Mechanistic insights into a NOx storage-reduction (NSR) catalyst by spatiotemporal operando X-ray absorption spectroscopy†

Lean-burn combustion engines, using fuels such as gasoline and diesel, provide greater fuel economy compared to conventional stoichiometric gasoline engines. However, the three-way catalysts used for conventional gasoline engines cannot purify NOx in the oxygen-rich exhaust emitted from lean-burn engines. Therefore, new catalytic approaches are needed to treat the NOx emissions from diesel and gasoline lean-burn engines. This situation has encouraged the development of new catalyst technology that can reduce NOx in excess oxygen, i.e., a NOx storage-reduction (NSR) catalyst for gasoline lean-burn engines, also known as a lean NOx trap (LNT). The NSR catalyst has been used practically in Japan in 1994. However, vehicle emission standards have been gradually strengthened on a global scale, leading to a demand for further improvements in NSR performance. Therefore, studies of lean NOx catalysts remain a challenging research task. 4,5 Basically, NSR catalysts are precious group metals (PGMs; Pt, Rh and Pd) together with supporting oxides such as Al2O3 and barium oxide/carbonate as a NOx storage material, that can promote catalytic oxidation and reduction processes (Fig. S1 in the ESI†). During the fuel-lean phase (oxygen excess), NOx is oxidized over the precious metal, and then stored as barium nitrate by reaction with the barium storage component. However, during the short fuel-rich phase (reductive atmosphere) of catalyst regeneration, the stored NOx is released and subsequently reduced to N2 over the precious metal. Depending on the driving conditions, a burst of NOx emission, often called an NOx spike or NOx puff, can be observed at the outlet of the catalyst bed during a rich period, because a portion of the stored NOx is not reduced to N2, leading to a NOx slip at the outlet. Several studies have been conducted on the NOx reduction mechanism, and significant progress has been made. However, the nature of NOx spike generation during the fast transient reaction remains very complicated, and requires further investigation. To elucidate the NOx spike generation mechanism, understanding the spatially resolved reactions occurring in the catalyst bed is necessary, because chemical species and catalytic active states have drastic gradients along the axial direction of the catalyst bed. Several research groups have conducted spatially resolved analysis of the NSR process using infrared (IR)/Raman spectroscopy that provides surface and bulk information about the Ba species, spatially resolved capillary-inlet mass spectrometry (SpaciMS) to monitor the composition of the gas phase, and IR thermography to obtain temperature measurements. Spatial analysis of the PGM active site is vitally important to understand the NSR process in the entire catalyst bed from front to rear. However, no studies have ever been conducted to track the PGM active state in the axial direction of the NSR catalyst. This is because the oxidation–reduction rate of the PGM, which is related to the catalytically active state, is very fast under the lean/rich perturbations at temperatures of about 200 °C and above, therefore, fast operando analysis methodology with millisecond temporal resolution is required to track the PGM active state. Therefore, a spatiotemporal operando X-ray absorption spectroscopy (XAS) method was applied to obtain spatially resolved monitoring of PGM in the axial direction as well as for the measurement of online outlet gas components to clarify the NOx storage and reduction processes, especially the NOx spike mechanism. A Toyota beamline (BL33XU) of SPring-8, combining a servo-motor-driven Si channel-cut

Lean-burn combustion engines, using fuels such as gasoline and diesel, provide greater fuel economy compared to conventional stoichiometric gasoline engines. However, the three-way catalysts used for conventional gasoline engines cannot purify NO x in the oxygen-rich exhaust emitted from lean-burn engines. Therefore, new catalytic approaches are needed to treat the NO x emissions from diesel and gasoline lean-burn engines. 1 This situation has encouraged the development of new catalyst technology that can reduce NO x in excess oxygen, i.e., a NO x storage-reduction (NSR) catalyst for gasoline lean-burn engines, 2 also known as a lean NO x trap (LNT). The NSR catalyst has been used practically in Japan in 1994. 3 However, vehicle emission standards have been gradually strengthened on a global scale, leading to a demand for further improvements in NSR performance. Therefore, studies of lean NO x catalysts remain a challenging research task. 4,5 Basically, NSR catalysts are precious group metals (PGMs; Pt, Rh and Pd) together with supporting oxides such as Al 2 O 3 and barium oxide/carbonate as a NO x storage material, that can promote catalytic oxidation and reduction processes (Fig. S1 in the ESI †). During the fuel-lean phase (oxygen excess), NO x is oxidized over the precious metal, and then stored as barium nitrate by reaction with the barium storage component. However, during the short fuel-rich phase (reductive atmosphere) of catalyst regeneration, the stored NO x is released and subsequently reduced to N 2 over the precious metal.
