- Email: [email protected]
Sensor and method for determining the ammonia loading of a zeolite SCR catalyst
Available online at www.sciencedirect.com
Sensors and Actuators B 130 (2008) 425–429
Sensor and method for determining the ammonia loading of a zeolite SCR catalyst D.J. Kubinski ∗ , J.H. Visser Ford Research and Advanced Engineering, MD 3083, P.O. Box 2053, Dearborn, MI 48121, USA Available online 14 September 2007
Abstract A sensor and a method for determining the NH3 loading of a base-metal zeolite catalyst are described. This catalyst material is used to reduce NOx emissions in diesel exhaust via reaction with injected NH3 , a process known as selective catalytic reduction (SCR). Some of the injected NH3 is stored on the catalyst surface before it reduces NOx. We demonstrate that the catalyst material itself can be used as part of a sensor measuring the level of stored NH3 . The AC conductivity of a thick film of the catalyst material was measured at 4 Hz and was found to increase when NH3 was introduced into the gas phase, eventually saturating presumably as the material reached its maximum NH3 storage capacity. The conductivity was monitored as the temperature was raised sufficiently to desorb (or oxidize) most of the stored NH3 . The resulting change in the conductivity upon heating was used as a measure of the NH3 loading level on the zeolite catalyst material. We demonstrate this parameter is dependent on NH3 concentration, NH3 loading time and gas temperature. © 2007 Elsevier B.V. All rights reserved. Keywords: Ammonia sensing; Zeolite; SCR catalyst; Desorption
1. Introduction The reduction of nitrogen oxide (NOx) emissions from diesel vehicles is made difficult due to the elevated O2 levels present in diesel exhaust. However, the urea/SCR (SCR = selective catalytic reduction) exhaust after-treatment system has been shown to be effective in reducing the NOx emissions, with low impact on fuel economy [1]. In this system NH3 , formed from the injection of aqueous urea, reacts selectively with NOx on the SCR catalyst to form N2 and H2 O. The SCR catalyst can be a base-metal zeolite material typically operating most efficiently above 200 ◦ C, and has the ability to store some of the injected NH3 [2]. The NH3 storage capacity for common SCR catalyst materials, however, is reduced with elevated temperature [2,3]. For optimal performance, the levels of NH3 injected into the exhaust must match the NOx emissions over a specific time interval. Under-injection of NH3 leads to a lower NOx conversion rate and the inability to meet the regulated emission standards. Over-injection gives yields unreacted NH3 , which is an undesirable emission. A catalyst fully saturated with stored NH3 has the positive benefit of yielding a high NOx conversion
∗
Corresponding author. Tel.: +1 313 322 3810. E-mail address: [email protected] (D.J. Kubinski).
0925-4005/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2007.09.007
rate. However, the maximum amount NH3 stored on the SCR catalyst decreases with temperature. Thus, a catalyst saturated with NH3 would liberate some of it upon excursions to high-exhaust gas temperatures, resulting for that case unwanted exhaust emission. The goal of the SCR catalyst system is to minimize both NOx and NH3 emissions by the proper control of the aqueous urea injection. In principle this can be achieved via control of the stored NH3 levels along the axial length of the catalyst, with high-NH3 loading in the upstream portion of the catalyst to ensure high NOx conversion and lower NH3 loading in the downstream portion to minimize NH3 emissions. One method to maintain an optimal NH3 -loading profile along the SCR catalyst is based on accurate models of parameters such as the NOx levels emitted by the engine, quantity of urea injection, temperature, gas flow, NOx–NH3 reaction rates and NH3 storage. This approach can be improved using NOx and/or NH3 sensors located downstream of the catalyst, providing feedback to help ensure that the slip of these gases are minimized. We demonstrate here a further possible improvement: By locally heating and desorbing the NH3 stored on a portion(s) of the SCR catalyst while simultaneously monitoring its ionic conductivity, it is possible to obtain an estimate of the quantity of stored-NH3 in that region. We demonstrate a sensor concept based on this temperature-programmed-desorption related technique.
426
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
Moos et al. reported that zeolites such as H-ZSM5 to be a surface proton conductor whose ionic conductivity is enhanced in the presence of NH3 , enabling them to be used as the basis of a selective NH3 sensor [4]. The fundamental sensing principle is that the mobility of H+ along the surface is enhanced by their reaction with NH3 to form NH4 + ions, which are bound less tightly and have lower activation energy for hopping along the zeolite surface [5]. In a similar fashion, we show here that the base-metal zeolite material used in the SCR catalyst can also be used to detect NH3 . We suggest that the amount of NH3 stored on the catalyst material can be estimated by monitoring its change in conductivity upon heating to temperatures sufficient to desorb or oxidize most of the stored NH3 . We demonstrate that by sacrificing continuous measurement of the NH3 -storage levels and measuring instead that thermally desorbed upon rapid heating, we obtain greater signal. Occasional measurements of the NH3 storage levels are expected to be sufficient to update the previously described urea-injection control models.
