edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 4 14 NONDESTRUCTIVE 3D X-RAY MICROSCOPY SPEEDS THROUGHPUT IN NEW FAILURE ANALYSIS WORKFLOWS Cheryl Hartfield, FASM Carl Zeiss Microscopy LLC, Dublin, California cheryl.hartfield@zeiss.com EDFAAO (2024) 4:14-19 1537-0755/$19.00 ©ASM International® INTRODUCTION The market for high performance computing (HPC) has surged due to the increasing use of artificial intelligence (AI) and machine learning across a growing number of industries. This is fueling the adoption of heterogenous integration and “chiplet” computing architectures, driving a need for tightly packaged interconnects in 2D and 3D to increase feature density (Fig. 1). As the size and complexity of the die and packaging increase, so do the characterization and failure analysis (FA) challenges. It is becoming more challenging to identify defects and their locations quickly and nondestructively to effectively get to the root cause of the failure.[1,2] New FA workflows that leverage 3D x-ray microscopy (XRM) along with femtosecond lasers (fs-laser) and AI training models[3] can speed up nondestructive fault detection with repeatable recipes. This article shows the application of 3D XRM to nondestructively detect non-optimized assembly processes that can influence local stresses and overall device reliability, making it useful for process development as well as FA. The example discussed is thermocompression bonding (TCB) at 20 µm bump pitches. ZEISS Versa 3D x-ray microscopes have submicron resolution and are well-suited for imaging these structures.[4] When used along with AI training models, 3D XRM can achieve analysis of highly integrated packaging structures with reasonable throughput for process validation and error correction guidance. A second example shows the effectiveness of 3D XRM to guide and verify the precise and targeted sample preparation of a 3D package using a fs-laser integrated FIB-SEM, with the goal to enable functional testing and fault isolation with very little damage to the package and IC. In both use cases, combining 3D XRM and deep learning achieves a robust and repeatable nondestructive analysis method for faster insights on highly integrated packaging structures. 3D DATA FOR COMPREHENSIVE CHARACTERIZATION OF TCB PROCESSES As interconnect pitch transitions below 100 µm, thermocompression bonding (TCB) is required. TCB enables Fig. 1 Fan-out and 2.5D interconnects achieve high interconnect densities, while the recent commercial adoption of hybrid bonding has enabled further exponential increases in interconnect density per area.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==