Dosimeter-Type NO_x Sensing Properties of KMnO₄ and Its Electrical Conductivity during Temperature Programmed Desorption

2.1. Sample Preparation and Characterization

The sensitive layer of the proposed NO_x dosimeter was prepared from 17 mol% KMnO₄ (Merck) deposited onto alumina, stabilized with 3% lanthanum (Puralox SCFa-140La3), serving as support oxide with surface area of 140 m²/g and mean particle diameter of 30 μm. The KMnO₄/La-Al₂O₃ powder was prepared by multiple infiltration of an aqueous solution of KMnO₄ into the La-Al₂O₃ powder, followed by drying at 100 °C and calcination at 600 °C for 5 h. Upon thermal decomposition KMnO₄ is known to form various potassium- and manganese-containing compounds, like K₂MnO₄ and K₃MnO₄, as well as manganese oxide MnO_x existing in different oxidation states [22,24–30]. The decomposition of KMnO₄ is reported by Boldyrev [25,26] to become noticeable in the temperature range from 205 to 280 °C. KMnO₄-impregnation and firing decreased the surface area of the powder to 100 m²/g (obtained by BET method (named after the originators of the method: Brunauer, Emmett, and Teller) relying on the adsorption of gases to determine the specific surface area of powders). The SEM (scanning electron microscope) analysis of the fired KMnO₄/La-Al₂O₃ powder is given in Figure 1 as a backscatter electron (BSE) image. The powder consists of spherical La-Al₂O₃-rich particles in the range of some tens of μm, partly embedded in a potassium and manganese comprising matrix, which was confirmed by energy-dispersive X-ray spectroscopy (EDX) measurements.

The KMnO₄/La-Al₂O₃ powder was mixed with an organic binder (KD2721, Zschimmer & Schwarz) in order to obtain a processable paste. The paste was deposited by spatula onto a 96% pure alumina substrate equipped with gold interdigitated electrodes (area 5 × 6 mm, finger width/distance 100 μm) and fired at 600 °C. The sample was pre-conditioned for several hours at temperatures up to 650 °C in NO_x containing oxygen-rich atmospheres.

The electrical properties of KMnO₄/La-Al₂O₃ were investigated in a test apparatus as sketched in Figure 2. Following installation in a quartz-tube furnace with inner diameter of 22 mm, the sample was heated to temperatures between 300 to 650 °C. The KMnO₄/La-Al₂O₃ layer was exposed to a 2 L/min lean gas flow (10% O₂, 50% N₂ humidified with a water bubbler at room temperature, and 5% CO₂ diluted in N₂ balance) with a gas exchange time of the system in the range of 8 s. The impedance Ẕ of the KMnO₄/La-Al₂O₃ sample was recorded by an impedance analyzer (Alpha High Performance Frequency Analyzer, Novocontrol).

2.2. Electrical Properties in the Unloaded State

The electrical properties of KMnO₄/La-Al₂O₃ in the unloaded state (after regeneration of the NO_x sorption sites) were evaluated from 300 to 650 °C by impedance spectroscopy in the frequency range of 1 Hz to 1 MHz. Plotting the impedances in the complex plane as Nyquist plots (real part Z′ and imaginary part Z″) yields near semicircular spectra at higher frequencies. This allows the bulk impedance to be modeled by a resistance R in parallel to a constant phase element CPE (R‖CPE). The corresponding impedance Ẕ_CPE expressed as a function of the model parameters n (ranging from 0 to 1) and Q as well as the angular frequency ω is given in Equation (1):

In Figure 3(a and b), examples for the corresponding Nyquist plots of the measured impedance data (dots) for 380 and 650 °C, together with the fitted R‖CPE curves in the upper frequency range (solid curves) and the corresponding fitting parameters, are displayed. While the KMnO₄/La-Al₂O₃ sample has a resistance of 180 kΩ at 380 °C, it decreases to 3.4 kΩ upon heating to 650 °C.

