Compact fiber-coupled dielectric-loaded plasmonic Mach-Zehnder interferometers operating at telecom wavelengths and controlled via the thermo-optic effect are reported. Two fabricated structures with Cytop substrate and a ridge made of PMMA or a cycloaliphatic acrylate polymer (CAP) were considered showing low switching power of 2.35 mW and switching time in the range of microseconds for a CAP ridge and milliseconds switching time for a PMMA ridge. Full output modulation is demonstrated for the structure with a CAP ridge and 40% modulation with a PMMA ridge.
©2012 Optical Society of America
Integrated optical devices and circuits are being increasingly used for light routing and switching in the rapidly developing area of broadband optical communications . Such devices are traditionally based on the guiding of light in a dielectric waveguide consisting of a core and a cladding, with the refractive index of the former being larger than that of the latter. Electromagnetic radiation propagating in and confined to the core (by virtue of total internal reflection) in the form of waveguide modes can be controlled with externally applied electrical signals via, for example, electro-, magneto-, and thermo-optic effects (depending on the dielectric properties and electrode configuration). Note that the controlling electrodes being usually metallic produce additional loss of radiation due to absorption. The effect of absorption could be minimized by increasing the electrode-waveguide separation but that would decrease the aforementioned (useful) effects as well, a circumstance that makes the positioning of electrodes in conventional waveguide modulators and switches a challenging design problem . Compared to dielectric waveguides, plasmonic devices can guide and manipulate optical signals on a subwavelength scale and below the diffraction limit of light. Among various SPP-based waveguide configurations, dielectric-loaded SPP waveguides (DLSPPWs)  represent an attractive alternative by virtue of being naturally compatible with different dielectric and industrial fabrication processes using large-scale UV lithography. DLSPPWs satisfy the important requirements of strong mode confinement, relatively low propagation loss, and straightforward integration with control electrodes enabling a thermo-optic control. The main advantage of the plasmonic technology is that gold stripes can be used both as supports of DLSPPWs and electrodes allowing for heating the plasmonic devices where the DLSPPW mode field reaches its maximum at the metal-dielectric interface [4–6]. The ability to transmit light and electricity in the same component simultaneously opens up interesting possibilities for hybrid electrical-optical integration .
The main problems in plasmonic technology are high losses related with metal-induced attenuation. This impact can be minimized by integration of short plasmonic waveguides with longer dielectric waveguides. In this way, the small size and low power switching capabilities of plasmonic can be blended with the low loss of dielectric waveguides and processing capacity of electronics, to provide miniaturized and power efficient photonic interconnect routers. Additionally, high propagation losses in the DLSPPWs due to absorption in metal are transferred into heat, which can be used to monitor the amount of power propagating in a plasmonic mode by measuring variations in the resistance of metal stripes .
Part of the challenge in developing the DLSPP based technology is minimizing the coupling losses from photonics platform to plasmonic waveguide. It was shown previously [9,10] that coupling losses of 3 dB was achieved in the first tries and it can still be further reduced to less than 1 dB by better design of photonic waveguide. The high coupling efficiency was confirmed experimentally  where the coupling loss was found to be 1 dB per transition between the SOI and DLSPP waveguides. The next challenge is to obtain higher modulation and reduction of power requirements as well as a decrease of switching time.
2. Fabrication and operation principle
The Mach-Zehnder interferometer is the most extensively studied thermo-optic switch because of its simplicity in design and fabrication and of the presence of the reference arm which is useful for compensation of common-mode effects. There are four main parameters which allows one to evaluate the MZI performance: switching time, power switching, modulation depth (visibility) and extinction ratio. High-performance MZI should present high-frequency operation (fast switching time), low power swiching requirement, 100% modulation and large extinction ratio.
