>> A Brief Introduction of Rare-Earth Doped Fibers
Rare-earth doped fiber is an optical fiber in which ions of a rare-earth element, such as neodymium, Erbium or holmium, have been incorporated into the glass core matrix, yielding high absorption with low loss in the visible and near-infrared spectral regions.
Fiber lasers and fiber amplifiers are nearly always based on glass fibers which are doped with laser-active rare earth ions (normally only in the fiber core). These ions absorb pump light, typically at a shorter wavelength than the laser or amplifier wavelength (except in upconversion lasers), which excites them into some metastable levels. This allows for light amplification via stimulated emission. Such fibers are often called active fibers. They are gain media with a particularly high gain efficiency, resulting mainly from the strong optical confinement in the fiber’s waveguide structure.
1. Why Rare Earth Ions?
Rare earth ions are good candidate for active ions in laser materials because they show many absorption and fluorescence transitions in almost every region of the visible and the near-infrared range.
Rare earths have other important characteristics in comparison to other optically active ions as well: the wavelengths of their emission and absorption transitions are relatively insensitive to host materials, the lifetimes of metastable states are long and the quantum efficiency tends to be high. These properties lead to an excellent performance of rare earth ions in many optical applications.
2. Common Types of Rare-Earth Doped Fibers
The technologically most important rare-earth-doped fibers are erbium-doped fibers for erbium-doped fiber amplifiers (EDFA) and ytterbium-doped fibers for high-power fiber lasers and amplifiers.
The following table shows the most common laser-active ions and host glasses and also typical emission wavelength ranges of rare-earth-doped fibers:
|Ion||Common host glasses||Important emission wavelengths|
|neodymium (Nd3+)||silicate and phosphate glasses||1.03–1.1 μm, 0.9–0.95 μm, 1.32–1.35 μm|
|ytterbium (Yb3+)||silicate glass||1.0–1.1 μm|
|erbium (Er3+)||silicate and phosphate glasses, fluoride glasses||1.5–1.6 μm, 2.7 μm, 0.55 μm|
|thulium (Tm3+)||silicate and germanate glasses, fluoride glasses||1.7–2.1 μm, 1.45–1.53 μm, 0.48 μm, 0.8 μm|
|praseodymium (Pr3+)||silicate and fluoride glasses||1.3 μm, 0.635 μm, 0.6 μm, 0.52 μm, 0.49 μm|
|holmium (Ho3+)||silicate glasses, fluorozirconate glasses||2.1 μm, 2.9 μm|
3. Host Compositions
The glass host composition affects the solubility of the rare earth dopant which, in turn, may affect the fluorescence lifetime, absorption, emission, and excited state absorption cross sections of the dopant transitions. These quantities affect the ultimate ability of the active material to provide gain.
Devices of general interest span rare earth concentrations of tens to several thousand parts per million (ppm), resulting devices of one to tens of meters long. For some applications dopant levels of 1 ppm and less are advantageous, resulting in devices several kilometers long.
For all designs, the rare earth should ideally be confined as a delta function in the center of the core for maximum gain per unit pump power. Practically, there is a necessary tradeoff between the confinement and the rare earth concentration. The more confined structures require a higher rare earth concentration for an equivalent length, eventually running into the clustering limit for the particular host glass composition. Clustering is to be avoided in that it induces fluorescence quenching and reduces the performance of the device.
>> Waveguide Designs
Let us consider the waveguide configurations used to bring the pump, signal, and active media together. The most widely used configuration consists of a rare earth doped fiber core, which allows the pump and signal to propagate together in a single mode fiber, there are a number of alternative fiber device geometries worthy of consideration too.
1. Core Confined
The most efficient conversion of pump to signal photons uses the design in which both pump and signal are confined in the fiber core. This configuration has been made especially attractive by the availability of commercial, low-insertion loss, low-reflectivity fiber couplers, which can be chosen to combine a variety of pump and signal wavelengths onto a common output fiber.
