Property optimization of Er/Yb-codoped LiNbO₃ crystals for LNOI lasers and amplifiers

Lithium niobate (LiNbO₃, LN) is regarded as one of the most promising material platforms for integrated photonics due to its excellent properties.^1–3 In recent years, with the breakthrough in micro–nano fabrication technology, revolutionary performance is expected in the form of lithium-niobate-on-insulator (LNOI) optical communication, as integrated photonic devices based on LNOI have been successfully demonstrated, such as high-performance electro-optic modulators,⁴ high-efficiency nonlinear optical frequency converters,⁵ and broadband frequency comb sources.⁶ Moreover, active devices based on rare-earth (RE) doped LNOI have garnered significant attention as an essential part of integrated photonics platforms.⁷ On-chip microcavity lasers⁸ and waveguide amplifiers⁹ based on Er-doped LNOI (Er:LNOI) have been successively reported, but the conversion efficiency of the Er:LNOI microcavity laser is still very low.¹⁰ 

One principal drawback of Er³⁺ as an optically active element is its low absorption band at 980 nm. A proven solution is to codope with Yb³⁺ since it has a large absorption cross section of around 980 nm and can enhance the pump efficiency through the energy transfer between Yb³⁺ and Er³⁺ ions.^11–13 On-chip microdisk lasers¹⁴ and waveguide amplifiers^15,16 based on Er/Yb-codoped LNOI have successfully demonstrated their superiority. However, due to the complexity of the mechanisms involved, the details of Yb³⁺ sensitization in Er:LNOI have not been clearly established. Furthermore, the presence of concentration saturation and the challenges of growing high-quality crystals with high doping levels are important considerations for practical applications.^17–19 Therefore, systematic experimental characterization is needed to provide significant guidance for the development of high-performance LNOI active devices.

In this paper, a series of Er/Yb-codoped LN crystals were grown along the c-axis by the Czochralski method. The structures of grown crystals were characterized by diffractometers. The absorption and emission characteristics in the near-infrared range of Er³⁺ and Yb³⁺ were analyzed by spectrometers. The transfer efficiencies between Yb³⁺ and Er³⁺ were calculated through the rate equations. In addition, the relative conversion efficiencies of different crystals were compared by varying the pump power levels. Overall, the influence of Yb³⁺ sensitization was discussed systematically, and the advantages of codoping were clarified.

Er/Yb-codoped LN single crystals were grown along the c-axis in the air by the Czochralski method using an intermediate frequency heater furnace, which can be divided into two series. One is LN crystals with a fixed Er³⁺ concentration (0.5 mol. %) and various Yb³⁺ concentrations (0, 0.5, 1.0, 1.5, and 2.0 mol. %), labeled as Er_0.5, Er_0.5Yb_0.5, Er_0.5Yb_1.0, Er_0.5Yb_1.5, and Er_0.5Yb_2.0, respectively. The other is LN crystals with a fixed RE³⁺ (RE³⁺ = Er³⁺ + Yb³⁺) concentration of 2.0 mol. %, which were labeled as Er_0.5Yb_1.5 (same as above), Er_1.0Yb_1.0 (1.0 mol. % Er³⁺ and 1.0 mol. % Yb³⁺), Er_1.5Yb_0.5 (1.5 mol. % Er³⁺ and 0.5 mol. % Yb³⁺), and Er_2.0 (2.0 mol. % Er³⁺ and 0 mol. % Yb³⁺), respectively. The pulling speed was 0.6 mm/h, and the rotation rate was 7 rpm during the crystal growth process.

The starting materials used for crystal growth were Li₂CO₃, Nb₂O₅, Er₂O₃, and Yb₂O₃ with a purity of 99.99%. All raw materials were mixed uniformly in a planetary ball mill and then sintered at 1100 °C to form polycrystalline blocks. After crystal growth, they were polarized at 1200 °C, then cut into z-cut thin wafers from the top and polished to optical grade.

The crystal structures were analyzed by using a Bruker-D8 ADVANCE diffractometer and a Rigaku SmartLab high-resolution diffractometer with copper targets. The near-infrared absorption spectra were measured using a Hitachi U-4100 spectrometer, and the photoluminescence spectra of these crystals were characterized using an FLS1000 photoluminescence spectrometer under 980 nm excitation.

