定义：
输出功率（>10W）很高的光纤激光器和光纤放大器。

第一台光纤激光器的输出功率只有几个毫瓦，近来光纤激光器发展很快，得到高功率光纤放大器，尤其是放大器的输出功率可以达到几十上百瓦，甚至在某些单模光纤中得到上千瓦。
这是由于光纤很大的表面积体积比（避免多余的热）和导波（波导）性质，避免了在很高温度下的热光效应问题。 
光纤激光器技术与其它的高功率固态体激光器、薄盘激光器等相比具有很大的竞争力。 

目录 

1. 双包层光纤和光束质量
2. 注入泵浦光
3. 激光器和放大器
4. 纳秒脉冲
5. 超短脉冲
6. 未来发展前景

双包层光纤和光束质量 

   图1：具有气体包层的双包层光纤结构图。
通常高功率光纤激光器和放大器都是采用稀土掺杂双包层光纤，又通过光纤耦合的高功率二极管棒或者其它激光二极管泵浦。泵浦管并不是进入光纤纤芯，而是进入内包层，也是在内包层产生激光。
产生的激光光束质量很好，甚至可以得到衍射极限的光束质量，需要采用单模的光纤。因此，光纤激光器输出光的亮度比泵浦光的亮度高几个数量级，尽管输出功率比泵浦光小。
（通常泵浦效率大于50%，有时甚至大于80%）因此这种光纤激光器可以用来作为亮度转换器，也就是提高光亮度的装置。 

对于特别高的功率，纤芯面积需要足够大（参阅大模式面积光纤），因为这时光强会很高，还有一个原因就是双包层光纤中包层与纤芯面积比值很大从而导致泵浦吸收很小。
当纤芯面积达到几千平方微米的量级，采用单模纤芯是可行的。采用多模光纤，模式面积比较大时也可以得到质量相当好的输出光束，这时光波主要是基模。
（通过将光纤绕成圈也可以在一定程度上一直高阶模式的激发，高功率时的强模式耦合情况除外）模式面积变得更大后，光束质量不再能保持衍射极限，但相比于例如工作于相似功率强度下的棒激光器，得到的光束质量仍是相当好了。 

注入泵浦光 
关于怎样注入非常高功率的泵浦光有几种方式可以选择。最简单的方式就是直接在光纤端口泵浦包层。这种方法无需特别的光纤部件，但是高功率泵浦光需要在空气尤其是空气玻璃界面传播，对灰尘或者不对准非常敏感。
许多情况下，会优先选择使用光纤耦合的泵浦二极管，这样泵浦光一直在光纤中传输。还有一种选择是将泵浦光输入到无源光纤中（未掺杂），将无源光纤缠绕在掺杂光纤周围，这样泵浦光被逐渐转移至掺杂光纤中。
有一些方法是采用专门的泵浦结合装置，将一些泵浦光纤和掺杂信号光纤融合在一起。还有其它的方法基于边泵浦的光纤线圈（光纤盘激光器），或者在泵浦包层刻槽这样泵浦光可以注入进去。后种技术允许泵浦光多点注入，因此可以更好的分配热负荷。 

图2：高功率双包层光纤放大器装置图，泵浦光通过自由空间进入光纤端口。气体玻璃界面必须严格对准和洁净。 
所有的注入泵浦光的方法之间的比较狠复杂，因为涉及到很多方面：转移效率，亮度损耗，易加工，灵活操作，可能的背向反射，从纤芯泄漏至泵浦光源的光，保持偏振的选择等。 

激光器和放大器 
理论上产生高功率激光最简单的方法是采用几个镜子直接构建一个腔激光器。然而，通常高功率光纤装置会与激光放大器结合起来，也就是采用MOFA结构（主振荡光纤激光器）。这种结构具有一些优点。
低功率的种子激光在线宽、激光噪声、波长调谐、脉冲产生等方面的性质比较容易控制，激光器腔内功率则更高一些（尽管作为高增益装置，光纤激光器允许很强的输出耦合）。另外，采用调制器方案更好，这样可以把各级的放大器连起来。
在有些情况，尤其是采用种子激光器的情况，通常采用多级放大器，这时在放大器链上模式面积和泵浦功率都逐渐增大。 

