1. Introduction

Thulium-doped fiber lasers operating in the 2 µm region receive growing attention nowadays for their potential in so called “eye-safe” applications such as laser surgery, biological imaging, atmospheric sensing or material processing [1–3]. Amongst materials for the fabrication of rare-earth-doped fibers, silica glass remains frequently used thanks to numerous advantages such as transparency up to NIR region, good thermal or chemical stability and high durability. Low price and good compatibility with common optical components also play a significant role in practical applications. However, several drawbacks put silica glass in disadvantage compared to other glass systems such as germanate, telluride or fluoride glass. Limited solubility of rare-earth ions in silica glass leads to the formation of clusters which causes concentration quenching and hinders luminescence properties. Silica glass also has a high phonon energy (≈1100 cm^-1) which increases the probability of non-radiative transitions and further limits radiative properties.

To overcome these shortcomings, silica glass needs to be modified with other oxides, i.e. Al₂O₃, P₂O₅ or Sb₂O₃ [4]. Previous studies have shown that zirconium oxide (ZrO₂) is a very perspective co-dopant of rare-earth ions in silica glass [5,6]. With an appropriate heat treatment, a phase separation can be induced in the binary SiO₂-ZrO₂ system which leads to the formation of ZrO₂ nanoparticles [7]. The presence of these nanoparticles prevents the clustering of rare-earth ions and improves luminescence properties. The zirconia particles also possess low phonon energy (≈470 cm^-1) which increases the number and probability of radiative transitions of rare-earth ions [8]. Furthermore, thanks to a good optical transparency, chemical durability and high refractive index, ZrO₂ appears as a suitable candidate for the fabrication of rare-earth doped optical fibers [5].

In this paper, we report on the experimental preparation and characterization of Tm-doped SiO₂-ZrO₂ fibers. Multiple fibers with various concentrations of zirconium and thulium in the fiber core were prepared by a modified solution doping method in order to study the effect of zirconia on the photoluminescence properties of Tm³⁺ ions and to find the optimal zirconia concentration. The fiber with the best optical properties was measured in a fiber laser setup and its laser characteristics were evaluated.

2. Experimental

2.1. Preparation of sols and optical fibers

Sols with varying concentrations of zirconium and thulium were prepared, using propan-1-ol (Lachema, p.a.), ethanol (Sigma Aldrich, 96% vol.), zirconium(IV)-butoxide (Ventron GMBG, p.a.) and thulium nitrate pentahydrate (Alfa Aeasar, 99.9%). Four sols were prepared, with zirconium concentrations of 0.16 mol.dm^-3, 0.1 mol.dm^-3, 0.09 mol.dm^-3 and 0.07 mol.dm^-3 and thulium molar concentrations corresponding to 3% of zirconium concentration in each respective solution. The process of preparation was as follows: 9.1 g of zirconium(IV) tert-butoxide and 0.23 g of thulium nitrate pentahydrate were dispersed in 200 ml of propan-1-ol and 3.7 ml of 96% ethanol. The volume of ethanol was calculated to achieve a H₂O/Zr-butoxide ratio of 0.5. The zirconium(IV) tert-butoxide was hydrolyzed by heating at 130 °C for 18 hours. The hydrolyzed sol was partially evaporated at a rotary evaporator to reduce its volume and reach a zirconium concentration of 0.16 mol.dm^-3. The less concentrated sols were prepared by diluting the 0.16M sol with propan-1-ol.

The preforms were prepared by a modified chemical vapor deposition (MCVD) method [9] combined with a modified solution-doping technique [10]. A set of porous silica layers was deposited onto the inner side of pure silica tubes (F300, Heraeus) with 18 mm outer diameter and 1.4 mm wall thickness. The porous layers were soaked with the prepared sols and left to dry on air for 24 h. The soaked tubes were further dried by oxygen for 3 h, sintered and collapsed into preforms.

Optical fibers were drawn at a 6 m high drawing tower of our own construction, equipped with a graphite resistance furnace. The fibers had cladding diameter of approx. 125 µm and core diameter of approx. 10 µm. The fibers were coated during drawing with UV-curable acrylate (Desolite 3473-3-14, DSM Netherlands).

