*3.1. Distributed Fiber Raman Amplifiers*

In a distributed FRA, pump at the optical frequency ω*<sup>P</sup>* and signal at the frequency ω<sup>s</sup> = ω*<sup>P</sup>* − ων are jointly launched into the same ordinary optical fiber. Raman optical amplification takes place when their frequency difference Ω = ω*<sup>P</sup>* − ω<sup>s</sup> lies within the bandwidth of the Raman-gain spectrum. The distributed FRA guides light at both the signal and pump wavelengths and it is normally single mode to ensure the best overlap of all traveling waves. A distributed FRA unit is a pumping unit at 14xx nm without gain medium and it, usually, consists of pump lasers, optical couplers to combine them, isolators and a few monitor taps, as well as the WDM coupler for signal and pump. Depolarizers are necessary when the pump is polarized, in order to suppress the polarization dependence of Raman amplification.

Figure 3 shows a configuration in which pumping light propagates bi-directionally in distributed FRA. Moreover, the system is provided with band pass filter (BPF) and arrayed waveguide grating (AWG) devices which act as demultiplexing unit in the receiving side.

As mentioned before, in a distributed FRA both forward and backward pumping schemes are possible. In the forward pumping, the signal is amplified near its fiber-input end, where the signal power is relatively high. Thus, this scheme corresponds to a booster amplifier. In the backward pumping, the optical signal is amplified near its fiber-output end before photodetection, where the signal power may be so weak as to be comparable to the noise light power. Since this scheme is used to improve detectivity or to reduce the minimum detectable signal power, it can be considered as a detection preamplifier [50,51].

We note that the process of Raman amplification takes place so rapidly that, unless the intensity noise of the forward pumping light is sufficiently small, the pumping light noise will be transferred to the signal light resulting in increasing transmission bit error rates. Since signals propagating in the opposite direction to the pump will average over the fluctuations, in order to mitigate unwanted

crosstalk between the pump and the signal, in many cases only backward pumping is used [52,53]. Typically, high-powered counter propagating Raman pumps are deployed in conjunction with discrete amplifiers, such as EDFAs.

**Figure 3.** Schematic representation of multiplexing/demultiplexing based distributed optical Raman amplifier.

Distributed FRAs offer several advantages: simplicity, because they provide direct optical amplification of signal light in the transmission fiber; flexibility in the use of signal wavelengths, because the Raman gain peak is dependent on the pump wavelength and not the emission cross section of a dopant; broad gain bandwidth, which allows to employ multiple pumps. However, one of the major benefits, only obtainable through the use of distributed amplification, is that signal gain may be pushed into a transmission span preventing the signal from decaying as much as it otherwise would have if no amplification was provided within the span. As a consequence, the signal-to-noise ratio does not drop as much as it would have in a system based on transmission through a passive fiber followed by a discrete amplifier (see Figure 4). Thus, lower signal powers can be used, nonlinear penalty are reduced and higher loss can be tolerated. This improved noise performance may be used in different ways. One way is to extend the reach between repeaters, another is to extend the total reach of a transmission system, and a third one is to improve the transmission capacity.

**Figure 4.** Ideal distributed amplifier: the loss is counterbalanced at every point along the span, leading to an improvement of the signal to noise ratio (SNR) respect to (ideal) discrete amplification.

Using distributed FRAs, the transmission fiber becomes an amplifier, giving an important advantage but, at the same time, the optimization of the fiber design with respect to amplifier performance becomes more difficult due to the need to optimize the fiber for signal transmission, too. Several parameters have to be taken into account to evaluate the performance of a distributed FRAs, such as linear noise accumulation, pulse distortion due to group velocity dispersion, effects due to Rayleigh scattering, nonlinear interactions between pump and signal channels, and pump–signal crosstalk. Considering that in distributed amplifiers the gain is gathered over tens of kilometers and that the Raman effect is a very fast process (response time in the order of femtoseconds), these noise sources become significant.

