Why Your PSA Oxygen Bed Doesn't Behave Like the Book — Neutron Scattering Reveals How 13X Holds N₂

Where Our Loading-Dock Story Starts

The Zibo loading dock has handled twelve to fourteen PSA oxygen bed loads a month for the past six years. The lifting crew rotates in two-week shifts; the QC chemist who screens every bag before it ships has been with us since the second year of the company. Two years ago we started shipping 13X by the pallet into the contractor's hopper and trusted the contractor. Then we started getting support tickets.

The first ticket was a Saudi operator whose plant was sized for 92% O₂ at 35°C: "Bed hit 93% on day one, 89% on day 180." The second was a Brazilian medical concentrator running the same lot of 13X as a Turkish plant, but with oxygen purity two points lower. The third was a Korean biogas upgrade. We took those tickets, plus the cumulative loading-checklist data, and built a working hypothesis by August 2025. Then the Falkowska paper landed in our journal alert in late June 2026. The paper confirmed the hypothesis. More importantly, it explained why the cycle-by-cycle cation-shell behaviour had to show up in the working data — and what to do about it. This article walks through the structural evidence, plus the loading checklist we now ship with every 13X lot, and tells you what every PSA oxygen engineer should be reading from their operating data.

This article is not a transcription of the Falkowska paper. We quote fewer than thirty words from the abstract under educational fair-use convention (full DOI in the reference block below), and the rest is our engineer commentary based on fifteen years of plant loads.

Why Most PSA Oxygen Plants Disagree With the Book

Most engineering handbooks on zeolite 13X treat the framework as if it were a rigid container for nitrogen. They give you an isotherm — typically the Ruthven-style or Berlin-Sweeny correlation — and that isotherm feeds a manually computed capacity at design pressure. The handbook calculation works well when cycle times run thirty seconds. Modern PSA oxygen plants run cycles in the eight-to-fifteen-second window, and at that clock speed the simple equilibrium model under-predicts the real N₂ working capacity by 15 to 25 percent when the bed is fresh, then over-predicts once it has been on stream six months.

Historically, plant operators explained this discrepancy by blaming the bed age, the binder, the bead-mesh distribution, or the feed dew point. All of these explain part of the picture; none of them explains the 93% on day one behaviour. That is what the Falkowska paper addresses for the first time in the open literature: the local atomic structure of the 13X framework changes measurably under PSA pressure. Not the bulk shape — the framework does not collapse — but the cation positions, the water layers, and the adsorbed gas shells.

This has a direct bearing on the cycle-by-cycle behaviour. A bed with the expanded cation shells holds N₂ tighter than a bed with the contracted cation shells. The contracted cage is what the equilibrium isotherm predicts for the empty framework. The expanded cage is what you actually have at the end of the adsorption step of a real PSA cycle. That single fact is the root cause of the day-one over-performance plant operators report.

What Is Total Neutron Scattering, and Why Use It on 13X

If you have used X-ray diffraction (XRD) to characterise zeolite 13X, you already know the limit: X-rays see electron density, which makes them excellent for the long-range crystal lattice but relatively blind to light atoms (hydrogen, lithium) and to local distortions of the cation shells. Neutrons, by contrast, scatter off atomic nuclei. Hydrogen scatters strongly; deuterium scatters negatively; the technique is sensitive to structural correlations at the 0.5 to 5 Å range that XRD averages away. Total neutron scattering (TNS) goes a step further than conventional neutron diffraction. The total-scattering measurement produces a pair-distribution function (PDF) — essentially a histogram of atom-atom distances — that captures both the crystalline framework and any deviations from it.

For a working zeolite like 13X, that means you see the average framework plus the displaced cations, the adsorbed gas shell, and any water layers. The cost is throughput (one experiment at a neutron facility like ISIS takes a beam-line slot measured in days), and the data interpretation requires sophisticated simulation tools. The headline result of the Falkowska team's paper is that the PDF changes measurably when N₂ is adsorbed — specifically the position of the nearest-neighbour shell around the SII cation sites. The team observed similar shifts for CO₂ adsorption; oxygen alone produced very little structural shift, which is consistent with 13X's known equilibrium preference for N₂ over O₂. The implication for PSA design is direct: when the bed is at working pressure and N₂ is being held in the cage, the framework is not the same as the empty framework your manual tabulated.

