定义：
通过主动或者被动锁模产生超短脉冲的激光器。

通过主动或者被动锁模产生超短脉冲的光纤激光器具有很多有利的特性：

• 稀土掺杂光纤的增益带宽较大，通常有几十纳米，能够产生飞秒脉冲。
• 有源光纤很高的增益效率使激光器工作在很低的功率处，并且允许腔内的光学器件具有一定的光学损耗。例如，在体激光器中，有些光学滤波器等不能用来进行色散补偿。
• 光纤激光器装置成本较低，并且尺寸小装置很坚固，尤其是当不采用自由空间器件时。
• 锁模光纤激光器主要是采用通信器件，这些器件已经发展的非常成熟，适宜应用于长距离工作并且成本适中。
• 激光器输出一般是光纤耦合的（例如，端口为光纤连接器），因此与通信系统兼容。
• 如果采用双包层光纤，那么可以得到很高的输出功率。

但是，有些皮秒和飞秒锁模光纤激光器和放大器，尤其是在脉冲能量，峰值功率和脉冲质量方面受限于光纤中强的非线性效应，有时也受色散影响。还有一个问题是如果光纤不是偏振保持的，那么其中的双折射无法控制。
有很多种超快光纤激光器，它们不仅辐射不同波长、脉冲长度和脉冲能量的脉冲，并且采用了不同的脉冲产生机制。下面的章节中讨论几种最重要类型的超快光纤激光器。

目录

• 1.5-μm飞秒掺铒光纤激光器
• 掺镱1-μm飞秒光纤激光器
• 高脉冲重复速率
• 色散和光纤非线性
• 孤子锁模光纤激光器受非线性效应的限制
• 脉冲展宽光纤激光器
• 自相似光纤激光器中的抛物线脉冲
• 其它工作在正常色散区域的光纤激光器
• 通过放大得到更高的脉冲能量
• 自启动问题
• 锁模光纤激光器中的饱和吸收器
• 结论

1.5-μm飞秒掺铒光纤激光器

掺铒光纤具有较宽的增益带宽，峰值增益位于1535 nm或者更长的例如1550 nm处，取决于光纤纤芯的组分和反转能级，而后者则取决于光纤长度，掺杂浓度和谐振腔损耗。 图1是一种很简单的掺铒光纤激光器。线性激光器谐振腔左端连着一个半导体饱和吸收反射镜（SESAM）作为被动锁模器，另一端就是裸光纤端口（菲涅尔反射率约为4%）。掺铒激光器由一个低功率的激光二极管泵浦，后者出射的光纤进入一个二色性光纤耦合器中。光纤长度决定了脉冲重复速率，而脉冲长度则是色散，非线性和增益的共同结果。

图1：简单的掺铒飞秒光纤激光器

更复杂的谐振腔可以得到更短的脉冲。例如图2中所示的8字形激光器装置。右侧的环是一个非线性放大环路反射镜。来自主谐振腔中的光分成两束相反方向的光进入环路中。顺时针方向以较低的功率先进入长的非线性光纤中，然后在掺铒光纤中被放大，而另一个方向的光则是先被掺铒光纤放大，产生的非线性相移更大。如果非线性相移为π（理想情况），那么两束光在耦合器干涉后的结果是进入主谐振腔的底部；进入另一个方向的光会被法拉第隔离器隔离。因此往返的增益在低功率时比较低，但是在一定功率时非常高（接近于脉冲的峰值功率）。这一装置类似于激光器增益与饱和吸收器的结合，使往返的脉冲与低功率的背景光分离开。人为的饱和吸收器产生一个在谐振腔中往返的脉冲，然后在左边低端端口处辐射一个脉冲列。

图2：8字形激光器装置，左侧为主谐振腔，右侧为非线性放大环路反射镜作为模式锁定装置。

在简单的情况下（例如，上面描述的8字形激光器），需要调节腔内色散用于孤子模式锁定，即往返的脉冲为准孤子脉冲，孤子参数与平均色散和非线性相移有关。尽管孤子工作状态下脉冲形状很规则，并且易于采用一些基本的定量描述，但是考虑到脉冲能量和脉冲功率，它具有一些限制，尤其是短脉冲情况下非线性效应变得非常强。通常来讲，脉冲长度不低于几百飞秒，脉冲能量在皮焦耳区域，平均能量只有几个毫瓦。通过结合光纤布拉格光栅来提高反常色散已经可以得到具有较高脉冲能量（远大于1 nJ）的皮秒脉冲，但是这时光纤中已经不再是孤子传输了。其它的模式锁定技术可以得到更高的脉冲能量和更短的脉冲长度，但是得到的脉冲形状更加复杂，并且通常脉冲质量也下降了。通过优化设计，掺铒飞秒光纤激光器可以得到的脉冲长度可以小于100 fs，有时甚至低于50 fs。典型的输出功率为几十毫瓦。如果采用不同的色散控制可以得到更高的脉冲能量，下面会有所讨论。
 