The electrical conductivity σ of the KMnO₄/La-Al₂O₃ specimen was estimated from the fitted R-values of the impedance in the R‖C dominated frequency range taking into account the electrode geometry. The electrode geometry is estimated from the capacitance of the uncoated structure, assuming a parallel-plate capacitor, as described in [31]. The resulting Arrhenius-like representation of σ as a function of inverse temperature 1/T in Figure 3(c) gives a thermal activation energy of the conductivity E_A of 0.8 ± 0.1 eV. This thermally activated conductivity leads to an almost two decades increase from ∼5·10⁻⁷ S/cm at 380 °C to 3·10⁻⁵ S/cm at 650 °C.

Information regarding the electrical conductivity of KMnO₄-based materials above the decomposition temperature in the literature is limited. In thermoelectric tests, KMnO₄ and its decomposition products were identified as n-type semiconductors by Boldyrev and Kabanov, and in the literature cited therein [26,28]. Upon thermal decomposition, the conductivity of KMnO₄ was found to increase, and depending on morphology, the conductivity was reported to range from 10⁻⁶ to 10⁻⁸ S/cm at 170 to 210 °C [26,28].

The lower conductivity of the KMnO₄/La-Al₂O₃ based material under investigation compared to the reported conductivity of pure KMnO₄ is attributed to the less conductive La-Al₂O₃ particles serving as support oxide in the applied sensitive coating. Recently published results on the K₂CO₃/La-Al₂O₃ system indicate a significant contribution of La-Al₂O₃ to the measured conductivity given its lower conductivity than that of pure K₂CO₃ [32].

3.1. Experimental Setup and Data Evaluation

To study the effect of NO_x in the low ppm range, the KMnO₄/La-Al₂O₃ sample was exposed to various NO and NO₂ concentrations, c_NO,in and c_NO2,in for defined time intervals t_NOx,in. NO_x was admixed to the 2 L/min lean base gas flow (10% O₂, 50% N₂ humidified with a water bubbler at room temperature, and 5% CO₂ diluted in N₂ balance). The outlet concentrations were determined by a chemiluminescence detector, as illustrated in Figure 2. In accordance to the reported catalytic activity of Mn-containing LNTs [23,33–36], as well as to results on LNT-based NO_x dosimeters [6,18], the NO_x sorption studies were performed at a sorption temperature T_sorption of 380 °C, with periodic heating to 650 °C for regeneration.

The sample impedance was recorded continuously during NO_x exposure. Since the fitted n-parameters of Ẕ_CPE (Equation (1)) of KMnO₄/La-Al₂O₃ in the high frequency range were found to be close to 1 (≈0.95), Q can be approximated by the capacitance C and the R‖CPE equivalent circuit model can be simplified to an R‖C circuit. Thus, R is calculated from the absolute value of the impedance |Ẕ| and the phase angle φ at a fixed frequency according to Equation (2).

From the Nyquist plots in Figure 3, 10 kHz (marked in red) was selected as an appropriate measurement frequency to monitor the temperature dependent electrical properties in the R‖C dominated range over time and was used, if not denoted otherwise. The absolute value of the relative resistance change due to NO_x exposure ΔR_rel is denoted as the sensor signal, and is defined by Equation (3), with R₀ being the base resistance in the NO_x unloaded state:

The analysis in terms of dosimeter-type sensing properties during NO_x sorption at constant flow rates is illustrated in Figure 4. During the NO_x loading stage, ΔR_rel is expected to increase in the presence of NO_x at T_sorption due to progressive NO_x accumulation, without recovery (Figure 4(a)). In the case of a constant flow rate, the cumulated NO_x exposure (or dose) A_NOx,in is given by the time integral of c_NOx,in as sketched in Figure 4(b), resulting in the unit ppm·s [6,16,37].

