Erbium-Doped Fibers Designed for Random Single-Frequency Lasers Operating in the Extended L-Band

Abstract

The paper presents the results of developing Er-doped optical fibers for creating random single-frequency lasers in the wavelength range of 1570–1610 nm. The possibility of broadening the luminescence band of Er³⁺ ions in silicate glasses in the long-wavelength region of the spectrum by introducing a high concentration of P₂O₅, as well as by additional doping with Sb₂O₃, is investigated. It is found that both approaches do not improve the dynamics of luminescence decay in the L-band. In addition, Er₂O₃-GeO₂-Al₂O₃-SiO₂ and Er₂O₃-GeO₂-Al₂O₃-P₂O₅-SiO₂ glasses were studied as the core material for L-band optical fibers. The developed fibers exhibited high photosensitivity and a high gain of 5 and 7.2 dB/m, respectively. In these fibers, homogeneous arrays of extended weakly reflecting Bragg gratings were recorded directly during the fiber drawing process. Samples of arrays 5 m long and with a narrow reflection maximum at ~1590 nm were used as the base for laser resonators. Narrow-band random laser generation in the wavelength region of 1590 nm was recorded for the first time. At a temperature of 295 K, the laser mode was strictly continuous wave and stable in terms of output power. The maximal power exceeded 16 mW with an efficiency of 16%.

1. Introduction

The development of single-frequency fiber lasers (SFLs) in the spectral range of 1570–1650 nm is extremely important for a number of important practical applications, such as precision gas analysis (CO₂, CH₄, H₂S, etc.), selective surgical action on biological tissues with a high content of hydrocarbon groups, LIDAR systems, distributed fiber sensors, and online monitoring of the state of fiber-optic communication lines [1,2,3]. As it is known, the key element of the fiber laser resonator is the active fiber, which acts as a gain medium. Indeed, the properties of the active fiber largely determine the characteristics of the entire laser system. The most common SFLs have a short resonator of several centimeters long, which is the easiest way to implement the selection of a single longitudinal mode [4,5,6]. To obtain generation in a short fiber resonator, it is necessary to ensure efficient absorption of pump radiation, which is usually achieved due to the high concentration of active ions in the core glass. In the wavelength range of 1570–1650 nm (L-band), erbium-doped fibers show the greatest promise in terms of laser generation [7]. However, in this spectral range, the gain of erbium fibers rapidly decreases with increasing wavelength [8,9], which makes it extremely difficult to develop lasers with a short resonator. In order to make the gain drop in the long-wavelength region more gradual, various methods for modifying the composition of the fiber core glass have been proposed. For example, in [10], the composition of Sb₂O₃/SiO₂ was studied, and it was shown that in an erbium fiber with the addition of antimony oxide in the core, the gain band in the long-wavelength region of the spectrum was extended to wavelengths of ~1620 nm. Another and quite effective approach was the development of a complex, multicomponent core glass matrix, in which the main dopant was phosphorus oxide—P₂O₅ [11,12]. Based on the developed fibers, lasers with a standard Fabry–Perot resonator were manufactured, capable of generating wavelengths up to 1630 nm. However, it should be noted that the length of the laser resonators was a few meters [11] or tens of centimeters [12]. In such relatively long resonators, obtaining and maintaining a single-frequency generation mode requires the mandatory inclusion of several additional selection elements in the optical circuit and therefore represents a rather complex technical task [13].

To achieve single-frequency lasing using a long section of a fiber, a random resonator can be used as an alternative to the classical Fabry–Perot fiber resonator or a ring resonator [14,15]. Generally, feedback in a random resonator is achieved due to Rayleigh scattering based on a set of reflective elements distributed along the entire length of the fiber [16]. A fairly simple and effective way to provide feedback is to use a uniform array of weakly reflective Bragg gratings recorded along the entire length of the fiber during its drawing [17]. If this array of gratings is recorded directly in the active fiber, a ready-made random laser resonator will be obtained, generating at the wavelength of maximum reflection of the whole array [18,19]. Moreover, under the action of pump radiation, dynamic inverse population gratings are formed inside such a laser resonator (Random-FBG), due to which automatic selection of one of the longitudinal modes is performed in the resonator; or in other words, conditions are created for generation in a single-frequency mode [20,21]. An undoubted advantage of the Random-FBG laser is the possibility of obtaining laser emission in spectral ranges where the gain is relatively low—at the edge of the luminescence bands of active ions. For example, in germanosilicate optical fibers doped with bismuth, the maximum of the luminescence band of bismuth active centers is located at approximately a wavelength of 1720 nm [22]. Quite recently, in a Random-FBG laser based on a bismuth-doped fiber, narrow-band laser generation at a wavelength of 1673 nm was obtained for the first time, at which the luminescence intensity is an order of magnitude lower than that at the maximum [23].

