Three Terms, Three Different Worlds
"Compressed air", "CDA" and "XCDA" get used interchangeably on site, but they control completely different levels of contamination. Confusing them means specifying and accepting the system wrong from the very first step.
| Term | What is controlled | Typical spec | Common basis | Where it is used |
|---|---|---|---|---|
| Compressed air | Liquid water, oil, rust | Dew point +3°C, oil ≤5 mg/m³ | In-house facility rules | Pneumatic tools, cylinders |
| CDA (clean dry air) | Moisture, oil, particles | Dew point −40°C, oil ≤0.01 mg/m³, particles ≤0.1 μm | ISO 8573-1 | General facility, cleanroom pneumatics |
| XCDA (extreme clean dry air) | Molecular contamination (AMC) | H₂O, CO₂ <100 ppt; TOC, acids, bases, refractory <10 ppt | Site-specific, usually referencing SEMI F21 AMC classes | Scanners, metrology tools, wafer and reticle purge |
The decisive difference is the unit in the spec column. The CDA world speaks in mg/m³ and μm; the XCDA world speaks in ppt. One part per trillion is roughly 2.5 millilitres — half a teaspoon — dissolved in an Olympic pool of 2,500 cubic metres. XCDA demands that half a teaspoon of contamination cannot be found anywhere in the pool.
One line to remember: CDA controls the dirt you can see and touch, XCDA controls the dirt you can only measure.
XCDA Starts From the Dirtiest Feedstock on Earth
A nitrogen purifier is fed 5N or 6N gas from a cylinder or a liquid nitrogen vaporiser. An XCDA purifier is fed the atmosphere. That is what makes it the hardest member of the gas purifier family — the raw material is itself the contamination source.
| Species | Ambient background (urban typical) | XCDA outlet threshold | Reduction required |
|---|---|---|---|
| H₂O (25°C, 60% RH) | ~1.9% (19,000 ppm) | <100 ppt | ~190,000,000× |
| CO₂ | ~425 ppm | <100 ppt | ~4,250,000× |
| Volatile bases (NH₃) | 1–10 ppb | <10 ppt | 100–1,000× |
| Volatile acids (SO₂, NOx) | 1–5 ppb | <10 ppt | 100–500× |
| Organics (TOC) | 100 ppb – several ppm | <10 ppt | >10,000× |
On top of that comes the oil the compressor contributes itself. The strictest oil-free grade in ISO 8573-1, Class 1, allows ≤0.01 mg/m³. That sounds clean, but converted to a volumetric concentration (estimating as a C12 hydrocarbon) it is roughly 1.4 ppb — still more than a hundred times the XCDA organics threshold of <10 ppt. The calibration basis differs (TOC is quoted as C₃H₈), but the order of magnitude is unambiguous: "oil-free compressor" does not mean "organics-free air".
This is why XCDA cannot be solved by a single box. It needs a chain of stages with clearly divided labour.
Why a −70°C Dew Point Says Nothing About Cleanliness
The most common piece of misdirection in a datasheet is using dew point as a proxy for purity. Dew point describes exactly one species, H₂O, and says nothing about CO₂, organics, acids or bases. And even for H₂O alone, the conversion is routinely misread.
| Atmospheric dew point (over ice) | H₂O volumetric concentration |
|---|---|
| −40°C | ~127 ppm |
| −60°C | ~10.7 ppm |
| −70°C | ~2.6 ppm |
| −80°C | ~540 ppb |
| −100°C | ~14 ppb |
| −120°C | ~150 ppt |
Look at the last two rows. A dryer advertising a "−100°C dew point" still leaves 14 ppb of H₂O at its outlet — that is 14,000 ppt, a factor of 140 above the XCDA limit of <100 ppt. To reach 100 ppt the atmospheric dew point has to fall to around −122°C, which is already beyond the credible measuring range of almost every dew point instrument on the market.
There is a subtler trap as well: pressure dew point and atmospheric dew point are not the same number. A "−40°C dew point" measured at 0.7 MPa(g) (about 8 bar absolute) corresponds to roughly 16 ppm by volume; expanded to atmospheric pressure that same gas has a dew point near −53°C. Always ask which one a datasheet means — the gap can be more than ten degrees.
Practical conclusion: ppt-class gas is not described by dew point, it is described by concentration. When someone discusses XCDA in terms of dew point, it usually means their instrument cannot see that range at all.
