1. Introduction

Fabrication and development of photonic integrated circuits (PICs) is of high interest since they can be used for integration of passive and active photonic devices on a single chip [1,2]. Among different materials which can be used for fabrication of PICs, single-crystalline lithium niobate (LN) is favorable for numerous optical and photonic applications due to its excellent electro-optical and nonlinear optical properties, wide transparency window, and good temperature stability [3–6]. However, bulk LN cannot be used for fabrication of very compact PICs since the required waveguides at the surface are typically realized by using diffusion techniques that suffer from weak optical confinement and high bending losses [7]. A solution for this challenge is the use of a thin film of LN embedded in materials with lower refractive index. LN-on-insulator (LNOI) substrates similar to silicon on insulator (SOI) are nowadays commercially available as single-crystalline X- or Z-cut LN thin films on a few-micrometer-thick insulating SiO₂ layer.

Based on the high second-order nonlinear susceptibility of LN, one of the interesting components for PICs are nonlinear optical elements for frequency conversion, such as second-harmonic generation (SHG) [8]. For efficient frequency conversion, the phase velocities of the interacting waves, i.e. the fundamental- and second-harmonic waves, should be matched, which can be achieved by quasi-phase matching (QPM) [9]. The most promising way to realize QPM in LN is the periodic inversion of the polarization direction of the LN crystal along the crystal z-axis, which can be achieved by different methods [10–14]. In the last decades several methods for domain engineering of LN were studied [15], such as proton exchange [16], electron beam irradiation [17,18], optical poling [19] and electric-field poling [20,21]. All of these processes target different applications and strongly vary with respect to process complexity and time, domain size, depth and uniformity.

In this work, we concentrate on the most common poling technique, which is electric-field poling (EFP). The EFP process is based on inversion of the LN crystal structure along its crystalline z-direction by applying an electric field that is stronger than the coercive field strength of LN. Commonly, EFP is used for Z-cut LN with the structured electrodes on one side of the bulk LN substrate and a flat homogeneous electrode on the opposite side. Applying an appropriate electric field yields a domain pattern that extends through the entire thickness of the LN substrate [21]. Whereas this method is well established and understood, it is not applicable for LNOI because its intermediate insulating SiO₂ layer prevents electrical current flow and, hence, undisturbed domain growth. Nonetheless, EFP can be applied at the surface of LNOI. In this case, the electrode structures are deposited on the surface of an X- or Y-cut LN with the electric field being directed along the in-plane z-axis. A voltage is then applied to produce the strong electric field (∼22 V/µm) along the surface [10–12].

Compared to other surface poling methods of LN like tip poling [22], the main advantages of surface EFP are the compatibility to standard lithographic processes enabling wafer scale production. This allows for high flexibility because different configurations of electrodes can be easily implemented. Finally, the process produces domain patterns that reach several micrometers beneath the surface, making it compatible with waveguide fabrication processes including diffusion waveguides as well as etched ridge waveguides in LNOI [4,6,23–26]. However, the domain switching necessary to achieve a well-defined poling pattern requires optimization of electrode configuration and geometry as well as optimization of the poling voltage pulse [27].

Although surface poling has been already demonstrated, a detailed analysis of optimal conditions to achieve regular poling domains is up to now missing. Filling this gap is important, as surface poling now becomes a crucial technique for realizing high-quality PICs in LNOI, where, although high-quality poled structures have been already demonstrated [12,23,28,29], an in-depth investigation of different poling conditions has not been published. In this work, we aim to optimize process conditions for surface EFP in order to produce a periodically poled pattern at the surface of congruent bulk Y-cut LN. We carry out our experiment on bulk LN due to its easier availability. We expect that the results of these studies can later be transferred directly to the surface poling of LNOI films. Since this work is a technology study, a specific sample design has not been realized yet. In the first step, we theoretically compare the electric fields of rectangular and tapered finger electrodes. We then proceed to present the experimental results for some of the studied electrode configurations. Afterwards, we study the influence of different characteristics of the electric poling pulse such as pulse shape, maximum poling voltage, and the duration of poling on the uniformity of the fabricated structures.