除了非线性放大环路反射镜，还有其他类型的饱和吸收器。一种常用的方法是采用非线性偏振旋转，其中随功率变化的偏振通过偏振光纤器件转化陈了随功率变化的透射率。图3给出了采用这种原理的光纤环形激光器装置，这一装置也可以用作线性光纤激光器谐振腔。这种技术类似于体激光器中的克尔透镜模式锁定，但是这一技术的确定在于温度变化也会引入偏振改变。这一问题如果采用保偏光纤和法拉第旋转器可以解决。这样得到的激光器在环境中是稳定的，因为温度引起的偏振变化被消除了。但是这种方法比较难实现全光纤装置，因此不是理想的稳定装置。 还有采用腔内调制器的主动锁模光纤激光器，例如，马赫-曾德尔类型。这种技术在光纤通信中非常重要，其中需要采用不同的光源。

图3：光纤环形激光器，利用非线性偏振旋转实现锁模

掺镱1-μm飞秒光纤激光器

有的锁模光纤激光器利用掺镱光纤，辐射的波长在1000-1100 nm。这种技术相对于掺铒激光器来讲存在一个很大的不同是石英光纤通常是处于正常色散区域。产生飞秒脉冲通常需要附加的反常色散，可以由光纤布拉格光栅或体光栅对得到。在1-μm波长区域光子晶体光纤中也可能发生反常色散。这种光纤可以用作腔内色散补偿的无源光纤或者有源光纤。因此可以将所有其中的技术应用于掺铒光纤激光器中。
掺镱飞秒光纤激光器可以得到与掺铒激光器类似的脉冲长度。得到的输出功率较小，有时较高。因此这种激光器在低脉冲能量区域可以与锁模体激光器相比拟。
 

高脉冲重复速率

在很多应用领域，例如，通信，需要数千兆赫脉冲重复速率的脉冲列。光纤激光器谐振腔如果只采用基模模式锁定来（即只有一个脉冲往返于谐振腔中）实现这一重复速率的话其长度太长。因此，通常需要谐波锁模，其中多个脉冲（甚至成百上千个）以特定的间隔往返于谐振腔中。很多技术用来保证模式间的等间隔和抑制超模噪声，但是这会提高装置的复杂性，并且激光器工作并不是很稳定。
 

色散和光纤非线性

与体激光器相比，光纤激光器谐振腔具有更强的色散和非线性效应（尤其是非线性克尔效应）。强的色散效应来自于玻璃的长度（大多数情况下谐振腔往返长度大于1m，体激光器约几毫米），并且由于采用的单模光纤具有很小的模式面积，光强更高，因此非线性效应进一步加强。强的非线性效应引入强的限制，下面会提到。
将具有不同符号色散的光纤结合一起可以补偿激光器谐振腔中的色散（参阅色散补偿）。但是，很难同时补偿二阶和高阶色散，这在产生超短脉冲非常重要。
有些情况下，激光器谐振腔需要很大的色散，这不能由色散光纤产生。啁啾光纤布拉格光栅和光纤环形器结合在一起可以解决。有时也采用体光栅对，但是自由空间光学引入对准、灰尘敏感性等问题，因此丧失了光纤激光器本身的优势。
 

孤子锁模光纤激光器受非线性效应的限制

在孤子锁模光纤激光器中，当谐振腔中往返一次的非线性效应比较强时准孤子脉冲变得不稳定。低于该阈值时，可以看到例如Kelly边带等。（锁模体激光器中几乎不会发生该效应。）该效应在短脉冲情况下更加明显，因为会得到短的孤子周期，或者单位长度非线性相移较大。但是，孤子脉冲越长脉冲能量越小。因此需要限制脉冲的能量使其远小于平均功率和输出功率。
如果采用大模式面积光纤，可以在一定程度上减小非线性效应。但是，相比于体激光器来说，非线性效应还是很强，因此得到的脉冲能量更低，通常不能达到有源光纤增益带宽设置的限制值。
 

