1. Introduction

While nonlinearity is preferably avoided in transmission systems, nonlinear optics has attracted significant interest for signal processing. The use of nonlinear effects has proved to be not only an elegant but also powerful technique to achieve all-optical ultra fast processing such as regeneration [1], wavelength conversion [2], multicasting [3] or time-division demultiplexing [4]. Considering future optical networks where a wide variety of modulation formats will be supported, it is important to look at nonlinear processes that can preserve both amplitude and phase. In that sense, four-wave mixing (FWM) is uniquely suited for processing of hybrid-modulated signals while offering ultrafast response and efficiency. FWM with record gain and bandwidth has been achieved in silica highly nonlinear fibers (HNLF) [5]. However novel platforms, such as silicon and soft glasses [6], have also emerged as potential candidates for nonlinear processing. The nanoscale engineering of silicon waveguides can be used to enhance nonlinearity, and chip scale all-optical functions have been demonstrated [7, 8]. While silicon is a solution for implementation of FWM on a monolithic optical device, it suffers from strong two-photon absorption (TPA) and free-carrier absorption (FCA) in the 1550 nm telecom band, presenting a fundamental obstacle and limiting parametric gain values. Chalcogenide glasses have raised attention due to their transparency deep in the infrared region and nonlinearity close to 1000 time higher than silica. Chalcogenide planar waveguides and nanowires have been used for wavelength conversion and all-optical signal processing [9, 10]. In addition, Chalcogenide microstructured fibers have also been of great interest [11] In addition, the ability to dispersion engineer the microstructured fiber through control of its geometry could enable phase matching between desired spectral bands for efficient nonlinear interaction in cm-long segment of fibers. Recently, four-wave mixing in a 43-cm long segment of suspended core chalcogenide fiber was demonstrated over a 5 nm bandwidth and nonlinearity as high as 31300 W⁻¹km⁻¹ was measured [12].

In this paper we present the experimental demonstration of four-wave mixing in a 2.5-cm long segment of Chalcogenide microstructured fiber. A continuous-wave (CW) pump at 1555 nm with 345 mW of power enabled the generation of idler waves over more than 70 nm. We also present the ultrafast wavelength conversion of picoseconds pulses from a 8 GHz pulsed signal.

2. Four-wave mixing experimental setup

The experimental setup is shown in Fig. 1 . Two tunable external cavity lasers (ECLs) are used as pump and signal. The pump laser (ECL1) was amplified by a high power erbium doped fiber amplifier (EDFA) and combined with the signal (ECL2) through a wavelength division multiplexer (WDM). The signal laser could be tuned over 70 nm for a wide wavelength range characterization of the fiber. A 50 GHz WDM was used to act both as a combining element and filtering element to eliminate excess amplified stimulated emission (ASE) from the EDFA. The pump and signal were coupled to the segment of Chalcogenide microstructured optical fiber (MOF) using a lens-tipped fiber (see Fig. 2(a) ). The Chalcogenide mixer output was fiber-coupled back to single mode fiber (SMF) using a second lens-tipped fiber. The output was then observed on an optical spectrum analyzer (OSA). After rough positioning of the input and output lens-tipped fibers, the alignments were achieved by adjusting the coupling stages with sub-μm precision while monitoring the output power through a 1% tap and a power meter. After final optimization, the coupling loss at each facet was estimated to be 6.5 dB for a total insertion loss of 13 dB.

 

Fig. 1 Experimental setup. ECL: external cavity laser; EDFA: Erbium doped fiber amplifier; PC: polarization controller; WDM: wavelength division multiplexer; MOF: microstructured optical fiber; OSA: optical spectrum analyzer; PM: power meter.

Download Full Size | PDF

 

Fig. 2 (a) Input lens-tipped fiber and Chalcogenide fiber observed under microscope. Magnified images of output facet observed from the CCD camera: (b) Illumination of the entire fiber from a wide input beam. Core, microstructure and claddings can be seen. (c) Optimized coupling through the core of the microstructured fiber.

