UNCORRECTED PROOF ARTICLE IN PRESS. 2 Grain boundary structures of atomic layer deposited TiN. 3 S. Li a, *, C.Q. Sun b, H.S.

Transcription

1 + model 1 Thin Solid Films xx (2005) xxx xxx 2 Grain boundary structures of atomic layer deposited TiN 3 S. Li a, *, C.Q. Sun b, H.S. Park c bstract a School of Materials Science and Engineering, The University of New South Wales, NSW 2052, ustralia b School of Electrical and Electronic Engineering, Nanyang Technological University, , Singapore c School of Materials Science and Engineering, Nanyang Technological University, , Singapore 9 Grain boundary plays an important role in determining the physical properties and chemical stability of the materials. In particular, the 10 structures of grain boundaries in atomic layer deposited TiN film may be one of the main factors to dominate the reliability and performance of 11 ULSI devices with multilayer structure of Cu-based interconnects. In this work, the characteristics of grain boundaries in the atomic layer 12 deposited (LD) ultralthin TiN films are investigated in detail. The results show that the small angle boundaries frequently appear in (200) epitaxy 13 growth areas while the large angle grain boundaries are often observed in the areas with (111) close-packed stacking configuration. The tilt and 14 twist boundaries are the two main structures of large angle grain boundaries. It is of interest to note that some large angle grain boundaries in the 15 LD TiN film are in an atomic scale. Such small and large angle boundaries in atomic scale lead to the pinhole free, high density and homogeneity 16 ultrathin polycrystalline LD TiN film. 17 D 2005 Published by Elsevier.V Keywords: Grain boundary; Titanium nitride Introduction 22 Cu is a favored material for the multilayered interconnects 23 of ultra large-scale integration devices because of its superior 24 electrical conductivity and electromigration resistance [1,2]. 25 However, Cu diffusion may cause the degradation of electronic 26 properties by deep-level trapping in silicon devices [3 5]. TiN 27 is one of the most commonly used material as a diffusion 28 barrier to prevent the Cu diffusion into Si and [6,7]. In 29 general, the high quality diffusion barriers for the multilayer 30 structure of the Cu-based interconnects are deposited by 31 conventional chemical vapor deposition, physical vapor depo- 32 sition, ionized metal plasma deposition techniques, etc. Since 33 the structure of the ULSI devices has been scaled down to 34 nanoscale recently, the development of deposition technique is 35 eagerly awaited to improve the step coverage of the diffusion 36 barrier for next generation devices. tomic layer deposition 37 (LD) method is a promising technique for these applications 38 because the saturation mechanism of LD results in an 39 inherent elimination of pinholes, excellent conformal coating characteristics as well as good thickness uniformity and homogeneity of the thin film produced [8,9]. The applicability of LD on the growth of TiN thin film has been successfully demonstrated [10 12]. lthough the surface reaction and physical model of the atomic layer deposition in the process have been systematically studied [8,13 15], the grain boundary structures of ultrathin TiN film deposited by LD technique are still not clear. In this work, the grain boundary structure of the ultrathin LD TiN film is studied in detail and the understanding provides the new insights into fundamental mechanism of TiN film formation in LD process, thus revealing the dominating factors, which determine the film growth and its reliability as a diffusion barrier. * Corresponding author. address: Sean.li@unsw.edu.au (S. Li). 2. Experiment 300 nm layer was deposited on Si substrate using plasma enhanced chemical vapor deposition (PECVD) and, subsequently, a 30 nm TiN film was deposited on the by LD technique with pure precursors of TiCl 4 and NH 3 at 390 -C inn 2 purging gas. Finally a 350 nm Cu film with column grain structure was deposited on the TiN layer by magnetron sputtering deposition to obtain the Cu/TiN/ /Si structure. High-resolution transmission microscopy (HRTEM) was /$ - see front matter D 2005 Published by Elsevier.V. doi: /j.tsf TSF-20868; No of Pages 5

