Open Access
Add to BibSonomyAbstract
Dielectric loaded surface plasmon waveguides (DLSPPWs) comprised of polymer ridges deposited on top of CMOS compatible metal thin films are investigated at telecom wavelengths. We perform a direct comparison of the properties of copper (Cu), aluminum (Al), titanium nitride (TiN) and gold (Au) based waveguides by implementing the same plasmonic waveguiding configuration for each metal. The DLSPPWs are characterized by leakage radiation microscopy and a fiber-to-fiber configuration mimicking the cut-back method. We introduce the ohmic loss rate (OLR) to analyze quantitatively the properties of the CMOS metal based DLSPPWs relative to the corresponding Au based waveguides. We show that the Cu, Al and TiN based waveguides feature extra ohmic loss compared to Au of 0.027 dB/μm, 0.18 dB/μm and 0.52 dB/μm at 1550nm respectively. The dielectric function of each metal extracted from ellipsometric spectroscopic measurements is used to model the properties of the DLSP-PWs. We find a fairly good agreement between experimental and modeled DLSPPWs properties except for Al featuring a large surface roughness. Finally, we conclude that TiN based waveguides sustaining intermediate effective index (in the range 1.05–1.25) plasmon modes propagate over very short distances restricting the the use of those modes in practical situations.
© 2017 Optical Society of America
1. Introduction
Material of choice for plasmonics studies are most often noble metals such as gold (Au) or silver (Ag). However, when practical applications of plasmonics are targeted, noble metals are not convenient due to their high cost and more importantly due to their ability to contaminate semiconductor materials. This last property makes noble metals incompatible with CMOS processes hindering the development of large scale, low cost plasmonic devices. The limitations of noble metals in practical plasmonic applications have been identified for quite some time and alternative CMOS compatible plasmonics metals have been proposed [1–3]. Mainly three metals have been considered: aluminum (Al) [4], titanium nitride (TiN) [5] and copper (Cu) [6,7].
In the context of plasmonics, the optical properties of Al, TiN and Cu have been investigated by means of different experimental techniques such as photoelectron emission microscopy [8], attenuated total internal reflection [9, 10], angle-resolved reflection spectroscopy [11] or ellipsometry [12, 13]. Among these techniques, only ellipsometry has been applied to the three CMOS compatible metals we target. By using the dielectric functions extracted from ellipsometric measurements, the properties of a given plasmon mode can be indirectly obtained by means of a further accurate modeling of the configuration of interest. In this respect, no direct characterization of the same plasmonic waveguide configuration using different CMOS compatible metals has been reported so far, impeding the direct evaluation and comparison of those CMOS metals in plasmonic functionalities. It is the purpose of this study to report on such a direct comparison by using a common plasmonic waveguiding platform implemented onto thin films of Au (considered as the reference metal) and the three aforementioned CMOS compatible metals. The waveguides we deploy in this study are the so-called dielectric loaded surface plasmon waveguides (DLSPPWs) [14–16]. This very versatile plasmonic waveguide configuration is well-known and has been demonstrated in very practical contexts such as high-bit rate optical transmission [17, 18] and thermo-optical components [19–21]. In our situation, the main interest of DLSPPWs resides in the fact that they can be readily fabricated on top of any plasmonic metal film provided that the fabrication process does not alter the quality of the metal surface.
The study is organized as follows. In section 2, we first briefly describe our optical characterization set-up and the fabrication process of the DLSPPWs. Next, we demonstrate in section 3 the fiber-to-fiber configuration for the propagation distance measurements along the plasmonic waveguides. The full optical characterization includes leakage radiation microscopy imaging for the evaluation of the effective index of the guided modes. In section 4, we compile the results obtained for CMOS metal based DLSPPWs. In particular, we compute the ohmic loss rate quantifying the excess of ohmic loss relative to gold per unit propagation length for the CMOS based DLSPPW modes. Finally in section 5, we modeled the properties of the DLSP-PWs on the basis of the metals dielectric function obtained from ellipsometric measurements. The differences between experimental and modeled properties of the DLSPPWs are discussed. Concluding remarks and perspectives are given in the last section.
2. Experimental background
2.1. Characterization set–up
The set–up we operate for the characterization of the CMOS metal plasmonic properties is shown in Fig. 1. The samples are mounted onto a two–axis translation piezo-electric stage installed onto an inverted microscope equipped with a 1.49 numerical aperture (NA) immersion oil objective. The waveguides implanted at the surface of the samples are excited by a slightly focused spot (spot diameter around 10 μm). The focused spot is produced by a fiber-focuser with a focal length of 10mm. The light source used for the excitation of the waveguides is either a tunable laser diode or an incoherent broadband amplified spontaneous emission (ASE) source. Both sources cover the C+L (1530nm–1625nm) telecom bands. After propagation, the guided light is coupled out and collected by a focuser of same type as the one used for the excitation. The collected light is analyzed with an Optical Spectrum Analyser (OSA) or an accurate powermeter when the broadband ASE or laser diode source is used respectively. In addition to the fiber-to-fiber measurements, our set-up allows us to perform leakage radiation (LR) [22, 23] characterizations of the plasmonic waveguides by using a sensitive InGaAs infra-red camera. The LR imaging combined the accurate positioning of the sample provided by the piezo–electric stage is needed to perform reliable fiber-to-fiber measurements. Finally, apart from direct space image, the LR configuration can be readily changed to observe the Fourier plane of the sample giving a direct access to the effective index of the waveguide mode of interest.
