1 Introduction

Emerging more stringent vehicle emissions rules for non-methane organic gases (NMOG) plus NO x represent an 80% reduction from Tier 2 Bin 5 levels [1]. Diesel and lean-burn gasoline engines have a higher efficiency than stoichiometric combustion; however, the excess O2 in the exhaust undermines the catalytic reduction of NO x to N2. Therefore, advanced NO x reduction technologies must be applied to exploit more efficient lean-burn engine technologies.

Currently, there are two commercialized technologies for DeNO x of lean combustion engine exhaust: NO x storage and reduction (NSR) [2, 3] and selective catalytic reduction (SCR) [4, 5]. NSR requires periodic lean-rich operation to enable NO x trapping and regeneration in the lean NO x trap (LNT). During the regeneration, the catalyst is briefly exposed to a rich feed, typically containing a reductant mixture of H2, CO, and hydrocarbons (HCs), which converts stored NO x to N2 and undesired by-products such as NH3 (toxic) and N2O (greenhouse gas). While NSR can be an effective solution for light- and medium-duty vehicles, its high precious group metal (PGM) loading, sulfur intolerance, limited catalyst durability, and by-product formation collectively limit NSR deployment. In contrast, the second technology, SCR, requires NH3 as reductant to reduce NO x over vanadium-based or metal (Fe, Cu)-exchanged zeolite catalysts. The NH3 is generated by thermal conversion of urea through controlled dosing. SCR technology has been widely adopted, especially since the advent of Cu-exchanged small-pore chabazite zeolite, which has unparalleled durability, activity, and selectivity. The requirement for the urea feed system generally limits SCR to large luxury cars and heavy-duty vehicles.

Several years ago, Toyota researchers [6,7,8,9,10] introduced a diesel DeNO x technology called “Di-Air” (diesel NO x aftertreatment by adsorbed intermediate reductants). Recent studies by Toyota and others have demonstrated significant NO x conversion enhancement, particularly at elevated temperatures (up to 800 °C) and space velocities (up to 120,000 h−1). Di-Air expands the operating temperature window giving high NO x conversion (from 500 to 800 °C). Published performance data suggest that the NO x reduction is promoted by a new reaction pathway involving strongly adsorbed intermediates such as R-NCO and R-CN generated from sequential HC oxidation by O2 and NO at fast injection frequency. More recent studies by our group [11,12,13] have shown that high frequency pulsing with propylene (C3H6) confirm the pioneering Toyota studies and reveal further NO x conversion enhancement at low temperature. Zheng et al. [12] showed using a dual-layer LNT + SCR monolithic catalyst that high frequency pulsing reduces the C3H6 light-off temperature to ~185 °C and proposed a corresponding mechanism. Other studies have advanced the understanding of Di-Air. Wang et al. [14] used a TAP reactor to investigate the effect of ceria component on NO x reduction and found that the oxygen anion defects derived from lattice oxygen are active in NO reduction under high temperature cycling. Reihani et al. [15] studied the reactivity of a range of reductants over a Pt/Rh LNT catalyst and found that CO yields the highest N2 selectivity. Finally, Ting et al. [16] developed a global kinetic model to simulate HC intermediate formation by rapid pulsing that will be published elsewhere.

Most previous studies of NSR have used H2 or CO as the reductant to study the NO x reduction mechanism over LNT catalyst [17,18,19], due to their comparative simplicity (versus hydrocarbons) and effective NO x reduction performance. However, real engine exhaust contains more complex reductants such as olefins, aromatics, and higher molecular weight alkanes. The complexity of the reductant type, HC intermediate formation, additional HC oxidation reactions, and potential coke formation are of interest to researchers. There have been some efforts to unravel the reaction pathways by the use of representative HCs such as C3H6 [20,21,22] and more complicated ones such as ethylene (C2H4), propane (C3H8), m-xylene (C8H10), and dodecane (C12H26) [15, 23]. Abdulhamid et al. [24] studied the effect of the reducing agent on NO x reduction over LNT catalyst and found that C3H8 is less active than C3H6. Al-Harbi et al. [23] reported that selected HCs (C3H6, C8H10, and C12H26) show comparable NO x reduction efficiency to H2 and CO at temperatures higher than 300 °C over a model LNT catalyst. At lower temperatures, HCs are generally less active than H2 due in part to HC activation as well as self-inhibition. Reihani et al. [15] reported that CO and C3H6 show the highest N2 selectivity under a 3-s pulsing period and 15% duty cycle.

In this study, we use three reductants (H2, C3H6, and C3H8) as model reductants, respectively, representing carbon-free, olefin, and alkane species to assess NO x reduction performance over a range of cycling frequencies. While aforementioned studies have made progress towards establishing a working mechanism for high frequency pulsing [6, 12], there have been few attempts to generalize this mechanism to explain differences across the reductant spectrum and mixtures. A specific goal of this study is to collect performance data of the three reductants with the aim to generalize the fast lean-rich cycling mechanism and provide improved strategies to enhance the conversion of NO x and less active alkane reductant.

2 Experimental

2.1 Catalyst

The LNT monolithic catalyst was provided by BASF (Iselin, NJ). The catalyst has a cell density of 400 cpsi (cell per square inch), a washcoat loading of 4.6 g/in3, and a PGM (Pt/Rh = 8) loading of 90 g/ft3. The washcoat also contained 15 wt% barium oxide (BaO) and 34 wt% ceria (CeO2) on a γ-Al2O3 support. The catalyst sample (~55 channels, 0.42 in diameter, 1 in length) was cut from a large cylindrical core using a dry diamond saw. The LNT catalyst was aged at 700 °C for 33 h in a closed benchtop muffle furnace (Thermo Scientific, FB1300).

