1. Introduction

Fiber lasers are attractive for practical applications due to their robustness, turn-on-key operation, and immunity to environmental conditions. They offer high power, excellent quality of the generated beam, and high power conversion efficiency. Yb and Er-doped laser already found their applications in telecom systems, laser machining, medicine as well as defense [1]. Fiber lasers and amplifiers based on thulium (Tm) doped silica fibers have gained importance due to the possibility of obtaining high output power and generation at 1700–2100 nm in the so-called eye-safe wavelength range. This wavelength range is attractive for applications in medicine, in free-space data transmission in the atmosphere (e.g. LIDAR, wind farms), material processing, and also as a pump for mid infrared sources [2].

There are several absorption bands available to excite the thulium ions in silica glass matrix with main absorption peaks at 460 nm, 670 nm, 790 nm, 1210 nm, and 1650 nm [3]. However, only pumping around 0.79 µm and 1.65 µm are efficient enough to be considered for high power, cladding-pumped silica fiber systems. Pumping at visible wavelengths has a limited conversion efficiency due to Stokes shift. On the other hand, using an absorption band of 1210 nm for pumping could theoretically reach the efficiency of 50–60%, but there are not suitable high power pump sources in this range. Only Raman fiber lasers can be considered for this purpose, but it requires a complex laser setup [4].

The most popular, straightforward, and cost-effective strategy for high power thulium-doped fiber lasers and amplifiers is to pump at 790 nm band, where high brightness laser diodes are commercially available. The commonly known advantage of 790 nm pumping is a cross-relaxation process in Tm doped silica, which allows obtaining two generated photons for one pump photon [5]. Therefore, instead of a theoretical maximum efficiency of about 40% due to Stokes shift, one can obtain an efficiency of 80%, expecting high power performance. However, to achieve a cross-relaxation effect and high efficiency in Tm doped silica, a high doping level, well above 2%wt, typically 3.5 wt% or more is required [2,5,6], which causes a significant increase of refractive index in comparison with undoped silica.

Due to the large contrast of refractive indices between active core and pure silica cladding only fibers with a small diameter of core can fulfill requirements of single mode propagation commonly expressed as normalized frequency V < 2.405. An application of an additional higher index area surrounding the core commonly termed a “pedestal” was proposed in 2006 by Tankala et al. [7] to reduce the contrast of refractive indices between the core and cladding and simultaneously allow increasing a diameter of the core, while single mode regime of operation is still maintained. Nowadays this approach is used commercially as well as in research as the basic solution for the development of large mode area single mode thulium doped fibers required for high power Tm doped fiber lasers [8–13]. However, the pedestal, despite having a lower effective NA for the fiber core, acts as a high NA waveguide itself playing a role not only cladding for the active core area but also an unwanted secondary core. The secondary core is formed because the pedestal area is surrounded by a lower refractive index area of pure silica, which is a standard fiber cladding. As a result, a fraction light beam that cannot be guided in the active core due to external perturbation or coupling issues is trapped in the pedestal area and can adversely affect laser generation efficiency as was previously pointed by Jollivet et al. [14].

The strategy to mitigate the influence of the pedestal on the efficiency and beam quality of Tm fiber laser was discussed in [15]. The authors suggested decreasing the interaction between the detrimental radiation propagating in the pedestal and the doped core by increasing the pedestal diameter and decreasing the diameter of the core. Although that approach delivered positive results, the increasing of the pedestal diameter extends the preform fabrication time and processing cost due to the higher number of deposition layers needed to be applied in used standard MCVD technology. On the other hand, decreasing the diameter of the core limits the scalability of laser output power due to nonlinear effects such as Stimulated Raman Scattering (SRS) or Stimulated Brillouin Scattering (SBS) [16].

