Supporting Information

Transcription

1 Supporting Information Real time observation of reconstruction dynamics on TiO 2 (001) surface under oxygen via an environmental TEM Wentao Yuan, Yong Wang, *, Hengbo Li, Hanglong Wu, Ze Zhang, *, Annabella Selloni, and Chenghua Sun *, Center of Electron Microscopy and State Key Laboratory of Silicon Materials, School of Materials Science and Engineering, Zhejiang University, Hangzhou , China Department of Chemistry, Princeton University, Princeton, NJ 08544, USA ARC Centre for Electromaterials Science, School of Chemistry, Monash University, Clayton, Victoria 3800, Australia * yongwang@zju.edu.cn, zezhang@zju.edu.cn, Chenghua.Sun@monash.edu 1

2 Contents of Supporting Information: Materials and methods Supporting Text: Figures S1-S17 References Movies S1-S8 2

3 Materials and methods: 1. Preparation of the anatase TiO 2 nanosheets: The anatase TiO 2 nanosheets with a side length of ~30 nm and a thickness of ~5 nm are prepared by a previously published hydrothermal route, with hydrofluoric as the capping agent. 1,2 In a typical synthetic procedure, 10 ml of tetrabutyl titanate and 1.2 ml hydrofluoric acid (50 wt %) were mixed in a Telfon-lined 50 ml autoclave and then kept at 200 C for 24 h. After the reaction, the white precipitates were collected by centrifugation and washed by deionized water and ethanol for several times, followed with a drying in an oven at 80 C overnight. To remove the fluorine ions absorbed on the {001}, the NaOH (1 mol/l) solution was used to wash the samples above. 2. Characterization The morphology and structure of the as-synthesized nanosheets was characterized by transmission electron microscopy (TEM, FEI F20). Surface compositions and the bonding states of the samples before/after NaOH cleaning were analyzed by X-ray photoelectron spectroscopy (XPS, Thermo Scientific ESCALAB 250Xi). 3. In-situ ETEM experiments: The in-situ observation of the reconstruction of anatase TiO 2 (001) surface was performed in an environmental transmission electron microscope (ETEM) with oxygen injection available (Hitachi H-9500). The TiO 2 nanosheets dissolved in ethanol solution were dispersed on a Mo micro-grid or a micro-chip and then loaded onto a TEM heating holder (Hitachi or Protochips). In-situ experiments in ETEM were carried out at 300 kv with an oxygen pressure of Pa at 500 C. An AMT digital camera and a DE-12 camera system (Direct Electron, LP) were employed to in-situ record the dynamic reconstruction behaviors. The frame rates are set for 6 fps and 10 fps on AMT digital camera and DE-12 camera, respectively. 3

4 4. Electron energy loss spectroscopy analysis: The electron energy loss spectroscopy (EELS) analysis was performed in a FEI Titan G ChemiSTEM operated at 200 kv with a nominal energy resolution of 1.1 ev. The TiO 2 nanosheets were loaded on a heating-chip (Protochips Inc.) without carbon film, which can be utilized to heat samples in our ETEM (Hitachi H-9500) and Titan G ChemiSTEM through Aduro sample holders (Protochips Inc.). Firstly, the sample were processed under the reconstruction conditions in ETEM for more than 60 mins (Hitachi H-9500, 500 C, oxygen pressure: Pa), allowing most of the samples to undergo reconstruction. Then the reconstructed samples were loaded into the Titan G ChemiSTEM for EELS characterization. To avoid the deposition of carbon in TEM column and to keep the reconstructed surface from contamination, the EEL spectra of samples were obtained at 500 C in vacuum ( Pa), through both imaging mode and STEM mode. Under the imaging mode, the spectra were collected with a magnification of 290 kx and an illuminated area diameter of ~150 nm (~100 pieces of nanosheets randomly distributed in the illuminated area). No objective aperture was used in the spectra collection and the collection semi-angle is expected larger than 100 mrad. Under the STEM mode, the convergence angle is 21 mrad and the collection angle is 19 mrad. The probe size is estimated to be ~0.14 nm in our experimental conditions and the beam current is ~0.12 na. Through both imaging mode and STEM mode, we did not detect significant signals of the possible contamination (fluorine and carbon). Indeed, one may see significant difference between the two images in Figure S7 where amorphous contamination can be easily removed by heating in oxygen. However, one cannot completely rule out the existence of a few single atoms (C or F) although EELS has been successfully employed to detect a single atom in recent work 3,4. Nevertheless, as we state in the later page, the reconstruction is an intrinsic behavior of TiO 2 (001) surface, which is well consistent with what observed in STM studies. 4

