A HEPA filter that needs to operate continuously at 500°C — does it leak anything dirty into its own downstream? That's not a philosophical question. It's a real engineering problem in the semiconductor industry.
Why Run This Test
Heat-resistant HEPA filter applications almost always have "something very sensitive on the downstream side":
- ▸Semiconductor oven exhaust (ASHER, bake oven) — exhaust eventually goes through facility treatment, but if the filter itself releases chemicals at high temperature, those can recirculate or contaminate downstream equipment
- ▸CVD / ALD process exhaust — high-temperature filtration after the process chamber, often connected to scrubbers or facility stacks downstream
- ▸Heat-treatment furnace purification — semiconductor packaging, panel sintering
- ▸Nuclear facility exhaust — high-temperature variants of BIBO systems, where downstream IS the atmosphere
The common thread: the filter is not just blocking dust — its own cleanliness is also part of the spec. That is why the industry runs "outgassing tests" — put the filter in a heated environment, bring it to target temperature, sample upstream and downstream air over time, and see what (if anything) the filter itself contributes.
Test Conditions
NIPPON MUKI ran a complete outgassing test on their ATMWC-20-P-F 500°C heat-resistant HEPA. Filter spec under test:
| Item | Spec |
|---|---|
| Model | ATMWC-20-P-F |
| Dimensions | 610 × 610 × 290 mm |
| Rated airflow | 20 m³/min |
| Rated ΔP | 245 Pa |
| Capture efficiency | 99.97% @ 0.3 μm |
| Continuous max temperature | 500°C |
| Frame | Special stainless steel |
| Media | Glass fiber |
| Sealant | Glass fiber |
| Separator | Special stainless steel |
| Gasket | Glass fiber |
A critical detail: before testing, the filter was thermally pre-conditioned with 10 cycles of 150°C ⇔ 500°C. This simulates the steady-state behavior of a filter that has been in service for a while and has already burned off any short-lived volatiles. Testing a brand-new filter would overstate the steady-state release.
Heating and sampling setup:
Chart 1: 500°C Outgassing Test — Apparatus & Temperature Curve
Filter is first "burned in" with 10 thermal cycles (150°C ⇔ 500°C), then mounted in a heat-resistant tunnel and held at 500°C for 24 h while sampling both upstream and downstream
Sampling: upstream and downstream air each pass through impingers filled with ultrapure water for 24 h, absorbing ions and metals. Ions analyzed by IC (ion chromatography), metals by ICP-MS. Quantification limits range 1–12 μg/m³.
The full test runs ~33 hours (1 h ramp + 24 h hold + 8 h cool-down). The 24 h at 500°C is the actual sampling window. Upstream and downstream air each pass through impingers filled with ultrapure water, capturing gas-phase ions and metals into the liquid. Then:
- ▸Ion Chromatography (IC) — anion / cation analysis
- ▸ICP-MS (Inductively Coupled Plasma Mass Spectrometry) — metals
Result: Only 1 of 8 Species Detected
Full measurement results:
Chart 2: Upstream / Downstream Concentrations of 8 Species
Every ion (acids, organic acids, base) downstream is below the detection limit — the only detected outgas is the metal boron (B), with downstream 783 vs upstream 719, a net 64 μg/m³ release
| Group | Species | LOD | Upstream | Downstream | Outgas (Δ) |
|---|---|---|---|---|---|
| Ions | NO₂⁻ | 1 | <1 | <1 | None |
| NO₃⁻ | 3 | <3 | <3 | None | |
| SO₄²⁻ | 4 | <4 | <4 | None | |
| CH₃COO⁻ | 12 | <12 | <12 | None | |
| CHOO⁻ | 3 | 10 | 6 | None | |
| NH₄⁺ | 2 | <2 | <2 | None | |
| Metals | B(Boron) | 2 | 719 | 783 | +64 |
| P(Phosphorus) | 4 | <4 | <4 | None |
Glass fiber (the standard HEPA media) contains 5–15% boron oxide (B₂O₃) as a glass-modifier that lowers the melting point. At sustained 500°C, B₂O₃ slowly volatilizes into gas-phase boron compounds (HBO₂, B(OH)₃). Other glass constituents — Na, K, Ca, Mg — have far higher melting points and stay locked in. So this report is empirical proof that "boron is the only thing to worry about for fiberglass HEPA at 500°C."
