Anda di halaman 1dari 4

Proceedings of 13th International Conference on Computer and Information Technology (ICCIT 2010) 23-25 December, 2010, Dhaka, Bangladesh

Analyzing Carbon Nanotube Interconnects in VLSI Application


1

Mahmudur Rahman , Ahrar Ahmed Chowdhury Department of Electrical Electronic and Communication Engineering, Military Institute of Science and Technology (MIST) Section-12, Mirpur, Dhaka-1216, Bangladesh mahmudndc@yahoo.com, ahrar_1988@yahoo.com Abstract of its sp2 bonding structure which helps to increase its conductivity in a high level. The sp3 bonding is less strong than this sp2 bonding and therefore a carbon nanotube has unique strength which made it as one of the strongest material in the world. In this paper we characterize metallic SWNT by Raman spectroscopy. Raman spectroscopy is a pioneer technique for characterizing carbon nanotubes very efficiently. Raman spectroscopy has superb non destructing property and it has very wide range resolution. In Raman technique sample infectivity risk is very low and it can work under very high temperature and high pressure process. The scattering of monochromatic radiation in Raman spectra gives an overview to the molecular structure of a carbon nanotube [2]. The difference in incident laser light energy and the scattered energy is related to the vibrational energy of molecules. For Stokes Raman scattering incident laser energy is greater than the scattered light energy and for anti stokes Raman scattering the scattered light energy is greater than the incident laser energy [2]. We observed stokes Raman scattering for different Raman samples of metallic SWNTs. The atomic vibrations related with the intrinsic properties of SWNTs cause different peaks in the Raman spectroscopy ,Radial breathing mode (RBM), disorder peak (D peak), graphitic peak (G). The G peak is very significant feature to attribute the bonding structure and electrical properties of SWNTs. The G peak originates from tangential vibration of carbon atoms in the SWNTs and it is present in most graphite like material. Lorentzian and BWF line shapes are two basic components to fit this tangential G band Raman spectra [3]. For semiconducting SWNTs lower frequency BWF line shape is non dominating but for metallic SWNTs BWF line shape is broad and dominating [4]. Higher frequency Lorentzian line shape is observed for both types of SWNTs. Metallic Carbon nanotubes have excellent properties of high current carrying capacity and mechanical stability and hence they are recently considered as a unique material for VLSI interconnects. Recent performance of CNT interconnects can potentially offer decreased in SWNT bundles as the number parallel

*1

Single Wall Carbon Nanotubes exhibit outstanding contributions in the recent VLSI interconnections. Interconnects analyzed in VLSI circuits depends on the electrical properties of carbon nanotubes. Metallic carbon nanotubes are very distinct for their ballistic conductivity in nano level interconnections. Different peaks are analyzed in Raman spectroscopy technique for characterizing metallic carbon nanotubes. The performance analysis of metallic carbon nanotubes is compared with the conventional Cu interconnects. In this study we analyzed resistivity and capacitance of carbon nanotubes interconnects which indicates carbon nanotubes interconnect are the most prominent solution for the future VLSI technologies. Keywords: Interconnects, Carbon Nanotube, Raman, VLSI, Resistance, Capacitance. I. INTRODUCTION Carbon Nanotube is a new form of pure carbon which is a rolled hexagonal carbon cylindrical structure. Name of Carbon nanotube is derived from its size and shape and it has diameter in nanometre scale. The diameter of a carbon nanotube can be ranged up to several nanometers. A carbon nanotube has the sp2 chemical bonding structure of carbon atoms. The bonding assortment of a carbon nanotube is very similar to a spined single layer graphene sheet. Different electrical and mechanical properties are distinguished for the various spinning alignment of a carbon nanotube. Single wall carbon nanotube (SWNT) and multi walled carbon nanotube (MWNT) are the two major classifications of carbon nanotubes. [1]The SWNT has a single wall of rolled grapheme sheet and MWNT has two or more walls of rolled grapheme sheet. The chirality of carbon nanotube is presented by indices (n, m). A chiral vector is C=na1+ma2 where a1 and a2 are two unit vectors which determine the rolling direction. SWNT exhibits different chiralities i.e. armchair, zigzag and chiral. An armchair carbon nanotube is fully metallic (n-m=0) but for moderate metallic n-m is multiple of 3 and the other chiralities of nanotubes are semiconducting nanotubes [1]. A metallic SWNT has delocalised electrons because

