Underfill Voids: Where They Come From and How to Find Them
Underfill is a filled epoxy that flows into the gap beneath a package and mechanically couples the component to the board. It reduces the strain that thermal cycling puts on the solder joints, and it does that only where it has actually reached.
A void is a region the resin did not fill, and its significance depends on where it is. A void in the middle of a large package changes very little, while a void that exposes a corner joint removes the reinforcement exactly where the strain is highest.
What Underfill Is For
The reinforcement works by transferring load from the solder joint into the board and the package body, which reduces the strain range the joint sees in each thermal cycle. The gain is large for a package with a low standoff, where the joint is small and the strain is concentrated.
That mechanism requires continuous material between the two surfaces. A void breaks the load path locally, and the joint beside it reverts to behaving as though it were not underfilled, which is why the acceptance criteria are written by position rather than by total void area.
Capillary Flow and What Stops It
The resin enters the gap by capillary action along one or more edges of the package and flows until the gap is full. The driving force is surface energy, and the resistance comes from viscosity, filler content and the gap height itself.
Flow stops when the driving force falls below the resistance. A gap of 0.03 mm fills slowly, and it becomes impossible to fill once contamination changes the surface energy of the board or the package underside.
Flux Residue and Surface Energy
Flux residue is the most common cause of incomplete filling, because it changes the surface that the resin has to wet. A no-clean residue in particular leaves a film that can be uneven, and the resin flows around it rather than through it.
Where underfill is planned, the assembly process should be chosen with that in mind: either a cleaned board with a verified cleanliness level, or a flux chemistry that the underfill supplier has tested for compatibility. Leaving the choice to whichever flux was already on the line is how voids appear in a process that was qualified with a different material.
Dispense Pattern and Volume
The pattern is chosen to give an even flow front. A single line along one edge flows the length of the package, an L pattern fills two edges and meets in the middle and a U pattern surrounds the package so that three fronts meet inside.
The volume has to fill the gap plus a fillet on the outside, and it is calculated from the package footprint, the standoff and the fillet radius. Too little leaves a short fill, while too much wicks onto neighbouring components or onto the pads of the next package, and the pattern itself is a process qualification rather than a setting.
Temperature, Warpage and Gap Height
Underfill is dispensed warm, commonly between 60 and 90 °C, because the viscosity falls and the flow is faster. The same temperature also softens the assembly, and a warped board changes the gap height from one corner of the package to another.
A gap that is thinner in the middle than at the corners fills from the corners and leaves the centre empty, which is a void that follows the package outline. Preheating the board to a uniform temperature and supporting it during dispense reduces that effect, and the warpage itself is the same phenomenon described in the notes on BGA warpage.
Vacuum Assistance
Vacuum-assisted dispensing pulls the resin into the gap and removes air ahead of the flow front, which improves filling in packages with a low standoff. The vacuum has to be released gradually, or the resin is drawn out again along with the air. A vacuum stage also changes the shape of the fillet, so acceptance criteria have to be set on parts built with the same equipment.
The technique adds equipment and cycle time, and it is normally reserved for packages that cannot be filled reliably by capillary flow alone. Where it is used, the vacuum profile becomes part of the process specification and is verified in the same way as the dispense pattern.
Finding Voids After the Fact
Ultrasound is the standard method, because it detects a resin and air interface through the package. The resolution depends on the frequency and on the filler in the resin, and a highly filled material gives a noisier image than an unfilled one. A repeatable void inspection depends as much on the coupling and the fixture as on the transducer, because a probe presented at a different angle returns a different image.
X-ray shows voids only where there is a difference in density great enough to register, so it is useful for large voids and poor for thin ones. Both methods are best used on a sectioned sample once, to establish what the image corresponds to, and then on production units. The void measurement conventions are described in the notes on BGA voiding and measurement.
Acceptance Criteria and Their Limits
A typical criterion limits the voided area to a percentage of the package footprint and prohibits voids that touch a corner joint or that bridge between two adjacent joints. The percentage is a rough measure, and the positional rules carry most of the meaning. A large void in the middle of a footprint is usually tolerated, while a small one over a corner joint is not, and the criteria should say so explicitly rather than leave it to the inspector.
The criteria should be derived from a reliability test rather than copied, because the tolerance to voids depends on the package size, the standoff and the thermal cycle the product will see. A criterion of ten percent is common, and it is meaningless without the positions it applies to. The method used to measure the voided area should also be named, because an ultrasound image and an X-ray image of the same void rarely give the same number.
Process Control and Records
Control is maintained through the dispense pattern, the substrate temperature, the pre-bake of the package and the cleanliness of the board. Each of those is recorded per lot, and the ultrasound image of a sample is kept with it. The pre-bake and the dispense temperature drift slowly, so a trend chart of the fillet width catches a change several lots before the voids appear in the sample image.
Where a void problem appears, the record shows which parameter changed. Most often it is a flux or a cleaning change rather than the underfill itself, because the dispense equipment is closely watched while the chemistry upstream is not. The general treatment of these defects is in the notes on coating and underfill application defects.
FAQ
Can underfill voids be eliminated completely? Not in practice. The aim is to keep them away from the joints that carry the highest strain and below a size that has been shown not to matter in testing.
Does a cleaned board always fill better? It fills more predictably. A verified cleanliness level removes the largest variable, and it is the usual step when a void problem cannot be traced to the dispense equipment.
Is X-ray good enough to find underfill voids? Only for large ones. Ultrasound resolves the thin voids that matter, and a sectioned sample should be used to establish what each image means before criteria are written.
