Optimization of laser release process for throughput enhancement of fan-out wafer-level packaging

Fan-out wafer-level-packaging (FOWLP) technology has been developed with various advantages, such as smaller form factor, lower cost, and simplified supply chain for heterogeneous integration. There have been several process schemes like chip-first or chip-last FOWLP integration discussed widely in conferences in recent years. One process in all of these process schemes has in common is the use of a temporary carrier for subsequent redistribution layer (RDL) formation, chip stacking and molding processes. Although the separation of a temporary carrier from the reconstituted wafer could be achieved without significant hurdles, there were few studies addressing optimization of carrier separation for throughput enhancement. Thus, this paper is designed to address the needs in optimizing carrier separation process based upon laser ablation technology. Two phases of experiments were designed to select the appropriate laser release layer and define optimal laser settings. The first experiment was used to evaluate correlation of the laser absorption, laser energy, and spot pitch versus completeness of laser ablation. The second experiment included RDL-first FOWLP integration. At first, 300-mm glass carriers (1000 mu m thick) with coefficient of thermal expansion of 8 ppm/degrees C were treated by selected laser release layers. After deposition of 0.05-mu m Ti/0.15 mu m Cu on the glass carrier, passivation of around 8 mu m was coated and patterned by lithography for electroplating Cu interconnections with a density of approximately 10% of the surface area. Subsequently, die bonding, build-up layers, or molding compound were applied on top to form a 200-mu m reconstituted wafer. The reconstituted wafer was then separated from the glass carrier through a laser ablation process using a 355-nm laser to determine optimal throughput. Experiments to study correlation of laser release layer with laser settings along with a demonstration of full RDL-first FOWLP integration are discussed thoroughly to address the need of throughput enhancement, which could serve as cornerstone for realizing cost-effective RDL-first FOWLP.