What is the Interaction between Atomically Dispersed Ni and Oxide

Materials Transactions, Vol. 50, No. 3 (2009) pp. 509 to 515 #2009 The Japan Institute of Metals

What is the Interaction between Atomically Dispersed Ni and Oxide Surfaces? Yuichiro Koike1 , Wang-Jae Chun2 , Kaoru Ijima3 , Shushi Suzuki1 and Kiyotaka Asakura1;4 1

Catalysis Research Center, Hokkaido University, Sapporo 001-0021, Japan Graduate School of Natural Sciences, International Christian University, Tokyo 181-8585, Japan 3 Department of Electronic Engineering, University of Yamanashi, Kofu 400-8510, Japan 4 Department of Quantum Science and Technology, Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Japan 2

Metal–oxide surface interaction is important in catalysis. We have studied Ni on TiO2 (110) and Al2 O3 (0001) by means of polarizationdependent total-reflection fluorescence extended X-ray absorption fine structure (PTRF-EXAFS) observations to elucidate the origin of metal– oxide interactions. Ni atoms interact with the dangling bond of the oxygen atoms. The Ni atoms are positively charged and bonded with more than two oxygen atoms. [doi:10.2320/matertrans.MBW200830] (Received November 4, 2008; Accepted December 24, 2008; Published February 12, 2009) Keywords: (polarization-dependent total-reflection fluorescence extended X-ray absorption fine structure) PTRF-XAFS, metal oxide surface, nickel cluster, metal–oxide interaction, TiO2 (110), Al2 O3 (0001)

1.

Introduction

Metal–oxide surface interaction is a crucial in understanding and controlling the catalytic properties of supported metal catalysts.1) The simplest interaction between a metal and an oxide is that between a single metal atom and the oxide surface. The interaction can be understood based on the bonding site and bonding features. In the case of semiconductors, a dangling bond plays an important role in the surface adsorption properties. However, the oxide surface is slightly more complicated than a semiconductor’s surface because of the presence of a positive ion (cation) and a negative ion (anion). For example, it is often believed that metal atoms are adsorbed on the cation site of the metal oxide.2) In this case, electron transfer is postulated to occur from the cation to the metal species.3) Moreover, oxygen defects play an important role in this type of interaction because the cation left behind has a dangling bond, where the electron is trapped to neutralize the surface charge.4) Subsequently, extensive electron transfer from the cation to the metal occurs.5) The other possibility is the formation of metal–oxygen (anion) bonds in which the metal is positively charged. In this study, we measured the total reflection fluorescence extended X-ray absorption fine structure (EXAFS) of Ni species atomically dispersed on Al2 O3 (0001) and TiO2 (110) surfaces to elucidate the basic principles of the metal–oxide surface interaction. We found that Ni strongly interacts with surface coordinatively unsaturated oxygen (anion) atoms making a covalent bond (less than 0.23 nm). The surface dangling bond at the oxygen site is important for the interaction with the Ni species. We should consider the polarity of dangling bonds on oxide surfaces. 2. 2.1

Experimental

Principle of polarization-dependent total-reflection fluorescence EXAFS EXAFS provides information about the local structure of the X-ray absorbing atoms even when they are highly

dispersed. Since the structure obtained from conventional EXAFS is averaged over all directions, it is difficult to elucidate the metal–oxide interface bonding information. When one uses polarized X-rays, such as synchrotron radiation, the EXAFS signal depends on the polarization and the bonding directions.6) When the angle, j , is defined as that between the polarization and j-th bonding directions, the EXAFS oscillation, ðkÞ, is expressed in the framework of single scattering and plane-wave approximation as X ðkÞ ¼ 3 cos2 j j ðkÞ; ð1Þ j