The powder XRD patterns of these doped LN crystals follow from Fig. 1. Compared with the standard PDF card (PDF No. 20-0631) of LN, all the diffraction peaks can be indexed to the known phase, and no new peaks appear, indicating the phase purity of the LN products. Consequently, Er³⁺ and Yb³⁺ ions have no influence on the structures of LN crystals by occupying the normal Li-site and Nb-site rather than the interstitial sites within the lattice. Meanwhile, there are variations in the relative intensities of these diffraction peaks, primarily due to the lattice distortion caused by the larger radii of Er³⁺ and Yb³⁺ ions.²⁰ The rocking curves of these crystals were measured, and the full widths at half maximum (FWHM) are around 20″–50″, indicating the good quality of as grown crystals. As an example, the rocking curve of Er_1.0Yb_1.0 crystal is shown in Fig. 2, and the inset displays the crystal picture, which is pink and contains no macroscopic defects, while the FWHM is about 29″.

The near-infrared unpolarized absorption spectra of Er/Yb-codoped LN crystals after reflection correction²¹ are depicted in Fig. 3(a), containing two absorption bands at 980 and 1530 nm, respectively. Compared with the even small absorption region at 980 nm for the singly Er-doped crystal Er_0.5, codoping with Yb³⁺ ions in LN leads to a much broader and more intense excitation band, which corresponds to the transition from the ground state ²F_7/2 to the excited state ²F_5/2 of Yb³⁺ ions. The fact, together with the efficient energy transfer from Yb³⁺ to Er³⁺ ions, forms the basis of Yb³⁺ sensitization in Er/Yb-codoped LN crystals, adequate to improve the pumping efficiency in the spectral range suitable for diode laser pumping.¹⁷ The 1530 nm absorption band is attributed to the ⁴I_15/2 → ⁴I_13/2 transition of Er³⁺ only. Due to the radiation trapping effect,²² a weaker absorption band is ideal. A weak absorption coefficient means a low level of active ions, which is disadvantageous for the emission process. Codoping with Yb³⁺ in Er-doped LN can also reconcile the contradiction.

Figure 3(b) illustrates the maximum absorption coefficients at 980 and 1530 nm bands of codoped LN crystals with various Yb³⁺ concentrations. The absorption around 980 nm can be considered to rely on Yb³⁺ primarily due to the weak contribution from Er³⁺. It is apparent that the value linearly depends on the Yb³⁺ concentration, indicating the content of Yb³⁺ in codoped LN increases linearly. However, the absorption coefficient around 1530 nm decreases slightly with the increase in Yb³⁺ content, which shows the doping level of Er³⁺ in LN is lowered. That is, the incorporation of Yb³⁺ will affect the segregation of Er³⁺. Actually, the two ions seem to be treated as one class of dopants due to the similarities of lattice position occupations and proprieties.^17,23

The near-infrared emission spectral features of Er/Yb-codoped LN crystals under 980 nm excitation follow from Fig. 4(a). The peak near 1060 nm on the left corresponds to Yb³⁺ (²F_5/2 → ²F_7/2), and the peak around 1530 nm on the right is associated with Er³⁺ (⁴I_13/2 → ⁴I_15/2). Except for the Er_0.5 crystal, the emission intensity of Yb³⁺ is far lower than that of Er³⁺, although the concentration of Yb³⁺ is the same or even several times larger than that of Er³⁺ in codoped LN crystals. Figure 4(b) displays the relationships among the maximum emission intensities around 1060 and 1530 nm of codoped LN crystals and the Yb³⁺ concentration. Unlike the linear increase in Yb³⁺ emission, the photoluminescence intensity of Er³⁺ approaches saturation when Yb³⁺ is larger than 0.5 mol. %. With a continuous increase in Yb³⁺ content, the emission intensity shows only a slight rise.

