Femtosecond fiber lasers hit power highs

Gregg Sucha and Heinrich Endert

New developments allow fiber lasers to leapfrog ultrafast solid-state lasers, eliminating the traditional trade-off between package size and power output.

In the ultrafast-laser arena, fiber lasers have usually been regarded as offering a compromise in terms of power versus packaging; that is, while fiber lasers offer a new platform for femtosecond pulse generation that is compact, simple, and well packaged, there has existed a seemingly inevitable trade-off in terms of power. In comparison to pre-existing low-power fiber lasers, solid-state laser counterparts such as Ti:sapphire lasers routinely provide up to 2 W of average power from a modelocked oscillator, with pulsewidths on the order of 100 fs.

This situation is changing, however. Recently, researchers have reported on an ultrafast fiber-laser system that produces average powers and pulsewidths of 13 W and 5 ps (before compression) and 5 W and 100 fs (after recompression), establishing a record in average power for any laser that produces pulses on the order of 100 fs.¹ Additionally, unlike the solid-state lasers, this power is achieved without the use of water cooling.

Cladding pumping eases pump-laser needs

High-power fiber lasers now rival solid-state lasers in terms of continuous-wave (CW) power and promise to revolutionize many different application fields. Developments in fiber-laser technology have raised their output powers to levels far beyond what was previously thought possible. One key technology is cladding pumping of double-clad fiber lasers, which greatly relaxes requirements on the pump lasers' transverse mode. Using this well-established technology, pump lasers are no longer required to have a single transverse mode; it is now possible to couple large amounts of optical power from broad-stripe diodes into the inner cladding, thus providing a method to obtain high average powers. In addition, ytterbium (Yb)-doped fiber lasers provide very high slope efficiencies (greater than 50%), and a cladding pumped Yb-fiber laser has been demonstrated to produce up to 100 W CW of average power.²

But while scaling up the average power of CW fiber lasers is relatively straightforward, achieving tens of watts of average power from fiber-based ultrashort-pulse laser systems presents significant technological challenges. Limits to the pulsed peak power are imposed by the large effective nonlinearity of the fiber core, which can cause severe pulse distortion or breakup if the system is not designed using the proper techniques to mitigate these effects. For example, high effective nonlinearity in fiber dictates that the fiber length must be minimized to reduce the pulse propagation length. Indirectly, this limitation also sets significant constraints on the fiber amplifier pumping, because the limited pulse-propagation length also limits pump-power absorption in a fiber amplifier. As a result of these limitations, previous generations of high-power femtosecond fiber systems have never exceeded on the order of 1 to 2 W of average power.

Key technologies

Recently, however, we have pushed back these barriers using a combination of key technology developments. First, amplifier fibers with larger cores have been used, which provide lower effective nonlinearity (because the core area is larger) thus allowing higher peak powers in the fiber (see Fig. 1). This might seem to be a problem at first glance, because femtosecond lasers require single-transverse-mode propagation, and large-core fibers generally operate with multiple transverse modes. However, with careful design and launching conditions, it is possible to obtain single-mode propagation of pulses in suitably designed large-size and multimode core fibers.³ One additional advantage of large-core fibers is that both the inner cladding and the aperture of the pump diodes can be scaled up; that is, higher-power, low-brightness pump diodes can be used. With core diameters up to 25 µm, the average output power can be scaled up significantly, yet the pulses propagate in a single transverse mode.

Additionally, the amplification of pulses in a fiber amplifier with positive group velocity dispersion (GVD) allows relatively high pulse energies to be amplified. This technique runs counter to accepted ways of thinking about pulse propagation in fibers, because in the recent past it has been considered desirable to take advantage of the solitonic pulse-shaping effects that occur in negative GVD fibers. However, solitonic pulse shaping and modulation instability (which also occurs in negative GVD fiber) set severe limitations on the peak power and can be quite deleterious at higher powers, causing severe pulse distortion. But amplification in positive dispersion fibers allows exploitation of the combined action of large gain, self-phase modulation, and positive dispersion to produce pulses with a nearly linear frequency chirp, even in the presence of large temporal and spectral broadening. In certain regimes, it is even possible to generate and propagate parabolic pulses, which have unique properties.⁴ In the present case, the amplified pulses are linearly chirped and have sufficient bandwidth to be recompressed to durations shorter than 100 fs.

