1. Introduction

Photorefraction in LiNbO₃ (LN) is a detrimental effect in the view of integrated optics. Although a family of Ti (vapor Zn)-diffused Er:LiNbO₃ waveguide lasers (amplifiers) and integrated devices have been demonstrated over the past years [1–3], the photorefractive effect not only affects the performances of these devices, but also limits both the pumping and operating wavelengths, and hence hinders further development of novel devices. Although doping with > 4.9 mol% Mg can eliminate the effect [4], heavy Mg doping causes difficulty in growing high optical-quality single-crystal when co-doped with the luminescent rare-earth ions, and decrease in both solubility and diffusivity as the rare-earth ions are incorporated by diffusion method. It is essential to seek other dopants that have lower threshold concentration of photorefractive damage. In addition to the much studied Mg²⁺, other dopants like the divalent Zn²⁺ [5], trivalent Sc³⁺ [6], Tm³⁺ [7], and In ³⁺ [8], and tetravalent Hf⁴⁺ [9], Zr⁴⁺ [10] and Sn⁴⁺ [11] can also suppress the photorefractive effect, and for a congruent crystal the threshold concentration is 7.5, 2.0, unknown, 3.0, 4.0, 2.0 and 2.5 mol%, respectively. Among them, the Sc³⁺ and Zr⁴⁺ display the lower threshold, 2 mol% only. The low threshold concentration favors to improve the optical quality of crystal and prompt the diffusivity and solid solubility of codoped rare-earth ions. Therefore, an LN crystal doped with > 2 mol% Sc³⁺ or Zr⁴⁺ would be more promising for developing a photorefractive-damage-resistant waveguide device.

As an alternative, a Zr⁴⁺-doped LN can be prepared by either the crystal growth technique or the method of Zr⁴⁺ diffusion doping. The refractive index of Zr⁴⁺-doped LN is the basic knowledge for applications. In particular, for the conventional Ti-diffused LN waveguide, which is usually fabricated by in-diffusion of a 6-10-µm-wide, 100-nm-thick Ti-strip at ~1050 °C for 9 h, the Ti⁴⁺-induced n_o and n_e increments at the 1.5 µm are only ~0.005 and ~0.012, respectively. It is unclear if the Zr⁴⁺ diffusion-doping contributes to the LN index. If so, it is crucial to know if the contribution is comparable to the Ti⁴⁺-induced increment. In the case of bulk Zr⁴⁺-doping, Nava et al. [12] has reported that the Zr⁴⁺ doping definitely changes the LN index and the change is comparable to the index increment of Ti:LN waveguide. It is unclear if this is also true for the Zr⁴⁺ diffusion doping case. It is essential to carry out a systematic study on Zr⁴⁺ diffusion-doping effect on the LN index. Present work focuses on this aspect of study and shows that the effect of diffusion-doping is small and clearly different from the bulk doping case.

2. Experimental methods

Twelve Z-cut congruent LN plates with 0.5-mm thickness and optical grade surfaces were used in present study. At first, the n_o and n_e values on surface of each as-grown plate were measured. Then, a ZrO₂ (99.99%) film with a thickness of (40-100) ± 2 nm was coated onto a half surface of each plate while the other part remaining uncoated for reference.

After the ZrO₂ film coating, the plates were annealed in the atmosphere of wet O₂ in order to suppress the Li₂O out diffusion. The diffusion temperature varied from 1000 °C to 1060 °C while the diffusion duration ranged from 2 h to 30 h. After diffusion, the n_o and n_e values at the doped and undoped surface parts of each plate were measured. The index was measured at the 1311 and 1553 nm wavelengths using a Metricon 2010 prism coupler. The measurements, which were repeated at five different places on doped or undoped surface part of each sample, give an averaged index value.

Secondary ion mass spectrometry (SIMS) was used to analyze the depth profile of the diffused Zr⁴⁺ ions in the twelve samples in total. The analysis was accomplished by a time-of-flight second ion mass spectrometry [ToF SIMS V]. A Cs⁺-beam (~45 μm in diameter) of 28-35 nA at 3 keV was used to sputter a crater of either 120 × 120 or 150 × 150 μm² and a pulsed Bi⁺ beam of 1 pA at 25 keV was used to analyze the negative secondary ions ⁷Li, ¹⁰⁹NbO, ³²O₂ and ¹²³ZrO₂ as a function of time. Since the Zr has a rather low sensitivity in the case of usual positive ion detection scheme, here the negative ion detection scheme was adopted to promote the Zr sensitivity.