5 Supporting Text: A. Details about the surface reconstruction: 1. Surface species When hydrofluoric acid is used as the capping agent for synthesizing TiO 2 nanosheets, species containing fluorine and other organic species can absorb on the surface of the TiO 2 nanosheets. These organic species mainly serve as stabilizer to keep highly reactive (001) surface with (1 1) configurations. After we use a NaOH solution to wash the nanosheets, the XPS spectra show that more than 62 % of the fluorine adsorbed on the {001} surface has been removed (seeing Figure S2). Although there are still fluorine and some organic species adsorbed on the TiO 2 nanosheets, we can rule out any significant influence of these species on our experiments. First, previous works reported that the surface fluorine can be removed by calcinating at a high temperature. 5 Second, all species adsorbed on the surface of TiO 2 nanosheets can be effectively removed with the aid of electron beam irradiation in oxygen environment at 500 C (refer to Movie S2 and Figure S7), and more discussion will be appeared in the later paragraph. Third, we also performed EELS analysis for the reconstructed surface. As shown in Figure S8, we did not detect the significant carbon or fluorine signals (main adsorbed species in our experiments; C K-edge: 284 ev and F K-edge: 685 ev) in the EELS profile for the reconstructed sample. Fourth, we note that the TiO 2 surface could be easily destroyed layer by layer at experimental temperature by electron beam radiation in nitrogen atmosphere ( Pa; 500 o C), as shown in Figure S15 and Movie S8. Contaminants on the sample surface, if any, will also be removed under the electron beam irradiation. The exposed surfaces should be clean surfaces (Figure S15e), where around two nanometer TiO 2 on the surface was removed. In such case, we can still observe the formation of surface reconstruction in the TiO 2 (001) surfaces, which indicates that the surface reconstruction is indeed an intrinsic behavior, not induced by contaminant adsorption. 2. Surface reconstruction and evolution In addition to the most stable (1 4) structure, a significant amount of (1 n) (n=3, 5) patterns usually form on the (001) surface, before complete equilibration. The reconstructed surface thus shows regions with mixed (1 n) (n=3, 4 and 5) periodicities, with a predominance of the (1 4) one. Figure S6 shows such a case. However, the (1 n) (n=3, 5) configurations are not as stable as the (1 4) one. They are rather active and often trigger a surface rearrangement by reacting with 5

6 other surface structures. Generally, this rearrangement results in a decrease of the (1 n) (n=3, 5) configurations and an increase of the (1 4) structure (exemplified by Figure S6 and Movie S3). B. Calculation of surface energies All calculations have been performed using the projector-augmented wave (PAW) method as implemented in the Vienna ab initio simulation package. 6 The Density Functional Theory (DFT) calculations have been conducted within the generalized gradient approximation (GGA), using the Perdew-Burke-Ernzerh of functional All structures were fully relaxed until both the total energy and the atom force converged to 10-4 ev and 0.01 ev/å, respectively. A plane-wave basis set with the kinetic energy cutoff of 400 ev is employed. The Brillouin zone has been sampled using Γ-centered Monkhorst-Pack k-point grids. For the surface energy calculations, one-side reconstructed slab models have been employed, with the bottom layer (O-Ti-O) fixed, as shown below. For our calculations, we employed slabs of four O-Ti-O trilayers with the atoms of the bottom O-Ti-O trilayer kept fixed in their bulk-like positions. We used a total of 60, 51, 51, 52, 52 atoms for the MF(Ti 20 O 40 ) AMR(Ti 17 O 34 ) ADM(Ti 17 O 34 ) AOM(Ti 17 O 35 ) ARM(Ti 17 O 35 ) models, respectively. All surface energies were calculated based on the following expression, γ Where, is the total energy of the slab, n is the total number of TiO 2 units in the slab model, m is the number of additional O atoms if the slab is nonstoichiometric (m=0 if the slab is stoichiometric); is the energy per bulk TiO 2 unit, is the oxygen chemical potential, is the surface energy of the slab side which is fixed. 11 The unrelaxed slab models are shown in Figure S3. With respect to ADM, the AOM and ARM models have one additional O atom which is located on the surface and in the subsurface, respectively. A vacuum space of 18 Å is used to avoid the interaction between the slabs. The optimization and single point 6

