Open Access
Add to BibSonomyAbstract
The implementation of power efficient and high throughput chip-to-chip interconnects is necessary to keep pace with the bandwidth demands in high-performance computing platforms. In recent years, considerable effort has been made to optimize inter-chip communications using traditional copper waveguides. Also, optical links are extensively investigated as an alternative technology for fast and efficient data routing. For the first time, we experimentally demonstrate simultaneous microwave and optical high-speed data transmission over metallic waveguides embedded in polymer. The demonstration is significant as it merges two layers of communications onto the same structure towards increased aggregated bandwidth, and energy-efficient data movement.
© 2015 Optical Society of America
1. Introduction
Demands for higher bandwidth in short-reach communications continue to grow, due principally to advances in integrated circuits which continue to become generally faster. For example, in todays computers, the communication between high-speed processors restricts the overall operational speed mainly due to inherent latency issues and energy dissipation [1–4]. To address this issue, multi-core and many-processor architectures are implemented to increase computational throughput. Consequently, this approach drives increased chip-to-chip (i.e., CPU/GPU-to-memory) input/output bandwidth demands. The International Roadmap of Semiconductors (ITRS) forecasts in year 2022 a need for approximately 3500 signal pins for high-end application specific integrated circuit (ASIC) processors having together an off-chip data rate of 230 Tbit/s, which corresponds to a data rate of approximately 66 Gbit/s per pin [3, 5]. In such a system, chip-to-chip interconnects are required to provide high bandwidth as well as high energy efficiency and low latency.
Over the past decade, optical interconnects have begun to replace electrical interconnects in short-reach applications, and are increasingly considered to be viable to address the bandwidth limitation of inter-chip communications as well. The deployment of optics in short-reach interconnects becomes viable if higher bandwidth density and at least comparable power efficiency is achievable (including optoelectronics) compared to electrical counterparts. In this regard, extensive research has been carried out to develop chip-scale CMOS compatible building blocks for optical links such as laser sources [6–8], modulators [9, 10], and photodetectors [11]. Further, the integration of electronics and optics would resolve the issue of limited bandwidth caused by transmission loss, impedance mismatching, crosstalk and electromagnetic interference (EMI) [12]. In this context, plasmonic structures [13–16] can potentially enable efficient and highly integrated optical interconnects for different applications (i.e., on-chip, off-chip, and chip-to-chip). The metallo-dielectric structures used in plasmonics provides a natural structure for guiding both light and electrical signals, which potentially leads to reduced energy consumption for opto-electronic circuitry if low loss waveguides can be implemented.
The surface plasmon polariton (SPP), a collective oscillation involving a charge density wave, is a transverse-magnetic wave propagating along an interface of two media (m1 and m2) at optical frequencies where the frequency-dependent permittivities (ε) are of opposite signs . At optical frequencies, a metal appears as a gas of almost free electrons (i.e., a cold plasma, satisfying and generally well-modelled by the Drude equation) as opposed to microwave frequencies where it appears essentially as a perfect conductor. Simultaneously, the other medium generally exhibits negligible loss . The conditions for SPP guidance are generally satisfied at optical frequencies (from the visible to the IR) by many metals and dielectrics. On a single metal-dielectric interface, SPPs are confined to the interface with fields that decay exponentially into both media. Structures that support SPPs can be thin metal films and strips, metal nanoparticles of different forms and sizes, and holes, slits, gaps, grooves or corrugations on metal films [17].
In a structure comprising a thin metal film of finite width embedded in a homogeneous dielectric, several symmetric and asymmetric supermodes (with respect to the metal film cross-section) are formed. The fundamental symmetric mode experiences lower attenuation due to less overlap with the lossy metal allowing propagation lengths up to centimeters. These supermodes are referred as long-range surface plasmon polaritons (LRSPPs). Improvements in propagation length occur at the expense of reduced confinement of the supermode compared to single-interface SPPs [18]. The fundamental LRSPP has a Gaussian-like modal field distribution [19] and can be excited efficiently via an end-coupling technique.
These features and other properties such as high surface sensitivity, electrical conductivity, and the ease with which the structures can be fabricated, e.g., with polymer, make them interesting for applications in biosensing [20–24], integrated optics [25–34], and optical interconnections [35–38]. In previous demonstrations [20, 34, 39] the ability of a plasmonic structure to simultaneously carry a DC or low-frequency current was exploited to drive the thermo-optic effect in a polymer cladding. Unless designed into an appropriate microwave structure (e.g., [25]), the electrical performance of a plasmonic structure is inherently bandwidth-limited and suitable only for propagating or applying low-frequency (low-speed) electrical signals. This is attributed to the fact that the microwave performance of a transmission line degrades for thicknesses below the skin depth due to high resistance. This has been demonstrated in [40] for a co-planar waveguide (CPW) transmission line.