State-of-the-art in the development of thulium doped fibers are presented in Table 1. The highest power reported so far from the Tm doped fiber MOPA amplifier operating in single mode regime were slightly over 1 kW. The 1-kW level was first time reached in MOPA at 2045nm in the setup pumped with 790 nm diodes, resulting in 53.2% slope efficiency [17]. Current record output power of 1150 W was reached in ultrafast MOPA at 1960nm center wavelength pumped at 793 nm with air-clad active fiber design and 61% slope efficiency [18]. Other recent demonstrations include a two-stage all-fiber MOPA configuration with a tunable wide-range 107 nm generation and maximum output power of over 1 kW at 2050nm [12]. The thulium doped fiber fabricated using nanoparticle doping technique to provide uniform high concentrations of active ions, allowing for high amplifier efficiency of up to 62% were reported [10]. In the case of a single cavity oscillator, the highest power was reported by Walbaum et al [8], where Tm doped fiber laser pumped at 790 nm produced 567 W at 1970nm. The laser efficiency was 49.4% concerning absorbed pump power with a relatively low beam quality of M²= 2.6.

Table 1. State-of-the-art in the development of thulium doped fibers

View Table

The shortest emission wavelength of 1628 nm was obtained in Tm/Ge co-doped silica fiber, where co-doping with passive germanium (Ge) ions increased the gain at the short wavelength of the thulium emission band [13]. Demonstrations of thulium doped fibers for laser or amplifier applications also include photonic crystal fibers with large 50 µm cores allowing ultrafast pulsed operation [19]. To extend the laser wavelength range towards mid-infrared (MIR), fibers co-doped with Tm and Ho ions were fabricated and used in free-space oscillator with tunable generation over a wide range of 1990 -2190 nm with maximum output power at 2100 nm [11]. The design of the large mode area fiber without a pedestal was presented in fiber of germanate glass synthesized by the conventional melting-quenching technique, which allowed for controlling the refractive index of the material intended for the core and cladding of the fiber [20]. The output power of the oscillator was 1.52 W, and due to the multicomponent glass used, the maximum power exceeding the tens of Watts is not expected. Apart from fibers with pedestals or photonic crystal fibers, other techniques have been demonstrated, e.g., structured-core design with thulium-doped nested-ring [21], or “pixelated” core design [22], though to our knowledge, for thulium-doped fiber fabrication has not been yet reported in the open literature. Alumino-phosphate silicate glass core [23], or inclusion of fluorine doped layers may also lead to a decrease of the average refractive index of the core but these techniques are not applicable or difficult to apply with thulium co-doping. The highest reported slope efficiency of thulium doped fiber laser pumped at 790 nm was > 72%, [24]. These results are obtained for multimode fiber with a small core with a diameter of 8 µm. The high efficiency is achieved due to the pedestal-free design of the Tm doped fiber, but it is not scalable toward large mode area (LMA) fibers required for high power lasers. The efficiencies of the high power thulium doped fiber lasers based on fibers designed with pedestal pumped at 790 nm slightly exceed 50% (Table 1) and there is still potential for significant improvement.

The pedestal also requires additional care when splicing to other fibers to apply, e. g., the all-fiber setup. The germanium-doped glass used to form the pedestal layer has a significantly lower melting temperature than the core and the cladding of the fiber. To achieve efficient light coupling between fiber with pedestal and standard fiber, a dedicated splicing procedure is required [14]. The use of fibers with pedestals brings several unfavorable technological issues that influence beam quality, efficiency, and maximum output power of the laser.

In this work, we propose a novel approach for the development of pedestal-free Tm doped large mode area fibers. We use a well-known stack-and-draw technique to form a fiber core based on thousands of rods, instead of assembly photonic crystal cladding, as it was reported previously [18]. We consider a fiber with a core composed of multiple identical sub-wavelength-sized Tm doped silica nanorods, which form the step index core (Fig. 1). Every nanorod has a central part doped with thulium surrounded by a ring of pure silica glass. This way we obtain ‘locally’ very high concentration of Tm ions required to enable a cross-relaxation process. Simultaneously, the nanostructure forms a uniform effective medium since individual rods are much smaller than the wavelength [26]. The refractive index of the fiber core is defined as the weighted average refractive index of both Tm doped and pure silica areas. As a result, we have locally a high concentration of Tm and a low effective refractive index in the core area. We note that the effective refractive index of the core can be engineered by modifying the ratio between Tm doped central area of individual nanorods and the passive silica glass area surrounding it, thus we can adjust the effective mode area and modal properties of the fiber. As a result use of a pedestal is not necessary to obtain single mode performance in large mode area fibers. We discuss the design of the fiber with a core of 30 µm, the characterization of the fabricated fiber and we demonstrate the generation in a proof-of-concept fiber laser system.

