1. Introduction
Electro-Optic (EO) waveguide modulators are essential in optical communications to convert electrical signals to optical signals. EO modulators produce cleaner modulated signals and provide faster modulation rates than direct modulation options [
1]. The majority of commercial EO modulators utilize lithium niobate which has a maximum r
33 tensor of 30.8 pm/V [
2] limiting the reduction in
Vπ (on-off voltage) for most modest laser power. To achieve devices with a low
Vπ, commercial lithium niobate devices must increase the interaction length (L) of the electro-optic effect. This results a figure of merit
Vπ-L being vital describe optimization of device size and
Vπ. Alternative to lithium niobate are electro optic polymers, which consist of a polymer matrix, commonly polycarbonate or PMMA, and nonlinear EO chromophore molecules [
3]. EO polymers with r
33 values in excess of 100 pm/V have been reported [
4,
5] allowing low voltage designs since the tensor is inversely proportional to the modulator on-off voltage. This high r
33 value also allows for low
Vπ-L values to be achievable.
For EO polymers to obtain an r
33, the chromophores must first be aligned through a poling process [
6]. Poling involves heating the EO polymer structure to its glass transition temperature while applying a strong electric field across the EO polymer. Various methods of efficient poling to reach maximum r
33 while avoiding dielectric breakdown have been demonstrated including corona poling [
7], and contact poling [
6]. Corona poling is not used due to the need for voltages in excess of 3 kV, and non-uniform polling of EO-Layers [
7].
Utilizing the contact poling method/process involves electrodes surrounding the EO-polymer to distribute an electric field. Contact poling can be used to pole a single film of EO polymer or full device stacks. The highest r
33′s reported have been for single films of EO-polymers maximizing the static field within the active layer by eliminating the clad resistive layers. Single films of the EO-polymer can be directly sandwiched between two electrodes to get a strong electric field across the polymer layer and provide efficient poling. Single EO-Polymer films can be poled and their r
33 can be measured using the teng-man measurement technique [
8]. However, this is not practical for waveguide devices as the electrodes would quickly attenuate any optical power. Using a full device stack of a cladding-core-cladding waveguide structure between the polling electrodes allows the waveguides to be fully fabricated and then poled. A major advantage of contact poling is that a device can be poled multiple times under various conditions and can be re-poled if the chromophore alignment relaxes. Relaxation of chromophore alignment can impact the lifespan of an EO polymer device [
9,
10]. The contact poling method results in the electric field being distributed between the cladding and core layers, requiring a higher applied voltage to get equivalent poling results. This higher voltage introduces additional risks to the poling process as the high voltage may cause arcing between electrodes or cause a dielectric break down in any of the device layers. A preferred method to ensure maximization of r
33 for a wave-guiding device structure is to pole utilizing the mid-process poling technique [
6]. Mid-process poling is performed after the core layer has been fabricated, allowing poling to be carried out on a partial stack of cladding-core layers, reducing the required applied voltage, and maximizing r
33.
The electrode size is also a significant factor in the poling process. Electrode size and position for an EO-Modulator are often constrained by the RF operational design frequency. Electrodes are patterned over the modulating part of the waveguide structure to pole the necessary section of the waveguide. Having the electrode over only the waveguide reduces the risk of dielectric breakdown. EO-Modulator designs may require a larger electrode for their end application, such as in the case of an EO antenna [
11]. Additionally large area poling can be used in poling full wafers before individual devices are diced. The use of a large poling electrode will result in the poling of a larger section of the core material but increases the risk of shorting through the device and rendering it unusable. Because a larger area is being poled there is higher probability of the poling E-field finding a material imperfection. Any imperfection such as a dust particle or an air bubble in the poled area can result in dielectric breakdown.
Inherent to the polling process of EO-polymers is the risk of dielectric breakdown. This risk is introduced by the combination of heating the polymer to the glass transition temperature which lowers the breakdown voltage while applying a high strength field to the device. For these reasons the effects of poling SEO100c at lower voltages and lower temperatures have been examined, minimizing of the risk of dielectric breakdown at the expense of a lower r33.
