Letter pubs.acs.org/NanoLett
Direct Imaging of Pt Single Atoms Adsorbed on TiO2 (110) Surfaces Teng-Yuan Chang,† Yusuke Tanaka,‡ Ryo Ishikawa,† Kazuaki Toyoura,‡ Katsuyuki Matsunaga,‡,§ Yuichi Ikuhara,†,§ and Naoya Shibata*,†,∥ †
Institute of Engineering Innovation, School of Engineering, the University of Tokyo, Yayoi 2-11-16, Bunkyo-ku, Tokyo 113-8656, Japan ‡ Department of Materials Science and Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan § Nanostructures Research Laboratory, Japan Fine Ceramics Center, 2-4-1 Mutsuno, Atsuta-ku, Nagoya 456-8587, Japan ∥ Japan Science and Technology Agency, PRESTO, 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan S Supporting Information *
ABSTRACT: Noble metal nanoparticles (e.g., gold and platinum) supported on TiO2 surfaces are utilized in many technological applications such as heterogeneous catalysts. To fully understand their enhanced catalytic activity, it is essential to unravel the interfacial interaction between the metal atoms and TiO2 surfaces at the level of atomic dimensions. However, it has been extremely difficult to directly characterize the atomic-scale structures that result when individual metal atoms are adsorbed on the TiO2 surfaces. Here, we show direct atomic-resolution images of individual Pt atoms adsorbed on TiO2 (110) surfaces using aberration-corrected scanning transmission electron microscopy. Subangstrom spatial resolution enables us to identify five different Pt atom adsorption sites on the TiO2 (110) surface. Combining this with systematic density functional theory calculations reveals that the most favorable Pt adsorption sites are on vacancy sites of basal oxygen atoms that are located in subsurface positions relative to the top surface bridging oxygen atoms. KEYWORDS: Heterogeneous catalyst, metal-oxide interaction, platinum, titanium oxide, single atom imaging, scanning transmission electron microscopy
T
adsorption.6,7 Enhanced bonding with oxygen vacancy sites has thus been considered to be of primal importance to account for Pt atom anchoring and subsequent nanoparticle nucleation and growth on TiO2 surfaces. While their importance is wellrecognized, previous studies focused only on the topmost Obr vacancies and not on other surface oxygen vacancies such as three-fold coordinated basal oxygen (Oba) vacancies. This may be partly due to the following two reasons: (1) the vacancy formation energy of Oba atoms is predicted to be much higher than that of Obr atoms,11 and (2) point defects on the TiO2 (110) surfaces imaged by STM have always been attributed to the Obr vacancies.9,10 However, the limited spatial resolution and complex image formation mechanism of STM may have
o fully understand the unique chemical and physical properties of metal nanoparticles dispersed on oxide surfaces, detailed atomic-scale knowledge of their structure and the nanoparticle/oxide-surface interactions is crucial.1−3 One system that has been studied extensively in this regard is Pt/ TiO2, which serves as a typical heterogeneous catalyst for CO oxidation. It has been reported that the stabilization of Pt atoms on TiO2 (110) surfaces, which have been intensively studied as model surfaces of TiO2, depends on their adsorption site and the formation of surface point defects such as oxygen vacancies.4−8 The topmost oxygen atoms, which are coordinated with the two Ti atoms below (so-called bridging oxygen, Obr), are regarded as the most favorable sites for vacancy formation and their vacancies have been experimentally observed by scanning tunneling microscopy (STM).9,10 Previous first-principles calculations predicted that these Obr vacancies are the most energetically favorable sites for Pt atom © 2013 American Chemical Society
Received: September 20, 2013 Revised: November 21, 2013 Published: December 19, 2013 134
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
Figure 1. Structural model illustrating the eight different adsorption sites on TiO2 (110) surface and HAADF STEM image of TiO2 observed along [110] projection. (a) The surface atomic structure of TiO2 (110) and possible Pt atom adsorption sites. The green balls show the eight different Pt atom adsorption sites; #1 (H1), #2 (atop Oba), #3 (H2), #4 (atop Ti6c), #5 (atop Obr), #6 (atop Ti5c), #7 (Obr vacancy), and #8 (Oba vacancy). (b) Atomic-resolution HAADF STEM image of TiO2 (110) surface observed along perpendicular [110] direction. The two bright contrast spots correspond to the Ti−O and Ti-only columns, respectively.
Figure 2. HAADF STEM images of Pt single atoms on TiO2 (110) surface. (a) Atomic-resolution HAADF STEM image of Pt supported TiO2 (110) surface. (b-f) The magnified views of Pt single atoms located on different surface sites found in a. As seen in the images, we can clearly distinguish five different Pt single atom attachment sites on the TiO2 (110) surface unit structure. The magnified images were low-pass filtered to remove noises.
