Co 2p Co(0) 778.3 Rh 3d Rh (0) 307.2 810 800 790 780 770 Binding Energy (ev) (a) 320 315 310 305 Binding Energy (ev) (b) Supplementary Figure 1 Photoemission features of a catalyst precursor which was prepared through oxidization with a following reduction at 400 o C in H 2. (a) Co 2p spectra and (b) Rh 3d spectra. The reduction at 400 o C in 5% H 2 resulted in the formation of thin layers of metallic Co with doped Rh atoms at a metallic state supported on cobalt monoxide. 1
K 2 (k) (Å -2 ) FT Magnitude (Å -3 ) 2.0 1.5 1.0 0.5 0.0-0.5-1.0-1.5-2.0 2.0 1.5 1.0 0.5 0.0 0 1 2 3 4 5 6 7 8 9 k (Å -1 ) (a) 0 1 2 3 4 5 r (Å) (b) Supplementary Figure 2 K-space (a) and r-space (b) of in-situ studies of EXAFS of catalyst precursor during calcination at 150 o C in O 2. Analysis of the data of k and r space of catalyst precursor experienced with annealing in 5% O 2 showed that (1) Rh is singly dispersed since CN(Rh-Rh) is 0, (2) Rh is bonded to oxygen atoms as CN(O-Rh) is 4-5, and (3) Rh does not bond to Co atoms as CN(Co-Rh) is zero. Supplementary Table 1 lists the coordination numbers and bond lengths of Rh atoms on the catalysts precursor calcinated in 5% O 2 upon Rh cations were introduced to the surface of Co 3 O 4. 2
Supplementary Figure 3 TEM studies of catalyst Rh 1 Co 3 /CoO. It is consistent with the value calculated from the XRD pattern reported in literature 1. (a) Large-scale image. (b) High resolution images of small area. (c) Enlargement of dashed box marked in Figure S3b. Interplanar distance of (111) planes was measured. 3
CoO (200) CoO (111) RhCo (111) 36 38 40 42 44 2 theta (degree) Supplementary Figure 4 Diffraction pattern of Rh-Co bimetallic nanoparticles supported on cobalt oxide. 4
Blue: oxygen vacancy Rh O Co (c) (d) (e) (f) (g) * (a) * * * * (h) (i) (j) (b) Supplementary Figure 5 Optimized structure models of an isolated site Rh 1 Co 3 and different adsorption modes of CO and NO on the surface of Rh 1 Co 3 /CoO(011) catalyst. (a) Optimized Rh 1 Co 3 site on CoO surface; here the blue ball marks the original position of an oxygen vacancy; the filled Rh atoms does not exactly locate at the oxygen vacancy. (b) Bonding environment of Rh atom, Rh 1 Co 3 ; Rh 1 atom bonds with Co(3), Co(5) and Co(7) and O(8) atoms; The distances between Rh 1 and Co(2), Co(4), and Co(6) are obviously larger than a typical Rh- Co bond. (c) Adsorption of only one NO molecule to the Rh 1 atom. (d) Adsorption of only one CO molecule on Rh 1. (e) Adsorption of two NO molecules on Co-Rh 1 -Co of Rh 1 Co 3. (f-j) Adsorption of both one NO and CO molecules on Rh 1 Co 3 /CoO. 5
Supplementary Figure 6 Optimized structural model of CO and NO on Co-Rh 1 -Co of Rh 1 Co 3 /CoO(001). 6
(iii) (v) (iv) (i) (ii) Supplementary Figure 7 Another reaction pathway including NO N+O and a following step N+NO N 2 +O and relative energies of reactant, transition state, and product. 7
(a) (c) (e) (b) (d) (f) Supplementary Figure 8 Computational results of NO molecule on Rh 1 Co 3 /CoO and Rh- Co bimetallic nanoparticle on CoO. (a) Optimized structures of NO on Rh 1 Co 3 /CoO. (b) Optimized structure of NO on Rh-Co bimetallic nanoparticle/coo. (c) and (d): Charge mapping of the optimized structures (a) and (b). (e) and (f): spin-polarized PDOSs projected on Rh-4d orbitals connected to N atom of the adsorbed NO and the N-2p and O-2p orbitals of adsorbed NO (c1 for Rh 1 Co 3 /CoO and c2 for Rh-Co nanoparticle/coo). In (e) and (f), the Fermi levels marked as dashed line are set at zero. 8
N 2 O on Rh 1 Co 3 /CoO x N 2 O on Rh 6 Co 6 /CoO x E ad (N 2 O)= -1.56 ev (a) E ad (N 2 O)= -0.85 ev (b) Supplementary Figure 9 Adsorption energies of N 2 O on singly dispersed Rh 1 Co 3 site on CoO (a) and on Rh-Co bimetallic nanoparticle supported on CoO (b). 9