3. Results and discussion

3.1 Determination of surface Zr⁴⁺ concentration C_Zr

First of all, we describe how to determine the Zr⁴⁺ concentration C_Zr at the diffusion surface. As a representative, Fig. 1(a) shows the depth profiles of ⁷Li, ¹⁰⁹NbO, ³²O₂, and ¹²³ZrO₂ signals detected from the sample coated with 80 nm thick ZrO₂ film and annealed at 1060 °C for 10 h. The red-ball curve represents the ZrO₂ profile. As expected, the substrate signals ⁷Li, ¹⁰⁹NbO and ³²O₂ show constancy with the depth. The ZrO₂ profile can be well fitted by a Gaussian function (this is also the case for the other samples),

 

Fig. 1 (a) Depth profiles of ⁷Li, ¹⁰⁹NbO, ³²O₂, and ¹²³ZrO₂ SIMS signals detected from a Z-cut LN plate coated with 80 nm ZrO₂ and annealed at 1060 °C for 10 h in wet O₂. The resultant C_Zr at surface is 4.1 ± 0.3 mol%. (b) Zr⁴⁺ diffusion-doping induced (a) Δn_e and (b) Δn_o measured as a function of surface Zr⁴⁺ concentration C_Zr.

Download Full Size | PDF

As the oxygen is homogeneous over the substrate, the ZrO₂ profile represents actually the Zr⁴⁺ profile. Based upon Eq. (1), the Zr⁴⁺ concentration profile can be expressed as

Table 1. Surface C_Zr, refractive indices (at 1553 nm) at Zr⁴⁺-doped and undoped surface parts of LN plates prepared by Zr⁴⁺ diffusion. The Δn_o and Δn_e values are also given.

View Table

When studied the bulk Zr⁴⁺-doped nearly stoichiometric LN, Kovács et al. [13] suggested that the tetravalent ion doped LN has a defect model similar to the widely accepted one for the divalent (Mg²⁺ [14]) and trivalent (Sc³⁺ [15]) ion doped LN. The model suggests that the dopant enters the Li site and eliminates the antisite as its concentration is lower than the threshold. Higher than the threshold, some dopants enter into the Nb sites. Although one cannot confirm that this is also the case for the diffusion doping, a high concentration (up to 16 mol% considered here) increases the probability that the Zr⁴⁺ enters into the Nb site. For a substrate element like Nb, Li or O, the SIMScannot measure a constituent atom alteration of only several percents because the Cs⁺ beam sometimes has several percents fluctuation in intensity. Thus, from the Nb curve in Fig. 1(a) one cannot deduce convincing information about the Zr⁴⁺ occupation in lattice.

3.2 Zr⁴⁺ diffusion-doping effect on surface refractive index

As the refractive indices at the 1311 nm wavelength show similar Zr⁴⁺ doping effect to those at the 1553 nm wavelength, here we only discuss the 1553 nm results. Table 1 brings together the index values at the 1553 nm. The error is ± 1 × 10⁻³. Obviously, the difference of n_o or n_e values at the doped and undoped parts of crystal surface reflects the Zr⁴⁺-doping effect, i. e., the above-mentioned Δn_o and Δn_e, which are also given in Table 1 for convenience. Figures 1(b) and 1(c) illustrate the Δn_o and Δn_e versus the surface C_Zr up to 16 mol%. For convenience, the error bar is indicated for each data. One can see that the Zr⁴⁺-doping contribution to the surface index is small for both n_e and n_o. The largest changing amplitude, which is only 0.7 × 10⁻³ for n_e and 0.3 × 10⁻³ for n_o, is within the experimental error of 10⁻³. We thus conclude that the Zr⁴⁺ diffusion-doping has little contribution to the LN index. This is clearly distinguished from the bulk-doping case. Nava et al. [12] has reported that in bulk-doping case the Zr⁴⁺ considerably affects the LN index. The n_o and n_e depend on the C_Zr in a non-monotonic manner. Both decrease with C_Zr when C_Zr < C_th ( = 2 mol%). Near the C_th, the Δn_o and Δn_e reach their respective extremes −4.5 × 10⁻³ and −5.5 × 10⁻³, respectively, which are definitely beyond the error given by them [12], 3 × 10⁻³, and are comparable to the index increment of the usual Ti:LN waveguide. As C_Zr > C_th, both n_o and n_e recover gradually, and Δn_o recovers to −1 × 10⁻³ and Δn_e to 2.5 × 10⁻³ at C_Zr = 3 mol%. As the C_Zr is further increased, both n_o and n_e increase continuously in appearance. For straightforward comparison of two cases of bulk- and diffusion-doping, in Figs. 1(b) and 1(c) we have set the maximum scales of Δn_o and Δn_e as their respective amplitude extremes in the bulk-doping case. One can see that the difference between the two cases is evident. As described above, the bulk-doping case shows definite effect of concentration threshold of photorefractive damage. To check if the effect also takes place in the diffusion-doping case, we have carefully investigated the region between 2 and 5 mol% by considering more C_Zr values. One can see from Figs. 1(b) and 1(c) that the threshold effect cannot be identified at all.

