1. Introduction

Serious photorefractive effect not only affects the performance of LiNbO₃(LN)-based passive and active devices, but also limits both the pumping and operating wavelengths and hence hinders the development of new devices. To suppress the effect, the LN crystal is doped with photorefractive resistant ions such as Mg²⁺ [1], Zn²⁺ [2], Sc³⁺ [3], Tm³⁺ [4], In³⁺ [5], Hf⁴⁺ [6], Zr⁴⁺ [7] and Sn⁴⁺ [8]. Among them, Sc³⁺ or Zr⁴⁺ doping shows the lowest concentration threshold C_th = 2 mol%, which is two and half times lower than that of widely studied Mg²⁺ doping, ~5 mol%. Low C_th is desired to improve the material homogeneity and optical quality of the LN crystal when codoped with Er³⁺, and to increase both the diffusivity and solid solubility of codoped Er³⁺. Ti-diffused LN (Ti:LN) waveguide is one of basic components for integrated optics. Ti:Zr:Er:LN waveguide would be a promising active component for developing various optical damage resistant active devices. The realization of such waveguide would enable to develop some new active devices such as 980-nm-pumped green upconversion and mid-infrared (2.7 µm) waveguide lasers, and various nonlinear optical devices based on periodically poled LN waveguide. These devices are not influenced by the photorefractive effect and can work at high optical intensity in both visible and infrared regimes. Towards the goal, one should develop at first the Ti:Zr:Er:LN strip waveguide. In the previous papers, Zr⁴⁺ diffusion properties [9], Ti⁴⁺/Zr⁴⁺ co-diffusion characteristics [10], and Ti:Zr:LN passive strip waveguide [11] have been reported. Here, we report the implementation of the Ti:Zr:Er:LN active strip waveguide for integrated optics.

2. Experiment

As an alternative to the incorporation method by crystal growth, Er³⁺ and Zr⁴⁺ can be incorporated into the LN by diffusion method [12]. To get the maximum optical gain and resist the optical damage, it requires that both Er³⁺ and Zr⁴⁺ profiles entirely cover the waveguide. Previous studies show that both Er³⁺ and Zr⁴⁺ diffuse much slower than Ti⁴⁺ [10]. By taking these two factors into consideration, we adopted a three-step process to fabricate a Ti:Zr:Er:LN strip waveguide, i. e., diffusion in sequence of Er³⁺, Zr⁴⁺ and Ti⁴⁺.

After fabrication, the Ti:Zr:Er:LN strip waveguide was fully characterized.

3. Results and discussion

The surface refractive indices at the doped and undoped surface parts were measured at the 1553 nm wavelength using a Metricon 2010 prism coupler. The measurement, which has an error ± 1 × 10⁻³, shows that the n_o (n_e) value is 2.2112 (2.1407) at the Zr⁴⁺/Er³⁺-codoped surface part and 2.2116 (2.1403) at the undoped part. The Er³⁺ and Zr⁴⁺ codoping contribution is on the order of 10⁻⁴ for both n_o and n_e. It is concluded that Er³⁺/Zr⁴⁺ codoping contribution to index of LN can be thought as null within the error, consistent with their single doping case [9,13]. The contribution is small compared with the Ti-induced ordinary (extraordinary) incrementΔn_o (Δn_e) in a usual Ti:LN waveguide [14], which is ~0.006 (0.012) at λ = 1.5 µm.

The Li-composition on the sample surface was evaluated on the basis of Li₂O-content-dependent birefringence [15] and the measured refractive index. After the Ti⁴⁺ diffusion, the Li₂O-contents at the Er³⁺/Zr⁴⁺-codoped and undoped surface parts are evaluated as 48.2 and 48.3 ± 0.1 mol%, respectively. We note that the Li₂O-contents at both the doped and undoped surface parts are the same within the error and the three annealing processes caused only slight decrease of Li₂O-content at both the doped and undoped parts due to Li out-diffusion. The situation is complicated in the waveguide layer because of co-presence of Ti⁴⁺, Er³⁺ and Zr⁴⁺. Moreover, the Ti⁴⁺ presence induces local index increase. In addition, the prism coupler cannot accurately measure a strip waveguide. All of these factors make it difficult to obtain the correct Li₂O content of strip waveguide. One has no choice but to evaluate it by referencing the situation on the Er³⁺/Zr⁴⁺-codoped surface part. Although the exact Li composition at the waveguide surface cannot be obtained, it should not have a large deviation from that on the Er³⁺/Zr⁴⁺-codoped surface part, i. e., the waveguide is in an environment of Li-deficient composition slightly away from the congruent point.

Figure 1(a) shows the morphology of the waveguide surface under the magnification of 1000 × . One can see that the waveguide surface shows slight unevenness.

 

Fig. 1 (a) Image of Ti:Zr:Er:LN strip waveguide. (b) Near-field patterns of TM and TE modes at 1.5 μm. (c) The waveguide and substrate emit green fluorescence under 980 nm excitation.

