1. Introduction

Over the past decade fiber laser technology has made dramatic progress in various operating regimes [1–3], being nourished by the inherent advantages of optical fibers as waveguiding gain media in terms of obtaining single-pass gain, beam quality, efficiency, heat dissipation, power-scalability, etc. Subsequently, fiber lasers have become more common and popular as coherent light sources in many areas of science and technology [1–3]. In particular, the research field of fiber-based supercontinuum (SC) generation is one amongst many good examples where the inherent compactness and low cost of ownership of fiber lasers have nicely been combined with the novel properties of photonic crystal fibers (PCFs) with dispersion-engineered profiles and high-nonlinearities [4–10].

In general, it is necessary to pump an optical fiber at wavelengths within an anomalous dispersion regime or close to its zero-dispersion wavelength in order to generate efficient SC [4–10]. While there have been a remarkable amount of SC generation results based on novel types of PCFs in a variety of regimes both in continuous-wave (CW) and in pulsed regimes [4–10], such sources are critically dependent on the development of PCFs which must have the right dispersion profile for the specific spectral band desired. Apart from the fact that PCFs are less commonly available compared to conventional types of optical fibers, the use of PCFs frequently leads to a significant level of difficulty when they are spliced to other optical fibers [11]. Thus, it must be of great interest if there is an alternative way to generate a broad continuum utilizing readily available, conventional types of optical fibers.

Hence, we would like to propose a novel method for generating a broadband continuum at a specific spectral range by exploiting stimulated Raman scattering (SRS) in conventional types of optical fibers, which in general have normal dispersion at wavelengths below ~1300 nm. In particular, we focus on the spectral range of 1200−1400 nm for which there is no efficient gain fiber like well-established neodymium-, ytterbium-, erbium-, and thulium-doped fibers [12–15]. (It is noteworthy to mention that bismuth-doped fibers can provide gain around this spectral range; however, the related research is still in an early stage [16].) In addition, this spectral range is important and useful for optical coherence tomography (OCT) applications because it provides a good trade-off among penetration depth, water absorption, and spatial resolution in biological tissue imaging [17, 18].

The motivations for this approach are also supported from the following aspects: While it is possible to generate SC utilizing a customized PCF with the right dispersion characteristics for obtaining the 1200−1400-nm spectral range when pumped with a source at ~1.1 μm, in general the level of requirement for the pump source and fiber (PCF) is relatively tight and high [7–10]. Furthermore, in many cases it is very hard to avoid the unwanted spectral conversion into the shorter wavelengths [7, 10] so that the overall conversion efficiency to the target spectral range cannot be as high as desired. On the contrary, the SRS process itself basically does not require specific dispersion characteristics of the nonlinear conversion fiber nor ultrashort-pulsed sources [7, 10]. That is, one can readily utilize conventional types of optical fibers for obtaining the spectral range of 1200−1400 nm even with a fiber-based amplifier source seeded by a well-controllable, nanosecond-pulsed diode laser at ~1.1 μm. In addition, the spectral conversion by SRS is advantageously unidirectional from the pump to the Stokes lines, which invariably leads to down-conversion except for some very special cases matched for the anti-Stokes generation [6], so that one can simply eliminate the unwanted loss due to the up-conversion to the short-wavelength side if the Raman-active fiber is pumped at ~1.1 μm.

However, the continuum spectra normally obtained through the SRS process in the normal dispersion regime tend to consist of discrete peaks of multi-orders of Stokes lines spaced by ~13.2 THz (in case of silica-based optical fibers) through cascaded Raman effects [19, 20] since soliton effects are in principle absent [4–10]. This may be a serious challenge to SRS-based continuum sources if normal-dispersion fibers are used. However, one can resolve this by utilizing a multi-wavelength pump [20–23]. This technique was initially developed to support wavelength-division-multiplexed (WDM) telecom systems operating at ~1.5-μm spectral range [21–23] where most optical fibers are already in the anomalous regime. Nevertheless, it can also be exploited for different spectral ranges having normal dispersion [20].