2.2. Preforms and fibers characterization methods

A Photon Kinetics A2600 refractive index profiler was used to measure the refractive index profile (RIP) of the preforms. To verify the longitudinal homogeneity of the preforms, RIPs at several longitudinal positions under nine angle projections were measured. Numerical aperture of the fibers was calculated by the Eq. (1) [11]:

Preform samples approx. 5 mm thick were cut and polished. The dopant concentrations were measured by electron probe microanalysis (EPMA) using Cameca SX-100 at approx. 20 points across the core.

SEM images of the preform core were taken using JEOL 7000F scanning electron microscope with acceleration voltage of 15 kV.

Absorption spectra were measured using a tungsten halogen lamp as a source and a Nicolet 8700 FTIR spectrometer (with resolution of 16 cm^-1) in the range from 1000 nm to 2200 nm. Standard cut-back method using Ando AQ1425 optical spectrum analyzer was used to determine the minimum of background losses.

The emission spectra were measured on Nicolet 8700 spectrometer in a setup with ThorLabs FPL1054S diode as a source at 1619 nm, an optical isolator, and a wideband coupler. The spectra were measured in the range of 1700–2300 nm with 8 cm^-1 resolution. The spectra were corrected on the insertion loss of the wideband coupler and the spectral response of the InGaAs detector.

Time-resolved luminescence was measured in a side detection setup with excitation at 1619 nm, using ThorLabs FPL1054S diode as the source and Hamamatsu InGaAs photodiode G12183-10K as the detector. The decay curves at various pump powers were normalized, and fitted by single and double exponential functions to analyze the decay characteristics and retrieve fluorescence lifetime values, which were subsequently extrapolated to zero pump power to eradicate the influence of amplified spontaneous emission (ASE) and other effects dependent on the concentration of Tm³⁺ ions, such as cross-relaxation [12].

Laser performance was measured in a laser setup depicted in Fig. 1. An in-house built erbium/ytterbium-doped fiber laser in a ring arrangement emitting at 1565 nm was used as a pumping source. The maximum pump power was limited at 0.9 W. The performance of thulium-doped fiber (TDF) was tested in a Fabry-Perot configuration in a cavity formed by high reflectivity fiber Bragg grating (HR) at 1950 nm (R∼99,5%) and perpendicularly cleaved fiber end (R∼3,5%). The laser spectrum was measured with a Miriad S3 spectrometer (spectral resolution 2 nm)

 

Fig. 1. Laser setup for the measurement of laser characteristics.

Download Full Size | PDF

3. Results and discussion

The prepared preforms were characterized with respect to the dopant concentrations and refractive index differences between optical fiber cores and claddings. The results are summarized in Table 1:

Table 1. Basic characteristics of the prepared preforms

View Table

Refractive index differences, numerical apertures, and dopant concentrations in the preforms increased gradually with dopant concentrations in the solutions. Refractive index profile and corresponding dopant concentration profiles of the 0.07M preform are shown in Fig. 2(a) for illustration. Other preforms displayed similar shapes of the profiles. The distributions of dopants were sufficiently homogenous across the preforms cores, the refractive index profiles followed the distributions of dopants. Refractive index and zirconium concentration data for all prepared preforms were compiled together and the resulting dependency of refractive index on zirconium concentration is shown in Fig. 2(b). As can be seen, the data show a clear linear dependency. The scattering of the data is caused by the error of measurement of the Photon Kinetics A2600 profiler (described in detail in [13]). Due to the high refractive index of ZrO₂, the refractive index difference was higher than in Al₂O₃ doped silica fibers of comparable concentrations [14].

 

Fig. 2. a) Refractive index and concentration profiles of the 0.07M preform; b) Dependency of refractive index on the zirconium concentration calculated from the experimental data.

