Continuous vs Batch Crystallizer: Selection Rules

Last updated: 2026-10-23

For most hydrometallurgical liquors above roughly 5 t/day (5.5 short tons/day) of crystal output, the continuous vs batch crystallizer decision defaults to continuous: a mixed-suspension mixed-product-removal (MSMPR) or Oslo circuit typically cuts crystallization cost per tonne by 20–35% versus batch operation (illustrative, 2026 equipment pricing). That default is overturned only by frequent product changeovers, campaign production, or liquor chemistry that demands seed-controlled batch trajectories.

Key Takeaways

• Continuous evaporation and crystallization circuits win on cost above ~5 t/day; batch wins on flexibility and changeovers.
• MSMPR crystallizers deliver steady-state crystal size distribution (CSD) control; Oslo fluidized-bed (FB) designs make the coarsest, narrowest crystals.
• Batch cooling crystallization gives trajectory control per cycle but a wider CSD spread (CV 0.6–1.0) and higher unit labor.
• Capital expenditure (CAPEX) per tonne/day favors continuous at scale; operating expenditure (OPEX) favors continuous with MVR heat integration.
• When product spec, feed composition, or both are still uncertain, run pilot crystallization before freezing the flowsheet.

How a Continuous Crystallizer and a Batch Crystallizer Actually Work

The continuous vs batch crystallizer distinction starts with how supersaturation is generated and consumed, because that single difference cascades into every downstream property: crystal size, purity, filterability, and the shape of your cost curve.

A mixed-suspension, mixed-product-removal (MSMPR) crystallizer is a continuously operated, agitated vessel that receives clarified feed liquor and withdraws product slurry at the same volumetric rate, holding magma density and supersaturation near steady state. Supersaturation is generated by flash evaporation under vacuum, by surface cooling, or by adiabatic evaporation, and it is consumed continuously on the circulating crystal population. The crystal size distribution (CSD) is then set by the population balance — nucleation rate, growth rate, and solids residence time, which is usually decoupled from liquor residence time by an internal clarification or thickening zone (Randolph & Larson, 1988). For the evaporation hardware that commonly feeds this duty, see our forced-circulation evaporator page.

An Oslo crystallizer is a gravity-circulation, fluidized-bed design that keeps a dense crystal bed suspended in supersaturation-lean liquor, so crystals grow without violent contact nucleation from the pump or agitator. Because the bed classifies itself — small crystals ride high, large crystals sit low and are withdrawn from the base — the product CSD is markedly coarser and narrower than MSMPR output of the same duty. Geometry, circulation ratio, and bed-height control are covered on the dedicated Oslo crystallizer page.

A batch cooling crystallizer is a sealed, agitated pressure or vacuum vessel that takes one charge of feed liquor, is seeded or nucleated deliberately, and is then cooled or evaporated along a programmed trajectory until the whole charge is discharged to a filter or centrifuge. One vessel processes one batch at a time; effective annual capacity equals charge volume multiplied by cycles per day, so cleaning and centrifuge turnaround belong in the capacity math, not in a footnote. Design details for both families are collected in our crystallizer design resource.

In short, continuous designs trade mechanical complexity for steady-state economics; batch designs trade economics for trajectory control. The choice between them is ultimately a choice of which failure mode you would rather manage: turndown drift and changeover cuts, or cycle-time overhead and per-batch variability.

