1722
J. Phys. Chem. A 2010, 114, 1722–1733
Gas-Phase Chemistry in Inductively Coupled Plasmas for NO Removal from Mixed Gas Systems Michelle M. Morgan, Michael F. Cuddy, and Ellen R. Fisher* Department of Chemistry, Colorado State UniVersity, Fort Collins, Colorado 80523-1872 ReceiVed: September 8, 2009; ReVised Manuscript ReceiVed: NoVember 24, 2009
Inductively coupled rf plasmas were used to investigate the removal of NO from a variety of gas mixtures. Laser-induced fluorescence and optical emission spectroscopy were employed to measure the relative gasphase density of NO as a function of the applied rf power, gas mixture, and catalytic substrate type. In general, the overall density of NO decreases as a function of applied rf power in both NO and N2/O2 plasmas, but the addition of gases such as H2O vapor and CH4, as well as the presence of Au-coated substrates, significantly affects the behavior of NO in these systems. Rotational and vibrational temperatures for NO were measured using laser-induced fluorescence excitation spectra and optical emission spectra. Results show NO vibrational temperatures are about a factor of 5 higher than rotational temperatures and indicate little dependence on applied rf power, feed gas composition, or overall system pressure. Possible mechanisms for the observed changes in [NO] as well as the rotational and vibrational temperature data are addressed. I. Introduction Nitrogen oxides (NOx) have been studied extensively because of the harmful reactions that occur when they are released into the atmosphere. Understanding the primary and secondary reactions involving nitric oxide and its byproduct deserves considerable attention because they are produced from anthropogenic sources such as vehicle exhaust and power plants. In a longitudinal study performed by the Environmental Protection Agency (EPA), the emissions of six primary air pollutants, NOx, carbon monoxide, lead, particulate matter, sulfur dioxide, and volatile organic compounds (VOCs), were tracked over an 18 year period.1 The results showed the emissions of many of these air pollutants were significantly reduced, which was attributed to the introduction of catalytic converters into diesel and regular unleaded gasoline fueled vehicles. During the same time period, however, NOx emissions increased by ∼10%, clearly suggesting a need for the development of new methods to diminish NOx emissions. One way to remove atmospheric pollutants from engine exhaust is to use a three-way catalyst (TWC).2 The primary role of catalytic converters is the transformation of regulated pollutants such as NOx, CO, hydrocarbons, and sulfur dioxide into less toxic compounds, as shown in reactions 1-3, demonstrating NO is ideally converted into N2 and O2. In a TWC, both catalysts are composed of metal oxides and other catalytic materials such as platinum, palladium, and rhodium.3-9 Platinum is an important component in both the oxidation and the reduction catalysts to aid in the removal of pollutants. Rhodium is specifically used in the reduction catalyst because it converts NO to N2. More recently, gold has been proposed as a possible catalyst material for TWCs,10-13 primarily for use in the oxidation catalyst; however, it may also aid in the desirable processes occurring in the reduction catalyst. Although current catalytic converters function adequately to meet the standards set by the EPA for NOx emissions, they cannot meet new emissions laws, which propose lower allowable concentrations * Corresponding author. E-mail:
[email protected].
for emitted pollutants. Thus, improvements in catalyst efficiency will need to be made.
2NOx f xO2 + N2
(1)
2CO + O2 f 2CO2
(2)
CxHy + (x + (y/4))O2 f xCO2 + (y/2)H2O
(3)
One possible approach to improving the performance of materials used in catalytic converters employs plasmas to either modify the catalyst surface or enhance removal or excitation of the pollutants themselves. Previous studies concentrated on the removal of nitrogen oxides using dielectric barrier and corona discharges, along with microwave and hollow cathode plasmas.14-21 Small amounts (typically hundreds to thousands of parts per million) of NO were added to these systems to mimic the amount of NO found in vehicle exhaust. Other plasma-catalytic studies focused on different variables such as catalyst preparation, type of catalyst used, and type of exhaust (e.g., diesel exhaust instead of gasoline).22-24 Generally, these studies focused on the overall processes occurring as a function of applied plasma power, and not on complete removal of NO. Notably, the feed gases used were those found in all exhaust systems, with the more complex gas mixtures roughly representing the ratios of the gases found in exhaust. Some of these studies reported on the species that were formed in the plasmas that could possibly initiate additional reactions, but did not explore the formation mechanisms or the parameter dependence of these species. There is also a dearth of information on the temperatures of plasma species in these systems. Although there are a few reports of kinetic and thermodynamic modeling of plasma decomposition of pollutants such as NO,25-29 the development of accurate computer models for NO removal will require knowledge of both internal and kinetic temperatures of plasma species as well as a thorough understanding of the gasphase and gas-surface reactions occurring in these systems. One goal of the study described here was to understand the fundamental gas-phase processes occurring between the various exhaust gas species using NO-based plasmas. For example, as the humidity in the system increases, the amount of NO removed
10.1021/jp908684c 2010 American Chemical Society Published on Web 01/05/2010
Gas-Phase Chemistry in Inductively Coupled Plasmas
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1723
Figure 1. Schematic diagrams of (a) the inductively coupled plasma reactor with expanded gas manifold used for the NO-OES studies, and (b) high vacuum apparatus used for LIF density and NO rotational temperature studies. In this partial schematic diagram for the imaging of radicals interacting with surfaces (IRIS) apparatus detailed elsewhere,42 the ICCD is located perpendicular to the molecular and laser beams.
