Atomic Structure and Local Electronic States of Single Pt Atoms

Nov 8, 2018 - Atomic Structure and Local Electronic States of Single Pt Atoms Dispersed on Graphene. Kenji Yamazaki , Yosuke Maehara , Chi-Cheng Lee ...
0 downloads 0 Views 6MB Size
Article Cite This: J. Phys. Chem. C XXXX, XXX, XXX−XXX

pubs.acs.org/JPCC

Atomic Structure and Local Electronic States of Single Pt Atoms Dispersed on Graphene Kenji Yamazaki,*,† Yosuke Maehara,† Chi-Cheng Lee,‡ Jun Yoshinobu,‡ Taisuke Ozaki,‡ and Kazutoshi Gohara*,† †

Division of Applied Physics, Graduate School of Engineering, Hokkaido University, Sapporo 063-8628, Japan The Institute for Solid State Physics, The University of Tokyo, 5-1-5, Kashiwanoha, Kashiwa, Chiba 277-8581, Japan



Downloaded via UNIV OF NEW ENGLAND on November 23, 2018 at 03:23:21 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: In single-atom catalysis, the atomic structure and local electronic states of single atoms on a supporting material remain a fundamental question. We experimentally and theoretically solved these problems for single Pt atoms dispersed on freestanding graphene using plasma sputtering. Electron microscopy revealed the atomic arrangements of Pt atoms binding to carbon atoms at the step edges of nanographene flakes. X-ray photoelectron spectroscopy revealed a large binding energy shift of the Pt 4f state of a single Pt atom. First-principles calculations elucidated that the Pt 5dxy-orbital in the step edge plays a crucial role in the formation of chemical bonds to C atoms and in the considerable charge transfer from Pt to C atoms, resulting in the large binding energy shift of the Pt 4f state.



INTRODUCTION The development of single-atom catalysts consisting of isolated single metal atoms dispersed on a support substrate has recently attracted significant interest,1−3 but it remains quite controversial whether a single isolated atom such as Pt on a supporting material is a superior catalyst. On one hand, the Pt atom as the ultimate small-size catalyst has been reported to have great potential due to its extremely high activity and selectivity.4−6 On the other hand, it has been shown that Pt single atoms behave as spectators in a specific reaction system.7 Thus, the chemical activity of single Pt atoms is still not very clear. To truly clarify this problem, it is essential to elucidate the atomic arrangement and local electronic states using a simple system of isolated single Pt atoms on a supporting material. Aberration-corrected scanning transmission electron microscopes (STEMs) with spectroscopy analysis of electron energyloss spectroscopy (EELS) have been used for characterizing the properties of isolated single atoms at an atomic scale.8−10 It has been reported that STEM-EELS measurement was applied to isolated single Pt atoms dispersed on graphene,11 but atomic arrangement and local electronic state in atomic resolution have not yet been elucidated. X-ray photoelectron spectroscopy (XPS) measurement has been widely used to characterize chemical and electronic properties on a larger scale.12 Elementspecific spectra and core-level shifts of metal nanoparticles have been observed since the 1970s.13 In the 1980s, positive core-level shifts were estimated and observed on Pt and C systems.14 In the 1990s, an XPS study of single Pt atoms on © XXXX American Chemical Society

silicon dioxide was conducted but the electronic properties and chemical activity remained unclear.15 In the 2000s, there was a perfusion of reports investigating the cause of the core-level shift, including studies on the effects of the support material,16 chemical doping,17 and the strong interaction between other elements.18 Recently operando XPS measurements during the chemical reaction of catalysts have also been made available.19 However, the details of the atomic structure and local electronic states of single Pt atoms are still not known. The observation region for atomically resolved images, such as in transmission electron microscopy (TEM)/STEM and scanning tunneling microscope, is on the nanometer scale. In contrast, spectroscopic measurement methods, such as XPS and X-ray absorption spectroscopy, require submillimeter or tens of micrometers diameters, even when using a synchrotron radiation source. Therefore, there is a discrepancy on the scale of almost 100−1000 times between atomically resolved imaging and spectroscopic measurement. To resolve this problem, there is an urgent need for well-defined samples of dispersed single metal atoms without sintering, as noted in a recent review.20



METHODS Graphene Synthesis and Transfer on a TEM Grid. Graphene was synthesized by the chemical vapor deposition Received: May 13, 2018 Revised: November 7, 2018 Published: November 8, 2018 A

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 1. Sputtering-time dependence of Pt on graphene. High-angle annular dark field (HAADF) images corresponding to sputtering times of 1, 5, 30, and 60 s from left to right. The insets in the images for sputtering times of 1, 5, and 60 s are enlarged views of a single Pt atom with a size less than 2 Å, a two-dimensional (2D) cluster with 8 atoms, and a crystalline three-dimensional (3D) nanoparticle with a (111) lattice fringe of 2.3 Å, respectively. The scale bar of each image is 5 nm. The graph below shows the sputtering-time dependence of energy-dispersive X-ray spectroscopy (EDX) counts (black line) and the cover ratio (blue line) for Pt atoms. The cover ratio was defined as the ratio of the bright area extracted from HAADF intensity corresponding to Pt atoms to the area of the entire field of view.

