Omniprobe's FIB and SEM Blog

Total Release™ Method for IC Process and Package Development

by Cheryl Hartfield

In the semiconductor industry, packaging development and Si technology development are often dependant upon each other.  This is due to the  requirement for the Si chip to actually be "packageable", and for packaging teams to develop new processes to handle "new silicon".  The interface between the package and the chip sees high stresses. The interface typically covers a large area, making the Total Release™ method for in situ lift-out an ideal approach to rapidly biopsy the area of interest. The Total Release™ method mills a minimum of material at two angles to create a wedge-shaped sample.  The sample is completely severed from the bulk, prior to attaching the probe and manipulating.  This provides rapid preparation enabling faster process development cycles.  A case history example is discussed below.

During the early 2000's, significant wirebonding and flip chip bumping challenges were experienced during the industry's adoption of low-K dielectrics.  A successful solution to the challenge required a combination of fab material process optimization, new Si design rules, and changes to standard assembly methods.  In situ lift-out by the Total Release™ method played a key role in enabling success.

Early wirebonding attempts resulted in significant "lifted ball" failures, with the wire pulling away the pad and some of its underlayers.  Optical and SEM images showed multiple layers in the Si BEOL layers failing.  Understanding the failure initiation was paramount to address the problem.  Modeling indicated highest stresses were at the edges of the ball bond.  Aluminum still remained at the edge and covered the failing layers beneath, making top down analysis by surface sensitive techniques such as Auger inadequate.  A FIB cross-section provides some insight, but lacks the contrast to adequately identify the specific interfaces that failed and differentiate between delamination and cohesive fracture.  Only a TEM image could provide enough contrast to clearly differentiate the involved layers.

A large area of ~65um in length had to be prepared for TEM, in order to capture the edges of the ball bond that likely contained the failure initiation site.  Conventional TEM preparation such as sawing, cleaving, and mechanical polishing, could not be applied to the fragile, damaged layers. FIB preparation was a necessity.  Due to the large area, a protective coating was applied to the surface of the sample ex situ; the lab's RIE plasma tool was leveraged to lay down a 1um thick layer of blanket TEOS.

The Total Release™ method was used to perform the lift-out.  By milling out a wedge-shaped slice, the area subjected to FIB milling was minimized, speeding up the preparation.  To securely stabilize the long sample on the grid, the sample was attached through-out its length.  The resulting TEM lamella was too large to grab in one TEM image.  Both edges of the ball bond landing site were preserved, and the resulting TEM image provided conclusive determination of the layers involved in the failure initiation.  This data led to process changes that improved the adhesion at this interface.

Flip chip bumps were likewise analyzed using the Total Release™ method for preparation.  To access the bump and under-bump metallization (UBM), cross-sectioning was required.  This can induce artifactual damage.  Studies showed the cross-section induced damage extended no more than ~3um past the surface, so preparing in situ lift-out samples 10um in depth past the sectioned edge allowed assessment of UBM damage in the absence of cross-section induced damage.

The examples here show the Total Release™ method for in situ lift-out can perform rapid preparation of large areas such as bond pads and flip chip bump UBM, even in the presence of severe pre-existing damage. This critical capability is in the feedback flow for process development, where rapid process cycles rule.  Total Release™ is an effective in situ lift-out technique to prepare high quality samples quickly.

The views and opinions expressed herein represent those of the author and are not necessarily those of Omniprobe, Inc., or its employees.  Any mention of product or company names are not endorsements.  Trademarks mentioned in this blog are the exclusive property of their respective owners.