hindered the detailed determination of surface oxygen vacancies. Diebold et al. reported two types of defect-originated image features on TiO2 (110) surfaces.10 They assigned the bright spots on the dark Obr rows to be the Obr vacancies, but the dark spots on the bright 5-fold Ti rows were not clearly identified. The presence of other oxygen vacancies and their effect on metal atom anchoring are thus still a matter of conjecture. In the present study, we use aberration-corrected scanning transmission electron microscopy (STEM) to directly image individual Pt atoms adsorbed on TiO2 (110) surfaces. Highangle annular dark-field (HAADF) STEM12 has proved to be an excellent technique for directly imaging individual metal atoms supported by TiO2 surfaces.13−16 Here, we show that the detailed Pt adsorption sites within the unit structure of the TiO2 (110) surface can be identified by HAADF STEM. The reason for the enhanced adsorption to these specific sites is discussed based on systematic density functional theory (DFT) calculations. We highlight the unexpected importance of Oba
vacancies, whose strong bonding to the Pt atoms may dictate the interaction between Pt atoms and TiO2 (110) surfaces. Figure 1a illustrates the atomic structure of TiO2 (110) surface and high-symmetry Pt adsorption sites (shown by green balls). Here, we extend the unit surface structure of TiO2 (110) two times to visualize the individual adsorption sites more clearly. The green balls labeled from 1 to 6 represent Pt atom adsorption sites on a stoichiometric TiO2 (110) surface. Oxygen vacancies are illustrated as translucent pink balls and the green balls labeled 7 and 8 thus represent Pt atom adsorption sites on the Obr and Oba vacancy sites, respectively. It should be noted here that the #7 and #8 sites appear to be the same as the #5 and #2 sites in the [110] projection but now with vacancies, and the Pt sites in Figure 1a have been centered to the vacancy sites, despite the potential for relaxation that accompanies the vacancy formation. Figure 1b shows a HAADF STEM image of a pristine TiO2 (110) surface observed along the [110] projection. The thin, electron transparent TiO2 specimen was annealed in air at 973 K for 1 h to produce an 135
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
Table 1. Binding Energies (in eV) of a Pt Atom at Each Adsorption Site on TiO2 (110) Surface Predicted by DFT Calculations site
#1
#2
#3
#4
#5
#6
#7
#8
notation Eb (eV)
H1 2.72
Oba 2.31
H2 1.85
Ti6c 1.76
Obr 1.47
Ti5c 1.31
Obr vacancy 3.64
Oba vacancy 3.82
Figure 3. DFT relaxed atomic structures, related HAADF STEM image simulations and experimental HAADF STEM images of Pt atoms on TiO2 (110) surfaces. (a−f) Pt single atom on site #1 through site #8 defined in Figure 1a. Top panels show the relaxed atomic structures predicted by DFT calculations. Middle panels show simulated HAADF STEM images based on the DFT relaxed structures. Bottom panels show corresponding experimental HAADF STEM images reproduced from Figure 2. Even after allowing structural relaxation, we cannot distinguish “#3, #4”, and “#5, #6, #7”, since HAADF STEM images are observed along [110] projection. However, due to the strong structural relaxation found in #8, we can clearly distinguish #8 from #2. Thus, we can distinguish five different Pt adsorption sites by the HAADF STEM images along [110] projection.
atomically flat TiO2 (110) surface as per previous studies.14,17−20 As indicated by the arrows, two different intensities of bright spots can be observed in the image. These two atomiccolumns are the so-called Ti−O column and Ti-only column, reflecting the atomic species present in each atomic column. Because of a greater atomic column occupancy along the [110] direction, the Ti−O columns have higher intensity than the Tionly columns. The oxygen columns in the troughs between Ticontaining atomic columns cannot be imaged because the signal from the light element columns is very weak at the HAADF detection angles. Figure 2a shows a typical HAADF STEM image of Pt atoms on a TiO2 (110) surface, which was obtained soon after highpurity Pt atoms were directly evaporated onto the TiO2 (110) surface. We have confirmed the evaporation of Pt atoms under the present experimental condition through large area STEMEDS analysis shown in the Supporting Information. The detailed sample preparation procedures are given in the Method Section in the Supporting Information. To minimize any effect of electron beam incidence on the Pt structures, we use a low-dose imaging condition with a probe current of about 10 pA and probe size of about 0.8 Å. We have checked that there is no obvious beam damage after several scans under the present imaging condition and thus consider our images to reflect stable Pt atom structures on the TiO2 (110) surface. Further, we tried to perform atomic-resolution STEM spectroscopy such as electron energy-loss spectroscopy and X-ray spectroscopy to obtain spectroscopic evidence of Pt in conjunction with the atomic-resolution HAADF STEM images. However, under the present low-dose condition such spectroscopic signals are too weak to be meaningful data. On
