PSA Nitrogen Generation: CMS vs 5A Zeolite — Mass Balance, Purity, and TCO

Why PSA Nitrogen Is a Different Problem From PSA Oxygen

Pressure swing adsorption is used to make both oxygen and nitrogen from compressed air, but the two applications sit on opposite sides of one thermodynamic question: which molecule does the sieve adsorb faster? In PSA oxygen, the sieve (LiLSX or 5A) must adsorb nitrogen faster than oxygen, because nitrogen is the impurity that has to be removed to enrich oxygen in the product. In PSA nitrogen, the sieve (CMS) adsorbs oxygen faster than nitrogen, so the nitrogen passes through the bed and comes out as product while the oxygen is held on the sieve temporarily.

This sign-flip in selectivity is not a minor chemical detail. It determines:

• Which sieve type is appropriate (kinetic CMS vs equilibrium 5A)
• How long each cycle can be (seconds for CMS, minutes for 5A)
• How the bed is regenerated (depressurization, vacuum, or both)
• What the air factor looks like at each purity level
• How sensitive the system is to water, CO2, and oil

Many published comparisons of "molecular sieve for nitrogen" make the error of treating 5A zeolite as a direct substitute for CMS. It is not. 5A and CMS occupy different roles in nitrogen service, and pretending they are interchangeable leads to plants that consume 1.5 to 2 times more air per Nm3 of product than a correctly designed CMS unit, or that require vacuum regeneration as a retrofit to reach commercial purity. The remainder of this article corrects that error step by step.

Kinetic Versus Equilibrium Selectivity: The Core Difference

Two physically different mechanisms separate air into N2 and O2:

Equilibrium selectivity is what zeolites do at high pressure and long contact times. The stronger-adsorbing molecule is held on the surface indefinitely until regeneration reverses the thermodynamics. Calcium A-type zeolite (5A) preferentially adsorbs nitrogen at equilibrium because the calcium cation polarizes the N2 quadrupole. The adsorption isotherm shows equilibrium loading of N2 roughly 1.5 to 2 times that of O2 on the same mass of 5A at 5 to 8 bar and 25 to 40 °C. To turn this equilibrium advantage into a PSA separation, you must run long adsorption half-cycles (5 to 15 minutes per bed), reach close to equilibrium, then regenerate by a deep pressure drop or vacuum. That is why 5A nitrogen PSA is rarely built industrially: the cycle is too slow, the bed is too large, and the air factor is too high.

Kinetic selectivity is what carbon molecular sieve (CMS) does at short contact times. CMS is a microporous carbon (typically coconut-shell-derived, activated to a precise pore-mouth size of 4.5 to 5.0 Angstrom) whose ultramicropores admit O2 (3.46 Angstrom kinetic diameter) faster than N2 (3.64 Angstrom). On a time scale of seconds to a few minutes, O2 loading on CMS exceeds N2 loading by a factor of 5 to 10, even though the equilibrium loadings would actually favor N2 if one waited for hours. The practical PSA cycle exploits this kinetic inversion: a 60 to 120 second half-cycle lets O2 enter the micropore network but does not let enough time pass for N2 to load. The N2 passes through as product; the O2 is held on the sieve and is released during depressurization.

Richter and colleagues at Bergbau-Forschung provided the foundational diffusion-coefficient data for O2 and N2 in CMS pellets in the 1980s and early 1990s, and Yang (Gas Separation by Adsorption Processes, Butterworth, 1987) and Sircar (Pressure Swing Adsorption literature reviews, Ind. Eng. Chem. Res.) generalized it into PSA cycle design. The CMS pellet of choice today is typically 1.0 to 1.6 mm (high-purity, short-cycle) or 1.5 to 2.0 mm (standard 99.0 to 99.5% duty), extruded or beaded carbon with controlled pore-mouth distribution, attrition below 0.1 wt% per ASTM D4058, and bulk density of 640 to 700 g/L.

Where 5A Zeolite Actually Fits in Nitrogen Service

5A zeolite is not the wrong material — it is the wrong material for the standard role people assign to it. In nitrogen service, 5A fits three legitimate niches:

1. Polishing beds for trace nitrogen removal from argon, helium, or hydrogen streams. Here the goal is to remove residual N2 from a non-air gas mixture at low pressure and very low concentration. 5A works because the partial pressure of N2 is small and the equilibrium advantage is enough at moderate cycle times.
2. Guard beds ahead of cryogenic nitrogen columns. Removing trace CO2 and water from feed air before it enters the cold box. This is a small adsorbent duty and 5A is well established for it.
3. Laboratory-scale nitrogen generators below roughly 5 Nm3/h where simplicity outweighs efficiency. A small two-bed 5A unit with 10-minute cycles and no vacuum is acceptable for an instrument lab or a school.

For any industrial duty above 5 Nm3/h, where cycle time, bed mass, energy, and air factor matter commercially, CMS is the standard answer. Throughout the rest of this article we will compare CMS baseline equipment to two design alternatives: a CMS plant with vacuum regeneration (VPSA) at the high-purity end, and a 5A nitrogen design at the long-cycle end. The 5A numbers are kept to remind the reader why they are not the standard industrial choice, and to keep the analysis honest.

