1. Introduction

High-power Yb-doped fiber (YDF) lasers are commonly used in scientific and industrial applications [1–3], where stability and reliability are very important. Over time, the photodarkening (PD) [4], i.e., the background loss of the laser fiber, increases, and the laser power decreases. This is a critical issue for the lifetime and modal stability of high-power fiber lasers [5–7]. PD has been investigated extensively in the last decade, but the physical origin of the pump-induced losses remains a subject of discussion [8–12].

PD is generally associated with dopant-related ultraviolet (UV) absorption bands and the generation of high-energy UV photons [8,9]. However, the origin of the UV absorption band observed in Yb³⁺/Al³⁺-doped fiber preform is controversial. Yoo attributed [9] the strong UV absorption band at 230 nm to the O vacancy center (ODC), while Engholm [8] considered that the UV absorption band originates from the charge-transfer (CT) transition. Although the assignment of the UV absorption band is controversial, the formation of Yb²⁺ [13–16] and aluminum-oxide hole centers (Al–OHC) [17] in the PD process is widely accepted. Considering that the energy (~5 eV) of the UV absorption band is significantly higher than that of the near-infrared (NIR) pump photons (~1.27 eV), the generation of high-energy UV photons via Yb³⁺ ion clusters is suggested [18,19]. Clusters can lead to cooperative up-conversion, which not only creates channels for energy loss but also increases the probability of producing high-energy photons. The energy needed for color-center formation can be obtained from a cooperative effect involving simultaneous excitation of ions in a cluster [20,21]. Currently, modified chemical vapor deposition (MCVD) combined with solution doping, which has the merits of high purity and ultralow optical loss [22], has been a standard method of laser fiber preform fabrication for industrial applications. However, it has limitations, such as imprecise control of the refractive index and cluster formation under high rare-earth doping [23,24], which may be not conducive to inhibiting PD.

It has been confirmed that the PD resistance of YDF can be improved via doping with P and/or Al, owing to the dissolution of clusters [18,25]. P doping has a greater effect on suppressing the PD because of the higher efficiency of P compared with Al in dissolving Yb³⁺ clusters [18]. However, excessive Al³⁺ increases the number of Al–OHC [26], which is another source of the PD of YDF [17], and co-doped P reduces the absorption and emission cross sections of Yb³⁺ ions [27]. Additionally, Ce doping can improve the PD resistance [28,29] but is not convenient for the regulation of the numerical aperture (NA). Hence, tradeoffs are often required for improving the PD resistance of the laser fiber.

To our knowledge, because of the possibility of controlling the composition and the homogeneity, the sol–gel process, which has outstanding dispersion homogeneity of dopant ions, is widely used to prepare amorphous or crystalline oxide [30,31]. Furthermore, because different species of chemicals can be mixed on the molecular level, the sol-gel method has been recently used to prepare rare earth-doped silica glass fibers [32] and large-mode area photonic crystal fibers [33–35]. Owing to the superiority of the distribution of the dopants, the sol-gel method is a possible solution for preparing PD resistance fibers.

Concerning the aforementioned issue, the major objective of this study is to investigate the PD resistance of a YDFs prepared via the sol-gel method by comparing its PD performance with a YDFs prepared via MCVD with conventional solution doping. The mechanism of the PD-resistance enhancement caused by the sol-gel preparation method was studied using cooperative luminescence spectra of Yb³⁺ ions, electron paramagnetic resonance (EPR) spectroscopy, and electron probe microanalysis (EPMA).

