In late 1970’s, a potential application of ion implantation was demonstrated in integrated optics. Good examples are the ion-implanted silica glass and LiNbO₃ waveguide [1, 2]. As an alternative method of fabricating waveguide structures, ion implantation shows impressive advantages over ion diffusion or ion exchange, which are both most frequently-used methods in forming waveguide in optical materials. The main advantage of ion implantation are: (1) a ‘clean’ process, no extrinsic ions are introduced in the guiding region; (2) waveguide can be formed at low temperature, which ensures a steady chemical composition in the waveguide region [3, 4]; (3) universal, as the ion implantation has been proved to enable the formation of waveguide in more than 60 materials by experiments [5, 6]; (4) controllable, the index profile of waveguide can be tailored by adjusting the implantation parameters in the approximate way. However, up to now, application of the ion implantation technology in integrated optics still cannot comparable to that in semiconductor industry. The limitation comes from two reasons: the high ion implantation dose (10¹⁶ ions/cm², compared to the 10¹⁴ ions/cm² in semiconductor industry) and the high waveguide loss (generally, the loss is 1 dB/cm for ion-implanted waveguides), which makes it give way to conventional but affordable diffusion and ion exchange methods. In year of 2001, we discovered that the waveguide can be formed not only by H and He ion implantation but also by relatively-heavier ions, such as Cu ions [7]. Contrasted to the high implantation dose of He or H implantation (more than 10¹⁶ions/cm²), the dose of implanted heavy ions is around 10¹³-10¹⁴ions/cm², in a dose range comparable to that in semiconductor industry. The decrease of ion dose indicates a dramatic drop of implant time by two orders of magnitude, meaning a significant cost cut in fabrication. Meanwhile, low implant dose causes less damage to the crystals structure, enabling low transmission loss in waveguide and nearly non-degraded optical parameters. The waveguide structures have been achieved using this method in many solids even including wide-gap semiconductors, ceramics and organic crystal [8–10]. Among them, single mode channel waveguides have been achieved in both congruent and stoichiometric LiNbO₃ single crystals experimentally [11, 12]. However, the loss in these waveguide is relatively high because of the characteristics from implanted waveguide: intrinsic absorption and scattering loss induced by implantation process and the light leakage loss from the index barrier. In general, annealing is the conventional method employed to reduce the loss in implanted waveguides, but the annealing conditions are often difficult to optimize because an appropriate annealing treatment must ensure to reduce the defects-related loss while not significantly reduce the height and width of damage-induced index barrier.

In the present report channel waveguides with different widths and a Y-branch beam splitter with splitting ratio of 50:50 at 1.5µm are fabricated and characterized in x and z-cut LiNbO₃. It is achieved by combining conventional photolithography pattern process and low dose O⁺ implantation, the same fabrication process as that in Si-semiconductor industry. By using a cumulative annealing treatment with stepwise temperature increment, the transmission loss of waveguide was reduced to 0.17dB/cm. The Y-branch beam splitter can provide equal splitting for both TM and TE mode. Y-branch splitter is a fundamental structure in many passive optical devices as well as in planar lightwave circuits (PLCs). The present Y-branch waveguide may achieve low-loss beam splitting function.

The x and z-cut LiNbO₃ samples with the size of 10 × 20 × 1mm³ were firstly patterned on the surface with Cu mask by photolithography, forming open channels of width from 4 to 15μm for the ion irradiation. Then the implantation was carried out at room temperature. The experiments deal with the same dose of oxygen ions, which is around 8.5 × 10¹⁴ ions/cm², but are carried out in two ways: one is a single implantation up to total dose with the same oxygen ion energy of 3.5 MeV; the other is a series of implantation step with ions in the sequence of energy of 4.0 MeV, 3.6 MeV and 3.0 MeV, accumulating to the same total dose. Thus, the formed waveguides has an only difference in the thickness of index barrier. The as-implanted samples were then measured using Metricon 2010 Prism Coupler at λ = 1.54 μm, and each of them showed only one distinct propagated TM mode. For TE mode, the m-line spectrum showed also a propagated mode but with poor light confinement. Since the waveguide properties depend only on the implanted-induced index profile, so the following results are all means on z-cut LiNbO₃ samples with TM polarization. Each sample was then annealed in air ambient from 200°C to 280°C with temperature increment of 15~20°C every step, duration is from 20 min to 3 hours depending on the monitoring of the loss measurement.

Laser light with wavelength around 1.54 µm was edge-coupled into the waveguides by a 10 × objective lens. A 100 × 0.9 N.A. objective lens was used to collect the light out of the waveguide end face. A camera connected with a computer was used to record the near-field optical intensity pattern from end face of crystal. The propagation losses of the waveguides were measured using the Fabry-Perot resonance method at 1.54 µm wavelength. The sample is heated slowly to a temperature change of several degrees Celsius, and the transmitted light intensity is recorded. The loss is derived from the intensity contrast [13].

The waveguides show different dark mode characteristics from that of waveguides formed by ion exchange or diffusion, as the latter guide the light due to the enhanced refractive index near sample surface. It is well known that the index profile in implanted waveguide has a typical “reduced index” layer, called “index barrier”, which acts as another waveguide boundary to confine light, shown in Fig. 1(a) . For waveguide with n_o index profile, optical confinement is achieved between sample surface and the optical barrier. For n_e, waveguide function comes from both the optical barrier and an increased index in waveguide region. In the present single-mode waveguide, the index profile in the waveguide by multi-energy ion implantation was reconstructed according to the damage profile, which is based on the theoretical model on index change in ion implant LiNbO₃ [14], the index profile in channel waveguide and light propagation direction is shown in Fig. 1(b).

