New insights into the NH3-selective catalytic reduction of NO over Cu-ZSM-5 as revealed by operando spectroscopy

To control diesel vehicle NOx emissions, Cu-exchanged zeolites have been applied in the selective catalytic reduction (SCR) of NO using NH3 as reductant. However, the harsh hydrothermal environment of tailpipe conditions causes irreversible catalyst deactivation. The aggregation of isolated Cu2+ brings about unselective ammonia oxidation along with the main NH3-SCR reaction. An unusual ‘dip’ shaped NO conversion curve was observed in the steamed zeolite Cu-ZSM-5, resulting from the undesired NH3 oxidation that produced NO. Here we gain further insights into the NH3-SCR reaction and its deactivation by employing operando UV-vis diffuse reflectance spectroscopy (DRS) and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) on fresh and steamed zeolite Cu-ZSM-5. We found that tetragonally distorted octahedral Cu2+ with associated NH3 preferentially forms during low temperature NH3-SCR (<250 °C) in fresh Cu-ZSM-5. The high coordination number of Cu2+ ensures the availability for high coverage of nitrate intermediates. Whilst in the steamed Cu-ZSM-5, [Cux(OH)2x−1]+ oligomers/clusters in pseudo-tetrahedral symmetry with coordinated NH3 accumulated during the low-temperature NH3-SCR reaction. These clusters presented a strong adsorption of surface NH3 and nitrates/nitric acid at low temperatures and therefore limited the reaction between surface species in the steamed Cu-ZSM-5. Further release of NH3 with increased reaction temperature favors NH3 oxidation that causes the drop of NO conversion at ∼275 °C. Moreover, competitive adsorption of NH3 and nitrates/nitric acid occurs on shared Lewis-acidic adsorption sites. Prompt removal of surface nitrates/nitric acid by NO avoids the surface blockage and tunes the selectivity by alternating nitrate–nitrite equilibrium. The formation of adsorbed NO2 and HNOx points to the necessity of an acid adsorbent in practical applications. The structural similarity under the NH3-SCR reaction and unselective NH3 oxidation confirmed the entanglement of these two reactions above 250 °C.

Ammonia Temperature-Programmed Desorption (NH 3 -TPD) was taken on Micromeritics Autochem II 2920 equipped with a Thermal Conductivity Detector (TCD). Typically, 60 mg fresh/steamed zeolite Cu-ZSM-5 was first dehydrated in He flow for 1 h at 600 °C with a heating ramp of 10 °C/min. Ammonia then flowed pass the sample at 100 °C, followed by flushing with He at 100 °C for 2 h. The TCD signal of ammonia desorption was recorded starting from 150 °C under a temperature ramping rate of 5°C min -1 in a 25 mL min -1 He flow and staying at 600 °C for 30 min.
Hydrogen Temperature-Programmed Reduction (H 2 -TPR) was performed on Micromeritics Autochem II 2920 equipped with a TCD. Typically, 100 mg fresh/steamed zeolite Cu-ZSM-5 was reduced in 5 vol% H 2 /Ar with the heating ramp of 10 °C/min. The TPR profile was recorded from 100 to 900 °C.
CO adsorbed Fourier-Transform Infrared (CO-FT-IR) spectroscopy was carried out on a Perkin Elmer 2000 FT-IR spectrometer equipped with a homemade transmission cell with CaF 2 window. A 10 mg sample was used to make the selfsupported pellet for both catalysts. For each experiment, the self-supported pellet was dehydrated at 300 C for 2 h under vacuum of 10 -6 mbar, followed by cooling down with liquid nitrogen. The reference spectrum was taken before dosing CO. The 10% CO/He (Linde, 99.998 %) was dosed in gradience to attain different surface coverage of CO. Spectra were recorded after CO dosage each time in a resolution of 4 cm -1 with 16 replications.

Catalyst Testing
Catalyst testing was performed in a fixed bed plug flow setup. All the required gases were provided by Linde and the flow rates of feed gases were controlled by Brooks flow meters. Typically, 50 mg sieved catalysts powders (0.125-0.425 mm) were closely packed in a quartz reactor. Prior to the NH 3 -SCR reaction, NH 3 oxidation or NO oxidation reaction, the packed zeolite was activated in 5 % O 2 /He at 550 °C for 1 h and was cooled to 150 °C to start the reaction.
For the standard NH 3 -SCR reaction, the catalyst was exposed to SCR feed composition of 1000 ppm NO, 1000 ppm NH 3 and 5% O 2 balanced by He with a Gas Hourly Space Velocity (GHSV) of 100,000 h -1 . The reaction was conducted in a desired temperature from 150 to 450 °C and was stabilized for 1 hour at each target temperature. The gas composition from outlet was determined by a FT-IR gas analyzer (Perkin-Elmer, Spectrum Two) that the real-time concentration of reactants (i.e., NO and NH 3 ) and products (NO 2 and N 2 O) could be recorded. After the reaction reach the steady state, the average concentration of outlet gas composition was used to determine conversion and yield at each reaction temperature. The production of N 2 was calculated based on N balance assuming that only gaseous reactants/products (NO 2 , NO, N 2 O, N 2 and NH 3 ) were formed in the reaction.
For a better understanding of catalytic behavior of NH 3 -SCR from the sense of side reactions, ammonia oxidation reaction and nitric oxide reaction were also performed over zeolite Cu-ZSM-5-fresh and Cu-ZSM-5-850stm. The reaction processes and the conversion calculations of NH 3 oxidation as well as NO oxidation were the same as the NH 3 -SCR reaction except for the compositions of gas feed. The gas feed composition was 1000 ppm NH 3 and 5% O 2 balanced by He for the NH 3 oxidation, and 1000 ppm NO and 5% O 2 balanced by He for the NO oxidation with a Gas Hourly Space Velocity (GHSV) of 100,000 h -1 .

