Ammonia Removal: Why Virgin Activated Carbon Barely Works

Virgin activated carbon holds roughly 0.25 to 2 mg of ammonia per gram. Acid-impregnated or surface-oxidised carbon holds 19 to 35 mg/g. That is a factor of about a hundred, measured across two NASA reports by overlapping authors — NTRS 20120004303 and NTRS 20150008167 — whose stated conditions differ in detail but which both ran at approximately 20 ppm ammonia, which is what makes the figures broadly comparable. And it is not a surface-area effect. If you are specifying carbon for ammonia duty on the strength of a BET number or a CTC percentage, those two figures will not tell you what the bed will do.

The short version

• NASA’s sorbent screening puts virgin Calgon BPL at 0.252–0.321 mg NH₃/g and virgin Norit DARCO at 1.959 mg/g, both at about 20 ppm inlet.
• The same programme measured 19.650 mg/g for Ammonasorb II, a phosphoric-acid-impregnated commercial carbon, and 34.6 mg/g for an oxidised PVDC-derived foam.
• A 2023 paper in Frontiers in Environmental Science states that physisorption of ammonia on unmodified carbon is so weak that capacity has “no obvious relationship with the surface area and pore structure”.
• The capacity is not recoverable. Ammonasorb II returned about 2.4 mg/g after one regeneration cycle, roughly 12 % of its initial loading, because the binding is chemisorption.
• Test conditions dominate the numbers. A published 40 mg/g figure was measured at 1,000 ppm NH₃; the NASA figures were at about 20 ppm. A 50× difference in inlet makes the two non-comparable.

Two NASA reports by overlapping authors, both at approximately 20 ppm ammonia but with differing stated conditions. Virgin carbons on the left, modified carbons on the right, and roughly a hundredfold between them.

Why virgin carbon fails on ammonia — a live technical argument

The orthodox explanation is molecular. Ammonia has a kinetic diameter of about 0.26 nm and is small, volatile and weakly polarisable, so the dispersion forces that hold a toluene or a chlorinated solvent onto a carbon surface are simply too weak to hold NH₃ at ambient temperature and trace concentration. It is not wrong so much as incomplete.

The NASA and Advanced Fuel Research authors contest it directly. In their 2015 paper for the International Conference on Environmental Systems, they note that it has been widely believed that unimpregnated carbon does not adsorb ammonia, and argue the real cause is textural rather than intrinsic: commercial activated carbons have broad pore-size distributions that are incompatible with efficient ammonia physisorption. Their data shows ammonia capacity correlating with pore volume specifically in the 0.9–2.2 nm range. On that reading, virgin carbon is not incapable of holding ammonia. It is simply that almost nobody makes a carbon whose porosity is concentrated where ammonia wants to sit — and when they did, the pore-engineered virgin carbon outperformed the commercial acid-impregnated benchmark.

Ammonia is about 0.26 nm across. The NASA argument is that only pores in the 0.9–2.2 nm band hold it usefully, and most commercial carbons are not built that way.

Both accounts converge on one conclusion, and it is the commercially useful one. From Frontiers in Environmental Science (2023), on unmodified carbon, the physisorption of ammonia is so weak that capacity has “no obvious relationship with the surface area and pore structure”.

Read that as a purchasing instruction. Surface area and CTC activity — the two numbers every carbon supplier leads with — do not predict ammonia performance. They predict physisorption capacity for large organic molecules, which is a different duty. For ammonia, the only number that means anything is a breakthrough capacity measured at a stated inlet concentration, humidity, temperature and endpoint. If a supplier cannot give you that, the datasheet does not answer the question you are asking.

The published data

Two NASA datasets and one peer-reviewed study converge on the same conclusion: at around 20 ppm NH₃ virgin activated carbons hold 0.25–1.96 mg/g, while acid-impregnated and surface-modified carbons hold 19.3–34.6 mg/g — an order of magnitude and more, on the same challenge.

Note that Ammonasorb II appears twice, at 19.650 and 23.7 mg/g, in two studies by overlapping authors on nominally the same product. That spread — about 20 % — is a useful calibration for how much weight any single published capacity figure deserves.