2. Simulation of the electrostatic fields

The conventional configuration for surface poling of LN consists of rectangular finger electrodes [11]. However, the requirements for surface poling are different with respect to conventional poling, as the direction of domain growth is oriented along the direction of the electrodes and not normal to it. Particularly the large fields induced around the corners of rectangular finger electrodes, which extend in the area between the electrodes, could prevent a homogeneous directional domain growth. It is shown that changing the geometry of the electrodes can improve the homogeneity of the poled areas [29], however, an optimized configuration is not yet found. For this reason, we first theoretically analyze the field strength distribution at the tip and present an approach to optimize it by adjusting the metallic electrode tip shape.

The electrostatic field distributions of differently shaped electrodes are numerically simulated using COMSOL Multiphysics. The material parameters for LN are chosen according to Teague et al. [30], using the clamped permittivity. It is assumed that the sample, i.e. LN and electrodes, is surrounded by an insulation oil with a permittivity of ɛ_oil = 2.7 like in the later experiment, where this is necessary to avoid electric breakdown of the surrounding air.

In the simulations and experiments presented here, the same periodicity of Λ = 8.3 µm is used, which approximately corresponds to the periodicity used for QPM type-II frequency down conversion of near-infrared light in waveguides [31]. The schematic of the electrodes’ structure on LN is shown in Fig. 1(a). Further, in the simulations we assume a distance between the electrodes of 50 µm, an electrode thickness of 0.2 µm, and a voltage between the electrodes of 1150 V. The voltage corresponds to an electric field strength of E_mean = 23.0 V/µm. This simplified simulation approach only gives an indication of the actual poling behavior, as there are several limitations: only the static electric field distribution caused by the electrode potential before the start of the inversion process is considered, while the actual domain inversion is a highly dynamic process. For instance, the potential influence of surface charges and internal charge displacements remains unconsidered. Additionally, the formation of the initial nucleation cells depends not only on the electric field strength, but also on the presence of material imperfections not considered in the simulation.

 

Fig. 1. (a) shows the configuration of the electrodes on the surface of a Y-cut LN crystal. The electrodes are along the z-direction of the crystal. (b) and (c) show the electric field strength E_z (V/µm) in the direction of the polarization axis (crystal z-direction) along the surface of a LN crystal for a rectangular and a tapered finger electrode respectively. For symmetry reasons only one half of the unit cell of an 8.3 µm periodic structure is plotted with a partial electrode visible in the lower left corner. Shown is the average electric field strength within a surface layer extending 0.7 µm down into the LN. (d) and (e) show the electric field strengths of (b) and (c) sorted in descending order along the z-direction. (f) is the electric field strength of the cross sections of subfigures (d) and (e) along l_z = 2 µm for rectangular and tapered finger electrodes (thick solid and dashed lines respectively). The dashed lines with squares and rhombus show the first derivative of the fields along the x-direction respectively.

Download Full Size | PDF

The crucial quantity for inverting the ferroelectric polarization state is the electric field component E_z along the polarization axis z. For obtaining a uniform pattern of periodically alternated ferroelectric orientation, a pronounced contrast of the electric field strength between sections to be inverted and non-inverted sections is required. Consequently, for obtaining a periodically poled pattern with a duty cycle of 50%, a strong drop of the field strength at half of the distance between the electrodes is desired. This consideration assumes that surface charges and internal charge displacements are negligible.

Among several simulated electrode structures, the most promising design was identified to be a tapered finger electrode, since the field strengths at the vertices of the tapered finger electrodes are smaller compared to rectangular finger electrodes. A range of different tapered lengths was simulated, the most promising electrode is found to have a tapered length around 2 µm and an electrode width around 3.5 µm. The proper width for rectangular finger electrode to have a duty cycle of 50% is found to be around 1.6 µm.

