In One Line: Turning "Pure Enough" Into "Process Grade"
Bottled water labelled "pure" is perfectly fine to drink. Rinse a wafer with it and the residual minerals leave watermarks that scrap the part. Gas works the same way — nitrogen stamped 99.999% (5N) is more than good enough for blowing down a tool enclosure, yet the same gas fed into an epitaxy furnace can ruin a wafer.
A gas purifier does something simple: it pushes already-pure gas up another two or three orders of magnitude, from "chemical grade pure" to "semiconductor process usable". It is not a filter — filters stop particles that have physical volume, while a purifier goes after molecular contaminants dissolved among the gas molecules themselves.
5N Does Not Mean Clean: What an N Grade Actually Contains
The "N" counts the nines. 4N is 99.99%, 5N is 99.999%, 6N is 99.9999%. Translated into impurity concentration it becomes far more intuitive: 5N means roughly 10 ppm of total impurities — ten parts in a million that are not the gas you ordered.
Fig. 1: The Gas Purity Ladder — What Does an "N" Grade Actually Mean?
Logarithmic axis, one decade per step. "5N" only guarantees total impurities below 10 ppm — three orders of magnitude away from the ppb levels processes require
How small is 1 ppb? About one second in 32 years, or 2.5 grams of salt in an Olympic pool. The N grade on the cylinder is a factory value; purity degrades through cylinders, piping and valves, which is why gas is purified again at the point of use.
The catch is that a process never cares about total impurities, only about which impurity. In 5N nitrogen, if those 10 ppm are mostly argon, most processes will not notice. If 2 ppm of it is water, the epitaxial interface will oxidise. That is why every purifier datasheet lists outlet concentrations for H₂O, O₂, CO, CO₂, CH₄ and NMHC individually instead of quoting a single purity figure.
A common purchasing misconception: "buy 6N gas and you can skip the purifier". 6N is a factory value describing the moment of filling, not the quality at the tool inlet.
The Gas Leaves Clean — The Problem Is the Journey
Even when the supplier meets the contract exactly, gas picks up impurities again on its way from the cylinder or bulk tank to the tool:
| Source of impurity | When it happens | Typical effect |
|---|---|---|
| Outgassing from pipe walls | Early life of new piping, during system warm-up | Moisture and hydrocarbons released at ppb–ppm level |
| Micro-leaks at valves and fittings | After long service, as seals age | Atmospheric O₂ and H₂O back-diffuse into the line |
| Air ingress during cylinder change | Every cylinder swap or manifold switchover | Short ppm-level spikes that must be purged out |
| Residual moisture after maintenance | After shutdowns and pipe rework | Wall-adsorbed water takes days to weeks to purge |
| Permeation through filters and hoses | Any section that is not all-metal | Continuous low-level ingress of atmospheric moisture |
Fig. 2: Where Does a Purifier Sit in the Gas Supply Chain?
Bulk purification at the facility header, fine POU purification at the tool — a two-stage split
Why two stages? Gas leaving the facility purifier still travels hundreds of metres of stainless piping through dozens of valves and fittings, picking up moisture and particles along the way. A POU purifier sits within a metre of the tool — the last line of defence.
In other words, "source purity" and "tool inlet purity" are two different numbers. A purifier at the point of use measures and controls the second one.
Six Impurities, Six Failure Modes
Each chemical formula on a purifier datasheet maps to a distinct way of losing product:
| Impurity | Mechanism | Typical consequence |
|---|---|---|
| H₂O moisture | Reacts with silicon and metals at high temperature | Interfacial oxidation, gate oxide thickness drift, electrolyte decomposition in batteries |
| O₂ oxygen | Directly oxidises process surfaces and solder joints | Film defects, solder voiding, rising metal layer resistance |
| CO / CO₂ | Decomposes into carbon and oxygen at high temperature | Carbon contamination of silicon, stacking faults in epitaxy |
| CH₄ / NMHC hydrocarbons | Cracks and deposits as carbon film | Surface carbon contamination, hazing of optics, analyser baseline drift |
| N₂ (inside H₂ or Ar) | Forms nitrides at high temperature | Shifts film composition and carrier concentration in GaN / III-V processes |
| Particles | Settle physically on the wafer surface | Pattern defects, shorts and opens — direct yield loss |
Note the difference in the last two rows: nitrogen is obviously harmless in a nitrogen process but is an impurity when carried in hydrogen or argon; particles are physical contamination handled by the high-precision filter built into the purifier (0.003μm is a common rating), a different mechanism from the sorption stages above.
