Introduction
Ask what destroys more upstream pipework than anything else and the honest answer is not sand, not fatigue, and not the sour-service cracking that gets all the specification attention. It is sweet corrosion: carbon dioxide dissolved in produced water, quietly thinning carbon steel from the inside.
CO₂ corrosion sits at the centre of one of the biggest cost decisions in facility design. Choose corrosion-resistant alloy (CRA) everywhere and the material cost of flowlines and piping can multiply several times over. Choose carbon steel where the conditions do not support it and the facility inherits a re-piping campaign, deferred production, and an integrity file nobody wants to own. Getting the prediction right — and understanding what it does and does not tell you — is what lets a project put expensive metallurgy only where it earns its keep.
This article covers how CO₂ corrosion works, how it is predicted, what actually controls the rate, and — the part that surprises people — how often plain carbon steel remains a defensible, economical choice.
The Mechanism — and Why Temperature Behaves Strangely
Dry CO₂ is harmless to steel. The problem starts when CO₂ dissolves in liquid water, forming carbonic acid. Although weakly dissociated, carbonic acid is more corrosive than a strong acid at the same pH, because the undissociated molecule provides a reservoir of hydrogen that keeps feeding the cathodic reaction at the steel surface. Iron dissolves, and the wall thins.
The signature of CO₂ attack is localised metal loss — smooth-bottomed pits and the stepped "mesa" attack that develops where flow disturbs partially protective films. Weld root erosion-corrosion at turbulence steps is a classic failure location.
Temperature behaves counter-intuitively. The bare-steel corrosion rate rises with temperature — reaction kinetics — up to a peak typically somewhere in the 60–90°C range. Above that, iron carbonate (FeCO₃, siderite) becomes sufficiently insoluble to precipitate as a dense, adherent scale that isolates the steel, and the rate falls. This is why a hot separator can outlive the cool flowline feeding it, and why a corrosion assessment done only at the design point, rather than across the operating envelope, can miss the worst-case coordinates entirely.
Predicting the Rate: de Waard–Milliams to NORSOK M-506
The starting point for nearly every screening study is the semi-empirical correlation of de Waard and Milliams, first published in 1975 and refined since:
log10(Vcor) = 5.8 − 1710 / T + 0.67 × log10(pCO2)
Vcor = corrosion rate (mm/year), bare steel, water-wetted
T = temperature (K)
pCO2 = CO2 partial pressure (bar)
The original nomogram deliberately errs conservative: it assumes fully water-wetted bare steel with no protective scale and no inhibition. Later revisions added correction factors for protective films, glycol, and in-situ pH.
NORSOK M-506 is the other workhorse — an empirical model fitted to laboratory data that adds wall shear stress and pH effects explicitly. It is valid roughly between 5 and 150°C, pH 3.5 to 6.5, and CO₂ partial pressures up to about 10 bar, and it responds strongly to the pH input — which makes getting the in-situ water chemistry right more important than the choice of model. Mechanistic models exist beyond these two, but for concept and FEED screening, de Waard–Milliams and M-506 bracket the problem well.
Two things matter more than which model you run. First, the models predict a worst-case bare-steel rate — the number is an input to a decision, not a prophecy. Second, the inputs dominate: CO₂ partial pressure (strictly, fugacity at high pressure), temperature, in-situ pH including bicarbonate buffering from formation water, water cut and wetting, and flow regime.
A rough screening rule that predates both models still earns its keep: below about 0.5 bar CO₂ partial pressure (≈7 psi), corrosion is usually manageable; between 0.5 and 2 bar it demands attention; above 2 bar (≈30 psi) expect severe corrosivity and plan accordingly.
A Worked Screening Example
Take a wet-gas flowline: 30 bara operating pressure, 5 mol% CO₂ in the gas, 60°C. The CO₂ partial pressure is 0.05 × 30 = 1.5 bar — in the "demands attention" band.