 

Fig. 1. A concept of pedestal-free Tm doped large mode area fibers. The fiber is composed of a low refractive index cladding (pure silica or fluorine doped silica) and the core area is composed of multiple identical sub-wavelength-sized structured nanorods. Every nanorod has a central part doped with Tm surrounded by a ring of pure silica glass. An effective refractive index of the core can be engineered by modifying the ratio between Tm doped central area of individual nanorods and the passive silica glass ring area.

Download Full Size | PDF

2. Development of pedestal-free Tm doped fiber

The concept of a nanostructured core of the optical fiber is based on the composition of glass nanorods made of two or more types of glasses with different refractive indices ordered in hexagonal lattice. The diameter of individual glass nanorods is the fraction of the light wavelength, usually below λ/2.5. Thus, the refractive index of the nanostructured core material can be treated as an effective medium for the considered wavelengths, since single elements of internal structure are much smaller than the wavelength and light cannot ‘see’ individual nanorods. The effective refractive index is calculated as an averaged refractive index over a wavelength-sized neighborhood and forms a continuous-like refractive index profile according to the designed distribution of glass nanorods [27]. In the case of a uniform mixture of two types of nanorods, the medium can be described with a single value of effective refractive index. This approach is based on the Effective Medium Theory (EMT) and Maxwell-Garnett mixing formula [26]. It was previously applied to develop other nanostructured optical fibers [28–30] and microoptical gradient index components [31,32].

The nanostructuring approach brings the availability of controlled distribution of separated thulium highly doped and undoped silica areas. Therefore the high refractive index of the thulium doped silica can be effectively balanced with pure silica passive area surrounding thulium doped areas to obtain the demanded effective refractive index of the nanostructured material. Implementation of such material within the fiber core allows creating the fiber with low difference of refractive indices between the core and cladding, forming a single–mode fiber without the need for a pedestal area around the core. Simultaneously, thulium doping level in doped areas (Tm doped nanorods) can achieve the level suitable for the cross-relaxation process, providing high performance efficiency in fiber laser setup.

The nanostructured core proof-of-concept fiber was designed as an effectively step-index fiber. It allows the use of a multiple stack-and-draw fiber development technique suitable for nanostructured large core fibers, as we previously showed [29]. We note that the nanostructuring approach allows developing of optical fibers with arbitrary 2D refractive index profiles using simulated annealing algorithms to determine the distribution of doped and undoped glass nanorods in hexagonal lattice [28,31,32]. However this case we use only one type of silica nanorods, where their central part is doped with thulium, while the remaining area is pure silica. The assembly of these types of nanorods into hexagonal lattice forms an effectively flat uniform refractive index in the core.