纳秒脉冲 
通常用于材料加工的高功率激光器为调Q激光器，产生很强的纳秒脉冲。在这个领域，光纤装置主要在峰值功率上受限制：需要远离各种非线性效应，自聚集效应将峰值功率限制在几个兆瓦，即使通过增大模式面积也不能再增大。
实际得到的峰值功率都小于1兆瓦，即使对于大模式功率的光纤也不例外。也就是说，如果脉冲长度为1ns，脉冲能量则小于几个毫焦耳。 
光纤装置的高增益也使灵活的MOPA装置易于实现，它采用增益开关激光二极管作为种子激光。这种方案不同于调Q激光器，它可以调整脉冲长度而不会改变脉冲重复率。 

超短脉冲 
光纤放大器大的增益带宽可以用来放大超短脉冲。然而，这时强的光纤非线性效应影响很大，因为能量很高的飞秒脉冲具有非常大的峰值功率。除了存在对光纤损伤的危险之外，强非线性效应引起强的脉冲畸变，另外很高的色散（包含高阶色散）也是一个问题。
尤其需要考虑脉冲质量时，这些因素都需要认真考虑。 
一种可能的方案是采用啁啾脉冲放大系统，放大器中脉冲长度被显著提高，因此非线性效应相应的被减小。另一个可供选择的方案是放大抛物线脉冲，上啁啾脉冲得到增益，放大器中光纤的非线性趋于自相似的方式，因此可能得到色散压缩的高质量脉冲。 
尽管上面提到的方案可以提高高功率超短脉冲的质量（至少对于几个兆赫兹的脉冲重复率和中等脉冲能量情况），这些结果通常是基于一些自由空间光学的装置，因此会失去一些光纤系统的优势。
一个很有意思的方向是发展全光纤的超短脉冲源，可以实现自由空间脉冲压缩（例如透射光栅），但是这样不需要将脉冲从自由空间注入到光纤中。 

未来发展前景 
尽管近来高功率光纤装置的发展已经非常迅速，仍然存在一些局限性阻碍了进一步的发展： 

• 高功率光纤装置的光强提高了很多。现在通常可以达到材料损伤阈值。因此，需要提高模式面积（大模式面积光纤），但是采用这种方法在需要得到高光束质量的情况时也存在局限性。 
• 单位长度的功率损耗已经达到100W/m量级，导致光纤中很强的热效应。采用水冷却的方法可以很大程度的提高功率。掺杂浓度低的更长的光纤冷却更容易，但是这会提高非线性效应。 
• 对于不是严格的单模光纤，当输出功率大于某个阈值时，存在模式不稳定性，通常为几百瓦特。模式不稳定性导致光束质量的突然降低，这是光纤中的热光栅（空间温度振荡很快）的作用。 
• 光纤非线性影响很多方面。即使在连续波装置中，拉曼增益非常高（即使分贝），以至于很大一部分功率被转移到波长较长的斯托克斯波上，这一波并不能被放大。单频工作则极大的受限于受激布里渊散射，当然也有一些测量方式在一定程度上可以抵消这一效应。在锁模激光器中的产生的超短脉冲，自相位调制会对其产生很强的光谱展宽效应。另外，还存在其它注入非线性偏振旋转的问题。 

由于存在以上的诸多限制，高功率的光纤装置通常不能被严格的看做可扩展功率的装置，至少在能达到的功率范围之外不能。
（之前的改进方法并没有采用单功率定标来实现，而是采用改进的光纤设计和泵浦二极管。）这在光纤激光器技术与薄盘激光器对比时有很重要的后果。在词条激光功率定标中有更详细的描述。 
即使不能真正的进行功率扩展，在改进高功率激光装置方面还有很多工作可以做。一方面是需要提高光纤设计，例如采用大的光纤模式面积和单模导引，通常采用光子晶体光纤来实现。
很多光纤器件非常重要，例如特殊的泵浦耦合器，光纤锥来连接具有不同模式大小的光纤和特殊的光纤冷却装置。一旦达到某个光纤的功率极限，合成光束是另一个选择，存在合适的光纤装置来实现这种技术。
对于超短脉冲放大器系统，存在许多方法减小或者甚至部分利用光纤的非线性效应，例如展宽光谱和其后的脉冲压缩。

Acronym: HPFL

Definition: fiber lasers and fiber amplifiers with high output powers of e.g.larger than 100 W

More general terms: fiber lasers and fiber amplifiers

Whereas the first fiber lasers could deliver only a few milliwatts of output power, there have subsequently been rapid developments which have lead to high-power fiber lasers and particularly amplifiers with output powers of tens or hundreds of watts, sometimes even several kilowatts from a single fiber. This potential arises from a very high surface-to-volume ratio (avoiding excessive heating) and the guiding (waveguide) effect, which avoids thermo-optical problems even under conditions of significant heating.