Fig. 1 Schematic view of the optical characterization set-up. The set-up combines leakage radiation microscopy imaging and a fiber-to-fiber configuration for cut-back method applied to short propagation distance waveguides. The two lenses L1 and L2 are used to form the image of the sample onto the InGaAs camera. A third lens L3 can be added in the optical path to form the image of the fourier plane of the sample onto the camera sensor.
2.2. Dielectric loaded surface plasmon waveguide fabrication
DLSPPWs have been extensively investigated in the recent years [24]. DLSPPWs are comprised of a metal film on which a dielectric material (most often a polymer) is deposited. In this study, we expose the well-known SU-8 resist by electron beam lithography to fabricate the DLSP-PWs. The process we use is compatible with any of the metal layers we consider in this work including reactive metals such as copper (Cu) or aluminum (Al). SU-8 resist is known to be highly sensitive and of negative tone when exposed to an electron beam. Owing to these two properties, extremely short exposure times (in the range of a few seconds) are needed to expose the typical pattern described in the following. All the samples used in this work were fabricated on a microscope cover glass (thickness 170 μm, roughness 0.22nm RMS) coated with a metal layer. Table 1 summarizes the deposition conditions of the metal layers used throughout this study. For TiN, the metal layer was obtained by radio frequency (RF) sputtering of a 99.99% pure Ti target in an argon/nitrigen gas mixture. The stoichiometry of the layer was checked by XPS analysis and the DC electric resistivity was found to be below 40μΩ.cm, close to optimum values reported in the literature [25, 26]. The thickness and roughness of all metal layers were measured by atomic force microscopy. The purpose of the fabricated samples is to evaluate the propagation distance of the DLSPPW modes sustained by waveguides lying at the surface of substrates coated with different metals. Bird-eye view scanning electron images of typical DLSPPWs fabricated on a gold coated substrate are displayed in Fig. 2. The patterns we consider are linear DLSPPWs of different lengths equipped with input and output grating couplers needed for the fiber-to-fiber measurements. DLSPPWs having a width [see inset in Fig. 2] ranging from W=300nm to W=1000nm (height h around 295nm) have been fabricated on each sample. The image displayed in Fig. 2(b) shows a detail of the grating used to couple the light in and out the DLSPPWs. For an average angle of the incident focused beam of 30°, the optimized grating has a period of 2.2μm and a filling factor of 0.45. The high magnification image shown in Fig. 2(c) taken at the corner of the taper connecting the grating to the DLSPPW reveals the rather rough side walls of the waveguides (average roughness grain size around 100nm).
Table 1. Deposition conditions and morphological properties of the different metals. (B.P.: Base Pressure, D.P.: Deposition Pressure, e-gun: electron gun, RF-PVD: radio frequency physical vapor deposition), Rgh: roughness. The thickness of each metal film is large enough to make leakage radiation losses negligible (see text).
Fig. 2 (a) Scanning electron microscope bird-eye view images of the dielectric loaded surface plasmon waveguides. (b) Detail of the output grating coupler corresponding to the white dashed perimeter shown in (a). (c) Zoomed image of the white dashed perimeter displayed in (b) showing the roughness of the taper side walls.
3. Optical characterization of the DLSPPWs
We focus on Au based DLSPPWs (considered as reference waveguides) for demonstrating our characterization procedure. Our procedure aims at measuring the effective index neff of the mode and its damping distance LSPP commonly defined as the propagation distance leading to the decay of the mode intensity. The mode effective index is given by where k′SPP is the phase constant of the mode and if λ0 denotes the free-space wavelength of the incident light. Figure 3(a) shows a typical LR image recorded for a 700nm-wide Au based DLSPPW excited at a free-space wavelength of 1550nm. As expected, the intensity of the DLSPPW mode decays at the typical scale of a damping distance LSPP. In principle, LSPP can be extracted directly from a LR image[see Fig. 3(a)]. However, owing to slight aberrations affecting the quality of the images at the edges of the microscope field of view, we only use LR images for LSPP measurements when the propagation distance is too small to perform reliable fiber-to-fiber characterizations. In this work, the main interest of LR imaging does not reside in the observation of the image plane of the sample but in the characterization of its Fourier plane. The Fourier plane of the sample can be imaged by adding a third lens [L3 in Fig. 1] in the optical path. It has been recognized that the Fourier plane of a propagating DLSPPW mode can be used to measure the effective index of the mode [23]. Fig. 3(b) displays the Fourier plane corresponding to the image of Fig. 3(a). On this Fourier plane image, one can observe the angular distribution of the incident spot, the momentum transfer provided to the incident beam by the input coupler [noted G in Fig. 3(b)] and the vertical line originating from the Fourier transform of the field distribution related to the DLSPPW mode propagation. Note that Fourier plane imaging is a convenient way to check the ability of a grating coupler to excite a given DLSPPW mode. Indeed, the excitation of the DSLPPW mode is expected to be optimum if the first order scattering of the incident beam overlaps the wave-vector spectrum of the propagating mode as it can be seen in Fig. 3(b). When saturating the color scale of the Fourier plane image (not shown), a narrow gold/air interface SPP arc appears on the image. This interface mode which is not intentionally excited is however of great interest to achieve the calibration of the Fourier plane. For free-space wavelengths in the C+L telecom bands, the effective index of the interface SPP differs only little (typically by about 3×10−3 to 5×10−3) from the refractive index of the superstate (air in our case) owing to the extension of the interface SPP field within this medium. Thus, by attributing an effective index of 1.004 to the interface SPP arc, the effective index of the DLSPPW mode can be obtained with relative error below 0.5%. In the case of the DLSPPW mode considered in Fig. 3, we obtain an effective index of (1.207±0.006). Knowing the effective index of the DLSPPW mode, we need to perform an accurate measurement of its damping distance.
