1. Introduction

Photonic crystal fibers and their applications [1] have attracted much attention in the recent decade. Hollow-core (air-guiding) photonic crystal fibers are of particular interest for applications such as gas spectroscopy [2, 3, 4], low-distortion data transmission [5], and compression of femtosecond fiber laser pulses [6, 7, 8]. The latter application especially requires high polarization maintaining properties, high anomalous dispersion and low third order dispersion in a wide spectral range, and low Kerr nonlinearity. The unique combination of above properties makes HC-PCFs the fibers of natural choice for femtosecond laser pulse compression, especially for the Yb fiber lasers operating in the wavelength range around 1 µm.

The HC-PCFs are normally used in conjunction with single-mode fibers, which requires low-loss splicing between the two fiber types. Apart from the low loss some applications, such as fiber lasers, also require a high polarization extinction ratio (PER) from such a splice. In recent publications, successful fusion splices between the SMFs and HC-PCFs were demonstrated, with splice losses in the range 1–2.3 dB [2, 3, 4, 6, 9, 10, 11]. However, to the best of our knowledge, successful low-loss PMsplicing between the PM-SMFs and HC-PCFs still remains a challenge, which clearly limits the use of the SMF-to-HC-PCF fiber assemblies to non-PM applications, thus excluding most laser applications where polarization cross-talk should be controlled or minimized. In this work we demonstrate a technique for reproducible low-loss PM splicing between SMFs and HC-PCFs, and its application in an environmentally stable, monolithic PM femtosecond fiber laser pulse compressor.

2. Fibers

As PM-SMF we used a standard Nufern PM Panda PM980-HP fiber with mode field diameter (MFD) of 6.6±1 µm and birefringence Δn=3·10^-4, measured at 980 nm wavelength [12].

As HC-PCF we used Crystal Fibre HC-1060-02 fiber [13], based on the 7-cell design, with an MFD of 6.5 µm measured at 1060 nm, featuring a high-transmission band at 1000–1100 nm wavelength. A SEM image of the fiber is shown in Fig. 1. Slight core asymmetry results in a spontaneous birefringence of the fiber [14, 15]. This birefringence was already employed for the compression of femtosecond pulses with a relatively high PER of 17 dB, although the light was coupled into the fiber using a bulk optic arrangement [8].

 

Fig. 1. SEM image of the HC-PCF fiber. Courtesy of B.J.Mangan, Crystal Fibre A/S.

Download Full Size | PDF

In order to characterize the birefringence of our HC-PCF we have spliced a piece of the fiber to the PM-SMF output of a modelocked Ytterbium all-PM-fiber laser with a splice loss of 0.42 dB and intentionally low PER of 0.19 dB, using the splicing technique described below. Our laser was operating at central wavelength of 1065 nm, emitting positively chirped laser pulses of 11 ps duration with 10 nm spectral bandwidth, and output PER of 24 dB. The principle of our laser is described in [16]. The central wavelength of the laser was chosen because of the relative flatness of the dispersion parameter D of HC-1060-02 PCF around 1065 nm [13].

We have cut-back the HC-PCF to a length of 9.8 m, and then performed autocorrelation measurements of HC-PCF-compressed laser pulses with a λ/2 plate and a polarizer placed between the collimated fiber output and the autocorrelator. Rotation of λ/2 plate at fixed polarizer position allowed us to control the polarization state of the measured signals. Given the intentionally low PER of PM-SMF-to-HC-PCF splice, the laser intensity was launched into both axes of the HC-PCF in nearly equal proportion. By rotating the λ/2 plate with respect to the principal axes of the HC-PCF, we were able to measure separately the autocorrelations of laser pulses polarized along each axis of the fiber, as well as the cross-talk signal between the axes, as shown in Fig. 2.

When the λ/2 plate was aligned so that both PM axes of HC-PCF were oriented at 45° in respect to the polarizer transmission axis, an autocorrelation featuring three peaks was observed, corresponding to the presence of two pulses in the measured signal. The autocorrelation intensity ratio is very close to 2 : 1 between the central and side peaks, and the time separation between them is ΔT=5.4 ps. We conclude that the measured signal consists of two pulses of orthogonal polarizations and nearly equal intensities, separated by the time interval of ΔT. Given the HC-PCF length of 9.8 m, this time interval corresponds to a fiber group birefringence of Δn=1.6510^-4. Such a high birefringence ensures strong PMproperties of the HC-1060-02 PCF, which was supported by the very high stability of the measured autocorrelations against even strong mechanical and temperature fluctuations induced on the fiber.