the other hand, the Z-contrast image characteristic of HAADF makes Pt (Z, 78) atoms stand out as brighter spots above the TiO2 support (Ti, Z = 22; O, Z = 8), and many Pt atoms are found to be single, isolated atoms. We found that the Pt single atoms do not always attach to the same sites but rather to several different sites on the TiO2 (110) surface. After extensively measuring the Pt single atom adsorption sites from several HAADF STEM images, we find that the stable Pt adsorption sites can be categorized into five sites. As examples, five magnified views from the areas enclosed by white squares in Figure 2a are shown in Figure 2b−f with white arrows indicating the position of the Pt atoms in each image. These images clearly show that Pt atoms stably attach to the different surface sites of TiO2 (110), even though these images are in projection along the [110] direction. Figure 2b shows a Pt atom locating near on a hollow site adjacent to a Ti−O column, suggesting the Pt atom sits on the #1 site (hollow site; H1) but with slight structural relaxation toward the near-by Ti−O column. Figure 2c shows a Pt atom locating near on an oxygen column adjacent to a Ti-only column, corresponding to the #2 or #8 sites (atop Oba or its vacancy sites) but evident structural relaxation toward near-by Ti-only column can be observed. Figure 2d shows a Pt atom locating on top of a Ti-only column, corresponding to the #3 or #4 sites (hollow site 2; H2 or atop Ti6c site). Figure 2e shows a Pt atom locating on top of a Ti− O column, corresponding to the #5, #6, or #7 sites (atop Obr or atop Ti5c or Obr vacancy sites). Figure 2f shows a Pt adsorption site locating near both the Ti−O and Ti-only columns, corresponding to the #8 site (atop Oba vacancy site) after significant structural relaxation along [001] direction, which will be discussed later. 136
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
For complementary information on the experimentally identified adsorption sites of Pt atoms on the TiO2 surface, energetically favorable sites were investigated using DFT calculations. Table 1 shows the calculated binding energies (Eb) of the six adsorption sites (#1−#6) on the stoichiometric (110) surface, and those of the Obr-(#7) and Oba-(#8) vacancy sites. The corresponding relaxed atomic structures and simulated HAADF STEM images are shown in Figure 3. Regarding the stoichiometric surface without oxygen vacancies, the #1 site is the most stable adsorption site with a binding energy of 2.72 eV and the #2 site is the second most stable site with a binding energy of 2.31 eV. The binding energies of other surface adsorption sites are much smaller than at these two sites. As compared in Figure 3, the experimental images and the simulated images based on the relaxed atomic structures show excellent agreement in these stoichiometric surface adsorption sites. Once the surface oxygen vacancies are created, they are far more favorable adsorption sites than those on the stoichiometric TiO2 (110) surface. As shown in the Table 1, the Eb of the #8 site becomes as high as 3.82 eV, even higher than that of the #7 site (3.64 eV). Thus, the Oba vacancy site is theoretically the most stable adsorption site among the eight possible adsorption sites considered here. As for the relaxed Pt structures, while the Pt atom adsorbed on the #7 site shows no obvious in-plane atomic shift relative to that on the #5 site, the Pt atom adsorbed on the #8 site exhibits a significant inplane atomic shift as compared to that on the #2 site. The Pt atom on the #8 site relaxes toward the nearest Ti5c atomic site (along the [001] direction). HAADF STEM image simulations based on the relaxed Pt structures, shown in Figure 3, show that the HAADF STEM can detect such Pt structural relaxation. The simulated image in Figure 3h is in good agreement with the experimental image shown in Figure 2f (duplicated in Figure 3h). It should be noted here that, whereas on the basis of the schematic in Figure 1a it might have been expected the #2 and #8 site cases to be indistinguishable by HAADF STEM imaging, the structural relaxation is such that they can, in fact, be distinguished. Thus, we conclude that the experimentally observed adsorption site in Figure 2f must be the Pt atom on the Oba vacancy site. While the Obr vacancy sites have hitherto been considered exclusively as the preferential sites for noble metal atom adsorption, our direct images and theory demonstrate that Oba vacancies also produce stable adsorption sites for noble metal atoms. In order to clarify the impact of Oba vacancy sites on Pt atom adsorption, we statistically estimate the Pt atom positions from several HAADF STEM images observed under similar imaging conditions. Figure 4 shows a histogram of the number of Pt single atoms on each adsorption site. It is clearly seen that the Pt atoms are most frequently found on the #8, which agrees well with the DFT energetics results. We also found that many Pt single atoms are on #1 and #2 sites, which are the two most stable Pt adsorption sites of the stoichiometric TiO2 (110) surface. Although Pt sites on #5, #6, and #7 cannot be discriminated and are thus summed up in the histogram, the #7 sites appear to have less impact on the Pt atom adsorption compared with the #8 sites, indicating the less impact of Obr vacancy sites on Pt atom adsorption. The statistical analysis thus clearly shows that the Oba vacancy sites have a significant impact on the Pt atom adsorption on the TiO2 (110) surface. It should be mentioned here that our calculated formation energy of the Oba vacancy (Ed) is slightly larger (by 0.15 eV)
Figure 4. Histogram of Pt single atom adsorption to each adsorption site. By counting the number of Pt single atoms on each adsorption site from several HAADF STEM images, it was found that Pt atoms were most frequently adsorbed on the #8 site.
