1 2 3 4 5 6 7 8 9 Journal of The Electrochemical Society, 156 3 1-XXXX 2009 0013-4651/2009/156 3 /1/0/$23.00 The Electrochemical Society Investigation of Carbon Nanotube Growth on Multimetal Layers for Advanced Interconnect Applications in Microelectronic Devices Nay Lin, a, * Huili Wang, a Pradeep Dixit, b, * Ting Xu, a Sam Zhang, a and Jianmin Miao a,z a Micromachines Centre, School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore 639798 b Packaging Research Center, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0250, USA 1 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 In this paper we present the microstructural study behind the growth of carbon nanotubes CNTs on the multimetal buffer layers due to its importance in microelectronics and microelectromechanical systems applications. Two different buffer layers, i.e., aluminum Al and titanium nitride TiN, were deposited on the conductive layers of tantalum/copper/tantalum. A 5 nm thick iron film was used as a catalyst layer to grow the CNTs. The fundamental mechanism behind the formation of catalyst nanoparticles on these two buffer layers, i.e., Al and TiN, was studied and analyzed by various characterization tools, such as atomic force microscopy, X-ray photoelectron spectroscopy, and scanning electron microscope. The formation of aluminum oxide nanoparticles during the CNT growth process was observed in the case of aluminum buffer layer. From the experimental results, it was concluded that TiN can be used as a stable buffer layer on the conductive metal lines. The CNTs growth on both buffer layers was found to be in random directions, which is due to the formation of bigger and less dense catalyst nanoparticles in comparison with the CNTs grown on the conventional buffer layer of thermally grown silicon dioxide on the silicon substrate, on which vertically aligned CNTs are grown. 2009 The Electrochemical Society. DOI: 10.1149/1.3060347 All rights reserved. Manuscript submitted October 29, 2008; revised manuscript received December 1, 2008. Published xx xx, xxxx. 25 Since their discovery by Sumio Iijima, 1 carbon nanotubes 26 CNTs have shown excellent thermal, 2 mechanical, 3 and electrical 27 properties 4 and thus are being extensively studied for various applications in nanoelectronics, 5 nano-electromechanical systems, 6 bio- 28 29 microelectromechanical systems, 7 biosensors, etc. CNTs have demonstrated scattering-free, ballistic electron transport and due to this, 30 31 a current density as high as 10 10 A/cm 2 can be achieved. The capability of carrying such a high current density has initiated the need 32 33 for fabricating CNT-based electronic devices such as resonators, 34 field-emission displays, etc. These CNT-based electronic devices 35 will be more compact in size and will have ultrahigh-processing 36 speed and lower time delay than the present copper-interconnectbased devices. Although high-aspect-ratio through-wafer copper 37 38 interconnection 8,9 was proposed for the next immediate generation 39 of three-dimensional packaging, CNT interconnection is likely to be 40 the final goal for future ultrahigh-current-density packaging technologies. 41 42 In past years, growth of CNTs by various methods, such as electric arc, 10 laser ablation 11 and chemical vapor deposition CVD, 12 43 44 have been reported. Due to its abilities of growing ultralong CNTs in 45 large numbers and selective area growth, the CVD process has 46 emerged as the main process to grow vertically aligned CNTs. Various research groups have reported the growth of single as well as 47 48 multiwalled CNTs with varying lengths, diameters, and orientations. 49 In past research publications, CNTs were grown on various metal 50 catalysts, such as iron Fe, nickel Ni, and cobalt Co, which 51 themselves were deposited on the silicon substrate by physical vapor 52 deposition methods such as sputtering or evaporation. In all these 53 cases, a silicon dioxide layer was grown on the silicon substrate by 54 the thermal oxidation method. The silicon dioxide layer acts as a 55 buffer layer between the silicon substrate and the metal catalyst and 56 prevents the diffusion of metal catalyst into silicon. In the absence 57 of any buffer layer, the metal catalyst will diffuse into silicon, which 58 in turn will affect CNT growth. 