2.2 Cyclic Experiments

The experimental setup was described in our previous studies [13, 22]. The flow reactor was fed a synthetic mixture prepared using a bank of ultra-high purity gases (Praxair, Inc.) metered by mass flow controllers (MKS Inc.). Water was injected by a syringe pump (Teledyne Isco, model 100DX) and vaporized by a heating system. All the lines were heated to 130 °C to avoid water condensation. A solenoid actuated four-way valve (Valco, Inc., micro-electric two position valve) provided the lean-rich cycling. The total flow rate was maintained at 3000 sccm (25 °C, 1 atm) comprising a main stream containing inert Ar flowing at 1200 sccm (STP). It was mixed with lean or rich components just upstream of the catalyst. In order to minimize upstream axial mixing of lean and rich feed, the switching valve was placed close to monolith reactor (~2 ft.). The tubular reactor system consisted of a quartz tube flow reactor and a Thermocraft™ tube furnace. Three K-type stainless steel sheathed thermocouples (Omega Engineering) inserted at the upstream, monolith, and downstream measured the feed gas (T f), catalyst (T c), and effluent (T e) temperature, respectively. A FTIR spectrometer (Thermo Scientific, 6700 Nicolet) monitored the species concentration of NO, NO2, N2O, NH3, CO, C3H6, C3H8, CO2, and H2O. A LabVIEW™ program was used to control overall system (MFCs and thermocouples) and acquire flow rate and temperature data.

Lean-rich cycling experiments were conducted over a model LNT catalyst. Table 1 described the feed compositions for the slate of experiments. Three different cycle times were employed with the same rich duty of 14%: 60 s lean/10 s rich, 30 s lean/5 s rich, and 6 s lean/1 s rich. Reductants included H2, C3H6, and C3H8. The feed stoichiometry was characterized using the stoichiometric number, S N, defined as

$$ {S}_{\mathrm{N}}=\frac{2\left[{\mathrm{O}}_2\right]+\left[\mathrm{NO}\right]}{\left[{\mathrm{H}}_2\right]+9\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]+10\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]} $$
(1)
Table 1 Feed gas compositions for the cycling experiments
Full size table

Table 1 reports the rich phase compositions, and all feeds had the same S N (8.06). In Feed #1 (H2 as reductant), the H2 and O2 concentrations were reduced by a half to avoid a flammable mixture. Feed #2 or #3 contained concentrations of C3H6 or C3H8 that gave the same S N value. In Feeds #4–#6, the concentration of C3H6 and C3H8 were varied to study the HC mixture effect of NO x reduction. Prior to each experiment, the catalyst was pre-treated by 5% O2 at 400 °C for 30 min, followed by exposure to cycling conditions at 30 s lean/5 s rich and 6 s lean/1 s rich for 15 min, respectively.

At each temperature, once a cyclic pseudo steady-state was reached (~30 min), the final 5 cycles were averaged to quantify the cycle-averaged NO x conversion, reductant conversions, and product yields. The NO x , C3H6, and C3H8 conversions are given by

$$ {X}_{{\mathrm{NO}}_x\ }\left(\%\right)=100\times \frac{{\left[{\mathrm{NO}}_x\right]}_0-\left[{\mathrm{NO}}_x\right]}{{\left[{\mathrm{NO}}_x\right]}_0} $$
(2)
$$ {X}_{{\mathrm{C}}_3{\mathrm{H}}_6\ }\left(\%\right)=100\times \frac{{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_0-\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}{{\left[{\mathrm{C}}_3{\mathrm{H}}_6\right]}_0} $$
(3)
$$ {X}_{{\mathrm{C}}_3{\mathrm{H}}_8\ }\left(\%\right)=100\times \frac{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_0-\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}{{\left[{\mathrm{C}}_3{\mathrm{H}}_8\right]}_0} $$
(4)

The product yields are given by

$$ {Y}_{NH_3\ }\left(\%\right)=100\times \frac{\int_0^{\uptau_{\mathrm{T}}}\left[{NH}_3\right]dt}{\int_0^{\uptau_{\mathrm{T}}}\left({\left[\mathrm{NO}\right]}_0\right)dt} $$
(5)
$$ {Y}_{{\mathrm{N}}_2\mathrm{O}\ }\left(\%\right)=100\times \frac{2{\int}_0^{\tau_{\mathrm{T}}}\left[{\mathrm{N}}_2\mathrm{O}\right]dt}{\int_0^{\tau_{\mathrm{T}}}\left({\left[\mathrm{NO}\right]}_0\right)dt} $$
(6)

Here, τ T is the total cycle time, and [ ] denotes the effluent concentrations of the various species at the exit of the catalyst. The subscript “0” denotes the inlet concentration of NO, C3H6, and C3H8.

2.3 Steady-State Experiments

HC oxidation steady-state experiments were carried out over the LNT catalyst at fixed S N of 1, 1.33, and 2. Each experiment was given sufficient time (at least 30 min) to achieve steady-state effluent concentrations. The C3H6 oxidation steady-state experiments consisted of three C3H6 concentration (1111, 1667, 2222 ppm) and fixed O2 concentrations (1%), H2O (3.5%), CO2 (5%), and the balance Ar over a range of temperatures (120–400 °C). Similarly, the C3H8 oxidation steady-state experiments consisted of three C3H8 concentration (1000, 1500, 2000 ppm) and fixed concentrations of O2 (1%), H2O (3.5%), CO2 (5%), and balance Ar over a range of feed temperatures (150–500 °C). The catalyst was pre-treated prior to each experiment, by 5% O2 at 400 °C for 30 min.