Here, we investigate the adaptive generation of a broadband Raman continuum in the spectral range of 1200−1400 nm via utilizing a conventional high-birefringence fiber (HBF) together with an all-fiber, dual-wavelength nanosecond-pulsed master-oscillator power amplifier (MOPA) source. The use of multi-level, triple-wavelength pulses has been investigated in a Raman amplifier, which was based on a highly nonlinear fiber (HNF) of 2 km in counter-propagating pump regime, controlling the Raman gain spectrum across the multiple Stokes lines [20]. However, to the best of our knowledge, there has been no report on the detailed SRS characteristics of HBFs and their applications to the “broadband continuum” generation in the nanosecond-pulsed regime. Therefore, we firstly discuss the unique SRS characteristics of the HBF in comparison with other types of optical fibers via investigating its spectral behavior for different polarization states of the input pump. Secondly, we investigate the impact of the inter-pulse time-delay and synchronization on the resultant SRS spectrum when dual-wavelength pump pulses at 1060 and 1080 nm are utilized. Building upon these results, we finally develop an all-fiber Raman continuum source having an average power of ~840 mW and a bandwidth of ~240 nm in full-width at half maximum (FWHM).

2. SRS characteristics of an HBF

Figure 1(a) shows the output spectra obtained through four different types of conventional optical fibers pumped by a nanosecond-pulsed source at 1060 nm having various peak powers up to a sub-kilowatt level, which include a standard single-mode fiber (SMF-28, Corning), a non-zero dispersion-shifted fiber (NZDSF, OFS), a large effective area fiber (LEAF, Corning), and an HBF (HB1500, Fibercore). One can see that discrete peaks of multi-orders of Stokes lines spaced by ~13.2 THz are formed through the fibers since all of them have normal dispersion characteristics at the pump wavelength up to ~1300 or ~1500 nm. However, one can notice that for the case of the HBF the spectral dips between the two adjacent Stokes lines are much shallower than those for the other fibers, as indicated by the downward arrows in Fig. 1(a). This spectral behavior of the fiber seems very useful for forming a continuum spectrum with a good spectral flatness if it is controllable.

 

Fig. 1 (a) Spectra of the Raman Stokes lines generated from various different Raman-active fibers. (b) Schematic representation of the walk-offs of the Stokes pulses against the pump pulses in an HBF for two different cases: The input polarization of the pump pulse is aligned along the slow axis in Case 1 and at 45° to the slow axis in Case 2, respectively.

Download Full Size | PDF

Motivated by this rather extraordinary feature of the HBF, we would like to understand and explain it qualitatively via Fig. 1(b), which depicts two cases that might happen in the HBF when pumped with a linearly polarized pulse, for example. (It should be noted that the spectrum shown in Fig. 1(a) was not necessarily generated by a linearly polarized pulse.) In the first case, we consider a pump pulse P_x linearly polarized along the slow axis (x-axis) of the HBF. While the pump pulse P_x can initially generate both Stokes pulses in the co- (S_x) and cross- (S_y) polarizations, P_x will predominantly amplify S_x rather than S_y through the SRS process because the Raman gain coefficient for the co-polarized Stokes is larger than that for the cross-polarized Stokes by approximately a factor of ten [24]. Thus, the impact of S_y is quite negligible. S_x might initially contain broad spectral components from the spontaneous Raman scattering process; however, the resultant spectrum of S_x after propagation through a distance will become narrowed down because of the successive SRS process although a slight spectral broadening would accompany because of the walk-off effect [25] between P_x and S_x via the chromatic group-velocity dispersion (GVD) of the HBF. It should be noted that S_x will travel slightly ahead of P_x, being up-chirped, i.e., having the leading edge red-shifted and the trailing edge blue-shifted as a result of self-phase modulation (SPM) together with the normal dispersion condition (which will be the case throughout this paper). Considering the pulse overlap between P_x and S_x, one can see that the trailing blue-shifted part will be better overlapped with P_x, thereby seeing higher Raman gain than the leading red-shifted part. It is noteworthy that in normal dispersion regime the trailing blue-shifted part will always see the higher Raman gain than the leading red-shifted part [25]. In addition, since S_x will also generate the next-order Raman Stokes via the cascaded Raman effect [19, 20], we can readily expect that the trailing blue-shifted part of S_x will undergo more power depletion because of seeing the higher Raman gain all the time, thereby giving rise to a “red-shift” in the resultant spectrum of S_x. Somehow, the amount of spectral gain broadening and shift is dependent on the amount of walk-off between P_x and S_x via the GVD of the HBF, which is in general quite limited in the cases of non-birefringence fibers as seen in Fig. 1(a).

