定义：
从激光器中提取腔的短脉冲或者超短脉冲的方法，采用的是激光器谐振腔中的光学开关。

腔倾斜是一种脉冲产生的技术，可以与Q开关结合或者与锁模结合，甚至同时与二者结合一起使用。任一情况下，基本的思路就是使激光器谐振腔的光学损耗越小越好，然后腔内形成很强的脉冲，在一次往返时间内采用一种光学开关，例如声光调制器或者普克尔斯盒，将该脉冲提取出来。 

腔倾斜产生纳秒脉冲 
最初腔倾斜是在Q开关激光器中引入的。其目的是为了消除Q开关中的一些基本的限制。尤其是，在Q开关激光器中脉冲重复速率增大的时候，脉冲长度也会增大，这是需要消除的；它来自于增益介质中储存能量较低时激光器增益也较低。并且，Q开关重复速率高也会引起脉冲损失。 
对腔倾斜的修正主要是谐振腔只包含高反射率反射镜（即没有部分透射输出耦合反射镜），输出耦合是由谐振腔中的光调制器控制的。通常是采用声光调制器，开的状态可以提取脉冲，然后将腔内光束衍射到输出端（见图1）。 

图1：腔倾斜激光器的示意图。需要提取脉冲时，声光调制器马上打开。其它情况下，光在谐振腔中循环，损耗很小。 
脉冲形成的过程如下： 

1. 首先，声光调制器的状态处于大多数谐振腔中的光都被耦合出去，因此由于器件在激光阈值以下因此不能产生激光。泵浦光源提供的能量大部分被储存在增益介质，与Q开关激光器类似。 
2. 然后调制器调整到另一个状态，其中光在激光器谐振腔中只经历小的寄生损耗。因此，腔内能量迅速增大，一般光往返几百次之后。在这一时间内，没有光耦合出去。 
3. 接下来，调制器快速改变到另一个状态，大多数光被耦合出去。在谐振腔中的光然后被提取出来。之后，另一个新的循环开始。 

与Q开关最显著的差别在于在输出脉冲产生之前，光脉冲能量先储存在腔内，而不是在增益介质中。最大的优势就是能量可以在小于一个往返时间内被提取出来，与腔内功率储存的时间无关。因此，脉冲长度是由谐振腔长度决定的（调制器的开关时间比谐振腔往返时间短），
因此与激光器增益无关。这表明几十非常高的脉冲重复速率的情况下，也可以得到几纳秒的脉冲长度。在这一区域，Q开关激光器会产生更长的脉冲，或者甚至不能在预定的时间间隙中产生一个脉冲。（当Q开关工作在很高频率时会产生脉冲损失。） 
脉冲重复速率较高时（高于上能态寿命的倒数），倾斜腔激光器非常稳定的工作，因为脉冲提取后还有一部分光在谐振腔内，这一部分光就可以作为下一个脉冲的种子光。这是适宜工作的范围。 
尽管倾斜腔的激光器设置在原理上与Q开关类似，但是二者的要求时不同的： 

1. 开关时间应该小于谐振腔往返时间，即不能超过几纳秒（Q开关在很低开关速度时也能工作）。若采用声光调制器，不仅需要微波光源可以快速开关，还需要调制器中的光束半径较小，因为开关时间受声波的渡越时间的限制。然而，小的模式面积会提高调制器中的光强，需要附加的腔内光学元件。 
2. 衍射效率需要较高，通常大于50%，从而大部分光可以在一个往返时间内被提取出来。因此，声光倾斜腔通常工作在很高的射频，在几百兆赫兹，Q开关的工作频率为几十兆赫兹，采用的相对较高的射频功率。 