NH3 , we define two operational modes: the NH3 loading mode and the measurement mode. Prior to the start of the NH3 loading mode, the sensor was free of stored NH3 . During the loading mode the temperature was held constant while a fixed concentration of gas-phase NH3 was exposed to the sensor, some of which was stored on the catalyst material. Constant temperature during loading mode, TL , was achieved by placing the sensor in quartz tube that was inserted into a tube furnace. In our experiments power was not applied to the sensor’s substrate heater during this mode. However, during the measurement mode a 9 V signal was applied to the substrate heater causing the temperature to rise sufficiently above TL to desorb or oxidize most of the NH3 that was stored during the loading mode. The amount of NH3 stored during the loading mode was determined from the change in the catalyst conductivity as the temperature was increased. We monitored this change in conductivity upon heating, testing for the influences of NH3 loading time, NH3 loading temperature, and NH3 concentration. 3. Results
2. Experimental The basic sensor configuration is shown in Fig. 1. The main component is a base-metal zeolite catalyst material (proprietary composition) that was deposited as a thick film on an alumina substrate. Contained within the alumina substrate is an embedded heater, enabling the temperature of the catalyst film to be raised to 600 ◦ C. The heated area is ∼5 mm by ∼5 mm. The catalyst film was also laid over a pair of metal electrodes, which were used to monitor its conductivity. The electrodes are configured as two parallel strips with ∼0.5 mm separation and which covered the entire heated region. The zeolite film was made by spreading a slurry containing a powder of the base-metal zeolite catalyst material mixed with a colloidal alumina solution (∼50 nm alumina particle size). The slurry was dried to a thick film using heated air. The alumina was added as a binder phase and was ∼20% of the final film by weight. Exact film thickness was not determined. In our tests, the conductivity was measured by monitoring the current induced upon applying a 5 VP–P sinusoidal signal at 4 Hz. The root mean square (rms) magnitude of the resulting current was measured using a Stanford Research Systems Model SR850 DSP lock-in amplifier. All data were taken in background gases of 5% O2 and 1% water, with the balance N2 , using a laboratory gas flow bench. Since we are describing monitoring the conductivity changes in the thick-film zeolite upon thermally desorbing the stored
Fig. 1. Cross-section of the SCR catalyst conductivity sensor.
Fig. 2 shows the rms current measured across the thick-film sensor as a function of time for both the NH3 loading mode and the subsequent measurement mode. For both of these time intervals the film was exposed to 400 ppm NH3 , which was introduced at t = 0. Prior to this time the sensor was free of stored NH3 . The first 40 min of Fig. 2 describe the loading mode, during which the temperature was held at ∼267 ◦ C. Note that during this time period the conductivity of the zeolite catalyst film slowly increased, approaching an asymptotic value after ∼30 min. The measurement mode was initiated at t = 40 min for the data shown in this figure. At that time 9 V was applied to the substrate heater. It is presumed that the rapid rise and subsequent decrease in the measured current during that time interval was a result of the rapid increase in the sensor temperature to values sufficient to
Fig. 2. The AC current as a function of time for the thick-film zeolite sensor. The composition of the background gas was 5% O2 and ∼1% water in N2 . Prior to t = 0, the sensor was free of stored NH3 . At t = 0, 400 ppm NH3 was introduced and kept on during the test. The sensor temperature was ∼267 ◦ C for the first 40 min. This time interval is labeled NH3 loading mode. After t = 40 min, the temperature of the sensor was raised sufficiently to desorb/oxidize most of the stored NH3 . This time interval is labeled measurement mode.
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
427
Fig. 3. The heating profile during the measurement mode used to remove most of the NH3 stored on the catalyst sensor material. The temperature at t = 0 was ≈267 ◦ C. The broken line gives the corresponding change in sensor current. These are the same data shown near t = 40 min in Fig. 2.
desorb most of the NH3 stored in the prior 40 min. The sensor’s current spike and corresponding temperature increase during the measurement mode are shown with greater clarity in Fig. 3. In this figure, t = 0 defines the time at which 9 V was applied to the substrate heater (occurring at t = 40 min in Fig. 2). The solid line shows that the temperature was increased to ∼520 ◦ C in 60 s, albeit in a non-linear fashion with the rate of temperature rise decreasing with time. The corresponding conductivity of the zeolite film was found to initially increase with raising temperature, reaching a maximum value at t ≈ 11 s that was almost five times the value measured prior to t = 0 (at the end of the loading mode). The corresponding temperature at the peak conductivity was ∼420 ◦ C. Further increase in the temperature after t ∼ 11 s resulted in a decrease in conductivity, presumably as most of the stored NH3 was either desorbed or oxidized. The influence of the duration of the loading mode on the measurement-mode response is demonstrated in Fig. 4. Shown is the rms sensor current versus measurement-mode time for cases of increasing loading-mode time intervals. For each of these cases the sensor was exposed to 500 ppm NH3 at ≈267 ◦ C during the loading mode, which ranged in time lengths from 0 to 40 min. The corresponding heating profile during the measurement mode was same for all the cases shown in Fig. 4, and was the same profile as shown in Fig. 3. It is noted in Fig. 4 that the peak height in current upon heating increased for longer NH3 loading times, reaching a maximum which did not change after 20 min of NH3 loading. Expect for the case of no NH3 loading, the curves converged to the same low value after 60 s of heating, a time which presumably corresponds to a temperature sufficient to desorb or oxidize most of the stored NH3 . It may be possible that a small amount of NH3 remained after 60 s, when the substrate temperature was ∼520 ◦ C. This may explain the slightly lower conductivity at that time for the case of no NH3 loading. It is our hypothesis that the area under each of the I(t ) curves of Fig. 4 can be used to approximate the level of NH3 storage. Each of the curves shown in Fig. 4 is integrated over the first 60 s of heating to give a plot of average current during the mea-
Fig. 4. Sensor current vs. measurement mode heating time. Shown are data for loading-mode time intervals ranging from 0 to 40 min. For all cases the substrate temperature was ≈267 ◦ C and NH3 level 500 ppm during the loading mode. The temperature was increased at t = 0 with the same profile as plotted in Fig. 3. The data were taken in 5% O2 and 1% water.