At a constant NO_x concentration, A_NOx,in scales linearly with t_NOx,in. The resulting characteristic line in Figure 4(c) correlates ΔR_rel with A_NOx,in. It has been shown in detail for a similar material in [6] that in the case of a linear correlation, the signal derivative of a NO_x dosimeter at a constant flow rate increases with the actual NO_x concentration.

3.2. Cumulative NO_x Detection at 380 °C

The presence of NO_x was found to decrease the resistivity of KMnO₄/La-Al₂O₃, with the electrical response continuing to satisfy the R‖C equivalent circuit (not shown). The temporal dependence of R on NO_x at 380 °C was studied by exposing the sample to pulses of NO and NO₂ for periods of t_NOx,in = 100 s with concentrations ranging up to 16 ppm. The pulse heights in terms of c_NO,in and c_NO2,in, together with the resulting sensor response, are displayed in Figure 5(a). The sensor response ΔR_rel (Equation (3)) increases stepwise in the presence of NO and NO₂ without any recovery at 0 ppm NO_x. The slope of ΔR_relvs. t increases with c_NOx,in. The characteristic line in Figure 5(b) is extracted from the measured data points and the course of the NO_x concentration according to Figure 4. ΔR_rel correlates almost linearly with the cumulated NO_x exposure A_NOx,in, independent of the NO_x species, up to at least 40% signal change with a NO_x sensitivity of 4.8%/1,000 ppm·s. The specimen thus provides comparable sensitivity to both NO and NO₂. Small deviations of the NO₂ related data points from linearity at the initial stage of exposure may originate from NO₂ adsorption on the inner surface of the feed lines. The sensing characteristics and in particularly the sensitivity of NO_x dosimeter with a comparable sensitive material were found to be dependent on the temperature as well as on the thickness of the sensitive layer [6,16].

Figure 5 indicates the strong and progressive sorption of NO and NO₂ onto KMnO₄/La-Al₂O₃ at 380 °C with corresponding impact on its electrical conductivity. As in K-Mn-containing LNTs [21,23], NO_x is expected accumulate on KMnO₄/La-Al₂O₃ by forming stable nitrates on the potassium sites generated upon KMnO₄ decomposition [22]. For LNTs it is well known that NO is first oxidized to NO₂ on redox active sites provided by e.g., precious metals, followed by chemical NO₂ storage by reaction with the alkaline (earth-) carbonates, mainly BaCO₃ or K₂CO₃, to form nitrates [38–40]. The observed increase in the conductivity of fully formulated LNTs in NO_x enables their application as total NO_x sensors [6,11,41] or for in-situ diagnostics of automotive catalysts [41–43].

The requirement of an incorporated oxidant, for the purpose of NO sorption, was verified by electrical means. Pure BaCO₃ or K₂CO₃, on the other hand, accumulates only NO₂, enabling conductometric NO₂ dosimetry, without NO cross-sensitivity [15,32]. Given the ability of KMnO₄/La-Al₂O₃ to detect either NO or NO₂, the oxidizing properties of KMnO₄/La-Al₂O₃ are demonstrated to be sufficient to convert NO to NO₂ prior to nitrate formation. This is consistent with MnO_x, as a product of KMnO₄ decomposition [25,27], being known as an effective oxidizing agent in NO_x reduction catalysts [22,24,33,34,36,44]. The contribution of MnO_x to the NO_x sorption capacity at 380 °C cannot be excluded [45–48].

The linear correlation between ΔR_rel and A_NOx,in in the low loading state of the NO_x dosimeter based on KMnO₄/La-Al₂O₃ in Figure 5 points on a sorption rate proportional to the NO_x concentration. This linearity provides a dual-mode functionality: while the sensor response corresponds directly to the cumulated NO_x exposure during the sorption period, the course of c_NOx,in can be determined via the signal derivative as described in [6,14]. Furthermore, these results demonstrate that decomposed KMnO₄ can be utilized in NO_x dosimeters and catalysts without any need for expensive precious metal additives due to its intrinsic oxidizing nature.

