Influence of an Al2O3 surface coating on the response of polymeric waveguide sensors

The responses of a polymer ridge waveguide Young interferometer with and without a bilayer of Al2O3/TiO2, fabricated by atomic layer deposition, are studied and compared when applied as an aqueous chemical sensor. The phase shift of the guided mode, as a result of the change in refractive index of the cover medium, is monitored. The results indicate that the over-coating affects the linearity of the sensor response. The effect of concentration on the linearity of the sensor response is investigated by applying different concentrations of water-ethanol solution. Although the performance of the sensor is improved by the additional layers, the study reveals a non-monotonic behavior of the device. We show that it comes mainly from the adsorption of ethanol molecules on the surface of the films. Such an understanding of the platform is crucial for sensing of analytes involving polar molecules.


Introduction
Optical sensors are compact, sensitive and immune to electromagnetic interference. Among the optical sensors, integrated photonic sensors based on on-chip waveguides and fibers are attractive for labeled and label-free detection. However, optical fiber sensors are limited in terms of materials, fabrication, geometries, miniaturization, and integration in a single chip [1,2]. Waveguides enable the use of various structures together with a large choice of materials facilitating the miniaturization of the devices depending on the sensing mechanism.
Polymer waveguides have potential for disposable and low cost sensors, especially for biosen-sor applications [3][4][5]. Typically, in a polymer waveguide, the evanescent tail of the modal field interacting with the cover medium (or analyte) is relatively weak because of the low refractive index contrast between the core and the over-cladding of the waveguide. Different solutions have been proposed for improving the sensitivity by increasing the portion of the evanescent tail of an optical mode of a low refractive index waveguide inside the cover medium, and therefore enhancing the interaction between light and matter. One technique is based on adding a high refractive index dielectric material, e. g., Ta 2 O 5 [6], SiN [7] or TiO 2 [8], or a metallic layer [9,10], on the low refractive index waveguide, which pushes the mode toward the boundaries of the waveguide. Another solution is the optimization of the waveguide structure by modifying the waveguide geometry, and therefore shifting the localization of the modal field closer to the surface or in the sensing region [11][12][13][14]. It should be noted that in the first technique, the use of additional layers on the waveguide structure may drastically change the surface chemistry and optical properties of the waveguide, which can modify the response of the sensor. The second solution may lead to an increase of the losses. In this article we investigate the limitations of the first method, often chosen for surface enhancement and functionalization. Titanium dioxide, TiO 2 , is a promising material for a variety of applications due to its high refractive index and transparency in a wide wavelength range [15,16]. In addition, TiO 2 is a bio-compatible material used in pharmaceutical and medical applications [17,18]. However, deposition techniques of TiO 2 have a major impact on the optical and chemical properties of the final product. Atomic Layer Deposition (ALD) is one of the most effective techniques to deposit a uniform layer of TiO 2 with a well-controlled thickness at rather low temperature [19,20]. The ALD technique is based on the reaction of the molecules of reductant and oxidant gases with the device surface, which results in a layer-by-layer deposition of the desired material [21]. Different precursors can be used for depositing TiO 2 films. These include TiCl 4 , TiI 4 , Ti(OEt) 4 , as a reductant, and H 2 O or H 2 O 2 , as an oxidant [22,23]. However, TiCl 4 and H 2 O, which are used in this study, are among the most common precursors used for depositing TiO 2 owing to the low deposition temperature and relatively high growth rate for achieving amorphous TiO 2 films. Recently, an increased interest has arisen for this material in biosensing mainly due to its potential for applications in enhanced Raman and fluorescent sensing [24,25].
Aluminum oxide, Al 2 O 3 , is one of the most robust and stable materials to be used in optical and electrical applications. It has a very good resistance to chemicals and has been used as a protective and a barrier layer [26,27]. Realizing a uniform, amorphous layer with a wellcontrolled thickness is crucial in many applications [28,29]. ALD processed Al 2 O 3 layers are usually grown by using trimethylaluminum (TMA) and H 2 O or O 3 precursors.