De Waard–Milliams at 333 K:
log10(Vcor) = 5.8 − 1710/333 + 0.67 × log10(1.5)
= 5.8 − 5.14 + 0.12
= 0.78 → Vcor ≈ 6 mm/year (6.1 unrounded)
Six millimetres a year of bare-steel attack. No corrosion allowance survives that, so carbon steel is only viable here with continuous inhibition — and this is where the arithmetic gets interesting, because what matters is not the inhibited rate alone but the inhibitor availability: the fraction of time the inhibitor is actually present and working.
Assume a well-selected inhibitor achieves 0.1 mm/year when present. The effective rate is the time-weighted blend of inhibited and bare rates:
Availability 90%: 0.90 × 0.1 + 0.10 × 6.1 ≈ 0.70 mm/y → 14 mm in 20 years
Availability 95%: 0.95 × 0.1 + 0.05 × 6.1 ≈ 0.40 mm/y → 8 mm in 20 years
Availability 98%: 0.98 × 0.1 + 0.02 × 6.1 ≈ 0.22 mm/y → 4.4 mm in 20 years
With a 6 mm corrosion allowance and a 20-year design life, only the 98% case works. The design conversation is therefore not "inhibitor or CRA" but "can operations demonstrate 98% injection availability for twenty years — pumps, storage, monitoring, management of upsets — and what does that cost against 13Cr?" That is the honest form of the question, and it is an operating-philosophy question as much as a materials one.
What Actually Controls Corrosivity
Beyond partial pressure and temperature, five factors decide whether the predicted number is real:
- Water wetting. No liquid water on the wall, no corrosion. Crude systems below roughly 30% water cut at reasonable velocity tend to stay oil-wet and behave far better than the models suggest; once water breaks out and wets the wall — low spots, low velocity, shutdowns — the bare-steel rate applies locally.
- In-situ pH. Formation water carries bicarbonate that buffers pH upward and promotes protective siderite; freshly condensed water is unbuffered, sits at pH 3.5–4, and is aggressive. A wet-gas line corroding at the top from condensing water can be in a different regime from its own bottom-of-line.
- Top-of-line corrosion. In stratified wet-gas flow, continuously injected inhibitor travels with the liquid at the bottom of the pipe and never reaches the condensing film at the top. Where condensation rates are high, mitigation is insulation, batch treatment, or volatile inhibitors — not more of the same inhibitor.
- Oxygen. CO₂ systems are anaerobic; even parts-per-billion oxygen ingress from injection packages or storage tanks multiplies rates and defeats inhibitor films. Many "CO₂ corrosion failures" are really oxygen excursions.
- Organic acids and flow. Acetic acid undermines siderite protectiveness at the hot end, and high wall shear or sand erosion strips films mechanically — corrosion and erosion velocity limits need checking together.
One more lever worth knowing: in wet-gas systems dosed with MEG for hydrate control, the glycol itself suppresses the CO₂ corrosion rate substantially — a genuine double duty that a screening study should take credit for.
The Materials Ladder
When the numbers say carbon steel cannot live, the selection ladder runs broadly:
- Carbon steel + corrosion allowance + inhibition — the base case. Allowances of 3–6 mm; viability decided by the availability arithmetic above and by whether the line can be inspected and pigged.
- 13Cr martensitic stainless (and super-13Cr) — the default CRA for sweet wells and flowlines, good to roughly 150°C in sweet service, with tight sour-service and chloride limits that must be checked against ISO 15156 the moment H₂S enters the picture.
- 22Cr / 25Cr duplex — strength plus chloride tolerance for topsides and subsea service, with defined H₂S ceilings.
- 316L and higher austenitics — instrument and utility scale rather than main process lines offshore, given chloride pitting limits.
- Clad or lined systems — carbon steel pressure envelope with alloy 625 weld overlay or CRA lining where severity is high but solid CRA is unaffordable; non-metallics (GRE, HDPE liners) increasingly own low-pressure produced-water duty.
