A Cu–Pd single-atom alloy catalyst for highly efficient NO reduction

A series of Cu–Pd alloy nanoparticles supported on Al2O3 were prepared and tested as catalysts for deNOx reactions. XRD, HAADF-STEM, XAFS, and FT-IR analyses revealed that a single-atom alloy structure was formed when the Cu/Pd ratio was 5, where Pd atoms were well isolated by Cu atoms. Compared with Pd/Al2O3, Cu5Pd/Al2O3 exhibited outstanding catalytic activity and N2 selectivity in the reduction of NO by CO: for the first time, the complete conversion of NO to N2 was achieved even at 175 °C, with long-term stability for at least 30 h. High catalytic performance was also obtained in the presence of O2 and C3H6 (model exhaust gas), where a 90% decrease in Pd use was achieved with minimum evolution of N2O. Kinetic and DFT studies demonstrated that N–O bond breaking of the (NO)2 dimer was the rate-determining step and was kinetically promoted by the isolated Pd.


Introduction
The reactions of nitric oxide (NO) have garnered intense interest from researchers in the human health, 1 and bioinorganic, 2 industrial, 3 and environmental chemistry elds. 4 Specically, NO removal has long been studied as an indispensable process for exhaust-gas purication. 5 Platinum-group metals (PGMs) such as Pt, Pd, and Rh are known to be efficient catalysts for the reduction of NO using CO, 6,7 H 2 , 8 NH 3 , 9 and hydrocarbons 10 as reductants. The recent challenges in this eld involve developing catalytic systems that function (1) at low temperatures under cold-start conditions, 11 (2) with minimum use of PGMs, 12,13 and (3) without emitting N 2 O, [14][15][16] which is a potent greenhouse gas. 17 These issues have been individually studied using different materials. The development of a single material that enables (1)-(3) is therefore highly desirable. To the best of our knowledge, no such material has been reported. In particular, achieving both (1) and (3) is difficult because N 2 O reduction to N 2 on PGMs requires relatively high temperatures (>300 C). 18 Therefore, an appropriate catalyst design is needed to obtain not only outstanding catalytic activity toward NO reduction but also high selectivity to N 2 with minimal incorporation of PGMs.
A promising approach that overcomes these challenges is the single-atom alloying concept, 19 which is relevant to single-atom chemistry. 20,21 The dilution of PGM atoms with less active metal atoms not only substantially reduces the use of PGMs but also enables drastic modication of the electronic and geometric structures for enhanced catalysis. 22 For example, the isolation of Pt or Pd with group 11 metals (Au, Ag, and Cu) enables molecular transformations that hardly proceed in the absence of single-atom alloying, such as selective hydrogenation, [23][24][25][26][27] formic acid dehydrogenation, 28 and hydrosilylation. 29 In these systems, the group 11 metals act as inert elements but modify the electronic and geometric factors of the PGM and, thus, its catalytic properties. Conversely, for NO reduction, the group 11 elements are known to be capable of NO activation. [30][31][32] Therefore, applying the single-atom alloying concept to NO reduction systems should provide an unprecedented synergistic effect for efficient NO conversion.
In this study, we focused on Cu as a main component because of its intrinsic activity toward NO reduction and its high earth abundance. We found that Cu-Pd/Al 2 O 3 (Cu/Pd ¼ 5) acts as a highly efficient catalyst for NO reduction at low temperatures (>150 C), without generating N 2 O emissions. Herein, we report both an innovative catalytic system for efficient NO reduction and novel catalytic chemistry of single-atom alloys.
