1. Introduction

Photonic crystals [1] are artificially created dielectric structures which are able to control the flow of light in specific directions and frequencies. Such structures allow us the flexibility to adapt specific applications such as low-loss waveguiding [2], nanoresonators [3] and even slowing light down [4]. Amongst the many applications of photonic crystals, micro/nanocavities [5–7] are one of the most distinctive and widely studied phenomenon which may confine light in a tiny space and yet achieve a low decay rate. Such features enable a wide range of possibilities, from quantum electrodynamics implementations [8], single photon sources [9] to add/drop optical multiplexer [10] and optical switching [11]. Due to its extremely compact size and flexibility of design, photonic crystal nanocavities are demonstrated to be a promising building block for future integrated photonic/optical nanodevices.

However, micro/nanofabrication of photonic crystal nanocavities remains a challenging task as such structures are extremely sensitive to tolerances in micro/nanofabrication which are typically difficult to control. Slight fabrication errors will incur losses and cause deviation from originally designed resonances. Hence, it is critical to achieve a reconfigurable device that is less affected by fabrication imperfections and ultimately functions as a tunable photonic device. Various innovative static-tuning approaches such as digital etching mechanism [12], nano-oxidation [13] and tuning by design methods may be helpful in fine-tuning of the resonant frequencies to overcome fabrication errors. Dynamic tuning approaches such as liquid crystal infiltration, thermo-optic tuning [14], liquid-infiltration [15], electro-optic tuning [16] to achieve tunable photonic devices have previously been reported. However many of such methods requires the integration of unusual materials that may not be readily amalgamated with conventional microfabrication techniques. Recently a few groups have proposed and demonstrated a promising approach to control and tune the resonance of nanocavity nano-mechanically utilizing tiny dielectric tips controlled by near-field scanning microscope (NSOM) or atomic force microscope (AFM) [17–20]. These excellent reports provide in-sights into achieving precise nano-mechanical out-of-plane tuning without the resonator suffering losses. Another simple but innovative approach to achieving large spectral shifts may be accomplished by utilizing coupled-cavities approach. With this approach, various groups have recently experimentally shown and studied the effects of coupled one-dimensional photonic crystal (1D PhC) nanocavities [21–25] utilizing novel vertical out-of-plane resonant scattering and NSOM techniques [26–29].

In order to achieve a sophisticated and fully on-chip integrated nanoscale displacement approach to control photonic crystal resonances without bulky NSOM or AFM systems, we need to look into both the micro and nanoscale such as employing the micro/nano-electromechanical systems (MEMS/NEMS). Sub-micron MEMS have been demonstrated to be promising solutions to enable fully-integrated tunable photonic devices, with vast development work reported on devices such as slab waveguide optical switches [30–32], complex micro-toroid resonators [33], waveguide phase-shifters [34] and add/drop demultiplexers [35] etc.

In this paper, we propose three different tip designs that are integrated with a novel retractable push-pull sub-micron MEMS mechanism to achieve low-loss and reversible control of resonance in a 1D PhC nanocavity. Such an approach is to the best of our knowledge the first report on a fully integrated on-chip photonic device equipped with a retractable in-plane nano-mechanical capability of achieving 1D PhC resonance control over a large range but yet robust against external vibrations and surface stiction-related problems that typically affect MEMS/NEMS structures. Since the proposed sub-micron MEMS device possesses relatively high natural vibration frequencies due to its small size (27µm width × 15µm length), switching speeds have been demonstrated to be at the range of few micro-seconds [34].

We believe that the proposed designs can be adapted for optical devices that require precise resonance fine-tuning, sensitive displacement sensors and those that require large fine-tuning range such as tunable optical filters.