where j ðkÞ is the EXAFS oscillation of the j-th bond. When the metal species is deposited on a single-crystal surface and the polarization direction is set to the surface normal, we can obtain structural information in the single-crystal direction. Furthermore, by changing the polarization directions against the surface normal, we can obtain a 3-D structure from the polarization-dependent EXAFS data. However, a single crystal has a very small surface area and extremely less metal deposition; thus, we cannot observe the metal species using the conventional EXAFS. In general, the fluorescence detection mode can be used for EXAFS signals from a dilute system. Soft X-ray surface EXAFS, such as that for C and S K-edges, was measured successfully7) using the fluorescence mode. However, in the case of a hard X-ray region, it is difficult to obtain the surface species because hard X-rays penetrate deeply into the bulk material, and X-rays scattered from the bulk result in a high X-ray background. Total reflection of X-rays is important to prevent such penetration, and it reduces the scattered X-ray background considerably.8,9) As a result, polarization-dependent totalreflection fluorescence EXAFS (PTRF-EXAFS) provides the 3-D local structure of the species deposited on single-crystal surfaces.10–12) Figure 1 shows the PTRF-EXAFS instrument.13) Because the chamber should be placed inside the hutch, a small room for X-ray protection, we should divide the instrument into three parts, a measurement chamber, a treatment chamber, and a transfer chamber. In the measurement chamber, a

510

Y. Koike, W.-J. Chun, K. Ijima, S. Suzuki and K. Asakura

Preparation Chamber PTRF-XAFS

19 SSD Transfer Chamber Measurement Chamber

Fig. 1 PTRF-EXAFS instrument.

six-axis goniometer is installed to measure three different orientations. The fluorescence X-rays can be detected with a 19-element solid state detector (SSD, Canberra, SSDGL0110S). In the preparation chamber, the sample can be heated up to 1500 K by a pyrolytic BN heater and sputtered with Arþ . XPS (X-ray Photoelectron spectroscopy) and LEED (Low Energy Electron Diffraction) optics are installed to characterize the surface. The transfer chamber can be connected to both of the other chambers. It has a small room separated from the other parts by a gate valve, where the sample can be kept at 108 Pa while it is moved from the preparation chamber to the measurement chamber using a Ti sublimation pump.14) 2.2 Sample preparation An -Al2 O3 (0001) (20 20 1 mm3 ) single crystal was purchased from Earth Jewelry Co. Japan. The sample was first treated at 1073 K for 3 h in air. Then, it was placed in the preparation chamber (base pressure 5 108 Pa) and cleaned by further annealing at 873 K. An optically polished rutile TiO2 (110) single crystal (20 20 1 mm3 , Crystal Base Co., Japan) was pretreated at 1273 K for 3 h in air to remove carbon contamination. It was then loaded into the UHV preparation chamber. The sample was further cleaned by several cycles of Arþ sputtering and annealing at 873 K. Finally, we obtained the TiO2 (110) and -Al2 O3 (0001) surfaces with a (1 1) LEED pattern, and little C1s peak was detected by XPS. Ni was evaporated on the TiO2 (110) surface by resistive heating of a W wire wrapped with a Ni wire. 3.

Results and Discussion

3.1 Ni on Al2 O3 Al2 O3 (0001) has the surface structure shown in Fig. 2. Al termination is revealed by LEED analysis and can be

Surface Al Top View

Surface O Shallow

2nd

layer Al Deep 2nd layer Al

Side View

0.084 0.000 –0.084 –0.133

Fig. 2 Al2 O3 (0001) surface structure.

understood in terms of the surface charge valence.15–17) The surface has three nonequivalent sites: threefold oxygen sites (called site-2 and site-3) and an exposed Al site (called site1). Site-2 and site-3 are distinguished by the second layer Al position. Al2 O3 has an Al2 O9 dimer unit where Al atoms were shifted outwards from the central position in two facesharing octahedra. The Al–Al distance is 0.266 nm. The [0001] direction normal to the surface is the same as the Al– Al direction. Thus, two types of truncation contribute to the (0001) surface. One is the truncation at the end oxygen triangle, and the other is that at the sharing oxygen triangle

What is the Interaction between Atomically Dispersed Ni and Oxide Surfaces?

0.2

0.2

0.0 χ(k)

χ(k)

0.0 E ⊥ surface

-0.2 2

4

6 -1 k / 10 nm

E// surface -0.2

8

2

10

6

k / 10 nm

(a)

0.2

8

10

-1

(b)

0.2

E ⊥ surface

E// surface

0.1 χ(k)

0.1 χ(k)

4

Polarization-dependent EXAFS oscillations of Ni on Al2 O3 (0001). The Ni coverage is 3 1013 cm2 .