solution containing Pd(NH 3 ) 2 (NO 2 ) 2 (Kojima Chemicals, 4.765 wt% in HNO 3 ) and/or Cu(NO 3 ) 2 $3H 2 O (Sigma-Aldrich, 99%), followed by stirring for 3 h at room temperature (50 ml H 2 O per gram of Al 2 O 3 ). The mixture was dried under a reduced pressure at 50 C, followed by reduction under owing H 2 (30 ml min À1 ) at 400 C (Pd) or 800 C (CuPd) for 1 h. The Cu/Al 2 O 3 (Cu: 6 wt%) and Cu-Pd/Al 2 O 3 (Cu: 6 wt%, Cu/Pd ¼ 3 and 5) catalysts were prepared by a deposition-precipitation method using urea. The g-Al 2 O 3 support was added to a vigorously stirred aqueous solution of Cu(NO 3 ) 2 $3H 2 O (50 ml H 2 O per gram of Al 2 O 3 ). Then, an aqueous solution of urea (Kanto, 99%) was added dropwise to the stirred mixture at room temperature (urea/Cu ¼ 30). The mixture was sealed with a plastic lm and kept with stirring at 70 C for 10 h. Aer completing the precipitation of Cu(OH) 2 , the supernatant was decanted and the resulting crude product was washed with deionized water three times, followed by drying under a reduced pressure at 50 C and calcination at 500 C for 1 h. For Cu/Al 2 O 3 , the resulting CuO/ Al 2 O 3 was reduced under owing H 2 at 400 C for 1 h. For Cu-Pd/Al 2 O 3 (Cu/Pd ¼ 3 and 5), the resulting CuO/Al 2 O 3 was used for successive impregnation of Pd in a similar fashion to that mentioned above. The resulting Pd-CuO/Al 2 O 3 was reduced under owing H 2 at 400 C for 1 h.

Reaction conditions
The catalyst (0.05 g) diluted with quartz sand (1.95 g, Miyazaki Chemical, 99.9%) was treated under owing hydrogen (50 ml min À1 ) at 400 C for 0.5 h prior to the catalytic reactions. NO reduction by CO was performed in a xed-bed continuous ow system by feeding NO (5000 ppm), CO (5000 ppm), and He (balance) with a total ow rate of 96 ml min À1 (GHSV ¼ 80 000 h À1 ). The gas phase was analyzed using an online thermal conductivity detection gas chromatograph (Shimadzu GC-8A, column: SHINWA SHINCARBON ST) located downstream. A stability test was done using twice the amount of catalyst (0.10 g) under similar conditions (GHSV ¼ 40 000 h À1 ). Aer a timeon-stream of 24 h, the catalyst was regenerated by owing hydrogen (50 ml min À1 ) at 400 C for 0.5 h, followed by continuing the catalytic run. A kinetic study was performed by changing the concentration of NO and CO between 0.3 and 0.6% with that of the counterpart xed at 0.5%. The reaction temperature was maintained at 150 C so that NO conversion did not exceed 30%, and the reaction rate (mol s À1 mol Pd À1 ) was calculated on the basis of NO conversion. NO + CO + O 2 and NO + CO + O 2 + C 3 H 6 reactions were performed under stoichiometric conditions as follows: NO (5000 ppm), CO (10 000 ppm), O 2 (2500 ppm), He (balance) with a total ow rate of 96 ml min À1 (GHSV ¼ 80 000 h À1 ), and NO (5000 ppm), CO (5000 ppm), O 2 (5625 ppm), C 3 H 6 (1250 ppm), and He (balance) with a total ow rate of 96 ml min À1 (GHSV ¼ 80 000 h À1 ), respectively.

Characterization
The crystal structure of the prepared catalyst was examined by powder X-ray diffraction (XRD) using a Rigaku MiniFlex II/AP diffractometer with Cu Ka radiation. High-angle annular dark eld scanning transmission electron microscopy (HAADF-STEM) was carried out using a JEOL JEM-ARM200 M microscope equipped with an energy dispersive X-ray (EDX) analyzer (EX24221M1G5T). STEM analysis was performed at an accelerating voltage of 200 kV. To prepare the TEM specimen, all samples were sonicated in ethanol and then dispersed on a Mo grid supported by an ultrathin carbon lm. The Fourier-transformed infrared (FT-IR) spectra of adsorbed CO were obtained with a JASCO FTIR-4200 spectrometer equipped with an MCT detector in transmission mode (resolution 4 cm À1 ). The samples were prepared as self-supporting wafers (2.0 cm diameter, <0.5 mm thickness) and were placed inside an IR cell with CaF 2 windows. A custom glass manifold was connected to the cell to control the gas for pretreatment and the amount of CO introduced. The cell was rst purged with He, and the sample was reduced under owing hydrogen (50 ml min À1 ) at 400 C for 30 min. Aer reduction, the wafer was cooled to 40 C under owing He. The wafer was exposed to CO (0.5%) and He (balance) with a total ow rate of 50 ml min À1 for 20 min. Aer the CO exposure, He was owed for 5 min to remove the gas phase and weakly adsorbed CO, followed by IR spectral measurements.