2.4 In Situ DRIFTS Experiments

In situ DRIFTS (Diffuse Reflectance Infrared Fourier Transform Spectroscopy) experiments were conducted using a FTIR spectrometer (Thermo Scientific, 6700 Nicolet) equipped with a MCT detector and a high temperature reaction chamber (Harrick Scientific Praying Mantis) outfitted with ZnSe windows. Details about the DRIFTS setup are reported elsewhere [25]. The powder sample was prepared by scraping the washcoat from the monolith wall and pressing the powder into a 70-mg pellet. The total flow rate for all experiments was maintained at 50 sccm (STP). The catalyst was pre-treated at 400 °C in a flow of 5% O2 in He for 1 h and then cooled to the target temperature to collect the background spectra. Then the catalyst was sequentially exposed to 500 ppm NO and 5% O2 for 1 h during which time DRIFTS spectra were collected by collecting 64 scans with 4 cm−1 resolutions in the range of 650–4000 cm−1. After pre-adsorbing NO x on the catalyst surface, 1000 ppm C3H6 or C3H8 and 1% O2 and balance Ar were introduced, and the catalyst was exposed to the gas mixture for half an hour. The DRIFTS experiments were carried out from 200 to 300 °C which corresponded to the light-off temperature range of C3H6 or C3H8.

3 Results

3.1 Impact of Reductant Type

Figure 1 compares the cycle-averaged NO x conversions as a function of feed temperature for the three reductants (1a. H2; 1b. C3H6; 1c. C3H8) in the left column, each for three total cycle times (60/10, 30/5, and 6/1) and fixed rich duty cycle (14%). Cycle-averaged NO x conversion as a function of cycle-averaged catalyst temperature for three reductants (1d. H2; 1e. C3H6; 1f. C3H8) is provided in the right column. Overall, H2 is the most effective reductant for low feed temperatures (≤250 °C), in line with previous studies [15, 23]. H2 and C3H6 show similar effectiveness above 250 °C with some differences noted. High NO x conversion (>75%) with C3H8 is encountered only at feed temperatures exceeding ~300 °C for the long cycle (60/10) and ~350 °C for the short cycle (6/1). The H2 data show an increase in the NO x conversion with decreasing cycle time over the entire range of feed temperatures. The NO x conversion increases by over 20% (absolute) when the cycle time is decreased from 70 to 7 s. During fast cycling (6/1), the NO x conversion approaches 95% over a range of feed temperatures (150–300 °C). These trends are consistent with the study of Reihani et al. [15]. Figure 1b shows a comparable NO x conversion enhancement with the cycle time decrease of up to 20% (absolute) when using C3H6 as the reductant. These results are consistent with those of Perng et al. [11] who reported ~90% NO x conversion when feeding a C3H6 containing mixture at 0.14 Hz injection frequency. The short cycle data show that the NO x conversion at low feed temperatures (<250 °C) increased by up to ~40% from those of the longer cycle time (70 s). Figure 1c shows that C3H8 is not nearly as effective reductant. Enhanced NO x conversion with increased injection frequency of propane is only achieved at high feed temperatures (>350 °C). In fact, in the intermediate temperature range (250–325 °C), the NO x conversion increases with increasing cycle time. This opposite trend indicates that the reductant type plays a prominent role with regard to the NO x conversion at least at intermediate temperature. The catalyst temperature range expanded to 550 °C due to the large exothermic heat effects of the reductant oxidation, i.e., more than 150 °C difference.

Fig. 1
figure 1

Cycle-averaged NO x conversion as a function of feed temperature using (a) H2, (b) C3H6, and (c) C3H8 as reductant or of catalyst temperature using (d) H2, (e) C3H6, and (f) C3H8 as reductant [conditions: lean: 300 ppm NO, 5% O2 (for H2) or 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 8.125% H2 or 1.8% C3H6 or 1.62% C3H8, 2.5% O2 (for H2) or 5% O2, 3.5% H2O, and 9% CO2]

Full size image

It is noteworthy that for the highest feed temperatures (T f >350 °C), the reductant identity is not as important. Figure 2 replots the same data as that in Fig. 1 to directly compare NO x conversion using three reductants (H2, C3H6, and C3H8) at three different cycle times (6/1, 30/5, and 60/10) as a function of the feed temperature (T f; left column) or cycle-averaged catalyst temperature (T c; right column). Within an intermediate feed temperature range (250–350 °C), the NO x conversion shows a strong dependence of reductant type for all three cycle times. H2 or C3H6 give comparable NO x conversion while C3H8 gives a much lower NO x conversion. In contrast, at high feed temperature (T f >350 °C), the difference in the NO x conversions obtained with each of the three reductants decreases. The corresponding dependence on cycle-averaged catalyst temperature is shown in the right column of Fig. 2d–f. These plots show a finer distinction between the three reductants as the cycle time is shortened from 60/10 to 6/1. For catalyst temperatures exceeding ~450 °C, the data show that propylene is the superior reductant for the 6/1 (Fig. 2d) and 30/5 (Fig. 2e) cycle timing. In fact, propane is the second most effective reductant although for the slowest cycling (60/10), the difference between propylene and propane is negligible. For all three cycle times, H2 is the least effective reductant, exhibiting a sharper falloff in the NO x conversion at the highest temperatures. While the difference between H2 are only several percentage points, these data suggest that the factor(s) beyond NO x storage site utilization are at work. We elaborate on this point later.

