1. Introduction

The important progress in the output power scalability of fiber lasers that has occurred during the recent years has been accomplished mainly due to the advances in design and fabrication of Ytterbium (Yb) doped fibers and the availability of high brightness pump diode lasers [1]. Yb-doped silica fibers are the preferred choice for a gain medium in high power sources because of their high threshold for optical damage, high glass stability, low optical losses, easy thermal management and high optical efficiencies due to the small quantum defect between the pump and the laser wavelength [2, 3].

Yb-doped fibers, like other rare-earth (RE) doped fibers, are typically prepared by modified chemical vapor deposition (MCVD) in combination with a solution doping technique [4]. In this process a porous silica layer is deposited inside a quartz glass tube, which is subsequently removed from the lathe and is immersed in a solution of RE-ions and co-dopants. To achieve complex core structures requiring several core layer depositions, this process is repeated iteratively multiple times, which reduces the usable length of the preform and results in a rather time consuming and expensive process. As a consequence, the number of core layers observed in commercial fibers is typically relatively small, limited to about 10 [5–7]. Moreover, the next generation of laser fibers will require further improvements in terms of doping concentration and its uniformity, accuracy in control of dopant and refractive index profiles while still maintaining low background losses. We believe that these improvements are likely to be more easily achieved by vapor phase techniques rather than solution doping based processes [8].

Recently, two modified solution doping techniques compatible with the MCVD process have been proposed [7, 9, 10]. They both eliminate repeating the quartz tube removal from and the reassembly in the lathe. A system based on the high volatility of lanthanide chelate complexes at moderately low temperatures allows high RE-doping when the chelates are heated in a crucible directly within the MCVD preform [9, 10]. Since the incorporation of RE-ions and co-dopants occurs simultaneously with the silica deposition, a further advantage is the possibility of reduced RE-ion clustering compared to the standard solution doping technique. An alternative to this technique consists of an in situ soaking of the porous soot with a thin layer of dopant solvent with the substrate in the lathe, and drying the solution by vaporization [7]. Multiple layers can be deposited in a reasonable timescale making this method well suited for fabrication of large-area cores required in high-power fiber laser applications.

A well-proven fabrication method named direct nanoparticle deposition (DND), developed at nLIGHT Oy (Finland), has been positioned as an alternative to the MCVD-method in manufacturing commercial laser fibers. The soot deposition is based on the combustion of gaseous and atomized liquid raw materials in an atmospheric oxyhydrogen flame [5, 6]. The size of the particles of doped silica can be adjusted between 10 and 100 nm providing a high flexibility in the glass composition with hundreds of accurately controlled layers, which allows tailoring the refractive index and the RE-doping profiles. Yet another new technique, based on the powder-sinter technology, has been shown to produce highly RE-doped bulk silica rods [11, 12]. An aqueous suspension of SiO₂ particles is mixed with a solution of RE-ions. This mixture is subjected to steps of dehydration, purification and sintering to obtain a core rod that is further processed to form a suitable preform for fiber production.

Despite the efforts on implementing an MCVD-compatible fabrication technique resulting in better cost-efficiency and fibers with better performance compared to the solution doping, it has not been reported that any of the proposed techniques have been widely implemented.

In this paper, we present a new fabrication method for Yb-doped fibers compatible with the MCVD technology; here the dopants are introduced by atomic layer deposition (ALD). ALD is a gas-phase vapor deposition process for thin film growth where the reactants are sequentially pulsed on the substrate surface [13, 14]. ALD is a self-terminating gas-solid reaction capable of coating complex shapes or even porous structures, for which other chemical vapor depositions (CVD) techniques would only clog the outer surface [15]. ALD also results in a highly uniform film thickness over large surface areas, which has been proved to be useful in a wide range of applications such as in production of electroluminescent displays and in the microelectronics industry [16]. The idea of combining the well-established MCVD fiber fabrication technology with the ability of ALD to coat porous structures was first used to fabricate an Erbium-doped fiber [17], as a proof of principle. We present the fabrication process for large mode area (LMA) Yb-doped fibers. The distribution of the dopants in the porous silica layer transversal to the MCVD soot was studied showing successful coating of the surface extending deep into the layer. Measurements of the refractive index profile (RIP) on a collapsed preform, as well as fiber results for the background losses, doping homogeneity and photodarkening are presented. Furthermore, we show the slope efficiency measurements for a cladding-pumped ALD Yb-doped fiber laser.

