Crystallization and Isolate Production

What this operation is — and explicitly what it is not

Crystallisation is a purification. A molecule already present in the feed is persuaded to leave solution as an ordered solid while everything else stays dissolved, and because a growing crystal lattice accepts the molecule that fits it and rejects the ones that do not, the solid that forms is purer than the solution it came from. That is the entire mechanism, and it is why crystallisation is the most powerful single purification step available to a cannabinoid processor: a good distillate at 90 percent CBD becomes a 99-plus percent solid in one operation. Nothing is converted. No molecule is made. Every gram of CBD in the isolate was CBD in the distillate, and every gram of CBD in the distillate was CBD or CBDA in the plant. No step on this page is a conversion, none of it is written so that it could become one, and the acid-catalysed chemistry that turns CBD into something else is not on this shelf at all — see extraction-methods for where that line is drawn and cannabinoids/isomers for the structural treatment. The isolate produced here is a purification product, and that is what makes it the cleanest, most specifiable and most boring material in the industry, which is exactly its commercial virtue.

Sources: Mullin JW 2001* · Myerson AS 2019*

Solubility, supersaturation and the driving force

Every solute has a solubility curve in every solvent: the maximum concentration that will stay in solution at a given temperature, which for nearly all organic solutes rises with temperature. Dissolve a solute at a high temperature to near its solubility limit and then cool the solution, and at some point the concentration present exceeds the concentration that can be held. That excess is supersaturation, and it is the thermodynamic driving force for crystallisation — the system now wants to deposit solid, and the amount by which it is supersaturated sets how hard it wants to. Between the solubility curve and the point of spontaneous nucleation there is a region, the metastable zone, where a solution is supersaturated but will not start crystallising on its own; it needs a surface to start on. That zone is where all controlled crystallisation is performed, because a solution held inside it deposits solid onto crystals that are already there, growing them in an orderly way, while a solution pushed past the metastable limit nucleates spontaneously everywhere at once and produces a mass of tiny, impure, solvent-trapping crystals or, worse, oils out. The craft of crystallisation is therefore the craft of generating supersaturation slowly enough and providing surfaces deliberately enough that growth beats nucleation.

• Supersaturation is the driving force; too little and nothing happens, too much and everything happens badly.
• The metastable zone is where controlled crystallisation lives.
• Growth on existing crystals gives purity; spontaneous nucleation everywhere gives fines and occlusion.

Sources: Mullin JW 2001* · Myerson AS 2019*

Solvent systems industry practice, not published data

The solvent has to dissolve the target well when warm and poorly when cold, dissolve the impurities well at all temperatures so they stay in the mother liquor, be removable to a residual-solvent specification, and be acceptable under the residual-solvent framework the product will be tested against. For CBD the solvents described in the trade literature are pentane, heptane, and pentane/ethanol mixtures. Pentane is a light non-polar alkane with a boiling point near 36 °C, which makes it very easy to remove from a cake and gives a steep solubility differential across a modest temperature range; the cost is extreme volatility and flammability. Heptane, boiling near 98 °C, is easier and safer to handle, dissolves a little more at a given temperature, and is harder to drive out of a crystal. A mixed system — typically a non-polar bulk with a small proportion of ethanol as a polarity modifier — is used to tune the solubility curve, to keep more of the impurity load in solution, and to slow the approach to supersaturation so that growth is controlled. All three of pentane, heptane and ethanol are Class 3 solvents under ICH Q3C, the lowest-concern class, which is a deliberate choice: the solvent used in the final purification step is the one most likely to end up in the product, so it should be the one with the most generous permitted exposure. Antisolvent crystallisation — dissolving in one solvent and adding a second in which the target is poorly soluble — is the other route to supersaturation and it is used in the same way, with the same seeding and rate discipline, except that the control variable is the addition rate rather than the cooling rate.

• Requirement: dissolves the target warm, not cold; dissolves the impurities always; comes out to specification.
• Pentane: steep differential, trivially easy to dry, extremely flammable and volatile.
• Heptane: safer handling, higher boiling, harder to drive out of the crystal.
• A small ethanol fraction tunes the curve and slows the approach to supersaturation.
• Class 3 solvents throughout — the final-step solvent is the one that ends up in the product.