Figure 2. Anchor sites of platinum atoms on freestanding graphene. (A) HAADF image used for analyzing the anchor sites of platinum atoms. Platinum was sputtered for 1 s. The lower panel shows an intensity profile along the yellow arrow in the HAADF image. The width of the line profile was 6 pixels. (B) Pseudocolor image according to the distribution of the HAADF intensity shown below. The image is averaged by a 5 × 5pixel filter (0.4 Å/pixel). The mixture of Gaussian distributions was estimated from the intensity distribution of the HAADF intensity using a statistical analysis of the EM algorithm.26 As a result, the distribution of intensity consisted of eight Gaussian distributions and the peak intensity of the eight distributions was proportional to the eight discrete numbers from 0 to 7 (see the inset and color bar). (C) Magnification of the region in the yellow square in (B). The panel below shows a line profile of HAADF intensity along the yellow arrow, including a platinum atom anchored by the step edge of the first and second graphene layers.

followed by ultrasonic treatment for 5 min. We loaded the Cu substrates into a 2 in. quartz chamber in a CVD furnace and heated them to 300 °C in a 50 sccm Ar flow to form a clean oxide layer. Next, the sample was annealed at 1000 °C for 30 min in a 48 sccm Ar and 2 sccm H2 mixed gas flow to remove

(CVD) method on a copper substrate (purity 99.8%; thickness 25 μm; Alfa Aesar #46365).21 Before the graphene synthesis, we immersed Cu foils into an acetic acid solution for over 48 h to remove a naturally formed Cu oxide layer. After the treatment, we rinsed the Cu substrate with pure water, B

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

1 shows the sputtering-time dependence of Pt atoms on graphene. In the high-angle annular dark field (HAADF) images across the top of the figure, as the sputtering time increases, the size of the bright regions also increases from single atoms to clusters of several atoms, resulting in crystalline nanoparticles characterized by lattice fringes. Dispersed single atoms with a size less than 2 Å are clearly discernible in the image of a sputtering time of 1 s. The atom was identified as Pt by energy-dispersive X-ray spectroscopy (EDX), and separation of Pt atoms of 99% was confirmed by measuring the nearest-neighbor distance between Pt atoms (Figure S3). The density of Pt atoms sputtered for 1 s was estimated approximately 3 × 1013 atoms/cm2. Therefore, the density for n s was n times the density for 1 s. Single Pt atoms were actively moving on the graphene surface due to the electron beam irradiation. Dispersed single Pt atoms were observed across the whole TEM grid. We confirmed that the dispersion of single atoms remained for 1 year under ambient conditions. The graph below shows the sputtering-time dependence of EDX count (black line) and cover ratio (blue line). The EDX count is proportional to the total number of Pt atoms. The total number of Pt atoms increases linearly, but the cover ratio becomes saturated after 10 s due to the thickness of Pt. This indicates that the sintering of single Pt atoms and the formation of a three-dimensional (3D) structure started from about 10 s in the density of 3 × 1014 atoms/cm2. Twodimensional (2D) planar clusters, which each contained about 10−20 Pt atoms and had a diameter of 1−2 nm in the density of 1.5 × 1014 atoms/cm2, were observed before the formation of 3D nanoparticles (Movie S1 and Figure S5). This shows that the Pt atoms grew to 3D nanoparticles via 2D clusters as the sputtering time increased. Although the atomic arrangement of Pt clusters composed of 10−20 Pt atoms are constantly changing during the continuous TEM observation, the numbers of Pt atoms in the cluster did not change significantly. Therefore, we assumed that this was a 2D cluster, as shown in Figure S5, although we also must acknowledge that the cluster deviated from perfect flatness in two dimensions. Anchor Site of Pt Atoms. The same region in the HAADF image was continuously observed to elucidate the anchor site of Pt atoms on graphene. As a result, the scanning electron beam irradiation gradually formed holes. The intensity of the holes was used as an absolute reference of HAADF intensity. Figure 2A is a HAADF image after continuous observation. Pt atoms are dispersed across the whole area except for the holes in the dark black region. Also, smoky or foggy contrasts are observed. Although similar contrasts have previously been considered to represent amorphous-like hydrocarbons or contaminations, the present observations supported small islands of multilayer nanographene flakes stacked on single-layer graphene because of the uniformity of their HAADF intensities.31 A line profile along the yellow arrow, including the hole, was measured and depicted below. The profile shows the discrete values of the intensities corresponding to the hole and to the first, second, and third graphene layers, indicated by 0, 1, 2, and 3, respectively. Figure 2B shows the pseudocolor image obtained by the HAADF intensity shown below.26 The Pt atoms extracted from Figure 2A are also drawn. There are many small islands composed of nanographene flakes. Clearly, no Pt is observed in the holes (black). In addition, very few Pt atoms are observed on the terrace of the first layer (red). Most Pt atoms are located at the