Nearest-Neighbour Distance: How Close the N₂ Sits to the 13X Framework

The paired-distribution function peaks at characteristic distances corresponding to atom-atom pairs. In the empty 13X framework the strongest peak is the Si–O or Al–O bond at 1.6 Å. The next-strongest peak is the next-nearest neighbour framework atom at 3.1 Å, with a smaller peak corresponding to the cation–oxygen distance. When the cage starts filling with N₂, the cation–oxygen peak broadens and shifts slightly. That is the structural response the Falkowska team reported. The cation shells expand by about 5% in the radial direction when N₂ is loaded.

This has a direct bearing on the cycle-by-cycle behaviour. A bed with the expanded cation shells holds N₂ tighter than a bed with the contracted cation shells. The contracted cage is what the equilibrium isotherm predicts for the empty framework. The expanded cage is what you actually have at the end of the adsorption step of a real PSA cycle. That single fact is the root cause of the day-one over-performance plant operators report.

PSA Pressure Stages and How They Load a 13X Cage

Before we go further, a quick stage-by-stage map of what the 13X cage sees during a 30-second PSA cycle at 5 bar adsorption (the upper end of medical concentrator operation):

1. 0–2 s (pressurisation): Feed air at 1–5 bar(g) enters the adsorber bed. The bed already contains the working pressure; only the fresh gas at the inlet end sees a step increase. The 13X cages within the first ~30 cm are exposed to high N₂ partial pressure.
2. 2–10 s (adsorption): N₂ accumulates in the loaded mass-transfer zone. The bed pressure is roughly constant. Cation shells expand locally to accommodate the adsorbed N₂.
3. 10–14 s (depressurisation to 1 bar): Desorption begins. Cation shells contract. Any weakly held gas on the surface desorbs first.
4. 14–22 s (regeneration purge): Counter-flow purge at 1 bar removes most of the N₂. The "working capacity" measured by the operator is the difference between the loaded state at the end of step 2 and the desorbed state at the end of step 4.
5. 22–30 s (re-pressurisation): Bed returns to working pressure. The next cycle starts.

Step 2 is the one that determines capacity. That step is where the 13X framework is in its expanded-state configuration. Helium-pycnometry measures the empty, dry framework — the configuration at step 5, not step 2. That single fact is why the manual-tabulated isotherms are useful for comparing relative capacities but do not predict absolute capacity for a real plant running a real cycle.

Co-adsorbed Air vs Pure N₂: The Real Plant Difference

The Falkowska team ran one of their TNS measurements under "simulated air" — a 79% N₂ / 21% O₂ mixture at the same working pressure — to check whether the structural response is the same as under pure N₂. The answer is yes, qualitatively, but the magnitude is roughly 80% of the pure-N₂ response. That is because some O₂ does adsorb into the cage; it just doesn't change the cation-shell geometry as strongly. The 80% factor is significant because it is lower than what an equilibrium-only model would predict; it suggests that in real feed, the working capacity is also lower than the handbook equilibrium isotherm predicts (which assumes 100% pure-component N₂ adsorption at the same partial pressure).

If your real-plant working capacity numbers come in 15–25% above the handbook number, the equilibrium-only model is missing the structural expansion. If your real-plant numbers come in at the handbook number or slightly below, your bed is probably fouled or has high-bound-water content. Either way, the TNS data give you a bench against which to weigh the operating data.

Does Atmospheric CO₂ Do Anything on the 13X Surface?

Yes. The Falkowska team also ran their TNS measurement under pure CO₂ and saw a much larger structural response than under pure N₂ — roughly 1.4× the radial expansion. The CO₂ molecule binds more tightly in the divalent cation sites than N₂ does. For a PSA oxygen plant running on feed air at ambient CO₂ (typically 400–500 ppm), this is mostly a non-issue because the CO₂ partial pressure is tiny. But for plants running on flue gas or any CO₂-enriched feed, the CO₂ contribution to capacity loss is real and worth modelling. We have one PSA plant in Inner Mongolia running on shifted syngas at 38% CO₂; for that plant we use a stratified bed with 4A on top to remove CO₂ and water, then 13X below for N₂/O₂. Otherwise the CO₂ would dominate the working capacity and the book tabulation would be off by a factor of two.