The Air/Material Balance: Where 1.27 Comes From

Let's strip the nitrogen PSA problem down to atoms. Dry atmospheric air is roughly:

• 78.08 vol% N2
• 20.95 vol% O2
• 0.93 vol% Ar
• 0.04 vol% CO2
• trace H2, Ne, He, CH4

PSA removes O2, H2O, and CO2 efficiently. It does not separate argon from nitrogen, so the dry N2+Ar product from a single PSA stage has a fixed Ar/N2 ratio of about 0.93/78.08 = 1.18 vol%. Under the industry-standard purity convention (N2+Ar reported together as "N2 purity"), a single-stage PSA can reach about 99.5% N2 with the residual impurity being unseparated O2 plus trace CO2 and water. Under the strict convention (Ar counted as an impurity), the single-stage ceiling is 78.08 / (78.08 + 0.93) = 98.82 vol% N2. Vendors almost universally quote the loose basis; the basis must be locked into the procurement specification to avoid comparing apples to oranges.

To go above the 99.5% N2 single-stage ceiling, additional separation work is required. A second CMS PSA stage, a VPSA stage, or a membrane stage can push O2 down further. None of these removes argon, so the residual ~1% Ar floor remains unless a downstream cryogenic or catalytic getter stage is added. That detail matters when a buyer asks a vendor for "99.9% N2 from a single PSA": the only honest answer is to ask whether the number is dry, CO2-free, includes argon, and is from one bed or two in series.

The air factor F is defined as Nm3 of feed air (at standard conditions, dry) per Nm3 of N2 product. In a perfectly ideal separator that removes only oxygen and returns all nitrogen plus argon, the minimum F is:

F_min = 1 / (x_N2 + x_Ar) = 1 / (0.7808 + 0.0093) = 1 / 0.7901 ≈ 1.266 ≈ 1.27 Nm3/Nm3

Anything better than 1.27 is physically impossible because the nitrogen and argon are both on the product side of the membrane analogy. Real PSA plants run higher because:

• Some N2 is lost into the depressurization and purge streams (typically 25 to 55% of the feed N2)
• Mass transfer zones at the inlet and outlet of the bed use up 10 to 30% of the bed length without contributing fully to the separation
• Pressure drop across the adsorbent bed consumes 0.3 to 1.0 bar of the available pressure swing

The community shorthand "carbon balance" is occasionally used for these material flows, but it is misleading because there is no chemical carbon being tracked (apart from CO2 in the feed, which is tiny). The correct engineering term is air/material balance, with CO2 tracked separately. We will use that terminology throughout.

Purity Versus Air Factor: The Curve That Decides Plant Economics

Every PSA nitrogen vendor has a curve that says "99.5% N2 needs 3.0 Nm3/Nm3; 99.99% needs 6.5 Nm3/Nm3." The shape of that curve — exponential, not linear — is the engineering reason N2 gets dramatically more expensive as purity rises.

These ranges are typical of CMS-based designs and represent the broad middle of the market, with vendor-specific designs landing above or below depending on cycle time, pressure ratio, and CMS grade. The CMS manufacturer, the bed length-to-diameter ratio, and the rotary-valve timing all move the absolute number by roughly plus or minus 15%. The shape of the curve is what the engineering team should trust; the absolute numbers should be checked against the vendor's performance curve.

A practical sanity check: at 99.5% N2, 3.0 Nm3/Nm3, and 7.5 bar adsorption pressure, the specific power to drive the compressor (assuming isentropic efficiency 70%, adiabatic compression to 8 bar from atmospheric) is about 0.34 kWh/Nm3. If the air factor is 5.0, the same calculation gives 0.57 kWh/Nm3. If the real plant data deviates more than 20% from the rough physics, the cycle is being penalized by pressure drop, dead volume, or excessive purge.

Reading Vendor Purity Conventions Without Getting Burned

Ask three nitrogen PSA vendors for a quote and you will get three different purity numbers for the same physical air feed. The reason is that "purity" has at least four working definitions in this market:

• Wet basis vs dry basis. Most PSA nitrogen product streams leave the bed at a pressure dew point of -30 to -50 °C. If the analyzer samples upstream of the dryer, you read the dry value. If the analyzer samples downstream, you must convert. The O2 reading does not move with drying, but the CO2 reading does.
• Including argon vs argon-free. Argon does not separate from nitrogen in PSA. A vendor quoting 99.9% N2 may actually be delivering 99.0% N2 + 0.93% Ar + 0.07% O2, with the 99.9% figure computed on an "N2-only" basis that ignores argon. Lock the basis into the procurement spec.
• Continuous vs spot measurement. A 99.5% N2 average with brief excursions to 98.5% during a switching valve transition is not the same product as a steady 99.5% N2.
• Reference conditions. 0 °C, 1 atm or 20 °C, 1 atm or 25 °C, 1 atm change the absolute Nm3 number by roughly 9%.

The defensive move for a buyer is to ask the vendor for the same air mass flow and the same purity on a single agreed basis (dry, includes argon, continuous, 0 °C 1 atm) and to lock that into the procurement specification. Aluminaworld can, on request, normalise multiple vendor curves to a single agreed dry-with-argon basis for buyer-side comparison, and can audit the analyzer chain at acceptance. If the vendor refuses to clarify the basis, treat the number as marketing.