The presented fabrication process can potentially result in an integrated ALD-MCVD system, which would allow several core layer deposition/doping cycles without the need to remove the preform from the MCVD system. This in turn would enable accurate in situ doping with significantly reduced fabrication time.

2. Fiber fabrication

We distinguish three phases in our preform fabrication process. In the first phase, a porous SiO₂ is deposited inside a silica tube using a conventional MCVD-system (Nextrom OFC12). The deposited soot layer has a thickness of 320 µm and the silica tube has an inner diameter of 19 mm. These dimensions were selected in order to achieve a fiber with a core diameter of about 20 µm and a total diameter of approximately 125 µm. The porosity was determined to be 82 ± 2%. The volume of the porous soot was calculated from the geometrical dimensions of a tubular sample and the soot layer thickness determined by optical microscopy. The equivalent solid glass volume was calculated from the weight of the soot in the sample and the known density of silica glass (2.2 g/cm³).

In the second phase, the MCVD-soot is transferred to a separate ALD equipment for doping with Yb₂O₃ and Al₂O₃. The equipment consists of a Beneq TFS 500 reactor with an HS 200 hot source and an ozone generator. The whole equipment is located in a clean room environment, and precautions such us sealing the soot-preform ends are taken to avoid contamination during the soot transferring. The soot-preform is positioned inside an in-house fabricated tube reactor with 230 mm of length and 25 mm of diameter integrated to the vacuum chamber of the TFS 500. This reactor ensures that the gas flow of precursors is well directed inside the soot and that purging of non-reacting chemicals is optimum.

In Yb-doping the porous soot surface is coated with Yb₂O₃. During the doping process the gaseous precursors react sequentially at the soot surface to form a thin film with a thickness less than a mono-layer in a one reaction cycle. Thus, an ALD cycle can be divided into four steps:

Figure 1 describes in detail the process steps for one ALD reaction cycle used in the deposition of Yb₂O₃ and Al₂O₃. Nitrogen is used as a carrier of the precursors and as a purging gas between precursor pulses. The ytterbium and oxygen precursors used areβ-diketonate Yb(thd)₃ (thd = 2,2,6,6-tetramethyl-3,5-heptanedionato) and ozone O₃, respectively. The porous soot layer constitutes a substrate with an extremely high aspect ratio. Previous studies have shown successful methods for conformal coating of nanoporous materials with a high aspect ratio (up to 10³:1) by increasing sufficiently the precursor exposure durations in comparison to the deposition on a flat surface [15]. Alternatively, in our method the pulsing of precursors in steps 1 and 3 is divided into two sub-steps (i.e. a micro-pulse and a micro-purge) that are repeated hundreds of times. This approach is necessary to enhance the penetration and evacuation of the precursor molecules into/out of the porous soot allowing homogeneous coating with the reactants also penetrating at the pores locating deep inside the substrate.

 

Fig. 1 ALD reaction cycle for deposition of Yb₂O₃ and Al₂O₃ in a porous silica soot.

