1. Introduction

Stimulated Raman scattering (SRS) has attracted much attention in optical fiber as it can turn otherwise passive optical fibers into broadband Raman amplifiers and tunable Raman lasers [1,2]. While this has been applied successfully by many groups to develop CW sources [3,4], significant efforts have also been devoted to investigating the use of SRS in the pulsed regime [5–12]. For ultrashort pulses (pulse width below 100 picoseconds), SRS is limited by the walk-off effect between the pump and Raman wavelengths which restricts the lengths of Raman fiber that can be used and thus requires relatively high peak powers [5,6]. Moreover, self-phase modulation (SPM) and four-wave mixing (FWM) can also become important and significantly affect the depletion of the pump power and the Raman pulse generation process [7,8].

On the contrary, in the quasi-CW pulse domain in which the pulse widths are between several to hundreds of nanoseconds, one can neglect the effect of group velocity mismatch between the pump and Raman wavelengths. Consequently a long Raman gain fiber can be used dramatically reducing the pulse peak powers requirements. Therefore higher order Raman Stokes components can be generated in either the near-infrared or visible spectral regions [9,10] using modest peak power pulses. For most commonly generated pulse shapes, e.g. Gaussian pulses, the effectively instantaneous Raman response time means that different parts of the pulse experience different levels of Raman gain. As a consequence the relative amount of light in successive Raman Stoke orders changes across the pulse giving rise to pulses with highly complex spectral-temporal profiles. This problem can be overcome by using rectangular shaped optical pulses such that a common frequency shift is obtained across the entire pulse envelope. Most nanosecond pulsed sources (e.g.Q-switched laser) generate Gaussian type pulses however it is possible to achieve nanosecond rectangular pulses with pulse energies in the mJ regime using active pulse shaping techniques in fiber based MOPA systems [13]. Using such a pulse shaping system and exploiting the uniform Raman gain experienced for rectangular pulse shapes we recently demonstrated the selective excitation of Raman Stokes lines up to the 3rd order pumped at 1060 nm and up to the 9th order pumped at 530 nm [11,12].

In this paper we demonstrate an extension of this approach to the use of pump pulses incorporating multi-amplitude-levels which enables the generation of Raman-shifted pulses in which the frequency changes in discrete steps across the pulse form in accordance with the pump amplitude profile. The technique should allow the realization of a new generation of wavelength agile fiber based laser systems operating across broad and interesting wavelength regimes.

To illustrate the approach we present the generation of pulses of several tens to hundred nanoseconds duration in which the wavelength changes sequentially from green (543 nm, 1st Stokes), to yellow (585 nm, 4th Stokes) to red (616 nm, 6th Stokes) using 530 nm pump pulses with a 3-step profile. Changing the ratio of the peak powers (step height) between the steps, and the temporal order of the step-changes allows the generation of sequences of colored pulses according to any desired combination of Raman Stokes Shift, providing dynamic and agile frequency tuning between well-defined wavelengths. The time averaged power distribution between the different Raman Stokes order components is also adjustable simply by varying the step widths.

The pump pulses were readily generated using a frequency-doubled fiber master oscillator power amplifier (MOPA). The configuration of the 1060 nm MOPA is similar to that reported previously [12,14,15] other than that we replaced all of the preamplifiers and the final stage amplifier with polarization-maintaining (PM) versions so that the whole system was fully PM. This provides advantages in terms of better long term stability and higher conversion efficiency. A Lithium Triborate (LBO) crystal was chosen for second-harmonic generation (SHG) because of its high damage threshold relative to other potential crystal choices such as PPLN or KTP. We used Pirelli Freelight fiber as the Raman gain medium and a length of 250 m was selected representing a reasonable compromise between the need to reduce attenuation and Kerr nonlinearities (e.g. SPM and FWM), whilst providing adequate nonlinear Raman gain at reasonable peak power levels.

2. System setup

A schematic diagram of the all-fiber PM MOPA, SHG (frequency doubler) and Raman converter is illustrated in Fig. 1 . The seed was a fiber pigtailed Fabry-Perot laser diode (Bookham CPE425) wavelength stabilized at 1060 nm with an external PM fiber Bragg grating. The optical output of the seed laser passed through an in-line high extinction ratio (~45dB) lithium niobate electro-optic modulator (EOM), which was driven by an arbitrary waveform generator (AWG) with 4 ns temporal feature resolution. The seed signal was adaptively shaped by the EOM and amplified by an all-fiber three-stage PM amplifier chain to deliver single mode, single polarization output of the required pulse shapes and power levels.

 

Fig. 1 Schematic of the Yb-doped all-fiber PM MOPA, SHG cavity and Raman converter.

Download Full Size | PDF

The first pre-amplifier stage is based on a core pumped, Yb³⁺-doped PM fiber (Nufern PM-YDF-5/130) with a mode field diameter and NA of 6.5 µm and 0.13 respectively. The amplifier was bi-directionally pumped by wavelength stabilized 975 nm single-mode laser diodes. The core absorption of the fiber was 680 dB/m at 975 nm. A 2.5 meter length of this fiber was used such that the amplifier provides maximum gain around 1060 nm. The output of the first stage amplifier was coupled into the second pre-amplifier stage via an optical isolator. The isolator limits both signal and ASE back-coupling between the two amplifier stages.