2 2 S. Li et al. / Thin Solid Films xx (2005) xxx xxx 62 employed to characterize the grain boundary structures of LD 63 TiN film and the cross-sectional HRTEM samples were 64 prepared with the conventional sandwich technique Results and discussion 66 TEM morphology in Fig. 1 shows that a homogeneous 67 ultrathin TiN layer with high density was deposited on the 68 amorphous layer by the LD method. Theoretically, LD 69 method deposits the TiN film with a pattern of stacking the 70 monoatomic layers in a close-packed configuration, thus 71 forming a high density epitaxy film. The stacking mechanism 72 of the high-density epitaxy film with close-packed configura- 73 tion has been explained successfully in elsewhere [8]. Itisof 74 interest to note that the stacking configuration on the lower 75 density (200) is also often observed in the LD film (Fig. 2) 76 and its stacking mechanism is still unclear. In general, the 77 chemistry of the surface reactions is far more complicated than 78 simply laying atomic layers on top of each other. surface 79 always creates a discontinuity of the bulk which means that 80 either there are dangling bonds on the surface or the 81 unsaturated bonds are terminated by surface reconstruction, 82 etc. [13]. The dangling bonds on the surface and the surface 83 reconstruction may result in the structure of initial as-deposited 84 monoatomic layer deviating from the ideally close-packed 85 configuration. In this case, the subsequent atom stacking on the 86 previous deposited monoatomic layer may result in two crystal 87 structures: (1) the atom stacking may eliminate the dangling 88 bonds in the underlying monoatomic layer, producing an 89 epitaxial layer with close-packed density; and (2) the subse- 90 quently deposited atoms may just simply lie on the top of the 91 reconstructed monoatomic layer, forming an epitaxial film with 92 a heredity density lower than the close-packed density, such as 93 TiN (200) configuration. These may be one reason why the 94 lattice fringes of the close-packed planes (111) and planes 95 (200) were frequently observed in the cross-sectional HRTEM 96 morphology of the ultrathin TiN film. In most cases, the ideal 97 saturation (close-packed) density for a full monoatomic layer is not easy to achieve due to surface reconstruction [8]. This may result in the variation of atomic stacking features intrinsically during the film growth, thus producing nano-polycrystalline structure. lthough the pseudo-single epitaxy film is expected to grow with LD method, the small and large angle boundaries indeed co-exist in the LD TiN film. From Fig. 2(a), it can be seen that the film grew epitaxially in the first ten monoatomic layers stacking on (200) planes (marked by ) and then split into two crystals (marked by and C) with a small angle of 3-. Itis believed that stacking error and geometric feature are two possible factors to cause such a small angle misorientation. Intrinsically, the bond dangling or surface reconstruction would occasionally cause some atoms to deviate from the position of ideally close-packed configuration. This may be the reason to cause a gap in the 11th monoatomic layer (marked by G 1 )to split the continuous monoatomic layer into two segments, forming a discontinuous monoatomic layer. Similar phenomena are also observed in the other areas of this sample. In this case, the subsequent atom stacking may eliminate the dangling bonds in the underlying monoatomic layer of per segment, continuously producing the epitaxial layers to form the epitaxy 12th discontinuous atomic layer. On the other hand, the surface profile may also play an important role in determining the structural characteristics of grain boundaries as schematically illustrated in Fig. 2(b). Geometrically the surface of any twodimensional convex surface with a small curvature can be resolved into two planar facets with a very small misorientation. The crystals, therefore, grew epitaxially on these planar facets respectively. In the first few atomic deposition cycles, the effects of such a small misorientation on film growth are negligible. The gap between the crystals was so small that it could be neglected, exhibiting a pseudo-epitaxy growth characteristic with multi-monoatomic layers, until the size of the gap was comparable with the size of the TiN atom, such as the gap between Crystals and C in the 11th monoatomic layer as shown in Fig. 2(a). Since the gap is smaller than the diameter of the atom, atoms cannot fill the gap