When the λ/2 plate was aligned so that only one of the PM axes of HC-PCF was co-aligned with the polarizer transmission axis, the autocorrelation traces consisted of only one peak. The width of the autocorrelations measured for the laser pulses travelling in two different polarization states was found to be different, which is caused by the polarization anisotropy of the dispersion parameter D of HC-1060-02. Given the fiber length, autocorrelation FWHM durations of 1118 fs and 592 fs, and autocorrelation convolution factors [17], we calculate this anisotropy to be ΔD=1.7 ps/(nm·km) at 1065 nm wavelength.

 

Fig. 2. Autocorrelations of laser pulses compressed in a HC-PCF spliced to the laser output PM-SMF with a splice loss of 0.42 dB and PER=0.2 dB. The autocorrelations were measured for polarization orientation angles θ=0⁰,45⁰, and 90⁰ with respect to the principle axis of the HC-PCF.

Download Full Size | PDF

The cut-back, performed to bring the HC-PCF to the optimal length of 9.8m for the best pulse compression, allowed us to measure the HC-1060-02 loss of 0.084 dB/m, in perfect agreement with the specified value [13].

3. Splicing technique

One of the key features in our splicing procedure is a continuous laser illumination of the splice. A long pigtail of PM-SMF fiber was spliced with one end to the isolated output of a PM laser, with a PER of 24 dB. The laser was run continuously during the splice procedure with an output power of 4 mW. This laser is a modelocked oscillator of the amplified system, which output was compressed as described above.

The normal-cleaved PM-SMF and HC-PCF were placed into the splicer (Ericsson FSU 995 PM), and manual-mode translational and rotational alignment was performed in order to maximize the laser transmission and PER through the fiber joint. The maximized PER of transmitted laser light was in the range 15.5–18 dB prior to the fusion process. HC-PCFs of lengths 1–10 m were used in our experiments.

After the fibers were translationally and rotationally aligned in the splicer, we performed a series of repeated fuses with the fuse parameters shown in the Table 1. The technical definition of the fuse parameters used here is the same as in [4, 9]. The fuse parameters were chosen in order to prevent or minimize the hole structure collapse in the HC-PCF during the fuse, which is one of the main reasons for high splice loss in PCFs (see e.g. [9]). When using longer fuse times and/or higher fuse currents, we have observed strong deterioration of the transmission through the fiber joint, as well as decrease of the PER.

 

Fig. 3. Typical transmission change during the splice procedure, resulting in low-loss (here: splice loss 0.52 dB, PER 18.1 dB) and high-loss (here: splice loss 1.26 dB, PER 14.2 dB) splices. 0th fuse - fibers are brought into close mechanical contact ensuring best transmission and PER. 1st, 2nd, and 3rd fuses are made with the parameters shown in Table 1.

Download Full Size | PDF

We did not introduce an offset between the PM SMF-to-HC-PCF fiber joint and the splicer electrodes. We also did not apply the pre-fusion, since we found that it will only increase the splice loss, most possibly due to collapse of the holes at the interface of HC-PCF before it is brought in contact with a bulk of PM-SMF. We employed the standard relatively well-used pair of electrodes in this work, although we found that an occasional double arc discharge can occur since the Fusion current 1 of 6 mA is about the weakest current that provides a stable arc in our splicer. The splicer control circuit may fail to provide an adequate feedback on this event. This can probably be rectified by using a brand-new pair of electrodes, like in the work [4].

We have found that a series of three repeated fuses, separated by the time interval of several seconds, is enough to ensure high splice transmission and PER, and high enough mechanical stability of the splice. The spliced fibers could always be safely taken out of the splicer, placed on a glass substrate, and protected with a lacquer or UV-curable glue, which did not affect neither the splice loss, nor splice PER. Thus, the spliced fibers were ready for use in applications. We note here that the repeated fuse approach was also used in [18] for controlled hole collapse in splicing solid-core PCFs to SMFs.

The splice loss was calculated from the laser transmission through the spliced SMF-PCF assembly, taking into account the attenuation in the given length of HC-PCF. We have chosen a splice loss of 0.8 dB to be our quality benchmark criterion. The PER was measured at the end of HC-PCF using a commercial polarization extinction ratio meter.