On the contrary, one can see a significant difference in the second case where the polarization of the pump pulse is aligned to somewhere in between the two principal axes of birefringence of the HBF, for example, at 45° to any of the two axes as shown in the lower part of Fig. 1(b). Each pulse component, P_x or P_y, travelling in one of the principal axes will contribute to producing Raman gain for both co- and cross-polarized Stokes pulses, S_x and S_y. The powers of S_x and S_y must be nearly even when they are initially generated. However, they will undergo asymmetric amplifications because of the consequence of the polarization-mode dispersion (PMD) as well as the GVD. As discussed above, one can easily see that the Stokes pulse in the fast axis S_y is formed ahead of P_x and P_y as well as S_x, considering the effects of the GVD and PMD. The walk-off between S_y and P_x is larger than that between S_x and P_x because of the additional PMD. In this regard, one can expect that the spectral gain broadening and shift of S_y (P_x → S_y) will considerably increase compared to the previous co-linear polarization case (P_x → S_x). On the other hand, S_x traveling in the slow axis will undergo relatively less spectral broadening and shift but more amplification, owing to the less temporal slippage with the pump pulses, P_x and P_y. In fact the PMD makes P_y catch up with S_x. Consequently, one can readily conclude that the output spectrum of S_y will be more broadened and red-shifted but will have less power in comparison with S_x. It is also noteworthy to mention that the peak gain coefficient for the cross-polarized Stokes generation is positioned closer to the pump wavelength (~100 cm⁻¹) than that for the co-polarized Stokes (440 cm⁻¹) [24], which further help the spectral broadening increase.

In addition to this Raman gain broadening effect, we do not completely rule out the contribution of parametric processes to the extraordinary spectral broadening via four-wave mixing (FWM) and cross-phase modulation (XPM) among the cross-polarized pump and Stokes pulses [26, 27] since the HBF preserves polarization and its GVD may be compensated by its PMD. In fact, we could observe a small growth of an anti-Stokes line at ~1010 nm, as shown in Fig. 1(a), although its effect was quite limited throughout the experiment. Thus, unlike non-polarization-maintaining (non-PM) fibers of SMF-28, NZDSF, and LEAF, the HBF can benefit from these polarization-dependent nonlinear processes in order to yield the significant spectral broadening and shift as observed in Fig. 1(a). Altogether, these effects must be very useful with regard to minimizing the number of multi-wavelength pulses required for obtaining a Raman continuum with a good spectral flatness [20–23]. Further experimental details of this unique asymmetric SRS property of the HBF will be discussed in the next section.

3. Experimental results

3.1. Experimental arrangement

The schematic of the broadband source pumped by an all-fiber, dual-wavelength MOPA is shown in Fig. 2 . The system consists of two single-mode, 14-pin butterfly laser diodes (LDs) with operating wavelengths of 1060 and 1080 nm, modulated by two custom-made synchronized LD controllers (Notice Korea). Thus, the two LDs are capable of generating pulses of ~2 ns, with different repetition rates and inter-pulse time-delays. The signals from both LDs pass through their own isolators and are combined into a fiberized 1060/1080 WDM combiner. The common port of the WDM combiner is then spliced to a tap coupler (99:1). The 99-% port seeds the first ytterbium-doped fiber amplifier (YDFA) of the system, while the 1-% port is used for monitoring purposes. The YDFA comprises 6 m of a double-clad ytterbium-doped fiber (YDF, CorActive) cladding-pumped with a 975-nm high-power multimode LD through a tapered fiber bundle (TFB). The YDF has a core of ~8 μm in diameter with a numerical aperture (NA) of 0.08 and a cladding of ~125 μm in diameter with an NA of 0.45, respectively. Its small-signal cladding absorption rate is ~6.8 dB/m at 976 nm. The output of the amplifier is spliced to an isolator to protect the first YDFA against any back reflection from the second YDFA. The signal is then launched into the second YDFA through another tap coupler (99:1) for monitoring purposes. The second YDFA comprises a 9 m of the same YDF used for the first YDFA and is also cladding-pumped with a 975-nm high-power multimode LD. Another isolator is spliced to the output end of the second YDFA, followed by a tap coupler (99:1) for monitoring the final output signal and a polarization controller (PC) for controlling the output polarization state, respectively. In addition, it is noteworthy that the inclusion of the two 1060/1140 WDM couplers between the first and second YDFAs and after the last tap coupler is to bypass the unwanted backscattered Stokes lines, i.e., mainly the first Stokes lines at 1100−1150 nm of the Raman continuum generated in the Raman-active fiber that is to be spliced to the output end of this dual-wavelength MOPA source. We initially investigate an HBF as the Raman-active fiber, motivated by the unique SRS property observed in Fig. 1.