Sources: Composite: extraction-equipment vendor documentation 2026* · International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use 2021* · Armarego WLF 2017* · Mullin JW 2001*

The operation, in order

The sequence is fixed and each step has a purpose that the next step depends on. First, start from high-purity distillate: crystallisation rejects impurities into the mother liquor, but a feed with a heavy impurity load gives a mother liquor so concentrated in impurity that the crystals grow through it and occlude it, so a second or third recrystallisation becomes necessary and the yield collapses. Second, dissolve the distillate completely in the warm solvent, with gentle heat and agitation, at a concentration close to but below saturation at the dissolution temperature; undissolved material acts as unintended nucleation sites and starts the crystallisation at a point you did not choose. Third, if the feed colour or impurity load warrants it, treat the hot solution — activated carbon or bentonite, then a hot filtration through a fine medium — because it is far easier to polish a solution than a solid, and a clarifying filtration also removes stray particulates that would otherwise nucleate. Fourth, cool into the metastable zone and seed. Fifth, cool further on a controlled profile while the crystals grow, with gentle agitation. Sixth, hold at the final temperature long enough for growth to finish and the mother liquor to reach equilibrium — cutting this short leaves product in solution. Seventh, filter. Eighth, wash the cake with cold solvent. Ninth, dry under vacuum to specification. Tenth, work up the mother liquor, which is a product stream in its own right.

• Feed quality first — crystallisation refines, it does not rescue.
• Full dissolution before cooling; undissolved solid is an uncontrolled seed.
• Clarify and polish hot, in solution, not later as a solid.
• Cool into the metastable zone, then seed, then continue cooling.
• Hold at final temperature — an early filtration leaves product in the liquor.

Sources: Mullin JW 2001* · Myerson AS 2019* · Composite: extraction-equipment vendor documentation 2026*

Seeding, and oiling out

Seeding is the single highest-leverage control in the whole operation. A small quantity of the existing pure crystal — commonly a fraction of a percent up to a few percent of the expected yield, added as a dry powder or as a slurry in cold solvent — is introduced when the solution has cooled into the metastable zone. What it does is take the decision about when and where nucleation happens out of the hands of chance: instead of the solution waiting until it is deeply supersaturated and then nucleating spontaneously in a burst, it begins depositing solid immediately onto the surfaces you provided, at a low supersaturation, growing crystals of known form. The consequences are a predictable batch time, a reproducible crystal size distribution, far less occluded solvent and impurity, and a much better filtration. Seeding also selects the crystal form, which matters wherever a compound has more than one, because the seed you add is the form that grows. The failure mode that seeding most reliably prevents is oiling out, and oiling out is the characteristic disaster of cannabinoid crystallisation. Because cannabinoids are viscous, low-melting, highly lipophilic molecules, a heavily supersaturated cannabinoid solution frequently separates as a second liquid phase — a sticky oil — rather than as a solid. Once that happens, the oil phase is a concentrated soup of the target together with every impurity that was in solution, it does not purify anything, it will not filter, it coats the vessel and the agitator, and the usual outcome is that the batch has to be redissolved and started again. The causes are always the same short list: cooling too fast, going too deep before nucleation begins, too high an initial concentration, too much impurity in the feed (impurities depress the melting point of the target and make the liquid phase more stable), and no seed. The remedy is the mirror image: seed early at low supersaturation, cool slowly, start from clean feed, and if oil appears, warm back up to redissolve it and re-approach the zone more slowly.

• Seed inside the metastable zone, at low supersaturation, with the crystal form you want.
• Seeding buys predictable batch time, reproducible size, less occlusion, better filtration.
• Oiling out is a second liquid phase, not a crystal — it purifies nothing and filters not at all.
• Causes of oiling out: fast cooling, deep supersaturation before nucleation, too concentrated, dirty feed, no seed.
• If it oils out, warm back to full solution and re-approach more slowly.

Sources: Mullin JW 2001* · Myerson AS 2019* · Composite: extraction-equipment vendor documentation 2026*