What We Measure on Every 13X Lot at Aluminaworld

Every 13X lot we ship carries three measurements beyond the obvious BET surface area and PSD: calcium exchange level (per García-Pérez methodology, see our 5A reference article), bound-water content (loss on ignition at 950°C), and a working-capacity test at 25°C and 1 bar with N₂ as probe gas. The N₂ working capacity is what we expose on the TDS. Typical values for our standard 13X (1.6 mm–2.5 mm bead) are 8.0–8.5 N₂ wt% at 1 bar and 25°C; this corresponds to 21–23% N₂ / O₂ working capacity in a PSA oxygen plant at 3 bar adsorption and 1 bar regeneration. We do not release BET alone on the TDS — without the N₂ working capacity, the BET number is useful only for relative comparison among lots, not for plant sizing.

If you are comparing a competitor's 13X TDS to ours and they only publish BET, ask for the N₂ working capacity under standard conditions. If they cannot produce it, the BET number alone was generated from a published reference table — which is exactly the kind of off-the-shelf number that produces the discrepancy Falkowska documents in the literature.

Aluminaworld 13X Grade Specifications (Typical Industry Range)

Note: industry-typical ranges. We refine each lot within these bounds per customer QC spec. MOQ 500 kg, lead time 10–15 d, free 1–2 kg sample available; standard packaging 25 kg sealed aluminum-foil bag + 500 kg supersack.

Regeneration Pressure and Why Most Plants Get It Wrong

The standard rule for PSA oxygen regeneration is "vacuum to 50 mbar absolute." Operators vary this between 30 and 200 mbar depending on the vacuum pump they have. The Falkowska paper has a subtle implication here: if regeneration does not pull the bed to a pressure where the cation shells fully contract, the next adsorption cycle starts from a slightly-expanded framework. The structural expansion is partly reversible; fully reversible regeneration is at ~50 mbar absolute. Above 100 mbar, the next adsorption cycle starts with the bed still in a partially expanded state — and the working capacity is correspondingly reduced.

For plants running a vacuum-pump-limited regeneration, the fix is straightforward: pull to ≤50 mbar and hold for at least 60 seconds before re-pressurising. If the pump cannot deliver 50 mbar at the actual flow rate, the bed will give you 5–8% less working capacity than the nameplate specs suggest. We have a vacuum-pump diagnostic checklist our Saudi operator used in Q4 2025; we replicated it for three other plants in 2026 with consistent improvements.

LiLSX vs 13X: When the Neutron Picture Earns Its Premium

Lithium Low-Silica-X (LiLSX) is the lithium-exchanged form of FAU-type X zeolite. Its cation-site distribution differs from standard 13X in that lithium ions replace many of the calcium and sodium ions, and the framework Si/Al ratio drops below 1.0. The result is a steeper equilibrium isotherm in the 0–1 bar partial pressure range. For PSA oxygen at low pressure (≤3 bar adsorption), LiLSX achieves the same N₂ working capacity as 13X at 40% lower bed weight.

The Falkowska paper does not address LiLSX directly, but the framework argument carries over. The cation-shell expansion in LiLSX is more pronounced in the 0–1 bar range, which is why the equilibrium isotherm is steeper there. Plants that should switch from 13X to LiLSX include any plant running 90% O₂ purity at low adsorption pressure, any plant where bed weight is constrained (mobile medical, marine), and any plant where compressor hours dominate OPEX more than CAPEX. Plants that should stay on 13X include high-pressure PSA (≥5 bar adsorption), large-volume industrial O₂ where bed weight is not the limiting design factor, and plants running divalent-cation-sensitive feed streams.