Depending on the driving conditions, a burst of NO x emission, often called an NO x spike or NO x puff, can be observed at the outlet of the catalyst bed during a rich period, because a portion of the stored NO x is not reduced to N 2 , leading to a NO x slip at the outlet. Several studies have been conducted on the NO x reduction mechanism, and significant progress has been made. [6][7][8][9][10][11][12][13][14] However, the nature of NO x spike generation during the fast transient reaction remains very complicated, and requires further investigation.
To elucidate the NO x spike generation mechanism, understanding the spatially resolved reactions occurring in the catalyst bed is necessary, because chemical species and catalytic active states have drastic gradients along the axial direction of the catalyst bed. Several research groups have conducted spatially resolved analysis of the NSR process using infrared (IR)/Raman spectroscopy that provides surface and bulk information about the Ba species, 6-9 spatially resolved capillary-inlet mass spectrometry (SpaciMS) to monitor the composition of the gas phase, [10][11][12][13] and IR thermography to obtain temperature measurements. 14 Spatial analysis of the PGM active site is vitally important to understand the NSR process in the entire catalyst bed from front to rear. However, no studies have ever been conducted to track the PGM active state in the axial direction of the NSR catalyst. This is because the oxidation-reduction rate of the PGM, which is related to the catalytically active state, is very fast under the lean/rich perturbations at temperatures of about 200°C and above, 15,16 therefore, fast operando analysis methodology with millisecond temporal resolution is required to track the PGM active state.
Therefore, a spatiotemporal operando X-ray absorption spectroscopy (XAS) method was applied to obtain spatially resolved monitoring of PGM in the axial direction as well as for the measurement of online outlet gas components to clarify the NO x storage and reduction processes, especially the NO x spike mechanism. A Toyota beamline (BL33XU) of SPring-8, combining a servo-motor-driven Si channel-cut monochromator with a tapered undulator, allowed rapid acquisition of high-quality data for quick scan XAS (QXAS), 17 along with an operando setup simulating automotive engine exhaust (Fig. S2a †). This study presents the importance of operando spatiotemporally resolved QXAS analysis for heterogeneous catalytic processes, and reveals the nature of the NO x storage and reduction process. In this study, a 0.5 wt% Rh/BaO (10 wt% as Ba)/γ-Al 2 O 3 catalyst was prepared. A capillary tube with a pellet catalyst sample (66 mg, sieve fraction 75-150 μm, ca. 8 mm in catalyst bed length) was set in a specially designed operando cell ( Fig. 1a and b). The NSR reaction was operated at 450°C, and a lean stream consisting of 0.07% NO, 7% O 2 , and He balance was introduced into the cell for 300 or 60 s, followed by a rich stream of 3% H 2 and He balance for 240 s. The gas flowing over the sample was quickly changed from a lean to a rich atmosphere using a gas-actuated switching valve (Fig.  S2b †). Measurement of inlet/outlet NO concentration confirmed that the amount of stored NO x during the lean 60 s supply corresponded to 0.050 mmol per g of catalyst (33% of the saturated NO x storage amount) and that during the lean 300 s supply, the amount corresponded to 0.151 mmol per g of catalyst (100% NO x storage amount). The experiments of 300 s and 60 s of lean supply are referred to as 100% and 33% NO x storage, respectively. Under the NSR operation, online mass spectra of the outlet gas species were obtained every 150 ms, along with XAS measurements.
To track the catalytically active state, Rh K-edge X-ray absorption near-edge structure (XANES) spectra were collected every 100 ms at four positions: 1 (A), 3 (B), 5 (C), and 7 (D) millimeters from the bed front ( Fig. 1a). High-quality data from the Rh K-edge XANES spectra were acquired using QXAS in transmission mode ( Fig. S2c †), which enabled tracking of temporal changes in the Rh oxidation state as Rh metal fraction. Fig. 1c and d show the Rh metal fraction at each position (A-D) after switching from lean gas to rich gas at 450°C.