Download Full Size | PDF

The fiber under test was a segment of AsSe Chalcogenide microstructured fiber provided by I. D. Aggarwal of Naval Research Laboratory under contract to Lockheed Martin Corporation. The fiber was fabricated using a direct extrusion technique. The structured preform was extruded and stretched into a microcane before drawing. The fiber was designed with the following characteristics: core diameter of 3.4 μm, pitch of 2.1 μm, hole diameter of 1.1 μm and an air-fill-fraction of 25%. An effective area A_eff ≅ 6 μm was provided by the vendor. Difficulty in the fabrication process resulted in a highly brittle structure. Defects such as hole collapses resulting in poor structure uniformity along the fiber length were observed, leading to high losses and decreased nonlinearity. To ensure good uniformity and light guidance in the core over the entire length of fiber used, only short segments could be experimentally used. Multiple segments of fibers were tested before selection. Parameters such as total coupling loss, loss through the segment and mode profile at the output of the fiber were used to facilitate the selection process. A 2.5-cm long segment, showing adequate uniformity and low losses was chosen for the demonstration. Testing of the segment and initial alignment was achieved by injecting amplified spontaneous emission (ASE) noise through the fiber tipped lens and observing the output of the fiber through an imaging lens and on an InGaAs camera. Observed magnified images of the output facet of the fiber can be seen in Fig. 2(b) and (c). Non-optimized positioning of the lens-tipped fiber resulted in the illumination of the entire fiber with most of the light travelling through the outside cladding (Fig. 2(b)). After optimization, light was coupled to the core of the fiber at the center of the hexagonal microstructure (Fig. 2(c)).

3. Four wave mixing results

In combination with available WDMs for optimum pump coupling and noise rejection, the pump position was optimized by trials and was ultimately positioned at 1555 nm. Previous results with a 1562.2 nm pump were obtained and described in [13]. The pump was amplified to 1.5 W of power, leading to a maximum pump power of 345 mW coupled in the Chalcogenide fiber. The signal wavelength was tuned over 70 nm between 1480 nm and 1550 nm with an optical output power maintained at 8.5 dBm. We calculated that 1.5 dBm of signal power was actually coupled to the microstructured fiber, the 7 dB total loss being attributed to the 6.5 dB coupling loss and 0.5 dB WDM loss. Before measuring four-wave mixing associated with the microstructured fiber, potential initial mixing associated with single mode fiber (SMF) was first measured. A 12.5 dB loss, to emulate the effect of the MOF, was created by removing the MOF and coupling fiber-tip to fiber-tip with a free-space gap. The optical spectrum at the output of the setup was then collected on the spectrum analyzer. Very low level of four-wave mixing (−60 to −50 dB of efficiency) was observed over a narrow 15 nm bandwidth. No FWM products in the SMF patch cord were generated beyond Δλ = 15 nm. A comparison between initial and MOF induced four wave mixing at the peak gain can be seen in Fig. 3(a) . This initial mixing is the result of the high pump power (1.5 W in fiber) and the signal seed co-propagating through the same length of SMF between the WDM and the input of the Chalcogenide MOF.

 

Fig. 3 (a) Optical spectra with and without Chalcogenide fiber, comparison at peak FWM. (b) Optical spectra at output of Chalcogenide fiber. Pump positioned at 1555 nm and signal swept between 1490 nm and 1550 nm. Idlers generated between 1559 nm and 1625 nm

Download Full Size | PDF

After insertion of the Chalcogenide MOF between the lens-tipped fibers, the optical spectra at the output of the fiber were once again collected for different signal wavelengths. The consistency between the subsequent measurements was monitored at the 1% tap. Readings on the power meter gave an indication of the amount of pump power coupled in the fiber at all times. While air flow and vibrations led to small fast constant power fluctuations within 1 to 2 dB, no significant changes in pump power were recorded while the measurements were performed. The results, shown for a signal tuned on the short wavelength side of the pump only, are plotted in Fig. 3(b). Additional four-wave mixing was clearly achieved in the Chalcogenide MOF: in comparison with the initial SMF mixing, idlers were generated with increased efficiency and over a wider bandwidth in the short 2.5-cm segment despite a 6.5 dB lower pump power. The conversion efficiency η_eff = P_idler(out)/P_signal(out), where P_idler(out) is the idler power and P_signal(out) is the signal power, was measured from the collected OSA spectra. The measurements were corrected to take into account the initial FWM generation in SMF. The resulting conversion efficiencies as a function of pump-idler wavelength separation (Δλ = λ_p - λ_idler) are plotted in Fig. 4 . Generation of the four-wave mixing product (idler) was achieved over more than 70 nm on the anti stoke side of the pump with a 3 dB conversion bandwidth of 40 nm, including the symmetric lobe to the short wavelength side. The generated idler output power was measured as a function of signal output power and the results are shown in Fig. 5(a) . As expected, a linear behavior with a one to one relationship was observed. Finally, the conversion efficiency as a function of pump power for a signal positioned 3 nm away from the pump was also characterized and is plotted in Fig. 5(b). Once again, linear behavior with a slope of 2 on the log-log scale was measured, as expected from the four wave mixing process. No onset of saturation, typically linked to a decrease in pump power coupling from damages to the fiber, was observed and higher conversion efficiencies could be expected from higher coupled pump power.

 

Fig. 4 Measured conversion efficiencies in MOF as a function of Δλ = λ_p - λ_idler, pump-idler wavelength seperation.

