To refine anisotropically all you need to do is delete “beq @ 0.23” from the end of the site line and type “adps” instead. After one cycle of refinement this line gets replaced by the correct adp macro. In TA all sites are forced to be positive definite by default. If you'd rather let adps go negative then think about why (my preference — John Evans 2009/08/27 17:44) then try adding the macro below:

      macro adp_no_limits
{
	move_to u11 min = Val - .05; max = Val + .05;
	move_to u22 min = Val - .05; max = Val + .05;
	move_to u33 min = Val - .05; max = Val + .05;
	move_to u12 min = Val - .05; max = Val + .05;
	move_to u13 min = Val - .05; max = Val + .05;
	move_to u23 min = Val - .05; max = Val + .05;
}

Be careful in e.g. multihistogram refinements. Here you can't just use the @ symbol for uij parameter names or they'll be different for each histogram. It's always sensible to double check that symmetry restrictions on sites are correct. Try table 8.3.1.1 in International Tables or the table on P104 in Terry Willis' text “Thermal vibrations in crystallography” linked here.

See also Matthew Rowles' macro on adps and the note about the error in the Willis table for space group 227.