 

Fig. 2 Schematic of the broadband continuum source based on a dual-wavelength MOPA seeded at 1060 and 1080 nm via a Raman-active fiber.

Download Full Size | PDF

In addition, the single-mode LDs operate in a pulsed-mode, generating pulse widths of ~2 ns at a repetition rate of 500 kHz throughout the whole experiment. In the given condition, the average output power of the combined two-LD signal is ~0.2 mW and is amplified up to ~70 mW through the first YDFA and up to ~1.2 W through the second YDFA, respectively. This yields a peak power of ~1.2 kW when the two-wavelength pulses are fully overlapped. This will be the operating condition throughout our experiment unless stated otherwise.

3.2. Analysis of the polarization-dependent SRS of an HBF

Since the all-fiber MOPA source is not based on PM fibers and PM components, an additional setup is prepared for the sake of analyzing in detail the polarization-dependent SRS in an HBF. The setup is as follows: The output end of the MOPA before the Raman-active fiber (see Fig. 2) is flat-cleaved, and its output beam is collimated in free space and passes through a broadband polarizing beam splitter (PBS, PBS-1), thereby resulting in a linearly polarized beam as shown in Fig. 3 . The polarizing angle of PBS-1 is fine-tuned via a rotation mount. Then, the output beam through the HBF is again collimated and passes through another broadband PBS (PBS-2). The transmitted beam from PBS-2 is coupled into a standard single-mode fiber (SMF-28) and monitored with an optical spectrum analyzer (OSA). The rotation of PBS-2 allows us to analyze the state of polarization (SOP) of the Raman continuum formed in the HBF. It is noteworthy that the HBF is fixed to a fiber holder with its slow and fast axes aligned parallel to x and y axes in terms of the laboratory coordinates. In addition, the polarization extinction ratios of both PBSs are better than 20 dB, and the reflected beams by PBS-1 and PBS-2 are terminated into individual beam dumps. The HBF (HB1500, Fibercore) investigated in our experiment has a core NA of 0.16, a mode field diameter of 7.9 μm at 1550 nm, and a cut-off wavelength of 1255 nm. Its length is ~60 m. In particular, its beat length is given by 1.9 mm at 1060 nm (2.7 mm at 1550 nm), yielding a PMD of ~1.88 ps/m at 1060 nm [28] while the chromatic dispersion parameter is given by β₂ = −13.7 ps²/km at ~1550 nm [29]. Given the dispersion parameters, we estimate that the walk-offs due to GVD and PMD are comparable, and the walk-off variation due to the PMD against the GVD is given by ± 113 ps in 60 m of the HBF. This variation may be regarded as relatively small in comparison with the duration of the input pump pulse, which is ~2 ns. However, it should be noted that if higher-order Stokes pulses are generated, they will undergo significant duration-narrowing because of the inherent nature of the nonlinear conversion process [30]. Therefore, we cannot simply underestimate the spectral gain-broadening due to the walk-off effect. Given that the walk-offs due to GVD and PMD are comparable, the maximum spectral shift can be as far as the amount of Stokes shift, although the overall shift will be limited because the walk-off is gradually built up over the fiber length. In addition, it should be noted that although the HBF supports higher-order modes (HOMs, including TE₀₁, TM₀₁, and HE₂₁) below the cut-off wavelength, we carefully arranged for the free-space optics, thereby exciting only the fundamental mode (HE₁₁) into the HBF at the pump wavelengths [31]. The total power launched into the HBF was reduced to roughly half the full output power of the MOPA mainly due to the insertion of PBS-1.