Compared with other codoped materials that can achieve a [Yb]/[Er] ratio up to 10 or even higher for greater gains,^24–26 the emission intensity around 1530 nm of Er/Yb-codoped LN only slightly increases when the [Yb]/[Er] ratio is more than 1.0, which is about 60% larger than the singly Er-doped LN. The results seem to indicate that the effect of Yb³⁺ sensitization in Er/Yb-codoped LN crystal is not good enough. Actually, there are several aspects to take into account. On the one hand, both Er³⁺ and Yb³⁺ ions enter the LN lattice and occupy the Li-site preferentially, competing with each other. Meanwhile, the limitation of solid solubility in LN restricts the total doping level.^17–19 On the other hand, the upconversion process will become stronger with increasing Yb³⁺ concentration, which is a disadvantage for the emission around 1530 nm.²⁷ The complex upconversion process is not considered in this paper, and there are several results available for reference that clarify that the upconversion process in Er/Yb-codoped LN can be adjusted by codoping with optical damage resistant additives.^28–30 

Figure 5 depicts a simple schematic of the energy transfer process between Yb³⁺ and Er³⁺ ions without upconversion. Both Er³⁺ and Yb³⁺ ions can be excited under the pump at 980 nm. The upper level ²F_5/2 of Yb³⁺ can either decay to the ground state ²F_7/2 or transfer its energy to the excited state ⁴I_11/2 of Er³⁺. Within the Er³⁺ ion, a rapid nonradiative transition between the ⁴I_11/2 and ⁴I_13/2 leads to a population build-up in the metastable ⁴I_13/2 level, generating the emission around 1530 nm by radiative transition to the ground state ⁴I_15/2.

It is apparent that the transfer efficiency decreases with increasing Yb³⁺ concentration. The calculated values are smaller than the experimental results independently estimated from the Yb^{3+ 4}F_5/2 decay rate,¹⁵ primarily due to the neglect of the upconversion process. Considering the limited emission intensity improvement and the difficulty of crystal growth for highly doping levels, it is not an ideal strategy for the Yb³⁺ concentration beyond 0.5 mol. % when codoped with 0.5 mol. % Er³⁺ in LN. Therefore, Er/Yb-codoped LN crystals may have better performance when the ratio of [Yb]/[Er] reaches about 1.0.

In general, the conventional Stokes luminescence component of about 1530 nm of Er³⁺ under 980 nm excitation is linearly dependent on the pump power for an unsaturated photoluminescence process.³³ Consequently, the slope magnitude of the emission intensity with respect to the pump power can reflect the relative conversion efficiency under the identical test conditions. Figure 7 registers the emission intensities of Er³⁺ in the doped LN crystals with various Yb³⁺ concentrations as functions of pump power levels at 980 nm. It is observed that the linear relationship between the emission intensity of each crystal and the pump power is well established. The relative conversion efficiencies of codoped LN with various Yb³⁺ concentrations are similar, increasing by about 55% compared with Er_0.5. Although the value of Er_0.5Yb_0.5 is a little bit lower than other codoped crystals, the total doping level is the lowest, which is beneficial for growing high-quality LN crystals. In addition, the absence of emission at low pump power suggests that the necessary population inversion of Er³⁺ in LN crystals can be achieved only when the energy reaches a certain level due to various losses.

Considering the difficulty of crystal growth for high-quality with large size and the solid solubility limitations of Er³⁺ and Yb³⁺ ions in LN crystals, it is crucial to properly allocate the doped ions to obtain the optimal optical properties, which will present significant guidance for the practical application. Therefore, the absorption and emission spectra of Er/Yb-codoped LN crystals with a fixed RE³⁺ concentration of 2.0 mol. % were tested under the same conditions. The relationships among the emission intensities and pump power levels are shown in Fig. 8. We have identified that the optimal doping concentration in singly Er-doped LN is about 2.0 mol. %.¹⁹ It is obvious that the relative conversion efficiencies of Er_1.0Yb_1.0 and Er_1.5Yb_0.5 crystals are similar to those of Er_2.0. The transfer efficiencies of Er_1.0Yb_1.0 and Er_1.5Yb_0.5 are also calculated as 25% and 12% (the absorption and emission spectra of singly Er-doped LN with concentrations of 1.0 and 1.5 mol. %, which are labeled as Er_1.0 and Er_1.5, are shown in Figs. S1 and S2 in the supplementary material), respectively, indicating that the Er_1.0Yb_1.0 crystal would have better performance. Moreover, the values of transfer efficiencies in both Er_1.0Yb_1.0 and Er_0.5Yb_0.5 are similar, confirming that optimal Yb³⁺ sensitization will occur in codoped LN when the ratio of [Yb]/[Er] reaches about 1.0.

