1. Introduction

Supercontinuum generation (SCG) is a phenomenon wherein an ultrashort pulse of light undergoes extreme spectral broadening to generate new frequency components as it propagates through a suitable nonlinear medium. Effects such as self-phase modulation, four-wave mixing (FWM) and Raman scattering [1] play an important role. SCG is a highly nonlinear process that can be improved by more precise engineering of the waveguide dispersion of the guiding structure, which results in greater control over the dynamics of the broadening process. This ‘management’ of the dispersion can to be carried out over very small lengths that could range from a few millimeters to even sub-millimeter length scales [2, 3].

We have used ‘holey’ fibers (HFs), which consist of a solid core surrounded by a lattice of air holes that constitute the cladding, and result in enhanced modal confinement and unique dispersion properties [4]. Supercontinuum generation has been explained on the basis of a soliton fission process that is highly nonlinear and as a consequence, the nonlinearity of the guiding structure as well as the dispersion, play crucial roles in the propagation dynamics of the solitons [5]. The nonlinearity, characterized through the nonlinear coefficient γ, as well as the group velocity dispersion (GVD) are parameters that can be controlled by tapering, or varying the transverse profile of the fiber along its length. Through tapering, a reduction in the core size of the fiber can be achieved which affects the nonlinearity as well as the zero-dispersion wavelength of the fiber in accordance with the eigenvalue equation for the propagation constants of the supported fiber modes of the fiber. Tapering also tunes the four-wave mixing and other phase-matching conditions required for the generation of ultra-broadband light [6]. This post-processing technique provides a way to engineer or ‘micromanage’ the dispersion and waveguiding properties of a holey fiber on millimeter length scales which facilitates careful and precise modification of the generated spectrum, giving rise to the term ‘dispersion micromanagement’ (DMM) [2].

We used a CO₂ laser as the heat source for our tapering setup. The tapering was carried out by a careful optimization of the pulling speed as well as the temperature of the fiber (controlled by the output power of the CO₂ laser) to ensure that adiabaticity was maintained so as to achieve low loss tapers. While tapering, care must be taken to preserve the microstructure of the HF, in order to maintain the desired waveguide properties. As mentioned earlier, SC generation results from an initial higher order soliton fissioning out into several fundamental solitons, and this occurs for the specific case where the input is in the anomalous dispersion regime [5,6]. A dispersive wave generation associated with a propagating soliton and the phase-matching of this wave with the Raman-shifted input soliton results in what is referred to as the ‘Cherenkov’ radiation (CR) [5,7,8]. The CR manifests as an isolated feature on the generated continuum spectrum and is commonly referred to as an anti-Stokes radiation (ASR) component. The nomenclature ASR stems from the fact that the resonant Cherenkov wavelength is blue-shifted with respect to the input pump. Previous experiments have shown that this feature exhibits high spectral coherence and exhibits very low broadband RF noise [2,9]. As such, this has potential as a versatile visible source for applications ranging from optical coherence tomography [10], high-precision optical frequency metrology [11] and CARS microscopy [12]. The DMM scheme utilizes the variation of GVD along the length of the fiber, as an extra degree of freedom for generating ASR pulses which can be centered to anywhere within the visible region of the spectrum and can also control the bandwidth of the ASR.

Several practical issues arise while using HFs. The air channels constituting the cladding are susceptible to contamination by substances such as cleaning solvents through capillary action, and ambient dust. For some biomedical applications, such as fluorescence detection [13], where there might be a need to immerse them into liquid media, it might be undesirable to allow the fluids to penetrate into the microstructure. Highly nonlinear HFs also have very small core sizes which make the fiber susceptible to damage from high power input pulses at the dielectric discontinuity between the fiber end face and air. Furthermore, when coupling light into such small-core fibers, positional stability becomes ever more demanding [10]. A commercial continuum generator is available [14] which incorporates end sealing - however, in our case, the end sealing of a DMM device is complicated because of the interactions with the DMM region. To address these issues, we have used a conventional fusion splicer and developed a defined sequence of steps to collapse the cladding capillaries of the HF at the input as well as the tapered output end. The parameters that were optimized to obtain such hole-collapsed ends included the magnitude of the arc current, the duration of the arc, and the positioning of the fiber with respect to the filament electrodes of the splicer. Figure 1 depicts the structure of these end-sealed DMM devices. The coreless section of glass provides the fiber with a modal expansion region as the holes that act to strongly confine the mode to the core are now eliminated, which intuitively, would facilitate easier coupling into the fiber [15].

