Acidic gas-phase contamination occupies an awkward position in a fab. It does not produce visible particle defects; it produces metal line corrosion, resist T-topping, wafer surface hazing and dopant concentration drift — the class of problem where no particle count explains why yield dropped. By the time the measurements point somewhere, several lots have usually already gone through.
The reflex answer is an activated carbon chemical filter. That is not wrong, but install both media in the same airstream, run them for two full years and analyse them on removal, and something counter-intuitive shows up: what decides the outcome is rarely how fast the medium captures. It is whether capacity was spent on things that were never the target.
AMC has no critical concentration, only a dose
One sentence in the 2020 Yield Enhancement chapter of the International Roadmap for Devices and Systems deserves to go on the wall: there is no certain critical contamination level for any given AMC species — concentration always has to be read together with contact time, and what ultimately matters is the dose, concentration multiplied by time (IRDS 2020 Yield Enhancement).
That overturns a very common line of reasoning on the floor: "we measure 0.05 ppb, the spec is 0.1 ppb, so we are fine." The conclusion only holds if contact time also falls inside the assumed range. The same reticle sitting in a stocker for three days and for three months accumulates entirely different surface loads at identical ambient concentration.
And because it is a dose, the value of a filter lies not in how high its efficiency reads when new, but in how many molecules get through in total across the whole change-out interval. A filter at 98% when new that falls to 60% within six months, and one that holds 85% for three years, can differ by an order of magnitude in cumulative penetration — and the second one has the less impressive datasheet.
This is also why chemical filter life and TCO have to be calculated from breakthrough curves rather than from the initial efficiency printed in a catalogue.
Which class acidic AMC belongs to, and where it comes from
Chart 2: Four AMC Classes (per SEMI F21)
Industry-standard classification. Each class damages the process in a different way.
| Class | Common species | Primary process damage |
|---|---|---|
| Acids (MA) | HCl, HF, H₂SO₄, NOx, SOx | Metal line corrosion, wafer surface oxidation, copper tarnishing |
| Bases (MB) | NH₃, Me₃N, NMP | Photoresist T-top (DUV resist surface failure) |
| Condensables (MC) | BHT, NMP, DOP (bp > 150°C) | Wafer surface hazing, optics contamination |
| Dopants (MD) | AsH₃, B₂H₆, BF₃, TEP | Alters dopant concentration, device parameter drift |
Concentration graded in ppt (parts per trillion). MA-1 = 1 ppt level, MA-10,000 = 10,000 ppt. Lower class numbers mean stricter requirements.
Molecular acids are one of the four AMC classes, a scheme formally adopted into the roadmap by ITRS in 2007 and still in use (details in the four SEMI F21 classes). IRDS later proposed reclassifying by chemical family, on the grounds that the old scheme grew out of industry jargon rather than rigour — "refractory", for instance, describes heat- and wear-resistant metals in metallurgy, but was borrowed in AMC work to mean substances that change the refractive index of optical elements.
On sources, the IRDS list is worth reading in full: outgassing from cleanroom construction materials, process equipment, processed wafers, wafer pods and environmental enclosures, inadequate exhaust and fugitive emissions from process chemicals — and the people in the room. Make-up air is a significant source too. The FOUP loses both ways: it outgasses itself, and it traps AMC carried out of steps such as etch and releases it slowly.
For acids there is one category that gets overlooked: acid precursors. Not acids themselves, but substances that become acids on a surface once they take up water.
| Precursor | What happens on the surface | How active |
|---|---|---|
| Sulphur dioxide SO₂ | Forms sulphurous acid with water, then oxidises to sulphuric acid | Most active; fast and strongly acidic |
| Nitrogen dioxide NO₂ | Forms nitrous and nitric acid with water | Less efficient, but continuous |
| PGMEA (ester solvent) | Hydrolyses into PGME and acetic acid | Latent; may only act once inside the filter |
Conversely, hydrogen sulphide and the nitrogen oxides are not acids themselves. Counting them into a "total acids" figure blurs the judgement: the question that matters is which specific acid this process, on this tool, is vulnerable to — not how many anions an impinger can be made to show.
The three mechanisms that hold a molecule
Chart 4: Three Sorption Mechanisms of Chemical Filters
Chemical filters capture AMC through physisorption, chemisorption, and ion exchange
Plain activated carbon relies mainly on physisorption, which weakens under high humidity. For acidic/basic AMC, impregnated carbon (with KOH/K₂CO₃/H₃PO₄) is used — chemical reaction locks molecules permanently.