3.3. NO_x Concentration Sensitivity at 650 °C

An important criterion for a useful sensor is the ability to refresh or regenerate the device following accumulation of the target gas analyte, which in this study is NO_x. The decreased thermodynamic stability of the formed nitrates upon heating limits the catalytic activity of LNTs [38–40,49] and alters the cumulative NO_x sensing characteristics of carbonates and LNT-based sensors [6,17,18,32]. According to Becerra et al.[22], nitrate and nitrite-like compounds formed on KMnO₄-based materials decompose in the temperature range of ∼550 to 670 °C. Hence, a thermal release of sorbed NO_x, leading to a recovery of the sorption sites of KMnO₄/La-Al₂O₃, seems feasible.

The effect of NO_x on the resistivity of KMnO₄/La-Al₂O₃ was studied at 650 °C to investigate this temperature as being suitable for regeneration. The sample was exposed to the NO_x concentration profile shown in Figure 6(a) with up to 8 ppm NO and 75 ppm NO₂. ΔR_rel is calculated from the impedance at 1 MHz due to the increased conductivity. Again, the conductivity of KMnO₄/La-Al₂O₃ increases in the presence of NO_x. But, as shown in Figure 6(b), at 650 °C, the value for ΔR_rel follows the course of c_NOx,in (instead of ∫c_NOx,in dt) being characteristic of a common concentration-detecting gas sensor response. Despite the corresponding concentration-related characteristic line in Figure 6(c), which gives a linear correlation, the low sensitivity of only 2.7%/100 ppm NO₂ limits the application of KMnO₄/La-Al₂O₃ as a NO_x sensing material operated at 650 °C.

The reversibly sensor response at 650 °C in Figure 6 indicates that the equilibrium of the NO_x sorption on the KMnO₄-based material is shifted to the side of the reactants and the resulting fast desorption goes along with the loss of NO_x accumulation capability. Hence, 650 °C seems an appropriate temperature to release formerly sorbed NO_x and to recover the sorption capacity, as well as the electrical properties of KMnO₄/La-Al₂O₃. The reversibility of the sensor response of KMnO₄/La-Al₂O₃ at 650 °C is consistent with results on an LNT-based NO_x dosimeter [17].

3.4. Efficiency of Thermal Regeneration

In the following test series, the efficiency of thermal regeneration, the reproducibility of the dosimeter-type NO_x sensing characteristics, and the influence of NO_x exposure time were studied. The same KMnO₄/La-Al₂O₃ sample was exposed to 8 ppm NO₂ or NO at 380 °C in periods of 250, 500, 750, 1,000, and 2,000 s. Between each NO_x exposure period, the sample was regenerated at 650 °C for about 5 min in the lean gas flow. The sensor responses as a function of t_NO2,in and t_NO,in are compared in Figure 7. The five NO₂-borne curves of ΔR_rel depicted in Figure 7(a) are almost identical in the corresponding overlapping time scales; please note that the data points corresponding to the longest NO₂ exposure of 2,000 s are partly masked by the other data curves. The corresponding NO curves up to 1,000 s in Figure 7(b) are overlapping as well. The sensor behaves linearly (following an initial incubation period) up to a resistance change of about 40%, with the slope of ΔR_rel in Figure 7(a) being nearly constant up to about 1,000 s (8,000 ppm·s NO₂). The nonlinearity at the beginning of NO_x exposure in Figures 5 and 7, i.e., the slight initial slope increase during the first 375 s in NO₂ (3,000 ppm·s NO₂), is assumed to be caused by NO_x (in particular NO₂) being adsorbed on the feed gas lines resulting in a delayed sensor response. Further NO₂ exposure leads to a decrease in the slope and ΔR_rel reaches a value of 70% after half an hour in 8 ppm NO₂. By definition, ΔR_rel cannot reach 100% as the conductivity increases, resulting in a flattening of the curve of ΔR_rel with continuing NO_x loading. It is expected that a greater sensitive layer thickness would increase the linear range to higher NO_x levels, but at reduced sensitivity (slope d(ΔR_rel)/dA_NOx,in), as reported for LNT-based dosimeters [16].