The substrate used was standard Si wafer that was covered with a spin-coated ~1.7μm-thick Cytop realized using a procedure similar to those reported in ref . Cytop grade CTL-809M and the corresponding solvent CT-SOLV180 used for dilution were obtained from AGC Chemicals Europe Ltd. The other processing resists used: AZnLOF 2070 and LOR-A were from Microchemicals Gmbh and MicroChem Corp., respectively. The investigated DLSPPW-based Mach-Zehnder interferometers (Fig. 1 ) were fabricated by a UV lithography process using a Süss Microtech MJB4 mask aligner in the vacuum contact mode and using consecutively two (commercial) masks. In the first step, a bilayer i-line (365nm) lithography process using AZnLOF resist (the imaging layer) on top of a LOR-A underlayer resist (facilitating the lift-off) followed by gold evaporation and lift-off in N-methyl-2-pyrrolidone was used to pattern the 50-nm-thick gold electrodes deposited on Cytop. In order to get a good adhesion of LOR on Cytop, we introduced a short oxygen plasma ashing step prior to LOR coating. In the second step, a spin-coated layer of polymer resist: PMMA or a cycloaliphatic acrylate polymer (CAP) was exposed through the second mask at a wavelength of 250nm and developed, thus defining the dielectric (polymer) ridges of the DLSPPW circuitry. The polymer ridges were ~1μm-wide in principal regions of the components and were on both sides connected via 25μm-long funnel structures with access (10μm-wide) polymer waveguides extending outside gold covered regions of DLSPPWs all the way up to the substrate edges (Fig. 1(a)), facilitating thereby the end-fire coupling of photonics waveguide with single-mode tapered optical fibers.
The final fabrication step was a cleavage of the sample perpendicular to the photonics waveguide resulting in ~2mm-long and ~10μm-wide photonic waveguide leading toward each side of the DLSPPW area. Fabricated MZI modulators were similar to some presented previously  based on the same mask design. The gold film was patterned so that gold stripes could be used both as DLSPPW supports and electrodes allowing for heating (with electrical currents) one MZI arm that was electrically isolated from the other arm by 1.5μm-wide gaps. Ultrasonic wire bonding was used to connect aluminum wires to ~200-μm-wide and 500μm-long bonding pads of the MZI modulator on the sample substrate and pre-structured electrodes on a carrier substrate. The overall resistance was measured to be ~17 Ω, with the heating (4μm-wide) electrode resistance being evaluated at ~5 Ω .
Compared to previously reported studies , magnesium fluoride MgF2 (n = 1.37) was replaced by another low refractive index material Cytop (n = 1.34), having a thermal conductivity much lower compared to MgF2. The previous studies showed that for a material with high thermal conductivity, heat dissipation mainly takes place in the substrate, which worsens the overall performances of the switches i.e. higher power consumption and lower switching speed. It has to be emphasized that low refractive index substrate material is required to ensure efficient guiding by photonic waveguides.
Two sets of samples were fabricated. The first one with a square PMMA ridge (w = h = 1 μm) (Fig. 1(b)) and the second one with a rectangular cycloaliphatic acrylate polymer (CAP) ridge (w = 1 μm, h = 0.6 μm) (Fig. 1(c)). Spectroscopic ellipsometry measurements showed that CAP films has a significantly higher thermo-optic coefficient TOC (∂n/∂T = −3·10−4 1/K) and refractive index at telecom wavelengths (n = 1.53) in comparison with PMMA films (∂n/∂T = −1·05−4 1/K and n = 1.493).
Fabrication of complex photonic waveguides integrated with plasmonic platform usually requires few lithography steps with very complex alignment setup. In our design, to ensure an easy fabrication procedure (one lithography step) the same polymer ridge (PMMA or CAP) was used as a part of photonics and plasmonics waveguides. To have a good confined photonics mode the thickness of the PMMA ridge (refractive index n = 1.49) should not be lower than 1 μm. For the CAP ridge (refractive index n = 1.53), however, the thickness of the polymer can be reduced to the 0.6 μm due to the higher refractive index of polymer.
The length of the heated MZI arm required to ensure the complete modulation is given by Lπ = λ/(2ΔTmax|∂n/∂T|), where λ is the light wavelength, ΔTmax is the maximum change in temperature determined by the difference between the glass transition temperature (Tg) of the polymer and the ambient temperature, |∂n/∂T| is the magnitude of the TOC. For PMMA, Tg ≈375 K corresponding to a practical maximum change in temperature ΔTmax = 65 K (during the experiments, a safety margin was adopted on applied temperature in order to prevent the structure from melting) and Lπ = 113.5 μm for λ = 1550 nm. Thus it is obvious that for the 46μm-long heated PMMA MZI arm one cannot expect the complete modulation. The maximum attainable modulation is ~40%. Compared to PMMA, CAP ensures around 3 times higher TOC and at the same time a higher Tg estimated at around 423 K, putting the length requirements for complete modulation to Lπ = 32.3 μm.