Figures 1(b) and 1(c) show the electric field strengths along the z-direction averaged within a 0.7 µm thick layer of the LN surface for the half of a single unit cell of the periodic rectangular and tapered finger electrode structures. Advantageous field distributions for periodic poling have two qualitative properties. First, for the x-coordinates where poling should be achieved, the field strength has to be above the coercive field strength for large domains of the z-coordinate. However, the absolute z-position where such a region of high field strength is achieved does not have a great impact. Furthermore, for x-coordinates where poling is not needed, the field strength should be low for all z-coordinates, with a sharp transition between both domains.

To assess which electrode design fulfills these conditions better, we use a different representation of the simulated field strengths, depicted in Figs. 1(d) and 1(e). There, for each x-position we take the vector of the calculated field strengths along the z-direction and sort it according to the field strength. For each vector, the largest value of E_z is then assigned to the smallest value of a new spatial coordinate, the effective length l_z. Smaller values follow for larger l_z, which is not equal to the z-coordinate. This representation allows to compare different electrode geometries, as it removes the z-dependence of the region of high field strength in the tapered geometry. Now we can estimate the impact of the poling field on the domain switching probability at the different x-positions according to the criteria formulated above. A high switching probability exists at x-positions where large field values are present for large values of the effective length. After the described sorting, the ideal field distribution would show high field strengths at high effective length l_z in the area to be poled (x < Λ/4) and negligible field strength in the areas that should remain unpoled (x > Λ/4). Especially, there should be a sudden drop of the field strength at x = Λ/4 = 2.075 µm [black dash-dotted lines in Figs. 1(b)–1(e)] with contour lines of equal field strengths being parallel to the z-direction.

Comparing subfigures 1(d) with 1(e) shows an improved field profile in the case of the tapered finger electrode, indicated by the straighter and more vertical contour lines around the intended poling boundary at x = Λ/4.

The advantage of the tapered electrode structure in Fig. 1(c) over the rectangular structure in Fig. 1(b) becomes more evident by comparing the cross-sections of the subfigures 1(d) with 1(e) along l_z = 2 µm (white dashed lines), which is illustrated in Fig. 1(f). For achieving a well-defined domain wall at x = Λ/4, indicated by the dash-dotted vertical line in Fig. 1(f), the electric field strength in this region (lines, left ordinate) should be close to the coercive field and the gradient of the electric field strength (symbols, right ordinate) should be as high as possible between these sections. Subfigure 1(f) compares the gradient of the field strength for both electrode structures, where the value for the tapered electrode is significantly larger around x = Λ/4. The exact x-position is slightly shifted above Λ/4, which means that a duty cycle of exactly 0.5 can be obtained only by slightly reducing the width of the tapered electrode below 3.5 µm. For the case of the tapered electrode, a second strong gradient occurred around x = 1 µm, but at this position the electric field strength exceeds the coercive field and the poling will take place anyway.

3. Fabrication

The microstructured electrode patterns are created by electron beam lithography and a subsequent lift-off process. The metal layer composing the electrodes consists of 170 nm Cr, which is important for increasing the number of nucleation sites [32], covered by 30 nm protecting Au. The dimensions of the finger electrodes are chosen based on the results explained in the previous part. For periodicity Λ = 8.3 µm, the tapered finger electrodes are 2 µm long and 3.5 µm wide, and the width of the flat finger electrodes is 1.6 µm.

Electric field poling pulses are created by a computer controlled high voltage amplifier (2HVA24-BP1-F-SHV-5KV, Ultravolts). Additional measurement circuits for high-precision voltage and current monitoring are used. For the crucial measurement of the comparatively low poling currents, an ultra-low bias current operational amplifier (OPA129, Texas Instruments) is used. This allows the measurement of the currents in the range of a few nanoamperes.

As the electrical breakdown strength of air is much lower (around 3 V/µm) than the coercive field strength (∼22 V/µm) of LN [33], during the poling process the sample is placed in silicone insulation oil with a dielectric breakdown strength exceeding 40 V/µm.