To develop nanostructured core fiber we used two aluminosilicate thulium doped preform rods fabricated with the MCVD technology extended with the nanoparticle-doping method [33], where the central part is doped with Al₂O₃ and thulium, while the remaining area is pure silica (Fig. 2). Nanoparticle doping offers the capability of reaching higher dopant concentrations in comparison to more commonly used solution doping [10]. Both preform rods were developed in the same technological process and they have similar profiles and doping levels. As the capillary for the preforms, we used the F300 Heraeus silica tube. The outer diameter of both of the MCVD preforms was 9.3 mm, and the typical concentration profile is shown in Fig. 2(b), which would correspond to absorption coefficient value of 800 dB/m at 793 nm in a fiber with complete overlap of the pump with the doped area. The doping level of both preforms exceeded the value of 10 000 ppm (i.e. 2.6 wt%) Tm³⁺ in the central part of the doped area which is not high enough for the efficient cross-relaxation effect in Tm doped silica required for high-efficiency laser performance. According to Jackson [5] the threshold of cross-relaxation effect in Tm doped silica is about 1.2 wt% of Tm³⁺. However, the efficiency of that effect increases with Tm doping level and it is commonly assumed that the doping level of at least 3.5 wt% is required for an efficient “two for one” process. The concentration of Tm and Al₂O₃ was measured with EPMA (electron probe micro analyzer) which enabled quantitative chemical analysis (WDS) at high sensitivity. [34]. The EPMA has a better limit of detection than typically offered by Energy dispersive X-ray (EDX) analysis. Concentration of Tm³⁺ was measured with 2% relative error. Considering the fill factor of the doped and undoped glass areas in MCVD preform the average absorption of the preform was estimated at 43.3 dB/m at 793 nm. The absorption was measured in individual rods with the central part doped with Tm and surrounded with the pure silica ring, which we later used to form a core of the final fiber. The absorption estimation was made with the formula ${\alpha _{abs}} = \; {\alpha _{glass}} \times {d^2}/{D^2}$, where α_glass is the absorption coefficient of Tm doped preform core, d and D are diameters of the doped core and whole MCVD preform, respectively.

 

Fig. 2. Refractive index profile (a) and concentration of Tm³⁺ and Al₂O₃ profile (b) of the thulium doped part of the silica rod made with MCVD technology used for the development of the nanostructured fiber. The diameter of the doped area was at 1.22 mm, and the outer diameter of the MCVD rod was 9.3 mm.

Download Full Size | PDF

We etched the preform rod down to 6.5 mm of outer diameter to achieve the expected average refractive index of the preform of 5.4 × 10⁻⁴ above the undoped silica one for the wavelength around 2 µm. This value was selected to ensure a single mode performance of the final fiber with a target core diameter of 40 µm based on normalized frequency V < 2.4 taking into account fiber cladding of pure silica. Note that the average refractive index of the preform rod is much lower than the maximum refractive index value of its central area with the refractive index of 22.6 × 10⁻³ higher than undoped silica (Fig. 2).

In the first stage of the fabrication process, we assembled 61 thulium doped silica rods into the hexagonal lattice with 9 elements on the diagonal. The structure was drawn at a fiber drawing tower to form an integrated intermediate preform with a hexagon cross-section. In the next step, the intermediate preform was cut into 109 hexagonal canes and stacked again in the hexagonal lattice inside the silica tube to form the final core preform. With this method, we achieved a preform of the core composed of 6649 (61 × 109) individual elements arranged in the hexagonal lattice to form a quasi-circular fiber core preform. The structure was drawn at the fiber drawing tower to form an integrated preform of the core (Fig. 3). The rugged edge between the core area and cladding is due to the stacking of hexagonal canes.

 

Fig. 3. SEM image of nanostructured core preform cross-section of 2.8 mm in diameter, the core structure (bright area) of 1.5 mm in diameter (a). Internal discrete structure of the core preform (b), the bright areas are thulium doped silica, and the dark area is undoped silica.

Download Full Size | PDF

The core preform was 2.8 mm in diameter, but we etched it down to 1.5 mm to eliminate even more the undoped silica area. The rugged layer between the core structure and cladding was also etched to provide a circular core for the final fiber (Fig. 4). It should be noted, that bending losses were too high for fiber with pure silica overcladding tube, as the numerical aperture of the core was only about 0.5 × 10⁻³. Therefore, by applying the rod-in-tube method, i.e. implementing the core preform inside the fluorine doped silica tube (F320-28 Heraeus glass tube), we fabricated the final fiber (Fig. 4) with a nanostructured core surrounded by fluorine doped silica cladding with depressed refractive index.

 

Fig. 4. The cross-section of the fabricated Tm doped fiber (a), the nanostructured core area with diameter of 30 µm (b).