Coatings around waveguide is now a common solution for improving photonic devices. In particular, applications of ALD TiO 2 as a gas sensor have been investigated theoretically and experimentally in different works [30,31]. Application of Al 2 O 3 as a humidity sensor has also been reported [32,33]. However, thorough investigations related to aqueous sensing with ALD processed TiO 2 and Al 2 O 3 have not yet been done. In this paper, we present a study on the influence of the ALD layers on a polymer waveguide in the case of aqueous homogeneous sensing of polar solutions. Our study case is a polymer waveguide based Young interferometer [8], sketched in Fig. 1. The chosen polymer is Ormocore from Micro Resist Technology GmbH [34]. It is known to be an excellent material for low loss optical waveguides and nano-imprinting [35,36]. Two cases are compared: 1) the interferometer without a coating; 2) the interferometer with a bi-layer of Al 2 O 3 /TiO 2 on the vertical sides only. The Al 2 O 3 being used as an etch stop-layer remains also on top of the waveguides and may affect the surface chemistry of the sensor. Therefore this layer has to be taken into account in the discussion and explanation of the experimental results, although its contribution on the fluctuations of the optical mode properties is relatively weak compared to the other layers.

Sensor layout, design, and fabrication
Polymer ridge waveguides were fabricated from Ormocore having a refractive index n = 1.54 at λ = 975 nm on a silicon wafer with a 3 μm thermal silicon dioxide layer. This polymer has a relatively high glass transition temperature of T g = 270 • C, which makes it stable upon thermal load during ALD processes. In order to optimize the sensitivity of the waveguide, the optical confinement factors, defined in Eq. (1), were calculated in different region of the waveguide using the Fourier Modal Method (FMM) [37]. We define Γ c as the confinement factor in the cover medium, i.e., the analyte, and Γ i , the confinement factor inside the waveguide. We studied, in particular the influence of the height of the ridge waveguide while assuming a constant width w = 1 μm. The thickness of the bias layer, which is formed during the nanoimprinting process, was set to 130 nm (see Figs. 1 and 2).
where E and H are, respectively, the electric and magnetic fields, and u z the unit vector along the z-axis. The electric field profiles for the quasi-TE fundamental mode are shown in Figs. 2(a) and (b), for the waveguide without and with the side coating, respectively. With the side coating, one can observe a longer evanescent tail of the electric field outside the waveguide, meaning larger interaction between light and analyte. The surface in contact with the analyte was the main topic of this study. The dimensions of the sensor were maintained unchanged to make comparison between structures with and without coating. Due to different modal properties the measurements were performed with different wavelengths. Figure 2(c) shows the variation of the optical field confinement inside the simple polymer waveguide for both quasi-TE and quasi-TM polarizations calculated assuming the refractive index of the cover medium to be 1.333 (water) at 633 nm wavelength. Note that for the coated and the non-coated samples the operating wavelengths are not the same. The reason is the increased size of the coated waveguide having the high refractive index layers on the sides. Increasing the wavelength, for coated waveguides, allows us not to modify the polymeric waveguide. Further explanations are given below.
By increasing the height of the ridge waveguide, the localization of the fundamental quasi-TE and quasi-TM modes inside the polymer ridge is increased, which leads to a decreased light intensity in the cover medium, see Fig. 2(c). To maintain a high optical confinement in the cover Electric field profiles of the fundamental quasi-TE mode of a polymer strip waveguide a) without a coating at λ = 633 nm and b) with a bi-layer of Al 2 O 3 /TiO 2 on the sides at λ = 975 nm. c) Field confinement factor (Γ c ) for quasi-TE and quasi-TM polarizations in the cover medium for a polymer ridge waveguide. d) confinement factors inside the waveguide and in the cover medium of the polymer ridge waveguide with a TiO 2 layer on the sides only, for the quasi-TE polarization. SEM images, as insets, show the cross sections of the polymer waveguides before and after coating. medium with low propagation losses, the height of the ridge waveguide was set to 290 nm and the effect of the TiO 2 layer was investigated for the quasi-TE polarization.
