In situ atomic-scale observation of oxygen-driven core-shell formation in Pt3Co nanoparticles

The catalytic performance of core-shell platinum alloy nanoparticles is typically superior to that of pure platinum nanoparticles for the oxygen reduction reaction in fuel cell cathodes. Thorough understanding of core-shell formation is critical for atomic-scale design and control of the platinum shell, which is known to be the structural feature responsible for the enhancement. Here we reveal details of a counter-intuitive core-shell formation process in platinum-cobalt nanoparticles at elevated temperature under oxygen at atmospheric pressure, by using advanced in situ electron microscopy. Initial segregation of a thin platinum, rather than cobalt oxide, surface layer occurs concurrently with ordering of the intermetallic core, followed by the layer-by-layer growth of a platinum shell via Ostwald ripening during the oxygen annealing treatment. Calculations based on density functional theory demonstrate that this process follows an energetically favourable path. These findings are expected to be useful for the future design of structured platinum alloy nanocatalysts.

Supplementary Figure 10|Oxidizing results of the disordered Pt 3 Co nanoparticle. a, CoO layers is found in a disordered Pt 3 Co particle after the oxygen annealing at 500 ºC for 30 minutes. Inset is the FFT pattern taken from the area indicated by the white box, demonstrating its disordered structure. b, Enlarged HRSTEM image illustrating the periodical unit cell of the oxide layer. c, Projection of CoO atomic model along <011> zone axis. Blue and red spheres represent the Co and O atoms, respectively.

Supplementary Note 1|Construction of structural models of Pt 3 Co nanoparticles
Model 0: A basic ordered Pt 3 Co model with no core-shell structure. An ordered intermetallic Pt 3 Co nanoparticle with no core-shell structure was first established. The diameter across two opposite (100) surfaces was 12.75 nm, the same dimension as the particle we observed in the experiment. The morphology of the particle was a truncated octahedron consisting predominantly of {100} and {111} facets. (xyz structure files of the models are available upon request to S. D.)

Supplementary Note 2|Final state of the Pt 3 Co nanoparticle during oxygen-driven core-shell formation
In the manuscript, Fig. 3 shows the continuous layer-by-layer Pt shell growth on (100) surfaces of a Pt 3 Co nanoparticle. At the time t=128 s, a 4-layer Pt-shell was formed and the measured diameter, d, was 13.57 nm.
Supplementary Fig. 5 exhibits the final state of this nanoparticle during oxygen-driven core-shell formation. Supplementary Fig. 5a, the high resolution STEM image taken at the time t=196 s, shows the 5-layer Pt shell on the (100) surface. The diameter at this time was measured to be 13.96 nm.
In the following time, the Pt shell stopped growth after the #5 Pt layer was formed. As shown in Supplementary Fig. 5b, the diameter, d, was still 13.96 nm in the next 30 minutes, and it remained unchanged upon further annealing in oxygen at 300 ºC.

Supplementary Note 3|Analysis of (111) surfaces of Pt 3 Co nanoparticles during oxygen annealing
The layer-by-layer growth on (100) surfaces was identified by direct observation along the [001] zone axis (Figs. 2, 3 in the manuscript). Although (111) facets are not parallel to the [001] zone axis, further analysis is able to reveal the status of (111) surfaces. The following analysis will be based on the geometry of a truncated octahedron. Therefore it is not necessary to distinguish Pt atoms and Co atoms in the figures and models.
Supplementary Fig. 6a is a truncated octahedron model projected along the [001] zone axis. Green spheres represent either Pt atoms or Co atoms for simplicity. For comparison, Supplementary Fig. 6b shows an extra atom layer (the red one) growing on a certain (111) surface. As illustrated in Supplementary Fig. 6, the length, L, along the [110] direction is defined as the projected distance from a reference point to the edge of the (111) surface. In Supplementary Figs. 6a and 6b, we set the octahedron center as the reference point (indicated by the blue sphere). Then it should been found that the measured length, L B , (in Supplementary Fig. 6b) will be longer than L A (in Supplementary Fig. 6a)  According to the measurements, it is clear that the distances L 0s , L 64s and L 128s are all the same length, 8.19 Å, corresponding to the projected distance between three (110) layers. No obvious structure change could be found in these high-resolution images. This result indicates that no growth or diffusion took place on the (111) top surface during the time period while a Pt shell was growing on the (100) surfaces.