Supplementary Figure 10 Optimized potential surface structure of (001) and (011) of CoO supporting Rh 1 Co 3 sites. The difference between the two surfaces is the coordination environment of Co cations. Each Co atom bonds to four O atoms on (011) surfaces but five O atoms on (001) surfaces. It is noted that both types of surfaces contain the adsorption site, Co-Rh-Co. Either (110) or (001) can be used as the surface structure of CoO of Rh 1 Co 3 /CoO in the DFT calculations. To test whether computation using CoO(001) could give substantially different computational results, the binding energy of two NO molecules on a Co-Rh-Co of CoO(001) and the binding energy of a CO molecule on a Co atom near to the Co-Rh-Co site on CoO(001) were calculated. The optimized adsorption modes and calculated the relative energies of CO and NO on the surface of Rh 1 Co 3 /CoO (001), were shown in Supplementary Figure 6. Very similar to (011) surface (Supplementary Figure 5), co-adsorption of two NO molecules on (001) is the most stable adsorption. 10
Supplementary Figure 11 Schematic for the calculation. (a) A nanocube of a pure metal. (b) Unit cell of a crystal with FCC lattice. (c) A bimetallic nanocube. 11
5 cobalt ions/1 nm 2 6.12 cobalt ions/1 nm 2 (a) (b) Supplementary Figure 12 Structural parameters of type B surface of Co 3 O 4. (a) Co 3 O 4 (1-10) surface and (b) Co 3 O 4 (001) surface. 12
Supplementary Figure 13 Structural parameters of surfaces of CoO. (a) CoO (1-10) surface and (b) CoO (001) surface. 13
Supplementary Figure 14 Illustration of the limitation of EDX line-scan technique in identifying whether Rh atoms are on surface or not. (a) Top view; (b) side view. The limited spatial resolution of EDX line-scan technique makes identification of location of Rh atoms challenging. 14
Supplementary Table 1 List of the calculated atomic ratio of all Rh atoms to all Co atoms of the topmost Co layer of CoO. Reaction condition of sample A Reaction condition of sample B Rh/Co first layer in CoO Upon vacuum drying of Rh(OH) z deposited on CoO. In this case, all Rh atoms are on the surface of cobalt oxide. During and after calcination in air at 150 o C 1.40% 1.31% During and after catalysis in NO+CO at 150 o C During and after catalysis in NO+CO at 400 o C 1.26% 1.25% 15
Supplementary Table 2 Comparison of catalytic activity and selectivity of the CoO, Rh 1 O n /CoO, and Rh 1 Co 3 /CoO measured under kinetics controlled regime (conversion<20%). 20ml/min of 10%NO, 30ml/min 10% CO, and 10ml/min 99.99% Ar were mixed and flew through a catalyst bed. The thickness of the catalyst bed is 3 mm. Conversion of NO on different catalysts Catalyst 10 mg CoO 5 mg Rh 1 O n /CoO 5 mg Rh 1 Co 3 /CoO Conversion at 110 o C 0 0 6.66% Conversion at 150 o C 2.97% 0.47% 12.4% Selectivity for production of N 2 on different catalysts Catalyst 10 mg CoO 5 mg Rh 1 O n /CoO 5 mg Rh 1 Co 3 /CoO Selectivity at 110 o C 0 0 97.9% Selectivity at 150 o C 0 0 100% Turn-over frequency for reduction of NO to N 2 on different catalysts. It is shown in the number of N 2 molecules produced per second from each Co atom (for CoO), each Rh atom (for Rh 1 O n /CoO), and each isolated bimetallic site Rh 1 Co 3 (for Rh 1 Co 3 /CoO. Catalyst 10 mg CoO 5 mg Rh 1 O n /CoO 5 mg Rh 1 Co 3 /CoO TOF at 110 o C 0 0 4.39 16
Supplementary Table 3 Adsorption energies of nitric oxide and bond length of Rh-N (D Rh- N) for NO molecules adsorbed on a Rh atom of the singly dispersed bimetallic site Rh 1 Co 3 /CoO and on a Rh atom of a Rh-Co bimetallic nanoparticle/coo. Catalyst Rh 1 Co 3 /CoO Rh-Co bimetallic nanoparticle/coo E ad -NO (ev) -3.05-2.50 D Rh-N (Å) 1.765 1.791 17
