Crystallization processes/techniques can be distinguished by the manner in which the supersaturation is created. The most frequently applied types of crystallization are:

• Evaporative crystallization
• Cooling crystallization from solution or the melt
• Reactive crystallization or precipitation.

The choice for a certain method depends on the properties of the compound to be crystallized, the feed and the thermodynamics of the system. For compounds with a low solubility the standard option is reactive crystallization/precipitation. As a rule of thumb, a compound has a low solubility when the equilibrium concentration in solution is lower than 10 g/kg. For compounds with a solubility higher than 200 g/kg, a choice has to be made between evaporative and cooling crystallization. Evaporative crystallization is typically the preferred option when the solubility is hardly dependent on T, which implies that cooling the saturated solution will not result in a significant amount of crystals. Kramer and Bruinsma[2] mentioned that evaporative crystallization is usually the preferred method when the amount of dissolved product decreases with less than 5 g/kg per °C (or dC/dT < 0.005 g/g.°C).

Evaporative Crystallization

In evaporative crystallization, the crystallization is resulting from the evaporation of the solvent. So, this process creates a vapor and a suspension of crystals in mother liquor. The heat of evaporation that has been added is in principle captured in the vapor stream. In principle this energy can be recovered, but the energy will be available at a lower T. The mother liquor will still contain the equilibrium concentration of product. It is possible to harvest this residual amount of product by recycling the mother liquor to the feed. The possibility to recycle the mother liquor will be restricted by the impurities. At a certain moment the concentration of impurities will become so high that they can influence the crystallization and/or the product purity. If that is the case, the mother liquor stream cannot be recycled any longer and the remaining mother liquor has to be discharged via a bleed or purge stream.

Cooling Crystallization

Cooling crystallization is attractive when the solubility of the product increases significantly with increasing temperature. In those cases, cooling crystallization is usually more energy friendly than evaporative crystallization. In a cooling crystallization process the feed is cooled in a heat exchanger, which can be situated inside the crystallizer or an external loop. The wall of the crystallizer can be used as internal heat exchanger, but the heat exchanger can also be integrated in the crystallizer in the form of cooled tubes or plates. Crystallization can take place when the liquid is cooled to a temperature below the equilibrium solubility. The lowest temperature in the system is at the surface of the heat exchanger. Therefore the cooling needs to be done carefully to prevent nucleation on the cold surface of the heat exchanger, which will lead to encrustation. Typically measures to prevent this unwanted phenomenon is to reduce the temperature difference between the coolant and the crystalizing solution, to increase the liquid velocity along the surface of the heat exchanger to level out the temperature difference over the length of the heat exchanger or to use a scraper to keep the surface of the heat exchanger free of solids. Alternative methods of cooling that not require a heat exchanger are flash cooling which involves (partial) evaporation of the solvent or direct cooling via insertion of a cold gas or coolant.

Melt crystallization can be regarded as a special form of cooling crystallization process. The main difference with cooling crystallization from solution is the absence of solvents, which implies that most melt crystallization processes are operated close to the melting point of the pure product. The feed for a melt crystallization process is an impure melt. Cooling this melt below the equilibrium temperature will typically result in the formation of a solid phase that is purer than the feed, while the impurities prefer to stay in the impure mother liquor. Section 2.2 will give more background information on melt crystallization, which is also known as fractional crystallization.