to form a continuous (200) pseudo-monoatomic layer. Hence, a discontinuous layer is produced. Similarly the subsequent deposition cycle grows an epitaxy 12th discontinuous atomic layer. For the gap, two configurations are possible in the subsequent deposition. If the gap is small enough, the atoms may either simply lie in it or just leave it open. The open gap can be clearly seen from the 15th to 18th atomic layers (arrowhead with G 2 ) while an atom lie on the site of the gap space in the 19th atomic layer (arrowhead with G 3 ), resulting in the changes of the atom stacking configuration in the subsequent deposition cycles. For example, the atoms on the 20th atomic layer in Crystal C lie in the space between the atoms instead of on top, producing stacking faults till an atom partially inserted into the gap of the 22nd atomic layer (arrowhead with G 4 ). In the subsequent deposition cycles, the (200) planes in crystals and C are no longer on the same level, and the atoms start to fill the gap space disorderly. It is difficult to define the structure of the gap between the two crystals, although the atoms in the gap can be clearly discerned. Cu TiN Fig. 1. TEM morphology showing a high density and homogeneity ultrathin TiN layer deposited on the surface of an amorphous layer by LD technique

3 S. Li et al. / Thin Solid Films xx (2005) xxx xxx 3 Fig. 2. (a) HRTEM image shows a small angle boundary caused by the atomic stacking error in the areas with (200) stacking configuration. (b) schematic illustration showing the formation mechanism of small angle boundary caused by surface profile of the substrate. 155 It is probably a transitional structure with a semi-crystalline 156 character that has a considerably high density. This atom 157 stacking feature dominates with further deposition giving rise 158 to the formation of two columns of crystals and C based on 159 the pseudo-epitaxy complex of crystal. 160 Furthermore, atomic stacking fault also strongly influences 161 the orientation of the epitaxy crystals. Fig. 3 shows a lattice image of the crystals deposited by LD technique. The film grew epitaxially in the first ten (200) atomic layers (marked by ). n incomplete (half) atomic layer (arrowhead with ) can be seen to form on the 10th atomic layer. Subsequent layers with (200) configuration (marked by C) above this incomplete atomic layer grew epitaxially with a small tilted angle of 7- to the (200) monoatomic layers in crystal D. This led to the formation of crystals C and D from the pseudo-epitaxy complex of crystal with different orientations. It may be another mechanism that the nano-polycrystalline structure formed during the LD process. The small angle grain boundary between crystals C and D is a strain free coherent boundary because the (200) planes of both crystals connected each other at the boundary. However, large angle grain boundaries are frequently observed in the areas with (111) close-packed configuration of the LD film. Fig. 4(a) shows a high-resolution morphology of a grain boundary between the crystals (crystals and ) with a 30- misorientation of TiN (111) planes. This boundary consisted of three regions with different characteristics marked by G 1,G 2 and G 3, respectively. Since the (111) lattice fringes of crystal terminated on the (111) lattice fringes of crystal with 30- in region G 1, the boundary in this region is incoherent. However, the (111) lattice fringes of crystals and connected each other in region G 3, exhibiting the D (200) oundary Fig. 3. HRTEM image shows that the stacking fault varied the orientation of the atomic layers, splitting one crystal into two. C

4 4 S. Li et al. / Thin Solid Films xx (2005) xxx xxx 187 characteristic of a strain free coherent boundary. The boundary 188 structure of region G 2 is complex, and it is a transitional 189 region from incoherent to coherent boundary structures. The 190 atomic image (arrowhead with h) shows that atoms in this 191 transitional region are self-organized to adjust their positions to 192 fit the structural change. Fig. 4(a) also shows that the grain 193 boundary formed in the order of firstly incoherent boundary, 194 then transitional structure and coherent boundary from the TiN/ 195 interface to TiN surface. It implies that the low energy 196 strain free coherent boundary is a favored boundary for the tilt 197 crystals (Fig. 4(b)) in the film deposited by LD technique, 198 although the tilt crystals nucleate with different orientations. 