Comparison of CO and CO₂ rf plasma treatment of SnO₂ nanoparticles for gas sensing materials

Their tunable nature and large parameter space make plasmas a useful material modification tool for a variety of applications (e.g., gas sensing, energy storage, biomedical devices, etc.).^1–4 In general, plasmas can functionalize, etch, or deposit a film on the surface of a material, modifying only the outer few layers of the overall substrate. Although plasma parameters also play a role, the plasma precursor often has the largest effect on determining which of these processes (or combination thereof) the plasma exhibits. The impact of the precursor is especially evident in systems where competing processes occur simultaneously. For example, in fluorocarbon plasmas, precursors decompose into species that contribute primarily to film deposition (CF_x radicals) and ones that primarily etch (F atoms). Depending on the precursor gas (and ultimately the dominating species), plasmas like C₃F₈ and hexafluoropropylene oxide deposit amorphous fluorocarbon films, whereas CF₄ and C₂F₆ plasmas result in etching of the material, presumably as a result of the differences in precursor F/C ratios.^5–6 

Another example of these competing processes is found in CO₂ plasmas. Upon plasma ignition, the CO₂ can decompose into CO (contributes to film deposition) and O atoms (an etchant in many systems).⁷ Exploring CO₂ plasmas and the balance between competing processes affords the opportunity to expand our understanding of the underlying chemistry and its impact on the dominant process(es). In particular, in this system, the etching process can be suppressed by employing CO as the precursor gas. CO undergoes minimal decomposition and the likelihood of two CO molecules forming CO₂ (+ O) that can then be decomposed again is low.^7–8 CO is often used as a precursor gas in plasma-enhanced chemical vapor deposition (PECVD) systems to deposit diamondlike carbon films on a variety of substrates.^8–12 Although there have been extensive studies on both CO and CO₂ plasmas individually, studies directly comparing these systems are limited. Kwon et al. examined the effect of additions of CO or CO₂ on the etching rate of tungsten and SiO₂ in a CF₄/O₂ plasma. The etch rate and selectivity of W increased with the addition of either CO or CO₂; however, this increased rate was only seen with additions of up to 1 and 5 SCCM of CO and CO₂, respectively.¹³ The authors state that the addition of small amounts of CO results in a significant decrease in CF_x radicals in the system, leading to enhanced etching. When more than 1 SCCM of CO is added, however, the etch rate decreased when compared with the system without CO, and the surface of the tungsten film had more carbon-containing residues. With the CF₄/O₂/CO₂ system, the addition of CO₂ allowed for the formation of more O and F atoms and a more efficient etching of the substrate. Although these results support the opposing roles of CO and O in a plasma, the CO and CO₂ were additives to a multigas etching system. Thus, elucidating the specific effects of these species on materials in the plasma is complicated, and a comparison of plasmas wherein CO and CO₂ are the major components of the precursor gas could provide valuable insight.

One place where understanding this etching/deposition balance is critically important is the application of CO or CO₂ plasmas in the context of surface modification of materials needed for high-end applications (e.g., sensors). Metal oxide gas sensors are highly sensitive to changes in surface chemistry and morphology as the sensing mechanism relies on gases interacting with the surface of the material to cause changes in bulk resistance. For n-type semiconductors like SnO₂, O₂ will adsorb onto the surface in an oxidizing interaction upon exposure to the atmosphere. This results in an observed increase in material resistance.^14–15 Conversely, reducing gases (like CO, ethanol, and benzene) interact with this adsorbed layer of oxygen and desorb, reducing the material resistance. Moreover, the reactivity of the adsorbed oxygen is highly dependent on the sensor operating temperature (T_S), making most SnO₂ sensors effective only at 300 °C or higher.^14–15 Clearly any changes to the surface chemistry or morphology through either etching or deposition could significantly affect how a target gas might interact with the metal oxide substrate as well as potentially lowering the effective T_S.