Design and Operating Parameters That Separate the Two

In any continuous vs batch crystallizer review, start from the parameter envelope below rather than from vendor brochures. The ranges are typical design envelopes for inorganic hydrometallurgical salts (illustrative design bases; project-specific values must be confirmed by liquor testing and pilot work), and the two numbers that dominate downstream performance are the median crystal size (D50) and the coefficient of variation (CV), the standard CSD width metric.

the Evaporator Crystallizer Systems for Mining Brines guide

Supersaturation discipline is the deepest difference. A continuous circuit holds the driving force slightly above the metastable limit everywhere in the vessel, all the time; nucleation is dominated by controlled secondary mechanisms and can be tuned with magma density and fines removal. A well-instrumented MSMPR uses that tuning continuously: raising magma density damps contact nucleation, while a fines-dissolution loop — withdrawing liquor loaded with crystals below a cut size, dissolving them, and returning clear liquor — trims the small end of the population without touching the coarse product. A batch vessel passes through a supersaturation peak once or twice per cycle, and the highest local value usually occurs at the heat-transfer surface, which is also where scale forms. Encrustation management therefore belongs in the selection conversation from day one — see our evaporator scaling control guidance for the fouling mechanisms and mitigation options.

Turndown behaves differently, too. Below roughly 40% of design rate, a continuous crystallizer simply stretches its residence time: the population ages, secondary nucleation shifts, and D50 wanders until the circuit is rebalanced. A batch train turns down for free — you just run fewer batches — but its operators, centrifuges, and cleaning labor stay on the payroll either way.

Finally, energy topology. Mechanical vapor recompression (MVR) is an evaporation energy-recovery method that recompresses the vessel’s own boiling vapor so its latent heat is reused instead of rejected to cooling water, and it pairs naturally with continuous crystallization because the vapor load is constant. A steady vapor load also means a steady compressor operating point, which keeps the compressor near its best efficiency and avoids the start-stop cycling that erodes seals and bearings. Batch vacuum crystallization draws a cycling vapor load that complicates recompression and usually pushes plants toward fresh steam or multiple-effect evaporation (MEE). The energy-system consequences are quantified in the cost section below.

Continuous vs Batch Crystallizer: Which Delivers the Better Crystal Size Distribution?

Crystal quality is where the crystallizer decision becomes contractual, because washing, dewatering, dissolution behavior, and downstream customer acceptance all key off the CSD — and off the mother-liquor inclusions trapped inside the crystals, which scale with peak supersaturation. A narrow, coarse CSD drains and washes cleanly; a fines-heavy CSD blinds filter cloths, carries mother liquor into the cake, and pushes impurity specs out of limits at the dryer rather than at the crystallizer, where the fix would have been cheaper.

how ASME VIII Div 1 vs Div 2 is decided

At steady state, an MSMPR produces a self-consistent CSD governed by the population balance: with typical inorganic growth rates of 10⁻⁸ to 10⁻⁷ m/s and residence times of 2–6 h, expect D50 of 150–600 µm with a CV of 0.4–0.6 (typical design envelope; Mersmann, 2001). The Oslo FB design exists precisely for specs beyond that envelope: by suppressing contact nucleation and classifying the bed, it routinely delivers D50 of 0.5–2 mm with CV of 0.2–0.4, at the cost of a taller vessel and tighter turndown.

Batch cooling crystallization, by contrast, lets the operator walk the metastable zone with a seed charge and a programmed cooling ramp. Well-executed seeded batches reach CV of 0.5–0.7; unseeded vacuum batches commonly sit at 0.8–1.0 with strong cycle-to-cycle drift, because natural cooling makes supersaturation peak early, nucleating a fines wall. That drift is the hidden cost: every filter cycle, wash ratio, and drying behavior requalification across batches is a plant-hygiene tax that never appears on the equipment datasheet. Measurement practice matters as much as the topology — insist on in-line or at-line particle sizing (laser diffraction or image analysis) with a documented sampling point, because sieve stacks under-report the fines fraction that actually determines centrifuge behavior, and a CV measured two different ways can differ by 0.1 or more on identical product.

Read the last column as rules, not preferences. For battery-grade nickel sulfate, precursor makers increasingly demand coarse, low-CV product, and the steady low supersaturation of a continuous Oslo circuit also lowers occluded mother liquor, improving rejection of cobalt, calcium, and magnesium — a purity argument, not just a CSD argument. For copper sulfate, the market accepts a wide CSD, so economics decide, and economics favor continuous at tonnage. For ZLD sodium sulfate, feed variability ends the debate in favor of continuous with a well-mixed surge tank. For lithium salts, choose batch first for the qualification phase: the morphology margin you get from hand-held seed control is worth more during customer sampling than any steady-state number.