Fig. 2
figure 2

Cycle-averaged NO x conversion as a function of feed temperature using H2, C3H6, and C3H8 as reductant (a) 6/1, (b) 30/5, and (c) 60/10 or of catalyst temperature using (d) 6/1, (e) 30/5, and (f) 60/10 [conditions: lean: 300 ppm NO, 5% O2 (for H2) or 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 8.125% H2 or 1.8% C3H6 or 1.62% C3H8, 2.5% O2 (for H2) or 5% O2, 3.5% H2O, and 9% CO2]

Full size image

Figure 3 shows the corresponding C3H6 and C3H8 conversions. The reductant conversions exhibit similar trends as the NO x conversions in Fig. 1. As the cycle time is decreased from 70 to 7 s, the effective C3H6 light-off temperature decreases from ~250 to ~225 °C. Moreover, complete conversion is sustained over a wide range of feed temperatures for the shortest cycle time. In contrast, for the longer cycle time, the conversion does not exceed 80% at the highest feed temperature considered (400 °C). For both the 35 and the 70 s cycle time, the C3H6 conversions are identical over the entire temperature range. In contrast, the C3H8 light-off temperature increases upon a reduction in the cycle time at intermediate temperature with the C3H8 conversions converging at higher feed temperature (350–400 °C). The different impact of cycle time for different reactants indicates that the Di-Air technology is strongly dependent on the reductant nature in an intermediate temperature range. The corresponding reductant conversion versus catalyst temperature data is provided in Fig. 3c, d. A detailed mechanism study is proposed in the next section.

Fig. 3
figure 3

Cycle-averaged reductant (C3H6, C3H8) conversions as a function of feed temperature using (a) C3H6 and (b) C3H8 as reductant or of catalyst temperature using (c) C3H6 and (d) C3H8 as reductant [conditions: lean: 300 ppm NO, 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 1.8% C3H6 or 1.62% C3H8, 5% O2, 3.5% H2O, and 9% CO2]

Full size image

A comparison of cycle-averaged by-product yields over the range of cycle times of the three reductants is shown in Figs. 4 (NH3) and 5 (N2O). In general, the data show that for a fixed cycle time that H2 has the highest NH3 yield at low temperatures (<250 °C) while C3H6 has the highest NH3 yield at elevated temperatures (>350 °C). The data reveal that NH3 yield decreases with increasing cycle frequency for each reductant. The magnitude of the NH3 yield decrease depends on the feed temperature and reductant. For example, the C3H6 data show negligible NH3 (~3%) slip for the faster cycle (6/1), as compared to much higher NH3 yield (~35%) at the conventional switching frequency (60/10). A similar trend was reported by Perng et al. [11] and Zheng et al. [12]. For H2 and C3H8, the NH3 yields converge at elevated temperatures for the 60/10 and 30/5 cycles. In contrast with C3H6, the yield diverges at the same temperature range. The duration of the rich regeneration significantly impacts the net NH3 generation, with a longer rich cycle being more favorable to NH3 formation. For example, at a feed temperature of 400 °C, the NH3 yield decreases from ~35 to ~15% and then to ~2.5% when the cycle time is shortened from 60/10 to 30/5 and then to 6/1. A similar decrease is observed for H2 and C3H8 at 400 °C. All reductants behave similarly at intermediate feed temperature (300 °C) with the NH3 yield decreasing from its maximum value at 60/10 to ca. 5% at 6/1. The corresponding NH3 yield versus catalyst temperature is provided in Fig. 4d–f.

Fig. 4
figure 4

Cycle-averaged NH3 yield as a function of feed temperature using (a) H2, (b) C3H6, and (c) C3H8 as reductant or of catalyst temperature using (d) H2, (e) C3H6, and (f) C3H8 as reductant [conditions: lean: 300 ppm NO, 5% O2 (for H2) or 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 8.125% H2 or 1.8% C3H6 or 1.62% C3H8, 2.5% O2 (for H2) or 5% O2, 3.5% H2O, and 9% CO2]

Full size image
Fig. 5
figure 5

Cycle-averaged N2O yield as a function of feed temperature using (a) H2, (b) C3H6, and (c) C3H8 as reductant or of catalyst temperature using (d) H2, (e) C3H6, and (f) C3H8 as reductant [conditions: lean: 300 ppm NO, 5% O2 (for H2) or 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 8.125% H2 or 1.8% C3H6 or 1.62% C3H8, 2.5% O2 (for H2) or 5% O2, 3.5% H2O, and 9% CO2]

Full size image

The dependence of N2O yield on cycle time is shown as a function of feed temperature for the three reductants in Fig. 5 (a. H2; b. C3H6; c. C3H8). Qualitative differences are noted for each reductant. For H2, the N2O yield decreases with increasing switching frequency. In contrast, increasing frequency increases the N2O yield for both C3H8 and C3H6, although the latter is only true at lower temperatures. At higher temperature, the N2O yield decreases with increasing frequency. This is particularly evident for C3H6 (Fig. 5b). The N2O yield is highest at T f = 225 °C, ~15% for the 6/1 cycle and decreases to ~5% for the 30/5 cycle. This trend suggests that the shorter storage and regeneration promote the formation of active intermediates which are beneficial to NO x conversion but somewhat increase N2O formation rates. The N2O yield as a function of catalyst temperature is provided in the right column of Fig. 5 (d. H2; e. C3H6; f. C3H8). At T c = 250 °C, H2 and C3H6 show the highest N2O yield while the maximum N2O formation is delayed to ~300 °C for C3H8. The comparison of N2O formation by three reductants at 6/1 and 60/10 is shown in Fig. 6. For each of the reductants, the N2O concentration exhibits a maximum versus temperature, an expected result. On the other hand, the onset temperature for N2O formation is similar for both propylene and hydrogen. Moreover, for the slower 60/10 cycling (Fig. 6b), the N2O yield is clearly higher for propylene and H2 than for propane which ignites at a higher temperature. For the 6/1 timing (Fig. 6a), N2O yields are more similar for all three reductants. Some of these trends are consistent with previous studies by Perng et al. [11] and Zheng et al. [12]. The formation of N2O is generally thought to occur by reaction between adsorbed NO and N species on reduced Pt [26]. Partridge et al. [27,28,29] reported that N2O formation strongly depends on the reaction between residual stored NO x and adsorbed reduction intermediates on the catalyst surface. Understanding the nuances of these trends is difficult without additional experiments, a subject for future study.