We have observed a typical behavior pattern in the transmission through the fiber joint during the repeated fuse procedure, leading to either low-loss or high-loss (i.e. splice loss exceeding 0.8 dB) splices. In Fig. 3 the typical transmission histograms of the low-loss (splice loss 0.52 dB, PER 18.1 dB) and high loss (splice loss 1.26 dB, PER 14.2 dB) splices are presented. The fuse number 0 represents a mechanical butt-coupling of the PM-SMF and HC-PCF in the splicer, optimized in terms of the laser transmission and PER. One can see that in the splice procedure leading to a low-loss splice the laser transmission gradually decreases after each fuse,

Table 1. Fuse parameters

View Table

whereas in the case of a high-loss splice a dramatic drop in transmission through the fiber joint is observed already after a first fuse in the sequence. The transmission histograms shown in Fig. 3 are normalized to the transmission through the best-optimized mechanical contact between the fibers. The mechanical contact has its own additional loss, which is not represented in the above histograms.

Since the laser providing the illumination of the fiber splice was run continuously during the splicing procedure, we were able to visually observe the escape of the laser light induced by the fuses. In Fig. 4 the visual images of the splices with various losses and PERs are presented. These splices have no visible tangs, chips or gap irregularities. In Fig. 4(a) an image of the splice with a loss of 0.31 dB and PER of 19.5 dB is shown. The hole structure in the PCF does not show visible signs of collapse. Escaping laser light is also not observable in this image. In Fig. 4(b) a splice with a loss of 0.6 dB and PER of 19.1 dB is shown. Even though the splice loss and PER are acceptably good, some escaped laser light is observed at the entrance into the PCF, and a slightly collapsed hole structure is observed with a collapse waist positioned at a distance of around 110 µm away from the fiber-to-fiber interface. In the splices shown in Figs. 4(a) and 4(b) the three-step splicing procedure with the fuse parameters from Table 1 was used. In the Fig. 4(c) a splice with a loss of >3 dB and PER <8 dB is shown. In this case we used longer fuse times than shown in Table 1. Significant escape of the laser light from the splice area, and major collapse of the hole structure is seen in the HC-PCF. This is a typical result of excessive heat applied to the fibers during the fusing. The collapse waist is found to be at approximately the same distance from fiber-to-fiber interface, as in the previous case.

We have observed that the PER of the low-loss splice improves from the initial values of 15.5–18 dB (PER at optimized mechanical contact) to 17–20 dB after the series of fuses is performed. We believe that Rayleigh scattering at the cleaved fiber facet is the reason for the higher polarization cross-talk before the splice. For the high-loss splice the PER may drop dramatically, which is accompanied by the fiber-joint transmission drop and presence of the laser light escape in the visual splice image.

The return loss of the PM-SMF-to-HC-PCF splices was measured with a setup using a fiber-optic circulator. The setup was calibrated with the 4% reflection from the normal-cleaved PM-SMF. We have found that the return loss from the PM-SMF-to-HC-PCF was identical to that of a normal-cleaved PM-SMF and amounts to -14 dB, which is natural for the interface between the normal-cleaved solid core and air-core fibers with similar core sizes. In [10] a successful approach was demonstrated to optimize the return loss of the non-PM SMF-to-HC-PCF splice, where the fibers were cleaved at 80 angle before fusing. Such an approach will unfortunately be difficult to implement in PM splicing, since both fibers will have to be cleaved at an angle where PM axes of both fibers are at the same orientation relative to the plane of this cleave (such as e.g. one of PM axes of both fibers must belong to the plane of their respective angle cleaves).

 

Fig. 4. Visual images of the splices made under laser illumination, coming from the PM-SMF. (a) Splice loss of 0.31 dB, PER of 19.5 dB; (b) Splice loss of 0.6 dB, PER of 19.1 dB. (c) Longer fuse times are applied, than shown in Table 1. Splice loss >3 dB, PER <8 dB. Circles mark the areas where escaping laser light is visible. Different apparent sizes of PM-SMF core areas are due to different orientations of stress rods relative to the camera.

Download Full Size | PDF

4. Discussion

Here we will analyze the possible mechanisms for loss and reduction of PER during fusion splicing of PM-SMFs and HC-PCFs. In general, the key reasons for the splice loss are mode mismatch, defect or collapse of the hole structure in the HC-PCF, and the presence of scattering particles around the fiber-to-fiber interface. All of the above are radiation loss channels, and may lead to the presence of escaped laser light on the splice image.