than that of the Obr vacancy, which implies predominant formation of Obr vacancies at the free TiO2 (110) surface. In this regard, the favorable Pt adsorption at the Oba-vacancy site may not be explained in terms of the stability of the Oba vacancy. However, the Pt adsorption at the Oba vacancies is rationalized by assuming possible adsorption processes: Pt atoms deposited on the surface with Obr vacancies do not necessarily occupy the Obr-vacancy sites directly but may kick out Oba atoms into the nearest neighboring Obr-vacancy sites and thereby situate on the resultant Oba-vacancy sites. Because the energy difference between Oba and Obr vacancies (ΔEd = Ed(Oba) − Ed(Obr)) is 0.15 eV as described above, the total energy gain upon the Pt adsorption at the Oba vacancy by the above process (ΔE = ΔEb − ΔEd where ΔEb = Eb(Oba) − Eb(Obr)) is found to be 0.03 eV. The positive ΔE value indicates that the process is exothermic, and therefore Pt adsorption at the Oba-vacancy sites is likely to occur. Our findings thus highlight the importance of the Oba vacancy on metal atom adsorption. The Obr vacancy has been extensively studied both experimentally and theoretically, but the Oba vacancy, which is located at the subsurface position relative to the protruding Obr atom, has been overlooked for years. This is probably due to both the limitations of microscope resolution and of cell dimensions in the previous theoretical calculations. By direct subangstrom resolution imaging of individual metal atoms on the supporting surface, our results rewrite our understanding of the vacancy formation and metal atom adsorption phenomena even in this most extensively studied oxide surfaces. 137
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
■
Letter
(20) Chang, T.-Y.; Ikuhara, Y.; Shibata, N. Appl. Phys. Express 2013, 6, 025503.
ASSOCIATED CONTENT
S Supporting Information *
Additional information includes sample preparation procedures, STEM image acquisition, EDS analysis, image simulation, and DFT calculations. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]. Author Contributions
T.Y.C. and N.S. designed the study and wrote the paper. T.Y. C. and R.I. performed the STEM experiments. Y.T., K.T., and K.M. performed theoretical calculations. Y.I. contributed to the discussion. N.S. directed the entire study. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS We thank S. D. Findlay at Monash University for his helpful discussions and suggestions and H. Oshikawa for his helps in STEM-EDS analysis. This study was supported in part by a Grant-in-Aid for Scientific Research on Innovative Areas “Nano Informatics” (Grants 25106002 and 25106003) from JSPS. N.S. acknowledges support from the PRESTO, JST and the JSPS KAKENHI Grant 23686093. A part of this work was conducted in Research Hub for Advanced Nano Characterization, The University of Tokyo, under the support of “Nanotechnology Platform” (Project No. 12024046) by MEXT, Japan.
■
REFERENCES
(1) Bell, A. T. Science 2003, 424, 1688−1691. (2) Haruta, M. CATTECH 2002, 6, 102−115. (3) Diebold, U. Surf. Sci. Reports 2003, 48, 53−229. (4) Sasahara, A.; Pang, C. L.; Onishi, H. J. Phys. Chem. B 2006, 110, 13453−13457. (5) Iddir, H.; Ö ğüt, S.; Browning, N. D.; Disko, M. M. Phys. Rev. B 2005, 72, 081407. (6) Iddir, H.; Skavysh, V.; Ö ğüt, S.; Browning, N. D.; Disko, M. M. Phys. Rev. B 2006, 73, 041403. (7) Ç elik, V.; Ü nal, H.; Mete, E.; Ellialtıoğlu, Ş. Phys. Rev. B 2010, 82, 205113. (8) Ç akir, D.; Gülseren, O. J. Phys. Chem. C 2012, 116, 5735−5746. (9) Fischer, S.; Munz, A. W.; Schierbaum, K. D.; Göpel, W. Surf. Sci. 1995, 337, 17−30. (10) Diebold, U.; Lehman, J.; Mahmoud, T.; Kuhn, M.; Leonardelli, G.; Hebenstreit, W.; Schmid, M.; Varga, P. Surf. Sci. 1998, 411, 137− 153. (11) Oviedo, J.; San Miguel, M. A.; Sanz, J. F. J. Chem. Phys. 2004, 121, 7427−7433. (12) Pennycook, S. J.; Jesson, D. E. Ultramicroscopy 1991, 37, 14−38. (13) Nellist, P. D.; Pennycook, S. J. Science 1996, 274, 413. (14) Shibata, N.; Goto, A.; Matsunaga, K.; Mizoguchi, T.; Findlay, S. D.; Yamamoto, T.; Ikuhara, Y. Phys. Rev. Lett. 2009, 102, 136105. (15) Ortalan, V.; Uzun, A.; Gates, B. C.; Browning, N. D. Nat. Nanotechnol. 2010, 5, 843−847. (16) Qiao, B.; Wang, A.; Yang, X.; Allard, L. F.; Jiang, Z.; Cui, Y.; Liu, J.; Li, J.; Zhang, T. Nature Chem. 2011, 3, 634−641. (17) Nakamura, R.; Ohashi, N.; Imanishi, A.; Osawa, T.; Matsumoto, Y.; Koinuma, H.; Nakato, Y. J. Phys. Chem. B 2005, 109, 1648−1651. (18) Lu, Y.; Jaeckel, B.; Parkinson, B. A. Langmuir 2006, 22, 4472− 4475. (19) Shibata, N.; Goto, A.; Choi, S.-Y.; Mizoguchi, T.; Findlay, S. D.; Yamamoto, T.; Ikuhara, Y. Science 2008, 322, 570−573. 138
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Direct Imaging of Pt Single Atoms Adsorbed on TiO2 (110) Surfaces Teng-Yuan Chang,† Yusuke Tanaka,‡ Ryo Ishikawa,† Kazuaki Toyoura,‡ Katsuyuki Matsunaga,‡,§ Yuichi Ikuhara,†,§ and Naoya Shibata*,†,∥ †