By adding thin TiO 2 layers only on the vertical sides of the ridge waveguide (for the fundamental quasi-TE mode case), one can obtain a rather high field in the cover medium, while maintaining a good confinement inside the waveguide. Adding the side layers makes the waveguide multimodal at λ = 633 nm. In order to maintain the single-mode performance of the waveguide, simulations and experiments were performed at 975 nm wavelength. Note that in this work, more than the sensitivity of the device, we were mainly studying the effect of the materials and geometry of the waveguides constituting the interferometer. Consequently, the polymer waveguide cross-section, and thus its exposed surface, should not change in size between the two structures. The thickness of the TiO 2 layer on the sides of the polymer rail was optimized and set to 90 nm, see Fig. 2 The studied sensor is a Young interferometer consisting of two polymer ridge waveguides of different lengths. Figure 1 shows a 3D schematic view of the sensor device and a cross-section of the waveguide. The two arms of the Young interferometer are separated by 50 μm from each other to avoid any cross-coupling. The waveguide arms of the interferometer are geometrically identical and the path length difference is 4 mm. This optical path length difference leads to a phase shift between the two arms. The sensing mechanism is based on the monitoring of the phase shift variation by observing the pattern created by the interference of the two output modes.
Nanoimprinting technique was employed to fabricate the polymeric structure. The fabrication process of the polymer waveguide is explained in details in reference [8]. After imprinting of the polymer waveguide, a bilayer of Al 2 O 3 /TiO 2 was coated on the waveguide by atomic layer deposition technique [19,21] using ALD Beneq TFS 200 equipment. The thickness of the Al 2 O 3 layer was set to 5 nm. This layer was grown using TMA and H 2 O as precursors at 120 • C. The TiO 2 layer was grown at 120 • C using TiCl 4 and H 2 O precursors. The purging time for these precursors was 750 ms and 1 s, respectively. In both cases, nitrogen was used as a purging and a carrier gas. The grown layers at this temperature are amorphous and smooth. The refractive indices of the grown Al 2 O 3 and TiO 2 layers were measured by ellipsometry at 975 nm wavelength as 1.63 and 2.33, respectively. The TiO 2 layer was further etched from the top of the waveguide by a fluorine based reactive ion etching using Plasmalab 80 equipment. Since our goal was to obtain a titanium dioxide layer only on the sides of the waveguides, the aluminum oxide layer was used here only as an etch-stop layer to protect the polymer waveguide from the TiO 2 -etching. It has no optical purpose, but affects the surface chemistry of the sensor. The fabricated Young interferometer was used for sensing the ethanol-water solutions of different concentrations. Measurements of samples without the ALD layers were performed at λ = 633 nm using free-space coupling from a stabilized HeNe laser (Thorlabs HRS015), whereas the samples with ALD layers were characterized at λ = 975 nm using end-fire coupling from a fiber pigtailed single mode laser diode (QFBGLD-980-5, Photonics LLC) to the sample. The input light was set, in both cases, to TE polarization and injected into the waveguides using polarization mainting tapered lens fibers. A flow cell was mounted on top of the chip and dionized water was flushed on the sample with a syringe pump (Nexus 3000, Chemyx Inc.) using the withdraw mode at the rate of 100 μL/min. Ethanol-water solutions with different concentrations of 1.67 · 10 −4 , 1.67 · 10 −3 , 1.67 · 10 −2 , 1.67 · 10 −1 wt.% were prepared and stored at room temperature. Both arms of our Young interferometer were exposed to the analytes. The interference of the two output modes, of different phases, lead to an interference pattern, which is monitored.
In order to measure the phase shift variations with the change in refractive index of the cover medium, the interference pattern was recorded with a CMOS camera at 3 s intervals. The measurement results were baseline corrected by measuring the drift from 0 until the time that the analyte was introduced to the sensing window. A complete detailed description of the experimental setup and data processing procedure is given in former studies [4,8,13].

Polymer ridge waveguide Young interferometer sensor
The first sensing measurements were performed with the polymer ridge waveguide Young interferometer. TE polarized light was coupled from a stabilized He-Ne laser (633 nm wavelength) to the input waveguide. The interference pattern was monitored by moving a 40× microscope objective 1.4 mm off from the focus of the waveguides output. Three different concentrations of the ethanol-water solution (0.00167, 0.0167, and 0.167 wt.%) were applied on the sensor. The measurements proceeded from the lowest to the highest concentration and were repeated for 3 times. In the measurements, 500 μL of analyte was applied in the sensing window and flushed away with deionized water.
Regarding the results shown in Fig. 3, the phase change increases after applying the ethanolwater solution to the measurement window and it takes about 2 min to reach a stable situation. After flushing water in the measurement window, the phase change returns to the base line after each concentration.