脉冲展宽光纤激光器

一种锁模光纤激光器中脉冲能量同时减小脉冲长度的方法是采用脉冲展宽光纤激光器，有时也称为色散管理光纤激光器。基本原理就是对光纤进行色散管理，这样在脉冲每次往返于谐振腔时被周期性展宽或者再压缩。由于谐振腔中的平均脉冲长度极大的增大，因此脉冲能量相应的增加，但是不会存在附加的非线性相移。采用这种方法可以得到的飞秒脉冲具有纳焦耳的能量。为了产生带宽限制脉冲，需要在光纤激光器谐振腔的某一合适位置进行输出耦合，或者采用附加的色散脉冲压缩。图4给出了一个环形激光器装置，利用了脉冲展宽原理，有时也利用线性谐振腔，还有时会用到所谓的σ谐振腔。 对于皮秒光纤激光器，也可以采用脉冲展宽技术。但是，得到变换极限脉冲不是特别现实，因为很大程度的展宽和压缩需要很大的色散。

图4：脉冲展宽光纤环形激光器装置

自相似光纤激光器中的抛物线脉冲

另一个方案是正常色散光纤激光谐振腔中抛物线脉冲放大过程。经过放大然后输出耦合后，利用光学滤波器重新设置脉冲带宽。由于脉冲在放大器光纤中自相似演化，这种激光器称为自相似光纤激光器。它通常辐射强啁啾脉冲，该脉冲在激光器谐振腔外被色散压缩，得到的脉冲长度低于100 fs。这种激光器具有很好的性能。但是，很难实现全光纤装置。
 

其它工作在正常色散区域的光纤激光器

除了自相似激光器之外，还有一大类全正常色散飞秒光纤激光器，其中避免了反常色散，但是脉冲演化不是自相似的。典型情况下，光谱滤波器决定了脉冲形成过程。产生的脉冲是强啁啾的，但是通常在激光器谐振腔外被强烈抑制。
这种技术受限因素目前还不是完全清楚，但是可以实现几十纳焦耳的脉冲能量，即使采用中等有效模式面积的光纤也可以实现。如果采用大模式面积光纤，脉冲能量可以大于100 nJ，甚至达到1 μJ。但是，也很难实现全光纤装置。
 

通过放大得到更高的脉冲能量

另一种获得高脉冲能量的方法是先采用一个相对低功率的光纤种子激光器，然后在光纤放大器中放大其输出（参阅主振荡功率放大器，主振荡光纤激光器）。在放大器中控制非线性相对比较简单，因为非线性相移比较容易处理。例如，在很高脉冲能量情况下采用啁啾脉冲放大器很合适。
需要注意许多商业超快光纤激光器通常包含一个锁模激光器和一个光纤放大器。
 

自启动问题

超快光纤激光器通常不会发生自启动模式锁定。主要原因是激光器谐振腔中的寄生反射很难抑制，包括其中的快速饱和吸收器用来进行模式锁定和长的激光器谐振腔。
 

锁模光纤激光器中的饱和吸收器

之前提到过，半导体饱和吸收反射镜（SESAMs）可以用在光纤激光器中进行被动锁模。但是SESAMs在光纤激光器中参数与体激光器中的差别很大。大多数情况下调制深度更高，主要是由于光纤激光器中的各种非线性效应和色散效应更强，有时也由于寄生反射效应。高的调制深度通常是利用多个量子阱吸收器或者多个厚的吸收层。但是，这会对装置带来附加的机械应力，因为吸收器层通常与布拉格反射镜结构不是晶格匹配的。这会引起吸收器损坏或者有效的装置寿命。
 

结论

尽管短脉冲产生过程采用光纤非常有利，但是其中的色散和高非线性效应限制了锁模脉冲激光器的性能，尤其是激光器的脉冲能量、脉冲长度以及脉冲质量。有很多方法处理这种问题。
在较低脉冲能量区域，锁模激光器由于其制备成本较低有很大的优势。其装置相对较小和坚固，前提是采用全光纤装置，仅采用保偏光纤的情况下。通常只有结合体光学器件才能得到最佳的性能，但是光纤光学的基本优势就丧失了。
光纤激光器在很多方面都与体激光器差别很大。并不仅仅是量上的差别；还有其相对较新的形成和影响产生脉冲的机制，还有其与体激光器不同的工作适用范围。为了得到很好性能和稳定的锁模光纤激光器，需要完全理解这些机制同时结合数值模型。实现某些指标的锁模光纤激光器的第一步就是明确最适合的工作原理。

