## Abstract

Electronics circuits keep shrinking in dimensions, as requested by Moore’s law. In contrast, photonic waveguides and circuit elements still have lateral dimensions on the order of the wavelength. A key to make photonics have a microelectronics-like development is a drastic reduction of size. To achieve this, we need a low-loss nanoscale waveguide with a drastically reduced mode area and an ultra-high effective refractive index. For this purpose, we propose here several low-loss waveguide structures based on graphene nano-ribbons. An extremely small mode area (~10^{−7}*λ*_{0}^{2}, one order smaller than the smallest mode area of any waveguide that has ever been reported in the literature; here *λ*_{0} is the operating wavelength in vacuum) and an extremely large effective refractive index (several hundreds) are achieved. As a device example, a nano-ring cavity of ultra-small size (with a diameter of ~10^{−2}*λ*_{0}) is designed. Our study paves the way for future VLSI (very-large-scale integration) optoelectronics.

© 2013 Optical Society of America

## 1. Introduction

Compared with microelectronics, today’s photonic waveguides and circuit elements are still too large in size. To make photonics have a microelectronics-like development, we need a low-loss nano-waveguide with a drastically reduced mode area (for small size in the transverse direction) and an ultra-high effective refractive index (so that the effective wavelength is small and consequently the device length can be short along the propagation direction). Graphene as a one-atom-thick platform for infrared and THz metamaterials [1] can play an important role in optical science and engineering. Since graphene was experimentally made for the first time from graphite in 2004, this newly available material, which is a single layer of carbon atoms, has attracted much attention due to its unique properties [2, 3]. It can serve as a platform for metamaterials and can support surface-plasmon polaritons (SPPs) at infrared or THz frequencies [4–9]. Some noble metals, such as silver and gold, are widely regarded as the best available plasmonic materials [10]. However, devices fabricated with these metals suffer from large Ohmic losses and cannot be tuned once the geometry of the structure is fixed. Compared with noble metals, graphene can be tuned flexibly via electrical gating or chemical doping [4]. It can support both TM and TE modes [4] and has strongly enhanced light–matter interactions [7]. Meanwhile, graphene exhibits relatively low Ohmic losses. The graphene layer has a single-atom thickness, which is much thinner than any metallic thin film that has ever been fabricated. Due to the above properties, graphene has been studied in many areas [11–15]. The SPP mode supported by graphene ribbons has already been studied, but the ribbon width is still large (on the order of micrometer) [9]. In this paper, the waveguide mode properties of graphene nano-ribbon (the width is smaller than 100 nm) is investigated, and low-loss plasmonic waveguides of ultra-small mode area are achieved.

## 2. The properties of graphene ribbons

As a unique property of graphene, the complex conductivity of a graphene layer can be tuned flexibly by controlling the chemical potential (via the applied electric field), chemical doping or ground plane evenness [3]. In this way graphene can behave as either a metal or a dielectric and support both TM and TE waveguide modes. The graphene considered throughout the present paper has a chemical potential *μ*_{c} = 0.15 eV, *T* = 3 K and scattering rate *Г* = 0.43 eV to achieve a conductivity value *σ _{g}* = 0.0009 + i0.0765 mS, which is capable of supporting TM SPP surface mode [5, 6, 8, 13] at 30 THz. The conductivity value

*σ*is derived using the Kubo formula model [4, 14]. The dispersion relations of TM SPP surface mode is expressed as

_{g}*β*

^{2}=

*k*

_{0}^{2}[1-(2/

*η*)

_{0}σ_{g}^{2}], where

*β*and

*k*are the wave numbers in the wave guide and the free space, respectively, and

_{0}*η*is the intrinsic impedance of free space. Thus, we get

_{0}*β*= (69.34 + i0.71)

*k*for an infinitely extended graphene sheet.

_{0}As the temperature increases, from the Kubo formula one sees that Re(*σ _{g}*) will increases. For example, Re(

*σ*) at 200 K goes up by 30% compared to that at 3 K. A larger Re(σ

_{g}_{g}) implies a smaller propagation length of graphene waveguide mode. At

*T*= 200 K, the propagation constant

*β*= (75.28 + i2.46)

*k*, (indicating that Im(

_{0}*β*) is three times large as compared to that at 3 K).

