Open Access
Add to BibSonomy
Abstract
Cyclic olefin copolymer (COC) is shown, via infrared variable angle spectroscopic ellipsometry (IR-VASE), to have low absorption along with low dispersion in the long-wave infrared (LWIR) band. The material is demonstrated as the dielectric standoff layer in an LWIR metasurface design consisting of metallic patches and a ground plane, which is fabricated via standard lithographic processes. The resultant metasurface is observed to display strong resonant behavior near 10 µm, without the absorption features typically observed in similar designs using previously studied polymeric materials. COC should be considered for use in future metasurface work where researchers wish to study the physics of LWIR metasurface behavior without the complications caused by absorptive loss in the dielectric layer(s).
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Metasurfaces are a class of artificial materials comprised of subwavelength structures which induce engineered light-matter interactions within geometrical thickness much smaller than the design wavelengths. With plentiful scientific applications, metasurfaces have been studied heavily in recent years [1,2]. A portion of work in metasurfaces has been carried out at long-wave infrared (LWIR) wavelengths (8-12 μm), where numerous filtering concepts have been demonstrated [3–6].
As metasurfaces continually trend larger in area, parallel fabrication processes as well as those which do not require high vacuum are desired. Considering a common configuration consisting of metallic elements stood off from a ground plane by a dielectric spacer layer, a relatively straightforward reduction in complexity and cost is replacement of the dielectric layer (typically deposited in vacuum) with a polymeric dielectric which can be spin-coated in ambient conditions. Prior work has shown that most materials considered for this type of design have significant absorption bands at LWIR [7,8]. For example, benzocyclobutene (BCB) is one spin-on dielectric polymer which has been used for LWIR metasurfaces. In a study focused on harmonic resonances in infrared metasurfaces, it was observed that the absorption bands in BCB overlapped with the spectral resonance features [6]. This complicated the interpretation of the fundamental and harmonic resonances displayed by the metasurfaces, motivating a transition to a different material system relying on ZnS as the dielectric layer [9]. One might consider polyethylene as a candidate polymer as it has been reported to have no absorption at LWIR [10,11]; however, spin coating of polyethylene thin films requires heated conditions and thus has not been widely adopted [12,13]. Certainly, various inorganic materials exist with windows of transparency in the long-wave infrared band, but these typically must be deposited under vacuum and tend to possess relatively high permittivity. Moreover, some of the materials (such as ZnSe) have handling concerns due to toxicity.
Here, cyclic olefin copolymer (COC) is shown as a possible material for consideration of use in LWIR metasurfaces. The optical properties of COC have been studied previously at visible and near-infrared (NIR) wavelengths [14] and recent work utilized COC for metasurfaces in the visible and NIR spectral ranges due its transparency and ease in fabrication processes [15]. The far-infrared optical properties of COC have recently been investigated and presented for use as windows and filters in this band [16]. Lately, COC has seen growing adoption as a dielectric material for THz metasurfaces [17–21]. The transparency and lack of dispersion in COC was highlighted in a recent review paper on dielectrics for THz metasurfaces, which discussed several examples where COC has been used [17]. Very recently, two switchable metasurface absorbers using VO2 and graphene have been numerically investigated, using COC as the dielectric standoff layers in each design [18,19]. Additionally, COC has recently been utilized as the substrate material for quasi-bound state in the continuum (quasi-BIC) THz metasurfaces, due to its low refractive index, low absorption, and flexibility [20,21]. The optical properties of COC at LWIR wavelengths have not been presented previously in the literature, although the LWIR spectral transmission through COC substrates [22] and free-standing films [23] has been noted in previous work.
In this report, the optical properties of thin-film COC are investigated and the complex relative permittivity is presented for 2-20 µm using infrared variable-angle spectroscopic ellipsometry (IR-VASE). The complex relative permittivity of the materials comprising optical metasurfaces explicitly impacts their performance and the values are key inputs to predictive modeling. The relative transparency of COC at LWIR wavelengths suggests the material is a suitable dielectric for use in LWIR metasurfaces. As a demonstration, an LWIR patch-type metasurface is designed using COC as the dielectric spacer layer between the patches and ground plane. The design is fabricated and strong metasurface resonance is observed near 10 µm without the complications of dielectric loss which is often observed with spin-on dielectric materials in the LWIR band.
