Reliability Studies of Excimer Laser-Ablated Microvias Below 5 Micron Diameter in Dry Film Polymer Dielectrics for Next Generation, Panel-Scale 2.5D Interposer RDL

摘要

This paper demonstrates the thermal cycling reliability of 4 μm diameter microvias using an ultra-thin dry film ABF, a non-photosensitive dielectric material. Such via scaling in conjunction with line scaling to achieve silicon BEOL-like RDL densities is required for the next generation of interposers. The dry film dielectric, ABF, is an epoxy-silica filler based material. This is an ideal material for a double-sided, panel-scale compatible electroless copper seed metal deposition process. The test vehicle consisting of daisy chain structures used for the reliability studies was fabricated by an excimer laser dual damascene process. The trenches for the daisy chain line and pad structures were first formed in a novel dry film ABF material. Microvias with diameter of 4 μm were then ablated in the film. The stepper system of the excimer laser allowed sub-micron alignment accuracy for the via structures. Two different capture pad structures were used to land the microvias. The 4 μm diameter microvias were landed in 4 μm width and 5 μm width capture pad structures. A panel-based electroless copper seed metal deposition process was used to form a conductive layer on the polymer film. The desmear process during the electroless deposition increased the microvia diameter to 5 μm and the capture pad widths to 5 μm and 6 μm respectively. The structures were filled by conventional electrolytic plating process and overburdened to a thickness of 5 μm. The panel-scalable Surface Planar DFS8910 tool was used to fly-cut 1 μm deep into the polymer and achieve the final circuitry. The challenges of this mechanical fly-cut process with filler based ABF materials and removal of complete electroless copper seed from the polymer anchors will be discussed. The resistance of the daisy chain structures containing an array of 400 microvias was measured after the planarization process. A yield of 88 % was achieved on a 300 mm wafer with 4 μm microvias and 5 μm capture pad structures with excellent daisy chain resistance. The samples were then exposed to: (A) 1000 liquid-to-liquid thermal shock cycles with a dwell time of 5 mins each at 125 °C and -55 °C and (B) 1000 air-to-air thermal cycles from -55 °C to 125 °C with a dwell time of 15 mins at each temperature node and a total cycle time of 1 hour. The resistances after thermal cycling tests showed an average increase of <; 5 %, well within the 10 % resistance change criteria.