Institute of Engineering Innovation, School of Engineering, the University of Tokyo, Yayoi 2-11-16, Bunkyo-ku, Tokyo 113-8656, Japan ‡ Department of Materials Science and Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan § Nanostructures Research Laboratory, Japan Fine Ceramics Center, 2-4-1 Mutsuno, Atsuta-ku, Nagoya 456-8587, Japan ∥ Japan Science and Technology Agency, PRESTO, 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan S Supporting Information *
ABSTRACT: Noble metal nanoparticles (e.g., gold and platinum) supported on TiO2 surfaces are utilized in many technological applications such as heterogeneous catalysts. To fully understand their enhanced catalytic activity, it is essential to unravel the interfacial interaction between the metal atoms and TiO2 surfaces at the level of atomic dimensions. However, it has been extremely difficult to directly characterize the atomic-scale structures that result when individual metal atoms are adsorbed on the TiO2 surfaces. Here, we show direct atomic-resolution images of individual Pt atoms adsorbed on TiO2 (110) surfaces using aberration-corrected scanning transmission electron microscopy. Subangstrom spatial resolution enables us to identify five different Pt atom adsorption sites on the TiO2 (110) surface. Combining this with systematic density functional theory calculations reveals that the most favorable Pt adsorption sites are on vacancy sites of basal oxygen atoms that are located in subsurface positions relative to the top surface bridging oxygen atoms. KEYWORDS: Heterogeneous catalyst, metal-oxide interaction, platinum, titanium oxide, single atom imaging, scanning transmission electron microscopy
T
adsorption.6,7 Enhanced bonding with oxygen vacancy sites has thus been considered to be of primal importance to account for Pt atom anchoring and subsequent nanoparticle nucleation and growth on TiO2 surfaces. While their importance is wellrecognized, previous studies focused only on the topmost Obr vacancies and not on other surface oxygen vacancies such as three-fold coordinated basal oxygen (Oba) vacancies. This may be partly due to the following two reasons: (1) the vacancy formation energy of Oba atoms is predicted to be much higher than that of Obr atoms,11 and (2) point defects on the TiO2 (110) surfaces imaged by STM have always been attributed to the Obr vacancies.9,10 However, the limited spatial resolution and complex image formation mechanism of STM may have
o fully understand the unique chemical and physical properties of metal nanoparticles dispersed on oxide surfaces, detailed atomic-scale knowledge of their structure and the nanoparticle/oxide-surface interactions is crucial.1−3 One system that has been studied extensively in this regard is Pt/ TiO2, which serves as a typical heterogeneous catalyst for CO oxidation. It has been reported that the stabilization of Pt atoms on TiO2 (110) surfaces, which have been intensively studied as model surfaces of TiO2, depends on their adsorption site and the formation of surface point defects such as oxygen vacancies.4−8 The topmost oxygen atoms, which are coordinated with the two Ti atoms below (so-called bridging oxygen, Obr), are regarded as the most favorable sites for vacancy formation and their vacancies have been experimentally observed by scanning tunneling microscopy (STM).9,10 Previous first-principles calculations predicted that these Obr vacancies are the most energetically favorable sites for Pt atom © 2013 American Chemical Society
Received: September 20, 2013 Revised: November 21, 2013 Published: December 19, 2013 134
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
Figure 1. Structural model illustrating the eight different adsorption sites on TiO2 (110) surface and HAADF STEM image of TiO2 observed along [110] projection. (a) The surface atomic structure of TiO2 (110) and possible Pt atom adsorption sites. The green balls show the eight different Pt atom adsorption sites; #1 (H1), #2 (atop Oba), #3 (H2), #4 (atop Ti6c), #5 (atop Obr), #6 (atop Ti5c), #7 (Obr vacancy), and #8 (Oba vacancy). (b) Atomic-resolution HAADF STEM image of TiO2 (110) surface observed along perpendicular [110] direction. The two bright contrast spots correspond to the Ti−O and Ti-only columns, respectively.
Figure 2. HAADF STEM images of Pt single atoms on TiO2 (110) surface. (a) Atomic-resolution HAADF STEM image of Pt supported TiO2 (110) surface. (b-f) The magnified views of Pt single atoms located on different surface sites found in a. As seen in the images, we can clearly distinguish five different Pt single atom attachment sites on the TiO2 (110) surface unit structure. The magnified images were low-pass filtered to remove noises.
