The default value null sets tolerance to 1e-6.
The source of this document is in NumericalLinearAlgebra.m2:215:0.