Download Full Size | PDF

The micro-pulse and micro-purge optimum durations were determined in a separate study (see Fig. 1), in which a set of a few centimeters long soot pieces were doped by using different parameters whilst the doping level was determined by a material analysis. A similar approach could be used to optimize the N₂ purge time as well, used in steps 2 and 4. Here the purge time was set to 30 min. to ensure a complete purging, resulting in a total cycle time of 138 min./cycle for Yb₂O₃ and 92.5 min./cycle for Al₂O₃. This entails that doping of a soot of 320 µm thick layer with for instance 4 cycles of Yb₂O₃ and 4 cycles of Al₂O₃ needs an approximate time of 15-16 hours. However, it should be noted that the total time for the ALD reaction cycle could be considerably reduced by optimization of the N₂ purge time. For example, in [15] only 4 min. of N₂ purge time was found to be sufficient for uniformly coating nanopores with a high aspect-ratio of 5 × 10³:1. The aspect-ratio of a porous silica soot with a layer thickness of 320 µm is estimated to be roughly 10³:1 [8]. Thus, using 4 min. of N₂ purge time in the process used in Fig. 1 is likely to be sufficient, and the cycle time could be reduced to 86 min./cycle for Yb₂O₃ and 40.5 min./cycle for Al₂O₃.

The Yb₂O₃ growth per cycle (GPC) was determined to be 0.15 Å/cycle, which is typical when using β-diketonate reactants due to the large size of the molecule [14]. More details about the equipment and the Yb-doping process have been reported elsewhere [18]. In a similar manner, Al-codoping is performed by deposition of Al₂O₃, using the recipe presented in Fig. 1 with Al(CH₃)₃ and O₃ used as precursors [19].

The distribution of the dopants in the porous soot layer was studied in the direction transversal to the soot preform length by a material analysis. Figure 2(a) shows a typical result for a cross section from the soot analyzed by a Scanning Electron Microscope (SEM) equipped with an Energy Dispersive Spectrometer (EDS). In this particular case the soot was heavily doped by using 8 cycles of Al₂O₃ (i.e. a 0.88 nm film) and 8 cycles of Yb₂O₃ (i.e. a 0.12 nm film), in this order. The Yb and Al concentrations in atomic percentage are plotted as a function of the soot depth (x-axis), with x = 0 µm referring to the inner surface layer of the SiO₂, which corresponds to the center of the core in a drawn fiber. The results confirm that ALD enables coating of the complex structure of porous glass with excellent penetration also to the region next to the cladding. Furthermore, the Yb/Al concentration ratio remains approximately constant, although the dopant concentration at the top layer x = 26 µm is about 50% lower compared to the concentration in the region closer to the cladding x = 300 µm (see Fig. 2(b), showing the normalized Yb/Al ratio versus the soot depth).

 

Fig. 2 (a) An analysis of the radial distribution of dopants in a porous soot measured by an SEM/EDS. The soot piece was cut out from the position corresponding to the outlet of the gases in the reactor during the ALD doping. The average concentrations of Yb and Al are 0.15 and 1.51 at%. (b) Normalized ratio Yb/Al with respect to the average.

Download Full Size | PDF

In the third phase of the fabrication process the soot is transferred back to the MCVD-system, where the sintering and collapsing of the soot is carried out by a standard procedure. The soot preform is first dried from adsorbed water by 20 passes of the hydrogen burner at 145 mm/min. with increasing temperature from 600 to 1000 °C. Following drying, the soot is sintered in one burner traverse (10 mm/min.) at an estimated temperature of 1900 °C. Finally, the tube is pre-collapsed and closed by repeated slow burner traverses (total time 1.5 hours) at temperatures around 2000-2200 °C. All steps are carried out with an addition of chlorine (20 volume%) to the inner gas flow in order to suppress OH-contamination in the glass.