The Purification Train: What Each of the Five Stages Does
Fig. 1: The Five-Stage XCDA Purification Train
Each stage handles only what it is good at — skip one and the next pays ten times over
The five stages are in series, not alternatives. The one most often underrated is stage 1: if pre-treatment lets oil aerosol through, the sorbent in stage 4 is poisoned for good. Regeneration will not bring it back — the bed has to be replaced.
The order cannot be rearranged, because every stage exists to unload the next one:
- 1Compression and pre-treatment — cyclone separation, coalescing filters, activated carbon for oil removal. The job here is not to produce clean gas, it is to protect what follows. Once oil aerosol slips through, the sorbent in stage 4 has its pore mouths coated and is poisoned permanently; hot regeneration will not drive it back out and the bed has to be replaced.
- 2Drying (temperature swing adsorption) — molecular sieve takes the bulk H₂O load from 1.9% down to the ppm range. This stage removes over 99.99% of the water molecules; without it, the deep adsorption beds downstream would saturate within hours.
- 3Catalytic oxidation — CO, H₂ and methane are small molecules that adsorbents barely hold at ambient temperature. A heated catalyst oxidises them into CO₂ and H₂O, species the downstream beds are good at. Classic "if you cannot beat it, convert it" engineering.
- 4Deep adsorption (multi-bed) — the stage that actually reaches ppt. Different layers handle different species: molecular sieve for H₂O and CO₂, modified activated carbon for organics and refractory compounds, impregnated media for SO₂ and NH₃. Multiple vessels run in parallel, alternating between ambient-temperature purification and high-temperature regeneration under automatic control, so supply never stops.
- 5Point-of-use filtration — however clean the first four stages are, the gas still travels through valves, tubing and the beds themselves, all of which shed particles. Final filtration sits as close to the tool as possible; a typical requirement is ≤1 pcs/m³ at 0.003 μm.
In other words, no single stage is "the XCDA purifier". XCDA is the output of the whole chain. Buying stage 4 alone and bolting it onto a facility CDA header rarely survives a year.
Six Specifications, Six Failure Modes
The spec table of an XCDA purifier looks like a column of numbers, but every row maps to a real tool failure.
| Specification | Typical threshold | Why it is controlled | What failure looks like |
|---|---|---|---|
| H₂O | <100 ppt | Water is the universal reaction medium and the carrier for acid, base and salt deposition | Optical hazing, corrosion of metal layers, drifting metrology baselines |
| CO₂ | <100 ppt | Participates in carbon deposition under deep UV, interferes with infrared measurement baselines | Falling lens transmission, inaccurate FTIR readings |
| Total organic carbon (as C₃H₈) | <10 ppt | Organic molecules adsorb onto optical surfaces and are photolysed by 193 nm light into a carbon film | Unstable exposure dose, shortened lens life |
| Volatile acids (as SO₂) | <10 ppt | Combine with moisture to form acids that attack metals and copper processes | Copper line corrosion, salt crystals becoming particles |
| Volatile bases (as NH₃) | <10 ppt | Neutralise the photoacid at the surface of chemically amplified resist | Post-develop CD deviation, T-top defects |
| Refractory compounds (as HMDSO) | <10 ppt | Siloxanes decompose under heat or light, leaving SiO₂ that cannot be burned or washed off | Permanently hazed optics, foreign material on the wafer |
| Particles | ≤1 pcs/m³ @0.003 μm | 3-nanometre particles are already fatal defects at advanced nodes | Random yield loss that is nearly impossible to trace |
Volatile bases deserve their own paragraph. Chemically amplified resist relies on photoacid diffusion to complete the reaction after exposure, and NH₃ is a base that neutralises the acid at the top surface. The result is an under-reacted resist top and a T-shaped profile after development — the industry calls it a T-top defect. The cruel part is the timing: during the queue time between exposure and post-exposure bake, ppb-level ambient NH₃ needs only a few minutes to ruin an entire lot, and nothing looks wrong to the eye. This is exactly why AMC control in the litho bay has been the number-one topic for years.
What "as SO₂" and "as HMDSO" Actually Mean
The "as" in a spec table does not claim the gas contains only that compound. It states the equivalence basis. Volatile acids may be a mixture of SO₂, HF, HCl and HNO₃; the analysis is calibrated against SO₂ and the total is reported as an equivalent SO₂ concentration. HMDSO (hexamethyldisiloxane) plays the same representative role for the siloxane family.