 

Fig. 1. Schematic image of (a) an untapered, unsealed HF; (b) an input and output end-sealed untapered HF; (c) an unsealed DMM HF; (d) an input end-sealed DMM, and; (e) an input and output end-sealed DMM.

Download Full Size | PDF

The effect of the hole-collapse on the mode propagating through tapered fibers has been explored for sensing applications by exploiting the evanescent nature of the field in the collapsed region [12]. The effects of partial hole collapse have also been investigated for it’s effect on modal expansion [20, 21]. Such a procedure for environmentally sealing a microstructured fiber has been reported before [17–19, 22]. However, to our knowledge, this is the first time that this technique has been applied to a very short, dispersion micromanaged fiber.

2. Experimental details and results

Our initial experiment involved the investigation of the coupling sensitivity of the end-sealed HF towards input alignment for the case of the ASR generation in the visible. We took a 2m length of untapered HF (NL-3.3-850, Crystal Fibre A/S), which had a zero dispersion wavelength at 850 nm, and hole-collapsed the input-end with the fusion splicer (Ericsson FSU 995 PM). The optimum magnitude of the arc current for sealing the front-end of the HF was determined to be 12 mA applied over a duration of 0.45 seconds. The fiber was positioned with a slight offset of ~20 μm with respect to the tips of the splicer electrodes. We then placed the fiber on a motorized stepper with a step size-resolution of 0.1 μm and coupled 100-fs pulses from a mode-locked Ti:Sapphire laser (Spectra Physics’ MaiTai) at a wavelength of 880 nm and generated continuum that spanned from 400 nm to 1200 nm. This nonlinear measurement method was adopted as it was considered relevant to the use of this device as a wavelength convertor. We used an interference filter to record the power contained in the red part of the continuum at 780 nm as a function of the transverse offset of the input end with respect to the pump beam [23]. Figure 2 shows the comparison of the results from the hole-collapsed fiber to those obtained with a conventionally ended fiber and the plot shows that there is indeed an enhancement of the input coupling function for the end sealed fiber implying reduced sensitivity towards input coupling. The FWHM of the input coupling function of the fiber increased from a value of 1.6 μm for the normal HF to 2.3 μm for the end-sealed HF. Although the plot shows coupled powers on a normalized scale, while carrying out the experiment, the maximum output power was identical for both cases to eliminate any inconsistency in measurement.

 

Fig. 2. Comparison of alignment sensitivity for normal-ended HF and end-sealed HF. Inset: Images of the end-facets of the hole-collapsed HF and the normal-ended HF

Download Full Size | PDF

The insets on the right of Fig. 2 show microscope images of the input facets for the sealed and the unsealed fibers. The maximum coupling efficiency was ~60% for both fibers, indicating that the end-sealing did not introduce significant losses for the HF. We experimentally determined that for our fiber, the length of the collapsed section of glass needed to be under ~280 μm for the input end in order to ensure dominant fundamental mode operation.

Our next step was to manufacture a DMM fiber with a sealed input end and investigate its performance. We took the same HF as before, i.e., with a core diameter of 3.3 μm and tapered it to about 2.3 μm over a length of 17 mm. This DMM was designed to generate ASR centered at 550 nm with a bandwidth of approximately 22 nm [2]. We wanted to ensure that we could integrate end-sealing with the DMM technology without significantly interfering with the nonlinear broadening processes occurring inside the device. We coupled 100 fs pulses at a wavelength of 920 nm and observed continuum that spanned from 350 nm to 1300 nm with an enhanced spectral feature at ~550 nm which was identified as the desired ASR (Fig. 3). The spectrum was captured on an Ando 6315 optical spectrum analyzer.

 

Fig. 3. (a). Output spectrum of the hole-collapsed DMM on a log scale. Inset: Filtered ASR on a linear scale. (b) Output spectrum on a linear scale. Dotted red line indicates the pump.