For the development of erbium SFLs designed for the L-band, the use of Random-FBG resonators therefore has good prospects. The design of a Random-FBG resonator potentially makes the generation threshold possible due to the long length of the fiber, even with a relatively low erbium concentration. For the successful practical implementation of such an approach, low optical losses in the fiber are extremely important, which depend on the purity and homogeneity of the core glass. The second, but no less important, criterion for fibers in the context of the task of developing a Random-FBG laser is the level of initial photosensitivity to UV radiation sufficient for inducing Bragg gratings. For example, aluminosilicate (Al₂O₃/SiO₂) and aluminophosphosilicate (Al₂O₃/P₂O₅/SiO₂) glasses are optimal matrices for doping with high erbium concentrations without forming ion clusters [24,25], which significantly worsen the gain characteristics of the fibers [26]. However, fibers with a core of Al₂O₃/SiO₂ or Al₂O₃/P₂O₅/SiO₂ compositions have extremely low initial photosensitivity to laser UV radiation [27], which significantly hinders the manufacture of Random-FBG laser resonators based on them. On the other hand, according to the same work [27], the addition of germanium oxide (GeO₂) in relatively small concentrations (5–10 mol%) and even to pure SiO₂ leads to an increase in the photosensitivity of the glass by an order of magnitude. It can therefore be expected that the GeO₂/Al₂O₃/SiO₂ (GAS) and GeO₂/Al₂O₃/P₂O₅/SiO₂ (GAPS) matrices should also be photosensitive.

The main goal of this work was to develop erbium-doped fibers for use in L-band Random-FBG laser resonators. The core glass should satisfy the following set of criteria: increased photosensitivity, transparency and homogeneity, and a high Er₂O₃ concentration limit. An erbium-doped fibers developed on the basis of such glass should demonstrate a sufficiently high gain in the L-band over fiber lengths of several meters. In addition, the refractive index difference between the core and the quartz cladding (Δn_core-clad) of the fiber should be in the range of approximately 0.005–0.015 to minimize splicing losses with standard fiber components used in the pumping and laser output circuit. Thus, based on the above requirements for the fiber core and relying on the analysis of literature data, potentially suitable glass compositions were selected for the studies in this work: GeO₂/P₂O₅/SiO₂ (GPS), Sb₂O₃/P₂O₅/SiO₂ (SbPS), Sb₂O₃/GeO₂/P₂O₅/SiO₂ (SbGPS), GeO₂/Al₂O₃/SiO₂ (GAS), and GeO₂/Al₂O₃/P₂O₅/SiO₂ (GAPS).

2. Materials and Methods

The preforms of the optical fibers were fabricated by MCVD method using HeraeusSuprasil F300 silica reference tubes. High-purity halides SiCl₄, GeCl₄, CCl₄, POCl₃, AlCl₃, SbCl₃, and Er(thd)₃ (thd = 2,2,6,6-tetramethyl-3,5-heptanedionate) were used as starting compounds. Containers with low-volatile AlCl₃, SbCl₃, and Er(thd)₃ were heated to 130, 120, and 160 °C, respectively, and the vapors were delivered into the glass deposition zone through separate heated lines to prevent vapor condensation and interaction of the precursors with each other. The core was fabricated using the technique described in detail in [19,28,29] with separate deposition of the porous matrix layer, its impregnation with Er₂O₃ from the gas phase, and fusion of the impregnated porous layer into transparent glass. This technique was used to fabricate a series of preforms with an Er-GPS core in the dopant concentration range of 0.007–0.038 mol% Er₂O₃, 1.6–2.7 mol% GeO₂, and 12.5–16 mol% P₂O₅. The introduction of a higher phosphorus concentration was limited by high stresses in the preform, which led to its cracking due to a large difference in the TEC of the GPS core and quartz cladding (tube) materials. To our knowledge, such a high level of doping of quartz glass with P₂O₅ has been achieved for the first time for MCVD technology. Preforms with GAS (5 mol% GeO₂ and 5 mol% Al₂O₃) and GAPS (4.5 mol% GeO₂, 4.5 mol% Al₂O₃ and 6 mol% P₂O₅) cores with Er₂O₃content increased to 0.07 mol% were also fabricated. When choosing the erbium oxide concentration, we were guided by the optimal resonator length, which for erbium Random-FBG lasers is in the range of about 5 to 8 m [18,30].

Antimony has a higher affinity for chlorine than for oxygen; therefore, unlike the other selected precursors, under MCVD conditions, the conversion of SbCl₃ to Sb₂O₃ does not occur due to the lower stability of the oxide. Moreover, in the presence of chlorine, the reverse reaction occurs—chlorination of Sb₂O₃, due to which a complete loss of antimony from the deposited porous layer (in the form of SbCl₃) is observed during its fusion in the presence of a chlorine agent (e.g., CCl₄). It was necessary to completely exclude contact of amorphous Sb₂O₃ with chlorine. Another difficulty in doping quartz glass with antimony is the high volatility of Sb₂O₃ under MCVD process conditions.

In this regard, the technique for antimony-containing glasses was changed. When the burner moved towards the flow of the vapor–gas mixture, a thin amorphous layer of soot (PS or GPS) was deposited. The impregnation and fusion stages of the soot layer were combined in one process. For this purpose, Er(thd)₃ and SbCl₃ vapors were mixed with oxygen and were delivered into the tube, and the layer was fused at a relatively low temperature. Due to the low thickness of the matrix layer and the high content of P₂O₅ in it (11–12 mol%), the fusion temperature typically did not exceed 1600 °C (i.e., the temperature was reduced by ~300 °C). In addition, water and chlorine formed during the conversion of Er(thd)₃ and SbCl₃ into oxides were almost completely bound to each other into hydrogen chloride, which left the deposition tube. This approach made it possible to simultaneously introduce high concentrations of antimony into the synthesized glasses and significantly reduce the amount of hydroxyl ions in them. Using this technique, preforms with cores of different compositions (SbS, SbPS, and SbGPS) containing from 0.6 to 3.8 mol% Sb₂O₃ were produced. The introduction of higher antimony concentrations, as in the case of the PS matrix, was limited by the destruction of the preform.