As a 2D material, graphene with infinitely-small thickness cannot be directly incorporated into the numerical simulation using conventional electromagnetic software such as COMSOL or CST. This difficulty can be overcome by treating the graphene layer as an ultra-thin material with an effective thickness *Δ* and effective bulk conductivity [4] *ε _{g,eq}* = -

*σ*/

_{g,i}*ω△*+

*ε*+ i

_{0}*σ*/

_{g,r}*ω△*, where

*σ*and

_{g,r}*σ*are the real part and imaginary part of

_{g,i}*σ*, respectively. Because the surface conductivity

_{g}*σ*remains constant when the effective thickness

_{g}*Δ*varies, the mode properties, e.g.,

*n*, are actually insensitive to the specific value of

_{eff}*Δ*, as long as

*Δ*is small enough (

*Δ*< 1 nm). In all our simulations, the thickness of graphene

*Δ*is assumed to be 0.4nm (very close to the typical value of 0.34 nm measured using the interlayer space of graphite [15]) to save the simulation time.

The waveguide mode of a single freestanding graphene ribbon is studied first. The 3D propagation of a guided wave on a typical graphene ribbon with width *w* = 40 nm is first shown in Fig. 1(a) to give the readers an intuitive idea on the mode confinement in both the lateral direction and the propagation direction. A discrete port is placed in front of the graphene ribbon to excite the guided mode. Although the length along the propagation direction is only 350 nm, more than 4 harmonic oscillations are supported, indicating the existence of waveguide mode with an ultra-large wave number. Meanwhile, the waveguide mode is also tightly confined in the lateral direction, as can be seen from the rapid decay of the optical field in the surrounding air. Since an ultra-small waveguide with some tightly confined energy is highly desirable for nanoscale photonic integration, it is worthwhile to study the energy profile of the waveguide mode. The electromagnetic energy density *W*(*x, y*) of the corresponding guided mode in Fig. 1(a) is shown in Fig. 1(b). Here *W*(*x, y*) is calculated by *W* = 0.5(*ε _{eff}ε_{0}*|

*E*|

^{2}+

*μ*|

_{0}*H*|

^{2}), where

*ε*is the effective permittivity and

_{eff}*ε*= ∂ (

_{eff}*ωε*)⁄∂

*ω*. Based on the dispersion model, our calculation gives

*ε*= 169.45 when the thickness

_{eff, g}*Δ*of the graphene layer is 0.4nm and

*ε*= 1. In Fig. 1(b), one sees that the electromagnetic energy is tightly confined inside the graphene due to the large magnitude of

_{eff,air}*ε*. It is also noted that a significant amount of energy is concentrated on the edges of the graphene ribbon. In our study, the microscopic details at the edges of the graphene ribbons are not taken into consideration, similar to the treatment in some other group’s work [9].

_{eff,g}The evolution of the waveguide mode in a graphene ribbon of finite width (micron dimension) has already been studied in [9]. It is found that an edge mode will appear due to the presence of the graphene edge. Here we want to further investigate the properties of graphene ribbon waveguides when the ribbon width is reduced down to nanoscale. It is expected that the ribbon width will greatly influence the behavior of the edge mode due to the strong interaction of ribbon edges.

The three waveguide modes supported by a freestanding graphene ribbon with width *w*< 270 nm is plotted in blue curves in Fig. 2(a). To find out the physical origin of the three modes, the edge plasmon mode supported by a semi-infinite graphene ribbon (EGSP) and the surface plasmon mode supported by an infinitely extended two-dimensional graphene (2DGSP) are also plotted in Fig. 2(a) in aquamarine blue and in green, respectively. Now it is clear that mode #1 and mode #2 originate from the coupling of EGSP and mode #3 originate from the evolution of 2DGSP. The effective index of mode #3 would decrease as width *w* gets smaller and finally becomes cutoff, leaving only EGSP modes (mode #1 and mode #2). The mode #1 and mode #2 arise from the symmetric and anti-symmetric hybridization of EGSP mode, as can be clearly seen from Fig. 2(b). The *E _{y}* component of mode #1 is symmetric with respect to the y axis, and

*E*component of mode #2 is anti-symmetric with respect to the y axis. The symmetric hybridized mode #1 has a higher refractive index than the mode #2. As the width further shrinks, these two modes will continue splitting: the effective index of mode #1 will increase and the effective index of mode #2 will decrease. Finally, the mode #2 is cutoff when

_{y}*w*< 50 nm. Eventually, the graphene ribbon can operate at single-mode region (red shaded region in Fig. 2(a)) if its width < 50 nm, leaving only the symmetric hybridized EGSP mode with very large refractive index. In contrast to the mode #2 and mode #3, the confinement of this remaining mode #1 will become tighter and tighter when the ribbon width further shrinks. The effective index of a graphene nano-ribbon is 117.5 when the ribbons width is 20 nm. It goes up by 30% compared to the

*n*of EGSP, and goes up by 70% compared to the

_{eff}*n*of 2DGSP. The increase of effective index of mode #1 is caused by the coupling of the edge modes, which is similar to the phenomenon occurring in metal nanowires [16].