978-1-4244-8494-2/10/$26.00 2010 IEEE

237

Raman Intensity

Raman Frequency (cm-1)


Fig. 1 G peak with first three are metallic (a), (b), (c) and the others (d),( e), (f) are semiconducting SWNTs. The dashed lines are showing BWF peaks and Lorentzian peaks. These fitted peaks are exploited to find the peak positions of the two intense G peaks. The lower frequency BWF peaks are broad and strong for the left three figures which are metallic SWNTs. For the right figures BWF peaks are narrowed and weak, hence these are semiconducting SWNTs.

walls are increased. Impurities (semiconducting mixtures of CNT) in SWNT bundles are however a challenge for the lower resistance metallic interconnects. Specific synthesis process is required for the growth of SWNTs and its bundles. But the synthesis of SWNT interconnects bundles is cumbersome because the fertility of catalysts is less for SWNTs. In this paper we first characterized the metallic SWNTs from the Raman G peak analysis. Six samples are observed in Raman spectroscopy techniques which are synthesized from Chemical vapour deposition method (CVD) method. Among these samples three of them were metallic SWNTs. The different properties (resistivity, capacitance, frequency response) of these metallic SWNTs are compared with Cu interconnects. II. EXPERIMENTAL Six samples of Raman spectra are collected from et. al Jorio [7]. All the samples were synthesized by CVD method on a slightly oxidized Si/SiO2 substrate which contained nanometer-size iron catalyst particles. The Raman spectra from isolated SWNTs on the Si/SiO2 substrate were obtained by using two Raman systems: a single monochromator spectrometer outfitted with a cooled charge coupled device detector , notch filters and

a Kaiser optical system and another one is modular research micro-Raman spectrometer. A backscattering technique was used to collect the Raman spectra. using a 1003 objective The laser spot were 2.41 eV ,2.54eV for semiconducting SWNTs and 1.58 eV , 2.41 eV fpr metallic SWNTs.2.41 eV! and 2.54 eV lines were produced from an Ar-ion laser, and the 1.58 eV were produced from a Ti:sapphire laser were used to obtain the Raman spectra. The laser power impinging on the substrate was 10 mW. III. CNT CHARACTERIZATION The different samples that observed in the Raman spectroscopy presented two types of Carbon Nanotubes, metallic SWNTs and semiconducting SWNTs. To characterize the particular metallic SWNTs from different types of samples we analyzed the typical G peaks in Raman spectroscopy. In Raman spectroscopy G peaks are visible at ~1600 cm-1.The visible G peaks have two intense features of low frequency G- peak and high frequency G+ peak. We elicited these two sub peaks of G peaks with the assistance of peak fitting process. In peak fitting process G- and G+ peaks are featured with particular Breight Wigner Fano line shape and lorentzian line shape respectively. G peaks of