Slow cool against crash: crystal habit and occluded solvent

The cooling profile decides what kind of solid you get. A slow, controlled cool — degrees per hour rather than degrees per minute, often with a programmed ramp and a hold — keeps supersaturation low throughout, so growth dominates over nucleation and the crystals grow large, well-formed and few. Large well-formed crystals have a low surface-area-to-mass ratio, which means little mother liquor adheres to them; they have regular faces that reject foreign molecules as they grow, which means high intrinsic purity; and they form an open, permeable filter cake that drains and washes readily. A crash cool — plunging the solution to low temperature quickly — generates high supersaturation immediately, nucleation happens everywhere at once, and the result is a mass of fine crystals or an amorphous solid. Fines have an enormous surface area holding mother liquor, they grow so fast that they entrap pockets of solution inside the crystal as inclusions, and inclusions are the reason a crashed batch can fail a residual-solvent panel no matter how long it is dried: the solvent is not on the surface where vacuum can reach it, it is inside the solid. Fines also form a dense, compressible cake that blinds the filter and cannot be washed effectively. The industrial rule of thumb, which is worth internalising, is that time spent on the cooling profile is bought back three times over in filtration rate, wash efficiency and drying time. Agitation belongs in the same discussion: gentle, continuous agitation keeps the solution uniform so that no local region becomes deeply supersaturated, keeps crystals suspended so they grow evenly on all faces rather than sitting on the bottom, and improves heat transfer to the jacket; but vigorous agitation causes secondary nucleation by attrition, breaking crystals and creating fines, and can shear a growing crystal population into exactly the fine mass you were avoiding. Gentle and continuous is the target.

Sources: Mullin JW 2001* · Myerson AS 2019*

Filtration, washing the cake, and drying to specification

Filtration separates the crystals from the mother liquor, and it is done cold, because the solid is only insoluble because it is cold — exactly the same rule as in winterization, and violated for exactly the same reason. A Buchner funnel or filter plate under vacuum at laboratory scale, a jacketed Nutsche filter, filter dryer or centrifuge at production scale; all of it pre-chilled. What comes out of the filter is not pure yet, because every crystal is wetted with a film of mother liquor that contains all the impurity the crystallisation rejected. Washing the cake is therefore not a finishing touch, it is the step that delivers the purity: a small volume of clean, cold solvent is displaced through the cake to push the mother liquor out and replace it with pure solvent. Cold matters because warm wash solvent dissolves the product it is supposed to be rinsing. Small volumes matter, applied as a displacement wash that flows evenly through the bed rather than as a flood that channels down one side; two or three small washes beat one large one. Then drying: the washed cake is wet with pure solvent, and that solvent is removed under vacuum, at a modest temperature, with a large surface area and enough time for diffusion out of the solid. This is the step that has to meet the residual-solvent specification, and it is a diffusion problem, so it is governed by cake thickness, temperature, vacuum depth and time — the same physics as the vacuum-oven discussion on rotary-evaporation. The specification comes from ICH Q3C and, for a laboratory running the panel, USP General Chapter 467; formulation/residual-solvent has the limits and the arithmetic. One warning that belongs here specifically: a crashed, fine, inclusion-bearing cake cannot be dried to specification at any reasonable temperature, because the solvent is inside the crystals and not on their surface. If a batch will not dry down, the problem is upstream in the cooling profile, and the fix is to redissolve and recrystallise properly rather than to dry it harder.

• Filter cold, with pre-chilled funnel, receiver and wash solvent.
• Wash with small volumes of clean cold solvent as a displacement, not a flood.
• The wash, not the filtration, is what delivers the purity.
• Dry under vacuum at modest temperature with a thin cake — it is a diffusion problem.
• A cake that will not dry to specification has inclusions; recrystallise, do not bake.

Sources: Mullin JW 2001* · Myerson AS 2019* · International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use 2021* · United States Pharmacopeia 2023* · Composite: extraction-equipment vendor documentation 2026*

The mother liquor is a product, not a waste industry practice, not published data

The mother liquor is everything the crystal rejected, dissolved in the crystallisation solvent, plus the fraction of the target that remained in solution at the final temperature. In a CBD isolate operation that means it carries the minor cannabinoids — CBC, CBG, CBDV, the varin series, whatever THC the feed contained within its limit — together with residual terpenoids, oxidation products, colour bodies and a substantial quantity of CBD itself, because no crystallisation is complete and the solubility at the final temperature sets the floor on recovery. Throwing it away discards a meaningful fraction of the batch value twice over: the CBD left in solution, and the minor cannabinoid fraction that is worth considerably more per gram than the isolate is. The standard workup is to recover the solvent by evaporation, then either recrystallise the residue to pull a second crop of isolate (lower purity, often recycled into the next batch rather than sold as isolate), or send the residue to chromatography, which is where minor cannabinoids are actually isolated — and mother liquor from an isolate plant is one of the better feedstocks for that, because the major component has already been largely removed. That is the economic connection between this page and chromatography, and it is a real one: the reason a CBD isolate operation can afford a preparative chromatography capability is often the mother liquor stream.