_{eff}The waveguide mode area is defined as *A _{eff}* = ∫∫

*W(x,y)dxdy*/

*W*, where

_{max}*W*is the maximum of the energy density of the whole waveguide cross section. The calculated mode area is displayed in Fig. 2(c) as a function of ribbon width. It decreases considerably as the width gets small. That is because the optical energy with the graphene ribbon dominates the whole energy due to the tight confinement, so a waveguide with smaller width can lead to a smaller mode area. In our calculation, an extremely small mode area of 1.3 × 10

_{max}^{−7}

*λ*[Fig. 3] is obtained at

_{0}^{2}*w*= 20 nm, which is the smallest among all the reported subwavelength waveguide (to the best of our knowledge). This mode area is much smaller than the smallest mode area that can be achieved using a thin metallic film. For example, a thin gold film with thickness

*t*= 4.6 nm and width

*w*= 20 nm can support a well confined surface plasmon mode. However, the confinement is quite poor compared with a graphene nano-ribbon. At

*λ*

_{0}= 1 μm, the effective mode index is Re(

*n*) = 4.487, and the mode area is calculated to be 1.5 × 10

_{eff}^{−4}

*λ*

_{0}

*, which is three orders larger than its graphene counterpart. Of course, the mode area supported by the gold film can further shrink by reducing further the film thickness. However, it is very difficult to fabricate a smooth metallic film with thickness*

^{2}*t*< 3 nm experimentally.

By placing the graphene nano-ribbon on a silicon substrate, the effective index of the waveguide mode can further increase (and the guided wavelength can be reduced further). However, the propagation loss also increases at the same time. A tight mode confinement is usually accompanied by a large propagation loss [17]. Here we propose to reduce the propagation loss by buffering the graphene ribbon with a silica layer, as shown in Fig. 3(a). *h _{SiO2}* is the height of the SiO

_{2}layer, and

*h*is the height of the Si layer. For this structure, we choose

_{Si}*w*= 20 nm,

*h*= 20 nm,

_{Si}*ε*= 11.9, and

_{si}*ε*= 2.09. If there is no SiO

_{SiO2}_{2}, the effective index is

*n*= 466.3 + 4.7i. Accordingly, the guided wavelength

_{eff}*λ*/

_{spp}= λ_{0}*Re(n*= 21.4 nm. The propagation length

_{eff})*L*defined as 1/Im(

_{m}*β*), where Im(

*β*) = Im(

*n*)

_{eff}*k*. Thus, the propagation length is 338 nm, and the figure of merit (FOM) [which is defined as the ratio of Re(

_{0}*n*) to Im(

_{eff}*n*) [18]] of this waveguide mode is 99.2. When a thin layer of SiO

_{eff}_{2}exists, a strong electric field is induced within the SiO

_{2}layer due to the high index contrast between Si and SiO

_{2}(as required by the continuity of the normal displacement field components). Consequently, the percentage of the optical energy inside the graphene layer decreases, which leads to a reduced propagation loss since the optical loss is entirely due to the damping inside the graphene. Taking

*h*= 5 nm as an example, we have

_{SiO2}*n*= 180.9 + 1.3i,

_{eff}*L*= 1256 nm, and

_{m}*λ*= 55.3 nm. Compared with the case when the SiO

_{spp}_{2}layer is absent, significant loss reduction is achieved. Although the Im(

*n*) of our slot structure is larger than that of the freestanding graphene ribbon (indicating short propagation length), Re(

_{eff}*n*) is also greatly enhanced due to the presence of high-index substrate. It turns out that our slot waveguide has a larger FOM compared to that of the freestanding case. Si has a high index and thus will pull the light away from the graphene to the SiO

_{eff}_{2}buffer layer. Since more energy is located in the region of the low refractive index SiO