2. Experimental

Two kinds of YDFs (named #M and #S) were prepared. #S was prepared via the sol-gel method combined with high-temperature sintering. Tetraethoxysilane, C₂H₅OH, AlCl₃•6H2O, and YbCl₃•6H₂O were used as precursors. Deionized water was added to sustain the hydrolysis reaction. All analytic reagents were mixed and thoroughly stirred at 30 °C to form an Al³⁺/Yb³⁺ homogeneously doped solution. The solution was heated from 70 to 1,100 °C under an oxygen atmosphere to produce dried powders, in which the organics were almost decomposed. The powders were melted in the vacuum state at 1,750 °C to form glass. Then, the glass was molded into a glass rod in a silica tube (F300) from Heraeus on an oxy-hydrogen flame to obtain a preform. After that, the fiber was drawn from the preform by rod in tube method at 2100 °C on the optical fiber draw tower, which is provided by Optogear OG-510. The organic coating, i.e. acrylate, was applied in the drawing process, and the coating was cured by UV-light. The detailed preparation process is presented in [36]. #M was prepared via MCVD with solution-doping processes. The following step-by-step process was performed to prepare the Yb³⁺-doped preform. First, a SiO₂ soot layer was deposited on the inside of a fused silica tube (F300). Then, the soot layer was soaked in an alcohol solution of YbCl₃•6H₂O and AlCl₃•6H₂O. The soot was vitrified, and the tube finally collapsed into a solid rod around 2,000 °C to obtain a preform. Finally, the preform was drawn into fiber by rod in tube method at 2100 °C, the drawing process is same as that of #S.

The #M and #S YDFs had nearly the same amounts of Yb₂O₃ and Al₂O₃. The Yb³⁺ and Al³⁺ contents were determined via inductively coupled plasma analysis. For the characterization of the PD behavior, fibers were drawn with core and inner cladding diameters of 10 and 200 μm, respectively, and NA is 0.08. Commercial Yb-doped silica fiber from Nufern (LMA-YDF-10/130-M, named #N) was used for comparison. The main characteristics of the three kinds of YDFs are listed in Table 1. They have the same core diameter of 10 μm. The core absorption coefficient of the three fibers at 974 nm was determined via the cut-back method. It is nearly the same for the three fibers, as shown in Table 1. Therefore, we assume that the #N fiber has a similar Yb³⁺ ion concentration to the #M and #S fibers. This excludes the effect of the Yb³⁺ doping concentration on the PD.

Table 1. Characteristics of investigated YDFs from different synthesis methods

View Table | View all tables in this article

Figure 1(a) is the schematic of the PD measurement. The induced loss at 633 nm was measured in situ during long-term Yb³⁺ excitation by 974 nm core pumping. To accelerate the PD, the inversion of fiber is saturated, at approximately 50%. The 974 nm pump light and 633 nm probe light were launched into the core through a custom 633/974 nm wavelength-division multiplexer (WDM) from the opposite direction. The short-pass filter at the end of the fiber, provided by THORLABS FESH0750 with cut-off wavelength of 750 nm, is mainly used to eliminate the residual pump light and luminescence to ensure reliability of the power-meter records.

 

Fig. 1 (a) Schematic of the PD measurement. (b) Experimental setup for core-pumped fiber lasers.

Download Full Size | PDF

The repeatability of the Yb³⁺ ion inversion was found to depend on the repeatability of the pump wavelength, assuming that the pump power is high enough to saturate the inversion of Yb³⁺ ions. To avoid the effect of temperature on the PD test, the experiment was performed in a thermostatic laboratory at 23 °C, and the temperature of the fiber was monitored by a temperature detector and found to remain at 22–23 °C throughout the experiment. The PD test shows that the wavelength offset of the laser diode (LD) pump laser is <0.1 nm at 22–23 °C (room temperature). The long-term power fluctuation of the probe laser power is <2%. The standard stretched exponential function is used to fit the measured data [37], as shown in Eq. (1).

The degree of PD can be characterized by the fitting parameters, i.e., the equilibrium loss α_eq, time constant τ, and stretching parameter β. To eliminate the error of the testing results due to the bleaching effect of the probe light [38], the power of 633 nm laser was fixed at 10 uW [39].

The experimental setup for the core-pumped fiber lasers is shown in Fig. 1(b). The laser cavity was built with two fiber Bragg gratings having 99.9% and 19.9% peak reflectivity at 1,035 nm spliced before and after the tested fiber, respectively. A WDM as a pump isolator connected to a 974 nm laser diode was used to filter out the reverse light.

