1. Introduction

The trend of transition from person-to-person communication to the internet of things has increased the communication capacity. This in turn has accelerated the demand for large-capacity optical communications. Although this demand has been realized by improving the spectral efficiency, it is approaching the nonlinear Shannon’s limit [1]. We are interested in bandwidth expansion to realize large-capacity optical communication. In the current optical communication system, a large-capacity could be realized by wavelength division multiplexing (WDM) using an optical amplifier. However, in the conventional erbium-doped fiber amplifier (EDFA), the gain band is limited to the wavelength range of approximately 1.55 µm. To expand optical communication bandwidth, it is strongly desired to have the optical amplifier operate in a wavelength band other than the 1.55 µm band. The low-loss window of optical fiber is in the 1.3–1.8 µm band. Making these wavelength ranges available will contribute to increasing the optical communication capacity by increasing the bandwidth per channel and the number of WDM channels. To expand the bandwidth of WDM transmission, several approaches have been proposed, such as hybrid Raman/EDFAs [2] and a combination of Tm-doped fiber amplifiers and EDFAs [3]. The wavelength range achievable with these approaches is limited because it depends on the nature of the available rare-earth elements and transmission fiber. In contrast, optical parametric amplifiers (OPAs) using nonlinear optical media are attractive because they can amplify any wavelength in principle by designing the phase matching characteristics. For example, transmission in the 1 THz band using OPAs based on highly nonlinear fiber (HNLF) [4] and transmission in the band at 10 THz using OPAs based on periodically poled LiNbO₃ (PPLN) waveguides [5] have been demonstrated. The phase matching of HNLF is determined by the dispersion characteristics of the fiber, and it is challenging to realize amplification at an arbitrary wavelength. In PPLN, it is easy to change the phase matching characteristics by engineering the χ⁽²⁾ grating; therefore, it has the potential for amplification in an arbitrary wavelength range. In addition, the use of idlers in OPAs in a wide wavelength range will be effective for flexible wavelength band operation [6,7].

Recent advances in periodically poled LiNbO₃ (PPLN) waveguide technologies have enabled various parametric interactions with high efficiency [8–10]. To expand the bandwidth of the optical amplifier, several groups have studied OPAs using PPLN waveguides [5,11–13]. Recently, we reported parametric wavelength conversion using multiple quasi-phase-matched (M-QPM) LiNbO₃ waveguides in a bidirectional configuration [14]. The M-QPM device enabled pump generation through second harmonic generation (SHG) and difference frequency generation (DFG)/OPA using individual phase-matching conditions [15,16]. The advantage of this configuration is that we can precisely design the wavelength range of parametric gain by designing the spacing between QPM wavelengths for SHG and DFG/OPA processes. A drawback of this configuration is that it compromises the parametric gain to obtain multiple QPM peaks. The goal of this study is to demonstrate the broadband OPA by varying the detuning of the QPM wavelength for the OPA process and the SH pump generation. For this purpose, we employed a PPLN waveguide, which exhibits a single QPM peak, for the OPA process to obtain a higher parametric gain. In addition, we employed an M-QPM device for pump generation through SHG process. In principle, various detuning can be obtained by employing two PPLN waveguides for SHG and OPA and changing the temperature of two waveguides. In reality, when the waveguide is assembled in a fiber pigtail module, the coupling efficiency between the waveguide and fiber may vary owing to thermal expansion of the module components under a significant temperature change. The M-QPM device allows us to generate an SH pump at various wavelengths by the QPM peak choice without temperature change. This configuration allows us to examine a wide range of detuning while minimizing the change in the coupling efficiency by combining moderate temperature changes of the PPLN waveguide for OPA. Taking advantage of this configuration, we demonstrated OPA/DFG in the wavelength range of 1.3 –1.8 µm.

2. Principle of operation

As described in a previous study, the gain band of the OPA process can be tuned by setting the pump wavelength to less than half the SHG QPM wavelength [8]. Figure 1 illustrates the calculated phase-matching curve for the OPA/DFG process with several detuning conditions. In this calculation, we assumed that the SHG QPM wavelength of the PPLN waveguide was 1535 nm, and we defined the detuning as the wavelength difference between the SHG QPM wavelength and twice the pump wavelength. We also assumed that the length of the PPLN waveguide was 50 mm, and we used the refractive index dispersion of bulk LiNbO₃ [17]. As depicted in Fig. 1, by varying the detuning from 0–6 nm, the parametric gain range can cover the low loss window of the optical fiber, which is 1.3–1.8 µm. Owing to the dispersion of LiNbO₃, the peak width of the phase-marching curve becomes narrower as the detuning increases. Although the calculation assumed a uniform waveguide, in reality, the effective refractive index of the waveguide varied owing to size variation. This means that the demand for uniformity of the PPLN waveguide will increase under wide detuning conditions.