59 The successful realization of CNT-based interconnects also requires on-chip copper interconnect lines and CNT interconnect con- 60 61 necting to on-chip copper interconnects. The most important process 62 in achieving the CNT interconnect is the growth of CNTs on metallization. A schematic diagram of CNT growth on a multimetal layer 63 is illustrated in Fig. 1. In this diagram, copper Cu acts as the 64 conductive metal line with a buffer layer to prevent the diffusion of 65 copper into silicon and also acts as an adhesion layer. The growth 66 mechanism of CNTs on metal buffer layers is different from that on 67 a silicon dioxide buffer layer. At the high process temperature 68 700 C there are relatively higher chances that the metal catalyst 69 may react with the other metals and form intermetallic compounds 70 IMCs. These IMCs are not desired for the satisfactory growth of 71 CNTs and must be avoided. Due to the continuously growing need 72 of CNT-based interconnects on copper conductive lines, it is necessary to understand the growth mechanism of CNTs on multimetal 74 73 layers. Unfortunately, there are not many published results available 75 which elaborate on CNT growth on multimetal layers. The growth 76 of CNTs on aluminum substrate 13 was reported in the literature. The 77 growth mechanism of CNTs on aluminum substrate was observed to 78 be a tip growth mechanism with an iron catalyst layer deposited by 79 spin-coating iron nitrate Fe NO 3 3 9H 2 O and C 2 H 2 as carbon 80 feedstock at 650 C at the CNT growth process. Titanium nitride had 81 been used as a diffusion barrier layer between aluminum and silicon 82 in the microelectronics industry for many years. More recently, there 83 were attempts to grow CNTs and carbon nanofiber on different 84 metal underlayers due to motivation for practical applications. 14,15 85 Although CNT growth on the Ti/Cu metal system was 86 demonstrated, 16 the length of the CNTs is limited due to the plasmaenhanced CVD process. 88 87 In this paper, CNT growth on different buffer layers SiO 2,Al, 89 and TiN was studied. The growth mechanism of CNTs on the two 90 * Electrochemical Society Student Member. z E-mail: mjmmiao@ntu.edu.sg Figure 1. Schematic diagram of CNT growth on copper conductive lines.
2 Journal of The Electrochemical Society, 156 3 1-XXXX 2009 Figure 3. Temperature profiles of annealing and CNT growth steps in the thermal CVD. represents the sample with the standard silicon dioxide buffer layer 126 used for growing CNTs. In order to realize a practical CNT-based 127 device, where other metal layers such as Cu conductive layer, 128 Ti/Ta adhesion/barrier layer, and Fe/Ni/Co catalyst layer are 129 also present, samples B and C were also investigated. Sample B 130 utilizes TiN as a buffer layer, while sample C considers Al. In both 131 cases, the conductive metal layers used were Ta/Cu/Ta. 132 Figure 2. Schematic diagram of samples used in experiments. 91 different buffer layers, i.e., Al and TiN on the Ta/Cu/Ta metal layers, is reported and compared with the CNTs grown on silicon oxide. 92 93 X-ray photoelectron spectroscopy XPS was used to study the 94 chemical state of the resulting catalyst nanoparticles. The surface 95 morphologies of the multimetal layers with and without catalyst 96 layers were studied by atomic force microscopy AFM and scanning electron microscopy 97 SEM. 98 Experimental 99 Sample preparation. The fabrication process can be summarized by the following steps: Silicon substrates 100, p-type, 100 101 0.1 10 cm resistivity, and 100 mm diam were cleaned in piranha 102 solution for 20 min at 120 C to remove any organic contaminations. 103 A1 m thick silicon dioxide layer was thermally grown on the 104 wafer in a furnace by wet oxidation at 1100 C. The deposition of 105 metal layers was performed using the magnetron sputtering process. 