applied sciences
Review Mini-Review: Modeling and Performance Analysis of Nanocarbon Interconnects
Wen-Sheng Zhao 1 , Kai Fu 1, Da-Wei Wang 2, Meng Li 3, Gaofeng Wang 1,* and Wen-Yan Yin 2,* 1 Key Lab of RF Circuits and Systems of Ministry of Education, School of Electronics and Information, Hangzhou Dianzi University, Hangzhou 310018, China; [email protected] (W.-S.Z.); [email protected] (K.F.) 2 Key Lab of Advanced Micro-Nano Electronic Devices and Smart Systems of Zhejiang Province, College of Information Science and Electronic Engineering, Innovative Institute of Electromagnetic Information and Electronic Integration, Zhejiang University, Hangzhou 310058, China; [email protected] 3 Beijing Aerospace Technology Institute, Beijing 100074, China; [email protected] * Correspondence: [email protected] (G.W.); [email protected] (W.-Y.Y.)
Received: 25 April 2019; Accepted: 24 May 2019; Published: 28 May 2019
Abstract: As the interconnect delay exceeds the gate delay, the integrated circuit (IC) technology has evolved from a transistor-centric era to an interconnect-centric era. Conventional metallic interconnects face several serious challenges in aspects of performance and reliability. To address these issues, nanocarbon materials, including carbon nanotube (CNT) and graphene, have been proposed as promising candidates for interconnect applications. Considering the rapid development of nanocarbon interconnects, this paper is dedicated to providing a mini-review on our previous work and on related research in this field.
Keywords: carbon nanotube (CNT); graphene nanoribbon (GNR); circuit modeling; interconnects; nanocarbon; electrothermal co-simulation
1. Introduction The breakthrough development of the semiconductor industry has revolutionized human society, from personal electronic gadgets, commercial and industrial equipment to military and aeronautical facilities. As predicted by Moore’s law, the number of transistors within a chip doubles about every two years, while the cost comes down [1]. According to the International Technology Roadmap for Semiconductors (ITRS) projection, a 10 nm minimum feature size could support a tera-scale chip with a trillion transistors by 2020 [2]. Such phenomenal progress has been achieved through scaling of digital integrated circuit (IC) feature size to smaller physical dimensions. The ongoing miniaturization of the IC feature size has had a significant benefit in increasing the transistor speed. However, different from the transistor, the interconnect performance would be degraded due to the reduced conduction area and increased scattering probability for electrons [3,4]. Under such circumstances, the chip performance is restricted by the shrinking interconnect dimensions on account of both the interconnect delay and the power dissipation. Moreover, the interconnect reliability has been becoming a more and more important problem, as the ampacity of conventional Cu wire cannot satisfy the increasingly stringent requirements [2,5,6]. Therefore, interconnects have become the major challenge in the design of modern ICs, thereby leading to the transition of IC technology from transistor-centric to interconnect-centric [7]. To address the interconnect challenges for next-generation ICs, diverse improvements of interconnect optimization and design methods have been reported, such as a low-k dielectric structure, three-dimensional integration, and inter-chip optical interconnects [8–10].
