Why AMC Needs a Whole Enclosure
HEPA filters reach 99.97% or better on particles at 0.1–0.3 μm, but they are almost useless against gaseous molecules. HCl has a kinetic diameter of about 0.3 nm and NH₃ about 0.26 nm — three orders of magnitude below a HEPA's most penetrating particle size. To the filter they effectively do not exist, and they pass straight through.
These molecular contaminants are collectively called AMC (Airborne Molecular Contamination). What they cause is not particle defects but photoresist T-topping, metal line corrosion, wafer surface hazing and dopant concentration drift — the class of problem where no particles can be found yet yield falls. For the species involved, the damage mechanisms and the concentration grading, The Invisible Killer: How Semiconductor Fabs Fight AMC (Airborne Molecular Contamination) and SEMI F21 AMC Classification: The Four Distinct Problems Behind "AMC" in a Semiconductor Fab already cover the ground. This article is about the equipment.
The only way to remove molecules is adsorption, and adsorption needs time — the molecule has to stay in the media pore long enough to be captured. That single requirement dictates the entire design of a chemical filter housing:
- ▸There must be enough media volume to give enough residence time
- ▸There must be a particulate filter upstream, or dust seals the pore openings of the activated carbon and the media fails before it is ever loaded
- ▸There must be a particulate filter downstream too, to catch the carbon fines abraded off the media by the airstream — otherwise the chemical filter becomes a particle source itself
- ▸There must be a fan able to overcome all three stages
Integrate those four things into a shell whose media can be replaced on a schedule, and you have a chemical filter housing. What separates it from "a chemical filter panel" is that it is a system that runs, needs maintenance, and has a pressure and airflow budget.
Three Airflow Arrangements: Recirculation, Once-Through, Standalone
The first cut in equipment selection is not airflow — it is the arrangement: where the air comes from and where it goes after treatment. Get this cut wrong and no amount of precision in the airflow and media calculations will save it.
Fig. 1: The Three Arrangements Differ Only in Where Air Enters and Leaves
The same media achieves very different results depending on the arrangement it sits in
Recirculation and standalone units are both closed loops, so concentration settles at an equilibrium set by emission rate ÷ airflow — more airflow only approaches it, never reaches zero. Only once-through can actually determine the concentration entering the room, because the source is outside and gets stopped ahead of the media. Where a plant has both acids arriving with the outside air and bases emitted indoors, these two arrangements are complementary rather than either/or.
Recirculation: the fastest remedy, but it only dilutes
The unit sits inside the process area, draws room air, and returns it treated to the same space. Because it does not touch outside air and does not change the room's air balance, it leaves the existing pressure regime alone — this is the fastest fix to deploy when an existing plant measures AMC out of specification, with no ductwork changes and no shutdown.
The cost is that it dilutes rather than blocks. With a continuous indoor source (developer evaporation, floor wax, ammonia carried in on cosmetics), concentration settles at an equilibrium set by emission rate ÷ airflow. More airflow pushes that equilibrium down but never to zero — and the relationship is not linear, so marginal returns fall away quickly at the top end.
Once-through (positive pressurization): the only arrangement that sets the inlet concentration
Treats outside air and supplies it one way into the room while holding positive pressure. This is the only one of the three that genuinely lowers the concentration entering the room, because the source is outside and gets stopped ahead of the media.
When it applies is clear-cut: the outside air itself carries the contamination. Next to a chemical plant or another fab's exhaust stack, coastal sites (salt spray and SOx), urban traffic emissions (NOx), or a plant whose own exhaust stack sits too close to its intake and short-circuits. In these cases recirculation is wasted effort — the outside keeps pouring in faster than any internal loop can catch up.
It is also the most expensive, because the volume to be treated is the entire makeup air quantity; airflow and energy use are the highest of the three.
Standalone air cleaner: no ductwork, at the cost of coverage
Essentially recirculation, but built as a floor-standing unit on castors — no ducting, just power. The advantage is mobility: roll it to whichever area measures out of specification, and use it to reinforce a local hot spot (near one particular tool, or a chemical change-out point) without touching the plant HVAC.
