- Wafer fab specs for filter cartridge metal extractables have been pushed down to ppt (parts per trillion) — over 1000x lower than the metal background of typical environmental water
- ICP-MS (Inductively Coupled Plasma Mass Spectrometry) is the gold standard for metal extractables testing — it can simultaneously detect 70+ elements with detection limits down to ng/L
- For the same 0.05 µm filter cartridge, PP can leach hundreds of ppb of Fe, while PTFE/UPE stays below 0.1 ppb — one material grade difference equals three orders of magnitude in metal cleanliness
- When reading a CoA, don't just look at total metals — the individual values for Na, K, Fe, Cu, Cr, Ni are what determine whether it gets onto the production line
- Pharmaceutical plants follow USP <232> / ICH Q3D, while semiconductor fabs follow the SEMI C series — these two standards differ by an order of magnitude or more and cannot be used interchangeably
- Why are wafer fabs willing to pay a 1000x premium for filter cartridges with ultra-low metal extractables?
- What is ICP-MS? Why is it the gold standard for metal extractables testing?
- Side-by-side: ICP-MS vs ICP-OES vs AAS
- Test workflow: from cartridge prewetting to final report
- Metal extractables performance comparison across materials (PP / PES / PTFE / UPE)
- Semiconductor vs pharmaceutical: two completely different acceptance standards
- How to read a CoA: the 5 elements buyers must focus on
- Common pitfalls
- Frequently Asked Questions
- References
Why are wafer fabs willing to pay a 1000x premium for filter cartridges with ultra-low metal extractables?
Take the same 10-inch 0.05 µm filter cartridge: a standard industrial PP version costs less than NT$1,000, while a semiconductor-grade UPE / PTFE cartridge can run NT$30,000 or even NT$100,000. Where does the difference come from? — metal extractables.
In 3 nm processes, the linewidth of a single logic gate on a wafer is smaller than a COVID virus. A single Na, K, Fe, or Cu atom landing on the silicon can affect electrical performance at best, or scrap an entire wafer batch at worst. Cu diffuses extremely fast in silicon, Na/K cause leakage in gate oxides, and Fe creates deep-level traps at P-N interfaces. So when a filter cartridge leaches these metals into photoresist, developer, HF, or KOH, it's effectively dosing the process chemicals with "targeted poison."
This is why the semiconductor industry is willing to pay a 100x price premium for filter cartridges: it's not that the cartridge itself is worth that much — it's that the wafer yield downstream is. A 1% yield drop on a 12-inch wafer batch causes losses far exceeding an entire year's filter procurement budget. The pharmaceutical industry is similar — excessive heavy metals in injectables or eye drops are a direct patient safety issue, and the FDA's Q3D standard tightly controls Pb, As, Cd, Hg, and other elements.
What is ICP-MS? Why is it the gold standard for metal extractables testing?
ICP-MS stands for Inductively Coupled Plasma Mass Spectrometry. Think of it as an "atom-by-atom scale": it first breaks the sample down into individual atoms, then weighs each one by mass number while counting how many there are.
Three operating steps
- Nebulization and sample introduction: the aqueous sample is first nebulized into micron-sized droplets, mixed with argon gas, and fed into the plasma torch.
- Plasma ionization: an RF coil generates a plasma torch in argon at 6,000–10,000 K (hotter than the sun's surface). At this temperature, almost all elements are stripped of one electron, becoming singly charged cations.
- Mass filtering and counting: ions are guided into a quadrupole mass spectrometer, separated one by one according to mass-to-charge ratio (m/z), and the detector counts the number of ions arriving per second (cps).
Why is it the gold standard?
- Ultra-low detection limit: most elements have an LOD (detection limit) in the 0.001–0.1 ppb (ng/L) range — 100–1000x lower than ICP-OES, and over 1000x lower than AAS
- Simultaneous multi-element analysis: a single injection delivers a full spectrum of 70+ elements (from Li-7 to U-238) in 1–2 minutes — no need to swap light source lamps one by one
- Wide dynamic range: linear from ppt to ppm across 9 orders of magnitude — major and trace elements measured on the same plate
- Isotope discrimination: can distinguish different isotopes of the same element (e.g., ⁶⁴Zn vs ⁶⁶Zn), particularly useful for tracking contamination sources
Side-by-side: ICP-MS vs ICP-OES vs AAS
All three metal measurement techniques are still in use, but they serve completely different roles. For semiconductor / pharmaceutical-grade filter cartridge QC, only ICP-MS is generally accepted.