Fig. 3 (a) Leakage radiation microscopy image of a 700×290 nm2 cross-section Au-based DLSPPW excited at 1550nm. (b) Imaging of the Fourier plane corresponding to the image shown in (a). The momentum transfer provided by the input grating leads to an efficient excitation of the DLSPPW mode (see text).
The procedure for the measurement of LSPP on the basis of fiber-to-fiber data is illustrated in Fig. 4. An incoherent broadband light source is used as the optical source while the light coupled out of the waveguide is collected by the output fiber-focuser and analyzed by means of an optical spectral analyzer with a selected wavelength resolution of 1nm. The raw spectra recorded for DLSPPWs of increasing length (per step of 10μm) are superimposed in Fig. 4(a). The highest intensity spectrum is obtained for the shortest DLSPPW (50μm) considered as the reference waveguide. Normalizing the raw spectra by the reference spectrum, the power transmitted along the DLSPPWs can be plotted, at a given wavelength, as a function of ΔL, the difference in length of the DLSPPW of interest relative to the reference DLSPPW. Fig. 4(b) shows the result of this procedure plotted on a logarithmic scale for a wavelength of 1550nm. The experimental results are well fitted by a linear law. From the slope of this fit, we find that the damping distance for this DLSPPW is LSPP=42.5μm at 1550nm. Following the same procedure for different wavelengths [see Fig. 4(c)], the spectral dispersion of LSPP is found to be for the DLSPPW mode of interest. The reliability of the fiber-to-fiber measurements greatly benefit from the LR imaging of the samples since it allows us to locate accurately the incident spot onto the input coupler of the DLSPPW under test. By adjusting the incident spot location and by optimizing the power collected by the output fiber, we have checked that two DLSPPWs of same length [the longest DLSPPW appear twice on each pattern, see Fig. 2] lead to the same transmission spectrum.
Fig. 4 Fiber-to-Fiber characterization of 700×290 nm2 Au based DLSPPWs. The waveguides are excited by a broadband ASE source. (a) Raw transmission spectra recorded for waveguides with lengths ranging from 50 to 150 μm per step of 10 μm. (b) For a wavelength of 1550nm, intensity collected at the output of the waveguides of different length normalized by the output intensity of the shortest waveguide plotted on a logarithmic scale. The dashed line is a linear fit. (c) Spectral dispersion of the propagation distance of the DLSPPW mode.
4. CMOS metal based DLSPPWs: Cu, Al, TiN
In this section we discuss the performances of Cu, Al and TiN-based DLSPPWs in comparison with Au-based waveguides. The results discussed next should be understood as typical plasmonic performances that are expected to be relevant in the context of practical applications. In particular, we evaporate the thin films following standard processes. More importantly, we intentionally performed the optical characterizations of the CMOS based waveguides several days after the metal deposition rather than on freshly evaporated films in order to account for aging effects expected for reactive metals such as Cu, Al or Ag [27]. Indeed, although stored in clean room environment after their deposition (temperature 20°C, humidity: 40–50%), we emphasize that we did not try to preserve the metal films from oxidization. For the purpose of a qualitative observation of CMOS metal based DLSPPWs, we display in Fig. 5, the LR images of DLSPPWs implemented onto different metal layers and excited at a free-space wavelength of 1550nm. For all images, the cross–section of the waveguide is around 300×300nm2 leading to a long propagating DLSPPW mode. The visual inspection of those images indicates clearly superior propagation distances of Au-based DLSPPWs. However, the propagation distance of a DLSPPW mode being dependent on its effective index, a more quantitative analysis must be conducted.
Fig. 5 Leakage radiation microscopy image at 1550nm of DLSPPW with a nominal width (design width) of 300nm and implemented on (a) Au (b) Cu, (c) Al and (d) TiN.