"<" means below the quantification limit (LOD) — not zero, but too low to quantify reliably. CHOO⁻ (formate) shows 10 upstream / 6 downstream — downstream is lower, meaning no outgassing (the filter actually absorbed a trace). NH₄⁺ also <2, so no base release from the glass fiber.
Seven species (six ions plus phosphorus) are all below the quantification limit downstream — effectively no outgassing. CHOO⁻ (formate) is even lower downstream than upstream — the filter actually adsorbed a trace, which is chemically reassuring.
The only detected outgas is metallic boron (B): upstream 719 μg/m³, downstream 783 μg/m³ — a net +64 μg/m³ contribution from the filter.
The high upstream baseline (719 μg/m³) reflects background boron in the test apparatus itself (heat-resistant tunnel, heater elements, lab environment all contain trace boron). The figure that matters is the downstream-minus-upstream delta — that's what the filter is actually contributing.
Why Only Boron?
Not surprising. Let's walk through it.
What's in HEPA glass-fiber media
HEPA/ULPA standard media is "microglass fiber." A typical E-glass formulation, by weight:
- ▸SiO₂ (silica) — 52–56%
- ▸CaO (calcium oxide) — 16–25%
- ▸Al₂O₃ (alumina) — 12–16%
- ▸B₂O₃ (boron oxide) — 5–10% ← the relevant one
- ▸MgO, Na₂O, K₂O — minor
B₂O₃ is a "glass modifier": it lowers the melting point and thermal expansion coefficient, making it possible to draw the glass into very fine fibers (essential for microglass-grade HEPA). Without B₂O₃, sub-1-μm fiber diameter is essentially impossible.
But B₂O₃ "evaporates" at high temperature
Here's the catch. B₂O₃ is a solid component of the glass, but above 450°C it starts to react with ambient water vapor:
- ▸B₂O₃ + 3 H₂O → 2 B(OH)₃ (boric acid, gas phase)
- ▸B₂O₃ + H₂O → 2 HBO₂ (metaboric acid, gas phase)
These gas-phase boron compounds are exactly the +64 μg/m³ detected downstream. The other glass constituents (Na, K, Ca, Mg, Si, Al) all have far higher melting points and stay put.
Plain analogy: B₂O₃ is the most "volatile" component in the glass — like sugar in a crème brûlée; bring up the heat and it's the first thing to smoke. The other ingredients are like starches: they only move at much higher temperatures.
Is 64 μg/m³ a Lot?
Depends on the comparison.
For general industrial exhaust (HVAC, fume scrubbing, lab ventilation), 64 μg/m³ of boron release is completely negligible.
For semiconductor, this number deserves a careful look:
Chart 3: Where Boron Comes From, Why It Matters, How to Control It
Glass-fiber HEPA media releases trace boron at 500°C — in semiconductor P-type doping, that is a ppt-level red line
Semiconductor industry typically requires total airborne boron <100 pg/m³ (picograms) — about 1/10,000 of what this test detected. For high-temperature exhaust applications (ASHER ovens, CVD exhaust, high-temp semicon process exhaust), if a fiberglass HEPA is used, the boron release must be evaluated against any potential recirculation path back to clean areas.
Boron is the most common P-type dopant in silicon semiconductors. Boron atoms in the silicon lattice donate "holes" and change conductivity. A doping concentration as low as 10¹⁵ atoms/cm³ (~50 ppt) is enough to flip the semiconductor type.
This means: for advanced-node fabs, total airborne boron must stay below ~100 pg/m³ (picograms) — about 1/10,000 of the level measured in this test.
In other words: 64 μg/m³ is irrelevant for general use; for an advanced fab cleanroom, it's a "do not let this near the clean area" number.