surement mode, IAverage , as a function of NH3 loading time. This plot is shown in Fig. 5. Also plotted in Fig. 5 are data for other loading-mode NH3 concentrations, which ranged from 100 to 500 ppm. The curves for both 400 ppm NH3 and 500 ppm NH3 loading mode concentrations reach asymptotic values for loading times greater than 30 min, presumably due to the saturation of NH3 stored on the catalyst material. Note that the asymptotic
Fig. 5. Average current during the first 60 s of the measurement mode as a function of the NH3 loading-mode time. Shown are data for NH3 loading concentrations of 100, 200, 400 and 500 ppm. For each case the loading-mode temperature was ≈267 ◦ C and the temperature was increased during the measurement mode using the same heating profile as shown in Fig. 3. The data were taken in 5% O2 and 1% water.
428
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
4. Discussion
Fig. 6. Average current increase during the measurement mode due to stored NH3 as a function of the NH3 loading time. Shown are data for NH3 loading temperatures ranging from ≈213 to ≈359 ◦ C. For each case, the NH3 concentration during loading time was 500 ppm. The data were taken in 5% O2 and 1% water.
values for IAverage are observed to increase with NH3 level. This trend is similar to that expected for the NH3 loading on the catalyst, with the saturation loading dependent on NH3 concentration [6]. The Langmuir adsorption isotherm predicts the saturation NH3 loading to scale as Keq CNH3 /(1 + Keq CNH3 ), where CNH3 is the gas-phase NH3 concentration and Keq is dependent on temperature. We examined the influence of NH3 loading-mode temperature, TL , on IAverage . We define the parameter IAverage (TL ) ≡ IAverage (TL ) − IAverage, no NH3 loading (TL ), which is the increase in IAverage (TL ) due to NH3 stored at loading temperature TL . Fig. 6 shows the trend of lower IAverage (TL ) for higher loading temperatures over the range TL ≈ 213 ◦ C to TL ≈ 359 ◦ C. For these data the NH3 concentration during the loading phase was 500 ppm. The results are consistent with studies that show the saturation NH3 loading on SCR catalysts to be dependent both on the NH3 level and temperature [2,3,6]. Not plotted were data taken at ≈174 ◦ C, which showed lower values of IAverage than those measured at 200 ◦ C. The cause for this requires further investigation. For clarity we plotted IAverage (TL ) in Fig. 6 instead of IAverage (TL ), so that all curves start at the origin on the lower left. Note that IAverage, no NH3 loading (TL ≈ 267 ◦ C) is the y-axis intercept in Fig. 5, which was found to be dependent on loading-mode temperature, increasing from 2.7 × 10−9 to 1.1 × 10−8 A over the loading-mode temperature range plotted. This dependence was presumably a consequence of two factors: the conductivity of the sensor in the absence of NH3 increasing with temperature and the average temperature during the measurement mode increasing with TL (For all cases 9 V was applied to the substrate heater to raise the temperature above TL . The temperatures obtained after 60 s of substrate heating varied by more than 100 ◦ C for the cases of TL ≈ 213 ◦ C and TL ≈ 359 ◦ C.).