Cycle Design: Adsorption, Equalization, Depressurization, Purge, Repressurization

A standard two-bed CMS PSA cycle is six steps per bed, repeated every 60 to 120 seconds total. The exact step names vary by control vendor, but the physical sequence is:

1. Adsorption: compressed air at 6 to 10 barg enters bed A, N2 passes through and is buffered in the product receiver at 5 to 9 barg, O2 is adsorbed onto the CMS. The product end of the bed is at near-product pressure.
2. Pressure equalization 1 (provide): at the end of adsorption, the feed end of bed A is connected to the product end of bed B. The residual high-pressure gas in bed A transfers to bed B, recovering compression energy.
3. Depressurization: bed A is vented toward atmospheric pressure through a silencer plus partial recycle back to the feed. O2 desorbs. About 15 to 30% of the feed N2 is lost in this step alone, with another 10 to 25% lost across equalization and purge.
4. Purge: a fraction of product N2 is fed countercurrent (from the product end) to bed A to push desorbed O2 out of the mass transfer zone. The purge-to-product ratio is typically 0.10 to 0.25.
5. Pressure equalization 2 (receive): as bed B reaches the end of its own cycle, the equalization loop is reversed and the residual high-pressure gas in bed B tops up the repressurizing bed A.
6. Repressurization: bed A is brought back to feed pressure, by product-side topping from the equalization step (cleanest) followed by feed-side topping to design pressure.

A two-bed implementation gives a continuous product stream because bed B is always running a step that produces N2 while bed A is regenerating. Three-bed designs typically add a second equalization step and recover 50 to 65% of the compression energy, at the cost of more valves and a slightly longer minimum cycle. Four-bed layouts can add a third equalization step and push recovery into the 65 to 75% range.

VPSA designs (vacuum swing adsorption) replace step 3 (depressurization to atmosphere) with a deeper step 3 (depressurization to roughly 200 to 400 mbar absolute using a vacuum pump) and a similar vacuum purge in step 4. Because the bed ends the cycle at sub-atmospheric pressure instead of 1 bar absolute, the partial pressure of adsorbed O2 at the end of regeneration is roughly 3 to 5 times lower, and the equilibrium-loading at the start of the next adsorption step is lower. The result is higher working capacity, smaller beds, and lower air factor, particularly above 99.9% N2. The trade-off is capital cost of the vacuum pump and the need for a vacuum-tight vessel.

Cycle Time and Bed Sizing: Why 60 to 180 Seconds

CMS is a kinetic adsorbent, which means cycle time is constrained by diffusion: at long cycle times, the kinetic advantage of CMS over 5A collapses because N2 has time to enter the micropores. The practical cycle window for CMS nitrogen PSA is 60 to 180 seconds, with shorter cycles favored at high purity (faster switching keeps the mass transfer zone from growing). For a 99.5% N2 plant, the typical total cycle is 90 to 120 seconds; for a 99.99% N2 plant, 60 to 90 seconds. The diffusion coefficient of O2 in a typical CMS pellet is roughly 30 to 100 times that of N2 under the same conditions (depending on CMS grade and temperature), and this large kinetic ratio is what defines the cycle-time window; the working-capacity ratio that drives bed sizing is a much smaller factor of about 5 to 10. If the cycle is too short, the mass transfer zone never fully develops and the bed is under-utilized; if the cycle is too long, the kinetic selectivity collapses to the equilibrium ratio and nitrogen starts to co-load. Most industrial CMS plants are designed around a 90-second cycle for 99.5% N2 and 60-second cycle for high purity, with the rotary valve timing set within plus or minus 5% of the design point. Outside that band, air factor drifts up and the bed either starves during adsorption or wastes energy during regeneration.

Bed sizing follows from required throughput, bulk density (640 to 700 g/L for CMS pellets), and breakthrough time at design air factor. The rule of thumb is:

• For 99.5% N2, the CMS inventory is 60 to 100 kg per Nm3/h of product
• For 99.99% N2, 100 to 200 kg per Nm3/h
• Bed length-to-diameter ratio typically 2:1 to 4:1, with multiple beds in parallel for large plants

Going below 60 seconds total cycle is mechanically possible but pushes the rotary valve into higher wear and limits the time available for pressure equalization. Going above 180 seconds starts to lose the kinetic advantage and the air factor drifts upward.

Pretreatment: The Train That Decides Whether the CMS Survives

The CMS bed is the most expensive replaceable adsorbent in the plant, and its end-of-life drivers are almost entirely upstream contamination: water, oil, and particulates. The pretreatment train is therefore the single most important engineering subsystem and the place where most plant failures originate.

The desiccant dryer stage is the one that actually protects the CMS from moisture damage. It is typically a twin-tower adsorption dryer using activated alumina (3 to 5 mm beads) as the working adsorbent, sometimes with a small 13X layer for additional water capacity. Activated alumina is preferred over silica gel for this duty because it dusts significantly less, it tolerates liquid water slug better, and it is regenerable at typical desiccant-dryer heater setpoints of 150 to 200 °C (heater-outlet). The CMS itself should see feed dew point below -40 °C (typically -40 to -70 °C at high-purity plants).

A separate activated carbon or coalescing filter is required upstream of the dryer to remove compressor oil aerosols. Synthetic compressor oils (polyglycols, diesters, PAOs) tend to polymerize inside the carbon micropores of CMS at the operating temperature and cannot be removed by regeneration. Once polymerized, the affected CMS micropores are irreversibly lost. This is the most common cause of premature CMS replacement in poorly maintained plants.