Supplementary Table 4 Bader charges for NO adsorption on a Rh atom of an isolated Rh 1 Co 3 on CoO and on a Rh atom of a Rh-Co bimetallic nanoparticle on CoO. Species Atom Bader charge (e) Rh 1 Co 3 /CoO Rh * 0.05 N 0.10 O -0.41 NO -0.31 Rh-Co bimetallic NP/CoO Rh * -0.13 * NO-coordinated Rh atom. N 0.22 O -0.53 NO -0.31 18
Supplementary Table 5 Effects of the U terms (2.5-3.1 ev) on the absolute value of the average magnetic moments M ave (in μ B per TM atom) of two Co atoms nearby Rh 1 in the structure of the i step and the calculated energy. U/eV (J=1 ev) 2.5 2.6 2.7 2.8 2.9 3.0 3.1 E/eV -343.299-342.362-341.409-340.553-339.659-339.049-337.948 M ave 2.391 2.396 2.407 2.419 2.434 2.486 2.449 Note: Here the U is the U eff =Ū-J. 19
Supplementary Table 6 Effects of the U terms (3.5-4.1 ev) on the absolute value of the average magnetic moments M ave (in μ B per TM atom) of two Co atoms nearby Rh 1 in the structure of the i step and the calculated energy. Ū/eV (J=1 ev) 3.5 3.6 3.7 3.8 3.9 4.0 4.1 E/eV -343.299-342.362-341.409-340.553-339.659-339.049-337.948 M ave 2.391 2.396 2.407 2.419 2.434 2.486 2.449 20
Supplementary Methods Identification of location of Rh atoms supported on surface of catalyst Rh 1 Co 3 /CoO. Rh atoms are located on the surface of oxide support. It is confirmed with the following experiments. One is the comparison of atomic ratio of all Rh atoms to all Co atoms of the topmost layer of Co atoms of CoO for sample A to that of sample B. Sample B was prepared with steps a-d in Figure 2. Sample B was the Rh 1 Co 3 /CoO catalyst exhibiting high conversion and selectivity for reduction of NO to N 2 at 110 o C. Sample A was prepared by introducing precursor of Rh to surface of well-prepared CoO nanorods through deposition precipitation with a following drying process in vacuum oven at 70 o C. It is noted that CoO was well prepared before the introduction of precursor of Rh. By tuning the concentration of Rh precursor in solution at this step of deposition precipitation, the atomic ratio of all Rh atoms to all Co atoms of sample A was remained as the same as that of sample B. For sample A, Rh atoms prepared on sample A existed in the format of Rh(OH) z and remained on the surface of CoO since there was no annealing and thus no thermal diffusion. Thus, all Rh atoms of sample A are on the surface of the substrate. The measured atomic ratio of all Rh to all Co atoms of the topmost layer of Co atoms of CoO with a size of 6-8 nm 6-8 nm 150-200 nm, is therefore 1.40%±0.07%. Sample B is in fact the catalyst of singly dispersed bimetallic sites. It was synthesized through the processes presented in Figure 2 of the main text. If Rh atoms of the catalyst were diffused to subsurface or even deep layers, the detected number of photoelectrons generated from Rh 3d (I Rh3d ) must have been decayed by following the Lambert-Beer s law,. In this equation, n is the location of Rh atoms presented; it is counted with number (n) of atomic layer along surface normal; d is the distance between two adjacent atomic layers along surface normal; is the mean free path of photoelectrons of Rh 3d while they travel through the cobalt oxide layers; it is about 1.6 nm for photoelectrons with KE of about 1180 ev travelling through cobalt monoxide. In sample A, there is no any loss of photoelectrons of Rh 3d since all Rh atoms are on the surface of cobalt monoxide. If all Rh atoms are anchored on the surface of CoO of sample B, the atomic ratio to all Rh atoms to all Co atoms of the topmost layer of Co of CoO should be about 1.40%±0.07% However, it will be smaller than 1.40%±0.07% of sample A (though