Filter Particle Release: How LPC Evaluates Cleanliness

A single 30 nm particle can scrap a wafer. This article walks through LPC laser scattering, particle release curves, semi fab acceptance criteria, and why pre-flush is mandatory.

Article Highlights · Key Points

A single 30 nm particle landing on an EUV wafer can scrap an entire die — LPC is the instrument purpose-built to hunt this kind of nanoscale dust
The Liquid Particle Counter (LPC) uses laser scattering / light obscuration to count particles down to ≥ 20 nm in real time
Every semiconductor filter cartridge must be pre-flushed before service, because the initial particle release curve is an exponential decay from a high peak — not a flat line
The dominance of UPE asymmetric membranes in advanced nodes was driven directly by LPC data
LPC catches particles, ICP-MS catches metallic ions, TOC catches organics — none of the three can be skipped

Table of Contents

Why a single 30 nm speck can decide a wafer's fate
What is LPC? How a laser "sees" a 30 nm particle
The particle release curve: why every cartridge needs pre-flush conditioning
LPC vs ICP-MS vs TOC: catching different dimensions of contamination
Particle release by membrane material (and UPE's nanoscale-scalpel status)
LPC acceptance criteria in semiconductor fabs (with real numbers)
How to read LPC data on a cartridge CoA
Common pitfalls (e.g. skipping pre-flush)
Frequently asked questions
References

Why a single 30 nm speck can decide a wafer's fate

The gate width of a 3 nm node is roughly 24 nm. In other words, a stray 30 nm particle is the equivalent of dropping a boulder larger than the gate itself onto the circuit. A contaminant of this size entering EUV lithography, CMP, wet cleaning, or high-purity chemical delivery at any step can directly cause shorts, opens, or pattern defects.

What's worse is that these nanoparticles don't come from the outside world — they come from the ultrapure water and chemicals the process itself uses, and from the filter cartridge itself. That's right: even a cartridge designed to "filter out impurities" can release particles of its own.

20–30EUV-process particle size of concern (nm)

<1Top-tier cartridge release spec (counts/mL @ ≥30 nm)

10²–10⁵Initial release of an unflushed cartridge

50–500 LTypical pre-flush volume

The semiconductor industry therefore developed a punishingly strict test method called the LPC (Liquid Particle Counter) particle release test. It doesn't just measure whether the cartridge can stop contaminants — it turns the spotlight on the cartridge itself: how many particles are you shedding? How big? How much flush water does it take to clean you up?

What is LPC? How a laser "sees" a 30 nm particle

LPC stands for Liquid Particle Counter. There are only two core operating principles:

1. Light scattering — for nanoscale particles

Liquid flows through a tiny flow cell illuminated by a high-power laser. Whenever a particle crosses the laser beam it scatters light, and a high-sensitivity photodetector (PMT or APD) on the side captures the scattered photons. Scattered signal intensity is proportional to the 6th power of particle size (Mie / Rayleigh regime). In other words, a 30 nm particle and a 60 nm particle produce signals that differ by a factor of 64, so the instrument can easily separate them.

2. Light obscuration — for micron-scale particles

Larger particles (> 1 µm) use the "light blocking" approach: the particle blocks part of the laser beam, the photodetector sees an instantaneous drop in intensity, and the magnitude of that drop is proportional to the projected area of the particle. Light obscuration is fast and supports high flow rates, so it is widely used to monitor the overall size distribution of process water.

Why is 20 nm the "physical limit" of LPC? When a particle is smaller than ~1/10 of the visible wavelength (i.e. < 50 nm for a 532 nm laser), the scattered signal sinks beneath the noise floor. Measuring 20 nm requires upgrading to a short-wavelength UV laser (355 nm) plus an extremely low-noise PMT. The only instruments on the market that can stably reach ≥ 20 nm are top-tier models such as the PMS (Particle Measuring Systems) UDI-20 series, RION KS-19F, and Beckman HSLIS.