X-ray absorption ne structure (XAFS) spectra were recorded on the BL14B2 station at SPring-8 of the Japan Synchrotron Radiation Research Institute. A Si(311) double-crystal monochromator was used. Energy calibration was performed using Pd foil. The spectra were recorded at the edges of Pd K in a transmission mode at room temperature. The pelletized sample was pre-reduced with H 2 at 400 C for 0.5 h, and then sealed in a plastic pack under a N 2 atmosphere together with an ISO A500-HS oxygen absorber (Fe powder). The obtained XAFS spectra were analyzed using Athena and Artemis soware ver. 0.9.25 included in the Demeter package. The Fourier transformation of the k 3 -weighted EXAFS from k space to R space was performed over a k range of 3.0-15Å À1 . Some of the Fouriertransformed EXAFS spectra in the R range of 1.2-3.0Å were inversely Fourier transformed, followed by an analysis using a usual curve tting method in a k range of 3-15Å À1 . The backscattering amplitude or phase shi parameters were simulated with FEFF 6L and used to perform the curve tting procedure. For Pd-Cu scattering, intermetallic Cu 3 Pd with a Pm 3m structure was considered for the FEFF simulation. The amplitude reduction factor (S 0 2 ) of Pd was determined to be 0.775 by tting the spectra of Pd black and then used for tting of other EXAFS spectra.

Computational details
Periodic DFT calculations were performed using the CASTEP code 33 with Vanderbilt-type ultraso pseudopotentials 34 and the Perdew-Burke-Ernzerhof exchange-correlation functional based on the generalized gradient approximation. 35 The planewave basis set was truncated at a kinetic energy of 350 eV, and a Fermi smearing of 0.1 eV was utilized. Dispersion correlations were considered using the Tkatchenko-Scheffler method with a scaling coefficient of s R ¼ 0.94 and a damping parameter of d ¼ 20. 36 The reciprocal space was sampled using a k-point mesh with a spacing of typically 0.04Å À1 , as generated by the Monkhorst-Pack scheme. 37 Geometry optimization was performed on supercell structures using periodic boundary conditions. The surface was modeled based on Cu(211)-(2 Â 3) (for NO and N 2 O related reactions), Cu(111)-(2 Â 2) (for CO oxidation), and Cu(111)-(3 Â 3) (for N 2 O decomposition) slabs that were six atomic layers thick with 13Å of vacuum spacing. The convergence criteria for structural optimization and energy calculation were set to (a) an SCF tolerance of 1.0 Â 10 À6 eV per atom, (b) an energy tolerance of 1.0 Â 10 À5 eV per atom, (c) a maximum force tolerance of 0.05 eVÅ À1 , and (d) a maximum displacement tolerance of 1.0 Â 10 À3Å . The transition state search was performed using the complete linear synchronous transit/quadratic synchronous transit (LST/QST) method. 38,39 Linear synchronous transit maximization was performed, followed by energy minimization in the directions conjugate to the reaction pathway. The approximated TS was used to perform QST maximization with conjugate gradient minimization renements. This cycle was repeated until a stationary point was found. Convergence criteria for the TS calculations were set to root-mean-square forces on an atom tolerance of 0.10 eVÅ À1 .

Results and discussion
The monometallic Cu (6 wt%) and Pd (2 wt%) and the bimetallic Cu-Pd (Cu: 6 wt%, Cu/Pd ¼ 1, 3, or 5; hereaer, Cu x Pd, x ¼ 1, 3, or 5) catalysts were prepared using g-Al 2 O 3 as a support by deposition-precipitation and/or impregnation methods. X-ray diffraction (XRD) patterns of the prepared catalysts revealed that Cu-Pd solid-solution alloy phases with bimetallic compositions similar to the metal ratio in the feed were formed (Fig. S1 † and Table 1). The crystallite sizes estimated using Scherrer's equation were 3-4 nm for all of the catalysts. Fig. 1a and b show a highangle annular dark eld scanning transmission electron microscopy (HAADF-STEM) image of Cu 5 Pd/Al 2 O 3 and the size distribution of the nanoparticles, respectively. A relatively narrow size distribution between 2 and 6 nm with an areaweighted mean diameter of 4.2 nm was obtained, consistent with the crystallite size estimated by XRD ( Table 1).