Continuous vs Batch Crystallizer CAPEX and OPEX: Which Costs More?

All figures below are illustrative equipment-only budgets — ex-works China, 2026 basis, 316L stainless steel for the sodium sulfate duty, 7,200 operating hours per year, power at 0.09 USD/kWh and steam at 25 USD/t. They are budgeting aids, not quotations; corrosive nickel duties requiring 904L or 254SMO raise both columns materially.

On CAPEX, batch is almost always cheaper to buy. A two-vessel batch train of 12 m³ vacuum-cooling crystallizers with condenser and utilities typically budgets at USD 1.0–1.5 million; a continuous forced-circulation MSMPR line with MVR skid, thickening zone, and automated controls for the same output budgets at USD 1.8–2.6 million. The premium is the compressor, the clarification internals, and the instrumentation that keeps the steady state steady.

On OPEX, the ranking inverts at any real tonnage, and the gap is driven by energy plus labor per tonne of product.

Worked-example basis: 10 t/day of Na₂SO₄ requires about 83 t/day of 12 wt% feed, leaving roughly 73 t/day of water to evaporate; the batch column assumes a 6 h effective cycle with two vessels online and one on wash. Below about 2 t/day of product, the same arithmetic flips: the continuous premium is never repaid, and the batch train’s CAPEX advantage stands. Between 2 and 5 t/day, the answer lives in your steam and power tariffs and your labor rates, which is exactly why we publish the basis instead of a verdict. For the energy-system choice behind the continuous column, compare MVR versus multiple-effect evaporation. Maintenance tells a similar story: continuous circuits carry compressor and pump wear parts on a planned schedule (budget 2–4% of equipment CAPEX per year, illustrative), while batch plants spread wear across far more frequent valve, agitator-seal, and gasket cycling — both figures belong in the OPEX line your finance team will ask about.

Continuous vs Batch Crystallizer: Which Should You Choose?

1. Above ~5 t/day of single-product crystal output, go continuous. Labor, wash and centrifuge cycle overhead, and un-recovered latent heat all scale per batch; at tonnage, the fixed batch overhead alone exceeds the continuous CAPEX premium within 2–4 years on the worked-example basis above.
2. Below ~2 t/day, or with more than three products on one line, go batch. Cleaning validation, changeover losses, and idle turndown outweigh energy savings; a continuous vessel that never reaches steady state loses its CSD advantage and keeps all of its complexity.
3. If the contract specifies D50 above 400 µm with CV below 0.4, specify an Oslo fluidized-bed continuous design — or a seeded batch with programmed cooling at small tonnage. An unseeded MSMPR cannot guarantee that spec, and an unseeded batch will nucleate a fines wall in the first hour; there is no retrofittable fix for either.
4. If the liquor scales — gypsum, calcium fluoride, or silica — settle the fouling strategy before choosing the topology. A scaling surface in a batch vacuum pan at least gives you a daily inspection window, which a sealed FC circuit forfeits; our evaporator scaling control guidance details how antifoulant dosing, turbulence design, and clean-in-place cycles shift this rule.
5. If the upstream leach is itself batch, either install an equalization tank sized for at least two batch volumes or make the crystallizer batch too. A continuous crystallizer fed a slug of off-spec liquor produces a contaminated product cut that no downstream wash removes; in our field experience, undersized buffer storage is the single most common self-inflicted wound in new crystallization plants — the surge tank is the cheapest crystallizer upgrade most projects skip.
6. For zero liquid discharge (ZLD) projects with drifting brine composition, default continuous with a well-mixed surge tank, not batch. ZLD brines change weekly as the mine plan changes; batch cycle plans written for last quarter’s liquor become this quarter’s off-spec product, while a continuous circuit with a large buffer simply re-finds its steady state. See our zero liquid discharge applications page for full-brine flowsheets.
7. If liquor chemistry, impurity rejection, or crystal habit is unproven, book a pilot campaign before committing CAPEX. Bench beakers rarely reproduce secondary nucleation or scale rates; a 2–4 week pilot on real liquor converts the rules above from guesses into design inputs — see our pilot testing services.