7 energy for the calculation of surface energies were obtained by standard DFT. The calculated diagram of surface energy is shown in Figure S4a while the relaxed slab models are presented in Figure S4b. Since the MF, AMR and ADM models are ideally stoichiometric, their surface energies do not depend on the oxygen chemical potential ( ). 11 On the other hand, the ARM and AOM models are not stoichiometric, and therefore their surface energies depend linearly on. Among the five investigated models, ADM and AOM have the lowest surface energies. For <-0.34 ev, the ADM model (surface energy of mev/å 2 ) is preferred, while the AOM model becomes more stable under oxygen rich conditions ( >-0.34 ev). This result indicates that the ADM configuration may exist at low (high temperature and/or low oxygen partial pressure) while the AOM configuration may dominate at higher (low temperature and/or high oxygen partial pressure). For the conditions of our experiment (temperature of 500 C and oxygen partial pressure of ~ Pa), we estimate = ev, as indicated by the green vertical dash-line in Figure S4a. It is evident that ADM has the lowest surface energy compared to the other models. Furthermore, the simulated HRTEM image based on the ADM model agrees well with our experimental data (Figure 1d). All our results thus indicate that the ADM model well describes the TiO 2 (001)-(1 4) reconstruction in our experiments. C. Effect of the electron beam and oxygen in the experiments: It is believed that electron beam (e-beam) plays a dual role in our experiments: on the one hand, it removes surface contaminants, and on the other hand, it would inevitably cause irradiation damage for the samples. (1) Removal of the surface contamination: When we analyze the formation process of the reconstruction (Movie S1), we initially see some amorphous species adsorbed on the TiO 2 surface (also refer to Figure 1a in the manuscript and Figure S7a). Under e-beam irradiation, the adsorbed organics species are rapidly removed (one more case can be found in Figure S7) mainly via the following three possible ways 12,13 : (a) through spattering effect (similar with the knock-on process), surface contamination can be directly removed; (b) the e-beam induces breakage of the bonds and further results in radiolysis of the contaminants; (c) the e-beam facilitates the oxidation of hydrocarbon species. In fact, the adsorbed species can be removed by oxidation also without the aid of e-beam irradiation, but they need relatively longer times (~10-30 s and > 100 s, with 7