We speculate that the integrated values of the curves shown in Fig. 4 give an estimate of the stored NH3 . The almost fivefold increase in maximum conductivity upon heating as shown in this figure demonstrates a signal advantage offered by this technique. Regarding the shape of these curves, the exact cause of the initial increase in conductivity upon heating is not fully understood. One contribution may be the temperature dependence of the NH3 sensing response itself. Another may be the nature of the storage sites on the zeolite. If we assume the enhanced conductivity due to NH3 is caused by an increase in mobility of the surface H+ ions as suggested in ref. [5], then moving stored NH3 from other sites to the H+ sites upon heating would be expected to increase the conductivity. This, however, requires further study. The decrease in conductivity upon heating past the temperature corresponding to the peak conductivity (t > ∼15 s in Fig. 4) is presumably a consequence of purging most of the stored NH3 by either desorption, oxidation or both. It is reasonable to question the ability of such a sensor and technique to accurately measure the stored NH3 levels, given that the conductivity change of zeolite-based sensors due to gasphase NH3 is known to be temperature dependent [5]. The curves shown in Figs. 5 and 6 might be made more accurate if a weighted integration of the curves in Fig. 4 was done such that for each time interval (and corresponding temperature) the temperature dependence of the NH3 response was factored out. In this way the integration of a temperature compensated conductivity change due to gas desorption could more accurately reflect the amount of stored NH3 . This correction will be applied in our future studies as detailed measurements of the temperature dependence of the NH3 sensitivity were not done for the base-metal zeolite catalyst material used in this study. We recognize the thick-film zeolite sensor used in this study was much too dense for practical purposes, resulting in a long time required to saturate the response. For example, Fig. 3 shows this took at least 20 min for 500 ppm NH3 loading. We speculate the slow diffusion was a consequence of making the zeolite film with too high an alumina binder fraction. We speculate the alumina grains, being much smaller than the zeolite grains, served to block the space between them resulting in longer paths for the NH3 to diffuse. Additional experiments done on films with only 10% alumina by weight show the loading time required to saturate the response to be decreased by almost a factor of 10. These results will be the focus of a future report. Ideally we desire to make a sensor with same diffusion-time dependency for the NH3 storage as that for the material coated onto the actual SCR catalyst. Future studies should also investigate the influence of water vapor and NH3 selectivity. Water concentration may be an important cross-sensitivity parameter as it has been shown to affect the response of a zeolite-based NH3 sensor [4]. For the data presented here, water was fixed at 1%. We did not test the selectivity of the sensor’s response to NH3 . However, our prior studies on another zeolite material, HZSM-5, demonstrated a selective NH3 response compared to that for CO, HC’s and NOx.
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
This concurred with the findings previously reported for that material [4]. We expect the zeolite material used here to behave similarly, but this like the water sensitivity should be investigated in future studies. Future studies will also focus on testing our hypothesis that the average current as depicted in Fig. 5 scales with the NH3 loading on the sensor material. This can be tested comparing IAverage measurements with NH3 adsorption measurements obtained via thermogravimetric analysis (TGA). Additionally, we want to obtain more quantitative IAverage versus loadingtemperature measurements. We recognize that the heating method used in this study requires improvement. For example, the results of Fig. 6 could be easier to interpret quantitatively if a constant temperature ramp was used (degrees/second), with the same final temperature for each case. In summary, we have demonstrated that an ammonia storage sensor for a zeolite-based SCR catalyst is possible using the SCR catalyst material itself as the sensing element. This was done by measuring its change in conductivity upon heating, in an attempt to purge it of stored NH3. References [1] C. Lambert, R. Hammerle, R. McGill, M. Khair, C. Sharp, Technical advantages of urea SCR for light-duty and heavy-duty diesel vehicle applications, SAE Technical Paper Series: Technical paper 2004-01-1292, Society of Automotive Engineers: Warrendale, PA, USA, 2004.
429
[2] P. Tennison, C. Lambert, M. Levin, NOx control development with urea SCR on a diesel passenger car, SAE Technical Paper Series: Technical paper 2004-01-1291, Society of Automotive Engineers: Warrendale, PA, USA, 2004. [3] L. Lietti, I. Nova, E. Tronconi, P. Forzatti, Transient kinetic study of the SCR-DeNOx reaction, Catal. Today 45 (1998) 85–92. [4] R. Moos, R. M¨uller, C. Plog, A. Knezevic, H. Leye, E. Irion, T. Braun, K. Marquardt, K. Binder, Selective ammonia exhaust gas sensor for automotive applications, Sens. Actuators B Chem. 83 (2002) 181–189. [5] U. Simon, U. Flesch, W. Maunz, R. M¨uller, C. Plog, The effect of NH3 on the ionic conductivity of dehydrated zeolites N beta and H beta, Micropor. Mesopor. Mater. 21 (1998) 111–116. [6] I. Nova, L. Lietti, E. Tronconi, P. Forzatti, Transient response method applied to the kinetic analysis of the DeNOx-SCR reaction, Chem. Eng. Sci. 56 (2001) 1229–1237.
Biographies D.J. Kubinski received his PhD in physics from Wayne State University in 1995. He has been with the Ford Research Staff since 1986. His present interests are in sensing technologies for automotive exhaust. J. H. Visser received MS and PhD degrees in electrical engineering from Delft University of Technology, Delft, The Netherlands, in 1985 and 1989, respectively. Since 1990, he has joined the Ford Research Staff where he has been working on sensors and their applications for automotive systems. His research interests include exhaust gas sensors (air-to-fuel ratio, NOx , NH3 , HC and CO), gasoline-vapor sensors, and sensors for measuring automotive liquids. He is currently a technical leader at Ford Research and Advanced Engineering and is the project leader of the sensors group in the Chemical Engineering Department.