hindered the detailed determination of surface oxygen vacancies. Diebold et al. reported two types of defect-originated image features on TiO2 (110) surfaces.10 They assigned the bright spots on the dark Obr rows to be the Obr vacancies, but the dark spots on the bright 5-fold Ti rows were not clearly identified. The presence of other oxygen vacancies and their effect on metal atom anchoring are thus still a matter of conjecture. In the present study, we use aberration-corrected scanning transmission electron microscopy (STEM) to directly image individual Pt atoms adsorbed on TiO2 (110) surfaces. Highangle annular dark-field (HAADF) STEM12 has proved to be an excellent technique for directly imaging individual metal atoms supported by TiO2 surfaces.13−16 Here, we show that the detailed Pt adsorption sites within the unit structure of the TiO2 (110) surface can be identified by HAADF STEM. The reason for the enhanced adsorption to these specific sites is discussed based on systematic density functional theory (DFT) calculations. We highlight the unexpected importance of Oba
vacancies, whose strong bonding to the Pt atoms may dictate the interaction between Pt atoms and TiO2 (110) surfaces. Figure 1a illustrates the atomic structure of TiO2 (110) surface and high-symmetry Pt adsorption sites (shown by green balls). Here, we extend the unit surface structure of TiO2 (110) two times to visualize the individual adsorption sites more clearly. The green balls labeled from 1 to 6 represent Pt atom adsorption sites on a stoichiometric TiO2 (110) surface. Oxygen vacancies are illustrated as translucent pink balls and the green balls labeled 7 and 8 thus represent Pt atom adsorption sites on the Obr and Oba vacancy sites, respectively. It should be noted here that the #7 and #8 sites appear to be the same as the #5 and #2 sites in the [110] projection but now with vacancies, and the Pt sites in Figure 1a have been centered to the vacancy sites, despite the potential for relaxation that accompanies the vacancy formation. Figure 1b shows a HAADF STEM image of a pristine TiO2 (110) surface observed along the [110] projection. The thin, electron transparent TiO2 specimen was annealed in air at 973 K for 1 h to produce an 135
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
Table 1. Binding Energies (in eV) of a Pt Atom at Each Adsorption Site on TiO2 (110) Surface Predicted by DFT Calculations site
#1
#2
#3
#4
#5
#6
#7
#8
notation Eb (eV)
H1 2.72
Oba 2.31
H2 1.85
Ti6c 1.76
Obr 1.47
Ti5c 1.31
Obr vacancy 3.64
Oba vacancy 3.82
Figure 3. DFT relaxed atomic structures, related HAADF STEM image simulations and experimental HAADF STEM images of Pt atoms on TiO2 (110) surfaces. (a−f) Pt single atom on site #1 through site #8 defined in Figure 1a. Top panels show the relaxed atomic structures predicted by DFT calculations. Middle panels show simulated HAADF STEM images based on the DFT relaxed structures. Bottom panels show corresponding experimental HAADF STEM images reproduced from Figure 2. Even after allowing structural relaxation, we cannot distinguish “#3, #4”, and “#5, #6, #7”, since HAADF STEM images are observed along [110] projection. However, due to the strong structural relaxation found in #8, we can clearly distinguish #8 from #2. Thus, we can distinguish five different Pt adsorption sites by the HAADF STEM images along [110] projection.
atomically flat TiO2 (110) surface as per previous studies.14,17−20 As indicated by the arrows, two different intensities of bright spots can be observed in the image. These two atomiccolumns are the so-called Ti−O column and Ti-only column, reflecting the atomic species present in each atomic column. Because of a greater atomic column occupancy along the [110] direction, the Ti−O columns have higher intensity than the Tionly columns. The oxygen columns in the troughs between Ticontaining atomic columns cannot be imaged because the signal from the light element columns is very weak at the HAADF detection angles. Figure 2a shows a typical HAADF STEM image of Pt atoms on a TiO2 (110) surface, which was obtained soon after highpurity Pt atoms were directly evaporated onto the TiO2 (110) surface. We have confirmed the evaporation of Pt atoms under the present experimental condition through large area STEMEDS analysis shown in the Supporting Information. The detailed sample preparation procedures are given in the Method Section in the Supporting Information. To minimize any effect of electron beam incidence on the Pt structures, we use a low-dose imaging condition with a probe current of about 10 pA and probe size of about 0.8 Å. We have checked that there is no obvious beam damage after several scans under the present imaging condition and thus consider our images to reflect stable Pt atom structures on the TiO2 (110) surface. Further, we tried to perform atomic-resolution STEM spectroscopy such as electron energy-loss spectroscopy and X-ray spectroscopy to obtain spectroscopic evidence of Pt in conjunction with the atomic-resolution HAADF STEM images. However, under the present low-dose condition such spectroscopic signals are too weak to be meaningful data. On
