Lithium niobate (LN) is an excellent crystal material that not only has unique properties in optoelectric, piezoelectric, and photoelastic but also exhibits mechanical and chemical stability as well. As an extremely versatile nonlinear optical waveguide, LN has been used in many fields. Its high electro-optic and nonlinear optical coefficients are used for various photonic applications, such as harmonic generation, optical parametric oscillation, and electro-optic devices. A combination of waveguide structures with its optical properties makes LN the most widely used material in integrated optics. The devices such as switches, interconnects, multiplexers, waveguide lasers, and nonlinear optical waveguides have been fabricated in LN. A waveguide structure is basic for all these devices. Among the several methods used to form waveguides in LN, titanium indiffusion and annealed proton exchange are the most popular [1

A. Sjoberg, G. Arvidsson, and A. A. Lipovskii, “Characterization of waveguides fabricated by titanium diffusion in magnesium-doped lithium niobate,” J. Opt. Soc. Am. B 5, 285–291 (1988). [CrossRef]

,2

J. L. Jackel, C. E. Rice, and J. J. Veselka, “Proton exchange for high-index waveguides in LiNbO₃ ,” Appl. Phys. Lett. 41, 607–608 (1982). [CrossRef]

]. As an alternative method, ion implantation provides another way to fabricate waveguides in LN. The first ion-implanted LN waveguide was demonstrated by high-dose He implantation in the late 1970s [3

P. D. Townsend and L. Zhang, Optical Effects of Ion Implantation (Cambridge University Press, Cambridge, UK, 1994). [CrossRef]

]. In the implanted waveguide the light is confined between the surface (air) and a low-index layer caused by implantation. It differs from an indiffused and exchanged waveguide in which the guiding lights are transmitted in a raised index layer. In contrast with a diffused or exchanged waveguide, an ion-implanted waveguide is formed in a nonequilibrium physical process. A beam of energized ions injects into the substrate and is slowed down by collision with the atoms in the substrate. The penetrated depth of ions depends on both the energy of ions and the density of the substrate. For a light-ion- (e.g., He or H) implanted waveguide, the low-index layer is formed at the end of an ion’s track, where ions aggregate and cause a reduction of index in the substrate because the material density is diluted. A guiding region between the surface (e.g., air) and the low-index layer (index barrier) is therefore formed [4

F. Lu, M.-Q. Meng, K.-M. Wang, X.-D. Liu, H.-C. Chen, and D.-Y. Shen, “Planar optical waveguide in Cu-doped potassium sodium strontium barium niobate crystal formed by megaelectron-volt He-ion implantation,” Opt. Lett. 22, 163–165 (1997). [CrossRef] [PubMed]

,5

M.-Q. Meng, F. Lu, K.-M. Wang, F.-X. Wang, W. Li, C.-Q. Ma, D.-H. Jiang, and Q.-M. Lu, “Property analysis of planar optical waveguide in NYAB formed by He⁺ ion implantation,” Electron. Lett. 33, 1045–1047 (1997). [CrossRef]

]. In an implanted waveguide the material properties in the guiding region remain almost unchanged because the guiding region itself is subject mainly to ionization processes that have a minimal effect on lattice order for most types of insulator, and any color centers produced are easily annealed [6

E. E. Robertson, R. W. Eason, M. Kaczmarek, P. J. Chandler, and X. Huang, “Ion beam manipulation of the photorefractive properties of strontium barium niobate planar waveguides,” Opt. Lett. 21, 641–643 (1996). [CrossRef] [PubMed]

–8

K. E. Youden, S. W. James, R. W. Eason, P. J. Chandler, L. Zhang, and P. D. Townsend, “Photorefractive planar waveguides in BaTiO₃ fabricated by ion-beam implantation,” Opt. Lett. 17, 1509–1511 (1992). [CrossRef] [PubMed]

]. Although undegraded optical properties can be expected in such kinds of waveguide, the transmission efficiency in the waveguide remains a problem. Because the low-index barrier has a limited thickness and the light in the guiding region could tunnel through the low-index layer, a high-transmission loss could result. To reduce the waveguide loss one can use either an annealing treatment or multiple energy implants, because postimplant annealing is an effective way to reduce absorption and multiple energy implants can be used to broaden the barrier and reduce the leakage of light if optical attenuation caused by narrow barrier tunneling is a problem. In 2001 Hu et al. determined that the extraordinary index of LN can be brought near to the surface of LiNbO₃ by low-dose Si, Cu, or C implantation [9