_{2}layer, the effective index of the waveguide mode Re(

*n*) decreases at the same time. A good effect is that the FOM of such a graphene ribbon waveguide increases to FOM = 142.9 (much larger than that for the structure without SiO

_{eff}_{2},). By changing

*h*, the effective index Re(

_{SiO2}*n*) varies from 160 to 400 [Fig. 3(e)]. It is worth noticing that there is a maximum FOM when the height of SiO

_{eff}_{2}varies. By comparing with the structure without SiO

_{2}, the maximum figure of merit Re(

*n*)/Im(

_{eff}*n*) ≈145 is obtained when

_{eff}*h*≈3.0 nm, which is 50% larger than that for the structure without SiO

_{SiO2}_{2}. For the structure with a maximum FOM, we have

*n*= 204.4 + 1.4i (i.e., the effective wavelength of the guided wave is

_{eff}*λ*= 48.9nm), and

_{spp}*L*= 1126 nm. In comparison, for a freestanding graphene waveguide with the same width, we have

_{m}*n*= 117.5 + 0.9i,

_{eff}*λ*= 89.1nm and

_{spp}*L*= 1785 nm. Since the effective wavelength in the waveguide on Si substrate is only about half of that in the freestanding graphene waveguide, more propagation cycles can be supported (before the light is very much attenuated) in our current design. Typically the size of an optical device is on the order of

_{m}*λ*. The graphene ribbon (with a silica buffer layer) on a silicon wafer provides a much smaller

_{spp}*λ*

_{spp}than the freestanding graphene with the same width. Here, high-index Si pulls the light away from the lossy graphene to the lossless SiO

_{2}layer and helps to improve FOM. It is noted that the buffering of our graphene ribbon increases slightly the device volume along the vertical direction, but the loss reduction is significant. This is very suitable for planar integration.

Finally, we would like to study the mode properties of a graphene waveguide formed by two graphene ribbons. Similar to the case of mode coupling in a conventional MIM (metal-insulator-metal) or IMI (insulator-metal-insulator) waveguide, it is expected that the coupling of the nano-ribbon waveguide can lead to further mode splitting. Thus, the waveguide constructed by two identical graphene ribbons with a nanometer gap is investigated in this work. Two kinds of coupling configurations, namely, the side-side configuration [Fig. 4(a)] and top-bottom configuration [Fig. 4(b)], are studied.

For the side-side configuration, *d* is the distance between the two graphene ribbons, and the width of graphene ribbons is chosen to be 20 nm. The hybridization of the waveguide modes will leads to the presence of both a symmetric mode [Fig. 5(a)] and an anti-symmetric mode [Fig. 5(b)], It is found that the symmetric mode can squeeze the optical energy effectively into the gap between the two ribbons, but the anti-symmetric mode only slightly modifies the profile of the optical energy density. For the symmetric mode, *n _{eff}* and the figure of merit increase when

*d*is reduced [Fig. 5(c)]. The waveguide mode area depends critically on the gap size [Fig. 5(d)]. A smaller gap will lead to a smaller mode area. The mode area is extremely small, only about 10

^{−7}

*λ*, which is an order smaller than the smallest mode area of any waveguide that has ever been reported in the literature [19].

_{0}^{2}For the top-bottom configuration, *d* is the vertical distance between the two ribbons, and the width of the graphene ribbons we use is also 20 nm. The mode with top-bottom configuration [20] is called “gap mode” in a very recent publication [21]. Here the name of “top-bottom configuration” is used for consistency and comparison with our side-side configuration. Again, there are both a symmetric mode and an anti-symmetric mode due to the presence of mode coupling. Figure 6(a)–6(f) shows the distributions of energy, electric field and magnetic field of the two modes. For the symmetric mode, more energy is confined in the area between the two ribbons. For the anti-symmetric mode, the field is mainly located at the top and bottom regions (the energy density, however, is still concentrated within the graphene ribbon due to the large effective permittivity *ε _{eff}*). Figure 6(g) displays Re(

*n*) of the two modes. The effective refractive index of the waveguide mode depends critically on distance

_{eff}*d*. If the quantum effect is not taken into account, Re(

*n*) can be extremely large, reaching a value of up to 1000 when

_{eff}*d*is smaller than 1 nm. The figure of merit is also a function of distance

*d*, and can be more than 180. However, the mode area remains more or less the same when

*d*changes due to the large amount of energy in graphene.