The economics are lifecycle, not tonnage: CRA buys freedom from inhibition logistics, monitoring, and availability risk. On a normally unattended platform or a subsea tieback, that freedom is usually worth the premium. Ten metres from a manned control room with a chemical injection skid, it often is not.
Where Carbon Steel Is Still Fine
The title promised this list, and it is longer than the alloy brochures imply:
- Dry gas downstream of dehydration. Meet the water dew point specification with margin — the whole point of TEG dehydration — and there is no free water and no corrosion; sales-gas pipelines run carbon steel for decades on this basis. The integrity condition is protecting the dew point, not the steel.
- Stabilised, dehydrated crude. Export-specification crude at low water cut in oil-wet flow barely touches the wall. The risk concentrates at dead legs, tank bottoms, and shutdown wetting — manage those locations, not the whole line.
- Low partial pressure systems. Below ~0.5 bar pCO₂ with modest temperatures, predicted rates often sit within an ordinary allowance even uninhibited. Run the number before specifying alloy.
- Inhibited systems with demonstrated availability. Where injection reliability, monitoring (coupons, ER probes, wall-thickness surveys at fixed locations), and a pigging routine are genuinely in place, inhibited carbon steel is the industry's standard answer for long flowlines.
- Short-life developments. An early production facility with a five-year horizon can carry the corrosion in the allowance itself. A 0.5 mm/year effective rate is 2.5 mm against a 6 mm allowance — a monitored, bounded risk that fast-track economics will take.
What unites these cases: the water is controlled, the mechanism is bounded, and there is a monitoring path to confirm the assumption. Carbon steel fails projects when it is chosen on optimism; it serves them when chosen on arithmetic.
Sweet, Sour, or In Between
CO₂ rarely travels alone. A little H₂S changes the film chemistry: iron sulphide precipitates ahead of iron carbonate and, at low levels, can actually reduce the general wall-loss rate — while introducing the entirely separate problem of sulphide stress cracking. As a rough screen, systems with a CO₂-to-H₂S partial pressure ratio above a few hundred behave sweet; below that, sulphide films start to control and the assessment changes character.
Keep the two damage mechanisms distinct in your head and in your reports. NACE MR0175 / ISO 15156 governs cracking resistance in sour service — it says nothing about how fast CO₂ removes wall. A material can be fully MR0175-compliant and still corrode at 6 mm/year; a system can be too sweet to need MR0175 and still eat an unprotected line. Two questions, two assessments.
Common Mistakes
- Using total pressure instead of CO₂ partial pressure — or, at high pressure, ignoring fugacity and over-predicting.
- Running the model with condensed-water pH for a formation-water system (or vice versa). The bicarbonate buffering difference can move the answer by an order of magnitude.
- Quoting the bare-steel rate as the design rate without stating the inhibition and availability assumptions that turn it into the effective rate.
- Claiming inhibitor availability the facility cannot demonstrate. If there is no monitoring, no low-flow alarm on the injection package, and no coupon programme, 98% is a hope, not a number.
- Forgetting the operating envelope. Turndown, shutdown wetting, seasonal cooling, and late-life water cut all move the coordinates — the corrosion assessment belongs to the whole life of the facility, not the design point.
- Conflating MR0175 compliance with corrosion control — see above; the specification that prevents cracking does not slow wall loss.
Conclusion
CO₂ corrosion is predictable — unusually so, for a damage mechanism. The models are mature, the controlling variables are measurable, and the mitigation options are well understood. What separates a sound material selection from an expensive one is honesty about the inputs: the real water chemistry, the real wetting regime, the real inhibitor availability the operation can sustain.
Run the screening early, at concept rather than detailed design, because the answer shapes line sizing, layout, chemical systems, and cost class. And before writing "CRA" across the line list, do the arithmetic: controlled water, bounded partial pressure, and demonstrated inhibition keep carbon steel — with its forgiving fabrication and familiar inspection — the right answer more often than the alloy price list would have you believe.