Fig. 3

0.0

0.0

–0.1

–0.1

–0.2

–0.2

4

6 8 –1 k / 10 nm

4

10

(c)

0.3 0.2 0.1 0.0 –0.1 –0.2 –0.3 –0.4 4

6 8 k / 10 nm–1

10

(d) 0.2

E ⊥ surface

E// surface

0.1 χ(k)

χ(k)

511

0.0 –0.1 –0.2

6 8 k / 10 nm –1

10

4

6

8 k / 10 nm

10

–1

Fig. 4 Simulated and observed EXAFS. (a) and (b) are based on the structure shown in Fig. 5, while (c) and (d) are based on the structure proposed by Ma.22)

of the Al2 O9 dimer unit. Al is located at 0.084 nm (under site-2) and 0.133 nm (under site-3) from the surface in the former and latter cases, respectively. Note that the next octahedral outside the Al2 O9 unit in the [0001] direction has no Al (denoted as a cross in the figure). When Ni was deposited on the Al2 O3 , we could easily observe Ni clusters that showed Ni aggregation. When we deposited Ni at 3 1013 atoms/cm2 , no Ni–Ni bond was observed, indicating that Ni was atomically dispersed. Figure 3 shows the EXAFS oscillations of the sample with the polarization vector parallel and perpendicular to the surface. The main oscillations damp quickly, indicating the presence of low-Z elements in the first nearest neighbor. According to one-shell curve fitting analyses assuming an oxygen shell as the nearest neighbor, a Ni–O distance of 0:195 0:003 nm is observed in both directions. The coordination number is 2:6 0:5 and 3:0 0:5 for the directions parallel and perpendicular to the surface, respectively. This result indicates that the location of Ni is not the atop site, which should have no oxygen atom in the parallel direction, nor any sites around the surface Al of site-1.

Considering the polarization dependence in coordination numbers, Ni atoms are situated on the threefold oxygen hollow site. The -Al2 O3 (0001) surface has two types of threefold oxygen sites, as shown in Fig. 2. We simulated the data based on the model structure and FEFF8.18,19) Details of the analyses were described elsewhere.20,21) We postulate an ideal bulk -Al2 O3 (0001) surface with relaxation of the surface oxygen atoms. We tried several structures, and finally we obtained one that reproduced the data well, as shown in Fig. 4. Its model structure is shown in Fig. 5. We included three types of Al atoms in this calculation: surface Al atoms and second- and third-layer Al atoms, which are denoted as Al1, Al2, and Al3, respectively. The Ni–Al distance is 0.278 nm, indicating that site-3 is the most plausible Ni location. The Ni atom is located at the threefold oxygen site with a Ni–O distance of 0.198 nm. The surface oxygen atoms are shifted 0.036 nm upwards from the original position. Site-3 corresponds to the truncated triangle at the center of the Al2 O9 unit. Ni is located at the other Al position inside the Al2 O9 unit. The surface oxygen atoms at site-3,

512


Top View

Al position of the truncated Side View

Al2O9 unit

0.198

Al2O9 unit Fig. 5

(a)

Model structure for Ni adsorption on an Al2 O3 (0001) surface.

(b)

Fig. 6 Anionic dangling bonds of Al2 O3 (0001) (a) and TiO2 (110) (b).

therefore, have a dangling bond directed to the center of the lost Al above the surface, as shown in Fig. 6. The Ni–O distance is determined to be 0.198 nm, which is similar to those found in Ni2þ compounds such as Ni(OH2 )6 2þ . This bond length indicates that the interaction between Ni and O should be covalent bonding with an overlapping of Ni 3d and O 2p orbitals accompanied by charge transfer. This is stronger than the Al–O interaction, and consequently Ni attracts the oxygen atoms, inducing surface relaxation. Ma et al. performed a DFT calculation on the same Ni–Al2 O3 (0001) system.22) They found that the adsorption site of Ni was site-3, which agrees with our EXAFS results. However, there are discrepancies in the Ni–O and Ni–Al distances, which were 0.227 nm and 0.263 nm in their work. When we calculated with their short Ni–Al distance of 0.263 nm, we

