journal of the electrochemical society, 156 0013-4651/2009 ...€¦ · for deposition of the tin...

PROOF COPY [JES-08-1905R] 066903JES

Investigation of Carbon Nanotube Growth on MultimetalLayers for Advanced Interconnect Applications inMicroelectronic DevicesNay Lin,a,* Huili Wang,a Pradeep Dixit,b,* Ting Xu,a Sam Zhang,a andJianmin Miaoa,z

aMicromachines Centre, School of Mechanical and Aerospace Engineering, Nanyang TechnologicalUniversity, Singapore 639798bPackaging Research Center, School of Electrical and Computer Engineering, Georgia Institute ofTechnology, Atlanta, Georgia 30332-0250, USA

In this paper we present the microstructural study behind the growth of carbon nanotubes �CNTs� on the multimetal buffer layersdue 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 ironfilm was used as a catalyst layer to grow the CNTs. The fundamental mechanism behind the formation of catalyst nanoparticles onthese two buffer layers, i.e., Al and TiN, was studied and analyzed by various characterization tools, such as atomic forcemicroscopy, X-ray photoelectron spectroscopy, and scanning electron microscope. The formation of aluminum oxide nanoparticlesduring the CNT growth process was observed in the case of aluminum buffer layer. From the experimental results, it wasconcluded that TiN can be used as a stable buffer layer on the conductive metal lines. The CNTs growth on both buffer layers wasfound to be in random directions, which is due to the formation of bigger and less dense catalyst nanoparticles in comparison withthe CNTs grown on the conventional buffer layer of thermally grown silicon dioxide on the silicon substrate, on which verticallyaligned 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.

Since their discovery by Sumio Iijima,1 carbon nanotubes�CNTs� have shown excellent thermal,2 mechanical,3 and electricalproperties4 and thus are being extensively studied for various appli-cations in nanoelectronics,5 nano-electromechanical systems,6 bio-microelectromechanical systems,7 biosensors, etc. CNTs have dem-onstrated scattering-free, ballistic electron transport and due to this,a current density as high as 1010 A/cm2 can be achieved. The capa-bility of carrying such a high current density has initiated the needfor fabricating CNT-based electronic devices such as resonators,field-emission displays, etc. These CNT-based electronic deviceswill be more compact in size and will have ultrahigh-processingspeed and lower time delay than the present copper-interconnect-based devices. Although high-aspect-ratio through-wafer copperinterconnection8,9 was proposed for the next immediate generationof three-dimensional packaging, CNT interconnection is likely to bethe final goal for future ultrahigh-current-density packaging tech-nologies.

In past years, growth of CNTs by various methods, such as elec-tric arc,10 laser ablation11 and chemical vapor deposition �CVD�,12

have been reported. Due to its abilities of growing ultralong CNTs inlarge numbers and selective area growth, the CVD process hasemerged as the main process to grow vertically aligned CNTs. Vari-ous research groups have reported the growth of single as well asmultiwalled CNTs with varying lengths, diameters, and orientations.In past research publications, CNTs were grown on various metalcatalysts, such as iron �Fe�, nickel �Ni�, and cobalt �Co�, whichthemselves were deposited on the silicon substrate by physical vapordeposition methods such as sputtering or evaporation. In all thesecases, a silicon dioxide layer was grown on the silicon substrate bythe thermal oxidation method. The silicon dioxide layer acts as abuffer layer between the silicon substrate and the metal catalyst andprevents the diffusion of metal catalyst into silicon. In the absenceof any buffer layer, the metal catalyst will diffuse into silicon, whichin turn will affect CNT growth.

