There is no one-size-fits-all answer to this question. Under the same heading of "wafer front-side cleaning", conventional flip-chip packaging and 2.5D / 3D advanced packaging land at opposite ends of the spectrum. The deciding variable is not the filament material but the feature size of the surface being cleaned.
Need the feasibility of a cleaning brush for semiconductor packaging assessed first? Send a drawing or a sample. YC Brushes assesses it from the drawing or sample and finalizes the design with a 3D-printed sample before volume production → Send your inquiry
Where a brush really sits in the wafer cleaning flow
Getting the positioning clear first keeps you from asking the wrong question. Wafer cleaning uses different methods at different process stages. They sit in different places and are not alternatives for the same step:
| Method | Characteristics | Position in the flow |
|---|---|---|
| O2 / Ar plasma cleaning | Non-contact | Surface preparation before die attach, wire bond and underfill |
| CO2 snow jet | Non-contact, non-destructive, residue-free | Specific surface cleaning applications |
| N2 ionized air blow-off | Non-contact | Commonly the last step after dicing and before pick-and-place |
| Megasonic + DIW | Wet process | Specific wet cleaning stages |
| Contact brush scrubbing | Contact | Post-CMP wet process |
The most important information in this table is actually in the last row: the established place of contact brushes in the semiconductor industry is the post-CMP wet process, where they treat the dielectric and metal layer surfaces after planarization. That is a "flat surface + wet environment" scenario. Carrying it over directly to cleaning before pick-and-place, which is a "bump structures + dry environment" scenario, means treating two completely different problems as one. In publicly available information we have not seen an established commercial brush module for the latter.
This does not mean it is not feasible. It means there is no ready-made answer to copy, and any solution has to be verified by actual testing.
Five feasibility levels by bump topology
The physical limits of a contact brush are simple: the filament tips must be able to get in between the structures, and the pressure applied must not exceed what the structures can tolerate. These two conditions divide feasibility into five levels.
| Level | Bump type | Typical height / pitch | Feasibility |
|---|---|---|---|
| 1 | Conventional C4 / flip-chip solder bump | 75–150 μm / 150–250 μm | Relatively low risk, but strict control is still required |
| 2 | Standard Cu pillar | 40–75 μm / 80–130 μm | Borderline; if a mechanical option is assessed, sponge type only |
| 3 | Fine-pitch Cu pillar | Below 40 μm / below 60 μm | Not feasible |
| 4 | Micro-bump (2.5D / 3D advanced packaging) | 10–25 μm / 40 μm or less | Strictly off-limits |
| 5 | Hybrid bonding (Cu-Cu direct bonding) | No bumps; Cu pads are coplanar | Off-limits; any contact is contamination |
Level 4 is strictly off-limits because of shear tolerance: the tolerance of a micro-bump is extremely low, and the pressure from a filament tip is enough to cause damage. In advanced packaging these structures rely on underfill to cushion lateral stress, but cleaning takes place before the underfill is applied, when the bumps are fully exposed and fragile.
Level 5 is even more direct: the bonding surface in hybrid bonding consists of coplanar copper pads, and any contact is contamination.
If the structure falls in Level 1, the parameters worth discussing are extremely soft filament (diameter below 70 μm), a linear sweep instead of rotation, and low contact pressure (on the order of 30–150 g/cm²). But this is a starting point for discussion, not a conclusion, and it still has to be verified on actual wafers.
What a brush can handle, and what it cannot
Even when the topology allows it, not every kind of contamination is suited to a brush.
Gaps it can fill: loose particles that nitrogen blow-off cannot carry away, and dust settled on top of the bumps. Removing this kind of contamination by mechanical contact is effective: for particles of 0.1–1 μm, the brush has to touch the particle to overcome the van der Waals force or remove embedded particles (Journal of the Semiconductor & Display Technology).
What it cannot handle:
- Strongly adhering organic residue: mechanical sweeping cannot remove it; solvent or plasma treatment is needed. This is a conclusion inferred from the removal mechanism.
- Sub-micron particles deep in the gaps between bumps: the filament tips physically cannot reach deep into the gaps between structures.
- Non-contact methods have limits of their own. A CO2 snow jet, for example, is non-destructive and residue-free, but its effect on sub-micron particles below 0.1 μm is limited (Academia.edu).
So the right question is not "can a brush replace the existing methods" but "can a brush cover the part the existing methods miss".
