异构集成 SOI化合物半导体光子学

Silicon has been shown to be an excellent material for  realising compact photonic integrated circuits (PIC)  operating in the telecom wavelength range (1100nm- 1600nm). Crystalline silicon has very low intrinsic  propagation losses and both large core fiber matched  waveguides and extremely compact so-called  photonic wire circuits have been demonstrated.  Through carrier injection or depletion the real and/or  imaginary part of the refractive index can be changed,  allowing for switching and modulation of light. Since  silicon is intrinsically transparent at these  wavelengths, integration with other materials is  required however for realizing efficient optical sources  and detectors. The latter are now typically based on  Ge, integrated with the silicon waveguides using  heteroepitaxial growth techniques. Although being  investigated intensively, these have had only limited  success thus far for integrating compound  semiconductors on silicon however, which is required  for realizing compact light emitters and amplifiers.  Therefore, several groups developed an alternative  integration scheme, based on waferbonding  technologies, which avoids the growth step but still  allows for a process compatible with waferscale  processing techniques. The latter is not only  important from a cost perspective, but it also allows  for accurate alignment of the active devices with  respect to the underlying circuits. This paper reviews  work of different groups in this field. In the following  section the integration approach is described in more  detail. Next, an overview of several fabricated  devices is given.

A classic approach in integrating III-V semiconductor  devices with silicon (and silica) PICs is based on flipchip bonding. In that case prefabricated optoelectronic devices (e.g. a DFB laser array) are  bonded upside down on a suitable substrate (e.g. a  recess in the PIC) . In some cases the substrate is  largely removed before bonding . No or little  comprises are required in the design of the  optoelectronic devices and they can be pretested.  Therefore, for small to medium volumes this is  certainly an attractive route, although reaching the  required placement accuracy (typically +/- 1um) is far  from trivial and the interface can exhibit considerable  loss. For large scale integration however, there is a  need for an integration approach compatible with  waferscale technologies as described above.

Fig1

The first electrically injected III-V on silicon lasers  were demonstrated in 2006. Fabry-Pérot  lasers were demonstrated using an approach  whereby most of the light is concentrated in the III-V  area of the laser section and coupled to the silicon  waveguide through a polymer taper once exciting the  laser facet. In  a hybrid integration platform was  demonstrated whereby most of the light is  concentrated in the silicon waveguide and the mode  is evanescently coupled to the waveguide region on  top. This platform was rapidly extended to more  complex laser designs. DFB-lasers as well as DBRlasers were demonstrated, with output powers of  respectively over 4mW and over 9mW. In  combination with quantum well intermixing  (prebonding) for defining regions with different band  gap also integrated sampled grating DBR-EAMs were  shown. See  for a recent review.

For short reach interconnect applications and alloptical signal processing however low power laser  sources are required. Therefore we developed  electrically injected microdisk sources, employing a  whispering gallery mode in a III-V semiconductor disk  structure. Initial devices had a threshold current of  0.5mA and where limited in output power . In  recent devices, having an improved heat sink, the  threshold current was reduced to 0.260mA and the  output power (CW operation) increased to 0.120mW . These devices can also be used as electro-optic  modulators , all-optical wavelength convertors  or as part of a compact multi-wavelength laser .

Heterogeneous integration of compound  semiconductors has proven to be a very powerful  platform for realising efficient sources integrated with  silicon waveguides. Further improvement will require  new device structure, e.g. using photonic crystal  structures for improved light confinement  or new  material types. Promising candidates are selfassembled materials such as colloidal quantum dots   or compound semiconductor nanowires .