Apply the rules in order. The first three resolve most greenfield decisions outright; rules 4 through 6 are the retrofit discipline, written because crystallization plants fail in the field on fouling, feed surges, and brine drift far more often than on thermodynamics. Rule 7 is the honesty check. If two rules conflict — a coarse CSD spec at 1.5 t/day across five products, say — the higher-numbered rule usually wins, because flexibility failures cost more per event than CSD shortfalls, and a specification that cannot survive its own liquor is a specification nobody can warrant.

What Do Real Project Numbers Show?

Three illustrative cases, each with its basis stated, show how the rules cash out.

First, a battery-grade nickel sulfate hexahydrate plant designed for 20,000 t/a product on an Oslo continuous circuit: design basis gives D50 of 480 µm, CV of 0.31, and thickener underflow at 35 wt% magma (illustrative design values, to be confirmed at pilot). The consequence that matters commercially is wash water below 0.15 m³ per tonne of product on a single-stage pusher centrifuge, which shrinks both the water balance and the effluent treatment load.

Second, a copper sulfate pentahydrate line at 4 t/day built as a two-vessel batch train, 12 m³ vessels, 8 h total cycle with vacuum flash cooling (illustrative heat balance: 25 kW jacket duty plus adiabatic flash at 12 kPa). Three batches per vessel per day yield the 4 t/day target with one operator per shift — but note that capacity planning had to include a 40-minute wash between charges, and the third vessel position reserved for future capacity was never installed because the two-vessel train met demand.

Third, a 10 t/day sodium sulfate ZLD duty: the MVR-driven continuous circuit carries a specific evaporation energy of 28 kWh/t water (vendor guarantee basis), against a batch design at 0.68 t steam/t water (design basis). At 0.09 USD/kWh, 25 USD/t steam, and 7,200 h/a, the annual energy spread is roughly USD 400k (illustrative) — the arithmetic behind Rule 1. The same project’s batch alternative also needed a third vessel and a duplicate centrifuge to cover wash downtime, which moved its true CAPEX from the attractive budget figure to within 20% of the continuous line before a kilowatt-hour was counted.

A caution from the literature as well as the field: published CSD and nucleation-kinetics data in journals such as Hydrometallurgy and Crystal Growth & Design rarely transfer unchanged to a new liquor, because secondary nucleation depends on impurity traces at the parts-per-million level. Treat every number in this section as design intent until your own liquor has said otherwise.

Frequently Asked Questions

Can a batch crystallizer make battery-grade nickel sulfate hexahydrate?
Yes, but only with disciplined seeding and a programmed cooling trajectory, and holding that cycle after cycle is the hard part. Modern cathode precursor specifications push CV and inclusion limits toward what steady-state continuous circuits do naturally. Batch suits qualification runs, customer sampling, and plants below about 2 t/day. Above that tonnage, an Oslo continuous design holds the CSD with far less operator attention, which is why most new battery-grade plants specify continuous crystallization.

What is the smallest economic size for a continuous crystallizer?
Around 3–5 t/day of crystal product is where continuous equipment typically repays its higher CAPEX through lower labor and energy per tonne (illustrative; the exact threshold depends on energy tariffs and duty hours). Below roughly 2 t/day, the vessel spends its life in deep turndown, the CSD drifts, and a batch train is cheaper to own and easier for a small operations team to run and maintain.