ALD-coated polymer ridge waveguide Young interferometer sensor
The responses for the different concentrations of the ethanol-water solution measured with the ALD-coated Young interferometer are shown in Fig. 4. The measurements started by applying the ethanol-water solution of concentration 0.00167 wt.%. In each measurement, 500 μL of analyte was injected in the measurement window and flushed with deionized water. The measurements were repeated for the solutions with a lower and a higher concentration, 0.000167, and 0.0167 wt.%, respectively. It is worth to note that in all measurements the trend of phase change is similar. However, the highest phase shift was observed in the first measurement performed with the ethanol-water solution of concentration 0.00167 wt.% [ Fig. 4(b)]. With all three concentrations, when the analyte is introduced to the measurement window, the phase shift drops slightly in the beginning. After about 90 s, the phase shift starts to increase but does not reach a stable state as long as the ethanol-water solution is covering the measurement window. After flushing the analyte with water, the phase change further increases for a few minutes and finally stabilizes. In none of the measurements, a return to the baseline was observed. The continuous increase of the phase shift under the presence of the analyte is problematic since the response of the sensor is not linear, unlike with the uncoated sensor.
The phenomenon above is attributed to attaching of the ethanol molecules to OH radicals or imperfections on the surface of the ALD-coated layers. These radicals might be originating either from the adsorbed water molecules on the TiO 2 surface during the ALD coating process ending with the reaction of H 2 O precursor with the surface, or due to imperfections that can be created during the etching process leading to adsorption sites for water or ethanol. Figure  5 illustrates the proposed mechanism of adsorption of ethanol molecules on the surface of the ALD layers. It has been shown that the first and second molecular layers formed at the surface because of the presence of OH groups may be stable while others are mobile and can be considered as a bulk liquid material [38]. Therefore, when the analyte is in contact with the sensor, a competition starts between water and ethanol molecules to link with the surface. Because of a stronger affinity of ethanol molecules with hydroxyl groups, a small drop in the phase change appears. The adsorption of ethanol molecules to the surface of the device causes the continuous increase of the phase change. Since the reaction sites are mobile, after flushing with water, the phase change starts to stabilize but the attached ethanol molecules do not desorb completely from the OH radicals on Al 2 O 3 /TiO 2 surface and the phase change does not return to the base level.  The strong adsorption of ethanol molecules to Al 2 O 3 and TiO 2 surfaces, mostly due to the presence of OH radicals and imperfections, explains the nonlinearity of our measurements: ethanol molecules can be flushed away by water only after a long time [see Fig. 4]. This nonlinearity of the results for higher concentrations of analyte is due to the saturation of the surface such that it cannot sense the increase of concentration of the ethanol in the analyte. The affinity of the polar molecules in the analyte for the sensor surface makes the device extremely sensitive to very small changes of the cover medium refractive index, which can be useful in qualitative bio-sensor applications. As it can be seen from Fig. 4(a) a refractive index change as small as 1 × 10 −7 RIU was measured with this device.
To validate our assumption about the influence of Al 2 O 3 /TiO 2 layers on the sensing measurement results, the surface properties of Al 2 O 3 and TiO 2 films, in particular the wettability, were studied through the measurement of the contact angle of the analyte on the ALD coated films.

Characterization of the ALD TiO 2 films
By comparing the coated and uncoated Young interferometers, it is evident that the surface plays a crucial role in the nonlinear response observed in the measurements. Two effects can explain the strong adsorption of ethanol molecules on the surface. The first one is due to the nature of the material itself (Ormocore, TiO 2 , or Al 2 O 3 ). The second is due to the post processing of the fabricated sample, which, in this case, is only the reactive ion etching (RIE) process used to remove the TiO 2 layer from the top of the waveguide structure. The RIE typically results in a surface roughness as well as activation of the surface due to the use of oxygen plasma during the process. Both phenomena increase the wettability of the sensor surface. The more the wettability increases, the more difficult it is to clean the samples after the measurement. Therefore, it leads to a non-return to the baseline after flushing with water, and to an almost never ending phase change when the sensor is in contact with the ethanol-water solution. Since the surface OH groups and imperfections, i.e., adsorbed oxygen due to the etching process, play an important role in the wettability of the materials by altering the surface energy, the contact angle of the liquid (water or analyte) is a relevant measure of the adsorption of the molecules on Al 2 O 3 and TiO 2 surfaces.