In this work, we present a novel metal waveguide structure which can simultaneously support LRSPPs at optical frequencies and microwave signals up to at least 40 GHz. We experimentally demonstrate, for the first time, the simultaneous transmission of high bit-rate optical (40 Gbit/s) and electrical (12 Gbit/s) signals over the same structure. No evidence of interference between the signals has been observed. This demonstration leads to a new paradigm for data interconnectivity where high-speed electrical and optical data transmissions are combined.
2. Simulation
Figure 1(a) and 1(b) show the schematic of the designed structure that propagates both optical and electrical signal over millimeters. Figure. 1(c) shows a 20×optical microscope image of a fabricated device, and Fig. 1(d) shows an optical microscope image of a 4.6 mm long die bearing 21 different devices. The structure comprises three gold strips with two 100×100 μm2 electrical pads at both ends. The structure is a co-planar waveguide (CPW) comprised of two thin ground strips (labeled G) and a thick signal strip (labeled S) in the center. The LR-SPPs modes propagate over the two thin ground strips while the electrical microwave signal propagates over the full CPW structure. In order to obtain a structure that provides low propagation loss, its optimization for optical transmission is first considered, and then its optimisation for microwave transmission is carried out.
Fig. 1 (a) and (b): Schematic of the structure of interest, and (c) 20× magnification microscope image of one of the fabricated structures at one end. The pad area is 100×100 μm2 with a 100 μm wide signal strip (Ws), and two 30 μm wide ground strips (Wg) separated by 2 μm gap from optical tapers. (d) Image of a 4.6 mm long fabricated die which includes 21 different microwave-optical transmission lines with straight optical reference waveguides in between.
LR-SPPs mode propagation over the structure is modelled using Lumerical MODE Solutions software to obtain the propagation loss and the mode field distribution. The software meshes the cross-section of the waveguide using finite differences, and then solves Maxwells equations by a sparse matrix technique in order to find the effective index and the mode fields. For the computations, the relative permittivity of the gold strip is assumed to be −116.58 + 11.46i (refractive index, n = 0.53+10.81i) at a wavelength of 1550 nm [41]. To investigate the optimum structure dimensions, the thickness t1 is varied from 20 to 35 nm, and the width Wg is varied from 2 to 30 μm in 1 μm increments. The dielectric selected is Cytop, an amorphous fluoropolymer with a refractive index of 1.3335 at 1550 nm [42], and the claddings are assumed thick enough (100 μm) to accommodate LR-SPP modes.
Figure 2(a) plots the computed propagation loss for different widths and thicknesses. As expected, the propagation loss decreases with decreasing thickness and width [43]. The lowest simulated propagation loss, obtained for a 20 nm thick gold film, corresponds to 0.02 dB/mm. This is achieved when the width is designed to be 2 μm. In Figs. 2(b) and 2(c), the vertical and lateral mode field diameters of the LRSPP mode are shown, respectively. The mode field diameter in this case is defined as the full-width electric field (Ey) spot size at the 1/e point. As shown in Fig. 2(b), the vertical mode field diameter becomes larger with reducing gold thickness and width, eventually reaching the field diameter of the background slab mode supported by the thick Cytop cladding in the absence of the metal strip. The inset shows a near-field image of the electric field (Re(Ey)) distribution of the 5 μm wide and 35 nm thick gold strip. The lateral mode field diameter [Fig. 2(c)], however, behaves differently. It initially gets smaller with reducing strip width from 30 μm, but then increases with further reduction of the strip width. This behaviour is attributed to the fact that the lateral mode field diameter is always larger than the width of the metal strip [44]. For sufficiently wide metal, the lateral mode field diameter becomes as wide as the width of the strip. This results in a linear-like trend for the lateral mode field diameter versus width. But for a narrower metal strip, the lateral field distribution becomes comparable to the field of the background slab mode (similar to the vertical mode field case), resulting in an increase of the lateral field diameter. Hence, there is a trade-off between propagation loss [Fig. 2(a)] and mode field diameter [Figs. 2(b) and 2(c)] in which lower propagation loss comes at the expense of larger mode field distribution.
Fig. 2 Computed (a) propagation loss, (b) vertical and (c) lateral mode field diameter as a function of gold ground strip width (Wg) and thickness (t1) for a single gold strip waveguide in a homogenous medium. The inset in (b) shows a near-field image of the electric field (Re(Ey)) distribution of the 5 μm wide and 35 nm thick gold strip.