Why does an Oslo crystallizer make coarser crystals than an MSMPR?
Because the fluidized bed removes the strongest nucleation sources: crystals grow in a mild, classified environment instead of colliding with an impeller or with each other at high shear. Secondary contact nucleation falls, so the available supersaturation is spent on growth rather than on new surface area, and the bed’s own classification withdraws only the coarsest fraction as product while fines circulate until they grow.

How long does one batch crystallization cycle take?
Typically 4–12 hours including filling, seeding, controlled cooling or vacuum evaporation, digestion, and discharge, with 6–8 h common for inorganic hydrate salts (illustrative). Cooling rate is the main lever: faster cooling raises supersaturation, nucleates fines, and shortens the cycle at the cost of CSD quality. Throughput planning must also include vessel washing and centrifuge turnaround, which together add 0.5–1.5 h per cycle in practice.

Do continuous crystallizers need seeding?
Only at startup, to establish the working crystal population; after that the steady-state population is self-sustaining and excess fines are dissolved in a fines-destruction loop. The operating lever is instead the solids residence time, usually controlled by a thickening or clarification zone inside the vessel. If feed composition swings hard enough to crash the population, an automatic seed injection system is cheap insurance against a multi-day recovery.

Which is cheaper to operate: continuous or batch crystallization?
Continuous, almost always at tonnage, because it pairs naturally with MVR or multi-effect heat integration and needs fewer operator-hours per tonne. In the worked example above, annual energy and labor for a 10 t/day sodium sulfate duty came out near USD 95k continuous versus USD 545k batch (illustrative). Batch OPEX improves with cheap site steam and low labor rates, but the ranking rarely reverses once output passes a few tonnes per day.

Can one crystallizer serve several different products?
Batch equipment is built for exactly that: drain, wash, validate, and reseed for the next salt, with changeover measured in hours. A continuous line can switch products too, but each changeover produces an off-spec transition cut, a CSD re-stabilization period of one to two residence times, and a full cleanout; scheduling economics deteriorate quickly beyond about three products per line. Wide product portfolios belong in batch vessels or dedicated continuous trains.

Conclusion: One Rule Set, Not a Coin Flip

The continuous vs batch crystallizer decision is a capacity-and-changeover calculation, not a matter of taste: default to continuous — MSMPR for tonnage, Oslo when the CSD spec is coarse and narrow — above roughly 5 t/day of single-product output, and default to batch below about 2 t/day, for campaign production, or during product qualification, with the 2–5 t/day gray zone resolved by your own steam, power, and labor tariffs. Every number in this article was labeled with its basis, and the ones that matter most to your project — supersaturation margins, scale rates, achievable CV on your actual liquor — can only be measured, not assumed. That measurement is inexpensive next to a mis-sized crystallization train: a short pilot campaign on real feed liquor converts this rule set into a flowsheet you can defend to your board. Start with our pilot testing services to lock the design inputs, then browse the resource library for crystallizer design guides, scaling control, and energy-system comparisons before you issue the specification.

References

1. Mersmann, A. (ed.). Crystallization Technology Handbook, 2nd ed. Marcel Dekker, New York, 2001.
2. Mullin, J. W. Crystallization, 4th ed. Butterworth-Heinemann, Oxford, 2001.
3. Randolph, A. D. and Larson, M. A. Theory of Particulate Processes: Analysis and Techniques of Continuous Crystallization, 2nd ed. Academic Press, San Diego, 1988.
4. Green, D. W. and Perry, R. H. (eds.). Perry’s Chemical Engineers’ Handbook, 9th ed., Section 18: Crystallization. McGraw-Hill, New York, 2019.
5. U.S. Geological Survey. Mineral Commodity Summaries 2025 (nickel, copper, lithium). Reston, VA, 2025.
6. Crystal Growth & Design. American Chemical Society (journal; secondary nucleation and CSD control studies).
7. Hydrometallurgy. Elsevier (journal; crystallization and impurity rejection in hydrometallurgical streams).

Changelog

2026-10-23 — First published.

Related reading

• Stage Efficiency
• What Is Hydrometallurgy? Definition, Process, Applications