Quantum dots (QDs)—nano-sized crystals that can emit light—can improve the performance of semiconductor lasers. The three-dimensional charge carrier confinement in QDs decreases a laser's sensitivity to temperature and lowers its operating current threshold. Realizing the full benefit of QD lasers requires integrating them with other devices on the same substrate.
However, conventional methods for fabricating integrated photonic devices may demand an arduous level of fabricating control, or may rely on active-region defects that can degrade semiconductor layer quality. Typically, layers with contrasting electronic properties are grown in regions formed by etching away existing semiconductor layers. The efficient connection of layers with different band gaps requires precise control of etch depths and regrowth thicknesses. Also, additional band-gap tuning after growth generally introduces defects into the active region.
To avoid these problems, we have been developing selective-area epitaxy (SAE), a new band-gap tuning technique that uses patterned, or masked, substrates to facilitate QD growth. The ‘selective areas’ are openings in the mask. In SAE, a semiconductor substrate is patterned with a layer of dielectric, such as silicon dioxide (SiO2), prior to growth. Under certain deposition conditions, epitaxial growth is inhibited on the dielectric surface, while adatoms migrate into the openings, where enhanced growth rates occur. The band-gap depends on epitaxial layer thickness and composition, which are controlled by local growth parameters that are tuned by varying the mask dimensions.
By carefully optimizing the growth parameters, SAE can facilitate controlled QD nucleation.1 The mask patterns in our work were pairs of SiO2 stripes of various widths separated by a 50μm-wide region. Figure 1 shows atomic force microscopy images taken in different regions of a patterned gallium-arsenide (GaAs) substrate on which indium gallium arsenide (InGaAs) was deposited. Figure 1(a) shows a region without patterning. The InGaAs deposited in this region didn't reach the critical thickness needed for QD nucleation. As a result, the growth proceeded in a two-dimensional, layer-by-layer mode that formed quantum wells (QWs). In contrast, Figure 1(b) shows QDs in a region where 50μm-wide SiO2 stripes were used. Here, an enhanced growth rate caused InGaAs deposition to exceed the critical thickness required for nucleation, so that QDs grew on the substrate.
Figure 1. Images showing controlled quantum dot (QD) nucleation from selective-area epitaxy. (a) New layer growth in this unpatterned region is too thin to nucleate QDs. (b) The enhanced growth rate in this patterned region between 50μm-wide SiO2 stripes leads to nucleation.
The QD band-gap energy can be tuned by varying QD size and indium content, which are themselves influenced by dielectric stripe width. A mask that varies in width would therefore yield QDs on the same substrate that have different band-gap energies. Figure 2 shows that band-gap energy for the 50μm-offset stripes decreases with increasing stripe width up to about 35μm, then generally levels off up to 70μm. The average QD size changes because of growth-rate variations, and the indium enrichment increases with increasing width because of the differences between indium and gallium adatom mobilities. The dependence of the QD size on SiO2 stripe width is complex. For a fixed separation between the stripes, the dot size first increases, and then decreases, with increasing stripe width.
Figure 2. The band-gap of selectively grown QDs varies with the width of the SiO2 patterning stripes.
The growth schemes described above can fabricate a QD laser integrated with a passive (QW) waveguide and QD lasers of various wavelengths. We fabricated a QD laser integrated with a passive waveguide using five layers of stacked QDs in the active region of the laser. The results are very promising, and include losses as low as 4cm-1 in the waveguide region. The low losses arise from a large differential band-gap of 140meV between the QDs in the laser's active region and the QWs in the waveguide.
SAE is a promising technology for fabricating the QD-based integrated devices of photonic integrated circuits. So far, our results demonstrate that SAE can selectively control the nucleation and band-gaps of self-assembled QDs. Having demonstrated the integration of a QD laser with a low-loss passive waveguide, our future efforts will concentrate on integrating multiple-wavelength QD lasers.