1. Introduction

Optical communications are essential in many areas. A significant advance that has made optical communications [1] possible is the development of low-loss passive optical fibers [2,3]. However, any dynamic functionality in a fiber-optic network generally requires discrete components, e.g. for switching, modulation, or tunable filtering. More recently, optical or photonic integrated circuits [4] have enabled active functionality in a compact chip-scale device. Although losses in semiconductor waveguides have been reduced substantially, the lowest loss is still achieved in silica [5] or silicon nitride [6] based photonic integrated circuits. These material systems, while enabling ultra-low losses, are generally not reconfigurable. Active functionality in silica or silicon nitride has been shown to a limited extent [7,8] using temperature tuning, although the thermo-optic effect is only of the order of Δn/ΔT = 10⁻⁵ which makes this approach [9] highly inefficient. While recent work has shown some electro-optic effects in strained silicon/silicon nitride [10,11], the origin and magnitude of these effects are still being studied [12,13]. Liquid crystals [14] on the other hand enable large Δn, although their response time is limited.

Recently, micro-electro-mechanical systems (MEMS) [15] and optical MEMS [16,17] have enabled reconfigurable photonic devices [18] including actuated-mirror phase shifters [19] and tunable filters [20]. These approaches have relied on dynamically controlling the physical path length and have achieved large phase shifts up to 3π [21], although they require a substantial device footprint. A different approach relies on changing the optical or effective path length by changing a material’s effective refractive index (n_eff) via evanescent field interaction. This approach does not require a change in the physical path length and thus has the potential for compact footprint devices and large-scale integration. Initial demonstrations utilized off-chip atomic force microscope tips [22] or silica fiber probes [23] to perturb the effective index of micro-/nanophotonic devices. More recently, a chip-scale approach utilized a silicon nitride waveguide side-coupled to an electrostatically-actuated perturber [24]. Actuation leads to a variable phase shift with modest effective index tuning (Δn_eff≈0.002) due to the tight optical mode confinement. In another implementation [25] an array of gold bridges suspended above a fixed gold layer serves as gap plasmon waveguides whose n_eff can be tuned by actuating the bridges and varying the air gap. The gap plasmons are probed with free-space-coupled light in order to measure the phase shift by imaging the interference pattern resulting in a measured Δn_eff≈0.03 [25]. However, plasmonic structures are challenging in terms of loss [26] and coupling to dielectric-based waveguides is often difficult, which may limit the integration of such structures in photonic integrated circuits.

In this work, we present a fully-integrated waveguide-coupled opto-electro-mechanical phase shifter architecture enabling an exceptionally large Δn_eff. Our approach is broadband, low-loss, and provides a direct means for coupling to other components on-chip and is therefore suitable for large-scale integration in photonic integrated circuits. The device combines a thin-core low-mode confinement engineered waveguide [27] with a suspended microbridge fabricated directly above the waveguide surface in a compact footprint device. The evanescent field interaction between the guided mode and the suspended microbridge leads to a controllable induced phase shift. We use electrostatic actuation in a metal-coated bridge to modify the waveguide-bridge separation and to dynamically control the phase shift. We also take advantage of forces induced by gradient electric fields [28–31] to actuate an uncoated dielectric bridge (no metal) to tune the effective index and phase while eliminating metal-induced losses. Both approaches are widely-applicable to micro-/nanophotonic waveguides to enable active functionality via large effective index tuning and phase shifts in structures using otherwise passive optical materials.

2. Approach

Figure 1 shows the device principle of operation. A thin silicon nitride (Si₃N₄) rib waveguide confines and guides light. Owing to the large aspect ratio, the optical mode is strongly confined along the waveguide width (w_rib = 2.5 μm with etch depth h_rib = 55 nm) but only weakly confined along the thickness (t_Si3N4 = 175 nm). The result is a large evanescent field at the waveguide surface that was maximized by choosing an appropriate Si₃N₄ layer thickness [27]. The fundamental quasi-TE waveguide mode (TE₀₀) has an effective index n_eff = 1.465 at λ = 1.5 μm, which is only slightly larger than the lower cladding index (n_SiO2 = 1.444) and is indicative of the weak mode confinement. At the same time, the measured waveguide propagation loss is modest (as low as 0.7 dB/cm [27]). A 300nm thick tensile SiN_X microbridge is suspended just above the waveguide surface with a nominal air gap of g = 190nm, as shown in Fig. 1(a). The microbridge is coated with a titanium/gold layer (Ti/Au = 5nm/35nm) and, together with an adjacent lower side electrode (Fig. 1(b)), forms a capacitive electrostatic actuator. By grounding the microbridge and applying a bias to the lower electrode the air gap separating the waveguide and microbridge perturber can be varied.

 

Fig. 1 a) Cross-section schematic showing a Si₃N₄ rib waveguide (core thickness t_Si3N4 = 175nm, rib width w_rib = 2.5 μm and etch depth h_rib = 55 nm), a metal/dielectric microbridge, and a lower electrode for actuating the microbridge’s vertical position, b) optical profile of a fabricated device with a flat microbridge and a lower side electrode, c) simulated effective index (n_eff) vs. air gap (g) and corresponding metal-induced loss, d) selected waveguide modes.