In the proposed structure, overlap of the optical modes propagating on the ground strips with the centre signal strip leads to increased optical loss, and thus needs to be controlled. According to our computations (see appendix A), at smaller separation distances (g), the lower the width and thickness of the metal ground strips (Wg, t1), the higher the effect of the signal strip on the propagation loss. The result suggests that in case of separation distances lower than 10 μm, wide ground strips (optical waveguides) are preferable in order to avoid additional optical loss.
For the optimization and investigation of the electrical microwave propagation, computations were carried out using ANSYS HFSS software. A 3D full-wave frequency domain electromagnetic field solver based on the finite element method (FEM) computes the electrical behaviour of structure. In our computations, the optimum dimensions of the structures cross-section are initially found, for which the characteristic impedance (Z0) is close to 50 Ω. Then, the S-parameters of the proposed structure are extracted by de-embedding the cross-section for different lengths and termination with 50 Ω load representing the measurement equipment ports. In these computations, the electrical properties of bulk copper, gold, and silicon are used from the software library. For the Cytop claddings, the relative dielectric constant used is 2 with a dielectric loss tangent of 0.0003 [42]. Figure. 3 gives results for the characteristic impedance of the structure as a function of frequency for different geometries.
As shown in Fig. 3, the characteristic impedance decreases with increasing frequency. Importantly for microwave transmission, the characteristic impedance decreases and approaches 50 Ω by decreasing the separation distance (g) from 15 to 5 μm and by increasing the ground strip width (Wg) from 5 to 30 μm. Moreover, comparing Figs. 3(a)–3(c) to Figs. 3(d)–3(f) shows that the wider the signal strip width the lower the impedance becomes. Furthermore, the S-parameters were computed as shown in Fig. 4 for different structures that have been fabricated and characterized. The structures have a 50 μm wide signal strip and 30 μm wide ground strips. As illustrated, the wider the gap between the strips, the higher the electrical transmission (lower loss). Also, it can be seen that the performance of the structure in terms of loss and bandwidth decreases by increasing the length.
Fig. 3 (a)–(f) Computed electrical characteristic impedance Z0 of the proposed structure as a function of frequency for different ground (Wg) and signal (Ws) strip widths, and separation (g).
Fig. 4 Computed electrical response of structures comprised of a 50 μm wide signal strip and 30 μm wide ground strips for different separation sizes (g) and lengths.
Based on the computational results, the available fabrication capabilities, and the available measurement equipment, structures were designed to enable simultaneous optical and electrical transmission. In the fabricated structures, the middle strip (electrical signal line) consists of a 1 μm thick copper (t2) film sandwiched between a 25 nm thick gold (t1) layer at the bottom and a 10 nm gold (t3) layer on top. The top gold layer serves as a protective layer to passivate the copper in order to prevent surface oxidization and degradation of copper conductivity. In the signal strip, mainly copper is used instead of gold to reduce fabricaton costs. The side strips (electrical ground lines) are thin gold films (25 nm thick) supporting LRSPP modes at optical frequencies. As shown in Fig. 5(a), the microwave mode is excited from the top via RF electrical probes configured as ground-signal-ground (GSG) with a pitch of 150 μm. The LRSPP mode is excited via an end-fiber coupled optical single-mode fiber. To maximize the coupling efficiency between the structure and the optical fiber, on-chip linear tapers are designed from 8 μm (Wt) to the width of the ground strips over a length of 600 μm [see Fig. 1(a)]. Also, 200 μm long straight segments (not shown in Fig. 1) are added to the end of the taper input to compensate for possible dicing inaccuracies. A small gap of 2 μm is inserted between the taper and the ground strip waveguides for electrical isolation. The total optical path is approximately 1.6 mm longer than that of the electrical transmission line because of the tapers and gaps. The width of the middle strip waveguide (Ws) is set to 50 μm, and the side strip waveguides have width (Wg) of 30 μm. The separation distance (g) is 15 μm. The electrical contact pads are 100 μm by 100 μm in area. The connections between the electrical pads to the electrical ground lines consist of five thin gold (25 nm thick) strips that are 5 μm wide and of length (Lconn) of 10 μm.
Fig. 5 (a) Illustration of the experimental set-up for the simultaneous excitation of LR-SPP and microwave modes. (b) Measured insertion loss versus waveguide length for 8 and 30 μm wide, 25 nm thick gold strip waveguides. The inset shows a far-field image of the LR-SPP mode output.