For this design, the launched pump threshold power Pth provides a reasonable figure of merit for the efficiency of a fiber laser or amplifier, a lower value being preferred. This quantity is proportional to
Aeff = effective core area
εp = fractional absorbed pump power
σe = stimulated emission cross-section
τf = pump fluorescence lifetime
The efficiency of this device, therefore, can be increased by diminishing the effective core area, increasing the pump absorption cross section, increasing the pump fluorescence lifetime, and increasing the stimulated emission cross section.
Furthermore, of these parameters, decreasing the mode field diameter (decreasing Aeff ) has the greatest effect on increasing the gain/pump power slope. Further improvement can be achieved by confining the rare earth to the central portion of the core, where the pump and signal intensities are generally highest.
The optimized waveguide design then requires consideration of both the device configuration and a number of material and waveguide properties determined by the fabrication methods used.
2. Double Clad
Another approach to achieving interaction of guided pump light with an active medium uses a single-mode guide or the signal surrounded by a multimode pump guide.
Pump light is launched from the fiber end into the undoped cladding, propagating in a zig-zag pattern through the doped core as it travels along the fiber. Configurations with the core offset in a circular cladding and a core centered in an elliptical cladding have been demonstrated. A high brightness neodymium fiber laser based on the latter design provided an output greater than 100 mW for an 807 nm diode array laser pump power of 500 mW.
The guiding geometry of this configuration was designed to maximize the use of power available from laser diode arrays, thereby producing higher-output powers.
3. Evanescent Field
The wings or evanescent field of the optical signal guided by a single mode fiber may be used to interact with an active material outside of the core region. One approach to enhancing this effect is to locally taper the guide, thereby causing the optical power to increase outside of the glass material bounds of the fiber over a length of several millimeters.
This tapering method has been used to demonstrate a 20 dB gain amplifier, for a pump power below 1 W, with a dye solution circulating around the tapered fiber region. Both the signal at 750nm and the pump at 650 nm were copropagating in the core.
Similarly, the active media can be incorporated in the cladding glass, as has been demonstrated for erbium (Er) or neodymium (Nd). Reported gain for an erbium-doped cladding structure was 0.6 dB for a 1.55 um signal, with a 1.48 um pump power of 50 mW.
The evanescent field may also be accessed by polishing away a portion of the fiber cladding, thus creating a structure similar to a D-shaped fiber. Pulsed amplification of 22 dB for one such dye evanescent amplifier has been achieved.
The pump power required for these devices to obtain a sizable gain far exceeds that needed for schemes in which active media are contained in the core. However, by using this evanescent interaction, active media, such as dyes, which cannot be incorporated into a glass, can be explored.
>> Characterization of Rare-Earth-Doped Fibers
In addition to all the properties of a passive (undoped) optical fiber, such as the guiding properties (effective mode area, numerical aperture, cut-off wavelength, bend losses), nonlinearities, etc., active fibers can be characterized with respect to several other properties:
- One of the most important parameters is the rare-earth doping concentration, most often specified in ppm wt (parts per million by weight). A higher doping concentration allows for efficient pump absorption in a shorter length and thus also reduces the effect of nonlinearities in high peak power devices. However, it can also lead to concentration quenching.
- Wavelength-dependent effective absorption and emission cross sections (and possibly ESA cross sections) together with the upper-state lifetime (and possibly lifetimes of intermediate levels) are required for calculating the wavelength tuning behavior, power efficiency, etc.
- Parameters for quantifying the speed of energy transfer processes are important particularly for codoped fibers.
As an alternative method, so-called Giles parameters can be specified, which depend on the doping concentration, effective mode area and effective cross sections.
For such characterization, a variety of measurement techniques are used. White-light absorption spectra can be used for finding absorption cross sections (for known doping concentrations). Emission cross sections are obtained from fluorescence spectra, with scaling e.g. via the reciprocity method or the metastable level lifetimes. Upper-state lifetimes are often obtained from fluorescence measurements with pulsed pumping, and ESA parameters can be obtained in experiments with a modulated pump power.
The resulting set of data can be used e.g. in laser and amplifier models based on rate equations. Such models allow one, e.g., to predict or check the performance of fiber laser or amplifier devices, the effect of possible modifications, etc.