the atomic ratio of all Rh atoms to all Co atoms of sample A is the same as that of sample B) if some of these Rh atoms of sample B are located in subsurface or deeper layers. As shown in Supplementary Table 1, the measured atomic ratio of all Rh atoms to all Co atoms of the topmost layer of Co of CoO of Rh 1 Co 3 /CoO during catalysis is about 1.25%. It is quite close to 1.40% Thus, certainly majority of Rh atoms of the catalyst Rh 1 Co 3 /CoO are in fact located on topmost layer of the catalyst. As discussed above the experiments using XPS showed that Rh atoms are still on the surface of cobalt oxide after experiencing the synthetic steps of Figure 2 in main text. A 2-D mapping could provide information to aid the understanding of the location of Rh atoms of the catalyst particle. Technologically, however to achieve a 2-D mapping of this sample could be extremely challenging. One reason is the quite low concentration of Rh atoms. The ratio of Rh atoms to all Co atoms on the surface of a cobalt oxide is very low, < 2%. Another reason is the location of the Rh atoms. The physical size of Rh 1 Co 3 along surface normal is only 0.5 nm or less, definitely less than 1 nm (as schematically shown in Supplementary Figure 14). Due to the 21
limited spatial resolution (actual resolution of EDX mapping is larger than 1-2 nm in most cases) and the small physical size of the singly dispersed bimetallic site, it is hard to get a clear EDS mapping image or a high-resolution EDX line-profile with enough resolution to judge whether the Rh atoms are on surface of CoO. In fact, there is another indirect evidence supporting the presentation of Rh atoms on the surface instead of sub-surface. This evidence is based on the principle that heterogeneous catalysis is performed on surface layer of a catalyst. If Rh atoms are in the subsurface and thus typically cannot participate into reaction on surface, the surface layer of CoO will be very similar to that of a bare CoO; then, its catalytic performance should be very similar to a bare CoO. In fact, the catalytic activity and selectivity of Rh 1 Co 3 /CoO (Figure 5a) are much higher than those of bare CoO (Figure 5c). From this point of view, Rh atoms of Rh 1 Co 3 play significant role in promoting catalytic activity and selectivity of Rh 1 Co 3 /CoO. Thus, Rh atoms of Rh 1 Co 3 /CoO must be on the topmost surface of the catalyst. Choice of U in DFT-U calculations. The U eff = U-J =3.0 ev value used in this work is 3.0 ev. It is deduced from the literature published by Wang, Maxisch, and Ceder (44), where U eff =3.3 ev was determined via fitting the oxidation energies of CoO within the GGA+U framework. While there is no universally accepted method for the choice of the value of U eff for the Hubbard correction, Ceder and co-workers carried out a systematic study of the oxidation energies of transition metals and suggested that the choice of U eff should be made to describe accurately the formation energies of different oxides. By using the latter approach, it was found that the U eff values are close to the empirical estimates of Wang, Maxisch, and Ceder (44). Using the linear response approach, Chen, Wu, and Selloni used U eff =4.4 and 6.7 ev for Co 2+ and Co 3+ in Co 3 O 4, respectively (46). The U eff value of 3.0 ev used here is higher than the 2.0 ev used by Walsh et al for Co doped ZnO (47), but is slightly lower than those for CoO described by Wang et al (44) and Chen et al (46) because in the partially reduced Co 2- in the Rh 1 Co 3 /CoO system there is weaker on-site repulsion in the less contracted 3d orbitals of Co ions (or atom) with lower oxidation state. In fact, 3.0 ev was also considered