1. Introduction

In recent years, the spread of communication services has been supported by increased optical communication capacity, which forms the backbone of transmission systems. The demand for an increased optical communication capacity continues to increase. Although increased capacity has been realized by improving the spectral efficiency using a multilevel modulation format, the spectral efficiency approaches the nonlinear Shannon limit [1]. Bandwidth expansion is a promising approach for increasing the capacity of optical communication. In current optical communication systems, a large capacity is realized by wavelength-division multiplexing (WDM) using an erbium-doped fiber amplifier (EDFA); therefore, the wavelength band used in long-haul optical communications is limited to the C/L band, which can be amplified using an EDFA. The amplification of wavelength bands other than the C/L band is desirable for bandwidth expansion. In contrast, highly efficient parametric interactions have been realized owing to recent advances in periodically poled LiNbO₃ (PPLN) waveguide technologies [2,3]. Several groups have reported optical parametric amplification (OPA) using PPLN waveguides [4–6]. In rare-earth-doped fiber amplifiers, the gain band is determined by the characteristics of the material; the advantage of PPLN-based OPA is that it is possible to amplify various wavelength bands by altering the phase-matching conditions.

In a previous study, we demonstrated OPA in the 1.3–1.8 µm range, which is a low-loss window for optical fiber, using two individual PPLN waveguides for second harmonic (SH) pump generation, and OPA [7]. In this study, we employed the 1.55 µm band pump, which has several advantages. It is also possible to use a high-power laser diode (LD) in 0.98 µm or 1.06 µm. However, no LD satisfies precise wavelength tunability, single mode operation, optical fiber output, and sufficient output power compared to the combination of 1.55 µm band LDs and EDFA. Furthermore, when using a 0.98 µm band or 1.06 µm LDs, the degenerate wavelength at which the signal and idler wavelengths match is approximately 2 µm, such that the wavelength of one of the signal and idler is longer than 2 µm [5]. Because the wavelength of one of the signals and the idler are outside the low-loss window of the silica fiber, such an output is unsuitable for fiber optic communication. This choice of pump wavelength narrows the range of fiber-optic communication applications. Conversely, the lowest loss wavelength of optical fiber is in the 1.55 µm band; by setting the degenerate wavelength to the 1.55 µm band, the signal/idler light can cover the low-loss window of silica fiber. For this design, a pump of approximately 0.775 µm was required. By changing the pumping wavelength by several nanometers, the gain band can be significantly changed from 1.3 to 1.8 µm [7].

Such a tunable 0.775 µm band pump can be obtained by combining an external cavity laser diode (ECL) and semiconductor optical amplifier (SOA). However, there are numerous issues, such as the stability of the coupling between the ECL and SOA by the spatial optical system, accuracy of the wavelength setting, and stability of the coupling to the optical fiber. With the current technology, such a combination is bulky and incompatible with optical communication equipment. Conversely, 1.55 µm band lasers are developed for WDM applications, and therefore commercially available tunable LDs have accurate wavelength tunability, single-mode operation, and optical fiber output delivery. Furthermore, by using an EDFA, it is possible to easily obtain the power level necessary to generate a pump through second-harmonic generation (SHG). Owing to recent improvements in the efficiency of PPLN, power loss during the SHG process has reached a level where there is no major concern.

In a previous study, the detuning between the phase-matching wavelengths of two PPLN waveguides was adjusted by changing the temperature to adjust the parametric gain band. Although this configuration enables the amplification of various wavelengths, a complicated procedure is required to set the operating conditions to obtain the desired gain band.

As an alternative approach, we propose a bidirectional configuration utilizing multiple quasi-phase-matched (M-QPM) LiNbO₃ waveguide for pump generation through SHG and OPA/difference frequency generation (DFG) processes [8]. In this configuration, the detuning between the phase-matching wavelengths for SHG and OPA/DFG was constant, enabling stable operation in the desired gain band. However, in the previous report, a three-peak M-QPM device was used; thus, the conversion efficiency was sacrificed because only two QPM peaks were used. In addition, the SH light was coupled to the optical fiber, reflected, and coupled to the LiNbO₃ waveguide, causing the SH power to experience a coupling loss between the waveguide and the fiber. Therefore, although we demonstrated the generation of an idler in our previous study, we did not achieve signal amplification.

