Supplementary Information. Probing of 2 dimensional confinement-induced structural transitions in amorphous oxide thin film

Transcription

1 Supplementary Information Probing of 2 dimensional confinement-induced structural transitions in amorphous oxide thin film Sung Keun Lee 1* and Chi Won Ahn 2 1School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea 2National Nanofab Center, Korean Advanced Institute of Science and Technology, Daejeon , Korea *Correspondence and requests for materials should be addressed to S.K.L ( sungklee@snu.ac.kr, web: ADDITIONAL MATERIALS AND METHODS TEM imaging. Cross-section TEM images of as-deposited alumina thin film with varying thickness were obtained. TEM specimens for 5 nm and 25 nm thin films were prepared using conventional low-angle Ar milling and then attached to 3mm Cu-ring grid, which fits in the TEM holder. Those for 10 nm, 60 nm, and 1.4 µm thin film were prepared using focused ion beam (FIB) workstation and then attached to Cu-ring grid. The cutting of the cross-section lamellae was done at 30 kv. Prior to FIB, the thin film surface (10 and 60 nm) was coated with C and then Au. The surface of 1.4 um film was coated with Pt (ref. 1 ). To reduce Ga incorporation and FIB-induced damage, the milling was finally polished at 5 kv using a current of 70 pa. The TEM images for 10 and 60 nm was obtained in a TEM [JEM- 2100F HR (FE-TEM, JEOL Ltd., Co.)] operated at 200 kev. The images for three representative films (5 nm, 25 nm, and 1.4 µm) were obtained in a TEM [JEM-ARM200F (JEOL Ltd., Co.)] operated at 200 kev. TEM image showed no clear evidence of the presence of crystalline phases in the as-deposited thin film (Figure S1). 1

2 Figure S1 Cross-section HRTEM image of as-deposited alumina thin film with varying nominal thickness from (A) 5 nm, (B) 10 nm, (C) 25 nm, and (D) 60nm. The TEM image of 1.4 µm thin film was previously reported 1. White arrow between amorphous Si 3N 4 and Al 2O 3 refers to interlayer. Wet-etching. Figure S2 shows the wafer that was etched following the procedures described in the main text. The middle transparent part etched by KOH was used for NMR analysis. The upper surface of the thin film was protected by O-ring and stainless steel jig. Etching selectively removed Si in the middle parts while the Si in the boundary was preserved (brownish part). The figure clearly shows that boundary of the wafer and thus the upper thin film sides are selectively protected upon wet-etching. We note that if the etching process would damage the thin film, the boundary part would have not been survived and 5 nm Al 2O 3 layers would have been completely removed immediately, leaving no NMR signal. This is clearly not the case here as also shown in figure S3 below. 2

3 Figure S2 Si wafer with Si 3N 4/5 nm Al 2O 3 thin film. Middle part of the Si was removed, yielding transparent Si 3N 4/5 nm Al 2O 3 thin film part used for NMR analysis. SI TEXT 27Al MAS NMR Background subtraction Figure S3 presents 27 Al MAS NMR spectra of the amorphous Al 2O 3 thin films (with varying thickness) and rotor background. The rotor background consists mostly of [6] Al. Figure 1 in the manuscript was obtained by subtracting the background signal from the spectrum. Note that additional spacer (with Al impurity) was added to fill the empty space in the NMR rotor due to small sample volume of 5nm thin film (~10 mg, rather than ~20 mg of thicker thin films) and thus the shape and the intensity of the backgrounds for 5nm thin films are slightly different from those rotors used for other thin films. The NMR spectra for 1.4 µm thin film and the background can be found at the previous report 1. 3

4 Figure S3 27 Al MAS NMR spectra of the background-subtracted amorphous Al 2O 3 thin film (black) and rotor background (red) with varying thickness as labeled. Quantification of 27 Al MAS NMR spectra for amorphous Al 2O 3 thin films In order to obtain quantitative fractions of [4,5,6] Al sites in the amorphous Al 2O 3 thin films, the 27 Al MAS NMR spectra for thin films were simulated using the Czjzek distribution functions for [4,5,6] Al sites using the DMFIT software 2. The Czjzek function takes into consideration quadrupolar peakshape with structural disorder and thus the distribution of electric field gradient tensors [see 3,4 for more detailed discussion on the model]. The method has been effective in providing quantitative fractions of [4,5,6] Al sites in diverse non-crystalline oxides (e.g., 4 ). Figure S4 shows the estimated fractions of [4,5,6] Al sites in the thin film with varying thickness. The NMR parameters used for simulations are also shown in table A1. Note that the quadrupolar coupling constant C q for [4,5,6] Al sites used for the simulation for bulk 4