For our double-ribbon configurations, the propagation length is not plotted due to the limitation of paper length. In fact, the propagation length *L*_{m} can be related to FOM as *L*_{m} = FOM**λ _{0}/*4πRe(

*n*). For example, in the top-bottom configuration, the propagation length is Lm = 1477.3 nm [while Re(neff) = 187.21, Im(n

_{eff}*) = 1.08 and FOM = 173] when*

_{eff}*d*= 5 nm.

Due to the extremely large wavenumber supported by this configuration, additional modes with high order oscillations are found in our calculation. If *d* is not too small, e.g. 20 nm, only two modes are supported. However, if the two ribbons get closer to each other, wavenumber *k* will grow significantly. Then more oscillations along the lateral direction will occur, resulting in more waveguide modes. For example, when *d* is 2 nm, one additional mode appears [Fig. 7]. This characteristic is potentially useful for the design of e.g. ultra-compact MMI (multimode interferometer) devices.

Since a single freestanding graphene ribbon can support a plasmonic waveguide mode with an extremely large wave number, this configuration has high potential for design of some ultra-compact optical structure. Here as an example we give a nano-ring cavity based on a single freestanding graphene ribbon. In our design, the width of the graphene ribbon is chosen to be 20 nm. It is well known that a round-trip phase accumulation of integral multiple of 2π can result in a resonant mode. With an inner radius *r* = 38 nm, we see a resonant peak at 30 THz (corresponding to a vacuum wavelength of about 10 μm) in our numerical simulation with CST software [Fig. 8]. The mode profile shows that it is a 4th-order ring cavity mode. The *Q* value (*Q* = *f*/*Δf*, *f* is the frequency, *Δf* is the linewidth) is 42.3. Considering the ring size is on the order of nanometer, the cavity size (compared to the vacuum wavelength) is surprisingly small. Since the response of graphene can be tuned easily by changing its chemical potential such as doping, the resonant frequency can be controlled in a flexible way.

## 3. Conclusion

In summary, the unique properties of plasmonic waveguides based on graphene nano-ribbons have been investigated. Compared with traditional plasmonic materials (noble metals), graphene has several advantages: 1) The properties of graphene can be tuned via electrical gating or chemical doping. 2) It has low Ohmic losses, and the FOM can be 145, which is much larger than that for a noble metal (for the noble metal, the FOM is usually smaller than 80 [22]). 3) The effective index of graphene can be ultra-large, which can make the effective wavelength of the guided wave ultra-small. 4) It is very thin (one-atom layer, which is far thinner than any fabricated metal layer) and can be very narrow (20 nm in this paper) for waveguide. Due to these superior properties, graphene nanoribbons waveguides have the potential for future optical devices.

The guided modes are tightly confined in both the lateral direction and the propagation direction. A single mode operation region has been identified if the ribbon width is smaller than 50 nm. Due to the tight confinement, a record small mode area (an order smaller than the smallest mode area of any waveguide that has ever been reported in the literature) and extremely high effective refractive index can be achieved. The low-loss waveguide structure with an embedded low index silica layer between the graphene layer and the silicon substrate has been proposed to reduce the propagation loss and increase the FOM of the plasmonic waveguide. It is very attractive for designing optical devices with high performance. The coupled configurations with two identical graphene ribbons also exhibit interesting properties. In particular, the side-side-coupling can further reduce the waveguide mode area, while the top-bottom-coupling can result in much larger effective indices than an isolated graphene ribbon. A nano-ring cavity of extremely small size based on a graphene ribbon waveguide has been shown to support a cavity resonance at far infrared range.

## Acknowledgments

The authors are grateful to the partial supports of Swedish VR grant (# 621-2011-4620) and AOARD, NSFC 61178062, and the National High Technology Research and Development Program (863 Program; No. 2012AA030402).

