When a fab engineer says "AMC is out of spec," they are actually talking about four completely different things — each requiring a different filter.
What Is AMC?
AMC (Airborne Molecular Contamination) is one of the trickiest enemies in semiconductor manufacturing. Unlike particle contamination, AMC is gaseous — no amount of HEPA / ULPA filtering will stop it, because these molecules are three orders of magnitude smaller than HEPA target particles.
An intuitive analogy: particle contamination is dust on a table — you can wipe it off with a cloth (HEPA). AMC is a smell in the air — you need activated carbon (chemical filters) to adsorb it.
In advanced nodes (below 7 nm), AMC tolerances have dropped to ppb (parts per billion) or even ppt (parts per trillion). A single fingerprint's organic volatiles can ruin an entire wafer lot.
SEMI F21's Four Categories
SEMI F21 divides AMC into four categories: MA, MB, MC, MD. This is not an academic taxonomy — it directly maps to different filter strategies, so understanding the categories means understanding selection.
SEMI F21 AMC Four Categories
Airborne Molecular Contamination (AMC) is classified into 4 categories by chemical nature — each has different sources, damage mechanisms, and filter solutions
| Code | Name | Example species | Common sources | Semiconductor impact |
|---|---|---|---|---|
| MA | Acids | HF, HCl, HNO₃, H₂SO₄, SOₓ, NOₓ | Process exhaust return, outdoor SOₓ/NOₓ, cleaning agents | Corrodes metal lines (Cu, Al), damages photoresist |
| MB | Bases | NH₃, NMP, TMAH, amines | Photoresist developer (TMAH), cleaning solutions, concrete, humans | T-topping (resist swelling), CD shift, yield loss |
| MC | Condensables | DOP, DBP, siloxanes, high-boiling VOCs | Sealants, plastic housings, lubricants, building materials | Forms thin film on wafer surface, affects etch uniformity |
| MD | Dopants | B (Boron), P (Phosphorus), As (Arsenic) compounds | Glass fiber (B₂O₃), HEPA filter media itself, adjacent furnace exhaust | Unintended doping → threshold voltage shift → device failure |
SEMI F21 defines the classification framework, not specific concentration limits. Each fab sets allowable limits (usually in ppb) per category based on its process node and yield targets. For example, advanced EUV processes may require MA < 0.1 ppb, while mature nodes may allow < 10 ppb.
| Category | Code | Representative Substances | Plain English |
|---|---|---|---|
| Acids | MA | HF, HCl, SOx, NOx | "Acid gases" that corrode metal interconnects |
| Bases | MB | NH₃, NMP, TMAH, amines | "Basic gases" that degrade photoresist |
| Condensables | MC | Organosiloxanes, DOP, plasticizer vapors | "Organic condensables" that form films on wafers |
| Dopants | MD | Boron (B), phosphorus (P) compound vapors | "Dopants" that alter semiconductor electrical properties |
Sources, Damage, and Impact by Category
MA (Acids)
Sources: Etch process exhaust backflow, outdoor SOx/NOx, cleaning-solution vapors
Damage: Corrodes copper/aluminum interconnects, attacks chrome films on reticles, spikes contact resistance
Real fab scenario: The CMP zone in a copper process emits HF-containing exhaust. If the MAU (makeup air unit) chemical filter fails, HF enters the litho bay through HVAC, etching the reticle chrome layer. At ppb levels, it goes unnoticed for days — typically discovered only when yield suddenly drops.
MB (Bases)
Sources: Photoresist developer (TMAH) vapor, ammonia in cleaning agents, amines exhaled by personnel
Damage: Neutralizes the acid-triggered reaction in chemically amplified resists (CAR), causing T-topping (line-width swelling at the top) — lethal at EUV's ultra-narrow features
Real fab scenario: Litho-bay TMAH tolerance is already < 1 ppb. An engineer entering the litho area with gloves but no face mask can emit enough trace ammonia to cause nearby wafer exposures to fail.
For more on NMP, TMAH, and boron control, see AMC NMP / TMAH / Boron Control.
MC (Condensables)
Sources: Sealants, plastic tubing, HEPA frame off-gassing, cosmetics/hand cream used by personnel
Damage: Condenses into nanometer-scale organic films on wafer surfaces, disrupting gate-oxide growth uniformity
Real fab scenario: A fab installed a new batch of HEPA frames with a different sealant brand. Three days later, furnace-zone gate-oxide thickness started drifting — root cause: the new sealant's organosiloxane off-gassing was 5× higher than the old brand.
This is why the [500°C heat-resistant HEPA outgassing test](/en/news/muki-500c-hepa-outgassing-test/) matters so much — glass-fiber HEPA itself is a potential MC / MD source.
MD (Dopants)
Sources: Borosilicate glass fiber (HEPA media itself), boron-containing cleaners, phosphoric-acid etch vapors
Damage: Boron/phosphorus deposits on the wafer surface directly shift MOSFET threshold voltage — the stealthiest contamination
Real fab scenario: Traditional HEPA uses borosilicate glass fiber as media. When used upstream of high-temperature furnaces (800°C+), the fiber itself releases trace boron vapor. This is why furnace zones have universally switched to boron-free glass fiber or PTFE membrane HEPA / ULPA.
Filter Strategy by Category
No single filter handles all four AMC types, because the adsorption/reaction mechanisms are entirely different.
