We introduce a method, which is free of specimen edge effect, for the determination of critical stress in material brittle fracture. Using this method, we have investigated the critical stress in silicon brittle fracture. Silicon wafers polished with silica gel showed an average fracture stress of 2.8 GPa, with a standard deviation of 1.2 GPa and a maximum observed value of 6.9 GPa. We have also investigated effects of various surface treatments on the critical stress. Mechanically lapped surfaces, with damage penetrating deeper than ∼3 &mgr;m, reduced the fracture strength to 0.3–0.4 GPa (damage surface placed in tension). Both polysilicon and quartz overlays reduced the silicon fracture strength only very slightly, while argon implantation (at 150 keV, with a dose of 1×1016cm−2) did not show a detectable effect. However, anneal of the argon implanted samples at 900 °C in nitrogen for one hour significantly increased the silicon fracture strength, conceivably as a consequence of the removal of existing surface flaws by an epitaxial regrowth of the implantation‐amorphized surface layer. The last finding is both of basic interest and of practical significance. Silicon wafers of (100) orientation showed a ∼50% higher fracture strength than wafers of (111) orientation.