I'll only mildly apologize for the large size and ugly figures, since it's free. Hope this is useful!
Albert
Although we don't know eta(x), we do know x(eta):
where {Xi} is the nodal coordinates of the element
So, the inverse:
is "easy" to evaluate.
For a trilinear hexahedron defined as:
Evaluating at the CENTROID (not at the Gauss points!!),
where all matrix entries are zero except:
For a 6-noded pentahedron (wedge):
so the centroid is again used to evaluate the eta's instead of the gauss point
For a 4 noded tetrahedron:
So, since a purely linear element has constant derivatives, the true jacobian is worthless for 4-node tetrahedra. Hence it is common to replace it with a volumic measure, to remain consistent with the other element's jacobians.