In case without demonstrating that the Zr⁴⁺ diffusion-doped LN samples and the Zr⁴⁺ bulk-doped crystals have the same properties, it would be no surprise to observe different refractive indices. Indeed, the contribution of a dopant to the substrate index depends on many other factors, in addition to the dopant concentration. These include the strain caused by dopant-induced lattice mismatch, the enhancement of electronic polarizability of crystal cell due to the large difference of electronic polarizability between the constituent cations and the dopant, as well as the elasto-optic effect caused by lattice contraction or expansion. One or all of these factors may contribute to the index, and some factors may have a positive contribution while the others may have a negative contribution. The contributions of these factors may change from one dopant to another. Even for the same dopant, the contributions may also change from one ion environment to another. Regarding the issue concerned here, the ion arrangement may be different from crystal growth to in-diffusion, for which the lattice is already present. The change of ion arrangement leads to the alteration of ion environment and hence the change of the contributions of the above-mentioned factors to the LN index. It is possible that the net contribution is definite in the bulk-doping case while minor in the diffusion-doping case.

3.3 Zr⁴⁺ diffusion-doping effect on surface Li-composition and Li₂O out diffusion issue

It is essential to know the Li⁺ composition at the Zr⁴⁺-doped and undoped surface parts of the studied samples and hence evaluate the Li₂O out-diffusion extent during Zr⁴⁺ in-diffusion into the bulk of crystal. Here, the Li₂O-contents at the doped and undoped surface parts of each plate were evaluated from the measured refractive indices using the Li₂O-content-dependent Sellmeier equation [16]. Evaluations from the indices at 1311 and 1553 nm give an averaged Li₂O-content. The error of refractive index, 10⁻³, yields a Li₂O content uncertainty of 0.1 mol%. For the as-grown congruent plate, the Li₂O content was evaluated as 48.5 ± 0.1 mol%, which is in excellent agreement with the nominal value. The agreement means that all of the Li₂O-content data evaluated hereafter are sound. After the diffusion process, the averaged Li₂O content at the Zr⁴⁺-doped surface is the same as that at the Zr⁴⁺-free surface within the ± 0.1 mol% error and this is the case for all the studied samples. Moreover, after the diffusion process the Li₂O contents on surfaces of all plates studied are also the same as that of the as-grown plates, implying that the Li₂O out-diffusion is ignorable for all the studied samples.

Finally, there is a point to be clarified regarding the little Zr⁴⁺ diffusion-doping effect on the LN index. One may query that the observed little effect may be due to the occurrence of such a circumstance: the Zr⁴⁺ diffusion-doping actually contributes negatively to the substrate index and the negative contribution is neutralized by the positive contribution due to the decrease of Li composition arising from the Li₂O out diffusion. In other words, the two contributions cancel each other out, resulting in a small Zr⁴⁺ diffusion-doping effect in appearance. Next, we exclude this possibility. The authors have done a large number of experiments on an ion diffusion-doping into LN. The previous studies have shown that the diffusion temperature is the major factor responsible for the Li₂O out diffusion [17]. Here we exemplify the Er³⁺ diffusion in the LN crystal. Under the same wet O₂ atmosphere and after the same 60 h duration, the Er³⁺ diffusion at 1130 °C resulted in 0.3 ± 0.1 mol% Li₂O-content reduction due to the Li₂O out diffusion, while the out diffusion was not measurable as the diffusion temperature is 1100 °C or lower. This is mainly because the Li diffusivity increases by about two times as the temperature goes from 1100 °C to 1130 °C [18]. Regarding the Zr⁴⁺-doped samples studied here, as the highest temperature adopted is only 1060 °C and the longest duration is only 30 h, it is reasonable that the Li₂O out-diffusion is not measurable for all the samples. Thus, the possibility that two contributions neutralize each other is small and the Zr⁴⁺ diffusion-doping effect on the LN index is indeed small.

4. Conclusion

In conclusion, Zr⁴⁺ diffusion-doping has little contribution to the refractive index of LN substrate. The conclusion is justified by excluding the possibility that the Li₂O out diffusion took place in the diffusion procedure. The contribution is small in comparison with the index increment of usual Ti:LN waveguide. The little effect of diffusion-doping is clearly different from the bulk doping case, in which both ordinary and extraordinary indices of the LN substrate show definite Zr⁴⁺ doping concentration and threshold effects. The difference is attributed to the different ion arrangements from the crystal growth to the in-diffusion. Future work should concentrate on evaluating the optical damage resistance and studying the electro-optic and nonlinear optical properties of the Zr⁴⁺ diffusion-doped crystal.