the other hand, the Z-contrast image characteristic of HAADF makes Pt (Z, 78) atoms stand out as brighter spots above the TiO2 support (Ti, Z = 22; O, Z = 8), and many Pt atoms are found to be single, isolated atoms. We found that the Pt single atoms do not always attach to the same sites but rather to several different sites on the TiO2 (110) surface. After extensively measuring the Pt single atom adsorption sites from several HAADF STEM images, we find that the stable Pt adsorption sites can be categorized into five sites. As examples, five magnified views from the areas enclosed by white squares in Figure 2a are shown in Figure 2b−f with white arrows indicating the position of the Pt atoms in each image. These images clearly show that Pt atoms stably attach to the different surface sites of TiO2 (110), even though these images are in projection along the [110] direction. Figure 2b shows a Pt atom locating near on a hollow site adjacent to a Ti−O column, suggesting the Pt atom sits on the #1 site (hollow site; H1) but with slight structural relaxation toward the near-by Ti−O column. Figure 2c shows a Pt atom locating near on an oxygen column adjacent to a Ti-only column, corresponding to the #2 or #8 sites (atop Oba or its vacancy sites) but evident structural relaxation toward near-by Ti-only column can be observed. Figure 2d shows a Pt atom locating on top of a Ti-only column, corresponding to the #3 or #4 sites (hollow site 2; H2 or atop Ti6c site). Figure 2e shows a Pt atom locating on top of a Ti− O column, corresponding to the #5, #6, or #7 sites (atop Obr or atop Ti5c or Obr vacancy sites). Figure 2f shows a Pt adsorption site locating near both the Ti−O and Ti-only columns, corresponding to the #8 site (atop Oba vacancy site) after significant structural relaxation along [001] direction, which will be discussed later. 136
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
Letter
For complementary information on the experimentally identified adsorption sites of Pt atoms on the TiO2 surface, energetically favorable sites were investigated using DFT calculations. Table 1 shows the calculated binding energies (Eb) of the six adsorption sites (#1−#6) on the stoichiometric (110) surface, and those of the Obr-(#7) and Oba-(#8) vacancy sites. The corresponding relaxed atomic structures and simulated HAADF STEM images are shown in Figure 3. Regarding the stoichiometric surface without oxygen vacancies, the #1 site is the most stable adsorption site with a binding energy of 2.72 eV and the #2 site is the second most stable site with a binding energy of 2.31 eV. The binding energies of other surface adsorption sites are much smaller than at these two sites. As compared in Figure 3, the experimental images and the simulated images based on the relaxed atomic structures show excellent agreement in these stoichiometric surface adsorption sites. Once the surface oxygen vacancies are created, they are far more favorable adsorption sites than those on the stoichiometric TiO2 (110) surface. As shown in the Table 1, the Eb of the #8 site becomes as high as 3.82 eV, even higher than that of the #7 site (3.64 eV). Thus, the Oba vacancy site is theoretically the most stable adsorption site among the eight possible adsorption sites considered here. As for the relaxed Pt structures, while the Pt atom adsorbed on the #7 site shows no obvious in-plane atomic shift relative to that on the #5 site, the Pt atom adsorbed on the #8 site exhibits a significant inplane atomic shift as compared to that on the #2 site. The Pt atom on the #8 site relaxes toward the nearest Ti5c atomic site (along the [001] direction). HAADF STEM image simulations based on the relaxed Pt structures, shown in Figure 3, show that the HAADF STEM can detect such Pt structural relaxation. The simulated image in Figure 3h is in good agreement with the experimental image shown in Figure 2f (duplicated in Figure 3h). It should be noted here that, whereas on the basis of the schematic in Figure 1a it might have been expected the #2 and #8 site cases to be indistinguishable by HAADF STEM imaging, the structural relaxation is such that they can, in fact, be distinguished. Thus, we conclude that the experimentally observed adsorption site in Figure 2f must be the Pt atom on the Oba vacancy site. While the Obr vacancy sites have hitherto been considered exclusively as the preferential sites for noble metal atom adsorption, our direct images and theory demonstrate that Oba vacancies also produce stable adsorption sites for noble metal atoms. In order to clarify the impact of Oba vacancy sites on Pt atom adsorption, we statistically estimate the Pt atom positions from several HAADF STEM images observed under similar imaging conditions. Figure 4 shows a histogram of the number of Pt single atoms on each adsorption site. It is clearly seen that the Pt atoms are most frequently found on the #8, which agrees well with the DFT energetics results. We also found that many Pt single atoms are on #1 and #2 sites, which are the two most stable Pt adsorption sites of the stoichiometric TiO2 (110) surface. Although Pt sites on #5, #6, and #7 cannot be discriminated and are thus summed up in the histogram, the #7 sites appear to have less impact on the Pt atom adsorption compared with the #8 sites, indicating the less impact of Obr vacancy sites on Pt atom adsorption. The statistical analysis thus clearly shows that the Oba vacancy sites have a significant impact on the Pt atom adsorption on the TiO2 (110) surface. It should be mentioned here that our calculated formation energy of the Oba vacancy (Ed) is slightly larger (by 0.15 eV)
Figure 4. Histogram of Pt single atom adsorption to each adsorption site. By counting the number of Pt single atoms on each adsorption site from several HAADF STEM images, it was found that Pt atoms were most frequently adsorbed on the #8 site.