The refractive index profile of the obtained preforms was measured using a Photon Kinetics 2610 preform analyzer. Scanning was performed in three projections at an angle of 120 degrees to check the circular symmetry and in several axial positions to determine the homogeneity of the core glass along the preform length. For the initial and rapid evaluation of the optical properties of the developed core compositions, multimode fibers with a standard outer diameter of 125 μm and a core diameter of ~20 μm were first drawn from the preforms. The elemental composition of the core glass of these fibers was determined using a JEOL 5910LV electron microscope. However, the sensitivity of the X-ray analyzer is insufficient for reliable detection of Er₂O₃ concentrations below 0.1 mol%. The Er₂O₃ content was determined from the intensity of the absorption peak of Er³⁺ ions with a maximum at a wavelength of 1535 nm (1 mol% corresponds to 780 dB/m [31]), measured in a multimode optical fiber. The purity and homogeneity of the synthesized core glass were estimated from the level of minimal optical losses measured in this optical fiber at a wavelength of 1200 nm (far from the absorption bands of Er³⁺ ions).

For the development resonators of the Random-FBG lasers, single-mode fibers with a standard diameter of 125 μm were used. Before drawing single-mode fibers, the original preforms were jacketed to secure a cutoff wavelength of ~950 nm. The process of fabricating random resonators consisted of creating extended FBG arrays directly during the fiber drawing process using a setup based on a pulsed excimer KrF laser (248 nm) [17]. The laser pulse repetition rate and the fiber drawing speed were synchronized so that each FBG with a fixed length of 10 mm was written in one laser pulse. The standard technique of UV irradiation of the fiber through a phase mask was used to write the FBG. In this work, phase masks with a period of 1070 and 1098.4 nm were used, forming FBG with a reflection peak at wavelengths of about 1550 and 1590 nm, respectively.

In this work, two types of FBG fiber arrays were used: half-filled (50%) with gratings with the reflection center at wavelengths of ~1550 nm and completely (100%) filled with gratings reflecting in the region of ~1590 nm. Grating arrays of the first type (50% filled) with a length of 20 m were used only for measuring the homogeneity and contrast of the recorded FBG. These measurements were performed using a Luna 4600 frequency reflectometer with an operating wavelength range of 1520–1570 nm. Grating arrays of the second type (100% filled) with an initial length of 20 m were used in the designed laser scheme for obtaining random laser generation in the L-band (specifically, at a wavelength of ~1590 nm).

Experiments on studying the laser properties of FBG arrays were preceded by a comparative analysis of their reflectivity spectra near a wavelength of 1590 nm. For this purpose, the fiber-optic scheme shown in Figure 1 was used. Radiation from a broadband superluminescent diode “3” was introduced into the FBG array sample “1” through a 50% splitter “2”. A small part of this radiation, reflected by the FBG array, passed again through the splitter and was directed to the input of the optical spectrum analyzer “4”. The spectrum of the reflected radiation was recorded with an optical resolution of 0.1 nm.

To determine the optimal length of a fiber laser resonator, it is extremely important to know the value of the small-signal gain in the active fiber. Thus, a negative or even zero value of the gain indicates the fundamental impossibility of achieving laser generation conditions at a given wavelength. In this work, the gain in erbium-doped fibers was measured in the wavelength range of 1525–1610 nm (combined C and L bands) using the technique used to characterize erbium-doped fiber amplifiers (EDFAs) [32,33]. Two independent APEX platforms AP1000-5 and AP1000-8 equipped with tunable semiconductor single-frequency laser units AP3350A (1525–1575 nm range) and AP3352A (1575–1610 nm range), respectively, were used as reference weak-signal sources. The studied optical fibers with an initial length of ~1 m were pumped at a wavelength of 976 nm by a diode laser with an output optical power of up to 600 mW.

The study of the laser characteristics of Random-FBG samples with a reflection maximum near the wavelength of 1590 nm was carried out under the same conditions, at room temperature (295 K), using the fiber-optic circuit shown in Figure 2. The IR pump radiation (974.5 nm) was introduced into the studied resonator sample “1” using multiplexer “2”. The transmitted (unabsorbed) pump radiation was output using the second similar multiplexer. The same multiplexers were used to output the laser radiation of ~1590 nm generated by a random resonator. Since the 1590 nm laser radiation could propagate both codirectionally and in the opposite direction to the pump radiation, two independent fiber-optic ports were used to record it, marked in Figure 2 by the symbols “F” (“forward”) and “B” (“backward”), respectively. Optical isolators “4” were installed in front of the ports in order to prevent the reflected signal from entering the laser circuit. An Ando AQ6317B optical spectrum analyzer with a maximum resolution of 0.01 nm was used to record the laser radiation spectrum. The laser and pump power levels were measured using a power meter. A scanning Fabry–Perot interferometer (VitaWave SCC-750) with a free dispersion region of 750 MHz and a resolution of 12 MHz was used for a more detailed study of the laser radiation mode composition. The use of an optical spectrum analyzer in combination with a scanning Fabry–Perot interferometer allows one to obtain complete information on the excited modes in fiber laser resonators with lengths from centimeters to meters, as well as to experimentally detect and confirm the establishment of a single-frequency generation mode. To record pulsed or continuous generation modes of the resonator, the output radiation was fed to the input of a high-speed photodetector based on an InGaAs pin photodiode, the signal from which was fed to a digital oscilloscope.