The active medium of the second-stage amplifier was a double-clad Yb³⁺-doped PM fiber (the same Nufern PM-YDF-5/130 but now operated in a cladding pump configuration). The cladding diameter of the fiber was 130 µm and the pump NA was 0.46. The measured absorption of the fiber at the pump wavelength (975 nm) for light launched into the cladding was 1.7 dB/m. A 6 m length of fiber was used such that the amplifier provides maximum gain at the signal wavelength. The gain medium was counter-pumped by a 975 nm fiber-pigtailed broad-stripe diode laser through a single (6 + 1) fused tapered fiber bundle (TFB).

The output from the 2nd pre-amplifier stage was spliced to the slow-axis of an in-line fast axis blocking PM isolator to reduce the non-polarized ASE and to increase the optical signal to noise ratio (OSNR) into the final amplifier. The resulting single polarization signal was then coupled into a large mode area (LMA), PM, double-clad, Yb³⁺-doped active fiber (PLMA-YDF-25/345 from Nufern). The fiber has a core diameter of 25 µm and a core NA of 0.06. The cladding diameter of the polymer coated fiber was 345 µm with a cladding NA of 0.45. The measured cladding absorption of the fiber at the pump wavelength (975 nm) was 2.6 dB/m. A 4 m length of this fiber was chosen such that the amplifier not only provides optimum signal gain but also absorbs most of the launched pump power. To minimize the splice loss between the passive single-mode PM fiber (SM98-PS-U25A from Fujikura) and the LMA active fiber the outer diameter of the LMA fiber was tapered down to 125 µm and a core diameter of 9 µm. This helps to reduce the mode field diameter mismatch between the two dissimilar fibers ensuring lower splice loss. Moreover tapering allowed us to obtain robust single mode operation even though the active fiber core can support several transverse modes. To prevent damage to the output, a 2 mm long pure silica mode-expanding end-cap was spliced to the fiber end which was angle-polished to avoid power being retro-reflected back into the fiber core. The amplifier was end-pumped using a 975 nm diode stack with a maximum pump power of 166 W. The diodes were water-cooled to ensure good wavelength stability. A simple lens combination was used to achieve ~80% coupling efficiency into the fiber. The signal and pump paths were split by dichroic mirrors.

The collimated 1060 nm MOPA output was frequency doubled using a 15mm long LBO crystal. The output was launched into the LBO via a focusing lens with a focal length of 100 mm. The diameter of the focused beam at the waist position was 70 µm corresponding to a Rayleigh range of 12 mm. The crystal was cut for noncritical phase matching at an operating wavelength of 1060 nm and was placed in an oven at a constant temperature of 155 °C for maximum frequency conversion. A free space isolator was used to protect the MOPA chain whilst a polarization beam splitter (PBS) was used to clean up the output polarization which was degraded slightly by the output isolator. The linear polarization of the output was rotated by a half-wave plate (HWP) to align to the principal axis of the LBO crystal for maximum conversion efficiency. The output from the LBO was passed through a dichroic mirror (DM) which rejects the unconverted 1060 nm pump light. The transmitted 530 nm light was collimated and coupled into a 250 m long Pirelli Freelight fiber using an aspheric lens with focal length of 11 mm to generate successive Raman Stokes lines.

3. Experimental results

To obtain the desired spectral-temporal Raman Stokes pulse profiles at the system output we used adaptive pulse shaping to generate multilevel pulse forms at the input to the Raman fiber. First we demonstrate the sequential generation of green (1st Stokes), yellow (4th Stokes) and red (6th Stokes) pulses using a 3-step input pulse. The MOPA was operated at a repetition rate of 50 kHz and the pulse width was set at 50 ns for each step (with a 4 ns transition between levels). In order to determine the required ratio of three-step heights, we first investigated the respective peak powers required to generate separate orders (1st, 4th and 6th) using simpler rectangular SHG pump pulses [12]. We then established the target multi-level SHG pump pulse shape and from this determined the fundamental MOPA output pulse shape required and from that the corresponding input optical pulse to be created using the EOM at the MOPA input. The measured MOPA output and SHG pulses are illustrated in Fig. 2(a) whereas the inset shows the required input pulse shape from the EOM. It is to be noted here that the step height ratios between the MOPA and SHG pulses are different due to the nonlinear power dependency of the SHG (quadratic when operating at relatively low power). The measured average output powers of the MOPA and SHG were 3.1 W and 275 mW respectively and their spectra are shown in Fig. 2(b) and 2(c) respectively. The MOPA was operated at a center wavelength of 1060.25 nm with a full width at half maximum (FWHM) spectral width of 0.35 nm. The OSNR was higher than 30 dB and the polarization extinction ratio was better than 20 dB. The center wavelength of the SHG was 530.12 nm with a FWHM of 0.17 nm.