Under the lean conditions, all Rh species in the catalyst bed were Rh oxide (Rh 2 O 3 ). After switching to the rich condition, for both 100% and 33% NO x storage, reduction in Rh occurred sequentially from the front A position to the rear D position. Thus, Rh at the downstream position (e.g., position B) was not reduced unless the reduction in Rh upstream (e.g., position A) was completed. The metallic Rh state is considered an active state for catalytic NO reduction [18][19][20] because Rh metallic clusters are required for the NO dissociation step. This indicates that NO reduction with H 2 over the Rh species occurred sequentially from upstream to downstream.
In Fig. 1c, "t1" and "t2" denote the onset time and the termination time, respectively, of Rh reduction, and the value obtained by subtracting t1 from t2 is defined as the "Rh reduction time" (calculation method in Fig. S3 †). Fig. 2a Fig. S5, † the greater the amount of stored NO x , the longer the Rh reduction time because the stored NO x prolongs the reduction of Rh oxide. Thus, the Rh reduction time during the rich period was proportional to the amount of NO x stored in the Ba compounds in the lean period. Results shown in Fig. 2a clearly indicate that NO x is fully stored  along the entire length of the catalyst bed after 300 s of lean supply (100% NO x storage), whereas for 33% NO x storage, NO x is stored only upstream at the A and B positions. Fig. 2b shows the evolution profiles of outlet NO, N 2 , N 2 O, and H 2 gases when the gas flow was switched from lean to rich conditions at 100% NO x storage. NH 3 production was not observed within the measurement time in Fig. 2b. After switching to the rich period of H 2 /He flow, a spike in NO emission was observed along with N 2 generation from the reaction of H 2 with the stored NO x , indicating that the stored NO x decomposed during the rich period to gas phase NO x and was subsequently released to the outlet of the catalyst bed without being reduced to N 2 . Such NO emission lasted for about 11 s until unreacted H 2 was evolved.
In addition to N 2 as a NO x reduction product, a trace amount of N 2 O was confirmed at around 3 s. Kubiak et al. investigated mechanistic aspects in the formation of N 2 O over Pt/BaO/Al 2 O 3 and Rh/BaO/Al 2 O 3 by transient microreactor experiments and operando IR spectroscopy. 21 They concluded that N 2 O formation involves the coupling of gaseous NO molecules with N-adspecies formed upon NO dissociation onto PGM sites. As for 100% NO x storage in our experiment, it is reasonable to consider that the concentration of gas phase NO became the highest at around 3 s, and trace N 2 O was generated accordingly.
The gas atmosphere along the axial direction inside the catalyst bed during the rich period was investigated. Since the reduction reaction rate of stored NO x is sufficiently faster than the gas flow velocity, NO reduction at the Rh metal sites proceeded sequentially from the upstream position with complete consumption of H 2 . Actually, at the elapsed time of 1 s (Fig. 1c), the Rh species at position A is reduced, but the Rh species in the region from position B to D remain as Rh oxide. Considering that the time for gas to flow through the catalyst bed was 60 ms for this experiment, the downstream region from B to D, after H 2 was completely consumed at the upstream of position A, is basically under an inert gas atmosphere (He plus N 2 ). A schematic for this process is shown in Fig. S6. † Luo et al. 11 and Choi et al. 13 reported that the stored NO x during a lean period is released to the outlet of the catalyst bed as gaseous NO x during an inert gas purge at temperatures of 400°C and above. In addition, Luo et al. 11 reported that an oxidizing atmosphere (oxygen excess) stabilizes barium nitrates, whereas they become less stable under an inert gas atmosphere, resulting in the formation of a NO x puff. Urakawa et al. 6 investigated the dynamic surface and bulk processes of Ba components during NSR reactions using spatiotemporal IR/Raman spectroscopy. Their results suggested that the surface Ba nitrite species decomposed during the rich period, releasing NO into the gas phase. According to their investigation, a portion of stored NO x in the region from B to D in this study became unstable under the inert gas atmosphere and was released to the outlet as a NO x spike emission.
All of these results led to the following conclusions about the NSR process for 100% NO x storage (Fig. 3a): Step 1: After the lean period of 300 s, NO x is fully stored along the entire length of the catalyst bed as barium nitrite or barium nitrate [BaĲNO x ) 2 )].
