By coherently combining the outputs from multiple lasers into one beam, scientists and engineers can generate highly coherent light at previously unavailable powers. Coherent beam combination is easiest to achieve for continuous-wave (CW) beams, and has led to high-power lasers for astronomical adaptive optics, potential military applications, and future free-space communications systems.
Although coherent combination of nanosecond-pulsed beams has been investigated, coherently combining pulsed beams is in general more difficult than for CW beams.1 However, this is where some of the benefits of coherent combination are most needed, as more conventional ways of boosting pulse peak powers can induce nonlinear optical effects that degrade the beam. For example, scaling up the peak power of a single femtosecond fiber amplifier creates nonlinearities; trying to avoid this by increasing the fiber diameter creates a less desirable, multiple-transverse-mode beam.
Researchers at the Laboratoire Charles Fabry de l’Institut d’Optique and the Office National d’Etudes et de Recherches Aérospatiales (both in Palaiseau, France) and Amplitude Systèmes (Pessac, France) have demonstrated coherent beam combining of two femtosecond fiber chirped-pulse amplifiers (CPAs), paving the way to creating very high peak and average powers with multiple combined devices.2 The experimental setup has a combining efficiency of 90% and produces an average power of 7.2 W before compression, with a 325 fs pulse duration after compression while maintaining a high-quality beam.
Phase-related parameters
The coherent combination of CW beams is relatively straightforward, with the critical requirement being the matching of the relative phase between the input beams. Femtosecond pulsed beams are another story entirely: Not only the relative phase but the group delay and group-delay dispersion must be matched as well.
As the researchers note, the sensitivity of the beam-combination setup to mismatches of the phase-related parameters rises as the desired pulse duration grows shorter. They have found that for pulses longer than about 100 fs, the group delay and group-delay dispersion can be statically rather than dynamically adjusted without adversely affecting the beam-combination efficiency (the relative phase must still be actively corrected).
A single-seed light source is used for the experiment; its output is split and fed to two separate amplifiers whose outputs are then combined. The seed laser is a bulk ytterbium:potassium yttrium tungstate (Yb:KYW) oscillator that produces 260 fs pulses at 35 MHz, with 1 W average power at a 1030 nm wavelength. A stretcher lengthens the pulses to 150 ps, and a half-wave plate and a polarizing beamsplitter divide the output into two beams with average powers that are individually adjustable (see Fig. 1).
The first beam is coupled into a fiber arm containing an integrated phase modulator and a 1.2-m-long, large-mode-area (LMA) fiber amplifier; the second beam is coupled into a second fiber arm that includes a section of singlemode fiber that tailors the group-velocity dispersion and at the same time functions as a coarse delay-matching element, with the output feeding into another 1.2-m-long LMA fiber amplifier.
Both fiber amplifiers are seeded with light at a 100 mW average power, a level low enough that they do not suffer nonlinear optical effects. The amplifier outputs are collimated and then brought together in a 50/50 beamsplitter. The portion of the two beams that are successfully combined passes to a pulse compressor, while the rest passes through the other face of the beamsplitter to a monitor photodiode; maximizing the coherent beam combination minimizes the signal received by the photodiode.
Phase control
The integrated phase modulator has a bandwidth greater than 1 GHz and is used to control the relative phase between the two amplifiers. The control signal includes a small 250 kHz voltage modulation that produces a 0.1 rad modulation of the relative phase; this is sensed by the photodiode and converted via a lock-in amplifier into an error signal fed back to the phase modulator (the integration time of the lock-in amplifier determines the maximum phase-noise-correction frequency, which is 25 kHz). The residual photodiode intensity fluctuations upon phaselocking correspond to a root-mean-square (rms) phase noise of only λ/40 over a 100 s period, with no readjustment of the delay line needed to keep the phase locked over hours-long periods.
In addition to minimizing the phase difference between the two beams, their arrival time, temporal shape, polarization, and spatial profile must be synchronized. Because both amplifiers produced light linearly polarized to better than 99%, polarization matching was a given. Optimum spatial beam overlap, however, required careful alignment of defocus, tilt, and beam profiles. The output beam was intermediate in shape between the two input beams (see Fig. 2).
The fiber amplifiers each produce 4 W average power; after coherent beam combining, the total output was 7.2 W, for a combining efficiency of 90%. The pulses were subsequently compressed to their final duration of about 325 fs at a compression efficiency of 60%. The pulse lengths for the individual fiber arms were 315 and 335 fs; the pulse length for the combined beam was 325 fs. Similarly, the recombined spectrum was intermediate between the slightly mismatched spectra for the individual arms.
The researchers believe their technique can be used to combine many amplified beams seeded by the same oscillator. For example, a setup very similar to theirs could be created to combine the output from a few tens of amplifiers, they note. “We are currently investigating the behavior of this same system with high accumulated nonlinear phase, and also building other architectures that allow passive control of the phase for the combination of several amplifiers (we have demonstrated the passive combination of two fiber amplifiers),” says Louis Daniault, one of the researchers.
REFERENCES
1. E.C. Cheung et al., Adv. Solid-State Photon., paper WA2 (2008).
2. L. Daniault et al., Opt. Lett., 36, 5, 621 (2011).