2. Device design

The overview of the proposed designs is shown in Fig. 1 , in which the 1D PhC is integrated onto a single mode TE nanowire waveguide suspended by low-loss elliptical suspensions [36]. The air-cladded nanowire waveguide has a width of 500 nm and thickness of 340 nm. The 1D PhC nanoresonator is based on a Fabry-Perot resonator that is achieved by introducing periodic air holes along the single mode waveguide with a central nanocavity defect to localize an optical mode within the bandgap of the PhC. The localized optical fields formed in this nanocavity region would then be easily perturbed. To achieve a resonant mode within the optical communications window, we chose a lattice constant of 370 nm, with a bragg-stack hole radius of 96 nm and a nanocavity length of 440 nm. The hole diameters are taper progressively down to a radius of 70 nm to achieve gradual confinement of the resonance. One additional out-taper hole with radius 70 nm is designed to enhance the transmission of resonance mode through mode matching. We theoretically calculated the resonance to be at λ_o = 1580 nm with a modest quality factor (Q) of approximately 4.8 × 10³ through 3D-FDTD. In this work, we focused on PhC designs with large transmissions with modest Q to ensure higher tolerance to fabrication imperfections and greater ease of in-coupling.

 

Fig. 1 SEM image of the nano-mechanically variable photonic devices. (a) Illustration of the proposed retractable nano-mechanical actuating mechanism with integrated low loss nanowire waveguides. Lighter color structures are driven with voltage where the darker structures are grounded. (b) coupled-nanocavity nano-tip (c) rectangular nano-tip (d) meniscus-like perturbing nano-tip. Inset in (c) & (d) is an enlarged schematic of the tip design for clarity.

Download Full Size | PDF

The retractable sub-micron MEMS push-pull structure, shown in Fig. 1(a) features two sets of electrostatic comb-drive actuators, each set for an opposite direction of displacement. As illustrated in the scanning electron microscope (SEM) image, the “Push” sub-micron comb-drive actuators displaces the tip closer to the 1D PhC nanocavity and likewise the “Pull” comb-drive actuators pulls the tip away from the 1D PhC nanocavity. This sub-micron MEMS structure can ensure mechanical stability and robustness of nano-mechanical perturbing members when in the optical near-field regime. The designed sub-micron actuators consists of 22 sets of push comb-drive fingers that are each 200 nm wide, 1.5 µm long and have a finger gap spacing of 200 nm. When displaced to a regime where the tip and the PhC are at the optical near-field regime, minute fluctuations in driving voltages or external vibrations may induce the tip surface to adhere to the PhC surface. Since the Van-de-Waals surface forces are much greater than the restoring spring force, the moveable member will be stuck to the PhC– rendering the device unusable. Hence, a second set of 31 pairs of Pull comb-drive actuators with identical dimensions as those of the Push actuators are incorporated to enable the retractable motion. Such push–pull mechanisms may also be configured with an identical number of comb-actuators to achieve a linear displacement to voltage relationship for higher accuracy control of nano-mechanical displacements [37]. In order to ensure a long stable travel range for the sub-micron gap between the comb-drive fingers, we designed tilted folded beam suspensions [38] with flexure beams of 150 nm in width and 9 μm in length to support the entire moving nano-mechanical structure.

Light is coupled into the waveguide and fed into the 1D PhC. As the PhC resonance is designed to be at the center of the bandgap, where only the resonance mode is able to transmit through the structure with high transmissions. The gap distance is defined by the air-gap distance between the nano-tip and the edge of the 1D PhC nanocavity. By nano-mechanically adjusting the gap distance between the 1D PhC and the perturbing structures, a resonance shift is expected due to interaction between the perturbing structures with the resonant mode. We introduced various perturbing tip designs as shown in the SEM images illustrated in Figs. 1(c)–1(d) to study the effects of various types of perturbing structures. The rectangular nano-tip as shown in Fig. 1(c) has a width W of 400 nm and length of 1.5 µm. This simple rectangular geometry design helps us to understand the in-plane perturbing characteristics on the 1D PhC nanocavity. As the rectangular nano-tip perturbs into the near-field of the resonant mode, we expect some scattering losses and broadening of resonance as previously reported [17–20]. Ideally, large rectangular tips may be designed to achieve higher spectral shifting, however this may induce impractical broadening of Q and reduction of transmitted power which potentially renders the PhC resonance useless. As the rectangular silicon nano-tip has a high refractive index, the evanescent field leaking from the resonance mode may possibly excite guided modes along the rectangular nano-tip, allowing energy leakage and thus degrading the Q and transmitted power. Intuitively, reducing W of the nano-tip may effectively reduce this energy loss. However this would also result in less spectral shifting.