Appl. Sci. 2019, 9, 2174; doi:10.3390/app9112174 www.mdpi.com/journal/applsci Appl. Sci. 2019, 9, 2174 2 of 20
Appl.and Sci. design2019, 9, 2174methods have been reported, such as a low-k dielectric structure, three-dimensional2 of 21 integration, and inter-chip optical interconnects [8–10]. Nanocarbon materials have attracted much attention since the carbon nanotube (CNT) was discoveredNanocarbon by an materials arc-discharge have attractedevaporation much method attention in 1991 since [11]. the Graphene, carbon nanotube a Nobel Prize (CNT) honored was discovereddiscovery, by has an further arc-discharge promoted evaporation the research method in this infield 1991 [12]. [11 It]. was Graphene, found that a Nobel nanocarbon Prize honored materials discovery,have many has furtherextraordinary promoted physical the research properties. in this For field example, [12]. It wasthe foundultrahigh that thermal nanocarbon conductivity materials of havenanocarbon many extraordinary materials can physical help heat properties. dissipation For example,in high-density the ultrahigh integrated thermal systems conductivity [13–15]. ofThe nanocarbonmaximum materialscurrent-carrying can help density heat dissipationof a CNT is inmore high-density than two orders integrated higher systems than that [13 –of15 Cu]. Thewires, maximumthereby mitigating current-carrying the electromigration-induced density of a CNT is more reliability than two ordersproblems higher [16]. than It is that natural of Cu to wires, apply therebynanocarbon mitigating materials the electromigration-inducedas an alternative option to potentially reliability problemsreplace Cu [ 16for]. interconnects It is natural and to applypassive nanocarbondevices in materialsICs [17–22]. as anIn recent alternative years, option there tohave potentially been many replace publications Cu for interconnects in the literature and devoted passive to devicesthe design, in ICs modeling/analysis, [17–22]. In recent years, and fabrication/integration there have been many publicationsof nanocarbon in theinterconnects. literature devoted This paper to theseeks design, to review modeling recent/analysis, research and efforts fabrication and progress/integration in the of nanocarbonnanocarbon interconnects.interconnects. In This particular, paper seeksthis topaper review focuses recent on research the modeling efforts andand progressdirect current in the (DC) nanocarbon performance interconnects. analysis Inof particular,nanocarbon thisinterconnects paper focuses for onnext-generation the modeling digital and direct ICs. current (DC) performance analysis of nanocarbon interconnects for next-generation digital ICs. 2. Graphene Nanoribbon (GNR) Interconnects 2. Graphene Nanoribbon (GNR) Interconnects Physically, graphene is a 2D monolayer of carbon atoms packed into a honeycomb lattice, and Physically, graphene is a 2D monolayer of carbon atoms packed into a honeycomb lattice, the quasi-1D graphene nanoribbon (GNR), as shown in Figure 1a, can be utilized as on-chip and the quasi-1D graphene nanoribbon (GNR), as shown in Figure1a, can be utilized as on-chip interconnects [23]. Depending on the edge shape, a GNR can be zigzag, armchair, or chiral (other interconnects [23]. Depending on the edge shape, a GNR can be zigzag, armchair, or chiral (other shapes). The zigzag GNR is always metallic, whereas the armchair GNR is metallic or shapes). The zigzag GNR is always metallic, whereas the armchair GNR is metallic or semiconducting, semiconducting, depending on the number of carbon atoms across its width. Different from the GNR, depending on the number of carbon atoms across its width. Different from the GNR, a CNT’s chirality a CNT’s chirality is defined by its circumferential edge shape. A single-walled carbon nanotube is defined by its circumferential edge shape. A single-walled carbon nanotube (SWCNT) can be (SWCNT) can be visualized as a seamlessly rolled-up GNR, on the basis of which a novel fabrication visualized as a seamlessly rolled-up GNR, on the basis of which a novel fabrication method has been method has been developed to unzip the CNT to form a GNR [24]. A multi-walled carbon nanotube developed to unzip the CNT to form a GNR [24]. A multi-walled carbon nanotube (MWCNT) is a (MWCNT) is a parallel assembly of coaxial SWCNTs, and the neighboring shells in an MWCNT are parallel assembly of coaxial SWCNTs, and the neighboring shells in an MWCNT are separated by the separated by the van der Waals gap. van der Waals gap.
(a) (b) (c)
FigureFigure 1. Schematics1. Schematics of (a )of monolayer (a) monolayer graphene graphene nanoribbon nanoribbon (GNR), ( b(GNR),) single-walled (b) single-walled carbon nanotube carbon (SWCNT),nanotube and (SWCNT), (c) multi-walled and (c) multi-walled carbon nanotube carbon (MWCNT). nanotube (MWCNT).