3. Results

3.1. Optical Fibers with Er-GPS Core

The high photosensitivity of GPS glasses has previously allowed the successful development of ytterbium- and erbium-doped optical fibers for random fiber lasers emitting near the maximum gain of active ions [19,30]. For this reason, we primarily investigated the possibility of manufacturing Er-doped optical fibers based on the GPS matrix. In contrast to ytterbium, among the rare earth elements used for active fibers, erbium dopant is the most sensitive to the process of active ion clustering due to the complex energy level scheme in which energy exchange (up-conversion) with nonradiative relaxation occurs between two closely located erbium ions. In this regard, it is very important to avoid even a minimal fraction of erbium clusters in glass when manufacturing Er-doped optical fibers. GeO₂ is an analog of SiO₂ and the contribution of 1.6–5 mol% GeO₂ to the solubility of Er₂O₃ in our case can be neglected. P₂O₅ is the best co-doping additive for Yb₂O₃ [29] but is not so effective for dissolving Er₂O₃ in silicate glasses [25]. Nevertheless, the presence of more than 10 mol% P₂O₅ in GPS glass should theoretically be quite sufficient to suppress the clustering of Er³⁺ ions at their concentrations below 0.01 mol%. We introduced the highest possible concentration of P₂O₅ up to 16 mol% into the GPS matrix not only for dissolving Er₂O₃ but also to make the decay of erbium luminescence in the long-wavelength range more significant.

Thus, in total four erbium-doped preforms based on the GPS matrix were manufactured: LD663-1, LD663-2, LD664, and LD666. The main data on the elemental composition of the core of these preforms are given in

Table 1. Main characteristics of Er-GPS preforms.

Preform ID	Δncore-clad	P2O5, mol%	GeO2, mol%
LD663-1	0.014	12.5	1.6
LD663-2	0.014	12.5	1.6
LD664	0.014	12.5	1.7
LD666	0.019	17.2	2.7

.

As an example, Figure 3 shows the data for the LD666 preform sample with the maximum content of dopants within the manufactured LD663–LD666 series. Some difference in the shape of the refractive index profile (Figure 3a) measured at different angles of rotation of the preform indicates a slight ellipticity of the core. The deviation from the circular geometry of the core is associated with a large difference in the temperature course of the viscosity among the silica cladding and the highly doped Er-GPS glass of the core at the stage of consolidation of the tube into a rod (collapse and sealing process). Due to the large difference in the properties of the core and silica cladding materials, it is very difficult to manufacture an Er-GPS preform with the ideal circular geometry. The average value of Δn_core-clad can be estimated as ~0.019, which, within the measurement error, agrees with the molar refractive indexes (C_mol%) of P₂O₅ and GeO₂ (0.9 and 1.45 × 10⁻³ × C_mol%, respectively), i.e., with the composition of Er-GPS glass.

The level of minimum (“background”) optical losses in a multimode fiber at a wavelength of ~1150 nm, drawn from the LD666 preform, was only 5–6 dB/km, which indicates high purity and homogeneity of the synthesized glass. In other samples of drawn multimode fibers of the Er-GPS series (LD663-1 and 2, LD664), the minimum losses were even lower (1–2 dB/km), which is due to a lower concentration of doping. As is known, the level of minimum optical losses in single-mode fibers increases slightly relative to multimode fibers with the same core glass composition. Eventually, the value of minimum optical losses in all single-mode Er-GPS fibers did not exceed 10 dB/km.

The spectral dependence of the gain in the LD663-1 and LD663-2 fibers, which differ only in the Er₂O₃ concentration, is shown in Figure 4. The gain has the highest value near the wavelength of 1535 nm, which coincides with the region of maximum absorption. As we move away from the maximum to the long-wavelength region, the gain in the active fibers smoothly decreases; however, in the wavelength range of 1580–1600 nm, the dependence of the gain on the wavelength has a more or less flat character and the same shape of the spectral curve for both fiber samples. In this case, the gain of the LD663-2 fiber has a low value of ~0.65 dB/m, and for the LD663-1 fiber, its value is almost twice as high as ~1 dB/m.

The contrast value in the arrays of low-reflecting gratings recorded during drawing in Er-GPS fibers was 41–43 dB (Figure 5). This indicates sufficient photosensitivity of Er-GPS glass to UV radiation. The reflection spectra of the grating arrays had a symmetrical shape, with the exception of the array recorded in the LD666 fiber. As can be seen from Figure 5, the reflection peak of the LD666 array relative to the LD664 array has a clear asymmetric broadening. The distortion of the spectrum is explained by the high fluctuation of the parameters of the recorded gratings, which is clearly visible on the reflectogram. The inhomogeneity of the FBG array in the LD666 fiber is due to the ellipticity of the core, in our opinion. During the drawing process, the optical fiber rotates around its axis at an arbitrary angle, and, depending on the angle of rotation relative to the laser beam direction, the conditions for guiding the refractive index grating in the core are changed. Since an array of several meters long contains several hundred gratings, the shape and width of the reflection spectrum will differ significantly from the reference due to accumulated inhomogeneities.