An intuitive approach to improve this design is to reduce the coupled guided modes along the nano-tip as inspired by an ellipsoidal reflector previously reported to have reduced emissions of a nanoresonator [39]. We heuristically designed a meniscus-like nano-tip design based on a series of numerical calculations using 3D–FDTD. The proposed meniscus-like nano-tip is shown in Fig. 1(d). The overall meniscus shape is 800 nm wide and consists of a convex with a radius R of 400 nm and concave profile that consists of a 500 nm wide plano edge at the centre as seen in the inset. The plano edge is designed to ensure that a larger area of tip perturbs into the evanescent fields. This design is able to effectively perturb the 1D PhC resonances and yet maintain a reasonable Q compared to the rectangular nano-tip. For good comparison with the rectangular tip, the gap distance is defined by the air-gap distance between the plano edge to the edge of the 1D PhC nanocavity. We may further increase the size of the meniscus-like structure to achieve larger spectral shifts; however this is a rather inefficient approach as compared to the coupled-cavity approach described below.

The coupled-cavity approach as a perturbing tip to achieve PhC resonance control through mode splitting from two identical 1D PhC nanocavities is also studied. The perturbing 1D PhC nanocavity structure as shown in Fig. 1(b) has an identical cavity length and tapering profile to the 1D PhC nanocavity. The out-tapered holes of the perturbing nanocavity are replaced with five additional Bragg stack holes of 96 nm radius to ensure that the coupled resonance energy does not leak into the perturbing structure. Likewise the gap distance is defined as the air-gap distance between the two nanocavities.

3. Fabrication and characterization

In the fabrication of the devices, we patterned the 1D PhC, sub-micron MEMS structures and waveguides through 100 keV electron beam lithography ELS-7000 (Elionix, Inc.) on a commercially-available SOI wafer (SOITEC, Inc.) The SOI wafer consists of a 340 nm silicon layer on a 1.0µm buried SiO₂ layer. The silicon device layer is p-doped to a resistivity of 14-22 Ω.cm which sufficiently conductive for electrostatic actuation. We first spin-coated a commercial electron beam resist ZEP520 on to the samples at a speed of 5000 pm forming a 350 nm - thick layer resist which acts as a mask. An electron beam with a registered dose of 320 µC/cm² is optimized to pattern the PhC, waveguides and sub-micron MEMS structures. Additionally, the flatness of the sample is ensured through cleaning of the backside and field correction to reduce field stitching errors due to the small field size. We then selectively etched the top silicon layer by utilizing inductive coupled plasma reactive ion etching (ICP-RIE) (Oxford Instruments Ltd.) with SF₆ and C₄F₈ chemistries to attain straight sidewall profiles. The parameters for the etching process is based on a modified time-multiplexed etching process with ICP power of 450W, RF power of 30W, gas pressure of 15mTorr and etching gas ratio of 1:2 (SF₆: C₄F₈).

The etch selectivity between the commercial electron beam resist and silicon is approximately 0.3 with an etch rate of approximately 200 nm/min. After etching, the polymer resist is removed from the SOI wafer utilizing a solvent dip followed by a typical piranha etch, which is a mixture of two parts sulfuric acid (H₂SO₄) and one part hydrogen peroxide (H₂O₂). Multiple cleaning of the samples with the piranha solution ensures the effective removal of polymeric residues in order to achieve low-loss waveguides. The sacrificial SiO₂ layer is then removed using an in-house designed HF vapor system having an etch rate of approximately 100 nm/min. The samples are finally cleaned again in an oxygen plasma RIE (Trion Technology, Inc.) dry etcher to remove any remnants of non-volatile residues left from sacrificial oxide etch.