3. Fabrication
The structure used in the experimental demonstration is fabricated with the following fabrication processing steps. First, an 8 μm thick layer of polymer (Cytop) is coated on a silicon wafer using multiple spin coatings and a solvent evaporation process to create the bottom cladding. The wafer is then patterned using a bilayer re-entrant photo-lithography process. A 25 nm thick layer of gold (Au) is then deposited using E-beam evaporation followed by wet solvent stripping to reveal the metal structures. The structures are then covered by a 7 μm thick layer of polymer (Cytop) for the top cladding. The second and third metal layers of the central signal strip waveguide and contact pads are patterned using a single-layer photo-lithography process. The patterns are then etched through the top Cytop cladding down to the metal level, followed by the deposition of 1 μm of copper and 10 nm of gold, with the excess metal lifted off. Finally, the wafer is diced into several dies of different lengths.
4. Measurements
4.1. Optical insertion loss measurement
Optical transmission measurements were obtained at a wavelength of 1550 nm, although longer wavelengths (up to 1675 nm) compatible with communication bands can be supported with lower propagation losses. In Fig. 5(b), the measured optical insertion loss at 1550 nm for two different waveguide widths (8 μm straight and 30 μm with two tapers) at three different lengths are shown. According to our computations (not shown), a gold metal strip 8 μm wide and 25 nm thick has the highest coupling efficiency to a standard single mode fiber (94%). Hence, an 8 μm wide straight waveguide is tested as a reference sample. The inset shows the far-field LRSPP mode measured using an infra-red camera. The background light is attributed primarily to uncoupled input light leaking through the polymer cladding surrounding the metallic waveguide. The slope of the curve is the attenuation and the intercept with the vertical axis is the input coupling loss. Thus, the measured attenuation for the 8 μm long straight waveguide is 1.71 dB/mm with a coupling loss of 2.42 dB/facet. Our computations predict an attenuation of 1.48 dB/mm and 0.25 dB of coupling loss per facet. Also, the measured attenuation for a 30 μm wide strip waveguide is 1.88 dB/mm with a coupling loss of 3.9 dB/facet, compared to computations of approximately 2 dB/mm of attenuation (see appendix A, Fig. 8(b), t1 of 25 nm and g of 15 μm), and a coupling loss of 0.25 dB/facet. Therefore, for 30 μm wide strips the expected total insertion loss for a length of 4.6 mm (3 mm waveguide with 1.6 mm tapers) is 9.25 dB. The difference between the measured value (12.6 dB, Fig. 5(b), first red data point) and modeling is attributed to the reason listed below, as well as fabrication and measurement errors. In the modeling, only the metal absorption based on the bulk properties of gold is considered as the sole source of loss. However, metal and edge roughness, as well as radiation losses in the tapers may contribute to the total insertion loss. The difference between predicted coupling loss and measured values is due to reflections from fiber/air and air/chip interfaces, and also to the non-ideal optical facets. Also, it must be mentioned that the effect of the contact pads were not considered in the optical modeling; based on the computations shown in Fig. 8(b), when a 30 μm wide and 25 nm thick LRSPP waveguide is at a 10 μm distance from a thick metal structure (signal strip or contact pads), the excess loss on the LRSPP is negligible (~dB). Additionally, comparison between measured coupling losses in straight waveguides and tapered waveguides suggests that the designed tapers at the input and output are not perfectly adiabatic. Also, in the measurement setup, background radiation from uncoupled input light to the LR-SPP cannot be completely eliminated.
4.2. Electrical S-parameters measurement
For electrical transmission characterization, an electrical input power of 10 dBm was provided by a 40 GHz vector network analyzer which was used to excite and capture the quasi-TEM mode propagating along the fabricated CPW. The waveguide was excited using electrical probes with a GSG configuration. The characterized waveguide was 3 mm long, comprised of a 50 μm wide signal strip with 30 μm wide ground strips and a separation of 15 μm. The overall width of the structure matches the 150 μm pitch of the GSG probes.
Figure 6(a) shows electrical transmission and reflection (S21 and S11) of the structure. The drop in transmission at frequencies below 10 GHz is attributed to the five short connections running from each electrical pad to the ground strips. These thin and short connections are of higher resistance compared to the electrical pads at low frequencies. As the frequency increases, the skin depth in the pads becomes smaller resulting in higher impedance, thus, less impedance mismatch with the connections. Accounting for the drop at low frequency, the 3-dB bandwidth of the structure is beyond 40 GHz. Also, the reflection response (S11) exhibits a similar trend and is in general agreement with modelling [Fig. 4]. In Fig. 6(b), the measured electrical transmission (S21) for waveguides of different lengths is given. The 3-dB bandwidth logically decreases to approximately 28, and 20 GHz as the length of the structure increases to 5 and 8 mm, respectively.
