Molecular Sieve Drying for Green Hydrogen Electrolyzer Output: -70 degC Dew Point Spec and Deoxo Bed Integration

Why a separate drying train is required after the electrolyzer

Hydrogen produced by either proton-exchange-membrane (PEM) or alkaline water electrolysis leaves the cell saturated with water at the cell operating temperature. A PEM stack at 60 to 80 degC delivers hydrogen at roughly 20 to 30 mol% water before any gas-liquid separation. An alkaline stack at 70 to 90 degC behaves similarly on the H2 side. After the standard gas-liquid separator and reflux condenser, residual water in the gas stream is still 1,000 to 3,000 ppmv, far above the limits set by ISO 14687 for fuel-cell-grade hydrogen and well above the -40 degC pressure dew point that is the standard for pipeline injection. A dedicated drying train, almost always built around molecular sieve, is therefore a required downstream subsystem of every green hydrogen plant.

The drying train must hit three constraints simultaneously: water concentration, oxygen concentration, and (for fuelling stations) trace hydrocarbon and CO limits. The first constraint is the one that drives the molecular sieve choice and the regeneration loop design. The second and third constraints are met by placing a deoxo catalyst downstream of the main dryer and a polishing sieve after the deoxo reactor to mop up the water the deoxo reaction generates. This guide walks through each subsystem, the standards that apply, and the field engineering decisions that matter most on a real plant.

Dew point targets: -40, -70, and -75 degC pressure dew point explained

The dew point target is set by the downstream use of the hydrogen and is the first design decision. Three standards cover the most common cases:

The -70 degC target corresponds to roughly 2.5 ppmv water at 30 barg and is the design point most PEM-fuelling-station dryers use. A two-bed molecular sieve dryer running 4A grade and regenerated by heated product hydrogen at 220 to 260 degC reaches the target with cycle times of 8 to 12 hours and a regeneration gas purge ratio of 4 to 6%. A heatless (pressure-swing) design can also meet the dew point but consumes more hydrogen because it cannot recover the heat of adsorption, so the purge ratio rises to 10 to 15%.

For LH2 storage or for fuel-cell systems with on-board reforming that are particularly sensitive to water, plants design to -75 degC. This adds an extra polishing stage and, in cold climates, a refrigerated pre-cooler upstream of the sieve train. Adding a third sieve stage has diminishing returns because the limiting factor becomes the bead's equilibrium water capacity at the regeneration temperature, not the number of stages.

Why 4A is the workhorse grade for hydrogen drying

The three common molecular sieve pore sizes for hydrogen service are 3A (3 angstrom), 4A (4 angstrom), and 5A (5 angstrom). 13X (10 angstrom) is occasionally used as a co-bed for CO2 removal upstream of the deoxo bed. The dominant 2024 to 2026 specification is 4A, with the following reasoning.

3A is preferred when the stream temperature is below about 10 degC, when CO2 is present at more than 100 ppmv, or when the dew point spec is -75 degC or colder. 4A is the default for everything else because it has the same equilibrium water capacity as 3A but a slightly faster adsorption rate and a lower cost per kilogram. 5A is the wrong choice for hydrogen drying because its pore admits methane, ethane, and other light hydrocarbons that the electrolyzer may carry in trace amounts, and these hydrocarbons are very hard to remove downstream of the sieve. 13X is reserved for the CO2 pre-bed when the inlet CO2 is above a few hundred ppmv, which is uncommon on a PEM electrolyzer but occasionally seen on an alkaline plant fed with imperfectly scrubbed CO2 from the oxygen side.

For a standard 100 Nm3/h PEM hydrogen plant, a typical 4A loading is 200 to 350 kg in the main bed and 60 to 100 kg in the polishing bed. Cycle time is 8 to 12 hours on heated purge or 4 to 6 hours on heatless. The total sieve inventory per plant is therefore 500 to 900 kg, depending on whether the design uses TSA or PSA regeneration and on whether the design includes an activated alumina pre-bed upstream.

Why the deoxo bed goes after the dryer, not before it

A common question from electrolyzer plant engineers is whether to put the deoxo catalyst upstream or downstream of the molecular sieve. The answer is firmly downstream, and the reason is the deoxo catalyst's water sensitivity. Palladium and platinum catalysts on alumina or hopcalite supports lose activity when the inlet water partial pressure exceeds roughly 100 ppmv, and they are typically warranted for stable operation only below 50 ppmv water. A two-stage molecular sieve dryer that achieves -70 degC dew point delivers about 2.5 ppmv water, well inside the safe operating window for the catalyst.

The other reason is that the deoxo reaction itself generates water. The chemistry is 2 H2 + O2 -> 2 H2O, so for every mole of oxygen removed the catalyst produces a mole of water. This water must be caught by a polishing molecular sieve downstream of the deoxo bed, otherwise the final product exits above the ISO 14687 Grade D 5 micromol/mol water limit. The standard train therefore looks like:

(1) KO drum and reflux condenser, (2) refrigerated after-cooler, (3) activated alumina pre-bed, (4) 4A molecular sieve main bed, (5) 4A molecular sieve polishing bed, (6) heater (if TSA) and deoxo catalyst bed, (7) 4A molecular sieve post-deoxo polishing bed, (8) product. Putting the deoxo catalyst before the main dryer would force the main dryer to handle both the original water load from the electrolyzer and the deoxo-generated water, which doubles the bed mass and the regeneration duty.

