Stevan Hunter

Copper Ball Bond over a Variety of Probe Marks in Two Pad Aluminum Thicknesses

In this experiment, Cu wire bonding takes place over a variety of probe marks on integrated circuit bond pads of 0.8μm and 3μm aluminum (Al) thicknesses. Probe marks were created by cantilever probe tips, typical of probes in manufacturing for bond pads of 75μm width or greater. Three probe cards were built to simulate different contact forces on the pads. Integrated circuit test wafers were probed using each probe card, one half wafer with smooth tips, and the other half wafer with roughened tips. This resulted in a variety of probe mark topographies to use in experimentation. Wire bonding over probe marks was done with 25μm bare copper (Cu) wire with a single bond recipe typical of manufacturing for 3μm pad Al. Data analysis was approached using design of experiment (DOE) factorial methods, bond shear force being the output of interest. Bond shear force values are affected by Al thickness, shear test direction relative to ultrasonic generator (USG) vibration direction, and also probe mark characteristics for Cu bonds on thick pad Al.

Microelectronics wirebond pull and shear test simulations using finite element method

Microelectronics wirebonding is used in the vast majority of semiconductor assembly and packaging operations worldwide. In the last few years, much of the manufacturing that traditionally had been gold (Au) wirebond has switched to using copper (Cu) wire. Once a wirebond “recipe” is established in manufacturing, reliability is checked by Bond Pull Strength test (Mil-Std 883G method 2011) and Bond Shear test (JEDEC JESD22 B116A). Both industry standards were established based on gold (Au) wirebonds. Cu wire is stronger and stiffer, producing higher values than Au in both pull and shear tests when well-bonded, so companies are comfortably complying with the old Au reliability limits while using Cu wire. But in fact, Cu intermetallic (IMC) growth and bonding is quite different than for Au, and failure modes in the shear test can be different for Cu than for Au. A new committee is active to revise the JESD22 B116A method to cover Cu wirebond as well as Au. BYU-Idaho students have responded to recent interest in switching from Au to copper Cu wire, modeling both the bond pull strength and bond shear tests using finite element methods. Simulations include the various ball bond wire types (Au, Cu, and silver (Ag) alloys), ball sizes, pad aluminum (Al) thicknesses, and IC bond pad structures in silicon wafers. Results show how stresses imparted to the bond pad in these reliability tests change when switching wire types, etc., helping to explain the physical results as well as to indicate potential issues in the interpretation of failure modes.

Simulation of ball bonding on various bond pad structures

Various integrated circuit bond pad structures in aluminum-silicon dioxide (Al-SiO2) metallization are modeled, with ball bonding stresses applied in static or dynamic simulations. This work follows from our EPTC 2012 paper. Of special interest is the apparent stress reduction in bond pads that mitigates top SiO2 film cracking during wirebond of ON Semiconductors more robust and circuit under pad (CUP) structures having thin top metal. Both gold (Au) and copper (Cu) materials are simulated as bond ball types. Static simulations are helpful to indicate the stress locations, while the dynamic simulations reveal how bonding stress values are affected by bond pad structures. Film thickness decrease directly increases stress coupling into bond pad sub-layers, making SiO2 cracks more likely during bonding. Dynamic simulations indicate that Al CUP features surrounded by SiO2 result in more stress reduction than removing the Al features of the first sub-layer. Bond pad stress increases more than 2 times when changing from gold (Au) to copper (Cu) ball bond, due to the materials properties change alone, not considering the increased ultrasonic power required for Cu bonding.