When studying gas adsorption on surfaces, Thiel and colleagues’ work has extensively explored the interactions between gaseous molecules with various transition metal surfaces.^16–20 Although these fundamental studies were done on clean, single crystal surfaces, these findings are significant as other work has found that these catalytic metals can enhance the response of SnO₂ sensors to some gases.^21–22 Therefore, understanding how these materials individually interact with gases is critical to understanding the synergistic effect of the composite material. For example, Columbia and Thiel report that when exposed to Pt(111), formic acid readily dissociates into CO₂ and H₂ in addition to simply adsorbing and desorbing from the surface.²⁰ Therefore, when studying gas sensors that combine catalytic metals with metal oxides (e.g., the addition of Pt to a SnO₂ gas sensor), these composite materials may complicate the sensing mechanism. In the above example, a Pt/SnO₂ sensor detecting formic acid will result in a reducing interaction; however, it is possible that the CO₂ formed from the dissociation of formic acid on the Pt surface can also generate a contradicting oxidizing interaction. Thus, it is vital to study the individual interactions of these materials and gases to understand unexpected gas sensor performance or even avoid these combinations when designing devices. Here, our work aims to complement gas adsorption and sensor studies done by Thiel and other researchers when working toward fabricating and interpreting the gas sensing performance of composite transition metal and metal oxide devices. By modifying the surface of SnO₂ sensors with plasma treatments, these studies allow for a better understanding of how changing these materials impact sensor performance and ultimately what surface characteristics are important in gas adsorption and detection.

Previous work in our laboratory modified SnO₂ nanoparticle sensors with various plasma systems (Ar/O₂, H₂, and H₂O_(v))^23–26 to improve sensor performance and elucidate the effects of plasma treatment on the material and the resulting sensing behavior. After O₂/Ar plasma modification, sensors (supported both via a traditional ZrO₂ wafer substrate and a novel complex paper substrate) showed an increase in adsorbed oxygen (O_ads) and demonstrated larger responses to select gases when compared with the untreated (UT) material.^23–25 SnO₂ treated with H₂ and H₂O_(v) plasmas, however, resulted in the SnO₂ being completely reduced to metallic Sn.^25–26 Interestingly, although these systems etched the material similar to the O₂/Ar plasma, this modification resulted in changes that detrimentally impacted gas sensing behavior. Thus, in addition to understanding the etching/deposition competition in CO and CO₂ plasmas, the present work also aims to demonstrate their utilization to improve the gas detection performance of SnO₂ nanoparticle sensors.

SnO₂ substrates were prepared similarly to previously published methods;²⁴ SnO₂ nanoparticles (NP) (diameter <100 nm, Aldrich) were mixed with methanol (ACS certified, Fisher Scientific) to form a slurry that was spread onto approximately 1 × 1 cm² ZrO₂ wafers (50 nm ZrO₂ on n-type 100 Si wafer, BioStar). The substrates were dried under ambient laboratory conditions for at least 4 h prior to use, to allow methanol to evaporate completely. Electrodes for gas sensing studies were made by attaching 1 cm Ag wires using Ag conductive paste (Sigma-Aldrich) and curing for 1 h at 120 °C.

All plasma treatments were performed with an inductively coupled radio frequency (rf) plasma glass barrel-style reactor, described previously.^23–24,27 Briefly, SnO₂ substrates were placed on a glass slide in the center of the coil region, where the plasma was ignited by an Advanced Energy MFX600 power supply at 13.56 MHz. Prior to plasma generation, either CO (≥99.0%, Sigma-Aldrich) or CO₂ (98.00%, General Air) was flowed through the reactor for 5 min with the total system pressure (p) at 150 mTorr. All materials were treated for 5 min followed by another 5 min of only gas flow to quench any active sites and limit undesired reactions with the atmosphere when removing the samples for analysis. The applied plasma power (P) for these treatments ranged from 25 to 100 W.