The contact angle of water and an ethanol-water solution (30 wt.%) was measured on different flat surfaces, representing the different possible surfaces of the real Young interferometer. The first set was composed of three glass substrates (BK7) coated with Ormocore, TiO 2 , and Al 2 O 3 . The second set was composed of the same layers etched with the same RIE process used during the fabrication of the real Young interferometer, i.e., a dry etching for 3 min in SF 6 /O 2 plasma. Figure 6  The Young interferometer studied in this paper is composed of Ormocore imprinted waveguides coated with a conformal thin layer of Al 2 O 3 and then on the side a thicker layer of TiO 2 . The polymer is not etched and is protected from the plasma by the Al 2 O 3 layer, which is used as an etch stop layer for TiO 2 . Al 2 O 3 is not etched by the RIE process but suffers from the surface activation by the oxygen plasma. According to our measurements by ellipsometry, the refractive index change of Al 2 O 3 after etching is negligible. This is mostly due to the change in surface bonding of the Al 2 O 3 layer, which however confirms the activation of the surface. The TiO 2 layer is etched and is also activated, although the etching process does not have a significant effect on the roughness of the TiO 2 layers (sides of the ridge). The surface energy and adsorption mechanism on an uncoated Young interferometer corresponds to Fig. 6(a) for water and Fig. 6(g) for the ethanol/water solution, while the surface properties of the coated Young interferometer can be explained by Figs. 6(e) and (f) for water, and Figs. 6(k) and (l) for the solution.
Although the wettability is higher for the solution than for water in the case of the uncoated sample, it is clear that it remains low in comparison to the coated sample, on which the contact angle is close to zero in both cases. Moreover, the wettability increases when water is replaced by the analyte and when the surface has been exposed to the plasma. This effect is evident in the cases of the ALD layers.
These results prove: 1) a higher affinity of the ethanol groups with the ALD layers than with the Ormocore, 2) a higher spreading of the solution around the waveguide, and 3) a more difficult cleaning of the surface from ethanol in case of a coated interferometer than an uncoated one.
Several ways can be envisioned to reduce or cancel the attraction of ethanol groups (or more generally polar molecules) to the TiO 2 layer in particular. For example, a change of the ALD precursors can be investigated. We have already remarked that an oxidant without hydrogen leads to a different behavior of the surface, although the trend remains similar because of the roughness induced by the etching. Another potential solution consists of the deposition of a very thin protective passivation layer on top of the whole device.

Conclusion
We have investigated the effect of the presence of ALD Al 2 O 3 /TiO 2 overlayers on the performance of a polymer waveguide based interferometric sensor. The characteristics of the device were studied in the case of the aqueous sensing of ethanol-water solutions of different concentrations. A refractive index change of 1 × 10 −7 RIU was detected with our device. The sensing results did not follow a linear monotonic trend with respect to the analyte concentration. Despite the nonlinearity of the results, the phase change rate is higher than with the uncoated sensor (close to 3 rad). This can enable a fast qualitative analysis of very small ethanol concentrations. We have also remarked that until the analyte is not flushed, the phase shift increases continuously, which means that even smaller concentrations can be measurable. Comparing the response of our device with a polymer ridge waveguide interferometer without the ALD layers confirms that the nonlinearity of the results is due to the presence of the ALD layers. To further investigate the observed characteristics, we measured the contact angles of water and an ethanol-water solution on the surface of the polymer and the ALD layers. These measurements confirmed the formation of hydroxyl groups due to the use of H 2 O as an oxidant precursor in the ALD process and activation and imperfections of the layers originating from the etching process as well the adsorption of ethanol molecules on the surface of the ALD layers. In this study, the sensing measurements were performed for an ethanol-water solution, but a similar behavior is expected for aqueous sensing of analytes containing polar molecules. The knowledge about the surface properties of the Al 2 O 3 /TiO 2 layers can facilitate the functionalization of these surface as well as the application of these materials for sensing of particular analytes. Finally, the deposition technique can be easily performed on more complex devices already in use in integrated optics such as Mach-Zehnder interferometers, for instance.