Sensors and Actuators B 130 (2008) 425–429
Sensor and method for determining the ammonia loading of a zeolite SCR catalyst D.J. Kubinski ∗ , J.H. Visser Ford Research and Advanced Engineering, MD 3083, P.O. Box 2053, Dearborn, MI 48121, USA Available online 14 September 2007
Abstract A sensor and a method for determining the NH3 loading of a base-metal zeolite catalyst are described. This catalyst material is used to reduce NOx emissions in diesel exhaust via reaction with injected NH3 , a process known as selective catalytic reduction (SCR). Some of the injected NH3 is stored on the catalyst surface before it reduces NOx. We demonstrate that the catalyst material itself can be used as part of a sensor measuring the level of stored NH3 . The AC conductivity of a thick film of the catalyst material was measured at 4 Hz and was found to increase when NH3 was introduced into the gas phase, eventually saturating presumably as the material reached its maximum NH3 storage capacity. The conductivity was monitored as the temperature was raised sufficiently to desorb (or oxidize) most of the stored NH3 . The resulting change in the conductivity upon heating was used as a measure of the NH3 loading level on the zeolite catalyst material. We demonstrate this parameter is dependent on NH3 concentration, NH3 loading time and gas temperature. © 2007 Elsevier B.V. All rights reserved. Keywords: Ammonia sensing; Zeolite; SCR catalyst; Desorption
1. Introduction The reduction of nitrogen oxide (NOx) emissions from diesel vehicles is made difficult due to the elevated O2 levels present in diesel exhaust. However, the urea/SCR (SCR = selective catalytic reduction) exhaust after-treatment system has been shown to be effective in reducing the NOx emissions, with low impact on fuel economy [1]. In this system NH3 , formed from the injection of aqueous urea, reacts selectively with NOx on the SCR catalyst to form N2 and H2 O. The SCR catalyst can be a base-metal zeolite material typically operating most efficiently above 200 ◦ C, and has the ability to store some of the injected NH3 [2]. The NH3 storage capacity for common SCR catalyst materials, however, is reduced with elevated temperature [2,3]. For optimal performance, the levels of NH3 injected into the exhaust must match the NOx emissions over a specific time interval. Under-injection of NH3 leads to a lower NOx conversion rate and the inability to meet the regulated emission standards. Over-injection gives yields unreacted NH3 , which is an undesirable emission. A catalyst fully saturated with stored NH3 has the positive benefit of yielding a high NOx conversion
∗
Corresponding author. Tel.: +1 313 322 3810. E-mail address: [email protected] (D.J. Kubinski).
0925-4005/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2007.09.007
rate. However, the maximum amount NH3 stored on the SCR catalyst decreases with temperature. Thus, a catalyst saturated with NH3 would liberate some of it upon excursions to high-exhaust gas temperatures, resulting for that case unwanted exhaust emission. The goal of the SCR catalyst system is to minimize both NOx and NH3 emissions by the proper control of the aqueous urea injection. In principle this can be achieved via control of the stored NH3 levels along the axial length of the catalyst, with high-NH3 loading in the upstream portion of the catalyst to ensure high NOx conversion and lower NH3 loading in the downstream portion to minimize NH3 emissions. One method to maintain an optimal NH3 -loading profile along the SCR catalyst is based on accurate models of parameters such as the NOx levels emitted by the engine, quantity of urea injection, temperature, gas flow, NOx–NH3 reaction rates and NH3 storage. This approach can be improved using NOx and/or NH3 sensors located downstream of the catalyst, providing feedback to help ensure that the slip of these gases are minimized. We demonstrate here a further possible improvement: By locally heating and desorbing the NH3 stored on a portion(s) of the SCR catalyst while simultaneously monitoring its ionic conductivity, it is possible to obtain an estimate of the quantity of stored-NH3 in that region. We demonstrate a sensor concept based on this temperature-programmed-desorption related technique.
426
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
Moos et al. reported that zeolites such as H-ZSM5 to be a surface proton conductor whose ionic conductivity is enhanced in the presence of NH3 , enabling them to be used as the basis of a selective NH3 sensor [4]. The fundamental sensing principle is that the mobility of H+ along the surface is enhanced by their reaction with NH3 to form NH4 + ions, which are bound less tightly and have lower activation energy for hopping along the zeolite surface [5]. In a similar fashion, we show here that the base-metal zeolite material used in the SCR catalyst can also be used to detect NH3 . We suggest that the amount of NH3 stored on the catalyst material can be estimated by monitoring its change in conductivity upon heating to temperatures sufficient to desorb or oxidize most of the stored NH3 . We demonstrate that by sacrificing continuous measurement of the NH3 -storage levels and measuring instead that thermally desorbed upon rapid heating, we obtain greater signal. Occasional measurements of the NH3 storage levels are expected to be sufficient to update the previously described urea-injection control models.