Who Uses Them, and Where
Fig. 3: Gas Purifiers Across Industries
Same nitrogen: a reflow oven needs ppm, epitaxy needs ppb — a 1000× difference that drives completely different equipment choices
| Industry / process | Main gases | What impurities do | Typical outlet requirement |
|---|---|---|---|
| Semiconductor epitaxy, etch, deposition | N₂, H₂, Ar, He | H₂O / O₂ cause interfacial oxidation and film defects | Each < 1 ppb |
| Advanced packaging reflow, wave solder | N₂ | O₂ causes joint oxidation, voiding, poor wetting | O₂ < 10–50 ppm |
| Display / LED MOCVD epitaxy | NH₃, H₂, N₂ | Moisture degrades crystal quality and light output | H₂O, O₂ < 1 ppb |
| Lithium battery dry room & formation | Dry air, N₂ | Water decomposes electrolyte, causes gassing and swelling | Dew point −40 to −60°C |
| Optical fiber preform deposition | O₂, He, Cl₂ | OH⁻ absorption peak raises fiber attenuation | H₂O at ppb level |
| Litho and metrology tool purge | XCDA clean dry air | AMC hazes optics and leaves wafer surface haze | H₂O / CO₂ < 100 ppt |
| Supercritical CO₂ wafer drying | CO₂ | NVHC precipitates on depressurisation; acids corrode metal | NVHC < 0.01 ppb |
| Laboratory instrument carrier gas | He, Ar, N₂ | Baseline drift, ghost peaks, worse detection limits | Each < 1 ppb |
Higher purity is not automatically better — it should match what the process actually tolerates. Over-specifying costs capital, consumables and regeneration energy; under-specifying shows up directly in yield. Ask the process owner for the tolerable impurity ceiling instead of copying another site's spec.
One point stands out: purity level is not an industry label, it is a process tolerance. For the same nitrogen, an advanced packaging reflow oven asks for O₂ below 50 ppm while an epitaxy furnace asks for below 1 ppb — nearly 50,000× apart. Specifying a reflow purifier to epitaxy standards spends money where there is no return.
The Two-Stage Layout: Facility Side and Point of Use
Large fabs normally split purification into two stages:
- 1Facility side: a large-flow fully automatic regenerative purifier on the main header, 10 to 50000 Nm³/h, with parallel adsorption beds — one supplying while another regenerates, running 24/7. It lifts the whole plant's baseline gas quality to ppb level.
- 2Point of use (POU): an inline purifier or heated getter purifier within a metre of each critical tool, 0.2 LPM to 4000 SLPM. It removes whatever the piping put back in.
Smaller plants and R&D labs usually implement only the POU stage — modest consumption and short piping mean a single inline purifier meets spec, with no case for building a full regeneration system for a couple of lines.
Three Signals That You Need a Purifier
- ▸The tool spec is stricter than the gas contract: the equipment maker asks for ppb-level inlet gas while the supply contract only guarantees ppm. Nobody closes that gap by default — that gap is where the purifier goes.
- ▸Yield dips line up with cylinder changes or maintenance: if defect peaks reliably follow a cylinder swap by a few hours, or a restart by a few days, impurity spikes are the likely cause, and no amount of recipe tuning will fix it.
- ▸The point of use reads worse than the facility header: 1 ppb of moisture at the header but 20 ppb at the tool inlet means contamination is generated in the piping, so the fix belongs at the point of use, not at the source.
FAQ
Q: Our gas is already 6N — do we still need a purifier?
It depends which impurity the process cares about. 6N means roughly 1 ppm of total impurities; if the process needs moisture below 1 ppb, that is still three orders of magnitude away. And 6N describes the moment of filling — after piping, valves and cylinder changes, the value at the tool inlet is always worse.
Q: How is a purifier different from a gas filter?
A filter mechanically intercepts particles that have volume; its pore size sets the smallest thing it can stop. A purifier captures molecules dissolved in the gas — water, oxygen, carbon monoxide, methane — by adsorption, catalysis or chemical bonding. They address contamination at completely different scales. Most purifiers include a filter, but a filter can never replace a purifier.
Q: Once the sorbent saturates, can it release impurities back into the gas?
Yes, and this is the failure mode to watch. Adsorption is a reversible equilibrium, so when a bed saturates — or when inlet concentration suddenly drops — previously captured impurities can desorb, briefly making the outlet worse than the inlet. Purifiers therefore need online monitoring and a managed replacement interval; they are not fit-and-forget devices.
Q: Can one purifier handle several different gases?
Not advisable. Sorbent formulations are designed for specific gas and impurity combinations. Using one column on both oxygen and hydrogen is not only inefficient but raises a real safety issue from cross-residues of oxidising and flammable gases. In practice it is one gas per unit, so the gas species must be defined at selection time.
Q: How do we verify it actually works after installation?
The direct method is simultaneous sampling at the purifier inlet and outlet with online trace analysers. Moisture is typically measured by CRDS (cavity ring-down spectroscopy) or capacitive dew point sensors, oxygen by electrochemical or fluorescence analysers, hydrocarbons by FID or GC-MS. The sample line itself must be purged to a stable background first, or you are measuring the tubing rather than the gas. For the principles and applicable ranges of these techniques, see comparison of online AMC monitoring technologies.
Q: Is maintenance expensive?
It depends on the type. Ambient inline purifiers are usually returned for regeneration every 1–2 years, with 20+ year equipment life; heated getter columns are non-regenerable and replaced on site roughly every 3–5 years; large automatic units regenerate themselves, so the cost sits in regeneration heating energy plus periodic sorbent replacement. Folding these into total cost of ownership is far more informative than comparing purchase prices.