There is no generally agreed minimum power level for the attribute “high power”. It may mean more than 100 W of average power, but some lasers are labeled as high-power levels even for 10 W only.

Fiber laser technology now competes strongly with other high-power laser technologies based on solid-state bulk lasers, such as thin-disk lasers.

Double-clad Fibers and Beam Quality

High-power fiber lasers and fiber amplifiers are nearly always realized with rare-earth-doped double-clad fibers, which are pumped with fiber-coupled high-power diode bars or other kinds of laser diodes. The pump light is launched into an inner cladding rather than into the (much smaller) fiber core, in which the laser light is generated. The laser light can have very good beam quality – even diffraction-limited beam quality if the fiber has a single-mode core. Therefore, the brightness of the fiber laser output can be orders of magnitude higher than that of the pump light, even though the output power is of course somewhat smaller. Double-clad fiber lasers can thus effectively be used as brightness converters. Typical optical-to-optical power efficiencies are above 50%, sometimes even above 80%; the overall electrical-to-optical power conversion efficiency of a diode-pumped fiber laser can be of the order of 50% for an Yb-based fiber laser pumped with efficient high-power diode lasers.

Figure 1: Structure of a modern double-clad fiber with an air cladding.

For the highest powers, the core area needs to be fairly large (→ large mode area fibers), because the optical intensities would otherwise become too high, and often also because a double-clad fiber with a large ratio of cladding to core area has a weak pump absorption. For core areas up to the order of a few thousand μm², it is feasible to have a single-mode core. Larger mode areas with still fairly good output beam quality are possible with a slightly multimode core, where most of the light propagates in the fundamental mode. (The excitation of higher-order fiber modes can be suppressed to some extent e.g. by coiling the fiber, except if one obtains strong mode coupling at high power levels [15].) For even larger mode areas, the beam quality can no longer be nearly diffraction-limited, but it can still be fairly good compared with, e.g., rod lasers operating at similar power levels.

The required fiber length is often several tens of meters, dictated by the need to absorb the injected pump light with sufficiently high efficiency. Although the fiber length can 'in principle be reduced e.g. by using a fiber core with a higher doping concentration, a large fiber core or a smaller pump cladding (for a smaller cladding/core ratio), there are various limits to that, e.g. arising from limits to the doping concentration, the beam quality of the pump source, beam quality requirements on the signal or thermal considerations.

Launching the Pump Light

There are several options for launching pump light at very high power levels. The simplest is to launch directly into the pump cladding at one or both fiber ends (see Figure 2). This technique does not require special fiber components; however, it needs the propagation of the high-power pump radiation through air (with free-space optics) and particularly through an air–glass interface, which is then very sensitive to dust and misalignment. In many cases, it is therefore preferable to use one of several techniques where one employs fiber-coupled pump diodes and keeps the pump light in fibers from there on (see Figure 3). One option is to launch the pump light into passive (undoped) fibers which are wound around the active fiber so that the light is gradually transferred into the active fiber (GTWave fiber). Other techniques are based on special pump combiner devices (fiber-optic pump combiners), where several pump fibers and a single active signal fiber are fused together. Yet other approaches are based on side-pumped fiber coils (fiber disk lasers) [3], or on grooves in the pump cladding through which pump light can be injected. The latter technique allows for multi-point pump injection and thus for a better distribution of the heat load.

Figure 2: A high-power double-clad fiber amplifier setup, where pump light launched through free space into the fiber end. The air–glass interface is critical in terms of alignment and cleanliness.

Figure 3: Set up of a high-power fiber laser. Light from eight fiber-coupled pump diodes is combined with two pump fiber couplers and sent into the active fiber from both directions. Fiber Bragg gratings are used to form the laser resonator.

A comparison of all the different pump launching techniques is complex, since it involves many aspects: not only the transfer efficiency, but also the loss of brightness, the ease of manufacturing, the flexibility of handling, possible back-reflections, leakage of light from the fiber core back to the pump source, the option to maintain the polarization, etc.