Download Full Size | PDF

We simulate [32] the waveguide’s TE₀₀ mode at λ = 1550 nm and find a strong dependence of the effective index on the air gap with n_eff increasing from 1.47 to 1.62 (Fig. 1(c)). The weak mode confinement results in an evanescent field that extends hundreds of nanometers above the waveguide surface and hence the waveguide’s n_eff is strongly affected by evanescent field interaction with the microbridge. The simulated loss induced by the Ti/Au metal layer on top of the microbridge is modest as shown in Fig. 1(c) (<0.08 dB for a L = 100 μm interaction length), likely because the thick SiN_X microbridge keeps the fundamental mode far away from the Ti/Au for gaps>150 nm (Fig. 1(d)). Finally, we note that the structure in Fig. 1 also supports a microbridge optical mode that is confined largely in the suspended SiN_X microbridge. However, this hybrid dielectric-metal plasmonic mode exhibits significant loss and is quasi-TM-polarized so that this mode is not expected to contribute to the device operation since we launch only horizontally-polarized light (see Appendix).

3. Experimental results

3.1 Device actuation

A fabricated phase shifter is shown in Fig. 2(a) (see Appendix for fabrication process). Suspended above the waveguide is the SiN_x/Ti/Au microbridge; the waveguide is continuous underneath the microbridge. Although not apparent in Fig. 2(a), the lower electrode has collapsed toward the substrate after sacrificial release (Fig. 2(b)). We use a white light optical profilometer to measure the bridge profile as we increase the bias voltage on the lower side electrode while keeping the bridge grounded. The beam deflection is measured as a function of bias (Fig. 2(c)) revealing that the beam pull-in voltage [33] has a 1/L_MEMS-dependence (Fig. 2(d)), where L_MEMS is the suspended microbridge length. Although our actuator geometry is not strictly parallel plate, the actuation is predominantly in the vertical direction since the large width-to-height ratio (w_MEMS/t_MEMS = 3 μm/350 nm) results in a bridge that is approximately 10x more compliant in the vertical direction compared to the lateral direction.

 

Fig. 2 Device actuation: a) fabricated microbridge suspended above a rib waveguide with lower side electrode (false color); the waveguide (red) is continuous underneath the microbridge, b) optical profilometer images of the device with no bias and at bridge pull-in towards the waveguide surface; the black regions in the images are the electrical probes used for biasing the device, c) measured linescan profile of a L_MEMS = 50 μm microbridge at different lower electrode bias voltages with the microbridge at ground, d) measured displacement vs. bias for various bridge lengths, L_MEMS. The pull-in voltage, at which the microbridge touches the waveguide, follows a 1/L_MEMS-dependence.

Download Full Size | PDF

3.2 Optical characterization

We measure the opto-electro-mechanical device response using the setup in Fig. 3(a). Lensed fibers couple light from a tunable laser to the device. The phase shift is measured using an unbalanced Mach-Zehnder interferometer (MZI) with a built-in path length imbalance of ΔL_MZI = 100 μm on one arm. The laser wavelength is tuned and the device transmittance is measured by collecting the output signal using a second lensed fiber and sending the signal to a photodetector. The bias on the side electrode is increased while grounding the microbridge, and for each voltage the MZI response is measured. As the microbridge is actuated down toward the waveguide surface the evanescent field interaction becomes larger and the waveguide n_eff increases resulting in a broadband MZI fringe shift (Fig. 3(a), inset). The total phase is ϕ_MZI ± ϕ_MEMS, where ϕ_MZI is the built-in phase shift due to the path length imbalance (ΔL_MZI) and ± ϕ_MEMS is the phase shift induced by the MEMS bridge (L_MEMS). The sign of ϕ_MEMS depends on the location of ΔL_MZI compared to L_MEMS. We measured several devices, some of which had the microbridge (L_MEMS) on the same waveguide segment as the built-in MZI path imbalance (ΔL_MZI) as shown in Fig. 3(a) and some of which had L_MEMS and ΔL_MZI on opposite waveguide arms. As shown in Fig. 3(b) and 3(c), the fringe shift direction depends on the relative location of the microbridge (L_MEMS) compared to the built-in MZI path imbalance (ΔL_MZI). For each spectrum ϕ_MEMS is measured and Δn_eff is extracted (Fig. 4).

 

Fig. 3 a) Experimental setup with 2.0 μm spot size lensed fibers for coupling to the device. The lower image shows a fabricated MEMS phase shifter, where an unbalanced Mach-Zehnder interferometer (MZI) is used to measure the phase shift. The left inset shows the MZI fringe shift vs. bias voltage indicating broadband operation (λ = 1490-1545 nm); DUT = device under test; D.C. = directional coupler; ΔL_MZI = Mach-Zehnder interferometer built-in path length imbalance; b) measured fringe shift for a device where the microbridge (L_MEMS) is on the opposite arm and c) on the same arm as the built-in path length imbalance (ΔL_MZI).

Download Full Size | PDF

 

Fig. 4 Measured phase shifting and effective index tuning: a) Measured and calculated phase shift vs. applied bias for a device with L_MEMS = 120 μm; b) measured and calculated Δ_neff vs. air gap g (L_MEMS = 120 μm), c) extracted effective index tuning (Δn_eff) vs. g for all measured devices; d) device figure of merit (FOM): measured transmittance (loss) as a function of induced phase shift for all measured devices; only modest losses are observed for phase shifts as large as 3π.