In Ref. [54] a 25.6 Tb/s experiment was demonstrated over a 240 km repeated span by employing 160 WDM channels with 50 GHz channel spacing in the C+L bands. The authors used backward Raman pumping to increase the received optical signal to noise ratio (OSNR) and simplify the optical repeaters, while an optical equalizer was used in the receiver to minimize channel distortion caused by narrow optical filtering.

#### *3.2. Discrete Fiber Raman Amplifiers*

Lumped or discrete FRAs use a fiber medium that is localized before or after transmission to fully or partially compensate for the transmission loss. Unlike a distributed FRA, all of the pump power is confined to the lumped element (see Figure 5). In this particular case, counterpropagating pump power is confined within the unit by the use of isolators surrounding the amplifier. Compared with Figure 3, no pump power enters the transmission line.

**Figure 5.** Discrete fiber Raman amplifier (FRA) scheme with pump power confined to the lumped element.

In discrete FRAs, the used fiber is a lot shorter (order tens instead of hundreds of km) compared to distributed FRAs. Since gains accumulate with fiber length, Raman amplification better fits in the distribuited FRAs class. Therefore, discrete FRAs are primarily used to increase the capacity of fiber-optic networks, opening up new wavelength windows, which are inaccessible by EDFAs. In FRAs, the signal is amplified at a wavelegth given by the frequency difference between the pump and the Stokes frequencies, thus by choosing the pump wavelength, gain at any wavelength can be obtained. Moreover, by combining multiple pump wavelengths, the shaping of the gain spectrum can be achieved, whereas by combining several pumps centered at different wavelengths, a flat gain in an ultrawide bandwidth is allowed.

The latter issue, i.e., a flat gain, is highly desirable in WDM systems in order to equally amplify all the channels. For this reason, flat gain of broad-band Raman amplifier, is considered one of the most important goal for the optimization of fiber communications. This task can be obtained by simultaneously using pumps at different wavelegths that generate Raman gain lightly shifted from each other so as to partly overlap each other leading to a composite gain (see Figure 6). The final Raman gain can be very flat over a wide range of wavelegth by accurately designing each pump wavelength and power. The gain-band expansion that can be achieved by multi-wavelength pumping is determined by the magnitude of the Stokes shift. A large Stokes shift is therefore desirable for a wideband FRAs using multi-wavelength pumping [55]. The maximum amplification bandwidth of a FRA is about 100 nm for silica fiber in the 1550-nm band, so the composite gain could cover a nearly 100-nm signal band [56]. They are particularly useful for amplifying (dense) WDM systems.

**Figure 6.** Multi-wavelength pumping technique for wideband Raman amplifiers.

Using a broadband pumping approach, amplifiers with gain bandwidths greater than 100 nm have been demonstrated [57] with average net gain of 2 dB gain for SMF, while for dispersion-shifted fiber and reverse dispersion fiber the net gain was of 6.5 dB. A 12- wavelength-channel WDM LD (high-power laser diode (LD) pumping unit), with a range of wavelength from 1405 to 1510 nm and combined by an Il-MZI-integrated, using a planar lightwave circuit technology, were used to enlarge the gain bandwidth of Raman amplifiers [57] However, due to a strong interaction between pumps, these broadband FRAs typically need of more power at the shortest wavelengths pumps which amplify the longer wavelegth pumps, leading to at an increase of the noise properties.

Dispersion-compensating Raman amplifiers integrate two crucial tasks, dispersion compensation and discrete Raman amplification, into a single component [58–60]. Dispersion compensating fiber (DCF) devices, based on typical dispersion for the propagating mode, are usually in the order of several kilometers and, as a result, show important insertion loss. Hansen et al. demonstrated that Raman gain produced in the DCF was as a simple and attractive method to achieve lossless DCF modules, without affect system performance [58]. Due to a small mode-field diameter, a very low pump powers are required, like those available from semiconductor laser diodes. The DCF shows a net gain higher than loss, resulting in a wider system margin and the possibility to introduce other elements such as optical add-drop multiplexers into the system. In Ref. [61] a polarization-combined four channel WDM 14xx nm laser diode unit was used as pumping light source to demonstrate a broadband Raman amplification in a DCF. A fully compensation of the losses in the DCF were obtained by means of the Raman gain over 50 nm bandwidth within +/− 0.5 dB variation.