Lasers and Amplifiers

The conceptually simplest option for generating high-power laser light is to build a fiber laser directly with some kind of mirrors on both ends. However, high-power fiber devices are very often built as laser-amplifier combinations, i.e., with a MOFA (master oscillator fiber amplifier) architecture. This concept has several advantages. It is easier to control the emission properties of a lower-power seed laser in terms of linewidth, laser noise, wavelength tunability, pulse generation, etc. Also, the fiber of an amplifier has to stand only a power about equal to the output power, whereas in a laser the intracavity power is higher (even though fiber lasers as high-gain devices allow for strong output coupling). Furthermore, it can be advantageous to use a modular design approach, where amplifier stages can be combined as required. In many cases, particularly when a low-power seed laser is used, one even uses several amplifier stages, typically with increasing mode areas and pump powers along the chain.

Nanosecond Pulses

Many high-power lasers for material processing are Q-switched bulk lasers, generating intense nanosecond pulses. The direct application of that technique to fiber lasers is very much limited in terms of achievable peak power, basically due to the strong fiber nonlinearities. However, the high gain of fiber devices allows the realization of very flexible MOPA devices with a gain-switched laser diode (→ picosecond diode lasers) as seed laser. This approach, unlike Q switching of a laser, allows one, e.g., to modify the pulse duration independently of the pulse repetition rate. Still, the possible pulse energies are far lower than possible with bulk lasers. For many applications, however, the possible performance is sufficient.

Ultrashort Pulses

The large gain bandwidth of fiber amplifiers allows for the amplification of ultrashort pulses. That can be exploited both in mode-locked fiber lasers and in ultrashort pulse fiber amplifiers.

Particularly in this pulse duration regime, substantial challenges arise from the strong fiber nonlinearity, because high-energy femtosecond pulses have huge peak powers. In addition to the risk of fiber damage, strong nonlinear distortions can result from the fiber nonlinearity, and the high level of chromatic dispersion (including higher-order dispersion) can also be problematic. Particularly if pulse quality matters, these matters need to be carefully considered.

One possible approach is to make fiber-based chirped-pulse amplification systems, where the pulse duration within the amplifier is strongly increased, so that the effect of nonlinearities is accordingly decreased. A less powerful alternative to this approach is the amplification of parabolic pulses, where up-chirped pulses experiencing the gain and nonlinearity of the amplifier fiber evolve in a self-similar fashion, and their close to linear chirp makes it possible to obtain a high pulse quality with dispersive compression.

An interesting direction is the development of all-fiber ultrashort pulse sources, possibly allowing for a final free-space pulse compressor (e.g. a transmission grating), but eliminating the need to launch pulses from free space into a fiber.

Limiting Factors

There are various possible limiting factors for the performance of high-power fiber amplifiers and lasers, which are briefly discussed in the following. Which of those factors limits the performance of a particular device, depends very much on the operation regime.

Available Pump Power

Some devices are limited by the amount of pump power which can be launched into the fiber. Note that even when using double-clad fibers, the beam quality of the pump source needs to be sufficiently high. Effectively, that means that only pump sources with sufficiently high radiance (brightness) can be used.

Fortunately, the continuous development of diode bars and broad area laser diodes results in devices with higher and higher radiance, so that the corresponding performance limits are extended more and more.

Heat Generation

A common problem for high-power lasers is the heat generation in the gain medium. Fiber amplifiers and lasers, however, are affected by this only at very high powers, basically because (a) the often quite high power conversion efficiency (resulting in a relatively low heat load), (b) the distribution of the heat load over a relatively long length (e.g. several meters) and (c) the waveguide function of the fiber core.

The heat load in a fiber may become a problem particularly in cases where the fiber length has to be minimized, e.g. in order to mitigate nonlinear effects. This means using fibers with relatively high doping concentration, allowing to turn over the required power within a shorter length. The temperature rise in the fiber core can then be quite substantial. Besides, the resulting thermal lensing may significantly modify the mode properties, e.g. reducing the effective mode area and increasing the number of guided modes.

The dissipated power per unit length often reached values of the order of 100 W/m, which causes significant heating for air-cooled fibers. Water cooling may be used for significant further increases in power, although it cannot remove the thermal lensing problem.

If the fiber is not strictly single-mode (which is often the case in high-power devices, particularly the last amplifier stage), the heating can also trigger another problem, called thermally induced transverse mode instability (TMI). That phenomenon usually occurs above a relatively well defined pump power level; a typical characteristic is that one then suddenly obtains a substantially reduced output beam quality, which is associated with a substantial transfer of power into higher-order guided modes. The distribution of optical power shows an oscillatory or chaotic behavior. The tendency for that mode instability grows for fibers with particularly large mode area.

