I have got the same problem for CNT with tersoff_1989
you should edit the SiC.tersoff potential file manually and make a rigid cut-off instead of smooth cut-off.
there are several variables in that file including : m, gamma, lambda3, c, d, costheta0, n, beta, lambda2, B, R, D, lambda1, A
you should zero the D=0 value for Si Si Si type. then your results will be correct and tensile stress decrease about 50%.
the reason was mention by belytschko 2002 in the following link:
http://prb.aps.org/abstract/PRB/v65/i23/e235430
regards,