What This Means for Buyers
The report carries three messages:
Layer 1: Reassurance for engineers
For most non-semiconductor heat-resistant applications (lab ventilation, pharma ovens, food drying, paint-line exhaust), this data confirms: a fiberglass HEPA running continuously at 500°C releases no ionic contamination — no acids (NO₃⁻, SO₄²⁻), no organic acids (CH₃COO⁻, CHOO⁻), no bases (NH₄⁺), no metallic phosphorus. At steady state, the filter is "clean."
Layer 2: Warning for semiconductor users
If you work in front-end semiconductor processes, this data is telling you: fiberglass HEPA cannot be used in high-temperature paths that recirculate to clean areas. Mitigations:
- ▸Switch to PTFE membrane media (no glass, no boron)
- ▸Switch to boron-free fiberglass (E-CR, ECR formulations)
- ▸Add a downstream chemical filter to scrub gas-phase boron
- ▸Demand outgassing reports from suppliers (and check whether burn-in was performed)
Layer 3: A selection mindset
Choosing a heat-resistant HEPA cannot stop at "99.97% efficiency." Self-cleanliness (outgassing grade) matters more than efficiency for some applications. Questions to ask suppliers before purchase:
- 1Does the glass formulation contain B₂O₃?
- 2Has a complete outgassing test been done? Which ions, which metals?
- 3How many burn-in cycles, at what temperature?
- 4Is there a residual-volatiles check before shipment?
If a supplier can't answer, reconsider.
FAQ
Q: Is the upstream baseline of 719 μg/m³ boron normal?
A: It's high but not surprising. The test rig itself (heat-resistant tunnel, heater, ducting) likely contains trace boron (in stainless or insulation), and the lab environment has some background. That's exactly why we focus on the downstream-minus-upstream delta — only that delta is the filter's own contribution.
Q: Why pre-condition with 10 cycles of 150°C ⇔ 500°C before testing?
A: To simulate steady-state behavior after a filter has been in service for a while. A brand-new filter releases the most volatiles on its first heat-up (manufacturing residues, easily-mobilized molecules). Testing a fresh unit would severely overestimate the long-term release. After 10 thermal cycles, short-lived volatiles are mostly gone, and what's measured is "stable long-term release" — far more useful for purchase evaluation.
Q: Why is CHOO⁻ (formate) lower downstream than upstream?
A: The filter is adsorbing a trace of this organic acid. Glass-fiber surfaces have silanol groups (Si-OH) that hydrogen-bond with certain polar organic molecules — physical adsorption. Not the design intent of a HEPA, but this little detail confirms that the filter at 500°C hasn't "broken down" — it still retains some adsorption capacity.
Q: Beyond ATMWC, are there other MUKI heat-resistant HEPA models?
A: NIPPON MUKI's heat-resistant HEPA line is graded by temperature: 250°C, 400°C, 500°C, and 800°C ranges. The ATMWC discussed here is the 500°C grade. Higher temperature ratings need progressively more specialized frame materials (stainless → heat-resistant alloy → ceramic) and cost accordingly. Spec to your maximum operating temperature including transients, not the average.
Q: Is UL900 fire certification the same as this heat-resistance test?
A: No. UL900 tests "flame spread and smoke generation in ambient HVAC duct" — concerned with what happens during a building fire. Heat-resistant HEPA is "designed to operate at high temperature" and lives next to a heat source by design — graded by EN 1822 / JIS efficiency standards plus material temperature rating. Different dimensions of the problem.
Standards & Background
- ▸EN 1822 — HEPA / ULPA efficiency grading (H10–H14, U15–U17)
- ▸ISO 29463 — international equivalent of EN 1822
- ▸JIS B 9901 / B 8330 — Japanese filter test methods
- ▸SEMI F21 — Airborne Molecular Contamination classification, including dopant species
- ▸ASHRAE 145.2 — chemical filter efficiency test method (relevant for "downstream chemical filter" evaluation)