To further investigate the properties of Er_1.0Yb_1.0 and Er_2.0 crystals, their absorption and emission spectra are compared, as shown in Fig. 9. It is observed that the absorption coefficient of Er_1.0Yb_1.0 near 980 nm is about twice that of Er_2.0 crystal, predicting a larger absorption cross section and a lower threshold power level, which are more favorable for the laser generation and amplification. In contrast, the absorption coefficient around 1530 nm for the Er_1.0Yb_1.0 is about half of that for the Er_2.0 crystal. On the one hand, lower absorption near 1530 nm can reduce the effect of radiation trapping on photoluminescence. There is a little distortion of the emission spectrum, allowing a more accurate performance prediction. On the other hand, limited by the technology of integration, the most common approach is to integrate active [optical pumped, Er(/Yb)-doped] and passive [unpumped, Er(/Yb)-doped] devices on the same substrate. Lower absorption means less transmission loss, which can significantly improve the whole performance of the integrated devices.

From Fig. 9(b), it is found that the emission peaks around 1530 nm of Er_1.0Yb_1.0 and Er_2.0 crystals basically overlap with each other. Considering the reduced transmission loss and lower threshold power, on-chip active devices based on Er_1.0Yb_1.0 will have better performance than those based on Er_2.0 crystal. In particular, for amplifiers, the weak absorption around 1530 nm will result in a large net gain. These conclusions demonstrate the superiority of Er/Yb-codoped LN again. In the longer wavelength range of 1550 nm, although the emission intensity of Er_1.0Yb_1.0 is a little smaller than that of Er_2.0 crystal, it is still an alternative working band.

In summary, a series of Er/Yb-codoped LN crystals with good quality were successfully grown by the Czochralski method. The absorption and emission spectra were analyzed in detail. The transfer efficiency and relative conversion efficiency were also displayed. It was concluded that the codoped LN exhibits better performance when it contains similar concentrations of Er³⁺ and Yb³⁺. Considering the challenges of growing high-quality crystals and the solid solubility limitation of RE³⁺, the optimal composition of codoped LN is ∼1.0 mol. % Er³⁺ and 1.0 mol. % Yb³⁺. This composition enables active devices to achieve lower threshold power, reduced transmission loss, higher transfer efficiency, and improved conversion efficiency. In particular, amplifiers will exhibit a remarkable gain effect due to the strong emission and weak absorption in the operating band around 1530 nm. Our results provided significant guidance for the application of Er/Yb-codoped LN crystals, especially for the development of integrated photonics based on LNOI.

See the supplementary material for the following: Fig. S1(a) shows the absorption spectra of Er1.0 and Er1.0Yb1.0, Fig. S1(b) shows the emission spectra of Er1.0 and Er1.0Yb1.0 under 980 nm excitation, Fig. S2(a) shows the absorption spectra of Er1.5 and Er1.5Yb0.5, and Fig. S2(b) shows the emission spectra of Er1.5 and Er1.5Yb0.5 under 980 nm excitation.

This work was supported by the National Natural Science Foundation of China (Grant No. 12034010), the Natural Science Foundation of Tianjin City (Grant Nos. 23JCZDJC00780 and 21JCZDJC00300), and the Program for Changjiang Scholars and Innovative Research Team in University (Grant No. IRT_13R29).

The authors have no conflicts to disclose.

Zhongzheng Zhang: Data curation (lead); Investigation (lead); Visualization (lead); Writing – original draft (lead). Yuqi Zhang: Data curation (equal); Investigation (equal). Zhaojie Wu: Data curation (equal); Investigation (equal). Xin Yuan: Investigation (equal). Shiguo Liu: Resources (lead). Hongde Liu: Supervision (equal); Writing – review & editing (equal). Dahuai Zheng: Supervision (equal); Writing – review & editing (equal). Yongfa Kong: Funding acquisition (lead); Supervision (lead); Writing – review & editing (lead). Jingjun Xu: Supervision (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.