Origin Of Growth Twins During Czochralski Growth Of Heavily Doped, Dislocation-Free Single Crystal Silicon

Low voltage power electronics are made from dislocation free silicon heavily doped with arsenic or antimony to provide low electrical resistivity. Attempts to grow crystals with decreased resistivity have led to a higher probability of twinning during growth, so that the crystal no longer possesses the required crystallographic orientation for device fabrication. The source of the twins must be identified so that crystal growth process conditions can be designed to eliminate this defect mechanism, allowing lower resistivity crystals to be grown reliably. In lightly doped crystals, twinning was ascribed to presence of carbon impurity or a low probability atomic stacking accident, neither of which should be affected by increased concentration of arsenic or antimony. Crystals that twinned during growth were characterized by resistivity, Laue back-reflection x-ray diffraction, optical and scanning electron microscopy, energy dispersive x-ray spectroscopy, spreading resistance, x-ray computed tomography and electron backscatter diffraction. The twin nucleation site of silicon crystals that were grown heavily doped with arsenic or antimony were compared to lightly doped crystals which twinned, and crystals that exhibited other defects. The initial twinning in the \u3c100\u3e orientation heavily doped crystals occurred from small gas bubbles bursting at a {111} facet at the three phase boundary, and forming a twin orientation domain on that facet. The gas bubbles likely consist of argon, the process gas used during solidification to remove silicon monoxide gas from the growth system. The higher levels of arsenic or antimony dopant may have changed the silicon surface tension, or provided additional impurities into the liquid silicon. Either effect may have changed the number or size of argon bubbles in the liquid silicon, leading to a higher incidence of gas bubbles near the {111} facet during solidification. Similar but smaller crater features were observed on two lightly boron-doped silicon crystals that twinned. Two other lightly doped crystals formed twins from carbon inclusions, consistent with carbon as a cause. Some heavily-doped twinned samples also show high concentrations of metals at the twin nucleation site, which could affect surface energy. Measurement of the geometry of crystal surface-to-facet radius eliminated a recently-proposed twin nucleation theory from consideration. Constitutional supercooling was demonstrated to not be a major contributing factor to twin nucleation. It was shown that deliberately introducing additional arsenic dopant during solidification would nucleate twins, but twins did not occur if only elemental carbon was introduced