Signs of laser generation at wavelengths of ~1590 nm were not registered in any of the Er-GPS series samples. Even at the maximum value of the input pump power of ~200 mW, only the luminescence signal of Er³⁺ ions was present in the emission spectrum. Figure 6 shows the luminescence spectra of the LD663-666 series optical fibers, normalized to the maximum of the signal intensity.

Analysis of the spectra in Figure 6 revealed a rapid decline in the long-wavelength luminescence edge in the fibers with a higher erbium concentration. Thus, the spectrum curve corresponding to the fiber with the lowest erbium concentration (LD663-2) lies above all the other samples. It is noteworthy that the long-wavelength erbium luminescence edge in the LD666 fiber has the fastest decline, despite the record high content of co-doping elements in its core, the concentration of which was significantly higher than in the other fibers of the series. This result once again confirms the low efficiency of P₂O₅ and GeO₂ for the effective dissolution of Er₂O₃ in silicate glasses. Thus, it was experimentally established that the composition of the Er-GPS core is not optimal for the manufacture of L-band laser resonators.

3.2. Optical Fibers with a Core Doped with Sb₂O₃

Next, the possibility of increasing the luminescence cross-section of Er³⁺ ions in the long-wavelength range by additional doping with Sb₂O₃ was investigated. We preliminarily manufactured a concentration series of preforms with a Sb₂O₃-SiO₂ core (Er₂O₃ was not introduced there), measured the refractive index profile, and analyzed the composition of the synthesized glass. Under the conditions of the MCVD process, the maximal concentration of Sb₂O₃ in the core was 3.8 mol% and was limited by preform cracking. By comparing the obtained data, the dependence of the Δn_core-clad value in quartz glass on the antimony content was obtained and an extremely high molar refractivity index of Sb₂O₃ was established: ~0.0079 per 1 mol% (Figure 7).

The value of minimum optical losses in a multimode fiber with an average concentration of antimony in this series (2 mol% Sb₂O₃) was 18 dB/km (Figure 8), which is 4 times higher than in the most concentrated Er-GPS fiber LD666 with the maximal content of the main dopants.

Two Er-doped preforms containing Sb₂O₃ (Er-SbPS) were also fabricated: LD677 and LD678. The elemental composition data of the core of these preforms are given in

Table 2. Main characteristics of Er-SbPS preforms.

Preform ID	Δncore-clad	P2O5, mol%	GeO2, mol%	Sb2O3, mol%
LD677	0.026	11.7	2.8	1
LD678	0.016	9.2	---	0.6

.

The effect of antimony doping into GPS glass can be assessed by comparing the waveguide characteristics of LD663-1 and 677-1 fibers. Both fibers had the same core diameter of ~5 μm, the same erbium concentration, and a similar matrix composition, differing essentially only in the presence of 1 mol% Sb₂O₃ in the core of LD677. As can be seen from Figure 9a,b, a relatively small addition of Sb₂O₃ significantly increases the refractive index of the core glass. It is important to note that the increase in Δn_core-clad is not additive and its value exceeds the total refractive contribution of dopants in the glass. It can be assumed that the network of multicomponent glass largely consists of complex structural complexes that bind antimony, phosphorus, and silicon atoms. For example, such complexes exist in the APS matrix, in which the refractive index is significantly lower than the refractive indices of pure aluminosilicate and phosphosilicate glasses [34].

From the point of view of using the fiber as the basis of a laser resonator, high Δn_core-clad is a disadvantage, since in this case, it is necessary to reduce the core diameter to maintain the single-mode regime, which ultimately leads to the complexity of connectivity with standard fiber components. In addition, the introduction of antimony reduces the viscosity of the glass and significantly complicates the manufacture of a fiber preform with good circular core geometry. As shown above, using the LD666 fiber as an example, deviations in the core geometry lead to a broadening of the reflection spectrum of the FBG array. From the comparison of the optical loss spectra of the LD663-1 and LD677 fibers (Figure 9c), it is evident that the addition of antimony to the GPS glass also has a negative effect on the level of “background” optical losses, which increases from 7.5 to 50–60 dB/km. This factor is important from the point of view of the emission characteristics of the fiber in a resonator with a length of several meters, since high “background” losses increase the laser generation threshold and reduce its efficiency.

Despite the identified disadvantages of the SbPS matrix, the main objective of this part of the research was to increase the erbium gain in the long-wavelength region of the spectrum. The spectral dependences of the gain for the LD677 and LD663-1 fibers are shown in Figure 10. The position of the gain maximum for both fibers is the same and is 1535 nm, which coincides with the maximum of the Er³⁺ ions luminescence signal in GPS glass. Near the maximum, the gain values in the LD677 fiber are on average 25% higher than in LD663-1 fiber, and the dynamics of the decrease to the wavelength of 1570 nm is smoother, which may be due to the addition of antimony. However, in the longer-wavelength range, the spectral dependences of the gain for both fibers practically overlap each other.