3. Modulation depth and extinction ratio
The light coupled to the dielectric waveguide from polarization maintaining tapered fiber was modulated in the plasmonic section by external driven voltage and then collected from the opposite side by the another polarization maintaining fiber and analyzed by a fiber-pigtailed ultrafast and high sensitive photoreceiver.
The operation of a thermo-optic MZI is based on changing the mode propagation constant in a heated arm resulting in the phase difference of two DLSPP modes interfering at the output Y-junction. The length of the heated MZI arm required to ensure complete modulation i.e. switching the light off in the Y-junction, is related to the phase difference π between the armsFig. 2 ). However, it can be realized with the CAP-loaded structure (Fig. 3 ) where the power needed to switch from maximum to minimum transmission was found to be Pπ ~8 mW with 20-65 μs switching time and Pπ ~9.8 mW for overdriven power resulting in 15-20 μs switching time and for the overall measured resistance of R = 17 Ω (metal electrodes, wires, electrode pads etc.). It does not represent however the true power requirements of the device, since the resistance of the heated arm of the MZI was found to be R ≈5 Ω and accounts only for 29.5% of the overall resistance. Base on this one can deduce a real switching power consumption of 2.35 mW and 2.8 mW respectively.
The performance of the MZI can be evaluated by using a visibility factor (modulation depth) which in the case of symmetric MZI should be as close to 1 as possible. The visibility can be found in the equation describing the transmission at the output of MZI:Fig. 3(b)) TOFF = 0.15 and TON = 0.59, giving a visibility factor V1 ≈0.59. The calculated beam transmission ratio k = 0.22 corresponds to a power difference between the two MZI arms of ~-6.4 dB. The calculations reveal that coupling losses of the plasmonic mode through the gap are ~-3.1 dB, so for two gaps we have losses of ~6.2 dB, which is in very good agreement with measurements. Thus the presence of the gaps fully accounts for the reduction of the visibility factor of the CAP-ridge MZI from VCAP≈1 (full modulation) to V1 = 0.59. To achieve a good modulation it is necessary to introduce a π-phase shift between two MZI arms and to get two interfering beams with the same intensity.
For the PMMA-ridge MZI structure the π-phase shift requirement cannot be fulfilled as was shown before. The calculated visibility factor V2 = 0.24 is much smaller compared to CAP (V1 = 0.59). Taking both these numbers into account it is possible to evaluate the visibility factor of the PMMA MZI structure related only to phase modulation VPMMA≈(V2∙VCAP)/V1 ≈0.40 where VPMMA and VCAP are the visibility factors obtained under the assumption of equal intensity in both MZI arms, and V1 and V2 are the experimental visibility factors of respectively the CAP and PMMA structures, where a ~6 dB power difference between the MZI arms is observed and, in consequence, evaluate a modulation depth VPMMA ~40% showing very good agreement with the calculated one.
Performances of different MZI can be compared, as well, in terms of extinction ratio (ER):
4. Switching on/off time and frequency response
The devices responses are shown in Fig. 2(b) and 3(b) for various values of applied driving power for PMMA and CAP ridges respectively. For MZI with a PMMA ridge, a 10% to 90% rise time of ~30 ms (switch off voltage) and 90% to 10% fall time of ~12 ms (switch on voltage) is observed, corresponding to a −3 dB frequency cutoff of ~26 Hz (Fig. 4(a) ).
Those results are much better than the previously reported where the switching on and off times was into seconds regime , however still far away from expectation. Significant improvement in the switching performance of the device was observed with CAP ridge. A rise time of ~65 μs (switch on voltage) and a fall time of ~20 μs (switch off voltage) were observed with a −3 dB frequency cutoff of ~15 kHz (Fig. 4(b)) for MZI power consumption of 2.35 mW. Further improvement in terms of the switching time can be achieved for overdriven power. So, for the power consumption of 2.8 mW the rise time of ~20 μs and fall time of ~15 μs were observed what corresponds to a −3 dB frequency cutoff of ~50 kHz (Fig. 4(b)).
The smaller ridge height (w = 0.6 µm) slightly contributes to the switching time reduction, but mostly, improvements can be attributed to the thermal properties of CAP. The specific heat coefficient which characterizes the amount of heat required to change the material’s temperature, is for CAP about 270 times smaller than for PMMA.