The comparison you must not make

A widely cited figure of 40 mg NH₃/g comes from the 2023 Frontiers study, for carbon modified with 10 M nitric acid at 90 °C. It is a real number, correctly reported. It was measured at 1,000 ppm NH₃ at room temperature and 200 mL/min.

The NASA figures above were measured at about 20 ppm.

That is a fiftyfold difference in inlet concentration. Adsorption capacity rises with partial pressure along the isotherm, so a capacity measured at 1,000 ppm will always exceed one measured at 20 ppm on the same material. Placing 40 mg/g in the same column as 23.7 mg/g and concluding that nitric acid beats phosphoric acid is not a conclusion the data supports. The same study, incidentally, reports the same carbon at only 15 mg/g when the operating temperature was raised to 100 °C — a reminder that temperature moves these numbers as hard as concentration does.

If you take one habit from this article, take this one: an ammonia capacity without an inlet concentration attached is not a specification. It is a marketing figure.

The Frontiers work is also worth citing for its mechanism, which is more interesting than simple acid–base neutralisation. The capacity gain came from carboxyl groups acting as primary sites, residual nitric acid co-adsorbing, and a circular hydrogen-bond network binding NO₃⁻, NH₄⁺ and carbonyl groups together at −1.63 eV. And unusually for an impregnation, BET rose — 525.20 to 581.27 m²/g — because acid etching opened porosity faster than the acid filled it.

Regenerability: the chemistry that gives you capacity takes away recovery

Chemisorption is what makes acid impregnation work, and it is also what makes the bed a consumable.

The NASA data is unambiguous. Ammonasorb II delivered 19.650 mg NH₃/g on first use and recovered only about 2.4 mg/g after a single regeneration cycle — roughly 12 %. The authors attribute it to severe irreversible sorption from strong chemical bonding to the phosphoric acid. Ammonia plus phosphoric acid gives ammonium phosphate, a non-volatile salt, and no vacuum or mild thermal cycle unmakes it.

The surprising part is what the same programme found on the other side of the comparison. The pore-engineered virgin PVDC foam reached 34.6 mg/g — higher than the acid-impregnated commercial benchmark — and retained roughly 47 % of capacity after vacuum regeneration, against Ammonasorb II’s 29 % in the same comparison. The virgin carbon beat the commercial impregnated product on both capacity and regenerability.

Two retention figures for Ammonasorb II are therefore in the public record, and they do not agree. The roughly 12 % above is 2.4 mg/g measured against the 19.650 mg/g first-use capacity reported at 20 ppm in NTRS 20120004303; the 29 % here is retention after vacuum regeneration against the 23.7 mg/g baseline measured in the ICES-2015 programme, NTRS 20150008167. They differ because they are different reports run under different conditions on nominally the same product — the same reason the two first-use capacities differ by about 20 % — and we publish both rather than choosing the one that reads better. That result should temper any assumption that impregnation is the only route to an ammonia sorbent; it also explains why the physisorption-is-hopeless framing is contested.

For anyone specifying a fixed bed, the practical consequence is simple. An acid-impregnated ammonia bed is sized on a single service life and changed out, not regenerated. Treat the impregnation loading as a stoichiometric budget: once the acid is neutralised, the bed is done.

Where ammonia actually matters

Wastewater and odour control. The Suez/Degrémont water handbook specifies impregnated activated carbon for hydrogen sulfide and ammonia removal in odour-control ventilation, at a passage speed of 500–2,000 m/h through a bed of about 1 m, at operating temperatures of 5–40 °C. Ammonia rarely arrives alone in this duty — it comes with H₂S and amines out of headworks, thickeners and dewatering — which is why the media selection is usually a multi-bed or blended-media question rather than a single-grade one.

Semiconductor cleanrooms and airborne molecular contamination. This is the demanding end. SEMI Standard F21-95 sorts airborne molecular contaminants into four classes — acids, bases, condensables and dopants — with ammonia and amines falling under bases; ISO 14644-8 is the other framework in use.