to be the optimal option for bulk Co metal by O Shea, Moreira, Roldan and Illas (48). Here how the U eff values affect the total energy and the magnetic moment on Co atoms in the original structure with O vacancy was considered. The effects of the U terms and the absolute value of the average magnetic moments in step i were listed in Supplementary Table 5 and Supplementary Table 6. As can be seen from the Supplementary Tables 5 and 6, the different U eff values have some influence on the total energy of this structure, but little effect on the relative values in reaction energy. The difference in energy calculated from U eff =2.9 and U eff =3.0 (only 0.61 ev) is the smallest one among all the differences. Especially, the absolute value of the average magnetic moment of two Co atoms, 2.486 in U eff =3.0 is close to the 2.65 of CoO in lattice (44). Thus, the choice of U eff =3.0 to describe the Co interacting with Rh is reasonable. Calculation of the number of the active sites Rh 1 Co 3 on surface of catalyst Rh 1 Co 3 /CoO. 10 mg catalyst of Rh 1 Co 3 /CoO was used for the catalytic measurement. The concentration of Rh is 0.25wt%. Rh 1 Co 3 is the active site. As Rh 1 Co 3 is singly dispersed, the number of all Rh atoms in the catalyst equals to the number of Rh 1 Co 3. It was calculated:. 22
Calculation of the number of Rh sites on surface of Rh-Co bimetallic NPs supported on CoO or SiO 2. To simplify the calculation, it is assumed that (1) the Rh-Co bimetallic nanoparticles are cubic (Supplementary Figure 11a), (2) Rh and Co pack into a FCC lattice (Supplementary Figure 11b), and (3) 50% of surface atoms of a cubic bimetallic nanoparticle are Rh atoms (Supplementary Figure 11c). The TEM studies showed the size of the alloy nanoparticles is about 1.5 nm. It is noted that Co atoms are dispersed on both bulk and surface of Rh-Co bimetallic particles. The ratio of Rh atoms on surface of a Rh-Co cube to all Rh atoms of the cube with a size of 1.5 nm can be calculated. Multiplying this ratio by the total number of all Rh atoms of the catalyst will give us the total number of Rh atoms on the surface of the catalyst participating into the catalysis. 1) The total number of Rh atoms in a 1.5 nm Rh-Co bimetallic nanoparticle can be calculated: = (Here l is the size of the cube in nm; 4 is the number of metal atoms in a FCC lattice cube; ½ is the overall atomic fraction of Rh. 112 is the total number of Rh atoms in a Rh 0.5 Co 0.5 cube with a size of 1.5 nm.) 2) The total number of all Rh atoms of surface can be calculated: = (Here 6 is the number of faces of a cube; (0.28nm) 2 is area of square unit cell of Rh atoms on (100) of a cube; ½ means that only half of the surface metal atoms are Rh atoms. 52 is the total number of Rh atoms on a Rh 0.5 Co 0.5 cube with a size of 1.5 nm.) 3) The ratio of R atoms on surface to the Rh atoms of a cube can be calculated:. 46.3% is the ratio of Rh atoms on surfaces of a Rh 0.5 Co 0.5 cube with a size of 1.5 nm to all Rh atoms of the cube. 4) The total number of all Rh atoms of 10 mg catalyst can be calculated: =2.85 (Here 5% is the weight percent of Rh atoms to support of CoO. 10 mg is the weight of support of the weight of the catalyst.). 5) The number of all Rh atoms on surface of Rh-Co metallic nanoparticle supported on CoO can be calculated: =2.85 10 18 0.463=1.33 10 18 Rh atoms Calculation of atomic ratio of Rh on surface to the Co atoms on the surface of a cobalt oxide nanorod (Rh 1 Co 3 /CoO catalyst). 1. Calculation of the total number of Rh atoms on surface of a Co 3 O 4 nanorod can be calculated with the following method. 1) Volume of a Co 3 O 4 nanorod (6 nm 6 nm 100 nm) V o =3,600 nm 3 23
