High-power supercontinua

Recently developed fully integrated all-fiber supercontinuum sources allow 5mW/nm spectral power density in the blue/uv, while cw pumped 50mW/nm is obtained in the near IR.
14 November 2007
Burly Cumberland, Andrey Rulkov, John Travers, and Sergei Popov

Supercontinuum generation (400nm-700nm) in glass was first reported by Alfano and Shapiro in 19701, achieved through the focusing of a high-power frequency-doubled Nd:glass laser into a bulk sample a few millimeters long. Optical fiber was first used as the supercontinuum generation medium pumped by a dye laser by Lin and Stolen in 19762, who noted the potential of the source for measurement in excited-state spectroscopy. For 25 years, time-resolved laboratory spectroscopy was the major application with various picosecond and femtosecond pulsed lasers used as the fundamental excitation source for the supercontinuum generation. Practically all possible nonlinear processes in fiber contribute to the supercontinuum, including self-phase modulation, four-wave mixing, Raman scattering, soliton generation, soliton collision and self interaction, with the dominant process or processes depending on the pump laser format.

In 1999 Ranka et al reported the use of photonic crystal fiber (PCF) as the nonlinear medium3 and stimulated a renaissance in research in supercontinuum generation. The introduction of PCF has allowed complete manipulation of the physical optical parameters of the fiber such that nonlinear phenomena can be observed in new wavelength regions that would have been impossible using standard fiber structures or known laser sources. This absolute control has meant that nonlinear optical response of the PCF to the propagating light, for example, can be enhanced or if required, totally eliminated, simply by appropriate fiber design. This has permitted innovation in an absolutely new family of fiber-based supercontinuum light sources and instruments that is poised to supersede conventional lasers in a vast number of important applications.

Despite the introduction of PCF, the average spectral power density within the continuum obtained with various laser pumping schemes remained modestly low, of the order of milliwatts, while the general experimental configuration was effectively unchanged from the original, favoring free space lens coupling between pump laser and fiber3-5. Consequently, instability and reproducibility were associated problems and although acceptable as a laboratory device such an arrangement could not meet the "hands off" requirement of a commercial applications tool.

The past decade has seen remarkable progress in the development of high-power fiber lasers (see for example www.ipgphotonics.com). Of the many systems, probably the Yb-doped fiber laser is the most mature and consequently represents the most attractive candidate for pumping and achieving a completely fiber-integrated supercontinuum source. In 2004, the Femtosecond Optics Group at Imperial College in collaboration with IPG Photonics reported the first all-fiber, fully integrated, high power supercontinuum source operating from 525 nm to beyond 1800nm6. A picosecond Yb-based MOPFA (master oscillator power fiber amplifier) with an average power of 8W was used as the pump and greater than 1mW/nm spectral power density was achieved throughout the supercontinuum. Optimization of the power-length product yielded little or no effect on the short wavelength extreme of the supercontinuum and although power scaling was possible to 20W average power and spectral power densities in excess of 5mW/nm, the supercontinuum did not extend significantly below 525nm. However, the spectral power density and spectral coverage together with the attractiveness of the fully fiber-integrated format stimulated several commercial companies to utilize the technique and configuration.

With picosecond pumping, the major nonlinear processes contributing to supercontinuum generation are soliton generation, soliton self interaction and four wave mixing. Consequently, it is vital to operate with the fundamental pump in the region of the zero dispersion of the nonlinear fiber. With PCF this is readily achievable at any wavelength in the near infra red. Self phase modulation and nonlinear spectral broadening of the pump extends into the region of normal dispersion where these components can act as an efficient pump for four wave mixing processes. For a PCF with a zero dispersion around 1040 nm and minimum of the pump wavelength of 960 nm, efficient phase matching can take place between Stokes and anti-Stokes components at 2100nm and 520 nm respectively. However, beyond 2100nm material and waveguide loss increases significantly and so longer wavelength generation is inhibited, which correspondingly prevents efficient four wave mixing of wavelengths below 520nm, as verified experimentally. As a consequence, reported supercontinuum spectra tend to operate above 500nm. There are, however, applications areas in the biosciences and other areas where extensive wavelength coverage in the blue/uv is required.