Further characterization may be required for quantifying effects such as photodarkening, which can sometimes seriously degrade the efficiency of active fiber devices.
>> Rare-Earth-Doped Fiber Amplifiers
1. A Brief Review of Rare-Earth-Doped Fiber Amplifiers
An important advance in optical fiber technology occurred with the development of fibers that amplify light through stimulated emission. These led to dramatic increases in the channel capacities of fiber communication systems in addition to providing the key components in many new forms of optical sources and signal processing devices.
Such fibers are made by incorporating various rare-earth ion dopants into the core material, the most successful of which has been erbium. Erbium-doped fiber amplifiers (EDFAs) in their usual configuration provide gain that maximizes at 1.53 um when the fiber is pumped by additional light input at either 1.48 or 0.98 um wavelength. Lengths of amplifying fibers are used as repeater sections in communications systems, replacing the expensive and complicated electronic units that were commonly used.
The primary motivation in using a fiber amplifier repeater is that the transmitted signal remains in optical form throughout the link rather than being transformed into an electrical signal and back to optical whenever a repeater stage is encountered. This property offers additional advantages, which include the ability to change system data rates as needed or to simultaneously transmit multiple data rates without need to modify the transmission span.
A further advantage (also true for Raman amplifiers) is that a single EDFA can provide gain for multiple wavelengths simultaneously. Such a task would otherwise require a separate electronic repeater for each wavelength. It is this feature that contributed to the realization of dense wavelength division multiplexed (DWDM) systems. For example, 80 wavelength channels having 50 GHz spacing can be accommodated within the conventional 1.53 to 1.56 um EDFA gain bandwidth. More recent efforts have resulted in the extension of EDFA gain into the longer wavelength (L band) range between 1.56 and 1.63 um.
Aside from systems applications, numerous device applications for signal processing in addition to the construction of erbium-doped fiber-based lasers have been demonstrated.
Other rare-earth dopants or dopant combinations have been used to produce fiber amplifiers that poses gain at other wavelengths in the visible and near-infrared. Examples of these include praseodymium-doped fiber amplifiers (PDFAs), which provide gain at 1.3 um and are pumped at 1.02 um. Ytterbium-doped fibers (YDFAs) amplify from 975 to 1150 nm using pump wavelengths between 910 and 1064 nm; erbium-ytterbium codoped fibers (EYDFAs) enable use of pump light at 1.06 um while providing gain at 1.55 um. Thulium-doped fluoride fibers (TDFAs) have been constructed for amplification at 0.8 um and 1.48 um.
2. Basic Theory of Amplification by Stimulated Emission
The mechanism of amplification by stimulated emission can be demonstrated using a simple material model, as shown in the following figure.
The material consists of Nt identical atoms per unit volume. Each atom has four possible energy states associated with, for example, four possible electron configurations. upward transitions between energy levels in a single atom occur through the absorption of an incident photon. In the absence of additional light, downward transitions occur either by nonradiative relaxation or radiative relaxation through spontaneous emission (the random emission of a single photon in any direction). When an additional photon is incident on the atom, a downward transition can be stimulated, resulting in the emission by the atom of a second photon, which propagates with the incident one, As more photons are generated, these in turn stimulate downward transitions in adjacent atoms; this cascading effect can ultimately result in substantial power gain, provided a sufficient number of atoms can be initially excited to the higher energy states and provided the number of downward transitions per unit time can be made to exceed the upward transition rate.
In the four-level model, light inputs at two different frequencies are separately responsible for absorption and emission. The light that is absorbed, known as the pump, is at the higher frequency; its presence induces transitions from level 1 to level 4, whose energy difference is hω2. The pump light is input at frequency ω2 to coincide with this resonance. Fast nonradiative transitions occur from level 4 to level 3, allowing a substantial number of atoms to assume the level 3 energy state. This build-up of “population” in level 3 is assured if relaxation processes from level 3 to level 2 or level 1 are either slow or are not allowed.