Because the absorption of the YDFs at wavelengths below 400 nm was extremely high, we had to use preform transverse slices instead of long-fiber samples for absorption-spectra measurements, and the slices are polished to 1 mm thickness. An absorption spectrum with a scanning step of 1 nm was recorded using a Lambda 950 UV–visible (VIS)–NIR spectrophotometer in the range of 200 to 1,100 nm. The cooperative luminescence spectra of Yb³⁺ ions were obtained using an FLSP920 spectrofluorometer (Edinburg Co., UK) with 976 nm pumping, and the emission spectra of Yb²⁺ ions was obtained by a Xe lamp under excitation of 330 nm and the detector is visible photomultiplier tube (PMT). The refractive index difference profiles of all fibers were tested using an Interfiber Analysis LLC, IFA-100 instrument at 633 nm. All of these tests were performed at room temperature.

EPR was implemented to identify Yb³⁺ clusters in the fiber. Approximately 100-mg-weight core glass was stripped out from the two preforms (#S and #M), and it was ground into powder for EPR tests. The two-pulsed (π/2 – τ–π – τ echo) echo field swept spectra were recorded at 4 K while the magnetic field was swept from 0 to 1,100 mT, using an E-580 BRUKER ELEXSYS EPR spectrometer operating at the X-band (9.25 GHz). The π/2 and π pulse lengths were 16 and 32 ns, respectively, with a time delay of τ = 136 ns between pulses. The Yb³⁺ distribution in the #M and #S fibers in the radial direction was characterized using EPMA (Shimadzu, 1720H).

3. Results and discussion

Figure. 2(a) shows the evolution of the induced loss at 633 nm for the three kinds of YDF fibers under the same testing conditions. The induced loss increases slowly for the #S fiber. Among the three fibers, #M exhibits the most rapid increase of the induced loss. ${\alpha }_{eq}$ was calculated by applying Eq. (1) to the tested curves; it was 87, 142, and 276 dB/m for the #S, #N, and #M fibers, respectively. This means that the PD is most serious in the #M fiber. The #S fiber has best PD-resistance behavior among the three kinds of fibers.

 

Fig. 2 (a) Evolution of the induced loss at 633 nm. (b) Output laser power versus the absorbed power for the two YDF fibers prepared via MCVD and the sol-gel method, as well as the #N fiber. Inset: the geometry of the fiber cross section. The length of the fiber test is 15cm.

Download Full Size | PDF

Figure. 2(b) shows the slope efficiency and threshold power before and after the PD of the #M, #S, and #N fibers. The corresponding results are shown in Table 2. Here, the fiber length is not optimized for the laser performance test. On the other hand, the grating wavelength (1,035 nm) selected may be not the best gain wavelength for the fiber; thus, the slope efficiency of the three pristine YDF fibers is <60%. Besides, due to the higher loss for fiber #S (0.3dB/m), the efficiency of fiber #S is lower than fiber #N, and the threshold of fiber #S is higher than fiber #N. To control the testing error, the same fiber was used to test the slope efficiency before and after the PD testing, and the slope efficiency variation range is within 2% after 3 times measurements. As clearly indicated by Fig. 2(b) and Table 2, after 65 h of PD, the #S fiber has the lowest reduction ratio of the slope efficiency among the three kinds of fibers. Additionally, its laser threshold increase ratio is the smallest. This indicates that the PD resistance of fiber #S prepared via the sol-gel method is better than those of fiber #M prepared via the MCVD method and the commercial fiber #N.

Table 2. Comparison of laser parameters of three kinds of YDFs before and after PD

View Table | View all tables in this article

UV–VIS–NIR absorption and emission spectroscopy, as well as EPMA and EPR analysis, was performed to explore the mechanism of the PD resistance of the #S fiber. Owing to the small core of the fiber, the fluorescence was very weak. The test was conducted using preform transverse slices of #S and #M fibers. No test was done on the preform of the #N fiber, because it is not available.