Cu-based and Al-based DLSPPWs have been optically characterized following the procedure described in the previous section for Au-based waveguides. For TiN-based DLSPPWs, fiber-to-fiber measurements were not possible due to short propagation distances. In this case, LR images were used for the evaluation of the damping distance with the non-linearity of the infrared camera corrected by means of a calibration procedure. Note that in the case of short propagation distances (LSPP around 15–20 μm) significantly smaller than the field of view of our microscope [see Fig. 5(d)], LR images are not subject to geometrical aberration and thus are convenient for the measurement of LSPP. The results of LSPP measurements obtained for DLSPPWs based on different metals as a function of the effective index of the guided mode are plotted in Fig. 6 for a free-space wavelength of 1550nm. As expected from DLSPPW properties, we note that LSPP increases for decreasing effective index whatever the metal. For quantifying the plasmonic properties of the CMOS metal based DLSPPWs, a careful analysis of the origins of the damping of the plasmon mode is needed. The energy dissipation of a DLSPPW mode originates from three dominant loss channels: (i) light absorption within the metal layer (ohmic losses), (ii) leakage radiations (LR) in the substrate through the metal layer (due to the leaky nature of the mode) and (iii) plasmon-photon decoupling due to waveguide surface roughness scattering and/or absorption within the polymer material. It is clear that only the intrinsic ohmic losses contributing to k″ohm are characteristic of the plasmonic properties of the metal based DLSPPWs. The two other contributions depend on the configuration and waveguide quality and can be viewed as artifacts. For the evaluation of the ohmic losses of the CMOS metal based DLSPPW, one has to separate the contribution of the intrinsic losses from the contribution of the two other loss channels. The damping distance of a DLSPPW mode can be expressed as where k″SPP denotes the attenuation constant of the mode. It is convenient to note the attenuation constant as k″SPP = k″Ω + k″LR + k″scat where each term in the sum refers to the respective contribution of the aforementioned loss channels and where we omit the dependence of each contribution upon the effective index neff of the mode. The impact of the LR losses on LSPP can be readily controlled by adjusting the thickness of the metal film. As discussed in further details in section 5, for all the samples we consider, the thickness of the metal layer, regardless the metal of interest, is chosen such that LSPP is not lower than 95% of the saturation damping distance obtained for an infinitely thick metal film (corresponding to vanishing leakage radiation losses). When the saturation condition of LSPP is verified, the contribution of LR losses can be safely neglected. The contribution of the scattering (and/or polymer absorption) loss channel is more difficult to evaluate and can be non-negligible for a good plasmonic metal featuring small ohmic losses. In this case, the damping distance LSPP of the DLSPPW is not strictly indicative of the plasmonic properties of the metal layer but more on the optical quality of the DLSPPWs. To overcome this limitation, we make the reasonable assumption (verified a posteriori in section 5) that k″scat does not depend on the metal since this contribution is directly related to the quality of the waveguide and not to the interaction of the DLSPPW mode with the metal layer. With this assumption (and neglecting LR losses), the attenuation constant of the DLSPPW mode can be noted k″SPP(metal) = k″Ω(metal) + k″scat. The plasmonic properties of a CMOS metal based DLSPPW relative to gold are now obtained by computing the difference which only depends on the ohmic losses of the CMOS metal based DLSPPWs and gold based DLSPPWs respectively but no more on the scattering channel. The physical interpretation of is straightforward by noting that the ratio of the DLSPPW mode intensity traveling in the x direction along a CMOS based waveguide to a Au-based waveguide is given by:
The ohmic loss rate (OLR) quantifying the excess of ohmic loss per micron for a given CMOS based DLSPPW mode relative to the corresponding Au based mode (of same effective index) in given by: where are the experimentally measured damping distances (expressed in microns) for the CMOS and Au based DLSPPWs respectively. Although the damping distances are a priori affected by the scattering loss channel, the ohmic loss rate is characteristic of the ohmic damping of the CMOS DLSPPW mode relative to the corresponding Au based DLSPPW mode. The is then a convenient parameter to quantify the plasmonic performances of CMOS metal DLSPPWs. From a practical point of view, is of interest for anticipating the damping distance of a CMOS metal based waveguide knowing the properties of the corresponding Au based DLSPPW. The ohmic loss rates given in table 2 are computed on the basis of the experimental results displayed is Fig. 6. In the range of effective index accessible with DLSPPWs, the OLR rate relative to gold for each CMOS metal fluctuates only moderately. The average ohmic loss rate indicates that for the Cu based DLSPPW, the mode intensity is reduced by a factor 2 (3dB) as compared to gold DLSPPW after a propagation distance of about 110 μm which is more than twice the typical damping distance . Thus for DLSPPWs, although we did not try to prevent oxidation, we observe that Cu performs almost as good as Au and can be considered as a good low cost alternative plasmonic metal at the telecom wavelengths. For Aluminum, the extra-loss rate is about 6.5 times larger than for Cu whereas for TiN the loss rate is more than 19 times larger than for Cu. Hence, Aluminum based DLSPPWs can only be used in configurations where extra-losses relative to Au of about 3dB for propagation lengths around 15 μm are sustainable. Finally, we note that for TiN, the situation is even more critical. Although some propagation is observed (for what we believe to be the first time) along TiN-based DLSPPWs, the intrinsic damping of TiN is too strong and limits severely the range of realistic applications that can be targeted with TiN based DLSPPWs. It is known that the properties TiN depend dramatically on the deposition conditions [28]. As discussed in details in the next section, the TiN layers used for our experiments are expected to be at the current state of the art in terms of optical properties.Table 2. Ohmic loss rate relative to gold for each CMOS metal for increasing effective index of the DLSPPW mode. At neff =1.2 for TiN, the damping distance needed for the computation of the is extrapolated from the experimental data shown in Fig. 6.