from automobile exhaust generally increases. Likewise, hydrocarbons are also found in exhaust fumes, and these can complicate the overall formation or removal of NO. Here, we use actinometric optical emission spectroscopy (OES) and laserinduced fluorescence (LIF) to examine the behavior of NO in an inductively coupled rf reactor as a function of applied rf power (P), gas mixture, and type of catalytic substrate. In addition to NO plasmas, we explore the formation of NO in N2/O2 plasmas and examined the effects of both water vapor (∼5%) and methane addition (∼3-5%) on removal of NO.30,31 Another goal of this study was to determine internal energies of NO and to examine the parameter dependence of the energy partitioning in these plasmas. To that end, we have measured the rotational and vibrational temperatures of gas-phase species using LIF and OES spectra, respectively. For comparison, the vibrational temperatures of N2 in some of the systems were also determined. These temperatures allow additional insight into the gas-phase formation mechanisms in the NO plasmas. II. Experimental Details General. The plasma reactor used for the OES experiments, Figure 1a, has been described in detail previously.32 Briefly, the feed gases enter into a glass tubular reactor where 13.56 MHz radio frequency (rf) power (RFPP 5S power supply) was
coupled to an eight-turn nickel plated copper coil to ignite the plasma and was tuned with a Jennings 100 pF variable capacitor. Gases entered the reactor directly prior to the induction coil and were removed using a Franklin Electric mechanical pump with a pumping speed of 4.2 L/s. The pressure in the chamber was monitored with an MKS Baratron capacitance manometer, which is insensitive to gas composition. The total reactor pressure, regardless of gas mixture, was held at 100 mTorr. The most significant change made to the reactor setup from previous studies was an expanded gas manifold, Figure 1a, comprising six gas lines. To ensure complete sealing, a Nupro butterfly valve was used for each gas line. Gas flow for five of the lines entering into the system was controlled by MKS mass flow controllers. The gases flowing into inlets 1-5 were nitrogen (General Air, 99.98%), oxygen (General Air, 99.5%), nitric oxide (Matheson Trigas, 99%), methane (Air products, 99.99%), and argon (General Air, 99.985%). In addition, water vapor was introduced to the gas manifold system via inlet 6, which had a Nupro-bellows-sealed metering value attached to a heated liquid bulb containing deionized water (18 MΩ cm). Prior to use, the water was subjected to several freeze-pump-thaw cycles to remove trapped atmospheric gases. Argon was added to the system as an actinometer and was maintained at 5-7% of the total reactor pressure for all experiments. Using N2 as the model,
1724
J. Phys. Chem. A, Vol. 114, No. 4, 2010
we have calculated the residence time in our reactor at 100 mTorr as ∼12 s. In some OES studies, two substrates were placed at 5 and 12 cm downstream from the downstream end of the rf coil, Figure 1a, to examine the effects of catalytic materials on the gas-phase chemistry.33 The substrates were silicon wafers, goldcoated Si wafers (5 nm Au), and platinum foil (Alfa Aesar, 99.99%, 0.025 mm). For Au-coated Si wafers, the Au was sputtered onto the Si substrate using a Hummer VII Sputtering System (Anatech Ltd.) immediately before placement in the plasma reactor. No major differences were observed between substrates placed at the two different locations; thus, data shown are for substrates placed 5 cm downstream. Gas-Phase Analyses. OES was used to characterize the excited-state species present in the relevant plasma systems. Optical emission spectra were acquired using a replaceable fused quartz window at the downstream end of the reactor, allowing for the coaxial collection of plasma emission for maximum signal intensities. Plasma emission was imaged onto the 10 µm entrance slit of an Avantes multichannel spectrometer fitted with four optical fibers fused into one cable. The spectrometer houses four fiber gratings yielding a 0.1 nm resolution, four 3648 pixel charge coupled device (CCD) array detectors, and a combined wavelength detection range from 187 to 1016 nm. Emission signal integration times ranged between 1 and 500 ms. Error limits reported represent the standard deviation of the mean of four measurements. As discussed in detail elsewhere,34-36 actinometry allows the determination of relative density for plasma species. In its simplest form, the ratio of the emission intensity from the excited-state species of interest to that of an actinometer is taken as the relative number density for the ground-state species of interest. Usually, a rare gas, such as Ar, is used as the actinometer. Thus, small amounts of Ar were added to our NOcontaining plasmas to provide a reference for calculation of the relative number density of NO, along with other emitting species produced in our plasmas. For all actinometric values reported, the Ar emission line at 750.4 nm was used for comparison. The limitations of OES for production of reliable density data have been discussed in detail previously.34 Specifically, the actinometer and the species of interest must both be excited primarily from the ground state; the excited species decay must occur via optical emission or optical emission accompanied by a parallel de-excitation pathway with a constant branching ratio; and the electron impact cross section for both species must have similar thresholds and shapes as a function of electron energy. There is good agreement between the shapes of the NO and Ar electron impact cross sections (with peaks occurring at ∼20-40 eV), and with the total cross section peak values ((∼40-45) × 10-19 cm2).37,38 Raw OES spectral lines in the A2Σ+-X2Π transition for NO were compared to simulated emission lines from the simulation tool LIFbase39 to calculate vibrational temperatures, ΘV, for NO. Vibrational temperatures for NO were obtained for plasmas with NO/Ar, N2/O2/Ar, NO/CH4/Ar, and NO/H2O(5%)/Ar gas mixtures, as well as each of these systems with Au, Pt, and Si substrates placed in the reactor. For comparison to other plasma species, OES spectra were also used to calculate ΘV for N2 using the simulation tool PGOPHER.40 Similar to the LIFbase package, this program models the vibrational spectra of a polyatomic molecule. The transition used here was the A3Σu+-X1Σg+ transition of N2. Laser-induced fluorescence (LIF)41 data for NO were collected using a high vacuum, plasma molecular beam apparatus, Figure