The UV–VIS–NIR absorption and VIS emission spectra obtained under excitation at 330 nm (λ_ex = 330 nm) for preform slices #M and #S are shown in Figs. 3(a) and 3(b), respectively. Excitation spectrum (λ_em = 525 nm) for #S sample is presented in the inset of Fig. 3(b), four excitation peaks located at 250, 300, 330 and 400 nm are observed, among which the strongest peak at 330 nm correspond to the 3.75 eV absorption band in Fig. 3(a). The absorption of an Al³⁺ single-doped silica preform slice (named #Al:SiO₂) prepared via the sol-gel method is also shown in Fig. 3(a), for reference. In Fig. 3(a), no obvious difference between #M and #S is observed in the range of 850 to 1,100 nm. However, there are obvious differences among the #Al:SiO₂, #M, and #S samples in the UV band (200–400 nm). For #Al:SiO₂, the UV absorption is very weak. For #S, a strong UV absorption band at 5.2 eV (236 nm) and a weak absorption band at 3.75 eV (330 nm) are observed in Fig. 3(a). For sample #M, a wide absorption band in the range of 200 to 300 nm is observed, but its absorbance is weak. It is evident that a strong UV absorption band is associated with the presence of Yb ions.

 

Fig. 3 (a) Optical absorption spectroscopy. Inset: an enlargement of the absorption spectroscopy in 300 to 500 nm region. (b) Photoluminescence emission spectra (λ_ex = 330 nm) for preform slices #M and #S. Inset: excitation spectrum (λ_em = 525 nm) for preform slice #S.

Download Full Size | PDF

Yoo attributed [9] the strong UV absorption band at 5.2 eV to the Yb-related ODC. Nd³⁺ and Er³⁺ are similar to Yb³⁺ in their atomic radius and valence state, but neither Nd nor Er-related ODC can be observed in Nd³⁺ or Er³⁺-doped silica glasses [40]; therefore, we consider that the strong UV absorption originates primarily from the CT transition from O^- ligands to the Yb³⁺ ions, which was first reported by Engholm et al. [8]. According to early studies [41,42], the CT absorption band is strongly dependent on the type of rare-earth ions and the structure of ligands. For Yb³⁺-doped silica glasses, the positions of the CT band depend on the second coordination sphere (Al, Si, P) of Yb³⁺ ions. For the Yb-O-Al, Yb-O-Si, and Yb-O-P coordination structures, the CT band is located at 5.2, 5.8, and 6.2 eV, respectively [8,43,44]. The existence of 5.2 eV band and the absence of 5.8 eV band for #S sample suggest that most of Yb³⁺ ions in #S are surrounded by Al-rich phase via Yb-O-Al linkage. The boarder UV absorption band for #M sample hints the more complex coordination structure, i.e. coexistence of Yb-O-Al and Yb-O-Si. Furthermore, the stronger UV absorption in #S is mainly due to the superposition of CT transitions of Yb³⁺ ions and 4f¹⁴→4f¹³5d¹ transitions of Yb²⁺ ions. This result is supported by the VIS emission spectra in Fig. 3(b).

Figure. 3(b) shows the VIS emission spectra obtained under excitation at 330 nm (λ_ex = 330 nm) for samples #M and #S. It is clearly observed that there is a significant emission peak at 525 nm under 330 nm excitation in slice #S. This emission originates from 4f¹³5d¹→4f¹⁴ transitions of Yb²⁺ ions [13]. Wang had demonstrated that Yb³⁺/Al³⁺ co-doped silica glass prepared via the sol-gel method may contain a small portion of Yb²⁺ ions owing to the non-oxidizing sintering [26]. However, no emission of Yb²⁺ ions is observed for slice #M in Fig. 3(b), owing to the oxidizing sintering condition in the MCVD process. Both the absorption spectra in Fig. 3(a) and the emission spectra in Fig. 3(b) suggest that Yb²⁺ ions exist in slice #S but not in slice #M.