The successful realization of CNT-based interconnects also re-quires on-chip copper interconnect lines and CNT interconnect con-necting to on-chip copper interconnects. The most important processin achieving the CNT interconnect is the growth of CNTs on metal-

lization. A schematic diagram of CNT growth on a multimetal layeris illustrated in Fig. 1. In this diagram, copper �Cu� acts as theconductive metal line with a buffer layer to prevent the diffusion ofcopper into silicon and also acts as an adhesion layer. The growthmechanism of CNTs on metal buffer layers is different from that ona silicon dioxide buffer layer. At the high process temperature��700°C� there are relatively higher chances that the metal catalystmay react with the other metals and form intermetallic compounds�IMCs�. These IMCs are not desired for the satisfactory growth ofCNTs and must be avoided. Due to the continuously growing needof CNT-based interconnects on copper conductive lines, it is neces-sary to understand the growth mechanism of CNTs on multimetallayers. Unfortunately, there are not many published results availablewhich elaborate on CNT growth on multimetal layers. The growthof CNTs on aluminum substrate13 was reported in the literature. Thegrowth mechanism of CNTs on aluminum substrate was observed tobe a tip growth mechanism with an iron catalyst layer deposited byspin-coating iron nitrate �Fe �NO3�3·9H2O� and C2H2 as carbonfeedstock at 650°C at the CNT growth process. Titanium nitride hadbeen used as a diffusion barrier layer between aluminum and siliconin the microelectronics industry for many years. More recently, therewere attempts to grow CNTs and carbon nanofiber on differentmetal underlayers due to motivation for practical applications.14,15

Although CNT growth on the Ti/Cu metal system wasdemonstrated,16 the length of the CNTs is limited due to the plasma-enhanced CVD process.

In this paper, CNT growth on different buffer layers �SiO2, Al,and TiN� was studied. The growth mechanism of CNTs on the two

* Electrochemical Society Student Member.z E-mail: [email protected] Figure 1. Schematic diagram of CNT growth on copper conductive lines.

Journal of The Electrochemical Society, 156 �3� 1-XXXX �2009�0013-4651/2009/156�3�/1/0/$23.00 © The Electrochemical Society

1

1

2

3

4

5

6789

10111213141516171819202122

2324

2526272829303132333435363738394041424344454647484950515253545556575859606162

63

646566676869707172737475767778798081828384858687888990



different buffer layers, i.e., Al and TiN on the Ta/Cu/Ta metal lay-ers, is reported and compared with the CNTs grown on silicon oxide.X-ray photoelectron spectroscopy �XPS� was used to study thechemical state of the resulting catalyst nanoparticles. The surfacemorphologies of the multimetal layers with and without catalystlayers were studied by atomic force microscopy �AFM� and scan-ning electron microscopy �SEM�.

Experimental

Sample preparation.— The fabrication process can be summa-rized by the following steps: Silicon substrates �100�, p-type,0.1–10 � cm resistivity, and 100 mm diam were cleaned in piranhasolution for 20 min at 120°C to remove any organic contaminations.A 1 �m thick silicon dioxide layer was thermally grown on thewafer in a furnace by wet oxidation at 1100°C. The deposition ofmetal layers was performed using the magnetron sputtering process.A tantalum layer of 20 nm thickness was used as an adhesion layerfollowed by a 1 �m thick copper layer. A 20 nm thick Ta layer wasdeposited again, which acts as a barrier layer and also preventscopper oxidation.

For deposition of the TiN layer, the wafers were loaded into adifferent magnetron sputtering chamber. The TiN sputtering processwas performed at room temperature and at a chamber pressure of0.93 Pa. The gas flow rates of argon and nitrogen were 50 and20 sccm, respectively. A plasma source power of 500 W was used ata substrate bias voltage of −20 V. The average thickness of 50 nmTiN was deposited for a duration of 6 min. For Al buffer layerdeposition, an electron-beam evaporation process was used. An Allayer of 5 nm and an Fe catalyst layer of 5 nm were depositedwithout breaking the vacuum. A very low chamber pressure of 2� 10−7 Torr was used to ensure the satisfactory uniformity of thefilm.

Three samples with different buffer layers were prepared to studythe formation of catalyst nanoparticles and their effects on CNTgrowth. Figure 2 shows a schematic description of the samplesalong with the thicknesses of individual metal layers. Sample A

represents the sample with the standard silicon dioxide buffer layerused for growing CNTs. In order to realize a practical CNT-baseddevice, where other metal layers such as Cu �conductive layer�,Ti/Ta �adhesion/barrier layer�, and Fe/Ni/Co �catalyst layer� arealso present, samples B and C were also investigated. Sample Butilizes TiN as a buffer layer, while sample C considers Al. In bothcases, the conductive metal layers used were Ta/Cu/Ta.