By applying the voltage to the electrodes pads the flow of an electric current through a metal releases heat which is then transferred to any materials in thermal contact with the metal. The amount of heat transferred to the ridge and the substrate depends on the surrounding materials thermal conductivity, contact area and the length of a dissipated power. For a square PMMA ridge (w = h = 1 µm) and h = 1.7 µm thick Cytop layer, only 10% of the dissipated power from the 4 µm wide gold electrode is transferred to the PMMA ridge. The coupling efficiency can be improved by a factor 2 if using a narrower gold stripe supporting the ridge, which could further decrease the switching times.
It has to be emphasized that fall time is associated with heating up a ridge and rise time with cooling it down. The heating up process depends on the substrate’s thermal conductivity coefficient. The lower it is compared to the ridge thermal conductivity coefficient, the faster the heating up process. Comparatively, the cooling down process is, at least, 2 times slower because the main heat dissipation process from the ridge takes place through the substrate.
5. Wavelength dependence
The presence of gaps on both sides of the heated part of the waveguide makes it similar to a Fabry-Perot cavity. It can be seen clearly (Fig. 5 ) that MZI transmission depends strongly on the wavelength exhibiting a periodic dependence with respect to the phase accumulated by the cavity mode per circulation. So, depending on the wavelength, the heating of the waveguide will cause an increase or decrease in the MZI transmission, because a thermally induced phase lag of the cavity mode can bring this mode closer or farther away from the condition of constructive interference with the second arm mode entering the coupling region.
The mode spacing between resonator modes in the Fabry-Perot cavity, namely the free-spectral range (FSR), allows the cavity length to be estimated from measured resonator transmission spectra. For a 46 μm long cavity (Fig. 1(a)) we measured a FSR of ~74 nm (Fig. 5(a) and 5(b)), in good agreement with the expected value from calculations, where the mode group index was almost constant for wavelength range λ = 1520 - 1630 nm and gives ng = 1.463 and ng = 1.486 for structure with PMMA and CAP ridge respectively.
Due to the thermo-optic effect in PMMA and CAP, the ridge index changes, resulting in an increase of the mode effective index neff and the group index ng with temperature, which blueshifts the resonance wavelength Δλ. This blueshift is related to the negative TOC of both polymers. Higher TOC of CAP and higher glass transition temperature ensure higher wavelength shift compared to PMMA (Fig. 5(a) and 5(b)). For a 46 μm long Fabry-Perot cavity and a maximum change in temperature ΔTmax = 65 K available for the PMMA ridge, the wavelength shift was estimated to be Δλ ≈7.4 nm. The maximum temperature increase corresponds to an applied electrical power P ≈14 mW. Thus, for P = 9.6 mW the wavelength shift is expected to be ~5nm, which fits very well with the experiment (Fig. 5(a)). For the structure with CAP ridge the wavelength shift of Δλ ≈36.5 nm is expected for ΔTmax = 80 K, equivalent to P ≈20 mW applied electrical power and to a blueshift of each resonance by ~1.825 nm/mW. Based on this, about ~12nm blueshift in resonance wavelength is expected when power is increased from 7.5 mW to 13.8 mW, which once again fits very well with measured spectra (Fig. 5(b)).
Compact fiber-coupled DLSPPW-based MZI thermo-optically modulated by an external voltage was presented. Two sets of samples with PMMA or CAP ridge supported by gold electrodes deposited on Cytop as a supporting material were fabricated and characterized, showing significant improvement in the performance of such MZI compared to previous studies where magnesium fluoride was used. This improvement is attributed to the lower thermal conductivity coefficient of CYTOP (0.12 W/m.K) compared to MgF2 (11.6 W/mK). In this way, the heating efficiency of the ridge is increased and the total power efficiency for switching purpose is decreased.
The experiments have shown π- phase shift with CAP-loaded MZI and around 0.4 π- phase shift with PMMA-loaded structure, resulting from a 3 times larger thermo-optic coefficient of CAP.
For the MZI with a CAP ridge, low switching power of 2.35 mW was obtained with 10 - 90% rise time of 65 µs and 90 – 10% fall time of 20 µs and an extinction ratio of 15 dB. Further improvement was observed in terms of the switching time (20 µs rise time and 15 µs fall time) for overdriven power of ~2.8 mW. The switching time was reduced by over 350 times compared to a structure with PMMA ridge with the same design, which should be attributed mainly to the thermal properties of CAP and partially to the smaller ridge height.
The presence of the Fabry-Perot cavity in the modulated arm of the Mach-Zehnder interferometer causes it to work as a wavelength-selective component. A thermo-optically controlled blueshift of the MZI resonance of about 7.4 nm for a temperature rise of about 65 K was observed for MZI with PMMA ridge and 36.5 nm for 80 K temperature rise for MZI with CAP ridge.