NH3 , we define two operational modes: the NH3 loading mode and the measurement mode. Prior to the start of the NH3 loading mode, the sensor was free of stored NH3 . During the loading mode the temperature was held constant while a fixed concentration of gas-phase NH3 was exposed to the sensor, some of which was stored on the catalyst material. Constant temperature during loading mode, TL , was achieved by placing the sensor in quartz tube that was inserted into a tube furnace. In our experiments power was not applied to the sensor’s substrate heater during this mode. However, during the measurement mode a 9 V signal was applied to the substrate heater causing the temperature to rise sufficiently above TL to desorb or oxidize most of the NH3 that was stored during the loading mode. The amount of NH3 stored during the loading mode was determined from the change in the catalyst conductivity as the temperature was increased. We monitored this change in conductivity upon heating, testing for the influences of NH3 loading time, NH3 loading temperature, and NH3 concentration. 3. Results
2. Experimental The basic sensor configuration is shown in Fig. 1. The main component is a base-metal zeolite catalyst material (proprietary composition) that was deposited as a thick film on an alumina substrate. Contained within the alumina substrate is an embedded heater, enabling the temperature of the catalyst film to be raised to 600 ◦ C. The heated area is ∼5 mm by ∼5 mm. The catalyst film was also laid over a pair of metal electrodes, which were used to monitor its conductivity. The electrodes are configured as two parallel strips with ∼0.5 mm separation and which covered the entire heated region. The zeolite film was made by spreading a slurry containing a powder of the base-metal zeolite catalyst material mixed with a colloidal alumina solution (∼50 nm alumina particle size). The slurry was dried to a thick film using heated air. The alumina was added as a binder phase and was ∼20% of the final film by weight. Exact film thickness was not determined. In our tests, the conductivity was measured by monitoring the current induced upon applying a 5 VP–P sinusoidal signal at 4 Hz. The root mean square (rms) magnitude of the resulting current was measured using a Stanford Research Systems Model SR850 DSP lock-in amplifier. All data were taken in background gases of 5% O2 and 1% water, with the balance N2 , using a laboratory gas flow bench. Since we are describing monitoring the conductivity changes in the thick-film zeolite upon thermally desorbing the stored
Fig. 1. Cross-section of the SCR catalyst conductivity sensor.
Fig. 2 shows the rms current measured across the thick-film sensor as a function of time for both the NH3 loading mode and the subsequent measurement mode. For both of these time intervals the film was exposed to 400 ppm NH3 , which was introduced at t = 0. Prior to this time the sensor was free of stored NH3 . The first 40 min of Fig. 2 describe the loading mode, during which the temperature was held at ∼267 ◦ C. Note that during this time period the conductivity of the zeolite catalyst film slowly increased, approaching an asymptotic value after ∼30 min. The measurement mode was initiated at t = 40 min for the data shown in this figure. At that time 9 V was applied to the substrate heater. It is presumed that the rapid rise and subsequent decrease in the measured current during that time interval was a result of the rapid increase in the sensor temperature to values sufficient to
Fig. 2. The AC current as a function of time for the thick-film zeolite sensor. The composition of the background gas was 5% O2 and ∼1% water in N2 . Prior to t = 0, the sensor was free of stored NH3 . At t = 0, 400 ppm NH3 was introduced and kept on during the test. The sensor temperature was ∼267 ◦ C for the first 40 min. This time interval is labeled NH3 loading mode. After t = 40 min, the temperature of the sensor was raised sufficiently to desorb/oxidize most of the stored NH3 . This time interval is labeled measurement mode.
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
427
Fig. 3. The heating profile during the measurement mode used to remove most of the NH3 stored on the catalyst sensor material. The temperature at t = 0 was ≈267 ◦ C. The broken line gives the corresponding change in sensor current. These are the same data shown near t = 40 min in Fig. 2.
desorb most of the NH3 stored in the prior 40 min. The sensor’s current spike and corresponding temperature increase during the measurement mode are shown with greater clarity in Fig. 3. In this figure, t = 0 defines the time at which 9 V was applied to the substrate heater (occurring at t = 40 min in Fig. 2). The solid line shows that the temperature was increased to ∼520 ◦ C in 60 s, albeit in a non-linear fashion with the rate of temperature rise decreasing with time. The corresponding conductivity of the zeolite film was found to initially increase with raising temperature, reaching a maximum value at t ≈ 11 s that was almost five times the value measured prior to t = 0 (at the end of the loading mode). The corresponding temperature at the peak conductivity was ∼420 ◦ C. Further increase in the temperature after t ∼ 11 s resulted in a decrease in conductivity, presumably as most of the stored NH3 was either desorbed or oxidized. The influence of the duration of the loading mode on the measurement-mode response is demonstrated in Fig. 4. Shown is the rms sensor current versus measurement-mode time for cases of increasing loading-mode time intervals. For each of these cases the sensor was exposed to 500 ppm NH3 at ≈267 ◦ C during the loading mode, which ranged in time lengths from 0 to 40 min. The corresponding heating profile during the measurement mode was same for all the cases shown in Fig. 4, and was the same profile as shown in Fig. 3. It is noted in Fig. 4 that the peak height in current upon heating increased for longer NH3 loading times, reaching a maximum which did not change after 20 min of NH3 loading. Expect for the case of no NH3 loading, the curves converged to the same low value after 60 s of heating, a time which presumably corresponds to a temperature sufficient to desorb or oxidize most of the stored NH3 . It may be possible that a small amount of NH3 remained after 60 s, when the substrate temperature was ∼520 ◦ C. This may explain the slightly lower conductivity at that time for the case of no NH3 loading. It is our hypothesis that the area under each of the I(t ) curves of Fig. 4 can be used to approximate the level of NH3 storage. Each of the curves shown in Fig. 4 is integrated over the first 60 s of heating to give a plot of average current during the mea-
Fig. 4. Sensor current vs. measurement mode heating time. Shown are data for loading-mode time intervals ranging from 0 to 40 min. For all cases the substrate temperature was ≈267 ◦ C and NH3 level 500 ppm during the loading mode. The temperature was increased at t = 0 with the same profile as plotted in Fig. 3. The data were taken in 5% O2 and 1% water.
