1. Introduction

Lithuium niobate (LN) is a versatile material used widely in optoelectronic, opto-acoustic, and piezoelectric applications [1–2]. The stripe-geometric lithium niobate waveguide is highly needed because it constitutes the basic structure for many optoelectronics devices, such as optical switch, modulator, coupler, multiplexer, etc. Stripe-geometric lithium niobate waveguides can be fabricated by etching, in-diffusion, or ion exchange. The ridge waveguide and Ti-diffused lithium niobate are the typical examples. As an alternative method, ion implantation is usually employed to form planar waveguides in lithium niobate [3–4]. Channel waveguides by He implantation was demonstrated in KNbO₃ and Ga₄GdO(BO₃)₃[5–7]. In both cases the implantation, sometimes with the aid of photolithography or the masking method, were carried out in two steps: a planar structure by relatively high energy ion implantation was formed in advance, followed by a low multienergy implantation to form the channel structure. Channel waveguide in KNbO₃ was also demonstrated by the single homogeneous He⁺ ion irradiation, in which a structured photoresist with wedged edges serves as an implantation mask [8]. In this article we report first the channel lithium niobate waveguide at 1.54 μm formed by one-step multienergy O²⁺ implantation. Because the refractive index barrier is caused by multienergy implantation, the waveguide with less leakage loss is obtained. For comparison, both planar and channel waveguides are formed simultaneously by employing the same implant parameters. The properties of waveguides are investigated using prism and end-face coupling methods. In the planar waveguide, only single guiding mode is demonstrated in the perpendicular direction to the surface of the sample. Two transverse modes and one transverse mode are observed in the n_e and n_o channel waveguides, respectively. The losses of the channel waveguides are measured and analyzed. This is a continuative work of our former research [9]; the experimental results provide verification to our analysis in the former report.

2. Experiment and analysis

Three units of energy O²⁺ at 3.0 MeV; 3.6 MeV; and 4.5 MeV, respectively, are implanted into X-cut and Z-cut lithium niobate with moderate beam doses. In order to get a flat damage profile at the end of the ion track, the ratio of beam dose for each energy is chosen at 0.45:0.55:1.2. The summed dose for each sample is from 4.4×10¹⁴ ions/cm² to 1.3×10¹⁵ ions/cm², which corresponds to the O²⁺ peak concentration from 1 to 3×10²⁰ ions/cm³. Implantation is carried out at room temperature, and the samples are titled 7⁰ off the normal direction of the sample surface in to avoid channelling effect. The channel structure is formed with the help of the lithography technique. The 15 μm channels [Au: Does this mean you have fifteen separate micrometer channels?], which are arranged at intervals of 50 μm, are exposed to the implantation. The remaining area is covered by photoresist. The thickness of the photoresist is estimated to be about 2 μm. According to the TRIM simulation, the LN damage profile caused by implantation is given in Fig. 1. It can be seen that the intensive-damage region is concentrated at the end of the ions track, at about 1.7μm below the sample. For the photoresist-covered area, the LN intensive damage lies just beneath the surface of LN because of the stopping effect of the photoresist on the impinging O²⁺ ion. At the joint between the exposed and covered areas, the damaged region may slightly interpenetrate because of the transverse straggling of implanted O²⁺ ions. According to our previous experiments and analysis [9–10], the extraordinary refractive index (n_e) of LN along the ion track is raised when the implanted O²⁺ ion dose is relatively low, which can cause an n_e index-raised guiding region. Meanwhile, a slight reduction of ordinary index (n_o) is induced in this region. At the intensive-damage layer, an index-reduced barrier is formed because of the reduction of material density and the partly amorphous nature of this layer. In consequence, two kinds of waveguides can be formed in such implanted LiNbO₃ in terms of refractive index: for n_e, a waveguide may be formed with the structure of an extraordinary index-raised guiding region sandwiched between the surface (air) and the low index barrier [9]; for n_o, a typical barrier-confined waveguide is formed. Such an implanted n_e channel waveguide has better optical confinement and a more-compact structure, because the index difference between the index-raised guiding region and the index-reduced barrier may be larger than or comparable to that of the in-diffusion and exchange waveguides.

 

Fig. 1. Simulated damage profile in LiNbO₃ caused by O²⁺ implantation according to TRIM code.

Download Full Size | PDF

By adjusting the thickness of the photoresist (corresponding to the location of index-reduced barrier), we formed the implanted n_e channel that has the configuration of an extraordinary index-raised channel “encapsulated” by the low-index barrier and air. The schematic of the implanted-channel waveguide formation is given in Fig. 2(a). The dark areas represent the damaged region. The dashed curve describes the profile of damage in a sectional view. The microscopic cross section of the channel waveguide formed by O²⁺ implantation is shown in Fig. 2(b).

 

Fig. 2. (a) Schematic illustration of the implanted-channel waveguide formation. (b) Microscopic cross section of the channel waveguide formed by O²⁺ implantation.