106 A tantalum layer of 20 nm thickness was used as an adhesion layer 107 followed by a 1 m thick copper layer. A 20 nm thick Ta layer was 108 deposited again, which acts as a barrier layer and also prevents 109 copper oxidation. 110 For deposition of the TiN layer, the wafers were loaded into a 111 different magnetron sputtering chamber. The TiN sputtering process 112 was performed at room temperature and at a chamber pressure of 113 0.93 Pa. The gas flow rates of argon and nitrogen were 50 and 114 20 sccm, respectively. A plasma source power of 500 W was used at 115 a substrate bias voltage of 20 V. The average thickness of 50 nm 116 TiN was deposited for a duration of 6 min. For Al buffer layer 117 deposition, an electron-beam evaporation process was used. An Al 118 layer of 5 nm and an Fe catalyst layer of 5 nm were deposited 119 without breaking the vacuum. A very low chamber pressure of 2 120 10 7 Torr was used to ensure the satisfactory uniformity of the 121 film. 122 Three samples with different buffer layers were prepared to study 123 the formation of catalyst nanoparticles and their effects on CNT 124 growth. Figure 2 shows a schematic description of the samples 125 along with the thicknesses of individual metal layers. Sample A Annealing of samples and CNT growth in thermal CVD. In 133 order to study the formation of catalyst nanoparticles and material 134 composition, the samples, consisting of the conductive layer, buffer 135 layer, and iron catalyst layer, were annealed in a thermal CVD 136 chamber. The samples were annealed at 700 C for 15 min in the 137 presence of H 2 100 sccm and Ar 400 sccm with a chamber pressure of 4.9 Torr. This annealing step induces the nucleation and 139 138 formation of catalyst nanoparticles which are needed to grow the 140 CNTs. 17 After the annealing step, the CNT growth step was performed. The temperature profile for annealing and CNT growing 142 141 process is illustrated in Fig. 3. Once the annealing step was over, 143 acetylene C 2 H 2 gas 100 sccm was introduced for 15 min at a 144 chamber pressure of 5.9 Torr. After the CNT growth process, the 145 temperature was ramped down to room temperature of 25 C at approximately 3 C/min. 146 147 Characterization of catalyst nanoparticles and CNTs. For the 148 XPS study of catalyst nanoparticles, a Kratos-Axis spectrometer 149 with monochromatic Al K 1486.71 ev X-ray radiation 15 kv 150 and 10 ma and hemispherical electron energy analyzer were used. 151 The morphology of the annealed catalyst layer was taken by AFM 152 Digital Instruments, Santa Barbara in the tapping mode. The 153 grown CNTs were characterized by the Hitachi scanning electron 154 microscope under an accelerating voltage of 15 25 kv. 155 Results and Discussion 156 CNT growth on silicon dioxide buffer layer (sample A). For 157 sample A, a catalyst layer Fe having a thickness of 5 nm was 158 deposited. After annealing in the thermal CVD chamber at 700 C, 159 the samples were characterized by SEM. Figure 4a shows the top 160 view of sample A after annealing, which has Fe catalyst on 1 m 161 thick thermally grown silicon dioxide layer. It can be seen that the 162 Fe film has been broken into nanoparticles of varying sizes and 163 shapes. The average size of these highly dense nanoparticles varies 164 between 30 and 100 nm. Such a highly dense nanoparticle array is 165 ideally suited for growing dense CNTs of relatively smaller diameter. When the CNTs were grown on this sample, a very satisfactory 167 166 CNT growth was observed. Figure 4b shows an SEM image of the 168 vertically aligned CNTs grown on this sample. The CNTs grown on 169 the sample are of multiwalled type and have a diameter varying 170 between 40 and 100 nm. The dense, vertically aligned CNT bundles 171 were grown on the annealed iron nanoparticles. Due to the high 172 density and the close packing of the nanoparticles, the CNT growth 173 was preferred in the vertical direction, as it is the only degree of 174 freedom available during CNT growth. The measured electrical resistivity of the CNTs is about 0.0097 cm. 17 175 176