surement mode, IAverage , as a function of NH3 loading time. This plot is shown in Fig. 5. Also plotted in Fig. 5 are data for other loading-mode NH3 concentrations, which ranged from 100 to 500 ppm. The curves for both 400 ppm NH3 and 500 ppm NH3 loading mode concentrations reach asymptotic values for loading times greater than 30 min, presumably due to the saturation of NH3 stored on the catalyst material. Note that the asymptotic
Fig. 5. Average current during the first 60 s of the measurement mode as a function of the NH3 loading-mode time. Shown are data for NH3 loading concentrations of 100, 200, 400 and 500 ppm. For each case the loading-mode temperature was ≈267 ◦ C and the temperature was increased during the measurement mode using the same heating profile as shown in Fig. 3. The data were taken in 5% O2 and 1% water.
428
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
4. Discussion
Fig. 6. Average current increase during the measurement mode due to stored NH3 as a function of the NH3 loading time. Shown are data for NH3 loading temperatures ranging from ≈213 to ≈359 ◦ C. For each case, the NH3 concentration during loading time was 500 ppm. The data were taken in 5% O2 and 1% water.
values for IAverage are observed to increase with NH3 level. This trend is similar to that expected for the NH3 loading on the catalyst, with the saturation loading dependent on NH3 concentration [6]. The Langmuir adsorption isotherm predicts the saturation NH3 loading to scale as Keq CNH3 /(1 + Keq CNH3 ), where CNH3 is the gas-phase NH3 concentration and Keq is dependent on temperature. We examined the influence of NH3 loading-mode temperature, TL , on IAverage . We define the parameter IAverage (TL ) ≡ IAverage (TL ) − IAverage, no NH3 loading (TL ), which is the increase in IAverage (TL ) due to NH3 stored at loading temperature TL . Fig. 6 shows the trend of lower IAverage (TL ) for higher loading temperatures over the range TL ≈ 213 ◦ C to TL ≈ 359 ◦ C. For these data the NH3 concentration during the loading phase was 500 ppm. The results are consistent with studies that show the saturation NH3 loading on SCR catalysts to be dependent both on the NH3 level and temperature [2,3,6]. Not plotted were data taken at ≈174 ◦ C, which showed lower values of IAverage than those measured at 200 ◦ C. The cause for this requires further investigation. For clarity we plotted IAverage (TL ) in Fig. 6 instead of IAverage (TL ), so that all curves start at the origin on the lower left. Note that IAverage, no NH3 loading (TL ≈ 267 ◦ C) is the y-axis intercept in Fig. 5, which was found to be dependent on loading-mode temperature, increasing from 2.7 × 10−9 to 1.1 × 10−8 A over the loading-mode temperature range plotted. This dependence was presumably a consequence of two factors: the conductivity of the sensor in the absence of NH3 increasing with temperature and the average temperature during the measurement mode increasing with TL (For all cases 9 V was applied to the substrate heater to raise the temperature above TL . The temperatures obtained after 60 s of substrate heating varied by more than 100 ◦ C for the cases of TL ≈ 213 ◦ C and TL ≈ 359 ◦ C.).
