GAC beds are usually sized on an empty bed contact time rule of thumb. The adsorption thermodynamics are well founded and worth understanding — but the single most important design lesson is that isotherm capacity, applied naively, overpredicts bed life by orders of magnitude.
Polanyi potential theory and micropore filling
Adsorption onto activated carbon is not surface coverage in the Langmuir sense; it is volumetric filling of micropores. The Polanyi adsorption potential is the work required to bring a molecule from bulk solution to the adsorbed state:
A = RT ln(Cs / Ce)Cs = aqueous solubility of the adsorbate; Ce = equilibrium concentration
The adsorbed volume then follows the Dubinin–Astakhov relation:
W = W0 · exp[ −(A / E)n ]W0 = micropore volume (cm³/g); E = characteristic energy (J/mol); n = heterogeneity parameter, typically 2–3
The practical value is that a single characteristic curve describes an adsorbent–adsorbate pair across temperatures and concentrations. Measure isotherms at two temperatures, convert loading to adsorbed volume by dividing by the adsorbate liquid density, and plot W against A. If the points collapse onto one curve, the theory holds and further isotherm work is unnecessary.
| System | W0 (cm³/g) | E (kJ/mol) | n |
|---|---|---|---|
| Bituminous GAC, general | ≈0.45 | ≈23 | 2–3 |
| Carbamazepine on bituminous GAC | ≈0.42 | ≈19.5 | ≈2.1 |
Breakthrough models: what they can and cannot tell you
Two models dominate column design. The Thomas model describes the breakthrough curve of a fixed bed:
Ct/C0 = 1 / { 1 + exp[ kThq0m/Q − kThC0t ] }kTh = Thomas rate constant; q0 = solid-phase capacity; m = carbon mass; Q = flow; t = time
The bed depth service time model linearises the same behaviour and is the better tool for scaling a pilot column to full size:
t = (N0 / C0v) · Z − (1 / kaC0) · ln(C0/CB − 1)N0 = volumetric bed capacity; v = linear velocity; Z = bed depth; CB = breakthrough concentration
BDST is genuinely useful because the slope is obtained from two pilot columns of different depth and transfers directly to full scale. Its weakness is the same as Thomas’s: both are single-solute models.
Why single-solute capacity is the wrong design basis
This is the part that catches people out, and it is worth working through explicitly.
- Bed volume: 500 m³/h × 0.25 h = 125 m³.
- Carbon mass at 500 kg/m³: 62,500 kg.
- Mass loading: 12,000 m³/d × 2.5 mg/m³ = 30 g/d.
- Single-solute capacity at an isotherm value of 120 mg/g: 62,500 kg × 120 mg/g = 7,500 kg of carbamazepine.
- Implied bed life: 7,500 kg / 30 g per day ≈ 685 years.
That answer is obviously wrong, and it is wrong for a physical reason, not an arithmetic one: the bed will never see 120 mg/g of carbamazepine because natural organic matter occupies the pore volume first.
Two effects destroy the single-solute prediction:
- NOM preloading. Background organic matter arrives at milligram-per-litre concentrations against the micropollutant’s nanograms to micrograms. It occupies and blocks pores long before the target compound approaches saturation.
- Competitive adsorption. Where NOM and micropollutant compete for the same sites, working capacity typically falls 30–60 per cent against clean-water isotherms.
Sizing on bed volumes
Practitioners size GAC in bed volumes treated to breakthrough because that is the quantity pilot work actually measures and the quantity that transfers between sites.
BV per day = 1440 / EBCT (minutes)Bed volumes of water treated per day at a given empty bed contact time
- Throughput: 1440 / 15 = 96 bed volumes per day.
- Pilot result for pharmaceutical breakthrough in a real secondary effluent: typically 10,000–30,000 BV.
- Bed life at 20,000 BV: 20,000 / 96 ≈ 208 days — about seven months.
That is consistent with operating experience, where GAC for micropollutants is changed or regenerated on a 6–18 month cycle. It is also three orders of magnitude away from the single-solute answer, which is the measure of how badly clean-water capacity misleads.
| EBCT | BV/day | Bed life at 20,000 BV | Suitability |
|---|---|---|---|
| 5 min | 288 | ≈70 days | Insufficient for most micropollutants |
| 10 min | 144 | ≈139 days | Minimum for pharmaceutical removal |
| 15 min | 96 | ≈208 days | Typical design point |
| 20 min | 72 | ≈278 days | Where consent is tight or NOM high |
The wider adsorption fundamentals are in activated carbon adsorption; the comparison with resin for PFAS duty in GAC vs ion exchange.
Regeneration economics
Thermal reactivation at around 850 °C under steam recovers better than 95 per cent of the original adsorption capacity, with a carbon loss of roughly 5–10 per cent per cycle. Above a virgin carbon price of about £150 per tonne, reactivation is the cheaper route, and it is substantially better on embodied carbon — relevant where the works has a net zero commitment. See net zero in water treatment.
Two practical constraints: reactivated carbon has a slightly different pore structure, so bed life shifts a little after the first cycle; and carbon that has adsorbed PFAS requires a reactivation furnace operating hot enough to destroy them, which not all facilities offer.
Design rules
- Size on bed volumes from pilot data, not on isotherm capacity. Clean-water capacity overpredicts bed life by orders of magnitude.
- Pilot with the actual water. NOM competition costs 30–60 per cent of capacity and is site-specific; no literature isotherm substitutes for it.
- Specify 10–20 minutes EBCT for pharmaceutical removal. Below 10 minutes the mass transfer zone does not fit in the bed.
- Use Polanyi–Dubinin to compare carbons and to extrapolate across temperature from limited isotherm work — that is what it is good for.
- Use BDST for scale-up from two pilot columns of differing depth; the slope transfers, the intercept carries the mass transfer information.
- Plan reactivation from the outset above about £150/tonne virgin carbon, and confirm the furnace suits the contaminants adsorbed.
Frequently asked questions
Why does isotherm capacity overpredict GAC bed life so badly?
Because isotherms are measured in clean water on a single compound. In a real effluent, natural organic matter arrives at milligram-per-litre concentrations against the micropollutant’s micrograms, and occupies the pore volume first. Working capacity falls 30–60 per cent, and the bed is exhausted for the target compound long before its own isotherm capacity is approached — the single-solute calculation can be out by three orders of magnitude.
What EBCT should I specify for pharmaceutical removal?
10–20 minutes, with 15 as a common design point. Below 10 minutes the mass transfer zone is longer than the bed, so breakthrough begins almost immediately and carbon is wasted. Above 20 minutes the additional bed life rarely justifies the vessel.
Is the Thomas model useless then?
No — it is the wrong tool for absolute prediction but the right one for interpretation. Fit it to pilot breakthrough data to extract effective capacity and rate constants under real competition, then use those fitted values for scale-up. The error is feeding it clean-water constants and believing the answer.
When is thermal reactivation worth it over virgin replacement?
Generally above a virgin carbon price of about £150 per tonne. Reactivation recovers better than 95 per cent of capacity at a 5–10 per cent mass loss per cycle, and carries a much lower embodied carbon burden. Check that the furnace runs hot enough for the contaminants adsorbed — PFAS in particular.