Disinfection By-Products: Formation Chemistry, UK Limits and the Control Hierarchy

What are disinfection by-products and which ones are regulated?

Free chlorine is a strong electrophile. In contact with natural organic matter (NOM) — humic and fulvic acids, algal exudates, amino acids — it substitutes halogen onto activated aromatic and aliphatic sites, ultimately cleaving the structure to release small halogenated molecules. Bromide, if present, is oxidised to hypobromous acid and is incorporated preferentially.

The regulated species are a small and somewhat arbitrary subset. Of the total organic halogen formed, THMs and HAAs together typically account for perhaps a quarter to a half; the remainder is unidentified. Compliance is therefore a proxy for a broader exposure, which is why the sector treats precursor reduction as the substantive objective and THM concentration as the reported indicator.

The formation kinetics: what actually drives concentration

THM formation is not a single reaction but a family of parallel and consecutive steps, so empirical power-law models fitted to bench and plant data remain the practical tool. They take the general form:

The exponents are the message. Formation is close to first order in organic carbon and only weakly dependent on chlorine dose. Reducing chlorine by 30% buys perhaps 5–8% less THM while directly eroding the microbial barrier; reducing TOC by 30% buys roughly 30% less THM with no microbiological penalty at all. The strongest single term is pH, and it is exceptionally cheap to adjust.

Sensitivity worked through. For a water at TOC 4.0 mg/L, pH 7.6, forming 78 µg/L TTHM at the far end of the network:

• Enhanced coagulation cutting TOC to 2.4 mg/L (40% removal): factor (0.6)^1.10 = 0.57, giving ≈45 µg/L.
• Reducing final pH from 7.6 to 7.0: factor (7.0/7.6)^1.6 = 0.88, a further 12% reduction — but check the consequences for plumbosolvency and for the calcium carbonate saturation state before doing it.
• Both together: 78 × 0.57 × 0.88 ≈ 39 µg/L, half the original.

Formation also continues in the distribution network for as long as a residual persists, with the classic asymptotic curve: rapid formation in the first hours, then a slow approach to the ultimate formation potential over several days. Compliance samples taken at the works are therefore not the binding case; the far end of the network is.

Precursor character: why SUVA is the first measurement

Not all organic carbon is equally reactive. Aromatic, humic material forms THMs readily; hydrophilic, low-molecular-weight material does not. Specific UV absorbance separates the two and, uniquely among cheap tests, predicts how the water will respond to coagulation.

Measuring SUVA before designing a DBP control scheme prevents the most common and most expensive mistake in this field: specifying enhanced coagulation for a low-SUVA water and being surprised when a substantial coagulant dose increase removes very little carbon while generating a great deal of additional sludge.

Bromide: the term you cannot treat

Bromide is oxidised by chlorine to hypobromous acid, which is a more effective halogenating agent than hypochlorous acid and is incorporated preferentially into NOM. Two consequences follow, and both are unfavourable.

• Mass shifts upward. Substituting bromine (80 g/mol) for chlorine (35.5 g/mol) raises the mass concentration of THMs for the same molar formation. A water that would form 60 µg/L of chloroform can exceed 100 µg/L as brominated species at identical molar yield.
• Toxicity shifts upward. Brominated and iodinated species are generally more cytotoxic and genotoxic than their chlorinated analogues, so the health-relevant burden rises faster than the regulated mass.

Bromide is conservative through conventional treatment: coagulation, filtration and GAC do not remove it, and only nanofiltration, reverse osmosis or ion exchange will. Where bromide is elevated by saline intrusion, road salt or an industrial discharge, the practical response is to reduce the organic precursor further, since bromide cannot be reduced. Ozone must also be handled with care in bromide-containing water because it generates bromate directly — see our ozonation guide.

The control hierarchy

Control options fall into four tiers, and they should be considered strictly in this order because the earlier ones are cheaper, more robust and carry no microbiological penalty.

1. Catchment and source management. Reduce NOM entering the works: source blending, selective abstraction depth, catchment measures on peat and agricultural land. Slow, cheap per unit, and permanent.
2. Precursor removal at the works. Enhanced coagulation (higher dose, optimised pH 5.5–6.5 for alum or ferric), magnetic or suspended ion exchange for DOC, GAC adsorption, or nanofiltration where the carbon is hydrophilic.
3. Disinfection strategy. Move the chlorine dose point downstream of the main precursor removal; use UV as the primary inactivation barrier with chlorine only for residual; consider chloramination for the distribution residual, which markedly reduces THM and HAA formation.
4. Distribution management. Reduce residence time through mains rezoning and reservoir turnover, control booster chlorination, remove dead legs. Formation continues wherever residual and precursor coexist.