We speculate that the integrated values of the curves shown in Fig. 4 give an estimate of the stored NH3 . The almost fivefold increase in maximum conductivity upon heating as shown in this figure demonstrates a signal advantage offered by this technique. Regarding the shape of these curves, the exact cause of the initial increase in conductivity upon heating is not fully understood. One contribution may be the temperature dependence of the NH3 sensing response itself. Another may be the nature of the storage sites on the zeolite. If we assume the enhanced conductivity due to NH3 is caused by an increase in mobility of the surface H+ ions as suggested in ref. [5], then moving stored NH3 from other sites to the H+ sites upon heating would be expected to increase the conductivity. This, however, requires further study. The decrease in conductivity upon heating past the temperature corresponding to the peak conductivity (t > ∼15 s in Fig. 4) is presumably a consequence of purging most of the stored NH3 by either desorption, oxidation or both. It is reasonable to question the ability of such a sensor and technique to accurately measure the stored NH3 levels, given that the conductivity change of zeolite-based sensors due to gasphase NH3 is known to be temperature dependent [5]. The curves shown in Figs. 5 and 6 might be made more accurate if a weighted integration of the curves in Fig. 4 was done such that for each time interval (and corresponding temperature) the temperature dependence of the NH3 response was factored out. In this way the integration of a temperature compensated conductivity change due to gas desorption could more accurately reflect the amount of stored NH3 . This correction will be applied in our future studies as detailed measurements of the temperature dependence of the NH3 sensitivity were not done for the base-metal zeolite catalyst material used in this study. We recognize the thick-film zeolite sensor used in this study was much too dense for practical purposes, resulting in a long time required to saturate the response. For example, Fig. 3 shows this took at least 20 min for 500 ppm NH3 loading. We speculate the slow diffusion was a consequence of making the zeolite film with too high an alumina binder fraction. We speculate the alumina grains, being much smaller than the zeolite grains, served to block the space between them resulting in longer paths for the NH3 to diffuse. Additional experiments done on films with only 10% alumina by weight show the loading time required to saturate the response to be decreased by almost a factor of 10. These results will be the focus of a future report. Ideally we desire to make a sensor with same diffusion-time dependency for the NH3 storage as that for the material coated onto the actual SCR catalyst. Future studies should also investigate the influence of water vapor and NH3 selectivity. Water concentration may be an important cross-sensitivity parameter as it has been shown to affect the response of a zeolite-based NH3 sensor [4]. For the data presented here, water was fixed at 1%. We did not test the selectivity of the sensor’s response to NH3 . However, our prior studies on another zeolite material, HZSM-5, demonstrated a selective NH3 response compared to that for CO, HC’s and NOx.
D.J. Kubinski, J.H. Visser / Sensors and Actuators B 130 (2008) 425–429
This concurred with the findings previously reported for that material [4]. We expect the zeolite material used here to behave similarly, but this like the water sensitivity should be investigated in future studies. Future studies will also focus on testing our hypothesis that the average current as depicted in Fig. 5 scales with the NH3 loading on the sensor material. This can be tested comparing IAverage measurements with NH3 adsorption measurements obtained via thermogravimetric analysis (TGA). Additionally, we want to obtain more quantitative IAverage versus loadingtemperature measurements. We recognize that the heating method used in this study requires improvement. For example, the results of Fig. 6 could be easier to interpret quantitatively if a constant temperature ramp was used (degrees/second), with the same final temperature for each case. In summary, we have demonstrated that an ammonia storage sensor for a zeolite-based SCR catalyst is possible using the SCR catalyst material itself as the sensing element. This was done by measuring its change in conductivity upon heating, in an attempt to purge it of stored NH3. References [1] C. Lambert, R. Hammerle, R. McGill, M. Khair, C. Sharp, Technical advantages of urea SCR for light-duty and heavy-duty diesel vehicle applications, SAE Technical Paper Series: Technical paper 2004-01-1292, Society of Automotive Engineers: Warrendale, PA, USA, 2004.
429
[2] P. Tennison, C. Lambert, M. Levin, NOx control development with urea SCR on a diesel passenger car, SAE Technical Paper Series: Technical paper 2004-01-1291, Society of Automotive Engineers: Warrendale, PA, USA, 2004. [3] L. Lietti, I. Nova, E. Tronconi, P. Forzatti, Transient kinetic study of the SCR-DeNOx reaction, Catal. Today 45 (1998) 85–92. [4] R. Moos, R. M¨uller, C. Plog, A. Knezevic, H. Leye, E. Irion, T. Braun, K. Marquardt, K. Binder, Selective ammonia exhaust gas sensor for automotive applications, Sens. Actuators B Chem. 83 (2002) 181–189. [5] U. Simon, U. Flesch, W. Maunz, R. M¨uller, C. Plog, The effect of NH3 on the ionic conductivity of dehydrated zeolites N beta and H beta, Micropor. Mesopor. Mater. 21 (1998) 111–116. [6] I. Nova, L. Lietti, E. Tronconi, P. Forzatti, Transient response method applied to the kinetic analysis of the DeNOx-SCR reaction, Chem. Eng. Sci. 56 (2001) 1229–1237.
Biographies D.J. Kubinski received his PhD in physics from Wayne State University in 1995. He has been with the Ford Research Staff since 1986. His present interests are in sensing technologies for automotive exhaust. J. H. Visser received MS and PhD degrees in electrical engineering from Delft University of Technology, Delft, The Netherlands, in 1985 and 1989, respectively. Since 1990, he has joined the Ford Research Staff where he has been working on sensors and their applications for automotive systems. His research interests include exhaust gas sensors (air-to-fuel ratio, NOx , NH3 , HC and CO), gasoline-vapor sensors, and sensors for measuring automotive liquids. He is currently a technical leader at Ford Research and Advanced Engineering and is the project leader of the sensors group in the Chemical Engineering Department.