Morgan et al. 1b, described in detail previously.42 Briefly, the desired gases enter through an inductively coupled plasma produced in a glass tubular reactor similar to that shown in Figure 1a. The pressure in the plasma source was maintained at 100 mTorr (as measured with a Baratron capacitance manometer prior to ignition of the plasma) for all experiments, regardless of gas mixture. The applied rf power range was 25-240 W for the LIF experiments, and the residence time was ∼8-10 s in the reactor source. The plasma expands into a differentially pumped vacuum chamber (base pressure ) ∼1 × 10-6 mTorr) and is collimated using a set of two narrow slits, 0.90-1.20 and 1.15-1.40 mm wide, respectively. As the plasma molecular beam enters the highvacuum region of the apparatus, it is an effusive molecular beam that consists of nearly all of the species that are present in the plasma. An excimer-pumped (Lambda Physik LPX210i, XeCl, 180 mJ/pulse, 35 Hz) tunable dye laser beam (Coumarin 460, Exciton) propagates in the principal scattering plane and intersects the molecular beam at a 45° angle. The laser is tuned to an absorption frequency of a particular species, thereby selecting one quantum state of a single species in the plasma molecular beam. Spatially resolved LIF signals are collected by a time-gated, intensified charge coupled device (ICCD) located perpendicular to both the molecular beam and the laser beam.42 Measurements were made as a function of plasma parameters, such as gas composition, reactor pressure, and P. Tunable laser light in the 226.100-227.000 nm range was produced by frequency doubling the output of the laser using a BBO1 doubling crystal. For NO density studies, a single wavelength (226.199 nm) was used to excite the NO. Instead of using an off-resonance wavelength, the laser light was blocked when collecting background images. Rotational temperatures were determined by collecting the total LIF intensity as a function of laser wavelength (226.100227.000 nm). Here, we used steps of 0.001 or 0.002 nm and smaller wavelength ranges to probe the A2ΣrX2Π transition of NO.43,44 The rotational excitation spectrum shown here was collected using a step size of 0.001 nm. Depending on the power output of the laser, the collection parameters of the ICCD camera were adjusted because under many conditions, the NO signal intensity surpassed the allowed intensities of the camera. The camera gate width was between 75 and 1000 ns, and the number of gates per exposure ranged from 75 to 450. The peak positions and intensities were compared to calculated peaks in the LIFbase simulation program39 wherein rotational temperature, ΘR, can be changed to determine the experimental ΘR(NO). At least three sets of peaks are compared to calculate error in ΘR. Surface Analyses. XPS analyses were performed on a Physical Electronics PHI-5800 ESCA/AES system. To minimize surface contamination from exposure to atmosphere, substrates were transferred from the plasma reactor to the XPS for analysis directly after treatment. Survey and high resolution spectra were collected using a monochromatic Al KR X-ray source (1486.6 eV), hemispherical analyzer, and multichannel detector. A photoelectron takeoff angle of 45° was used for all measurements. Survey spectra were collected with a pass energy of 93.90 eV, and high resolution spectra were acquired at an analyzer pass energy of 23.50 eV. Percent compositions were calculated from high resolution spectra, and values reported are the mean and standard deviation of three measurements from one to two samples. Curve fitting of the high resolution spectra of plasma treated substrates was performed using Gaussian functions. Binding energy scales were referenced to the C1s peak at 284.8 eV.
Gas-Phase Chemistry in Inductively Coupled Plasmas
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1725
Figure 2. Raw OES spectra of plasmas with different feed gases: (a) 95/5 NO/Ar; (b) enlarged NO peak series for the spectrum shown in (a); (c) 70/24/6 N2/O2/Ar; and (d) 94/6 NO/Ar with a Au-coated Si wafer. Each spectrum was acquired at a total system pressure of 100 mTorr with P ) 50 W. Major species include NO, N2, O (4), and Ar (]). The Ar line marked in both (a) and (c) is the line used for actinometric measurements at 750.4 nm.
The morphology of Au-coated films was measured using an atomic force microscope (Nanosurf Easyscan 2) in static force mode. Silicon tips (Nanosensors, PPP-NCLR) with a tip radius 200 W, however, the relative intensity decreases in the system with water vapor, whereas with methane, the signal intensity plateaus at P > 125 W. Notably, the additives do decrease [NO] at all P, relative to the system without any additives. Similarly, Figure 5b compares [NO] in the N2/O2/ H2O/Ar and NO/H2O/Ar plasma systems, where the behavior of [NO] is significantly different between the systems. The NO signal as a function of P in the NO/H2O/Ar system looks very similar to that in the same system without any H2O vapor added, Figure 4, whereas the NO emission intensity increases somewhat with P in the N2/O2/H2O mixture. Given the differences in the behavior of [NO] shown in Figure 5, we explored the effects of P on the emission intensities of other species in systems with H2O vapor. Figure 6a and b shows the effects of P on the intensity of NO, N2, OH, HR, and O in the NO/H2O/Ar and N2/O2/H2O/Ar systems, respectively. Similar to NO, the signal intensity of N2 decreases as P increases. In contrast, the signal intensities for HR and O increase with P, likely due to increased fragmentation at higher P. The behavior for OH is relatively flat at P < ∼100 W, but then increases dramatically at higher P. The emission behavior for all species in these two systems is very similar, with the primary difference being the emission intensities are significantly lower in the N2/ O2/H2O/Ar system. The effects of 5% methane addition were also examined in the NO/Ar and N2/O2/Ar systems, Figure 7. Similar to what was observed with H2O vapor addition, the P dependence of
Gas-Phase Chemistry in Inductively Coupled Plasmas
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1727
Figure 6. Relative intensity of plasma species (x) as a function of P for (a) NO/H2O and (b) N2/O2/H2O plasmas. The error bars are one standard deviation of the mean of four measurements.
[NO] in the two systems is very different, Figure 7a, with the intensity of NO decreasing as a function of P with CH4 addition. In contrast to the H2O additive systems, however, the emission behavior of other plasma species is much more complex and lends insight into the underlying chemical processes occurring in this system, Figure 7b and c. In both systems, the carboncontaining species have relatively low emission and virtually no P dependence. This is somewhat surprising, especially for the CH signal, as increased power generally leads to increased fragmentation of hydrocarbons.45 The atomic species O and HR also behave similarly in both systems, increasing with applied rf power for P e 150 W, and reaching a plateau above 150 W. One difference between these two systems is that there appears to be much more dissociation of the feed gases in the N2/O2/ CH4/Ar system, as evidenced by the more dramatic decrease in [N2] with increased P and the higher concentrations of HR. Figure 8 shows the relative intensity of NO in the N2/O2/ H2O/Ar plasma system with 5-20% added water vapor. For all [H2O], the dependence of NO emission intensity is very similar with a peak in signal strength occurring at P ) ∼200 W. As the concentration of water vapor in the system increases, the intensity of NO decreases, with nearly a factor of 2 reduction from 5% to 20% added water vapor. Because the trends do not change with [H2O], 5% water vapor was used in most of the studies reporting the addition of water vapor. Figure 9 shows the effects of three different substrates, Si, Pt, and Au, on the relative intensity of NO. The P-dependent behavior of [NO] is not completely independent of the substrate placed in the reactor. Relative to the results with no substrate in the system, the Au substrate appears to be better at removing NO from the system at high P. The other two substrates do not improve the removal of NO from the plasma in comparison to
Figure 7. Relative intensity of NO as a function of P (a) for NO/CH4 and N2/O2/CH4 plasmas. Relative intensity of the emission from various species in (b) NO/CH4 and (c) N2/O2/CH4 plasmas. The error bars are one standard deviation of the mean of four measurements.