To address these drawbacks, we devised a reflective module in which a LiNbO₃ waveguide with two QPM peaks and a built-in mirror for the SH pump were assembled. These features enabled us to generate an SH pump and efficiently use it for the OPA process in a single device. Using this module, we demonstrated OPA in the S-band in a compact configuration.

We confirmed that the idler light generated in the module retained phase information from the signal light. The signal quality was evaluated based on the phase transparency of the wavelength-conversion process.

2. Reflective multiple QPM LiNbO₃ waveguide module

We designed an M-QPM device with two phase-matching peaks to utilize different phase-matching conditions for pump generation via SHG and amplification/wavelength conversion using OPA/DFG. The M-QPM device is based on continuous-phase modulation of a periodic χ⁽²⁾ grating [9]. Figure 1 shows the phase modulation function and the calculated phase- matching curve of the device. The efficiency in the phase-matching curves was normalized by that of an unmodulated structure with the same interaction length. The phase modulation function was optimized to maximize the efficiency under the following phase-matching conditions:

Here, Δβ is the difference in the propagation constants of the interacting waves, Λ₀ is the poling period, and Λ_ph is the phase modulation period. Λ₀ was 17.4µm, Λ_ph was 32 mm, and waveguide length was 48 mm. The designed χ⁽²⁾ grating structure was fabricated on a Zn:LiNbO₃ substrate using electrical poling [10]. The substrate was processed into a ridge waveguide using direct bonding and dry etching techniques [11].

 

Fig. 1. (a) Phase modulation function and (b) calculated phase matching curve.

Download Full Size | PDF

The waveguide is assembled in a reflective fiber pigtail module. Figure 2(a) shows the schematic structure: a high-reflection (HR) coating was applied to the edge of the waveguide such that a 1.55 µm fundamental light could be transmitted and only the SH light was reflected. Two fiber pigtails were assembled to allow the input and output of the signal, idler, and pump. Four dichromatic mirrors were used to separate fundamental and SH waves. The SH-waves were output as a spatial beam through a window installed in the module. The waveguide temperature was adjusted using a Peltier device built into the module. The insertion loss between the two fiber pigtails was −2.7 dB at 1.49 µm.

 

Fig. 2. (a) Schematic structure and (b) evaluation of waveguide module of the multiple QPM device.

Download Full Size | PDF

This module was designed for the following purposes: A pump in the 1.55 µm band was input from the left fiber port as shown in Fig. 2(a). The pump was coupled to a LiNbO₃ waveguide through a lens and converted using SHG. The SH light is reflected by a HR mirror at the end of the waveguide and propagates through the waveguide in the opposite direction. The signal was input from the right port as shown in Fig. 2(a) and was coupled to the waveguide through a lens. The OPA/DFG process between the reflected SH pump and signal light amplified the signal and generated an idler. The reflected and leaked SH pumps were divided using a dichromatic mirror and the outputs from the windows in the module.

The phase-matching curve of the waveguide and characteristics of the HR mirror were evaluated as follows. First, the tunable pump light was input from the right-hand side port as shown in Fig. 2(b), and the SH light emitted in the same direction as the pump was measured. Next, the pump was input from the left-hand side port as shown in Fig. 2(b), and the SH light reflected at the edge of the waveguide and propagated through the waveguide was measured. In addition, the pump light was input from the left-hand side port, and the SH light converted in the waveguide and leaked through the mirror at the end of the waveguide was measured as shown in Fig. 2(b).

Figure 3 shows the phase-matching curves obtained using each method. The temperature of the waveguide was set to 36 °C and a tunable pump generated from an ECL was amplified using an EDFA and injected into the module through an isolator. The power input to the module was estimated to be 22.3 dBm. As shown in Fig. 3, two phase-matching peaks at approximately 1552 nm were observed for each measurement. When the pump is input from the right port, SH peak powers of 16.4 and 17.0 dBm were obtained at wavelengths of 1551.3 and 1551.8 nm, respectively. From the measured values, the normalized SHG efficiencies, including the coupling loss between the fiber and waveguide, were estimated to be 151%/W and 174%/W, respectively, at each peak.

 

Fig. 3. Measured phase-matching curves of the multiple QPM module.