Download Full Size | PDF

Figure 4(a) shows the measured MZI phase shift as a function of bias voltage for a device with L_MEMS = 120 μm. A phase shift Δϕ>2π is measured for V_bias = 4.2 V. From the measured phase we can extract the change in effective index as Δϕ = 2πΔn_effL_MEMS/λ, where λ is the wavelength and the displacement calibration curves measured in Fig. 2 were used to convert the applied bias voltage into an effective air gap (see Appendix). This conversion allows us to calculate the expected Δn_eff and Δϕ. The measurements and calculations in Fig. 4(a) and 4(b) show excellent agreement indicating that the modeling in Fig. 1 accurately describes our device behavior. In Fig. 4(c) we show the measured and calculated Δn_eff for five phase shifters with bridge lengths L_MEMS = 50-120 μm. The maximum extracted Δn_eff > 0.02, which is among the largest measured MEMS-based Δn_eff reported to date [24,25], although it is still significantly smaller than the maximum Δn_eff≈0.1 predicted by the simulation in Fig. 1(c); this is likely due to the fact that the g>100 nm for the measurements in Fig. 4(b) and 4(c). Our phase shifter figure of merit (FOM) is the measured transmittance vs. the induced phase shift, which we plot in Fig. 4(d). The transmittance is obtained by monitoring the peak transmission (i.e. at the MZI fringe maxima) as we increase the applied bias. The devices exhibit a FOM = 0.75 dB/π (1.5 dB/2π in Fig. 4(d)). We attribute the reduction in transmittance at larger phase shifts due to the strong coupling to the microbridge and the increased optical mode overlap with the Ti/Au metal layer at small air gaps (see Fig. 1(d)). Finally, we note that due to the nonlinear microbridge displacement with applied bias (see Fig. 2d) as well as the exponential decay of the evanescent optical field the devices exhibit a tradeoff between higher stability (smaller Δϕ) and larger Δϕ (at the expense of lower stability and higher phase noise).

3.3 Device actuation via gradient electric fields

In order to decrease potential metal-induced optical losses we fabricated and measured a MEMS phase shifter that relies on a different type of actuation mechanism that is similar to dielectrophoresis (DEP). Although DEP [34] is a technique commonly used to displace and sort microparticles and cells in microfluidics [35] (similar to the gradient optical force [36] used in optical tweezers [37]), it has only recently been used to actuate MEMS/NEMS devices [28–31]. Our actuator design consists of a suspended microbridge on top of a rib waveguide, identical to previous devices. However, the microbridge is now left uncoated (i.e. no metal). Two Ti/Au (5nm/55nm) side electrodes are patterned on both sides of the microbridge on top of the sacrificial SiO₂ layer, as shown in Fig. 5(a) (see Appendix for fabrication details). Since the microbridge has been etched partially into the underlying SiO₂, the side electrodes are beneath the microbridge. By applying a bias across the electrodes, the microbridge experiences a gradient electric field (Fig. 5(b)) and hence a vertical force that displaces the microbridge towards the waveguide surface (Fig. 5(c)). The gap between the two electrodes is designed to be 4 μm for ease of fabrication – significantly larger than the 1 μm actuation gap for the devices in Figs. 2-4. The electric field is therefore modest and we require significantly larger actuation voltages compared to devices using electrostatic actuation. It should be noted that simple design changes, e.g. reducing the side electrode separation, will lead to enhanced electric field gradients and a reduced actuation voltage.

 

Fig. 5 Device actuation via gradient electric field forces: a) Optical micrograph image of a fabricated waveguide device in which the suspended SiN_X microbridge is uncoated (i.e. no metal); two side electrodes separated by 4 μm are patterned on either side of the microbridge; the electric field lines during actuation are drawn to illustrate the device operation (false-color); the side electrodes are lower than the microbridge leading to a gradient electric field that actuates the suspended microbridge down towards the waveguide; b) white light profilometer image before actuation, c) microbridge profile measured at different bias voltages showing beam actuation via the gradient electric field (L = 40 μm microbridge test structure).

Download Full Size | PDF

The optical phase shift was measured using the same approach as before. Figure 6(a) shows a representative MZI fringe shift measured for different applied bias voltages (L_MEMS = 100 μm). A second device with L_MEMS = 40 μm was measured and the phase shift as a function of applied bias for both devices is shown in Fig. 6(b). The two devices achieve phase shifts of Δϕ≈3π/2 and Δϕ>π/2 at bias voltages of 36V and 60V, respectively. An accurate displacement vs. bias calibration could only be performed on a test bridge without a waveguide underneath (L_MEMS = 40 μm). Nonetheless, the calculated Δϕ and Δn_eff show general agreement with measurements for the L_MEMS = 40 μm device (Fig. 6(b) and 6(c)). More importantly, the phase shifters show substantially lower loss compared to the devices in Fig. 4 with a figure of merit FOM = 0.07dB/π (L_MEMS = 100 μm, 0.1 dB/1.5 π in Fig. 6(d)) and FOM = 0.53dB/ π (L_MEMS = 40 μm, 0.4 dB/0.75 π in Fig. 6(d)).The reduced loss is a direct consequence of the absence of metal in the suspended microbridge perturber and also results from the slightly thinner microbridge that minimizes coupling of optical power to the microbridge (see Appendix).

 

Fig. 6 Measured phase shifting with gradient electric field actuation: a) Measured MZI fringe shift for different gradient electric fields (L_MEMS = 100 μm); the fringes have been offset by 0.2 V increments; b) measured Δϕ for two devices (L_MEMS = 100 μm and L_MEMS = 40 μm) along with the calculated phase shift vs. bias for one device (L_MEMS = 40 μm); c) measured and calculated Δn_eff vs. air gap (L_MEMS = 40 μm); d) figure of merit (FOM): measured transmittance vs. Δϕ for both devices. The microrbridge is no longer protected by a top metal coating during sacrificial release and is etched substantially. The post-release microbridge thickness was measured to be 254 nm +/− 12 nm thick and the simulations were performed with the modified thickness. In contrast, the devices in Figs. 1-4 assume a 300 nm thick microbridge.