For optical emission spectroscopy (OES) measurements, the CO or CO₂ pressure was lowered to 135 mTorr to allow for the addition of 15 mTorr of Ar to the system while maintaining p = 150 mTorr. A fused quartz window was installed on the reactor, located in the middle of the coil region, and the substrates were placed directly in line with the window. An Avantes AVASPEC-3648-USB2-RM multichannel spectrometer and accompanying software were used to collect spectra both with and without a substrate using a 50.00 ms integration time and 600 averages. Relative species densities (denoted as [X], where X is the excited state species) were calculated by averaging the actinometric intensity ratio using the 750.3 nm excited state Ar line as the actinometer. [O] and [CO] were determined using the 844.7 and 483.5 nm emission lines, respectively. All calculated values represent values after 2.5 min of plasma exposure; however, spectra were collected every 30 s during the 5 min plasma on time and were consistent throughout the entire plasma on time.

(a) X-ray photoelectron spectroscopy (XPS). All samples were analyzed within 24 h of plasma treatment with a PHI-5800 ESCA system with a monochromatic Al Kα X-ray source (1486.6 eV photons) and spectra were fit with casaxps v2.3 software. For the high-resolution (HRes) XPS spectra, individual binding environments used a Gaussian–Lorentzian (30:70) fit and the FWHM were constrained to ≤2.0 eV. Peak positions were charge-corrected by setting the higher energy Sn3d peak to 486.6 eV and a Shirley background was used for all HRes spectra except for Sn_3d, which used a Tougaard background. All reported values represent the average and standard deviation from three sampling areas on three samples (i.e., n = 9). Atomic ratio percentages for the CO P = 25 W samples were calculated by omitting the Na_KLL contributions from the O_1s and Sn_3d peaks, when present.

(b) Fourier-transform infrared (FTIR) spectroscopy. FTIR measurements were collected on a Thermo Scientific Nicolet 6700 FTIR spectrometer (Madison, WI). KBr pellets were made with ∼0.3 g of FTIR grade KBr (≥99% trace metals basis, Sigma Aldrich), and an analysis of the CO plasma-treated samples were done immediately after plasma modification. The instrument was purged with N₂ gas for >4 h before analysis and the baseline was corrected using the included software (Omnic v8.2).

Gas sensing studies were executed similarly to previously published methods.^16,18 For static sensor response (also referred to herein as “response” or “temperature dependent response”), T_S started at room temperature (25 °C), was increased to 50 °C, followed by additional increases every ∼10 min in 50 °C increments to 300 °C. After ∼20 min at 300 °C, the temperature was lowered by 50 °C increments every ∼10 min (time starting once T_s cooled to the set temperature) to 50 °C and then to 25 °C (i.e., a final 25 °C increment). Sensor response was determined by calculating the ratio of the sensor resistance in air (R_air) to the sensor resistance in the target gas (R_gas) and reported as an average at each T_S (n = 3 for each gas). For response and recovery studies, the sensor was held at a constant T_S and the resistance of the sensor was monitored continuously as a function of time. Target gas flow was turned on for 5 min and then flow was stopped and air was flowed into the sensing chamber for 10 min. During the first ∼3 min of air flow, an increased flow rate was used to quickly flush the chamber of the target gas and then was lowered to 25 SCCM for the remaining target gas off time.

During plasma modification of the SnO₂ NP, excited state species were monitored using OES. Representative raw spectra of the four systems (CO and CO₂ plasmas, with and without SnO₂) at P = 25 W can be seen in Fig. S1,⁴³ for the entire spectral range studied (see supplementary material for raw OES spectra).⁴³ Peaks from the CO Ångström (450–560 nm) and 3rd positive (283–330 nm) bands clearly dominate all four spectra, along with Ar lines (695–911 nm) from the inclusion of ∼10% Ar to the system. Although the CO₂ plasma was generated from a feedgas mixture containing only CO₂ and Ar, the most intense peaks in the spectra are attributed to excited state CO. It is well understood that CO₂ decomposes via Eq. (1),^7,28–31