| Item | ICP-MS | ICP-OES (ICP-AES) | AAS (Flame / Graphite Furnace) |
|---|---|---|---|
| Detection principle | Plasma ionization + mass spectrometry | Plasma excitation + emission spectroscopy | Atomic absorption spectroscopy |
| Typical LOD | 0.001–0.1 ppb (ng/L) | 0.1–10 ppb | Flame: 1–100 ppb; Graphite furnace: 0.05–1 ppb |
| Simultaneous elements | 70+ elements simultaneously | 30–60 elements simultaneously | One element at a time |
| Analysis speed | 1–3 min / sample | 2–5 min / sample | 3–8 min / element |
| Dynamic range | 9 orders of magnitude | 5–6 orders of magnitude | 2–3 orders of magnitude |
| Isotope discrimination | Yes | No | No |
| Instrument cost | USD 200,000–500,000 | USD 70,000–150,000 | USD 30,000–80,000 |
| Cost per sample | Higher (argon + standards) | Medium | Low |
| Typical applications | Semiconductor, pharmaceutical trace, environmental trace | Water, soil, alloys | Routine water testing, education |
| Suitability for cartridge QC | Standard for SEMI / USP <232> | Marginal for high-concentration samples | Insufficient sensitivity |
Test workflow: from cartridge prewetting to final report
This isn't a "feed in a sample, get a number out" affair. From unboxing the cartridge to producing the report, every step can shift results by at least one order of magnitude. Below is a typical SEMI-grade metal extractables test workflow:
Step 1 · Pre-conditioning / Flushing
Flush the cartridge with ultrapure water (UPW, <18.2 MΩ·cm, TOC <5 ppb) at a specified flow rate to remove loose particles and surface metals left from shipping, packaging, and manufacturing. Semiconductor-grade cartridges typically require 3–10 cartridge volumes at 2–10 LPM, flushing until downstream metal readings stabilize.
Step 2 · Extraction
After flushing, switch to the "extraction fluid" for soaking or recirculation. The extraction fluid is selected based on the target process:
- Aqueous chemicals → extract with UPW
- Acidic chemicals (HF, HNO₃, HCl) → extract with 1% HNO₃ or the target process acid
- Alkaline chemicals (KOH, TMAH) → extract with NH₄OH or the target process base
- Organic solvents (photoresist, thinner) → extract with the corresponding solvent, then acid digest before analysis
Typical extraction conditions: room temperature to 50 °C, static or recirculating for 24–48 hours. Semiconductor manufacturers sometimes run aggravated tests (70 °C × 168 hr) to simulate worst-case scenarios.
Step 3 · Sample + Blank
Take samples in parallel: cartridge extract (sample) and the same batch of unfiltered extraction fluid (blank). The blank's metal readings are subtracted directly from the sample, so the blank's cleanliness sets the effective floor of the entire test.
Step 4 · ICP-MS analysis
Before injection, dilute with 2% HNO₃ (or run undiluted depending on the case), and add internal standards (commonly ⁴⁵Sc, ⁷²Ge, ¹¹⁵In, ²⁰⁹Bi) to correct for matrix effects. The analytical panel covers at least:
Step 5 · Data processing and reporting
Results are expressed as ng/cm² of membrane area or ppb (ng/g of extract). The CoA lists each element's value, blank, and LOD (< LOD is typically reported as "ND" or "< 0.05 ppb"). Semiconductor customers compare against their in-house incoming spec, and any out-of-spec result triggers an immediate rejection.
Metal extractables performance comparison across materials (PP / PES / PTFE / UPE)
For the same pore size and same brand, different materials can yield metal extractables differing by three orders of magnitude. Below are representative values commonly cited in industry technical literature (units: ppb, 24 hr extraction in 18 MΩ UPW):
| Material | Typical applications | Total Metals | Fe | Na | Cu |
|---|---|---|---|---|---|
| PP (polypropylene) | Industrial prefiltration, chemical | 50–500 ppb | 10–200 ppb | 20–300 ppb | 0.5–10 ppb |
| PES (polyethersulfone) | Pharmaceutical, biotech sterile filtration | 5–50 ppb | 1–10 ppb | 2–20 ppb | 0.1–1 ppb |
| PTFE (high-purity grade) | Semiconductor, strong acids/bases | 0.5–3 ppb | 0.05–0.5 ppb | 0.1–1 ppb | < 0.05 ppb |
| UPE (ultra-high-molecular-weight PE) | Photoresist, HF, IPA | < 1 ppb | < 0.1 ppb | < 0.5 ppb | < 0.02 ppb |
Visual comparison (Fe extractables, log scale)
Semiconductor vs pharmaceutical: two completely different acceptance standards
Even though both industries care about "metal extractables," they look at completely different numbers and different regulations. The most common rookie mistake is mixing the two standards.
Key differences summary
- Different elements of concern: semiconductor focuses on Na / K / Fe / Cu / Ni / Cr (electrical impact on silicon); pharmaceutical focuses on Pb / As / Cd / Hg (toxicological risk)
- Different units: semiconductor uses ng/cm² and ppt-level numbers; pharmaceutical uses µg/day and ppm-level PDEs
- Different extraction conditions: semiconductor uses target process chemicals (HF, IPA, etc.); pharmaceutical uses simulating solutions (pH 3, pH 9, 50% ethanol, etc.)