The reproducibility of the sensor response in Figure 7 indicates that the sorption sites of the KMnO₄/La-Al₂O₃-based dosimeter material can be recovered by releasing sorbed NO_x thermally. Heating up to 650 °C restores the NO_x sensing characteristics at 380 °C, independent of the former NO_x exposure duration. The base resistance in the unloaded state R₀ was found to decrease slightly with time without impacting the NO_x sensitivity. This might be attributed to small morphological changes during thermal aging, which, however, are too small to be seen by SEM. It is noteworthy to mention that the dosimeter principle avoids such long term signal drifts by definition, since the zero level of ΔR_rel is reset after each regeneration step. This is one of the key advantages of conductometric dosimeters compared to classical semiconductor gas sensors.

4.1. eTPD Setup and Data Evaluation

eTPD on KMnO₄/La-Al₂O₃ is performed with the experimental arrangement shown in Figure 2. The eTPD related data and their evaluation are summarized in Figure 8.

To recover the NO_x sensing characteristics of KMnO₄/La-Al₂O₃ in between the NO_x sorption intervals in the lean gas flow in Figure 7, the sample was heated from 380 °C (T_sorption) to 650 °C (T_desorption) with a heating rate of 74 °C/min from 425 to 635 °C, while monitoring the impedance at 10 kHz. The temperature increase started at t_heat, 50 s after the end of the preceding NO_x dosing interval. The resulting NO_x desorption curve (Figure 8(a)) is displayed as a NO_x concentration c_released. At t_start = t_heat + 50 s an increase of NO_x is observed in the outlet until t_end. As illustrated in Figure 8(b), at a constant flow rate, the time integral of c_released, evaluated in the time interval of t_end – t_start = 300 s, reflects the released NO_x amount A_released. A_released is expected to be proportional to the quantity of sorbed NO_x, if the sorption sites are fully recovered by heating.

The conductance G of KMnO₄/La-Al₂O₃, with G = 1/R, was found to be affected by the temperature and the NO_x loading level, both changing during thermal regeneration. As sketched in Figure 8(a), NO_x release results in a convergence of G to G₀ = 1/R₀, G₀ being the temperature dependent conductance in the unloaded state. The time integral of the conductance upon heating, relative to those of G₀, is evaluated as the cumulative electrical response F_G. F_G is calculated according to Equation (4) and is shown in Figure 8(b) as the area between the curves corresponding to the two loading states. In the ideal case, F_G would be a measurand for A_released (Figure 8(c)) and would depend linearly on the cumulative sorbed NO_x amount.

4.2. Evaluation of the Released Amount

In Figure 7, the resistance responses of KMnO₄/La-Al₂O₃ during an NO₂ sorption series at 380 °C with various exposure periods t_NO2,in are reported. The subsequent regeneration by heating to 650 °C to release the formerly sorbed NO_x can be analyzed in terms of TPD. The corresponding outlet NO_x concentrations c_released in the lean 2 L/min gas flow, with a resolution of 0.1 ppm given by the CLD, for t_NO2,in up to 1,000 s are compared in Figure 9. After 1,000 s in 8 ppm NO₂ (red curve, as indicated in Figure 9), KMnO₄/La-Al₂O₃ starts to release NO_x at about 400 °C. c_released increases with temperature, and at about 550 °C, a maximum is reached at about 1.3 ppm. Shorter NO₂ exposure periods, representing a lower amount of NO_x loading, yield lower peak heights of c_released. At 650 °C, c_released reaches zero for all curves, indicating the end of NO_x release. Additionally, both peak maximum and desorption onset are shifted to lower temperatures with increasing t_NO2,in. The latter points to a lower stability of the sorbed NO_x with increased NO_x loading. Concerning the low values of c_released, it should be considered that the evolved NO_x is diluted in the 2 L/min lean gas flow and that the sensitive KMnO₄ coating amounts only to an area of about 30 mm² (5 × 6 mm).