The physical mechanism underlying the thermally induced mode instability is relatively complicated. It involves the generation of a long-period grating in the fiber [14] which couples light from the fundamental mode to higher-order modes. At the same time, the grating is generated by modal interference, which is leading to a spatially varying intensity pattern and (via Kramers–Kronig) to a refractive index pattern. An important aspect is a slight difference in optical frequency between the involved modes, which leads to a easily detected beat note and to a movement of the mentioned refractive index grating [15].

Nonlinear Limitations

Various high-power fiber devices are essentially limited in performance by consequences of fiber nonlinearities of different types:

• Particularly single-frequency devices suffer from stimulated Brillouin scattering (SBS). The problem is that the intense signal light in an amplifier, for example, leads to a high Brillouin gain for counterpropagating light with a slightly lower optical frequency. Once that nonlinear gain reaches approximately 90 dB, quantum noise is amplified to a substantial power level. The resulting counter propagating wave then extracts power from the signal, which in turn reduces the nonlinear gain; that interaction usually leads to chaotic power fluctuations. There are various mitigation techniques. For example, it is common to apply a first phase modulation to the input signal, which can be removed after the amplifier, and reduces the Brillouin gain by increasing the signal linewidth beyond the Brillouin gain bandwidth. One may also optimize the fiber design, particularly concerning the effective mode area and used length, and exploit a temperature gradient along the fiber.
• Similarly, a nonlinear gain arises from stimulated Raman scattering (SRS), where however various parameters are very different from those of Brillouin scattering: the gain coefficient is much lower, while the nonlinear gain bandwidth is far higher, and the nonlinear gain also occurs for co-propagating light. SRS is often the dominant limiting factor for the amplification of ultrashort pulses, where SBS is not relevant due to the broader signal bandwidth and the rather limited spatial overlap region for counterpropagating short pulses.
• A rather hard limit for the possible peak power in a fiber is set by nonlinear self-focusing. Above a certain critical power, that effect leads to a catastrophic collapse of the beam profile, which is usually followed by instant destruction of the fiber material. Interestingly, the critical power is not increased by using a fiber with larger effective mode area, even though the optical intensity for the undisturbed mode is reduced by that. For silica fibers, having a relatively low nonlinear index, the critical power is relatively high (several megawatts), but still a serious limitation in some cases with ultrashort pulse amplification. Even when applying chirped-pulse amplification, the pulse energy is limited to the order of 10 mJ, because there are practical limits to the usable chirped pulse duration.
• There can also be the problem of laser-induced damage at fiber ends. That problem can be substantially mitigated by using core-less end caps, in which the beam can expand to a substantially larger area until it reaches the more sensitive glass–air interface.

Note that the optimization of high-power file devices often involves a trade-off between different kinds of limitations; best results are usually obtained by finding an optimal compromise.

Prospects for Further Improvements

Even though the progress in the development of high-power fiber devices has been tremendous in recent years, various kinds of more or less severe limitations are now encountered, which are expected to slow this progress:

• In most cases (particularly for pulsed operation), fiber devices are not limited by available pump power, or at least the pump power could be increased further with already available laser diodes.
• The optical intensities in high-power fiber devices have been enormously increased. Now they are often close to the damage threshold of the material. Therefore, increased mode areas (→ large mode area fibers) are used. However, it seems that the limits of this approach have also been more or less reached, at least if high output beam quality is required.
• There still seems to be some potential for mitigating the thermally induced mode instability, but probably not a possibility to eliminate that limitation altogether.
• Fiber nonlinearities are usually the limiting factor for devices involving short or ultrashort pulses. For various devices (particularly for mode-locked fiber lasers), improved mitigation methods have led to substantial performance improvements; it is unclear whether substantial further advances along that route are possible.

Due to the explained limitations, high-power fiber devices can – contrary to common opinions – usually not be considered as power-scalable in a reasonable sense, at least not beyond the already achieved performance level. Previous advances have also largely not been achieved with simple power scaling, but with a combination of other methods, such as using improved pump diodes and fiber designs, and optimizing parameters to ideally deal with existing trade-offs. See the article on power scaling of lasers for more details.

Once the limit of the power per fiber has been reached, beam combining is a further option, albeit at the cost of substantially reducing the system complexity.

Although the mentioned optimization techniques have allowed for often astonishing performance of fiber-based high-power laser sources, these achievements are usually based on setups which involve a lot of free-space optics and hence lose many of the specific advantages of all-fiber systems. However, it may be possible to develop more all-fiber systems which still offer good enough performance for many applications.

Bibliography