Download Full Size | PDF

4. Discussion

In terms of effective index tuning, our waveguide-coupled devices substantially exceed the Δn_eff measured in previous silicon nitride optomechanical phase shifters [24] and are comparable to previous work using gap plasmon waveguides [25]. We attribute the improved Δn_eff to the thin waveguide core design, which maximizes the evanescent field at the surface [27]. The large evanescent field maximizes the sensitivity to the presence of the dielectric microbridge perturber and enables a large Δn_eff to be achieved. In our measurements, the microbridge is actuated down towards the waveguide resulting in an increase in Δn_eff. We note that other actuator geometries can be envisioned in which the air gap separating the microbridge and waveguide is increased during actuation, for example by placing a side actuation electrode above the waveguide [38]. The increased gap decreases Δn_eff so that bi-directional tuning can be envisioned by using both types of actuators in a single device.

The gradient electric field actuated devices in Figs. 5-6 have shown that metal-coated bridges are not required for actuation and phase shifting. The advantage with this approach is the strong reduction in optical loss at small gaps due to the absence of metal. However, for sufficiently thick microbridges the waveguide optical mode can still couple to the microbridge (Appendix) leading to power transfer and, effectively, loss. An additional improvement to the phase shifters would therefore be to adiabatically taper [39] the width of the microbridge along its length to prevent any significant power coupling to the microbridge.

The power requirements for electrostatic actuators are inherently low since they consume virtually no power in the steady-state. The energy required to tune the phase shifter can be estimated from the device capacitance as 0.5 × C_MEMS × V_bias², where C_MEMS is capacitance of the MEMS electrostatic actuator. Using simple models based on microstrip transmission lines [40–42] we estimate a device capacitance of C_MEMS≤1 pF. Assuming a worst-case operation at 10 MHz continuous switching, a device capacitance C_MEMS = 1 pF and a 5 V switching voltage the switching energy is 12.5 pJ which gives an operating power of P = 12.5 pJ/100 ns = 125 μW/π-phase shift. This is still significantly lower power than thermo-optic phase shifters, which require in excess of 10 mW/π-phase shift [43]. Additional design optimization to reduce the device capacitance will further reduce the switching power. Furthermore, many applications require lower switching rates leading to a further reduction in power. Finally, a key advantage compared to thermo-optic phase shifters is the minimal steady-state electrical power required in electrostatic actuators. In contrast, thermooptic approaches require electrical power both during switching and at steady-state.

Comparing the two actuation schemes, we find that the devices actuated by a gradient electric field require much larger actuation voltages (Figs. 5-6) compared to the electrostatically-actuated devices (Figs. 2-4). A detailed modeling study for the two actuation mechanisms is found in ref [31]. The gradient electric force scales linearly with the microbridge thickness whereas the electrostatic force scales linearly with the microbridge width [31]. In general, electrostatic actuation of micromechanical structures is therefore more efficient since the spring constant of a microbridge has a linear width-dependence but a highly non-linear dependence on the thickness. Nonetheless, we emphasize that the actuation voltage required for the gradient electric field devices can be lowered substantially by reducing the separation between the side electrodes in order to increase the electric field.

In addition to enabling lower-power phase shifting compared to thermo-optic approaches, an optomechanical phase shifter also potentially enables much faser device speeds. In Fig. 7(a) and 7(b) a L_MEMS = 75 μm phase shifter similar to the devices in Fig. 4 is biased at V_dc = 4.5 V while a small-signal of amplitude V_ac = 100 mV is used to excite the microbridge. For sinusoidal actuation the device response can exceed frequencies of 1 MHz (Fig. 7(a)), and actuation with a square-wave drive indicates rise- and fall-times of around τ = 400 ns (Fig. 7(b)). This is substantially faster than most thermo-optic phase shifters while doing so without the need for a high-Q optical cavity. Figure 7(c) shows the fundamental mechanical resonance measured with a network analyzer demonstrating a f_0,m = 2.3 MHz and a mechanical quality factor Q_0,m = 103 (in air). The rise-/fall-time is essentially dominated by the mechanical resonance, i.e. τ = 1/f_0,m. Simulations show that by simply shortening the microbridge length the mechanical resonance frequency can exceed 10 MHz, which would enable switching speeds of τ<100 ns. By moving to other materials or device geometries, mechanical resonance frequencies approaching 1 GHz (i.e. nanosecond-range response times) may be possible in the future [29,44,45], although there will be a trade-off between higher-frequency operation and total phase shift due to the generally smaller microbridge displacements at higher frequencies.

 

Fig. 7 Measured frequency and temporal response for a L_MEMS = 75 μm phase shifter a) MZI response to a 1MHz sinusoidal drive (V_dc = 4.5V, V_ac = 100mV, λ_laser = 1502.0nm TE), b) MZI response to a 200 kHz square wave drive (V_dc = 4.5V, V_ac = 100mV, λ_laser = 1502.0nm TE), c) fundamental mechanical resonance mode measured with a network analyzer (V_dc = 4.5V, RF drive = −30dBm, λ_laser = 1525.4nm TE), d) COMSOL simulations of the fundamental mechanical resonance mode for a t_Si3NX = 300nm thick bridge with 35nm thick gold layer (blue) and for a t_Si3NX = 250nm bridge with no metal (red); the mechanical resonance measured in (c) is also shown for reference.

Download Full Size | PDF

As shown in Fig. 1, our weakly-confined waveguide design results in an evanescent field that extends hundreds of nanometers above the waveguide surface. This has implications not only for tuning n_eff but also for any optical forces that may arise. Although optical forces [36] in optomechanical structures are an active field of research, we expect the optical force to be modest for g>100 nm since the force is proportional to dn_eff/dg (see Appendix). The weak mode confinement design may therefore be advantageous in terms of device stability.

Although all of the measurements presented here were performed with the electric field in the wafer plane (TE-polarization), this is not a strict limitation of our optomechanical phase shifter architecture. For the metal-coated microbridge devices in Figs. 1-4 we use TE-polarization to prevent coupling to the hybrid metal-dielectric plasmonic mode (Appendix), which experiences higher loss. In contrast, the pure dielectric microbridges (Figs. 5 and 6) are not limited to a particular polarization and can be designed for operation with TM-polarized light.