A large r
33 of 226 pm/V has been reported in a TiO
2/SEO100c/TiO
2 multilayer structure measured using the teng-man method [
8]. An r
33 of 166 pm/V has been reported for a directional coupler using SEO100c with a proprietary cladding material [
4]. This paper reports and quantifies the research of r
33 measurements poled under increasingly stronger electric fields of the EO polymer SEO100c for application designers. Of particular interest is the achievable r
33 in a practical waveguide EO modulator, constructed using commercial polymers under various poling conditions. A polymer modulator made from commercially available products demonstrates the progress that has been made in EO polymer design.
3. Waveguide Fabrication
The first step in the fabrication process was using e-beam deposition to deposit 10 nm of chrome adhesion layer followed by 100 nm of gold on a silicon wafer to form the bottom electrode. A 4.3 μm layer of Norland Products NOA 73 was spun on using a static spinning technique, then UV cured for 30 min with low intensity UV to form a smooth bottom cladding layer.
To create the core layer, a vial with 0.5 g of SEO100c had 7.45 g of dibromomethane added. The solution was then mixed on a rotator at 35 rpm for 8.5 h. Once mixed, the SEO100c was passed through a 0.2 μm nylon filter and allowed to settle for 30 min. After settling, the SEO100c was deposited by a static spinning process with a spin speed of 1000 rpm for 40 s producing a 2.7 μm-thick core layer. The wafer was then baked in a vacuum oven at 75 °C for 18 h.
A layer of Futurrex’s PR1-1000a photoresist was spun on, photolithography was performed to define 6 µm wide rib waveguides, and the wafer was developed in Futurrex’s RD6. The PR-1 photoresist was used because it was known to not chemically damage the SEO100c core layer [
14]. Following development, it was etched for 125 s in an O
2 environment using a reactive ion etching process. The etched structures were measured at an etch depth of 667 nm. Using harsher chemicals such as acetone or resist remover potentially degrade the EO-polymer, therefore the PR-1 was removed by flood exposure to UV followed by development in RD6.
A second 4.3 μm cladding layer of NOA73 was spun on and UV cured for 30 min. The top electrode was formed by e-beam deposition of 10 nm of chrome followed by 100 nm of gold. PR-1 was spun on, and the top electrodes were patterned using photolithography. The top electrodes were then wet etched using type TFA gold etchant and CR7 chrome etchant to etch the layers of the top electrode. The PR-1 was again exposed to UV light and developed in RD6. A shadow mask was placed over the wafer and the bottom electrodes were accessed by RIE in an O
2 environment. Etching through the polymer layers was performed by alternating periods of etching with cool off periods to prevent the wafer from overheating. The completed stack is shown in
Figure 1.
Devices were then diced apart in single passes using a nickel bonded blade with a spindle speed of 30,000 RPM with a feed rate of 40 mils/s. These dicing parameters were selected to produce clean and consistent end faces.
4. Poling of Electro-Optic Polymers
Poling an EO-polymer to align the chromophores, is a critical step in creating an active EO polymer. The data sheet provided by Soluxra LLC specifies the maximum alignment for SEO 100c occurs at an electric field strength of 100 V/μm across the electro optic layer which would produce an r
33 of 140 pm/V [
15]. In the test device the electrodes had 3 layers between them, a bottom cladding layer 4.3 μm thick of NOA 73, a 2.7 μm core layer of SEO 100c, and a top cladding layer 4.3 μm thick of NOA 73. The resistivities at 135 °C of NOA 73 and SEO 100c are known to be 3.3 × 10
10 Ωcm and 3.0 × 10
10 Ωcm, respectively [
6]. Equation (1) shows how the voltage across the core layer of SEO100c was calculated.
where
Vcore is the voltage applied to the core,
Vtotal is the voltage applied to the full stack.
ρcore and
ρclad are the resistivities of the core and cladding layers. Finally,
dcore and
dclad are the thicknesses of the core and cladding layers, 2.7 µm and 4.3 µm, respectively. Applying the voltage divider rule, the strength of the electric field across the EO-polymer was determined to be 21% of the total electric field strength. Although this is a low value it can be increased in future devices by utilizing thinner cladding layers, or through substitution of the cladding material to a material with a lower resistivity.