- Different frequencies: semiconductor may test every batch; pharmaceutical tests primarily during vendor qualification + periodic requalification
How to read a CoA: the 5 elements buyers must focus on
A CoA (Certificate of Analysis) often lists 30+ element values. With limited time, focusing on these 5 will help you avoid 90% of pitfalls:
| Element | Why it matters | Typical SEMI-grade limit | Typical pharmaceutical-grade limit |
|---|---|---|---|
| Fe (iron) | Most ubiquitous contamination source on the line (piping, agitators, blades); deep-level traps in silicon | < 0.5 ppb | < 50 ppb |
| Na (sodium) | Glass, human contact, resin residues; causes leakage in gate oxides | < 1 ppb | < 100 ppb |
| Cu (copper) | Common in BEOL processes; diffuses extremely fast in silicon — even nanoscale segregation affects electrical performance | < 0.1 ppb | < 10 ppb |
| Ni (nickel) | Leaches from stainless steel equipment; MOSFET interface traps | < 0.2 ppb | < 20 ppb |
| Cr (chromium) | Leaches from stainless steel equipment; Q3D Class 3 in pharma — toxicology requires control | < 0.2 ppb | < 11 µg/day (Q3D parenteral) |
Common pitfalls
Frequently Asked Questions
If ICP-MS can detect down to ppt (parts per trillion), why are some elements still out of reach?
Mainly due to polyatomic interference. ICP-MS uses an argon plasma, so Ar-based ions (⁴⁰Ar⁺, ⁴⁰Ar¹⁶O⁺, ⁴⁰Ar³⁵Cl⁺, etc.) interfere with elements at neighboring mass numbers, such as ⁵⁶Fe, ⁵²Cr, ⁷⁵As, ⁸⁰Se. The fix is a collision/reaction cell (KED or DRC mode) that uses He / H₂ / NH₃ to knock out polyatomic ions, allowing Fe / As / Se to also reach ppt levels.
Why do some cartridge suppliers only provide total metals without individual elements?
Usually for two reasons: (1) they only run ICP-OES internally, which can't reach the required LOD for individual elements; (2) some elements exceed limits but the total still falls within spec, so they list only the total to mask it. Professional buyers should require an 18+ element ICP-MS panel with individual values and LODs — never accept a total-only report.
Why is the semiconductor industry willing to spend so much on metal extractables testing?
Because the value of a single 12-inch wafer (including invested process costs) can reach USD 5,000–20,000, and a 25-wafer batch is over USD 100,000. If cartridge-leached Cu causes a 5% yield drop on the entire batch, the loss far exceeds an entire year's filter budget. An ICP-MS test costs about USD 200–500 — practically free insurance against the loss.
Does the pharmaceutical industry need to reach ppt levels?
Mostly no. USP <232> / ICH Q3D PDEs are typically in the µg/day range, which translates to roughly ppb-level cartridge extractables — sufficient for the application. Pharma cares about "toxicological risk to patients," not "electrical damage". So PES / hydrophilic PTFE are typically adequate for pharmaceutical use, with no need to upgrade to semiconductor-grade UPE.
Should the extraction fluid be UPW or the actual process chemical?
It depends on what you care about most. UPW extraction is a neutral condition reflecting the cartridge bulk's "easily leachable metals" — the industry benchmark. Process chemical extraction (HF, KOH, IPA, etc.) reflects the actual extractables under use conditions and is closer to the field, but every chemical change requires retesting. High-end semiconductor customers typically demand both.
Once a cartridge has been initially flushed, do metal levels stabilize?
Not completely, but they drop sharply. The typical curve: a two-order-of-magnitude drop in the first 30 min, slow decay over 1–4 hr, and approaching steady state after 24 hr. So when high-end customers commission cartridges in the field, they often run a 4–24 hr "on-site flushing" and only connect to the production line after the effluent meets spec.
Can ICP-MS measure anions (Cl⁻, F⁻, SO₄²⁻) at the same time?
No. ICP-MS detects cations primarily; anions require IC (ion chromatography). A complete semiconductor spec typically includes ICP-MS (cations + trace metals) + IC (anions) + TOC — three instruments, none replaceable.
References
- Entegris — Filtration Technical Literature (includes UPE / PTFE cartridge ICP-MS extractables data)
- Cytiva — Extractables and Leachables Services (BPOG protocol)
- Pall Corporation — Extractables and Leachables Testing
- SEMI Standards — C68, C70, C79 series (process chemicals elemental specifications)
- USP — Elemental Impurities <232> / <233> / <2232>
- ICH Q3D(R2) Guideline for Elemental Impurities
- Agilent — ICP-MS Application Notes (high-purity semiconductor chemical analysis)
- Thermo Fisher Scientific — ICP-MS for Semiconductor Applications
- BioPhorum (BPOG) — Extractables Testing Protocol for Single-Use Components