Supplementary Note 4|Ex-situ results of the oxygen annealed Pt 3 Co nanoparticles
Ex-situ oxygen annealing was performed on the original Pt 3 Co/C sample in a furnace tube, while nitrogen annealing was also carried out for comparison. All the conditions (temperature, heating rate and annealing time) were similar to those used in the in-situ experiment. Either the oxygen or the nitrogen annealing was first performed for 30 minutes under 720 °C (at a very high heating rate) and then for an additional 15 minutes under 300 °C.
Aberration-corrected scanning transmission electron microscopy (AC-STEM) was performed for the characterization of the ex-situ annealed samples by using a JEM-ARM300F Grand ARM transmission electron microscope. Here, Supplementary Figs 7a-c present the typical high-angle annular dark field (HAADF) images of the oxygen annealed Pt 3 Co nanoparticles. As shown in Supplementary Fig. 7a, the alternating (200) planes of bright-dark contrast corresponding to the ordered intermetallic Pt 3 Co structure can be observed. The fast Fourier transform (FFT) pattern clearly illustrates the (100) superlattice spots, which characterized the intermetallic L1 2 phase. Meanwhile, it is clear that (100) and (111) surface region exhibits a brighter contrast in the HAADF image, demonstrating the structure of the Pt-rich shell on the ordered Pt 3 Co core. In addition, more examples in Supplementary Figs. 7b and 7c also show the ordered core-shell structure of the Pt 3 Co nanoparticles through the ex-situ O 2 annealing treatment.
For comparison, Supplementary Figs. 7d-7f present the HAADF images of the N 2 annealed Pt 3 Co sample. In Supplementary Fig. 7d, the alternating (200) planes of bright-dark contrast can also be found, indicating the transition from disordered to ordered structure under the high temperature of 720 ºC. However, as shown in Supplementary Fig. 7d Furthermore, energy dispersive X-ray spectroscopy (EDS) was also performed on these two ex-situ annealed samples. Supplementary Fig. 8 is the EDS element maps of Pt 3 Co nanoparticles, which were treated by O 2 and N 2 annealing, respectively. Results from the oxygen annealed sample (Supplementary Figs. 8a-8d) reveal Pt surface enrichment since the profile of the Pt map is bigger than that of the Co map. Intensity profiles, collected from the rectangular box in Supplementary Fig. 10a and aligned to the maximum value of Pt and Co intensities clearly reveal the Pt segregation at the surface region of the oxygen annealed Pt 3 Co nanoparticles. However, in contrast, Pt surface enrichment is not found in the N 2 annealed sample ( Supplementary Figs. 8e-8h). It is obvious that the profiles of the elemental maps of Pt and Co are almost the same in Supplementary Figs. 8f and 8g. The homogeneous distribution is confirmed in the intensity profiles, which were from the rectangular box in Supplementary Fig. 10e, and no Pt segregation is observed at the surface region of the nanoparticle. Therefore, these ex-situ results demonstrate the oxygen-driven effect on the ordered core-shell formation of Pt 3 Co nanoparticles. Meanwhile, the results also confirm that our in-situ findings also apply under similar ex-situ conditions, ruling out the electron beam effect in the in-situ experiment.

Supplementary Note 5|XPS analysis of the oxygen annealed Pt 3 Co nanoparticles
X-ray photoelectron spectroscopy (XPS) was performed on the ex-situ oxygen annealed Pt 3 Co sample and the as-prepared one. The experiment was carried out on a Kratos Analytical AXIS Supra spectrometer utilizing monochromatic Al Kα radiation (1486.7 eV, 250 W) under ultra-high vacuum conditions (∼10 −9 Torr). The electron inelastic mean free path () is about 2.0 nm, and the highest signal contribution originates from the topmost surface region (65% of the signal comes from first 2.0 nm).
The Pt/Co ratios at the surface region are calculated from the corresponding Co 2p and the Pt 4f peak areas, where the Shirley background were subtracted, by the equation where X i is the molar fraction, I i is the integrated area of the XPS peaks, and  i is the photoelectron cross-section (Scofield factor) for the element i (i=Pt, Co). The calculated Pt/Co ratios are presented in Supplementary Table 1. The surface Pt/Co ratio of the as-prepared sample (2.88) is very similar to the bulk value (3.0). In contrast, the sample treated by oxygen annealing shows a Pt/Co ratio of 4.81, indicating a Pt-rich composition at the surface region.
Moreover, Supplementary Fig. 9 shows the normalized Pt 4f photoelectron spectra of these two kinds of Pt 3 Co samples. The binding energies were referred to the Au 4f 7/2 signal at 84.0 eV from the sample holder, which was in electrical contact with the samples. In the low energy band (Pt 4f 7/2 ), a small positive shift is observed in the Pt binding energies of O 2 annealed sample (71.25 eV) compared to that of the as-prepared Pt 3 Co sample (71.15 eV). According to a theoretical explanation 1 , this change in electronic structure properties is due to the variation in the surface atomic distribution of Pt in the Pt 3 Co nanoparticles. Since a strain is introduced on the Pt shell due to a smaller lattice parameter of the Pt 3 Co core, the shell Pt atoms are squeezed closer, leading to the increased overlap of the d-orbitals, and consequently the band broadens. Then the center of the d-band moves to low energy in order to maintain the same filling degree 2 . As the density of states at the Fermi level decreases, it is increasingly more difficult to ''ionize'' the metal, resulting in the increase of Pt XPS binding energies 3 . Such a phenomenon is also found in the published literature relating to the Pt-Co core-shell structure 4 . Particularly, Zhang et al. 5 also observed a similar positive shift (0.1 eV) of Pt 4f 7/2 binding energy when the surface of Pt 3 Co nanoparticles was enriched with Pt through an annealing treatment.