Download Full Size | PDF

After implantation all the samples are annealed at 250°C for 20 min in air ambient. Waveguide measurements are carried out with prism coupling and end-face coupling. For the planar waveguide structure, a single mode at 1540 nm is demonstrated by dark-mode measurement; the mode profile is shown in Fig. 3. The location of a possible index barrier is situated at about 1.7 μm beneath the sample surface (see Fig. 1). This value is smaller than the typical thickness of the diffusion single-mode LN waveguide (usually 3–5 μm), which implies the possibility that a more-compacted single-mode waveguide can be formed by implantation. The effect of compressed waveguide size can be attributed to the big index difference between the guiding region and the index barrier.

 

Fig. 3. Mode profile of single-mode planar waveguide at 1540 nm by prism-coupling measurement.

Download Full Size | PDF

The measurement for the channel waveguide is carried out by use of the end-face coupling method. For mode distribution measurement, a BT&D EFA 1000 erbium fiber amplifier is used. A tunable bandpass filter is connected to the EFA to select a wavelength of 1540 nm. The output light of the waveguide is coupled out by a 100× microscope objective. The mode pattern is detected by a computer-compatible video camera. For the n_e and n_o channel waveguides, the typical output optical intensities exhibit Gaussian profiles, which are given in Fig. 4. As can be seen, two transverse waveguide modes can be supported in each n_e channel waveguide because of the relatively large index difference and the broad width of channels (~15μm). When we slightly adjust the location of input focus on the input end face of each n_e channel waveguide, the output intensity pattern experiences a process of deformation from an asymmetric single transverse mode to good symmetric double modes, as shown in Fig. 4(a). The intensity pattern of the modes shows good consistency with the waveguide channel structure, i.e., the intensity profile exhibits a nearly uniform ellipsoidal shape for the optical confinement at the down side of the pattern. This optical confinement comes from the index interface between the raised index n_e in the guiding region and the reduced n_e in the index barrier. Near the sample surface, the optical intensity is concentrated because of the intensive optical confinement from the air-LiNbO₃ interface. It can be expected that if we shrink the channel width a single-transverse-mode waveguide at 1540 nm can be obtained. For the n_o barrier-confined waveguide, only a single transverse mode is observed, as shown in Fig. 4(b). This can be simply attributed to the small index difference between the guiding and barrier regions, since the ordinary index in the guiding region is reduced after implantation. Comparing the mode patterns from the two kinds of waveguides, it can be found that the n_e waveguide provides a better optical confinement.

 

Fig. 4. (a) The typical intensity pattern of double transverse modes from the n_e channel waveguide output; (b) the typical intensity pattern of a single transverse mode from the n_o channel waveguide output.

Download Full Size | PDF

The channel waveguide loss at 1540 nm is measured before and after annealing. The light source used (1540 nm) is from Santec TSL-80 tunable semiconductor laser. Since the measurement setup we used is more suitable for the single-mode waveguide, the n_o single-mode waveguide is measured first. The measured waveguide loss is in the range of 0.9–1.2 dB/cm, depending on the width of each waveguide. This value is quite low in comparison with our previous LiNbO₃ samples. As we know, high loss in implanted waveguides is caused mostly due to the high implant dose and the optical leakage from the index barrier. Because the higher the implant dose is, the more defects are produced in the guiding region, which may cause enhanced absorption and scattering loss. In the present experiment, the highest O²⁺ dose is only 1.3×10¹⁵ ions/cm²; the magnitude is at least one order lower than that of a typical He-implanted waveguide. With moderate annealing, those defects can be removed easily. Meanwhile, the index barrier become broad due to three units of energy O²⁺ implantation, this widened barrier plays an important role in reducing the leakage loss. The loss in the n_e channel waveguide is also measured using the same setup. Being as similar as the n_o channel waveguide, a high optical attenuation is observed in the as-implanted samples. After annealing at 250°C for 20 min, the waveguide loss decreases dramatically, i.e., it reaches as low as 0.7 dB/cm. Because the n_e channel waveguide supports double transverse modes, the measured result is only an estimate. However, it can be expected that the lower loss can be obtained in the single-mode n_e channel waveguide. To investigate and improve the light propagation in the implanted channel waveguides, we are carrying out more experiments to optimize the implantation and annealing condition.

3. Summary

Implanted planar and channel waveguides at 1540 nm are formed by one-step MeV multienergy O²⁺ implantation. The single-mode planar waveguide, as well as the spatially distributed, single perpendicular mode and double- or single-transverse modes in channel waveguides are demonstrated through prism coupling and end-face coupling. According to our simulation and analysis, the thickness of the waveguide is approximately 1.7 μm—smaller than the size of typical diffusion LiNbO₃ waveguide. The output intensity patterns show good symmetry in profile, indicating a quite uniform waveguide structure. Low loss of 0.9 and 0.7 dB/cm is obtained in the channel n_o and n_e waveguides, respectively. Since the waveguide loss is governed by factors such as the implant parameters and annealing conditions, improved experiments are being carried out. As is known, implantation has many unique advantages for forming waveguides, especially in optical crystals [3]. This method provides the possibility for forming a more-compact waveguide structure and may be generalized to the other optical crystals such as KTP, to form the nonlinear optical channel structure.