An MTJ in an MRAM array or TMR read head is disclosed in which a capping layer has a bilayer configuration with a non-magnetic NiFeX inner layer on a NiFe free layer and a Ta layer on the NiFeX layer to improve dR/R and minimize magnetostriction. Optionally, a trilayer configuration may be employed where the Ta layer is sandwiched between an inner NiFeX layer and an outer Ru layer. The X component in NiFeX is preferably an element having an oxidation potential greater than Ni or Fe such as Mg, Hf, Zr, Nb, or Ta. NiFeX is preferably formed by co-sputtering a NiFe target with an X target at a forward power of about 200 W and 50 W, respectively. In an MRAM structure, the Mg content in NiFeMg may be increased to >50 atomic % to improve the gettering power of removing oxygen from the free layer.