238

semiconducting SWNT are firmly responsible due to the symmetric lorentzian line shape. However a strong G+ peak is delineated for the semiconducting samples of SWNTs. However a very minute G- peak is also observable in semiconducting Raman peaks which is quite negligible. In the contrary a strong G- peak is observed for the metallic SWNTs. This is due to the origin of Breight Wigner Fano line shape in the continuum of the metallic energy levels. Particular Phonon-plasmon coupling is an intrinsic feature of metallic SWNTs. Due to this coupling the Breight Wigner Fano (BWF) lineshape is broad and as a consequence G- peak is dominating and strong for metallic SWNTs. We analysed six samples of carbon nanotubes in the Raman spectroscopy technique. For the first three samples we observed strong and broad lower frequency G- peaks. And the other three samples Gpeaks is weak and non-dominating. This particular behaviour of the peaks we tentatively stated the first three are metallic SWNTs and other three are semiconducting SWNTs. IV. PERFORMANCE ANALYSIS OF INTERCONNECT A. Interconnect Resistance The interconnect resistance of CNT is exceeding one of the superior alternative comparing with other conventional interconnects. It has excellent property of low resistance in the VLSI applications. Resistance of CNT interconnect is strongly dependent on its length. It maintains an inverse relation with length. The intermittent distance between two collusions for an electron is the mean free path and CNT interconnect has a mutual relation with this mean free path. CNT has mean free path of typically 1 m [8, 9, 10]. For CNT lengths less than 1 m (L< CNT), a ballistic electron transport occurs within the nanotube and the resistance is very low for the very short length (6.45 K). The compatible equation for this length is [11] h R = R f = 2 = 6.45K (1) 4e Here h is the plancks constant and e is the electron charge. However, for CNT lengths greater than 1 m (L >CNT) resistance increases with length and hence the new equation for the length is [12] Rf h 1 R=( 2 ) = (2) 4e CNT CNT For SWNT bundle we get a much propitious resistance in interconnect. In a bundle there exist several numbers of parallel paths for carrying the electron so the equivalent resistance is much smaller for a bundle. The resistance can be given as R (3) Rbundle = SWNT nCNT In this equation nCNT is the total number of CNTs forming bundle. nCNT also represent the number of the

parallel path so the conductivity increase with the number of SWNT in a bundle. The resistance derived above is much smaller than the observed d.c. resistance of a CNT (at low bias) in practical. This is due to the presence of imperfect metalnanotube contacts which give rise to an additional contact resistance. h 1 R=( 2 ) + 2Rc (4) 4e CNT In this equation Rc is the contact resistance at the one end which given rise for the practical interconnect resistance. Since interconnect is formed between two ends of contacts. So we propose total number of resistance should double (2Rc). The resistance of these contacts is often so high that it veils the observation of CNT internal resistance. The observed resistance for CNTs practically so high that it can be in the range of 100 K [13, 14]

Resistivity ( -cm)

Length (m)
Fig. 2 Characteristics of Resistivity of different interconnectdepending on length [15].

In this figure its is observable that there are two regions that, short length (< 30 m) and long length ( >30m) interconnects. The different curves presents interconnects resistivity for Cu and CNTs. The Fm is the fraction of metallic CNTs in SWNT bundles. Considering short length (10 m), for the curve Fm=1/3, the metallic proportion is less so resistivity is high (~3.4 -cm) and for the curve Fm=1, the metallic proportion is very high so resistivity is very low (~1.1 -cm). Comparing SWNT bundle interconnect of Fm=1 with the Cu interconnect it is quite observable that SWNT bundle has lower resistivity than Cu (2 cm). This low resistivity is very effective for the VLSI application. For the metallic SWNT the optimum point is ~23 m. After this point the resistivity decreases than the Cu interconnects. So for practical approximation the length of metallic MWNT should be much lower which is constraint for interconnect design. For short length region the rate of change of resistivity of metallic interconnect is much higher (~3.42 -cm/m) and for

239

long length the rate of change is very small (~0.0016 cm/m). B. Interconnect Capacitance In interconnects there are two types of capacitance, electrostatic capacitance (Ce) and quantum capacitance (Cq). The electrostatic capacitance occur between the ground substrate and the interconnect CNT. This parasitic capacitance largely depends on the interconnect distance and the gap between interconnect and substrate material. Electrostatic capacitance is the intrinsic plate capacitance for isolated CNT. Electrostatic capacitance is described by the equation 2 (5) Ce = y ln d Here d is the diameter of the CNT maintaining y distance from ground plate. For d=1nm and y=1m Ce=30aF/um When current flows in nanotube a quantum static energy is stored which is accounted for quantum capacitance. The frequency of interconnects depends on these capacitance. For the large value of Ce the propagation delay increase and the information carried by interconnects become slow. Parasitic capacitance is small for SWNT compare to the CNT bundles. In CNT bundles the individual capacitance of each SWNT is adjoined as all SWNTs are parallel connected in nanotube bundle. V. CONCLUSION In this paper we have discussed the different properties of CNT interconnects. A metallic SWNT interconnect is very effective in the field of VLSI for its outstanding ballistic conductivity, low resistance and low capacitance. To characterize the metallic interconnects from different samples of CNTs synthesized form CVD method; we utilized G peak analysis technique in the Raman spectroscopy. These estranged metallic interconnects were then studied for the VLSI interconnections. Resistance for short length (mean free path< 1m) CNT interconnects is very useful due to their very low value comparing with Cu interconnects. Capacitances for CNT interconnects are also depicted in respect of speed and propagation delay. Finally we concluded CNT interconnects would embed in the future VLSI technology. VI. ACKNOWLEDGMENT Authors are very grateful to Proffesor M. S. Dresselhaus, Department of Physics, Massachusetts Institute of Technology (MIT), Cambridge and Asst. professor Ado Jorio, Department of Physics, University Federal de Minas Gerais (UFMG), Brazil for sharing their experimental data to support this work. Finally authors are thankful to Dr. Sharif Mohammad Mominuzzaman, Professor, Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology for his helpful support throughout the work.