5 alumina is systematically larger (~0.7-1 MHz) than those reported from our previous 2D NMR study of the 1.4 µm amorphous alumina thin film 1,5. The current study with 1D MAS NMR study yields C q values of 7.8 MHz for [4] Al, 7.2 MHz for [5] Al, and 5.2 MHz for [6] Al; while previous 2D 3QMAS study reported C q values of 7.1 MHz for [4] Al, 6.2 MHz for [5] Al to 4.5 MHz for [6] Al. Similar difference between C q values from 1D 27 Al MAS and 2D 27 Al 3QMAS NMR spectra has been reported for 27 Al NMR parameters for Ca-aluminosilicate glasses: C q of [5] Al site in CaAl 2SiO 6 glass (~ 7.8 MHz) from 1D 27 Al MAS NMR is larger than these obtained from the 2D 3QMAS NMR study (~4-5 MHz) 4,6,7. These difference results from the fact that the 2D 3QMAS study at 9.4 T may not excite the Al sites with larger C q effectively and lineshape analysis of 1D MAS NMR can be model dependent and may systematically overestimate C q values 4,6,7. Although the results with 1D MAS NMR spectra provide quantitative fraction, we note that the estimated absolute [5] Al fraction can vary slightly depending on the methods used to extract the quantitative fractions. Taking into consideration uncertainty in NMR parameters used for the simulation, there is uncertainty in the absolute fraction of each Al site used in the current study. However, as shown in figure S3 (and figure 1), the [5] Al contribution to total peak intensity clearly decrease with decreasing thickness and the relative difference in fraction [5] Al between the films with varying thickness is unlikely to be variant. 5

6 Figure S4 Estimated fractions of [4,5,6] Al sites in in amorphous Al 2O 3 thin film as a function of deposition thickness (d) from 1.4 µm to 5 nm. Table S1 NMR parameters used for simulations and retrieved fractions of Al sites in amorphous alumina thin film with varying thickness Al site Chemical Shift δ Cs (ppm) Variation in Δδ Cs (ppm) Quadrupolar coupling constant C q (MHz) Intensity (%) [4] Al 73.5± ± ± ±5 5 nm [5] Al 41.3± ± ± ±5 [6] Al 13.8± ± ± ±5 [4] Al 69.5± ± ± ±5 10 nm [5] Al 38.7± ± ± ±5 [6] Al 10.8± ± ± ±5 [4] Al 68.2± ± ± ±5 60 nm [5] Al 40.0± ± ± ±5 [6] Al 10.0± ± ± ±5 [4] Al 69.5± ± ± ±5 250 nm [5] Al 41.3± ± ± ±5 [6] Al 9.4± ± ± ±5 [4] Al 68.5± ± ± ± µm [5] Al 41.3± ± ± ±3.5 [6] Al 9.4± ± ± ±3.5 6

7 Insights from at initio calculation of stability of Al coordination cluster with [3] O In order to evaluate the relative stability of Al clusters in the amorphous Al 2O 3 network, we performed ab initio calculations of Al clusters with [3] O using the Gaussian03 (ref. 8 ). The optimized model Al-clusters include those involving [3] O with two [5] Al and one [4] Al (2 [5] Al- [3] O- [4] Al) and with two [4] Al and one [6] Al (2 [4] Al- [3] O- [6] Al) (figures S5 and S6). [3] O- [4,5,6] Al bond length within each coordination polyhedron (with tetrahedral, trigonal bipyramidal, and octahedral symmetry) are set to be identical. Without these Al symmetry constraints, clusters with [3] O are not sustained, forming [2] O. This indicates that the stability of [3] O depends on the presence of networks around it. The clusters were charge-balanced by terminal hydrogen. The geometry of the clusters was optimized with flexible (figure S5) and fixed the dihedral angle (figure S6) involving the terminal H. These clusters were optimized at the B3LYP hybrid density functional theory and the 6-311G(d) and G(2d,p) basis sets (to check basis set effect). Single point energy of these optimized cluster were calculated using B3LYP/6-311+G(2d,p), which has reproduced experimental bond energies for oxide glass clusters relatively well 9,10. Equilibrium Geometries. Figures S5 (with flexible dihedral angle for terminal hydrogen) and S6 (with fixed dihedral angle for terminal hydrogen) show the optimized model Al clusters with [3] O. The [4] Al-O, [5] Al-O, and [6] Al-O bond distances are Å, Å, and Å, respectively. While [4] Al-O distance is slightly longer than that simulated from aluminosilicate clusters with only [2] O (1.758 Å), Al-O are in general consistent with previously reported average bond lengths in the glasses at high pressure obtained using quantum simulations (e.g., 1.854, and Å for [5] Al-O, and [6] Al-O, respectively) 9,10. Table S2 shows that calculated single point energy and optimized internal variables of the model clusters. Relative stability of Al clusters. The single point energy difference between clusters (2 [5] Al- [3] O- [4] Al and 2 [4] Al- [3] O- [6] Al) give insights into ε {=1/2*[*E(2 [5] Al)-E( [4] Al+ [6] Al)]} in Eq. 1. The calculated energy differences between clusters per Al vary ~ from 19 kj (clusters 7