Precipitation

In precipitation the supersaturation is created by the mixing of two streams. The most common forms of precipitation are: (1) reactive crystallization, (2) pH shift crystallization and (3) anti-solvent (or extractive) crystallization. In reactive crystallization, the solid phase is formed due to a reaction between components that were originally present in the (two) feed streams, for instance A (aq) + B (aq) ® AB (s). pH-shift crystallization often makes use of the change in the chemical state/charge of ionic species with pH. Well known examples of ions that change state with pH are for instance carbonate and phosphate. In a pH shift crystallization, typically a reaction takes place between an acid and a base. Such a reaction is usually very fast and can be quite exothermic, especially when the reactants are present in relatively high concentrations. In an anti-solvent crystallization the anti-solvent is usually mixed with a (concentrated) solution. The anti-solvent is typically well-mixable with the solvent, the crystalizing product has a low(er) solubility in the anti-solvent and for economic and environmental reasons it should be possible to recover the anti-solvent. Precipitation can be fast or slow. A precipitation is called fast of one of the underlying processes in crystallization, e.g. the nucleation, is fast compared to the mixing. A fast conversion step does not necessarily imply that the residence time in all precipitation processes are short. Long(er) residence times are for instance required when a coarser product or a different polymorph needs to be made than in the primary conversion/precipitation step. The size of a precipitate can be increased by ripening or ageing of the suspension. Ripening typically runs at relatively low supersaturation, which prolongs the residence time. Precipitation is also known for its characteristic that metastable intermediates can be formed around the inlets of the feed streams. These intermediates can be amorphous or (pseudo)polymorphs of the desired, thermodynamically stable end product. The transition from the metastable phase to the thermodynamically stable product requires a liquid mediated recrystallization step, which runs at a relatively low supersaturation and thus requiring relatively long residence times. An advantage of such a slow re-crystallization is that it is typically accompanied by a strong increase of the purity of the product.

Fractional Crystallization

Fractional crystallization is a term that is used to describe a process where repeated crystallization steps are used to increase the purity of the product and/or to increase the yield of the process. Applications can be found in metal (metal refining/zone melting), oil & gas (e.g. oil dewaxing), food (e.g. palm oil fractionation and freeze concentration) and in chemistry/chemical industries (e.g. paraffin wax de-oiling or the ultra-purification of chemicals). In a fractional crystallization process the crystals formed in the first stage are separated from the mother liquor with devices like filters, centrifuges or wash columns and remolten to be used as feed in a second crystallization stage. Alternatively, the crystals are not remolten, but re-suspended in a liquid with a higher product concentration at a higher operating temperature. After re-suspending, the crystals may be allowed to grow if that is needed/beneficial. The choice between re-melting or re-suspending depends on the quality of the crystals. Re-suspension is the preferred option when the crystals already have more or less the right internal purity, shape and size. If one or more of these parameters significantly differs from the specifications of the end product re-melting/re-crystallization is usually the best option. This second crystallization stage will operate at a higher temperature than the first stage due to the higher purity of the feed in the second stage. If the mother liquor still contains a significant amount of product, it may also be used as feed for an additional crystallization stage. This stage will then be operated at a lower temperature than the first crystallization, because the mother liquor has a lower concentration of product than the feed of the first crystallization stage. In the described process the solids and liquid move in different directions: the crystals go up in temperature while the mother liquor faces a decrease in temperature. Therefore, this staged crystallization processes is also known as a counter-current cascade. A well-known example of a process deploying the counter-current cascade principle is the TSK-CCCC process[3]. The crystal transport/separation in this process is done by cyclones, while the liquid transport is achieved by gravity.

In principle, fractional crystallization can refer to crystallization from solution as well as to crystallization from the melt. In practice, it is used more in relation with melt crystallization, which probably explains why the term fractional crystallization is sometimes used as a synonym for melt crystallization. Melt crystallization is particularly interesting for the separation and ultra-purification of organic chemicals due to their favourable melting points. Ulrich and Glade[4] found out that 71% of the organic chemicals in the Merck catalogue from 1991 had a melting point between 0 and 200°C.

Fractional/melt crystallization is a separation method that can be an attractive alternative for distillation for the separation of organic mixtures. Distillation is by far the most frequently used separation process in chemical industry, but there are also distinct disadvantages such as thermal degradation of product, excessive consumption of energy, poor performance in case of azeotropes. In 1989, TNO [5]compared in a desk study the energy consumption of conceptual separation processes using distillation with that of melt crystallization for the purification of 5 bulk chemicals: benzene, styrene, phenol, caprolactam and dimethyl terephthalate (DMT). For all products melt crystallization required much less energy than distillation: 27% less for benzene; 32% less for styrene, 63% less for phenol; 50% less for caprolactam and an impressive 96% less for DMT. Moreover, distillation requires large reflux ratios and a large number of stages when the feed contains compounds with similar boiling points/volatility. These disadvantages can be overcome using melt crystallization, making use of the fact that for organic chemicals the melting enthalpy is typically only 1/3 of the enthalpy of vaporization, its high selectivity and its relatively low operating temperature. Melt crystallization avoids the use of solvents, which implies that volumes of process streams can be kept small and that no processes/equipment are needed for recovery and recycling of the solvents.