 

Fig. 1. Phase matching curves for OPA/DFG process with several detuning conditions.

Download Full Size | PDF

We conducted an experiment to demonstrate the broadband OPA with several detunings by using the experimental setup illustrated in Fig. 2. An external cavity laser diode (ECL) was used to generate pump light in the 1.535 µm band. The pump was amplified using an EDFA and it was injected into a multiple-QPM device. The M-QPM device was assembled in a fiber pigtail module, which is capable of temperature control using a Peltier device. The pump light was converted to a 0.768 µm band second harmonic pump. Signal light was generated from another ECL. The SH pump was combined with the signal and injected into a PPLN waveguide, which was also assembled in a fiber pigtail module. The M-QPM and PPLN waveguides and their modules were fabricated by an in-house process. The periodical χ⁽²⁾ gratings were fabricated by electrical poling of an Zn:LiNbO₃ substrate, and they were processed into waveguides using direct-bonding and dry etching techniques [9]. The fiber pigtail modules were assembled using laser welding. The signal was amplified through the OPA process, and the idler was generated through the DFG process. The amplified signal and idler were observed using an optical spectrum analyzer (OPA). Because the measurable wavelength of the OSA is up to 1700 nm, a multi-channel analyzer (MCA) was used to observe idlers with longer wavelengths. The temperature dependence of the SHG QPM wavelength was approximately 0.15 nm/°C. The M-QPM device exhibited three QPM peaks with a spacing of 0.8 nm. The combination of temperature control and choice of the QPM peak allowed the detuning to be adjusted over a wide range.

 

Fig. 2. Experimental setup.

Download Full Size | PDF

In some experiments, the signal light was modulated using a LiNbO₃ IQ modulator, and the bit error rate (BER) of the idler was measured to examine the deterioration of the signal quality in the wavelength conversion process. In this case, the signal light wavelength was set to 1483 nm and a 10 Gbaud quadrature phase shift keying (QPSK) modulation was applied. The idler at a wavelength of 1590 nm was separated by a band-pass filter (BPF), and the optical power was adjusted using an attenuator (ATT). For simplicity, one of the quadrature signals was detected using a differential phase-shift keying (DPSK) receiver. The DPSK receiver consisted of a preamplifier with an L-band EDFA, BPF, delayed interferometer, and balanced photodiode (PD). The BER was measured using an error detector.

To examine the parametric gain spectra before conducting an experiment of OPA/DFG, the spectra of parametric fluorescence were observed by injecting only the SH pump light into the PPLN waveguide. Figure 3 depicts the measured spectra under several detuning conditions. Dotted lines were measured using the MCA. The output power of the EDFA was set to 33 dBm. To achieve the wavelength range of OPA/DFG, the temperature of the M-QPM module was fixed at 36 °C, and the PPLN module was tuned from 54 °C – 80 °C. From the measurement of the SHG phase matching curve, we confirmed that the change in the coupling efficiency between the PPLN waveguide and the optical fiber in this temperature range was within 0.5 dB. Beyond this temperature range, the coupling efficiency dropped sharply due to the thermal expansion of parts in the module. This is because the specific module used in this study is assembled by setting the temperature at 60 °C, which is close to its operating temperature. Pump at 1535.7 nm was used to obtain detunings of 0.2 nm and 0.35 nm and pump at 1534.9 nm was used to obtain other detunings. The SH power generated from the M-QPM module was in the range of 22.1 –22.8 dBm and it depended on the choice of phase matching wavelength. Wide detuning was enabled by using an M-QPM device and changing the pump wavelength while keeping the temperature range of the PPLN waveguide within the range where the coupling efficiency does not change significantly.

 

Fig. 3. Optical spectra of parametric fluorescence with several detuning condition. Dotted lines were measured by MCA.

Download Full Size | PDF

The broader peaks of the idler with wavelengths longer than 1700nm are due to the limited resolution of the MCA. The sensitivity of the MCA was calibrated based on the signal intensity measured using the OPA. As the detuning was increased, the peak wavelength of the signal and idler changed to shorter and longer wavelengths, respectively.

Based on the spectra of parametric fluorescence, the signal wavelength range was determined, and the OPA/DFG experiment was conducted with several signal wavelengths. Figures 4 and 5 depict the superimposed spectra of the signal and idler with detunings of 0.2 nm and 2.2 nm, respectively. The black dotted lines depicted in Figs. 4 and 5 represent the signal spectra without SH pump injection. We observed the amplification of the signal and generation of an idler with a higher intensity than that of the input signal. Conversions were obtained from 1.50–1.57 µm with a detuning of 0.2 nm and from 1.41 –1.68 µm with a detuning of 2.2 nm.