 

Fig. 3 Schematic prepared for analyzing the polarization-dependent SRS in an HBF. AL: aspheric lens; RM: rotation mount.

Download Full Size | PDF

We first operated the MOPA source with a single-wavelength input either at 1060 or 1080 nm, varying the input polarization angle via PBS-1 with respect to the slow-axis (x-axis). It should be noted that in this case we temporarily took PBS-2 out of the setup in order to monitor the whole output. Since the peak power of the input pulse was sufficiently high, one can easily see the nonlinear spectral evolution of the output spectrum by cascaded SRS as shown in Fig. 4 . The spectral powers are normalized to the maximum power levels, and for visual aids “spectrum-prints” plotted in color-coded lines in dB scales are also accompanied in the bottom. The results indicate that when the input polarization was aligned either to the slow or fast axis of the HBF, the Raman Stokes lines had relatively deeper spectral valleys. (See the cases of the angle of PBS-1 aligned at 0° or 90°). In these cases, the differences between the adjacent peaks and valleys of the spectral power were given by at least 20 dB. However, one can notice that when the input polarization was aligned at 45° to the slow axis, the spectral power level at the valleys significantly increased by more than 5 dB in most cases, and the Raman Stokes lines were red-shifted relative to the cases aligned at 0° or 90°. This experimental observations support our explanation offered in Section 2 with Fig. 1(b), verifying that the spectral broadening and red-shift of the Raman Stokes lines occur when cross-polarized Stokes pulses of non-negligible power are provided together with the pump pulse.

 

Fig. 4 Measured optical spectra (upper) and their spectrum-prints (lower) of the Raman Stokes lines generated through 60 m of an HBF for different angles of rotation of PBS-1 for input wavelengths of (a) 1060 nm and (b) 1080 nm, respectively.

Download Full Size | PDF

While we could observe the polarization-dependent spectral broadening through a linearly polarized pump pulse, one may notice that the generated spectrum in this condition is significantly different from what we could observe in Fig. 1(a). We attribute this to the difference in the polarization and modal conditions between them. That is, the result shown in Fig. 1(a) was obtained through all-fiberized coupling between the MOPA and the HBF whereas the result shown in Fig. 4 was obtained through free-space coupling with a PBS-1. It is noteworthy that the input pump pulse in the case of Fig. 1(a) might well be elliptically polarized since the MOPA system was based on non-PM fibers while it was seeded by linearly polarized input pulses. In addition, in the case of Fig. 1(a), the output pigtail fiber (HI1060, Corning) of the MOPA system was fusion-spliced to the HBF, so that a fraction of the input power could be in HOMs (TE₀₁, TM₀₁, and HE₂₁). This could give rise to rather complicated, additional spectral broadening via parametric process based on FWM and XPM among cross-polarized HOM pump and Stokes pulses [26, 27, 32]. This effect may further be boosted up if their chromatic GVDs are compensated by the PMD. However, we did not observe any strong parametric process throughout the experiment shown in Fig. 4, where we carefully controlled the launching condition in order that only the fundamental mode (HE₁₁) was excited in the HBF [31].

Next, we analyzed the SOP of the output beam by restoring the insertion of PBS-2 in the setup. We initially investigated the co-linear polarization case in which the input polarization was matched to either of the two principal axes of birefringence of the HBF. Figure 5 represents the corresponding spectrum-prints for the input wavelengths of 1060 and 1080 nm, respectively. The angle of rotation on the vertical axis shown in Fig. 5 indicates the polarizing angle of PBS-2 measured from the slow axis (x-axis) of the HBF. One can see that as discussed in Fig. 1(b), the Stokes lines are predominantly in the same polarization as that of the input pump with extinction ratios of more than 30 dB. In addition, the extinctions ratios did not show any noticeable spectral dependence.