A chemical filter has only three routes for capturing molecules: physical adsorption, chemisorption via an impregnated agent, and ion exchange. The difference is not which is strongest, but selectivity and reversibility.
Physical adsorption relies on pores and van der Waals forces and turns nothing away. Chemisorption relies on the agent coated on the pore walls, and its selectivity depends on the formulation (covered in an introduction to chemical filter impregnation). Ion exchange relies on functional groups bonded to a polymer backbone, and reacts only with ions of the opposite charge.
The three inevitabilities of activated carbon
Activated carbon is a microporous carbon material, usually made from natural organic feedstock or coal, and not a clean, inert graphite structure. Impregnation only loads an agent into the pores; the pores remain pores. Three side effects follow that cannot be designed away:
- ▸Physical adsorption always happens — volatile organics and esters will be taken in, and they occupy capacity
- ▸Redox reactions always happen at the surface — nitrogen oxides get converted and stored
- ▸Adsorption is reversible — incoming strong acids displace weak acids already held, most visibly as pH falls
None of these is obvious in a laboratory test. All three show up after two years in the field. They are not quality problems; they are the price of the mechanism.
Ion exchange resin: selectivity built into the material
Ion exchange resin is a porous polystyrene/divinylbenzene polymer with functional groups bonded at the chain ends. Acid-removal grades carry anionic groups — the more common product on the market is actually the cation type for base removal, aimed at ammonia and amines. The backbone itself is clean and inert and there are no adsorption pores, so the characteristics are a mirror image of carbon: esters and VOCs are not physically adsorbed, nitrogen oxides are not stored, and acids are chemically bonded rather than held loosely, so changing conditions do not release them back into the airstream.
How each protection target shifts on ion exchange media
Relative performance (0–100) of the same targets on two media types
Acid targets move up while every non-target drops to the floor — by design, not by defect. Activated carbon pores take in everything, so capacity is shared with organics and nitrogen oxides; on ion exchange resin only the anionic functional groups react, leaving the whole capacity for acids. The trade-off is that ammonia, chlorine, ozone and organophosphate dopants are not addressed at all, so sites that must control those still need carbon alongside.
The price is written on the same chart: every non-target drops to the floor. Chlorine, ozone and other oxidants, ammonia and amines, organophosphate dopants — ion exchange does not address any of them. That is design intent rather than defect, but it means selection has to confirm first that the site has no such requirement. Get it wrong and the result is not slightly worse performance; it is none.
When the filter becomes the source
The failure mode that gets missed most often is the filter turning from a barrier into an emitter.
After five years, is the filter absorbing acetic acid or releasing it?
Coater/developer temperature-humidity inlet filter, acetic acid inlet at 4 ppbv
Activated carbon draws the process solvent PGMEA into its pores, where prolonged heat and humidity hydrolyse it, splitting off acetic acid that is then released downstream — the filter turns from a barrier into a source, measured at a net emission of 4.14 ppbv that attacks the chiller heat exchanger directly. Five hours of clean-air purging only brings the release down to 0.51 ppbv, which means the acetic acid is not surface residue but is still being generated. Ion exchange resin has no adsorption pores and does not take up esters, so there is nothing stored to cleave; after five years it is still removing.
If the inlet filter on the temperature and humidity control section of a coater/developer is carbon-based, years of taking up PGMEA from the lithography area end with heat and humidity hydrolysing it, cleaving off acetic acid and releasing it downstream. At that point the filter's outlet concentration is higher than its inlet — it is manufacturing the contaminant.
The awkward part in the field is that a differential pressure gauge sees none of it. Pressure drop reflects particulate blockage only and does not change because of a chemical reaction. Catching it needs outlet sampling, which is why online AMC monitoring is increasingly standard in advanced fabs.
Ion exchange resin has no adsorption pores and does not take up esters, so there is nothing stored that could be cleaved. That brings a maintenance benefit: a sudden rise in outlet concentration is always saturation or bypass leakage, never the medium giving back what it held. One branch of the fault tree disappears.