When not to go to a brush supplier
We once received an inquiry about wafer front-side cleaning in a packaging and testing process, and the customer wanted us to quote directly. We had three options at the time: decline outright, quote as asked, or first explain the current state of the industry and the technical risks and then talk. We chose the third: we explained that no comparable module appears in publicly available information, set out the feasible range of bump topologies, and asked in return five technical questions that had to be clarified first.
The reason is a practical one: if we quote as asked and the sample sheds filament or damages bumps on the customer's production line, the responsibility lies entirely with the supplier, and that loss far exceeds the value of the order. Laying the uncertainty out in the open protects both sides.
In the following three situations, we recommend not starting your assessment from a brush solution:
- The structure falls at Level 3 or further down the table (fine-pitch Cu pillar, micro-bump, hybrid bonding). It is physically not feasible, and no material or process can get around that.
- The contamination is strongly adhering organic residue. What is needed is chemical or plasma treatment; mechanical sweeping does not work.
- The residue gap left by the existing non-contact methods has not been confirmed yet. If you do not know what the existing flow misses, there is no way to judge what the brush is supposed to cover, and no basis for designing a verification method.
Five things that must be confirmed before any assessment
Before any quote, these five pieces of information decide whether a solution holds up:
| Item | Why it must be confirmed first |
|---|---|
| Bump type, height and pitch | Directly decides which of the five levels applies, and also the upper limit of the filament diameter |
| Nature and source of the contamination | Loose particles, settled dust and organic residue each need a different mechanism |
| Existing cleaning flow and its residue gap | A brush fills a gap, so you need to know what gap |
| Mechanical parameters | Down force, linear speed and quick-release interface, provided by the equipment side |
| Verification method and acceptance criteria | Shear testing and particle adder measurement on actual wafers are required; appearance alone is not enough |
The last item is the most critical. There is no clear data in publicly available information on how filament tip pressure affects the shear tolerance of solder bumps, which means any solution must be verified by actual testing: the customer provides test wafers, and shear and particle adders are measured, instead of judging from a specification sheet.
Why this article does not give you a clear recommended solution
Because for this application the honest answer is "it depends, and in most cases we do not recommend it".
A brush plays a gap-filling role in the semiconductor cleaning flow. Its applicable range is narrow, and the more advanced the packaging technology, the less applicable it is. For Level 4 and 5 structures, this is not a problem that price or specifications can solve; it is a physical limit.
Only if your structure falls in Level 1 and you have confirmed that the existing flow leaves a residue gap that can be removed mechanically is it worth going on to discuss materials and mechanics. Before that, the first thing to do is clarify the nature of the contamination and the gap in the existing flow.
FAQ
Why can a brush be used post-CMP but not before pick-and-place?
The surface conditions are completely different. Post-CMP treats a surface after planarization and is a wet process; a wafer before pick-and-place has bump structures and is in a dry environment. A filament behaves differently, and carries different risks, on a flat surface and on a surface with raised structures. In publicly available information we have not seen an established commercial brush module for the latter.
Why are micro-bumps strictly off-limits?
Because their shear tolerance is extremely low, and the contact pressure of a filament tip is enough to cause damage. In advanced packaging these structures rely on underfill to cushion lateral stress, but cleaning takes place before underfill, when the bumps are fully exposed. This is a physical limit, not something a material or process can get around.
Can a contact brush replace plasma or nitrogen blow-off?
No. It fills a gap; it does not replace them. A brush deals with the loose particles that nitrogen cannot carry away and the dust settled on top of the bumps. It does not work on strongly adhering organic residue, which needs solvent or plasma. When assessing, first confirm what kind of residue gap the existing flow leaves.
Can it be used on conventional C4 bumps?
The risk is relatively low, but strict control is still required. The direction worth discussing is extremely soft filament (diameter below 70 μm), a linear sweep and low contact pressure. But this is a starting point for discussion, not a conclusion: there is no clear data in publicly available information on the relationship between filament tip pressure and the shear tolerance of solder bumps, so it must be verified on actual wafers.
What information should I prepare before an assessment?
Five items: bump type and dimensions, nature and source of the contamination, the existing cleaning flow and its residue gap, mechanical parameters (down force / linear speed / quick-release interface), and the verification method and acceptance criteria. Without the first three, any quote is only a guess.
References
- Journal of the Semiconductor & Display Technology
- Academia.edu - Surface cleaning with the CO2 snow jet
Further reading
- How to choose cleanroom brushes: 3 red lines where the class decides the material
- Compounded or surface-coated conductive fiber? Service life and selection explained
- How to choose anti-static brush resistance: a three-grade selection guide
- What to prepare for a custom brush quote: the three spec elements
- Industrial brush application cases
- The complete guide to custom brushes