Download Full Size | PDF

As the refractive index of the fluorine tube is -4.06 × 10⁻³ below pure silica one, according to vendor specification, and the refractive index of the nanostructured core is estimated to be 0.54 × 10⁻³ above pure silica, we expect that the developed optical fiber has a refractive indices difference between core and cladding of 4.6 × 10⁻³. The calculated numerical aperture of the fiber is NA = 0.115 ($NA = \; \sqrt {{n_{co}}^2 - {n_{cl}}^2}$). This value is comparable with commercially available large mode area Tm doped fibers with pedestal (Table. 1). The normalized frequency calculated as $V = \frac{{\pi d}}{\lambda }NA$ was V = 5.6 at the 2 µm wavelength, which indicates the guiding of a few modes.

The fiber in the cross-section was slightly elliptical due to the fiber drawing processing conditions. The nanostructured core had a diameter of 30 and 31 µm in perpendicular directions, and the overall fiber was 110 µm and 115 µm in diameter. The core ellipticity does not influence on optical or structural properties of the fiber, since is very small concerning the total diameter and dedicated operation wavelength around 2 µm. The calculated diameter of individual Tm doped nanorods areas in the core was 71 nm, and the distance between doped nanorods was 352 nm if we consider only geometric changes due to elongations and no diffusion effects. The diameter of individual glass nanorods corresponds to λ/5 at operation wavelength around 1.8 µm, therefore it meets the he λ/2.5 criterion in excess and it can be treated as a uniform effective medium although internal nanostructure exists. Unlike SEM images of the core preform, shown in Fig. 3, the individual thulium doped areas in the final fiber core were not observed due to the limited resolution of the method based on back-scattered electrons (BSE) used in SEM investigation [34] and low contrast between doped and undoped silica glasses. The temperature of the fiber drawing process was relatively low for silica, i.e. 1910 °C to limit diffusion processes between doped and undoped glass areas. Similar drawing parameters were set during the fabrication of silica fiber made with germanium nanostructured rods, in which the doped and undoped areas were distinguished [30]

The fiber was coated with low index polymer to achieve a double-clad structure, commonly used in fibers for laser applications [35]. The refractive index of polymer at 792 nm is 1.38. It corresponds to the numerical aperture of NA = 0.44 for clad pumping.

3. Characterization of pedestal-free Tm doped fiber

The refractive index profile of the fabricated fiber was measured using the commercially available optical fiber analyzer (IFA-100, Interfiber Analysis Inc.) at the wavelength of 979 nm [36]. The measured difference between effective refractive indices of the nanostructured core and fluorine doped cladding is -4.85 × 10⁻³, as shown in Fig. 5, which corresponds well to the designed value presented in the previous chapter. It should be noted that fluctuations of refractive index in cladding area visible in Fig. 5 are because the cladding of the fiber was made of a few fluorine tubes pushed into each other during the fiber preform assembling process. The variation of refractive index corresponds to the variation of fluorine doping level in the individual tubes. This approach limited the possibility to choose the shape of the cladding to a circular one, however, D- shape or polygonal cladding would be favorable for fiber laser [19].”

 

Fig. 5. Measurements of 1D refractive index profile of the Tm doped nanostructured fiber (a); 2D RIP profile (b).

Download Full Size | PDF

We also measured the spectral absorption in the cladding of the fiber. The tested fiber similarly to fiber described in the previous section was coated with the same low index polymer. The cladding absorption was measured in a 572 cm long fiber sample laid freely on the optical table. The fiber was pigtailed from both sides by 105/125 µm core multimode fiber and illuminated by a halogen lamp. Spectral attenuation of the radiation in fiber cladding was estimated by the assumption that the optical power was homogeneously distributed across the fiber cross-section. We used two optical spectrum analyzers ANDO AQ6317B and FT-IR spectrometer Nicolet 8700. The measured absorption in fiber cladding is shown in Fig. 6. The additional peak around 1700nm is due to polymer coating absorption. We note that absorption spectra in the broad spectral range were similar in all cases, especially at pumping wavelength at 793 nm considered within this work. The thulium absorption was measured and the absorption spectra in broad spectral range were found similar to the spectra reported in literature in thulium-doped alumino-silicate fibers [37]. The estimated cladding absorption in the peak at 793 nm was 3.9 dB/m.