The limit is coverage. It relies on natural room mixing to spread the treated air, so the effect weakens with distance and the real effective range depends on the room's airflow organisation. In a cleanroom with strong unidirectional flow, a badly placed unit may do nothing at all, or even disturb the existing flow pattern.
In practice these are used together
Rarely does a site pick just one. The typical configuration is: once-through handles the acids arriving with the outside air, recirculation handles the bases emitted indoors, and standalone units reinforce local hot spots. They address different sources, so they are complementary rather than either/or.
The way to decide which to add is to measure inlet and room concentration at the same time: inlet already out of specification → once-through; inlet fine but room out of specification → recirculation; overall fine but one area out of specification → standalone.
Tying In to Process Equipment
"Install it next to the tool" can mean four completely different things, differing by two orders of magnitude in airflow — and with opposite requirements on positive versus negative pressure.
Fig. 2: Four Tie-In Points, Each With Different Airflow and Pressure Requirements
"Next to the tool" can mean four different things — and the tie-in point changes the answer by two orders of magnitude
| Tie-in point | How it connects | Typical airflow | Pressure requirement | Where treated air goes | Matching model |
|---|---|---|---|---|---|
| ① Local point exhaust | Capture hood over a chemical bath or wet process tank, short duct to a drum scrubber | 170 – 1,700 CMH | Negative-pressure extraction | Discharged outdoors | BS-DS |
| ② Process tool exhaust | Downstream of the tool exhaust port, between it and the facility exhaust header | 510 – 18,350 CMH | Negative; must match the tool exhaust pressure | Discharged outdoors | BS-DBS / BS-HDBS |
| ③ Room-level recirculation / makeup | Floor-standing unit inside the process area, or short duct to the room supply/return grille | 850 – 6,860 CMH | Holds room positive pressure | Returned to the same clean zone | BS-APS / BS-PPS |
| ④ Facility supply header | In series on the MAU/AHU supply section, after the medium filter and before the terminal HEPA | 1,700 – 68,000 CMH | Positive, supply side | Sent to clean zones plant-wide | BS-SAH |
The tie-in point decides more than airflow — it also decides where the treated gas goes. Local and process exhaust are always negative-pressure extraction, treated and discharged outdoors; they must never be returned indoors. Room-level and facility header tie-ins are on the supply side, where treated air goes back into the clean zone. Swapping these two is the most serious design error possible — it means feeding high-concentration process exhaust into the cleanroom.
The critical point is that the supply side and the exhaust side must never be mixed up. ① and ② are exhaust side: negative-pressure extraction, treated and discharged outdoors, handling gas orders of magnitude more concentrated than room air. ③ and ④ are supply side: positive pressure, with treated air returning to the clean zone. Fitting an exhaust-side model on the supply side means feeding high-concentration process exhaust into the cleanroom — the most serious design error possible, and because the media sits in between, it may not be obvious for some time.
Four things to confirm with the equipment vendor before tying in
- 1Exhaust volume and available pressure — the tool exhaust port's design airflow, and how much extra downstream pressure drop it will tolerate. A chemical filter housing readily runs 800–2,500 Pa, which many tool exhaust fans cannot sustain; a booster fan is often needed.
- 2Flange size and type — round or rectangular, and the bolt pitch. A rectangular-to-round transition is easy to fabricate, but it has to be in the overall dimensions from the start, or the space turns out to be short on site.
- 3Gas composition, concentration, temperature and humidity — these determine the media formulation, and acid/base mixtures additionally need layer separation. High temperature reduces adsorption capacity, and above 70% relative humidity the organic adsorption capacity of activated carbon drops noticeably as water competes for adsorption sites.
- 4Interlock signalling — whether the tool should interlock to a stop, or raise an alarm, when the housing fan stops. This is almost always remembered the week before acceptance, by which time the control panel is already built.
Sizing the Airflow
There are three sizing methods for three different purposes. Using the wrong one is off by several times — the most common mistake at the selection stage.