could not reproduce the EXAFS oscillation perpendicular to the surface, as shown in Fig. 4. This difference may come from the electronic state of Ni. The Ni atom in their calculation had a neutral charge and weak interaction with the surface, but in the real system, more charge transfers to the surface from the Ni. The atomically dispersed Ni structure is not entirely stable and aggregates around 5 1013 cm2 corresponding to less than 0.05 ML. This may be due to the surface charge balance. Ni has a slight positive charge, and the Al2 O3 surface should be negative in order to balance the charge. As a result, an upward dipole moment is created and increases with Ni loading. The dipole moment with the same direction is not stable due to the electrostatic repulsive force between the dipole moments. Subsequently, Ni aggregates to form a Ni cluster, accompanied by the return of electrons from the surface to the Ni cluster to reduce the surface dipole moment. Consequently, the Ni cluster forms with rather small coverage. We conclude that Ni interacts with oxygen dangling bonds to form Ni–O covalent bonds accompanied by electron transfer from Ni to oxygen. Note that the surface Al has dangling bonds, but those dangling bonds do not form a bond with Ni. We propose here that there should be cationic and anionic dangling bonds on the oxide surface that can accept anionic and cationic foreign atoms, respectively. 3.2 Ni on TiO2 (110) Figure 7 shows the surface structure of the TiO2 (110). Three types of atoms occur on the surface. One is protruding oxygen atoms running along the [001] direction, called bridging oxygen (OB ); the second is exposed Ti4þ between the bridging oxygen atoms, called fivefold Ti; the third is inplane oxygen atoms next to fivefold Ti. Bridging oxygen is believed to be easily removed from the surface, leaving a point defect site. Consequently, two Ti cations underneath are exposed to the surface.23) Other defect sites such as surface fivefold and subsurface sixfold Ti vacancies can be neglected because the TiO2 is an n-type semiconductor. Figure 8 shows Ni polarization-dependent EXAFS spectra with a coverage of 1 1013 atoms/cm2 . When 3 1013 atoms/cm2 Ni is deposited on the surface, which is the same Ni amount as on the Al2 O3 (0001), we observed Ni–Ni bonds clearly on the TiO2 (110) surface, indicating that either the number of adsorption sites on this surface is smaller than on the Al2 O3 (0001), or the interaction between TiO2 (110) and Ni is weaker.24) The amplitudes of all observed EXAFS oscillations decay quickly, indicating the presence of oxygen atoms, not Ti or Ni atoms, as the nearest neighbor. Ni is present in an isolated form. Preliminary curve-fitting analysis shows that the Ni–O distance is about 0.20 nm (0.199 nm for the [001] and [11 0] directions and 0.204 nm for the [110] direction).24) The bond length suggests that the bonding character is a covalent bond with electron transfer from Ni to oxygen. We have further analyzed EXAFS by comparing the observed EXAFS data with simulated EXAFS based on FEFF8. We postulate that Ni is adsorbed on the anion sites of the unreconstructed terrace (sites 1–6 in Fig. 7). Atop and bridging sites are rejected because the Ni–O should not be


513

A

Side View

5-fold Ti

4

Top View

OB

in-plane O

2 3 5

1 6

C Fig. 7

TiO2 (110) surface structure: side view (upper) and top view (lower). Large and small balls are oxygen and Ti atoms, respectively.

0.5 E // [001]

0.4 0.3

kχ(k)

0.2

E // [110]

0.1 0.0

-0.1

E // [110]

-0.2 -0.3 -0.4 -0.5 2

4

6

k/ 10 nm

8

-1

10

Fig. 8 PTRF-EXAFS spectra for Ni species on TiO2 (110) with three different polarizations.