The energy spacing between levels 3 and 2 is hω1. The model assumes that relaxation of population from level 3 to level 2 can occur through stimulated or spontaneous emission; the latter occurs with characteristic relaxation time τ. The 3 –> 1 transition is assumed forbidden. From level 2, fast nonradiative decay again occurs to level 1. Because levels 4 and 2 both relax quickly, the populations of these levels are both essentially zero, meaning that the total atomic population Nt is divided in some proportion between levels 3 and 1, so that Nt = N1 + N3. This also means that the level 3 population can exceed that of level 2, resulting in a population inversion between these two levels. The result is that net gain can occur for light at frequency ω1 because the rate of stimulated downward transitions between 3 and 2 exceeds the upward transition rate between these levels.
To determine the gain as a function of the various input and medium parameters, rate equations for the population densities of the important energy levels must be solved. Consider a fiber whose core is doped with erbium or another substance for which the excitation dynamics can be described by the four-level model. The rate equations that describe the populations of levels 1 and 3 are, respectively,
The pump and signal powers Pp and Ps, respectively, are expressed in terms of photon flux densities [photons (sec – m2)-1] by dividing both quantities by the energy per photon hω2 or hω1 and the fiber core area Ac (erbium is assumed to be present in the core only). σap and σes are the absorption and emission cross-sections, respectively, expressed in m2; these, when multiplied by the appropriate photon flux densities, yield the probability of excitation or de-excitation of a single atom in a specified time period. Multiplying the cross sections by the associated number densities of the ground or excited states (N1 or N3) yields the exponential absorption absorption or gain coefficients for the pump and signal powers, respectively, that reside in the core (where absorption and gain exist).
The above two equations are most easily solved in steady state, in which all time derivatives are zero. The resulting expressions for N1 and N3 in terms of Nt are, respectively,
The saturation powers for the pump and signal are defined respectively as
We note that in the absence of signal power, Psatap is the pump power required to equalize the two populations.
The signal and pump powers grow or attenuate with distance in the fiber according to the respective equations
The equations are coupled because N1 and N3 both depend on Pp and Ps. An analytic solution can be obtained for the special case in which Pp << Psatap and Ps << Psates. As a result, N1 ≈ Nt and N3 ≈ Nt(Pp/Psatap). Under these conditions, the second equation above is readily solved to yield
The last equation can then be solved by assuming weak absorption for the pump, such that Pp(z) ≈ Pp(0). Then, using the foregoing approximation for N3, we obtain
This result, although greatly simplified, demonstrates that at a given pump power level, the available gain is appreciable if (1) the absorption and emission cross sections are high and (2) if the lifetime of the metastable state (level 3) is long.
>> Rare-Earth Doped Fiber Fabrication Methods
1. Low-Loss Communication Fiber
The standard method of fabricating doped silica fiber fall into two basic categories, both based on the reaction of halides, such as SiCl4, GeCl4, POCl3, SiF4, and BCl3, to form the desired mix of oxides.
Category 1 reacts the chlorides in a hydrogen flame and collects the resulting soot on a mandrel for subsequent sintering to a transparent glass. Processes based on this method are commonly referred to as vapor axial deposition (VAD) and outside vapor deposition (OVD).
Category 2 reacts the chlorides inside a substrate tube that becomes part of the cladding, simultaneously reacting, depositing, or sintering as a torch plasma fireball or microwave cavity traverses the tube. Processes based on this method are commonly referred to as modified chemical vapor deposition (MCVD), plasma chemical vapor deposition (PCVD), and intrinsic microwave chemical vapor deposition (IMCVD).
All these methods create a preform, or large-geometry equivalent, which is desired in the fiber. The preform is then drawn into an optical fiber by heating one end to the softening temperature and pulling it into a fiber at rates of 1-10 m/s.
Index-raising dopant ions, such as germanium, phosphorus, aluminum, and titanium, and index-lowering dopants such as boron and fluorine, are introduced into the reaction stream as halide vapors carried by oxygen at a temperature near 30°C. The halide compounds of rare earth ions are, however, generally less volatile than the commonly used chlorides and fluorides of the index-modifying dopants, thereby requiring volatilizing and delivering temperatures of a few hundred degrees. This requirement has stimulated the vapor and liquid phase handling methods to be discussed.