Fig. 3: Three Ways to Size Airflow — the Wrong One Is Off by Several Times
First decide whether the goal is lowering concentration, countering a known source, or holding positive pressure
Recirculation or standalone units: no defined source, just bring the overall concentration down
Airflow Q = room volume V × air changes per hour (ACH)
- Litho bay 12 m × 8 m × 3 m = 288 m³
- Take 8 ACH for AMC control (strict litho areas can exceed 15 ACH)
- Q = 288 × 8 = 2,304 CMH
- → Select BS-APS-2000V (3,398 CMH) for margin
ACH is a rule of thumb, not a guarantee. Applying it without ever measuring concentration only proves air is moving, not that the target is met.
Emission rate measured: answers how much airflow it takes to reach a target concentration
Airflow Q = emission rate G ÷ (target concentration − inlet concentration)
- NMP emission rate G = 40 mg/h
- Target indoor concentration 0.05 mg/m³, inlet 0.01 mg/m³
- Q = 40 ÷ (0.05 − 0.01) = 1,000 CMH
- → Select BS-APS-1000V (1,699 CMH)
The denominator is target minus inlet. If inlet already sits near the target, the denominator approaches zero and required airflow approaches infinity — at that point the fix is switching to once-through to lower the inlet, not adding airflow.
Once-through positive pressurization: airflow follows the room’s air balance, not concentration
Airflow Q = (process exhaust + room leakage) × (1 + positive-pressure margin)
- Process exhaust 2,000 CMH, door and gap leakage 400 CMH
- Take a 12% positive-pressure margin
- Q = (2,000 + 400) × 1.12 = 2,688 CMH
- → Select BS-PPS-1000V (3,500 CMH)
Sum the exhaust of every tool, including standby tools that are normally off. Miss one and the room cannot hold positive pressure.
AMC concentrations are usually quoted in ppb, so convert to mass concentration before using the dilution method: at 25°C and 1 atm, C(mg/m³) = C(ppm) × molecular weight ÷ 24.45. For NH₃ (MW 17), 1 ppm ≈ 0.695 mg/m³ and 1 ppb ≈ 0.695 µg/m³. Whichever method you use, check afterwards that face velocity falls inside the model’s design range — the right airflow forced into an undersized housing still fails, because residence time comes up short.
Whichever method is used, finish with a face velocity check: divide the selected airflow by the model's face area and confirm the velocity falls inside the design range. The point of this step is that the right airflow does not guarantee sufficient residence time — force a large airflow into a small housing and face velocity climbs, molecules are swept out of the media before they can be captured, and concentration still will not come down.
Rough boundaries for residence time: at ppb inlet concentration, modules are enough (0.06–0.19 s); at the ppm range, consider deep beds (0.8 s and up); process exhaust in the tens of ppm needs multiple deep beds. Face velocity and residence time for each model are tabulated on the AMC Chemical Filter Housing product page.
Select the fan against the final pressure drop, not the clean value. The changeout ΔP is generally set at 1.5–2 times the clean value, and the fan static pressure must still deliver design airflow at that point — otherwise airflow falls below design partway through the filter's life, and it does so silently.
Installation Points
What installation problems have in common is that none of them can be fixed by later adjustment. Once the duct elbow is welded and the unit is hoisted upstairs, any change means taking the equipment apart again.
Fig. 4: Six Installation Errors That Mean Rework Once the Unit Arrives
All of these should be caught at the drawing stage — fixing them on site costs far more than checking beforehand
| Common error | What happens | Correct practice |
|---|---|---|
| Not enough service clearance | Side-access doors will not open; changing media means pulling the whole unit out first | Leave 800 mm or more at the front for the doors to swing fully open, and 300 mm or more at each side and the rear |
| Elbow butted against the inlet or outlet | Airflow skews, so one side of the media saturates while the other is barely used — effective life is cut short | Leave at least 1.5 duct diameters of straight run before and after the inlet and outlet |
| Ducting hard-connected to the housing | Fan vibration travels down the duct into the structure, causing noise and fatigue cracks | Fit a flexible connector between them, and confirm its material resists the gas being treated |
| ΔP gauge tapped at the wrong points | It spans only the pre-filter or the whole unit, so the real chemical stage ΔP cannot be read | Tap across the chemical media stage, with a separate set for the pre-filter |
| Housing corrosion-protected but ducting is not | When handling acids, the ducting corrodes through first and the leak is outside the housing | Confirm the corrosion rating of ducting, exhaust outlet, flexible connector and hangers together |
| Supply phase sequence and voltage unverified | The fan runs backwards, airflow drops to about a third, and the ΔP reading still looks plausible | Confirm phase sequence before energising, and measure actual airflow at commissioning rather than just checking the motor turns |
The first four are about airflow and space and are visible on the drawings; the last two are about materials and electrical and need to be confirmed before ordering. What they share is that none can be fixed by later adjustment — the duct elbow is already welded, the unit is already hoisted upstairs, and any change means taking the equipment apart again.