Annealing of samples and CNT growth in thermal CVD.— Inorder to study the formation of catalyst nanoparticles and materialcomposition, the samples, consisting of the conductive layer, bufferlayer, and iron catalyst layer, were annealed in a thermal CVDchamber. The samples were annealed at 700°C for 15 min in thepresence of H2 �100 sccm� and Ar �400 sccm� with a chamber pres-sure of 4.9 Torr. This annealing step induces the nucleation andformation of catalyst nanoparticles which are needed to grow theCNTs.17 After the annealing step, the CNT growth step was per-formed. The temperature profile for annealing and CNT growingprocess is illustrated in Fig. 3. Once the annealing step was over,acetylene �C2H2� gas �100 sccm� was introduced for 15 min at achamber pressure of 5.9 Torr. After the CNT growth process, thetemperature was ramped down to room temperature of 25°C at ap-proximately 3°C/min.

Characterization of catalyst nanoparticles and CNTs.— For theXPS study of catalyst nanoparticles, a Kratos-Axis spectrometerwith monochromatic Al K� �1486.71 eV� X-ray radiation �15 kVand 10 mA� and hemispherical electron energy analyzer were used.The morphology of the annealed catalyst layer was taken by AFM�Digital Instruments, Santa Barbara� in the tapping mode. Thegrown CNTs were characterized by the Hitachi scanning electronmicroscope under an accelerating voltage of 15–25 kV.

Results and Discussion

CNT growth on silicon dioxide buffer layer (sample A).— Forsample A, a catalyst layer �Fe� having a thickness of 5 nm wasdeposited. After annealing in the thermal CVD chamber at 700°C,the samples were characterized by SEM. Figure 4a shows the topview of sample A after annealing, which has Fe catalyst on 1 �mthick thermally grown silicon dioxide layer. It can be seen that theFe film has been broken into nanoparticles of varying sizes andshapes. The average size of these highly dense nanoparticles variesbetween 30 and 100 nm. Such a highly dense nanoparticle array isideally suited for growing dense CNTs of relatively smaller diam-eter. When the CNTs were grown on this sample, a very satisfactoryCNT growth was observed. Figure 4b shows an SEM image of thevertically aligned CNTs grown on this sample. The CNTs grown onthe sample are of multiwalled type and have a diameter varyingbetween 40 and 100 nm. The dense, vertically aligned CNT bundleswere grown on the annealed iron nanoparticles. Due to the highdensity and the close packing of the nanoparticles, the CNT growthwas preferred in the vertical direction, as it is the only degree offreedom available during CNT growth. The measured electrical re-sistivity of the CNTs is about 0.0097 � cm.17

Figure 2. Schematic diagram of samples used in experiments.

Figure 3. Temperature profiles of annealing and CNT growth steps in thethermal CVD.

2 Journal of The Electrochemical Society, 156 �3� 1-XXXX �2009�2

91

929394959697

98

99100101102103104105106107108109110111112113114115116117118119120121122123124125

126

127128129130131132

133134135136137138139140141142143144145146147

148149150151152153154155

156

157158159160161162163164165166167168169170171172173174175176



CNT growth on Ta/Cu/Ta metal layers with TiN buffer layer(sample B).— When CNTs were grown on sample B, which has TiNbuffer layer, some interesting mechanisms were observed. Figure 5ashows the AFM profile of the sample with an annealed TiN surfacewith underlying Ta/Cu/Ta metal layers. AFM analysis of the surfaceprofile of sample B shows that the grain structure of TiN before theannealing is almost the same as the sputtered TiN thin film, as re-ported in the literature.18 The average grain size of the sputtered TiNfilm is about 20–30 nm. Figure 5b shows the AFM profile of the

same sample after the annealing step. The formation of nanopar-ticles on the TiN layer can be observed from the picture. It can beobserved that the size of nanoparticles formed is larger than thatrequired for catalytic growth of CNTs ��100 nm�. The density ofparticles having a diameter of approximately 20–80 nm is very low,as it can be observed in the figure. This is the reason only the forestof CNTs were grown on the TiN barrier layer rather than a verticallyaligned CNT bundle as widely reported,19 because the formation ofa vertically aligned CNT bundle is due to the vertical direction ofCNT growth being the only possible degree of freedom for highlydense catalyst nanoparticles.