Download Full Size | PDF

The presence of ZrO₂ nanoparticles in the amorphous matrix of the preforms was confirmed by SEM imaging. As can be seen in Fig. 3, the size of the nanoparticles was greatly dependent on the concentration of zirconia. In the highest doped 0.16M fiber, ZrO₂ formed large nanoparticles with size up to 50 nm in diameter. Due to the presence of large nanoparticles, the core of the 0.16M preform exhibited a dense white coloration. On the other hand, the lowest doped 0.07M preform contained high density of small nanoparticles with diameter up to 10 nm and the preform core was transparent.

 

Fig. 3. SEM visualization of ZrO₂ nanoparticles, a) 0.16M preform, overall view, b) 0.16M preform, detail, c) 0.07M preform, overall view, d) 0.07M preform, detail.

Download Full Size | PDF

Absorption spectra of the prepared fibers are shown in Fig. 4. Optical properties of the fibers are greatly dependent on the size of the nanoparticles. In general, the attenuation of the fibers can be divided into two main components – background (intrinsic) losses and extrinsic attenuation caused by optically active dopants such as Tm³⁺ ions or OH-groups [15]. Basic optical attenuation arises from structural imperfections and density fluctuations in the matrix which are caused by the presence of nanoparticles. The Rayleigh scattering on the incorporated particles is one of the primary sources of attenuation in nanoparticle doped fibers and increases with the size of the nanoparticles and refractive index difference between host matrix and nanoparticles. The effect of nanoparticle size on the attenuation is clearly seen from the absorption spectra. The attenuation of the highest doped 0.16M fiber with largest nanoparticles could not be measured due to the extreme attenuation and impermeability of the fiber. In the other fibers, the Rayleigh scattering losses gradually diminished with decreasing concentration of zirconium and were the lowest in the 0.07M fiber with small nanoparticles.

 

Fig. 4. Measured absorption spectra of the 0.1M, 0.09M and 0.07M fibers.

Download Full Size | PDF

The extrinsic losses are caused mainly by optically active dopants such as rare-earth ions or OH-groups. The optical attenuation caused by Tm³⁺ ions corresponds to their absorption. Two absorption bands assigned to Tm³⁺ ions were present in the spectra – ³H₅ (1211 nm) and ³F₄ (1592 nm for 0.07M fiber). The maximum of the ³F₄ absorption band was slightly red-shifted for the 0.09M and 0.10M samples due to the overlap with Rayleigh scattering.

A higher limit estimate of the background losses in the 0.07M fiber was determined from the absorption spectrum at 1330 nm and was 0.9 dB/m. Standard cut-back method was used to find a minimum of background losses, which was determined to be 0.098 dB/m at 935 nm. The background losses of fibers doped with rare-earth ions are usually in the range of 0.05–0.5 dB/m [16]. The higher than usual losses in the 0.07M fiber are most likely caused by the presence of nanoparticles, i.e. high refractive index difference between nanoparticles and host matrix, or OH- impurities originating from the initial solutions.

The 0.07M fiber exhibited highest quality and lowest losses and was therefore selected for measurement of photoluminescence properties and laser performance. The luminescence properties were measured using 1619 nm diode as an excitation. The emission spectrum of the ³F₄↔³H₆ transition recorded on Nicolet 8700 spectrometer is shown in Fig. 5. The emission spectra had the maximum at 1779 nm, the spectral shape corresponds well to thulium spectra reported elsewhere in literature [12,17].

 

Fig. 5. Measured emission spectrum of the 0.07M fiber

Download Full Size | PDF

Luminescence decay of the ³F₄↔³H₆ emission band was measured in a side detection setup in order to suppress the influence of amplified spontaneous emission (ASE) [12]. An example of a fluorescence decay curve for 16 mW input power is shown in Fig. 6(a). As can be seen, the fluorescence decay exhibited a significant deviation from single exponentiality. The decay curve was accurately fitted by double exponential function (adjusted R-squared = 0.99997) which returned the values of τ_SLOW= 600 ± 1 µs and τ_FAST= 222 ± 1 µs. Deviations from single exponential decay have been previously reported numerous times for Tm³⁺ ions [18,19] and have been generally attributed to several phenomena. First explanation would be inter-ionic energy transfer processes such as cross relaxation or pair-induced quenching. These processes are observed mainly in high Tm³⁺-doped fibers because the probability of inter-ionic energy transfer increases with the decreasing distance between Tm³⁺ ions and the formation of clusters. In the case of the 0.07M fiber, low Tm³⁺ concentration of 260 ppm and the presence of ZrO₂ nanoparticles should be beneficial in terms of suppressing the clustering of Tm³⁺ ions, however Simpson et al [20] have previously shown that energy transfers resulting from the formation of clusters may occur even in Al₂O₃ co-doped silica fibers with Tm³⁺ concentrations as low as 200 ppm. Moreover, the value of 260 ppm represents the overall concentration of Tm³⁺ in the fiber, but the local concentration in the nanoparticles can be higher. Therefore, the effects of Tm³⁺-Tm³⁺ energy transfer cannot be ruled out.