By this approach a soot piece of 120 mm in length, and previously doped with 4 cycles of Al₂O₃ (0.44 nm film) and secondly with 4 cycles of Yb₂O₃ (0.06 nm film), was transformed into a preform (preform A). Note that, compared to Fig. 2 the numbers of cycles used in the Preform A are reduced to half for both co-dopants, leading to a reduction in the Yb-concentration by the same factor. Figure 3(a) shows the refractive index profile (RIP) of the core measured for different values of the preform position, with z = 0 mm indicating the soot end at the inlet of the gases in the reactor during the ALD doping. The core RIP was calculated as an increase with respect to the cladding index. An undesired gradient in the refractive index along the length of the preform is observed, with maximum and minimum peak values of 2.5 × 10⁻³ (z = 0 mm) and 1.4 × 10⁻³ (z = 120 mm), respectively. We attribute this effect to a distortion of the Al₂O₃ deposition caused by small leakages of ambient air into the ozone line. The ambient air contains H₂O molecules, and the Al(CH₃)₃/H₂O process is a well-known alternative ALD process for deposition of Al₂O₃ [14]. Therefore, the un-controlled amount of H₂O reacts with the Al-precursor Al(CH₃)₃ and leads to a non-uniform Al₂O₃ deposition. This increases the refractive index especially at the inlet position. Unlike the Al-precursor, the Yb-precursor Yb(thd₃) does not react with the H₂O molecules and thus a uniform Yb₂O₃ deposition is achieved. The ozone line was repaired before the doping of another soot preform, named preform B, whose RIP is shown in Fig. 3(b). As a result, the variation of the mean refractive index along the 130 mm long preform was reduced to only ± 1.5%.

 

Fig. 3 Refractive index profiles (RIPs) at different positions along the preform. Refractive index change is calculated as an increase with respect to the refractive index of silica. (a) Preform A (b) Preforms B and C.

Download Full Size | PDF

A material analysis by EDS carried out for a cross section of the preform A showed that the distribution of Yb and Al in the radial direction follows the refractive index profile seen in Fig. 3(a), and that the Yb/Al ratio is approximately constant also after the collapsing process.

Moreover, RIPs of preforms A and B feature a depression in the index at the center of the core. This well-known ‘center dip’ in the refractive index is attributed to vaporization of dopants during the preform collapsing stages. Similar RIPs have been observed in Yb-doped phosphorous-silicate and germanium-silicate preforms fabricated by MCVD with solution doping [20, 21]. A new collapsing process was developed to obtain a more radially uniform dopant distribution. The Cl₂-concentration was reduced from 20 to 5 volume%, the number of drying steps was reduced from 20 to 10, and the sintering/collapse stages were carried out with no admixture of chlorine. The soot C in Fig. 3(b) was collapsed using this new process and it is observed that the ‘center dip’ of the RIP is successfully eliminated, although still a small non-uniformity in the deposition of Al₂O₃ was detected in the first tens of millimeters. It should be noted that the ALD results presented in this paper are all produced using soot pieces cut out from the same master soot preform, which is beneficial for comparison purposes, e.g., by ensuring a similar soot porosity.

3. Fiber results

The preform A corresponding to the results presented in Fig. 3(a) was drawn into a fiber with a length of 1 km. This fiber has an 18 µm core diameter and a total diameter of 125 µm, and a numerical aperture (NA) of the core varying from 0.07 to 0.08 along the fiber. To enable cladding pumping, the round-shape fiber was coated with a low index acrylate to result in an NA of 0.46 for the cladding. In order to determine the background attenuation the cut back method was implemented. Figure 4(a) shows the core attenuation spectrum. At 1200 nm, outside the Yb-absorption region, the fiber attenuation is observed to be only 20 dB/km, although a water peak is observed at 1380 nm.

 

Fig. 4 (a) Background attenuation spectrum. (b) Yb and Al concentrations along the fiber length.