The incomplete formation of Fe catalyst nanoparticles of beanshape can also be seen in Fig. 5b. With the same experimental pa-rameters and catalyst thickness varying from 2 to 10 nm on the ther-mally grown silicon dioxide layer, the vertically aligned dense CNTsbundles were grown. An annealing temperature of 700°C was usedfor 15 min for the Fe catalyst layer on the TiN buffer layer. Theabove experiments show that an annealing temperature of 700°C isnot enough for complete formation of nanoparticles, which may bedue to the difference in adhesion properties between the Fe/SiO2interface and the Fe/TiN interface. Figure 6 shows the top view ofCNTs grown on the TiN buffer layer. It can be seen that the densityof CNTs is much lower in comparison with the results usually re-ported in literature with CNT growth on a stable buffer such assilicon dioxide. The CNTs grew in random orientations, forming alayer of porous CNT-coated surface approximately 2 �m thick.

CNT growth on Ta/Cu/Ta metal layers with an aluminum bufferlayer (sample C).— Figure 7 shows an AFM surface profile ofsample C with an Al buffer layer and Fe catalyst after annealing.Formation of nanoparticles with sizes ranging from 20 to 200 nmcan be seen. The density of nanoparticles is much lower than that ofcatalyst nanoparticles on the buffer layer such as silicon dioxide.According to the available experimental data, the nanoparticles con-sist of aluminum oxide, iron catalyst, and possibly iron oxide, andthe source of oxygen is postulated to be from the atmosphere andoxidized iron catalyst. In order to determine the chemical state of theAl and Fe catalyst layer, XPS examination was performed with an-nealed samples having an Al buffer layer on the Ta/Cu/Ta metallayer. An XPS spectrum of the sample is shown in Fig. 8. One of thecurves shows the peak before etching in argon plasma and the othershows the peak after etching for 960 s. XPS experimental resultsshow that the aluminum oxide particles were formed during theannealing process in the presence of hydrogen flow. The source ofoxygen is the oxidized iron catalyst layer. Although there is a reduc-tion reaction inside the reaction chamber with the presence of hy-drogen gas flow, oxygen from the iron oxide layer may have at-

Figure 4. SEM images of �a� the top view of iron catalyst particles on thesilicon dioxide layer after annealing at 700°C and �b� CNTs grown on thesilicon oxide buffer layer.

Figure 5. �Color online� AFM profile of annealed TiN surface, �a� withoutFe catalyst and �b� with Fe catalyst.

Figure 6. CNTs grown on the TiN buffer layer.

3Journal of The Electrochemical Society, 156 �3� 1-XXXX �2009� 3

177

178179180181182183184185

186

187188189190191192193194195196197198199200201202203204205206207208209210211

212213214215216217218219220221222223224225226227228229230231



tracted the underlying Al layer, forming stable ceramic aluminumoxide particles which act as the stable support for CNT growthduring the recrystallization of Al and Fe nanoparticles. Although it ispossible to have a very thin native aluminum oxide layer on thesamples during Al/Fe deposition and after taking the samples out ofthe vacuum chamber, the thickness is limited to less than 1 nm,enabling the metallic Al to reflow during annealing.

Figure 9 shows the CNTs grown on the Al buffer layer withTa/Cu/Ta metal layers. The density of CNTs is very low and CNTsare randomly oriented. This is due to the formation of low-densitycatalyst nanoparticles allowing the CNTs to grow in random orien-tation. In thermal CVD growth of CNTs, vertically aligned CNTswere obtained due to the densely grown CNTs guiding themselvesvertically. The only possible direction for CNT growth to continue isthe vertical direction, because growth in the lateral direction is im-possible due to the presence of other CNTs on the side. However,with the low density of catalyst, no such self-guiding mechanism ispossible because CNTs continuing to grow will also have the freespace to grow laterally until coinciding with others. Along with theabsence of another guiding mechanism �e.g., electric field in plasma-enhanced CVD�, CNTs grown on the low-density catalyst layer arerandomly oriented. In contrast to this situation, experiments with