4.3. Optical and electrical modulated data transmission
Finally, we demonstrate the feasibility of simultaneous transmission of optical and electrical digital signal through the proposed structure without interference between the two signals for 3 mm length. Figure. 7 shows BER measurements of 40 Gbit/s optical and 12 Gbit/s electrical for simultaneous and independent propagation. The transmission bit-rate is limited by the experimental setup. As observed, the electrical signal does not interfere with the optical signal, as confirmed by measuring the BER of the optical signal in the presence and absence of the electrical signal (simultaneous and independent transmissions); the simultaneous and independent transmissions overlap perfectly. The inset shows the transmitted optical eye at 40 Gbit/s when the electrical data is present at 12 Gbit/s. Also, we observe transmission of 12 Gbit/s electrical data with low BER (<10−9) in the presence of 40 Gbit/s optical signals and that no interference between the two co-propagating signals is noted. The inset shows the electrical eye in presence of optical data. The solid lines in Fig. 7 are exponential fit curves to measured values. Also, it must be mentioned that the sensitivity of our BER tester in our experimental setup is 10 mVpp (roughly −36 dBm).
Fig. 7 BER measurements of simultaneous and independent 40 Gbit/s optical and 12 Gbit/s electrical signals. The insets show captured eye diagrams. The solid lines are exponential fit curves to measured values.
These results confirm the possibility of transmitting optical and microwave signals simultaneously on the same structure. Based on the measured RF bandwidth for a 3 mm long structure [Fig. 6(a)], we assume that a 50 Gbit/s electrical data rate is possible. In the case of the optical channels the bit-rate is limited by the speed of the optoelectronics at each end of the link. Assuming also a data rate of 50 Gbit/s per optical channel, the aggregate bandwidth would be 150 Gbit/s because there are two optical (LR-SPP) channels (one on each ground strip, 100 Gbit/s) and one microwave channel over the CPW (50 Gbit/s). Now, considering that the lateral footprint of the tested structure, excluding the contact pads and any area required to fit the optical and electrical components, is 140 μm, yields a bandwidth density of approximately 1 Gbit/s/μm. This density compares favorably to recent advances in electronic transmission line designs [45, 46] where full-swing RC lines produce a bandwidth density of the same order (1 Gbit/s/μm).
Generally, electrical interconnects become difficult to design as data rates begin to exceed about 10 Gbit/s, due to frequency dependent losses, crosstalk and frequency resonance effects [47]. The proposed structure in this work is capable of transmitting electrical signals up to data rates which more complex design, signaling and equalization approaches are not required. The aggregate bandwidth can be further increased by co-propagating optical and electrical signals. Moreover, The bandwidth density of the proposed solution can be improved significantly, in two ways: Firstly, based on computations, we find that the lateral size of the structure can be reduced to a few tens of microns if the requirement for impedance matching is relaxed (see appendix B), which can be the case for monolithic integration of the structure with ancillary electronics. Secondly, the optical bandwidth can be increased by increasing the speed of the optoelectronics (channel data rate) and, in principle, by wavelength division multiplexing.
5. Conclusion
In conclusion, we have proposed a novel metallic waveguide structure for the simultaneous transmission of optical and microwave data signals, and demonstrated this capability by the simultaneous error-free transmission (BER<10−9) of optical and electrical data at 40 Gbit/s and 12 Gbit/s, respectively, over a 3 mm long structure without any observable interference. The measured channel losses for the optical and electrical signals are approximately 12.6 (at 1550 nm) and 3.4 dB (at 12 GHz), respectively. The overall bandwidth density of the tested structure is approximately 1Gbit/s/μm which is already comparable to existing state-of-the-art electronic interconnects. Additionally, the projected bandwidth density requirement of ASIC chips in 2022 is estimated to 0.7 Gbit/s/μm assuming 95 μm pitch in a flip-chip ball grid array packaging technology [5]. The structure is capable of far greater bandwidth densities via optimisation of its design, and by increasing the speed of the optoelectronics or adding optical carriers in a wavelength multiplexing scheme. The structure is therefore capable of addressing the bandwidth density challenge in present and future chip-to-chip communications, leading to increased module bandwidth for high-performance computing platforms among other applications.
Appendix A: Optical loss computations in the presence of a thick signal strip
In the proposed structure, overlap of the optical modes propagating in the ground strips with the middle signal strip leads to increase optical loss which needs to be investigated. In Fig. 8, the optical loss of a thin strip of two different widths (Wg=5 and 30 μm) is computed for various thicknesses (t1=20, 25, and 30 nm). The effect of the 50 μm wide (Ws), 500 nm thick (t2) middle signal strip is studied for various separation distances (g). In these computations, the top cladding of 15 μm of Cytop and the thick silicon substrate are considered.