Download Full Size | PDF

The dopant concentration distribution along the fiber is plotted in Fig. 4(b). Yb-concentration in atomic% was derived from Yb-absorption measurements carried out by the cut back method using a white light source, matching single-clad passive fibers to couple the light into the sample core, and an optical spectrum analyzer. Furthermore, high index oil and bend losses applied to the fiber sample were used to eliminate higher order modes and cladding modes that could disturb the measurement. The average Yb-absorption value measured at the 920 nm peak is 107 dB/m (which corresponds to an Yb concentration of 0.07 atomic%) with a maximum variation along the fiber of 4% from its average value (see Fig. 4(b)). Note that we already referred to this non-uniformity in the Al-doping in the previous section. The Al-concentration was determined from the mean refractive index increase Δn of the core, i.e. Δn = [Yb] × (molar refractivity of Yb) + [Al] × (molar refractivity of Al). The Δn was measured from the preform (data corresponds to the RIPs showed in Fig. 3(a)), and the molar refractivity of Yb₂O₃ and Al₂O₃ are 7.8 × 10⁻³ index change/mol% and 2.2 × 10⁻³ index change/mol%, respectively. The molar refractivities were calculated by fitting the index profiles for a number of preforms with Yb and Al concentrations determined by EDS. Additionally, the fluorescence lifetime was measured along the fiber giving an average value of 0.824 ms with a variation of ± 1%.

Photodarkening, the long-term time dependent increase of loss by pump irradiation [22], was studied on short fiber samples by cladding pumping at 915 nm while a core-propagating probe centered at 600 nm was detected by the lock-in method. A detailed description of the employed photodarkening setup has been previously presented [23]. Furthermore, the undesired decrease in the Al-concentration along the fiber was utilized to characterize the dependence of the photodarkening propensity with the doping level of Al. Figure 5(a) shows the temporal evolution of the absorption coefficient (Δα_PD) at 600 nm measured for fiber samples with different Al content. The length of the fiber samples was 2 cm and the pump power was 10 W at 915 nm. These conditions ensure a uniform population inversion along the sample, which was estimated to be approximately 70% by simulation. The Δα_PD after 40 hours of pumping is shown in Fig. 5(b) as a function of the Al-concentration.

 

Fig. 5 (a) Photodarkening experiments at 10 W of pump power (70% population inversion) characterized by the absorption coefficient change (Δα_PD) at 600 nm for different levels of the Al content. (b) Δα_PD after 40 hours of pumping versus the Al concentration expressed in atomic%.

Download Full Size | PDF

For comparison purposes, another MCVD soot preform was prepared using the recipe of the soot from section 2. The soot was subsequently doped by Yb and Al using the solution doping method, it was sintered and collapsed into a preform that was finally drawn into a fiber with core and cladding diameters of 20 µm and 140 µm, respectively. The collapsing process followed the recipe used for the preform A (see Fig. 3(a)), featuring a similar RIP with a ‘center dip’. Measurements of the background attenuation were performed showing a loss level of 150 dB/km at 1200 nm.

This fiber was used in photodarkening measurements for benchmarking. The fiber parameters and the measurement conditions are summarized in Table 1 . Taking into account that the fibers have different Yb-concentrations, a direct comparison of photodarkening propensity requires adjusting the pump power to reach similar density of excited Yb-ions ([Yb*]). The [Yb*] was calculated as the product of the density of Yb-ions and the population inversion. As a consequence of the very similar shape in the radial distribution of the Yb-concentration between the compared fibers (measured by EDS and confirmed by the RIP measurements), the [Yb*] profiles in the two measurements are also matched in the radial direction, and not just their average values. Note also that in Yb-doped fibers co-doped with Al only, within the range of Al concentrations used in this work, the Yb absorption cross section depends very little on the Al content. The situation is different with fibers co-doped also with P [20]. This method has been previously used in photodarkening benchmarking of fibers with different Yb-concentrations and it has been validated [24–26]. After 70 hours of pumping, still before saturation of the losses, the Δα_PD was significantly lower for the ALD fiber (see Fig. 6 and Table 1). Furthermore, a stretched exponential fit to the photodarkening curves estimates that prolonging the experiments until saturation (Δα_sat) could only increase the loss difference [27].