Figure 8. Relative intensity of NO in N2/O2/H2O/Ar with varying amounts (5-20%) of water vapor as a function of P. The error bars are one standard deviation of the mean of four measurements.
the system without any substrates present. Although the trends for removal of NO are similar, Au-coated substrates are the best choice of the substrates examined, despite the observation that the NO signal does not completely disappear with any of the substrates.
1728
J. Phys. Chem. A, Vol. 114, No. 4, 2010
Morgan et al.
Figure 9. Relative intensity of NO in NO/CH4 with no substrate (red b), Au-coated Si wafer (green 9), Si wafers (yellow 2), and Pt foil (blue 1) as a function of P. The error bars are one standard deviation of the mean of four measurements. Substrates were placed 5 cm downstream from the coil.
Figure 10. Power dependence data for various species in a 94/6 NO/ Ar gas mixture with a freshly treated (5 cm downstream from the coil) Au-coated Si wafer and a 1 week aged sample, which was used in the previous experiment. Error in these measurements was extremely small; thus, error limits are shown, but are smaller than the size of the individual data points.
One issue with any catalyst is the effect of aging. Specifically, upon exposure to atmosphere or other contaminants, catalytic surfaces can become “poisoned” and lose catalytic activity. Here, we explored our Au-coated substrate’s ability to retain its catalytic properties using a 94/6 NO/Ar plasma, Figure 10. In these experiments, a freshly sputtered Au-coated Si wafer was placed in the reactor (5 cm downstream from the coil), the plasma was ignited, and spectra were collected. The substrate was removed from the reactor and left, covered, under ambient laboratory conditions. One week later, the same wafer was used under the same plasma conditions. Clearly, the Figure 10 data demonstrate that the amount of NO removed from the system as P is increased is significantly lower with the aged substrate. This effect suggests exposure to the plasma alters the Au surface in such a way to diminish its catalytic properties. As plasmas can affect both the chemical and the physical properties of surfaces, we used XPS and SEM to explore possible compositional and morphological changes to Au-coated substrates exposed to NO/Ar plasmas. Untreated substrates had an elemental composition comprising Au (∼50%), O (∼32%), C (∼16%), and Si (∼2%), with the trace amounts of Si likely arising from the underlying Si substrate. Upon exposure to any of our plasmas, the XPS elemental composition changes, with significantly more Si (∼25-30%) and O (∼40-45%) observed on the substrate. This suggests one of two things is occurring during plasma treatment: (1) the plasma is etching the walls of the glass reactor and SiO2 is redepositing on the substrate; or
Figure 11. SEM images (50 000×) of Au-coated Si wafers (a) untreated, and placed 5 cm downstream from the coil and treated with 100% NO plasmas for 30 min with (b) P ) 50 W and (c) P ) 300 W.
(2) the plasma is serving to remove the Au-coating, thereby exposing the underlying substrate. Similar chemical compositions were observed for substrates exposed to NO-based plasmas, suggesting the effect on the surface of each substrate was independent of gas mixture. Note that other, similar studies using Pt foil and Al2O3 substrates did not show additional quantities of Si on the surface after plasma treatment. This further indicates sputtering of the reactor walls is not likely to be the source of the Si we observe in the XPS analysis of the Au substrates. Figure 11 contains SEM images for Au-coated Si substrates, one untreated, Figure 11a, and two treated in 100% NO plasmas with different P, Figure 11b and c. These images reveal some changes in the morphology of the substrates upon plasma treatment, regardless of P. Specifically, the treated substrates appear to have less gold on the surface, as evidenced by the apparent island formation seen in Figure 11b and c. This
Gas-Phase Chemistry in Inductively Coupled Plasmas
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1729
Figure 12. Vibrational temperatures for NO as a function of P measured in plasmas without substrates, and with Au, Pt, or Si using mixed gases of the following composition: (a) NO/Ar, (b) N2/O2/Ar, (c) NO/CH4/Ar, and (d) NO/H2O(5%)/Ar. Error bars are estimates of error from PC values acquired in the simulation process (see text).
suggests that the second mechanism proposed above may be the more plausible, as the “bare” spaces seen in these images are likely the underlying Si substrate. Interestingly, AFM data showed no significant difference in the roughness factor or the peak-to-peak ratios for these substrates, which would indicate damage to the surface. NO Vibrational Temperatures. The vibrational temperature, ΘV, of NO was determined from OES spectra using the simulation tool LIFbase.39 Figure 12 contains ΘV values obtained from comparison of the simulation to experimental OES spectral lines. The peak correlation (PC) between experimental and simulated peaks was maximized in determining each value of ΘV, yet remained in the range 0.8 < PC < 1.0 for each OES spectrum analyzed. Such PC values are not unprecedented as Howle et al. report peak correlations of 0.75 for a similar analysis of CH and CN emission from photodissociation of CH3CN.48 The lower PC values can be attributed to spectral overlap between NO emission lines and emission from other NO transitions as well as emission from other plasma species. Thus, the error for our ΘV values was derived from the PC value, as it provides a reasonable estimate of the error in the simulation of the experimental spectrum. As is evident from Figure 12, there is little to no dependence of ΘV upon gas mixture or substrate, but there appears to be a slight P dependence. For comparison, the OES spectra were also used to determine ΘV(N2) using an alternate spectral simulation program, PGOPHER.40 For N2, we find there is little dependence on plasma or substrate type, similar to what was observed for ΘV(NO). The ΘV values determined for N2, however, appear to be much cooler, at about 400-500 K. It should be noted that these ΘV values are for NO* and N2* found in our plasmas and are not necessarily representative of all plasma species. This is discussed further below. LIF and NO Rotational Temperatures. LIF spectroscopy provides information on the gas-phase density of ground-state
Figure 13. NO LIF intensity as a function of percent nitrogen in N2/ O2 plasmas using P ) 50 and 100 W. The error bars are one standard deviation of the mean of three measurements.
