Introduction Over the past several decades there has been growing interest in the development of devices for optical wavelength conversion. In Raman converters, a long-wavelength Stokes input is in the presence of a pump transformed into a short-wavelength anti-Stokes output via a Raman-resonant third-order nonlinear process called coherent anti-Stokes Raman scattering (CARS) [1]. In Kerr-based parametric converters, a signal input is in the presence of a pump converted to an idler output via the third-order nonlinear process of non-resonant four-wave-mixing (FWM) [2]. Both types of conversion can occur in silicon waveguides, and as such have great potential for application in optical communication systems.
Introduction Over the past several decades there has been growing interest in the development of devices for optical wavelength conversion. In Raman converters, a long-wavelength Stokes input is in the presence of a pump transformed into a short-wavelength anti-Stokes output via a Raman-resonant third-order nonlinear process called coherent anti-Stokes Raman scattering (CARS) [1]. In Kerr-based parametric converters, a signal input is in the presence of a pump converted to an idler output via the third-order nonlinear process of non-resonant four-wave-mixing (FWM) [2]. Both types of conversion can occur in silicon waveguides, and as such have great potential for application in optical communication systems.
The possibility to perform broad wavelength generation and conversion is of great interest in classical and quantum optics. With the introduction of integrated nonlinear silicon photonics, these tasks have been made even more viable thanks to the combination of the high nonlinear refractive index of silicon with the high confinement provided by the photonic waveguides. Exploiting this platform, several solutions for wavelength conversion have been investigated, for both optical processing [1] and applications in the Mid Infrared (MIR) [2]. The possibility to perform broad and controllable wavelength generation can be exploited also for creating integrated sources of quantum states of light based on single photons [3] or correlated photon pairs [4]. Both the generation and conversion can be accomplished by means of Four Wave Mixing (FWM). FWM is a nonlinear optical process in which two pump photons at frequency are converted into signal and idler photons at frequencies and ω/ respectively (with ωI/ωS). In the particular case of wavelength conversion, FWM is stimulated by a seed wave at the same wavelength of the signal (idler) that will be converted to the wavelength of the idler (signal). FWM has to satisfy the energy conservation, according to which the relation ωP+ ωP= ωS+ ωIholds, and its efficiency is ruled by the phase matching condition. This latter condition is related to the phase mismatch among the waves involved in the FWM process, that is evaluated as Δk = kp+ kp−kS−kI, where kP, kS, kIare the wavevectors for the pump, the signal and the idler, respectively. When Δk = 0, the phase matching condition is fulfilled and the efficiency of FWM is maximized. What is important for wavelength generation and conversion is the spectral position of the phase matching condition, which determines the wavelength at which the maximum efficiency of the conversion or generation process is achieved. Because of this, techniques for the control of the phase matching condition have been developed, focusing on the possibility to tailor the Group Velocity Dispersion (GVD) through a proper engineering of the waveguide geometry [5]. While these techniques consider only the first order waveguide mode, here we propose to exploit the higher order modes propagating inside a multimode waveguide. The dispersion of the higher order modes can be used to tune the spectral position of the phase matching condition for the FWM process, achieving controllable and large spectral translation in both Stimulated Four Wave Mixing (sFWM) and Spontaneous Four Wave Mixing (SFWM).
We propose an efficient wavelength conversion scheme that is based on either Raman-resonant four-wave mixing or non-resonant Kerr-induced four-wave mixing in a silicon nanowire ring, and that does not require dispersion engineering of the nanowire. We rely on the spatial variation of the Raman and Kerr susceptibilities around the ring to quasi-phase match the wavelength conversion processes for TE polarized fields. The flexibility of this quasi-phase-matching scheme can lead to wavelength conversion efficiencies from -26.7 dB to values larger than 0 dB, and in certain circumstances makes it possible to outperform conventional phase-matched conversion in a dispersion-engineered silicon ring by factors exceeding 6 dB.