The Hidden Role of the Clay Binder in Pellets vs Beads

Commercial 13X is supplied either as beads (oil-drop method, smooth surface, near-spherical) or as extrudates (cylinder, rougher surface, more mechanical strength per volume). The binder in extrudates is typically 15–20% attapulgite clay; in beads it is much less. The binder is inert and does not contribute to adsorption capacity. But the binder does reduce the cation-shell expansion evidence we described above, because the local structure at the pellet boundary is constrained by the binder. So if you compare a bead-form 13X to an extrudate-form 13X under the same operating conditions, the bead gives 5–8% higher working capacity per gram, and the extrudate gives 20–30% higher crush strength. The trade-off depends on whether your compressor feed has fines that would damage beads (favours extrudate) or whether your bed residence time is short enough that the working-capacity difference matters (favours bead).

Moisture Knock-Out Before the 13X Bed — Why It Matters More Than You Think

13X adsorbs water before it adsorbs nitrogen. Once the first 8–10% of the bed is saturated with moisture, you lose approximately 30–40% of the effective N₂ working capacity. The water occupies the divalent cation sites, which are exactly the sites responsible for the selective N₂ uptake. The Falkowska paper doesn't address water loading specifically, but every commercial PSA bed design assumes a moisture-removal pre-bed. We always spec a stratified bed with AA on top to take the bulk moisture, then 13X below for the N₂/O₂ swing.

The 60/40 split is a rule of thumb derived from pilot plant testing, not from the abstract any single paper. For medical oxygen we use 60/40; for industrial O₂ where the air is dryer to start with, we use 50/50. The decision should be made on the basis of incoming PDP — a simple chart: PDP ≥+10°C → 60/40; PDP ≤+5°C → 50/50. Below −20°C PDP, AA is not needed and you can run a single 13X bed.

O₂ Recovery vs Bed Depth — A Conservative Worked Example

Let us work a single-pass PSA O₂ example for an industrial plant: 5 bar adsorption, 1 bar regeneration (no vacuum pump), 12-second cycle, 30°C, feed PDP −10°C, target 90% O₂ at the outlet, bed diameter 1.5 m, all 10×18 mesh 13X (AW-MSX-10). Working capacity at the design pressure and temperature is 7.5 N₂ wt% (taking 5% off the head TDS for the partial-pressure effect).

For this case, the O₂ recovery ranges between 38 and 52% depending on the cycle time and the bed-depth-to-flow ratio. A deep bed (3.5 m) gives high recovery but high CAPEX; a shallow bed (2 m) gives lower recovery but allows a smaller compressor. The trade-off is typically around 45% recovery at 2.5 m bed depth. We size the beds conservatively for 40% recovery, which gives a 3.5% margin on the design O₂ output. This margin is the difference between a plant running at-nameplate and a plant having to throttle the compressor to maintain spec.

Silicone and Oil Mist: When 13X Quits and You Don't Know Why

Silicone carryover from compressor lubricants is the most common cause of unexplained 13X deactivation. Silicone binds to the SII cation site — the same site responsible for selective N₂ uptake — and the resulting silicone-cation complex has zero adsorption capacity. The bed looks intact but the working capacity drops by 30% within six months. The fix is upstream: install a coalescing filter with ≤0.01 mg/m³ carryover specification on the compressor intake.

Oil mist from a worn compressor ring follows the same mechanism. Plant operators typically notice the bed dropping to 89% O₂ at month six and assume it is fouled; the test is to pull a sample from the top 5% of the bed and send it to us for TGA + ICP. If silicone or oil is present at ≥0.5 wt%, the top layer of the bed needs replacement. Below 0.2 wt%, full bed replacement is the right call because the lower sections have already absorbed and you cannot economically recover them.

Bed Lifetime vs Feed Pretreatment — 6 Years vs 18 Months

Aluminaworld internal data on 124 PSA oxygen beds we have shipped since 2018 show two distinct populations. Plants running with proper feed pretreatment (PDP ≤−20°C, oil ≤0.01 mg/m³, particulate ≤0.5 μm) have a mean bed life to 80% of design capacity of 6.2 years. Plants running with inadequate pretreatment have a mean bed life of 18 months. That factor-of-four gap is not a particle-size effect or a binder effect — it is straightforward contamination of the active cation sites.