Fig. 6
figure 6

Cycle-averaged N2O yield as a function of catalyst temperature for at (a) 6/1 and (b) 60/10 s using three different reductants [conditions: lean: 300 ppm NO, 5% O2 (for H2) or 10% O2, 3.5% H2O, and 9% CO2; rich: 300 ppm NO, 8.125% H2 or 1.8% C3H6 or 1.62% C3H8, 2.5% O2 (for H2) or 5% O2, 3.5% H2O, and 9% CO2]

Full size image

Figure 7a, b shows the steady-state HC conversion as functions of feed temperature for three concentrations of C3H6 and C3H8. In these experiments, the catalyst was exposed to a feed consisting 1% O2 and 2222 ppm C3H6 or 2000 ppm C3H8 to maintain the same stoichiometric number (S N = 1). The two additional HC concentration levels correspond to S N = 1.33 and 2 based on 1% O2 in the feed mixture. The C3H6 light-off curve (Fig. 7a) shows an opposite trend, which indicates that C3H6 oxidation is self-inhibiting on the LNT catalyst. In contrast, Fig. 7b shows that the C3H8 light-off temperature decreases with an increase in the C3H8 concentration. This indicates that the overall reaction order is positive with respect to C3H8. As discussed in our previous works [12, 30], faster switching enables improved utilization of C3H6 due to increased contact with the oxidant O2. This leads to a higher C3H6 conversion and may mitigate the inhibition by surface C3H6, especially at low temperatures. In addition, the ceria provides an enhanced oxygen supply to the Pt crystallites from proximal oxygen storage sites [6, 12]. Thus, more frequent switching enhances the utilization of oxygen buffer (ceria). C3H8 does not exhibit the same inhibition behavior. As a result, a prolonged cycle is beneficial to the conversion of both C3H8 and NO x . We expand on these points below. The corresponding HC conversions versus catalyst temperature is provided in Fig. 7c, d.

Fig. 7
figure 7

(a) C3H6 conversion and (b) C3H8 conversion as a function of feed temperature or (c) C3H6 conversion and (d) C3H8 conversion as a function of catalyst temperature during steady-state operation [condition: varied concentration of HC, 1% O2, 3.5% H2O, 5% CO2 bal. Ar]

Full size image

3.2 DRIFTS Measurements

Toyota researchers [8] have proposed a mechanism for the rapid cycling Di-Air technology that involves the generation of HC-based intermediates (e.g., R-NCO and R-CN) that react selectively with NO x , producing N2. More specifically, Bisaiji et al. [7] reported based on reactor and surface IR measurements that the NO x conversion enhancement involves active organic intermediates (CxHyOxNt) generated from adsorbed NO x and partially oxidized HC during short rich phase duration. The mechanism is distinguished from the conventional NO x storage and reduction pathway that involves the conversion of stored nitrites and nitrates. Recent work by Zheng et al. [12] reported that a dual-layer LNT + SCR catalyst enhanced NO x conversion at low feed temperature. The enhancement was attributed to the formation of a reactive intermediate such as a nitrogenate which reacts selectively with NO x . The contrasting features obtained with H2, C3H6, and C3H8 during cyclic switching operation and the steady-state in situ DRIFTS measurements were presented in this section to help provide a deeper understanding of the underlying mechanism.

In situ DRIFTS measurements were carried out using the washcoated NSR catalyst in powder form to identify the evolution of surface organic intermediates during transient storage and reduction to identify any differences between the two hydrocarbon reductants. The DRIFTS experiments were carried out in the temperature range of 200–300 °C to better understand the NO x conversion differences when using propylene and propane. At higher temperatures in that range, the C3H8 is found to overcome the dehydrogenation kinetic barrier to form olefinic species and corresponding HC intermediates, such that differences between the alkane and olefin are negligible. However, at lower temperature, the effect of cycle time for NO x reduction enhancement depends on the reactivity of the two hydrocarbons.

Figure 8a, b compare the temporal evolution of DRIFTS spectra in the 1300–1100 cm−1 range when exposing the NO x pre-stored catalyst, initially exposed to NO x (500 ppm NO and 5% O2 for 1 h), to a feed mixture containing 1000 ppm C3H6 (Fig. 8a) or 1000 ppm C3H8 (Fig. 8b) in the presence of 1% O2 at 200 °C (It is noted that this feed is slightly lean.). The two major peaks evident before the admission of the reductant at 1220 and 1270 cm−1 are assigned to barium nitrite (Ba(NO2)2) [31] and barium nitrate (Ba(NO3)2) [12], respectively. Both peaks decrease upon exposing the catalyst to 1000 ppm C3H6 with 1% O2 (net lean). The sharp decrease of the 1220 cm−1 peak suggests that the barium nitrite is easily reduced at 200 °C; in comparison, the 1270 cm−1 peak assigned to the bulk nitrate is more slowly consumed. A somewhat different trend is observed when exposing the NO x pre-stored NSR catalyst to 1000 ppm C3H8 with 1% O2 at 200 °C (Fig. 8b). While the barium nitrite peak (1220 cm−1) decreases as it did with the C3H6 feed, the nitrate (1270 cm−1) peak increases during the C3H8 exposure. This suggests that the surface nitrites are oxidized by O2 rather than reduced by the less active C3H8. The low reactivity of propane makes more O2 available to react. With propane not very reactive at 200 °C, more O2 is available to react according to

$$ {\mathrm{NO}}_{2\left(\mathrm{s}\right)}^{-}+0.5{\mathrm{O}}_2\to {\mathrm{NO}}_{3\left(\mathrm{s}\right)}^{-} $$
(7)
Fig. 8
figure 8

DRIFTS spectra taken over LNT catalyst after 1 h pre-adsorbed NO x and subsequently exposed to a flow of (a) C3H6 or (b) C3H8 + 1% O2 for 30 min at 200 °C within wave number range of 1300–1100 cm−1 and (c) C3H6 or (d) C3H8 + 1% O2 for 30 min at 300 °C within wave number range of 1400–1200 cm−1