Fig. 8 Computed propagation loss versus separation distance (g) and thickness (t1), for ground strips (a) 5 μm and (b) 30 μm wide (Wg). The signal strip width (Ws) and thickness (t2) are 50 μm and 500 nm, respectively. Cytop thickness is assumed to be 15 μm on thick silicon substrate. Dashed lines show propagation loss when Cytop thickness is 100 μm.
As illustrated in Fig. 8 in presence of signal strip, the smaller the separation distance (g < 10 μm), the higher the propagation loss. Also, comparing Figs. 8(a) and 8(b) shows that increase in loss at smaller gap sizes is more significant for 5 μm wide ground strips (Wg) compared to 30 μm strips which is attributed to high mode overlap with signal strip. The dashed lines show propagation loss of their respective strip embedded in a thick homogenous Cytop cladding (100 μm) in presence of signal strip. The difference between dashed and solid lines shows the effect of 15 μm thick Cytop and silicon substrate on optical propagation loss.
Appendix B: Electrical simulation for narrow signal strip
In the fabricated structure, width of the signal strip was chosen such that the reflection is minimized when the line is terminated with 50 Ω load (experimental setup). However, this limitation results in a large width of the signal strip (50 μm), thus a wide footprint and a lower bandwidth density. Computations show that if this restriction is relaxed, a narrow signal strip can be used. Figure. 9 shows results for an identical 3 mm long line as the tested structure except with 1 μm wide signal strip and gap size (g) of 4 μm.
As illustrated, when the proposed structure is terminated with 50 Ω loads [Fig. 9(b)], significant reflection and poor transmission occurs, as opposed to the case when ports are terminated with 90 Ω loads as shown in Fig. 9(c).
References and links
1. R. G. Beausoleil, “Large-scale integrated photonics for high-performance interconnects,” ACM J. Emerg. Technol. Comput. Syst. 7(2), 6 (2011). [CrossRef]
2. F. Y. Liu, D. Patil, J. Lexau, P. Amberg, M. Dayringer, J. Gainsley, H. F. Moghadam, X. Zheng, J. E. Cunningham, A. V. Krishnamoorthy, E. Alon, and R. Ho, “10-Gbps, 5.3-mW optical transmitter and receiver circuits in 40-nm CMOS,” IEEE J. Solid-St. Circ. 47(9), 2049–2067 (2012). [CrossRef]
3. D. A. B. Miller, “Device requirements for optical interconnects to silicon chips,” Proc. IEEE 97(7), 1166–1185 (2009). [CrossRef]
4. S. Borkar and A. A. Chien, “The future of microprocessors,” Commun. ACM 54(5), 67–77 (2011). [CrossRef]
5. International Roadmap of Semiconductors (ITRS), ITRS 2012 Edition, 2012 Overall Roadmap Technology Characteristics (ORTC) Tables, Performance and Packaged Chips Trends, http://www.itrs.net.
6. P. Michler, A. Kiraz, C. Becher, W. V. Schoenfeld, P. M. Petroff, Lidong Zhang, E. Hu, and A. Imamoglu, “A quantum dot single-photon turnstile device,” Science 290(5500), 2282–2285 (2000). [CrossRef] [PubMed]
7. A. J. Shields, “Semiconductor quantum light sources,” Nat. Photonics 1(4), 215–223 (2007). [CrossRef]
8. K. Iga, “Surface-emitting laser-its birth and generation of new optoelectronics field,” IEEE J. Sel. Top. Quantum Electron. 6(6), 1201–1215 (2000). [CrossRef]
9. Q. Xu, B. Schmidt, S. Pradhan, and M. Lipson, “Micrometre-scale silicon electro-optic modulator,” Nature 435(7040), 325–327 (2005). [CrossRef] [PubMed]
10. G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, “Silicon optical modulators,” Nat. Photonics 4(8), 518–526 (2010). [CrossRef]
11. J. Michel, J. Liu, and L. C. Kimerling, “High-performance Ge-on-Si photodetectors,” Nat. Photonics 4(8), 527–534 (2010). [CrossRef]
12. Y. Takagi, A. Suzuki, T. Horio, T. Ohno, T. Kojima, T. Takada, S. Iio, K. Obayashi, and M. Okuyama, “Low-loss chip-to-chip optical interconnection using multichip optoelectronic package with 40-Gb/s optical I/O for computer applications,” J. Lightwave Technol. 28(20), 2956–2963 (2010). [CrossRef]
13. H. A. Atwater, “The promise of plasmonics,” Sci. Am. 296(4), 56–62 (2007). [CrossRef] [PubMed]
14. M. L. Brongersma and V. M. Shalaev, “Applied Physics: The Case For Plasmonics,” Science 328, 440–441 (2010). [CrossRef] [PubMed]
15. W. L. Barnes, A. Dereux, and T. W. Ebbesen, “Surface plasmon subwavelength optics,” Nature 424(6950), 824–830 (2003). [CrossRef] [PubMed]
16. J. J. Ju, S. Park, M. S. Kim, J. T. Kim, S. K. Park, J. M. Lee, J. S. Choe, and M. H. Lee, “Data Transfer for Optical Interconnects using Long Range-SPP Transmission Lines,” in Optical Fibre Communication Conference, paper OW3E-6 (Los Angeles, CA, USA, 2012).