Supplementary Note 6|Size distribution of Pt 3 Co nanoparticles before and after oxygen annealing
Size distributions of at least 500 Pt 3 Co nanoparticles were determined by in-situ diameter measurements. Supplementary Fig. 3 shows the size histograms before (black) and after oxygen annealing (red).
Statistics for particle sizes are listed in Supplementary Table 2.The diameter of Pt 3 Co nanoparticles before annealing ranged from 1.49 nm to 15.60 nm, and the mean size was about 4.81 nm. After in-situ annealing for 30 minutes (under 760 Torr oxygen, 720 ºC), the diameter was found to range from 2.21 nm to 19.21 nm, and the mean size was about 6.63 nm. These results show that this annealing treatment in an oxygen atmosphere did not induce severe sintering or coalescence of Pt 3 Co nanoparticles.

Supplementary Note 7|Consideration of the effect of oxygen annealing on the amorphous carbon support of the Pt 3 Co/C sample
Considering the fact that bulk carbon can burn and may be converted into gas (CO or CO 2 ) upon sample heating within the oxygen environment, it is necessary to both (1) examine the possible effect of this reaction on gas composition and (2) directly assess the stability of the carbon support.
First, we estimate the moles of carbon and oxygen in the cell under our experimental condition. Supplementary Fig. 2 is a schematic diagram showing the cross section view of the Protochips Atmosphere TM gas cell. The dimensions of the cell are about 3cm×1cm×5m. The area for sample loading is 100 m×100 m, and the gap between the chip and the window is about 5 m.
If we assume that the in-cell gas can be treated as an ideal gas 6 , following the ideal gas law, where p is the gas pressure, V is the gas volume, n is the number of moles, R is the universal gas constant 0.08206 (atm•L)/(mol•K), and T is the temperature, the number of oxygen moles (n O ) in the gas cell under our experimental condition (V=1.5×10 -6 L, p=1.0 atm, T=993 K) is The typical volume of catalyst sample is less than Assuming the sample is all amorphous carbon, then the carbon moles of the support (n C ) is 3 16 C sample 11 C 3 C 2 10 1 10 1.7 10 mol 12 10 where  C is the mass density of amorphous carbon (2.0 g/cm 3 ), and M C is the molar mass of carbon (12 g/mol). This means the oxygen moles always far exceeds the carbon moles in the cell under our experimental condition. Even if the carbon support can react fully with the oxygen gas, the carbon oxide (CO or CO 2 ) produced is only of order 0.1% of all the gas in the cell. It can thus be ensured that the gas composition in the cell is oxygen with purity over 99.9% during the experiment.
However, by direct observation of the carbon support, we can demonstrate its stability during the annealing experiment. Supplementary Fig. 14a, b are the low magnification STEM images showing the Pt 3 Co/C sample before and after the oxygen annealing process, respectively. False color images are used to enhance the contrast of the carbon support in the gas cell. The light blue part besides the nanoparticles is the carbon support. From these two images, it can be seen that the profile of the carbon support was essentially unchanged after the oxygen annealing, while some of the nanoparticles exhibited different contrast due to the tilting of their orientations. Supplementary Fig. 14c, d show the detailed characterization of the carbon support before and after the oxygen annealing, respectively. The green area represents the carbon support and the blue particles are the Pt 3 Co catalyst. Red area here represents the background (the cell window) on which some tiny and separated green spots should be the noise. Although some of the particles exhibit a slight degree of sintering after 30-minute oxygen annealing, the carbon support shows no noticeable change, according to these two images. Therefore, no obvious loss of the amorphous carbon can be found according to the in-situ observation. The carbon support of the Pt 3 Co/C catalyst is thus quite stable during the 30-minute static oxygen annealing (under 760 Torr oxygen, 720 ºC).
Overall, we conclude that the carbon support is stable during the 30-minute oxygen annealing, and its possible combustion would not, in any case, significantly affect the oxygen environment inside the gas cell.

HAADF-STEM image simulation
After the atomic models (see Supplementary Note 1) were established, a (100) surface slab (11.59 nm × 11.59 nm × 1.37 nm) was extracted from each model for the image simulation.
HAADF-STEM image simulation was performed using the QSTEM simulation package 7 . The simulation was carried out using a 512×512 pixel area and a single slice thickness of 1.93 Å. The microscopy parameters used for the simulations were the same as those for imaging.