5. Conclusion

In conclusion, we have demonstrated a broadband opto-electro-mechanical phase shifter enabling Δϕ>2π phase shifts via large-scale effective index tuning (Δn_eff = 0.01-0.1) in a L_MEMS≈100 μm device. Our waveguide-integrated MEMS-based phase shifter not only enables a large tunable Δn_eff comparable to liquid crystal [14] and gap plasmon [25] based approaches, the vertical architecture results in an extremely small device footprint occupying an area only slightly larger than the waveguide itself. The evanescent coupling between a standard dielectric waveguide and a MEMS microbridge perturber coated with metal (as in Figs. 3 and 4) may also provide a scalable approach to coupling to plasmonic or hybrid metal-dielectric modes or potentially gap plasmons [25]. Furthermore, actuation via gradient electric fields (as in Figs. 5 and 6) is versatile and can be implemented in any suspended dielectric structure. Both approaches may find wide application in many different areas to provide active functionality to many passive optical materials via broadband Δn_eff tuning at low electrical powers.

Appendix

A1. Variables.

See Table 1 for variables.

Table 1. Variables.

View Table

A2. Device fabrication

We start with commercial Si(100) wafers of 325 μm thickness with 5,000 nm thermal oxide (SiO₂) and 175 nm silicon nitride (LPCVD Si₃N₄). The waveguides and Mach-Zehnder interferometers (MZI’s) are defined by electron-beam lithography using MaN2403 resist followed by an ICP/RIE dry etch to a depth of 55 nm using SF₆/C₄F₈ chemistry.

Next, the micro-electro-mechanical (MEMS) layers are deposited using PECVD (SiO₂=230 nm and tensile SiN_X=310 nm; thicknesses verified by ellipsometry) after which Ti/Au=5 nm/35 nm is evaporated. The tensile stress in the as-deposited SiN_X is found to be σ≈100 MPa by performing test depositions and measuring the wafer curvature before and after deposition. We emphasize that the intrinsic stress calibration does not include the effects of the Ti/Au metal layers. For this reason, we also fabricated a number of test microbridges and measured their profile after sacrificial release. With an optimized process we can release microbridges up to L_MEMS=300 μm with a flat profile.

The MEMS bridges and side electrodes are patterned using MaN2403 resist. For ease of alignment the MEMS bridges are designed to be 3.0 μm wide, which is slightly wider than the rib waveguides, which have W_rib=2.5 μm. The pattern is then transferred into the Ti/Au layer using Ar-milling and into the SiN_X layer using an ICP/RIE etch with SF₆/C₄F₈ chemistry. We etch 110 nm into the SiO₂ sacrificial layer (measured by contact profilometry). The devices in Figs. 1-4 and Fig. 7 are cleaved to expose the waveguide facets and are then released using buffered-HF (BHF) followed by critical point drying. After sacrificial release the microbridges remain flat; however, the side electrode, which acts like a compliant cantilever, reproducibly collapses towards the substrate due to the built-in strain gradient [46] between the SiN_X and Ti/Au layers. This enables the vertical electrostatic actuator to be realized using only a single and self-aligned lithography step. The actuator process results in a 230nm vertical separation (not considering the waveguide rib) and a 1 μm lateral gap between the microbridge and side electrode.

Before sacrificial etching, the devices in Figs. 5-6 have their Ti/Au layers removed using gold etch TFA and piranha clean (H₂SO₄:H₂O₂). The side electrodes are patterned using electron-beam lithography with ZEP-520A resist followed by Ti/Au=5 nm/55 nm evaporation and lift-off. The devices are then released in BHF followed by critical point drying. The bridge thickness is reduced to 254 nm±12 nm (measured using contact profilometry on collapsed cantilevers) since the SiN_x is not protected by Ti/Au during the release etch.

A3. Data analysis

A LabVIEW program extracts the average phase over the wavelength scan and records the average fringe spacing or FSR (free-spectral range) for each bias voltage. The MZI’s BAR output is I_BAR(λ)∝1+cos(ϕ_MZI ± ϕ_MEMS), where ϕ_MZI is the phase introduced by the MZI’s built-in path length imbalance and ϕ_MEMS is the phase contribution from the MEMS microbridge perturber. The sign of ϕ_MEMS depends on the location of the microbridge (L_MEMS) relative to the built-in phase shift (ΔL_MZI).

The effective index tuning (Δn_eff) is obtained from Δϕ=(2π/λ)∫_x=0-L Δn_eff[m(x)]dx, where λ is the laser wavelength, L=L_MEMS is the microbridge length, Δn_eff[m(x)] is the change in effective index due to the evanescent field interaction between the optical mode and the microbridge, and m(x) is the microbridge deflection or mode profile that represents the microbridge-waveguide gap (height) at position x. A typical microbridge profile during actuation is shown in Fig. 8a from which we find that a cosine gives a good approximation to the measured bridge profile. Therefore, Δϕ=2πΔn_eff(g)L_MEMS/λ_avg, where λ_avg is the center wavelength of the MZI spectrum and Δn_eff(g) is the effective index tuning assuming a constant average air gap g (Fig. 8b). The assumption of a constant average air gap is consistent with previous work [24] and does not affect the accuracy of the extracted Δn_eff.

 

Fig. 8 a) Measured beam bending and cosine fit for an L_MEMS = 50 μm device, b) model showing relation between microbridge deflection (Δz), minimum waveguide-microbridge gap (g_min) and average gap (g).

