Thus the a priori bound for the error is,

or

For the a posteriori bound,

where . Therefore,

or

Now the error estimate is given by

is the solution of

Consequently,