## References and links

**1. **L. Ju, B. Geng, J. Horng, C. Girit, M. Martin, Z. Hao, H. A. Bechtel, X. Liang, A. Zettl, Y. R. Shen, and F. Wang, “Graphene plasmonics for tunable terahertz metamaterials,” Nat. Nanotechnol. **6**(10), 630–634 (2011). [CrossRef] [PubMed]

**2. **A. K. Geim and K. S. Novoselov, “The rise of graphene,” Nat. Mater. **6**(3), 183–191 (2007). [CrossRef] [PubMed]

**3. **K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, “Electric field effect in atomically thin carbon films,” Science **306**(5696), 666–669 (2004). [CrossRef] [PubMed]

**4. **A. Vakil and N. Engheta, “Transformation optics using graphene,” Science **332**(6035), 1291–1294 (2011). [CrossRef] [PubMed]

**5. **S. A. Mikhailov and K. Ziegler, “New electromagnetic mode in graphene,” Phys. Rev. Lett. **99**(1), 016803 (2007). [CrossRef] [PubMed]

**6. **M. Jablan, H. Buljan, and M. Soljacic, “Plasmonics in graphene at infrared frequencies,” Phys. Rev. B **80**(24), 245435 (2009). [CrossRef]

**7. **F. H. L. Koppens, D. E. Chang, and F. J. García de Abajo, “Graphene plasmonics: a platform for strong light-matter interactions,” Nano Lett. **11**(8), 3370–3377 (2011). [CrossRef] [PubMed]

**8. **G. W. Hanson, “Dyadic Green’s functions and guided surface waves for a surface conductivity model of graphene,” J. Appl. Phys. **103**(6), 064302 (2008). [CrossRef]

**9. **A. Y. Nikitin, F. Guinea, F. J. García-Vidal, and L. Martín-Moreno, “Edge and waveguide terahertz surface plasmon modes in graphene microribbons,” Phys. Rev. B **84**(16), 161407 (2011). [CrossRef]

**10. **P. R. West, S. Ishii, G. V. Naik, N. K. Emani, V. M. Shalaev, and A. Boltasseva, “Searching for better plasmonic materials,” Laser Photon. Rev. **4**(6), 795–808 (2010). [CrossRef]

**11. **S. Thongrattanasiri, F. H. L. Koppens, and F. J. García de Abajo, “Complete optical absorption in periodically patterned graphene,” Phys. Rev. Lett. **108**(4), 047401 (2012). [CrossRef] [PubMed]

**12. **L. Wu, H. S. Chu, W. S. Koh, and E. P. Li, “Highly sensitive graphene biosensors based on surface plasmon resonance,” Opt. Express **18**(14), 14395–14400 (2010). [CrossRef] [PubMed]

**13. **E. H. Hwang and S. Das Sarma, “Dielectric function, screening, and plasmons in two-dimensional graphene,” Phys. Rev. B **75**(20), 205418 (2007). [CrossRef]

**14. **B. Wunsch, T. Stauber, F. Sols, and F. Guinea, “Dynamical polarization of graphene at finite doping,” New J. Phys. **8**(12), 318 (2006). [CrossRef]

**15. **G. Eda, G. Fanchini, and M. Chhowalla, “Large-area ultrathin films of reduced graphene oxide as a transparent and flexible electronic material,” Nat. Nanotechnol. **3**(5), 270–274 (2008). [CrossRef] [PubMed]

**16. **E. Verhagen, M. Spasenović, A. Polman, and L. K. Kuipers, “Nanowire plasmon excitation by adiabatic mode transformation,” Phys. Rev. Lett. **102**(20), 203904 (2009). [CrossRef] [PubMed]

**17. **S. A. Maier, *Plasmonics: Fundamentals and Applications* (Springer, 2007).

**18. **P. Berini, “Figures of merit for surface plasmon waveguides,” Opt. Express **14**(26), 13030–13042 (2006). [CrossRef] [PubMed]

**19. **Q. Huang, F. Bao, and S. He, “Nonlocal effects in a hybrid plasmonic waveguide for nanoscale confinement,” Opt. Express **21**(2), 1430–1439 (2013). [CrossRef] [PubMed]

**20. **An early version of the present paper was put online in May, 2013, at ArXiv. S. He, X. Zhang, and Y. He, “Graphene nano-ribbon waveguides,” ArXiv preprint 2013, DOI: arXiv: 1305.6500, http://arxiv.org/abs/1305.6500.

**21. **Y. Francescato, V. Giannini, and S. A. Maier, “Strongly confined gap plasmon modes in graphene sandwiches and graphene-on-silicon,” New J. Phys. **15**(6), 063020 (2013). [CrossRef]

**22. **R. Buckley and P. Berini, “Figures of merit for 2D surface plasmon waveguides and application to metal stripes,” Opt. Express **15**(19), 12174–12182 (2007). [CrossRef] [PubMed]