 

Fig. 1 (a) a schematic illustration of index profile in typical ion-implanted LiNbO₃ waveguide; (b) the n_e index profile in 7-µm width channel waveguide.

Download Full Size | PDF

Absorption and scattering can give rise to loss in ion implanted waveguide due to implantation-induced lattice disorder. However, since the implantation dose is very low (~10¹⁴ions/cm²), induced defects can be very low as well. It can be expected that by applying an annealing treatment at moderate temperature, a reduced waveguide loss and a recovery of lattice order can be achieved simultaneously. Another loss factor is light leakage, because light power can tunnel through the index barrier and be radiated into the substrate. An effective method to prevent light from leaking through index barrier is to broaden the width of the barrier, which can be met simply by multi-energy ion implantation. It has been demonstrated in our experiment that the loss of waveguide by multi-energy implantation is almost half the value of those by single-energy ion implantation. By comparing the near field patterns of output lights from channel waveguides, a better optical confinement was observed in multi-energy implanted waveguide.

After appropriate annealing, one of the measured Fabry-Perot transmission resonances/fringes in channel waveguide is shown in Fig. 2 (left). The measured losses of waveguides with different waveguide width are shown in Fig. 2 (right). The lowest loss 0.17dB/cm comes from the 7-µm wide channel waveguide.

 

Fig. 2 (left) the measured Fabry-Perot transmission resonances/fringes from 7-µm channel waveguide; (right) the measured losses of waveguides with different waveguide width under annealing condition of 200°C, 20min; 250°C, 40min and 270°C min.

Download Full Size | PDF

For analyzing and evaluating the waveguide characteristics, we measured and simulated the light propagation in the implanted channel waveguide, which has the index profile given in Fig. 1(b). Figure 3 (left) shows a measured near field pattern of output light from a 7-µm-wide channel waveguide. By using Beam Propagation Method, the simulated transverse field profile of TM mode after light propagates at 2000µm is given in Fig. 3 (right). The calculated field profile has a good consistence with the experimental result. Since the index profile of each channel waveguide has a gradual decrease in both side boundaries, the measured field intensity doesn’t show a sharp decline as in simulated results.

 

Fig. 3 (left) the measured near field pattern of output light from a 7-µm-wide channel waveguide in z-cut LiNbO₃;(right) the simulated transverse field profile of TM mode after light propagate at 2000μm.

Download Full Size | PDF

The waveguide loss of 0.17dB/cm is the best result among all of the single mode channel waveguides formed by implantation, which is entirely comparable with that of waveguide devices currently used in market. Although both TE and TM mode can be transmitted in the same waveguide simultaneously, it is apparent that TM mode has a much low waveguide loss. The result is consistence with our former analysis and is preferable to application, since in the z-cut LiNbO₃ a larger optic-electric coefficient is available for TM mode. The low propagation loss and low material composition contamination (‘clean process’) will highlight the implantation as a competitive method in forming waveguide for integrated photonic applications.

A useful waveguide device is the beam splitter, which divides the optical power in an input waveguide equally and sends them to out waveguides. Actually, such a “Y-branch” 
structure is the basis of many optoelectronic devices, such as modulators and couplers. Similar implantation parameters were employed to fabricate Y-branch waveguide structure. The angular separation of two divergent channels is 1.1°, and the separation of two parallel branches is 30μm. The width of each channel is 13 μm. The near field pattern of output light from Y-branch structure is shown in Fig. 4(a) and 4(b), the measured splitting ratio is about 50:50. It is clear that light energy was divided into two parts by the Y-branch and guided out. A similar calculation is done to simulate the light propagation in Y-branch and the transverse field profile of TM mode after light propagate at 4500µm is shown in Fig. 4 (c) and (d). Since the waveguide is multi-mode, so the amplitude oscillation in Fig. 4(d) is due to mode beating.

 

Fig. 4 (a) and (b) are the measured output near field pattern from Y-branch structure; (c) and (d) are the simulated output near field pattern and the light propagation path (waveguide width in (d) is normalized to 2μm for illustration). In the simulation, waveguide width of 13μm, angular separation of two divergent channels of 1.1° and the separation of two parallel branches of 30μm are used.

Download Full Size | PDF

We adjust the simulation parameters to test the splitting efficiency of the Y-branch. If keeping the waveguide width and index profile unchanged, the index contrast in waveguide determines the function of beam splitter. In the present experiment, Y-branch structure with an average index contrast higher than 2% can achieve beam splitting effectively. According to our previous researches, 2% index contrast can be easily obtained in many implanted waveguides when appropriate implantation parameters are applied. It can be expected that a curved waveguide structure, even micro-ring [15], with low loss may be achieved by using the low dose implantation following with annealing treatment. This approach has a potential application in integrated optics and optoelectronics devices.

In conclusion, a multi-energy ion implantation technique is introduced to fabricate low-loss channel waveguides in LiNbO₃ substrate. The ion implantation dose is very low (~10¹⁴ions/cm²), which will facilitate the application of ion implantation technique in integrated optics. The loss of the channel waveguides can be as low as 0.17dB/cm. A waveguide splitter is successfully fabricated. Splitting of light energy is observed.