Moisture Damage: Reversible or Irreversible?

There is a common myth in the field that "any water on CMS is fatal." It is not. The reality is more nuanced:

Physically adsorbed water at modest loadings (up to about 5 to 8 wt%) is in many cases recoverable. A controlled dry purge with dry N2 or dry air at 60 to 120 °C for several hours can desorb the water from the micropores and bring CMS capacity back to within 80 to 95% of the original. This is the routine regeneration behavior of a desiccant dryer tower: the CMS is exposed to brief wet peaks and recovers every cycle.

Liquid water slug (a drain failure, a refrigerated-dryer bypass, an accidental condensate flood) is more serious but not always fatal. If the wetting event is short and the bed can be brought back to dry conditions within 24 hours, thermal reactivation at 150 to 200 °C under dry N2 purge often recovers 70 to 90% of the original capacity.

Oil or hydrocarbon contamination is the irreversible case. Synthetic compressor oils and certain hydrocarbons polymerize inside the CMS micropores at the PSA operating temperature and cannot be removed by any reasonable regeneration. The CMS has to be replaced, and the cause has to be fixed first to avoid the same outcome two years later. This is why oil removal upstream of the desiccant dryer is a critical subsystem and not an optional filter.

Procurement specifications should therefore state:

• Maximum feed pressure dew point: -40 °C at all operating loads (with explicit alarm at -30 °C)
• Maximum feed oil content: 0.01 mg/Nm3 (with alarm at 0.05 mg/Nm3)
• Maximum particulate: Class 2 per ISO 8573-1:2010
• Liquid water alarm and drain fail alarm as safety interlocks

What a 5A Nitrogen PSA Looks Like (and Why It Loses)

It is worth quantifying the alternative to drive the point home. A 5A zeolite nitrogen PSA at equilibrium cycle would operate as follows:

• Adsorption half-cycle: 5 to 15 minutes per bed
• Adsorption pressure: 6 to 10 barg (similar to CMS; the 5A working capacity for N2 increases with feed pressure, and the cycle is constrained by equilibrium kinetics, not by an absence of a pressure-driven capacity gap)
• Regeneration: countercurrent depressurization to 100 to 200 mbar absolute, because the equilibrium advantage of 5A for N2 over O2 needs deep vacuum to reverse
• Air factor for 99% N2: 4.5 to 7 Nm3/Nm3 (1.5 to 2.5 times higher than CMS at the same purity)
• Specific power for 99% N2: 0.6 to 0.9 kWh/Nm3

For a 99.9% N2 duty the 5A numbers get worse, not better, because each step up in purity costs more air at equilibrium than at kinetic selectivity. The conclusion is the same as the headline: 5A is the wrong tool for industrial nitrogen PSA. It is the right tool for narrow polishing and guard duties, and we will come back to those in a moment.

International Standards Governing CMS Nitrogen PSA

Procurement, safety, and performance specs all hang on standards. The principal documents that apply (or are commonly invoked) to a CMS nitrogen PSA plant are:

• ISO 8573-1:2010 Compressed air — Contaminants and purity classes. Used to specify dryer outlet dew point (Class 2 or 3), particulate (Class 2), and oil (Class 1).
• ISO 1217 Acceptance tests for compressors and specific power measurement. Used to verify vendor claims about compressor efficiency.
• ISO 14175 Gas classification for welding end-uses. Applies only when the N2 product is being used as a backing or shielding gas in welding; it does not classify the PSA adsorbent itself.
• ASME Boiler and Pressure Vessel Code Section VIII, EN 13445, and PED 2014/68/EU for pressure vessels including the adsorber shells, receivers, and buffer tanks.
• IEC 61511 Functional safety — Safety instrumented systems for the process industry sector. Applies where shutdown interlocks are classified as safety functions (high-pressure trip, O2 analyzer trip on product).
• ISO 9277 BET surface area measurement of porous solids, used in the CMS Certificate of Analysis.
• ISO 13320 Particle size distribution by laser diffraction, used for CMS lot acceptance.
• ASTM D4058 Standard test method for attrition and abrasion of catalysts and catalyst carriers (used here as an agreed comparative method for adsorbent attrition). Do not invent clauses or pin the lot acceptance to a clause number that does not exist.

CMS and Aux Adsorbent Specifications: What To Ask The Vendor

A useful procurement specification is one that is verifiable lot by lot. The following table is a starting point; specific projects may need to tighten or relax individual rows.

Lock the test methods into the purchase order. A CoA that says "passes internal QC" without a method reference is not verifiable and is not negotiable in a supply contract dispute.

Activated Alumina and 13X in the Pretreatment Train

Aluminaworld's published product catalog covers molecular sieves (3A, 4A, 5A, 13X) and activated alumina. Several of these grades have direct roles in a nitrogen PSA pretreatment train even though they are not the CMS itself.

Activated alumina in the desiccant dryer. The desiccant air dryer upstream of the CMS is the workhorse moisture-removal stage. Activated alumina (2.5 to 5 mm beads or 3 to 5 mm beads, BET 280 to 360 m2/g) is the standard desiccant. It has a higher equilibrium water capacity than silica gel at low dew points and tolerates liquid water slug better. Regeneration is at 150 to 200 °C heater-outlet with a small fraction of dried air, and cycle time is typically 4 to 8 hours. The typical dryer bed sizing for a 100 Nm3/h plant at -40 °C pressure dew point is 200 to 350 kg of activated alumina per tower.