the oxide on the Cu substrate. Finally, graphene was grown at 1000 °C for 180 min in a 97.5 sccm Ar, 2 sccm H2, and 0.5 sccm CH4 mixed gas flow. The sample was then rapidly cooled to room temperature. We dipped the H2SO4 and H2O2 mixed solution (H2SO4/H2O2 = 3:1) to remove the misplaced graphene grown on the back of the Cu substrate. After etching the Cu substrate with 100 mM ammonium peroxodisulfate solution for 4 h, the specimen was rinsed thoroughly with distilled water and then transferred onto a carbon-supported Cu TEM grid. The graphene was verified as a single layer by electron diffraction,22 as shown in Figure S1. Plasma Sputtering for Single-Atom Dispersion. Sputtering in plasma is well established to form thin films, but this method has also been used to synthesize nanoparticles. Several studies have also attempted to downsize the resulting nanoparticles.23−25 In this study, we used this method to disperse single atoms on a substrate, as shown in Figure S2. Gas molecules in a vacuum chamber were ionized by accelerated electrons, which were extracted from a target material under sufficient applied voltage. Then, the positively ionized atoms of the gas molecules in plasma sputtered the target material. The time schedule for the applied voltage is shown in the figure below, where ta and tr are the applied period and subsequent rest period, respectively. Although nanoparticles and thin films can be easily produced by increasing the sputtering time n, dispersed platinum single atoms were obtained under the conditions of air, V = 200 V, P = 4.5 Pa, ta = 1 s, tb = 10 s, and n = 1, respectively. We deposited a single atom, clusters, and nanoparticles of Pt on graphene to control the sputtering time n. TEM and STEM Observation. We used an aberrationcorrected TEM/STEM (Titan Cubed 60−300, FEI) to verify that the graphene was a single layer and to clarify the atomic arrangements of single Pt atoms, as shown in Figures 3, S1, and S5−S7. The TEM was operated at 80 kV with a monochromator. We also used an aberration-corrected STEM (ARM-200F, JEOL) at 80 kV to obtain the images shown in Figures 1, 2, S3, and S4. XPS Measurement. XPS measurement was conducted on a JEOL JPS-9200. We used Mg Kα X-rays with a power of 300 W. The pressure in the spectrometer was 1 × 10−7 and 8 × 10−7 Pa during the measurements. The samples were held in a stainless-steel container and then loaded into separate differently pumped fast entry chambers for more than 1 h. Then, they were loaded into the analyzer chamber. The area analyzed was 1 mm in diameter. Theoretical Calculation. The first-principles calculations were performed using OpenMX code,27 where the normconserving relativistic pseudopotentials and optimized pseudoatomic basis functions were adopted.28−30 The atomic positions in all of the considered structures shown in Figures S10−S17 are relaxed at experimental lattice constants until the forces were less than 0.0003 hartree/b without taking spin− orbit coupling into account. Computational details of theoretical calculations are shown in the Supporting Information.



RESULTS AND DISCUSSION Dispersion of Pt Atoms on Freestanding Graphene Using Plasma Sputtering. Platinum (Pt) atoms were dispersed on freestanding graphene using plasma sputtering. Details of the sample preparation are provided in the Supporting Information (see also Figures S1 and S2). Figure C

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 3. Atomic arrangements of single Pt atoms on graphene. (A) A high-resolution TEM image taken with a 1 s exposure time. Platinum was sputtered for 5 s. (B) A HAADF image taken with a scanning time of 4.98 s/frame. The same specimen was used although the observation region was different from that in (A). By measuring the intensity, as shown in Figure 2, we confirmed that the regions labeled 0, 1, 2, 3, and 4 in the image were the hole (vacuum), the first layer, the second layer, the third layer, and the fourth layer, respectively. (C−E) The enlarged views indicated by the arrows (a)−(c) in (A). (F−H) The simulated TEM images by multislice calculation based on the atomic arrangements in the bottom panels. (I−K) The atomic arrangements of carbon and Pt atoms. In the figures, the platinum atom is white and the carbon atoms of the first, second, and third layers are red, green, and blue, respectively. The connected lines between the nearest-neighbor atoms are drawn in these atomic arrangements for clarity.

step edges between layers. Figure 2C is a magnification of the yellow dashed square region in Figure 2B. In this region, there are 11 Pt atoms. All of them are located at step edges, such as notations 1−2 (first and second), 2−3 (second and third), 3− 4 (third and fourth), and 5−6 (fifth and sixth). The line profile of 1−2 along the yellow arrow, including Pt, was measured as shown in the graph below. The intensity profile shows the difference between the first and second layers on both sides of the large peak corresponding to the Pt atom. In a similar way, we identified anchor sites of all single Pt atoms in Figure 2B. As a result, 223 (93%) and 17 (7%) of the 240 Pt atoms were anchored to the edge and terrace, respectively. The edges could be classified into two types, free edges of the hole in the first layer and step edges between the layers, and there were 3 (1%) and 220 (92%) edges of either type, respectively. Considering that the holes in the first layer were intentionally poured as an absolute reference of HAADF intensity, we conclude that the dispersed Pt atoms are anchored dominantly at the step edges of nanographene flakes. The extent to which the electron beam damaged the sample could not be clearly determined, but step-edge structure was also observed before the hole was formed, as shown in Figure S4. Atomic Arrangement of Single Pt Atoms. Highresolution imaging using TEM was performed to clarify the atomic arrangement of Pt atoms adsorbed on graphene. Although the atomic arrangement above the third layers was hardly realized due to the complicated contrast and atomic migration, typical examples could be clarified. Figure 3A shows an example of a TEM image. There is a wide area of terrace with hexagonal lattice. This area was confirmed as the first layer, as shown in Figure S6. Areas with complicated patterns resembling contaminants are also observed. Figure 3B shows a HAADF image observed in the same specimen. There are many bright spots of Pt atoms anchored at the step edge of