VII. REFERENCES [1] M. S. Dresselhausa, G. Dresselhausb, R. Saito, A. Jorio, Raman spectroscopy of Carbon Canotubes, Elsevier, Physics Reports 409 4799, 2005. [2] (2010) The Andor Technology Website. [Online]. Available: http://www.andor.com/learning/applications/Rama n_Spectroscopy/ [3] S. D. M. Brown, A. Jorio, P. Corio, M. S. Dresselhaus, G. Dresselhaus, R. Saito, and K.Knepp, Origin of the Breit-Wigner-Fano lineshape of the tangential G-band feature of Metallic Carbon Canotubes, Phys. Rev. B, vol. 63, 155414, 2001, [4] A. Jorio. A. G. Souza Filho, G. Dresselhaus, M. S. Dresselhaus, A. K. Swan, M. S. nl, B. B. Goldberg, M. A. Pimenta, J. H. Hafner, C. M. Lieber, and R. Saito, G-band resonant Raman study of 62 isolated Single-Wall Carbon Nanotubes, Phys. Rev. B, vol. 65, 155412, 2002. [5] B. Q. Wei, Reliability and Current Carrying Capacity of Carbon Nanotubes, Appl. Phys. Lett., Vol. 79, No 8, pp. 1172-1174, 2001. [6] P. L. McEuen, Single-Walled Carbon Nanotube electronics, IEEE Trans. Nanotechnology, Vol. 1, No. 1, pp. 78-85, 2002 [7] A. Jorio, A. G. Souza Filho, G. Dresselhaus,4 M. S. Dresselhaus, A. K. Swan, M. S. U nlu, B. B. Goldberg, M. A. Pimenta, J. H. Hafner,C. M. Lieber,and R. Saito, G-band resonant Raman study of 62 isolated Single-Wall Carbon Nanotubes physical review b, volume 65, 155412 [8] N. Srivastava and K. Banerjee, A comparative scaling analysis of metallic and Carbon Nanotube Interconnections for nanometer scale VLSI Technologies, Proc. 21st Intl. VLSI Multilevel Interconnect Conf., 2004, pp. 393-398. [9] J. Kong, Quantum Interference and Ballistic Transmission in Nanotube Electron Waveguides, Physical Review Letters, Vol. 87, No. 10, 106801, 2001. [10] J.-Y. Park, Electron-Phonon Scattering in Metallic Single-Walled Carbon Nanotubes, Nano Letters, Vol. 4, No. 3, pp. 517-520, 2004. [11] S. Datta, Electrical Resistance: An Atomistic View, Nanotechnology, Vol.15, pp. S433-S451, 2004. [12] S. Datta, Electronic Transport in Mesoscopic Systems, Cambridge Univ. Press, 1995. [13] Th. Hunger, Transport in ropes of Carbon Nanotubes: contact barriers and Luttinger Liquid theory, Physical Review B, Vol. 69, 195406, 2004. [14] W. Liang, Fabry-Perot interference in a Nanotube Electron Waveguide, Nature, Vol. 411, pp. 665669, 2001. [15] Hong Li and Kaustav Banerjee, High-frequency analysis of Carbon Nanotube Interconnects and implications for on-chip inductor design, IEEE Transactions On Electron Devices, Vol. 56, No. 10, october 2009

240

Anda mungkin juga menyukai