H. Hu, F. Lu, F. Chen, B.-R. Shi, K.-M. Wang, and D.-Y. Shen, “Monomode optical waveguide in lithium niobate formed by MeV Si⁺ ion implantation,” J. Appl. Phys. 89, 5224–5226 (2001). [CrossRef]

,10

H. Hu, F. Lu, F. Chen, B.-R. Shi, K.-M. Wang, and D.-Y. shen, “Extraordinary refractive-index increase in lithium niobate caused by low dose ion implantation,” Appl. Opt. 40, 3759–3761 (2001). [CrossRef]

]. In the meantime, single-mode waveguides were obtained in the implanted samples. If such a single-mode waveguide could be supported completely by a raised index layer, a no-leakage implanted waveguide should be expected. However, even though we made several attempts to anneal the samples and to reduce absorption loss, the waveguide was still highly lossy. We did not observe any bright line when we used end-face coupling. An index-raised waveguide can be deduced from the measured mode effective extraordinary index, which is higher than that of a substrate. However, we cannot determine whether a low-index barrier would still play a part in confining light, since the index profile in such a waveguide cannot be constructed if we depend only on the measured waveguide mode. In an attempt to investigate the properties of an index-raised waveguide and to reduce waveguide loss, we performed the following experiment. Z-cut LN was implanted with O²⁺ ions at three different energies to broaden the possible index barrier. The total dose of ions is at the magnitude of 10¹⁴ ions/cm², approximately 2 orders of magnitude lower than that of conventional He-implanted waveguides. A limited O²⁺ ion range was defined in advance to make sure that only one mode can be supported in the guiding region. After implantation, the pronounced increased extraordinary index was obtained in some of the samples. The index barrier was formed at the end of the O²⁺ ion track and was widened by multienergy implantation. In the as-implanted sample one dark mode was obtained by prism-coupling measurements. A bright line was observed only in the annealed sample by use of end-face coupling. The possible index profile in the waveguide was reconstructed according to our theoretical model and the results were discussed.

Z-cut LN, with one polished upper surface and two polished end faces, were implanted by O²⁺ at room temperature. The energies of O²⁺ ions are 1.0, 1.5, and 2.0 MeV at doses of 1.2×10¹⁴, 1.5×10¹⁴, and 4×10¹⁴ ions/cm², respectively. During implantation the LN samples were tilted 7O off normal to avoid channeling. After implantation, one sample was annealed in the ambient air at 250 °C for 30 min to remove the point defects and to decrease absorption loss. We then used prism coupling and end-face coupling to investigate the waveguide structure. A single narrow and deep dip in the dark-mode profile was observed by prism coupling for both as-implanted and annealed samples. The coexistence of a shallow and broad dip in the tail of the mode profile indicates that the waveguide exhibits some characteristics of a barrier-confined waveguide. Figure 1 shows the dark-mode profile obtained with prism coupling measurements (Model 2010, Metricon Corporation, Pennington, New Jersey). The inlet in the figure shows a comparison of the mode profile before and after annealing. The peak of the dip drifts slightly to the left after annealing, which means that the refractive index rises slightly after thermal annealing.

End-face coupling was performed with a He-Ne laser at 633 nm. No detectable output was obtained from the as-implanted sample, which is attributed to the high absorption and scattering loss caused by the implantation process. A bright line was observed only in the annealed sample, which is shown in Fig. 2. Although the bright line does not show uniform brightness in the distribution, which is due mostly to inferior polish of the output end face of the sample, the brightness of the output line still indicates that the transmission efficiency in the waveguide is quite good.

Since the index profile in the waveguide cannot be constructed on the basis of one measured mode, we tried to describe the index profile with our theoretical model [8

K. E. Youden, S. W. James, R. W. Eason, P. J. Chandler, L. Zhang, and P. D. Townsend, “Photorefractive planar waveguides in BaTiO₃ fabricated by ion-beam implantation,” Opt. Lett. 17, 1509–1511 (1992). [CrossRef] [PubMed]

]. According to the model, enhancement of the extraordinary index induced by the implantation can be a synthesized effect of four factors: degradation of spontaneous polarization, modification of molar polarization, molar volume change, and the optoelastic effect. Among these factors, spontaneous polarization dominates the increment of extraordinary index. Figure 3 is the lattice damage distribution in LN caused by O²⁺ implantation according to the transport of ions in matter code. Substituting known parameters of LN into parameters such as the electro-optic coefficient, the reconstructed extraordinary index profile is given in Fig. 4, where S is a variable parameter that represents the effect from the lattice damage. The results show that the extraordinary index profile is sensitive to the lattice damage. For a slightly damaged lattice (when the O²⁺ dose is low and S is small), the extraordinary index is enhanced in the whole guiding region, similar to the waveguide formed by indiffusion or the exchange method. As the ion dose increases, the index at the end of the O²⁺ track begins to decrease, mostly because the density of LN was diluted because of the aggregation of O²⁺ ions. When the O²⁺ dose is high enough, the index at the end of the ion track descends so significantly that it might be lower than that of the substrate. In this case, a low-index layer is formed to serve as an index barrier for a waveguide.

