Silicon Interposer Warpage Study for 2.5D IC without TSV Utilizing Glass Carrier CTE and Passivation Thickness Tuning

Abstract

Warpage control is a key process character for 2.5D IC without Through-Silicon Vias (TSV). For low cost package trend, TSV is not used for advanced 2.5D IC package. Yield is still an issue for 2.5D IC package, Die-Bonding (DB) last process is applied to make sure top die is bonded on good interposer. Therefore, no top die will be sacrificed for low interposer process yield, cost could be more reduced. Re-Distribution Layers (RDL) is first process for DB last interposer, then reversing wafer to do silicone removing, contact pad opening, micro pad plating, die bonding, and compound molding. Because interposer thickness is about 30~50um for 2.5D IC without TSV after silicone removing process, carrier bonding is used before silicone removing. Passivation layer is composited of organic material which coefficient of thermal expansion (CTE) is usually larger than 80ppm. Interposer warpage is substantially increased after silicone removing, which leads that carrier is a major substrate to support RDLs. Glass carrier is discussed in this study and glass CTE and passivation layer thickness are also tuned to optimize interposer warpage after silicone removing. There are two RDL layers and two PBO as passivation material in interposer. Warpage variation is a key evaluation character during this study. From our studies, glass CTE is increased from 3.2ppm to 5.7ppm, and interposer warpage value is reduced and variation could be reduced about 100um at Si removing stage, passivation thickness is increased from X to X+5um, and interposer warpage value is increased and variation could be increased about 50um at Si removing stage. Bi-layer radius formula is used to describe the warpage behavior of under bump metallization (UBM), Bonding, and Si removing stages and is well matched NTI package warpage behavior at UBM and Si removing stages. For Bonding warpage behavior, bi-layer radius theory is interfered by non-perfect adhesion at glue layer, but warpage variation trend is still matched with warpage experiment.