The energy-dispersive X-ray spectroscopy (EDS) analysis of a single nanoparticle revealed that the Cu and Pd atoms comprising the nanoparticle were homogeneously dispersed (Fig. 1c). The high-resolution HAADF-STEM image shows an fcc crystal structure viewed along the [100] direction, consistent with the formation of a solid-solution alloy (Fig. 1d). Moreover, the isolation of Pd atoms by Cu was indicated by the presence of atoms with distinct Z contrasts. Note that the corresponding HAADF-STEM image of Cu/Al 2 O 3 showed a weak and uniform Z contrast compared with that of Cu 5 Pd/Al 2 O 3 (Fig. S2 †).
The degree of Pd isolation was further investigated by Fourier-transform infrared (FT-IR) and extended X-ray absorption ne structure (EXAFS) analyses (Fig. 2). As shown in Fig. 2a, the FT-IR spectra of CO adsorbed onto Pd/Al 2 O 3 exhibited absorption peaks assigned to the stretching vibration of C]O   adsorbed on top (2086 cm À1 ), bridge (1975 cm À1 ), and hollow sites ($1880 cm À1 ). 40 Similar absorption peaks were observed for CuPd/Al 2 O 3 , suggesting that the Pd-Pd ensembles largely remain even aer 1 : 1 alloying. For Cu-rich samples, an absorption band assignable to CO adsorbed on metallic Cu was also observed at 2100 nm À1 . 41 The peak intensities for the bridge and hollow CO substantially decreased and disappeared in the spectra of Cu 3 Pd/Al 2 O 3 and Cu 5 Pd/Al 2 O 3 , respectively, indicating that Pd atoms at the surface were isolated upon 5 : 1 alloying. There remained a weak absorption band for linear CO on Pd in the spectrum for Cu 5 Pd/Al 2 O 3 , suggesting that the isolated Pd atoms are also present at the surface. Fig. 2b shows the Fourier transforms of the Pd K-edge EXAFS spectra of the Pd-based catalysts (the X-ray absorption near edge structure spectra, raw EXAFS oscillations, curve-tting, and summary of EXAFS curve tting are shown in Fig. S3-S5 and Table S1, † respectively). CuPd/Al 2 O 3 showed both Pd-Pd and Pd-Cu bonds, while Cu 5 Pd/Al 2 O 3 exclusively showed Pd-Cu bonds, suggesting that the Pd atoms in the bulk were also isolated by Cu upon 5 : 1 alloying. Thus, small Cu-Pd nanoparticles with a single-atom alloy structure were successfully formed on the Al 2 O 3 support. Considering the limited sensitivity of EXAFS and FT-IR, we cannot completely exclude the presence of Pd-Pd interaction. However, only a small number of Pd-Pd sites, if any, seem not to contribute to the overall catalytic performance.
We next tested the catalytic activity of Cu x Pd/Al 2 O 3 in NO reduction by CO (GHSV ¼ 80 000 h À1 ), as a model reaction for exhaust-gas purication. Fig. 3a shows the NO conversion to N 2 (C N 2 ) for the prepared catalysts as a function of reaction temperature. Here, C N 2 was obtained by multiplying the NO conversion and the N 2 selectivity (Fig. S6 †). Pd/Al 2 O 3 gave the lowest C N 2 , because of the poor N 2 selectivity <40% (Fig. S6b †).