Journal of The Electrochemical Society, 156 3 1-XXXX 2009 3 Figure 6. CNTs grown on the TiN buffer layer. Figure 4. SEM images of a the top view of iron catalyst particles on the silicon dioxide layer after annealing at 700 C and b CNTs grown on the silicon oxide buffer layer. 177 CNT growth on Ta/Cu/Ta metal layers with TiN buffer layer 178 (sample B). When CNTs were grown on sample B, which has TiN 179 buffer layer, some interesting mechanisms were observed. Figure 5a 180 shows the AFM profile of the sample with an annealed TiN surface 181 with underlying Ta/Cu/Ta metal layers. AFM analysis of the surface 182 profile of sample B shows that the grain structure of TiN before the 183 annealing is almost the same as the sputtered TiN thin film, as reported in the literature. 18 The average grain size of the sputtered TiN 184 185 film is about 20 30 nm. Figure 5b shows the AFM profile of the Figure 5. Color online AFM profile of annealed TiN surface, a without Fe catalyst and b with Fe catalyst. same sample after the annealing step. The formation of nanoparticles on the TiN layer can be observed from the picture. It can be 187 186 observed that the size of nanoparticles formed is larger than that 188 required for catalytic growth of CNTs 100 nm. The density of 189 particles having a diameter of approximately 20 80 nm is very low, 190 as it can be observed in the figure. This is the reason only the forest 191 of CNTs were grown on the TiN barrier layer rather than a vertically 192 aligned CNT bundle as widely reported, 19 because the formation of 193 a vertically aligned CNT bundle is due to the vertical direction of 194 CNT growth being the only possible degree of freedom for highly 195 dense catalyst nanoparticles. 196 The incomplete formation of Fe catalyst nanoparticles of bean 197 shape can also be seen in Fig. 5b. With the same experimental parameters and catalyst thickness varying from 2 to 10 nm on the ther- 199 198 mally grown silicon dioxide layer, the vertically aligned dense CNTs 200 bundles were grown. An annealing temperature of 700 C was used 201 for 15 min for the Fe catalyst layer on the TiN buffer layer. The 202 above experiments show that an annealing temperature of 700 C is 203 not enough for complete formation of nanoparticles, which may be 204 due to the difference in adhesion properties between the Fe/SiO 2 205 interface and the Fe/TiN interface. Figure 6 shows the top view of 206 CNTs grown on the TiN buffer layer. It can be seen that the density 207 of CNTs is much lower in comparison with the results usually reported in literature with CNT growth on a stable buffer such as 209 208 silicon dioxide. The CNTs grew in random orientations, forming a 210 layer of porous CNT-coated surface approximately 2 m thick. 211 CNT growth on Ta/Cu/Ta metal layers with an aluminum buffer 212 layer (sample C). Figure 7 shows an AFM surface profile of 213 sample C with an Al buffer layer and Fe catalyst after annealing. 214 Formation of nanoparticles with sizes ranging from 20 to 200 nm 215 can be seen. The density of nanoparticles is much lower than that of 216 catalyst nanoparticles on the buffer layer such as silicon dioxide. 217 According to the available experimental data, the nanoparticles consist of aluminum oxide, iron catalyst, and possibly iron oxide, and 219 218 the source of oxygen is postulated to be from the atmosphere and 220 oxidized iron catalyst. In order to determine the chemical state of the 221 Al and Fe catalyst layer, XPS examination was performed with annealed samples having an Al buffer layer on the Ta/Cu/Ta metal 223 222 layer. An XPS spectrum of the sample is shown in Fig. 8. One of the 224 curves shows the peak before etching in argon plasma and the other 225 shows the peak after etching for 960 s. XPS experimental results 226 show that the aluminum oxide particles were formed during the 227 annealing process in the presence of hydrogen flow. The source of 228 oxygen is the oxidized iron catalyst layer. Although there is a reduction reaction inside the reaction chamber with the presence of hy- 230 229 drogen gas flow, oxygen from the iron oxide layer may have at- 231