从以上可以看到倾斜腔相比于Q开关的优势在于其高脉冲重复速率，尤其是同时还需要很短的脉冲长度的情况下。另一方面，这一技术会引入对模式面积和开关速度额外的限制。 

腔倾斜产生超短脉冲 
倾斜腔也可以用在锁模激光器上来产生具有很高脉冲能量的超短脉冲。由于是纳秒器件，谐振腔损耗需要保持很小，因此循环的脉冲变得很强。当达到最大强度，腔内脉冲可以通过光开关耦合出去。需要的开关速度比纳秒器件还要高。因此，需要采用声光调制器和偏振器的结合。另外，开关需要与循环脉冲合成：快速的光电二极管控制弱脉冲通过高反射腔反射镜的泄露，驱动电子学采用信号在合适的时间控制开关，因此当脉冲在谐振腔另一端时开关就会起作用。这种情况下的脉冲长度不受限于往返时间，往返时间长也是有好处的，因为这会降低对开关时间的要求。 

图2：倾斜腔皮秒激光器的装置图。倾斜腔包含一个普克尔斯盒，一个四分之一波片和一个薄膜偏振片（TFP）。 

倾斜腔产生超短脉冲通常用在锁模固态激光器中，例如钛蓝宝石激光器或者二极管泵浦掺钕或者掺镱激光器。得到的脉冲能量通常比普通锁模激光器高一个数量级（通常在1微焦量级），脉冲重复速率在几百千赫兹到几兆赫兹。倾斜腔激光器的平均功率通常远大于
普通的锁模激光器，这主要是由于寄生损耗效应。 

图3：倾斜腔用作脉冲拾取器。 
倾斜腔激光器的设计，尤其是用于超短脉冲产生，是很重要的问题。除了需要快速的开关，打开普克尔斯盒或者倾斜腔引入的光学非线性和色散都会引入问题。腔内脉冲能量的急剧变化也与孤子锁模不兼容。因此，不应该将倾斜腔看做一个简单的放置到锁模激光器中。
整个系统需要仔细设计，需要考虑到倾斜腔的非线性效应。 
在高脉冲能量情况下，需要较低的脉冲重复速率。然而，这时倾斜腔就不能很好的工作：在两个提取的脉冲之间腔内功率几乎消失，这一过程的稳定性就降低了。类似的，也很那将倾斜腔与锁模，Q开关同时结合起来，因为Q开关会在腔内快速形成高功率，
而锁模过程需要更长的时间来形成稳定的脉冲。在低重复速率和更高脉冲能量时，采用正反馈放大器更好，这时脉冲形成和放大过程不相干。 

非线性倾斜腔 
之前提到的所有的倾斜腔技术都是采用电控制的光学开关，也可以通过非线性相互作用实现倾斜腔。其中，倾斜腔激光器包括一个非线性晶体，满足相位匹配实现和频产生，通过另一个激光器通过Q开关反射脉冲进入晶体，实现倾斜腔。能量以和频的短脉冲的形式提取出来。
脉冲长度主要由Q开关入射脉冲的长度决定。在脉冲之间，非线性晶体只在谐振腔中产生非常小的计生损耗，然后形成很高的腔内循环能量。需要注意的是，腔内功率限制了产生脉冲的峰值功率，因为每一个输出光子需要另一个激光器中的光子进行转化。 
如果脉冲能够通过连续泵浦倾斜腔激光器的弛豫振荡合成，可以进一步增强功率效率。当然，非线性倾斜腔的原理也可以通过脉冲泵浦或者Q开关来实现。

Definition: a method for extracting intense short or ultrashort pulses from a laser, using an optical switch in the laser resonator

Cavity dumping is a technique for pulse generation which can be combined either with Q switching or with mode locking, or sometimes even with both techniques at the same time. In any case, the basic idea is to keep the optical losses of the laser resonator as low as possible for some time, so that an intense light pulse builds up in the resonator, and then to extract this pulse within about one cavity round-trip time using a kind of optical switch (“cavity dumper”), such as an acousto-optic modulator or a Pockels cell.

