Laser Assisted Bonding Technology Enabling Fine Bump Pitch in Flip Chip Package Assembly

Abstract

Advanced silicon (Si) node technology development is moving to 10/7nm technology and pursuing die size reduction, efficiency enhancement and lower power consumption for mobile applications in the semiconductor industry. The flip chip chip scale package (fcCSP) has been viewed as an attractive solution to achieve the miniaturization of die size, finer bump pitch, finer line width and spacing (LW/LS) substrate requirements, and is widely adopted in mobile devices to satisfy the increasing demands of higher performance, higher bandwidth, and lower power consumption as well as multiple functions. The utilization of mass reflow (MR) chip attach process in a fcCSP with copper (Cu) pillar bumps, embedded trace substrate (ETS) technology and molded underfill (MUF) is usually viewed as the cost-efficient solution. However, when finer bump pitch and LW/LS with an escaped trace are designed in flip chip MR process, a higher risk of a bump to trace short can occur. In order to reduce the risk of bump to trace short as well as extremely low-k (ELK) damage in a fcCSP with advanced Si node, the thermo-compression bonding (TCB) and TCB with non-conductive paste (TCNCP) have been adopted, although both methodologies will cause a higher assembly cost due to the lower units per hour (UPH) assembly process. For the purpose of delivering a cost-effective chip attach process as compared to TCB/TCNCP methodologies as well as reducing the risk of bump to trace as compared to the MR process, the laser assisted bonding (LAB) chip attach methodology was studied in a 15x15mm fcCSP with 10nm backend process daisy-chain die for this paper. Using LAB chip attach technology can increase the UPH by more than 2-times over TCB and increase the UPH 5-times compared to TCNCP. To realize the ELK performance of a 10nm fcCSP with fine bump pitch of 60μm and 90μm as well as 2-layer ETS with two escaped traces design, the quick temperature cycling (QTC) test was performed after the LAB chip attach process. The comparison of polyimide (PI) layer Cu pillar bumps to non-PI Cu pillar bumps (without a PI layer) will be discussed to estimate the 10nm ELK performance. The evaluated result shows that the utilization of LAB can not only achieve a bump pitch reduction with a finer LW/LS substrate with escaped traces in the design, but it also validates ELK performance and Si node reduction. Therefore, the illustrated LAB chip attach processes examined here can guarantee the assembly yield with less ELK damage risk in a 10nm fcCSP with finer bump pitch and substrate finer LW/LS design in the future.