Regeneration loop design: heatless PSA versus heated-purge TSA

Two regeneration schemes are used on green hydrogen dryers. The first, called heatless or pressure-swing adsorption (PSA), swings the bed from the high adsorption pressure (typically 30 barg) to a low regeneration pressure (typically 4 to 7 barg) without heating. The pressure swing plus a small fraction of product hydrogen flowing counter-current through the bed desorbs the water. The trade-off is that the regeneration gas purge ratio is high: typically 10 to 15% of the recovered H2 stream must be used as purge, and this purge gas is either vented or recycled through the electrolyzer. On a 100 Nm3/h plant, a heatless dryer consumes 10 to 15 Nm3/h of product hydrogen as purge.

The second scheme, called heated-purge or thermal swing adsorption (TSA), adds an electric or steam heater and a heat exchanger to the regeneration loop. The bed is regenerated at the same pressure but at 220 to 280 degC, which is well above the desorption temperature of water on 4A molecular sieve. The purge ratio falls to 3 to 6% because the heat input does most of the desorption work, and the cycle time can be extended to 8 to 12 hours. The trade-off is capital cost and the need for stainless steel piping downstream of the heater, because hot hydrogen at 250 degC attacks carbon steel and can cause decarburization and embrittlement in long-term service.

For fuelling stations where product hydrogen has the highest value and where lost purge hydrogen represents a direct revenue loss, TSA is the dominant choice even at higher capital cost. For industrial pipelines where hydrogen is sold at lower pressure and where the purge gas can be returned to the electrolyzer at near-ambient pressure, heatless PSA remains common. Either scheme can meet the ISO 14687 Grade D water limit when sized correctly, but TSA offers more headroom for the -75 degC specification.

Liquid carryover and the activated alumina pre-bed

Liquid water is the worst-case event for a molecular sieve bed because it locally saturates the beads, causes caking, and on regeneration produces sharp thermal gradients that fracture beads and generate fines. Three protections are mandatory.

The first is a stainless steel mesh demister pad in the knockout drum, sized for 99% removal at the design gas velocity (typically 0.5 to 1.0 m/s through the pad). The second is a refrigerated or fan-cooled after-cooler set 5 to 7 degC above the gas inlet temperature to drop the bulk of the water by condensation before the demister. The third, and the one that catches the operating errors, is an activated alumina pre-bed installed immediately upstream of the molecular sieve main bed.

The pre-bed is sized to handle 5 to 10 minutes of saturated inlet flow in an upset without saturating the downstream sieve. For a 100 Nm3/h hydrogen stream this typically means 60 to 120 kg of 5x7 or 7x14 mesh activated alumina with BET surface area above 280 m2/g. The pre-bed is regenerated on the same cycle as the main sieve and consumes the bulk of the water swing, which is much cheaper to replace than the high-grade molecular sieve. The pre-bed also buffers the sieve from sudden pressure swings that follow electrolyzer load changes, which are the most common source of liquid carryover in operating plants.

Sizing the beds: velocity, L/D ratio, and approach temperature

Three engineering parameters decide whether a molecular sieve dryer meets its dew point target on a real plant.

The first is approach velocity, the superficial gas velocity through the bed, which should be 0.10 to 0.18 m/s for hydrogen service. Below 0.08 m/s the bed is oversized and the regeneration gas cannot effectively sweep the desorbed water. Above 0.20 m/s the bed is undersized and the residence time is too short to reach equilibrium, producing a high effluent water concentration. The second is the L/D ratio (bed length to bed diameter), which should be 2.5 to 4.0 for hydrogen. Lower than 2.5 produces channelling at the vessel walls; higher than 4.0 produces excessive pressure drop and uneven regeneration. The third is approach temperature, the difference between the gas inlet temperature and the dew point of the inlet stream, which should be 5 to 10 degC. A larger approach temperature is fine but inflates the bed size; a smaller approach temperature risks breakthrough.

Pressure drop across the bed is calculated with the Ergun equation and should be below 0.3 bar for the entire bed at design flow. Higher pressure drop means higher blower or compressor power and a more frequent sieve replacement schedule, because high pressure drop correlates with fines accumulation and bed compaction over time. Buyers specifying a molecular sieve for a green hydrogen dryer should ask the supplier for both the equilibrium water capacity curve at 30, 60, 90, and 120 degC and the pressure drop curve at 0.10, 0.15, and 0.20 m/s approach velocity. The curves should be lot-specific and tied to the sieve's mesh size and binder composition.