Cavity Dumping for Nanosecond Pulses

Originally, cavity dumping was invented in the context of Q-switched lasers. The purpose is to eliminate some basic limitations of Q switching. In particular, it can be disturbing that the pulse duration achievable with a Q-switched laser increases when the pulse repetition rate is increased; this is a consequence of the lower laser gain for a lower stored energy in the gain medium. Also, Q switching with high repetition rates may lead to pulse dropout.

The modification for cavity dumping is essentially that the resonator contains only highly reflecting mirrors (i.e., no partially transmissive output coupler mirror), and output coupling is controlled with the optical modulator in the resonator. Typically, this is an acousto-optic modulator, which is turned on briefly for pulse extraction and then diffracts the intracavity beam into the output (see Figure 1).

Figure 1: Schematic setup of a cavity-dumped laser. The acousto-optic modulator (AOM) is turned on only briefly when a pulse is extracted. At other times, the light can circulate in the resonator with low losses.

The pulse formation then works as follows:

• First, the modulator is kept in a state where most of the light in the resonator is coupled out, so that lasing cannot occur because the device is below the laser threshold. The energy provided by the pump source is then largely stored in the gain medium, as in a Q-switched laser.
• The modulator is then switched to a state where light circulating in the laser resonator experiences only the small parasitic losses. As a result, the intracavity power builds up quickly – typically, within a few hundred resonator round-trip times. No light is coupled out during that period.
• Then, the modulator is again quickly switched into the state where most of the light is coupled out. The light in the resonator is thus extracted within about one round-trip time. After that, a new cycle can begin.

A notable difference from Q switching is that before generation of the output pulse the energy is stored in the intracavity light field, rather than in the laser gain medium. The perhaps most important advantage is that this energy can be extracted within just one round-trip time, independent of the time required for building up the intracavity power. Therefore, the pulse duration achievable is more or less determined by the resonator length (provided that the switching time of the modulator is not longer than the resonator round-trip time), and is decoupled from the laser gain. This means that even for very high pulse repetition rates of e.g. several megahertz, pulse durations of a few nanoseconds are still achievable. In that regime, a Q-switched laser would generate much longer pulses, or would even not be able to generate one pulse for each intended time slot. (Pulse dropouts can occur when the Q switch is operated at too high frequencies.) Similarly, the method can be used for lasers with low-gain materials, where short pulses would be difficult to achieve with Q switching.

For high pulse repetition rates (well above the inverse upper-state lifetime), a cavity-dumped laser operates particularly stably, since some light always remains in the laser resonator after pulse extraction, and this light can act as a seed for the next pulse. This is the preferred operation regime.

Although the laser setup for cavity dumping is in principle similar to that for Q switching, the requirements particularly concerning the Q switch are different:

• The switching time should not be longer than the resonator round-trip time, i.e., it should usually not exceed a few nanoseconds (while Q switching works well with much lower switching speeds). For acousto-optic modulators, this requires not only that the RF source can be quickly switched, but also that the beam radius in the modulator is fairly small, since the switching time is limited by the transit time of the acoustic wave through the laser beam. Unfortunately, a small mode area increases the optical intensities in the modulator and may require additional intracavity optics.
• The diffraction efficiency should be high, ideally well above 50%, so that most of the light can be extracted within one resonator round-trip. For that reason, acousto-optic cavity dumpers are often operated with higher RF frequencies of several hundred megahertz, as compared with tens of megahertz for Q switches, and use relatively high RF powers. This is acceptable in terms of heat generation, since the duty cycle of the RP supply is relatively low, while a Q-switch needs to be driven over much longer times.

Since the energy is stored in the light field rather than in the gain medium, cavity dumping can even be applied to some semiconductor lasers, in particular to vertical external-cavity surface-emitting lasers (VECSELs).

From these considerations it follows that cavity dumping is superior to Q switching mainly for high pulse repetition rates, particularly if short pulse durations are required at the same time. On the other hand, this technique introduces additional constraints e.g. on mode areas and switching speeds.