observed in the [11 0] direction for these sites, contrary to the observed results. The Ni–O bonds in all directions are visible only at sites 3 and 4. However, we cannot reproduce the observed EXAFS data based on these models, as shown in Fig. 9. The poor reproduction of the experimental spectra results from the short Ni–Ti distance (Ti atom under the bridging oxygen shown by A or fivefold Ti in Fig. 7), which

always appears to be less than 0.25 nm. Adsorption site 4 was reported to be the most stable site based on the DFT calculation for the adsorption of Pd on TiO2 (110).25) We should mention that the dangling bond of oxygen is directed to the surface normal from the bridging oxygen, and the inplane oxygen atoms are coordinatively saturated. Thus, sites 3 and 4 are unable to make covalent bonds, but participate mainly in ionic bonds or van der Waals interaction. The Ni–O distances should be longer than that observed in this work. We have concluded that no stable adsorption sites are present on the terrace sites. Another possibility for a Ni adsorption site is a step edge. Figure 7 shows that TiO2 (110) has step edges, most of which run parallel to h001i or h11 ni (n ¼ 1; 2; 3 . . ..).26) We found that the h11 ni step edge can afford a suitable adsorption site that is demarcated by ellipse C in Fig. 7, and consists of two oxygen atoms: a bridging oxygen on the lower terrace (OL ) and an in-plane oxygen (OU ) on the upper terrace.26) Note that both oxygen atoms have dangling bonds, as shown in Fig. 6(b). We can reproduce the observed spectra well, as shown in Fig. 9(b), when we put Ni atoms on this site. The detailed structure of the model is shown in Fig. 10. We estimated the density of the adsorption sites as 2{3 1013 sites/cm2 based on a reported STM image.23,27) This value is larger than the Ni coverage studied here. The respective Ni–OL and Ni–OU distances are 0:205 0:003 nm and 0:200 0:002 nm. The Ni–OL distance is longer than that of Ni–OU . This asymmetric structure can be explained by

514


(a)

Fig. 9

(b)

Simulated XAFS data based on the TiO2 (110) surface: (a) Ni is adsorbed on site-3, and (b) Ni is located as shown in Fig. 10.

0.195 nm

0.200 nm

O Ti Ou

Ni

0.198 nm 0.205 nm

OL

Fig. 10 Ni structure on the TiO2 (110) surface.

the bulk structure of the rutile TiO2 , in which corresponding Ti–O bonds have the same anisotropy; Ti–OL and Ti–OU are 0.198 and 0.195 nm, respectively.24,28) Metal atoms are located at the site to which the dangling bonds of surface oxygen atoms are directed. The oxygen dangling bond on the terrace is located only at the bridging oxygen, directed to the [110] direction, which can be an adsorption site for Ni. However, the single Ni-bridging oxygen bond is not strong enough to fix it on the atop site. Consequently, Ni can hop to the next bridging oxygen atom and diffuse along the [001]

direction. The Ni atoms can find other Ni to form clusters before reaching the step edges if the surface density of Ni is high.27,29) At less than the critical coverage, Ni atoms can reach the step edge site without colliding with other Ni atoms. Here, two Ni–O bonds can fix the Ni atoms. This suggests that two bonds are necessary to stabilize the atomically dispersed Ni species. This is demonstrated by Cu species deposited on a thiophene carbonic acid-modified TiO2 (110) surface, where monoatomically dispersed Cu is stabilized on the surface by two covalent bonds of Cu–O and Cu–S,30) while Cu is easily aggregated otherwise. Atomically dispersed Ni is scarcely obtained on this surface because no stable adsorption site exists on the terrace site and the number of step edges is small. In this study, we have obtained three important results regarding the interaction between Ni and oxide surfaces. First, metal atoms such as those of Ni are stabilized by the interaction of surface oxygen atoms. Second, the stabilization requires more than two covalent bonds. Third, the adsorption site should have dangling bonds of oxygen atoms, while Ni does not adsorb on the cationic dangling bonds. We should consider the polarity of the dangling bond. Oxide surfaces contain cationic dangling and anionic dangling bonds. Ni preferentially becomes positively charged, and is thus adsorbed on the anionic dangling bond. Further theoretical calculation is necessary to explain the idea of the cationic and anionic dangling bonds and their adsorption properties.