nanographene, as shown in Figure 2. Two-dimensional clusters are also observed. By comparing Figure 3A with Figure 3B, it becomes clear that the area with the complicated pattern in Figure 3A is a laminated structure of nanographene flakes containing most of the Pt atoms. It can also be seen in both images that few or no Pt atoms are observed on the terrace site of the first layer. The three panels of Figure 3C−E are enlarged views of the three regions indicated by arrows (a)−(c) in Figure 3A. The simulated TEM images in Figure 3F−H are in good agreement with the experimental images in the corresponding panels above. The bottom panels of Figure 3I−K show the atomic arrangements of carbon and Pt atoms. The analysis method is shown in Figure S6. In Figure 3I, two Pt atoms sit near the bridge site between two carbon atoms. In Figure 3J, two Pt atoms are observed, one in the divacancy of two carbon atoms on the terrace and the other within the five- and six-membered rings at the free edge of the hole. In Figure 3K, a Pt atom can be seen within the five- and six-membered rings at the step edge between the first and second layers. The first layer showed a pristine hexagonal lattice, whereas the second layer had a polycrystalline structure, e.g., five- and seven-membered rings. The other examples at the step edge are shown in Figure S7. Single Pt atoms are dispersed across the whole grid, but only a small region can be observed using TEM. Therefore, we conducted XPS measurements to obtain information on the entire dispersed single Pt atoms. XPS Measurement of the Pt Atoms Supported on Graphene. Figure 4 shows XPS spectra of Pt 4f peaks with different sputtering times on graphene. The detailed procedures of XPS measurement are provided in the Supporting Information, as shown in Figures S8 and S9. The binding energy of the spectra is referenced to the C 1s binding energy of sp2 bonding of graphene at 284.5 eV.32 We successfully observed the peaks even when the sputtering D

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 4. XPS spectra of Pt 4f core level on graphene. (A) Series of spectra at different sputtering times. These solid colored lines represent the experimental data after background subtraction and charge collection. We separated the single-atom component (dashed lines) and the other components (dotted lines) from the raw data. (B) The atom ratio of the single-atom component and other components. The atom ratios of single atoms were obtained by dividing the number of single atoms from the total amount of Pt from each HAADF image. The atom ratios of single atoms were also obtained from the peak area ratio of each after dividing the XPS spectrum into single-atom component and other components. (C) Peak positions of the single-atom component and the other components at each sputtering time. The peak position of the single-atom component was almost constant.

atom, and we extracted this component from the other spectra with different sputtering times, as shown in Figure 4A because the single Pt atoms are certainly observed, as shown in Figure 1. The ratio of the single-atom component was plotted to compare the peak intensity and peak position between the single atom and the others in Figure 4B,C. The approximated curve in Figure 4B was in good agreement with the ratio of a single atom, which was calculated from the HAADF observations. This indicates that the XPS results, which provide averaged information across a large area, reflect the local structure of Pt on graphene. The peak position of the single-atom component in Figure 4C was 73.0 ± 0.4 eV at each sputtering time. The peak position of the other components began to decrease from 30 s and became close to that of bulk Pt. This result indicated that the 3D structure was mainly formed at around 30 s. The details of the peak fitting method are discussed in the Supporting Information, including with respect to other parameters, such as the full width at halfmaximum in Table S1, the ratio of the peak intensity and the interval of the peak position between Pt 4f7/2 and 4f5/2. Such a positive peak shift is often thought to be due to binding with oxygen. Some previous papers have interpreted the peak around 73.0 eV as a signal of Pt−O bonding.37 However, we have not obtained data correlated with Pt and oxygen in EDX and XPS measurement, as shown in Figures S3 and S8. X-ray absorption fine structure measurement of electron transfer by single atomization of Pt on a carbon material has been

time was 1 s. In addition, a higher binding energy peak was observed for shorter sputtering times. The peak positions of Pt 4f7/2 and 4f5/2 spectra at 1 s were 73.0 ± 0.4 and 76.2 ± 0.4 eV, respectively. The peaks were located 2.0 eV higher than the Pt 4f peak of single-crystal Pt, which is known to appear at ∼71.0 eV.33 In our XPS measurement of Pt(111), the peak position of Pt 4f7/2 was 71.0 eV, as shown in Figure S9. Positive corelevel shifts have been observed for metal nanoparticles on other substrates, and these were often interpreted in terms of final state effects. The higher energy shift of Pt 4f has also been observed on carbon material composites.34,35 In our experiments, we observed a larger peak shift (∼2.0 eV) compared to that in the previous study using a carbon support (0.6−1.2 eV). As already mentioned, we prepared a 99% single-atom dispersion of Pt atoms on graphene at a sputtering time of 1 s and we also confirmed that the dispersed Pt atoms were anchored dominantly at the step edges of nanographene flakes. Although the +2 eV core-level shift of Pt 4f has been observed with single Pt atoms on carbon nitride, oxidized Si carbon nitride,36 and oxidized Si,15 we revealed that the single Pt atoms on the step edges of CVD-grown graphene observed in the experiments show a large core-level shift. The peak position of Pt 4f is shifted with longer sputtering time, and the peak position of the 60 s sample was 71.2 eV. It was in almost the same position as that of the Pt(111) single crystal.33 Here, we fitted the spectra of 1 s by a pair of Gaussians. These fitted curves were defined as a component of a single E

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 5. Electronic states of single Pt atoms on graphene. Calculated isosurface of charge density for a single Pt atom (A) on the terrace of graphene sheet and (C) anchored at the step edge, where an isovalue of 0.12 e/b3 was used. (B) The local density of states (LDOS) of Pt d orbitals and the two nearest-neighboring C atoms in the model (A) and (D) LDOS of Pt 5d orbitals and three neighboring C atoms in the model (C), where components in d orbitals of the Pt atom that are strongly hybridized with C atoms are indicated.