 

Fig. 6. Measurements of spectral absorption of the Tm doped nanostructured fiber.

Download Full Size | PDF

We also measured the background loss of developed fiber with a single mode laser diode at 1310 nm outside the absorption bands of Tm ions with the standard cut-back method. The measured fiber was bent with a large radius of 0.5 m to eliminate bending losses. The 16 m long fiber sample was cut sequentially down to 6.5 m. We measured the loss of 0.20 dB/m, which is relatively high, but it allows obtaining the laser generation. High background losses are related mainly to the technological regime of assembly of preform in research grade labs without clean room and quality of the used material.

4. Laser performance of Tm-doped nanostructured fiber

We tested the thulium-doped nanostructured fiber in a laboratory laser setup shown in Fig. 7. The resonator had a reflectivity of ca. 3.5% on both ends formed by Fresnel reflection at the perpendicularly cleaved investigated fiber. As a pump source, we used a 30W multimode laser diode operating at central wavelength of 793 nm (spectral width of 2.3 nm) pigtailed to a 105 µm core fiber with the numerical aperture of NA = 0.22. The laser diode was driven using a standard Laser Diode Driver with a built-in diode temperature controller (Wavelength Electronics). We used a pair of identical aspherical lenses L1, L2 of f = 11 mm and NA = 0.25 (Thorlabs C220TMD-B) to couple efficiently the pump beam into the internal cladding of the examined fiber. The lens L1, and L2 were covered with an anti-reflection coating at pump wavelength. We used dichroic mirrors DM1, DM2 (>99% reflectance at the wavelength range 1900-2300 nm and 94.6% transmission at the pump wavelength 793 nm) to separate the laser outputs from the pump beam. The laser output beams were collimated with lens L2 at the first output and with uncoated lens L3 (Thorlabs A240TM) at the second output. The transmission losses of lens L2 and L3 at the laser wavelength, including the Fresnel reflection from the surface of the lens are considered in the calculation of the laser output power. The tested thulium doped fiber was coiled with the radius of 10 cm.

 

Fig. 7. The test laser setup; M₁,M₂ – dichroic mirrors; L₁, L₂, L₃ aspherical lens

Download Full Size | PDF

Figure 8 shows the laser output powers versus the absorbed pump power for fiber lengths of 5.8 m, 6.8 m, 7.8 m, and 8.8 m. We found out that the optimum fiber length in terms of slope efficiency (SE) was about 7.8 m. Absorbed pump power (x-axis in Fig. 8) was calculated from the difference between the input and unabsorbed radiation that was measured at the output end of the tested fiber.

 

Fig. 8. The measured output power versus absorbed power for the laser setup with nanostructured fiber with various lengths for individual (Pout1, Pout2) and both (Pout1+ Pout2). The optimum fiber length of the laser is 7.8 m. Pout1, Pout2 denotes power measured for input and output end of the fiber.

Download Full Size | PDF

The maximum slope efficiency measured in the fiber laser setup was 29.0%. 4.2 W was the maximum power of the laser output and it was limited by the available pump power. Saturation of output power is not observed even at the highest achievable pump power The laser action threshold was around 7.27 W of absorbed power. In the laser setup, we measured both the laser output power and the residual pump power P_RP, which allowed calculating the pump power absorbed by the fiber. In calculations, we also took into account the transmission losses of the residual pump power P_RP caused by dichroic mirror DM2 at 793 nm wavelength.

The achieved laser efficiency is lower than the reported state-of-the-art Tm doped fibers presented in Tab. 1. The relatively low efficiency in this laser is related to too low concentration of thulium ions in part of the elements so they do not fully benefit from the 2 for 1 cross-relaxation effect. The cross relaxation effect is performed when Tm³⁺ concentration in the core is above 3.5 wt%, i.e. approximately above 13 400 ppm ions (in dependence on the composition of other components). In addition high losses in the developed fiber reduce a gain efficiency.