Acknowledgment The work is financially supported by a Grant-in-Aid for Scientific Research on the Priority Area ‘‘Atomic Scale Modification (474)’’ from MEXT. REFERENCES 1) G. Ertl and H. Knozinger (Eds.): Handbook of Heterogeneous Catalysis, (Weinheim, VCH, 1997). 2) C. Xu, X. Lai, G. W. Zajac and D. W. Goodman: Phys. Rev. B 56 (1997) 13464. 3) J. A. Horsley: J. Chem. Phys. 76 (1982) 1451. 4) M. Sterrer, E. Fischbach, T. Risse and H. J. Freund: Phys. Rev. Lett. 94 (2005). 5) A. Del Vitto, G. Pacchioni, F. O. Delbecq and P. Sautet: J. Phys. Chem. B 109 (2005) 8040. 6) Y. Iwasawa (Eds): X-Ray Absorption Fine Structure for Catalysis and surface, (Singapore, World Scientific, 1996). 7) J. Stohr: X-ray absorption Principles, Applications, Techniques of EXAFS, SEXAFS, and XANES. ed by D. C. Koningsberger and R. Prins, (New York, 1988) p. 443. 8) L. G. Parratt: Phys. Rev. 95 (1954) 359. 9) S. M. Heald, E. Keller and E. A. Stern: Phys. Lett. 103A (1984) 155. 10) K. Asakura, W. J. Chun and Y. Iwasawa: Topics in Catalysis 10 (2000) 209. 11) W.-J. Chun, K. Asakura and Y. Iwasawa: J. Phys. Chem. 102 (1998) 9006. 12) K. Asakura: Cataly. Surv. Asia 7 (2003) 177. 13) W. J. Chun, Y. Tanizawa, T. Shido, Y. Iwasawa, M. Nomura and

515

K. Asakura: J. Synchrotron. Rad. 8 (2001) 168. 14) Y. Koike: Doctoral Thesis, (Hokkaido University, 2006). 15) J. Ahn and J. W. Rabalais: Surf. Sci. 388 (1997) 121. 16) P. Guenard, G. Renaud, A. Barbier and M. Gautier-Soyer: Surf. Rev. Lett. 5 (1997) 321. 17) C. F. Walters, K. F. McCarty, E. A. Soares and M. A. Van Hove: Surf. Sci. 464 (2000) L732. 18) J. J. Rehr, R. C. Albers and S. I. Zabinsky: Phys. Rev. Lett. 69 (1992) 3418. 19) K. Ijima, Y. Koike, W.-J. Chun, Y. Satio, Y. Tanizawa, T. Shido, Y. Iwasawa, M. Nomura and K. Asakura: Chem. Phys. Lett. 384 (2004) 134-38. 20) W.-J. Chun, K. Asakura and Y. Iwasawa: J. Phys. Chem. 102 (1998) 9006. 21) W. J. Chun: doctorol Thesis, (Tokyo University, Tokyo, 1998). 22) Q. Ma, K. Klier, H. Cheng, J. W. Mitchell and K. S. Hayes: J. Phys. Chem. 105 (2001) 2212. 23) U. Diebold: Surf. Sci. Reports 48 (2003) 53. 24) Y. Koike, K. Ijima, W. J. Chun, H. Ashima, T. Yamamoto, K. Fujikawa, S. Suzuki, Y. Iwasawa, M. Nomura and K. Asakura: Chem. Phys. Lett. 421 (2006) 27. 25) J. F. Sanz and A. Marquez: J. Phys. Chem. C 111 (2007) 3949. 26) U. Diebold, J. Lehman, T. Mahmoud, M. K. Kuhn, G. Leonardelli, W. Hebenstreit, M. Schmid and P. Varga: Surf. Sci. 411 (1998) 137. 27) K. Fujikawa, S. Suzuki, Y. Koike, W.-J. Chun and K. Asakura: Surf. Sci. 600 (2006) L117. 28) E. A. Grant: Rev. Mod. Phys. 31 (1959) 646. 29) Y. Koike, K. Fujikawa, S. Suzuki, W. J. Chun, K. Ijima, M. Nomura, Y. Iwasawa and K. Asakura: J. Phys. Chem. C 112 (2008) 4667. 30) W. J. Chun, Y. Koike, K. Ijima, K. Fujikawa, H. Ashima, M. Nomura, Y. Iwasawa and K. Asakura: Chem. Phys. Lett. 433 (2007) 345.