Table 1. Fiber parameters and measurement conditions of the photodarkening benchmarking

View Table

 

Fig. 6 Photodarkening benchmarking between an ALD-doped fiber and an MCVD-solution doping fiber with similar doping. [Yb*] is tuned to be approximately the same. Photodarkening measurements were fitted by a stretched exponential.

Download Full Size | PDF

To characterize the efficiency of the ALD fiber, we tested two pieces with different Al concentration (i.e. 0.30 and 0.47 atomic%) in a linear cavity fiber laser. The laser configuration is shown schematically in Fig. 7(a) . The Fabry-Perot cavity was formed by the 4% Fresnel reflection of the flat perpendicular cleave and a high reflectivity dichroic mirror. To separate the signal from the pump another tilted dichroic mirror was used at the input. A 15 m long fiber was cladding pumped by a free-space coupling of a pump laser diode emitting at 915 nm with a coupling efficiency of ~85%.

 

Fig. 7 (a) Fiber laser configuration (b) Laser output power as a function of the absorbed pump power. Inset: laser spectrum for a laser output power of 6.5 W.

Download Full Size | PDF

The laser signal output power is shown as a function of the absorbed pump power in Fig. 7(b) for the two fiber pieces used. The maximum output power is limited to 6.5 W by the available pump power. Note that the fiber has a round-shape inner cladding, which leads to a relatively low cladding pump absorption because of the limited pump-to-core overlap. In order to investigate the efficiency of the active fiber itself, we evaluated the laser performance by using the absorbed pump power that is calculated by the difference between the non-absorbed pump power and the launched pump power. The latter power is determined by the measurement of the pump power at the input after 20 cm of the fiber and the calculated absorbed pump power in the 20 cm fiber piece. The laser spectrum is shown in the inset of the Fig. 7(b) for the fiber piece with 0.30 atomic% of Al and a value of the laser output power of 6.5 W.

4. Discussion

The laser characteristics of the fiber show good efficiency with values for the slope between 78% and 80%, with the fiber laser operating at 1080 nm and a pump wavelength of 915 nm. The fiber background attenuation was observed to be low, at a level that is comparable to fibers fabricated by the conventional solution doping method. Furthermore, our method of incorporating the Yb-ions showed excellent homogeneity in its distribution along the 1 km long fiber. This is due to the outstanding uniformity of the ALD-grown films that is achieved also in large substrates.

The radial distribution of the dopants in the porous soot layer presented in Fig. 2(a) shows non-uniformity in the doping. Taking into account that the deposition in ALD mode entails a uniform coating of the substrate surface with films of Yb₂O₃ and Al₂O, the differences in the radial distribution may be indicative of a radial variation in the surface-area to volume ratio. Nevertheless, a more direct measurement method is needed to validate this assumption, e.g., a study of the variation of the surface area in the radial direction.

The photodarkening measurements, presented in Fig. 6 for the two fiber samples at the same level of [Yb*], indicate that the ALD-doped fiber suffers from less photodarkening in comparison to the fiber doped with solution doping. A more quantitative comparison could be achieved if both fibers would have even more similar dopant concentrations, and not only the same [Yb*]. This should be considered for future experiments. It should also be noted that the Al concentration is significantly higher in the solution doping fiber, which should actually reduce the photodarkening propensity as shown in Fig. 5(b) and in ref [25]. We hypothesize that the good photodarkening resistance of the ALD fiber is a consequence of the reduction in the Yb-ion clustering due to the homogeneous Yb-ion incorporation by the ALD method before sintering and collapsing.

The plot of the photodarkening losses after 40 hours of pumping versus the Al-concentration can be fitted approximately by a linear curve indicating that photodarkening precursors can be reduced almost linearly in relation with the Al content. A similar result was previously reported for Yb-doped fibers fabricated by the MCVD and solution doping method using a broader range of Al concentration levels [25].