2. Ellipsometry experiment and results
The optical properties of the thin film COC material were determined via IR-VASE using a J. A. Woollam instrument and associated WVASE software. (For a technical description of the apparatus, see [24]). In this work, thin films of COC were deposited using mr-I T85-1.0 (from Micro Resist Technology), which is a COC-based thermal nano-imprint lithography resist with target thickness on the order of 1 µm. The substrates for all the data presented here were silicon wafers coated with optically thick (∼230 nm) Au. Using a typical IR-VASE measurement configuration, we initially observed that the ∼1 µm target thickness of the mr-I T85-1.0 formulation yielded results with several absorption features that were indistinguishable from instrumental noise. Thus, a more rigorous approach was taken. In the measurement settings, the number of scans to average was increased to 250 to improve signal-to-noise ratio and the resolution was set to 4 cm-1 to resolve fine spectral features. For two angles of incidence (70 and 75 degrees), each automated measurement took approximately 12 hours to complete.
The strategy for modeling of complex relative permittivity (ε′,ε′′) for materials via ellipsometry is to first determine the sample thickness as a substrate (optically thick) or use a Cauchy dispersion model to determine thin-film thickness over a transparent spectral range. Then, the spectral pseudo-permittivity from the raw data are calculated using a point-by-point fit, which enforces Kramers-Kronig causality. Next, the lineshape of the pseudo-permittivity is fit using physical oscillator models, which have absorptive resonances in ε′′ and dispersion fluctuations in ε′. The low-frequency offset of ε′ (the “DC offset”) and resonances far outside of the measured spectral range are often added to fit ε′ correctly. The parameters are iterated upon until the mean squared error between the measured and modeled data are minimized. Here, the Au substrate layer was measured before application of the COC material and it was accurately modeled as a single Drude oscillator. The COC material model on top of the Au substrate layer was developed by determining the film thickness in a spectral region void of absorption features, then fitting Gaussian oscillators (amplitude, spectral location, and bandwidth) and the low-frequency offset of ε′ to match the pseudo-permittivity lineshape.
For the COC material model, multiple samples were fabricated to improve data accuracy. Since sample thickness corresponds directly to the amplitude of absorptive features derived from ellipsometry, generating a single material model from multiple samples greatly increases confidence in the reported results. Three samples of different thicknesses were fabricated by varying the speed of the spin-coater: 903 nm (3000 rpm), 1479 nm (1000 rpm), and 2425 nm (500 rpm). The thickest (2425 nm) COC sample was analyzed first and the results were checked against the two thinner COC samples to ensure that the Gaussian oscillators and low-frequency offset could be applied consistently across the three sample measurements. Only absorption features that exist for all three data sets (with amplitude in the raw data that decreases monotonically with decreasing film thickness) are included in our results.
The derived optical properties of thin film COC from this study are shown in Fig. 1, in terms of wavenumber (cm-1) for consistent frequency spread across the 2-20 µm band of interest. The major spectral features appear in the 2800-3000 cm-1 range (∼3.3-3.6 µm) which bring about strong absorption peaks in ε′′ and corresponding rapid fluctuations in ε′ (anomalous dispersion). Another relatively large spectral feature appears near 1450 cm-1 (∼6.9 µm) and a number of relatively small features appear in the 700-1400 cm-1 (∼7.1-14.3 µm) range. While small spectral features present in the LWIR range (8.0-12.0 µm) indicate that the COC films are not free from absorption, the band averaged ε′′ is calculated to be relatively low at 0.012 with a maximum of 0.028 near 933 cm-1 (∼10.7 µm). It is noted that a slightly higher peak in ε′′ (0.036) appears just outside of LWIR between 7.9 and 8.0 µm. Additionally, dispersion is quite small with ε′ having an average of 2.355 with a minimum of 2.354 and a maximum of 2.366 within the LWIR band. For further examination or use in electromagnetic modeling, the tabulated results of Fig. 1 can be found in Data File 1 [25].
Fig. 1. Real (blue dashed, left axis) and imaginary (red, right axis) parts of the relative permittivity of COC as derived from experimental ellipsometry measurements. See Data File 1 [25] for underlying values.