 

Fig. 4. Optical spectra of the signal and idler with detuning of 0.2 nm.

Download Full Size | PDF

 

Fig. 5. Optical spectra of the signal and idler with detuning of 2.2 nm.

Download Full Size | PDF

We repeated the experiment with several detuning conditions. Figure 6 depicts the wavelength dependence of the OPA gain at the signal and DFG conversion efficiency under several detuning conditions. Dotted lines were measured by MCA. OPA/DFG in the range of 1.3 –1.8 µm were demonstrated by varying the detuning. The choice of multiple QPM peaks for the SHG process and temperature control enabled us to demonstrate broadband operation. Maximum gain and conversion efficiency were 15.0 dB and 14.7 dB, respectively. The gain and conversion efficiency decreased as the wavelengths deviated from the degeneracy point. This phenomenon occurred because of the tolerance of the phase-matching curve in the OPA/DFG process, which became narrower as the signal/idler wavelength was away from the degeneracy point, as depicted in Fig. 1.

 

Fig. 6. Wavelength dependence of OPA gain and DFG conversion efficiency with several detuning conditions. Dotted lines were measured by MCA.

Download Full Size | PDF

In an actual device, the local phase-matching wavelength is distributed along the waveguide owing to the distribution of the waveguide size. Because the phase matching tolerance in the OPA/DFG process is loose when the signal/idler wavelengths are close to the degeneracy point, the influence of the waveguide distribution seems to be negligible. In contrast, the phase matching tolerance is narrower when the signal/idler wavelengths are away from the degeneracy point, so that the gain and conversion efficiency is lowered because of the accumulation of the phase mismatch of the signal/idler wave generated at each part of the waveguide.

Figure 7 depicts the relationship between the detuning and peak wavelengths of the OPA gain and the DFG conversion efficiency. It is confirmed that OPA/DFG in 1.3 –1.8 µm band could be realized by setting the detuning from 0–6 nm. The detuning dependence of the peak wavelength of the parametric fluorescence is also plotted in Fig. 7. Good agreement between them proves that parametric fluorescence measurements are useful in estimating the gain peak. The solid line depicted in Fig. 6 indicates the theoretical predictions calculated by using the refractive index dispersion of bulk LiNbO₃. The slight deviation from the theoretical prediction in large detuning occurred because the waveguide dispersion was ignored in the calculation. The temperature change sensitivity in the detuning is determined by the temperature dependence of the QPM wavelength, which is approximately 0.15 nm/°C. For example, the gain peak wavelength slope versus detuning is 41 nm/nm at 1450 nm according to Fig. 7. This indicates that a temperature accuracy of 0.16 °C is required to maintain the gain peak wavelength with a 1 nm accuracy. Let us consider the response time when changing the gain band by a detuning control. For example, to change the detuning by 1 nm, which corresponds to changing the gain peak wavelength by 41 nm, the PPLN module temperature must be changed by 6.6 °C. Depending on the performance of the temperature controller, such temperature changes will require several minutes. By optimizing the combination of the phase matching wavelength and operating temperature of the M-QPM and PPLN devices, we can expect amplification in a wider wavelength range with less temperature change.

 

Fig. 7. Peak wavelength pf OPA gain and DFG conversion efficiency as a function of detuning.

Download Full Size | PDF

Figure 8 depicts the BER performance of the idler at a wavelength of 1590 nm. As a reference, we also conducted a back-to-back experiment using a 1590 nm laser diode, the same IQ modulator, and a DPSK receiver. As depicted in Fig. 8, the wavelength conversion process imposes a negligible power penalty up to a BER of 10⁻⁹. Because the differential detection employed in this work is resistant to phase noise, the phase noise of the pump laser may affect the signal quality of the idler.

 

Fig. 8. BER performance of the idler and back-to-back experiment.

Download Full Size | PDF

3. Conclusions

In this study, we demonstrated the OPA/DFG in the 1.3 µm –1.8 µm band by using SH pump detuned from half of the SHG QPM wavelength of PPLN. The use of the M-QPM device for SH pump generation and temperature control of the PPLN waveguide enabled OPA/DFG with a wide range of detuning conditions. A maximum gain of 15.0 dB and conversion efficiency of 14.7 dB were obtained. With a large detuning condition, the gain and conversion efficiency decreased because of the decrease in the phase-marching tolerance. It was confirmed that the idler generated through the wavelength conversion process had sufficient quality for data transmission. Although the SH pump power was compromised by using the M-QPM device to obtain a wide range of detuning, the gain and conversion efficiency could be further improved by using a PPLN device with an appropriate phase-matching wavelength for the SHG process. We believe that the results of this study enable optical transmission outside the 1.55 µm band and are useful for expanding the capacity of future optical communication systems.