Download Full Size | PDF

A4. Displacement vs. bias calibration

The microbridge displacement vs. bias calibration is performed using a white-light optical profilometer (Zygo). A bias voltage is applied to the respective MZI-MEMS bridge and the beam deflection is measured. The maximum deflection is then extracted from a linescan across the bridge surface to obtain the displacement vs. bias calibration. We measured the displacement vs. bias for a variety of MEMS actuator geometries. By performing a curve fit to the data, we can convert the applied bias to a microbridge displacement; this lets us plot both measured and simulated Δn_eff vs. air gap on the same plot (as in Fig. 4b and 4(c) and Fig. 6(c)). Similarly, we can convert from a simulated gap to an applied bias in order to plot the measured and simulated phase shift vs. bias on a single plot (as in Fig. 4(a) and Fig. 6(b)).

For the gradient electric field device (Figs. 5-6) it is difficult to measure the bridge deflection accurately since the high reflectivity of the two side electrodes, the bridge transparency, and the rib waveguide underneath the microbridge result in a reduced signal from the microbridge surface which makes optical profilometry a non-ideal method for calibrating the displacement in these devices. For this reason, the displacement calibration for the gradient electric field actuation devices (Fig. 6) could only be performed on a test bridge without a waveguide underneath. In contrast, the displacement calibration for the electrostatic devices (Fig. 4) were performed on MEMS bridges with waveguides (i.e. opto-electro-mechanical phase shifters).

A5. Optical simulations: other modes and device configurations

A5.1 Microbridge modes

The device in Fig. 1 also supports a mode that resides predominantly in the MEMS microbridge (Fig. 9). However, the loss is significantly higher than for the waveguide mode due to the larger mode overlap with the Ti/Au(5nm/35nm) layers. The field component is predominantly E_y which indicates that this is a TM-polarized mode. We therefore expect that this mode does not contribute substantially to the device operation since we perform our measurements with TE-polarization (launch horizontally-polarized light).

 

Fig. 9 Microbridge hybrid dielectric/metal plasmonic mode: simulated effective index and metal-induced loss for the microbridge mode in which the light is confined in the microbridge. The E_y-field component dominates.

Download Full Size | PDF

A5.2 Effect of microbridge thickness

We performed simulations for various microbridge thicknesses for the device in Fig. 5 (microbridge with no Ti/Au metal) to account for the partial etch of the microbridge during sacrificial release. We find that for thinner microbridges (e.g. t_MEMS=200 nm) the effective index decreases compared to thicker bridges (e.g. 250 nm). Nonethless, there is still substantial effective index tuning (Fig. 10a). For bridges of 300 nm thickness the structure supports both waveguide and microbridge modes and the device more closely resembles a directional coupler. The simulations show that as the microbridge-waveguide separation (g) is decreased the waveguide mode n_eff decreases since this mode is pulled upwards and experiences a larger overlap with the air region above the waveguide; in contrast, the microbridge mode n_eff increases with decreasing g (Fig. 10b). Furthermore, for small g the waveguide mode is no longer supported and all the optical power resides in the microbridge mode. The simulations in Fig. 10b suggest that device configurations can be envisioned in which actuation of a microbridge perturber down towards the waveguide can lead to a decrease in effective index so that bi-directional tuning may be possible by using thin (as in Fig. 10a) or thick microbridges (as in Fig. 10b).

 

Fig. 10 a) Effective index tuning vs. air gap for microbridge thicknesses t_MEMS = 200 nm and 250 nm. The geometry resembles the devices in Figs. 5 and 6 in which the MEMS microbridge consists of a SiN_X layer and no TiAu metal; for these structures there is no microbridge optical mode supported, b) effective index tuning for a microbridge thickness t_MEMS = 300 nm in which both microbridge and waveguide modes are supported; this structure more closely resembles a directional coupler in that the preferred optical mode is the microbridge mode. For g<300 nm the waveguide mode is not supported and only the TE microbridge mode exists.

Download Full Size | PDF

A6. Optical forces

Optical gradient forces are an active area of research. However, in an opto-mechanical phase shifter any optical force can lead to device instability or other undesirable device performance. Although we have not measured any optical forces in our devices, we can estimate the induced gradient optical force from [47]: F_optical=(n_group/n_eff)(L_MEMS/c)(dn_eff/dg)P, where P is the optical power. In Fig. 11 we plot the calculated optical force per unit power and unit length for our device based on the dn_EFF/dg simulations from Fig. 1. We find that the optical force is substantially smaller compared to similar waveguide structures made from silicon [47]. We attribute the smaller force to the weak mode confinement in our Si₃N₄ waveguide design so that the evanescent field extends hundreds of nanometers away from the waveguide surface leading to a smaller dn_Eeff/dg and hence a smaller F_optical.

 

Fig. 11 Simulated optical gradient force. Silicon waveguide devices experience a larger force due to the larger dn_eff/dg (Si waveguide: from ref [47].).

Download Full Size | PDF

Acknowledgments

We thank NRL/ONR for funding this work under a base 6.1 program (ONR 61153N). We also thank the NRL Nanoscience Institute (NSI) staff for cleanroom access and fabrication assistance. MWP thanks J.B. Khurgin (JHU) for helpful discussions.