Figure 2 shows the setup used for poling. A small oven was used to heat the device. A high voltage DC power supply had the positive lead connected to top electrode via a probe and the negative lead connected to a pico-ammeter through a HV protection circuit, then to the bottom electrode of the device. The current data from the ammeter and temperature data from a k-type thermocouple were digitally logged during the poling process.
The device was placed in the oven, elevated to allow heating from both the top and the bottom, and connected with probes to a high voltage power supply. The oven was turned on and the device began to be heated to reduce thermal shock. When the temperature reached 50 °C. A DC voltage was applied across the device and heating continued until the oven reached 135 °C. The hold temperature of 135 °C ensures that the glass transition temperature of 135 °C is not exceeded while maximizing r
33. At the glass transition temperature chromophores can move freely to orientate themselves in the direction of the applied electric field. Exceeding the glass transition temperature will result in dielectric breakdown, causing the device to electrically short, destroying the device. As soon as 135 °C was reached the heating element of the oven was switched off and the device was allowed to cool. Once the oven had cooled to 100 °C the oven was opened to facilitate rapid cooling. The voltage was turned off when the device was below 50 °C and then allowed to cool to room temperature. Leaving the voltage on while the device is cooling prevents the relaxation of chromophore alignment while the device is at an elevated temperature.
Figure 3 shows the temperature and current data collected from poling with increasing poling voltage. The poling current is not correlated with the electro-optic coefficient, but it does indicate that charge moves more freely at elevated temperatures. Seeing a temperature-dependent current spike as the polymer approaches the glass transition temperature is a positive indication that the device is heating during the poling process as the resistivity of the polymer layers decrease with temperature.
5. r33 Measurement Setup
The r
33 values of the poled devices were measured. A Mach-Zehnder interferometer (MZI) was constructed with optical fiber and the device was inserted into one leg of the MZI, the testing setup is shown in
Figure 4. A 100 mW diode laser at 1545 nm was used for testing of the device. This power was split with 90% of the power to the device leg (to overcome waveguide losses and coupling losses) and 10% of the power within the reference leg. For the device leg, a polarization controller was inserted before the splitter, to ensure the light launched into the device was TM polarized. TM light was used because the fabrication process placed the electrodes above and below the waveguides. The polarization of the light is important as the r
33 only effects the index of refraction in the direction the EO-polymer was poled. In order for this effect to be observed the light must be TM polarized. The TM polarized light was coupled into the single mode waveguide with a 6 µm PM fiber mounted on a three-axis XYZ micrometer stage. The output of the device was collected by a 9 µm PM fiber mounted on an identical three-axis XYZ stage. The reference leg first passed through a variable attenuator to allow for the power to be matched to that of the device leg. The variable attenuator can cause a change in polarization that must be corrected through a second polarization controller to match the device leg’s polarization. Both legs were then recombined in a 50/50 combiner and went to an optical power meter connected to an oscilloscope. A function generator was used to apply a triangular wave to the device. This signal was monitored on the oscilloscope as the optical power modulation was observed.
Data was collected for frequencies of 40 Hz, 100 Hz, 500 Hz, and 1k Hz using the oscilloscope.
Figure 5 shows an example modulation to measure
Vπ.
Vπ was measured for each set of data and r
33 values were calculated using Equation (2) [
16].
where
λ is the wavelength (1545 nm),
d is the thickness of the core layer (2.7 µm),
n is the index of refraction of the core (1.65) [
6,
13] and
Γ is the overlap integral is 1 for vertically sandwiched electrodes because the entirety of the waveguide is covered by the uniform electric field.
Vcore is the percentage of
Vπ applied across the core layer, found to be 21% of the total field as described in Equation (1).
L is the modulation length over which the voltage is applied which varied from 1.73 to 3.63 cm.
Vπ is the change in voltage applied to the device required to change from maximum to minimum optical power. As previously described,
Vπ was measured directly. From that measurement, r
33 could be calculated using Equation (2). It is important to note that because only the electric field in the EO- polymer will contribute to the modulation, the voltage across the 2.7 μm of the core layer was found using Equation (1) and the resistivities of NOA 73 and SEO 100c. The poling, measurement, and calculation processes were repeated for multiple devices at various poling voltages ranging from 50 to 800 V.