formation of bonding between Pt and C atoms largely increases the binding energies of Pt 4f states. Motivated by the atomic structure shown in Figure 3K obtained from the analysis of the experimental TEM image, we considered a single Pt atom anchored at step edges of graphene flake(s) on a graphene sheet, as shown in Figure S16, where the Pt atom is coordinated by three carbon atoms. The calculated binding energies of the Pt 4f7/2 states for the models in Figure S16 were found to be within 72.06−72.26 eV, which are in reasonably good agreement with the experimental value of 73.0 ± 0.4 eV, whereas the insulator treatment39 gives a better agreement, as listed in Table S3. The agreement in the binding energies of Pt 4f states between XPS measurement and the DFT calculations provides solid support for the formation of single Pt atoms at the edges in the samples. To elucidate the physical origin of the large shift in binding energies of Pt 4f states, we focus on the population of 5d orbitals of Pt atom, as listed in Table S2, which summarizes all computational results in this study, since the population of the 6s orbital of Pt atom might not be directly relevant to the binding energies of Pt 4f states due to the delocalization of the 6s orbital. It is found that the populations for the ground and excited states of the step-edge model (Pt1flakeA on graphene) are 8.47 and 9.22, respectively, which are smaller than those (9.14 and 9.65) of the physisorption model of a Pt atom (Pt1-graphene). The reduction of populations in the 5d orbitals implies considerable charge transfer from Pt atom to C atoms in the step-edge model, which might be responsible for the large shift of binding energies. An evident correlation between the population of the Pt 5d orbitals and the binding energies of 4f states can be confirmed in Figure S18. The correlation can be understood in such a way that the large population of the Pt 5d orbitals destabilizes and stabilizes electrons and the created core hole in the 4f states in the ground and excited states, respectively. As shown in Figure 5A,C, the formation of bonds between Pt atom and C atoms can be seen in isosurface maps

reported, but it cannot be distinguished whether this transfer is due to carbon or oxygen.18 We verified and analyzed the relationships of Pt−C bonds and the core-level shift of Pt 4f states in our experimental conditions by theoretical calculation. First-Principles Calculation of Absolute Core-Level Binding Energy. To support the large binding energies of Pt 4f states of a single Pt atom observed in our XPS measurements and elucidate the physical origin, we performed theoretical calculations on the absolute binding energies of core levels of Pt 4f states on the basis of density functional theories (DFTs) within generalized gradient approximation38 in combination with a recently developed method to directly compare calculated binding energies of core levels with experimental data in the absolute scale.39 All geometrical structures we investigated were fully optimized. The computational details and all calculations are presented in the Supporting Information. The accuracy of our method was confirmed by calculating the binding energies of Pt 4f7/2 states of the face-centered cubic (FCC Pt) and PtO2 bulks, which are found to be 71.08 and 74.31 eV (Figure S10), respectively, well compared to the experimental values of 71.033 and 74.1 eV.40 With the computational method, we first examined binding energies of Pt 4f states in a single Pt atom physisorbed on the bridge site in graphene sheet, as shown in Figure S11. It turned out that the calculated binding energy for the Pt 4f7/2 state (69.01 eV) is much smaller than that observed in our XPS measurements (73.0 ± 0.4 eV), suggesting that the physisorption of a Pt atom on graphene can be excluded from our further consideration. As model cases of a chemisorbed Pt atom, we calculated a single Pt atom chemisorbed at a mono and divacancy in graphene sheet, as shown in Figure S12, and obtained the binding energies of 72.77 and 71.74 eV for the Pt 4f7/2 states, respectively. The binding energies are larger than those of FCC Pt and well compared to the XPS measurements. Compared to the case of the physisorption of a Pt atom, it would be concluded that the F

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C of charge density of the step-edge model (Pt1flakeA on graphene), whereas for the physisorption model (Pt1graphene), no evident bonds can be seen. The chemical bonds between Pt atom and C atoms in the step-edge model can also be confirmed by the local density of states (LDOS), as shown in Figure 5D. The LDOSs of neighboring C atoms significantly overlap with those of Pt 5d orbitals in the stepedge model (Pt1 flakes on graphene), further supporting the formation of chemical bonds between Pt atom and C atoms, whereas the neighboring C atoms do not largely contribute to local electronic states consisting of Pt 5d orbitals near the Fermi level in the physisorption model (Pt1-graphene), as shown in Figure 5B. The Pt 5dxy orbital in the step-edge model plays a crucial role in forming the chemical bonds because of directional lobes in the orbital toward C atoms, resulting in large splitting of bonding and antibonding states. As already discussed, the total population of Pt d orbitals is 9.14 in the physisorption model of a Pt atom (Pt1-graphene), which can be regarded as nearly fully occupied orbitals, more than 90%. On the other hand, the C atom has four valence electrons nominally. Therefore, the populations of the bonding states are largely contributed by electrons of Pt 5d orbitals more than C atoms when the Pt 5d orbitals form the chemical bonds with the neighboring C atoms. This is a kind of donation of electrons from the Pt atom to the neighboring C atoms, leading to reduction of populations of Pt 5d orbitals. In fact, the population of Pt 5dxy-orbital in the step-edge model is found to be 1.28, which is much smaller than those (1.67− 1.90) of the other d orbitals, as listed in Figure 6A. Thus, it is concluded that the formation of bonds reduces the populations of Pt 5d orbital due to the unbalance of original populations in Pt 5d orbitals and valence orbitals of carbon atom. The