Fig. 6 Damping distances as a function of effective index of DLSPPWs. Dashed lines are guidelines to the eyes.
5. Ellipsometric characterizations and comparison with DLSPPWs properties
In this section, we investigate the plasmonic properties of our metal layers extracted from ellipsometric measurements and we compare those results to experimental data obtained directly with the DLSPPWs. Ellipsometry has been recently used to study the optimum deposition conditions of plasmonic metals such as Au, Ag, Cu and Al [12]. Unlike the ellipsometric measurements performed in [12], our ellipsometric data were recorded on the face of the metal layer exposed to air and featuring the RMS roughness given in table 1. The optical constants of the materials were determined by variable-angle spectroscopic ellipsometry (J.A. Woollam VASE). The ellipsometric measurement was performed in the spectral range of 500–2300nm (10nm steps) with incident angles of 75°, 80° and 85°. The optical model consisted of ambient air, a surface roughness layer and the material of interest. The roughness layer was modeled on the basis of the Bruggemann effective medium approximation assuming 50% metal and 50% air. The thicknesses of the roughness layer were fixed to the values determined by AFM measurements (table 1). The complex dielectric function of the metal was parametrized by applying the Drude-Lorentz oscillator model:
with the angular frequency ω and the dielectric constant at infinite frequency ε∞. The Drude term is characterized by the plasma frequency ωp and the damping frequency γd. The Lorentz oscillators are described by the amplitudes Aj, the damping frequencies γj (broadening) and the center frequencies ω0,j. The metal layer was considered as an infinitely thick substrate. The optical constant of the material of interest was retrieved by fitting the Drude-Lorentz model to the ellipsometric data. The results of the Drude-Lorentz fits of the complex dielectric function are compiled in table 3 and the corresponding dielectric functions are displayed in Fig. 7.Table 3. Fitting parameters of the spectroscopic ellipsometric measurements
Fig. 7 Dielectric function for the metal of interest extracted from spectroscopic ellipsometry measurements. The solid lines correspond to the imaginary part of the dielectric function ε″. The dashed lines show the real part of the dielectric function ε′.
At the free-space wavelength of 1550nm, the dielectric function of the different metals are given in table 4. By comparing the experimental DLSPPW properties to the properties modeled on the basis of the experimental dielectric functions, we can check the reliability of the ellipsometric data fitting procedure. We use the effective index method (EIM) for modeling the DLSPPWs. Although approximate, this method has been proven to be convenient for DLSPPW at least for sufficiently wide waveguides [15]. The remaining unknown for applying the EIM is the refractive index of the exposed SU-8. The actual refractive index of the resist is of moderate interest in this context because this parameter cannot account for the scattering loss originating from the roughness of DLSPPW side walls. In this study, we adopt a practical point of view and we attribute in our model an effective extinction coefficient for the resist by direct comparison with the Au-based DLSPPW experimental results. The real part of the refractive index of SU-8 at 1550nm is around n=1.57 [29]. By fixing n at 1.57 and by taking the extinction coefficient at k=1×10−3 the results displayed in Fig. 8 are obtained. In Fig. 8(b), the effective indexes of the DLSPPW modes returned by the EIM are in fair agreement with the experimental ones whereas in Fig. 8(c) the agreement between computed and experimental values for the damping distance is restricted to the widest waveguides. The same trend is observed in Fig. 8(d) showing the damping distance against the effective index of the DLSPPW modes. This behavior does not arise from the approximate nature of the EIM since similar results are obtained with rigorous finite element modeling. One can identify two origins for the mismatch between the modeled and experimental damping distances at low effective index. First, we note that for small width-to-height aspect ratio waveguides, the SU-8 DLSPPWs feature rounded edges deviating significantly from a rectangular cross-section. Second and more importantly, the extinction coefficient attributed to the SU-8 resist to account for the scattering loss channel along the waveguide is likely to depend on the modal field distribution of the DLSPPW mode. As a result, the agreement between the experimental and the computed damping distances can only be obtained in a limited range of waveguide width. In all the following, the comparison between computed and experimental results will be restricted to the highest effective index DLSPPW modes. For the 800nm-wide DLSPPW corresponding to the widest single-mode DLSPPW, the computed damping distance increases by less than 1.5% if the thickness of the Au film is increased from 87nm to 200nm. This result indicates a posteriori that for the experimental Au film thickness of 87nm, the leakage radiation losses of the DLSPPW mode are negligible compared to the other loss channels (ohmic+scattering), a condition that is obviously mandatory for extracting relevant information about the plasmonic performances of the underneath metal from DLSPPW properties. The saturation condition of the damping distance is also verified for all CMOS metal based samples considered in this work.
Table 4. Refractive index and corresponding dielectric function at 1550nm for each metal of interest.