2.1.2.1. Multilayer Multilayer Graphene Graphene Nanoribbon Nanoribbon (MLGNR) (MLGNR) Interconnects Interconnects InIn comparison comparison to to the the CNT, CNT, the the GNR GNR can can be be easily easily controlled controlled horizontally horizontally and and is moreis more compatible compatible withwith conventional conventional lithography lithography [24 [24].]. By By using using the the tight-binding tight-binding approximation, approximation, the the conductance conductance ofof a a monolayer GNR has been calculated as a function of width W, Fermi level E , and specularity monolayer GNR has been calculated as a function of width 𝑊, Fermi level F𝐸 , and specularity parameterparameterp for 𝑝 edge for edge diffuse diffuse scattering scattering in Reference in Reference [25]. It [2 was5]. It demonstrated was demonstrated that the that atomically the atomically thick monolayerthick monolayer GNR could GNR outperform could outperform Cu wire forCu widths wire for below widths 8 nm. below Figure 8 2nm.a shows Figure the 2a schematic shows the Appl. Sci. 2019, 9, 2174 3 of 21 Appl. Sci. 2019, 9, 2174 3 of 20 schematicdiagram of diagram a monolayer of a monolayer GNR interconnect GNR interconnect with its equivalent with its equivalent circuit model. circuit In model. the model, In theR model,c is the
𝑅contact is the resistance, contact resistance, and it can and be calculatedit can be calculated by: by:
𝑅 =𝑅 +𝑅 (1) Rc = Rmc + RQ (1) where 𝑅 is the imperfect contact resistance, and it highly depends on the fabrication technology. where Rmc is the imperfect contact resistance, and it highly depends on the fabrication technology. 𝑅 is the quantum contact resistance, and it is calculated as 𝑅 =ℎ ⁄ 2𝑒 𝑁 , where ℎ is the = 2 Planck’sRQ is the constant quantum and contact 𝑒 is resistance,the electron and charge. it is calculated The number as R ofQ conductingh/ 2e Nch channels, where hinis athe monolayer Planck’s GNRconstant can andbe calculatede is the electron by [26,27]: charge. The number of conducting channels in a monolayer GNR can be calculated by [26,27]: 1 1 𝑁 = + X 1+e 1 ⁄ X 1+e 1 ⁄ (2) Nch = + (2) (En EF)/(kBT) (En+EF)/(kBT) n 1 + e − n 1 + e where 𝑘 is the Boltzmann’s constant and 𝑇 is the temperature. For a given width of the metallic armchairwhere kB GNRis the with Boltzmann’s 𝑊 > 10 constant nm and 𝐸 and >0.1T is eV the, 𝑁 temperature. can be approximated For a given as width 1.2𝑊𝐸 of the, where metallic 𝑊 𝐸 andarmchair are GNR in withunitsW of> nanometers10 nm and E andF > 0.1electroneV, N chvolts,can berespectively approximated [28]. as The1.2WE totalF, whereresistanceW and of EaF monolayerare in units GNR of nanometers is given by: and electron volts, respectively [28]. The total resistance of a monolayer GNR is given by: ℎ 𝐿 1 𝑅 =𝑅 +𝑅 𝐿= X 1 + ! 1 − (3) 2𝑒h 𝜆L − Rtotal = Rc + RSL = 1 + , (3) 2e2 λ i eff,i where 𝐿 is the interconnect length, and 𝜆 , is the effective MFP (mean free path) for the 𝑖th L i subband.where is The the interconnectMFP is determined length, andby variousλeff,i is the scattering effective mechanisms, MFP (mean free and path) it is fora function the th subband. of the specularityThe MFP is parameter determined 𝑝 by [29]. various Herein, scattering the fully mechanisms, diffusive GNR and itcan is a be function represented of the specularityby 𝑝=0, p p = p = meanwhile,parameter 𝑝=1[29]. Herein, represents the that fully the di ffGNRusive is GNRfully specular. can be represented by 0, meanwhile, 1 represents that the GNR is fully specular.