Experiments are carried out to characterize and calibrate the static displacement to voltage relationships of the designed sub-micron MEMS structures. The device characterization is performed by measuring the gap between the perturbing structures under the SEM at selected driving voltages as shown in Fig. 2 . As the fabricated sub-micron MEMS structures are identical for all tip designs, we showed that the quadratic nature of the voltage to gap displacement characteristics agrees quite well with theoretical estimations [38]. The insets in Figure. 2 show the SEM images of the various tip designs at a near-field gap distance of 70 nm. We also tested the retraction capabilities of a stiction incurred device under the SEM. A reverse force is exerted by electrically driving the “Pull” actuators. As shown in Fig. 2(d) the anchors of the “Pull” fingers are darken due to driven voltages. A 23 V actuation voltage in conjunction with the tilted folded beam restoring forces is sufficient to retract the device against surface stiction forces as shown in Fig. 2(e).

 

Fig. 2 SEM characterization of the static displacements as a function of applied voltage of the retractable nano-mechanical actuator. Inset shows the various designs at 70 nm gap (a) coupled-nanocavities, (b) meniscus-like nano-tip, (c) rectangular nano-tip. SEM demonstration of a (d) stiction incurred nano-mechanical structure and (e) retracted mechanical structure

Download Full Size | PDF

4. Experimental results and discussion

In the optical characterization setup, we used a tunable infrared laser source (ANDO AQ4321) that is coupled into a polarization controller and output into a polarizing maintaining single mode (PM-SM) fiber. We first launch light from the PM-SM fiber onto a near-IR linear polarizer which is aligned to the waveguide TE direction. Power is monitored using an InGaAs power meter. The polarization controller and PM-SM fiber are then finely adjusted to achieve the desired TE polarized light. The polarizer and power meter is then removed to directly end-fire light into the silicon device. The output signal is collected at the other end of the sample by using a multi-mode fiber connected to an optical spectrum analyzer (ANDO AQ6317) for optical characterization.

Optical characterization of the sub-micron MEMS PhC device is performed over a spectral range from 1520 to 1620 nm with a spectral resolution of 0.025 nm step measurement. The results of all these devices are normalized to that of an identical 500 nm wide waveguide without PhC patterns and perturbing tip structures.

The transmission results of the resonant 1D PhC nanocavities without driving voltages are shown in black in Figs. 3(b) -3(c) & Fig. 4(b) . Some variations in the central resonance wavelength and Q are to be expected due to fabrication tolerance errors. The 1D PhC resonance of the rectangular nano-tip device demonstrates a central resonance at 1597 nm with an experimental waveguide loaded Q_rectangular of 7 × 10². whilst the corresponding values for the device with the meniscus-like nano-tip device are around 1601 nm with an experimental waveguide loaded Q_meniscus of 1.3 × 10³. The 1D PhC resonance of the coupled nanocavity possesses a resonance of around 1598 nm with an experimental waveguide loaded Q_cc of 1.1 × 10³. The modest experimental waveguide loaded Q are to be expected due to the limitations in our process equipment. Figure 3(a) shows the effect on the spectra of the devices as the nano-tip perturbs the centre of the 1D PhC nanocavity with decreasing gap displacements. As the gap between the tip and the nanocavity is reduced from 750 to 400 nm – which corresponds to 0 V to 22 V driving voltages- no detectable spectral shift is observed. For the rectangular nano-tip, as the gap is reduced to 230 nm gap, we observe a minute shift of 0.04 nm in the wavelength but no change in the Q. At a gap of 190nm, there is an increase of wavelength shift to 0.11 nm and a slight decrease in Q to 6.35 × 10². Finally, as the nano-tip displaces to a distances at a near-field gap of 90 nm a 1.1 nm spectral shift is observed. However, at such small distances, we detected degradation of the experimental Q to approximately 3 × 10². The 3D FDTD results (solid lines) are also shown in Fig. 3(a) where we numerically predict the resonance shift with the absence of transmission reduction. However, it would be impossible for our device to achieve spectral shift beyond 90 nm air gap as beyond that, displacing the tip would yield a hardly distinguishable resonance peak along with the signal noise due to transmission losses as previously described. This result is expected as we previously discussed, the high refractive index nano-tip effectively causes energy leakage of the resonant mode.