Common instruments and specifications

Model	Smallest measurable size	Flow rate	Typical application
PMS UDI-20	20 nm	10 mL/min	EUV photoresist / DI water inline
PMS Chem 20	20 nm	10 mL/min	Strong acids/bases / solvent inline monitoring
RION KS-19F	20 nm	10 mL/min	UPW / cartridge QC
RION KS-41B	40 nm	10 mL/min	Process chemicals
PMS APSS-2000	0.5 µm	50 mL/min	Injectable batch release
Beckman HIAC 9703+	1.0 µm	10–60 mL/min	Pharmaceutical USP <788>

Data are typically reported as counts/mL @ ≥ X nm, meaning "the number of particles per millilitre at or above X nanometres." This wording matters — for the same cartridge, counting ≥ 30 nm versus ≥ 50 nm easily yields numbers an order of magnitude apart. Failing to read the size threshold carefully means you are comparing two different metrics.

The particle release curve: why every cartridge needs pre-flush conditioning

Installing a brand-new cartridge, opening the valve, and immediately taking the water to use is one of the most unforgivable mistakes in a semiconductor fab. Why? The curve below explains it.

Figure 1 · Typical particle release curve of a semiconductor cartridge (log–log schematic)

When a new cartridge first comes online, particle release can reach 10⁵ counts/mL @ ≥ 30 nm. These particles have three sources:

Membrane manufacturing residue: polymer debris on the membrane surface, unwashed wetting agents (IPA, glycerol), and particles carried in by process water
Mechanical settling: pleat surfaces, housing seams, O-rings, and adhesive lines releasing micro-debris when first pressurized
Metallic components: support nets, core tubes, and end-cap interfaces inside the cartridge leaching ions on first contact with liquid (this fraction is also caught by ICP-MS)

As cumulative flush volume builds, the curve drops exponentially and ultimately approaches an asymptote — that asymptote is the cartridge's true baseline cleanliness. The lower the asymptote and the less flush volume needed to reach it, the better the cartridge. A top-grade UPE cartridge can drop below 1 count/mL within 50 L; a cheap PP cartridge may still hover above 100 counts/mL even after 500 L of flushing.

Field practice: fabs do not wait for a cartridge to "clean itself." Standard practice is to build a pre-flush recirculation loop, continuously flush at low flow rate with same-grade UPW or sample fluid, and sample the LPC every 30 minutes until three consecutive readings fall within spec (qualification pass). The whole procedure normally takes 4–24 hours.

LPC vs ICP-MS vs TOC: catching different dimensions of contamination

Many engineers assume that good LPC numbers mean everything is fine — a common misconception. LPC only measures undissolved solid particles, but real-world contamination has two more battle fronts:

Metric	What it catches	Units	Damage to the process
LPC	Solid microparticles in the liquid (polymer debris, inorganic particles)	counts/mL @ ≥ X nm	Pattern defects, shorts, opens, electrical noise
ICP-MS	Dissolved metallic ions (Na, K, Fe, Cu, Cr, Al, …)	ppt (pg/g)	Reduced carrier lifetime, gate leakage, Cu diffusion contamination
TOC	Dissolved total organic carbon (wetting-agent residue, plasticizers, microbial metabolites)	ppb (µg/L)	Surface organic film contamination, CMP instability, photolithography residue
Non-volatile residue (NVR)	Non-volatile residue left after evaporation	µg/L	Integrated contamination indicator covering particles + dissolved species

All three metrics must pass simultaneously before a cartridge can be considered truly clean. High-end CoAs (Certificates of Analysis) from Entegris, Pall, Cobetter, and similar suppliers report all three. If a CoA only gives you LPC, be careful — metal release or TOC may be its weakest link.

How the three relate: on the same membrane, high LPC indicates loose structure; high ICP-MS indicates leaching from metal hardware or the membrane scaffold; high TOC indicates poorly rinsed wetting agents. Only when all three are low does the entire chain — raw materials, membrane casting, assembly, and packaging — prove itself.