Definition: ultrafast fiber lasers which are actively or passively mode-locked for generating ultrashort pulses

More general terms: mode-locked lasers, fiber lasers

Fiber lasers have a number of qualities which make them attractive for ultrashort pulse generation in the picosecond or femtosecond region via active or passive mode locking:

• The gain bandwidth of rare-earth-doped fibers is large – typically tens of nanometers – which allows the generation of femtosecond pulses.
• The high gain efficiency of active fibers makes it possible to operate such lasers with fairly low pump powers and/or to tolerate intracavity optical elements with relatively high optical losses. For example, certain optical filters or arrangements for dispersion compensation could not be used in bulk lasers.
• Fiber laser setups can be fabricated with low cost and can be very compact and rugged – particularly if free-space optics are not used.
• Mode-locked fiber lasers can largely be based on telecom components, which have been very carefully developed for reliable long-term operation and have a moderate cost.
• The laser output is naturally fiber-coupled (e.g. delivered to a fiber connector) and thus compatible with telecom systems.
• With double-clad fibers, there is a potential for high output powers.

On the other hand, the performance of picosecond and femtosecond mode-locked fiber lasers and amplifiers – particularly in terms of pulse energy, peak power and pulse quality – is in many cases severely limited by the strong nonlinearities of fibers (see below), and sometimes also by the chromatic dispersion. Also, there are difficulties for realizing very high pulse repetition rates in the multi-gigahertz region, where one requires harmonic mode locking, which is difficult to optimize for high stability and reliability; most femtosecond fiber lasers work somewhere between 10 MHz and some hundreds of megahertz. Other problems can result from uncontrolled birefringence of fibers, if they are not polarization-maintaining.

There is a large variety of ultrafast fiber lasers, not only emitting at different wavelengths and with different pulse durations and pulse energies, but also exploiting different physical mechanisms of pulse generation. The following sections discuss the most important types of ultrafast fiber lasers.

1.5-μm Femtosecond Erbium Fiber Lasers

Erbium-doped fibers, as otherwise used mainly for erbium-doped fiber amplifiers, exhibit a broad gain bandwidth, with a maximum either at 1535 nm or at some longer wavelength such as 1550 nm – depending on the composition of the fiber core and on the inversion level, which itself is determined by fiber length, doping concentration and resonator losses.

Figure 1: Simple erbium-doped femtosecond fiber laser.

A particularly simple type of erbium-doped fiber laser [4] is shown in Figure 1. The linear laser resonator is terminated by a semiconductor saturable absorber mirror (SESAM) as a passive mode locker on the left-hand side and a bare fiber end (with roughly 4% Fresnel reflection) on the right-hand side. The erbium-doped fiber is pumped with a low-power laser diode, the output of which is introduced into the resonator with a dichroic fiber coupler. The fiber length determines the pulse repetition rate, and the pulse duration of e.g. some hundred femtoseconds is determined by the interplay of dispersion, nonlinearity and the gain.

Figure 2: Figure-eight fiber laser setup, with a main resonator on the left-hand side an a nonlinear amplifying loop mirror as the mode-locking element.

Shorter pulses can be obtained from more sophisticated resonators. An example is the “figure-eight” (or figure-8) fiber laser setup [2] as shown in Figure 2. The ring on the right-hand side is a nonlinear amplifying loop mirror. Light coming from the main resonator (the left ring) is split into two counterpropagating components in the loop mirror. As the light propagating in clock-wise direction first passes a long nonlinear fiber with low power and is then amplified in the erbium-doped fiber, whereas light propagating in the other direction is amplified first, the nonlinear phase shift is larger for the latter component. If the difference in nonlinear phase shifts is π (the ideal case), the two components will interfere in such a way at the coupler in the middle that all the light is sent towards the bottom of the main resonator; light sent in the other direction will be eliminated by the Faraday isolator. The round-trip gain is thus very small for low powers, but much larger in a certain range of powers (ideally arranged to be near the peak power of the pulses). Effectively this setup acts like the combination of some laser gain with a saturable absorber, favoring the peak of a circulating pulse against the low-power background light. This artificial saturable absorber generates a single pulse circulating in the resonator and thus a pulse train emitted at the lower left port.