In the poling pulses studied here, the voltage is changing according to freely adjustable sequences of linear functions. The best poling results are obtained by the application of the poling voltage pattern depicted in Fig. 2. It consists of a short rectangular poling pulse with an electric field strength E > E_coercive and length Δt₄ with additional trapezoidal initial (Δt₁ = 30 ms, Δt₂ = 10 ms, Δt₃ = 0.5 ms,) and stabilization (Δt₅ = 0.5 ms, Δt₆ = 10 ms, Δt₇ = 60 ms,) sections of a lower voltage below the coercive field strength. This poling pulse scheme has been established for bulk Z-cut LN poling and shows significantly better results than simple trapezoidal pulses. Also tested were the effects of nucleation pulses, which are short voltage spikes slightly below the coercive field strength of LN as depicted in Fig. 2, applied prior to the actual poling process. These pulses are meant to increase the amount of nucleation cells for domain growth in the vicinity of the electrodes (see the discussion below).

 

Fig. 2. Schematic diagram of the nucleation pulses and the poling pulse (blue shaded area) with a trapezoidal stabilization pulse.

Download Full Size | PDF

4. Results and discussion

Electric field poling is a very dynamic process and many different parameters during the poling process could affect the performance and homogeneity of the final result. In this study we focus on the effect of electrode tip shape, poling pulse parameters, and nucleation pulses, which we think are the main contributing factors on the quality of the poling. We investigate the effect of changing these parameters to achieve the best poling result.

Before we discuss our experimental results, we briefly introduce the methods we used to visualize and analyze the generated poling patterns. First, we used several non-destructive methods for the visualization of inverted ferroelectric domains in LN. For verifying whether successful domain inversion has occurred, a first picture of the created domain patterns can be obtained by polarization contrast microscopy (PCM). However, PCM is not able to clearly show the domain walls and it is not suitable for a detailed domain pattern characterization.

As another method for non-destructive domain inspection, we use second-harmonic imaging microscopy (SHIM) [34–36]. In this method, we irradiate the sample with a focused laser beam that generates a second-harmonic signal at the focus position, which can be spectrally filtered and used for imaging. We carry out this measurement point by point by scanning the sample with respect to the laser focus to generate a second-harmonic intensity map for the sample. This enables domain visualization due to an increased second-harmonic signal in the vicinity of the ferroelectric domain boundaries [34].

The methods described up to now do not allow for reliable quantitative analysis of the poled structures, as their spatial resolution is too low and the measured signals cannot be directly related to the domain boundaries at the sample surface. Images with sufficiently high spatial resolution can be achieved by selective etching of the poled surfaces and subsequent inspection with scanning-electron microscopy (SEM) or optical microscopy. The selective etching relies on an etch-rate contrast of oppositely poled LN domains in hydrofluoric acid (HF). In this way, the ferroelectric domain patterns on Z- and Y-cut LN surfaces can be revealed [37], although the samples are altered in this process.

Therefore, samples were etched only upon finalizing all poling experiments with different poling pulse shape and electrode shape configurations. For intermediate characterization to determine specific experimental parameters based on results of first experiments, the non-destructive techniques for domain visualization were used. For analysis, the etched domain patterns are typically imaged via optical microscopy along a length of 20 periods and the edges created by selective etching are detected. Generally, the pattern quality is better in the region close to the positive electrodes. Therefore, only a partial area of the domain pattern is considered, which extends from 2 to 10 µm from the positive electrode and forms a narrow stripe that perpendicularly crosses the domain walls. Its width is just large enough to allow the patterning of a ridge or diffusion waveguide subsequently.

In order to capture the quality of the realized poling patterns in a single number, we calculate the structure factor of the realized grating relevant for first-order quasi-phase matching. This corresponds to the amplitude of the first term of the Fourier series describing the spatial profile of the nonlinear susceptibility along the propagation direction. This estimation assumes that the domain walls visible at the surface extend perpendicular into the structure. Since we are interested only in a thin layer at the surface of the structure, this assumption is justified. We compare the calculated structure factors to the ideal ones obtained for a duty cycle of 50% by considering their ratio and call this ratio the grating quality. Since it considers numerous poled domains of a realized structure, this value averages over many realizations of the same poling process.

