定义：
放大超短光脉冲的放大器。

超快放大器是应用于放大超短脉冲的光放大器。一些超快放大器被用来放大高重复率的脉冲列，得到很高的平均功率而脉冲能量仍然处于中等水平，另外一些情况下，低重复率的脉冲得到更大的增益，得到很高的脉冲能量以及相对很大的峰值功率。将这些强脉冲聚焦到一些目标物上时，得到非常高的光强，有时甚至大于1016 W/cm2。 
作为例子，考虑锁模激光器的输出情况，脉冲重复率为100MHz，长度为100 fs，平均功率为0.1W。于是脉冲能量为0.1W/100MHz=1nJ，峰值功率小于10kW（与脉冲形状有关）。一个高功率放大器，作用于整个脉冲上，可以将其平均功率提高到10W，于是脉冲能量提高到100nJ。另外，可以在放大器之前采用一个脉冲拾取器将脉冲重复率减小到1kHz。如果高功率放大器还是将平均功率提高到10W，那么这时脉冲能量则有10mJ，峰值功率可以达到100GW。 

超快放大器的特殊要求 
除了光放大器的通常技术细节之外，超快装置还面临额外的问题： 

• 尤其对于高能系统，放大器的增益必须非常大。在上面讨论的离子中，需要增益高达70dB。由于单通放大器受限于增益，通常采用多通道工作。采用正反馈放大器可以实现非常高的增益。另外，也经常采用多级放大器（放大器链），其中第一级提供高的增益而最后一级则优化得到高脉冲能量和有效能量提取。 
• 高的增益通常也意味着对背向反射光（正反馈放大器例外）更敏感，并且更趋于产生放大的自发辐射（ASE）。在一定程度上，ASE可以在两级放大器之间放置光开关（声光调制器）得到抑制。这些开关仅在放大脉冲峰值附近非常短的时间间隔内打开。然而，这一时间间隔相比于脉冲长度仍然很长，因此抑制脉冲附近的ASE背景噪声不太可能。光学参量放大器在这一方面性能更好，因为它们只在泵浦脉冲通过时提供增益。不会对背向传播的光进行放大。 
• 超短脉冲具有显著的带宽，可以通过放大器中的增益变窄效应来减小，因此得到的放大脉冲长度变长。当脉冲长度小于几十个飞秒时，需要超宽带放大器。增益变窄在高增益系统中尤其重要。 
• 尤其对于高脉冲能量的系统，各种非线性效应会使脉冲时间和空间形状发生畸变，甚至会由于自聚焦效应损坏放大器。一种有效的抑制这一效应的方法是采用啁啾脉冲放大器（CPA），脉冲首先被色散展宽到例如1ns的长度，然后被放大，最后色散压缩。另一种不太常见的替代方法是采用分脉冲放大器。另一个重要的方法，是提高放大器的模式面积从而来减小光强。 
• 对于单通放大器，只有当脉冲长度足够长从而脉冲通量达到饱和通量量级，但不会导致强非线性效应的情况下才有可能得到有效的能量提取。 

一些可能性能指标的例子 
对超快放大器的不同要求体现在脉冲能量，脉冲长度，重复率，平均波长等的不同。相应的，需要采取不同的装置。下面给出了一些有典型的不同类型系统得到的性能指标： 