reduction of populations in Pt 5d orbitals would be a dominant mechanism why a single Pt atom anchored at step edges, as shown in Figure 3, has the large binding energy of Pt 4f states. As shown in Figure 6B, the binding energies of Pt 4f7/2 states against the number of Pt−C bonds further support the general trend that forming chemical bonds to carbon atoms increases the binding energies of Pt 4f7/2 states. We further investigated Pt clusters placed on graphene sheet and anchored at a step edge, as shown in Figure S17, where the geometrical structures were fully optimized starting from 3D and 2D structures, respectively. Since we did not explore a wide variety of other initial structures, the obtained structures shown in Figure S17 might be metastable structures. Nevertheless, it is interesting to see that the planar structure anchored at a step edge was obtained even after geometry optimization, which seems to be consistent with the experimental observation that 2D planar nanoclusters of Pt with a diameter of 1−2 nm are obtained just before the formation of 3D nanoparticle. A similar trend on stability of planar structures for small Pt clusters has also been reported by another DFT calculation.41 We see that Pt atoms in the cluster tend to have smaller binding energies of Pt 4f states except for the anchored Pt atom, as listed in Table S2. The shift to smaller binding energies of Pt 4f states is consistent with the experimental observation.



CONCLUSIONS AND DISCUSSION In this paper, we revealed experimentally and theoretically the atomic structure and local electronic states of single Pt atoms dispersed on graphene using plasma sputtering. First, we reported the achievement of a uniform dispersion of single Pt atoms on graphene by simple plasma sputtering. It was confirmed that our method could disperse single Pt atoms over an entire 3 mm graphene-transferred TEM grid. In addition, by increasing the sputtering time, 2D clusters were found between single atoms and 3D nanoclusters. Next, aberration-corrected TEM/STEM imaging and multislice calculation were used in a structure analysis to reveal the atomic arrangement of Pt single atoms on graphene. We also conducted XPS measurements using the same sample and observed a large core-level shift for single Pt atoms. Finally, theoretical calculations based on DFT, recently developed as a general method for obtaining absolute binding energies of core levels in metals and insulators,39 were in good agreement with the experimental results. We also elucidated the physical origin of the large binding energy of the Pt 4f state of a single Pt atom. As shown in Figure S6, TEM analysis in which carbon atoms are substituted with oxygen atoms shows that Pt atoms are bonded to three carbon atoms as Pt−C. However, the TEM analysis does not provide information on the whole sample. Therefore, we conducted XPS measurements to obtain information on all of the dispersed single Pt atoms. In addition, we performed DFT calculations on the absolute binding energies of core levels of Pt 4f states to determine whether the three-coordination structure of Pt−C at the step edges observed in the experiment gives a large core-level shift. The calculation results strongly support the experimental results of TEM and XPS. The DFT calculations were performed at zero temperature, which means the geometry optimization was performed without considering any temperature effect, whereas in the experiment, the data was obtained at a finite temperature above absolute zero. Since the calculation results are in good agreement with the experiments,

Figure 6. Populations of d orbitals and binding energy of 4f states in single Pt atoms. (A) Decomposed occupation numbers of Pt 5d orbitals in the ground state via the Mulliken population analysis. In the atomic picture, the 5d orbitals are expected to be occupied by nine electrons. (B) Relationship between the binding energy of the Pt 4f7/2 state and the numbers of Pt−C bonds, where the excited states were treated as metal in the calculations of the binding energy. The number of bonds is counted solely on the basis of the consideration of three sp2 bonds per C atom in graphene. The missing C−C or C−H bond surrounding the Pt atoms is counted as one Pt−C bond. G