As previously stated, the physical oscillator models in ellipsometric analysis are explicitly responsible for the lineshape of ε′ and ε′′. The amplitude, spectral location, and bandwidth of the Gaussian oscillators used in the spectroscopic ellipsometry model are presented in Table 1, arranged in order of decreasing frequency (cm-1) and normalized to the strongest oscillator at 2946.8 cm-1. As is often the case in spectroscopy, the lineshape cannot generally be modeled by single oscillators and thus a summation of overlapping oscillators is used to model the experimental ellipsometric data. It is noted that many of the oscillators shown here correlate to absorption peaks which have been shown in previous FTIR studies involving COC [22,23]. In prior work involving deposition of SiO2 on COC films, COC absorption peaks near 2946 cm-1 and 2868 cm-1 were attributed to stretching vibration modes of –CH2 and –CH3 while the peak at 1453 cm-1 was attributed to the –CH3 wagging mode in the polymer backbone [22]. A study focused on polypropylene/COC blends identified and assigned modes for several absorption bands in the spectra of COC that correspond to the oscillators observed here: 2941 cm-1 (SP3C asymmetric stretching vibration), 2866 cm-1 (SP3C symmetric stretching vibration), 1452 cm-1 (-CH2- bending vibration), 1294 cm-1 (stretching vibration of C-C backbone), 1254 cm-1 (CH bending vibration) and 934 cm-1 (CH bending vibration) [23].
Table 1. Amplitude, wavenumber, and bandwidth of oscillator models used in the COC model
3. Metasurface design, fabrication, and results
One common LWIR metasurface configuration operates as a reflective band-reject filter and consists of a patterned metallic patch geometry separated from a metallic ground plane by a dielectric standoff layer. The metallic patches exhibit a dipolar resonance which primarily depends on the patch length and the standoff layer is tuned to an appropriate thickness which greatly enhances this resonance, leading to a spectral absorption band centered at the design wavelength. The standoff layer confines the resonant electric fields in a dipolar distribution between the patches and ground plane and the optical properties of the dielectric material greatly influences the ultimate performance of the filter [26]. Such a design is presented here, a simple square-patch metasurface, to illustrate the utility of COC as a potential dielectric standoff layer in LWIR metasurfaces.
Finite element method electromagnetic modeling and simulation was carried out using the radio frequency (RF) module of COMSOL Multiphysics 5.5. The Floquet method was employed to apply the appropriate boundary conditions for an infinite planar array while only solving for the unit cell. The metasurface was simulated at normal incidence using linearly polarized excitation. (Due to the square symmetry of this design, the orthogonal linear polarizations induce an equal response at normal incidence). Au was used as the metallic material in the patches and ground plane due to its relatively high conductivity versus other metals at LWIR. The frequency dependent relative permittivity of Au from previous work [9] was implemented in the model and the relative permittivity of COC was initially assumed to be 2.34 without loss across the wavelength domain of interest. The design was optimized via a series of parametric sweeps of the patch size, periodicity, and COC layer thickness based on the strength of the resonant behavior – equating here to absorption – near 10 μm. The final design consists of 2.6 μm Au patches with 5 μm periodicity, deposited on a dielectric spacer layer of COC above a Au ground plane. The thicknesses of the patch, dielectric, and ground plane are 75 nm, 175 nm, and 150 nm, respectively.
The demonstration metasurface was fabricated via various standard processes. First, a 150 nm layer of Au was deposited via sputtering to form the ground plane. The mr-I T85-1.0 COC material was spin coated using a Laurell tool at 3000 rpm for 90 seconds and then baked at 180° C for 60 seconds. This procedure yielded a thickness of approximately 1 μm, which was then reactive-ion etched (RIE) using an SF6/O2 mixture in an Oxford Plasmalab system to obtain the desired 175 nm film thickness. Lithographic patterning was carried out using the following metallization and liftoff technique. PMMA A4 (Microchem) was used as a positive-tone electron beam lithography (EBL) resist. The resist was deposited via spin-coating at 3000 rpm for 90 seconds, followed by a 180° C post bake on a hotplate, yielding an approximate thickness of 200 nm. The resist was exposed using an FEI Quanta 650 tool, with a beam current of 30 pA. The exposed resist was developed using MIBK:IPA (1:3) at room temperature for 50 seconds. The patch metallization was carried out via sputter deposition of 75 nm Au (with a Ti adhesion layer). Liftoff of the resist was conducted using PG Remover (Microchem) heated to 60° C for approximately 2 hours with a brief (2-3 second) sonication, followed by an IPA rinse and drying with N2. An SEM micrograph showing the resultant patch metasurface is shown in Fig. 2.