The reproducibility of the dosimeter-type NO_x sensing characteristics at 380 °C (Figure 7) and the missing NO_x accumulation at 650 °C (Figure 6) reveal that the NO_x sorption capacity of KMnO₄/La-Al₂O₃ can be recovered by heating to 650 °C. Therefore, the quantity of sorbed NO₂ on KMnO₄/La-Al₂O₃ can be estimated from the subsequently thermally released NO_x amount A_released, being the area under the desorption peak as shown in Figure 8(b). Figure 9(b) reveals that A_released and hence the amount of NO_x sorbed in KMnO₄/La-Al₂O₃ increases almost linearly (with only a small offset) with NO₂ exposure, reflected by t_NO2,in. Consequently, NO₂ is sorbed on KMnO₄/La-Al₂O₃ with a time constant sorption rate during the 8 ppm NO₂ exposure periods. After 1,000 s in 8 ppm NO₂, resulting in a cumulated NO₂ exposure of 8,000 ppm·s, about 150 ppm·s NO_x are released. This indicates that only about 1.9% of the NO₂ in the passing gas flow is sorbed in the KMnO₄ based sensitive layer. The gas velocity of 5.3 m/min together the sensitive area length of 6 mm amounts to a residence time of about 70 ms, being comparable to those in catalysts [37]. However, the huge gas volume above the sensitive layer in the 22 mm diameter quartz tube inhibits full NO_x storage in this NO_x dosimeter setup. The small offset of A_released in Figure 9(b) amounts to about A_offset ≈ 15 ppm·s. If one divides this value by the integration time of 300 s, one obtains an average concentration of 0.05 ppm NO_x. A closer look at Figure 9(a) reveals that this is (roughly) the offset of the NO_x concentration measurement by the CLD with a resolution of 0.1 ppm. As a conclusion, the offset of A_released can be attributed to an integration error. The analysis of the corresponding data after NO exposure yield the same qualitative results (data not shown here) but with a smaller offset. Hence, a further explanation might be the partial overlap of c_released with the preceding decay of c_NO2,in due to NO₂ adsorption in the feed lines.

The observed sensor response ΔR_rel of KMnO₄/La-Al₂O₃ during NO_x sorption at 380 °C (Figure 7) obviously corresponds to the amount of loaded NO_x. In Figure 10, ΔR_rel, caused by 8 ppm NO or NO₂ for up to 1,000 s, is related to A_released, which is obtained from the subsequent regeneration shown in Figure 9. For NO exposure, as well as for NO₂ exposure, ΔR_rel increases linearly with A_released. Hence, in the investigated range, ΔR_rel serves as a linear measure for the NO and NO₂ loading levels of KMnO₄/La-Al₂O₃, and due to the constant NO_x sorption rate (Figure 9(b)), also for the cumulated NO_x exposure (NO_x dose). Thereby, the conductivity of KMnO₄/La-Al₂O₃ is slightly more sensitive to NO compared to NO₂. From a catalytic point of view, it would be expected that NO₂ in the gas flow influences the material's properties more than NO, since NO₂ can be sorbed directly as nitrate, whereas NO needs to be oxidized first [23,36]. However, the manganese oxide components of the decomposed KMnO₄ might become reduced upon oxidizing NO, thereby affecting the conductivity of KMnO₄/La-Al₂O₃ and hence the NO sensitivity. The delay in the sensor response resulting in an x-axis intercept in Figure 10 is expected to be caused by the already discussed inaccuracy of the determination of A_released by integration of small values of evolved NO_x (A_Offset ≈ 15 ppm·s in Figure 9(b)). In addition, NO_x (in particular NO₂) adsorption to the feed lines lowers the sensor response but increases the analyzed value for the desorbed amount.