The defect mechanism is specific and worth understanding, because it explains why the limits are so low. In chemically amplified deep-UV resist, exposure generates acid photolytically, and that acid drives the subsequent development chemistry. Airborne ammonia from the cleanroom neutralises it at the resist surface, producing the classic T-topping profile. As Horiba’s process monitoring literature puts it, the interaction is device-defect relevant because it “deteriorates line width and line structures”. Nanometres of line width lost to parts per billion of a gas.

On the numbers, we will be precise about what we can and cannot verify. The 1997 National Technology Roadmap for Semiconductors set airborne amine contamination levels for the following fifteen years at about one part per billion. That figure is documented. We could not verify any specific modern named ITRS or SEMI numeric ammonia limit, so this article does not state one. What is documented is the direction of travel: industry commentary describes AMC requirements shifting from parts per million to parts per billion and, for some species, to parts per trillion. If you are quoting a fab, get the limit from the fab.

Suracsh grades for ammonia

Two grades are published for this duty, both coconut-shell based.

Now the honest part, and it follows directly from this article’s own argument. Suracsh does not publish an ammonia capacity figure for either grade. And the figures that are published — CTC activity, iodine number, apparent density — are exactly the numbers that this article has just spent two thousand words explaining do not predict ammonia performance. CTC is a mesopore and large-micropore probe measured by mass gain from a saturated carbon tetrachloride stream; iodine number is a relative porosity index from aqueous solution. Neither is an ammonia measurement. The 8–12 % impregnation figure on GAM is the most useful number on the sheet, because it bounds the stoichiometry.

Also relevant is SURSORB K, the EN 14387 K-class grade — the class covering ammonia and organic ammonia derivatives — which shares the range’s moisture of 10 ± 5 % and ball pan hardness min 98. The impregnant is not disclosed on the datasheet, and we could not verify from any accessible source what impregnant commercial Type K filter carbons actually use, so this article makes no claim about it. What EN 14387 does define is a breakthrough test rather than a material composition, and that is covered in what EN 14387 actually tests, alongside the published SURSORB ABEK breakthrough table, where ammonia at 1,000 ppm inlet with a 25 ppm outlet limit is one of the three tested gases.

What to ask for

If you are buying carbon for ammonia duty, the specification questions are narrow:

1. What is the breakthrough capacity, in mg NH₃/g or g/cc, at your inlet concentration — not at 1,000 ppm?
2. At what relative humidity and temperature, and to what outlet endpoint?
3. What is the impregnation loading, and what acid?
4. Is the bed sized as a consumable? On chemisorption, assume yes.

Everything else on the datasheet describes a different duty.

The full specification for every grade named here is published in HTML at suracsh.us, and every datasheet downloads without a form. Where a capacity figure is not published, the sheet says nothing rather than something approximate — and this article says the same.

Sources

• Wójtowicz et al., ICES-2015-179, NASA NTRS 20150008167 — ammonia sorbents, capacities and regenerability
• NASA NTRS 20120004303 — ammonia capacity of Calgon BPL, Norit DARCO and Ammonasorb II at 20 ppm
• Li et al., Frontiers in Environmental Science (2023), doi:10.3389/fenvs.2023.976113 — acid-modified carbon for NH₃
• General Carbon, “A Guide to Impregnated Activated Carbon” — phosphoric acid for ammonia
• Suez Water Handbook — odour control ventilation design
• Semiconductor Digest, “Airborne molecular contamination in cleanrooms” — SEMI F21-95 classes and the 1997 NTRS amine target
• Horiba — monitoring ammonia AMC in semiconductor fabrication, and the T-topping defect mechanism
• Semiconductor Digest, “Can filtration keep up with AMC control limits?”
• Desotec — physisorption is reversible, chemisorption often permanent
• Phosphoric-acid-impregnated activated carbon for airborne molecular contaminants in cleanroom make-up air units
• Thommes et al., Physisorption of gases (IUPAC Technical Report), Pure Appl. Chem. 87(9–10):1051 (2015)
• ASTM D3467 — Carbon Tetrachloride Activity of Activated Carbon
• ASTM D4607 — Iodine Number of Activated Carbon

Questions