#### **4. Fiber Raman Lasers (FRLs)**

FRLs are similar to ordinary lasers. They basically involve three key components: A low-effective-area and high-nonlinearity optical fiber for providing Raman gain, mirrors for feedback and components for isolation between the pump laser (usually a rare-earth-doped fiber laser) and the Raman resonator. A first analogy is that, in FRLs, lasing occurs when the Raman-active gain medium is placed inside a cavity, for example, between mirrors reflecting the first Stokes wavelength. A second

analogy is that, in FRLs, the threshold power is obtained when Raman amplification during a round trip is as large as to compensate the cavity losses. However, there are also some important differences between FRLs and traditional lasers. A first one is that an amplifier medium based on Raman gain is used rather than on stimulated emission from excited atoms or ions. A second difference is that the required wavelength for pumping Raman laser does not depend on the electronic structure of the medium, so it can be chosen to minimize absorption. In addition, it is worth highlighting the unique property of Raman laser, i.e., they can be operated at several wavelengths simultaneously. Higher-order Stokes wavelengths can be generated inside the active medium at high pump powers and they can be properly dispersed spatially in association with separate mirrors for each Stokes beam [62].

The first continuous-wave Raman laser in fiber was demonstrated in 1976 [63,64]. Pumping at the watts level, a distributed Raman gain, higher than the linear attenuation plus losses at splice/connector points, was induced in a single-mode glass fiber, placed inside a Fabry-Perot cavity. SRS process occurred in fiber resulting in an intense output, due to wavelength-selective feedback for the Stokes light of the Fabry-Perot cavity, which was formed by two partially reflecting mirrors. Higher-order Stokes wavelengths were generated inside the fiber at high pump powers, too. The spatial dispersion of various Stokes wavelengths through an intracavity prism allowed tuning of the laser wavelength.

A single wavelength shift FRL is a fiber resonator at the Stokes wavelength, in which SRS shifts the spectrum of the propagating pump radiation through an optical fiber towards lower frequency Stokes components. Raman lasing is obtainable in conventional single-mode telecom fibers (e.g., SMF-28) as well as other passive fibers, by trapping Stokes components by reflectors and by pumping the laser by a high-power rare-earth-doped fiber laser (commonly ytterbium). In an all-fiber configuration, enough positive feedback for lasing at the first Stokes wavelength is easily achieved. Directionality to the laser emission is provided by a high-reflecting fiber Bragg grating (reflectivity > 99%) on the input side and by a low-reflectivity output coupler (at Stokes wavelength) on the output side. Fiber Bragg gratings (FBGs) are narrow-band reflectors, acting as the mirrors of the laser cavity and inscribed directly into the core of the optical fiber used as the gain medium. However, we note that in Raman laser, when Raman gain of the active medium is very large, the laser can build up from noise to a substantial signal without any feedback. In addition, a single wavelength shift FRL, with a broadband flat Raman gain profile can be obtained using multiple pump wavelengths. Figure 7 shows a schematic diagram of a specific Raman fiber laser with one wavelength shift. In this particular case, the laser uses a 1117 nm ytterbium-doped fiber laser and converts it to 1178 nm.

**Figure 7.** Schematic representation of a Raman fiber laser with one wavelength shift. (HR—high-reflectivity fiber Bragg grating, OC—output coupler—a low-reflectivity fiber Bragg grating).

A multiple wavelength shifts FRL takes advantage of SRS cascading. The pump light gives rise to the "first-order" laser light in a single frequency-shifting step, which remains trapped in the laser resonator. Afterwards, the "first-order" laser light can be pushed to very high power levels becoming itself the pump for the generation of "second-order" laser light, which is shifted by the same vibrational frequency of first order. Using this technique, conversion of the pump light (typically around 1060 nm) to an "arbitrary" desired output wavelength through several discrete steps can be performed by using a single laser resonator. In Figure 8 is schematically reported the wavelength conversion over two or more Stokes shifts obtained by using a cascaded Raman resonator with fiber Bragg gratings incorporated in it, in order to realize nested cavities at each of the intermediate wavelengths. These intermediate wavelengths are chosen close to the peak of the Raman gain of the respective preceding wavelength; finally, a low-reflectivity output coupler terminates the wavelength conversion. The resulting output is at the designed final wavelength while only small fractions are at the intermediate wavelengths.