Step 2: During the first second of the rich period, the stored NO x at the front position near A is reduced to N 2 by H 2 over the metallic Rh species, and N 2 is released to the outside of the catalyst bed. At the same time, a portion of the stored NO x at the back position of B-D (e.g., the surface nitrites which are weakly adsorbed) is decomposed under the inert gas, and released into the gas phase. Therefore, NO spike emission and N 2 generation begin to occur almost simultaneously. Besides, a trace amount of N 2 O is generated in the presence of gas phase NO at around 3 s.
Step 3: Step 2 continues for ∼11 s until H 2 reaches the end of the catalyst bed. At this point, all of the stored NO x is reduced by H 2 , which is consistent with the termination of Rh reduction at position D (Fig. 1c).
Next, the 33% NO x storage case was examined by obtaining outlet gas profiles (Fig. 2c). The amount of N 2 generation from 33% NO x storage (0.024 mmol g −1 ) was less than that for the 100% NO x storage case (0.059 mmol g −1 ), by a value that corresponds to the difference between the two NO x storage amounts (0.050 and 0.151 mmol g −1 ). H 2 emission was observed at approximately 5 s into the rich period, which is consistent with the time when reduction of Rh species at position D was completed (Fig. 1d). Note the delay of 2.5 s for NO emission after N 2 generation in the 33% NO x storage case. A similar delay was not observed for 100% NO x storage.
The delay of the NO spike was investigated (Fig. S7 †). According to the Rh reduction behavior in Fig. 1d during the first second of the rich period, a H 2 atmosphere exists upstream of position A; in contrast, the atmosphere downstream from B to D contains an inert gas. The NO x stored near position B under the inert gas is decomposed to the gas phase, resulting in desorbed NO x flowing downstream relatively slowly because the gas phase NO x is weakly trapped in the downstream C and D regions at vacant sites on the BaO surface. This slow travel via repetition of desorption/re-trapping results in a 2.5 s delay based on the onset time of N 2 emission.
Incidentally, the ratio of the NO x spike emission amount to the amount of stored NO x for 33% NO x storage, calculated from the evolution profiles of the outlet gases (Fig. 2c), was 4%, which is significantly less than the 21% found for 100% NO x storage (Fig. 2b). This difference is interpreted as follows. Since no re-trap sites are available downstream in the 100% NO x storage case, the decomposed NO x exits unreduced from the catalyst bed. In contrast, for the 33% NO x storage case, some of the downstream re-trapped NO x species are reduced to N 2 at the Rh sites by H 2 coming from upstream, and the travelling NO x species that the H 2 does not reach are released to the outlet.
Thus, the NSR process for 33% NO x storage is interpreted as follows (Fig. 3b): Step 1: After the 60 s lean period, NO x is stored as BaĲNO x ) 2 upstream of the A and B positions. Downstream, from C to D, vacant sites (BaO) exist for NO x trapping.
Step 2: During the first second of the rich period, the NO x stored at the front position near A is reduced to N 2 by H 2 over the metallic Rh species, and released to the outside. At the same time, a portion of the NO x stored toward the back of B is decomposed under the inert gas and released into the gas phase. The decomposed NO x species are re-trapped downstream in the C and D regions.
Step 3: At 2.5 s into the rich period, the reduction front reaches position B. The decomposed NO x species travel farther downstream, repeating the desorption/re-trap process on the vacant sites. The traveling NO x is released to the outlet as an NO spike unless H 2 gas reaches the traveling NO x .
Step 4: Step 3 continues until the H 2 reaches the D position, and then H 2 emission is observed at approximately 5 s into the rich period.

Conclusions
A spatiotemporal operando XAS technique was used to investigate the dynamic behavior of the NO x storage and reduction process of a Rh/BaO/Al 2 O 3 catalyst. The Rh K-edges XAS in the axial direction of the catalyst bed were monitored every 100 ms, which was combined with online mass spectrometry. Using this methodology, the fast transient phenomena occurring inside the NSR catalyst were tracked and the NO x storage distribution during the lean period was clarified, along with the cause of the NO x spike emission.
In this study, a simplified model gas containing NO + O 2 and H 2 was used, in order to gain mechanistic insights into the principal NSR process. However, an actual engine exhaust includes H 2 O, CO 2 , CO and so on. Using the spatiotemporal operando XAS methodology, we plan to investigate the effect of coexisting gases and reductants on the catalytic active site. Furthermore, combining SpaciMS with the current XAS set-up can give a deeper understanding of the overall mechanism occurring in the NSR convertor. We hope to establish an ultimate spatiotemporal operando system in the near future.

Conflicts of interest
There are no conflicts to declare.