Compared to other thermo-optic switches (photonic and plasmonic), our MZI offers the following advantages: small footprint (115x15 μm2) with a heated arm length of 46 μm, easier on-chip integration, and simple fabrication compatible with different dielectric materials and with industrial fabrication using UV-lithography.
Recently, a new promising configuration for guiding well confined SPP modes over millimeter long distance was introduced theoretically [13,14] and confirmed experimentally , namely long-range DLSPPWs (LR-DLSPPWs), where the gold electrode is completely embedded in supporting materials, enabling more efficient heating of materials and further reduction of power needs for switching purpose. Additionally, significant reduction of propagation losses is achieved, opening the possibility to realize low-loss and compact plasmon-based photonic components and circuits.
This work was supported by the European Union (EU) project FP7-249135 (PLATON) and by the Danish Council for Independent Research (contract no. 09-072949).
References and links
1. Plasmonic Nanoguides and Circuits, S. I. Bozhevolnyi, ed. (Pan Stanford Publishing, 2008).
2. G. Coppola, L. Sirleto, I. Rendina, and M. Iodice, “Advance in thermo-optical switches: principles, materials, design, and device structure,” Opt. Eng. 50(7), 071112 (2011). [CrossRef]
3. T. Holmgaard and S. I. Bozhevolnyi, “Theoretical analysis of dielectric-loaded surface plasmon-polariton waveguides,” Phys. Rev. B 75(24), 245405 (2007). [CrossRef]
4. A. V. Krasavin and A. V. Zayats, “Passive photonic elements based on dielectric-loaded surface plasmon polariton waveguides,” Appl. Phys. Lett. 90(21), 211101 (2007). [CrossRef]
5. T. Nikolajsen, K. Leosson, and S. I. Bozhevolnyi, “Surface plasmon polariton based modulators and switches operating at telecom wavelengths,” Appl. Phys. Lett. 85(24), 5833–5835 (2004). [CrossRef]
6. G. Gagnon, N. Lahoud, G. A. Mattiussi, and P. Berini, “Thermally activated variable attenuation of long-range surface plasmon-polariton waves,” J. Lightwave Technol. 24(11), 4391–4402 (2006). [CrossRef]
7. T. W. Ebbesen, C. Genet, and S. I. Bozhevolnyi, “Surface-plasmon circuitry,” Phys. Today 61(5), 44–50 (2008). [CrossRef]
8. A. Kumar, J. Gosciniak, T. B. Andersen, L. Markey, A. Dereux, and S. I. Bozhevolnyi, “Power monitoring in dielectric-loaded surface plasmon-polariton waveguides,” Opt. Express 19(4), 2972–2978 (2011). [CrossRef] [PubMed]
9. J. Gosciniak, V. S. Volkov, S. I. Bozhevolnyi, L. Markey, S. Massenot, and A. Dereux, “Fiber-coupled dielectric-loaded plasmonic waveguides,” Opt. Express 18(5), 5314–5319 (2010). [CrossRef] [PubMed]
10. 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(2), 1207–1216 (2010). [CrossRef] [PubMed]
11. R. M. Briggs, J. Grandidier, S. P. Burgos, E. Feigenbaum, and H. A. Atwater, “Efficient Coupling between Dielectric-Loaded Plasmonic and Silicon Photonic Waveguides,” Nano Lett. 10(12), 4851–4857 (2010). [CrossRef]
12. R. Daviau, A. Khan, E. Lisicka-Skrzek, R. Niall Tait, and P. Berini, “Fabrication of surface Plasmon waveguides and integrated components on Cytop,” Microelectron. Eng. 87(10), 1914–1921 (2010). [CrossRef]
14. J. Gosciniak, T. Holmgaard, and S. I. Bozhevolnyi, “Theoretical Analysis of Long-Range Dielectric-Loaded Surface Plasmon Polariton Waveguides,” J. Lightwave Technol. 29(10), 1473–1481 (2011). [CrossRef]
15. V. S. Volkov, Z. H. Han, M. G. Nielsen, K. Leosson, H. Keshmiri, J. Gosciniak, O. Albrektsen, and S. I. Bozhevolnyi, “Long-range dielectric-loaded surface plasmon polariton waveguides operating at telecommunication wavelengths,” Opt. Lett. 36(21), 4278–4280 (2011). [CrossRef] [PubMed]