8 and without e-beam irradiation, respectively). (2) Irradiation damage of the TiO 2 nanocrystals: The metal oxides become unstable when they are exposed to the E-beam irradiation in vacuum. In particular, the oxygen atoms in TiO 2 crystals can be ejected under continuous E-beam irradiation, by the Knotek-Feibelman mechanism. 12,13 Here, a similar process has been observed, notably that the TiO 2 crystal undergoes gradual dissolution under vacuum conditions ( Pa; 500 o C) as shown in Figure S14. However, if we introduce oxygen into the TEM chamber (O 2 pressure: Pa; 500 o C), we do not observe significant E-beam irradiation damage (Movie S1-S7). Based on these findings, we can rule out the effect of the e-beam irradiation in the formation process. As the 300 kv electrons indeed result in a serious irradiation damage on TiO 2 crystal without oxygen protection, we further explored the situation at a lower voltage (100 kv). At 100 kv, we observed the reconstruction (please refer to the larger TiO 2 nanocrystal in Figure S17) and the e-beam irradiation (refer to the smaller TiO 2 crystal in Figure S17). However, compared with the situation at 300 kv, the radiolysis of the sample at 100 kv is not that significant and takes longer time (radiolysis time: < 10 min (300 kv), > 20 min (100 kv) ). Although the radiolysis is reduced at lower voltage, the quality of the images becomes worse with respect to the 300 kv. Therefore, 300 kv is still recommended. The oxygen also plays important roles in our experiments. First, it protects our samples from the E-beam irradiation damage. One series of controlled trials well illustrates this point. When we turn off the oxygen in our experiments ( Pa; 500 o C), the TiO 2 nanosheets are quickly damaged by the E-beam irradiation, as shown in Figure S14. And likewise, in N 2 atmosphere, (as shown in Figure S15 and Movie S8) the TiO 2 nanosheets also undergo rapid dissolution (nitrogen pressure: Pa; 500 o C). It is thus reasonable to infer that the oxygen atmosphere provides an appropriate environment for us to dynamically observe the formation and evolution of the TiO 2 reconstruction in ETEM. Second, the oxygen can react with the adsorbed hydrocarbon species on the surface with the aid of e-beam irradiation. Although we cannot rule out the possibility that environmental oxygen is involved in the reconstruction, our experiments suggest the presence of oxygen is not a necessary condition for the reconstruction. In fact, the reconstruction has been observed also under vacuum conditions and in pure N 2 atmosphere (Figure S16). Thus we conclude that the (1 4) reconstruction is an 8

9 intrinsic behavior of the anatase (001) surface. Movie S8 provides a powerful evidence, that in N 2 atmosphere (nitrogen pressure: Pa; 500 o C), TiO 2 surface could be easily destroyed layer by layer at experimental temperature by e-beam irradiation. So contaminants on the sample surface, if any, can also be removed readily, leading to exposed clean surfaces, as shown in Figure S15, where around two nanometer TiO 2 on the surface was removed. In such case, we can still observe the formation of surface reconstruction in the TiO 2 (001) surfaces, which indicates that the surface reconstruction is indeed an intrinsic behavior, not induced by contaminant adsorption or oxygen. D. Ti or TiO x on the outmost layer? The (1 4) surface reconstruction on anatase TiO 2 (001) has been widely studies by STM and several models were proposed to explain such reconstruction [ref 13,15,17 and 18 in the manuscript]. Based on our experimental conditions (500 o C and Pa oxygen), the ADM model is verified to be the most stable one, which shows the outmost layer is TiO x. However, it should be noted that it is rather difficult to distinguish the Ti or TiO x in the TEM images based on phase contrast. First, note that the titanium atoms can be easily oxidized and the surface oxidation of the pure metal titanium can even occur below 300 C 14, so it is reasonable to expect the outmost layer is TiO x at our experimental conditions (500 C and oxygen environment). Second, the STM studies and DFT calculations have confirmed that the (1 4) surface reconstruction is based on the ADM model, with which the simulated HRTEM images match well with the experimental ones; Third, to further confirm our conclusion, we perform DFT-based molecular dynamics calculations (refer to Figure S9). It is assumed Ti exists as TiO x in our original model, which is based on several tests on the stability of Ti under oxygen. Single Ti atom (highlighted by blue circle) on TiO 2 has been examined when oxygen is introduced. Both optimization at 0 K and DFT-based molecular dynamics (773 K, 1.8 ps) confirmed that Ti will be oxidized rapidly, as shown in Figure S9, indicating that one-coordinated Ti is not stable. Based on the above statement, we believe the outmost layer in our experiments is indeed TiO x. 9

10 E. Calculation of the energy profile associated with the 3d-to-4d transition The energy profile shown in Figure 4d was calculated based on a slab with the periodical boundary condition, as shown in Figure S13, in which there is only one unit along the row. So it described the infinitely long row. During the calculations, geometries and energies were further corrected using the DFT plus Hubbard model with U=4.0 ev. 10