The conventional MCVD-solution doping is a multi-step process consisting of deposition of the porous silica layer, cutting and soaking in water containing RE-salts, reassembling of the soot in the lathe, and sintering and collapsing. By repeating iteratively those steps (excluding collapsing) a doped core with multiple layers can be achieved. However, the maximum number of layers is typically restricted to around 10 in commercial fibers, limiting the accuracy and flexibility of the doping and the refractive index profile required for fabrication of advanced fibers. Additionally, the throughput time is fairly long due to the multi-step process, which makes the fiber development work slow and expensive. On the contrary, ALD is a gas-phase doping technique that could easily be integrated with MCVD. Such an in situ doping technique will circumvent the repeated preform removing and reassembling steps used in solution doping and therefore allow for larger number of layers, increase the usable length of the preform and result in a higher yield. Additionally, a multi-layer process would entail deposition of thinner layers of porous SiO₂. In ALD-doping the required exposure time of the precursors has been shown to be proportional to L², with L being the pore length [15]. Therefore, the overall doping time of a soot with 320 µm thick SiO₂ layer doped with 4 cycles of Al₂O₃ and 4 cycles of Yb₂O₃ could potentially be well reduced from 15 to 16 hours to a fraction of an hour by iterative deposition of e.g. 10 layers of SiO₂. Furthermore, an integrated ALD-MCVD process has the potential to minimize contaminations of e.g. water, since the as-deposited soot is doped from the gas phase without prior exposure to ambient atmosphere.

Moreover, the ALD method has been proven to be suitable for thin film deposition of other RE-ions in their oxidized state, such as Er, Tm, Nd and Bi. Additionally, typical fiber co-dopants like Ge, P and Ce can also be incorporated by ALD [14]. Al and P are important because they are typically used to improve the RE-ion solubility, and Ce (and also P) has been shown to be useful for improving the photodarkening resistance of the laser fibers [25, 28]. On the other hand, however, these co-dopants are responsible also for increasing the fiber background loss attenuation, complicating the control of the refractive index or reducing the Yb absorption and emission cross section [25, 28]. For that reason it is our opinion that ALD, compared to the other methods, is of particular importance for avoiding the heavy use of co-dopants, since it has the potential to considerably promote diffusion of the RE-dopants and therefore reduce the formation of clusters.

5. Conclusions

We have presented ALD as a new fabrication method of producing Yb-doped fibers. We have shown that ALD is an effective method of incorporating Yb-ions, in the form of thin films, at the surface of the porous silica layer of an MCVD-soot. The study of the radial dopant distribution in a section of a soot prior to sintering revealed that ALD is capable of doping the full depth of a 320 μm thick silica soot layer. The ALD recipe optimized for enhancing the doping of porous soot layers was presented in detail and it could be also useful in other applications that require RE-doping of substrates with extremely high aspect ratio, such as nano-cavities.

By optimization of the ALD-reactor, we succeeded in obtaining a mean refractive index along the preform with a variation of only ± 1.5% over a length of 130 mm. Furthermore, the collapsing process was also modified to eliminate the center dip of the RIP. By combining the benefits of these two improvements it is possible to achieve an accurate control of the refractive index profile, meeting the needs of future advanced high power fiber lasers.

The fiber results showed that we obtained a homogeneous Yb-doping along the fiber axis over a 1 km long fiber, and a good performance in terms of background losses. Furthermore, we have shown that the ALD fiber has less photodarkening propensity compared to a fiber with similar doping fabricated by the MCVD-solution doping technique with the same MCVD recipe. The study of the photodarkening propensity versus the Al-content revealed that photodarkening diminishes with the enhancement of the co-dopant level, as has previously been shown with fibers prepared with the solution doping method. Also, the ALD fiber was tested in a fiber laser configuration and it showed a slope efficiency of 80%.

Finally, the versatility of the ALD method enables the development of a fabrication process that integrates ALD with the MCVD system without the need to remove the preform from the MCVD system. This potentially allows accurate in situ doping with significantly reduced fabrication time, which in turn results in reduced production cost of Yb-doped fibers.