As seen in Fig. S1 (see supplementary material for raw OES spectra),⁴³ other studies have demonstrated that the CO molecule in both CO₂ and CO plasmas does not efficiently decompose further into atomic C and O.^7–8,30,32 From Fig. 1, which contains expanded views of key spectral areas for the four systems, the main difference between the excited state gas-phase species within these two plasmas is the presence of atomic O (777 and 844 nm). In addition, the CO₂⁺ emission band can also be seen only in the spectrum of the CO₂ plasma [Fig. 1(a)], a feature found in most CO₂ discharges resulting from the A²Σ⁺–X²Π transition of CO₂⁺.³³ 

Using rare gas actinometry, relative species densities were calculated for CO and O (Fig. 2). At all powers investigated, [CO] was higher in both CO plasma systems when compared with the CO₂ systems [Fig. 2(a)]. Under most conditions, the relative species densities with and without an SnO₂ NP substrate are within error of each other, indicating that the types of gas-phase species present are not strongly influenced by the SnO₂ NP material in the plasma and the majority of the excited state species are products of the plasma feed gas. One exception is the [CO] in the CO plasma at 50 and 75 W where the presence of SnO₂ in the plasma leads to a more excited state CO in the gas phase. The [O] was also monitored as a function of P [Fig. 2(b)] and, as similarly seen in the study by Kwon et al., the [O] in the CO₂ plasma is significantly larger than in the CO plasma where [O] ≤ 0.10.¹³ As previously noted, atomic O arises primarily from a decomposition of CO₂ via Eq. (1), but in the CO plasma, the O mainly comes from a decomposition of CO to C and O. Similarly, Yasuoka et al. identified small amounts of atomic O in their CO emission spectra, but these lines were barely distinguishable from spectral noise.⁸ It is also possible that some of the O could be generated from etching the SnO₂ NP or the glass reactor walls. As with the [CO], however, the SnO₂ is likely not contributing because in both the CO₂ and CO plasmas, the [O] is the same (within error) with and without the NP. The numerical values of all relative species densities can be found in Table S1 (see supplementary material for relative species densities).⁴³ Note that under some conditions, small amounts of Sn were observed in the gas phase, an additional indicator of some etching of the NP. This was also seen in previous studies from our laboratory, wherein SnO₂ NP were treated in a H₂ plasma at 100 and 150 W and p = 80 mTorr.²⁶ Similar to the CO₂ system, the H₂ plasma etches the SnO₂ lattice, leading to the generation of excited state Sn and O-containing products observed with OES.

After plasma exposure, NP substrates were analyzed using XPS to understand the impact of plasma treatments on the surface chemistry of these materials. Figure 3 shows a representative HRes spectra of the CO-treated SnO₂ (p = 150 mTorr, P = 25 and 100 W); the calculated XPS atomic compositions for all plasma-treated SnO₂ can be found in Table I, along with previously published results for the UT material.²⁴ After CO plasma treatment at P = 25 W, there is clear evidence of carbon film deposition on the surface, resulting in the increase from ∼8.6% to ∼70.7% C. This is accompanied by a drastic decrease in the percent Sn and O at the surface. The resulting film from this treatment primarily consists of sp³ carbon, but there are also some oxygen functional groups on the surface. Centered at ∼533 eV, the O—C peak is ∼50% of the O composition.