The figure-eight laser as described above is necessarily a ring laser since for passive mode locking one requires a reduction of loss with increasing optical power, and that is possible only when closing the resonator with the other output of the NALM. However, one can extend the scheme with a non-reciprocal phase shifter, which may be constructed with two Faraday rotators and a waveplate [52]. In that way, the reflection characteristics of the loop mirror can be changed such that one obtains reduced loss for increasing optical power, so that one can use a linear resonator outside the ring; the laser can then be called a figure-nine laser. A modified form applies the non-reciprocal phase outside the nonlinear loop mirror, which in that case has a kind polarizing beam splitter [54].

Even in a ring laser, it can be beneficial to include a non-reciprocal phase shifter for better self-starting behavior by obtaining a large positive slope of reflected vs. incident pulse energy on the loop.

Soliton mode locking is conceptually simple, but does not allow for high pulse energies.

In relatively simple cases, the intracavity chromatic dispersion is adjusted for soliton mode locking, i.e., the circulating pulse is a quasi-soliton pulse, with soliton parameters according to the average dispersion and nonlinear phase shift. Although soliton operation can provide clean pulse shapes and is comparatively easy to describe with some basic quantitative guidelines, it has severe limitations concerning pulse energy and peak power, essentially due to nonlinear effects which become too strong at short pulse durations (see below). Typically, pulse durations do not get below several hundred femtoseconds, pulse energies are in the picojoule region, and the average powers are only a few milliwatts. Higher pulse energies (well above 1 nJ) in picosecond pulses have been obtained by incorporating a fiber Bragg grating for increasing the anomalous dispersion [10], but this is actually no more true soliton propagation in the fiber. Other mode-locking techniques (see below) allow for higher pulse energies and shorter pulse durations, but at the cost of significantly more complicated physics of pulse shaping, and often a reduced pulse quality. With careful optimization, erbium-doped femtosecond fiber lasers can reach pulse durations below 100 fs, sometimes even below 50 fs [30]. Typical output powers are of the order of some tens of milliwatts. Higher pulse energies are possible with different kinds of dispersion management, as discussed below.

Figure 3: Fiber ring laser, mode-locked with nonlinear polarization rotation.

In addition to nonlinear amplifying loop mirrors, there are other kinds of artificial saturable absorbers. A popular choice is to exploit nonlinear polarization rotation (polarization additive pulse mode locking), where a power-dependent polarization change is converted into a power-dependent transmission through a polarizing optical element. Figure 3 shows a ring laser setup based on this principle, which can be applied also to linear fiber laser resonators. The technique is similarly versatile as Kerr lens mode locking of bulk lasers, but has the disadvantage that it normally suffers from polarization changes which can be introduced by, e.g., temperature variations. That problem can be avoided by using a modified approach based on polarization-maintaining fibers in combination with Faraday rotators [6]. That approach leads to an environmentally stable laser in the sense the temperature-induced polarization changes are eliminated. However, the technique is less easily implemented in the form of an all-fiber setup, and is thus also not ideal for stable devices.

There are also actively mode-locked fiber lasers, using an intracavity modulator, e.g. of Mach–Zehnder type. This technique is important particularly in the area of optical fiber communications, where synchronization of different sources is required.

Ytterbium-based 1-μm Femtosecond Fiber Lasers

Mode-locked fiber lasers can also be based on ytterbium-doped fiber for emission somewhere in the range of roughly 1000–1100 nm. An important difference for this technology, as compared with erbium-doped lasers, is that the dispersion of silica fibers is normally well in the normal dispersion regime. The generation of femtosecond pulses often requires additional anomalous dispersion, which can be generated with, e.g., a fiber Bragg grating or a bulk grating pair. Anomalous dispersion in the 1-μm wavelength region can also occur in photonic crystal fibers. Such fibers can be used as passive fibers for intracavity dispersion compensation or as active fibers. They allow one to apply basically all the techniques originally developed for erbium fiber lasers.

Ytterbium-based femtosecond fiber lasers achieve similar pulse durations as their erbium-doped cousins. The output powers achievable are also similar, or somewhat higher. Such lasers may thus be seen as competitors for some of the mode-locked bulk lasers in the regime of low pulse energies.