17. S. A. Maier, Plasmonics: Fundamentals and Applications (Springer Science and Business Media, 2007).
18. P. Berini, “Long-range surface plasmon polaritons,” Adv. Opt. Photonics 1(3), 484–588 (2009). [CrossRef]
19. R. Charbonneau, C. Scales, I. Breukelaar, S. Fafard, N. Lahoud, G. Mattiussi, and P. Berini, “Passive integrated optics elements based on long-range surface plasmon polaritons,” J. Lightwave Technol. 24(1), 477–494 (2006). [CrossRef]
20. T. Zhang, G. Qian, Y. Y. Wang, X. J. Xue, F. Shan, R. Z. Li, J. Y. Wu, and X. Y. Zhang, “Integrated optical gyroscope using active Long-range surface plasmon-polariton waveguide resonator,” Sci. Rep. 4, 1–6 (2014).
21. J. Homola, “Surface plasmon resonance sensors for detection of chemical and biological species,” Chem. Rev. 108(2), 462–493 (2008). [CrossRef] [PubMed]
22. O. Krupin, C. Wang, and P. Berini, “Selective capture of human red blood cells based on blood group using long-range surface plasmon waveguides,” Biosens. Bioelectron. 53, 117–122 (2014). [CrossRef]
23. L. O. Diniz Jr., E. Marega, F. D. Nunes, and B. H. V. Borges, “A long-range surface plasmon-polariton waveguide ring resonator as a platform for (bio) sensor applications,” J. Opt. 13(11), 115001 (2011). [CrossRef]
24. Y. H. Joo, S. H. Song, and R. Magnusson, “Long-range surface plasmon-polariton waveguide sensors with a Bragg grating in the asymmetric double-electrode structure,” Opt. Express 17(13), 10606–10611 (2009). [CrossRef] [PubMed]
25. A. Melikyan, L. Alloatti, A. Muslija, D. Hillerkuss, P. C. Schindler, J. Li, R. Palmer, D. Korn, S. Muehlbrandt, D. Van Thourhout, B. Chen, R. Dinu, M. Sommer, C. Koos, M. Kohl, W. Freude, and J. Leuthold, “High-speed plasmonic phase modulators,” Nat. Photonics 8(3), 229–233 (2014). [CrossRef]
26. P. Berini and I. De Leon, “Surface plasmon-polariton amplifiers and lasers,” Nat. Photonics 6(1), 16–24 (2012). [CrossRef]
27. P. Neutens, P. Van Dorpe, I. De Vlaminck, L. Lagae, and G. Borghs, “Electrical detection of confined gap plasmons in metalinsulatormetal waveguides,” Nat. Photonics 3(5), 283–286 (2009). [CrossRef]
28. R. F. Oulton, V. J. Sorger, D. A. Genov, D. F. P. Pile, and X. Zhang, “A hybrid plasmonic waveguide for sub-wavelength confinement and long-range propagation,” Nat. Photonics 2(8), 496–500 (2008). [CrossRef]
29. A. Akbari, R. N. Tait, and P. Berini, “Surface plasmon waveguide Schottky detector,” Opt. Express 18(8), 8505–8514 (2010). [CrossRef] [PubMed]
30. L. Chen, X. Li, G. Wang, W. Li, S. Chen, L. Xiao, and D. Gao, “A silicon-based 3-D hybrid long-range plasmonic waveguide for nanophotonic integration,” J. Lightwave Technol. 30(1), 163–168 (2012). [CrossRef]
31. J. T. Kim and S. Park, “Vertical polarization beam splitter using a hybrid long-range surface plasmon polariton waveguide,” J. Opt. 16(2), 025501 (2014). [CrossRef]
32. J. Lee and M. A. Belkin, “Widely tunable thermo-optic plasmonic bandpass filter,” Appl. Phys. Lett. 103(18), 181115 (2013). [CrossRef]
33. D. C. Zografopoulos, R. Beccherelli, A. C. Tasolamprou, and E. E. Kriezis, “Liquidcrystal tunable waveguides for integrated plasmonic components,” Photon. Nanostruct.: Fundam. Appl. 11(1), 73–84 (2013). [CrossRef]