NO in our plasmas for comparison to the relative densities acquired by OES. LIF excitation spectra were also used to determine ΘR(NO) in plasmas containing different gas mixtures. Ground-state NO is formed in N2/O2 plasmas, Figure 13. As expected, the LIF intensity decreases as the amount of O2 in the system is reduced. Interestingly, the peak intensity occurs at very low levels of N2 (i.e., 10-20%), suggesting that it is the amount of oxygen in the system that affects formation of NO. There is significantly more NO produced at the higher P with %N2 ) 20-50%, however, clearly indicating the rf power serves to dissociate the feed gases, allowing the formation of NO. The effect of adding H2O and CH4 on [NO] in N2/O2/NO plasmas is shown in Figure 14. With one exception, there is a slight decrease in the intensity of the NO LIF signal as a function of P for all systems, similar to what was observed in the OES
1730
J. Phys. Chem. A, Vol. 114, No. 4, 2010
Figure 14. Normalized LIF intensity data for NO as a function of P for N2/O2/NO plasma systems containing different amounts of water vapor: 3.0 mTorr CH4 (1), 4.7 mTorr H2O (red 2), 10.7 mTorr H2O (green 9), 14.8 mTorr H2O (blue b), and 20.1 mTorr H2O (pink [). The error bars are one standard deviation of the mean of three measurements. Total reactor pressure was 100 mTorr for all experiments.
Figure 15. Fluorescence excitation spectrum for NO (a) acquired using a 50 W 100% NO plasma molecular beam, and (b) a calculated spectrum with ΘR ) 335 K for the NO A2Σ r X2Π transition.
data for the N2/O2/NO plasma, Figure 4. In the CH4 system, there is a slight reduction in the NO as P is increased, whereas the NO intensity in the H2O system is relatively constant with P at higher [H2O]. The exception to this trend is the data acquired using the lowest [H2O], 4.7 mTorr. These data, however, resemble the OES data for the same plasma system, Figure 5b. Additional LIF density data for NO in other plasmas also follow trends similar to those observed with the NO OES data. Overall, these data suggest that the OES and LIF densities track with each other in all of the systems studied here. Figure 15 shows an experimental fluorescence excitation spectrum for NO obtained using a 50 W 100% NO plasma along with a calculated spectrum for NO A2ΣrX2Π assuming ΘR ) 335 K. The excellent agreement in line positions between the experimental and calculated spectrum verifies the fluorescing
Morgan et al. species is indeed NO. A comparison of the relative line intensities using a linear least-squares method yields ΘR ) 337 ( 29 K for NO under these conditions. With the same method, ΘR(NO) values for plasmas with different gas mixtures and at different P were determined, Table 1. Interestingly, there is no real dependence on either gas composition, overall system pressure, or P, with ΘR(NO) ) ∼330 K under all conditions. For the 100% NO and NO/Ar (12/88) plasmas, there does appear to be a slight increase in ΘR(NO) with increased P, but the increase is within experimental error of values measured at the lowest P. Moreover, this trend does not hold true at the highest P for the NO/Ar system. This type of behavior, the lack of parameter dependence for ΘR, has been observed previously in our laboratories for SiH in SiH4 and Si2/H6 systems and indicates that ΘR of the plasma species (in this case NO) is thermalized.49,50 IV. Discussion The data presented here are primarily focused on the gasphase composition of NO-based plasmas. Initially, we sought to determine if NO destruction could be controlled or enhanced using plasmas. In NO and NO/Ar plasmas, we found that [NO] generally decreased as a function of P, as expected from increased dissociation when more energy is available to the system. With addition of water and methane vapor, the absolute amount of NO decreased significantly, but the relative intensity of NO increased with P. This may be because the addition of these gases promotes formation of NO through recombination reactions. For the N2/O2 systems, we found that at higher P, the formation of NO was promoted, suggesting that at high P, both N2 and O2 are more readily dissociated and that formation of NO occurs at a faster rate than recombination back to N2 and O2. This also explains the observed decrease in [N2] and the concomitant increase in [O] in these plasma systems. These additives made the systems more complex, resulting in an increase in [NO] in the N2/O2 gas mixtures, suggesting additional NO formation reactions51 compete with NO destruction processes when H2O or CH4 are present in the mixture. The introduction of water vapor or methane to these plasmas introduces complexity and a wider range of plasma species that can significantly affect [NO] through gas-phase reactions. For example, in plasmas with H2O(g) added, we see OES signals arising from NO, N2, OH, HR, and O. Although N and O can recombine to form NO, process 4, the rate constant for this reaction (k ) 1.45 × 1011 cm3 mol-1 s-1) is significantly lower than that for process 5, forming NO and H (k ) 3.05 × 1013 cm3 mol-1 s-1).27
N(g) + O(g) f NO(g) N(g) + OH(g) f NO(g) + H(g)
(4) (5)
Although it might be argued that process 4 is unlikely to happen at the relatively low pressures used in our plasmas, this process cannot be completely discounted in accounting for formation of NO.27 Additional reactions involving nitrogen atoms have also been proposed as important to either the destruction of NO, process 6, or the formation of NO under conditions where there is an excess of oxygen species (e.g., the N2/O2 systems or those with H2O), processes 7 and 8.20
N(g) + NO(g) f N2(g) + O(g)
(6)
N(g) + O2(g) f NO(g) + O(g)
(7)
O(g) + N2(g) f NO(g) + N(g)
(8)
Reactions 6 and 7 have room temperature rate constants of 6.8 × 10-12 and 6.0 × 10-11 cm3/s, respectively,52 and that for
Gas-Phase Chemistry in Inductively Coupled Plasmas
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1731
TABLE 1: NO Rotational Temperatures (K) for Different Mixturesa P (W) 25 50 75 100 125 150 200 240
NO (50) 337(29) 330(26) 317(13)
NO (100) 356(13) 343(12) 366(15) 366(20) 368(34) 366(30)
NO/Ar (12/88)
NO/Ar (50/50)
NO/Ar (80/20)
318(10) 313(9) 321(32) 313(13) 339(25) 348(13) 348(13) 314(16)
322(13)
320(10) 329(25) 325(23) 323(8) 330(28) 322(18)
325(15) 320(18) 332(10)
NO/H2O (60/40)
NO/CH4 (30/20)
327(24)
318(20)
332(10)
332(10)
328(20) 317(8)
328(10)
N2/O2/NO (70/20/10)
N2/O2 (90/10)
350(37) 328(24) 312(10) 328(20) 317(15) 323(28)
331(36) 323(40) 322(15) 325(18) 320(13) 320(12)
a
Values shown in parentheses under the gas mixture column labels indicate the partial pressure of each gas. The ΘR values given are the mean of several values; error limits are shown in parentheses and are the standard deviation of the mean.