Full size image

Several studies [31,32,33] have reported that the nitrate route, in which NO oxidizes first to NO2 and is subsequently stored in the form of nitrates, occurs along with the nitrite route at low temperatures. Thus, the low reactivity of the C3H8 is unable to keep up with ongoing NO x storage. Lietti et al. [34] investigated the nitrite route over a model Pt/Ba/Al2O3 catalyst by operando FTIR spectroscopy. They found that the nitrite route is dominant at low temperature (150 °C) and proposed that nitrate species are generated from the oxidation of surface nitrites. Figure 8c, d shows the temporal DRIFTS spectra in the 1400 to 1200 cm−1 spectral range under the same conditions as in Fig. 8a, b but at a higher feed temperature (300 °C). The two major peaks at 1250 and 1270 cm−1 decrease after exposure to HCs, which indicates that at higher temperatures, both reductants (C3H6 and C3H8) participate in the stored NO x reduction.

Figure 9 compares the temporal DRIFTS spectra in the 1700–1500 cm−1 range for the same experimental conditions as in Fig. 8. After exposure to 500 ppm NO and 5% O2 for 1 h, three major peaks are observed: 1540, 1560, and 1580 cm−1. These are assigned to cerium nitrate (Ce(NO3)3) [31], aluminum nitrate (Al(NO3)3) [31], and bidentate nitrate (AlO(NO)O) [35], respectively. Upon introducing the HC + O2 mixture, the three peaks decrease, faster at 300 °C than at 200 °C. This trend indicates that both C3H6 and C3H8 are reactive. In the same experiments, several peaks are observed during the C3H6 exposure, corresponding to 1590, 1640, 1660, and 1680 cm−1. These peaks, which are assigned to OCO [35], C=O in formyl species [35], C=O [12], and C=N [12], are more prominent at 300 °C than at 200 °C. These species were identified in an earlier DRIFTS experiments reported by Zheng et al. [13] carried out under similar reaction conditions. Toyota researchers have also confirmed the formation of gas phase intermediates including acetaldehyde, hydroxylamine among other species, using proton transfer reaction mass spectrometry (PTR-MS) [6]. These earlier works and the current DRIFTS data suggest the formation of surface oxygenates and nitrogenates during C3H6 exposure. In contrast, the same oxygenate peaks are not detected when the pre-stored NO x catalyst is exposed to the C3H8 + O2 mixture (Fig. 9b, d). These findings suggest that the formation of active oxygenates contribute to the enhancement of NO x reduction with propylene at intermediate temperatures. While their detection does not prove their active involvement in the NO x reduction but rather as spectators, the difference in the propylene and propane DRIFTS data is noteworthy.

Fig. 9
figure 9

DRIFTS spectra taken over LNT catalyst after 1 h pre-adsorbed NO x and subsequently exposed to a flow of (a) C3H6 or (b) C3H8 + 1% O2 for 30 min at 200 °C and (c) C3H6 or (d) C3H8 + 1% O2 for 30 min at 300 °C within wavenumber range of 1700–1500 cm−1

Full size image

Figure 10 compares the temporal DRIFTS spectra within the 3100–2850 cm−1 range when the NO x pre-stored catalyst is exposed to a feed containing either 1000 ppm C3H6 (Fig. 10a) or 1000 ppm C3H8 (Fig. 10b) along with 1% O2 at T f = 200 °C (net lean). The emergence and growth of the 2960 cm−1 peak are observed during the C3H6 + O2 exposure (Fig. 10a), which is assigned to the C-H stretch [36]. In Fig. 10b, multiple peaks (2900, 2960, and 2980 cm−1) are observed during the continuous feed of C3H8 + O2. The peaks at 2900 and 2980 cm−1 are assigned to C-H stretches in a formate (HCOO−) group [37] and C=CH group [36], respectively, suggesting that adsorbed C3H8 undergoes a dehydrogenation process to olefinic species. To confirm the dehydrogenation process, in Fig. 11, 1000 ppm C3H6 was fed to a clean LNT catalyst to monitor the surface peak evolution. Four peaks in the 3050–2850 cm−1 range evolved during C3H6 exposure. The evolution of peaks at 2980 and 2900 cm−1 demonstrate C3H8 dehydrogenation over LNT catalyst. As far back as 1940, Haensel [38] reported that Pt-based catalyst is active for alkane dehydrogenation. James et al. [39] demonstrated that a Pt-based catalyst promotes alkane dehydrogenation as a result of its proclivity to catalyze C-H bond scission with limited C-C bond scission. The slow C3H8 dehydrogenation process may explain the low activity of C3H8 and inferior NO x reduction performance at intermediate temperatures.

Fig. 10
figure 10

DRIFTS spectra within wave number range of 3100–2850 cm−1 over LNT catalyst taken at 200 °C after 1 h pre-adsorbed NO x and subsequently exposed to a flow of (a) C3H6 or (b) C3H8 + 1% O2 for 30 min

Full size image
Fig. 11
figure 11

DRIFTS spectra within wave number range of 3050–2850 cm−1 over LNT catalyst taken at 200 °C after exposed to a flow of C3H6 for 20 min

Full size image

The DRIFTS measurements reveal complex chemistry during exposure of the NSR catalyst to hydrocarbons. While the HC exposure time in the DRIFTS experiments is more protracted than in the actual cycling process, the surface species data suggest the formation and accumulation of adsorbed intermediate oxygenates and nitrogenates. As stated previously, the DRIFTS data do not prove the involvement of the detected species, but the contrasting behavior of propylene and propane suggest a contributing role of surface intermediates. That said, a working mechanism for fast switching over the LNT catalyst by using different HCs is proposed to explain the opposite experimental trends using different HCs during the cycling experiments. Figure 12 is a proposed schematic of the cyclic process which builds on earlier mechanisms proposed by our group [12, 13]. Previous works have applied high-frequency propylene addition into a feed containing NO + O2 over a LNT catalyst, forming hydrocarbon intermediates (surface oxygenates and nitrogenates) through propylene partial oxidation. These intermediates may react with stored NO x or accumulate and store over the catalyst to further reduce NO x during the early lean phase. As the surface intermediates deplete during the late lean phase, NO is oxidized to NO2 and stored over barium sites on the catalyst. When using C3H8 as the reductant, dehydrogenation must occur before the oxidation chemistry occurs. At intermediate temperatures, the NO x reduction process is limited by C3H8 dehydrogenation. Therefore, longer cycle time enables more time for propane activation and dehydrogenation at lower temperatures.