TheMFDs of the fibers [12, 13] suggest that the mode-matching will not be a reason for high loss under the assumption of Gaussian single modes and co-aligned fiber cores.

As we have mentioned above, the heat load applied to the HC-PCF may create the conditions for a minor collapsing of the hole structure, which deteriorates the photonic bandgap (PBG) effect, and thus the guidance in the air core. Coupling of part of the light launched from the SMF into the holey cladding of the HC-PCF will lead to radiation loss, observable as leaking light in the splice image (see Figs. 4(b,c)). We note here, that minor collapse of the hole structure of HC-PCF is not specific only to the lossy splices, and was also observed at some splices with low loss and high PER.

The possible reason for the hole-structure collapse waist being positioned at a distance of around 110 µm away from the fiber-fiber interface is that during the fuse, the overlap of 4 µm (see Table 1) ensures that the struts between the air-holes are quickly attached to the bulk of the SMF, which prevents the following surface tension from pulling the air-hole structure away from the interface. In the work [2] the recess of the hole structure resulting from the fuse, due to the action of surface tension along the struts connecting the air holes, was identified as a main source of higher splice loss. This is most likely not the case in our work, since here it would result in a repeatable, and not an occasional high loss, as we observe.

We thus believe that neither mode mismatch, nor minor collapsing of the hole structure in the PCF is the main reason for the occasional higher splice loss in our experiments.

Cleaving the HC-PCF may result in occasional creation of a small glass particle, such as a fragment of a strut between the air holes in the PBG structure, which may find itself inside an air-core of the fibers. Such a particle will clearly act as a scattering center, especially after the fuse is applied, and the melted particle will locally distort the core. We believe that creation of such strongly scattering particles inside the hollow core of the PCF can be one reason for the dramatic drop in the transmission through the fiber joint observed after the first fuse in the splice sequence in case of lossy splice (see Fig. 3). Another reason for an occasional high loss can be bad local attachment of the struts between the air-holes to the bulk of SMF, leading to the local PCF structure recess, also resulting from the cleave defect.

In order to check the above hypothesis we have performed the series of 5 cleaves of HC-PCF fiber and inspected the cleaved facets with an SEM. In 2 out of 5 cleaves we were able to find the cleaving defects - in one case the fragment of the hollow-core wall was chipped off, and in one case a row of about 10 holes in the PBG structure was damaged. The SEM images of defect fiber facets are shown in Fig. 5. The thickness of the wall fragments is found to be around 200 nm, and the lateral sizes can reach 12–15 µm. Whereas the chipped-off wall fragment will clearly create a strong scattering center in the fiber core, the damaged row of holes in the PBG structure will prevent good attachment of the hole structure of the PCF to the bulk of the SMF. The latter will create the conditions for the hole structure recess during the fuse [2], which is otherwise avoided in our splicing approach.

 

Fig. 5. Defects of HC-PCF cleave: (a) Fragment of a hollow core wall is chipped off. (b) A row of holes in PBG structure is damaged. Courtesy of L. Wei, DTU Fotonik.

Download Full Size | PDF

Altogether, we have performed a series of 10 splices using the technique described above, of them 8 - with a maximized PER. The splices with a loss exceeding 0.8 dB, which was a splice quality benchmark in this work, could be predicted at an early stage from monitoring the transmission through the fiber joint during the splicing process, and from the appearance of escaping laser light in the visual image of the splice.

In Fig.6 we show the observed dependency of splice PER on splice loss. It is seen from this figure that higher PER can be ensured when the splice loss is minimized. The negative correlation between the PER and splice loss supports our hypothesis, that it is the presence of local PCF structure defect that is responsible for the dramatic reduction of splice properties. Indeed, the presence of a strong scatterer, such as a small glass fragment fused into the air-core of the fiber, will not only create propagation loss (thus making the scattered laser light visible on the splice image), but will also enhance the polarization cross-talk between the fiber axes, therefore compromising the resulting PER. Rayleigh scattering is the key reason for the polarization cross-talk in birefringent fibers [19]. The local PCF structure recess will also lead to local fiber distortion, and therefore to increased polarization cross-talk. We speculate that higher splice loss of 2.3 dB, mentioned in work [6], could possibly be the reason for the conclusion about the non-PM properties of the HC-PCF similar to the one used here.

 

Fig. 6. Symbols - observed dependency of PER on splice loss. Solid line - linear regression with the slope of -4.56. Dashed lines mark the positions of mean values for splice loss of 0.62 dB and PER of 19 dB for defect-free splices (splice loss <0.8 dB).