In the following, by using the calculated grating quality, we present the influence of the different parameters on the poling result. At the end of this part, the depth of the poled area into the LN is investigated.

4.1 Influence of the nucleation pulses

First, we investigate the influence of the nucleation pulses applied before the actual poling pulse on the overall domain pattern uniformity. Figures 3(a) and 3(b) show optical microscope images obtained with PCM of samples which were poled without and with applying the nucleation pulses, respectively, while the other parameters remained the same. Here, the electrodes appear yellow whereas the poled domains can be seen as vertical stripes on the grey LNOI substrate. Figure 3(b) obtained for poling with nucleation pulses shows a better result because the domains are homogeneously appearing at all electrode tips and are not merged, which demonstrates the importance of applying nucleation pulses. We found field strengths of 17-18 V/µm, a pulse duration of 1 ms and an amount of 500 pulses to be suitable to achieve the best results compared by PCM. Consequently, for all further experiments nucleation pulses with these parameters are applied before the actual poling pulses.

 

Fig. 3. Optical microscope image of electrically poled LN with the electrode period of 8.3 µm and the distance between positive and negative electrodes of 50 µm. Fabricated sample for poling (a) without applying the nucleation pulses and (b) with prior application of nucleation pulses.

Download Full Size | PDF

4.2 Influence of the electrode tip shape

In order to verify experimentally whether tapered finger electrodes actually lead to a better domain homogeneity and periodicity as suggested by our initial simulations, we compared Y-cut samples with rectangular and with tapered finger electrodes when poled with identical parameters. Furthermore, for the negative terminal we use either a sequence of finger electrodes similar to the one at the positive electrode or a single flat electrode. Nucleation pulses are applied before poling. The maximum voltage of the poling pulse is set to 1175 V, corresponding to a field strength of 23.5 V/µm, and the duration of the main poling pulse is Δt₄ = 1 ms. Figure 4 shows representative results of measured domain boundaries along the surface from the SEM images taken after poling and subsequent selective etching, where subfigures 4(a),4(c)/4(b),4(d) show rectangular/tapered finger electrodes combined with flat / finger [(a,b)/(c,d)] electrodes, respectively.

 

Fig. 4. The influence of the electrode shape on the inverted domains is studied for otherwise identical poling conditions, where the blue shaded areas are inverted domains deduced from selective etching. (a) and (b) show the result for rectangular finger electrodes on one side, while the other sides are flat and finger electrodes respectively. In (c) and (d) the corresponding results for tapered finger electrodes rather than rectangular finger electrodes are shown. (e) presents the calculated grating quality for the samples (a) to (d). In subfigures (a) to (d), the positive and negative electrodes are at the top and bottom, respectively.

Download Full Size | PDF

Comparing the plots in Fig. 4, it can be concluded that all structures with the exception of the one with rectangular finger electrodes on both sides have similar grating qualities, plotted in Fig. 4(e). The significantly smaller value for the latter case is due to the occurrence of domain merging seen in Fig. 4(c), which is almost completely suppressed by the usage of tapered finger electrodes in Fig. 4(d). Also, both samples with a flat negative electrode show a good pattern quality. However, they show an increased amount of needle-like inverted domains in areas which should not be inverted. As these randomly appearing domains hint towards a more unstable poling process harder to control and optimize, for all following experiments the structure with tapered finger electrodes for the positive and negative electrodes is used.

4.3 Influence of the poling pulse

Next, we investigate the dependence of the poling result on the parameters of the applied poling pulse. We use the poling scheme sketched in Fig. 2, consisting of the application of nucleation pulses followed by a short poling pulse with a poling duration of a few milliseconds sitting on a fixed background. The maximum poling voltage and the poling duration are identified to be the most important parameters and, thus, shall be investigated in detail.

We use Y-cut samples with tapered finger electrodes at both sides with a distance of 50 µm and a period of 8.3 µm for all tests, which corresponds to the geometry used in the simulations described above. For all samples, we used 500 nucleation pulses realizing field strengths of 17 V/µm and 1 ms duration followed by the poling voltage pattern shaped as in Fig. 2.

