| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Deep Learning Accelerated Lab-Scale X-Ray Computed Tomography of Low-Melting-Point Solder Alloys Used in Heterogeneously Integrated Semiconductor Packages |
| Author(s) |
Eshan Ganju, Nikhilesh Chawla |
| On-Site Speaker (Planned) |
Eshan Ganju |
| Abstract Scope |
In recent years, laboratory-scale X-ray microscopy systems have increased in popularity. However, the limitations in flux within these lab-scale systems frequently result in time-consuming data acquisition, hindering the pace of scientific discovery. Deep learning (DL) methods show immense promise in enhancing low-dose lab-scale X-ray tomography data and semantically segmenting the data to extract microstructural information rapidly. This study presents an in-depth analysis of the deep learning approach to expedite the acquisition of time-resolved (4D) lab-scale absorption contrast tomography (ACT) data of low temperature solder alloys (SnBiIn) used in heterogeneously integrated semiconductor packages at different levels of aging. We used low and high-dose ACT datasets to train Generative Adversarial Networks (GANs) based models to refine the low-dose data to approach the quality of the high-dose datasets and segment the different phases within the dataset. We also present a quantitative comparison of the image quality and time savings achieved using DL-based approach. |