Download Full Size | PDF

We also looked at the noise characteristics for the device. Conventional ultra-broadband light generating devices, such as untapered HFs, exhibit significant spectral intensity fluctuations and broadband amplitude noise in the radio frequency range, making the resulting continuum unstable for different spectral regions. Noise seeded from the pump’s shot noise gets amplified by the nonlinear processes that occur as the pulse propagates and broadens along the length of the fiber [24, 25]. DMM devices overcome this noise limitation by essentially having small lengths which results in a truncation of all the noise-amplifying processes. When we taper the HF over small length scales, we essentially introduce a phase-matching condition that varies along the fiber as a function of the changing core size [2, 26].

Previously published results demonstrate a comparison of noise figures between DMM devices and other fibers [25]. We investigated the effects of end-sealing on the noise performance of the DMM by spectrally filtering out the ASR and recording it on an RF spectrum analyzer via a fast photodetector of 1 GHz bandwidth. Our results show an imperceptible rise of the noise floor of the ASR above the baseline, leading us to conclude that the structural modifications to the DMM did not compromise the stability of the device (Fig. 4).

 

Fig. 4. Broadband noise for the filtered ASR at 550 nm. Inset: picture of the ASR.

Download Full Size | PDF

By carrying out the first few experiments, we were satisfied that the DMM device could perform as per its design without any significant change. The important question now was whether we could now carry out the same procedure at the back-end of the device while maintaining its desirable properties. Previous theoretical studies as well as experiments [2] have implied that the processes occurring at the end section of the fiber, comprising of a few millimeters, are crucial to the ASR generation process and hence, extra care needs to be taken in order to ensure that the dispersion map of the device is maintained. The length of the green ASR-generating DMM was now extended slightly by about a millimeter; however, the diameter was kept roughly constant over the extra section. This device was now 18 mm long and had a final core diameter of 2.2 μm (Fig. 5). The sealing at the output end was carried out by reducing the arc current from 12 mA to 10 mA as now the output cross section was narrower than that of the input and hence required lower heat. This procedure yielded a short collapsed region of pure silica resulting in a sealed section for the DMM which was ~50 μm long. The length of the sealed segment was limited by the extent of the fusion electric arc’s effect in terms of the related heat diffusion, as also the exact position where the fiber was cleaved. The sealed length affects the generated spectrum especially for the sealing at the back-end as the intensity drops rapidly as the mode expands (since it is now unguided) and depletes the nonlinearity, thus affecting the ASR generation process. Once again, 100 fs pulses at a wavelength of 915 nm from a mode-locked Ti:Sapphire were launched into the DMM, which now had both the input and output ends hole-collapsed.

 

Fig. 5. (a). Physical profile for the DMM generating ASR centered at 550 nm. (b) Microscope images of the hole-collapsed input and the output ends of the DMM

Download Full Size | PDF

Figure 6 shows the spectrum obtained from an unsealed DMM, a DMM with the input end sealed, a DMM with both the input and output ends sealed, and a DMM which was not optimally sealed. The comparison shows that for the DMM that was sealed successfully, as shown in Fig. 6(c), the ASR could still be generated, albeit with a slight loss of efficiency and bandwidth which can be explained on the basis of a perturbation of the dispersion profile at the output end by the extra processing.

 

Fig. 6. Continuum spectra from (a) DMM with both ends sealed; (b) DMM with the input end sealed; (c) DMM with unsealed ends; (d) DMM with non-optimized sealing at both ends. Dashed lines indicate the region of the ASR.

Download Full Size | PDF

When we cleaved the fiber incorrectly at the output end, i.e., when the sealed section was too long, we were no longer able to generate the desired ASR in the spectrum as is depicted in Figure 6(d). Further experiments to examine the specific effects of such types of processing on the nonlinear phenomena occurring inside the DMM are being carried out and will be reported elsewhere.

3. Conclusions

We developed a systematic method to hole-collapse the input and output ends of a dispersion micromanaged holey fiber. We investigated the sensitivity of coupling towards input alignment and concluded that the modified input face resulted in a more stable configuration with greater ease of coupling. We also looked at the continuum profile and broadband noise characteristics of the DMM and we concluded that the processing to the end facets of the fiber resulted in a device that was more robust and environmentally stable without compromising its spectral performance.