than that of the Obr vacancy, which implies predominant formation of Obr vacancies at the free TiO2 (110) surface. In this regard, the favorable Pt adsorption at the Oba-vacancy site may not be explained in terms of the stability of the Oba vacancy. However, the Pt adsorption at the Oba vacancies is rationalized by assuming possible adsorption processes: Pt atoms deposited on the surface with Obr vacancies do not necessarily occupy the Obr-vacancy sites directly but may kick out Oba atoms into the nearest neighboring Obr-vacancy sites and thereby situate on the resultant Oba-vacancy sites. Because the energy difference between Oba and Obr vacancies (ΔEd = Ed(Oba) − Ed(Obr)) is 0.15 eV as described above, the total energy gain upon the Pt adsorption at the Oba vacancy by the above process (ΔE = ΔEb − ΔEd where ΔEb = Eb(Oba) − Eb(Obr)) is found to be 0.03 eV. The positive ΔE value indicates that the process is exothermic, and therefore Pt adsorption at the Oba-vacancy sites is likely to occur. Our findings thus highlight the importance of the Oba vacancy on metal atom adsorption. The Obr vacancy has been extensively studied both experimentally and theoretically, but the Oba vacancy, which is located at the subsurface position relative to the protruding Obr atom, has been overlooked for years. This is probably due to both the limitations of microscope resolution and of cell dimensions in the previous theoretical calculations. By direct subangstrom resolution imaging of individual metal atoms on the supporting surface, our results rewrite our understanding of the vacancy formation and metal atom adsorption phenomena even in this most extensively studied oxide surfaces. 137
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
Nano Letters
■
Letter
(20) Chang, T.-Y.; Ikuhara, Y.; Shibata, N. Appl. Phys. Express 2013, 6, 025503.
ASSOCIATED CONTENT
S Supporting Information *
Additional information includes sample preparation procedures, STEM image acquisition, EDS analysis, image simulation, and DFT calculations. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: [email protected]. Author Contributions
T.Y.C. and N.S. designed the study and wrote the paper. T.Y. C. and R.I. performed the STEM experiments. Y.T., K.T., and K.M. performed theoretical calculations. Y.I. contributed to the discussion. N.S. directed the entire study. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS We thank S. D. Findlay at Monash University for his helpful discussions and suggestions and H. Oshikawa for his helps in STEM-EDS analysis. This study was supported in part by a Grant-in-Aid for Scientific Research on Innovative Areas “Nano Informatics” (Grants 25106002 and 25106003) from JSPS. N.S. acknowledges support from the PRESTO, JST and the JSPS KAKENHI Grant 23686093. A part of this work was conducted in Research Hub for Advanced Nano Characterization, The University of Tokyo, under the support of “Nanotechnology Platform” (Project No. 12024046) by MEXT, Japan.
■
REFERENCES
(1) Bell, A. T. Science 2003, 424, 1688−1691. (2) Haruta, M. CATTECH 2002, 6, 102−115. (3) Diebold, U. Surf. Sci. Reports 2003, 48, 53−229. (4) Sasahara, A.; Pang, C. L.; Onishi, H. J. Phys. Chem. B 2006, 110, 13453−13457. (5) Iddir, H.; Ö ğüt, S.; Browning, N. D.; Disko, M. M. Phys. Rev. B 2005, 72, 081407. (6) Iddir, H.; Skavysh, V.; Ö ğüt, S.; Browning, N. D.; Disko, M. M. Phys. Rev. B 2006, 73, 041403. (7) Ç elik, V.; Ü nal, H.; Mete, E.; Ellialtıoğlu, Ş. Phys. Rev. B 2010, 82, 205113. (8) Ç akir, D.; Gülseren, O. J. Phys. Chem. C 2012, 116, 5735−5746. (9) Fischer, S.; Munz, A. W.; Schierbaum, K. D.; Göpel, W. Surf. Sci. 1995, 337, 17−30. (10) Diebold, U.; Lehman, J.; Mahmoud, T.; Kuhn, M.; Leonardelli, G.; Hebenstreit, W.; Schmid, M.; Varga, P. Surf. Sci. 1998, 411, 137− 153. (11) Oviedo, J.; San Miguel, M. A.; Sanz, J. F. J. Chem. Phys. 2004, 121, 7427−7433. (12) Pennycook, S. J.; Jesson, D. E. Ultramicroscopy 1991, 37, 14−38. (13) Nellist, P. D.; Pennycook, S. J. Science 1996, 274, 413. (14) Shibata, N.; Goto, A.; Matsunaga, K.; Mizoguchi, T.; Findlay, S. D.; Yamamoto, T.; Ikuhara, Y. Phys. Rev. Lett. 2009, 102, 136105. (15) Ortalan, V.; Uzun, A.; Gates, B. C.; Browning, N. D. Nat. Nanotechnol. 2010, 5, 843−847. (16) Qiao, B.; Wang, A.; Yang, X.; Allard, L. F.; Jiang, Z.; Cui, Y.; Liu, J.; Li, J.; Zhang, T. Nature Chem. 2011, 3, 634−641. (17) Nakamura, R.; Ohashi, N.; Imanishi, A.; Osawa, T.; Matsumoto, Y.; Koinuma, H.; Nakato, Y. J. Phys. Chem. B 2005, 109, 1648−1651. (18) Lu, Y.; Jaeckel, B.; Parkinson, B. A. Langmuir 2006, 22, 4472− 4475. (19) Shibata, N.; Goto, A.; Choi, S.-Y.; Mizoguchi, T.; Findlay, S. D.; Yamamoto, T.; Ikuhara, Y. Science 2008, 322, 570−573. 138
dx.doi.org/10.1021/nl403520c | Nano Lett. 2014, 14, 134−138