• 掺镱光纤放大器可以放大将100MHz的10ps的脉冲列放大到平均功率为10W。（具有这种性能的系统有时被称为超快光纤激光器，尽管它其实是一个主振荡功率放大器装置。）采用大模式面积的光纤放大器可以比较容易得到峰值功率为10kW。但是如果是飞秒脉冲，这种系统就会具有非常强烈的非线性效应。开始是飞秒脉冲，然后进行啁啾脉冲放大，可以很容易得到几个微焦耳的能量，或者在极限情况下可以大于1mJ。一个替代的方法是在具有正常色散的光纤中放大一个抛物线脉冲，然后进行色散压缩脉冲。 
• 一个多通的体放大器，例如基于钛蓝宝石的放大器，可以提供大的模式面积，得到的输出能量在1J的量级，脉冲重复率比较低，例如10Hz。将几个纳秒的脉冲展宽是必要的，则是为了抑制非线性效应。后来被压缩到比如20fs，峰值功率可以达到几十太瓦（TW）；最先进的大系统可以得到峰值功率大于1PW，也就是皮瓦量级。小一些的系统，可以产生例如10 kHz的能量为1 mJ的脉冲。多通放大器的增益通常在10dB量级。 
• 正反馈放大器中可以得到几十dB的高增益。例如，采用钛蓝宝石的正反馈放大器，可以将1 nJ的脉冲放大到1 mJ。另外，还需采用啁啾脉冲放大器来抑制非线性效应。 
• 采用基于掺镱的薄盘激光头的正反馈放大器，可以将长度小于1 ps的脉冲放大到几百微焦耳，而无需采用CPA。 
• 采用调Q激光器产生的纳秒脉冲泵浦的光纤参量放大器可以将展宽的脉冲能量放大到几个毫焦耳。在单通道工作的情况就可以达到几个分贝的高增益。而对于特殊的相位匹配结构，增益带宽非常大，因此进过色散压缩后可以得到非常短的脉冲。 

商用超快放大器系统的性能指标通常远低于科学实验中得到的最好性能。许多情况下，主要原因是实验中采用的装置和技术通常由于缺乏稳定性和坚固性而无法应用于商业装置中。例如，复杂的光纤系统包含光纤和自由空间光学之间的多个跃迁过程。可以构建全光纤放大器系统，但是这些系统很那达到采用了体光学元件的系统性能。还有一些其它的情况，光学器件工作在它们的损伤阈值附近；但是对于商用装置，则需要更高的安全保证。另一个问题是需要采用一些特殊的材料，非常难以获得。 

应用 
超快放大器具有很多应用： 

• 许多装置被用于基础研究。它们可以提供强脉冲用于强的非线性过程，例如高次谐波产生，或者将粒子加速到很高的能量。 
• 大型的超快放大器在研究中被应用于激光诱导核聚变（惯性约束核聚变，快点火）。 
• 能量在毫焦耳的皮秒或者飞秒脉冲在精密加工领域很有利。例如，很短的脉冲可以实现非常精细准确的切割薄金属板。 

超快放大器系统由于其复杂性和价格昂贵很难应用于工业中，有时也因为缺乏坚固性。在这种情况下，需要在技术上更进步的发展来改善这种状况。

Definition: amplifiers for ultrashort (femtosecond or picosecond) optical pulses

Alternative terms: ultrashort pulse amplifiers, femtosecond or picosecond amplifiers

More specific terms: regenerative amplifiers, multipass amplifiers, optical parametric amplifiers, optical parametric chirped-pulse amplifiers, chirped-pulse amplifiers

Ultrafast amplifiers are optical amplifiers which are applied to ultrashort pulses. Some of these devices are used for amplifying high repetition rate pulse trains, leading to a high average power whereas the pulse energy remains moderate. In other cases, a much higher gain is applied to light pulses at lower repetition rates, leading to high pulse energies and correspondingly huge peak powers. Enormously high optical intensities, sometimes above 10¹⁶ W/cm², are achieved by focusing such intense pulses tightly on some target.

The following kind of pulse generation devices may be involved:

• The seed laser may be an internal or external mode-locked laser, or in some cases a gain-switched laser diode (→ picosecond diode lasers).
• Particularly for pre-amplification with high gain, but moderate power levels, one may either use a fiber amplifier or a bulk regenerative amplifier.
• Further boosting of the output energy and peak power is possible with a single-pass or multipass amplifier.
• There are also optical parametric amplifiers, which can produce very high gain and substantial output energies, but require energetic pump pulses rather than a continuous or quasi-continuous pump source.