• Carries the minor cannabinoids, the colour bodies, and a real quantity of unrecovered CBD.
• Recover the solvent, then either take a second crop or send it to chromatography.
• It is the natural feedstock for minor-cannabinoid isolation because the major component is gone.

Sources: Composite: extraction-equipment vendor documentation 2026* · Hazekamp A 2004*

Achievable purity, and the yield conversation contested industry practice, not published data

The normal commercial specification for CBD isolate is 98 to 99.9 percent CBD by mass, with 99 percent and above routine from good distillate and a single well-run crystallisation, and 99.5 percent and above achievable with a careful wash or a recrystallisation. That purity is not difficult in the sense of requiring exotic equipment; it is difficult in the sense of requiring feed quality, a controlled cooling profile and an honest wash. The yield conversation is separate and it is where operations deceive themselves. Single-pass crystal yield is limited by the solubility of the target in the solvent at the final temperature, so a fraction of the CBD always stays in the liquor; deeper final cooling and a leaner solvent ratio raise the yield and simultaneously raise the risk of fines and oiling out, so the yield-versus-purity trade is real and it is usually resolved by taking a good first crop at high purity and working the liquor up separately rather than by squeezing the first crop. A processor should report isolate purity and crystal yield and liquor cannabinoid content together, because a 99.9 percent isolate at 50 percent yield with an unworked liquor is a worse outcome than a 99.2 percent isolate at 80 percent yield with the liquor recovered.

Contested — caveat. The 98-99.9 percent range is the commercial specification band as sold and as tested; it is not a claim about any particular process achieving it. Reported yields vary widely with solvent, ratio, final temperature and feed purity, and most published figures come from vendors rather than from controlled studies.

Sources: Composite: extraction-equipment vendor documentation 2026* · Mullin JW 2001*

Why CBD crystallises readily and most cannabinoids do not contested

CBD is a mild oddity in its own chemical family: it is a crystalline solid at room temperature with a melting point in the region of 65 °C, whereas Δ9-THC, CBG, CBC and most of the other neutral cannabinoids are viscous oils or low-melting glasses that show no inclination to crystallise at all. The structural reason is molecular shape and packing. CBD is an open, comparatively symmetric, relatively rigid diphenol — two hydroxyl groups on an unfused resorcinol ring, a terpene ring that is not closed onto it — and that geometry lets CBD molecules pack into an ordered lattice with hydrogen bonding between the phenolic hydroxyls holding it together. THC, by contrast, has its pyran ring closed onto the aromatic ring, leaving a single free hydroxyl, a fused tricyclic shape with a stereocentre and an awkward, non-planar profile; it packs poorly, hydrogen-bonds less extensively, and prefers to remain a supercooled liquid. The consequences for a processor are practical and important. First, isolate production as a routine operation is essentially a CBD phenomenon among the neutral cannabinoids: you cannot simply apply this page to CBG or CBC and expect a crop, and attempts to crystallise most minors run into oiling out immediately because for those molecules the liquid phase is genuinely the stable one at the temperatures involved. Second, the route to isolating minor cannabinoids is therefore chromatographic rather than crystallographic, which is why preparative chromatography rather than crystallisation is the tool described for minors and for reference-grade material. Third, the minors that can be crystallised tend to be ones that also pack well — CBG is reported as crystallisable with more difficulty, and the acid forms are a separate story entirely. And fourth, this is why an isolate market exists for CBD at commodity prices while minor cannabinoids are priced per gram like reference standards: the purification economics are completely different molecules apart.

• CBD: open, symmetric, two free phenols, hydrogen-bonded lattice, melts near 65 °C, crystallises readily.
• THC and most neutral minors: fused, hindered, one free hydroxyl, poor packing, stay as oils or glasses.
• Therefore isolate production is largely a CBD operation, and minors go to chromatography instead.
• This is the structural reason minor cannabinoids cost what they cost.

Contested — caveat. Melting points reported for CBD vary between roughly 62 and 68 °C across sources depending on polymorph and purity, and the crystallisability of individual minor cannabinoids is documented unevenly — much of what circulates about CBG and CBN crystallisation is trade experience rather than published crystallographic work. The structural explanation given here is the standard reading of the molecular geometry, not a measured result.