11 Supporting Figures: Figure S1. TEM images of the as-synthesized anatasetio 2 nanosheets. (a) Low-magnification TEM images of TiO 2 nanosheets. (b) HRTEM images of the TiO 2 nanosheets, viewed from the [010] direction. The inset shows the morphology model of the TiO 2 nanosheet, which is dominated by the (001) surface. Figure S2. X-ray photoelectron spectra of the as-synthesized/naoh-washed TiO 2 nanosheets. (a) Full Spectrum. (b) Spectra of F adsorbed on the surface of the nanosheets. (The green circles are the raw data and the red/blue lines represent the fitted lines, respectively.) 11

12 Figure S3. Unrelaxed slab models of the reconstructed anatase TiO 2 (001)-(1 4) surface. (a) The microfacet model (MF); (b) The added row model (ARM); (c) The added and missing rows model (AMR); (d) The add oxygen model (AOM); (e) The ad-molecule model (ADM). Figure S4. Surface energies (mev/å 2 ) vs oxygen chemical potential ( ). (a) The DFT calculated surface energies of the models proposed previously vs oxygen chemical potential ( ) or O 2 pressure. (b) Relaxed slab models of the reconstructed anatase TiO 2 (001)-(1 4) surface. The vertical dashed green line at = ev corresponds to our typical experimental conditions (oxygen pressure of Pa and temperature of 500 o C). 12

13 Figure S5. HRTEM images of the bulk-terminated (001) surface. (a) An enlarged view of the bulk-terminated (001) surface shown in Figure 1a. (b) The intensity profiles along the colored dashed lines in (a). Figure S6. Atomic formation and evolution process of the (1 n) reconstructions. (Movie S3, The oxygen pressure is kept at Pa and the temperature is kept at 500 o C.) (a-f) Sequential TEM images of the anatase TiO 2 (001) surface during the formation and evolution of the reconstruction, acquired at 0 s, s, s, s, s and s. Figure S7. The sequential images show the reconstruction formed after the removal of the species adsorbed on surface. (The oxygen pressure is kept at Pa and the temperature is kept at 500 o C.) The layered adsorbed species disappeared under the e-beam irradiation and the (1 4) reconstruction shows up after the disappearance of the adsorbed species. 13

14 Figure S8. EELS profile of the TiO 2 nanocrystals after surface reconstruction. (The dispersion was set for 0.05 ev per channel. carbon K-edge: 284 ev and fluorine K-edge: 685 ev) The spectra (a) and (b) were collected in imaging mode with a current density of 1500 e/å 2 ; the collection times are 2 s and 20 s for (a) and (b), respectively. The spectra (c) and (d) were collected under the STEM mode and the corresponding collection positions are shown in the insets of (c) and (d) (the scale bar in the inset of (c) indicate 10 nm); The probe current is ~0.12 na and the collection time is 5 s. Figure S9. DFT-based molecular dynamics calculations of single Ti atom on anatase TiO 2 (001) surface (Ti: gray and O: red). 14

15 Figure S10. Atomic evolution of the reconstructed (001) surface. (Movie S5, the oxygen pressure is kept at Pa and the temperature is kept at 500 o C) (a) Sequential HRTEM images of the dynamic structural evolution of (1 n) reconstructed (001) surface, at atomic scale. (The red arrows indicate the unstable double-row states during the evolution.) (b) Statistical diagram of the locations of the TiO x rows in Movie S5. The five rows (from top to bottom) correspond to the five vertical bars (from right to left) in (a), where each bar indicates a ridge on the (001) surface. 15

16 Figure S11. Sequence of TEM images of an oblique nanosheet showing the stabilities of 4d and 5d configuration. (Movie S6, the oxygen pressure is kept at Pa and the temperature is kept at 500 oc.) It can be seen that during the whole process, the TiOx rows in 5d configuration, marked by dotted yellow ellipse, are more reactive than those in 4d configuration. 16

17 Figure S12. Series of TEM images show a typical migration of TiOx row during the structure evolution. (Movie S7, the oxygen pressure is kept at Pa and the temperature is kept at 500 oc) The morphology model of the TiO2 nanosheet is shown in the inset of (a). The dotted yellow ellipses in (b-m) indicate the evolution regions, which encompass an unstable (1 x 5) configuration. The (1 n) (n=3, 5) configurations are eliminated through the continuous migration of TiOx rows from the right (inner part) to the left (surface edge). 17