Fig. 8 (a) Configuration for the modeling of the Au based DLSPPWs. hSU8=290nm is fixed at the experimental value of 290nm measured by atomic force microscopy. The thickness of gold is 87 mn and the dielectric function of gold is given in table 4. (b) Comparison of the experimental and modeled effective index of the DLSPPW mode as a function of the waveguide width. (c) Comparison of the experimental and modeled damping distances as a function of the DLSPPW width. (d) Change of damping distance of the Au based DLSPPW mode as a function of its effective index.
The Au layer is then replaced by a CMOS metal film with a thickness equal to the experimental one. The damping distances [plotted in Fig. 9] as a function of the DLSPPW mode effective index have been obtained by using the dielectric functions given in table 4. For Cu, we note once again a good agreement between EIM results and experimental data for large effective index modes. This agreement is clearly degraded when considering Al results since the asymptotic damping distance computed for large effective index modes is more than 25% larger than the experimental value. For TiN, the agreement is again only fair with a relative error between computed and experimental values of LSPP for the highest index in the range of 30%. However, owing the very short propagation distance of the TiN-based DLSPPW modes, the deviation between the computed and experimental damping distance is only 1.5μm within the accuracy of our LR image based measurement procedure of LSPP. From the results compiled in Figs. 9(a)–(c), we note that a significant deviation between experimental and computed results occurs for Al, the metal featuring the largest surface roughness (3.6 nm RMS). One could suggest that the discrepancy between ellipsometric and DLSPPW based characterizations of Al thin films is due to the growing of an alumina layer at the surface of the metal film. However, this alumina layer (typical thickness of 2–3 nm) is expected to grow during the first hours after the metal film is exposed to the ambient conditions and remains stable at a scale of at least one month [31]. The self-protective alumina layer is thus unlikely to explain the difference between ellipsometric and DLSPPW results that we attribute to the roughness of the metal film, although the fitting procedure of our ellipsometric data did account for such a roughness. In fact, the imaginary part of the dielectric function of Al seems to be slightly underestimated. By increasing ε″ of Al by 25% from 43 to 54 (at 1550nm) returns a EIM asymptotic value of LSPP only 1.2μm above the experimental value.
Fig. 9 Comparison of modeled (dashed line) and experimental (dots) characteristics of the CMOS metal based DLSPPWs. For each metal, the thickness of the metal layer and the polymer waveguide (hSU−8) used for the modeling are equal to the experimental values (a) Copper (hSU−8=290nm), (b) Aluminum (hSU−8=295nm), (c) TiN (hSU8=295nm). The modeled data are obtained using the dielectric function of TiN given in table 4. (d) Comparison of experimental and computed data by using the dielectric function of TiN extracted from [5] for the dashed curve (EIMA) and from Ref [30] for the dash-dotted curve (EIMB).
Our approach for modeling the scattering losses based on the extinction coefficient of the SU-8 resist gives us the possibility to quantify the contribution of this channel to the damping distance the waveguides. If we denote as and the damping distances obtained when the extinction coefficient of SU-8 is respectively set to k=1×10−3 and k=0, the damping distance due to the scattering loss channel alone is given by:
By applying eq. 4 to a 800nm-wide wide DLSPPW (sustaining a DLSPPW mode with an effective index around 1.2), we find , and . This result supports the assumption used in section 4 stating that for a given DLSPPW mode, the scattering losses do not depend significantly on the plasmonic metal. Finally, our results indicate that TiN performs rather poorly for plasmon modes of intermediate effective index (in the range 1.05–1.2). However, given that the optical properties of TiN layers are highly dependent on the deposition conditions, it is necessary to verify our conclusion for TiN dielectric functions recently reported in the literature. At a free-space wavelength of 1550nm, it has been found that TiN deposited by sputtering has a dielectric function of ε =−72+i22 [5] whereas it changes to ε =−22+i40 [30] when an atomic layer deposition (ALD) set-up is operated. Those dielectric functions have been used with the EIM to obtain the damping distance curves plotted in Fig. 9(d) and denoted EIMA and EIMB respectively. By comparison with our experimental results, we note that recently discussed TiN fabrication optimization processes lead to either degraded or marginally improved performances of the DLSPPWs, suggesting once again that TiN can only be hardly implemented with DLSPPW modes. However, if the plasmonic mode of interest is of low effective index (such as in the case of interface SPP), TiN can still be of some practical interest, in particular at telecom frequencies, at the cost of further optimization of its optical properties.6. Conclusion
In summary, we have characterized the plasmonic properties of three CMOS compatible metals Cu, Al and TiN at telecom wavelengths in comparison with Au considered as the reference metal. Our approach relies on a plasmonic waveguiding platform comprised of polymer waveguides loaded on top of each metal film. By operating simultaneously a leakage-radiation microscope and a fiber-to-fiber configuration mimicking the standard cut-back method, we have measured the damping distance of the dielectric loaded surface plasmon waveguide modes together with their effective index. This approach is only applicable to Au, Cu and Al leading to DLSPPW modes with sufficient propagation length. In the case of TiN, because of their short propagation distances, the DLSPPWs are investigated by the analysis of their leakage radiation images.