Cavity Dumping of Ultrashort Pulses

Cavity dumping can also be used in mode-locked lasers for generating ultrashort pulses with higher pulse energies. As for nanosecond devices, the resonator losses are kept small for most of the time, so that the circulating pulse can become intense. When it has reached the maximum intensity (and stable pulse parameters), that intracavity pulse can be coupled out with an optical switch. The switching speed required is then even higher than for nanosecond devices. Therefore, an electro-optic modulator in combination with a polarizer is usually required. In addition, the switching must be synchronized with the circulating pulse: a fast photodiode monitors the weak pulses leaking e.g. through a highly reflecting cavity mirror, and the driver electronics use this signal to fire the optical switch at the right time, so that the switching occurs while the pulse is at the other end of the resonator. The pulse duration achievable is in that case, of course, not limited by the round-trip time; a long round-trip time can even be beneficial because it reduces the requirements concerning the switching time.

Figure 2: Schematic setup of a cavity-dumped picosecond laser. The cavity dumper contains a Pockels cell, a quarter-wave plate, and a thin-film polarizer (TFP).

Cavity dumping for ultrashort pulses is mostly used with mode-locked solid-state bulk lasers, such as titanium–sapphire lasers or diode-pumped neodymium-doped or ytterbium-doped lasers. The achieved pulse energy is typically about an order of magnitude higher than with an ordinary mode-locked laser (i.e., typically of the order of 1 μJ), and the pulse repetition rate can be hundreds of kilohertz or several megahertz. The average output power of cavity-dumped lasers is usually significantly smaller than for a mode-locked laser without cavity dumper; this is due to the effect of parasitic resonator losses.

Figure 3: A cavity dumper for use as a pulse picker. For maximum speed, the nonlinear crystal has been directly attached on the electronic driver. The photograph has been kindly provided by EKSPLA Lasers.

The design of a cavity-dumped laser, particularly for ultrashort pulse generation, is not a trivial issue. Apart from the aspect of fast and well synchronized switching, problems can arise e.g. from ringing of the Pockels cell or from the optical nonlinearity and chromatic dispersion introduced by the cavity dumper. The greatly varying intracavity pulse energy is also not compatible with soliton mode locking. Therefore, one should not consider a cavity dumper as a simple add-on to an independently designed mode-locked laser. Instead, the system as whole should be properly designed, taking into account e.g. the effect of the nonlinearity of the cavity dumper on the pulse formation from the very beginning.

For higher pulse energies, lower pulse repetition rates may be considered. However, cavity dumping then no longer works very well: when the intracavity power really vanishes between two extracted pulses, the stability of the process is degraded. For similar reasons, it is also not easy to combine cavity dumping with mode locking and simultaneous Q switching: whereas Q switching can quickly build up a high power in the resonator, the mode-locking process requires much more time to form stable pulses. For low repetition rates (e.g. for 10 kHz or lower) with correspondingly higher pulse energies, regenerative amplifiers are preferable, where the pulse formation and the amplification process are decoupled.

Nonlinear Cavity Dumping

Whereas all previously discussed techniques for cavity dumping are based on some electrically controlled optical switch, it is also possible to do cavity dumping with a nonlinear interaction [8, 11]. Here, the cavity-dumped laser contains a nonlinear crystal, which is phase-matched e.g. for sum frequency generation, and cavity dumping is realized by sending pulses from another laser, which is Q-switched, into that crystal. The energy is then extracted in the form of a short pulse at the sum frequency. The pulse duration is essentially determined by the duration of the Q-switched input pulse. Between the pulses, the nonlinear crystal causes only low parasitic losses in the laser resonator, and thus allows the build-up of a high intracavity circulating power. Note that the circulating intracavity power limits the peak power of the generated pulse, since each output photon requires one photon from each laser to be converted.

If the pulses could be properly synchronized with the relaxation oscillations of the continuously pumped cavity-dumped laser (via a suitable choice of repetition rate), the power efficiency could be enhanced further. Of course, the principle of nonlinear cavity dumping can also be realized with pulsed pumping or Q switching [11].

Bibliography