Yasuoka et al. examined diamondlike carbon films grown on Si using a CO rf plasma (p = 2.6–80 Pa, P = 25–200 W).⁸ Both FT-IR spectra and Rutherford backscattering with elastic recoil detection analysis (RBS/ERDA) reveal that the films deposited from CO plasmas were a mixture of C and O (∼76% and ∼20%, respectively, from RBS/ERDA results), very similar in composition to the films grown in this study. Yasuoka et al. did detect ∼4% H incorporation in their films; however, they attributed this to contamination in the reaction chamber as they also analyzed films deposited from C₂H₂ plasmas using the same reactor. Another study deposited compositionally similar films using an atmospheric dielectric barrier discharge CO plasma.³⁴ Gravimetric analysis of the deposited films showed that the samples comprised C (41%), O (51%), and H (3%). Geiger and Staack determined that their hydrogen contamination was the result of atmospheric water absorbing and reacting with the freshly grown film during the scraping process and that the films are “extremely hydroscopic”.³⁴ As the presence of hydrogen in films cannot be explicitly determined with XPS, we did try to limit postdeposition contamination. Specifically, all samples were allowed 5 min of CO gas flow following plasma exposure to quench any remaining active sites on the material with only the precursor gas. The success of this step is difficult to determine from these XPS studies alone as both CO plasma deposited films had most of the C bound as aliphatic C—C/C—H, so it is possible that our films also have some undetected hydrogen incorporation. Preliminary FTIR studies (Fig. S2),⁴³ indicate that both the UT and 25 W CO plasma-treated KBr pellets have similar spectra, making it difficult to definitively determine if our films have some hydrogen incorporation (see supplementary material for FTIR spectra).⁴³ 

As the O—C and SnO₂ lattice O (O_lat) peaks overlap, it is difficult to discern if the O_ads peak is also present in these CO plasma-treated samples, but this is unlikely because film deposition covers the vast majority of the SnO₂. In some of the 25 W-treated samples, the spectra reveal a small amount of Na (1.7%) and corresponding contributions from the Na Auger peaks can be seen in the HRes Sn_3d and O_1s spectra. It is unlikely that this results from sample handling contamination as Cl was not detected in the survey scans (Fig. S3) (see supplementary material for XPS spectra).⁴³ Instead, we believe that the Na arises from the walls of the plasma reactor, which primarily consist of Si, O, and Na, and becomes incorporated into the carbon films during plasma exposure. A similar situation was previously reported in our lab with Si contamination of O₂ plasma-treated TiO₂ NP (p = 200 mTorr, P = 75–220 W).³⁵ The Si in these samples ranged from ∼4 to 21%, and although the NP were on a F:SnO₂ glass substrate, it was determined that the Si was coming from the reactor walls and was dispersed throughout the top 100 nm of the TiO₂ films.

From a compositional perspective, the CO 100 W plasma-treated samples, however, are very similar to the UT material with no increase in percent C on the surface (Table I). This result is somewhat unexpected, given that the depositing behavior usually observed in CO rf plasmas and the analysis of the 25 W CO plasma-treated NP.^8,12,34 Examining the excited-state gas-phase data (Fig. 2), the lack of film deposition aligns with the observed [O] and [CO] trends. As discussed in the Introduction, the CO molecule is believed to be the main species involved in carbon film deposition, whereas atomic O is an etchant. When both species are present, the system could clearly have competing etching and deposition processes, with the overall behavior relying on the dominating species. In the 25 W CO plasma, [CO] is the highest of all conditions investigated and [O] is the lowest, Fig. 2. Thus, deposition is the dominant process under these conditions. At the higher powers, [CO] generally decreases and [O] increases, indicating that the CO plasma transitions from film deposition to an etching dominated system. Some deposition still occurs in the 100 W CO system, as the C_1s spectra on these samples have ∼56% of the C as C—O or C=O, which differs greatly from the adventitious carbon on the UT material, which primarily consists of aliphatic carbon (i.e., C—C/C—H binding environment).²⁴ The thin carbonaceous deposit was clearly not fully etched away by the oxygen in the system. From the O_1s spectra, there are two regions corresponding to O_ads and O_lat, similar to the untreated NP.²⁴ As stated previously, the amount of O_ads is thought to be important for the gas sensing performance of these materials; thus, the O_ads/O_lat ratio was calculated for these samples. For the 100 W-treated samples, the O_ads/O_lat increased from ∼0.254 to 0.411 when compared with the UT SnO₂, suggesting a significant increase in O_ads.