2) Number of Co atoms of a Co 3 O 4 nanorod (Here, V o =3,600 nm 3,,, ) Thus, the total number of Co atoms of a Co 3 O 4 nanorod is 164970. 3) Atomic ratio of Rh atoms to all Co atom of catalyst precursor after annealing in O 2 at 250 o C can be calculated with the following steps. (a) Mole of Rh The weight percentage of Rh to Co 3 O 4 is 0.25% based on inductively coupled plasma (ICP) measurements. Then, the mol of Rh atoms in 1 gram of catalyst precursor: = 2.44 10-5 mole (b) Mole of Co atom in 1 gram of catalyst precursor: = 1.25 10-2 mole (c) Atomic ratio of all Rh atoms to all Co atoms 2.44 10-5 / 1.25 10-2 = 0.195 % 4) Number of Rh atoms in a Co 3 O 4 nanorod Since the atomic ratio of Rh to Co is 0.195%, the number of Rh atoms (N Rh ) supported on a Co 3 O 4 nanorod is 322 (=0.195% 164970). As Rh atoms are locate on surface of Co 3 O 4, there are 322 Rh atoms located on surface of a Co 3 O 4 nanorod. 2. Total number of Co atoms on the topmost surfaces of each Rh 1 /Co 3 O 4 nanorod can be calculated with the following method. 1) The surface area of a nanorod (6 nm 6 nm 100 nm) with exposed (1-10), (001), and (110) faces. Surface area of two (1-10) surfaces: 6 nm 100 nm 2 = 1200 nm 2. Surface area of two (001) surfaces: 6 nm 100 nm 2 = 1200 nm 2. Surface area of two (110) surfaces: 6 nm 6 nm 2 = 72 nm 2. 2) As shown in Supplementary Figure 12, each square nanometer (1 nm 2 ) of (1-10), (001), and (110) has 5, 6.12, and 5 cobalt atoms, respectively. So the total number of cobalt atoms on the surface of a nanorod is N(cobalt ions) = 5 1200 + 6.12 1200 + 5 72 = 13704. That is, there are approximately 13704 cobalt atoms on the topmost surfaces of a Co 3 O 4 nanorod [N surface (Co) = 13704]. 3. Ratio of Rh atoms (on surface) to all Co atoms of the topmost layer of Co atoms of a Co 3 O 4 nanorod can be calculated with the following method. As the Rh atoms were introduced upon the synthesis of Co 3 O 4 nanorods, it is assumed that all Rh atoms were anchored on the surface of the Co 3 O 4 nanorods. The total number of Rh atoms on surface of a Co 3 O 4 nanorod and the total number of all Co atoms on the topmost surface layer of a nanorod were calculated above. Then, the ratio of Rh atoms to all Co atoms of the topmost of Co atoms of a Co 3 O 4 nanorod is 24
That is, there is one Rh atom among every 42 Co atoms of the topmost surface of Co 3 O 4 nanorods. If some Rh atoms are buried in the subsurface or deeper layers, the atomic ratio of Rh to Co will be lower than 2.35%. 4. Ratio of Rh atoms (on surface) to all Co atoms of the topmost surface of a CoO nanorod can be calculated with the following method. As shown in Supplementary Figure 13, each square nanometer (1 nm 2 ) of (1-10), (001), and (110) have 7.77, 10.99, and 7.77 cobalt atoms, respectively. For a CoO nanorod with a size of about 6 nm 6 nm 100 nm: Surface area of two (1-10) surfaces is 6 nm 100 nm 2 = 1200 nm 2. Surface area of two (001) surfaces is 6 nm 100 nm 2 = 1200 nm 2. Surface area of two (110) surfaces is 6 nm 6 nm 2 = 72 nm 2. The total number of cobalt atoms on the surface of a nanorod is N(cobalt ions) = 7.77 1200 + 10.99 1200 + 7.77 72 = 23071 That is, there are approximately 23071 cobalt (II) cations on the topmost surface of a CoO nanorod [N surface (Co) = 23071]. Then, the ratio of Rh atoms to Co atoms of the surface of each CoO nanorod is Thus, there is one Rh atom among every 70 Co atoms of the topmost surface of CoO nanorods. The measured ratio of all Rh atoms on surface to all Co atoms of the topmost Co atoms of CoO nanorods are 1.25%-1.40% in each step of the preparation of Rh 1 Co 3 /CoO. They are consistent with the calculated atomic ratio, 1.40%. 25