Crystals consist of regularly ordered rows of atoms. In general, when two different crystals are joined, there are problems. For one, they may be ordered differently. Then, at the interface between the crystals, the atoms may not match up properly and some atoms may be left with incomplete bonds. For semiconductor crystals, this causes serious problems. Semiconductor devices depend on the very subtle control of these crystal's electrical properties. Incompletely bonded atoms upset these properties. OK, so why even try to combine two different semiconductor crystals? Because the combined use of different crystals allows us to fabricate many faster, more powerful devices.

How does one overcome this problem of imperfect interfaces between crystals? There had been one standard solution: use two crystals that, although made of different atoms, just happen to have exactly the same atomic arrangement and spacing. A few such combinations exist. A couple of these form the basis for semiconductor light emitting diodes and lasers.

No such fortuitous partner exists for Si, the most technologically significant semiconductor. Ge is chemically compatible, but its atoms are larger and thus spaced farther apart. If Ge, or GeSi alloy crystal is grown on top of Si, one expects the configuration shown at the upper left. Both layers are free of strain, but at the interface many atoms have 3 rather than the standard 4 crystal bonds. Those atoms with 3 bonds ruin any electrical device fabricated near this interface. In the early 1980's we overcame this problem. By using our very clean and well-controlled MBE growth techniques, we showed how GeSi could be grown at unexpectedly low temperatures. At these temperatures, in contradiction to earlier theories and experiments, we showed that the GeSi overlayers would compress in order to achieve perfect interfacial bonding. This is depicted at the upper right and now goes by the general name "strained layer epitaxy." This breakthrough provided the basis for the devices described in subsequent images, for most of my fourteen patents, and for commercial devices such as those now going into commercial production at IBM and elsewhere.