Centrifugal Extractor Selection: Sizing and Stages

Last updated: 2026-11-04

Centrifugal extractor selection hinges on three numbers: phase disengagement under about 60 seconds, combined throughput below roughly 100 m3/h, and solvent inventory that is too expensive to hold in open tanks. This guide turns those numbers into rotor volume, G-force, and stage count, then shows where a mixer-settler or column still wins on cost inside any solvent extraction flowsheet. Miss one number and no vendor datasheet will rescue the project.

Key Takeaways

• Pick a centrifugal extractor when phases disengage in under about 60 seconds.
• Size on rotor volume, G-force (100-2,000×g typical), and stage efficiency (85-95%).
• A 5-stage extraction means 5 machines piped in series with interstage pumping.
• Solids tolerance is the hard limit: crud-heavy feeds belong in mixer-settlers.
• Budget 2-4x the mixer-settler purchase price and roughly half the footprint.

What Is a Centrifugal Extractor and How Does It Separate?

A centrifugal extractor is a rotating liquid-liquid contactor that mixes two immiscible phases in an annular or central mixing zone and then separates the dispersion inside a spinning rotor, discharging each phase under its own weir. The rotor replaces gravity with centrifugal force — typically 100-2,000 times the acceleration of gravity (100-2,000×g, indicative across designs) — so droplets that would need minutes to settle in a tank travel a few centimeters in seconds. An annular centrifugal extractor is a configuration that feeds both phases into the annulus between a fixed housing and the spinning rotor, mixes them by shear, and lifts the dispersion into the rotor for separation; a rotary drum type instead mixes internally and separates along the drum wall.

The dispersion band is a transition layer that forms inside the rotor between the heavy-phase layer and the light-phase layer, where droplets are still breaking and coalescing. Its thickness is the machine’s throttle. Push throughput, or drop bowl speed, until the band reaches the heavy-phase weir and organic starts leaving with the raffinate. Watch this band at the inspection port: it announces flooding before the assays do.

Inside the rotor, the interface radius is set by the two weir radii and the density difference between the phases — not by an operator’s valve. Heavy phase climbs the outer wall to its weir; light phase is skimmed near the axis. Change solvent and you change the density difference, so a machine holding a stable interface on one system can sit at an entirely different radius on another. That is the first reason test work on the real phases, at real temperature, matters before rotor geometry is frozen.

Residence time is where the physics departs most sharply from a mixer-settler. A mechanical stage of a centrifugal extractor holds each phase for tens of seconds — commonly 10-60 s of mixing plus 5-30 s of separation (typical, design-dependent) — against the 2-5 minutes of mixing and 30-90 minutes of settling a conventional mixer-settler stage requires. That compression is why total solvent inventory can fall by 60-80% (indicative) for the same duty: less solvent idling in the circuit, a smaller fire and toxicity envelope, less working capital parked in tanks, and a circuit that responds to a flowsheet change in minutes rather than shifts.

Because each machine contributes roughly one equilibrium stage, cascades are built by piping units in series rather than by baffling one open box. For the wider family of contactors and where each one fits, browse our solvent extraction resource hub — the choice among mixer-settler, centrifugal, and column is a flowsheet decision.

Centrifugal Extractor Selection Parameters: What Sizes the Machine?

Centrifugal extractor selection is ultimately a sizing exercise against seven parameters, and the table below is the checklist we work through on every inquiry. All ranges are typical and indicative — every value must be confirmed by test work on your own phases.

G-force is the headline number, but it is a compromise, not a maximum. Raise G and droplets settle faster, so the rotor shrinks — but the mixing shear usually rises too, droplets leave the mixing zone finer, and the raffinate entrainment climbs. Most hydrometallurgical duties settle out between 300 and 600×g; pushing toward 1,000×g and beyond is justified for tight emulsions or very high-value liquors, and paid for in bearing life and power. The number that actually sizes the purchase, though, is combined throughput per rotor volume: divide total phase flow by the 20-60 m3/h per m3 band, and the rotor volume — and therefore the machine class and most of the price — falls out immediately.

Phase ratio flexibility is the parameter projects forget until commissioning. Because the interface radius is a geometric consequence of weir radii and density difference, a machine tuned for O/A 1:1 will not hold its interface at O/A 4:1 unless the design anticipated it. Quality designs accept the 1:5 to 5:1 band on the same hardware; anything wider needs a weir change or a variable-weir design, and that is a spec-sheet question worth asking every vendor.

Stage efficiency is a fraction that compares a real mechanical stage’s mass transfer performance to an ideal equilibrium stage — typically 85-95% per machine — so the number of units equals ideal stages divided by efficiency, rounded up. There is no “half stage” and no shared wall between stages: a 5-stage extraction means 5 physical units in series, each with its own motor, its own interstage piping, and either its own pump or an elevation scheme that lets phases flow to the next unit. Budget that piping and those pumps as part of the contactor cost from day one — on a 6-stage cascade they are not rounding error.

One warning that only surfaces in pilot campaigns: 80mm-class pilot rotors do not always scale linearly to industrial machines for systems with slow coalescence. A small rotor gives droplets high G but a very short settling path and a short time under load, and it usually runs on fresh, clean organic; an industrial rotor gives a longer path but gentler effective settling per millimeter, and plant organic ages. Sigma-based scale-up has over-predicted industrial capacity by a factor of 2 or more on exactly these systems. Run the confirmation on real, aged phases — a campaign on our pilot-scale annular centrifugal extractor exists for this reason, and the 80mm rotary drum centrifugal extractor is the bench counterpart we use for phase-disengagement screening.

Where Does Centrifugal Extractor Selection Beat the Alternatives?

Good centrifugal extractor selection is scenario work: name the scenario, name the winner. Below are the five we meet most often, each with an explicit conclusion — because “it depends” is not a design deliverable.

Scenario 1: high-value metal, fast disengagement, tight footprint. Nickel, cobalt, rare earth, and battery-recycle liquors typically disengage in well under a minute, the metal value justifies tight control, and the plant often sits indoors or on expensive ground. Verdict: the centrifugal extractor wins. You buy fewer square meters, a fraction of the solvent inventory, and stage-by-stage instrumentation that a heap of open tanks cannot match.

Scenario 2: large-tonnage base-metal SX with crud-bearing PLS. Copper and zinc circuits move hundreds to thousands of m3/h, and their loaded liquors carry clay, silica, and crud that would pack a rotor in weeks. Land is cheap, phases are dilute, and settlers are forgiving. Verdict: mixer-settler preferred, full stop. The mixer settler design guide covers that sizing path.

Scenario 3: six or more stages at modest flows. Each centrifugal unit is a pressure-boundary machine with a motor, a seal, and interstage pumping, so a long cascade multiplies cost and failure points faster than a train of open boxes that share walls. Verdict: at 6+ stages with modest flows, a column or a mixer-settler cascade is normally cheaper, provided floor space and solvent inventory allow it.

Scenario 4: rapid ramp-up, hazardous solvent, remote or indoor site. If the diluent is flammable, the reagent is toxic, or the plant must start producing within months of mechanical completion, the small holdup and short inventory of centrifuges pay directly. Verdict: centrifugal wins, and the faster commissioning is worth money — count it in the comparison.

Scenario 5: strict raffinate entrainment spec. When the raffinate feeds ion exchange, an electrowinning dam, or an effluent stream with an organic-in-water limit, centrifuges discharge cleaner than gravity settlers under comparable chemistry. But read the baseline first — entrainment in mixer settlers explains what gravity sets leave on the table, and mixer settler stage efficiency shows how much of the gap is chemistry rather than hardware. Verdict: centrifugal wins on entrainment, unless the upstream crud load disqualifies it under Scenario 2 logic.

One structural note that surprises first-time buyers: because every centrifugal stage is a discrete machine, the cascade needs a valve-and-header philosophy decided early — dedicated standby per stage, one shared standby on a common header, or no standby at all for a short train. Each choice changes piping, elevation, and price, and retrofitting a standby later is far more painful than choosing one on paper.

What Goes Wrong in Service — and How Do You Fix It?

Most centrifugal extractor selection mistakes surface in the first six months of operation as one of the six failure modes below. Each is written the way shift reports read: phenomenon, then cause, then fix.

1. Organic in the raffinate. Phenomenon: a visible sheen or measured organic entrainment climbing above roughly 50 ppm (typical alarm band) on the heavy-phase discharge. → Cause: the dispersion band has been pushed to the heavy-phase weir by over-throughput, low G-force, or fine droplets stabilized by upstream reagent carryover, humates, or silica. → Fix: cut combined flow 10-20% or raise bowl speed one step, verify the interface radius against the design density difference, and treat the chemistry — stabilized fines cannot be spun out at any economical G.

2. Vibration that grows over weeks. Phenomenon: bearing-housing vibration creeping from under 2 mm/s toward 4.5 mm/s and beyond (indicative alarm limits). → Cause: crud solids depositing asymmetrically on the rotor wall — a crud excursion leaves a few hundred grams of hardened, phase-dispersed solids turning at 900 rpm. → Fix: open, clean, inspect, and balance-check the rotor before it returns to speed, every time. A rotor that is cleaned but not rebalanced after a crud excursion has torn up its own bearings within days; balancing discipline after any excursion is the most skipped, most expensive-to-skip procedure on this machine.

3. Interface loss after a solvent or feed change. Phenomenon: both weirs discharging mixed phase shortly after a reagent or diluent change. → Cause: the density difference moved, so the interface radius moved out of the design band; the weirs are fixed geometry. → Fix: return to the design O/A ratio, re-verify densities at operating temperature, and fit adjusted weirs if the new regime is permanent.

4. Cross-contamination at startup. Phenomenon: first hours of product off-spec after every shutdown. → Cause: interstage lines and rotor sumps drain unevenly between phases, and residual organic (or aqueous) flushes forward into the wrong stage. → Fix: a written drain-and-flush sequence per unit and a staged startup at low flow until each interface is established — improvised startups always cost a shift’s production.

5. Mechanical seal weeping. Phenomenon: process liquor or dye-trace fluid at the seal faces. → Cause: faces scored by crystal grit during a solids excursion, or a flush plan starved by a clogged filter. → Fix: scheduled seal inspection, verified clean flush supply, and never run to failure — a seal failure at speed throws liquor, and the cleanup dwarfs the seal.

6. Capacity shortfall against the pilot. Phenomenon: flooding at 70-80% of the pilot-predicted rate. → Cause: the classic pilot-scale nonlinearity — an 80mm-class rotor over-predicts slow-coalescing systems, and aged plant organic coalesces worse than the fresh pilot solvent. → Fix: re-rate G upward within the rotor’s mechanical limit first, then buy capacity; and re-run disengagement test work on aged organic before ordering more machines, because the second campaign usually explains the first.

A pattern runs through all six: the machine rarely fails alone. Chemistry pushed it, geometry answered, and the fix usually spans both. Diagnose the flowsheet before the bearings.

What Do Centrifugal Extractors Cost to Buy and Run?

Against an equivalent mixer-settler train, expect a purchase price of roughly 2-4x per unit of throughput (indicative, for mid-size stainless machines), partially offset by an installed footprint 30-50% smaller, solvent inventory 60-80% lower, and building volume that shrinks with it. The capex gap narrows as stages multiply, because open settlers share walls and centrifuges share nothing — which is the arithmetic behind the 6-stage verdict in the previous section. All figures below are indicative planning ranges on a 1-2 m3-class rotor, 300-600×g, installed but excluding civil works; treat them as budget-grade only, to be confirmed by quotation.

Materials of construction drive more of the price spread than size does, and the corrosion duty — sulfate or chloride, warm or hot — decides the material long before throughput does.

Energy is modest but not free. A 300-600×g class machine drawing a few kilowatts per unit consumes on the order of 0.2-0.5 kWh per m3 of combined phase throughput (indicative; assumption: mid-size rotor at design flow, including mixing energy) — comparable to mixer-settler mixing power at small scale, and cheaper than it where settlers need large slow agitators plus interstage pumping across long distances. At part load, efficiency degrades faster than on a settler because the rotor still spins at full speed; a plant expecting years of 30% turn-down should price that into the operating comparison.

Maintenance is the real operating difference. Plan an annual overhaul window covering mechanical seals, bearings, gasket kits, and a rotor balance check, with an indicative maintenance cost of 3-6% of purchase price per year (assumption: clean service; crud-heavy service can double it, and the vibration item above is why). Seals are consumables in any honest budget, and rotor balancing after every crud excursion is a fixed labor line, not an occasional surprise. None of this is secret — it is simply omitted from first-pass economics, and it is what turns a 2x premium into a 3x total-cost premium when the feed is dirtier than the datasheet assumed.

Worked Sizing Example: 7 Steps From Duty to Machine Count

Here is a complete, redoable centrifugal extractor selection for a nickel-cobalt purification duty, with every number traceable and internally consistent. The seven steps below are the same ones we run before any quotation.

1. Define the duty and flows. Aqueous feed is 25 m3/h of purified leach liquor; the extraction requires O/A 0.8, so organic flow is 20 m3/h and combined throughput is 45 m3/h. Record both phases’ densities and viscosities at operating temperature — they enter steps 4 and 5.
2. Set residence times from bench data. Shake tests show 90% of equilibrium reached in 30 s of mixing. Apply a 1.5x margin: design mixing residence 45 s. Phase disengagement on the same tests completes in well under 15 s, so design separation residence 20 s.
3. Convert residence to machine volume. Mixing volume = 45 m3/h × (45/3600) h = 0.5625 m3. Separation volume = 45 × (20/3600) = 0.25 m3. Total internal volume = 0.8125 m3, which selects a 1.0 m3-class rotor. Check against the throughput band: 45 m3/h ÷ 1.0 m3 = 45 m3/h per m3, inside the 20-60 indicative band.
4. Fix the G-force and speed. Choose 350×g at the rotor wall (mid-band for SX). For a 0.8 m rotor diameter (r = 0.4 m): ω² = 350 × 9.81 / 0.4 = 8,584 rad²/s², so ω = 92.6 rad/s, which is 92.6 × 60 / (2π) ≈ 885 rpm — well within a standard drive range.
5. Confirm the dispersion band on real phases. A pilot run on aged plant organic at 45 m3/h equivalent shows a stable band with roughly 40% margin to the heavy-phase weir at 350×g. This step is where the 80mm-vs-industrial nonlinearity gets retired with data instead of hope.
6. Count the stages. McCabe-Thiele on the bench isotherm gives 3 ideal stages for the extraction strip of the flowsheet. At 85% stage efficiency: 3 / 0.85 = 3.53, rounded up to 4 mechanical units in series.
7. Decide sparing and piping. Four duty units plus one shared standby on a common header, interstage pumping per unit with valved bypasses, and elevation set so the header drains completely on shutdown. The cascade is now a hardware list, not a concept.

The output: four 1.0 m3-class units at 350×g and 885 rpm, one shared standby, and a piping philosophy decided before the purchase order. Every number here came from three bench tests, one pilot campaign, and arithmetic a reviewer can check line by line — which is exactly what a sizing document is for.

FAQ

How many centrifugal extractor stages does a typical hydrometallurgical circuit need?

Most purification duties need 3-6 mechanical stages: 2-4 ideal stages from the McCabe-Thiele diagram divided by 85-95% stage efficiency and rounded up. An extraction scrub strip, for example, that resolves to 3 ideal stages needs 4 machines at 85% efficiency. Count units, not boxes — each mechanical stage is one physical rotor, so the stage count drives cascade piping, sparing, and price directly.

What G-force is right for solvent extraction duties?

Hydrometallurgical duties typically run 300-600×g at the rotor wall, inside the wider 100-2,000×g design range. Higher G shortens settling time but increases droplet shear and bearing loads, so the optimum is the lowest G that holds a thin, stable dispersion band at design flow. Set it from pilot data on your actual phases at operating temperature, then verify band margin after the first month of plant operation.

Can a centrifugal extractor handle suspended solids or crud?

Only in small quantities — below roughly 0.1-1% w/w depending on design, and crud is the worst case because it deposits on the rotor wall as an asymmetric layer. Gravity settlers tolerate crud by design; centrifuges accumulate it until vibration or flooding forces a washdown. If the feed carries meaningful crud, either install upstream polishing filtration or select a mixer-settler, and say so in the selection document.

Why is rotor rebalancing needed after a crud excursion?

Crud solids deposit unevenly on the rotor wall, and even a few hundred grams of asymmetric deposit at 900 rpm creates rotating forces that the bearings see continuously. Vibration then climbs over days to weeks, finishing in bearing damage or worse. Cleaning removes the deposit, but only a balance check proves the rotor is safe to return to speed. Treat rebalancing after every crud event as mandatory, not optional, maintenance.

How is the phase ratio controlled without changing hardware?

Within the design’s flexibility band — typically 1:5 to 5:1 on the same hardware — the ratio is set by flow control on each phase feed, while the interface radius follows from weir geometry and density difference. Beyond that band, the interface leaves the design position and both weirs can discharge mixed phase. Wider ratios need adjusted or variable weirs specified at purchase, which is why the expected ratio range belongs in the datasheet from day one.

Do pilot-scale results scale directly to industrial rotors?

Not always. An 80mm-class pilot rotor gives droplets high G over a very short path and short time, usually on fresh organic, while industrial rotors present a longer, gentler settling field and aged solvent. For slow-coalescing systems this combination has over-predicted industrial capacity by a factor of 2 or more. Scale up with disengagement data on aged, real phases and confirm the dispersion band margin on the pilot before freezing machine count.

What maintenance interval should a plant budget for?

Plan an annual overhaul covering mechanical seals, bearings, gaskets, and a rotor balance check, with day-to-day attention to vibration trending and seal weeping. Indicatively, budget 3-6% of purchase price per year in clean service; crud-heavy or chloride service can push that higher. Machines in continuous duty rarely stop for attention between annual windows — but when a crud excursion happens, the balance check cannot wait for the schedule.

Conclusion

Centrifugal extractor selection is not a catalog exercise; it is three numbers — disengagement time, throughput per rotor volume, and stage count — plus the honesty to admit when crud, scale, or a six-stage cascade hands the duty back to a mixer-settler or a column. Size the rotor from residence time, fix G-force from pilot data on aged phases, convert ideal stages to machines at real stage efficiency, and budget the seals, bearings, and balancing discipline from the first estimate. If the scenario work above puts your duty in the centrifugal column, send us the flows: request a quote with your phase flows, densities, and disengagement data, and our engineers will return a sized machine list with the arithmetic shown — or a candid recommendation that a settler serves you better. Either answer saves a project; we would rather be candid early than expensive later.

References

• Ritcey, G.M. and Ashbrook, A.W., Solvent Extraction: Principles and Applications to Process Metallurgy, Part I and II, Elsevier.
• Lo, T.C., Baird, M.H.I. and Hanson, C., Handbook of Solvent Extraction, Krieger Publishing.
• Rydberg, J., Cox, M., Musikas, C. and Choppin, G.R., Solvent Extraction Principles and Practice, 2nd ed., Marcel Dekker.
• Thornton, J.D., Science and Practice of Liquid-Liquid Extraction, Oxford Science Publications.
• Perry’s Chemical Engineers’ Handbook, 9th ed., McGraw-Hill (liquid-liquid extraction equipment section).
• SME Mineral Processing and Extractive Metallurgy Handbook, Society for Mining, Metallurgy and Exploration, Englewood, CO.
• ALTA Metallurgical Services proceedings (ALTA SX/ISX sessions), Perth.
• Hydrometallurgy journal, Elsevier.

Changelog

• 2026-11-04: Initial publication.