By concatenating lengths of PCF where the zero dispersion wavelength of the fiber segments decreased with propagation along the fiber, the four-wave mixing process can be enhanced, ensuring that short wavelength radiation generated in the preceding segment can act as the pump in the normal dispersion, nearer to the zero dispersion wavelength and requiring a shorter wavelength Stokes component for efficient four wave mixing. In 2005, the Femtosecond Optics Group successfully demonstrated this technique using only two stages of PCF in a fused, step dispersion profile fiber7. The supercontinuum extended by nearly 200 nm down to 440 nm and spectral power densities in excess of 2mW/nm were achieved in the blue.

This technique can be further refined such that as the pump pulse propagates in the PCF the sequentially generated short wavelength component of the supercontinuum shifts in synchronism with the zero dispersion of the fiber, ensuring optimum four-wave mixing. This is achieved by tapering the core of the PCF and is undertaken as the fiber is pulled from the perform at the manufacturing stage, through varying the speed of the pull. What distinguishes this process from conventional tapers manufactured using filament and arc splicers or hydrogen flame is that long lengths (tens to hundreds of meters) of controlled tapering can be achieved which is essential to optimize the phase matched four wave mixing processes associated with the fundamental pump pulses in the picosecond range. In association with Professor Jonathan Knight's group at the University of Bath, who manufactured the precision tapered fibers, the Femtosecond Optics Group demonstrated high power supercontinuum generation from 340 nm to beyond 2200nm8. Figure 1 below shows the blue/uv components of two supercontinua generated in two tapers with slightly different tapered profiles, illustrating that up to 4mW/nm could be generated down to 400 nm while as low as 340 nm 2mW/nm can be achieved, with such power densities and above also achievable to the long wavelength region. Naturally, other nonlinear processes contribute to the supercontinuum generation process, in particular, soliton trapping of the dispersive waves can significantly enhance the blue/uv. Through optimization of the tapered fiber dispersion profile, extension to 300 nm can be readily achieved as can scaling of the spectral power density to 10mW/nm.

Figure 1. UV/Visible spectral region of supercontinuum generated in two tapered PCFs

It is not necessary to use pulsed pumps in order to generate supercontinuum in optical fiber. CW pumping schemes can be used and this is particularly attractive since cw based fiber lasers with power scaling from many tens to hundreds of watts have remained extremely compact and cost effective. Supercontinuum generation using both conventional and photonic crystal fiber with cw fiber-laser excitation was first demonstrated by the Femtosecond Optics Group in 2002. Naturally the dominant mechanism for supercontinuum generation differs to that for picosecond pumping, but had been well established from the early soliton Raman generation experiments of the mid 1980s9. For supercontinuum generation around the pump wavelength of a cw Yb fiber laser, PCF is essential. It is important for the pump to lie in the region of anomalous dispersion such that modulational instability leads to the rapid formation of many soliton-like structures, these solitons experience gain, temporally narrowing and consequently they experience self-Raman interaction, spectrally shifting to longer wavelengths. Soliton Raman gain and soliton-soliton collision also contribute to the supercontinuum generation process. As a result of the large number of interacting solitons, the generated supercontinuum tends to be relatively flat in intensity, however, because of the soliton-Raman nature of the generation process, the supercontinuum develops only to wavelengths longer than the pump. Other nonlinear processes can generate anti-Stokes components, however, the wavelength extent is less significant than the Stokes components. Pumping using Yb:Er based fiber lasers at 1550 nm in conventional dispersion shifted fiber has permitted supercontinuum generation from 1550 nm to beyond 2000 nm in continua with better than 0.7 dB peak-to-peak flatness and 16 mW/nm spectral power density10.