Download Full Size | PDF

The near-field pattern of guided mode was captured by end-fire coupling method. Polarized light from a 1547 nm single-frequency laser was launched into one strip waveguide via endface coupling between a polarization-maintaining single-mode fiber and one strip waveguide. The magnified near-field image of guided mode was projected onto an infrared camera through a 40 × microscope objective lens. A polarization controller was used to control the state of polarization of the light launched into the waveguide. The results show that the 8-μm-wide strip waveguide supports transverse magnetic (TM) or transverse electric (TE) single-mode propagation at the 1.5 μm wavelength. Figure 1(b) shows the TM and TE mode patterns. For convenience, we consider an x-y Cartesian reference frame fixed at the center of the waveguide surface with x axis along the width direction and y axis pointing to the depth direction. Analysis shows that the light intensity of either the TE- or TM-mode follows the Gauss profile A_xexp[-2(x/W_x)²] in the x direction and Hermite-Gauss profile A_yy²exp[-2(y/W_y)²] in the y direction. The mode size W_x × W_y is 4.9 × 4.3 (4.5 × 3.9) μm² for the TM (TE) mode. The waveguide shows slight polarization dependence and the TM mode size is larger. This is expected for an X-cut/Y-propagation Ti:LN waveguide because Δn_e is larger than Δn_o.

The waveguide loss was evaluated from the insertion loss measured at λ = 1547 nm. It includes Er³⁺ ground state absorption, waveguide loss, Fresnel reflection loss and coupling loss between the waveguide and the coupling fiber. The fiber-to-fiber insertion loss of the 16 mm long, 8-μm-wide waveguide was measured to be 9.7/8.8 dB for the TM/TE polarization. From the known mode sizes, the coupling loss between the 8-μm-wide waveguide and the single-mode fiber (with a mode field diameter 10.3 μm) can be determined as 1.9/2.3 dB for the TM/TE mode at each side of the waveguide. With ignored etalon effect, the Fresnel reflection loss is evaluated from the refractive index values at the 1.5 μm wavelength. It has a value of 0.3 dB in total at two sides of the waveguide for either the TE or TM mode. The Er³⁺ ground state absorption loss at the 1547 nm wavelength is evaluated from the known Er³⁺ absorption cross section values and Er³⁺ concentration profile given below. The Er³⁺ absorption cross section under the TM/TE polarization, named σ^TM(λ)/σ^TE(λ), has a value of (1.0/0.6 ± 0.05) × 10⁻²⁰ cm² at the 1547 nm wavelength [16]. The Er³⁺ absorption loss, named Loss_abs, is given by Loss_abs = 4.343αL [dB], where L is the waveguide length, α is the waveguide-depth-dependent Er³⁺ absorption coefficient in unit of cm⁻¹ and is given by α(y) = σⁱ(λ)C_Er(y), where i = TE or TM, and C_Er(y) is the Er³⁺ concentration depth profile given below. Because the C_Er varies with y, here we consider its average over the depth range in which the guided light is effectively confined. Evaluation shows that the Er³⁺ absorption loss is 5.2/3.2 dB for the TM/TE mode over the 16 mm long waveguide. The waveguide loss is then evaluated as 0.3/0.5 dB/cm for the TM/TE mode. It is slightly larger than that of Ti:Er:LN waveguide, which can be as low as 0.1 dB/cm [17]. With optimized waveguide fabrication parameters, the loss figure can be lowered.

The surface and/or depth profiles of Er, Zr, Ti, Nb and O ions were analyzed by a time-of-flight secondary ion mass spectrometry (ToF SIMS V, ION-TOF GmbH). Figure 2(a) shows the Ti⁴⁺ profile (red balls) on the waveguide surface. The surface Ti⁴⁺ profile can be fitted by a sum of two error functions (see the blue plot). The fitting expression and parameter values are indicated (W is the initial Ti-strip width). The diffusion width d_x is 5.4 ± 0.1 µm, yielding a lateral Ti⁴⁺ diffusivity 0.97 ± 0.04 μm²/h. Figure 2(b) shows the depth profiles of Nb, O, Er, Zr and Ti ions in the waveguide. The Ti, Zr and Er profiles can be fitted by a Gauss function. The fitting expressions are indicated. The Ti, Zr and Er concentration profiles can be written as C_j(y) = C⁰_jexp[-(y/d_j)²], where j = Ti, Zr or Er; C⁰_j is the surface concentration and d_j is the 1/e diffusion depth of the ion j that has a value of d_j = 3.8, 9.5 and 6.2 ± 0.1 μm for Ti, Zr and Er, respectively, which yields a bulk diffusivity of 0.48 ± 0.02, 0.38 ± 0.01 and 0.064 ± 0.002 μm²/h in order. From the law of mass conservation, C⁰_j is evaluated as (24.5 ± 1.0), (4.5 ± 0.1) and (1.0 ± 0.1) × 10²⁰ ions/cm³ for Ti, Zr and Er, respectively.

 

Fig. 2 (a) Surface Ti⁴⁺ profile, (b) depth profiles of Nb, O, Er, Zr and Ti.