Fig. 12
figure 12

Working mechanism for fast injection over LNT catalyst by using C3H6 or C3H8 as reductant

Full size image

This mechanism discussion applies to the intermediate temperature range (T f = 250–325 °C) when there is a large difference in the NO x conversion when using propylene and propane. As noted earlier, the high temperature data show less impact of the reductant identity on the NO x conversion. The overriding effect of faster cycling is clearly better utilization of stored NO x , an effect that is largely independent of the reductant type. Shakya et al. [40] showed in a combined experimental and modeling study that shorter cycle time leads to improved NO x storage utilization and higher cycle-averaged conversion. Conventional NSR process has identified two types of NO x storage sites: proximal sites and distal sites [41]. Proximal sites are used to describe the barium adjacent to Pt, while the distal sites are used for the bulk barium far away from Pt. During fast switching, the shorter lean duration limits the access primarily to the proximal sites which may also reduce the NO x slip from LNT catalyst during distal sites saturation at longer lean duration. The recent modeling work of Ting et al. [42] has demonstrated that fast switching enables better utilization of the proximal sites which may enhance NO x storage efficiency. There are two additional factors that contribute to the conversion enhancement under fast cycling, high temperature conditions. The first involves adsorbed HC intermediates pathway put forward by Toyota researchers [6,7,8]. The generation of intermediates containing nitrogen persists on the surface long enough to be oxidized to N2. That propylene and propane give higher NO x conversion than H2 at elevated temperature is, although circumstantial, not direct evidence. Dasari et al. [43] showed a secondary peak during conventional NSR using CO as the reductant and proposed that this results from the oxidation of surface isocyanates in the absence of H2O. Mráček et al. [29] showed evidence for another N2 forming pathway during the lean part of the NSR cycle but the intensity of this pathway decreased with increasing temperature. This might argue against the HC mechanism. While their study considered conventional NSR timing (60/5), they suggest that the reaction pathway may contribute to NO x conversion at shorter cycle times. At this point, direct evidence for the involvement of a HC pathway has not been provided.

3.3 HC Mixture Effects

In practice, the rich feed will not contain a single reductant and in fact will be comprised of not only CO and H2 but also a mixture of hydrocarbons. It is therefore of interest to examine the NSR catalyst performance when using a feed containing a mixture of C3H8 and C3H6. Table 1 reports the compositions of the feeds which include C3H8-only (Feed #3), equimolar C3H8 + C3H6 (C3H8—50; Feed #4), 75% C3H8 (C3H8—75; Feed #5), and 87.5% C3H8 (C3H8—87.5; Feed #6). The mixtures were kept at the same cycle-averaged S N (=8.06).

Figure 13 shows the cycle-averaged NO x and C3H8 conversions and catalyst temperatures as functions of the feed temperature using different HC reductant mixtures at a fixed 6/1 switching frequency. Increasing the amounts of C3H6 in the mixture increases the NO x conversion over the entire feed temperature range. Only at the highest feed temperature (350 °C), comparable NO x conversions are obtained by all four different reductant mixtures. At low feed temperature, the addition of C3H6 has a non-linear effect. For example, Fig. 13b shows that at 275 °C, the C3H8 conversion increases from ~10 to ~50% for a feed containing 87.5% C3H8 + 12.5% C3H6. The corresponding NO x conversion increases from ~10 to ~70%. The enhancement is attributed to the exothermicity of the C3H6 oxidation. An inspection of the corresponding average catalyst temperature (Fig. 13c) indicates the importance of thermal effects. The catalyst temperature increases by ~150 °C for the equimolar feed compared to using C3H8 as the sole reductant. Thus, since C3H6 is more easily oxidized, it assists in thermally promoting the less active C3H8 which leads to the earlier light-off of C3H8. The temperature rise is sufficient to lower the mixture light-off temperature from that of the less active C3H8. The corresponding NO x and C3H8 conversions versus catalyst temperature are provided in Fig. 13d, e.

Fig. 13
figure 13

Comparison of cycle-averaged (a) NO x conversion, (b) C3H8 conversion and (c) catalyst average temperature as a function of feed temperature or (d) NO x conversion, and (e) C3H8 conversion as a function of catalyst temperature using different HC mixture as reductant [conditions: lean (6 s): 300 ppm NO, 10% O2, 3.5% H2O, and 9% CO2; rich(1 s): 300 ppm NO, 8530 ppm C3H6 and 8530 ppm C3H8 or 4150 ppm C3H6 and 1.246% C3H8 or 2050 ppm C3H6 and 1.435% C3H8, or 1.62% C3H8, 5% O2, 3.5% H2O, and 9%CO2]

Full size image

Figure 14 shows a different view of the impact of C3H6 addition. The NO x conversion, C3H8 conversion, and average catalyst temperature rise are shown as a function of the C3H6 feed fraction in the HC reductant mixture for feed temperatures in the 250–300 °C range. For T f = 250 °C, the NO x conversion increases from a negligible level to ~85% with the replacement of 50% of C3H8 with C3H6. The increase is even sharper at 275 and 300 °C. For example, at T f = 300 °C, NO x conversion increases from 25 to 80% when C3H6 mole fraction in the HC mixture is increased from 0 to 12.5%. When the C3H6 concentration exceeds 50%, the NO x conversion approaches 85% for all three feed temperatures. Figure 14b shows the C3H8 conversion as a function of C3H6 concentration in the HC mixture at different feed temperatures. The C3H8 conversion increases to 75% at T f = 300 °C, which is ~50% higher than using pure C3H8 reductant. With 50% C3H6 in the HC mixture, NO x reduction performance is almost identical to that using pure C3H6 as reductant. Therefore, enhancement of NO x and C3H8 conversion is contributed by both chemical and thermal effects.