Figure. 4(a) shows cooperative luminescence spectra of preform slices #M and #S with 976 nm excitation. Cooperative luminescence corresponds to the simultaneous de-excitation of two Yb³⁺ ions, which results in the emission of one photon with a shorter wavelength. It has been shown that the intensity of cooperative luminescence is proportional to the distance between the Yb³⁺ ions, which can be used to probe clusters in Yb³⁺-doped glasses [45]. As shown in Fig. 4(a), three distinct peaks of the cooperative luminescence spectrum are apparent at 488, 502, and 512 nm, respectively, which result from convolution of the sharp peak at 976 nm and the relatively broader peak at 1020 nm in the near-infrared emission spectrum of isolated Yb³⁺ ion [46]. The cooperative luminescence intensity for the preform slice #S is obviously lower than that of the preform slice #M. This may indicate that the degree of Yb³⁺ ion clusters in slice #S is less than that in slice #M.

 

Fig. 4 (a) Cooperative luminescence spectra (λex = 980 nm) of preform slices #M and #S. (b) Electron spin echo-detected EPR spectra for preforms #M and #S. Inset: EPR spectra obtained with a magnetic field of 0–400 mT.

Download Full Size | PDF

EPR research on Yb³⁺-doped silica glass has shown that the two-pulsed (π/2 – τ–π – τ echo) echo field swept spectra intensity at a near-zero magnetic field correlates with the Yb³⁺ ion clusters [47]. To further compare the clustering degree of Yb³⁺ between fibers #S and #M, the electron spin echo-detected EPR spectra of preforms #M and #S were considered, as shown in Fig. 4(b). The electron spin echo-detected EPR spectra are composed of broad asymmetric lines with main peak located at 500–600 mT, the broad aspect of the spectra is correlated to the topological disorder of glass structure. The position of main peak is considered as “fingerprint” data sensitive to the local environments of the rare-earth ions [48,49]. Our pervious study [49] shows that the main peaks corresponding to the Yb-O-P and Yb-O-Al coordination spheres are located at 430 and 550 mT, respectively. Due to the measure condition in this work is identical with our previous work [49], we can speculate that the Al dominates the next nearest neighbor atom of Yb via Yb-O-Al linkage in these two samples according to the position of main peak in electron spin echo-detected EPR spectra. This result further supports that the 5.2 eV band originates from the CT transition for the situation of Yb-O-Al coordination sphere, and the boarder electron spin echo-detected EPR spectra for #M is attributed to the inhomogeneous broadening of multiple Yb site, i.e. coexistence of Yb-O-Al and Yb-O-Si.

Experiments and simulations confirm that there is no EPR line corresponding to any Yb³⁺ isotope at zero magnetic field if the glass is only doped with Yb³⁺ ions, and a reasonable explanation for the observation of an echo at zero magnetic field in the glasses is the existence of pairs or more complex clusters of Yb³⁺ ions [47]. As clearly shown in the inset of Fig. 4(b), the intensity of fiber #M is stronger than that of fiber #S at zero magnetic field, which means that fiber #S has less clusters of Yb³⁺ ions than fiber #M and that the dispersion of Yb³⁺ is better in the fiber prepared via the sol-gel method. These results are consistent with the cooperative luminescence spectra shown in Fig. 4(a).

To understand the dispersion homogeneity of Yb³⁺ ions in core glass, the EPMA mapping of two fibers was performed. Figures 5(a) and 5(d) show the EPMA mapping of the fiber cross section for Yb³⁺ distribution in #S and #M, respectively. The red dots indicate that the Yb³⁺ concentration is high, and the blue ones indicate that the Yb³⁺ concentration is low. The red dots are concentrated in the center of the core of fiber #M, but the red dots are evenly distributed throughout the core of fiber #S. From the inset of Figs. 5(a) and 5(d), the EPMA radial line scanning of fiber for Yb³⁺ distribution shows that the fluctuation of Yb³⁺ concentration of #S is weaker than that of #M. This clearly indicates that the dispersion homogeneity of Yb³⁺ ions is better in fiber #S than in fiber #M. Besides, the Al³⁺ distribution in fiber #S and #M are presented in Figs. 5(b) and 5(e), it shows that the enhancement of Al³⁺ concentration from the edge of the core to the center of the core in fiber #M, instead, Al³⁺ ions are uniformly distributed throughout the core of fiber #S, it can also be verified by the EPMA radial line scanning of fiber for Al³⁺ distribution from the inset Figs. 5(b) and 5(e). In addition, the radial refractive index difference profiles of fiber #S and #M at 633 nm are shown in Figs. 5(c) and 5(f). The refractive index difference of the core of fiber # S is almost constant, while it is slightly increased along the radial direction in fiber #M. It demonstrates that the fiber #S is with better dopant uniformity.