Rc/2 RSdz LKdz LMdz Rc/2 CQdz
CEdz
Side contact N=1+Inter(H/δ) Top contact Contact R layer W δ MLGNR W SLGNR Rperp MLGNR H
Ground plane Ground plane Ground plane (a) (b) (c)
FigureFigure 2. 2. SchematicsSchematics of of(a) (monolayera) monolayer GNR GNR interconnect, interconnect, (b) top-contacted (b) top-contacted MLGNR, MLGNR, and (c and) side- (c) contactedside-contacted MLGNR MLGNR interconnect. interconnect.
However,However, it it is is worth worth noting noting that that the the physical physical properties properties of monolayer of monolayer GNR GNR is susceptible is susceptible to the to substratethe substrate influence, influence, which which can canbe beavoided avoided for for the the case case of of multilayer multilayer GNR GNR (MLGNR). (MLGNR). Moreover, Moreover, MLGNRMLGNR has has superior superior characteristics characteristics for for signal signal propagation propagation as as the the multilayer multilayer structure structure can can effectively effectively reducereduce the the interconnect interconnect resistance resistance [29–32]. [29–32]. Figu Figureres 22b,cb,c showshow thethe schematicschematic diagramsdiagrams ofof MLGNRMLGNR interconnectsinterconnects with with top top contacts contacts and and side side contacts, contacts, respectively. respectively. In In the the figure, figure, 𝛿δ denotes the the van van der der Waal’sWaal’s gap betweenbetween thethe adjacentadjacent graphene graphene layers, layers,H 𝐻is the is the thickness thickness of theof the MLGNR MLGNR interconnect, interconnect, and andthe numberthe number of graphene of graphene layers layersN = 1𝑁=1+Inter+ Inter(H/δ) ,𝐻𝛿 where⁄ , where “Inter (“Inter)” represents ⋅ ” represents that only that the only integer the · integerpart is considered.part is considered. ToTo date, most of the experimentsexperiments onon multilayermultilayer graphene graphene have have employed employed top top contacts, contacts, as as shown shown in inFigure Figure2b [2b33 –[33–35].35]. For For the top-contactedthe top-contacted MLGNR MLGNR interconnect, interconnect, Kumar, Kumar, et al. [et36 ]al. developed [36] developed a resistor a resistorcircuit model circuit with model the considerationwith the consideration of in-plane resistance of in-planeRlayer resistanceand perpendicular 𝑅 and resistance perpendicular between resistanceadjacent layers betweenRperp adjacent. It was found layers that 𝑅 the. perpendicularIt was found that resistance the perpen makesdicular the conductance resistance amakes nonlinear the conductancefunction of the a nonlinear number function of graphene of the layers. number Further, of graphene Pan, etlayers. al. [37 Further,] conducted Pan, et a benchmarkal. [37] conducted study, awhich benchmark indicated study, such top-contactedwhich indicated MLGNR such cantop- exhibitcontacted advantages MLGNR under can certainexhibit circumstances.advantages under With certain circumstances. With the increasing interconnect length, however, the in-plane resistance increases, while the perpendicular resistance decreases continually. It can be seen in Figure 3, that as Appl. Sci. 2019, 9, 2174 4 of 21
Appl.the increasingSci. 2019, 9, 2174 interconnect length, however, the in-plane resistance increases, while the perpendicular4 of 20 resistance decreases continually. It can be seen in Figure3, that as the length exceeds a certain value, the theinfluences length exceeds of the contact a certain type value, and interlayer the influences coupling of the can contact be ignored, type and interlayer the top-contacted coupling MLGNR can be ignored,can be treated and the as top-contacted a side-contacted MLGNR one [28 can]. be treated as a side-contacted one [28].