Download Full Size | PDF

Figure 3(a) shows the 1.5 μm amplified spontaneous emission (ASE) spectra of the Ti:Zr:Er:LN waveguide pumped at 980 and 1480 nm [Fig. 1(c) shows the waveguide and substrate emit green fluorescence under the 980 nm excitation]. As the ASE is actually the amplified fluorescence of Er³⁺ in a waveguide, a comparison of the ASE spectrum with the typical emission spectrum of Er³⁺ ions in an LN crystal may help us to judge the crystalline phase of Er³⁺ ions present in the waveguide studied. Comparison shows that both are similar, indicating that the presence of the Er³⁺ ions in the waveguide is in the form of LN crystalline phase. In addition, the 1530 nm fluorescence decay was measured under the 980 nm excitation. The inset in Fig. 3(a) shows the measured (red balls) decay of the 1530 nm fluorescence for the Er³⁺ ions in the Zr:Er:LN substrate. The green plot denotes the mono-exponential fit. The measured and fitted decays overlap well with a lifetime τ_f = 2.8 ± 0.1 ms, which is similar to the value of bulk Er³⁺-doped LN [18]. It is concluded in combination with the ASE spectra that Zr-codoping does not affect the 1.5 μm spectroscopic property of Er³⁺.

 

Fig. 3 (a) ASE spectrum of Ti:Zr:Er:LN strip waveguide pumped at 980 and 1480 nm. Inset shows the measured and fitted decay of the 1.5 μm fluorescence under the 980 nm excitation. (b). Pump power dependence of net gain of Ti:Er:LN and Ti:Zr:Er:LN strip waveguides.

Download Full Size | PDF

Next, we discuss the optical damage resistant issue of the waveguide. To obtain an optical damage resistant waveguide, waveguide layer must be entirely covered by Zr⁴⁺ profile and the C_Zr at the 1/e waveguide depth should be above the threshold C_th ( = 2.0 mol%). For the waveguide studied here, d_Zr ( = 9.5 μm) is much larger than d_Ti ( = 3.8 μm), i. e., the Ti⁴⁺ profile is entirely covered by the Zr⁴⁺ profile. The C_Zr at the surface and at the 1/e Ti⁴⁺ concentration depth are 2.4 mol% and 0.9 mol%, respectively, implying that the waveguide should have a certain resistance to optical damage. The resistance was verified by examining the stability of the 1547 nm small-signal output under high 980 nm pump level. Figure 3(b) shows the net gain as a function of 980 (red balls) or 1480 nm (blue balls) coupled pump power. For comparison, the result of a usual 1.8 cm long Ti:Er:LN waveguide without Zr⁴⁺ codoping is also given (black balls). The net gain is obtained from the measured signal enhancement [19]. For the usual Ti:Er:LN waveguide, the signal drops significantly as long as the 980 nm pump power exceeds 40 mW due to the serious optical damage. For the Zr⁴⁺-doped waveguide, however, the signal is stable up to the available coupled 980 nm pump power of 120 mW. The gain increases stably with the pump power. The stable output means that the photorefractive effect is effectively suppressed. As the optical damage is relatively weak at λ = 1480 nm, stable output is observed in the case of 1480 nm pumping too.

Present method of suppressing the photorefractive effect is superior to others reported previously, such as 1) device operation at >90 °C [20], which increases complexity of experimental setup, 2) light propagation along the optical axis of LN [21], which fails to use the larger electro-optic coefficient r₃₃, and 3) use of ZnO-diffused waveguide codoped with >4.9 mol% MgO [22], for which heavy MgO doping causes material inhomogeneity and difficulty in growing high-quality single-crystal in the case of codoping with rare earth ions.

Finally, attention is paid to the amplification characteristic of the studied waveguide. We note from Fig. 3(b) that the gain in the case of 980 nm pumping is 1.1 dB/cm for the available coupled pump power 120 mW. Due to the smaller absorption coefficient at the 1480 nm, the maximum net gain in the case of 1480 nm pumping is 0.2 dB/cm. The smaller gain is related to the following factors. 1). The gain does not saturate yet at the 120 mW pump power. 2). The signal wavelength considered here, 1547 nm, is away from the 1530 nm, where the gain peaks. 3). The Er³⁺ has a lower concentration at waveguide surface, only ~0.5 mol%. 4). The waveguide has a larger loss. With optimized fabrication parameters, a higher gain is expected.

4. Summary

We have fabricated Ti:Zr:Er:LN strip waveguide by sequential diffusion of Er³⁺, Zr⁴⁺ and Ti⁴⁺. The waveguide is optical damage resistant, supports TE and TM single-mode propagation at 1.5 μm, shows slight polarization dependence, and has a loss of 0.5/0.3 dB/cm for the TE/TM mode. A net gain of 1.1/0.2 dB/cm is obtained for available 980/1480 nm pump power of 120 mW. With optimized fabrication condition, higher gain and resistance to optical damage are expected. The waveguide is then promising for active integrated optics.