Particle release by membrane material (and UPE's nanoscale-scalpel status)

Membranes made from different polymers can differ in LPC performance by orders of magnitude. The chart below compares typical release levels at the same 0.05 µm rated pore size and the same sample volume (sampled after a 200 L flush):

UPE (asymmetric)< 1 counts/mL @≥30nm

UPE (symmetric)~ 5 counts/mL

PTFE hydrophilic-modified~ 30 counts/mL

Nylon 6,6~ 200 counts/mL

PES~ 500 counts/mL

PP (meltblown)> 5000 counts/mL

UPE (Ultra-high Molecular Weight Polyethylene) asymmetric membranes are the ceiling for advanced-node cartridges. They have three structural advantages:

Ultra-high molecular weight (> 3,000,000 g/mol) gives a stable structure that barely sheds debris
Asymmetric pore distribution (dense top layer + open bottom layer) keeps the tightest retention layer extremely thin while still delivering good flux
The casting process is solvent-free and self-wetting (no wetting agent required), so TOC residue is extremely low

EUV photoresist filtration, advanced CMP slurry, and ultrapure-water point-of-use polishing almost universally use UPE asymmetric membranes. Entegris' Impact series, Pall's Photokleen / Optimizer series, and Cobetter's Aletheia series are representative examples.

LPC acceptance criteria in semiconductor fabs (with real numbers)

LPC specs vary dramatically by process node, chemical, and fab. Below are widely referenced industry benchmarks (each fab's internal spec is normally tighter):

Application	Node	LPC spec	Notes
EUV photoresist filtration	3 nm / 5 nm	< 10 counts/mL @ ≥ 30 nm	Measured at POU; some fabs require < 5
193i photoresist filtration	7 nm / 10 nm	< 50 counts/mL @ ≥ 30 nm	Includes thinners / developers
CMP slurry POU	≤ 14 nm	< 100 counts/mL @ ≥ 50 nm	Slurry itself contains particles; effective abrasive count must be subtracted
UPW polish	All advanced nodes	< 1 count/mL @ ≥ 50 nm	Per SEMI F63 / F75
Wet-clean chemicals (SC1/SC2/HF)	≤ 14 nm	< 30 counts/mL @ ≥ 30 nm	Strong acid/base; PFA / PTFE housing required
Cartridge initial release (post pre-flush)	Photo bay	< 10 counts/mL @ ≥ 30 nm	Three consecutive stable readings required to qualify

Since 2024, leading fabs have begun pushing toward ≥ 20 nm specs, requiring EUV photoresist loops to hit LPC < 5 counts/mL @ ≥ 20 nm — essentially at the physical limit of LPC. Worldwide, fewer than 10 cartridge SKUs can pass this bar.

How to read LPC data on a cartridge CoA

Top-tier cartridges ship with a CoA. When reading the report, check these fields:

LPC

Size threshold

Look at ≥ 20 nm / ≥ 30 nm / ≥ 50 nm values separately. The same cartridge might show < 1 at ≥ 50 nm but 80 at ≥ 30 nm.

LPC

Cumulative flush volume

"Measured after 50 L of flush" versus "after 500 L" makes a huge difference. The CoA must state the test conditions (test volume, flow rate, pressure).

LPC

Instrument model

Numbers from a PMS UDI-20 and a RION KS-41B cannot be compared directly. Only when the CoA names the instrument can you understand its sensitivity and sizing calibration baseline.

Metals

ICP-MS element list

Critical elements Na, K, Ca, Mg, Fe, Cu, Cr, Ni, Al, and Zn each reported in ppt. Cu is a death sentence in fabs — anything > 50 ppt should be rejected.

Organics

TOC (total organic carbon)

Top grade should be < 5 ppb; self-wetting UPE membranes can reach < 1 ppb. A CoA without TOC is missing one of three legs.

Lot

Lot traceability code

The serial number must trace back to the membrane casting lot, cartridge assembly date, and QC operator. This is the lifeline when investigating an incident.

Common pitfalls (eight out of ten engineers have made these)

Pitfall 1: skipping pre-flush and putting a new cartridge straight into the process. The most lethal mistake. A cartridge without pre-flush can release up to 10⁵ counts/mL initially — that's like dirtying high-purity chemicals hundreds of times before sending them downstream. No matter how new or expensive the cartridge is, qualification flush is mandatory.

Pitfall 2: using ≥ 0.5 µm data to satisfy a ≥ 30 nm requirement. Excitement at "< 1 count/mL" on a CoA evaporates when you read the parenthetical "@ ≥ 0.5 µm" — that has nothing to do with a 30 nm process. Mismatched size thresholds make the data meaningless.