 

Fig. 5 Spectrum-prints of the Raman Stokes lines generated through the HBF when the input polarization was aligned either to the (a) x-axis or (b) y-axis for input wavelengths of 1060 (upper) and 1080 nm (lower), respectively. The angle of rotation denotes the angle of PBS-1 (shown in Fig. 3) with respect to the slow axis (x-axis).

Download Full Size | PDF

On the contrary, one can see significant difference in the polarization-dependent spectral behavior for the case of the input polarization aligned at 45° to the slow axis (x-axis) of the HBF, as shown in Fig. 6 . For this experiment, we monitored the spectral power through PBS-2, while rotating its polarizing angle from the x-axis. It is noteworthy that there are two noticeable features: One is that the overall Raman conversion strength was maximized when the Stokes lines were polarized in the slow axis (0° in terms of the rotation angle), i.e., S_x, as denoted in Fig. 1(b). The other is that although the Stokes lines polarized in the fast axis ( ± 90° in terms of the rotation angle), i.e., S_y, as denoted in Fig. 1(b), had less power than S_x, their spectral broadening and red-shift were much more significant: One can see that the spectral dips become shallower and shifted towards longer wavelengths as the angle of PBS-2 increased from 0° to 90°. These experimental observations again fully support our explanation offered in Section 2. That is, for the pump excitation at 45° to the slow axis (x-axis), S_y will represent more spectral broadening and red-shift while S_x will represent more overall Raman conversion. In fact, these consequences are due mainly to the effects of the combination of the PMD and GVD of the HBF as discussed in Fig. 1(b).

 

Fig. 6 Spectrum-prints of the Raman Stokes lines generated from the HBF when the input polarization was aligned at 45° to the slow axis (x-axis) for the input wavelengths of (a) 1060 and (b) 1080 nm, respectively. The angle of rotation denotes the angle of PBS-2 (shown in Fig. 3) with respect to the slow axis (x-axis).

Download Full Size | PDF

3.3. Impact of the inter-pulse time-delay between the dual-wavelength pump pulses

Since we use the dual-wavelength pump pulses, it must be worthwhile to investigate the impact of the inter-pulse time-delay between the dual-wavelength pump pulses when they are turned on at the same time. This analysis will allow us to understand the pulse-to-pulse Raman interaction and its influence on the spectral shaping of the resultant Stokes lines.

Figure 7 shows the evolution of the output spectra for different values of the inter-pulse time-delay where we kept the input polarization of the pump pulses aligned either to the slow axis (x-axis) or fast axis (y-axis). In such conditions, the SRS will predominantly occur between the pump and Stokes pulses in the same polarization, as discussed in Fig. 5, as well as in Fig. 1(b), so that we can see the impact of the inter-pulse time-delay separately from the other effects, e.g., SRS effects between cross-polarized pumps and Stokes. We varied the time-delays from −2.5 ns to + 2.5 ns, where the positive sign indicates that the 1060-nm pulse was ahead of the 1080-nm pulse. The minimum resolvable step of the time-delay was 0.1 ns. One can see that for the cases where the time-delays are close to zero [see Fig. 7(e)], the pulse-to-pulse Raman interaction became stronger with an increase of more than 10 dB of the spectral power level in the range between 1200 nm and 1300 nm, compared to the cases having the time-delays of ± 2.5 ns. Actually, this consequence can be explained in the similar manner with the case previously discussed in Fig. 6, if we regard the cross-interaction between the two pulses of different wavelengths as the cross-interaction between the two pulses of different polarizations. In fact, the 1060-nm pump pulse can work as an auxiliary pump for the Stokes lines generated by the 1080-nm pulse, and vice versa. Therefore, one can readily expect that the overall SRS will be boosted up as the overlap between the two pulses is increased. On the other hand, if there is a certain level of time-delay between the two pulses which is shorter than the individual pulse width, an intermediate part of one pulse, which may be leading or trailing, will eventually see the peak of the other pulse, so that the spectral components in the slope of the one pulse will consequently see “boosted” Raman gain. Thus, this walk-off effect will also lead to spectral broadening [25].