So when comparing two datasheets, check that the calibration basis matches before comparing the numbers. Comparing values quoted on different bases is a common source of wrong conclusions.
Where XCDA Is Actually Used
- ▸Air bearings and lens purge on scanners — a scanner stage floats on air bearings to position at high speed with almost no friction, which takes a large and extremely stable gas supply. Oil and particles in that gas land directly on the bearing surfaces and the optics, which is the main reason XCDA flow specifications extend to the 20,000 Nm³/h class.
- ▸Optical paths in metrology tools — overlay and film-thickness measurement systems shift systematically once a molecular film forms anywhere in the beam path. That shift never raises an alarm; it just sends process engineers chasing a phantom signal for weeks.
- ▸Wafer and reticle transport and storage — FOUPs, reticle pods and stockers are continuously purged with XCDA or high purity nitrogen to hold a low-AMC micro-environment. Wafers usually spend more time waiting between process steps than being processed, and that idle time is when contamination accumulates.
How XCDA Relates to AMC Chemical Filters
The two are often treated as alternatives, but they defend different paths.
| AMC chemical filter | XCDA purifier | |
|---|---|---|
| Path defended | Air in the room (make-up, recirculation, process outgassing) | Gas in the pipe, delivered to the tool |
| Inlet concentration | ppb to ppm | ppb to percent (ambient air) |
| Outlet target | sub-ppb | ppt |
| Replacement model | Whole media set replaced at saturation | Beds regenerated at high temperature and reused |
| Failure signature | Slow rise in environmental monitoring | Online outlet readings spiking or stepping up |
The correct mental model is layered defence: chemical filters hold down the background of the entire room, XCDA holds down the specific stream that blows onto the wafer and the optics. When the room background is high, the inlet load on the XCDA purifier rises with it and regeneration cycles shorten — the two are coupled, not interchangeable. For the design logic on the filter side, see chemical filter selection and impregnation.
FAQ
Q: Can we turn CDA into XCDA by adding a chemical filter to the header?
No — the gap is in orders of magnitude, not in degree. A chemical filter typically takes ppb down to sub-ppb; XCDA needs ppt. Worse, filters have almost no capacity for H₂O and CO₂, which happen to be the two largest impurities in air. Without a drying stage and regenerable beds, the media saturates very quickly.
Q: XCDA or high purity nitrogen purge — which should we use?
It depends on the purpose. If oxygen has to be excluded (oxidation or flammability control), nitrogen is mandatory. If the goal is simply a clean, AMC-free environment without creating an asphyxiation hazard, XCDA fits better — it is still air, so it is markedly safer in areas people can enter. Many fabs use both: nitrogen inside sealed enclosures, XCDA in open or semi-open areas.
Q: Do we really need <10 ppt? Is that not over-engineering?
It depends on the node and the point of use. Within one fab, scanners and metrology tools need the full specification while ordinary pneumatic components are fine on CDA. The pragmatic approach is tiered supply: CDA on the facility header, XCDA only at the tools that need it, rather than driving the whole plant to ppt. For flow calculation and selection order, see the gas purifier selection guide.
Q: Does the outlet get dirty when a bed regenerates?
A well-designed multi-vessel system produces a brief concentration disturbance at switchover, which should be absorbed by downstream buffering and the final adsorption stage — online monitoring should show no obvious step. If every switchover produces a regular spike in the data, it usually means insufficient regeneration gas flow, regeneration temperature not reaching setpoint, or a valve sequencing problem. This is one of the things to watch closely during acceptance.
Q: Can we use ordinary stainless steel tubing for an XCDA system?
At ppt level the tubing is itself a contamination source. The rough inner surface of ordinary stainless steel adsorbs moisture and organics, takes a very long time to dry down after start-up, and may never reach target at all. These systems normally use electropolished 316L (EP 316L) with tight control of weld quality and dead-leg length. The best purifier in the world is wasted behind non-compliant tubing.
Q: How do we know whether our existing system actually meets XCDA?
You cannot read it off a nameplate — it has to be measured. But ppt-level measurement is a discipline of its own: choose the wrong instrument or fail to purge the sample line long enough and the numbers describe your sampling system rather than your gas. That subject has its own article: XCDA measurement, sampling and acceptance.