**Figure 8.** Schematic representation of a Raman fiber laser with multiple wavelength shifts by using a cascaded Raman resonator (RIG—Raman input grating set, ROG—Raman output grating set).

Tunability is an important property for lasers. The first tunable FRL, using a linear cavity configuration, was reported by Jain et al. [65]. A tuning range of 8 nm with an efficiency of 9% were demonstrated. However, the bulk optics (prisms and rotating mirrors), placed inside the laser cavity for tuning, introduced mechanical instabilities as well as coupling losses, which gave rise to high power thresholds and low efficiencies. In order to eliminate losses and to make a compact device, the approach based on bulk optics was replaced by an all-fiber design. In such a design, cavity mirrors were integrated within the fiber by the development of tunable fiber Bragg gratings [66]. In 1988, the first FRL based on fiber Bragg gratings was reported [67].

Since the Raman gain value of germanosilicate fiber is about eight times higher than for silicate fiber, these fibers are extensively used in FRLs. Due to their peak Raman shift of 440–490 cm−1, in an FRL with a 1064 nm pump source, the third and the sixth Stokes orders produce outputs at 1239 nm and 1484 nm, respectively. So, in order to make a FRL at 1484 nm, one would need six cascaded cavities. In Ref. [68], a GeO–doped high-power and widely tunable all-fiber Raman laser was demonstrated. The FRL, using a linear cavity configuration and a purely axial compression of the FBGs, was continuously tuned over 60 nm, from 1075 to 1135 nm. A Stokes output power up to 5 W, for 6.5 W of launched pump power was provided. A high efficiency over the whole tuning range, from 76.1 to 93.1% and laser thresholds varying from 0.78 to 2.59W were obtained, respectively.

It is worth noting that the peak Raman shift of low-loss phosphosilicate (P2O5-SiO2) fiber is 1330 cm<sup>−</sup>1, so when they are pumped by a 1064 nm pump source, the first Stokes (S1) and the second Stokes (S2) orders occur at 1239 nm and 1484 nm, respectively. This means that to make a FRL at 1484 nm, only two cascaded cavities would be required, thereby greatly increasing the FRL efficiency. In Ref. [69], a CW high-power Raman flber laser was reported, which was formed by a CW 8.4 W/ 1064 nm Yb-doped double-clad fiber laser as a pump, a 700 m phosphosilicate fiber and cascaded cavities with two pairs of FBG mirrors for the first and the second Stokes orders. A maximum output power of 2.24 W and slope efficiency of 32.8% at 1484 nm were demonstrated. In Ref. [70], Babin et al. demonstrated a simple all-fiber widely tunable phosphosilicate FRL of high efficiency. The laser had more than 50 nm tuning range and generated up to 3.2 W of output power with 72% maximum slope efficiency. The output power was almost constant in the range 1258–1303 nm.

Typically, the operating wavelengths of Raman laser are in the 1–1.5 μm region of the spectrum and have been created using silica glass as the base material with the desired application directed toward telecommunications. However, there are a large number of applications requiring high power laser with wavelengths longer than 2 μm. As a result of the versatility offered by FRLs, these types of sources are attractive for this purpose [71]. Chalcogenide glass fibers have a Raman gain coefficients approximately two orders of magnitude greater than the gain coefficients of silica. For wavelengths longer than 1.8 μm, the development of Raman fiber lasers based on chalcogenide glasses has now become technically achievable [72]. In Ref. [73] a Raman fiber laser, using a nonoxide glass as the gain medium and a 2 μm fiber laser pump source was demonstrated. Due to the low optical losses and high nonlinear index of commercially available glassy As2Se3 fiber, a combined output of 1 W at the first and second Stokes wavelengths of 2062 and 2074 nm, were generated respectively.