 

Fig. 7 Spectrum-prints of the Raman Stokes lines generated from the HBF with the dual-wavelength pulses at 1060 nm and 1080 nm when the input polarization was aligned to (a) the slow axis (x-axis) or (b) fast axis (y-axis), varying the inter-pulse time-delay from −2.5 ns to + 2.5 ns. The time traces of (c)−(g) depict the input dual- wavelength pulses with different inter-pulse time-delays.

Download Full Size | PDF

3.4. Spectral shaping with an HBF in conjunction with an HNF

Based on the experiments described in the previous sections, we found that the Raman Stokes generation in an HBF is correlated with the arrangement for the input pump in terms of its polarization and inter-pulse time-delay as well as its peak power, yielding a considerable level of change in the resultant spectra of the Raman Stokes lines. This actually opens up the possibilities to use HBFs for tailoring the Raman continuum profile through controlling those input pump parameters. Here, we explore this novel option for generating a broadband continuum with a good flatness in the 1200−1400-nm spectral range.

We first investigated the HBF (60 m) used for the fundamental analyses in the previous sections. This fiber was now fusion-spliced to the dual-wavelength MOPA source, as shown in Fig. 2. In addition to the descriptions for the experimental setup given in Section 3.1, it is noteworthy that the seed pulses were synchronized with no intra-pulse time-delay and their power levels were initially coarse-tuned to yield even power levels for both 1060- and 1080-nm pulses. It should be noted that the wavelength choice of the two pump pulses was determined mainly by the availability of the seed LDs at the time of the experiment. Assuming that the wavelength of the first pump pulse is fixed at 1060 nm, the wavelength of the second pump pulse must ideally be tuned at 1086 nm in order that the latter generates its first Stokes line at 1140.5 nm, which is exactly matched with the center wavelength of the first- and second-order Stokes lines generated by the former. However, as presented Fig. 1(a), we could see a significant level of extraordinary spectral broadening between the two adjacent orders of Stokes lines when we utilized the HBF, so that the wavelength choice of the two pump pulses (1060 and 1080 nm) was reasonably acceptable for effectively filling up the spectral gaps among the consecutive Stokes lines even in the current form. Nevertheless, one can think of fine-tuning the wavelengths of the dual-pump pulses in order to obtain a further enhancement in the optical spectrum of the output signal.

The resultant output spectra obtained through the HBF for three different input pump conditions are shown in Fig. 8 . The blue and red traces are for the cases before and after optimizing the SOP of the input pump, respectively. It should be noted that the SOP of the input pulse was adjusted via the fiberized PC shown in Fig. 2. (The input pulses to the HBF might well be elliptically polarized and contain a fraction of power in higher-order modes.) One can see that both the level of spectral power and the spectral flatness in the range of 1200 to 1300 nm were significantly improved by the adjustment. The green trace is for the case when the peak power of the 1080-nm seed pulse was individually reduced by ~10%, where one can see the further improvement both in the spectral power level and in the spectral flatness. In this case, the resultant Raman continuum became as broad as ~150 nm (FWHM) centered at ~1250 nm. The average output power was ~930 mW, yielding the input-to-output conversion efficiency of ~78%.

 

Fig. 8 Output spectra and their spectrum-prints of the Raman continuum generated through the HBF: The blue (A) and red traces (B) are for the cases before and after optimizing the polarization state of the input pump, and the green trace (C) is for the case when the peak power of the 1080-nm seed pulse was individually reduced by ~10%.

Download Full Size | PDF

However, we observed that the spectral power of the Raman continuum dropped by more than 10 dB at wavelengths longer than 1320 nm, regardless of the optimization of the input pump. In order to enhance the spectral power in the longer wavelengths, it was necessary to increase either the power of the pump lines or the length of the Raman-active fiber. The maximum output power of the MOPA source was actually limited by the safe operation range of the fiberized components, including the output isolator and tap/WDM couplers at the output end, which were graded to operate below ~1.5 W. Consequently, we considered splicing an additional length of another Raman-active optical fiber to the HBF. We chose an HNF (HNDS1310B, Sumitomo) since it has a relatively high nonlinear coefficient (γ = 15 W⁻¹km⁻¹ at 1300 nm) while it has a moderate normal chromatic dispersion at 1300 nm (D = −8 ps/nm/km at 1300 nm). In addition, its cut-off wavelength is given by 1020 nm. Next, its length was optimized to ~10 m via the cut-back method in order to obtain the Raman continuum well-centered at 1200−1400 nm. (It should be noted that if the HNF is too long, the spectral range of the continuum will exceed over 1500 nm at the expense of the spectral power at ~1200 nm, which is not desirable in our case.)