Download Full Size | PDF

We have also performed the measurements on the reciprocal HC-PCF-to-PM-SMF splice loss. From the series of experiments we found the reciprocal loss to be 2.19±0.33 dB, ranging 1.77–2.57 dB. This is considerably higher than that of the forward loss. We explain such a high reciprocal loss by the fact that the 7-cell HC-PCF is not strictly single-mode, and thus the mode evolution at the interface and upon propagation in this fiber will distort the initially launched mode. The same effect of higher reciprocal loss was observed and explained in the work [4]. The exact numerical analysis of the HC-PCF-to-PM-SMF splice loss would involve integration over the mode structure of HC-PCF, where the exact intensity distribution between different modes is unknown [4]. Such an analysis is a subject for a separate theoretical and experimental investigation.

5. Conclusions

In conclusion, we have demonstrated a successful technique of reproducible low-loss polarization-maintaining fusion splicing between the PM-SMFs and HC-PCFs, using a repeated fuse approach under laser illumination coming from the PM-SMF. The combination of moderate fusion currents and short fusion times during the fuses, separated by a time interval of several seconds, prevents a significant collapse of the hole structure, and allows for good attachment of the HC-PCF strut structure to the bulk of PM-SMF.

The series of 8 splices where both splice loss and PER were optimized, resulted in 6 successful splices with a splice loss below 0.8 dB, which was our quality benchmark. The losses of successful splices varied in the range 0.31–0.77 dB, with an average splice loss of 0.62±0.24 dB. The PER of successful splices varied in the range 17.3–19.7 dB, with an average PER of 19±0.68 dB. In Fig. 6 one can see that the scattering of individual splice loss and PER values for defect-free splices (i.e. splice loss <0.8 dB) in respect to their respective mean values is rather low, and most of the individual values are grouped close to their respective mean value. The return loss of the PM-SMF-to-HC-PCF splice was found to be -14±0.2 dB, which is expected from the ideal interface between glass and air. Such splice properties make the PM-SMF-to-HC-PCF assemblies relevant for use in PM fiber laser applications, such as monolithic fiber laser pulse compressors, thus enabling direct end-of-the-fiber femtosecond pulse delivery, as we demonstrated in this work.

In spite of a higher HC-PCF-to-PM-SMF reciprocal splice loss, caused by the multi-mode nature of the HC-PCF, and the return losses of both forward and reciprocal splices of -14 dB, complex PM-SMF–HC-PCF–PM-SMF fiber assemblies could be used for intracavity dispersion control in uni-directional ring-cavity lasers, where the sub-cavity effects can be neutralized by an intracavity isolator.

Fusion splicing is a stochastic process, with many interplaying factors, such as individual cleave quality, possible fluctuations in fuse currents and times, local fiber structure fluctuations, fiber alignment accuracy etc. In general, we found that by monitoring the transmission through the fiber joint during the splice (see Fig. 3), and by inspecting the visual images of the splice (see Fig. 4), one can predict the lossy splice already at first stage of the repeated fuse process. All splices with a splice loss exceeding 0.8 dB demonstrate a significant drop in the PM-SMF-to-HC-PCF joint transmission already after the first fuse is applied. In this case we also always observed noticeable escape of the laser light at the entrance of the HC-PCF core, even though a significant collapse of the hole structure was not always apparent. In case of a low-loss splice there is a slight gradual decrease in transmission after each fuse, and the escape of the laser light is either not observable, or small.

Generally, our splicing approach does not lead to significant collapse of the air hole structure. Thus, we believe that the key factor leading to occasional higher splice loss and polarization cross-talk in our fusion splicing approach is the PCF structure defects resulting from the cleave. Typical defects are: glass strut fragments situated in the hollow core of the fiber and acting as strong scattering centers; and damaged holes in the PBG structure preventing good attachment of HC-PCF structure to the bulk of PM-SMF during the fuse and leading to the local PBG structure recess. Presence of such defects, leading to low splice quality, normally reveals itself after the first fuse was applied to the fiber joint. The observed rate of success of low-loss PM-SMF-to-HC-PCF splice of 7/10 correlates well with the observed rate of success of defect-free HC-PCF cleave of 3/5.

A clear negative correlation between the splice loss and PER was found, which demonstrates the importance of low-loss splicing for the good polarization-maintaining properties of the resulting fiber assembly.