The most underrated of these is service clearance. A chemical filter housing is not install-and-forget equipment; its media is a consumable, and the pre-filter may need changing every few months. Three hundred millimetres omitted at the design stage costs several times the labour at every media change thereafter, and sometimes a crane.
Maintenance: the Failure a ΔP Gauge Cannot See
The most counter-intuitive thing about a chemical filter housing is that chemical media saturation does not show up on ΔP.
A loaded particulate filter raises pressure drop, because accumulated dust narrows the flow path. But adsorption saturation just means the pores are occupied by molecules — flow resistance barely changes. The ΔP gauge reads normal the whole way, until someone measures outlet concentration and discovers the media failed long ago.
Fig. 5: Reading ΔP Against Outlet Concentration (Media Saturation Never Shows on ΔP)
Watching only the ΔP gauge misses the most important failure mode
The particulate filter is dust-loaded; the chemical media is still working
→ Change the pre-filter only; leave the chemical media alone
Pre-filter loaded and chemical media saturated — usually run past its interval
→ Replace pre-filter and chemical media together, and review the change interval
Both signals normal
→ Keep logging ΔP on schedule and sampling periodically
ΔP looks normal but adsorption is already saturated — the quadrant most often misread
→ Replace the chemical media. If this recurs often, residence time or media quantity was under-designed
The top-right quadrant is the dangerous one: the ΔP gauge reads normal, the logs look fine, but the chemical media is already loaded and outlet concentration is climbing. This is why a chemical filter housing cannot inherit the maintenance plan written for particulate filters — catching this quadrant requires periodic outlet sampling, or sending media out to have its remaining adsorption capacity measured.
So the maintenance plan for a chemical filter housing cannot be copied from the particulate filter playbook. In practice:
- ▸The ΔP gauge governs the upstream and downstream particulate filters; here the pressure trend is read exactly as for any filter
- ▸Outlet concentration sampling governs the chemical media, either sent out periodically or tracked with online instrumentation. For the trade-offs between monitoring technologies, see AMC Monitoring Technology Compared: IMS, SAW, CRDS, GC-MS — Which One to Use?
- ▸Residual capacity testing calibrates the changeout interval. Sampling the media at the first changeout verifies whether the estimated life was accurate, giving the later intervals a real basis
Chemical media life cannot be inferred from a calendar — it depends on inlet concentration and cumulative treated airflow. The same media set installed at two plants whose inlet concentrations differ tenfold will have lifetimes that differ tenfold. On the cost side, Chemical Filter Lifecycle & TCO: From Breakthrough Prediction to Optimal Replacement Strategy and Regenerable Chemical Filters: A Complete Guide to Regeneration vs. Disposable Options discuss where the break-even between disposable and regenerable media falls.
Summary
The selection order for a chemical filter housing is fixed: arrangement (where air enters and leaves) → airflow (with the right method) → media and residence time → pressure drop and fan. The usual consequence of reversing it is picking a model to suit existing ductwork, then finding residence time is short, deepening the media, watching pressure drop climb, and being forced to reselect the fan and re-run the power feed.
The trade-off between the three arrangements compresses into one line: recirculation treats indoor emission, once-through treats what the outside air brings in, standalone units treat local hot spots. Deciding which to use means measuring inlet and room concentration, not going on intuition.
On maintenance, remember the counter-intuitive part: a ΔP gauge covers the particulate filters and cannot see the chemical media. Plants that watch only ΔP find out long after the media has failed.