The laser operated in the free-running laser mode as indicated from the laser spectrum examples in the range 1880-1925nm (Fig. 9). To achieve narrow-line stable laser emission the frequency selective elements, e.g. Bragg grating, are required [8].

 

Fig. 9. The spectrum of laser generation for 6.8 m and 8.8 m long sections of nanostructured fibers.

Download Full Size | PDF

To verify the modal properties of the lasing beam we measured the beam quality parameter M² using the standard method according to ISO 11146 norm (Fig. 10) [38]. We obtained M_x²= 1.24 and M_y²= 1.30 for the generated spectrum at maximum 4W power output. The M² parameters were determined based on measurements carried on for a beam generated at the single fiber output (P_out2).The achieved good beam quality of the laser beam indicates that generation was mainly obtained in the fundamental mode with a noticeable, but small fraction of power localized in higher modes (M²= 1.3). Higher order modes, theoretically supported in the fabricated fiber, were effectively suppressed with the bending of the fiber in the laser setup. The bending radius of the tested fiber in measurement setup was 10 cm. Bend-induced single mode performance is commonly used in fiber lasers [5].

 

Fig. 10. Measurement of the M² parameters for laser beam generated in Tm doped nanostructured fiber. M_2x = 1.24 and M_2y = 1.30 are obtained for 4W laser output power.

Download Full Size | PDF

5. Conclusions

We experimentally verified a new concept of pedestal–free Tm doped active fiber using a concept of the nanostructured core fibers. In this concept, a core of the fiber is composed of discrete nanorods which form effectively continuous refractive index distribution, since nanorods have diameters much smaller than the wavelength of the propagating light beam.

The main advantage of the concept is the possibility of independent control of local thulium concentration and effective refractive index of the fiber core. Instead of a uniform distribution of thulium in all core area we proposed to locally concentrate thulium ions in the form of nanorods. This way we can obtain a high local concentration of thulium to use a cross-relaxation effect for high gain efficiency. Simultaneously, we can control the effective refractive index of the core by changing the distance between nanorods when filling with pure silica glass passive areas. Adjustment of size, doping level of nanorods, and distance between them allows controlling the difference of refractive indices between the core and cladding without the use of any pedestal area.

We have verified experimentally the performance of fabricated fiber for low-power operation up to 4W of laser power output due to limited available pump power. We have shown a generated multiline laser action in the range 1880-1925nm (no mode selection with Bragg grating is applied) in a bend-induced single mode regime (measured beam quality of M_x²= 1.24 and M_y²= 1.30). We have obtained a relatively low slope efficiency of 29%. This is related to the relatively low concentration of thulium ions in the nanorods, just close to the limit of the presence of the cross-relaxation effect. Moreover, a combination of low average concentration of thulium ions at fiber cross-section and high background losses didn’t allow obtaining higher slope efficiency in a longer section of fibers (optimum fiber length was 7.8 m).

Our results show that nanostructuring of optical fiber offers an additional degree of freedom in the development of fiber lasers. Local concentration of active ions allows to control independently local gain and effective refractive index in the core. Moreover, the arbitrary distribution of active nanorods allows shaping gain distribution in the core as well as its refractive index profile in 2D cross-section. Nanorods of various types of glasses (passive and active ones) can be arbitrarily distributed in hexagonal lattice forming the fiber core. Since individual rods meet the criterion of an effective medium, they can create effectively continuous material of the fiber core with arbitrary designed gradient index or gain distribution at its cross section [30,32]. With this method, very large mode area fibers can be developed as we have previously shown for passive fibers [39]. Finally, we note that nanostructured core fibers are all-solid structures without any air holes in their structure. Therefore their further integration with standard fiber components is relatively straightforward, as we have previously demonstrated for germanium doped silica nanostructured fibers [40].