The experimental and simulated spectral reflectivity at LWIR wavelengths for the metasurface design presented here is shown in Fig. 3. The experimental data was measured using a Perkin-Elmer microscope FTIR which samples a 100 µm diameter circular sample area. The measurement is unpolarized, with the instrument collecting an annular field of view near normal incidence. (For a technical description of the instrument, see [27]). As previously stated, the simulated results come from infinite array modeling in COMSOL; the modeled results in Fig. 3 reflect the COC relative permittivity shown in Fig. 1. The experimental and simulated results show relatively good agreement, with a strong absorptive resonance near 10 µm and trending toward high reflectivity away from that resonance. The difference between the Q-factors of the two resonances plotted is most likely caused by losses arising from residual roughness in the metallic films and roughness in the patch edges seen in Fig. 2, which were not accounted for in the COMSOL model [28,29]. Overall, both show a functional shape which is characteristic of a metasurface resonance, which is mostly unaffected by the absorptive features seen in many lossy LWIR materials. The small absorption features observed in COC via ellipsometry appear as slight dips or kinks in the reflectance lineshape of both the experimental and simulated data. Comparing the designed and fabricated metasurface reflectance demonstrates that the COC dielectric material can indeed be employed in this manner due to its high LWIR transparency.
Fig. 3. Measured (solid) and simulated (dashed) spectral reflectivity results for the optical metasurface discussed here.
4. Discussion and conclusions
COC was shown to be a useful dielectric material for metasurface design and fabrication at LWIR. Figure 4 (top) supports this conclusion via parametric simulation of the patch size of the example metasurface previously discussed. Here, all other parameters are held constant while the patch size is varied from 2.0 to 3.5 µm in 0.5 µm increments and it is seen that the resonance shifts depending on the patch size as expected. Figure 4 (bottom) shows the same parametric sweep; however, the COC complex relative permittivity values are replaced with those of an alternative dielectric material, BCB. As mentioned previously, BCB is a spin-on dielectric polymer which has been often used in LWIR metasurface fabrication. The material has approximately the same average real relative permittivity over the LWIR spectral range as COC (2.35). The BCB relative permittivity was previously derived via ellipsometry and is dominated by absorptive features near 8 µm, 9.5 µm, 12 µm and 12.4 µm [6]. Comparing Fig. 4 top and bottom, it is clear that the absorptive features and accompanying fluctuations in the real part of the relative permittivity of BCB are complicating factors which affect the spectral properties of the intended design. Most notably, the fundamental resonance no longer clearly shifts according to the patch size. It is noted that a previous study found BCB to be one of the more favorable dielectrics for infrared devices such as metasurfaces, because of its favorable processing conditions and dielectric properties, as compared to several organic and inorganic materials [7].
Fig. 4. (Top) Simulated spectral reflectivity of the demonstration metasurface design with parameterization of the patch width as 2.0 µm (blue line), 2.5 µm (orange line), 3.0 µm (yellow line), and 3.5 µm (violet line). (Bottom) Simulated spectral reflectivity of the demonstration metasurface design with parameterization of the patch width as 2.0 µm, 2.5 µm, 3.0 µm, and 3.5 µm, replacing the COC layer with BCB.
As regards the benefits and limitations of COC for use in LWIR metasurfaces, the observed transparency is an important benefit in reducing the complications brought about by absorption bands shown by most dielectrics at these wavelengths. While this report highlighted the use of COC in studying the fundamental physics of metasurface resonance, it is also noted that the material could be used to improve the Q factor and/or efficiency of LWIR devices by decreasing the losses in the standoff layer. We note that small amounts of absorption were indeed observed in the ellipsometric analysis of COC at LWIR. Although these did not have a significant impact of the performance our demonstration metasurface (as evidenced by Figs. 3 and 4), such a statement cannot be made in general for all LWIR metasurfaces as each metasurface design induces unique concentrated electromagnetic field distributions in the dielectric layer(s) which will each have unique performance regarding dielectric absorptions. The LWIR peak in ε′′ occurs near 933 cm-1 (∼10.7 µm), arguably having the most impact on potential metasurface performance in this range. Additionally, relatively strong absorption bands were observed at MWIR wavelengths, which would hamper designs seeking broadband performance over both bands.