References and links

1. G. Keiser, Optical Fiber Communications, 3rd ed. (McGraw Hill, 2003).

2. S. E. Miller, T. Li, and E. A. J. Marcatili, “Research toward optical-fiber transmission-systems,” Proc. IEEE 61(12), 1703–1751 (1973). [CrossRef]  

3. E. A. Marcatili, “Bends in optical dielectric guides,” AT&T Tech. J. 48, 2103 (1969).

4. T. L. Koch and U. Koren, “Semiconductor photonic integrated-circuits,” IEEE J. Quantum Electron. 27(3), 641–653 (1991). [CrossRef]  

5. H. Lee, T. Chen, J. Li, O. Painter, and K. J. Vahala, “Ultra-low-loss optical delay line on a silicon chip,” Nat. Commun. 3, 867 (2012), doi:. [CrossRef]   [PubMed]  

6. J. F. Bauters, M. J. R. Heck, D. D. John, J. S. Barton, C. M. Bruinink, A. Leinse, R. G. Heideman, D. J. Blumenthal, and J. E. Bowers, “Planar waveguides with less than 0.1 dB/m propagation loss fabricated with wafer bonding,” Opt. Express 19(24), 24090–24101 (2011). [CrossRef]   [PubMed]  

7. E. J. Klein, D. H. Geuzebroek, H. Kelderman, G. Sengo, N. Baker, and A. Driessen, “Reconfigurable optical add-drop multiplexer using microring resonators,” IEEE Photonics Technol. Lett. 17(11), 2358–2360 (2005). [CrossRef]  

8. R. Amatya, C. W. Holzwarth, H. I. Smith, and R. J. Ram, “Efficient thermal tuning for second-order silicon nitride microring resonators,” in 2007 Photonics in Switching, (2007), pp. 149–150.

9. K. Ikeda, R. E. Saperstein, N. Alic, and Y. Fainman, “Thermal and Kerr nonlinear properties of plasma-deposited silicon nitride/ silicon dioxide waveguides,” Opt. Express 16(17), 12987–12994 (2008). [CrossRef]   [PubMed]  

10. R. S. Jacobsen, K. N. Andersen, P. I. Borel, J. Fage-Pedersen, L. H. Frandsen, O. Hansen, M. Kristensen, A. V. Lavrinenko, G. Moulin, H. Ou, C. Peucheret, B. Zsigri, and A. Bjarklev, “Strained silicon as a new electro-optic material,” Nature 441(7090), 199–202 (2006). [CrossRef]   [PubMed]  

11. M. Cazzanelli, F. Bianco, E. Borga, G. Pucker, M. Ghulinyan, E. Degoli, E. Luppi, V. Véniard, S. Ossicini, D. Modotto, S. Wabnitz, R. Pierobon, and L. Pavesi, “Second-harmonic generation in silicon waveguides strained by silicon nitride,” Nat. Mater. 11(2), 148–154 (2011). [CrossRef]   [PubMed]  

12. J. B. Khurgin, T. H. Stievater, M. W. Pruessner, and W. S. Rabinovich, “On the origin of the second-order nonlinearity in strained Si-SiN structures,” J. Opt. Soc. Am. B 32(12), 2494–2499 (2015). [CrossRef]  

13. M. Borghi, M. Mancinelli, F. Merget, J. Witzens, M. Bernard, M. Ghulinyan, G. Pucker, and L. Pavesi, “High-frequency electro-optic measurement of strained silicon racetrack resonators,” Opt. Lett. 40(22), 5287–5290 (2015). [CrossRef]   [PubMed]  

14. S. R. Davis, G. Farca, S. D. Rommel, S. Johnson, and M. H. Anderson, “Liquid crystal waveguides: new devices enabled by > 1000 waves of optical phase control,” Emerging Liquid Crystal Technologies V 7618 (2010).

15. J. W. Judy, “Microelectromechanical systems (MEMS): fabrication, design and applications,” Smart Mater. Struct. 10(6), 1115–1134 (2001). [CrossRef]  

16. M. C. Wu, A. Solgaard, and J. E. Ford, “Optical MEMS for lightwave communication,” J. Lightwave Technol. 24(12), 4433–4454 (2006). [CrossRef]  

17. X. H. Ma and G. S. Kuo, “Optical switching technology comparison: Optical MEMS vs. other technologies,” IEEE Commun. Mag. 41, S16–S23 (2003).

18. Q. Xu, L. Chen, M. G. Wood, P. Sun, and R. M. Reano, “Electrically tunable optical polarization rotation on a silicon chip using Berry’s phase,” Nat. Commun. 5, 5337 (2014), doi:. [CrossRef]   [PubMed]  

19. D. T. Fuchs, H. B. Chan, H. R. Stuart, F. Baumann, D. Greywall, M. E. Simon, and A. Wong-Foy, “Monolithic integration of MEMS-based phase shifters and optical waveguides in silicon-on-insulator,” Electron. Lett. 40(2), 142–143 (2004). [CrossRef]  

20. C. K. Madsen, J. A. Walker, J. E. Ford, K. W. Goossen, T. N. Nielsen, and G. Lenz, “A tunable dispersion compensating MEMS all-pass filter,” IEEE Photonics Technol. Lett. 12(6), 651–653 (2000). [CrossRef]  

21. T. Ikeda, K. Takahashi, Y. Kanamori, and K. Hane, “Phase-shifter using submicron silicon waveguide couplers with ultra-small electro-mechanical actuator,” Opt. Express 18(7), 7031–7037 (2010). [CrossRef]   [PubMed]  

22. W. C. L. Hopman, K. O. van der Werf, A. J. Hollink, W. Bogaerts, V. Subramaniam, and R. M. de Ridder, “Nano-mechanical tuning and imaging of a photonic crystal micro-cavity resonance,” Opt. Express 14(19), 8745–8752 (2006). [CrossRef]   [PubMed]  

23. P. T. Rakich, M. A. Popovic, M. R. Watts, T. Barwicz, H. I. Smith, and E. P. Ippen, “Ultrawide tuning of photonic microcavities via evanescent field perturbation,” Opt. Lett. 31(9), 1241–1243 (2006). [CrossRef]   [PubMed]  

24. M. Poot and H. X. Tang, “Broadband nanoelectromechanical phase shifting of light on a chip,” Appl. Phys. Lett. 104(6), 061101 (2014), doi:. [CrossRef]  

25. B. S. Dennis, M. I. Haftel, D. A. Czaplewski, D. Lopez, G. Blumberg, and V. A. Aksyuk, “Compact nanomechanical plasmonic phase modulators,” Nat. Photonics 9, 267–273 (2015).