Download Full Size | PDF

As shown in Fig. 3, in addition to the two main peaks near 1552 nm, side peaks were observed from 1549 nm to 1550 nm. These side peaks are larger than those of the calculated phase-matching curve shown in Fig. 1(b). Moreover, the phase-matching curve of the actual device showed a large asymmetry with respect to the wavelength compared to the calculated curve. The difference between the phase-matching curves of the actual and calculated devices is due to the non-uniformity in the thickness and width of the waveguide. If the uniformity of the waveguide is improved, the unwanted side peaks can be suppressed, and the conversion efficiency at the main peaks can be improved.

When the reflected SH was measured by inputting the pump from the left port, the SH was 5 dB lower than that when input from the right port. This lower intensity was due to the propagation loss of the SH light and reflection loss at the edge of the waveguide. When we measured the leaked SH by inputting the pump from the left port, the SH was three orders of magnitude smaller than the reflected SH, indicating that the HR mirror exhibited ideal reflection characteristics. A possible cause of the reflection loss is the imperfections in the angle of the facet, which needs to be improved in the future.

3. Evaluation of OPA/DFG using reflective M-QPM module

We evaluate the OPA/DFG performance and signal quality of the idler using a reflective M-QPM module. The experimental setup is shown in Fig. 4. An ECL generated 1.55 µm pump light, which was amplified with an EDFA and injected into the multiple QPM module through an isolator and circulator. Through the SHG process, the SH pump light is generated and reflected in the backward direction. In this process, a phase-matching condition corresponding to the shorter-wavelength QPM peak shown in Fig. 3 was utilized. A 1.50 µm-band signal was generated by another ECL and modulated in the format of 20 Gbit/s QPSK using an IQ modulator. The modulated signal was injected into the module using a circulator and propagated with the reflected SH pump. Through the OPA/DFG process between the signal and SH pump, the signal was amplified and 1.60 µm idler light was generated. In this process, a phase-matching condition corresponding to the longer-wavelength QPM peak, shown in Fig. 3, was utilized. In this configuration, two QPM conditions were used for the SHG and OPA/DFG processes. This allowed us to detune the SH-pump wavelength from the QPM condition for the OPA/DFG process and the parametric gain in the desired wavelength range using a single device. The amplified signal and idler were outputted using a circulator. The output intensity was monitored using an optical spectrum analyzer. Only the idler was extracted from the module output using a bandpass filter. After adjusting the optical power with an attenuator, the signal was input to a differential phase-shift keying (DPSK) receiver using an L-band EDFA as a preamplifier, and the bit error rate (BER) was measured using an error detector. For comparison, we measured the BER in a back-to-back configuration, using a distributed feedback laser diode with the same wavelength as the idler for the transmitter. The OPA demonstrated in this study utilizes the second-order nonlinear optical coefficient d₃₃ of LiNbO₃, which exhibits a polarization dependence, where the gain can be obtained only for polarization along the z-axis of the LiNbO₃ crystal. Because this study is a proof of concept using the reflective configuration, the experiment was conducted without eliminating polarization dependence. To eliminate the polarization dependence, it is necessary to perform polarization diversity. One possible method is to separate the orthogonally polarized components using PBS, amplify each using two individual OPAs, and combine them as described in [6].

 

Fig. 4. Experimental setup for the evaluation of OPA/DFG performance.

Download Full Size | PDF

An OPA/DFG experiment was conducted using several signal wavelengths to examine the dependence of the parametric gain on signal wavelength. Figure 5 shows the superimposed spectra of the signal and the idler. The black dotted lines in Fig. 5 represent the signal spectra without SH pump injection. The pump power input into the module was estimated to be 31.9 dBm. The pump wavelength was tuned to match the shorter-wavelength QPM peak and the SH pump power was maximized. The waveguide temperature was set to 40 °C and the reflected SH output power was 25.9 dBm.

 

Fig. 5. Spectra of the signal and idler when the pump was tuned to the QPM wavelength.