Sources: Mullin JW 2001* · Adams R 1940* · Gaoni Y 1964* · Composite: extraction-equipment vendor documentation 2026*

THCA diamonds: the same physics, a different legal category industry practice, not published data

The crystals sold as diamonds are THCA, the acid form, and their production is the same physics described on this page applied to a different molecule. THCA is a crystalline solid, more readily crystallised than neutral THC precisely because the carboxylic acid group gives it strong hydrogen-bonding capability and a shape that packs, which is the same structural argument as for CBD. The operations are recognisably identical: a supersaturated solution of THCA in a solvent or in its own terpene fraction, a controlled approach to supersaturation, nucleation and slow growth over days to weeks, then separation of the crystals from the liquid phase — which in the diamonds-and-sauce presentation is deliberately kept and sold alongside the crystals as the terpene-rich mother liquor rather than worked up. It is also the clearest illustration of why decarboxylation belongs where you put it and not elsewhere: THCA diamonds exist because the material was never decarboxylated, and heating them converts them to Δ9-THC, which is what happens when they are consumed. The only difference that matters between a THCA crystallisation and a CBD crystallisation is regulatory: CBD isolate from compliant hemp is a hemp product in most jurisdictions, and THCA crystal is a controlled-substance product in most, testing at or near the theoretical maximum total THC. The chemistry, the equipment and the craft are the same. Processing decisions about which one an operation can make are licensing decisions, not technical ones.

Sources: Composite: extraction-equipment vendor documentation 2026* · United States Department of Agriculture 2021* · Wang M 2016*

Safety, as part of the operation

Crystallisation handles the most hazardous solvents on this shelf at the coldest temperatures, so the discipline is specific. Pentane boils at about 36 °C and has a flash point near minus 49 °C, which means its vapour is ignitable at any temperature you will ever encounter, it boils at slightly above room temperature so a warm room pressurises a closed container, and its vapour is denser than air and travels. Pentane work requires ventilation designed for a heavier-than-air vapour, a genuinely ignition-free environment with appropriately rated electrical fittings, bonding and grounding of every metal vessel and every transfer (static from pouring a non-conductive solvent is a documented ignition source, and pentane is about as non-conductive as solvents get), and no open transfers of any volume. Heptane is less volatile but still a flammable liquid under the code and gets the same handling in kind if not in degree. Cold: crystallisers and cold rooms run from minus 20 °C to minus 80 °C, and cryogenic-rated gloves and no bare-skin contact with cold metal are required, with the same caution as on winterization that cold burns accumulate without the warning that heat gives. Warming a sealed cold vessel is a pressure hazard with a low-boiling solvent inside and must never be done — a vessel taken out of a minus 40 °C freezer and closed will pressurise as it warms. Dissolution at elevated temperature is a hot-flammable-solvent operation, so the heat source is a controlled bath with no open element and no flame anywhere, the vessel is refluxed or closed to a condenser rather than open to the room, and a solvent whose boiling point is near the dissolution temperature is under enough vapour pressure to push a stopper out. Vacuum filtration and vacuum drying bring the evacuated-glassware inspection rules from short-path-distillation. And two specifics: filter cakes and dried isolate are fine organic powders which can carry a dust-explosion risk in bulk handling and are a nuisance respiratory exposure at any scale, so handle them with extraction and containment; and no peroxide-forming ether — diethyl ether, diisopropyl ether, THF — should be used as a crystallisation solvent and then evaporated to dryness, because concentrating peroxides into a solid residue is the worst version of that hazard.

• Pentane is ignitable at every ordinary temperature, boils near 36 °C, and its vapour sinks and travels.
• Bond and ground everything; non-conductive solvents generate static on transfer.
• Never seal a cold vessel containing a low-boiling solvent and let it warm.
• Hot dissolution: controlled bath, no flame, no open element, condenser rather than an open vessel.
• Cryogenic gloves for freezer and cold-room work; no bare skin on cold metal.
• Evacuated glassware rules as on short-path work: inspect for star cracks, only rated vessels.
• Dried isolate is a fine organic dust — containment and extraction, and mind the dust-explosion question in bulk.
• No peroxide-forming ether evaporated to dryness, ever.

Sources: National Research Council (US) 2011* · Kelly RJ 1996* · National Fire Protection Association 2024* · Occupational Safety 2024* · Armarego WLF 2017*