8 with fixed dihedral angle for terminal hydrogen) to 30 kj/mol (clusters with flexible dihedral angle for terminal hydrogen) (Table S2). The calculation confirms that [5] Al with [3]O is indeed a higher energy cluster as also predicted from the our modeling approach. While the current approach with fixed symmetry constraints is by no means complete (distortion of Al polyhedron is expected in the real system), it gives a clear and sufficient insight into the relative stability of Al coordination polyhedra with [3] O. The prevalence of [5]Al in bulk amorphous alumina is thus likely due to the fact that volume of amorphous networks in an extended length scale is required to sustain the high-energy clusters with [5]Al and [3] O. With decreasing thickness of the thin film, the [5] Al- [3] O species become less stable without additional strain reservoir around them. Figure S5 The optimized geometry of Al clusters with flexible dihedral angle involving terminal hydrogen [2 [5] Al- [3] O- [4] Al cluster (right) and 2 [4] Al- [3] O- [6] Al cluster (left)]. Red dotted lines indicate hydrogen bonding. Figure S6 The optimized geometry of Al clusters with fixed dihedral angle involving terminal hydrogen. 8

9 Table S2 Optimized geometries, single point energies, and simulation conditions for the Al clusters. # Al Clusters Basis Sets Equilibrium Geometry Energy Energy/Al/T g E Clusters used Single point energy Geometry optimization [4] Al-O (Å) [5] Al-O (Å) [6] Al-O (Å) kj J/K J/K 1 2 [5] Al [4] Al* G(2d,p) G(2d,p) (1-5)** 2 2 [5] Al [4] Al* G(2d,p) 6-311G(d) (2-6) 3 2 [5] Al [4] Al** G(2d,p) G(2d,p) (3-7) 4 2 [5] Al [4] Al** G(2d,p) 6-311G(d) (4-8) 5 2 [4] Al [6] Al* G(2d,p) G(2d,p) [4] Al [6] Al* G(2d,p) 6-311G(d) [4] Al [6] Al** G(2d,p) G(2d,p) [4] Al [6] Al ** G(2d,p) 6-311G(d) B3LYP energy level of theory is used for the simulations. *optimized with flexible Al-O-H dihedral angle (Figure S5) **optimized with fixed Al-O-H dihedral angle (Figure S6) ***(1-5) refers to (energy of cluster 1 - energy of cluster 5) T g value of 973 K was used REFERENCES 1 Lee, S. K., Lee, S. B., Park, S. Y., Yi, Y. S. & Ahn, C. W. Structure of amorphous aluminum oxide. Phys. Rev. Lett. 103, (2009). 2 Massiot, D. et al. Modelling one- and two-dimensional solid-state NMR spectra. Mag. Res. Chem. 40, (2002). 3 Le Caer, G. & Brand, R. A. General models for the distributions of electric field gradients in disordered solids. J. Phys. Cond. Matt. 10, (1998). 4 Neuville, D. R., Cormier, L. & Massiot, D. Al environment in tectosilicate and peraluminous glasses: A Al-27 MQ-MAS NMR, Raman, and XANES investigation. Geochim. Cosmochim. Acta 68, (2004). 5 Lee, S. K., Park, S. Y., Yi, Y. S. & Moon, J. Structure and disorder in amorphous alumina thin films: Insights from high-resolution solid-state NMR. J. Phys. Chem. B. 114, (2010). 9

10 6 Xue, X. Y. & Kanzaki, M. Structure of hydrous aluminosilicate glasses along the diopside-anorthite join: A comprehensive one- and two-dimensional 1 H and 27 Al NMR study. Geochim. Cosmochim. Acta 72, (2008). 7 Park, S. Y. & Lee, S. K. Structure and disorder in basaltic glasses and melts: Insights from high-resolution solid-state NMR study of glasses in diopside-ca-tschermakite join and diopside-anorthite eutectic composition. Geochim. Cosmochim. Acta 80, (2012). 8 Gaussian 03, Revision C.02 (2003). 9 Lee, S. K., Kim, H. N., Lee, B. H., Kim, H. I. & Kim, E. J. Nature of Chemical and Topological Disorder in Borogermanate Glasses: Insights from B-11 and O-17 Solid- State NMR and Quantum Chemical Calculations. J. Phys. Chem. B 114, (2010). 10 Lee, S. K., Musgrave, C. B., Zhao, P. & Stebbins, J. F. Topological disorder and reactivity of borosilicate glasses: Ab initio molecular orbital calculations and 17 O and 11 B NMR. J. Phys. Chem. B 105, (2001). 10