Fig. 14
figure 14

Comparison of cycle-averaged (a) NO x conversion, (b) C3H8 conversion, and (c) catalyst average temperature as a function of C3H6 concentration in HC reductant at T f = 250, 275, and 300 °C [conditions: lean(6 s): 300 ppm NO, 10% O2, 3.5% H2O, and 9% CO2; rich(1 s): 300 ppm NO, 1.8% C3H6 or 8530 ppm C3H6 and 8530 ppm C3H8 or 4150 ppm C3H6 and 1.246% C3H8 or 2050 ppm C3H6 and 1.435% C3H8, or 1.62% C3H8, 5% O2, 3.5% H2O, and 9% CO2]

Full size image

Notwithstanding the impressive NO x conversion enhancement achieved with fast cycling, previous studies have shown that undesired by-product N2O yield may increase as well [11, 12]. To examine this potential drawback of the Di-Air process, Fig. 15 shows the impact of the cycle timing on the N2O yield versus feed temperature for several different HC feeds. Figure 15 compares N2O yield as a function of feed temperature for three different scenarios. In general, the highest N2O yields are obtained during fast cycling at lower feed temperatures and with HC feed mixtures having higher C3H6 fractions. For example, Fig. 15a shows a ~30% N2O yield at T f = 225 °C for an equimolar feed. This is the highest N2O yield among all the feeds. In contrast, when using C3H8 as the sole reductant, the N2O yield shows a maximum of only 8% at the higher feed temperature of 300 °C. Figure 15b shows for a feed containing 12.5% C3H6, the N2O yield is less sensitive to changes in the cycle time. The data in Fig. 15a, b suggest that a higher C3H6 concentration in the HC mixture promotes the N2O formation. Finally, Fig. 15c compares the N2O yields for different reductant mixture compositions for the shortest cycle (6/1). It is clear that an increasing feed fraction of C3H6 leads to higher N2O yield. Moreover, the maximum N2O yield for each feed occurs at higher feed temperature as the C3H6 fraction decreases. The corresponding N2O yield versus catalyst temperature is provided in Fig. 15d–f. Collectively, N2O formation is a complex function of the feed gas composition and cycle timing. A thorough understanding of the effects will require combined experiments and modeling.

Fig. 15
figure 15

Comparison of cycle-averaged N2O yield as a function of feed temperature by varying lean-rich switching frequency using reductant (a) 50% C3H6 and (b) 12.5% C3H8 in HC mixture; (c) comparison of N2O yield as a function of feed temperature at 6–1 s by varying HC mixture. Comparison of cycle-averaged N2O yield as a function of catalyst temperature by varying lean-rich switching frequency using reductant (d) 50% C3H6 and (e) 12.5% C3H8 in HC mixture; (f) comparison of N2O yield as a function of catalyst temperature at 6–1 s by varying HC mixture [conditions: lean(6 s): 300 ppm NO, 10% O2, 3.5% H2O, and 9% CO2; rich(1 s): 300 ppm NO, 1.8% C3H6 or 8530 ppm C3H6 and 8530 ppm C3H8 or 4150 ppm C3H6 and 1.246% C3H8 or 2050 ppm C3H6 and 1.435% C3H8, or 1.62% C3H8, 5% O2, 3.5% H2O, and 9%CO2]

Full size image

4 Conclusions

The effects of three reductants (H2, C3H6, C3H8) on lean NO x reduction using rapid lean-rich switching frequency were systematically evaluated over a model LNT catalyst.

  • At intermediate feed temperatures (250–325 °C), the NO x conversion increases when the switching frequency is reduced using alkane reductant (C3H8). This trend is opposite to that obtained with the olefin (C3H6). The opposing trend is attributed to the positive reaction order with respect to C3H8 and the slower dehydrogenation process. The absence of DRIFTS detected surface intermediates when using propane underscore the differences between the olefin and alkane.

  • High switching frequency significantly increases NO x conversion over LNT catalyst at elevated temperatures (T c >400 °C) in which the impact of reductant is less. Up to a ~20% (absolute) NO x conversion enhancement is achieved by shortening the cycle time by a factor of 10, from 70 to 7 s at fixed rich phase duty fraction. The enhancement is contributed to the more frequent regeneration of NO x storage sites, exothermic heat effects of the reductant oxidation, and the generation and conversion of active hydrocarbon intermediates.

  • The enhancement of NO x and C3H8 conversions when C3H6 is introduced into HC mixture can be attributed to chemical effects, cycle time, thermal effects, intermediates formation, improved NO x storage utilization, and catalyst temperature rise contributed by active C3H6 oxidation in the HC mixture.

Research is ongoing at the University of Houston to determine the underlying mechanism for NO x enhancement by fast cycling. The reactivity and DRIFTS spectrum differences between propylene and propane show evidence for enhancement, at least in part, by chemically based enhancement at intermediate temperatures. On the other hand, the strong exotherms associated with the oxidation chemistry and the apparent convergence in NO x conversion at higher temperature suggest that transport and NO x storage site utilization effects are also important.