4 Journal of The Electrochemical Society, 156 3 1-XXXX 2009 Figure 9. SEM image of CNTs grown on the aluminum buffer layer. Figure 7. Color online AFM profiles of Al 2 O 3 nanoparticles after annealingat700 C. 232 tracted the underlying Al layer, forming stable ceramic aluminum 233 oxide particles which act as the stable support for CNT growth 234 during the recrystallization of Al and Fe nanoparticles. Although it is 235 possible to have a very thin native aluminum oxide layer on the 236 samples during Al/Fe deposition and after taking the samples out of 237 the vacuum chamber, the thickness is limited to less than 1 nm, 238 enabling the metallic Al to reflow during annealing. 239 Figure 9 shows the CNTs grown on the Al buffer layer with 240 Ta/Cu/Ta metal layers. The density of CNTs is very low and CNTs 241 are randomly oriented. This is due to the formation of low-density 242 catalyst nanoparticles allowing the CNTs to grow in random orientation. In thermal CVD growth of CNTs, vertically aligned CNTs 243 244 were obtained due to the densely grown CNTs guiding themselves 245 vertically. The only possible direction for CNT growth to continue is 246 the vertical direction, because growth in the lateral direction is impossible due to the presence of other CNTs on the side. However, 247 248 with the low density of catalyst, no such self-guiding mechanism is 249 possible because CNTs continuing to grow will also have the free 250 space to grow laterally until coinciding with others. Along with the 251 absence of another guiding mechanism e.g., electric field in plasmaenhanced CVD, CNTs grown on the low-density catalyst layer are 252 253 randomly oriented. In contrast to this situation, experiments with CNTs grown on a silicon dioxide layer showed highly dense nanoparticle formation and vertically aligned CNT growth, as discussed 255 254 in the previous section. From the above experimental results, the 256 growth mechanism of CNTs in the present case is similar to that of 257 CNTs with an Al/Fe bimetallic layer catalyst 13 with a tip growth 258 mechanism. Figure 10 shows a schematic diagram of the growth of 259 CNTs on Ta/Cu/Ta metal layers with CNTs grown on the aluminum 260 oxide particles. Figure 11 shows CNTs grown on patterned 261 Ta/Cu/Ta metal layers on which an Al buffer layer and an Fe catalyst layer were selectively deposited. Figure 11b shows a detailed 263 262 view of CNTs grown 4 m thick on the multimetal layers. Due 264 to the possible implication of a buffer layer, the type and conductivity of the CNTs may differ from those grown on the silicon oxide 266 265 layer, and further research is needed to investigate the type and 267 electrical properties of those CNTs. 268 Conclusion 269 CNT growth on Ta/Cu/Ta metal layers was studied with different buffer layers, namely, Al and TiN. In order to compare the 271 270 growth of CNTs on TiN and Al buffer layers with those grown on 272 silicon dioxide buffer layers, we studied the formation of catalyst 273 nanoparticles on the silicon dioxide layer. Highly dense catalyst particles were observed and vertically aligned CNTs were grown, as 275 274 widely reported in literature. From experiments with a TiN buffer 276 layer, we found a stable TiN layer after annealing without recrystallization. The grain size of the TiN layer was the same as that of 278 277 sputtered TiN thin film reported in the literature. Study of the size 279 and density of nanoparticles shows that the larger and incomplete 280 formation of nanoparticles after annealing is attributed to the smaller 281 Figure 8. Color online XPS spectrum of aluminum oxide before and after argon plasma etching in XPS. Figure 10. Schematic diagram of the CNT growth mechanism on an Al buffer layer. Al was formed to Al 2 O 3 nanoparticles after annealing.