Originally, Yb-based systems did not exhibit such extensive wavelength coverage11 and although supercontinua of greater than 300 nm width were obtained, allowing optical coherence tomographic measurements with a spatial resolution of better than 4 µm in tissue, the upper spectral extent was limited to about 1380 nm. This was associated with the water absorption loss arising from water attached to the air-glass surface of the air holes of the fiber. Even though manufacturing processes were improved in later generations of PCF, the loss associated with the extended length of fiber essential for supercontinuum generation using modest cw pump powers, inhibited long wavelength development beyond the water peak.

The problem of the distributed loss inhibiting the soliton Raman generation process can be simply solved by reducing the fiber length while maintaining the power-length experimental factor which consequently requires an increased pump power. This has recently been demonstrated by the Femtosecond Optics Group. Fusion splicing 25 m of PCF with a zero dispersion at 860nm to the output of a 50W Yb-doped fiber laser has resulted in a 30W average power supercontinuum that extends from 1100 nm to 1700 nm with an unprecedented spectral power density of 50mW/nm in the supercontinuum up to 1380 nm. Figure 2 shows the spectral output of this cw-pumped supercontinuum. The long wavelength limit in this case is set by the second zero dispersion of the PCF, which is at 1567nm, beyond which the fiber is normally dispersive and cannot sustain the self-Raman-shifting soliton components. Beyond this, dispersive wave components can contribute a limited extension and possible Raman orders to the continuum.

Figure 2. Generated cw pumped supercontinuum with >50mW/nm 1100nm - 1380 nm

Through enhanced fiber design it should be possible to achieve a supercontinuum extending beyond 1700nm and generate spectral power densities of 100mW/nm over the range 1000nm to 2000nm in very compact packages.

As a result of their pioneering advances in the development of high power fiber-integrated supercontinuum sources, the potential of commercial development and new proposals for blue/uv generation and extension to the near and mid-infrared, the Femtosecond Optics Group were recently awarded the Imperial College Research Excellence Award that will allow the group to expand and develop these versatile, widely applicable and diverse sources.

Burly Cumberland, Andrey Rulkov, John Travers, Sergei Popov
Femtosecond Optics Group, Imperial College London
London, UK
www.imperial.ac.uk/research/photonics/femto/

References

1. R. Alfano and S.L. Shapiro, Phys Rev Lett 24, 1217 (1970).
2. C.Lin and R.H. Stolen, App Phys Lett 28, 216 (1976).
3. J.K. Ranka, R.S. Windeler and A.J. Stentz, Optics Lett 25, 25, (2000).
4. S. Coen, A.H.L. Chau, R. Leonhardt, J. D. Harvey, J.C. Knight, W.J. Wadsworth and P. S. Russell, Opt Lett, 26, 1356 (2001).
5. L.Provino, J.M. Dudley, H. Maillotte, N. Grossard, R.S. Windeler and B.J. Eggleton, Elect Lett 37, 558 (2001).
6. A.B. Rulkov, A.G. Getman, M.Y. Vyatkin, S.V. Popov, V.P. Gapontsev and J.R. Taylor, Post Deadline Paper CPD7 Conference on Lasers and Electro-Optics, CLEO 2004.
7. J.C. Travers, S.V. Popov and J.R. Taylor, Opt Lett, 30, 3132 (2005).
8. A. Kudlinski, A.K. George, J.C. Knight, J.C. Travers, A.B. Rulkov, S.V. Popov and J.R. Taylor, Opt Exp 14, 5715 (2006).
9. A.S. Gouveia-Neto, A.S.L. Gomes and J.R. Tayor, Opt Lett 12, 1035 (1987).
10. S.V. Popov, A.B. Rulkov and J.R. Taylor, Paper TuA6 Advanced Solid State Photonics, Santa Fe, NM, (2004).
11. J.C. Travers, R.E. Kennedy, S.V. Popov, J.R. Taylor, H. Sabert and B. Mangan, Opt lett 30, 1938 (2005).
PREMIUM CONTENT
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research