High Pulse Repetition Rates

In various applications areas, such as telecommunications, pulse trains with multi-gigahertz pulse repetition rates are required. Fiber laser resonators are usually too long to achieve such repetition rates with fundamental mode locking, i.e. with a single pulse circulating in the resonator. By using fibers with particularly high doping concentration one can achieve relatively high repetition rates [60] – but unfortunately so far not in the 1.5-μm spectral region relevant for telecom applications.

Therefore, harmonic mode locking is often required, where multiple (sometimes hundreds or even thousands) of pulses circulate with a well-defined spacing. Various techniques have been developed to guarantee the equal spacing and suppress supermode noise, but these tend to increase the complexity of the setup and often do not allow very robust operation of the laser.

Dispersion and Fiber Nonlinearity

Compared with a bulk laser, a fiber laser has much stronger dispersive and nonlinear effects (particularly via the Kerr effect) in its resonator. Strong dispersive effects arise from the long path lengths in the glass (in most cases well above 1 m per resonator round trip, compared with a few millimeters in a bulk laser), and nonlinear effects are in addition strengthened by the small mode areas of the single-mode fibers used, which leads to high optical intensities. Particularly the strong nonlinear effects introduce strong performance limitations, as discussed in the following sections.

At the least the fiber dispersion in the laser resonator can be compensated (→ dispersion compensation) e.g. by combining fibers with dispersion of different signs. However, it is not always easy to compensate both second-order and higher-order dispersion, as is necessary for the generation of very short pulses.

In some cases, fairly large amounts of dispersion are required in laser resonators, which cannot be generated with dispersive fibers. Chirped fiber Bragg gratings in combination with optical circulators can then be a solution. Bulk grating pairs are also sometimes used, but the required free-space optics introduce alignment issues, dust sensitivity, etc., and thus destroy some of the inherent advantages of fiber lasers.

Nonlinear Limitations of Soliton Mode-locked Fiber Lasers

In the case of a soliton mode-locked fiber laser, the quasi-soliton pulses become unstable if the nonlinear effects per resonator round trip become too strong. Below that threshold, characteristic effects such as Kelly sidebands can be seen. (Mode-locked bulk lasers nearly never exhibit this effect.) This is particularly the case for short pulse durations, because these lead to a short soliton period or, in other words, to a strong nonlinear phase shift per unit length. On the other hand, longer fundamental soliton pulses have lower pulse energies. One therefore often has to limit the pulse energy to values far below what would be possible in terms of average pump and output power.

To some extent, the nonlinearity can be decreased by using large mode area fibers. However, the nonlinearity still remains very strong compared with that in bulk lasers, so that the pulse energies achievable are far lower and often do not reach the limit set by the gain bandwidth of the active fiber.

Stretched-pulse Fiber Lasers

Figure 4: Setup of a stretched-pulse fiber ring laser.

A technique for raising the possible pulse energy and/or for lowering the pulse duration of a mode-locked fiber laser is used in stretched-pulse fiber lasers [5, 8, 9, 13, 19, 22], also sometimes called dispersion-managed fiber lasers. The basic idea is to use dispersion management with fiber spans of alternating dispersion, so that the pulses are periodically stretched and recompressed in every resonator round trip. As the average pulse duration in the resonator can be strongly increased, the pulse energy can be accordingly increased without obtaining excessive nonlinear phase shifts. Nanojoule energy femtosecond pulses can be generated in this way. For generating bandwidth-limited pulses, the output coupling needs to be arranged at an appropriate location in the fiber laser resonator, or external dispersive pulse compression must be applied. Figure 4 shows a ring laser setup based on the stretched-pulse principle, but linear resonators and so-called sigma (σ) resonators [13] are also used.

For picosecond fiber lasers, the stretched-pulse technique can also be used. However, it is then hardly practical for obtaining transform-limited pulses, since the large stretching and compression would then require very large amounts of dispersion.

Parabolic Pulses in Similariton Fiber Lasers

Another approach is the application of parabolic pulse amplification within a fiber laser resonator with normal dispersion. After amplification and output coupling, the pulse bandwidth is then reset with an optical filter. As the pulses undergo a self-similar evolution in the amplifier fiber, such lasers are called similariton fiber lasers. They often emit strongly chirped pulses, which can be dispersively compressed outside the laser resonator, resulting in pulse durations well below 100 fs.