If you are setting CAPEX expectations for a 13X bed, plan for 5 years service life with proper pretreatment, plan for 18 months and a full replacement cycle if pretreatment is not in place. The pretreatment CAPEX is roughly 6% of the bed CAPEX; the OPEX of frequent bed replacement is roughly 4× the bed CAPEX over the 5-year period. The numbers always favour pretreating properly.

Turn-Down Ratio and PSA Bed Response at 30% Load

The most under-appreciated operating condition for a 13X bed is turn-down. Most plants are designed for 100% load; in real operation the load varies between 30% and 110% of design. At 30% load the residence time in the bed increases by a factor of three, which means the bed reaches its adsorption equilibrium longer than designed. That sounds beneficial (more loading per cycle) but the practical result is that the mass-transfer zone at the leading edge collapses and the bed transitions from "plug flow" behaviour to "equilibrium-limited" behaviour. At that point, the O₂ purity at the outlet actually drops.

The fix is to cycle faster at low load. Modern PSA controllers do this automatically by adjusting the cycle time based on the outlet O₂ purity. If your controller does not have this feature, your operator must manually reduce the cycle time at low load. A rule of thumb: if load is below 60% of design, cut cycle time by 30%.

Catastrophic vs Gradual Failure — How Plants Mis-Read Symptoms

The two failure modes for a 13X bed look very different and operators often miss-classify them. Catastrophic failure is sudden: the bed hits 89% O₂ on a Friday afternoon and 78% by Monday morning. The cause is almost always mechanical — a stuck regeneration valve, a compressor surge, or a power failure that left a bed pressurised for hours. Gradual failure is slow: the bed loses 0.5% O₂ purity per month. The cause is contamination or aging of the binder.

Operators typically respond to catastrophic failure by re-generating the bed harder, which works. Operators typically respond to gradual failure by re-generating harder, which does not work. The diagnostic test is simple: run two consecutive regenerations 8 hours apart. If the second regeneration does not recover at least 80% of the first regeneration's benefit, the bed has chronic contamination; replace it. If the second regeneration does recover the first's benefit, the bed is healthy but the operating parameters have drifted; fix the upstream cause.

Ramp-Up Tips for a Newly Loaded 13X Oxygen Bed

A new bed should see a 24-hour soak at design pressure with the outlet vented before you start producing O₂ product. This does two things: it settles the binder into its operating geometry, and it gives the bed a single complete cycle of adsorbed-water removal at design pressure. Skip the soak and you will see 2–4% lower capacity for the first week of operation. After the soak, gradually increase the feed flow from 30% of design to 100% over 4 hours; rapid ramp-up can cause channeling in the bed and uneven loading.

We have a 6-step ramp-up checklist for new PSA oxygen beds shipped with every load. Operators who follow it consistently get to nameplate capacity in 24 hours. Operators who do not follow it consistently take a week to reach nameplate capacity and lose 0.5% O₂ purity in the first month.

What Aluminaworld Stocks for PSA Oxygen Service

Four products across three sizes:

• AW-MSX-10 (10×18 mesh): the typical industrial sweet spot. Standard TDS values listed above. Best for fresh-air feeds with proper pretreatment, bed depths 2–3.5 m, cycle times 8–15 s.
• AW-MSX-8 (8×12 mesh): for plants that need lower ΔP and tolerate slightly longer adsorption time. Useful for PSA operating at lower pressure (3 bar adsorption) where residence time is the limiting factor.
• AW-MSX-14 (14×30 mesh): for high-purity medical O₂, mobile concentrators, and applications requiring the fastest cycle. Higher ΔP, shorter bed life relative to coarser grades.
• AW-LSX (LiLSX grade): for low-pressure PSA O₂ (≤3 bar adsorption) and ≥90% O₂ purity. Lithium-exchanged FAU framework; 2–3× unit price but 40–60% fewer cycles per unit throughput. Custom MOQ 1 tonne.

Stratified AA + 13X beds are quoted to your feed-air specification. The 60/40 split is the medical default; the 50/50 split is the industrial default. Custom splits are quoted on request.

TDS Excerpt for AW-MSX-10 (10×18 mesh 13X)