Nonlinear Fiber Optics - 4 ed. Agrawal

312 Chapter 8. Stimulated Raman Scattering
output in the form of solitons of widths ∼100 fs, but at a wavelength corresponding
to the first-order Stokes wavelength. Furthermore, the wavelength can be tuned over a
considerable range (∼10 nm) by using the time-dispersion technique discussed in Section
8.2.2. A ring-cavity configuration, shown schematically in Figure 8.23, is commonly
employed. A dichroic beam splitter, highly reflective at the pump wavelength
and partially reflective at the Stokes wavelength λ s , is used to couple pump pulses into
the ring cavity and to provide the laser output.
In a 1987 experimental demonstration of the Raman soliton laser [174], 10-ps
pulses from a mode-locked color-center laser operating near 1.48 μm were used to
pump the Raman laser synchronously. The ring cavity had a 500-m-long, polarizationpreserving,
dispersion-shifted fiber having its zero-dispersion wavelength λ D near 1.536
μm. Such a fiber permitted the pump and Raman pulses to overlap over a considerable
portion of the fiber as the pump and Raman wavelengths were on opposite sides of
λ D by nearly the same amount (λ s ≈ 1.58 μm). Output pulses with widths ∼300 fs
were produced with a low but broad pedestal. In an attempt to remove the pedestal,
the ring cavity of Figure 8.23 was modified by replacing the fiber with two fiber pieces
with variable coupling between them. A 100-m-long section provided the Raman gain
while another 500-m-long section was used for pulse shaping. The SRS did not occur
in the second section because the coupler reduced pump power levels below the
Raman threshold. It was possible to obtain pedestal-free pulses of 284-fs width when
wavelength separation corresponded to 11.4 THz (about 90 nm). However, when the
wavelengths were 13.2 THz apart (corresponding to maximum Raman gain), considerable
pulse energy appeared in the form of a broad pedestal. This complex behavior can
be attributed to the XPM effects.
In later experiments, 100-ps pulses from a mode-locked Nd:YAG laser operating at
1.32 μm were used to synchronously pump a Raman soliton laser [175]–[177]. This
wavelength regime is of interest because conventional fibers with λ D ∼ 1.3 μm can be
used. Furthermore, both the pump and Raman pulses are close to the zero-dispersion
wavelength of the fiber so that they can overlap long enough to provide the required
Raman gain (the walk-off length ≈300 m). Pulse widths as short as 160 fs were obtained
in an experiment that employed a 1.1-km fiber that did not even preserve wave
polarization [175]. Output pulses contained a broad pedestal with only 20% of the
energy appearing in the form of a Raman soliton. In another experiment, a dispersionshifted
fiber with λ D = 1.46 μm was used [176]. Raman solitons of about 200-fs
width were then observed through the second- and third-order Stokes lines, generated
near 1.5 and 1.6 μm, respectively. This process of cascaded SRS has also been used
to generate Raman solitons near 1.5 μm by pumping the laser with 1.06-μm pump
pulses [178]. The first three Stokes bands then lie in the normal-GVD regime of a conventional
fiber (λ D > 1.3 μm). The fourth and fifth Stokes bands form a broad spectral
band encompassing the range 1.3–1.5 μm that contains about half of the input energy.
Autocorrelation traces of output pulses in the spectral region near 1.35, 1.4, 1.45, and
1.5 μm showed that the energy in the pedestal decreased as the wavelength increased.
In fact, output pulses near 1.5 μm were nearly pedestal free.
Even though Raman soliton lasers are capable of generating femtosecond pulses
that are useful for many applications, they suffer from a noise problem that limits their
usefulness. Measurements of intensity noise for a synchronously pumped Raman soli-