4.3.1 Influence of the maximum poling voltage

Poling pulses achieving a maximum field strength between 22.0 and 24.0 V/µm and otherwise identical parameters are applied. Figure 5 shows the measured poling current for different field strengths.

 

Fig. 5. Measured current at the moment when poling occurs (Δt₄) for different maximum poling voltages. In this example, the electrodes are tapered at both sides with periodicity of 8.3 µm and the distance between the electrodes of 50 µm. The dashed lines show the time period Δt₄ where the voltage above the coercive field strength is applied. The inset shows the current and voltage signal during a complete poling process, the time span depicted in the main image is marked by orange shading.

Download Full Size | PDF

Comparing the poling current of different samples shows a substantial increase for field strengths above 23.0 V/µm. Here, continuous channels between the two electrodes are established, which allow for charge transport. As this happens due to the poled domains, the higher currents are a signature for a more complete poling of domain area between the electrodes. This is reinforced by analyzing the domain structure obtained by selective etching, which is plotted in Figs. 6(a)–6(e) together with the result of SHIM measurements for different poling field strengths. For field strengths below 23.0 V/µm, the domains, which start to grow from the upper electrode, do not reach the lower electrode. Thus, no channels for the flow of charges are created. Increasing the field strength to 24.0 V/µm results in domains that reach the opposite electrode, but also start to broaden and merge. This also indicates that the critical field strength to obtain a uniform periodic structure is between U_p = 23.0 V/µm and 24 V/µm.

 

Fig. 6. Influence of maximum poling voltage on the domain inversion for samples with periodicity of L = 8.3 µm. (a) to (e) show the inverted domains for maximum poling voltage Up = 22.0, 22.5, 23.0, 23.5, and 24.0 V/µm, respectively. The intensity is measured by SHIM and the black lines are from the SEM images after etching. (f) shows the calculated grating quality for samples (a) to (e).

Download Full Size | PDF

In order to further compare these domain structures, the grating quality is calculated in the area between −2 and −10 µm. The comparison of the resulting grating qualities in Fig. 6(f) shows that the best domain structure is obtained by the pulse with a poling field strength of 23.5 V/µm, well above the coercive field strength. Altogether, the results indicate that the poling pattern quality sensitively depends on the poling voltage.

4.3.2. Influence of the poling duration

In this section, we describe the results of experiments where the duration Δt₄ of the main poling pulse is varied from 0.5 ms to 2.5 ms, while the mean field strength is fixed to 23.0 V/µm, within the optimum parameter range found earlier. Apart from this, the same poling scheme as in the experiment before is used. The poling domain boundaries obtained by etching as well as overlaid SHIM images for the samples poled with 0.5, 1.0, 1.5, 2.0, 2.5 ms duration are shown in Figs. 7(a)–7(e) respectively.

 

Fig. 7. Influence of poling duration on the domain inversion. The periodicity of the electrodes L is 8.3 µm. The measured signals from SHIM (intensity) and the inverted regions after etching and following SEM (lines) are shown in (a) to (e) for poling duration of 0.5, 1.0, 1.5, 2.0, and 2.5 ms respectively. The maximum poling voltage U_p is kept at 23.0 V/µm. (f) shows the calculated grating qualities for samples (a) to (e).

Download Full Size | PDF

The comparison of the corresponding grating qualities shown in Fig. 7(f) indicates that the best domain pattern quality is obtained from a pulse with a duration of 2.5 ms [see Fig. 7(e) for poling pattern]. The achieved poling pattern from this experiment has a corresponding quality of more than 90%, which is close to the ideal case.

4.4 Depth of the inverted domains

After establishing parameters for an optimal poling process, we are finally measuring the extension of the poled domains into the substrate in order to establish whether this is large enough to pole through typical film thicknesses in LNOI. In the previous experiments, we used Y-cut substrate to be able to investigate the quality of the poling in the whole area between the electrodes. To characterize the poling depth as described below, now we use X-cut lithium niobate to enable etching around the poled volume.