The resultant Raman continua generated in the HBF-HNF are shown in Fig. 9 where we considered for two different intra-pulse time-delays of ~7 and ~0 ns, respectively. We only fine-tuned the polarization of the input pump lines, keeping their seed powers unadjusted. We finally obtained Raman continua having a bandwidth of ~240 nm (FWHM) for both cases. Actually, the case with no intra-pulse time-delay between the 1060- and 1080-nm pulses could lead to a broader spectrum slightly shifted to the longer wavelengths, as discussed in the previous section in Fig. 7. The total output power measured after the whole Raman-active fibers was ~840 mW, yielding the input-to-output conversion efficiency of ~70%. The efficiency drop was attributed to the additional SRS process in the HNF as well as the splicing loss between the HBF and HNF.

 

Fig. 9 Output optical spectra and their spectrum-prints of the Raman continuum generated through the HBF (60 m) + HNF (10 m) for two different inter-pulse time-delays: (a) For Δt = ~7 ns and (b) Δt = ~0 ns.

Download Full Size | PDF

5. Conclusion

We have investigated the SRS in an HBF for various conditions of the input pump pulse, paying particular attention to its spectral broadening and shaping. We found that the PMD of the fiber in conjunction with its GVD plays a crucial role via the interaction between the pump and Stokes pulses, which is a unique property that cannot be seen with other non-birefringence optical fibers. That is, the walk-off, i.e., the inter-pulse time-delay between the pump and Stokes pulses incurred by the PMD and GVD of the HBF gives rise to extraordinary temporal shifts of the Raman-gain peaks, thereby leading to the correspondent broadening of the overall Raman gain spectrum [25]. In addition, the extraordinary spectral broadening and shift can also be boosted up in case parametric processes via FWM and XPM among cross-polarized HOM pump and Stokes pulses are involved. This spectral broadening effect can further be exploited via varying the inter-pulse time-delay if a multi-wavelength pump regime is considered. When we used the input pump comprised with dual-wavelength pulses at 1060 and 1080 nm with a variable inter-pulse time-delay, we could observe another significant level of spectral broadening in the resultant Raman Stokes lines.

Exploiting these unique properties, we have developed a Raman continuum source based on an all-fiber MOPA synchronously seeded at 1060 and 1080 nm with a variable inter-pulse time-delay followed by a 60-m Raman-active fiber of an HBF. This source could emit a Raman continuum with a good spectral flatness having an average power of ~930 mW and a bandwidth of ~150 nm (FWHM) centered at ~1250 nm. It should be noted that the spectral shape and flatness were formed by the adaptive control of the polarization, inter-pulse time-delay, and peak powers of the dual-wavelength pulses generated from the MOPA. With the inclusion an HNF of 10 m in addition to the HBF, we generated a Raman continuum having an average power of ~840 mW and a bandwidth of ~240 nm (FWHM) in the 1200−1400-nm spectral range. Both results represented good input-to-output conversion efficiency of in excess of 70%, having the spectrum well confined in 1200−1400 nm. Broadband continuum sources in this spectral range are, in particular, of great interest for OCT imaging applications [17, 18]. For example, the 240-nm bandwidth demonstrated here is capable of yielding an axial resolution of ~3 μm when utilized as a light source for a spectral-domain OCT system [17]. In this regard, Raman continuum sources based on nanosecond-pulsed MOPAs combined with HBFs or other conventional types of optical fibers could be attractive alternatives to SC sources based on the ultrashort-pulsed lasers and/or PCFs which may suffer from strong spectral modulation or low optical conversion efficiency.