Pitfall 3: looking only at LPC and ignoring ICP-MS / TOC. A cartridge with beautiful LPC numbers but constant metal leaching produces "batch-correlated gate leakage" — the kind of ghost story whose root cause cannot be traced. All three metrics must be reviewed together.

Pitfall 4: comparing data across instruments from different vendors. RION and PMS use slightly different sizing calibrations (PSL latex sphere calibration curves differ), so values cannot be matched 1:1. To compare cartridges, use the same instrument with the same standard suspension.

Pitfall 5: assuming a fresh cartridge will hit spec immediately. Every cartridge change must be followed by a full qualification flush of the entire piping segment. Replacing a cartridge means restarting the full startup procedure, not just "pull the old one, plug in the new one."

Pitfall 6: treating LPC as a "retention efficiency" test. LPC measures how much the cartridge releases, not retention rate. Retention is determined by challenge testing (latex bead / bacterial challenge). The two are completely different concepts.

Frequently asked questions

Do LPC values change with flow rate?

Yes. Higher flow rate (more shear) makes the cartridge structure release more particles, so LPC tests must fix the flow rate (typically 1–4 LPM for a 10-inch cartridge). Fab specs require the cartridge to meet spec at the actual operating flow rate, not just barely pass under reduced test conditions.

Why do some cartridges show worsening LPC the longer they are pre-flushed?

The normal curve is monotonically decreasing. A V-shape or upward trend indicates one of three things: (1) the cartridge structure is breaking under pressure and the membrane is shedding; (2) upstream piping or O-ring contamination is reaching the LPC sample port; (3) the instrument itself is contaminated (residual PSL calibration suspension in the flow cell). Stop immediately, inspect upstream tubing, and replace the sampling line if necessary.

How does LPC relate to SEMI F63?

SEMI F63 is the "Guide for Particle Measurement in Ultrapure Water for the Semiconductor Industry," covering how to measure UPW particle levels with LPC, sampling methods, and instrument calibration intervals. It is not a cartridge test standard — it is plant-level water quality monitoring. Cartridge testing is governed by supplier internal SOPs or by the customer (fab) incoming-QC spec.

How is a liquid particle counter calibrated?

Using monodisperse PSL latex spheres (Polystyrene Latex Spheres, with standard sizes such as 30 nm / 50 nm / 100 nm / 0.5 µm). Latex of known size is added to UPW and run through the LPC; the sizing channel thresholds are adjusted so that the count matches theoretical concentration. The calibration interval is typically 6 months or 2,000 hours, depending on fab spec.

Is UPE membrane really that good? When should it not be used?

UPE is the ceiling for EUV photoresist, CMP, and UPW, but it has weaknesses: temperature limit of about 80 °C, and poor resistance to strong oxidizers (concentrated nitric acid, aqua regia, hot SC1). For those duties, switch to PFA / PTFE housings paired with PTFE membranes. There is no universal material — membrane choice always depends on fluid and temperature.

Can LPC data serve as a cartridge lifetime indicator?

Partially. When LPC values begin climbing and can no longer be flushed back to the original baseline, the cartridge structure is degrading (shedding, collapse, loose adhesive). However, cartridge life is primarily judged by differential pressure (ΔP) — once ΔP reaches the design limit (typically 1.5–2.0 bar) the cartridge must be replaced. LPC is a supporting indicator, not the primary lifetime metric.

Do household RO / drinking-water filters need LPC testing?

No. Household water purification follows NSF/ANSI 42 / 53, defined in the ≥ 0.5–5 µm range — orders of magnitude looser than the semiconductor 30 nm bar. Household equipment never operates at the level where particle release is even relevant; its concerns are chlorine, lead, and heavy metals. LPC is a luxury measurement reserved for fabs, biotech, and pharmaceutical applications.

References

Struggling to interpret LPC data — or planning to step up to ≥ 30 nm-grade filtration?

Share your process node, chemicals, flow requirements, and current ΔP curve. The JIUNYUAN engineering team will help you interpret the CoA, design a pre-flush procedure, and recommend cartridges in the right material.

Contact JIUNYUAN Engineers →