We now discuss the possible situation of the index profile in a waveguide. If the index profile in a waveguide can be assumed such as is the case in the first situation, where S is small (S<0.5) and the extraordinary index is raised throughout the ion range, such as that shown in Fig. 4, a hypothetic asymmetric slab waveguide might be a reasonable structure. An ideal model is shown in Fig. 5(a).

According to waveguide theory, the cutoff and guidance conditions of the TM mode in an asymmetric dielectric slab waveguide are

where d is the depth of the waveguide; n ₀, n ₁, and n ₂ are the refractive index in air, the guiding region, and the substrate, respectively; k ₀ and k are the wave vectors in air and in the guiding region, respectively; and k=(k 0 2n 1 2 -β ²)^1/2 □ p=(β ²-k 0 2n 2 2 )^1/2 □ q=(β ²-k 0 2n 0 2 )^1/2.

According to the measured TM mode, the effective extraordinary index is approximately 2.2116, which is larger than that of the LN substrate (2.2028). If we assume that n ₀=1 (air), n ₁=2.2150 (the index in the guiding region should be slightly larger than the measured effective extraordinary index; see the simulated result in Fig. 4), and n ₂=2.2028 (LN substrate), by substituting m, n ₀, n ₁, n ₂, and k into Eq. (2), the depth of a single-mode waveguide can be calculated at approximately 3λ, where λ=633nm. The result means that, if the waveguide is a complete index-raised layer, the depth of the guiding region should be at least 1.9µm. This value is much bigger than the simulated depth of the waveguide in Fig.4, which is only approximately 1.4µm.

If the second situation is reasonable (such as when S>1 in Fig. 4), the light is still confined between the surface (air) and the low-index barrier. If we assume that the index barrier is lower than that of a LN substrate by 0.5%, which is reasonable for most of the implanted waveguide with low dose implantation, the possible index profile can be described as Fig. 5(b). Because the barrier is thick enough compared with the depth of the waveguide, it is also reasonable to neglect the tunneling effect and to assume that an asymmetric slab is still available. In this case we determined that the depth of the waveguide is approximately 1.3λ, at approximately 0.82µm. The result is consistent with the simulated profile in Fig. 4, where the possible waveguide depth is approximately 0.8µm. The model in Fig. 5(b) is also supported by our waveguide measurement. If the waveguide has a structure such as that shown in Fig. 5(a), there should be no loss due to leakage in the waveguide. Broadening the index barrier would not have any effect on waveguide loss. However, we found that the bright output line can be obtained only in a multienergy implanted waveguide. By prism coupling a single-mode waveguide can also be found in the single energy implanted waveguide, but no bright line is observed. So the current low-loss waveguide can be attributed to the broadened index barrier, which plays a significant role in reducing leakage loss.

We have formed a single-mode waveguide in LiNbO₃ by O²⁺ implantation at three different energies. We measured the waveguide by using both prism coupling and end-face coupling methods. We obtained a bright output line only in the annealed sample. The index profile in the waveguide is constructed according to our theory model. The analysis results indicate that the probable index profile in the waveguide should be a structure of the index-raised guiding region sandwiched between surface (air) and a low-index barrier. Since a broadened index barrier could prevent the light from leaking, a low-loss waveguide with an index-raised layer can be achieved by multienergy implantation. Because the implanted waveguide can be tailored by modulating the implantation parameters, the versatility of this method facilitates processes that are not available for conventional methods. In addition, the structure of an index-raised layer plus barrier confinement has another unique advantage in the fabrication of channel waveguides. Instead of conventional multistep implantation [11

B. Vincent, A. Boudrioua, J. C. Loulergue, P. Moretti, S. Tascu, B. Jacquier, G. Aka, and D. Vivien, “Channel waveguides in Ca₄GdO(BO₃)₃ fabricated by He⁺ implantation for blue-light generation,” Opt. Lett. 28, 1025–1027 (2003). [CrossRef] [PubMed]

], the current method provides the possibility that channel waveguides can be fabricated simply by implantation.