Download Full Size | PDF

 

Fig. 5 (a) Measured signal-idler power relationship. (b) Conversion efficiency as a function of pump power

Download Full Size | PDF

The nonlinear coefficient γ, and group velocity dispersion D were determined from the measured data. The power of the FWM term (P_FWM) is given by Eq. (1) [14], the FWM efficiency η which depends on the phase matching between the pump, signal and idler is given in Eq. (2), and the phase mismatch factor Δk expressed for the pump degenerate case is given in Eq. (3):

In these equations, P_P is the input pump power, P_S is the signal power, L is the fiber length, α is the propagation loss, Δω = |ω_P−ω_S| is the frequency detuning with ω = 2πc/λ, and c is the speed of light. For a signal positioned close to the pump, η = 1. The data from Fig. 5(b) was thus used in conjunction with Eq. (1) to determine γ. After fitting, a value of γ ≈1500 W⁻¹km⁻¹ was extracted, in agreement with the fiber values given by the manufacturer.

The wavelength detuning data, as plotted in Fig. 4, was then used in conjunction with the derived nonlinear coefficient value γ for the estimation of D. A value of α = 2 dB/m was used. The experimental data was fitted using Eqs. (1), (2) and (3) [14]. The fitting parameter D ≈-260 ps/nm/km was determined from the experiment. The conversion bandwidth${\Omega }_{FWM}\approx {\left[\raisebox{1ex}{$(8{\pi }^2\text{c)}$}\!\left/ \!\raisebox{-1ex}{$({\lambda }^2\left|D\right|L)$}\right.\right]}^{1/2}$, is dependent on the magnitude of the group velocity dispersion in the small-gain limit. Waveguides with lower magnitude of D would therefore have a large conversion bandwidth.

4. Wavelength conversion setup and results

The experimental setup was finally modified (see Fig. 6 ) to verify that the four-wave mixing process in the Chalcogenide microstructured fiber does not significantly degrade an optical signal by demonstrating wavelength conversion of a high speed signal. The signal CW seed was replaced by a cavity-less pulse source driven by a CW external cavity laser as described in [15]. The 8 GHz, 1.4 ps full width half maximum (FWHM) pulse train was positioned at 1566 nm. At the output of the microstructured fiber, the generated idler was separated from the pump and the signal by a cascade of WDMs, necessary to extinct the strong pump. The idler was then amplified with an EDFA before visualization on an ultrafast optical sampling oscilloscope (500 GHz bandwidth). The wavelength conversion was observed on the optical spectrum analyzer for pump ON and pump OFF (Fig. 7(a) ). A conversion efficiency of −37dB was measured, in accordance with the CW characterization described previously. Self phase modulation (SPM) on the signal was observed due to high peak power of the signal which propagated with a 15 dBm average power. SPM was primarily generated in the relatively long (more than 10 m) single mode fiber linking the source to the Chalcogenide fiber. The high quality 8 GHz pulsed signal waveform at the output of the Chalcogenide fiber for pump OFF, can be seen in Fig. 7(b). As expected, no idler was generated when the pump was turned OFF (Fig. 7(c)). As the CW pump was turned ON, idler generation was achieved and the pulsed output was observed on the oscilloscope (Fig. 7(d)). Due to low conversion efficiency and an additional 5 dB loss from the cascade of WDM, the power level of the idler at the input of the optical amplifier before the optical sampling scope was low which resulted in excess noise generation. However, clear wavelength conversion was achieved and a pulsed idler was generated with similar characteristics as the input signal. We expect the performance to significantly improve by adequate idler filtering (i.e. loss optimization), signal power/pump positioning optimization and improved microstructured fiber fabrication.

 

Fig. 6 Experimental setup for wavelength conversion of ps pulse train

Download Full Size | PDF

 

Fig. 7 (a) Optical spectra of the input pulsed train (dashed line) and the Chalcogenide fiber output (full line). Temporal waveform of (b) signal at output of MOF, pump off; (c) idler at output of MOF, pump off, (d) idler at output of MOF, pump ON

Download Full Size | PDF

5. Conclusions

We demonstrated continuous-wave four-wave mixing in a short segment of Chalcogenide microstructured fiber over a broad bandwidth. Idler waves were generated over more than 70 nm with 345 mW of CW pump power in only 2.5-cm of fiber, allowing for conversions across the S, C and L bands. The wavelength conversion capability of the fiber was tested with a 1.4 ps FWHM 8 GHz pulsed signal. The generation of a wavelength converted idler was observed despite non-optimized operating conditions. Indeed, given the quality of the Chalcogenide sample and its non-optimized dispersion characteristic, the obtained results clearly indicate the high potential of such platform for ultrafast signal processing and significant progress can be expected as the fabrication process matures.