Some examples for typical pulse energies and peak powers:

• A mode-locked laser may generate 100-fs pulses with a repetition rate of 100 MHz and an average power of 0.1 W. The pulse energy is then only 0.1 W / 100 MHz = 1 nJ; nevertheless, due to the short pulse duration the peak power is nearly 10 kW (somewhat depending on the pulse shape).
• A high-power amplifier, applied to the full pulse repetition rate, may raise the average power by 20 dB to 10 W, thus increasing the pulse energy to 100 nJ and the peak power to the order of 1 MW. Further substantial increases of peak power are difficult, because the average power would become very high.
• A pulse picker before the amplifier may be used to reduce the pulse repetition rate to 1 kHz. If the high-power amplifier can still raise the average power to 10 W, the pulse energy is now 10 mJ, and the peak power nearly 100 GW. The required amplifier gain is now as high as 70 dB.
• Systems with pulse pickers may even be operated with much lower pulse repetition rates, e.g. of 10 Hz. One may then reach even higher pulse energies and peak powers while keeping the average power level quite moderate. However, one then needs a huge amount of amplifier gain.

Special Requirements for Ultrafast Amplifiers

Beyond the general technical issues of optical amplifiers, ultrafast devices meet a number of additional challenges:

• Particularly for high-energy systems, the amplifier gain must be huge. As shown in the examples discussed above, a gain of many tens of decibels is often needed. As single-pass amplifiers are limited in terms of gain, multipass arrangements are frequently used. Very high gain is possible with regenerative amplifiers. Also, it is common to use multiple amplifier stages (→ amplifier chains). For example, one may use a regenerative amplifier plus one or two multipass amplifiers. The last stages are optimized for high pulse energies and efficient energy extraction.

In many types of laser amplifiers, back-reflected light could also be strongly amplified!

• A high gain also implies a high sensitivity against any back-reflected light (except for regenerative and parametric amplifiers). Therefore, one often requires Faraday isolators between the amplifier stages. In addition, it is often necessary to carefully avoid extensive optical feedback from the application into an amplifier device, because the use of a Faraday isolator after the last amplifier stage may not be feasible.

ASE suppression in high-gain amplifier systems often requires multiple measures.

• The high gain also implies a strong tendency for amplified spontaneous emission (ASE). To some extent, ASE in laser amplifiers can be suppressed with optical switches such as acousto-optic modulators between the amplifier stages. These switches are opened only for short time intervals around the peaks of the amplified pulses. However, these time intervals are usually still long compared with the pulse duration, so that it is not possible to suppress some ASE background around the pulses. Optical parametric amplifiers are better in this respect, as they provide gain only during the passage of the pump pulse. Also, there is no amplification for light propagating in the backward direction.

Broadband amplification is essential for keeping the pulses short.

• Ultrashort pulses have some significant optical bandwidth, which can be reduced by the effect of gain narrowing in an amplifier, so that the amplified pulses become longer. For pulse durations below a few tens of femtoseconds, ultrabroadband amplifiers are required. Note that gain narrowing is particularly important in high-gain systems.

Optical nonlinearities are often a major challenge and limit the achievable performance of ultrafast amplifiers.

• Particularly for systems with high pulse energies, various optical nonlinearities can distort the temporal and spectral pulse shape, or may even lead to laser-induced damage of the amplifier via self-focusing. An effective technique for mitigating such effects is chirped-pulse amplification (CPA), where the pulses are first dispersively stretched to e.g. 1 ns duration, then amplified, and finally dispersively recompressed. A novel and not yet common alternative is divided-pulse amplification. Another important measure is to increase the mode areas in the amplifiers so as to reduce the optical intensities.

Efficient energy extraction without reaching the optical damage threshold?

• For a single-pass amplifier, efficient energy extraction is possible only if the pulse duration is long enough to reach a pulse fluence of the order of the saturation fluence without causing strong nonlinear effects.

The energy storage in laser amplifiers is a crucial feature for ultrafast amplifiers.