AMC Category × Filter Strategy
Different AMC types require different chemical media — no single filter handles all four categories
| AMC Category | Recommended media | Removal mechanism | Selection focus |
|---|---|---|---|
| MA | Chemically impregnated activated carbon (acid-specific) + KOH / Na₂CO₃ | Chemisorption (neutralization) | Note: HF requires a dedicated formulation; general impregnated carbon is poor for HF |
| MB | Phosphoric / citric acid-impregnated activated carbon | Chemisorption (acid-base neutralization) | High NH₃ fluctuation needs more carbon; TMAH is a large molecule — match pore size |
| MC | Virgin (non-impregnated) activated carbon or synthetic sorbent | Physical adsorption (Van der Waals forces) | Siloxanes adsorb irreversibly — carbon life is short, frequent replacement needed |
| MD | PTFE-media HEPA (boron-free) + impregnated carbon (boron-specific) | Prevent HEPA from releasing boron + chemisorb boron compounds | Advanced nodes (< 7 nm) have almost universally switched to PTFE HEPA over fiberglass |
In practice, semiconductor fabs install multiple chemical filter stages in the MAU/AHU sequence, addressing MA → MB → MC → MD in order. Advanced fabs also add a final chemical filter stage before the FFU as a "last line of defense."
| AMC Category | Filter Strategy | Media / Adsorbent |
|---|---|---|
| MA (Acids) | Chemical filter (base-impregnated) | Activated carbon impregnated with KOH or Na₂CO₃ |
| MB (Bases) | Chemical filter (acid-impregnated) | Activated carbon impregnated with H₃PO₄ or citric acid |
| MC (Organics) | Chemical filter (high-surface-area carbon) | Virgin or specialty-impregnated activated carbon |
| MD (Dopants) | Specialty chemical filter + HEPA source control | Impregnated carbon + boron-free HEPA |
Note: MA uses base-impregnated carbon; MB uses acid-impregnated carbon — swapping them makes things worse. This is why you cannot evaluate a chemical filter by just asking "does it have activated carbon" — you must ask what it is impregnated with.
For more on chemical filter selection logic, see AMC Chemical Filter Deep Dive.
Why One Filter Cannot Handle Everything
Four reasons:
- 1MA and MB adsorbents are mutually exclusive: Acid gases need base-impregnated carbon; basic gases need acid-impregnated carbon. One carbon bed cannot be both acidic and basic
- 2MC needs large-pore surface area: Organic condensable molecules are larger, requiring meso-/macro-pore carbon structures different from the micro-pore carbon used for MA/MB
- 3MD is not solved by chemical filters alone: Dopant control starts at the HEPA itself (boron-free fiber / PTFE membrane), not by adding a chemical filter downstream
- 4Lifetimes differ widely: MA/MB chemical filter life depends on impregnation loading and challenge concentration; MC life depends on carbon weight; MD life depends on HEPA fiber type. Replacement cycles differ for all four
Most advanced fabs install 2–4 stages of differently formulated chemical filters along the MAU → FFU path, each targeting a different AMC category.
Advanced-Node Special Requirements
In EUV lithography (below 7 nm) and high-k dielectric processes, AMC control enters another order of magnitude:
- ▸MB tolerance < 0.1 ppb (vs < 10 ppb in legacy nodes — a 100× tightening)
- ▸MC tolerance < 1 ppb
- ▸MD (boron) tolerance < 0.01 ppb
These requirements mean:
- ▸Chemical filter replacement frequency increases dramatically (from every 6 months → every quarter or even monthly)
- ▸Monitoring must be real-time online (weekly grab-sample analysis is too slow)
- ▸HEPA media selection shifts from "is it boron-free" to "every single component's off-gassing must be controlled"
For chemical filter efficiency testing standards, see ASHRAE 145.2 Chemical Filter Test to understand how to read breakthrough curves.
FAQ
Q: What is the relationship between SEMI F21 and ASHRAE 145.2?
A: SEMI F21 defines "what to filter" (AMC classification and tolerance levels). ASHRAE 145.2 defines "how to test the filter" (chemical filter breakthrough-curve test method). They are complementary — use F21 to determine which AMC categories your fab must control and at what concentrations, then use 145.2's method to verify your selected chemical filter can last long enough.
Q: My fab runs 28 nm — do I need this level of detail?
A: 28 nm AMC tolerances are much more relaxed than 7 nm (typically MA/MB at 10–50 ppb), but you still need to classify. In practice, a 28 nm fab needs at least separate MA + MB chemical filters. MC and MD can be decided based on environment. If your fab has furnace processes (oxidation, diffusion), MD control remains important.
Q: Do MA and MB neutralize each other if both are present?
A: In theory, acids and bases neutralize. But at ppb-level concentrations, neutralization reaction rates and completeness are very low. Moreover, the neutralization products (salt particles) become particle contamination. You cannot skip either chemical filter by counting on mutual neutralization.
Q: How often should chemical filters be replaced?
A: No universal answer — it depends on challenge concentration, filter capacity, and allowable breakthrough. General guideline: advanced fabs (below 7 nm) every 3–6 months, mature nodes (28 nm and above) every 6–12 months. The most accurate approach is to predict lifetime using ASHRAE 145.2 breakthrough curves, combined with real-time online monitoring.
Q: How is AMC concentration measured?
A: By category: MA/MB commonly use ion chromatography (IC) or real-time gas monitors (e.g. ppb-level NH₃ detectors); MC uses GC-MS or FID; MD (boron) uses ICP-MS. Advanced fabs have deployed real-time monitoring points at MAU outlets, FFU outlets, and tool inlets — offline grab sampling is no longer sufficient.
