Reliable joining of armor to the cooling substrate for plasma facing components is of interest for the divertor, first wall and startup limiter in ITER and compliant interlayers between the armor and heat sink have been suggested for reducing stresses. This paper summarizes some results on tungsten (W) armor from a broader study that included 1mm or 2mm interlayer (and no interlayer) between W or beryllium (Be) armor joined to a copper alloy (Hycon-3) heat sink. The analyses were done with a 2-D finite element model. The layout was created in PATRAN and the calculations were performed with ABAQUS. Generalized plane strain elements and temperature dependent material properties were used for all parts of the model (armor, interlayer, channel and back). Strain hardening of the W armor, interlayer materials (soft copper and Cu50-W50) and Hycon channel was also included. The Cu50-W50 was a hypothetical material with the values of its properties set at midway between those of W and Cu. The thermal history of the samples modeled began with fabrication (stress free state) at 550 C followed by cooling to 25 C and then application of a heat flux of 5 MW/m² on the plasma facing surface with 160 C water at 3.8 MPa and 10 m/s in the coolant channel. The stresses and plastic strains versus the position along the joint at the base of the armor for the various configurations are compared as are bar charts that show the stresses at several discrete locations around the sample. The W50-Cu50 interlayer was sufficiently strong that the stresses in the armor mimicked those for W bonded directly to the channel (no interlayer) in most cases. Both the no interlayer and the W-50-Cu50 interlayer cases exhibited high (~ -390 MPa) X stresses in the armor at the center of sample and high Y stresses near the edge of the joint. In comparison, the X and Y residual stresses with the 1mm soft copper interlayer were roughly half the values of those without the interlayer. The pattern of the stresses is somewhat complex. Plastic strains near the edge of the joint were 4-6.5% here as compared with values of 20% and higher found in earlier cases with perfectly plastic materials. The calculated stresses and strains near the edge of the joint must be evaluated with some reservations. The calculational problem is well known ("stress singularity" problem - the equations at the corner of the joint and the free surface cannot be satisfied no matter how fine a mesh is used) and the calculated value depends upon the mesh size and is not therefore independent of the modeling approach.