2. Rare Earth Vapor Phase
Methods to deliver rare earth vapor species to the reaction/deposition zone of a preform process have been devised for both MCVD and VAD or OVD techniques. The configurations employed for MCVD are shown in the following figure, for which rare earth dopants are delivered to an oxidation reaction region along with other index-controlling dopants. The low vapor pressure rare earth reactant is accommodated either by taking the vapor source close to the reaction zone and immediately diluting it with other reactants or by delivering the material as an aerosol or higher vapor pressure organic compound.
The heated frit source was made by soaking a region of porous soot previously deposited on the upstream inner wall of an MCVD tube with a rare earth chloride-ethanol solution. On heating to 900°C and being allowed to dry, the sponge became a vapor source. Two other source methods use the heated chloride directly as a source after dehydrating. The dehydration is necessary in that most rare earth chlorides are, in fact, hydrated. Dehydration may be accomplished by heating the material to nearly 900°C with a flow of Cl2, SOCl2, or SF6. The attraction of the heated source injector method is that the rare earth reactant source is isolated from potentially unwanted reactions with the SiCl4, GeCl4, or POCl3 index-raising reactants.
A variation of the heated chloride source method requires a two-step process referred to as transport-and-oxidation. Here, the rare earth chloride was first transported to the downstream inner wall by evaporation and condensation, followed by a separate oxidation step at higher temperatures. The resulting single-mode fiber structure consisted of a P2O5/SiO2 cladding and a Yb2O3/SiO2 core, one of the few reported uses of a rare earth dopant as an index-raising constituent. A 1-mol% Yb2O3/SiO2 core provided the 0.29% increase in refractive index over the near silica index cladding.
The aerosol delivery method overcomes the need for heated source compounds by generating a vapor at the reaction site. A feature of this method is the ability to create an aerosol at a remote location and pipe the resulting suspension of liquid droplets of rear earth dopant into the reaction regions of the MCVD substrate tube with a carrier gas. The aerosols delivered this way were generated by a 1.5 MHz ultrasonic nebulizer commonly used in room humidifiers. Both aqueous and organic liquids have been delivered by this technique, allowing the incorporation of lead, sodium, and gallium, as well as rare earths. Given that most of the aerosol fluid materials contain hydrogen, dehydration after deposition is required for low OH content.
Vapor transport of rare earth dopants may also be achieved by using organic compounds that have higher vapor pressures than the chlorides, bromides, or iodides. These materials can be delivered in lines heated to 200°C, rather than the several-hundred degree requirements for chlorides. The application of this source to MCVD has been reported using three concentric input delivery lines. Multiple rare earth doping and high dopant levels are reported with this method, along with background losses of 10 dB/km and moderate OH levels of near 20 ppm.
Rare earth vapor, aerosol, and solution transport may also be used to dope preforms fabricated by the OVD or VAD hydrolysis processes. Such doping may be achieved either during the soot deposition (see the following figure) or after the soot boule has been created (two more pictures below). The introduction of low vapor pressure dopants to VAD was initially reported using a combination of aerosol and vapor delivery. The incorporation of cerium, neodymium, and erbium has been accomplished in the OVD method by introducing rare earth organic vapors into the reaction flame.
Cerium, for example, was introduced as an organic source, cerium β-diketonate. The high vapor pressure of this compound allowed delivery to the reaction flame by a more traditional bubbler carrier system with heated delivery lines. Another high vapor pressure organic compound used as the rare earth chelate. Here a 1.0-wt% Nd2O3 double-clad fiber was fabricated for high-output powers, with background losses of 10 dB/km. Concentration of Yb2O3 as high as 11 wt%, as required for the double-clad laser, were also achieved by this method.
