Optics & Photonics News

Optical mode [not to scale] n contact III-V Mesa n- InP n- InP/InGa AsP SL p- InGaAs p- InP cladding p- AlGaInAs SCH AlGaInAs MQWs Current Proton Proton Si substrate Buried oxide p contact
III-V
region
SOI
region

(Left) Schematic of hybrid silicon laser. (Right) Silicon wafers containing 2-mm-thick InGa AlAs layers bonded to SOI wafers.

In addition, SiO2 serves as a good waveguide cladding material and rare earth-doping host material. This allows silicon waveguides with propagation loss that is typically one order of magnitude lower than its compound semiconductor counterparts. It also has enabled the exciting demonstration of silicon Raman lasers. However, it wasn’t until recently that the challenge of electrical pumping was addressed.

Because of these benefits of silicon wafers, researchers have worked to epitaxially grow compound semiconductors on silicon. However, the large lattice and thermal mismatch between silicon and III-Vs typically results in a large dislocation density, which limits the laser efficiency and reliability. There have been advances in epitaxial techniques that have promise for new direct bandgap material systems and lattice matching to silicon, such as GaNAsP. Nevertheless, the mismatch of lattice constants and thermal expansion differences remain big challenges.

An exciting new result is the recent demonstration of the first germanium laser with room-temperature operation. However, its indirect bandgap has resulted in relatively low material gain, limiting operation to optically pumped structures; sufficient strain should make Ge direct bandgap with potentially higher gain.

Leveraging mature complementary metal-oxide semicon-
ductor (CMOS) technology is another primary motivation
for developing silicon-based lasers. The past half-century has
ushered in a CMOS technology revolution—in design tools,
circuit architecture, substrate manufacturing and epitaxy,
device processing, packaging, testing and quality control. These
advances pave the way to high-volume, low-cost manufacturing
once a good laser structure on silicon substrates can be devel-
oped. Good examples are planar lightwave circuits, which
emerged over the past two decades, and silicon Raman lasers,
which were developed recently. The latter may be useful for
generating laser wavelengths out into the far infrared and for
wavelength converter and amplifiers. However, since it requires
an external pump laser, the Raman laser is not practical for
interconnect applications.

Early on, we found
the thermal expansion
mismatch between
InP and silicon to
be the key hurdle
for transferring
InP epitaxial films
to silicon, since
conventional direct
wafer bonding
techniques were done
at high temperatures.

Invention of hybrid silicon lasers

Wafer bonding is widely used to make the silicon-on-insulator (SOI) wafers used in most silicon photonic devices. One of us (Bowers) had the idea to make hybrid silicon lasers by wafer bonding a crystalline III-V layer to an SOI substrate in much the same way that a silicon layer is bonded to a silicon wafer to make SOI substrates. The idea was to keep the optical mode in the SOI waveguide and amplify the evanescent tail with a forward biased gain region for a laser or amplifier.

Another of us (Mario Paniccia at Intel) had been developing many elements of photonic integrated circuits (PICs) on SOI but didn’t have an electrically pumped laser. Thus, a partnership was formed between Bowers’ group at the University of California, Santa Barbara (UCSB), and Paniccia’s at Intel.

A reverse biased III-V layer could also be used for modulators or photodetectors. Hence, it seemed to us that all of the necessary active elements needed for a photonic integrated circuit could be realized in this way. The first step was to obtain a high-quality, crystalline layer of III-V material on an SOI substrate.