26. J. B. Khurgin, “How to deal with the loss in plasmonics and metamaterials,” Nat. Nanotechnol. 10(1), 2–6 (2015). [CrossRef]   [PubMed]  

27. M. W. Pruessner, D. Park, T. H. Stievater, D. A. Kozak, and W. S. Rabinovich, “An optomechanical transducer platform for evanescent field displacement sensing,” IEEE Sens. J. 14(10), 3473–3481 (2014). [CrossRef]  

28. Q. P. Unterreithmeier, E. M. Weig, and J. P. Kotthaus, “Universal transduction scheme for nanomechanical systems based on dielectric forces,” Nature 458(7241), 1001–1004 (2009). [CrossRef]   [PubMed]  

29. Y. I. Sohn, M. J. Burek, V. Kara, R. Kearns, and M. Loncar, “Dynamic actuation of single-crystal diamond nanobeams,” Appl. Phys. Lett. 107(24), 243106 (2015), doi:. [CrossRef]  

30. S. Schmid, M. Wendlandt, D. Junker, and C. Hierold, “Nonconductive polymer microresonators actuated by the Kelvin polarization force,” Appl. Phys. Lett. 89(16), 163506 (2006). [CrossRef]  

31. S. Schmid, C. Hierold, and A. Boisen, “Modeling the Kelvin polarization force actuation of micro- and nanomechanical systems,” J. Appl. Phys. 107(5), 054510 (2010). [CrossRef]  

32. C. Mutliphysics, COMSOL Multiphysics, 2014.

33. P. M. Osterberg and S. D. Senturia, “M-TEST: A test chip for MEMS material property measurement using electrostatically actuated test structures,” J. Microelectromech. Syst. 6(2), 107–118 (1997). [CrossRef]  

34. R. Pethig, “Review article-dielectrophoresis: status of the theory, technology, and applications,” Biomicrofluidics 4(2), 022811 (2010). [CrossRef]   [PubMed]  

35. C. Q. Yi, C. W. Li, S. L. Ji, and M. S. Yang, “Microfluidics technology for manipulation and analysis of biological cells,” Anal. Chim. Acta 560(1-2), 1–23 (2006). [CrossRef]  

36. M. L. Povinelli, M. Loncar, M. Ibanescu, E. J. Smythe, S. G. Johnson, F. Capasso, and J. D. Joannopoulos, “Evanescent-wave bonding between optical waveguides,” Opt. Lett. 30(22), 3042–3044 (2005). [CrossRef]   [PubMed]  

37. A. Ashkin, J. M. Dziedzic, and T. Yamane, “Optical trapping and manipulation of single cells using infrared laser beams,” Nature 330(6150), 769–771 (1987). [CrossRef]   [PubMed]  

38. M. W. Pruessner, D. Park, T. H. Stievater, D. A. Kozak, and W. S. Rabinovich, “An opto-electro-mechanical phase shifter,” in CLEO: Laser Science to Photonic Applications (OSA, 2015), paper STu1F.5.

39. T. J. Seok, N. Quack, S. Han, R. S. Muller, and M. C. Wu, “Large-scale broadband digital silicon photonic switches with vertical adiabatic couplers,” Optica 3(1), 64–70 (2016). [CrossRef]  

40. T. G. Bryant and J. A. Weiss, “Parameters of microstrip transmission lines and of coupled pairs of microstrip lines,” IEEE Trans. Microwave Theory Mt16, 1021 (1968).

41. Y. S. Sohn, J. C. Lee, H. J. Park, and S. I. Cho, “Empirical equations on electrical parameters of coupled microstrip lines for crosstalk estimation in printed circuit board,” IEEE Trans. Adv. Packag. 24(2), 521–527 (2001). [CrossRef]  

42. W. Q. Che, L. M. Gu, and Y. L. Chow, “Formula derivation and verification of characteristic impedance for offset double-sided parallel strip line (DSPSL),” IEEE Microwave Wireless Comms. 20(6), 304–306 (2010). [CrossRef]  

43. M. R. Watts, J. Sun, C. DeRose, D. C. Trotter, R. W. Young, and G. N. Nielson, “Adiabatic thermo-optic Mach-Zehnder switch,” Opt. Lett. 38(5), 733–735 (2013). [CrossRef]   [PubMed]  

44. X. M. Henry Huang, C. A. Zorman, M. Mehregany, and M. L. Roukes, “Nanoelectromechanical systems: Nanodevice motion at microwave frequencies,” Nature 421(6922), 496 (2003). [CrossRef]   [PubMed]  

45. A. Gaidarzhy, M. Imboden, P. Mohanty, J. Rankin, and B. W. Sheldon, “High quality factor gigahertz frequencies in nanomechanical diamond resonators,” Appl. Phys. Lett. 91(20), 203503 (2007). [CrossRef]  

46. M. W. Pruessner, T. King, D. Kelly, R. Grover, L. Calhoun, and R. Ghodssi, “Mechanical property measurement of InP-based MEMS for optical communications,” Sensors Actuat. A 105(2), 190-200 (2003). [CrossRef]   [PubMed]  

47. M. Li, W. H. P. Pernice, C. Xiong, T. Baehr-Jones, M. Hochberg, and H. X. Tang, “Harnessing optical forces in integrated photonic circuits,” Nature 456(7221), 480–484 (2008). [CrossRef]   [PubMed]