reaction 8 is 7 × 10-12 cm3/s.53 Reaction 7 has been suggested as the primary route for relaxation of vibrationally excited N2 molecules in N2/O2 plasmas. Given the relatively high vibrational temperatures measured for N2 in our system, this reaction is likely to be key to the production of NO, especially in the N2/O2 systems. Interestingly, the data in Figures 5 and 6 suggest addition of water promotes formation of NO at high P, perhaps via reaction 5 or via the loss of singlet oxygen atoms via reaction with H2O to form OH radicals.26 This is consistent with the results of Fresnet et al. who found that when H2O was added to their phototriggered N2/NO discharge the NO removal efficiency decreased.54 Similar results were obtained when a hydrocarbon, C2H2, was added to their system.25 This decrease in NO breakdown was attributed to the reaction of singlet metastable states of N2, designated collectively as N2(a′). Without H2O (or a hydrocarbon) in the system, NO can be effectively removed via reactions with N2(a′), reaction 9.
NO + N2(a′) f N + O + N2
(9)
N2(a′) + H2O(C2H2) f products
(10)
The decrease in NO removal when H2O (C2H2) is added to the system occurs because reaction 10 removes N2(a′) from the system, thereby decreasing the effect of reaction 9. The consequence of the loss of singlet metastable states in the plasma is an overall decrease in the frequency of reaction 9. Our results suggest a similar effect may be occurring in our plasmas, especially at high P, conditions that should lead to greater concentration of N2(a′). Similar trends were observed with methane addition, although the presence of carbon-containing species (CO and CH) did not appreciably affect removal of NO. Interestingly, our results contrast somewhat with the results of Hueso et al. who used similar gas mixtures but a microwave plasma rather than an rf plasma.20,21 Specifically, addition of methane to a microwave plasma containing N2 and NO resulted in significant concentrations of CN*, C2*, and NH*. We see no clear evidence for formation of any of these species in our OES spectra, which is likely a result of the very different excitation mechanisms at work in the two different plasma systems. One possible explanation for the increase in [NO] in the N2/O2/CH4 system at higher P is that OH is formed under these conditions. Note that at P > 100 W, the relative intensity of OH increases, Figures 6 and 7. Thus, reaction 5 may also be affecting the density of NO in systems with a hydrocarbon present. Similar to the H2O(g)-containing plasma, however, the decrease in NO destruction may also result from a decrease in the N2(a′) density in the plasma. When Au-coated substrates were placed in the reactor, the relative NO intensity was slightly lower for most gas mixtures. Moreover, the NO density with Au-coated wafers was lower
than with uncoated Si wafers or with Pt foil in the reactor. It is notable that the efficacy of the Au substrates decreases with aging of the material. This could be a result of the buildup of adventitious carbon on the surface between uses. If true and not preventable in actual catalytic converters, gold would not be a viable or useful catalyst as it would be too expensive to replace often. Although the type of substrate present did affect the removal of NO from these systems, the more significant effect came from increased rf power. This is most likely because higher P effectively dissociates molecules in the system, and, as noted above, even though recombination reactions can serve to re-form NO, dissociation can lead to formation of alternate gas-phase species. Clearly, control of the process gases and their relative concentrations, along with the overall power applied to the system and the type of catalyst used, is required to achieve viable plasma processes for removal of NO. Given that these plasma systems are very complex and the individual chemical reactions at work are not fully understood, one possible approach to fully understand the plasma chemistry would be a computational model of the system. This necessarily would require knowledge of all of the species present in the plasma as well as their internal and kinetic temperatures. Thus, it is useful to discuss the ramifications of our ΘR(NO) and ΘV(NO) data on formation mechanisms in the plasma. Notably, there was virtually no dependence of ΘR on the parameters of reactor pressure, P, or gas mixture. This lack of dependence on P indicates that increasing the rf power does not lead to rotational heating of the NO radicals. When combined with the observation that NO intensity changes as a function of P in plasmas with different feed gas compositions, this is even more striking. Specifically, in the 100% NO plasmas as well as the NO/Ar systems, the OES and LIF intensities of NO* and NO decrease significantly with P. This indicates that at higher P, a portion of the energy dissipated in the plasma likely results in dissociation of NO. Note that formation of NO* in these plasmas likely occurs via direct electron impact of the NO being fed into the system and/or via a collision with an excited Ar atom (in the NO/Ar systems). In contrast, an increase in [NO] with applied rf power is observed in the systems where either water vapor or methane was added, suggesting that formation of NO is occurring via recombination reactions such as processes 4 and 5. In the N2/O2 system, NO can only be formed via a bimolecular collision, either between N2 and O2 or between the atomic species, processes 11 and 4, respectively.