3. Rare Earth Solutions
One of the first reported means for incorporating low-volatility halide ions into high-purity fiber preforms used a liquid phase “soot impregnation method”. A pure silica soot boule was first fabricated by flame hydrolysis, with a porosity of 60-90% (pore diameter of 0.001 – 10 um). The boule was immersed in a methanol solution of the dopant salt for 1 hour and then allowed to dry for 24 hours, after which the boule was sintered in a He/O2/Cl2 atmosphere to a bubble-free glass rod (see the previous figure). The dopant concentration was controlled by varying the ion concentration in the solution. This general technique, later referred to as molecular stuffing, has been used to incorporate Nd and Ca in silica.
A variation of this solution-doping technique combining MCVD and the solution doping has more recently been reported (see the figure below). Here, an unsintered (porous) layer of silica is first deposited inside a silica tube by the MCVD process. This layer is doped by filling the tube with an aqueous rare earth chloride solution; this solution is allowed to soak for nearly 1 hour, and then the solution is drained. The impregnated layer is dried at high temperatures in the presence of flowing chlorine/oxygen mixture. Index-raising dopants such as aluminum have also been incorporated by this method.
Although this process would seem to be inherently less pure, it has produced doped fibers with background losses of 0.3 dB/km. This general method has also been extended by replacing aqueous solutions with ethyl alcohol, ethyl ether, or acetone solvents for Al3+ and rare earth halides. Solubilities vary widely among the rear earth nitrates, bromides, and chlorides, and all are useful.
Fibers made with these nonaqueous solvents contained a relatively low OH impurity level as evidenced by the less than 10 dB/km absorption at 1.38 um. Aqueous solution methods may also produce low OH fibers with proper dehydration techniques.
As erbium-doped silica amplifiers were developed, it became clear that confinement of the dopant to the central region of the core was very important, as were uniformity and homogeneity of the deposit. To this end, another MCVD dopant method was developed, referred to as sol-gel dipcoating. The process coats the inside of an MCVD substrate tube with a rare earth-containing sol, which subsequently gels, leaving a thin dopant layer. (see the following picture). Both rare earth and index-raising dopants may be combined. The coating sol is formed by hydrolyzing a mixture of a soluble rare earth compound with Si(OC2H5)4 (TEOS). The viscosity of the gel slowly increases with time as hydrolysis polymerizes the reactants. Deposition of the film then proceeds by filling the inside of the MCVD support tube with the gel, followed by draining. The gel layer thickness is controlled by the viscosity of the gel which, in turn, is determined by its age and the rate at which the gel liquid is drained. Film thickness of a fraction of a micrometer is typical, thereby allowing a well-confined dopant region. The coated tube is returned to the glass-working lathe for subsequent collapse.
4. Rod and Tube
The first optical fiber was made by drawing a preform assembly made of a core rod and cladding tube of the proper dimensions and indices. Recent adaptations of this method have been demonstrated for making compound glass core compositions. To retain the overall compatibility with communication-grade doped silica fiber, a small compound glass rod is inserted into a thick-walled silica tube. The combination is then drawn at the high temperatures required by the silica tube. As a result, a few of the less stable constituents of the compound glass are volatilized. In spite of this, lengths of fiber can be drawn that are long enough for practical use.
With interest in distributed Er-doped amplifiers, a need arose for a method to produce uniform and very low dopant levels. Although the solution doping and outside process methods were successful in controlling these low levels of dopant, a new, rod-and-tube like technique was also devised to meet this challenge. Here rare earth was introduced into an MCVD preform as the core of a fiber with a 150 um outside diameter and a 10 um core diameter (see the figure below). The “seed” fiber was inserted in the bore of an MCVD preform before the last collapse pass. During the final collapse pass, the fiber becomes a diffusion source for the dopant ions in the center of the preform core.
Fibers fabricated by this method have shown losses as low as 0.35 dB/km and well-controlled erbium levels of 0.01 ppm Er3+, corresponding to a ground-state absorption level of only 1 dB/km at 1.53 um. Ina sense this resembles a miniature rod-and-tube process except that the rod is effectively dissolved in the host core, as evidenced by the change in fluorescence spectrum from the seed composition to the core composition.