 

Fig. 5 EPMA mapping of fiber cross sections for (a) Yb³⁺ distribution in fiber #S, (b) Al³⁺ distribution in fiber #S, (d) Yb³⁺ distribution in fiber #M, (e) Al³⁺ distribution in fiber #M, and radial refractive index difference profiles of (c) fiber #S and (f) fiber #M at 633 nm. Inset of (a), (b), (d), (f): EPMA radial line scanning of fiber for Yb³⁺ and Al³⁺ distribution.

Download Full Size | PDF

According to the above discussion, a model describing the formation and annihilation of color centers caused by PD and photobleaching (PB) in YDF is proposed, as shown in Fig. 6. Previous studies show that the mechanism of PD is related to the CT band [8,50]. Although the single photon energy (~1.2 eV) of the laser in the Yb³⁺-fiber is far lower than that of the CT band, it is reasonable to presume that high-energy UV photons can be generated via the simultaneous excitation of a cluster composed of several closely spaced Yb³⁺ ions. For example, the cooperative up-conversion luminescence of two Yb³⁺ ions produces a 2.5-eV VIS photon, as shown in Fig. 4(a). The accumulation of the energy of four or five laser photons of Yb³⁺ ions is enough to excite the 5.2-eV CT band. The photochemical reaction of the PD process can be described as follows:

 

Fig. 6 Model of color-center generation and annihilation in the PD process of the Yb³⁺/Al³⁺ co-doped silica fiber.

Download Full Size | PDF

Al-OHC is an aluminum-oxide trapped hole center, although the Si–O trapped hole center (Si-OHC) is expected to form in the PD process. However, neither emission signals nor EPR signals of Si-OHC were detected in our previous works [29,49]. This might be because the negatively charged [AlO_4/2]^- tetrahedral has the priority to trap the positively charged holes. Our early studies demonstrate that Yb³⁺ ions in aluminosilicate glass are surrounded by [AlO_4/2] and [SiO₄] tetrahedrals [39].

Compared with fiber #M, the lower clustering degree (see Figs. 4 and 5) in fiber #S leads to the lower probability of high-energy UV photon formation, resulting in a lower Al-OHC content. On the other hand, the inherent Yb²⁺ ions (see Fig. 3) in fiber #S promote the photochemical reaction of Eq. (2) in the reverse direction, enhancing the PB effect and further reducing the Al-OHC content. This result is supported by Kirchhof [51], who found that the reductive treatment of a YDF leads to an enhanced PD resistance. It is well-known that the Al-OHC is primarily responsible for the PD effect. The lower Al-OHC content in fiber #S indicates the better PD resistance compared with fiber #M.

4. Conclusions

The PD behaviors of YDFs prepared via the sol-gel method and MCVD with the conventional solution-doping method were investigated. For the YDFs obtained via the sol-gel method, the PD-resistance behavior was greatly improved, besides, under 974 nm pumping, the induced loss caused by PD was smaller, and the laser slope efficiency decreased less after PD testing. The up-conversion emission spectrum, EPMA results, and low-temperature EPR spectra confirm that Yb³⁺ ions were more uniformly dispersed in the YDFs prepared via the sol-gel method than in the YDFs obtained via the MCVD method. In addition, according to optical absorption spectroscopy and photoluminescence emission spectra, small amounts of Yb²⁺ ions were found in a Yb³⁺/Al³⁺ co-doped silica preform prepared via the sol-gel method. Both the existence of Yb²⁺ and the homogeneous dispersion of Yb³⁺ ions benefit the PD resistance of YDFs. A model based on the energy-transfer process between Yb³⁺ ions and color centers has been proposed to explain the balance of the PD and PB for a YDFs under 974 nm excitation. This work suggests that the sol–gel method may be effective for preparing PD-resistant YDFs.