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

(3) Mitchell, S.; Thomas, J. M.; Pérez-Ramírez, J. Single Atom Catalysis. Catal. Sci. Technol. 2017, 7, 4248−4249 see references therein. . (4) Tao, F.; Dag, S.; Wang, L.-W.; Liu, Z.; Butcher, D. R.; Bluhm, H.; Salmeron, M.; Somorjai, G. A. Break-Up of Stepped Platinum Catalyst Surfaces by High CO Coverage. Science 2010, 327, 850−853. (5) Qiao, B.; Wang, A.; Yang, X.; Allard, L. F.; Jiang, Z.; Cui, Y.; Liu, J.; Li, J.; Zhang, T. Single-Atom Catalysis of CO Oxidation Using Pt1/ FeOx. Nat. Chem. 2011, 3, 634−641. (6) Jones, J.; Xiong, H.; DeLaRiva, A. T.; Peterson, E. J.; Pham, H.; Challa, S. R.; Qi, G.; Oh, S.; Wiebenga, M. H.; Pereira Hernandez, X. I.; et al. Thermally Stable Single-Atom Platinum-on-Ceria Catalysts via Atom Trapping. Science 2016, 353, 150−154. (7) Ding, K.; Gulec, A.; Johnson, A. M.; Schweitzer, N. M.; Stucky, G. D.; Marks, L. D.; Stair, P. C. Identification of Active Sites in CO Oxidation and Water-Gas Shift over Supported Pt Catalysts. Science 2015, 350, 189−192. (8) Suenaga, K.; Okazaki, T.; Okunishi, E.; Matsumura, S. Detection of Photons Emitted from Single Erbium Atoms in Energy-Dispersive X-Ray Spectroscopy. Nat. Photonics 2012, 6, 545−548. (9) Ramasse, Q. M.; Seabourne, C. R.; Kepaptsoglou, D.-M.; Zan, R.; Bangert, U.; Scott, A. J. Probing the Bonding and Electronic Structure of Single Atom Dopants in Graphene with Electron Energy Loss Spectroscopy. Nano Lett. 2013, 13, 4989−4995. (10) Lin, Y. C.; Teng, P. Y.; Yeh, C. H.; Koshino, M.; Chiu, P. W.; Suenaga, K. Structural and Chemical Dynamics of Pyridinic-Nitrogen Defects in Graphene. Nano Lett. 2015, 15, 7408−7413. (11) Stambula, S.; Gauquelin, N.; Bugnet, M.; Gorantla, S.; Turner, S.; Sun, S.; Liu, J.; Zhang, G.; Sun, X.; Botton, G. A. Chemical Structure of Nitrogen-Doped Graphene with Single Platinum Atoms and Atomic Clusters as a Platform for the PEMFC Electrode. J. Phys. Chem. C 2014, 118, 3890−3900. (12) Peters, S.; Peredkov, S.; Neeb, M.; Eberhardt, W.; Al-Hada, M. Size-Dependent XPS Spectra of Small Supported Au-Clusters. Surf. Sci. 2013, 608, 129−134. (13) Kim, K. S.; Winograd, N. X-Ray Photoelectron Spectroscopic Binding Energy Shifts Due to Matrix in Alloys and Small Supported Metal Particles. Chem. Phys. Lett. 1975, 30, 91−95. (14) Mason, M. Electronic Structure of Supported Small Metal Clusters. Phys. Rev. B 1983, 27, 748−762. (15) Eberhardt, W.; Fayet, P.; Cox, D. M.; Fu, Z.; Kaldor, A.; Sherwood, R.; Sondericker, D. Photoemission from Mass-Selected Monodispersed Pt Clusters. Phys. Rev. Lett. 1990, 64, 780−783. (16) Nakamura, J.; Kondo, T. Support Effects of Carbon on Pt Catalysts. Top. Catal. 2013, 56, 1560−1568. (17) Zhang, X.; Lu, Z.; Xu, G.; Wang, T.; Ma, D.; Yang, Z.; Yang, L. Single Pt Atom Stabilized on Nitrogen Doped Graphene: CO Oxidation Readily Occurs via the Tri-Molecular Eley−Rideal Mechanism. Phys. Chem. Chem. Phys. 2015, 17, 20006−20013. (18) Liu, J.; Jiao, M.; Lu, L.; Barkholtz, H. M.; Li, Y.; Wang, Y.; Jiang, L.; Wu, Z.; Liu, D.; Zhuang, L.; et al. High Performance Platinum Single Atom Electrocatalyst for Oxygen Reduction Reaction. Nat. Commun. 2017, 8, No. 15938. (19) Koitaya, T.; Yamamoto, S.; Shiozawa, Y.; Takeuchi, K.; Liu, R. Y.; Mukai, K.; Yoshimoto, S.; Akikubo, K.; Matsuda, I.; Yoshinobu, J. Real-Time Observation of Reaction Processes of CO2 on Cu(997) by Ambient-Pressure X-Ray Photoelectron Spectroscopy. Top. Catal. 2016, 59, 526−531. (20) Liu, J. Catalysis by Supported Single Metal Atoms. ACS Catal. 2017, 7, 34−59. (21) Yamazaki, K.; Maehara, Y.; Gohara, K. Characterization of TEM Moiré Patterns Originating from Two Monolayer Graphenes Grown on the Front and Back Sides of a Copper Substrate by CVD Method. J. Phys. Soc. Jpn. 2018, 87, No. 061011. (22) Meyer, J. C.; Geim, A. K.; Katsnelson, M. I.; Novoselov, K. S.; Booth, T. J.; Roth, S. The Structure of Suspended Graphene Sheets. Nature 2007, 446, 60−63. (23) Torimoto, T.; Okazaki, K.; Kiyama, T.; Hirahara, K.; Tanaka, N.; Kuwabata, S. Sputter Deposition onto Ionic Liquids: Simple and

we think that the temperature effect may change only slightly the structural features, leading to that the conclusions remain unchanged. Single-atom dispersion on graphene by plasma sputtering is possible not only for Pt but also for other elements, such as Au, Ir, W, Ta, Pd, Ru, Ni, Fe, etc. The new findings and methods in this paper are surely expected to open up many new fields, including the single-atom catalysis in the future.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.8b04529. Diffraction pattern in the observed area, EDX spectrum in a vacuum, illustration of plasma sputtering, identification of isolated single Pt atoms, atomic arrangement of single Pt atom at step edge between the first and second layers, another XPS spectra, details of theoretical calculations (PDF) Two-dimensional planar clusters of Pt on graphene (Movie S1) (AVI)



AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (K.Y.). *E-mail: [email protected] (K.G.). ORCID

Kenji Yamazaki: 0000-0002-1736-5845 Yosuke Maehara: 0000-0003-0199-4978 Chi-Cheng Lee: 0000-0002-3895-9802 Jun Yoshinobu: 0000-0001-7774-8701 Kazutoshi Gohara: 0000-0002-0730-4969 Author Contributions

K.Y., K.G., and J.Y. conceived the idea and designed the study. K.Y. and Y.M. performed the experiments of sample preparations, TEM/STEM observations, and XPS measurements. Y.M. and K.G. analyzed TEM/STEM data. K.Y., K.G., and J.Y. analyzed XPS data. C.-C.L. and T.O. performed the theoretical calculations. K.Y. and K.G. and C.-C.L. and T.O. wrote the experimental part and theoretical part, respectively. All authors contributed and commented on the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was partly supported by a JSPS Grant-in-Aid for Scientific Research (grant numbers: 26105009, 15K17642, and 17H05212). A part of this research was supported by the Nanotechnology Platform Program of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan. T.O. and C.-C.L. were partly supported by Priority Issue (creation of new functional devices and highperformance materials to support next-generation industries) to be tackled by using Post ‘K’ Computer, MEXT, Japan.