100 1000 μm ) Ω 10 100 μm
L=10 μm
1 Resistance (k Top-contacted Side-contacted 0.1 20 40 60 80 Number of Layers FigureFigure 3. EffectiveEffective resistances resistances of of top-contacted top-contacted and and side-contacted side-contacted MLGNR MLGNR interconnects interconnects with Fermi levellevel E𝐸F ==0.60.6 eV,eV, specularity parameter p𝑝=1= 1,, andand cc-axis-axis resistivityresistivityρ 𝜌c ==0.30.3 Ω Ω⋅cmcm [28[28].]. ·
Figure4 shows the multi-conductor transmission line model and the simplified equivalent single-conductor (ESC) transmission line model of a side-contacted MLGNR interconnect. In the multi-conductorFigure 4 shows model, the the multi-conductor mutual capacitance transmission and inductance line model between and the the adjacent simplified graphene equivalent layers single-conductorhave been considered. (ESC) transmissi The MLGNRon line interconnects model of area side-contacted assumed to be MLGNR decoupled, interconnect. because severalIn the multi-conductordecoupled graphene model, layers the havemutual been capacitance grown experimentally and inductance in between References the [ 33adjacent,34], and graphene demonstrated layers have been considered. The MLGNR interconnects are assumed to be decoupled, because several theoreticallyAppl. Sci. 2019, 9, in 2174 Reference [39]. In this case, the resistance of the MLGNR interconnect is a parallel5 of 20 decoupled graphene layers have been grown experimentally( in) References [33,34], and demonstrated = PN i,i connection of the resistance of each layer, i.e., RS i=1 RS . The per-unit-length (p.u.l.) magnetic theoreticallyand 100 times in larger Reference than the [39]. minimum-sized In this case, the gate, resistance respectively, of the at MLGNR these levels. interconnect The imperfect is a parallel contact inductance LM and electrostatic capacitance CE are determined , by the MLGNR geometry and its connectionresistance isof neglected,the resistance and ofthe each Fermi layer, level i.e., is 𝑅 set = as∑ 0.6𝑅 eV.. TheAs shownper-unit-length in Figure (p.u.l.) 5, the magneticMLGNR surrounding dielectrics. By applying the boundary conditions, the equivalent kinetic inductance LK inductanceinterconnects 𝐿 with and fullyelectrostatic specular capacitance edges show 𝐶 su areperior determined performance by the overMLGNR Cu wiresgeometry at both and theits and quantum capacitance CQ in the ESC model can be obtained by recursive schemes [40,41]. As the surrounding dielectrics. By applying the boundary conditions, the equivalent kinetic inductance 𝐿 kineticintermediate inductance and global is much levels. larger Such than an the advantage magnetic inductance,can be enhanced the equivalent with the kinetic feature inductance size scaling in and quantum capacitance 𝐶 in the ESC model can be obtained by recursive schemes [40,41]. As the thedown. ESC However, model can when be expressed the specularit as [28y]: parameter decreases to 0.8, the electrical performance of the kineticMLGNR inductance interconnects is much will larger degrade than theto themagnetic comparable inductance, level theof equivalentCu wires. kineticTherefore, inductance it can bein theconcluded ESC model that canthe MLGNRbe expressed interconnects as [28]: have 20the potential to outperform conventional Cu wires at LK nH/µm (4) an advanced technology node, and more attention≈ 3NWE20 shouldF be paid to the edge quality. 𝐿 ≈ nH⁄ 𝜇m (4) 3𝑁𝑊𝐸
(1,1)The equivalent(1,1) (1,2) quantum(2,3) capacitance(N-1,N) is mainly(N,N) determined by the mutual coupling one between RS dz L K dz LM dz LM dz L M dz LM dz the1 adjacent layers. As the number of layers is larger than eight, the equivalenti quantum capacitance (1,1) C dz RSdz LKdz LMdz CQdz (2,2) L(2,2)dz L(2,3)dz L(N-1,N)dz L(N,N)dz Q canR beS dz givenK by: M M M v 2 CEdz (1,2) (2,2) C E dz z z+dz CQ dz (N,N) (N,N) (N,N) iin1 R dz L dz L dz i iout S K 𝐶M ≈ 120𝑊𝐸 1(2,3) + 1+ aF⁄ 𝜇m (5) N C E dz 2.76𝐸 Rc/2 Rc/2 C(N,N)dz Q (N-1,N) vin v vout C E dz (N,N) z=0 z=L Based on the extracted parameters,C E dz the 50% propagation delay can be calculated by [42]: