CO₂ EOR and Huff & Puff

Carbon dioxide is the most successful gas-injection agent in the industry, and among enhanced recovery methods generally it is the one whose activity has kept growing while others declined.^[1] It earns that position through a combination no other cheap gas matches: at reservoir conditions it dissolves readily into oil, swells it, cuts its viscosity, lowers interfacial tension, and — above a threshold pressure — becomes fully miscible with it, erasing the interface that traps oil in pores. But owning a good molecule is not a recovery process. The process is how you deliver it, and there are two: a continuous flood through an injector-producer pattern, and cyclic injection into one well — huff and puff. This article covers the physics they share and the very different engineering they demand.

What CO₂ Actually Does to Oil

Five mechanisms operate, and their relative importance shifts depending on pressure and on which architecture you choose.

• Swelling. CO₂ dissolves into the oil phase in large volumes, expanding it. Swollen oil occupies more pore space, which pushes some of it out of the corners and dead ends where it was stranded, and raises relative permeability to oil.
• Viscosity reduction. The dissolved CO₂ thins the oil, sometimes dramatically for heavier crudes. Thinner oil moves more easily toward the well.
• Extraction of intermediates. Dense CO₂ strips the C₂–C₆ fraction out of the oil, building a CO₂-rich, intermediate-enriched bank at the front. This is not a side effect — it is the mechanism by which miscibility develops.
• Interfacial tension reduction. Lower IFT means a higher capillary number, which means less residual oil left behind at the pore scale. At full miscibility, IFT goes to zero and the concept of residual oil to the displacing phase disappears entirely.
• Solution gas drive on blowdown. When pressure is dropped, dissolved CO₂ comes out of solution and drives oil toward the wellbore. Marginal in a continuous flood; central in huff and puff.

One physical fact underpins all of it: carbon dioxide has a critical temperature near 88 °F (31 °C).^[5] Essentially every oil reservoir is hotter, so CO₂ is never a liquid down there — it is a supercritical dense fluid, with a density approaching that of a liquid and a viscosity closer to that of a gas. That split personality is exactly why it displaces oil so well at the pore scale and sweeps it so badly at the reservoir scale.

Miscibility and the MMP

The single most consequential number in any CO₂ project is the minimum miscibility pressure. Above it the displacement is miscible and pore-scale efficiency approaches complete; below it the process still works, but through the weaker mechanisms of swelling and viscosity reduction alone.

Miscibility with CO₂ is almost never first-contact. It develops through multiple contacts: as the injected fluid moves through the rock it repeatedly vaporises intermediate components from the oil, so the front becomes progressively more like the oil it is displacing until the two compositions merge and the interface vanishes.^[5][6] The theory of how those composition paths develop is a subject in its own right.^[7]

Miscibility is not a property of CO₂. It is a property of CO₂, this oil, this temperature and this pressure — achieved by repeated contact, not by mixing.

Two practical consequences follow. First, MMP must be measured, not assumed — correlations against oil gravity, temperature and C₅₊ composition are useful for screening, but a slim tube test or a rising-bubble apparatus on the actual crude is what a project decision rests on. Second, there is no benefit in pressure far above the MMP: once miscibility is achieved, pushing higher adds compression cost without adding displacement efficiency.^[5] The target is comfortably above MMP and below fracture pressure — a window that, in shallow or hot reservoirs, sometimes does not exist at all.

Architecture One — The Continuous Flood

The classical CO₂ flood injects continuously into a pattern of injectors and produces from surrounding producers. It contacts a large reservoir volume, and where conditions suit it, it is the highest-recovery option available.

Its weakness is not displacement — it is sweep. Supercritical CO₂ is roughly an order of magnitude less viscous than the oil it displaces, so the mobility ratio is deeply unfavourable and the front is unstable: it fingers. It is also much less dense than oil and brine, so it overrides, rising to the top of the interval and leaving the lower part unswept. And it will find any high-permeability streak and channel through it. The result is the recurring paradox of CO₂ flooding: excellent efficiency in the rock it touches, and a disappointing fraction of the reservoir touched.

The standard countermeasure is WAG — water alternating gas — injecting slugs of water between slugs of CO₂. The water reduces the effective mobility of the gas phase and helps hold the front down, at the cost of some loss in contact efficiency and a more complicated operation. Foam, gels and polymer-assisted variants attack the same conformance problem from other directions.^[2][3]

The other defining feature of a mature flood is recycling. Produced gas comes back CO₂-rich and must be separated, dehydrated, recompressed and reinjected. Late in project life, the recycle stream dwarfs the fresh purchase, and the plant handling it — not the reservoir — becomes the limiting asset.

Architecture Two — Huff & Puff

Cyclic CO₂ injection uses one well for all three roles. There is no pattern, no injector-producer pair, and no requirement that CO₂ travel anywhere except into the rock immediately around the wellbore and back out again.

The mechanism mix is genuinely different from a flood. There is no displacement front to speak of; the recovery comes from CO₂ entering the oil near the well, swelling and thinning it, and then from the blowdown when the well is opened and dissolved gas expands to drive oil back. In very tight rock, the rate-limiting step is molecular diffusion of CO₂ into the matrix — which is precisely why soak time matters, and why huff and puff behaves so differently in a shale than in a conventional sand.

What the design levers actually are

• Injected volume per cycle — sets how far CO₂ penetrates from the well. Too small and the treated radius is trivial; too large and much of the gas simply comes back.
• Soak time — the pause that lets diffusion and dissolution do their work. Too short wastes the injected gas, too long wastes calendar time and deferred production. This is the parameter most often set by trial across the first cycles.
• Number of cycles — the practical question is when to stop, because the incremental oil per cycle falls.
• Drawdown on the puff — the blowdown is a driving mechanism, so how quickly pressure is dropped matters, tempered by sanding and asphaltene concerns.

Choosing Between Them

The decisive question is not which recovers more oil — a flood does, when it works — but whether the reservoir will let CO₂ travel from one well to another at all. In shale and tight oil that condition is simply not met: permeability is measured in nanodarcies, wells communicate through fracture networks if at all, and a conventional pattern flood has nowhere to sweep. Cyclic injection into the stimulated volume around a single horizontal well sidesteps the requirement entirely, which is why huff and puff has become the default gas-EOR concept in unconventional plays.^[9]

Huff and puff also earns its place in conventional settings for an unglamorous reason: it is a cheap experiment. A single-well cycle tests whether this oil responds to CO₂ at all, at a fraction of the cost and lead time of a pattern, and the answer informs whether the flood is worth committing to.

Documented Field Applications

SACROC — the project the whole industry learned from

Commercial CO₂ flooding has a single starting point: injection began at the SACROC unit in Scurry County, West Texas, in January 1972, and it is still running more than half a century later — the oldest continuously operated CO₂-EOR project there is.^[11][12] The field was discovered in 1948 and had already been through primary depletion and waterflood before CO₂ was chosen as the tertiary mechanism. Between 1972 and 2009 the project took more than 175 million tonnes of CO₂, and as of 2019 the unit still produced over 28,000 bbl/d by CO₂-EOR.^[12]

Two details from that history land squarely on the arguments made earlier in this article.

• Supply defined the project, not the reservoir. The initial CO₂ came from gas plants roughly 220 miles south, delivered through pipelines built specifically for it; the operators later switched to a natural CO₂ source in Colorado in 1996 partly because the gas-plant supply was not consistent enough.^[11] That is the supply-chain problem of the earlier section, played out over decades.
• The tertiary prize is real but bounded. Published performance at SACROC frames the recovery ladder as roughly 20 % primary, 20 % secondary by waterflood, and about 10 % tertiary by CO₂, with one documented project delivering 2.7 MMbo incremental — slightly under 10 % of its OOIP.^[13]

The more instructive part of the recent record is what happened to areas that had been written off. Two marginal projects, deemed uneconomic for tertiary recovery because of poor-quality pay, were reconsidered on the strength of optimised injection profiles for conformance control, upgraded gas processing and more reliable production equipment. Injection started in 2017 and both met their targets, returning about 8 % tertiary recovery against the roughly 10 % typical of higher-graded projects.^[13] Conformance and facilities decided viability — exactly where this article argued CO₂ projects are won or lost.

Where the idea came from, and where it went next

The first field experiment predates SACROC: a CO₂ injection at the Mead Strawn field near Abilene, Texas, in 1964, which recovered over 50 % more oil than secondary waterflooding.^[14] At the other end of the timeline, the Weyburn–Midale project in Saskatchewan, on injection from 2000, became one of the first international demonstrations that capture, utilisation and storage could be run and monitored together — the associated-storage argument made earlier, executed at field scale.^[15]

The counter-evidence: immiscible CO₂ is rare for a reason

The MMP discussion has a blunt statistical shadow. Immiscible CO₂ floods are a small minority of projects — on one count 7 out of 124 — and their contribution is smaller still: in 2010, immiscible operations produced about 19,200 bbl/d out of roughly 303,000 bbl/d of total CO₂-EOR production.^[14] Below the MMP the process still works, but the record shows it works at a scale that rarely justifies the infrastructure. Where immiscible CO₂ has succeeded at size — the Bati Raman heavy-oil project being the largest — it is on oil that miscible flooding could never have handled anyway.^[14]

Half a century of SACROC says the physics was never the hard part. Supply, conformance and gas-handling capacity decided which barrels were produced.

Screening, and What Actually Kills Projects

The published screening criteria for CO₂ miscible flooding point consistently in the same direction: relatively light oil (broadly above about 22–27 °API, and better lighter), low to moderate viscosity, and — the binding constraint — sufficient depth to hold a pressure above the MMP without fracturing the formation.^[1] Remaining oil saturation must be worth chasing, and the reservoir should not be so heterogeneous or so fractured that conformance is hopeless before the first barrel of CO₂ is bought.

But screening criteria describe where the physics can work. Projects fail elsewhere:

• CO₂ supply. The commonest reason a technically sound project never starts. Volume, purity, delivery pressure, contract duration and distance to a pipeline are project-defining, and impurities cost real money by raising the MMP.
• Conformance. Fingering and override are not edge cases; they are the expected behaviour, and the design must assume them.
• Asphaltene precipitation. CO₂ destabilises asphaltenes in some crudes, plugging pores and tubing. Cheap to test for beforehand, expensive to discover afterwards.
• Corrosion. CO₂ plus water makes carbonic acid. Metallurgy, inhibition and monitoring are a permanent operating cost, not a one-off design item.
• Recycle capacity. In a mature flood most injected gas is recycled gas; if the plant cannot handle the volume, the reservoir's capability is irrelevant.

There is one further dimension that has changed the arithmetic. CO₂ injected for recovery largely stays in the reservoir, so associated storage is an inherent by-product of the process. Where carbon has a price or a compliance value, a project that once had to justify itself on incremental barrels alone can now be assessed on both — which alters the economics of exactly the marginal, mature fields that are otherwise hardest to justify.

Validation

The recovery mechanisms, mobility and conformance arguments, and the fractional-flow framing of gas injection follow Lake and co-authors^[2] and Green and Willhite,^[3] the two standard EOR texts. The screening ranges and the observation that CO₂ flooding is the one EOR method whose activity has grown continuously come from Taber, Martin and Seright.^[1] The mechanism of oil displacement by CO₂, the existence of a miscible-displacement pressure and the finding that pressure well above it brings no further benefit are from Holm and Josendal.^[5] Slim tube determination of the MMP follows Yellig and Metcalfe;^[8] the effect of methane and nitrogen impurities on MMP, including the worked magnitudes quoted in Fig. 01, is from Stalkup's monograph on miscible displacement.^[6] The compositional theory of multiple-contact miscibility follows Orr.^[7] Application to shale and tight oil, where diffusion-limited cyclic injection replaces pattern flooding, follows Sheng.^[9] Documented field experience — what these processes actually did rather than what they were designed to do — is compiled across more than 250 pilot and field applications by Sheng.^[4] The underlying PVT behaviour, swelling tests and gas-injection fundamentals follow Ahmed.^[10]

A CO₂ flood fails at the reservoir scale far more often than at the pore scale. The molecule is rarely the problem; getting it to the oil is.

Frequently Asked Questions

What is CO₂ huff and puff?

Cyclic injection into a single well: CO₂ is injected (the huff), the well is shut in so the gas can soak into the rock and dissolve into the oil, then the same well is produced (the puff). One well does all three jobs — no pattern, no interwell sweep, no requirement that CO₂ cross the reservoir. Incremental oil is largest in the first cycle and declines thereafter.

What is minimum miscibility pressure and why does it matter?

The lowest pressure at which CO₂ and the reservoir oil become miscible through repeated contacts, so the interface between them disappears and pore-scale displacement approaches complete. Above MMP the flood is miscible and highly efficient; below it, recovery relies on the weaker swelling and viscosity mechanisms. MMP is measured in a slim tube and rises with temperature, with heavier oil, and with methane or nitrogen in the injected stream.

When is huff and puff better than a continuous CO₂ flood?

When a flood cannot sweep: very low permeability rock such as shale and tight oil where wells do not communicate; reservoirs too small or compartmentalised for a pattern; limited or intermittent CO₂ supply; and single-well pilots testing the response before capital is committed. A continuous flood recovers more overall when the rock is permeable and continuous enough for CO₂ to travel injector to producer.

What are the main risks in a CO₂ EOR project?

Conformance above all — CO₂ is far less viscous and less dense than oil, so it fingers, channels and overrides rather than sweeping evenly. Then asphaltene precipitation in susceptible crudes, corrosion wherever CO₂ meets water, and the cost and reliability of supply, compression and recycling. Economics are usually decided by CO₂ cost and recycle volume rather than by reservoir physics.