As a spin-on material, mr-I T85-1.0 has an obvious fabrication advantage over comparable vacuum-deposited dielectric layers. The material is useable as-is, eliminating the need to dissolve COC pellets in solvent chemicals, which has been noted by other researchers [16]. Additionally, mr-I T85-1.0 was easily etched by a standard RIE process and has shown compatibility with a standard EBL process. COC, by way of mr-I T85-1.0, should be strongly considered for future experimental work in LWIR metasurfaces.
Acknowledgments
A portion of this work was carried out with equipment and support of the Rice Shared Equipment Authority. The authors express gratitude to Dr. Tom Tiwald of J. A. Woollam Co., Inc. for helpful discussions regarding experimentation and fitting of the IR-VASE data.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are available in Data File 1, Ref. [25].
References
1. N. Meinzer, W. L. Barnes, and I. R. Hooper, “Plasmonic meta-atoms and metasurfaces,” Nat. Photonics 8(12), 889–898 (2014). [CrossRef]
2. S. S. Bukhari, J. Vardaxoglou, and W. Whittow, “A metasurfaces review: definitions and applications,” Appl. Sci. 9(13), 2727 (2019). [CrossRef]
3. S. Liu, M. B. Sinclair, T. S. Mahony, Y. C. Jun, S. Campione, J. Ginn, D. A. Bender, J. R. Wendt, J. F. Ihlefeld, P. G. Clem, J. B. Wright, and I. Brener, “Optical magnetic mirrors without metals,” Optica 1(4), 250–256 (2014). [CrossRef]
4. B. Debbrecht, M. McElhiney, V. Clary, C. Cullen, M. S. Mirotznik, and B. G. DeLacy, “Cavity-based aluminum nanohole arrays with tunable infrared resonances,” Opt. Express 25(20), 24501–24511 (2017). [CrossRef]
5. J.-Y. Jung, K. Song, J.-H. Chio, J. Lee, D.-G. Choi, J.-H. Jeong, and D. P. Neikirk, “Infrared broadband metasurface absorber for reducing the thermal mass of a microbolometer,” Sci. Rep. 7(1), 430 (2017). [CrossRef]
6. J. D’Archangel, E. Tucker, E. Kinzel, E. A. Muller, H. A. Bechtel, M. C. Martin, M. B. Raschke, and G. Boreman, “Near- and far-field spectroscopic imaging investigation of resonant square-loop infrared metasurfaces,” Opt. Express 21(14), 17150–17160 (2013). [CrossRef]
7. W. R. Folks, J. Ginn, D. Shelton, J. Tharp, and G. Boreman, “Spectroscopic ellipsometry of materials for infrared micro-device fabrication,” Phys. Status Solidi C 5(5), 1113–1116 (2008). [CrossRef]
8. R. D. Raspberry, Y. Lee, J. C. Ginn, P. F. Hines, C. L. Arrington, A. E. Sanchez, M. T. Brumbach, P. G. Clem, D. W. Peters, M. B. Sinclair, and S. M. Dirk, “Low loss photopatternable matrix materials for LWIR-metamaterial applications,” J. Mater. Chem. 21, 13092 (2011). [CrossRef]
9. E. Tucker, J. D’ Archangel, and G. Boreman, “Near- and far-field investigations of dark and bright higher order resonances in square loop elements at mid-infrared wavelengths,” Opt. Express 25(5), 5594–5608 (2017). [CrossRef]
10. V. M. Zolotarev, B. Z. Volchek, and E. N. Vlasova, “Optical constants of industrial polymers in the IR region,” Opt. Spectrosc. 101(5), 716–723 (2006). [CrossRef]
11. W. R. Folks, S. K. Pandey, G. Pribil, D. Slafer, M. Manning, and G. Boreman, “Reflective infrared ellipsometry of plastic films,” Int J Infrared Milli Waves 27(11), 1553–1571 (2007). [CrossRef]
12. O. Mellbring, S. K. Oiseth, A. Krozer, J. Lausmaa, and T. Hjertberg, “Spin coating and characterization of thin high-density polyethylene films,” Macromolecules 34(21), 7496–7503 (2001). [CrossRef]
13. E. Lock, S. Walton, and R. Fernsler, “Preparation of ultra thin polystyrene, polypropylene and polyethylene films on Si substrate using spin coating technology,” NRL/MR/6750—08-9092 (Naval Research Laboratory, 2008).