34. G. Gagnon, N. Lahoud, G. A. Mattiussi, and P. Berini, “Thermally activated variable attenuation of long-range surface plasmon-polariton waves,” J. Lightwave Technol. 24(11), 4391–4402 (2006). [CrossRef]
35. N. Kinsey, M. Ferrera, G. V. Naik, V. E. Babicheva, V. M. Shalaev, and A. Boltasseva, “Experimental demonstration of titanium nitride plasmonic interconnects,” Opt. Express 22(10), 12238–12247 (2014). [CrossRef] [PubMed]
36. B. Banan, M. S. Hai, E. Lisicka-Skrzek, P. Berini, and O. Liboiron-Ladouceur, “Multichannel Transmission Through a Gold Strip Plasmonic Waveguide Embedded in Cytop,” IEEE Photon. J. 5(3), 2201811 (2013). [CrossRef]
37. J. T. Kim and S. Y. Choi, “Graphene-based plasmonic waveguides for photonic integrated circuits,” Opt. Express 19(24), 24557–24562 (2011). [CrossRef] [PubMed]
38. J. T. Kim, J. J. Ju, S. Park, M. S. Kim, S. K. Park, and M. H. Lee, “Chip-to-chip optical interconnect using gold long-range surface plasmon polariton waveguides,” Opt. Express 16(17), 13133–13138 (2008). [CrossRef] [PubMed]
39. G. Giannoulis, D. Kalavrouziotis, D. Apostolopoulos, S. Papaioannou, A. Kumar, S. Bozhevolnyi, L. Markey, K. Hassan, J. C. Weeber, A. Dereux, M. Baus, M. Karl, T. Tekin, O. Tsilipakos, A. K. Pitilakis, E. E. Kriezis, K. Vyrsokinos, H. Avramopoulos, and N. Pleros, “Data transmission and thermo-optic tuning performance of dielectric-loaded plasmonic structures hetero-integrated on a silicon chip,” IEEE Photon. Technol. Lett. 24(5), 374–376 (2012). [CrossRef]
40. M. Ramazani, H. Miladi, M. Shahabadi, and S. Mohajerzadeh, “Loss measurement of aluminum thin-film coplanar waveguide (CPW) lines at microwave frequencies,” IEEE Trans. Electron Dev. 57(8), 2037–2040 (2010). [CrossRef]
41. P. B. Johnson and R. W. Christy, “Optical constants of the noble metals,” Phys. Rev. B 6(12), 4370–4379 (1972). [CrossRef]
42. Asahi Glass Company, Technical Information Cytop, distributed by Bellex International Corporation, http://www.bellexinternational.com.
43. P. Berini, “Plasmon-polariton waves guided by thin lossy metal films of finite width: Bound modes of symmetric structures,” Phys. Rev. B 61(15), 10484–10503 (2000). [CrossRef]
44. I. Breukelaar, R. Charbonneau, and P. Berini, “Long-range surface plasmon-polariton mode cutoff and radiation in embedded strip waveguides,” J. Appl. Phys. 100(4), 043104 (2006). [CrossRef]
45. J. W. Poulton, W. J. Dally, X. Chen, J. G. Eyles, T. H. Greer, S. G. Tell, and C. T. Gray, “A 0.54 pJ/b 20Gb/s ground-referenced single-ended short-haul serial link in 28nm CMOS for advanced packaging applications,” In IEEE International Solid-State Circuits Conference Digest of Technical Papers, 404–405 (San Francisco, CA, USA, 2012)
46. T. O. Dickson, Y. Liu, S. V. Rylov, B. Dang, C. K. Tsang, P. S. Andry, J. F. Bulzacchelli, H. A. Ainspan, X. Gu, L. Turlapati, M. P. Beakes, B. D. Parker, J. U. Knickerbocker, and D. J. Friedman, “An 8x 10-Gb/s source-synchronous I/O system based on high-density silicon carrier interconnects,” IEEE J. Solid-St. Circ. 47(4), 884–896 (2012). [CrossRef]
47. M.A. Taubenblatt, “Optical interconnects for high-performance computing,” J. Lightwave Technol. 30(4), 448–457 (2012). [CrossRef]