Ultrafast laser amplifiers are usually continuously pumped, except for very low pulse repetition rates below the inverse upper-state lifetime. In any case, the required pump power is far lower than the output peak power. This is essentially because one can exploit the energy storage in the laser gain medium.

For optical parametric amplifiers, not offering energy storage, the pump pulse duration is often similar to the duration of the amplified pulses. In some cases, however, the pump pulses can be significantly longer, because one can exploit the group velocity mismatch in the nonlinear crystal.

Examples of Possible Performance Figures

The requirements on ultrafast amplifiers in terms of pulse energy, pulse duration, repetition rate, mean wavelength, etc., are very diverse. Correspondingly, very different types of devices are used. The following list gives some typical examples of the performance reached by different types of systems:

• An ytterbium-doped fiber amplifier may amplify a 100-MHz train of 10-ps pulses to an average power of 10 W. (A system with such output performance is sometimes called an ultrafast fiber laser, although it is actually a master oscillator power amplifier device.) The peak power of roughly 10 kW can be relatively easily handled with a large mode area amplifier fiber. For femtosecond pulses, however, such a system would exhibit severe nonlinear effects. Starting with femtosecond pulses and applying chirped-pulse amplification, several microjoules can easily be reached, or in extreme cases > 1 mJ. An alternative is the amplification of parabolic pulses in a fiber with normal dispersion, and subsequent dispersive compression.
• A multipass bulk amplifier, based e.g. on Ti:sapphire and realized with a large mode area, may provide an output energy of the order of 1 J at a low repetition rate of e.g. 10 Hz. Strong pulse stretching to a duration of e.g. a few nanoseconds is necessary in order to limit nonlinear effects. After compression to e.g. 20 fs, the peak power is tens of terawatts (TW); the most advanced large systems already reach peak powers above 1 PW, i.e. in the petawatt domain. Smaller systems generate e.g. 1-mJ pulses at 10 kHz. The gain of a multipass amplifier is usually of the order of 10 dB.
• A much higher gain of several tens of decibels is possible with a regenerative amplifier. Made with Ti:sapphire, e.g., such a device may amplify 1-nJ pulses to 1 mJ. Again, chirped-pulse amplification has to be applied in order to limit nonlinear effects.
• Using a regenerative amplifier based on an Yb-doped thin-disk laser head, pulses with durations somewhat below 1 ps can be amplified to hundreds of microjoules without applying CPA.
• An optical parametric amplifier, pumped with nanosecond pulses from a Q-switched laser, can amplify stretched pulses to energies of several millijoules. A high gain of tens of decibels can be achieved in a single pass. For special phase-matching configurations, the gain bandwidth can be very large, so that very short pulse durations are possible after dispersive compression.

The performance figures of commercial ultrafast amplifier systems often greatly lag behind the best performance values achieved in scientific experiments. In many cases, a main reason is that experimental systems rely on techniques or components which cannot be used in commercial devices due to a lack of stability and robustness. For example, complex fiber-based systems have been demonstrated which contain multiple transitions between fibers and free-space optics. All-fiber amplifier systems can be made, but these do not reach the performance levels of systems containing bulk-optical elements. In other cases, optical components are operated close to their damage threshold; for commercial devices, a larger safety margin is required. Another problem can be the use of very special parts, which are difficult to procure.

Applications

Ultrafast amplifiers find diverse applications:

• Many devices are used in fundamental research. They can provide intense pulses for highly nonlinear processes such as high harmonic generation, or for the acceleration of particles to high energies.
• Large ultrafast amplifiers are under consideration for the use in laser-induced nuclear fusion (inertial confinement fusion, fast ignition).
• Intense picosecond or femtosecond pulses are interesting for laser material processing. Millijoule pulse energies are required in some cases, while laser micromachining often works with microjoule or even nanojoule pulse energies.

Short pulse durations make it possible to achieve very precise cutting of thin metal sheets, for example.

Industrial applications are often made difficult by the complexity and cost of ultrafast amplifier systems, sometimes also by a lack of robustness. Further technological progress may improve the situation.

Bibliography