The performed complex studies show that additional doping of Er-GPS glass with ~1 mol% Sb₂O₃ leads to a significant increase in the level of “background” losses in the fiber as well as a rise in the difference in the refractive indices of the core and cladding, which complicates the use of the fiber in telecommunication laser circuits. At the same time, although the addition of antimony slightly slows down the gain decay at the edge of the C-band, near the wavelength of 1560 nm, it does not affect the gain characteristics of the fiber in the L-band of wavelengths.

In conclusion of this section, we consider the characteristics of another Er-fiber doped with Sb₂O₃—LD678. GeO₂ was not introduced into the core of this preform, and the amount of antimony and phosphorus oxides was reduced (Table 2). This fiber was manufactured to evaluate the effect of antimony dopant on the photosensitivity and transparency of phosphosilicate glass in the absence of GeO₂. As can be seen from Figure 11, despite the relatively small addition of Sb₂O₃, Δn_core-clad has increased significantly in the fiber. In LD678, the “background” optical losses have also increased relative to the Er-GPS fibers of the LD663-666 series but to a lesser extent than compared to the Er-SbGPS fiber LD678. The smaller increase in the value of losses in the LD678 sample can be explained by the simpler matrix composition and at the same time the lower concentration of antimony oxide.

Despite the relatively low concentration of Sb₂O₃, the value of the recording contrast of gratings in the array based on the LD678 fiber was ~50 dB, which indicates the high efficiency of antimony as a dopant that increases photosensitivity to UV radiation. It should be noted that the laser generation threshold at a wavelength of ~1590 nm could not be achieved in resonators based on LD677 and 678 fibers, which is most likely due to the low value of the gain coefficient of ~1 dB/m.

3.3. Fibers with Er-GAS and Er-GAPS Core

The gain of erbium-doped fibers in the L-band can be increased by increasing the Er₂O₃ concentration but only under the condition of minimizing the clustering of Er³⁺ ions. The most effective matrices for suppressing clustering are those containing aluminum oxide—aluminosilicate and aluminophosphosilicate [24]. Based on this fact, in addition to the previously manufactured samples, two additional preforms with increased erbium concentration were fabricated: LD681 and LD682 based on GAS and GAPS matrices, respectively. Al₂O₃ was used as a co-doping additive, as the dopant most effectively dissolving Er₂O₃ in silica, and the GeO₂ additive was introduced to impart higher photosensitivity to the glasses.

As can be seen from

Table 3. Main characteristics of the fabricated Er-GAS and Er-GAPS preforms.

Preform ID	Δncore-clad	GeO2, mol%	Al2O3, mol%	P2O5, mol %
LD681	0.017	5	5	---
LD682	0.006	4.5	5	6

, both the preforms contained approximately the same amount of Al₂O₃ and GeO₂. However, the refractive index of the LD682 core turned out to be almost 3 times lower than in LD681, which is due to the presence of phosphorus in the LD682 core and the binding of all Al³⁺ ions in the AlPO₄ group. According to [34], AlPO₄ complexes do not contribute to the increase in the refractive index of doped glass, unlike Al₂O₃, which has a relatively high refractive index (~0.0023 per 1 mol% Al₂O₃).

In fact, in the LD681 and LD682 samples, the erbium concentration was significantly increased. At the same time, as can be seen from Figure 12, the “background” optical losses in single-mode fibers (cutoff wavelength ~0.95 μm) remained low (~5 dB/km), at the level of low-doped Er-GPS fibers, which indicates high transparency of the synthesized glasses. Obviously, the introduction of a high concentration of Al₂O₃ made it possible to avoid clustering of Er³⁺ ions, despite their high concentration in relation to the samples previously studied in this work. As can be seen from Figure 12, the intensity of the Er³⁺ absorption peak in fibers with GAS and GAPS matrices was more than 30 dB/m, which is at least 3 times greater than in fibers with Er-GPS (LD663, LD664) and Er-SbGPS (LD677, LD678) glass cores. These two key factors can explain the dramatic increase in the gain in the 1590 nm wavelength region, compared to the samples of the previous series: 7 dB/m in LD681 and 5 dB/m in LD682. The contrast of the grating recording in 50% filled arrays made on the basis of LD681 and LD682 was about 50 dB. This value indicates the high initial photosensitivity of Er-GAS and Er-GAPS glasses to UV radiation and is quite sufficient for using the grating array as a random resonator of several meters long.

3.4. Implementation of Single-Frequency Random L-Band Lasers

The basis for random resonators was extended arrays of low-reflecting gratings with 100% filling, since the level of the return signal in them is maximum and 2 times higher than in arrays with a recording density of 50%. Experiments to study the generation properties of the manufactured arrays were proceeded by a comparative analysis of their reflection spectra near a wavelength of 1590 nm. The reflection spectra of 5 m long samples, measured with an optical resolution of 0.1 nm, are shown in Figure 13.