In order to measure the depth of inverted domains in a surface-poled substrate, a trench is prepared by focused ion beam milling to expose the crystal z-face. This is schematically shown in Fig. 8(a). In particular, the trench is milled at a distance of around 10 µm from the positive electrodes and perpendicular to the z-axis in the poled X-cut substrate, which is shown in the SEM image in Fig. 8(b). The domains are then made visible by etching with HF, where we use that the surfaces of LN normal to the z-direction show an etch rate contrast for differently oriented domain polarities. Figure 8(c) shows the shaped ferroelectric domains after etching. The shape of the inverted domains, inversely shaped at the opposite (towards + Z/-Z) side, is clearly seen. The domains extend 1 µm below the surface as indicated by the white dashed lines. This is somewhat less than expected from the simulations, which close to the electrode show that a field above the coercive field strength is present down to 2 µm below the surface. However, our previous experiments showed that larger field strengths are needed to achieve poled domains extending towards the other electrode, hence a decreased thickness of the poled layer with increased distance from the electrode is consistent with our findings. We additionally carried out SHIM measurements of the sample with focus positions at different depths from the sample surface to reconstruct a cross-sectional view of the domain profile. The measurements confirm that at the same distance of 10 µm from the positive electrodes the depth of the inverted domain is approximately 1 µm.

 

Fig. 8. X-cut LN does not show an etch rate contrast with HF etching for different domain orientations. To visualize the domains a trench is therefore cut in the X-cut surface to expose the LN Z-face at the sidewalls. The Z-face shows significant etch rate contrast which can be used to measure the depth of the inverted domains. The ferroelectric pattern was created by tapered finger electrodes with a period of Λ = 16.6 µm and a distance of 50 µm. Before the HF wet etching, a 2 µm wide trench was created by focused ion beam milling. (a) Schematics of the sample and the trench. (b) SEM image of selectively etched X-cut LN sample with the trench. (c) SEM image of a cross-section of the trench that was prepared after etching, parallel to and through one of the inverted domains. The inverted domain protrudes from the left into the trench and appears as inversely shaped groove at the opposite side. The etched pattern indicates a homogenously inverted zone with a thickness of around 1 µm which is in good agreement with the electric field calculations shown in (d).

Download Full Size | PDF

5. Conclusion and outlook

We investigated the capability of surface electric field poling (EFP) for creating ferroelectric domain patterns for quasi-phase matching (QPM) nonlinear frequency conversion processes in Y-cut bulk lithium niobate (LN). Although many works on EFP of LN address the periodic poling at the surface of LN and magnesium-oxide- (MgO-) doped LN for different configurations [23,26,27,29,38,39], our result is a detailed systematic study of surface poling of Y-cut LN. Samples were fabricated by electron beam lithography and subsequent lift-off of the metal layer forming the electrodes. We investigated the influence of different parameters during the poling process on achieving a uniform periodic domain pattern.

We introduced tapered finger electrodes and found that they are more appropriate than the rectangular finger electrodes for producing electric field distributions for achieving highly uniform 1-dimensional QPM structures. Our experimental results for different electrode tip configurations were consistent with the theoretical predictions. We also investigated other important variables in domain engineering, namely poling voltage and duration. We have seen that just a slight change of these parameters significantly influences the performance of the manufactured structures. As best poling pulse parameters for our electrode distance and geometry we found a main poling pulse situated on a high-voltage background just below the coercive field strength, where the main pulse has a duration of 2.5 ms and ensures a field strength of 23.5 V/µm within the material to be poled. Furthermore, this poling pulse should be preceded by nucleation pulses.

Our setup and experiments can be applied with slight changes for fabrication of nonlinear integrated optics elements in LN on insulator (LNOI), which is an important substrate material for photonic integrated circuits (PIC). Strong light confinement in small modal volumes of the integrated devices yields a large efficiency of nonlinear optical processes. Especially the periodically poled LNOI devices significantly contribute to nonlinear optics applications [10,12,23]. Therefore, we will extend our systematic study of surface poling also to thin film substrates.