Two fronts: source and path
Chart 2: Two Fronts of AMC Control — Source Reduction × Path Interception
Either alone leaks. A complete AMC strategy fights on both fronts at once
Source control: cut emission at origin
- ▸Low-outgassing materials, paints, floor adhesive
- ▸Local exhaust on process tools
- ▸FOUP / gloves / packaging screening
- ▸Operator workflow discipline
Path filtration: catch before the wafer
- ▸MAU / OAU chemical filters
- ▸Molecular-sieve sorbents at FFU return
- ▸Mini-environment / point-of-use filtration
- ▸Continuous IMS / GC-MS monitoring + replacement cycle
Filter media must match the target: KOH / Na₂CO₃-impregnated carbon for acids, H₃PO₄ for bases, un-impregnated high-surface-area carbon for organics.
IRDS describes AMC control as two things running in parallel: preventing AMC release, for which tight source control is mandatory, and continuous removal through air recirculation. The order cannot be reversed — whatever can be stopped at source should not be left to the filter.
Permeation through tubing and acid generated as media age are both source problems, and a new filter will not solve either. It is worth asking once in every selection meeting: is this acid coming in from outside, or are we producing it ourselves?
Selection: when to use which
| Site condition | Recommendation | Why |
|---|---|---|
| Strong acids only (HF, HCl, HNO₃, SO₂) | Ion exchange | All capacity goes to acids; better life and peak resilience |
| VOCs, esters and condensables also in scope | Activated carbon | Physical adsorption is required; ion exchange has no such mechanism |
| Ammonia and amines in scope | Acid-impregnated carbon or cation-type resin | Anionic groups do not react with bases |
| Chlorine, ozone or other oxidants present | Activated carbon | Ion exchange has no corresponding mechanism |
| Outdoor-air section, construction-phase diesel fumes | Activated carbon, including treated grades | The load at this stage is mostly organics and nitrogen oxides |
| Final barrier close to the process (FFU, tool inlet, stocker) | Ion exchange | Narrow target, long life and low secondary contamination required |
In practice most plants combine the two by zone and by stage rather than choosing one: carbon at the outdoor-air and exhaust sections for organics and oxidants, and an ion exchange chemical filter close to the process as the last barrier against acids. Either one alone leaves a gap.
For fuller background on AMC damage and case histories, see how a bottle of ammonia can ruin a wafer.
Frequently asked questions
Q: Can ion exchange filters replace activated carbon entirely?
No, and this is the first thing to confirm when selecting. Ion exchange handles acids only; it does nothing for ammonia and amines, chlorine, ozone, volatile organics or organophosphate dopants. If the site has those requirements too, the two have to be staged together. Read the other way, if the target really is strong acids, using carbon means spending most of the capacity on things that are beside the point.
Q: Can a differential pressure gauge tell me when to change a chemical filter?
Only half of it. Rising pressure means the upstream particulate filter is loading with dust, and that is what needs changing. But chemical media saturation does not show up in pressure drop at all — a flat pressure reading with a rising outlet concentration is what failure of the chemical stage looks like. This is the single most common misreading on the floor. Pinning down the change-out point needs periodic outlet sampling, or sending media out for residual capacity testing.
Q: Is a "total acids" specification meaningful?
Only to a limited extent. The IRDS position is that AMC has to be judged as a dose rather than a single concentration figure, and most acid-related defect risks are tied to a particular process step or a particular tool. Writing a specification around total acids tends to buy a filter that is average against every acid and outstanding against none. The more useful question is: which acid is this tool most vulnerable to, at what target concentration, and over what contact time?
Q: Will ion exchange media give acids back the way carbon does?
Mechanically, no. Physical adsorption on carbon is reversible, and incoming strong acids displace weak acids captured earlier. On ion exchange the acid forms a chemical bond with the functional group and is not released back into the airstream by shifts in temperature, humidity or pH. The practical consequence is that when outlet concentration rises abnormally, "the medium is giving it back" can be ruled out immediately, leaving only saturation and bypass to check.
Q: Should the make-up air unit and the FFUs use the same chemical medium?
Usually not. The outdoor-air section faces volatile organics, nitrogen oxides and seasonal sulphur dioxide — a complex, high-volume load that suits activated carbon. FFUs sit close to the process, where the target is narrow and the requirement is long life with no secondary emission, which favours ion exchange. Running one medium throughout the plant means paying for it at one end or the other: either short life, or a gap that never gets covered.
