One of the most powerful techniques in fiber optical communication systems is wavelength division multiplexing (WDM). By utilizing the large (~300 nm), low-loss (0.2 - 0.4 dB/km) transmission bandwidth, a single fiber can transmit many wavelengths. One fiber can potentially support transmission of tens of terabits per second of information over thousands of kilometers, to meet the exponentially-growing capacity demand. One of the key components for WDM systems is the optical amplifier; currently the most widely used optical amplifier is the erbium-doped fiber amplifier (EDFA). However, its bandwidth and operating wavelength are limited. For example, the conventional EDFA (C-band) can amplify signals, roughly between 1530 nm and 1562 nm. Even with the longer wavelength window (L-band), it can extend only up to 1605 nm, that is, 75 nm in total. To mitigate the bandwidth limitation of EDFAs, alternative optical amplifiers have been investigated, and one of the most promising candidates is the fiber optical parametric amplifier (OPA).

Fiber OPAs are based on the third-order nonlinear susceptibility c⁽³⁾ in fiber. When an optical signal at angular frequency w_s co-propagates in a fiber with a strong pump at w_p, the signal is amplified through a parametric process. OPAs can exhibit large bandwidth, and may find applications as optical amplifiers for WDM transmission. Another wavelength, called the idler, is generated at w_i = 2w_p - w_s. The idler contains the same modulation information as the input signal, with an inverted spectrum. This phase-conjugated idler can be used not only for wavelength conversion in WDM networks, but also for mid-span spectral inversion (MSSI) which can combat fiber dispersion, and even some of the detrimental fiber nonlinearities such as self-phase modulation (SPM), cross-phase modulation (XPM) or stimulated Raman scattering (SRS).

In order to use an OPA as a practical optical amplifier, key issues must be addressed, namely, bandwidth, signal gain, noise figure, pump efficiency, and polarization sensitivity. Prior to this work, the record OPA performance reported in the literature was 200 nm bandwidth, 49 dB gain and 4.2 dB noise figure in three separate experiments. We have proceeded further by improving OPA design, and with better knowledge of the dispersion profile of the gain medium in OPA itself, we were able to achieve 360 nm bandwidth. In another experiment, we have achieved a record OPA continuous-wave (CW) gain of 60 dB, by using a novel two-segment design. We have also demonstrated the first low-noise-figure (sub-4 dB) optical parametric wavelength converter with high conversion gain (40 dB).

In a well-designed OPA, most of the pump energy should be transferred to the signal and the idler. Hence a measure of the efficiency of such a device is the pump depletion that is obtained. We have been able to achieve 92% pump depletion, with a 23 dBm CW pump in an 11 km dispersion-shifted fiber (DSF) - the highest depletion ratio ever demonstrated in fiber devices. Furthermore, OPAs have not been practical as they were not polarization-independent in their simplest form. In our work, polarization-independent OPAs, in both one-pump and two-pump configurations, were proposed, investigated and demonstrated. With a 1 km highly-nonlinear dispersion-shifted fiber (HNL-DSF) and a 25 dBm CW pump, a fiber OPA with gain of 15 +/- 0.5 dB over a 20 nm range has been demonstrated.

In addition to using OPA as an amplifier or a wavelength converter, new and novel applications of OPA have been investigated in this work. For example, we demonstrated the first CW fiber optical parametric oscillator (OPO). It uses the OPA as a gain medium inside an optical cavity. Here, we achieved an internal conversion efficiency of 30%, compared to the maximum theoretical value of 50%, by using a 1 km HNL-DSF and two fiber Bragg gratings (FBGs). Pump power threshold was 24 dBm; output wavelength could be tuned over 80 nm by tuning the pump.

In order to deploy OPA in mid-span spectral inversion (MSSI), the generated idler must not be spectrally broadened, which is common in one-pump OPAs, due to the required pump dithering for stimulated Brillouin scattering (SBS) suppression. We have demonstrated a novel all-optical technique to cancel the idler broadening by using two pumps phase-modulated 180± out of phase, one of them being obtained by four-wave mixing in an auxiliary fiber. The resulting OPA idler quality is comparable to that of the output signal.