N2(g) + O2(g) f 2NO(g)
(11)
Because the creation mechanisms in the various plasma systems studied here are different, it would be reasonable to expect that the internal temperatures of molecules formed in these processes
1732
J. Phys. Chem. A, Vol. 114, No. 4, 2010
would be different. As noted above, we have previously reported a similar lack of dependence on gas composition and P with SiH radicals formed in SiH4 and Si2H6 plasmas.49,50 With SiH, available data indicated the rotational-rotational and rotationaltranslational energy transfer reactions were near unity. Thus, Kessels et al. concluded that ΘR(SiH) was thermalized to the average gas temperature before the SiH was extracted from the plasma into the molecular beam. Similarly, we believe that ΘR(NO) is representative of the gas temperature in our plasmas, ∼330 K, slightly above room temperature. Energy partitioning into the vibrational modes of NO can lend further insight into the overall character of the molecule and the plasma system. In the calculations performed here, ΘV was determined to lie in the range from 1400 to 1900 K for all systems studied. This contrasts sharply with the measured ΘR(NO) values, which are considerably lower. Because of the way in which the spectra are simulated, which necessarily excludes overlap from other transitions, this is likely the lower limit range of ΘV. Indeed, the expected minimum value of ΘV in a nonthermal NO plasma is roughly 2300 K.55 A value less than this “minimum” indicates that vibrational-translational energy transfer occurs more rapidly than the chemical reaction forming NO.55 This aligns well with previous results of Saupe et al., who estimated ΘV ≈ 2500 K from a Treanor distribution for NO in optical pumping experiments,56 and the harmonic oscillator-rigid rotor approximation for ΘV in an ideal gas of NO is >2700 K.57 Although our measured values are significantly lower than these estimated values, the trends seen as a function of P and upon addition of substrates are likely accurate and indicate that a surface undergoing plasma modification does not influence the internal energy of vibration in the NO molecule. In contrast, values of ΘV for N2 were also calculated and found to be a factor of 5 lower. This may indicate that any vibrationally hot N2 that originally existed in the plasma was quickly consumed in the production of NO (e.g., via reaction 11). As a final note, ΘR(NO) is lower than that measured for any other plasma species we have studied thus far, including OH in 100% H2O plasmas,58 NH in NH3 plasmas,59,60 and SiF in SiF4 plasmas.61 In many of these systems, we also measured the translational (or kinetic) temperature of the molecules, ΘT, using LIF. Similarities and differences between ΘR and ΘT for a specific plasma species have led to the hypothesis that differences in internal and translational temperatures may be strongly linked to the mechanism for formation of individual plasma species as well as to the mechanism for energy dissipation in plasmas. For NO, the overall lack of parametric dependence of ΘR over a wide range of conditions suggests complete thermalization occurs, regardless of the formation mechanism. Not having data for ΘT(NO) in these systems, however, prevents us from drawing firmer conclusions regarding either energy dissipation or formation mechanisms. Moreover, having ΘR data for only a single species in some of the more complex plasma systems explored here (e.g., N2/O2/H2O(CH4)/Ar plasmas) also limits our ability to discuss these plasma mechanisms at any deeper level. This is clearly an area for future studies exploring ΘR and ΘT for multiple species in NO-based plasma systems. V. Summary The gas-phase densities of both ground-state and excited NO have been measured as a function of the applied rf power in a variety of NO- and N2/O2-based plasmas. In general, the overall density of NO decreases as a function of P, but the addition of gases such as H2O vapor and CH4, as well as the presence of
Morgan et al. Au-coated substrates, affects the P-dependent behavior of NO in these systems. Rotational temperatures for NO indicate no dependence on P, feed gas composition, or overall system pressure, suggesting the internal energy of NO has been thermalized and represents the overall gas temperature in these systems. Vibrational temperatures for NO measured from OES spectra are considerably hotter, but still lower than predicted, suggesting vibrational-translational energy transfer occurs more rapidly than NO formation reactions in the mixed gas plasma systems studied. Acknowledgment. This work was supported by the ACSPRF (#44510-ACS). We thank Dr. Patrick McCurdy for assistance with the acquisition of SEM and XPS data. References and Notes (1) EPA. How nitrogen oxides affect the way we live and breathe. EPA-456/F-98-005, 1998. (2) Shelef, M.; McCabe, R. W. Catal. Today 2000, 62, 35. (3) Di Monte, R.; Fornasiero, P.; Kaspar, J.; Rumori, P.; Gubitosa, G.; Graziani, M. Appl. Catal., B: EnViron. 2000, 24, 157. (4) Maunula, T.; Ahola, J.; Salmi, T.; Haario, H.; Harkonen, M.; Luoma, M.; Pohjola, V. Appl. Catal., B: EnViron. 