199 This phenomenon may be related to the growth energy of the 200 monoatomic layer. In Fig. 4(a), it can be clearly seen that the 201 width of the boundaries is in an atomic scale regardless of what 202 structures the grain boundaries have. This kind of atomic-scale 203 boundary can effectively restrict the boundary diffusion, which 204 is the major interdiffusion mechanism in semiconductor 205 devices. However, the other type of boundary, such as twist 206 boundary, can also be observed in the film deposited by LD 207 technique. Fig. 5 shows the (111) lattice fringes of two TiN 208 grains partially over lapped, evidence of the twist boundary. 209 It is of interest to note that the small angle boundaries 210 frequently appear in (200) epitaxy growth areas while the large 211 angle boundaries are often observed in the areas with (111) (a) (b) 2 nm (111) 2 nm close-packed stacking configuration. This phenomenon may be attributed to nature of TiN lattice structure, which atomic distance on (100) plane is greater than that on the close-packed plane of (111). In this case, it can be inferred that the atoms in (200) have much greater deviation from the lattice point caused by the bond dangling and surface reconstruction. Therefore, the stacking errors resulted by atom displacements are easy to produce in (200) plane, thus forming the small angle boundaries as shown in Figs. 2 and 3. In the (111) plane, all atoms are close-packed together and the spatial size for atom displacement is very limited. Therefore, the small angle boundaries are seldom observed in the areas where the crystals grow along the (111) planes. Rotation axis G 1 θ G 2 G 3 β (a) (b) Rotation axis (111) Fig. 5. (a) HRTEM image showing a twist grain boundary in the ultrathin TiN film. (b) schematic illustration shows the structure of twist grain boundary in (a). θ Conclusion 257 Fig. 4. (a) HRTEM image showing that the grain boundary consists of three regions with different structures. (b) schematic illustration shows the structure of tilt grain boundary in (a). The characteristics of grain boundary structures in the TiN film deposited by LD technique are studied in detail. The result shows that the small angle boundaries are often observed in (200) epitaxy growth areas while the large angle boundaries are frequently appear in the areas with (111) close-packed stacking configuration. It is believed that the bond dangling and surface reconstruction may be the reason to intrinsically cause the small grain boundaries in the film, which the atoms stack on (200) planes. However, the surface profile of the

5 S. Li et al. / Thin Solid Films xx (2005) xxx xxx amorphous substrate is an extrinsic mechanism to dominate the 268 formation of large angle boundaries on the areas with (111) 269 close-packed stacking configuration. The tilt and twist grain 270 boundaries with the width in atomic scale are two main 271 characteristics of large grain boundaries. The coherent bound- 272 ary should be the favored boundary in the connection of the tilt 273 grains. The small and large angle boundaries in atomic scale 274 should be the dominant factors to produce the pinhole free, 275 high density and homogeneity ultrathin polycrystalline LD 276 TiN film. 277 References [1] Z.J. Radzimski, W.M. Posadowski, S.M. Rossnagel, S. Shingubara, J. 280 Vac. Sci. Technol., 16 (1998) [2] J.R. Lloyd, J.J. Clement, Thin Solid Films 262 (1995) [3] J. Torres, ppl. Surf. Sci. 91 (1995) 112. [4]. roniatowski, Phys. Rev. Lett. 62 (1989) [5] S.M. Sze, Semiconductor Devices, Physics and Technology, New York, Wiley, [6].E. Kaloyeros, E. Eisenbraun, nnu. Rev. Mater. Sci 30 (2000) 363. [7]. Kumar, H. akhru, C. Jin, W.W. Lee, T.M. Lu, J. ppl. Phys. 87 (2000) [8] T. Suntola, Thin Solid Films 216 (1992) 84. [9] M. Ritala, M. Leskelä, Nanotechnology 10 (1999) 19. [10] M. Ritala, M. Leskelä, J. Dekker, C. Mutsaers, P.J. Soininen, J. Skarp, Chem. Vapor. Depos. 5 (1999) 7. [11] M. Juppo, M. Ritala, M. Leskelä, J. Electrochem. Soc. 147 (2000) [12] H. Jeon, J.W. Lee, Y.D. Kim, D.S. Kim, K.S. Yi, J. Vac. Sci. Technol., 18 (2000) [13] T. Suntola, ppl. Surf. Sci 100/101 (1996) 391. [14] C.H.L. Goodman, M.V. Pessa, J. ppl. Phys. 60 (1986) R65. [15] M. Leskelä, M. Ritala, J. Phys. IV (France) 9 (1999) Pr