The weakest reflection signal was recorded in samples LD663-1 and LD663-2, and the most intense one was recorded in the sample LD682. Based on this, it can be concluded that the average FBG contrast in the LD682 fiber was also higher than in the other studied fibers. This can be explained by the extremely high photosensitivity of Er-GAPS glass to UV radiation in a series of studied samples, due to the presence of phosphorus and an increased amount of germanium in the matrix at the same time. Similarly, in samples LD677 and LD678 with the addition of Sb₂O₃, the signal reflected from the array of gratings also has a comparatively high intensity, which indicates a positive effect of this dopant on the resulting photosensitivity. The shift of the Bragg peak position of the gratings in these samples was by a few nanometers toward the long-wavelength region relative to samples LD663-1, LD663-2, and LD682. It is obviously explained by higher Δn_core-clad values.

As was already mentioned above, in samples LD663-1, LD663-2, LD677, and LD678, it was not possible to reach the lasing threshold in the 1590 nm region, even at the maximum values of the input pump power of ~200 mW and the resonator length increased to 10 m. Only the characteristic luminescence of Er³⁺ ions was observed in the emission spectra. The main reason for the absence of lasing signs was the insufficient gain of the fibers in the spectral region of ~1590 nm, where the reflection peaks of the FBG arrays recorded in these fibers are located. On the contrary, in the LD681 and LD682 samples, which have a higher gain, laser generation was already recorded at a pump power of ~10 mW (LD681) and ~20 mW (LD682). The optical spectrum of laser radiation of the 5 m long LD682 sample, measured at a pump power of 30 mW, is shown in Figure 14.

The laser emission spectrum shows a single narrow peak at a wavelength of 1587.65 nm, the position of which completely coincided in wavelength with the maximum reflection of the FBG array. The ratio between the maximum intensity of the laser emission and the spontaneous luminescence level was about 50 dB. The measured spectral half-width of the peak was ~0.02 nm (insert in Figure 14) and was limited by the resolution limit of the used spectrum analyzer. The presence of such a narrow spectral peak in the absence of side-lasing satellites is one of the main features of the single-frequency laser generation mode.

In addition, the mode composition of the LD682 resonator radiation was investigated using a scanning Fabry–Perot interferometer. Since this laser has a cavity length of 5 m, the simple calculation gives the longitudinal mode-spacing value of about 20 MHz. Therefore, if the measured laser were able to generate in several modes simultaneously, several resonance peaks with a separation distance of 20 MHz would be visible within the free spectral range of the interferometer. Since no polarization-maintaining (PM) fibers were used in this work, the degree of polarization of the laser radiation was expected to be elliptical. Figure 15 shows the interferograms measured without a polarizer at the laser output (a) and measured with a polarizer (b). When using a polarizer, only the main resonances are visible in the free dispersion region of the interferometer (750 MHz), which corresponds to the passage of single-frequency and linearly polarized radiation through the interferometer. In the absence of a polarizer, the interferogram shows a signal from two orthogonally polarized resonator modes, the distance between which is ~300 MHz. This result is in good agreement with other works in which non-polarization-maintaining fibers were used to fabricate the laser resonator and pump circuit.

In accordance with [21,22], the single-longitudinal mode operation is established in the random laser cavity almost instantly. Therefore, the laser output emission spectrum consists of a single narrow line. Based on the other previous work [19], we estimate the laser natural linewidth is less than 1 kHz. However, the direct experimental measurement of the linewidth for a laser with a ~1590 nm emission wavelength is pretty difficult because it requires non-standard fiber-optic components.

It is important to note that laser generation occurred both in the same direction as the pump radiation input and in the opposite direction. Figure 16 shows the dependencies of the radiation power of the LD681 (a) and LD682 (b) resonator samples on the pump power input to the resonator. As can be seen from the figure, the opposite direction had a clear advantage in terms of the generation threshold and differential efficiency. The LD682 resonator demonstrated the highest differential efficiency (16%). In terms of output power (more than 15 mW), the LD682 random laser is comparable to commercially available single-frequency fiber and diode lasers of the telecommunication range.

4. Discussion

In the context of the problem of creating L-band optical fibers, our research in this work was aimed at finding glass compositions for the Er-doped fiber core that meets a whole set of specified properties. The most difficult task was to increase the luminescence intensity of Er³⁺ ions in the long-wavelength range far from the maximum (~1530 nm) and achieve gain values of at least several units per meter at the wavelengths of ~1590 nm. As one of the solutions, two co-doping elements were considered: P₂O₅ and Sb₂O₃. It is logical to assume that a noticeable effect could be expected in the case of higher concentrations of these oxides.

We managed to introduce up to 16 mol% P₂O₅, and, as far as we know, this is a record-breaking limit value for preforms manufactured using MCVD technology. However, no noticeable effect in increasing the gain in the L-band for Er-GPS optical fibers was registered, despite the high phosphorus concentration and high optical transparency of the synthesized glasses. At a relatively high content of 12–16 mol% P₂O₅, the shape of the erbium luminescence band spectrum did not change, and the observed rapid decline in its long-wavelength edge with an increase in the Er₂O₃ concentration indicates an extremely low efficiency of P₂O₅ as a co-doping element for suppressing the clustering of Er³⁺ ions. Apparently, an even higher concentration of P₂O₅ is required to achieve the result, namely, at a level of 40 mol%—as it was demonstrated in the phosphate glass studied in [12]. It is worth noting separately that for a random resonator based on an array of low-reflecting gratings, the circular geometry of the fiber core is very important. The ellipticity of the core leads to significant fluctuations in the grating parameters along the grating array length and, as a result, to distortions and broadening of the reflection spectrum of the entire array. Insufficiently selective reflection at the wavelength makes it difficult to obtain laser generation. Under MCVD process conditions, it is extremely difficult to consolidate a tube preform into a solid circular geometry rod due to the large difference in the temperature of the viscosity of the highly doped Er-GPS core and silica cladding. It can be stated that the potential of the Er-GPS matrix as a potential application to random L-band lasers is limited by the capabilities of the MCVD technology and can be fully revealed by manufacturing fibers with a higher concentration of phosphorus and erbium using other technological methods.