1997, 12, 287. (5) Mok, Y. S.; Ravi, V.; Kang, H. C.; Rajanlkanth, B. S. IEEE Trans. Plasma Sci. 2003, 31, 157. (6) Orlando, T. M.; Alexandrov, A.; Lebsack, A.; Herring, J.; Hoard, J. W. Catal. Today 2004, 89, 151. (7) Wojciechowska, M.; Lomnicki, S. Clean Technol. EnViron. 1999, 1, 237. (8) Yeom, Y. H.; Henao, J.; Li, M. J.; Sachtler, W. M. H.; Weitz, E. J. Catal. 2005, 231, 181. (9) Yoon, S.; Panov, A. G.; Tonkyn, R. G.; Ebeling, A. C.; Barlow, S. E.; Balmer, M. L. Catal. Today 2002, 72, 251. (10) Bond, G. C. Platinum Met. ReV. 2007, 51, 63. (11) Ding, X.; Li, J.; Hou, J. G.; Zhu, Q. J. Chem. Phys. 2004, 121, 2558. (12) Mallick, K.; Witcomb, M. J.; Scurrell, M. S. Appl. Catal., A: Gen. 2004, 259, 163. (13) Corti, C. W.; Holliday, R. J.; Thompson, D. T. Top. Catal. 2007, 44, 331. (14) Penetrante, B. M.; Bardsley, J. N.; Hsiao, M. C. Jpn. J. Appl. Phys. 1997, 36, 5007. (15) Rajanlkanth, B. S.; Srinivasan, A. D.; Ravi, V. IEEE Trans. Dielectr. Electr. Insul. 2005, 12, 72. (16) Sun, Q.; Zhu, A.; Yang, X.; Niu, J.; Xu, Y. Chem. Commun. 2003, 1418. (17) Tas, M. A.; van Hardeveld, R.; van Veldhuizen, E. M. Plasma Chem. Plasma Process. 1997, 17, 371. (18) Tsai, C.; Yang, H.; Jou, C. G.; Lee, H. M. J. Hazard. Mater. 2007, 143, 409. (19) Gorry, P. A.; Whitehead, J. C.; Wu, J. Plasma Process. Polym. 2007, 4, 556. (20) Hueso, J. L.; Gonzalez-Elipe, A. R.; Cotrino, J.; Caballero, A. J. Phys. Chem. A 2005, 109, 4930. (21) Hueso, J. L.; Gonzalez-Elipe, A. R.; Cotrino, J.; Caballero, A. J. Phys. Chem. A 2007, 111, 1057. (22) Mok, Y. S.; Koh, D. J.; Shin, D. N.; Kim, K. T. Fuel Process. Technol. 2004, 86, 303. (23) Tonkyn, R. G.; Barlow, S. E.; Hoard, J. W. Appl. Catal., B: EnViron. 2003, 40, 207. (24) Kwak, J. H.; Szanyi, J.; Peden, C. H. F. J. Catal. 2003, 220, 291. (25) Kovacs, T.; Turanyi, T.; Foglein, K.; Szepvolgyi, J. Plasma Chem. Plasma Process. 2006, 26, 293. (26) Orlandini, I.; Riedel, U. J. Phys. D: Appl. Phys. 2000, 33, 2467. (27) Sathiamoorthy, G.; Kalyana, S.; Finney, W. C.; Clark, R. J.; Locke, B. R. Ind. Eng. Chem. 1999, 38, 1844. (28) Orlandini, I.; Riedel, U. Combust. Theory Modell. 2001, 5, 447. (29) De Bleecker, K.; Herrebout, D.; Bogaerts, A.; Gijbels, R.; Descamps, P. J. Phys. D: Appl. Phys. 2003, 36, 1826. (30) Fresnet, F.; Baravian, G.; Magne, L.; Pasquiers, S.; Postel, C.; Puech, V.; Rousseau, A. Appl. Phys. Lett. 2000, 77, 4118. (31) Lee, H. M.; Chang, M. B.; Yang, S. C. J. EnViron. Eng. 2003, 129, 800. (32) Bogart, K. H. A.; Dalleska, N. F.; Bogart, G. R.; Fisher, E. R. J. Vac. Sci. Technol. A 1995, 13, 476. (33) Chatterjee, D.; Deutschmann, O.; Warnatz, J. Faraday Discuss. 2001, 119, 371. (34) Trevino, K. J.; Fisher, E. R. Plasma Process. Polym. 2009, 6, 180.
Gas-Phase Chemistry in Inductively Coupled Plasmas (35) Gottscho, R. A.; Donnelly, V. M. J. Appl. Phys. 1984, 56, 245. (36) Malyshev, M. V.; Donnelly, V. M. J. Appl. Phys. 2000, 88, 6207. (37) Boffard, J. B.; Chiaro, B.; Weber, T.; Lin, C. C. At. Data Nucl. Data 2007, 93, 831. (38) Schappe, R. S.; Edgell, R. J.; Urban, E. Phys. ReV. A 2002, 65, 042701. (39) Luque, J.; Crosley, D. R. LIFBASE: Database and spectral simulation (version 1.5); SRI International Report MP 99-009, 1999. (40) Western, C. M. PGOPHER, a Program for Simulating Rotational Structure; University of Bristol, http://pgopher.chm.bris.ac.uk. (41) Grill, A. Cold Plasma in Materials Fabrication: From Fundamentals to Applications; IEEE Press: New York, NY, 1994. (42) McCurdy, P. R.; Bogart, K. H. A.; Dalleska, N. F.; Fisher, E. R. ReV. Sci. Instrum. 1997, 68, 1684. (43) Stillahn, J. M.; Trevino, K. J.; Fisher, E. R. Ann. ReV. Anal. Chem. 2008, 1, 261. (44) Cremaschi, P.; Johnson, P. M.; Whitten, J. L. J. Chem. Phys. 1978, 69, 4341. (45) Zhou, J.; Fisher, E. R. J. Phys. Chem. B 2006, 110, 21911. (46) Liu, D.; Fisher, E. R. J. Vac. Sci. Technol., A 2007, 25, 1519. (47) Pastol, A.; Catherine, Y. J. Phys. D: Appl. Phys. 1990, 23, 799. (48) Howle, C. R.; Arrowsmith, A. N.; Chikan, V.; Leone, S. R. J. Phys. Chem. A 2007, 111, 6637. (49) Kessels, W. M. M.; McCurdy, P. R.; Williams, K. L.; Venturo, V. A.; Barker, G. R.; Fisher, E. R. J. Phys. Chem. B 2002, 106, 2680.
J. Phys. Chem. A, Vol. 114, No. 4, 2010 1733 (50) Zhou, J.; Zhang, J.; Fisher, E. R. J. Phys. Chem. A 2005, 109, 10521. (51) Pejakovic, D. A.; Marschall, J.; Duan, L.; Martin, M. P. J. Thermophys. Heat Transfer 2008, 22, 178. (52) Herron, J. T. J. Phys. Chem. Ref. Data 1999, 28, 1453. (53) Gordiets, B. F.; Ferreira, C. M.; Guerra, V. L.; Loureiro, J. M. A. H.; Nahorny, J.; Pagnon, D.; Touzeau, M.; Vialle, M. IEEE Trans. Plasma Sci. 1995, 23, 750. (54) Fresnet, F.; Baravian, G.; Magne, L.; Pasquiers, S.; Postel, C.; Puech, V.; Rousseau, A. Plasma Sources Sci. Technol. 2002, 11, 152. (55) Fridman, A. Plasma Chemistry; Cambridge University Press: New York, 2008. (56) Saupe, S.; Adamovich, I.; Grassi, M. J.; Rich, J. W. J. Chem. Phys. 1993, 174, 219. (57) McQuarrie, D. A. Statistical Mechanics; University Science Books: Sausalito, CA, 2000. (58) Bogart, K. H. A.; Cushing, J. P.; Fisher, E. R. J. Phys. Chem. B 1997, 101, 10016. (59) Butoi, C. I.; Steen, M. L.; Peers, J. R. D.; Fisher, E. R. J. Phys. Chem. B 2001, 105, 5957. (60) McCurdy, P. R.; Butoi, C. I.; Williams, K. L.; Fisher, E. R. J. Phys. Chem. B 1999, 103, 6919. (61) Zhang, J.; Williams, K. L.; Fisher, E. R. J. Phys. Chem. A 2003, 107, 593.
JP908684C