Seven QC checks for incoming molecular sieve lots

Most field failures on green hydrogen dryers trace back to incoming material that did not meet specification. Eight checks catch the majority of these failures. The first is equilibrium water capacity per ISO 17607 or per the supplier's gravimetric method at 30 degC and 50% RH, target above 20 wt% for 4A bead. The second is bulk density, target 720 to 760 kg/m3 for 4A 8x12 bead, because underdense lots indicate cracking or poor packing that hurts cycle life. The third is crush strength per ASTM D4179, target above 5 N/bead average and above 1.5 N/bead minimum on the bottom screen. The fourth is attrition per ASTM D4058, target below 0.2 wt% loss at 30 minutes. The fifth is particle size distribution per ASTM D2862, confirming the 8x12 or 4x8 mesh falls within 5% of the advertised cut.

The sixth is moisture content of the as-shipped drum by Karl Fischer, target below 1.5 wt%, because pre-loaded water means lower working capacity on first installation and longer commissioning time. The seventh is sodium content by ICP-OES on an acid-digested sample, target below 0.5 wt% Na2O, because high sodium can volatilize during high-temperature regeneration and poison downstream deoxo catalyst. The eighth and final check is that the Certificate of Analysis is lot-specific and matches the lot number printed on the drum label; generic CoAs are a common source of out-of-spec material reaching site. For buyers specifying multiple lots over a multi-year contract, a lot retention program (1 kg retained sample per drum, sealed in foil, stored for 12 months) allows post-event analysis if a downstream failure is attributed to the sieve.

Common operating problems and their root causes

Four problems appear repeatedly on operating plants. The first is high outlet dew point despite a fresh bed, which almost always traces to insufficient regeneration temperature or insufficient regeneration time. The second is a slowly rising outlet dew point over months, which usually indicates fouling by carryover oil from the compressor or by trace glycol from the upstream heat exchanger. The third is a sudden pressure drop increase across the bed, which points to fines accumulation and requires a bed top-up or replacement. The fourth is a high deoxo catalyst outlet temperature that walks up over time, which indicates catalyst poisoning by sodium or sulfur and requires the upstream sieve to be checked for sodium slip.

For each problem, the corrective action is mechanical: raise regeneration temperature by 10 to 20 degC, install or service the upstream knockout drum, replace the sieve, or change the sieve specification to a higher-sodium-tolerance grade. The least useful response is to lower the operating pressure or to extend the cycle time, both of which only mask the underlying problem. A well-designed green hydrogen dryer with correctly specified 4A molecular sieve, a properly sized activated alumina pre-bed, and a clean regeneration loop will deliver -70 degC pressure dew point for 3 to 5 years between sieve replacements.

Internal links and product references

For the molecular sieve and activated alumina products that fit the typical green hydrogen drying train, see:

• Type 4A / 5A / 13X Molecular Sieve (8x12 and 4x8 mesh bead, Na2O below 0.4 wt%, ISO 17607 water capacity above 21 wt%)
• Activated Alumina (5x7 and 7x14 mesh bead for pre-bed duty, BET above 320 m2/g, crush strength above 8 N/bead)
• Silica Gel (indicating grade for knock-out drum and pre-cooler protection, beads 3 to 5 mm)
• Molecular Sieve Powder (used in some designs for the post-deoxo polishing stage where smaller bed volumes are preferred)

Related blog articles

• Molecular Sieve 13X for Cryogenic Air Separation Pre-Purification (parallel engineering on CO2 + H2O removal before the cold box)
• How to Calculate Adsorbent Bed Pressure Drop: Ergun Equation, Real-World Examples, and Particle Size Trade-Offs (calculation method shared with hydrogen dryers)
• Molecular Sieve 13X for CO2 Capture from Flue Gas: Pressure Swing Adsorption vs Chemical Absorption TCO (PSA vs TSA trade-offs, directly relevant to TSA vs heatless choice)
• Dual-Bed Silica Gel + Molecular Sieve Design for Industrial Drying: When One Tower Is Not Enough (parallel trade-off for the activated alumina pre-bed + molecular sieve main bed)

Conclusion and next steps

A correctly specified molecular sieve drying train downstream of a PEM or alkaline electrolyzer is a low-risk subsystem when the design follows the standard pattern: knockout drum, activated alumina pre-bed, two-stage 4A molecular sieve, deoxo catalyst, and a post-deoxo polishing sieve. The key engineering decisions are the dew point target (-70 degC is the most common), the choice between heatless PSA and heated-purge TSA regeneration, and the management of liquid carryover during electrolyzer pressure swings. The seven QC checks listed above catch the majority of field failures.

For a green hydrogen plant in the planning or commissioning stage, Aluminaworld can supply the molecular sieve, activated alumina pre-bed, and silica gel indication material on a single PO, with lot-specific CoA, ISO 17607 water capacity data, and ASTM D4179 crush strength data attached. Lead time is 15 to 25 days from PO for standard 4A 8x12 mesh bead in 25 kg drums or 1000 kg supersacks.

Next step: For a sizing review of an existing or planned hydrogen drying train, send the gas flow rate, pressure, target dew point, and intended regeneration scheme to barry@aluminaworld.com or message us on WhatsApp. Free sample (1 kg 4A 8x12 bead) is shipped within 5 business days for any project in the design or pilot stage.