Download Full Size | PDF

Signal amplification was observed as shown in Fig. 5. We also observed conversion of the 1.49 µm band to the 1.62 µm-band idler. To the best of our knowledge, this is the first demonstration of an OPA using a single reflective device for both SHG and OPA. In the previous study, we reported DFG from the 1.4 µm to the 1.6 µm band using a 3-peak M-QPM module equipped with fiber for SH wavelength in a bidirectional manner [8]. We achieved conversion efficiency of −3.8 dB but amplification of the signal was not achieved. In contrast, in this study, a maximum gain of +9.0 dB and a conversion efficiency of +8.2 dB were achieved. This difference is due to both the improvement in waveguide efficiency owing to the reduction in the number of peaks in the M-QPM waveguide and the reduction in the loss of the SH pump by the elimination of coupling to an optical fiber.

Because of the uncertainty of the transmission loss of the waveguide, coupling loss between the fiber and waveguide, and reflection loss of the SH at the edge of the waveguide, it is difficult to estimate the theoretical gain from the measured values of the SHG conversion efficiency and insertion loss of the module. However, it would be useful to estimate the gain and DFG efficiency by assuming a nominal SHG conversion efficiency. If we neglect the SH pump depletion and transmission loss in the waveguide, the parametric gain for the signal and DFG efficiency are expressed as follows [12]:

Here, P_SH is the SH pump power, P_s is the input signal power, P_o is the output signal power, and P_i is the idler output power. The normalized gain coefficient γ is expressed in the following approximation, assuming a low DFG efficiency for an idler with a low SH pump power:

Here, η is the normalized DFG efficiency. We neglected the difference between the photon energies of the signal and idler to obtain a general idea. In this case, η can be approximated using the normalized SHG efficiency. The performance of a wavelength converter, such as a PPLN waveguide is evaluated by measuring the normalized SHG conversion efficiency under relatively low-power input conditions. As shown in Eqs. (2) and (3), the gain/conversion efficiency of the OPA/DFG is nonlinear with respect to the pump power. It is difficult to predict the gain directly from the normalized SHG efficiency. However, as shown in Eq. (4), the DFG conversion efficiency is proportional to the pump power under relatively low-power conditions; therefore, the parametric gain coefficient γ can be calculated from the normalized DFG efficiency, which can be approximated by the normalized SHG efficiency. If γ can be calculated, we can predict the gain and conversion efficiency, which will be useful for evaluating the validity of the OPA performance.

Figure 6 shows the calculated gain and DFG efficiency as functions of the SH pump power. If we neglect the transmission loss of the signal in the waveguide and assume that the coupling loss between the waveguide and fiber on both sides is the same, the normalized DFG efficiency of the waveguide is estimated to be 320%/W. Assuming this efficiency, an SH pump power of 29.6 dBm is required to obtain a signal gain of 9.0 dB. The actual reflected power was 25.9 dBm; therefore, assuming the above efficiency, the SH light waveguide loss was estimated to be 3.7 dB, and the mirror reflection loss was estimated to be 1.3 dB. Assuming that the normalized DFG efficiency of the waveguide was 500%/W, an SH pump power of 27.7 dBm was required to obtain a signal gain of 9.0 dB. Assuming this efficiency, the waveguide loss of the SH light was estimated to be 1.8 dB, and the reflection loss of the mirror was estimated to be 3.2 dB. As aforementioned, the estimation of the DFG efficiency varies depending on the values of the loss parameters, but it can be of the order of 320 to 500%/W. The SHG efficiency estimated from the data shown in Fig. 3 ignores the depletion of the fundamental light; therefore, it may have been underestimated. A separate SHG efficiency measurement of the bare waveguide before it was assembled in the module was 420%/W, which agrees well with the above estimate. As shown in Fig. 6, the gain increases exponentially by increasing the SH pump power and improving the waveguide efficiency. Thus, further improvements are expected in the future.

 

Fig. 6. Calculated gain as a function of SH pump power.

Download Full Size | PDF

Under the experimental conditions shown in Fig. 5, the maximum DFG efficiency was obtained when the idler wavelength was 1.62 µm, which is outside the gain band of the L-band EDFA, and it is not possible to evaluate BER as it is. Therefore, to shift the wavelength of the idler into the gain band of the L-band EDFA, we performed a similar experiment by detuning the pump wavelength by 0.2 nm from the phase-matching peak. Figure 7 shows the spectra obtained under these conditions. We observed the conversion of the 1.50 µm band to the 1.60 µm-band idler in this case. The peak signal gain was 1.5 dB, and the peak conversion efficiency was −2.6 dB. As the pump was detuned from the QPM peak where the maximum SH pump power was obtained, the parametric gain was reduced. Sufficient idler output was obtained even under these conditions, and experiments were conducted with modulated signals to evaluate the signal quality of the idler.