13X as a CO2 and moisture guard bed ahead of the CMS. Where the feed air has elevated CO2 (urban air, near combustion sources, indoor air intakes), a small 13X guard layer at the feed (inlet) end of the CMS bed, or as a separate small guard vessel between the desiccant dryer and the CMS, captures CO2 and residual moisture that would otherwise load and foul the CMS over time. The 13X removes CO2 preferentially because of its ~10 Angstrom pore opening, and loads very slowly because the partial pressure of CO2 in air is only about 420 ppm. A 5 to 10 cm 13X layer ahead of the CMS, or a small separate guard vessel, extends CMS life on dirty air by 1 to 2 years.

Activated carbon / coalescing filter for oil. Independent of the desiccant dryer, a separate activated carbon or fine coalescing filter is the only reliable stage to remove compressor oil aerosol down to the <0.01 mg/Nm3 level. Activated carbon is preferred for vapor-phase oil removal; coalescing filters are preferred for liquid aerosol. Either is acceptable as long as it is upstream of the desiccant dryer and rated for the working pressure.

Selection Guide: Which Adsorbent Per Duty

Putting the above together, a buyer-side decision tree looks like:

• Industrial N2 at 95 to 99.5% (general plant use, food-grade, tire inflation, wine preservation) → CMS in two-bed PSA with atmospheric regeneration, activated alumina dryer upstream. No vacuum pump needed.
• Pharma / electronics N2 at 99.9 to 99.99% → CMS with atmospheric regeneration plus VPSA option; or two CMS stages in series; activated alumina dryer plus 13X CO2 guard bed; oil-free compressor mandatory.
• Semiconductor / specialty gas at 99.999% N2 → CMS first stage to 99.9% then VPSA or membrane second stage; or CMS first stage plus catalytic O2 getter for residual 100 ppm; treated as a near-cryogenic equivalent.
• Trace N2 removal from argon or helium → 5A zeolite polishing bed (very different cycle, much smaller plant) — not a CMS duty.
• Lab-scale N2 below 5 Nm3/h → small membrane N2 unit or a bench-scale CMS unit; 5A only acceptable if size and energy are not constrained.

10-Year TCO Comparison: Illustrative Worked Example

This example is labelled illustrative, not derived from a specific customer plant. It uses transparent formulas and assumptions you can adapt to your own electricity price, capital cost, and duty cycle.

Assumed duty: 50 Nm3/h N2 at 99.5% purity, 6,000 operating hours per year, electricity at $0.10/kWh, adsorption pressure 7.5 barg, desorption pressure 1.05 bara, no vacuum pump. Two designs are compared: (A) CMS two-bed PSA, atmospheric regeneration; (B) 5A two-bed PSA, vacuum regeneration (because atmospheric regeneration cannot reach 99.5% on 5A at reasonable cycle).

The conclusion: even though 5A is roughly one-third the price per kilogram, the higher air factor, the larger bed, the vacuum pump capex, and the faster replacement schedule more than offset that advantage at 99.5% N2 and 50 Nm3/h. The CMS plant is cheaper to buy and dramatically cheaper to run over 10 years.

For larger plants the conclusion gets stronger: at 500 Nm3/h, the air-factor and energy differential between CMS and 5A at 99.5% N2 scales to about $150,000 per year of electricity. Over 10 years that is $1.5 million in operating cost, against an adsorption-stage capex differential of a few hundred thousand dollars.

Procurement Acceptance Specification: How To Lock It In

The most common procurement failure is a contract that quotes a performance number ("delivers 99.5% N2") without a measuring protocol, an acceptance test window, and a remedy for failure. A defensible acceptance spec for a new nitrogen PSA should include:

• Feed air conditions (pressure, temperature, humidity, oil content, particulate class)
• Product N2 flow at reference conditions (typically 0 °C, 1.013 bar, dry)
• Product purity, with explicit basis (dry, includes argon, continuous, 99.5 vol% min)
• Specific power at design air factor (vendor warrants kWh/Nm3)
• Air factor at design purity (vendor warrants F)
• Pressure dew point of product (vendor warrants -40 °C min)
• Noise level at 1 m (typically ≤80 dBA)
• Vibration and valve cycle warranty (rotary valves typically rated 50 million cycles)
• Acceptance test on site, witnessed by buyer, with 24 to 72 hour steady-state run at full load

A useful penalty / remedy clause pays the vendor a portion of holdback until the acceptance test passes; this is standard in the global compressor and PSA market.

Operating Monitoring: KPIs That Tell You When To Change The Bed

Operators need three continuous numbers to decide when a CMS bed is reaching end of life. None of them is "time on stream."

• Product O2 trend at design cycle time. A clean CMS plant runs steady at the design purity for years. A creeping O2 trend (slowly rising baseline with the same cycle time) is the leading indicator of capacity loss. Plan CMS replacement when the trend reaches 80% of the alarm setpoint.
• Bed pressure drop. A clean CMS bed has a stable pressure drop during the adsorption half-cycle. A rising pressure drop indicates fines generation (attrition), bed compaction, or plugging of the top layer by desiccant fines from upstream. Investigate before changing the bed; sometimes the upstream cause is the real fix.
• Desiccant-dryer outlet dew point. A stable -40 to -50 °C pressure dew point is the protective envelope. Drift toward -20 °C is the early warning that the activated alumina is exhausted and that the CMS will start loading water. Replace the desiccant before the alarm trips.