Such lasers can reach good performance parameters. However, it is difficult to realize them in all-fiber technology.

Other Fiber Lasers Operating in the Normal Dispersion Regime

In addition to similariton lasers, there is a wider class of all-normal-dispersion femtosecond fiber lasers, where anomalous dispersion is also avoided, but the pulse evolution is generally not self-similar [32]. In typical cases, the action of a spectral filter plays an essential role in pulse formation. The pulses generated are strongly chirped, but can often be well compressed outside the laser resonator.

A special form is the Mamyshev oscillator [55]. Such a device contains two different bandpass filters with non-overlapping pass bands. Without the strong spectral broadening before those filters, a circulating pulse with any center wavelength would necessarily have a very large round-trip loss. Thus, the combination of those filters effectively leads to a kind of artificial saturable absorber, which forces the laser to operate in a pulse generation regime.

The limits of such techniques have so far not been fully explored, but pulse energies of tens of nanojoules are certainly possible [26], even using fibers with moderate effective mode areas. With large mode area fibers, pulse energies above 100 nJ [43] or even of the order of 1 μJ [48] can be obtained. With a Mamyshev oscillator, approximately 1 MW peak power has been reached [55]. However, all-fiber realizations are difficult to achieve.

Higher Pulse Energies by Subsequent Amplification

Another possibility for obtaining higher pulse energies is to use a relatively low-power fiber seed laser and amplify its output in a fiber amplifier (→ master oscillator power amplifier, master oscillator fiber amplifier). The management of nonlinearity is easier in an amplifier, as nonlinear phase shifts are less critical, and techniques are available to deal with them. For example, chirped-pulse amplification is suitable for very high pulse energies.

Note that many commercial ultrafast fiber lasers actually contain a mode-locked laser and a fiber amplifier.

Self-starting Issues

Ultrafast fiber lasers often do not exhibit self-starting mode locking. The main reasons are parasitic reflections within the laser resonator, which are difficult to suppress, the use of fast (artificial) saturable absorbers for mode locking, and the often long laser resonator, leading to a low ratio of pulse duration to round-trip time.

Saturable Absorbers for Mode-locked Fiber Lasers

As mentioned above, semiconductor saturable absorber mirrors (SESAMs) can be used for passive mode locking of fiber lasers. However, the parameters of SESAMs for fiber lasers are often different from those required for bulk lasers. In particular, the modulation depth has to be substantially higher in most cases, essentially since various nonlinear and dispersive effects are stronger in a fiber laser, and sometimes also because of parasitic reflections. The higher modulation depth is usually obtained by employing multiple quantum well absorbers or multiple thicker absorber layers. Unfortunately, this often leads to excessive mechanical strain within the device, as the absorber layers are often not exactly lattice-matched to the Bragg mirror structure. This can lead to problems with absorber damage or limited device lifetime.

Conclusions

Although fibers have a number of beneficial properties for short pulse generation, the dispersion and particularly the high nonlinearity of fibers severely limits the performance of mode-locked fiber lasers particularly in terms of pulse energy, pulse duration, and often also pulse quality. Various methods are used to deal with these issues. Often, these involve relatively complicated physics (e.g. arising from highly nonlinear pulse propagation), which can make it substantially more difficult than for bulk lasers to find well working or even optimized laser designs.

For lower pulse energies, mode-locked fiber laser products can be attractive due to their potentially low fabrication cost. Their setups can be fairly compact and robust, provided that an all-fiber setup containing only polarization-maintaining fibers can be used. Maximum performance is often achievable only by incorporating bulk-optical components, but the specific advantages of fiber optics are then often lost. In particular, it is problematic to require coupling of free-space laser beams into single-mode fibers, where micrometer-precise alignment is required and a single dust particle may seriously affect the performance.

Fiber lasers are in many respects very different from their bulk laser counterparts. These are not only quantitative differences; partly there are entirely new mechanisms forming and influencing the generated pulses, and qualitatively different operation regimes which are not required for bulk lasers and more difficult to analyze and optimize. A good understanding of such mechanisms, complemented with numerical modeling, is required in order to realize mode-locked fiber lasers with high performance and good stability as required for an industrial product. A first step towards the realization of a mode-locked fiber laser meeting certain specifications is to identify the most appropriate principle of operation for given output specifications.

Bibliography