This thesis describes the implementation of the all-optical wavelength converters based on quasi-phase-matched (QPM) lithium niobate (LiNbO3) crystals and some related optical nonlinear materials for optical fiber communication systems and other applications. Such devices would be the enabling technology to realize high-capacity and transparent communication networks required in wavelength-division-multiplexing systems and highly efficient laser sources at desired wavelengths. The wavelength converters discussed in this thesis are based on the three-wave mixing processes attributed to the second-order nonlinear interactions (x2,). The main challenge with a LiNbO3 QPM wavelength converter is to implement a highly efficient device, which is capable of operating at a power of hundreds of mW. For this purpose, several wavelength conversion schemes have been demonstrated and theoretically investigated, including the fiber ring resonators and a novel cascaded χ(2).∖ The former based on the conventional cascaded x(2) contains erbium-doped fiber amplifiers and other passive components exclusively as pump sources (no additional pump source is required) with excellent characteristics such as high black-box efficiency. While the encouraging properties of the latter are those pump wavelengths can be set outside the communication window, and same efficiency can be achieved by employing two pump sources with lower output power as compared with the conventional cascaded one. By taking into account the higher-order dispersion effects, pulse propagation in QPM second harmonic generation waveguides is also analyzed. In addition, the photorefractive effect (PRE) in QPM waveguides is systematically studied by using a pump-probe method, since the LiNbO3 crystals are extremely vulnerable to high-power shorter-wavelength light illumination. Both temperature and wavelength dependent results are helpful in further understanding the physical insight of the PRE, and providing a guideline for practical application of the QPM LiNbO3 waveguides.
Silicon photonics has drawn a lot of attention over the last decades, mainly in telecom-related application fields where the nonlinear optical properties of silicon are ignored or minimized. However, silicon’s high χ(3) Kerr optical nonlinearity in sub-micron-scale high-confinement waveguides can enable significant improvements in traditional nonlinear devices, such as for wavelength conversion, and also enable some device applications in quantum optics or for quantum key distribution. In order to establish the viability of silicon photonics in practical applications, some big challenges are to improve the optical performance (e.g., optimize nonlinearity or minimize loss) and integration of optics with microelectronics. In this context, we discuss how electronic PIN diodes improve the performance of wavelength conversion in a microring resonator based four-wave mixing device, which achieves a continuous-wave four-wave mixing conversion efficiency of −21.3 dB at 100 mW pump power, with enough bandwidth for the wavelength conversion of a 10 Gbps signal. In the regime of quantum optics, we describe a coupled microring device that can serve as a tunable source of entangled photon pairs at telecommunications wavelengths, operating at room temperature with a low pump power requirement. By controlling either the optical pump wavelength, or the chip temperature, we show that the output bi-photon spectrum can be varied, with implications on the degree of frequency correlation of the generated quantum state.
Parametric up- and down-conversion require phase matching between the generating and generated wavelengths, introducing strict requirements on the refractive index dispersion of materials used for this nonlinear processes. Optical microfibers can exploit the different modal overlap with core and cladding materials to phase match different modes at the required wavelengths. Intermodal phase-matching has been successfully exploited in optical microfibres to generate second- and third-harmonics, both in straight waveguides and related resonators. The simultaneous generation of guided higher harmonics also allowed to exploit other nonlinear effects like four wave mixing to generate light at short wavelengths in the UV part of the spectrum. Parametric down-conversion allows to generate entangled photons, but with efficiencies significantly smaller than their up-conversion processes because of their reliance of spontaneous photons generation from vacuum. The use of more complex architectures, based on multiple coupled waveguides, allows to achieve a quasi-phase matching condition, thus a sizeable improvement on the overall conversion efficiency.
In this talk, I will discuss two applications of nonlinear quantum photonics in silicon nanowires: correlated photon pair sources and a single-photon wavelength converter. Correlated photon pairs can be generated in a nonlinear silicon waveguide through the spontaneous four-wave mixing (SFWM) process, where two photons from a bright pump laser are annihilated, and a two photons are spontaneously created at different wavelengths (the signal and idler). However, indistinguishable photon pairs are of particular interest for two-photon interference experiments, meaning that the signal and idler should have the same wavelength and be entirely identical to each other. Indistinguishable pairs can be generated with the SFWM process. where two pump lasers create a degenerate signal and idler at the frequency mid-way between them. Since the two photons are indistinguishable, it is not trivial to divide them into separate waveguides so as to manipulate them individually. I will present recent results concerning a silicon quantum splitter, where degenerate pairs are created in a ring resonator which is pumped in both directions. The ring is coupled to a sagnac loop interferometer, which combines the two directions at a 50 : 50 coupler. Here, a quantum interference effect (reverse-Hong Ou Mandel interference) splits the photons to two output waveguides. This device is ultra-compact and the use of a sagnac loop means that the phase is automatically set correctly for the interference. I will also discuss progress on single photon wavelength conversion in silicon nanowires using the four-wave mixing Bragg scattering (FWMBS) process. Here, two intense pump beams are used to modulate a weak signal, causing it to shift to the idler by the frequency difference between the pumps. In principle, this can allow coherent, noiseless frequency conversion of single photon signals. Fig. 1(b) shows output spectra from a 6mm silicon nanowire when the frequency shift is varied 100 GHz to 400 GHz. Multiple peaks are visible, corresponding to up-conversion and down-conversion of the input signal.