After CO₂ plasma modification, the SnO₂ NP HRes spectra for the 25–75 W plasma-treated samples (Fig. 4) look very similar to the CO 100 W-treated and UT samples. For these conditions, the XPS atomic compositions of Sn and O are within experimental error of the UT SnO₂ NP; however, the O_ads/O_lat increases after CO₂ plasma exposure. This suggests that O within the lattice of the material was etched and then during postplasma gas flow or exposure to the atmosphere, more O adsorbed to these vacant sites. Note that for some spots on the 100 W CO₂-treated materials, an additional CO_x functionality was observed in both the Sn_3d and O_1s HRes spectra. This additional region is likely an indication that along with etching the SnO₂ lattice, the CO₂ plasma resulted in some irregular functionalization of the surface.

Although CO₂ plasmas are not commonly used to modify metal oxides, they are often used for functionalization of polymer membranes for wastewater treatment.^36–39 Compared with an O₂ plasma, CO₂ modification allows for the implantation of O-containing functional groups with less morphological damage and longer treatment stability. In one study,³⁸ XPS analysis revealed that postplasma treatment, polypropylene, polycarbonate, and polysulfone membranes all had increased oxygen content resulting from additional carbonyl and carboxylic acid groups. Water contact angle measurements also indicated O-functionalization and a more hydrophilic surface, with either a significant decrease in contact angle or complete absorption of the water drop during analysis. Clearly, CO₂ plasma exposure readily incorporated oxygen into these soft, polymeric materials. In this work, however, functionalization of SnO₂ was observed only sporadically and under the harshest condition (i.e., the highest P and p), as seen in Fig. 5. One consideration regarding these differences in functionalization comes from a consideration of the energy needs in each case. Functionalization of the polymeric membranes first requires chain scission, which is easiest at aliphatic C sites (bond energies of C—C and C—H are ∼360 and ∼410 kJ/mol, respectively).⁴⁰ In contrast, with the SnO₂ NP substrates, the scission of an Sn—O bond requires a substantially higher amount of energy (a bond energy of 548 kJ/mol⁴⁰). Thus, only within plasmas at P = 100 W do species within the plasma have enough energy to allow for some functionalization similar to that observed with the CO₂ plasma-treated polymers.

Again, the trends observed in the material characterization studies are reflected in the gas-phase OES data. At P ≤ 75  W, the [CO] and [O] are fairly independent of P, with a much larger [O] than [CO]. Thus, even with some CO radicals present to be used in film deposition, the etching behavior of the atomic oxygen dominates enough to not only clean adventitious carbon from the SnO₂, but to also remove any plasma deposited film. Only at P = 25 W did the presence of an SnO₂ substrate increase [O], indicating that some of the oxygen from the material lattice was removed, becoming excited-state O_(g) in the plasma. In addition, this was the only condition wherein Sn was detected by OES, indicating that the plasma was not selectively etching O from the lattice, but rather removing both Sn and O simultaneously. Thus, the 25 W CO₂ plasma-treated sensors had one of the largest increases in O_ads/O_lat compared with the UT NP. The slight increase of this ratio with the 50 W- and 75 W-treated sensors can be attributed to the cleaning of carbon from the SnO₂ rather than etching of the lattice. Finally, in the 100 W system, the [O] was the largest of all the parameters studied and etched the SnO₂ drastically enough such that additional functional groups could be incorporated into the material or adsorbed onto the surface postplasma exposure. Either of these processes increased the amount of O_ads, resulting in the largest increase of O_ads/O_lat.

For this study, the plasma-modified SnO₂ substrates were tested as gas sensors to determine the effect of these treatments on device performance, as gas detection with these materials relies heavily on surface chemistry and morphology. The response (where response = R_air /R_gas at each T_S) of the UT SnO₂ to CO, ethanol, and benzene is shown in Fig. 6(a) at a variety of T_S. As the UT sensors have the largest response to benzene at T_S = 200 °C, these were the conditions selected for further response and recovery studies [Fig. 6(b)]. In these dynamic studies, the UT sensor showed no response when benzene gas flow was turned on (indicated by the green solid lines) or recovery behavior when benzene flow was turned off (indicated by the red dashed lines). A similar lack of response or recovery behavior of these UT sensors was also seen with the other analytes. This is consistent with previous results from our laboratory for Ar/O₂ plasma-treated SnO₂ NP sensors,¹⁶ demonstrating that as constructed here, these UT materials are not ideal gas sensors.

