to form photo-induced oxygen ions: ). The net result of these adsorbtion and desorbtion processes of oxygen molecues is that the photoinduced holes acumulate into the surface increasing the width of the hole-accumulation layer (Figure 5e). Consequently, the resistance of CaFe 2 O 4 decreases in this reaction. The reaction can be explained as in the following equation: Sensors 2020, 20, x 7 of 12 When the light illuminates the materials (Figure 5b), electrons are excited from the valence band to conduction band and electron-hole pairs are generated. The holes react with the pre-chemisorbed O ( ) to form oxygen molecules which will be desorbed from the surface of CaFe2O4  Figure 6 shows a scheme of the sensing mechanism when the material is exposed to the target gas (ethanol vapors) under illumination together with the dynamic response of the sensor (green line) and the corresponding energy band diagram for each situation. When the sensor is exposed to the ethanol vapors, the ethanol molecules are absorbed on the surface and react with photo-induced  Figure 6 shows a scheme of the sensing mechanism when the material is exposed to the target gas (ethanol vapors) under illumination together with the dynamic response of the sensor (green line) and the corresponding energy band diagram for each situation. When the sensor is exposed to the ethanol vapors, the ethanol molecules are absorbed on the surface and react with photo-induced oxygen ions to form water vapor (H 2 O) and CO 2 consuming photo-induced oxygen ions from the surface by releasing electrons (Figure 6b). The reaction can be described as follows: oxygen ions to form water vapor (H2O) and CO2 consuming photo-induced oxygen ions from the surface by releasing electrons (Figure 6b). The reaction can be described as follows: The released electrons will return to the valence band and cause a decreasing concentration of oxygen ions in the surface, resulting in electron-hole compensation and eventually narrowing the hole-accumulation layer. This narrowing process results in an increased resistance when a reducing gas is introduced [50], as shown in Figure 6e. When ethanol vapors are removed from the chamber, the remaining ethanol molecules adsorbed in the surface of the material will eventually desorb through reactions (5) and (6) and be replaced again by adsorbed oxygen molecules returning to the original situation (increase the concentration of hole and the width of the HAL and also resulting in a decrease the electrical resistance of CaFe2O4 (Figure 6c,f). However, the energy needed to either desorb ethanol from the surface or induce reactions (5) and (6) could be higher than that provided by the incident photons, and therefore some of the adsorbed ethanol molecules (or acetaldehyde from reaction (5)) will remain attached to the surface resulting in a state slightly different from the original with a different resistance. Figure 6. The dynamic response (the green line ) toward reducing gas (i.e., ethanol vapors) under visible light irradiation that corresponded to: (a) photo-activation before target gas exposure has been introduced, (b) target gas interacts with photo-induced oxygen ion and attaches on the surface (photo-adsorption process), (c) photo-desorption process when target gas is detached from the surface, (d) energy band diagram photo-activation before interaction with target gas, (e) photo-adsorption process causes a decreasing the width of HAL and increasing of the electrical resistance, and (f) photo-desorption process causes an increasing the width of HAL and resulting in a decrease the electrical resistance of CaFe2O4. Figure 6. The dynamic response (the green line) toward reducing gas (i.e., ethanol vapors) under visible light irradiation that corresponded to: (a) photo-activation before target gas exposure has been introduced, (b) target gas interacts with photo-induced oxygen ion and attaches on the surface (photo-adsorption process), (c) photo-desorption process when target gas is detached from the surface, (d) energy band diagram photo-activation before interaction with target gas, (e) photo-adsorption process causes a decreasing the width of HAL and increasing of the electrical resistance, and (f) photo-desorption process causes an increasing the width of HAL and resulting in a decrease the electrical resistance of CaFe 2 O 4 .
The released electrons will return to the valence band and cause a decreasing concentration of oxygen ions in the surface, resulting in electron-hole compensation and eventually narrowing the hole-accumulation layer. This narrowing process results in an increased resistance when a reducing gas is introduced [50], as shown in Figure 6e. When ethanol vapors are removed from the chamber, the remaining ethanol molecules adsorbed in the surface of the material will eventually desorb through reactions (5) and (6) and be replaced again by adsorbed oxygen molecules returning to the original Sensors 2020, 20, 850 9 of 12 situation (increase the concentration of hole and the width of the HAL and also resulting in a decrease the electrical resistance of CaFe 2 O 4 (Figure 6c,f). However, the energy needed to either desorb ethanol from the surface or induce reactions (5) and (6) could be higher than that provided by the incident photons, and therefore some of the adsorbed ethanol molecules (or acetaldehyde from reaction (5)) will remain attached to the surface resulting in a state slightly different from the original with a different resistance.

Conclusions
CaFe 2 O 4 nanoparticles have been synthesized by a sol-gel auto-combustion method resulting in unconventional metal oxides with bandgap of around 1.9 eV, which is suitable for visible light spectra. Light-activated room-temperature gas sensors based on this material has been tested and validated. The maximum responses toward ethanol vapor and recovery time were 3.6% at 100 ppm,~18 min and~41 min, respectively. The maximum response corresponds to the maximum absorption spectra (425-455 nm) of CaFe 2 O 4 based on results of optical absorption UV-vis diffuse reflectance spectra. The energy provided by the LEDs is sufficient to activate the desorption process between the ethanol and the surface of CaFe 2 O 4 . In addition, it has been confirmed that visible light activation contributed to breaking the bonding of target gases from the surface sensing.