The system consists of a Yb-fiber-based ultrashort-pulse seed laser, a short length of a single-mode fiber for stretching the pulses, and a Yb-fiber power amplifier stage (see Fig. 2). The seed laser operates at 1055 nm and delivers 2-ps pulses at a repetition rate of 50 MHz at an average power of 300 mW. The seed pulses are linearly chirped and have a bandwidth of approximately 20 nm. The power amplifier consists of a 4.3-m length of Yb-doped fiber with a 25-µm-diameter core and 300-µm-diameter inner cladding. The cladding-to-core area ratio is about 144, thus enabling efficient pump absorption in this short-length fiber amplifier. The fiber amplifier is pumped from both ends with two fiber-bundle-coupled laser-diode sources operating at 976 nm. An estimated 14 W of pump power is coupled into each of the amplifier ends, of which 97% is absorbed in the amplifier. The slope efficiency of this large-core fiber amplifier with respect to absorbed pump power is approximately 50%.

A maximum amplified power of 13 W (diffraction-limited) was obtained at the fiber output—with a resulting pulse energy of 0.26 µJ—and the width of the amplified pulses increased from 2 to 5 ps as a result of an induced chirp from the nonlinearity and positive dispersion. These pulses were recompressed down to 100 fs after passing through a conventional diffraction grating compressor at an output wavelength of 1055 nm (see Fig. 3). The grating throughput efficiency was 40%, giving 5 W average power. The use of properly optimized gratings could provide up to 85% throughput, giving more than 10 W of average power with this particular system.

Power is scalable

The use of fiber-based laser architectures no longer sets limits on the average power of ultrafast laser systems, but in fact allows old limits set by solid-state lasers to be exceeded. Additionally, the fiber-based laser systems do not require water cooling, giving them a large advantage over high-power solid-state lasers. However, the key technologies that have pushed the power higher are not limited to the roughly 10 W demonstrated here. The new ultrafast fiber laser technology is readily scalable to significantly higher average powers. The 13-W levels obtained in this work were limited by the available pump laser power, not by the nonlinear effects; even the simple addition of higher-brightness pump diodes could double the output power without requiring reconfiguration. Fiber core diameters can be increased without losing the ability to support primarily single-mode pulse propagation. By scaling up the present design, it should be possible to obtain 100-fs pulses with an average power of up to 100 W.

The production of such high levels of average power will make ultrafast fiber laser technology the workhorse femtosecond laser system of the future and will greatly enhance the usefulness of ultrafast lasers in many scientific, industrial, and medical applications. Additionally, frequency conversion such as second-harmonic generation, optical parametric generation, and sum-frequency mixing will provide high powers at visible wavelengths, making such laser systems useful for full-color (red-green-blue) displays, for example. The higher output powers will greatly increase the throughput of certain applications which, when combined with robust turnkey operation, will bring them out of the laboratory and into widespread use.

REFERENCES

1. A. Galvanauskas and M. E. Fermann, Proc. CLEO 2000 (San Francisco, CA), postdeadline paper CPD3.
2. V. Dominic et al., Proc. CLEO 1999 (Baltimore, MD), paper CPD-11.
3. M. E. Fermann, Opt. Lett. 23(1), 52 (1998).
4. M. E. Fermann et al., Proc. CLEO 2000 (San Francisco, CA), paper CME2.

GREGG SUCHA is a scientist and HEINRICH ENDERT is vice president of marketing and sales at IMRA America, Ann Arbor, MI 48105; e-mail: [email protected].