Search

In the framework of High Performance Computing and Datacom, silicon photonics interposers propose an interesting approach, while providing new challenges. This paper demonstrates such an integration and focuses on TSV Mid integration impact on sensitive photonic structures such as ring modulators, focusing on two specific technological aspects: substrate thinning and TSV integration. It is shown that thinning down to 100 microns and integrating TSV do not impact photonic performances more than wafer level process variability. Finally, thanks to these results, 3D Si photonics design opportunities will be discussed.

Performance degradations of Silicon Photonics (SP) high speed photodetector represent a major issue for the reliability of these devices. An explanation of these degradations is presented based on both electrical characterization and device modelling. The observed degradations of both dark current and responsivity can indeed be modeled by a single carrier lifetime degradation, attributed to an increase of the surface recombination rate, impacting an unexpected large contribution of diffusion in the photocurrent. The results obtained with this model are experimentally confirmed by extracting the activation energy of the dark current, before and after stress. The improved physical understanding of the degradation is expected to lead to shorter test protocols for SP devices.

In this paper, the reliability of germanium photodiodes of the PIC25G technology for silicon photonic applications is experimentally studied. Using advanced characterization technics, it is shown that the dark current, the photonic current and the cut-off frequency of the photodiode can be degraded during voltage stress of 106 s, which could ultimately induce some device performance drift. The causes of these degradations are presently attributed to interface defects between germanium and SiO 2 , until more detailed investigation are pursued.

International audience; Silicon photonics technology emerged as a promising solution to address the technical challenges related to 100 Gb/s and 400 Gb/s optical link. Enabling the development of silicon photonics products requires the development of optical passive libraries integrated within conventional CAD tools used in the CMOS design flow. The optimization and modeling of silicon photonics optical passive is therefore a key point that can be addressed by leveraging methodologies that have been previously set up for optimizing RF passive in CMOS and BiCMOS technologies. In this paper, the relevance of such an approach is evaluated: the combination of FDTD electromagnetic simulations and a Design Of Experiments (DOE) prototyping have been used for optimizing scalable Grating Couplers (GCs) targeting Wavelength-Division Multiplexing applications (WDM). The obtained models for the GC have been successfully qualified experimentally.

One of the technological challenges of direct observation of extra-solar planets remains to develop a modal filter operating from 6 to 20 μm. In the present paper we present one of the two candidate technologies for the fabrication of such modal filters, together with Fiber Optics: Integrated Optics. The solution based on alltelluride buried channel waveguides was considered. In the so-called waveguides, the vertical confinement of the light is achieved in a higher-index core layer consisting in a 15 μm-thick Ge 17 Te 83 film which is deposited onto a lower-index Te 75 Ge 15 Ga 10 substrate, and further covered by a lower-index superstrate consisting in a 15 μm-thick Ge 24 Te 76 film. Concerning the horizontal confinement of the light, it is obtained by etching the core layer. As this stage, the all-telluride buried channel waveguides are proved to transmit light from 2 to 5 μm and to behave as channel waveguides with a satisfactory confinement of the light. It confirms the validity of the technology and the good quality of the input and output facets. Tests between 6 and 20 μm are being performed and will allow giving a confirmation of the potential of Te-based integrated optics components for nulling interferometry.

Summary form only given. In numerous laser and optical amplifier engineering optimization problems, the prime design criteria are the device output power and the overall gain. The amount of pump power available is usually limited and fibre parameters such as core dimensions and numerical aperture are set for achieving minimum coupling losses. With this respect we developed optical amplifier design curves that could summarize in one plot existing relationships of the most pertinent design parameters. The model is based on [1]. These diagrams show the required active medium length as a function of both Ytterbium and Erbium concentrations for achieving a given output power. Such a representation provides a quick, visual recognition of the optimal length, as can be observed in figure 1a). No unique optimal solution exists but instead several length-concentrations triplets. The generated “maps” allow a clear identification of the different operating regimes of the amplifier. For instance, it can be observed that at low Er concentrations no particular advantage is obtained if Yb concentration is increased. These conclusions are not quite obvious when using other representations. Regarding the selected concentration ranges, higher values are limited such that clustering effects be limited. Lower bounds are chosen in order to prevent obtaining excessively long devices. We used this approach for designing a core pumped fiber amplifier based on Yb3+-Er3+ co-doped phosphate glass which was then developed in house. The pump and signal powers were chosen to be Pp =166 mW and Ps = 3.16 mW (+5 dBm). The desired output power was Pout = 45 mW. Both Ytterbium and Erbium concentrations have been selected to be 2.5×1026 ion/m3, thus leading to a required length of 27 mm for the active medium, as indicated by a cross point in figure 1a). Single mode laser diodes working at 975.6 nm (pump) and 1535 nm (signal) were spliced to the inputs of a WDM combiner. Its output was butt-coupled to the active fiber. A photograph of the in-house fabricated amplifier section is shown in figure 1b). The optical output power at 1535 nm was 13.8 mW. The difference with the expected value can be partly related to the presence of green emission, attributed to an up-conversion process. In terms of gain, the amplifier provided 6.39 dB for an input signal of +5 dBm. The gain per unit length was thus 2.37 dB/cm, which is higher than or near to values reported in [2,3] (0.94 dB/cm and 3.3 dB/cm). In addition, a gain of 11 dB (with a -30dBm signal) was observed when the pump power was increased to Pp = 479 mW, which leads to a 4.07 dB/cm gain, a value that is comparable to [4,5] (4.2 dB/cm and 5.2 dB/cm). Later simulations showed that at this pump level a notable increase in output power can be obtained if 10 to 20 mm longer fibers are used.

Summary form only given. The development of WDM)telecommunications systems has driven the rapid development of erbium doped fiber amplifier and very narrow band lasers. Recently, new efficient waveguide amplifier on phosphate glass has been demonstrated. In this paper, we present the first steps of fabrication and analysis of 10 monolithically integrated DFB lasers on a chip. The starting point of our work is an Er/Yb codoped phosphate glass, on which buried channel waveguides have been realized using ion exchanged techniques. Recent progress in that technology has led to a gain of 3dB/cm. The waveguides present a mean index variation of 2.10/sup -2/ above substrate refractive index value of 1.52. For operation at 1.55 /spl mu/m, the grating period predicted by the Bragg phase matching condition is d = /spl lambda//2n/sub e/ = 0.505 /spl mu/m related to the mean effective index n/sub e/ of the waveguides. In this work, a corrugate 1 cm long grating was made on the top of the wafer. Interferometric exposure of photoresist with an Ar/sup +/ laser operating at 488 nm was used to print the grating. After development, the grating pattern is transferred on the glass surface through a reactive ion etching process. The glass is etched sufficiently to get more than 60% of reflectivity at Bragg wavelength. Under pumping at 980 nm, single longitudinal mode operation around 1.55 /spl mu/m is demonstrated for all the ten waveguides of the chip.