CNTs grown on a silicon dioxide layer showed highly dense nano-particle formation and vertically aligned CNT growth, as discussedin the previous section. From the above experimental results, thegrowth mechanism of CNTs in the present case is similar to that ofCNTs with an Al/Fe bimetallic layer catalyst13 with a tip growthmechanism. Figure 10 shows a schematic diagram of the growth ofCNTs on Ta/Cu/Ta metal layers with CNTs grown on the aluminumoxide particles. Figure 11 shows CNTs grown on patternedTa/Cu/Ta metal layers on which an Al buffer layer and an Fe cata-lyst layer were selectively deposited. Figure 11b shows a detailedview of CNTs grown ��4 �m thick� on the multimetal layers. Dueto the possible implication of a buffer layer, the type and conductiv-ity of the CNTs may differ from those grown on the silicon oxidelayer, and further research is needed to investigate the type andelectrical properties of those CNTs.

Conclusion

CNT growth on Ta/Cu/Ta metal layers was studied with differ-ent buffer layers, namely, Al and TiN. In order to compare thegrowth of CNTs on TiN and Al buffer layers with those grown onsilicon dioxide buffer layers, we studied the formation of catalystnanoparticles on the silicon dioxide layer. Highly dense catalyst par-ticles were observed and vertically aligned CNTs were grown, aswidely reported in literature. From experiments with a TiN bufferlayer, we found a stable TiN layer after annealing without recrystal-lization. The grain size of the TiN layer was the same as that ofsputtered TiN thin film reported in the literature. Study of the sizeand density of nanoparticles shows that the larger and incompleteformation of nanoparticles after annealing is attributed to the smaller

Figure 7. �Color online� AFM profiles of Al2O3 nanoparticles after anneal-ing at 700°C.

Figure 8. �Color online� XPS spectrum of aluminum oxide before and afterargon plasma etching in XPS.

Figure 9. SEM image of CNTs grown on the aluminum buffer layer.

Figure 10. Schematic diagram of the CNT growth mechanism on an Albuffer layer. Al was formed to Al2O3 nanoparticles after annealing.

4 Journal of The Electrochemical Society, 156 �3� 1-XXXX �2009�4

232

233234235236237238239240241242243244245246247248249250251252253

254

255256257258259260261262263264265266267268

269

270271272273274275276277278279280281



density of nanoparticles available for CNT growth. This causesgrowth of CNTs in random directions rather than in the verticaldirection. From the experiments with an Al buffer layer, the forma-

tion of aluminum oxide nanoparticles was observed together with Feand Fe2O3 in the XPS analysis. We found that the CNT growthmechanism with the Al buffer layer is due to the formation of Al2O3nanoparticles on the underlying metal layers during the CNT growthprocess. It can be concluded that the growth of randomly orientedCNTs on the multimetal layer with Al and TiN buffer layers on themetal layers is due to the sparse nanoparticle formation of the CNTgrowth process during annealing as compared to the CNT grown ona silicon dioxide buffer layer. A randomly oriented CNT-coated sur-face was observed on the samples with Al and TiN buffer layers onthe underlying metal layers, although the TiN buffer layer may besuitable due to its electrically conductive properties in practical ap-plications.

Acknowledgments

The authors acknowledge support by the Agency for Science,Technology and Research �A*STAR�, Singapore, under SERC grantno. 042 114 0042 and the industrial sponsorship by Delphi Automo-tive, Singapore.

Nanyang Technological University assisted in meeting the publicationcosts of this article.

References1. 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. Enoks-

son, 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�.

Figure 11. SEM picture showing �a� the CNTs grown on the Al buffer layeron a Ta/Cu/Ta multimetal system and �b� a magnified view of CNTs grownon a selective area.

5Journal of The Electrochemical Society, 156 �3� 1-XXXX �2009� 5

282

283284

285

286287288289290291292293294295296297

298

299300301302

303304

305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336

AQ:#1

AQ:#2



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

NOT FOR PRINT! FOR REVIEW BY AUTHOR NOT FOR PRINT!


journal of the electrochemical society, 156 0013-4651/2009 ...€¦ · for deposition of the tin...

Documents