Combining the classical TPD method, with the dosimeter-type electrical response of KMnO₄/La-Al₂O₃, demonstrates the possibility of sensing NO_x exposure and of electrically monitoring the NO_x loading level of the NO_x sorbent in-situ, both with linear correlation at low loading.

4.3. Electrical Information upon Thermal Regeneration

The conductance upon releasing NO_x provides information about the amount of previously sorbed NO_x. This may also be useful for NO_x dosimetry. Figure 11a depicts the courses of the conductance G during thermal regeneration after exposure to 8 ppm NO₂, for the different loading states indicated by its specific NO_x exposure period t_NO2,in. The course of the temperature is shown for comparison (black dots). G₀ reflects the conductance in the NO_x unloaded state corresponding to t_NO2,in = 0.

Being thermally activated, G₀ increases by nearly two orders of magnitude, which agrees with Figure 3. At 380 °C, the conductance in the partly NO_x loaded state G is higher than G₀. The difference between G and G₀ corresponds to the cumulative NO_x response, ΔR_rel (Figure 7). With progressive temperature, log G increases like log G₀. The difference between log G and log G₀ increases with t_NO2,in, indicating a correlation with the NO_x loading level. Between about 480 and 530 °C, the curves of G start to converge to those of G₀. Finally, above about 620 °C (230 s) all curves of G coincide with G₀ indicating that the unloaded state is recovered. A more detailed analysis reveals that the inflection point in the course of log G corresponds to the onset of NO_x release shown in Figure 9(a) as c_released. The temperature of the minimum in the slope of log G coincides with the temperature of the maximum of c_released. Both are being shifted to lower temperatures, the higher the former loading level was. Hence, the convergence of the curves of log G to the reference in the unloaded state can be attributed to thermal NO_x release from KMnO₄/La-Al₂O₃, which decreases the temperature-dependent conductivity to the unloaded value.

The comparison of the curves of the conductance G during regeneration (Figure 11(a)) suggests that the deviation of the course of log G from log G₀ might reflect the amount of previously sorbed NO_x. In fact, Figure 11(b) reveals a linear correlation between F_G calculated according to Equation (4) and the preceding sorption interval t_NO2,in. Accounting for the time constant NO and NO₂ sorption rate in the low loading state (exemplarily shown for NO₂ in Figure 9(b)), F_G is also a linear function of A_released as shown in Figure 11(c) for NO and NO₂ exposure, respectively. Therefore, besides of ΔR_rel, the cumulated electrical response F_G of KMnO₄/La-Al₂O₃ during regeneration may also be a suitable sensor signal for the cumulated NO_x exposure and the in-situ loading level. Again, the resulting sensitivity to NO is slightly higher than those to NO₂. Furthermore, the sensor response exhibits an offset, in particular, for NO₂. Besides of the small integration error A_offset when determining the area under the low level concentration curve during desorption in Figure 9a, these offsets likely originate from NO₂ adsorption in the feed lines.

Considering the electrode geometry, the conductivity σ can be calculated from the conductance and the data from Figure 11(a) can be plotted in an Arrhenius-like representation in the area of a constant heating rate as depicted in Figure 12. The data points of the electrical characterization of the KMnO₄/La-Al₂O₃ sample after equilibration at various temperatures shown in Figure 3(c) are added to Figure 12 as black dots. Concerning the thermally activated conductivity in the unloaded state, the direct comparison reveals that the eTPD data agree well with those obtained in the equilibrated state. This confirms the recovery of the sorption sites by heating to 650 °C. The NO_x saturated state of the carbonates was found to give a more pronounced transition in the curve of the equilibrated temperature-dependent conductivity upon thermal decomposition [32].