Journal of The Electrochemical Society, 156 3 1-XXXX 2009 5 282 density of nanoparticles available for CNT growth. This causes 283 growth of CNTs in random directions rather than in the vertical 284 direction. From the experiments with an Al buffer layer, the formation of aluminum oxide nanoparticles was observed together with Fe 285 and Fe 2 O 3 in the XPS analysis. We found that the CNT growth 286 mechanism with the Al buffer layer is due to the formation of Al 2 O 3 287 nanoparticles on the underlying metal layers during the CNT growth 288 process. It can be concluded that the growth of randomly oriented 289 CNTs on the multimetal layer with Al and TiN buffer layers on the 290 metal layers is due to the sparse nanoparticle formation of the CNT 291 growth process during annealing as compared to the CNT grown on 292 a silicon dioxide buffer layer. A randomly oriented CNT-coated surface was observed on the samples with Al and TiN buffer layers on 294 293 the underlying metal layers, although the TiN buffer layer may be 295 suitable due to its electrically conductive properties in practical applications. 296 297 Figure 11. SEM picture showing a the CNTs grown on the Al buffer layer on a Ta/Cu/Ta multimetal system and b a magnified view of CNTs grown on a selective area. Acknowledgments 298 The authors acknowledge support by the Agency for Science, Technology and Research A * 299 STAR, Singapore, under SERC grant 300 no. 042 114 0042 and the industrial sponsorship by Delphi Automotive, Singapore. 301 302 Nanyang Technological University assisted in meeting the publication costs of this article. References 1. S. Iijima and T. Ichihashi, Nature (London), 354, 56 1991. 2. P. Kim, L. Shi, A. Majumdar, and P. L. McEuen, Phys. Rev. Lett., 87, 215502 2001. 3. R. S. Ruoff, D. Qian, and W. K. Liu, C. R. Phys., 4, 993 2003. 4. B. Q. Wei, R. Vajtai, and P. M. Ajayan, Appl. Phys. Lett., 79, 1172 2001. 5. J. Li, Q. Ye, A. Cassell, H. T. Ng, R. Steven, J. Han, and M. Meyyappan, Appl. Phys. Lett., 82, 2491 2003. 6. M. Dequesnes, S. V. Rotkin, and N. R. Aluru, Nanotechnology, 13, 120 2002. 7. N. Sinha, J. Ma, and J. T. W. Yeow, J. Nanosci. Nanotechnol., 6 3, 2006. 8. P. Dixit and J. Miao, J. Electrochem. Soc., 153, G771 2006. 9. P. Dixit, J. Miao, and R. Preisser, Electrochem. Solid-State Lett., 9, G305 2006. 10. A. A. Puretzky, D. B. Geohegan, X. Fan, and P. Cook, J. Appl. Phys., 70, 153 2000. 11. Y. Zhang, H. Gu, and S. Iijima, Appl. Phys. Lett., 73, 3827 1998. 12. Y. Huh, M. L. H. Green, Y. H. Kim, J. Y. Lee, and C. J. Lee, Appl. Surf. Sci., 249, 145 2005. 13. C. Emmenegger, J. M. Bonard, P. Mauron, P. Sudan, A. Lepora, B. Grobety, A. Zuttel, and L. Schlaphach, Carbon, 41, 539 2003. 14. M. S. Kabir, R. E. Morjan, O. A. Nerushev, P. Lundgren, S. Bengtsson, P. Enokson, and E. E. B. Campbell, Nanotechnology, 16, 458 2005. 15. M. S. Kabir, R. E. Morjan, O. A. Nerushev, P. Lundgren, S. Bengtsson, P. Enoksson, and E. E. B. Campbell, Nanotechnology, 17, 790 2006. 16. M. Y. Chen, C. M. Yeh, C. J. Huang, J. Hwang, A. P. Lee, and C. S. Kou, J. Electrochem. Soc., 153, 11 2006. 17. T. Xu, Z. H. Wang, J. M. Miao, X. F. Chen, and C. M. Tan, Appl. Phys. Lett., 91, 042108 2007. 18. J. E. Sundgren, B. O. Johansson, H. T. G. Hentzell, and S. E. Karlsson, Thin Solid Films, 105, 385 1983. 19. Z. Linbo, J. Xu, Y. Xiu, Y. Sun, D. W. Hess, and C. P. Wong, Carbon, 44, 253 2006. 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 AQ: #1 AQ: #2
NOT FOR PRINT! FOR REVIEW BY AUTHOR NOT FOR PRINT! AUTHOR QUERIES 066903JES #1 AQ: Please insert pg. no. for Ref. 7 on line 315. #2 AQ: Please check accuracy of Ref. 10 on line 318