Underfill Selection and Cure for Flip Chip and BGA Assemblies

Solder joints under a large area array package are asked to survive thermal cycles that strain them in shear. The die and the substrate expand at different rates, and the outermost joints absorb most of the mismatch because they sit farthest from the neutral point. Underfill is the material that addresses that problem. It flows into the gap under the package by capillary action, fills the space between the joints and mechanically couples the die to the board so the strain is shared instead of concentrated. This article covers material selection, flow behaviour, cure and the practical consequences.

What Underfill Does Mechanically

The failure mode it targets is fatigue cracking of the solder joint at the package interface, driven by the difference in expansion between silicon and the board. Without reinforcement, the joint must accommodate the whole mismatch through plastic deformation, and the number of cycles it can survive falls quickly as the package gets larger and the joints get smaller.

Filling the gap with a material that bonds to both surfaces changes the load path. Part of the mismatch is transferred into the underfill, which is compliant enough to deform without cracking, and the joint sees a smaller strain amplitude. The improvement is typically an order of magnitude in thermal cycles, which is why the material is standard on large flip chip devices and on many fine pitch ball grid arrays.

Capillary Flow and Gap Geometry

Capillary flow depends on the gap height, the surface energy of the two surfaces and the viscosity of the material. A narrower gap produces a stronger capillary force but a higher resistance to flow, so there is an optimum band and the process is sensitive to the actual standoff the package achieves. A ball array that collapses to a smaller gap than expected will fill slowly or incompletely.

Temperature is the main control during dispensing. Warming the board lowers viscosity and increases flow speed, which is why boards are typically preheated before underfill is dispensed. The penalty is that warm material cures faster, so the working window between warming and dispensing has to be respected rather than extended for convenience.

Material Properties That Decide the Result

The coefficient of thermal expansion is the first property to match. A material with an expansion close to that of the solder, and well below that of the board, shares the strain without pulling the joints apart at low temperature. Fillers such as silica are added to reduce expansion and to raise the modulus, and the filler content is what limits how fine a gap the material can flow into.

Other properties matter in specific cases. The glass transition temperature must sit above the operating range so the material does not soften when the product is hot, the modulus has to suit the joint geometry, and the adhesion must survive moisture and reflow. Where a product sees repeated reflow, a material that reflows softening or that absorbs moisture quickly will fail even if its thermal expansion is ideal.

Dispense Patterns and Process Control

The dispense pattern determines whether the material fills the gap completely. A single line along one edge is the simplest and works on small packages, while an L or U shaped pattern along two or three edges is used on larger ones to shorten the flow path. The objective is that the fronts meet inside the package, because the last place to fill is the most likely place to trap a void.

Flow time is measured rather than assumed. The operator dispenses on a sample, waits for the X-ray inspection result, and adjusts the volume and the preheat until the fillet is continuous. Because flow time changes with temperature and with material batch, the check has to be repeated after a supplier change, not just at the start of a program.

Cure Profile and Its Effect on Stress

The cure profile determines the final properties and also the residual stress that the material locks in. A slow ramp allows the material to flow and to relax before gelation, while a fast ramp can trap voids and leave a partially cured region in the centre of the package where the temperature lags the edge.

The centre lag is the practical issue. Large packages take time to reach temperature, so a profile that cures the edge correctly may leave the middle under-cured, and the resulting mechanical property difference is invisible from the outside. Measuring the actual temperature under the package, using the same thermal measurement discipline applied in reflow oven profile verification, is the only way to confirm that the whole volume reached the required cure.

Rework and Repair Considerations

Underfill makes rework harder by design. Removing a package requires heating the joints above liquidus and dealing with a cured material that bonds the die to the board, so the usual sequence is to cut or mill the material away, remove the package, clean the site and then decide whether the board is reusable. On dense assemblies the answer is often no.

Where rework is expected, two approaches reduce the cost. A reworkable chemistry softens at a defined temperature so the package can be lifted with controlled force, and a local keep out area limits the material to the package so the surrounding components are not affected. The trade is slightly lower mechanical performance, which is acceptable only if the thermal cycle requirement is known. Reliability testing that demonstrates the resulting life is described in thermal cycling test design.

Verification and Defect Signature

Verification uses X-ray for voids and for missing fillets, and acoustic microscopy for delamination between the material and the die or the board. A white or cloudy fillet usually indicates incomplete flow or moisture absorption, and a fillet that is present on one side only indicates that the fronts met before the package was filled.

The gopcb assembly group records the dispense pattern, preheat temperature, flow time and cure profile for every product that uses underfill, because those four parameters predict almost every defect that occurs later. When a package with an apparently good fillet fails thermal cycling, the record is what shows whether the material or the profile changed.

Verification of Material and Supplier Changes

A material change is a process change, and it should be treated as one. Two underfills that meet the same specification sheet can differ in flow speed, filler size and cure behaviour by enough to change the dispense volume and the flow time the process was built around. Receiving inspection should include a flow test on a sample board rather than only a certificate review.

Supplier changes also affect the shelf life. Materials absorb moisture from the air, and a cartridge that has been open for a week behaves differently from a fresh one, with slower flow and a greater tendency to trap voids. Recording the batch and the open date with each build lets a flow problem be traced back to the material rather than blamed on the dispenser.

FAQ

Is underfill needed under every ball grid array? No. Small packages with a low thermal mismatch index usually survive without it. The decision is based on package size, joint count and the expected thermal cycles.

Can underfill be applied before reflow? There are materials designed for that sequence, but they require a matched process and remove the ability to inspect or rework the joints. Most production uses the post reflow route.

How is incomplete fill detected? By X-ray for voids and by acoustic microscopy for delamination. Visual inspection of the fillet is necessary but not sufficient, because the fillet can be continuous while the centre of the package is empty.