 

Fig. 7. Spectra of the signal and idler when the pump was detuned by 0.2 nm from the QPM wavelength.

Download Full Size | PDF

Figure 8 shows the spectra of the modulated signal and idler. The signal exhibited a sideband with a sinc-function shape owing to the QPSK modulation, and the idler also exhibited a similar sideband. This is evidence that the phase information of the signal is transferred to the idler through the DFG process. Figure 9 shows the BER performance of the idler as a function of the received power. No power penalty was observed in the BER characteristics of the idler compared with that of the back-to-back measurement. In the reflective configuration used in this study, there is a concern that the signal waveforms may deteriorate owing to unwanted reflections inside the module. From the results of the BER measurements, we conclude that the idler has a signal quality comparable to that of the signal. The theoretical noise figure of a non-degenerate OPA is 3 dB; in practice, it is necessary to consider the coupling loss between the fiber and waveguide and the propagation loss of the waveguide. Evaluation of the noise figure requires precise measurement, which will be a challenge in the future [13].

 

Fig. 8. Spectra of (a) the modulated signal and (b) idler.

Download Full Size | PDF

 

Fig. 9. BER performance.

Download Full Size | PDF

In this study, the maximum conversion efficiency was obtained at an idler wavelength of 1.62 µm; however, there is no suitable preamplifier for this wavelength at present. Because of the lack of a preamplifier, we changed the idler wavelength for the BER measurements. However, this does not imply that OPA/DFG at 1.62 µm has no practical significance even if there is no suitable preamplifier at present.

The problem of not having a preamplifier for 1.62 µm can be resolved in the future if the gain of OPA becomes sufficient for use as a preamplifier. Furthermore, it is possible to use OPA/DFG to convert a wavelength of 1.62 µm to a wavelength, such as C-band or S-band, where mature preamplifiers are available. Such use of OPAs/DFGs will be useful for expanding the bandwidth and improving the flexibility of optical communication systems.

To obtain a separate signal and idler from the degenerate wavelength in the χ⁽²⁾ based OPA, it is necessary to slightly detune the SH pump from half the degenerate wavelength to a shorter wavelength. We detuned the SH pump using an M-QPM device. Although we demonstrated OPA for the 1.48 µm band signal and 1.62 µm band idler in this study, a different gain band can be realized by changing the peak spacing, hence the phase modulation period of the M-QPM device and there is no need to change the waveguide design. In particular, the advantage of our approach is that the interaction between the signal and idler over a wide wavelength range can be obtained by making slight changes to the χ⁽²⁾ grating design. In addition, the wavelength of the pump needs to be changed only slightly, and existing optical communication components, such as an EDFA can be used.

To obtain similar changes in spacing between the signal and idler wavelengths in OPA based on χ⁽³⁾ media, such as Si, SiN₄ waveguides, or highly nonlinear fibers without changing the wavelength of the pump, precise control of the waveguide dispersion is required to obtain phase matching [13–16]. A specific waveguide design is required to amplify a given wavelength. It is also possible to change the signal/idler wavelength by changing the pump wavelength; however, in that case, we require a pump with a nonconventional wavelength that cannot be obtained with an EDFA.

The configuration developed in this study generates a pump in the 0.775 µm band within the LiNbO₃ waveguide using optical communication components, such that it is potentially compatible with optical communication systems.

4. Conclusions

We demonstrate an OPA using a reflective multiple-QPM LiNbO₃ waveguide module. Two-peak QPM devices and reflection mirrors were used at the edge of the waveguide to enable pump generation via the SHG process and amplification/wavelength conversion via the OPA/DFG process in a single device. We demonstrated amplification of a 1.48 µm signal and conversion to a 1.63 µm idler. Phase-transparent conversion of the signal with QPSK modulation was confirmed. BER measurements of the idler confirmed that the idler preserved the phase information of the signal.