Cu/Al 2 O 3 exhibited a higher C N 2 than Pd/Al 2 O 3 because of its much higher N 2 selectivity (Fig. S6b †). CuPd/Al 2 O 3 showed a C N 2 trend similar to that of Cu/Al 2 O 3 because of the consequence of increased NO conversion and decreased N 2 selectivity (Fig. S6 †). Thus, the 1 : 1 alloy of Cu and Pd gave an insufficient catalytic performance for selective NO reduction. Interestingly, however, both NO conversion and N 2 selectivity increased when the alloying ratio was increased to 3 : 1 and 5 : 1 (Fig. S6 †), which resulted in great enhancements in C N 2 (Fig. 3a). NO was completely converted to N 2 over Cu 5 Pd/Al 2 O 3 without generating N 2 O emissions even at 200 C, where Pd showed a C N 2 of only 5%. Notably, on going from CuPd to Cu 5 Pd, the catalytic activity increased even though the Pd content was decreased to 1/5 (Table 1). Therefore, a specic synergistic effect between Cu and Pd likely contributed to the unique properties of the singleatom alloy catalyst. We emphasize that using an excess amount of Pd/Al 2 O 3 (0.50 g) with 10 times the equimolar Pd included in Cu 5 Pd/Al 2 O 3 (labeled as Pd Â 10) still resulted in poor performance (Fig. 3b), highlighting the outstanding performance of the single-atom alloy catalyst. We also tested the long-term stability of Cu 5 Pd/Al 2 O 3 in NO reduction by CO under standard conditions (GHSV ¼ 40 000 h À1 ), where 100% C N 2 was maintained at 175 C. Although a number of bimetallic catalysts for NO reduction have been reported thus far, [12][13][14][15][16][42][43][44][45][46][47][48] to the best of our knowledge, the present work represents the rst success in complete NO x removal at a temperature less than 200 C. At 150 C, although C N 2 decreased slightly at the initial stage because of N 2 O formation, it recovered aer a short H 2 treatment. This result implies that the accumulation of oxygen species at the catalyst surface triggers the loss of N 2 selectivity and that the catalytic performance could be recovered under rich conditions. We next examined the catalytic performance of Cu 5 Pd/Al 2 O in NO reduction in the presence of O 2 and O 2 + C 3 H 6 ; these conditions more closely resemble those encountered in practical use. Cu/Al 2 O 3 delivered poor performance under NO + CO + O 2 conditions. By contrast, Cu 5 Pd/Al 2 O exhibited much higher performance than Cu/Al 2 O and Pd/Al 2 O. Notably, Cu 5 Pd/Al 2 O still exhibited a performance better than or comparable to "Pd Â 10" even in the presence of O 2 or O 2 + C 3 H 6 , respectively ( Fig. 3b and S7; † a comparison with temperature dependence and T 50 is presented in Fig. S7 †). Furthermore, N 2 O evolution was sufficiently suppressed over Cu 5 Pd, where the C N 2 O (NO conversion to N 2 O, Fig. 3b and S7 †) was much lower than those for Pd. Thus, the single-atom alloy catalyst enabled not only a decrease in the noble metal use to 1/ 10 but also highly selective NO x removal. In the reactions conducted in the presence of O 2 and O 2 + C 3 H 6 , reaction temperatures greater than 200 C were needed to achieve sufficient catalytic performance (Fig. 3 and S7 †). A possible explanation is that the number of active sites for NO reduction decreased because of the involvement of other reactions such as CO and/or C 3 H 6 oxidation. This journal is © The Royal Society of Chemistry 2019 Next, to clarify the nature of the synergistic effect, we conducted a mechanistic study based on kinetic analysis and density functional theory (DFT) calculations. First, the apparent activation energy (E A ) for NO reduction by CO was estimated for the representative catalysts. The corresponding Arrhenius-type plots and the resulting E A values are shown in Fig. 4 and Table 2, respectively. Cu 5 Pd gave an E A value lower than those of Pd and Cu, which is consistent with the observed catalytic activity. In addition, we estimated the reaction orders for NO and CO pressures (P NO and P CO , respectively) to consider the rate-determining step (RDS). Both Cu and Cu 5 Pd showed negative and positive orders for P NO and P CO , respectively (see Table 2 and Fig. S8 † for details). Unlike the case for Pd-based catalysts, 49 bond dissociation of N-O has been speculated to occur via (NO) 2 dimer formation on Cu surfaces. [50][51][52] Therefore, in the present study, an extended Langmuir-Hinshelwood model that includes (NO) 2 dimer formation, the subsequent N-O scission (N 2 O formation), and N 2 O decomposition (N 2 formation) was considered. We solved the rate equation of each step regarded as the RDS by using the overall site balance and equilibrium constants for the other steps (see the ESI, † kinetic analysis). In most cases, the reaction order for P NO is positive, which is inconsistent with the observed experimental results. Conversely, when the N-O scission of the (NO) 2 is considered as the RDS, the orders for P NO and P CO range from À2 to 0 and from 0 to 2, respectively, in agreement with the experimental values. Thus, our kinetic study suggests that the N-O scission was the RDS in NO reduction by CO. Upon incorporation of Pd atoms into pure Cu, the order for P NO becomes less negative, while that for P CO decreases substantially. This result indicates that the strong adsorption of NO inhibits CO adsorption onto Cu, while the latter is enhanced in the presence of Pd. We performed DFT calculations for the relevant elemental steps on pure and Pd-doped Cu surfaces. On the basis of the literature, 53 the step site of the (211) surface was considered the active site for N-O scission (Fig. S9 †). The corresponding energy diagrams are shown in Fig. 5.
