Browse Prior Art Database

Statistical Methods to Improve Chip Structures for Better CPI

IP.com Disclosure Number: IPCOM000214095D
Publication Date: 2012-Jan-10
Document File: 2 page(s) / 22K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a set of statistical techniques to identify particular components of the Back End of Line (BEOL) of semiconductor devices that strongly correlate with the likelihood to form white bumps. The invention is an algorithm that receives white bump and BEOL data and then produces a small number of significant BEOL components that can explain most of the white-bump data.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 2

Statistical Methods to Improve Chip Structures for Better CPI

In recent technology nodes, chip-package interactions (CPI) have led to the fracturing and delamination of brittle structures in the Back End of Line (BEOL) of semiconductor devices (i.e., white bumps). The location on the device of those problems is not completely random, and it is somewhat correlated with particular structures in the BEOL or in the laminate designs. It is quite difficult to actually identify which structures are responsible for the increased propensity to fail, due to the high complexity of the BEOL.

Disclosed here is a set of statistical techniques to identify particular components of the BEOL that strongly correlate with the likelihood to form white bumps. These techniques only use data about the number of white bumps observed at each C4 location and data about the structure of the BEOL (in the form of feature densities at the C4 locations) to identify significant components.

The invention is an algorithm that receives white bump and BEOL data and then produces a small number of significant BEOL components that can explain most of the white-bump data. It works as follows:

1. For a given semiconductor device, organize the white bump data in a column vector W, with component W(i) being the number of white bumps observed on the ith C4 in an experiment where a number of devices were assembled and inspected for white bumps.

2. For the same device, let the matrix D represent design features of the device BEOL. Columns of D are vectors the same size as W, with ith component corresponding to the design feature at the ith C4 location. Examples of design features include copper or dielectric densities, effective mechanical properties computed from the latter, etc.

3. The...