 

Fig. 6. a) measured decay curve of the 0.07M fiber under 16 mW excitation with single and double exponential fits, b) fluorescence lifetimes of the 0.07M fiber from single and double exponential fits of the decay curves as a function of excitation power.

Download Full Size | PDF

Another factor to consider is possible location of Tm³⁺ ions in different optically active sites in the matrix as suggested by Moulton et al [21] for Al₂O₃ doped silica fibers. It has been reported that rare-earth ions are preferentially partitioned in the silica-poor phase (nanoparticles) during phase separation [22]. However, Tm³⁺ ions can generally be located in a Zr-rich (ZrO₂ nanoparticles) or Zr-poor (silica matrix) environment with both sites displaying different decay rates which may cause deviations from single exponentiality. Tm³⁺ ions dispersed in an environment with higher phonon energy (silica matrix) would manifest faster decay (characterized by τ_FAST) than Tm³⁺ ions embedded in the ZrO₂ nanoparticles (characterized by τ_SLOW).

The decay curves were measured for multiple input powers and fitted by double exponential functions. The values of slow and fast components of the luminescence lifetime were extrapolated to zero pump power, which returned the values of τ_SLOW= 686 ± 1 µs and τ_FAST= 304 ± 1 µs, as shown in Fig. 6(b). Despite lower accuracy, the decay curves were fitted by single exponential function to compare the luminescence decay of the 0.07M fiber with literature, which returned the lifetime value of 420 ± 10 µs after extrapolation to zero pump power. This value corresponds well to the higher end of fluorescence lifetimes found in silica glass which usually vary between 220 and 420 µs depending on the Tm³⁺ concentration [23,24].

The 0.07M fiber was further tested as a gain medium in a fiber laser. Er/Yb-doped fiber laser emitting at 1565 nm was used as a pump source. The fiber was spliced into the laser setup and the dependence of the laser output power (P^out) on the absorbed pump power (P^{abs. pump}) was measured. Two main parameters were obtained: laser threshold, calculated as the intercept of the linear fit from the dependence of laser output power on absorbed pump power, and slope efficiency, a characteristic parameter that is defined by Eq. (2):

 

Fig. 7. a) Slope efficiency and laser threshold as functions of active fiber length, b) Output power of TDFL using 0.07M fiber (1.8 m) as a function of absorbed pump power; laser spectrum at 530 mW output power as inset

Download Full Size | PDF

4. Conclusions

Tm-doped SiO₂-ZrO₂ fibers were prepared using the MCVD method combined with modified solution-doping technique. ZrO₂ nanoparticles were formed during the preform processing, their size was dependent on the Zr concentration. Optimal Zr concentration was found to be approx. 1 at. %. Silica fiber doped with 1 at. % of zirconium and 260 ppm of thulium had sufficiently low background losses and exhibited strong emission around 1.8 µm. Lifetime of the ³F₄↔³H₆ transition of Tm³⁺ ions extrapolated to zero pump power was 420 ± 10 µs. For the first time in literature, we have demonstrated laser operation in a Zr/Tm-codoped silica fiber. In a fiber laser setup, the fiber had a threshold for laser operation of 233 mW and a slope efficiency of 72.7%. MCVD combined with modified solution doping proved to be a feasible method of preparation for Tm-doped SiO₂-ZrO₂ optical fibers for use in fiber lasers.