NO adsorption was weakened by the addition of Pd, which is consistent with the change in the reaction orders ( Table 2). Dimerization occurs at the terrace site adjacent to the step site, with a negligible energy barrier. The subsequent N-O scission is triggered by capture of an oxygen atom by the edge Cu atoms, resulting in the formation of an on-top N 2 O with E A values of 59.8 and 47.9 kJ mol À1 for pure and Pd-doped Cu, respectively. The calculated E A values agree with the experimental values ( Table 2). The lower E A for the Pd-doped Cu appears to originate from the destabilized adsorption of the (NO) 2 dimer by Pd.
We also considered the CO oxidation process (CO + O / CO 2 ), which is necessary for the catalytic cycle (Fig. S10 †). The CO + O reaction over pure and Pd-doped Cu(111) surfaces gave E A values of 60.9 kJ mol À1 and 34.1 kJ mol À1 , respectively ( Table  2 and Fig. S10 †). These values are very similar to or lower than those for N-O scission, which is consistent with the RDS being the scission of N-O. We also simulated the N 2 O decomposition process (N 2 O / N 2 + O) on Cu(211) and (111) surfaces to understand the intrinsic high N 2 selectivity of Cu ( Fig. 6 and S11; see S12 † for the pictures of the optimized structures).
The monodentate linear N 2 O was bent to form a bidentate N 2 O at the edge site of the Cu(211) plane without an energy barrier. The bidentate N 2 O was subsequently decomposed into N 2 and O with a low E A of 33.4 kJ mol À1 , indicating that the N 2 O once formed could be smoothly decomposed into N 2 to afford high N 2 selectivity. Although the Cu(111) surface was also active  for N 2 O decomposition in a similar fashion, the energy barrier was higher than that of Cu(211) (51.6 kJ mol À1 , Fig. S11 †). Because large Cu-Cu ensembles are present on the surface of the Cu and Cu-rich catalysts (Cu 5 Pd and Cu 3 Pd), N 2 O decomposition could also be enhanced on these catalysts. However, for CuPd, this effect is limited because of the dilution of Cu-Cu ensembles and the increase of Pd-Pd ensembles. Thus, our calculation rationalized the substantial enhancement in catalytic activity on the basis of the formation of the Cu-Pd singleatom alloy and the origin of the excellent selectivity for N 2 formation. The elucidated mechanism differs completely from those proposed for other bimetallic alloy systems. For example, for the Pt-Co system, Sato et al. reported that alloying with Co makes Pt electron-rich, which enhances back donation to adsorbed NO, inducing bond breaking. 13 Therefore, Co likely acts as a promoter for Pt. By contrast, in our system, the isolated Pd acts as an efficient promotor for Cu.

Conclusion
We prepared a series of Cu-Pd/Al 2 O 3 catalysts for selective NO reduction at low temperatures. Alloying of Pd with a large amount of Cu (Cu/Pd ¼ 5) isolates Pd and drastically improves both the catalytic activity and N 2 selectivity, affording outstanding catalytic performance. In the NO reduction by CO, NO is completely converted to N 2 even at 175 C, with long-term stability for at least 30 h. The high catalytic performance is also achieved in the presence of O 2 and C 3 H 6 , where the amount of Pd needed for a comparable performance can be reduced to 1/ 10, with minimum evolution of N 2 O. For Cu/Al 2 O 3 and Cu 5 Pd/ Al 2 O 3 , the N-O bond scission of the (NO) 2 dimer is the RDS in NO reduction by CO. This step is kinetically facilitated by the isolated Pd atoms. N 2 O decomposition to N 2 smoothly proceeds on the Cu surface, which contributes to the excellent N 2 selectivity observed for Cu and Cu-rich catalysts. The key to this efficient catalysis is the sufficient isolation of Pd atoms by Cu, highlighting the importance of catalyst design based on singleatom alloy structures. The insights gained in this study provide not only a highly efficient deNO x system with substantially reduced noble-metal content, but also open a new path for the chemistry of single-atom alloys.

Conflicts of interest
There are no conicts to declare. This journal is © The Royal Society of Chemistry 2019