The use of an alternative co-doping with Sb₂O₃ resulted in a significant increase in the refractive index of the core and an even greater non-additive growth of Δn_core-clad in case of the simultaneous presence of GeO₂ and P₂O₅ in the core glass. This is presumably due to the more complex structure of multicomponent glasses. Since the optimal Δn_core-clad in Er-based optical fibers from the point of view of acceptable losses during splicing should not exceed ~0.015, the amount of Sb₂O₃ in the synthesized glasses was reduced to 1 mol%. Indeed, the addition of 1 mol% Sb₂O₃ to the GPS matrix contributed to a 25% increase in the gain in the C-band, but we did not record any significant difference in the wavelength range of 1580–1610 nm. At the same time, the level of minimum optical losses in this fiber increased 5 times, to a level of ~50 dB/km, which undoubtedly worsened the generation characteristics of an extended random resonator (the length was 5 m). Nevertheless, it is worth noting the extremely high efficiency of antimony in imparting photosensitivity to laser UV radiation in silicate glasses. Only 0.6 mol% Sb₂O₃ in the SbPS matrix was sufficient for recording a high-contrast FBG in one laser pulse, which was later confirmed by observing an intense reflection peak of the FBG array at a wavelength of ~1590 nm.

The combination of properties of the GPS, SbPS, and SbGPS matrices developed as the core material for Er-doped fibers did not allow for achieving laser generation at a wavelength of 1590 nm. First of all, due to the low gain, the value of which did not exceed 1 dB/m. Therefore, a different approach to the problem of increasing the gain was proposed—to develop a matrix in which the total intensity of the Er³⁺ ion luminescence increased, but the shape and maximum of the band would not change. To do this, we increased the erbium concentration in the core (up to about 0.07 mol%), at which the absorption coefficient at the maximum (~1530 nm) increased from 6–7 to 32–34 dB/m, and Al₂O₃ was used as the most effective dopant to suppress the clustering of erbium ions. Moreover, a deliberately higher concentration of Al₂O₃ ~5 mol% was introduced into the core than that required to dissolve erbium with a content of 0.1 mol%. Nevertheless, the problem of aluminosilicate and aluminophosphosilicate matrices is the extremely low initial photosensitivity to laser UV radiation, which makes it impossible to record an FBG array with sufficient contrast. To increase the photosensitivity, a ~5 mol% GeO₂ was additionally introduced into the core. Thus, GAS and GAPS matrices were developed and fibers with a relatively high erbium content in the core were manufactured on their basis.

Both manufactured fibers (Er-GAS and Er-GAPS) had a low level of “background” optical losses (like Er-GPS fibers), high contrast of the recorded FBG arrays, and, most importantly, increased gain in the 1590 nm region—5 (Er-GAPS) and 7.2 dB/m (Er-GAS). It can be stated that we managed to synthesize photosensitive, homogeneous, high-purity glasses in which the clustering of Er³⁺ ions is suppressed even at a higher erbium content relative to the other synthesized glasses. The lower value of the gain coefficient of the Er-GAPS fiber is due to the complete binding of Al₂O₃ into AlPO₄ complexes. This dopant is also widely used in the manufacture of Er-doped fiber optics, but its efficiency is significantly inferior to that of Al₂O₃. On the other hand, the Δn_core-clad value in the Er-GAPS fiber was only 0.006 and was only due to the refractivity of 4.5 mol% GeO₂ and 1 mol% P₂O₅ that was unbound in AlPO₄. Single-mode Er-GAS and Er-GAPS fibers had approximately the same cutoff wavelength of ~0.94 μm, but the core diameter differed in these fibers by almost 2 times: 3.8 and 7 μm, respectively. The advantage of the GAPS matrix is the ability to vary the light-guiding parameters of the erbium fiber over a wide range for optimal matching in numerical aperture with other components of fiber circuits by changing the concentrations of doping components. In addition, due to the simultaneous doping with germanium and phosphorus oxides, the GAPS matrix demonstrated record photosensitivity among the studied fibers, which is one of the key factors affecting the efficiency of random fiber resonators in the form of a recorded array of Bragg gratings.

5. Conclusions

The main practical result of our research was the first demonstration of random-cavity fiber lasers based on Er-GAS and Er-GAPS optical fibers capable of generating single-frequency radiation in the 1590 nm wavelength region. The laser generation threshold in the Er-GAS resonator was lower (13 mW), which is due to the higher radiation power density in the core of a smaller diameter. At the same time, the Er-GAPS resonator design provided higher values of laser efficiency (16%) and output power level (16 mW), comparable to single-frequency C-band diode lasers.