Single-frequency laser sources have a great number of applications, but are difficult to implement, suffer from poor robustness, poor quality (linewidth and stability) and are expensive. In recent years, there has been much interest in rare-earth doped distributed feedback (DFB) fiber lasers, due to their very low noise and high efficiency [74]. The oscillating wavelength can be accurately defined by a Bragg grating written directly into the fiber and inserting a π phase-shift (PS) at, or close, to the center of the Bragg grating structure. Of course, the operation of π-PS-DFB fiber lasers has been mostly limited to the spectral bands of rare-earth dopants [75].

A Raman gain based DFB fiber laser would have a number of potential advantages. First, the possibility of generating narrow linewidth low-noise oscillation in wavelength bands outside the one of rare-earth doped materials. Second, Raman based fiber laser systems do not suffer from issues associated with high-concentration rare-earth doped fibers, which limit their efficiency due to thermal effects. Using Raman gain in DFB fiber lasers was proposed for uniform FBGs by Perlin and Winful [76] for the first time, while a Raman DFB fiber laser based on a relatively short (<20cm), π-shift fiber Bragg grating were proposed by Hu et al. [77]. Westbrook et al. have realized the first functional DFB using the Raman gain in fiber, but with a very high threshold (40 W) was observed at 1584 nm. The authors showed that watt-level thresholds can be achieved by using a small-core high-numerical aperture (NA) fiber [78]. Shi et al. demonstrated that a polarized pump and an ultralong FBG (30 cm) could yield sub-watt threshold at a lasing wavelength of 1120 nm [79].

In reference [80] highly efficient Raman DFB fiber lasers with CW output power up to 1.6 W was demonstrated, for the first time. The DFB Bragg gratings were written directly into two types of commercially available passive germano-silica fibers. Two lasers of 30 cm length were pumped with CW power at 1068 nm up to 15 W. The threshold power was ~2 W for a Raman-DFB laser written in standard low-NA fiber, and only ~1 W for a laser written in a high-NA fiber, both of which oscillate in a narrow linewidth at ~1117 nm and ~1109 nm, respectively. The slope efficiencies were ~74% and ~93% with respect to absorbed pump power in the low-NA fiber and high-NA fiber respectively. Recently a Raman-DFB with lower thresholds at 1120 nm and 1178 nm was demonstrated [81]. Despite these developments, the performance in terms of slope efficiency has remained poor (1–7% for single-ended output) compared to the expected simulated efficiencies (40–50%).

By exploiting the multiple scattering of photons in a disordered gain medium, a random laser can be obtained allowing a coherent light source without a traditional cavity [82]. For different scientific and medical applications, ultrafast Raman fiber lasers are an interesting option. An efficient way to obtain a high power ultrafast Raman fiber laser is pulse pumping, but it requires a real-time synchronization between the pump pulses and laser cavity. In order to overcome this limitation, recently, random fiber lasers with distributed Raman gain and Rayleigh feedback in standard telecommunication optical fibers were demonstrated [83,84]. Their advantages are a simple and flexible design, quasi-CW operation, narrow spectrum generation, high beam quality and pump energy conversion efficiencies comparable to the ones of conventional cavity lasers [85]. In recent years, many works have been reported on the mechanism [84,85], the improvement of performance including power scaling [86,87], wavelength

tuning [88,89], and Q-switching [90,91]. Nowadays, random fiber is a unique laser technology with various advantages in the fields of fiber sensing and optical communications [92,93].

A promising approach for the further development of devices based on SRS is represented by Photonic crystal fiber (PCF). The most important advantage of this technology is the reduction of the fiber length and the power levels required. The first preliminary results in photonic crystal fiber for Raman amplification and for an all optical Raman modulator was presented by Yusoff et al. [94]. An analysis of the Raman properties in triangular photonic crystal fiber was reported by Fuochi et al. [95], which investigated the influence of presence of the germanium doped core on the Raman properties. A model of PCF amplifiers was developed by Bottaccini et al. [96], which demonstrated that germanium doped PCF can enhance the gain for the same pump power. Finally, SRS in ethanol core PCF was demonstrated by Yiou et al. [97]. The approach of PCF with liquid could open new perspectives for SRS nonlinear properties with applications in optical sensing.