 

Fig. 3 (a) Experimental and simulated resonance shift of perturbing tips as a function of offset distance & driving voltage. E-fields plots of both nano-tip designs are shown in the inset. Transmission spectra of the resonance shift with static electrical actuations for (b) rectangular nano-tip (c) meniscus-like nano-tip. Solid dots denotes the experimental results, solid lines denotes the fitted lorentzian peak. (d) Experimental Q of menicus-like tip and rectangular tip as a function of offset distance

Download Full Size | PDF

 

Fig. 4 (a) Experimental and simulated resonance shift of coupled nanocavities as a function of offset distance & driving voltage. E-field plots of even and odd modes are shown as inset (b) Transmission spectra of the resonance shift with static electrical actuations (c) Experimental Q of odd and even modes as a function of offset distance

Download Full Size | PDF

In Fig. 3(a), we also showed the resonance tuning characteristics of the meniscus-like nano-tip. In this cas, at a gap of 220 nm, we first obtained a 0.095 nm wavelength shift. Further increasing the driving voltage to 24.25 V, we reduced the gap distances to approximately 160 nm and achieve a 0.34 nm spectral shift. Finally, at a near-field distance of 85 nm gap at approximately 24.75 V, we observed a 2.25 nm spectral shift which agrees well with the FDTD results. Furthermore, as shown in Fig. 3(b) is the experimentally loaded Q of the shifted spectra is approximately 1 × 10³ which meant that it did not degrade significantly from its initial value of Q_meniscus = 1.3 × 10³. The meniscus-like tip cannot displace beyond 80nm displacement gap as the concave apex would contact with the edge of 1D PhC nanocavity. A summary of the Q degradation of both nano-tip designs are shown in Fig. 3(d).

We also illustrate the mode diagrams of the 1D PhC resonance when perturb by the rectangular and the meniscus-like nano-tip design at a distance of 50nm in Fig. 3(a). Some field energies are observed to be leaking out from the rectangular tip which further validated the loss mechanism of such designs. The meniscus-like tip showed very minimal energy decay due to such a loss mechanism. This result can be further verified from the theoretical Q calculated using FDTD as shown in the inset. At a gap distance of 50nm gap, the rectangular nano-tip showed a theoretical Q of 3.82 × 10² while the meniscus-like nano-tip showed a theoretical Q of 3.5 × 10³.

It is clear from the first two designs that fine resonant control of PhC nanocavities can be achieved through a tip perturbative approach. However to achieve large spectral range tuning, resonance splitting- where the resonances are split by coupling two identical cavities may be a better approach. The splitting induces two modes to be formed, namely the even and odd modes. With an initial resonant wavelength of 1598 nm, we observe in Fig. 4(a) that by adding a driving voltage of 23 V, which corresponds to a 225 nm gap, is sufficient to achieve a ± 3 nm shift respectively for both even and odd modes. Further increasing the driving voltages to 23.5 V -corresponding to a 180 nm gap would initiate the unequal spectral split between the even mode (4 nm) and the odd mode (−3.5 nm). Finally at driving voltages of 24.4 V which corresponds 70 nm gap distances, an 18 nm shift for the even mode is induced, whilst the odd mode remains little changed at a −3.9 nm. The mode fields of the even and odd modes at 50nm gap distances are shown as insets in Fig. 4(a). The even mode is shown to be a slotted-like mode with energies strongly confined in the air-region, whereas the odd mode is shown to be a dielectric-confined mode as has been reported previously [26–29]. Figure 4(b) shows the experimental spectral at various driving voltages. The spectral results shown in this plot demonstrates that by mechanically shifting the structure to a 200 nm gap distance with a 24.0 V driving voltages, both the experimental loaded Q for the even and odd mode is approximately around to 1.5 × 10³ and to 1.9 × 10³ respectively. At 24.3 V which corresponds to a 70 nm gap distances, the waveguide loaded Q for even and odd modes are approximately 1.05 × 10³ and 1.6 × 10³ respectively. We summarized the experimental Q of both even and odd modes of the coupled cavities structure in Fig. 4(c). It is noticeable that the odd mode shows slightly higher Q compared to the even mode. However minimal degradation in Q is seen when both cavities are displaced to a near-field gap. Hence we conclude that such resonance control methods are capable of achieving large resonance shifts without significant reduction in Q.