18 Figure S13. Slab model used to calculate the energy profile for the transition from 3d to 4d based on a (12 1) supercell. (a & b) Side and top view of TiO 2 with 3d pattern and (c & d) Side and top view of TiO 2 with 4d pattern. The reconstructed TiO x row is shown as big balls, and the other part shown as small ball and stick. Figure S14. In situ observation of the reconstructed (001) surface in vacuum. (We turn off the oxygen when the reconstruction is formed. The oxygen pressure is kept at Pa and the temperature is kept at 500 o C) (a) The TEM image of the reconstructed (001) surface. (b) The reconstructed surface becomes disordered under the e-beam irradiation. (c, d) With further e-beam irradiation, the reconstruction disappears and the TiO 2 nanosheet becomes highly damaged. 18

19 Figure S15. In situ observation of the reconstruction in N 2 atmosphere. (Movie S8, nitrogen pressure: Pa; 500 o C) (a-f) Sequential TEM images show a TiO 2 nanosheet with (1 4) reconstructed (001) surface becoming gradually decomposed under E-beam irradiation. During the radiolysis, the (1 4) reconstruction continues to spontaneously form on the (001) surface at each time. The dashed blue line indicates the initial height of the sample. 19

20 Figure S16. Stability of the reconstruction under different conditions. (a) Snapshot of the reconstructed surface without introducing oxygen (Pressure: Pa, 500 o C). (b) The reconstruction formed in N 2 atmosphere (Nitrogen pressure: Pa, 500 o C). (c) A reconstructed sample, after 24 hours in atmosphere (Snapshot in vacuum condition, pressure: Pa, 20 o C). Figure S17. The reconstructed TiO 2 samples under 100 kv e-beam irradiation without oxygen protection. ( Pa; 500 o C) 20

21 References: 1 Han, X. G.; Kuang, Q.; Jin, M. S.; Xie, Z. X.; Zheng, L. S. J. Am. Chem. Soc. 2009, 131, Wu, X.; Chen, Z. G.; Lu, G. Q.; Wang, L. Z. Adv. Funct. Mater. 2011, 21, Egerton, R. F. Electron energy-loss spectroscopy in the electron microscope Third Edition. (Springer, 2011). 4. Senga, R.; Suenaga, K. Nat. Commun. 2015, 6, Liu, G.; Yang, H. G.; Pan, J.; Yang, Y. Q.; Lu, G. Q.; Cheng, H. M. Chem. Rev. 2014, 114, Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140, Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54, Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, Kresse, G.; Joubert, D. Phys. Rev. B 1999, 59, Reuter, K.; Scheffler, M. Phys. Rev. B 2001, 65, Egerton, R. F.; Li, P.; Malac, M. Micron 2004, 35, Egerton, R. F.; McLeod, R.; Wang, F.; Malac, M. Ultramicroscopy 2010, 110, Lu, G.; Bernasek, S. L.; Schwartz, J. Surface Science 2000, 458,

22 Caption for Movies: Movie S1. The in-situ formation of the (1 4) reconstruction on anatase TiO 2 (001) surface observed from the [010] direction. Movie S2. The in-situ formation of the (1 4) reconstruction on anatase TiO 2 (001) surface observed from the [010] direction. Movie S3. The in-situ formation and evolution of the reconstruction on anatase TiO 2 (001) surface observed from the [010] direction. Movie S4. The in-situ observation of the reconstructing anatase TiO 2 (001) surface from the [010] direction (3d to 4d). Movie S5. The in-situ observation of the reconstructing anatase TiO 2 (001) surface from the [010] direction (5d to 4d). Movie S6. The in-situ observation of the reconstructing anatase TiO 2 (001) surface from a top view. Movie S7. The in-situ observation of the reconstructing anatase TiO 2 (001) surface from a top view. Movie S8. In situ observation of the dissolution of anatase TiO 2 nanosheet in N 2 atmosphere. 22