Chloramination illustrates the trade-off well. It suppresses THM and HAA formation substantially, but it is a weaker disinfectant, it can promote nitrification in the network with its own compliance consequences, and it favours the formation of NDMA. It solves one problem by accepting a different set.

Worked example: keeping CT while cutting formation

A works treats 20 Ml/d of upland reservoir water: TOC 4.5 mg/L, SUVA 4.2, bromide 30 µg/L, chlorine contact tank volume 1,600 m³. Network THMs reach 92 µg/L in late summer — compliant, but with no margin.

Step 1: verify the disinfection barrier. Contact time must be assessed on t₁₀, not on nominal retention.

• Nominal HRT = 1,600 m³ ÷ (20,000/24 m³/h) = 1.92 h = 115 min.
• Tracer testing gives a baffle factor t₁₀/T = 0.55, so t₁₀ = 63 min.
• At a residual of 0.8 mg/L leaving the tank, CT = 0.8 × 63 = 50 mg·min/L — ample for the required Giardia and virus inactivation at this temperature and pH.

Step 2: attack the precursor. With SUVA 4.2 the NOM is humic and highly amenable to coagulation. Jar testing at optimised coagulation pH raises DOC removal from 32% to 55%, cutting TOC entering chlorination from 4.5 to 2.0 mg/L. Predicted THM factor (2.0/4.5 × nothing else changed)^1.10 = (0.444)^1.10 = 0.41, so 92 → ≈38 µg/L.

Step 3: verify nothing else broke. Enhanced coagulation raises sludge production by roughly the ratio of coagulant dose, lowers alkalinity and may require pH correction after filtration; residual aluminium must be checked at the new pH; and the filters must handle the additional floc load. The engineering here is in the balance, not the headline dose — see our coagulation and flocculation guide for the jar testing protocol.

Step 4: consider the cheap hydraulic gain. Improving the contact tank baffle factor from 0.55 to 0.70 through added baffles raises t₁₀ to 80 min, which allows the same CT at a residual of 0.63 rather than 0.8 mg/L — a 21% lower dose and a proportionate reduction in formation, for the cost of some stainless steel. Identifying that opportunity requires a tracer test or a hydraulic model, as discussed in our article on residence time distribution.

Chlorine decay and the network as a reactor

Once water leaves the works, the distribution system behaves as a long, poorly characterised plug-flow reactor with wall reactions. Chlorine decays approximately first order:

The engineering tension is immediate: enough residual at the extremity of the network to maintain microbiological protection means a high dose at the works, which is exactly the condition that maximises DBP formation over the intervening residence time. Booster chlorination resolves it by re-dosing partway along, allowing a lower works dose and a flatter residual profile — lower peak concentration for the same minimum residual, and therefore less total formation.

Modelling this properly requires a calibrated network hydraulic model with a water quality module, decay coefficients measured on the actual water and pipe material, and residence times derived from real demand patterns rather than design flows. Without it, the sampling programme is the only evidence, and it will always be sparse relative to the number of network locations. The chlorine chemistry itself is covered in our chlorination guide.

Practical programme for a works with a DBP margin problem

1. Characterise the precursor. DOC, UV₂₅₄, SUVA and bromide, seasonally — upland waters vary by a factor of two or more between winter and late summer.
2. Establish the formation profile. Simulated distribution system testing, plus network sampling at the extremities, not just at the works.
3. Fit a site-specific model. Use your own data to fit the power-law exponents; published coefficients indicate the shape of the response, not its magnitude.
4. Audit the contact tank. Tracer test it. A poor baffle factor forces a higher residual for the same CT and is often the cheapest thing to fix.
5. Optimise coagulation before adding a barrier. Where SUVA is above 4, enhanced coagulation is almost always the lowest whole-life cost route.
6. Only then consider GAC, ion exchange or membranes, sized against the precursor that coagulation cannot reach.
7. Re-verify the microbial barrier after every change. No DBP measure may be implemented that has not been shown to preserve inactivation.

Structured that way, most works find sufficient margin without capital works. Where they do not, the choice between adsorption and membrane polishing is a treatability question best answered with a pilot-supported process design rather than a specification.