For the quantitative comparison of the CMOS metals plasmonic properties, we define the ohmic loss rate (OLR) for each DLSPPW mode. The OLR is characteristic of the ohmic loss contribution of each CMOS metal based waveguide mode relative to the corresponding mode sustained by the Au based DLSPPW. This OLR can be computed directly from the damping distances measured experimentally. At a free-space wavelength of 1550nm, we show that the ohmic losses experienced by the DLSPPW mode implemented on Cu, Al and TiN are on average increased respectively by 0.027 dB/μm, 0.18 dB/μm and 0.52 dB/μm relatively to gold. Those ohmic loss rates are expected to be rather insensitive to the fabrication quality of the DLSPPWs.
In addition to the optical characterizations of the DLSPPWs, the metal films have been investigated by spectroscopic ellipsometry. Knowing the thickness and the roughness of each film measured by atomic force microscopy, the dielectric function of each metal has been obtained by fitting the ellipsometric data with a model taking into account the roughness of each layer. The experimental dielectric functions have been used to model the DLSPPWs with the effective index method. For low roughness films (Cu, TiN) the experimental and computed properties of the DLSPPWs are in fairly good agreement whereas significant deviations are observed for rough layers such as Al. Finally, we show that, TiN-based DLSPPWs feature very short propagation lengths (in the range of only few free-space wavelengths) making the use of those modes rather unrealistic in the context of true practical applications.
Funding
H2020-EU.2.1.1 project PlasmoFab grant number 688166.
References and links
1. 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, 795–808 (2010). [CrossRef]
2. A. Boltasseva and H. A. Atwater, “Low-loss plasmonics metamaterials,” Science 331, 290–291 (2011). [CrossRef] [PubMed]
3. G. V. Naik, V. M. Shalaev, and A. Boltasseva, “Alternative plasmonic materials: Beyond gold and silver,” Adv. Mater. 25(24, 3264–3294 (2013). [PubMed]
4. R. Zektzer, B. Desiatov, N. Mazurski, S. I. Bozhevolnyi, and U. Levy, “Experimental demonstration of cmos-compatible long-range dielectric loaded surface plasmon waveguides (lr-dlsppws),” Opt. Express 22, 22009–22017 (2014). [CrossRef] [PubMed]
5. N. Kinsey, G. V. Ferrera, M. Naik, V. E. Babicheva, V. M. Shalaev, and A. Boltasseva, “Experimental demonstration of titanium nitride interconnects,” Opt. Express 22, 12238–12247 (2014). [CrossRef] [PubMed]
6. C. Delacour, S. Blaize, P. Grosse, J.-M. Fedeli, A. Bruyant, R. Salas-Montiel, G. Lerondel, and A. Chelnokov, “Efficient directional coupling between silicon and copper plasmonic nanoslot waveguides: Toward metal-oxide-silicon nanophotonics,” Nano Lett. 10, 2922–2926 (2010). [CrossRef] [PubMed]
7. D. Y. Fedyanin, D. I. Yakubovsky, R. V. Kirtaev, and V. S. Volkov, “Ultralow-loss cmos copper plasmonic waveguides,” Nano Lett. 16, 362–366 (2016). [CrossRef]
8. H. S. Lee, C. Awada, S. Booutami, F. Charra, L. Douillard, and R. Espiau de Lamaestre, “Loss mechanisms of surface plasmon polaritons propagating on a smooth polycrystalline cu surface,” Opt. Express 20, 8974–8981 (2012). [CrossRef] [PubMed]
9. D. Steinmüller-Nethl, R. Kovacs, E. Gornik, and Rödhammer, “Excitation of surface plasmon on titanium nitride films: determination of the dielectric function,” Thin Solid Films 237, 277–281 (1994). [CrossRef]
10. N. C. Chen, W. C. Lien, R. C. Liu, Y. L. Huang, Y. R. Lin, C. Chou, S. Y. Chang, and C. W. Ho, “Excitation of surface plasma wave at tin/air interface in the kretschmann geometry,” J. Appl. phys. 109, 043104 (2011). [CrossRef]
11. H. Gao, J. Henzie, M. H. Lee, and T. W. Odom, “Screening plasmonic materials using pyramidal gratings,” PNAS 105, 20146–20151 (2008). [CrossRef] [PubMed]
12. K. M. McPeak, S. V. Jayanti, S. J. P. Kress, S. Meyer, S. Iotti, A. Rossinelli, and D. J. Norris, “Plasmonic films can easily be better: Rules and recipes,” ACS Photonics 2, 326–333 (2015). [CrossRef] [PubMed]
13. A. Lehmuskero, M. Kuittinen, and P. Vahimaa, “Refractive index and extinction coefficient dependence of thin al and ir films on deposition technique and thickness,” Opt. Express 15, 10744–10752 (2007). [CrossRef] [PubMed]
