Divided-pulse Amplification | previous | next | feedback |
(Acronym: DPA)
Definition: a method of amplifying intense ultrashort pulse while avoiding excessive nonlinear effects
Divided-pulse amplification is a novel technique [2], introduced for mitigating the problem of excessive nonlinear phase shifts in amplifiers for ultrashort pulses. Such phase shifts can result from the high peak power of amplified pulses when these propagate e.g. in the gain medium of an amplifier device. They can result in significant spectral broadening and distortion of the pulses, or even in optical damage of the amplifier.
Principle of Divided-pulse Amplification
The basic principle of divided-pulse amplification is to divide each pulse into a sequence of several (or even many) pulses before amplification, and to recombine the pulses after amplification. In principle, this pulse division could be achieved by using beam splitters (e.g. partially reflective dielectric mirrors), but the recombination would then be very difficult. Therefore, the proposed technique for splitting and recombining is one which was invented much earlier [1]. A linearly polarized pulse is split into two pulses by sending it through a birefringent crystal, with an angle of 45° between the original polarization direction and the optical axis of the crystal. The crystal length is chosen so that the group velocity mismatch leads to a relative time delay larger than the pulse duration. The method can then be repeated with additional crystals, which are chosen to be longer than the first one – ideally, the crystal length is doubled in each step, obtaining an equidistant pulse train. The orientation of the optical axis also always changes by 45° for each crystal.
Figure 1: Setup of an amplifier based on divided-pulse amplification. The pulses make a double pass through a fiber amplifier, and the Faraday rotator (FR) separates the amplified pulses from the input pulses at a polarizing beam splitter (PBS), even if the fiber is not polarization-maintaining.
Recombination (after amplification) is then possible e.g. by sending the pulses in the reverse direction through the same sequence of crystals, after rotating their polarization by 90° (e.g. with a Faraday rotator). An alternative is to use a second set of crystals. Interestingly, it is not necessary to match the crystal lengths to a fraction of a wavelength, which would be difficult to achieve; it is only that the recombined pulses will then not be linearly polarized. It may actually occur that the pulse distortions in the amplifier also affect the polarization of the recombined pulses.
As the number of pulses is doubled with each crystal, 10 crystals can generate, e.g., 1024 pulse copies. A 1024-fold reduction in peak power may often be sufficient; note that picosecond pulses need a weaker reduction in peak power than femtosecond pulses with the same pulse energy, because their peak power is lower from the beginning, and in some cases also because narrow-band amplifiers provide higher gain per unit length.
Comparison with Chirped-pulse Amplification
An alternative (older and more common) method for mitigating the problem of excessive nonlinear phase shifts is chirped-pulse amplification, where the pulses are dispersively stretched, then amplified (while having a moderate peak power), and then recompressed in a dispersive compressor such as a pair of diffraction gratings.
The two methods differ in various respects:
- For relatively long (picosecond) pulses, chirped-pulse amplification (CPA) requires impractically large amounts of dispersion. This problem does not exist with divided-pulse amplification (DPA).
- For fairly short (femtosecond) pulses, the effect of pulse broadening via chromatic dispersion of the birefringent crystals can become problematic with DPA, even though the use of optimized materials may allow for significant further improvements. (For shorter pulses, shorter crystals may be used, but this does not fully compensate the effect of shorter pulses, since the sensitivity of pulses to dispersion scales with the inverse square of the pulse duration. The effect of optical nonlinearities scales in a more benign manner.)
- The loss of pulse energy can be significantly lower with DPA than with CPA with a pair of diffraction gratings.
- The difficulty in obtaining linearly polarized recombined pulses with DPA may be a significant disadvantage in many cases. Also, the technique is less practical with amplifiers which require linear polarization, since the different polarization components then have to be processed separately. In these respects, CPA is better, except perhaps if the amplifier used is not polarization-maintaining.
- Concerning alignment, CPA can be difficult when a pair of diffraction gratings with large dispersion is used. This problem does not exist with DPA.
In conclusion, it appears that chirped-pulse amplification is generally more suitable for pulse durations far below 1 ps, whereas divided-pulse amplification offers advantages for longer pulses. The decision regarding a particular method can, however, also depend on other circumstances such as details concerning polarization.
Bibliography
| [1] | H. E. Bates, R. R. Alfano, and N. Schiller, "Picosecond pulse stacking in calcite", Appl. Opt. 18 (7), 947 (1979) |
| [2] | S. Zhou, F. Wise and D. G. Ouzounov, "Divided-pulse amplification of ultrashort pulses", Opt. Lett. 32 (7), 871 (2007) |
See also: amplifiers, ultrashort pulses, nonlinearities, nonlinear pulse distortion, chirped-pulse amplification, Spotlight article 2007-03-11