REFERENCES

(1) Flytzani-stephanopoulos, M.; Gates, B. C. Atomically Dispersed Supported Metal Catalysts. Annu. Rev. Chem. Biomol. Eng. 2012, 3, 545−574. (2) Liang, S.; Hao, C.; Shi, Y. The Power of Single-Atom Catalysis. Chem. Cat. Chem. 2015, 7, 2559−2567. H

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C Clean Synthesis of Highly Dispersed Ultrafine Metal Nanoparticles. Appl. Phys. Lett. 2006, 89, No. 243117. (24) Jeff, R. C.; Yun, M.; Ramalingam, B.; Lee, B.; Misra, V.; Triplett, G.; Gangopadhyay, S. Charge Storage Characteristics of Ultra-Small Pt Nanoparticle Embedded GaAs Based Non-Volatile Memory. Appl. Phys. Lett. 2011, 99, No. 072104. (25) Ramalingam, B.; Mukherjee, S.; Mathai, C. J.; Gangopadhyay, K.; Gangopadhyay, S. Sub-2 Nm Size and Density Tunable Platinum Nanoparticles Using Room Temperature Tilted-Target Sputtering. Nanotechnology 2013, 24, No. 205602. (26) Dempster, A. P.; Laird, N. M.; Rubin, D. B. Maximum Likelihood from Incomplete Data via the EM Algorithm. J. R. Stat. Soc., Ser. B 1977, 39, 1−38. (27) The code, OpenMX, pseudoatomic basis functions, and pseudopotentials are available at http://www.openmx-square.org. (28) Theurich, G.; Hill, N. A. Self-Consistent Treatment of SpinOrbit Coupling in Solids Using Relativistic Fully Separable Ab Initio Pseudopotentials. Phys. Rev. B 2001, 64, No. 073106. (29) Morrison, I.; Bylander, D. M.; Kleinman, L. Nonlocal Hermitian Norm-Conserving Vanderbilt Pseudopotential. Phys. Rev. B 1993, 47, 6728−6731. (30) Ozaki, T. Variationally Optimized Atomic Orbitals for LargeScale Electronic Structures. Phys. Rev. B 2003, 67, No. 55108. (31) Yamashita, S.; Koshiya, S.; Ishizuka, K.; Kimoto, K. Quantitative Annular Dark-Field Imaging of Single-Layer Graphene. Microscopy 2015, 64, 143−150. (32) Siokou, A.; Ravani, F.; Karakalos, S.; Frank, O.; Kalbac, M.; Galiotis, C. Surface Refinement and Electronic Properties of Graphene Layers Grown on Copper Substrate: An XPS, UPS and EELS Study. Appl. Surf. Sci. 2011, 257, 9785−9790. (33) Cameron, S. D.; Dwyer, D. J. Surface Core Level Shifts in Pt3Ti(111). Surf. Sci. 1986, 176, L857−L862. (34) Siburian, R.; Kondo, T.; Nakamura, J. Size Control to a SubNanometer Scale in Platinum Catalysts on Graphene. J. Phys. Chem. C 2013, 117, 3635−3645. (35) Wei, H.; Huang, K.; Wang, D.; Zhang, R.; Ge, B.; Ma, J.; Wen, B.; Zhang, S.; Li, Q.; Lei, M.; et al. Iced Photochemical Reduction to Synthesize Atomically Dispersed Metals by Suppressing Nanocrystal Growth. Nat. Commun. 2017, 8, No. 1490. (36) Chen, Z.; Mitchell, S.; Vorobyeva, E.; Leary, R. K.; Hauert, R.; Furnival, T.; Ramasse, Q. M.; Thomas, J. M.; Midgley, P. A.; Dontsova, D.; et al. Stabilization of Single Metal Atoms on Graphitic Carbon Nitride. Adv. Funct. Mater. 2017, 27, No. 1605785. (37) Dvořaḱ , F.; Camellone, M. F.; Tovt, A.; Tran, N.-D.; Negreiros, F. R.; Vorokhta, M.; Skála, T.; Matolínová, I.; Mysliveček, J.; Matolín, V.; et al. Creating Single-Atom Pt-Ceria Catalysts by Surface Step Decoration. Nat. Commun. 2016, 7, No. 10801. (38) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868. (39) Ozaki, T.; Lee, C.-C. Absolute Binding Energies of Core Levels in Solids from First Principles. Phys. Rev. Lett. 2017, 118, No. 026401. (40) Bancroft, G. M.; Adams, I.; Coatsworth, L. L.; Bennewitz, C. D.; Brown, J. D.; Westwood, W. D. ESCA Study of Sputtered Platinum Films. Anal. Chem. 1975, 47, 586−588. (41) Xiao, L.; Wang, L. Structures of Platinum Clusters: Planar or Spherical. J. Phys. Chem. A 2004, 108, 8605−8614.

I

DOI: 10.1021/acs.jpcc.8b04529 J. Phys. Chem. C XXXX, XXX, XXX−XXX