Underfill gets scheduled like a finishing operation — bump the wafer, attach the die, then fill the gap. But whether underfill can flow at all is decided by geometry that was frozen at the bump decision, months earlier. That is why underfill problems feel like they arrive from nowhere: by the time they show up, the thing that caused them cannot be changed.
The two processes
Capillary underfill is dispensed along the edge of an already-attached die and pulled into the gap by capillary action. Nothing pushes it. The flow is driven entirely by the geometry of the space it is flowing into, and it has to travel from the die edge to the centre without stalling and without trapping air between bump rows.
No-flow underfill inverts the order. The material is deposited on the substrate first, and the die is placed into it — the bumps push through the underfill to reach the pads, and reflow and cure happen together. There is no flow distance to worry about because nothing has to flow anywhere.
What decides between them
1. Standoff height, which is a bump decision
Capillary flow needs a predictable gap. A copper pillar array gives one by construction — the post holds the die at the height you plated it. A collapsed C4 array at tight pitch gives a gap that varies with how each individual joint happened to collapse, and a flow front that meets varying resistance stalls unevenly.
This is the coupling that matters: the bump choice and the underfill choice are the same decision, made months apart by different people. Choosing C4 for cost at a pitch where collapse is marginal, and then discovering the underfill will not flow cleanly, is a common and expensive sequence.
2. Flow distance, which is die size
Capillary underfill has to reach the centre of the die from its edge. Large die make that a long journey through a crowded gap, and the middle is exactly where an incomplete fill is hardest to detect and most damaging — it is the region under the highest thermomechanical stress.
3. Bump density
Every bump is an obstacle to the flow front. A dense array at fine pitch is a maze, and a maze is where a flow front splits, rejoins, and traps air. No-flow avoids the entire class of problem by never asking anything to navigate.
Why no-flow is not simply better
Because it moves the difficulty rather than removing it. The die now has to be placed through a viscous material, which affects placement accuracy and demands that the bumps displace underfill cleanly enough to make metallurgical contact at every joint. Trapped underfill between a bump and its pad is an open circuit, and it is inside the joint where inspection is hardest.
No-flow also couples the cure to the reflow, so one thermal profile has to satisfy both the solder metallurgy and the underfill chemistry. Capillary keeps those independent, which is a real advantage when either one is fussy.
What we run, and how it gets checked
We run both capillary and no-flow underfill, alongside C4 solder bump, copper pillar and gold stud interconnect. Void content in the underfill and in the joints is verified by X-ray when void measurement is required by the part or the specification, with automated optical inspection and metrology alongside it, and cross-section where a programme needs destructive confirmation. No process is void-free by assertion; the useful question is what void content was measured and against what limit.
Where to start
Bring these three and the underfill process picks itself:
- Die size — specifically the longest distance underfill would have to travel from edge to centre.
- Bump pitch and array density, which set how obstructed that path is.
- Whether the standoff is controlled or collapsed, which is really a question about which bump you have chosen.
More on the interconnect options on our flip-chip assembly page, and the bump decision itself is covered in C4 versus copper pillar — worth reading first if that choice is still open.
Answered.
What is the difference between capillary and no-flow underfill?
Capillary underfill is dispensed at the edge of an already-attached die and drawn into the gap by capillary action, so it has to flow the whole distance to the die centre. No-flow underfill is deposited on the substrate first and the die is placed into it, with the bumps pushing through to reach the pads and reflow and cure happening together — so nothing has to flow anywhere.
When should you use no-flow underfill?
When capillary flow is the risk: large die where the edge-to-centre distance is long, dense fine-pitch arrays that obstruct the flow front, or collapsed bump geometries where the standoff varies joint to joint. The trade is that the die must be placed through viscous material and the bumps must displace it cleanly enough to make contact at every joint.
Why does bump choice affect underfill?
Because capillary flow needs a predictable gap. A copper pillar holds the die at a standoff you control by construction; a collapsed C4 array at tight pitch gives a gap that varies with how each joint collapsed, and an uneven gap makes the flow front stall unevenly. The bump decision and the underfill decision are effectively the same decision, which is why making them months apart causes trouble.
How are underfill voids detected?
Void content is verified by X-ray when void measurement is required by the part or the specification, alongside automated optical inspection and metrology, with cross-section available where a programme needs destructive confirmation. The useful question about any underfill process is what void content was measured and against what limit, not whether it is void-free.
Do you run both underfill processes in the US?
Yes — capillary and no-flow underfill, alongside C4 solder bump, copper pillar (C2) and gold stud interconnect, in Halethorpe, Maryland, on US soil at prototype and low volume.