Tutorial #11
George Riley
September 2001

GENERAL

Under-Bump Metallization (UBM) is to a flip chip bump what a foundation is to a house: the essential base that supports the whole structure. Like a foundation, too, the UBM is mostly out of sight, and little thought about — unless there are problems with it. UBM, for most flip chip processes, is the first step in bumping the integrated circuit (IC) bond pads.

The final metal layer of most IC bond pads is aluminum, providing a satisfactory surface for conventional wire bonding. Unfortunately, this surface is inhospitable to most conductive bumps. Aluminum forms an oxide immediately upon exposure to air, and this native oxide is an electrical insulator. A wire bond in its formation scrubs through the insulating oxide to weld with the underlying metal. Bumps need an alternative strategy for making electrical connection.

Aluminum is not a readily solderable surface, neither wettable nor bondable by most solders. Aluminum may corrode over time when not protected from the environment. Consequently, successful bumping must first replace the oxidized aluminum surface with a more hospitable material, the UBM.

This UBM must meet several requirements. It must provide a strong, stable, low resistance electrical connection to the aluminum. It must adhere well both to the underlying aluminum and to the surrounding IC passivation layer, hermetically sealing the aluminum from the environment. The UBM must provide a strong barrier to prevent the diffusion of other bump metals into the IC. The UBM must be readily wettable by the bump metals, for solder reflow.

Meeting all these requirements generally requires multiple layers of different metals, such as an adhesion layer, a diffusion barrier layer, a solderable layer, and an oxidation barrier layer. Like a happy family, the UBM layers must be compatible metals which in combination have low internal mechanical stresses. The composite UBM should result from processes that are relatively simple, inexpensive, and easily reproducible in volume production.

The design of the UBM should provide a metal layer that adequately defines and limits the applied bump (hence an alternative name, the “bump-limiting metal”) while overlapping the die passivation.

Figure 1 shows a typical design layout for UBM relative to the original pad.

UBM PROCESSES

The IC pad aluminum oxide may be removed by sputter etching, plasma etching, ion etching, or by a wet chemical treatment. Successive UBM layers may be vacuum deposited by evaporation or sputtering, or be chemically plated. UBM deposition processes generally require wafers of IC die, rather than individual die.

The wide and growing array of UBM compositions (e.g. Cr:Cr-Cu:CU, Ti:Ni-V, Ti:Cu, Ti:W:Au, Ni:Au) results in part from the desired bump material and characteristics, in part from the intended end use application, and in part from the experience and the invested capital base (or, the hopes, fears, prejudices, and pocketbook) of the manufacturer. The required metallurgies are part art, part science, and part accumulated experience.

About 75% of UBM currently produced consists of multi-metal layers evaporated or sputtered in a vacuum system. A typical process sequence would be:

1. Sputter etch the native oxide to remove oxide and expose fresh aluminum surface.

2. Deposit 100 nm Ti / Cr / Al as the adhesion layer.

3. Deposit 80 nm Cr:CU as the diffusion barrier layer.

4. Deposit 300 nm Cu / Ni:V as the solder-wettable layer.

5. Deposit 50 nm Au as the oxidation barrier layer (optional).

Electroless nickel UBM is unique in being a wet-chemical process that requires neither vacuum processing nor masking.

Figure 2 shows a wafer with electroless nickel UBM on the bond pads.  The aluminum oxide is chemically etched, and replaced with zinc seed crystals. Nickel is electrolessly plated over the zinc as a barrier layer and solderable layer, and immersion gold is plated over the nickel as an oxidation barrier.

The bump normally entirely covers the UBM, concealing it from view. The size of the UBM controls the spread, and therefore affects the resulting height, of a solder bump.

Figure 3 is a cross-section of a solder bump showing the underlying electroless nickel-gold UBM overlapping the die passivation.

The robustness of a particular UBM depends on its composition, deposition process, and the skills of the manufacturer. Low stress, stable films are desired, and the deposition process governs the initial tensile or compressive stresses within the layers. However, crystalline changes and grain growth within in the metals can change the characteristics of the film throughout its lifetime. For example, the aluminum, nickel, and copper constituents can undergo substantial re-crystallization at temperatures below 150C, changing the structure and stresses. Higher temperatures during reflow or subsequent processing can cause dissolution of UBM materials.

There is no perfectly optimal UBM, and practical UBM formulations require tradeoffs between competing constraints. For example, high temperature or high power applications would benefit from a high copper content, but copper dissolves readily and has a high re-crystallization rate. Electroless nickel UBM is a relatively low cost wet chemical process, but electroless nickel has almost ten times the re-crystallization rate of sputtered nickel. The entire UBM/bump system must reflect a balancing of capabilities and costs.

SUMMARY

A growing number of suppliers can provide UBM, bumping, or technology transfer and licensing of those skills. While most flip chip users need not and probably should not start out by depositing their own UBM, they should be aware of the tradeoffs to be evaluated in seeking a source to provide UBM. They should also be aware that the related metallurgy depends heavily on experience, and that producing good, reproducible, stable UBM requires more than just a chemistry set and a cookbook.

ACKNOWLEDGEMENTS

Material provided by Pac Tech, Flipchip Technologies, and Unaxis Semiconductors, and an informative and helpful discussion with Mike Varnau, Technical Fellow and Department Manager at Delphi Automotive Systems, are gratefully acknowledged.