Thin Evaporated Metal Substrates Al, Cu, Au, Mg

In comparison with native bulk copper and bulk aluminum substrates we have also prepared layers 100 nm thick of pure aluminum, copper, gold, or magnesium evaporated on silicon wafers. Due to the contact with ambient air, the native layers of oxides and adsorbates form on top of the metals. On these metal film substrates, EP plates 1 mm thick were prepared by curing the DGEBA-DETA mixture (100 14) at ambient temperature. 

As depicted in Pigs. 9.11 and 9.12, the BM investigations reveal broad, mechanically stiffened interphases in the EP adjacent to all the metal films. The v level of these interphases is almost the same but the position of the peak maximum and the half-width at half-maximum (HWHM) depend on the kind of metal. The EP interphases on aluminum and copper are almost twice as wide as on magnesium and gold. In contrast to the polished metals considered above, the adhesion strength is good enough to maintain the EP-metal joint in the mechanical test. The samples fail between the metal film and the silicon wafer. 

50 Tm) has formed with the maximum sound velocity at the Mg interface. AU of this evidence indicates that the inhibition of the curing process (70% turnover of epoxy groups for RT curing) has been partly released within the mechanical interphase. But in the case of Mg the mechanical interphase remains even after the post-curing, and the height of the interphase has increased. 

For the Mg samples we investigated additionally the influence of the Si wafer on the post-cured state by removing the EP layer with the Mg from the wafer after post-curing (Fig. 9.15). This leads to an astonishing result After release from the rigid wafer, the formerly stiffened interphase became more compliant than the bulk inside the EP layer. 

The width of the negative mechanical interphase exceeds 100 pm. What makes the results even more strange is that the minimum of the negative interphase is about 2% below the related interphase value of the RT-cured sample. It is obvious that DETA losses during post-curing due to diffusion are not able to explain a decrease in the sound velocity data below those values measured for the sample cured at RT. 

---