Operando Spectroscopies
Operando UV-Vis Diffuse Reflectance Spectroscopy (DRS) was also performed during the catalyst testing and achieved by the specially designed quartz fixed-bed reactor with a UV-Vis transparent window. A high-temperature UV-vis optical fiber probe connected to an AvaSpec 2048L spectrometer was employed to collect the UV-Vis DRS spectra every 2 min.
Operando Diffuse Reflectance Infrared Fourier transform Spectroscopy (DRIFTS) was performed making use of 10-20 mg catalyst, which was loaded in a sample holder in a high temperature Harrick reaction chamber with ZnSe windows and was pressed into a pellet with flat surface. The reactant gases were flowed through the pellet from bottom to top. Samples were calcined in 5 % O 2 /He at 550 °C before the reaction. The gas feed composition and reaction procedure were the same as the catalytic test on a fix bed reactor when performing the NH 3 -SCR reaction. NO or NH 3 cutoff experiment was also performed at 250 °C. The DRIFTS experiment protocol is briefly illustrated in Scheme S1. The DRIFTS data were collected continuously by a Bruker Tensor II spectrometer with a MCT detector with 32 accumulation times and resolution of 4 cm -1 . The effluent gas composition was determined by a FT-IR gas analyzer (Perkin-Elmer, Spectrum Two).
Scheme S1. Procedure of operando Diffuse Reflectance Infrard Fourier Transform Spectrsocopy (DRIFTS) experiments at 250 °C, with 1000 ppm of NH 3 and/or 1000 ppm NO balanced by 5% O 2 /He.Gases flowed in the sequence of NH 3 +O 2 , NH 3 +NO+O 2 and NO+O 2 .       The LMCT band usually presents as a broad and intense band as the charge transfer is fully allowed between metal and ligand. 7 The LMCT band lying above 30000 cm -1 in the hydrated Cu-ZSM-5 originated from a O 2to Cu 2+ charge transfer, which superimposes on the fundamental absorption edge of the zeolite matrix. 8,9 The LMCT band of ammoniated Cu 2+ is located at a higher wavenumber of around 42000 cm -1 along with a shoulder at around 32000 cm -1 , suggesting the existence of one or more H 2 O ligands in the ammoniated Cu 2+ complex.    The d-d transition band of Cu 2+ was fitted using Gaussian functions. By inspecting the evolution of the d-d transition band under reaction condition via Principal Component Analysis (PCA), the d-d transition band was fitted using three Gaussian peaks, whose peak positions were restrained to the regions 10200-10800 cm -1 , 13000-14000 cm -1 , and 16000-18000 cm -1 . In addition, the peak width was confined as well based on the initial fitting result of a reference spectrum (UV-Vis diffuse reflectance spectrum of fresh zeolite Cu-ZSM-5 in He at 150 °C). All spectra were fitted using the same fitting parameters. The R square of fit reached the level of 0.9999 for every fitted spectrum. The fitting curves and individual fitted components of fresh and 850 °C steamed zeolite Cu-ZSM-5 are respectively shown in Figures S13-14 to indicate the reliability of the fitting.    The corresponding operando UV-Vis diffuse reflectance spectra of the c) fresh and d) 850 °C steamed zeolites Cu-ZSM-5 were also collected in the reaction, and cycle 2 represent the reversed cycle. The comparable catalytic performance and spectroscopic observations in the two NH 3 -SCR cycles strongly suggest that the evolution of species observed during the first cycle was reversible, and not due to irreversible deactivation and degradation of Cu species (which may nonetheless occur over multiple catalytic cycles or prolonged testing). The stretching vibration of O-H in the hydroxyl group in a Cu-exchanged zeolite is well documented. 10 The absorption band at 3738 cm -1 , 3659 cm -1 as well as 3602 cm -1 is assigned to O-H vibration in external silanol, [CuOH] + and Brønsted acid respectively. The stronger acidity presents a lower frequency due to a more ready-to-donate proton and thus a longer O-H bond. The N-H bond of adsorbed NH 3 has a lower stretching vibrational frequency than that of a hydroxyl group in zeolites. NH 3 is chemisorbed on the catalyst surface via the lone electron pair from the N side of NH 3 . A few N-H bands are observed in 3400-3100 cm -1 region, which are typical N-H asymmetric and symmetric stretching ν s (NH 3 ) modes as well as the splitting of ν s (NH 3 ) due to a Fermi resonance with the overtone of asymmetric NH 3 deformation. 11,12 It is noted that upon the consumption of the NH 3 bands the [CuOH] + developed first, followed by the depletion of Brønsted acid in the NH 3 oxidation. The interaction of NH 3 with the Brønsted acid site generates NH 4 + that could be tracked by the N-H bending vibration at ~1460 cm -1 . Finally, the broad feature at around 3000 cm -1 indicates the formation of H-bonded NH 3 with a surface hydroxyl group.

Operando Diffuse Reflectance Infrared Fourier Transform Spectroscopy Experiments
When the NH 3 -SCR gases were fed, Cu cations solvated with NH 3 to form the stable ammoniated-Cu 2+ , which is mobile and allowed to free the Cu ions from their charge balanced positions. The elimination of this Cu-framework interaction allowed the framework T-O-T network to relax to the original vibration. Therefore, the perturbed framework vibrational band vanished along with the NH 3 adsorption ( Figure S17-20). By exposure to SCR gas composite, the Cu moieties interacted with reactants such as O 2 , or the intermediates such as NO x -, which caused the slight shift of the bands.     The experimental procedure can be found in Scheme S1. The colors indicate spectral changes from blue to yellow following the reaction time in each subfigure.