Figure 5 shows the time-resolved transient responses of the meniscus-like tip-tuning mechanism and the coupled nanocavity tuning mechanism. The measurement is performed on the identical device using a CW IR laser for both devices respectively. The transient experimental setup is identical to that described previously except that the output is connected to a high-speed InGaAs photodetector with 150 MHz response time. The nano-mechanical structure is driven with a bias voltage of 21 V and a 1 kHz square wave-form amplitude from 0 V to 3.25 V. We first drive the nanomechanical structure with the biased voltage at which point the red-shifted resonance is then identified utilizing the optical spectrum analyzer. The bias red-shifted resonance Q is found to be in the range of 1.2 × 10³. We then tune the CW IR laser to the half height of the resonance peak which corresponds to an output power of approximately 1 µW or detector output of several mV as can be seen in the inset of Fig. 5(a). As the nano-structure is driven by a square waveform voltage, rising voltage would cause the nano-tip to displace into the near-field regime and further red-shifts the resonance, causing a decrease in the transmission from the edge of the resonance wavelength. Vice versa, the falling voltage would induce the nano-tip to return to its initial position and blue-shifts the resonances. Figure 5(a) shows the response when the meniscus-like nano-tip displaces away from the near-field regime, thus demonstrating an increase in the transmissions. The initial motion exhibits an over-damped response with the rise time of approximately 2 µs for the nano-tip structures. The rectangular nano-tip was not investigated as it has a similar nano-tip structure and MEMS dimensions as compared to the meniscus like tip.

 

Fig. 5 Transient characteristics of the nano-mechanical modulated PhC resonator with a square-wave driven signal under atmospheric conditions of (a) meniscus-like nano-tip, (b) coupled-nanocavity. Inset in (a) shows the chosen half height location of resonance peak which would encounter transmission increase when actuating voltage is reduced. Resonance red-shifts as actuating voltage is increased and blue shifts with reducing actuating voltage.

Download Full Size | PDF

Figure 5(b) describes the transient response of the coupled-nanocavity structure, with similar tuning configuration to that described for the nano-tip. Similarly the nano-mechanical structure is driven with a 1 kHz square wave-form amplitude from 0 to 3 V with a bias voltage of 20 V. Similar techniques are used to obtain the transient response. The even mode resonance peak is chosen for its higher sensitivity to displacement. The bias red-shifted resonance Q of the even mode is found to be in the range of 1.3 × 10³. It can be deduced from the absence of overshoot and oscillatory motions that such nano-mechanical systems are also over-damped. The rise time was approximated at about 18 µs which is significantly longer than that of the meniscus like nano-tip even though the actuating structures are similar. The slight increase in rise time may be attributed to the larger size of the perturbing cavity which potentially encounters larger squeeze film air-damping when compared to the meniscus-like nano-tip. It has been previously shown [40,41] that the dynamic response and resonance of micro/nanostructures are dominated by air-damping of such structures. We believe that the response time of such nanomechanical structure may be improved significantly by further decreasing the mass of the sub-micron MEMS structures and increasing the spring constant of the nanobeam suspensions. The dominating viscous air-damping that typically affect such MEMS structures can be mitigated by repeating the transient optical experiment under vacuum conditions.

5. Conclusions

In summary, we have carried out numerically and experimentally studies on three new tip designs that are integrated with a push-pull sub-micron MEMS mechanism to control resonance of 1D PhC cavity with minimal loss. The in-plane rectangular nano-tip perturbation approach demonstrated small tuning properties and incurred vast broadening of the Q and decreasing transmitted power. The meniscus-like perturbating tip resulted in 2.25 nm resonance shifts with minimal transmitted power loss and deterioration of Q. It is demonstrated that 1D PhC nanocavities are highly sensitive to geometries and size of the perturbing tip. Such designs may potentially be utilized for post-fabrication fine-tuning of 1D PhC nanocavity resonance or high-precision wavelength tunable optical filters. The coupled–nanocavities approach demonstrates a relatively large resonance shift of 18 nm with minimal degradation in Q. Such devices may be useful for tunable photonic devices requiring large tuning ranges such as photonic switches or tunable optical filters. We also demonstrated the feasibility of using a retractable structure to easily recover nano-mechanical perturbation structures that had suffered stiction when operating in the near-field regime.