Representative raw resistance plots of the CO₂-and CO plasma-treated materials are shown in Fig. 7. With the CO₂ sensors, we observe small fluctuations in the resistance even when T_S is held constant, which is typical of these sensors.²⁵ In Fig. 7(a), the sensor clearly demonstrates the expected behavior, where the resistance decreases as T_S is increased from 25 to 300 °C in the first half of the gas exposure (from time 0 to ∼93 min). As T_S is decreased back to 25 °C, the resistance returns approximately to 175 kΩ. This pattern is also seen with the other CO₂ plasma-treated sensors in air, CO, and ethanol. The CO plasma-treated sensors, however, do not follow this expected behavior. For example, from ∼13 to 31 min (when T_S = 50 °C) the sensor resistance increases from ∼249 to ∼300 kΩ, whereas typical fluctuation is ∼10 kΩ. In addition, the final resistance (when T_S returns to 25 °C) is at least half that of the resistance at the start of the experiment. Both behaviors persist with the other CO modified sensors and with the other gases used in these sensing studies. These data align with XPS data that suggest surface carbon arises from a plasma deposited film, and this film prevents interactions of the analyte gas with the metal oxide surface during gas sensing experiments. Although film formation from CO plasma exposure was detrimental to sensor performance in this work with SnO₂ gas sensors, this is not true for all situations. When the deposited film acts as the active sensing material, plasma deposition is especially beneficial when working with a morphologically complex supporting substrate. For example, an O₂ rf plasma was used to graft polyaniline onto polyester fabric, creating a successful room temperature NH₃ gas sensor.⁴¹ The polyaniline sensors achieved a sensitivity [(R_gas − R_air)/R_air] of 2.49 to 40 ppm NH₃ with response and recovery times of ∼126 and ∼162 s, respectively. In other applications, contexts such as industrial coatings, CO plasmas are often used to deposit diamondlike carbon films via plasma-enhanced chemical vapor deposition.^8–10,41

The unstable sensor resistance in all gases (air, CO, benzene, and ethanol) did not allow for the calculation of response values for the CO plasma-treated sensors, so only the CO₂ plasma-treated sensor response values are reported (Fig. 8). The sensors treated with lower power plasmas (P = 25–75 W) demonstrated an increase in response to benzene and CO when compared with the UT sensors. The 25 W CO₂-treated sensor had the largest response to benzene of ∼35 at T_S = 250 °C, followed by the 75 W-treated sensor (CO response ∼16 at T_S = 150 °C), and finally, the 50 W-treated sensor (benzene response ∼6 at T_S = 150 °C). From the XPS analysis, the 25 W sensors also had the largest O_ads/O_lat, followed by the 75 W and finally 50 W sensors. These results highlight the importance of O_ads during gas detection and the relationship between the material surface and observed device performance. The response of the 100 W-treated sensors to all the analytes is much less than that of the UT sensors. This indicates that the partial functionalization of this material changed the type of interaction of the analyte gas with the SnO₂ surface. Instead of all these gases interacting with the O_ads on the SnO₂, the gases may be adsorbing directly onto the surface of the metal oxide in an oxidizing interaction. This would result in the target gases causing a decrease in material resistance and ultimately a sensor response <1 to these gases. The switching of sensor behavior to CO and benzene was seen in previous work in our lab studying O₂/Ar- and H₂O_(v)-treated SnO₂ gas sensors.²⁵ 