The investigation of NO_x sorption on KMnO₄/La-Al₂O₃ demonstrates that eTPD enables one to correlate the analyte-induced electrical response quantitatively with the actual analyte loading state. This is achieved by observing the electrical properties and the gas desorption characteristics simultaneously. The eTPD method might enhance the understanding of the analyte sorption related electrical properties of functional materials applied for gas sensing or catalysis. Approaches on interpreting the conductivity during thermally releasing gas species are reported in the literature as well, but—to our knowledge—only without a simultaneous quantitative analysis of the desorption peak and hence with a missing correlation with the actual loading state under identical conditions. Colin et al.[50] and Fortin et al.[51] modeled the electrical influence of chemisorbed gases on semiconductors upon heating and verified it for the system oxygen-CdSe. The slope of the conductivity is reported to give information on desorption or binding energies of the species [51–53]. Rossé et al.[53] explains in detail the course of the resistivity in the Arrhenius-like representation during TPD being dependent on the heating rate and the amount of chemisorbed species. Additionally, the recovery of the initial loading state is described as a convergence to the unloaded reference. This description agrees fully with the interpretation of the results on NO_x loaded KMnO₄/La-Al₂O₃ in Figure 11. However, applying eTPD, these electrical results were additionally verified by the analysis of the simultaneous desorption peak (Figure 9(a)). Yamazoe et al.[54] and Rodríguez-Gonzáles et al.[55] compared the conductivity with the evolution of desorbed gases as well, but these tests were performed on multiple samples in different setups. The Simon group [55–57] investigated the temperature dependent NH₃ loading level of zeolites being active for the selective catalytic reduction (SCR) of NO_x. The conductivity upon heating reveals information on the conduction mechanism, the NH₃ desorption temperature as well as the SCR active temperature region allowing for in-situ reaction monitoring. Kubinski et al.[58] demonstrated that the average resistance during the thermally-induced NH₃ release from an SCR zeolite catalyst correlates with the former NH₃ exposure. This enables in-situ monitoring of the amount of sorbed NH₃ with a higher sensitivity compared to those in the NH₃ sorption mode [58].

The equivalency of the two conductivity-related sensor responses ΔR_rel and F_G of KMnO₄/La-Al₂O₃ as a measure for the cumulated NO_x exposure is demonstrated in Figure 13 as a monotone and almost linear correlation, independent of the type of exposed NO_x species. Both values can be applied as NO_x dosimeter-type sensor responses and correlate linearly with the quantity of sorbed NO_x, enabling in-situ monitoring of the loading state, although they are analyzed upon NO_x sorption at 380 °C (ΔR_rel) and upon NO_x release by heating up to 650 °C (F_G), respectively. Hence, dependent on the application and the information of interest, two different sensing modes are feasible with the proposed impedimetric NO_x dosimeter based on KMnO₄. In both cases, NO_x is accumulated in the sensitive layer at sorption temperature and thermally released during periodic regeneration intervals. However, in the first method, the change in the conductance during NO_x accumulation in the low loading state is monitored as a continuous and linear measure for the cumulated NO_x exposure as well as for the amount of sorbed NO_x. Concentration information can be obtained time-continuously from the signal derivative. On the contrary, in the second method, the integrated difference between the conductivity during NO_x release upon heating and the conductivity in the unloaded state serves as the measurand. Unfortunately, no time-continuous information on the NO_x concentration can be obtained.

Simultaneous NO_x detection in the sorption and release mode may be realized on one single sensor platform, with multiple independently heated sensitive layers, as described in [17]. A combination of both sensing modes to extract further information will be the focus of further research. The redundant sensing information is expected to enable a plausibility consideration of the time resolved sensor signal during NO_x sorption and of the regeneration success. Additionally, the linear measurement range for the NO_x exposure is expected to be enhanced in the regeneration mode. Another important issue of gas sensors is the sensitivity to other gases as well as poisoning of sensitive layers, e.g., by SO₂ [59,60]. Since interfering gases might affect NO_x sorption and release as well as the temperature-dependent conductivity differently, a combination of the electrical responses upon NO_x sorption and release might be particularly useful.