A 10-minute scheduled walk-down once per shift, plus weekly review of the three KPIs above, is enough to predict CMS end of life to within roughly 6 months at typical industrial plants. Skipping the walk-down is the single most common root cause of unscheduled CMS replacement.

7 Common Mistakes When Specifying a Nitrogen PSA

1. Treating CMS and 5A as interchangeable in nitrogen service. Different mechanism, different cycle, different air factor. The air factor penalty alone can double the operating cost of a 5A plant vs a CMS plant at the same purity.
2. Specifying "99.9% N2" without an argon basis. Without argon in the statement, a 99.9% number can hide an actual N2 + Ar purity closer to 99.0%. Lock the basis into the contract.
3. Letting the desiccant dryer drift. The most common root cause of premature CMS replacement. The dryer is a cheaper, easier-to-replace adsorbent; never let its exhaustion compromise the CMS.
4. Choosing the wrong compressor lubrication strategy for the purity target. Mineral-oil-lubricated compressors with a properly sized coalescing plus activated-carbon filter train upstream of the desiccant dryer are routinely used for non-food, non-pharma industrial nitrogen PSA at 95 to 99.5% purity. Oil-free scroll or oil-free screw compressors are strongly recommended above food-grade and pharmaceutical purity (and required for USP <467>- and EMA Annex 1-compliant product), because synthetic compressor oils polymerize inside the CMS micropores and are not removable by regeneration.
5. Skipping pressure equalization. Removing the equalization step "to simplify the controls" costs 10 to 20% of the air factor. The 5 to 10% capex saving is not worth the lifetime operating cost.
6. Skipping the redundant analyzer. A single O2 analyzer with no cross-check is a single-point failure. For PSA nitrogen in the 0.01 to 1% O2 range, use a paramagnetic primary with an electrochemical or fuel-cell cross-check, or two paramagnetic analyzers from different manufacturers, on the high-purity product stream.
7. Buying on per-kg CMS price. The right comparison is per Nm3 of N2 delivered over the bed lifetime. The cheapest CMS per kg is rarely the cheapest per Nm3.

Energy and Carbon: The Quiet Case For PSA Nitrogen

Industrial nitrogen has three main supply modes: PSA on-site generation (this article), cryogenic on-site generation (large plants, typically >500 Nm3/h), and merchant liquid N2 delivered by road tanker. Each has a different carbon and cost profile. Roughly:

• PSA at 99.5%: 0.30 to 0.45 kWh/Nm3 N2
• Cryogenic on-site at 99.999%: 0.25 to 0.40 kWh/Nm3 N2 with much higher capital cost, only economic above 500 Nm3/h
• Merchant liquid N2: 0.8 to 2.0 kWh/Nm3 embodied when transport energy is amortized (truck delivery plus boil-off losses)

For a 50 to 200 Nm3/h industrial plant, on-site PSA is typically 30 to 60% lower in energy and carbon than merchant liquid delivery, and lower in capital cost than cryogenic on-site. The energy penalty for raising purity from 99.5% to 99.99% in PSA is roughly 1.5 to 2.0x in kWh/Nm3, so purity should be sized to end-use need, not aspirational.

Compressed-air system audits show that 30 to 50% of the energy in a typical industrial compressor room is wasted on leaks, pressure set too high, and untreated air being used where treated air is not needed. Before buying a new PSA, an air-leak survey and a pressure profile audit on the existing compressor room can reduce the N2 plant size needed and pay for the PSA in 12 to 24 months at 50 Nm3/h.

Common CMS Defects And How To Diagnose Them

After 15 years of supplying adsorbents and reviewing field returns, the typical failure modes on CMS beds break into roughly five buckets. Diagnosing the right bucket drives the right corrective action.

Loss of N2 capacity without rising pressure drop. This is usually moisture or oil poisoning. Pull a bed sample and send it for thermogravimetric analysis (TGA) and BET. If moisture is up, run an offline thermal reactivation. If oil hydrocarbons are present, replace the bed and fix the upstream cause.

Rising pressure drop across the bed. Three candidates: (a) fines from attrition (check upstream dust filter, sample the bed top); (b) compaction from cyclic loading (recheck bed support, hold-down, and design velocity); (c) fines from upstream desiccant dryer that have migrated through the after-filter (replace the after-filter and possibly the desiccant).

Premature O2 breakthrough. Check the cycle time and the equalization settings first. If unchanged, the CMS is at end of life or the feed conditions have drifted (higher temperature, higher dew point, higher CO2). Restore design conditions and re-baseline.

Visible dust in product stream. Almost always the after-filter is exhausted, or the bed top has fluidized. Replace the after-filter, sample the bed top, and consider hold-down redesign.

Co-elution of CO2 with O2. CMS does not remove CO2 well. If CO2 rises in the product, check the 13X guard bed (if used) or the upstream CO2 source. A small CO2 removal stage in front of the dryer usually fixes this.