14. G. Khanarian, “Optical properties of cyclic olefin copolymers,” Opt. Eng. 40(6), 1024 (2001). [CrossRef]
15. J. Cai, C. Zhang, and W.-D. Li, “Dual-color flexible metasurfaces with polarization-tunable plasmons in gold nanorod arrays,” Adv. Opt. Mater. 9(4), 202001401 (2020). [CrossRef]
16. E. J. Wollack, K. L. Denis, A. Barlis, M.-P. Chang, A. S. Kutyrev, K. H. Miller, and P. C. Nagler, “Far-infrared properties of cyclic olefin copolymer,” Opt. Lett. 45(3), 780–783 (2020). [CrossRef]
17. R. T. Ako, A. Upadhyay, and W. Withayachumnankul, “Dielectrics for terahertz metasurfaces: material selection and fabrication techniques,” Adv. Opt. Mater. 8(3), 1900750 (2020). [CrossRef]
18. Y. Liu, R. Huang, and Z. Ouyang, “Terahertz absorber with dynamically switchable dual-broadband based on a hybrid metamaterial with vanadium dioxide and graphene,” Opt. Express 29(13), 20839–20850 (2021). [CrossRef]
19. M. Zhang and Z. Song, “Switchable terahertz metamaterial absorber with broadband absorption and multiband absorption,” Opt. Express 29(14), 21551–21561 (2021). [CrossRef]
20. Y. K. Srivtava, R. T. Ako, M. Gupta, M. Bhaskaran, S. Sriram, and R. Singh, “Terahertz sensing of 7 nm dielectric film with bound states in the continuum metasurfaces,” Appl. Phys. Lett. 115(15), 151105 (2019). [CrossRef]
21. T. C. Tan, Y. K. Srivastana, R. T. Ako, W. Wang, M. Bhaskaran, S. Sriram, I. Al-Naib, E. Plum, and R. Singh, “Active control of nanodielectric-induced THz quasi-BIC in flexible metasurfaces: a platform for modulation and sensing,” Adv. Mater. 33(27), 2100836 (2021). [CrossRef]
22. K.-S. Ma, F. Reza, I. Saaem, and J. Tian, “Versatile surface functionalization of cyclic olefin copolymer (COC) with sputtered SiO2 thin film for potential BioMEMS applications,” J. Mater. Chem. 19(42), 7914–7920 (2009). [CrossRef]
23. A. Gopanna, R. N. Mandapati, S. P. Thomas, K. Rajan, and M. Chavali, “Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy and wide-angle X-ray scattering (WAXS) of polypropylene (PP)/cyclic olefin copolymer (COC) blends for qualitative and quantitative analysis,” Polym. Bull. 76(8), 4259–4274 (2019). [CrossRef]
24. J. A. Woollam, “IR Vase Brochure,” https://www.jawoollam.com/download/pdfs/ir-vase-brochure.pdf.
25. J. D’ Archangel, “IR COC permittivity derived from thin film ellipsometry,” figshare (2021), https://doi.org/10.6084/m9.figshare.16614013.
26. S. Ogawa and M. Kimata, “Metal-insulator-metal-based plasmonic metamaterial absorbers at visible and infrared wavelengths: a review,” Materials 11(3), 458 (2018). [CrossRef]
27. XXI. Perkin-Elmer, “Product Note: FT-IR Microscopy,” https://www.perkinelmer.com/CMSResources/Images/46-73845PRD_MultiScopeFTIRMicroscopeSystem.pdf.
28. F. Benson, “Waveguide attenuation and its correlation with surface roughness,” Proc. IEE - Part III: Radio and Communication Engineering 100(64), 85–90 (1953). [CrossRef]
29. G. Gold and K. Helmreich, “Surface impedance concept for modeling conductor roughness,” in 2015 IEEE MTT-S International Microwave Symposium (2015), pp. 1–4.