 

Fig. 2 (a) 3-step output pulses of MOPA, SHG and Raman lines (inset: initial pulse shape from the EOM); (b) Spectrum of MOPA output; (c) Spectrum of SHG output.

Download Full Size | PDF

Figure 3(a) shows proof of the sequential excitation of 1st (543 nm), 4th (585 nm) and 6th (616 nm) Raman Stokes lines. The extinction ratio of the selected Raman lines was better than 10 dB relative to their nearest neighbors (in terms of the time average powers) whilst the 1060 nm pump depletion was ~15 dB. The spectrum broadened and became asymmetric for higher order Stokes due to the presence of several competing nonlinear processes [16,17]. The measured peak powers of 1st, 4th and 6th Raman Stokes were 3 W, 9.5 W, 22 W respectively. The higher peak powers for the higher order Stokes lines is predominantly determined by the higher threshold power although the fact that the transmission loss in the Raman fiber decreases at longer wavelengths also has an impact. The pulse shape of the Raman Stokes output is shown as the green plot in Fig. 2(a) and blue plot in Fig. 3(b). Comparing with the SHG pulse in Fig. 2(a), the overall pulse width of the Raman Stokes shifted output pulse is slightly longer because of the walk-off among the Raman Stokes and pump wavelengths. We then used a grating to diffract the output from the Raman fiber to establish the temporal distribution of different Stokes orders. The results are shown in Fig. 3(b) where it is seen that each of the orders are confined to specific temporal regions across the pulse as expected. Note that the majority of residual 530nm pump remains at the trailing edge of the overall pulse.

 

Fig. 3 (a) Spectrum of Raman pulses with simultaneous excitation of green, yellow and red light (inset: picture of the Raman output diffracted by a grating); (b) Overall and separate 3-step output pulses of Raman output.

Download Full Size | PDF

Similarly we can generate a combination of two Stokes lines using 2-step pulses as shown in Fig. 4 . The time average power distribution of the different Raman Stokes can be shown to be adjustable by varying the step widths (the peak power is obviously defined by the requirement for a fixed level of gain to achieve complete Raman transfer from the fundamental). The blue plots in Fig. 4(a) and 4(b) show the temporal and spectral profiles of the 1st and 4th Stokes orders with average output powers of 7 mW and 22 mW respectively for equal pump pulse step durations of 50 ns. In order to obtain the same average output power at both Stokes lines, the duration of the pump pulse corresponding to the 1st order stokes line was tripled compared to that of the 4th order Stokes line (150 ns and 50 ns respectively) as depicted in Fig. 4(a) so that both the pump pulse components have equal energies. This resulted in equal output powers of 21 mW for both the Stokes lines as indicated by the pink plot in Fig. 4(b). This also shows the flexibility of varying the temporal orders of the 1st and 4th Stokes component, as well as the possibility of incorporating a user defined temporal separation between the different colored pulses. Given a specified fiber length, changing the ratio of the peak powers (step height) between the steps, and the order of the step-changes allows the generation of sequences of colored pulses according to any desired combination of Raman Stokes Shift, providing dynamic and agile frequency tuning between well-defined wavelengths.

 

Fig. 4 (a) 2-step output pulses of Raman conversion with different pulse energy and separation; (b) Spectra of the Raman pulses: exciting green and yellow simultaneously (inset: picture of the Raman output diffracted by a grating).

Download Full Size | PDF

We have also investigated potential routes for scaling up the Stokes shifted average output powers. This can be done either by increasing the pulse duration (at fixed pulse repetition rate) so as to increase the energy per pulse at a given Stokes line, or by shortening the Raman gain medium at a fixed pulse duration and repetition rate to increase the peak power required to achieve energy transfer to a given Raman order. Figure 5 illustrates an example of scaling using the later approach where the 1st and 4th order Stokes lines were excited in a 100m length of Freelight fiber (the output pulse shapes were kept the same as for the 250 m long fiber case). To obtain the same sequential pattern of excitation of Raman Stokes orders, 284 mW of 532 nm pump was required for the 100 m long fiber giving a total average Stokes shifted power of 115 mW. This is to be compared with values of 34 mW of output for 116 mW of input when the 250 m fiber was used. The numbers are consistent with theoretical predictions when fiber attenuation in the visible and pump coupling loss are taken into account. Note that the long wavelength spectral components appear broadened and more asymmetric for the shorter fiber length which we believe to be induced by the relatively stronger SPM and XPM in this case [16,17].

 

Fig. 5 Spectra of the 1st and 4th Raman Stokes with 2-step output pulses in different lengths of Raman gain fiber.

Download Full Size | PDF

4. Conclusions

In summary, we demonstrate the generation of pulses of several tens to hundreds of nanoseconds in duration in which the wavelength changes sequentially from green, to yellow to red using 530nm pump pulses with a 3-step profile or the combination of any two using 2-step pulses by exploiting SRS in optical fiber. The multilevel pump pulses were generated using a frequency-doubled fiber MOPA. The technique should allow the realization of a new generation of wavelength agile fiber based laser systems operating across broad and interesting wavelength regimes. Such sources could find potential applications in the fields of spectroscopy, metrology and biology to name but a few.