What Aluminaworld Supplies For A Nitrogen PSA Plant

Aluminaworld's published product catalog focuses on molecular sieves (3A, 4A, 5A, 13X) and activated alumina. Several of these grades fit directly into the pretreatment and guard duties of a CMS nitrogen PSA plant:

• Activated alumina for the desiccant air dryer (the workhorse bed that protects the CMS from moisture). Available 2.5 to 5 mm and 3 to 5 mm beads, BET 280 to 360 m2/g, attrition below 0.5 wt% per ASTM D4058.
• 13X molecular sieve for the CO2 and moisture guard bed at the feed (inlet) end of the CMS, or as a polishing layer ahead of a cryogenic column. Available 1.6 to 2.5 mm beads, BET above 600 m2/g.
• 5A molecular sieve for trace N2 polishing in argon / helium streams (a genuine 5A duty), or for the laboratory-scale nitrogen case where simplicity dominates.
• 3A and 4A molecular sieves for instrument-air drying and for drying the regenerant stream of the desiccant dryer.
• Particulate and oil-removal filter elements as accessories through partner supply.

Aluminaworld does not manufacture CMS in-house; for the CMS itself we are happy to refer the buyer to qualified CMS producers (such as the major carbon-adsorbent suppliers in Europe and Asia) and to supply the complete auxiliary adsorbent package, the dryer sizing, and the CoA documentation that the CMS vendor will typically request. If your procurement team is specifying a turnkey nitrogen generator, contact us early in the project and we will work out the auxiliary adsorbent BOQ against your PSA vendor's BOP.

Evaluating a CMS Vendor: 8 Questions To Ask

1. What is the kinetic O2 working capacity of the supplied grade, measured between the adsorption pressure the design uses (typically 4 to 8 barg) and the desorption pressure (1.0 bar for atmospheric regeneration, 200 to 400 mbar absolute for VPSA) at 25 °C, with the test apparatus, sample mass, and equilibration time named — and what is the corresponding equilibrium O2 capacity at the same conditions?
2. What is the attrition per ASTM D4058, and on what sample mass?
3. What is the bulk density band and the PSD band per ISO 13320?
4. What reference PSA plants has this exact grade run in, and for how long?
5. What is the recommended regeneration profile (cycle time, equalization steps, purge ratio)?
6. What pretreatment does the warranty assume (dew point, oil, particulates)?
7. What is the warranty remedy if the bed fails earlier than design life?
8. What is the lot sample retention policy (typically 6 to 12 months)?

PSA Versus Membrane And Cryogenic: Where CMS Plants Win

The three industrial nitrogen supply technologies compete on cost, purity, and scale. A short buyer-side summary:

• Membrane N2 is the lowest capital cost, simpler to operate, but limited to about 95 to 99.5% N2 and relatively high specific power above 99% purity. Best fit for small, intermittent, lower-purity needs.
• PSA CMS nitrogen (this article) covers the middle of the market at 95 to 99.99% N2 with the lowest operating cost at moderate purity. Best fit for 20 to 1000 Nm3/h continuous duty at 99.0 to 99.9% N2.
• Cryogenic on-site is justified only above roughly 500 Nm3/h of high-purity N2 (above 99.99%) where scale and steady demand make the higher capex pay back.
• Merchant liquid N2 remains the right answer for very small peak demands and for users who need periodic high-purity cylinders or LN2.

Field Engineer Pre-Commissioning Checklist

When a nitrogen PSA is being commissioned, the auxiliary adsorbent package (desiccant dryer, 13X guard, oil removal) is the section most often skimped. The following 12 checks cover the points that decide long-term CMS health:

1. Aftercooler drain functioning and routed to a safe drain.
2. Pre-filter installed with correct flow direction; pressure drop indicator at 0.5 bar typical change-out.
3. Refrigerated dryer outlet dew point confirmed below +5 °C at full load.
4. Oil-removal filter (coalescing or activated carbon) installed upstream of the desiccant dryer; outlet analyzer reading below 0.01 mg/Nm3.
5. Desiccant dryer outlet dew point confirmed below -40 °C at full load.
6. Desiccant dryer regeneration heater cycling within vendor spec; no hot spots.
7. After-filter (0.01 μm) installed between the dryer and the CMS inlet.
8. CMS bed bed-support screen, hold-down, and top layer correctly installed per drawings.
9. First adsorption cycle watched end-to-end with the vendor commissioning engineer.
10. Product O2 analyzer calibrated against a known reference gas (typically 99.5% N2 ± 1% certified).
11. Product pressure regulator set to design pressure; product flow measured with calibrated meter.
12. Trip alarms tested and interlock logic exercised (high O2, high bed delta P, dryer failure).

A Field Walk-Through: What End-Of-Life Looks Like on a Real Plant

It is helpful to anchor the engineering numbers above with one realistic end-of-life story, kept deliberately generic so that no customer is identifiable. The plant runs a 200 Nm3/h nitrogen line supplying a tire-inflation room; the product is 99.5% N2 on a dry basis, argon included; the desiccant dryer is sized at roughly 600 kg of activated alumina per tower (which lines up with the rule-of-thumb of 200 to 350 kg per 100 Nm3/h of air feed plus margin); the CMS bed holds roughly 13,000 kg of CMS pellets across two parallel beds of 6,500 kg each (consistent with the 60 to 100 kg CMS per Nm3/h rule-of-thumb for 99.5% N2), commissioned six years ago; the compressor is oil-injected screw with a downstream coalescing filter.