Figure 9 shows the ethanol response and recovery behavior of a P = 25 W CO₂ plasma-treated sensor. T_S = 250 °C was used for these experiments as the static studies revealed that the sensor demonstrated the largest response to ethanol and benzene at this temperature. After the first exposure, the sensor exhibits a distinct decrease in resistance; however, this decrease in signal gets smaller with subsequent ethanol exposures. In addition, there seems to be very little signal recovery once the ethanol flow is turned off and the air flow turned back on. Note, however, that the resistance stabilizes after the first exposure, indicating that the sensor could be reused once. Subsequent ethanol exposures demonstrate that repeated use after that may result in false readings due to the instability of the sensor resistance. This behavior indicates that after the initial interaction with ethanol, the surface of the SnO₂ becomes poisoned and does not allow for the reversable adsorption and desorption of gaseous species, ultimately stopping the gas detection process.

Recent work by Hu et al. examined the H₂ sensing performance of Ar plasma-treated ZnO–SnO₂ nanofibers.⁴² Similar to our work, the authors sought to improve sensor performance by etching the lattice of the metal oxides to form O vacancies and ultimately increase O_ads on the surface. With Ar plasma treatment, they increased the sensor response to 100 ppm H₂ from 10 to 18, lowered the T_S by 30 °C, and decreased the response time from 69 to 24 s. Also, their plasma-treated sensors demonstrated good response and recovery behavior after four 500 ppm H₂ exposures; however, increasing the treatment time had an adverse effect on sensor repeatability. Although a direct comparison between this work and the ZnO-SnO₂ sensors cannot be made, both sensors had similar responses to their target gases. The CO₂ plasma-treated sensors studied here, however, were able to achieve these results at a lower T_s than the 300 and 330 °C used with the ZnO–SnO₂ sensors. Although this work only has a small (50 to 150 °C) decrease in operating temperature, this is a promising indication that CO₂ plasma modification can be used to improve the gas sensing performance of SnO₂ NP.

Plasma systems with competing deposition and etching processes are important tools for surface modification of a range of high-performance materials. Here, CO and CO₂ plasma treatment of SnO₂ NP sensors were compared and evaluated as a modification strategy to improve the gas sensing performance of these devices. From OES measurements, excited state species in both systems were similar, but additional signals from CO₂⁺ and O were detected in the CO₂ plasma. Calculated relative species densities show that in general, the CO₂ plasmas have a higher [O] and the CO plasmas have a larger [CO], suggesting that etching dominates in the former, whereas deposition dominates in the latter. From the XPS analysis of the NP, the CO plasma indeed deposited an amorphous carbon film, whereas the CO₂ plasma primarily etched the SnO₂ lattice and increased the amount of O_ads.

When utilizing these treated materials as a gas sensor, the film deposited on the CO-modified sensors proved to be detrimental to the gas sensing process by impeding analyte-SnO₂ interaction and not allowing for the calculation of the sensor response to any of the selected gases. CO₂ plasma-modified sensors treated at P = 25–75 W had much higher responses to benzene and CO compared with the UT sensors, and the 25 W-treated sensor demonstrated response and some recovery behavior during dynamic gas sensing studies. It should be noted, however, that the 100 W CO₂ treatment resulted in nonuniform functionalization of the NP, greatly reducing the response of these sensors compared with the UT ones. Further work with higher P CO₂ plasma treatments is needed to determine if the addition of CO_x groups to the SnO₂ lattice directly results in producing oxidizing interactions with these reducing gases and a sensor response <1. Ultimately, this work identified that the etching versus deposition processes in the CO and CO­₂ systems primarily depends on the behavior of the dominating species (i.e., CO or O). Perhaps more importantly, these data demonstrate that for gas sensing applications, the etching behavior of the CO₂ plasma is more desirable.

This work was funded by the National Science Foundation (No. NSF CBET-1803097). The authors would like to thank the staff of the Colorado State University Analytical Resources Core for assisting with the materials characterization data. All work was completed at/in association with Colorado State University.

The data that support the findings of this study are available within the article and its supplementary material or from the corresponding author upon reasonable request.