14. B. Steinberber, A. Hohenau, H. Ditlbacher, A. L. Stepanov, A. Drezet, F. Aussenegg, A. Leitner, and J. R. Krenn, “Dielectric stripes on gold surface as surface plasmon waveguides,” Appl. Phys. Lett. 88, 094104 (2006). [CrossRef]
15. T. Holmgaard and S. I. Bozhevolnyi, “Theoretical analysis of dielectric-loaded surface plasmon-polariton waveguides,” Phys. Rev. B 75, 245405 (2007). [CrossRef]
16. A. V. Krasavin and A. V. Za, “Passive photonic elements based on dielectric-loaded surface plasmon polariton waveguides,” Appl. Phys. Lett. 90, 211101 (2007). [CrossRef]
17. S. Papaioannou, D. Kalavrouziotis, K. Vyrokinos, J.-C. Weeber, K. Hassan, L. Markey, A. Dereux, A. Kumar, S. I. Bozhevolnyi, M. Bauss, T. Tekin, D. Apostolopoulos, H. Avramopoulos, and N. Pleros, “Active plasmonics in wdm traffic switching applications,” Sci. Rep. 2, 652 (2012). [CrossRef] [PubMed]
18. M. G. Nielsen, J.-C. Weeber, K. Hassan, J. Fatome, C. Finot, S. Kaya, L. Markey, O. Albrektsen, S. I. Bozhevolnyi, G. Millot, and A. Dereux, “Grating couplers for fiber-to-fiber characterizations of stand-alone dielectric loaded surface plasmon waveguide components,” J. Light. Technol. 30, 3118–3125 (2012). [CrossRef]
19. J. Gosciniak, S. I. Bozhevolnyi, T. B. Andersen, V. S. Volkov, J. Kjelstrup-Hansen, L. Markey, and A. Dereux, “Thermo-optic control of dielectric-loaded plasmonic waveguide components,” Opt. Express 18, 1207–1216 (2010). [CrossRef] [PubMed]
20. J.-C. Weeber, K. Hassan, L. Saviot, A. Dereux, C. Boissière, O. Durupthy, C. Chaneac, E. Burov, and A. Pastouret, “Efficient photo-thermal activation of gold nanoparticle-doped polymer plasmonic switches,” Opt. Express 20, 27636–27649 (2012). [CrossRef] [PubMed]
21. K. Hassan, F. Leroy, G. C. des Francs, and J.-C. Weeber, “Dihedron dielectric loaded surface plasmon athermal polarization converter,” Opt. lett. 39, 697–700 (2014). [CrossRef] [PubMed]
22. A. Drezet, A. Hohenau, D. Koller, A. Stepanov, H. Ditlbacher, B. Steiberger, F. Aussenegg, A. Leitner, and J. Krenn, “Leakage radiation microscopy of surface plasmon polaritons,” Mat. Sci. Eng. B 149, 220–229 (2008). [CrossRef]
23. J. Grandidier, S. Massenot, G. Colas-des Francs, A. Bouhelier, J.-C. Weeber, L. Markey, A. Dereux, J. Renger, M. U. Gonzàlez, and R. Quidant, “Dielectric loaded surface plasmon waveguides: Figures of merit and mode characterization by image and fourier plane leakage microscopy,” Phys. Rev. B 78, 245419 (2008). [CrossRef]
24. A. V. Krasavin and A. V. Zayats, “Active nanophtonic circuitry based on dielectric-loaded plasmonic waveguides,” Adv. Optical Mater. 3, 1662–1690 (2015). [CrossRef]
25. P. Patsalas, C. Charitidis, S. Logothetidis, C. A. Dimitriadis, and O. Valassiades, “Combined electrical and mechanichal properties of titanium nitride thin films as metalization materials,” J. Appl. Phys. 86, 5296–5298 (1999). [CrossRef]
26. Y. L. Jeyachandran, S. K. Narayandass, D. Mangalaraj, S. Areva, and J. A. Mielczarski, “Properties of titanium nitride films prepared by direct current magnetron sputtering,” Mat. Sci. Eng. A 445–446, 223–236 (2007). [CrossRef]
27. P. Dawson, B. A. F. Pyugranier, and J.-P. Goudonnet, “Surface plasmon polariton propagation length: A direct comparison using photon scanning tunneling microscopy and attenuated total refelction,” Phys. Rev. B 63, 205410 (2001). [CrossRef]
28. V. Naik, J. L. Schroeder, X. Ni, A. V. Kildishev, T. D. Sa, and A. Boltasseva, “Titanium nitride as a plasmonic material for visible and near-infrared wavelengths,” Opt. Mater. Express 2, 478–489 (2012). [CrossRef]
29. X.-b. Wang, J. Sun, C.-M. Chen, X.-Q. Sun, F. Wang, and D.-M. Zhang, “Thermal uv treatment on su-8 polymer for integrated optics,” Opt. Mater. Express 4, 509–517 (2014). [CrossRef]
30. J. A. Briggs, G. V. Naik, T. A. Petach, B. K. Baum, D. Goldhaber-Gordon, and J. A. Dionne, “Fully cmos-compatible titanium nitride nanoantennas,” Appl. Phys. Lett. 108, 4941413 (2016). [CrossRef]
31. C. Langhammer, M. Schwind, B. Kasemo, and I. Zorić, “Localized surface plasmon resonances in aluminum nanodisks,” Nano. Lett. 8, 1461–1471 (2008). [CrossRef] [PubMed]