In year four the operator notices a slow drift: the N2 product purity, measured downstream of the product buffer, has crept from 99.6% to 99.4% over six months. The cyclic valve timing has not changed. The dryer outlet dew point reads -38 °C consistently. The plant increases the adsorption half-cycle by 4 seconds in an attempt to recover purity, succeeds for two months, then begins to lose purity again as the cycle-time compensation hits the rotary-valve minimum. The diagnostic is unambiguous: the CMS bed is reaching end of useful life. Air factor has quietly risen from 3.0 to about 3.4 Nm3/Nm3 to maintain the design flow, which is the trade-off the plant is paying in compressor energy.

Two proposals go on the table. Plan A: replace the CMS only. Plan B: replace the CMS plus the desiccant dryer activated alumina. The activated alumina is on a 6 to 8 year life and is showing some channeling at the top of the bed. The operator chooses Plan B; plant downtime is 36 hours; the new bed delivers 99.7% N2 within an hour of startup, and air factor returns to 2.95 within a week of operation. The cost is the CMS change-out plus about 30% of the dryer activated alumina plus the labor of a single shift. The decision saves roughly $30,000 a year in compressor electricity for the next five years relative to running on the deteriorating bed, and it forecloses the risk of an unplanned CMS event during a peak demand week.

This is what end-of-life looks like in practice: a gradual drift in the KPIs that operators are instructed to watch, a trade-off between air factor and purity, and a coordinated change-out of the most-stressed adsorbent beds in the train. There is no calendar date. There is no vendor warranty without a verified performance curve. The article's recommendation to monitor three KPIs continuously is not a paperwork chore; it is the difference between a planned 36-hour shutdown and an unscheduled 5-day outage during a peak demand month.

References And Authoritative Sources

The recommendations and numbers in this article draw on the following non-proprietary sources. No fabricated URLs or clause numbers are included.

• Yang, R.T., Gas Separation by Adsorption Processes, Butterworth (1987) — foundational PSA cycle analysis.
• Sircar, S., "Pressure Swing Adsorption" reviews, Industrial & Engineering Chemistry Research — applications and cycle thermodynamics.
• Richter, E. and colleagues, Bergbau-Forschung CMS diffusion-coefficient studies (Carbon, mid-1980s to early 1990s) — original CMS kinetic-selectivity data for O2 and N2.
• ISO 8573-1:2010, Compressed air — Contaminants and purity classes.
• ISO 1217, Acceptance tests for compressors.
• ISO 14175, Gases for welding — classification.
• ISO 9277, BET surface area of porous solids.
• ISO 13320, Particle size distribution by laser diffraction.
• ASTM D4058, Attrition and abrasion of catalysts and catalyst carriers.
• ASME Boiler and Pressure Vessel Code, Section VIII.
• EN 13445, Unfired pressure vessels.
• PED 2014/68/EU, Pressure Equipment Directive.
• IEC 61511, Functional safety — Safety instrumented systems for the process industry.

Related Reading On Aluminaworld

Several earlier Aluminaworld articles cover the auxiliary adsorbents and procedures referenced above. They are useful for cross-checking the pretreatment train, the bed sizing workflow, the regeneration best practices, and the procurement spec:

• 5A vs 13X for PSA Hydrogen: Pore Size and Capacity
• Molecular Sieve for O2 Concentrator: LiLSX vs 5A vs 13X Performance Comparison
• Molecular Sieve Attrition Rate: Method, Measurement, and Limits
• Molecular Sieve Regeneration Best Practices: Temperature, Purge, and Time
• How to Calculate Molecular Sieve Replacement Frequency
• How to Verify Molecular Sieve Quality: A Pre-Shipment Checklist

Frequently Asked Questions

Next Steps: Talking to Aluminaworld About Your Nitrogen Project

If you are designing, specifying, or troubleshooting a PSA nitrogen plant, the article above gives you the mass balance, the purity versus air factor trend, the pretreatment train, the standards, and a transparent 10-year TCO comparison. The next step is to translate these to your specific duty: required N2 flow, required purity, feed air conditions, electricity price, and the capex ceiling your project team is working to. We can help on the adsorbent side of those numbers, and we are happy to coordinate with your PSA vendor or EPC on the auxiliary bed BOQ.

For CMS nitrogen pretreatment, desiccant dryer sizing, 13X guard beds, or full BoQ support for a new generator, contact us via:

• WhatsApp: +86 133 2522 2240 (fastest, 12-hour reply) — message: "Hi Aluminaworld, I need help with CMS PSA nitrogen pretreatment sizing and media qualification."
• Email: barry@aluminaworld.com
• R&D sample: 5 kg activated alumina or 13X guard bed, 7-day lead time, full CoA included
• Bulk order: 500 kg MOQ, 15-day production, FOB/CIF/CFR from Qingdao Port (80 km from our factory)

Aluminaworld has supplied molecular sieve and activated alumina to industrial gas, food, pharmaceutical, electronics, and welding customers across 60+ countries for 15 years under ISO 9001 quality control. Pretreatment design for CMS PSA nitrogen is one of our standard technical briefs; please send us your flow, purity, and feed dew-point targets and we will respond with a sized adsorbent BoQ and a lifecycle cost model within two working days.