There are two different methods for melt crystallization which differ in the way that the solid crystals are growing. In a layer melt crystallization the crystals grow as a layer on a cold wall of a heat exchanger. The liquid phase is either stagnant in static layer crystallization or mixed (mechanically or as a falling film) in dynamic layer crystallization. The heat exchanger can have different shapes like plates or tubes. When the layer has reached the required thickness, the remaining mother liquor is drained and the layer is molten off with or without layer purification steps like sweating and/or washing. In suspension melt crystallization the crystals grow freely suspended in the liquid phase. The liquid/suspension is often cooled with a heat exchanger, but in this case it is designed and operated to prevent that the solid is forming on the wall. Although, layer crystallization and suspension crystallization are based on the same physical principles there are characteristic difference between the options. Layer crystallization is typically a batch process and high growth rates in the order of 10-5-10-6 m/s are required to compensate for the relatively small surface area on which the growth takes which is in the order of 50-100 m2/m3. Suspension crystallization is typically a continuous process and the large surface area of the suspended crystals, which can be as high as 5000-10000 m2/m3, implies that the required production rate can be reached at relatively low growth rates in the order of 10-7-10-8 m/s. A consequence of a low growth rate is that the kinetic incorporation of impurities in the crystalline phase is strongly suppressed, which explains why the solid material (crystals) made in a suspension growth process is significantly purer than the solid layer in the layer growth process. Therefore a layer growth process will require more stages than a suspension growth process to compensate for the lower purification efficiency per stage. The specific energy consumption of layer grow is significantly higher than for suspension growth due to the higher number of stages and the subsequent heating and cooling cycles inherent to a batch process. In an EU-funded project the purification of caprolactam and naphthalene was investigated for both the layer and the suspension melt crystallization. For the same separation target, the suspension growth option required 66% less energy than the layer growth process for the purification of naphthalene, while for caprolactam the difference was even larger, namely a 75% lower energy consumption for the suspension-based process

The main requirement for deploying the larger purification and energy efficiency of suspension melt crystallization is that it is possible to grow crystals of sufficient size and suitable shape for the solid-liquid separation that follows the crystallization step. A Hydraulic Wash Column, that combines continuous solid-liquid separation with a highly efficient counter-current washing step at throughputs as high as 5-20 tonnes of product per hour per m2 wash column, requires crystals with an average crystal size of 50-100 µm. Wash column operation is easy with hard crystals with a reasonable 3-dimensional shape (cubes, spheres, needles, …), but can be challenging for soft and/or plate-like, or hair-shaped crystals as these properties strongly reduce the porosity of the crystal bed. Frequently mentioned points of attention for suspension melt crystallization is the handling of slurries and scale up. Of course slurry handling requires more attention than liquid handling, but there are many industrial processes and products where slurry handling is done without any problem at small (e.g. pharmaceuticals) and very large scale (e.g. sugar, soda ash). Scale up of scraped heat exchanger suspension crystallizers and the Hydraulic Wash Column is not a problem: the tubular design of the Armstrong scraped Surface Continuous Crystallizer has a beneficial surface to volume ratio and the modular design allows for an easy expansion with growth in demand. The scale up principle of the Hydraulic Wash Column, where the number of filter tubes is increased proportional with the cross-sectional surface area is straightforward and proven up to a diameter of 55 cm. The production capacity of this column was more than 5 tonne per hour. A further scale up to 1.13 m, resulting in a HWC with a cross-sectional surface area of 1 m2, is considered feasible, which will push the production capacity up to 20 tonne per hour for a single HWC.