Crystallization theory for practice introduction

The result of any crystallization process is a crystalline matter with a certain crystal size distribution (CSD), certain crystal habit and purity. These properties very often are quality requirements and welldefined by market demands. Furthermore, the crystallization process itself requires a minimum CSD, as the suspension leaving the crystallizer still needs to be separated. The separation achieved and the product purity obtained improves with compact crystal habit, and coarser crystal size. Therefore, the quality and efficiency of crystallization as unit separation process is strongly dependent on crystal size and crystal habit. These properties also affect agglomeration during storage, dust generation and bulk density and thus influence the handling of the crystalline product downstream from the crystallizer: Crystallization theory addresses these critical aspects of the formation of crystalline solids. While crystal size and shape can be regulated by factors outside the crystallizers influence (impurities, solvent, etc), of the parameters that the crystallizer can control, the most important is the appropriate handling of supersaturation (the driving force of crystallization). Supersaturation can be produced by evaporation or cooling (Fig. 1). In case of flat solubility (low dependence with respect to temperature) the evaporation is the optimum choice, while solubility with strong temperature dependence is best suited to cooling crystallizers.

Figure 1 Crystallization process Higher supersaturation leads to greater crystal growth (Fig. 2). However, supersaturation also controls nucleation (Bo), and this limits the freedom of choice of a level of supersatu-ration in the crystallizer: spontaneous (primary) nucleation is caused by exceeding a critical degree of supersaturation and secondary nucleation can vary with the presence of crystals within the metastable solubility field. Secondary nucleation is fed by crystal-to-crystal and crystal-to-wall collisions, and therefore can be controlled by the mechanical energy input to the crystallizer.

Figure 2 Kinetics Usually, low nucleation will produce coarser crystals. Fig. 3 illustrates this principle. In a very simplified example, assuming a product crystal mass of 10 g, 10 particles of one-gram crystals are generated if there were only 10 nuclei at the start of crystallization. By comparison, if there were 104 nuclei present at the start of crystallization, they will result in an equal number of 1 mg crystals, totalling the same 10 g product mass. However, the degree of possible supersaturation is limited, and the result is that significantly more time is necessary to crystallize ten one-gram crystals than one-mg crystals. This relationship is indicated by the time arrows in Fig. 3. These principles lead to the following basic crystallizer design rules:

The supersaturation in crystallizers must be within the metastable range; High supersaturation is important for efficient crystal growth rates; Secondary nucleation should be minimized by limiting the mechanical energy input.

Figure 3 Influence of nucleation on crystal size Crystallizers adhering to these principles are the well-known basic types illustrated in Fig. 4.

Figure 4 Basic types of crystallizers

A basic crystallizer design feature is the way supersaturation is controlled and kept within the metastable range. The method is independent of the crystallization process. Fig. 5 illustrates this method for a vacuum-cooled crystallizer. This principle is also valid for va-cuum evaporation, surface cooling and reaction crystallization.

Figure 5 Supersaturation Control in a Vacuum Cooled Crystallizer The unit illustrated is a Forced-Circulation (FC) crystallizer which is equipped with an ex-ternal circulation loop. The simplified solubility graph shown includes the metastable zone (dashed line). The events occurring in one circulation loop are indicated by numbers in both the solubility diagram and the FC crystallizer sketch. The suspension (3) in the crys-tallizer is mixed with feed liquor (1). In cooling crystallization the feed is hotter and more concentrated than the circulated suspension. As a consequence, the resulting mixture (2) is more concentrated and hotter than the original suspension. The mixture is pumped to the Vapor/Liquid separator vessel (4) which is maintained at the boiling point of the suspension. There, the supersaturated solution temperature drops as it flashes to its the boiling point adiabatically. The boiling creates supersaturation C, which is represented by the line 3 - 4. This C is consumed by crystal growth, and the solution returns to its steady state. Read More The generated supersaturation (4) can be adjusted by varying the recirculation through the crystallizer. As more solution is recirculated, points (2) and (4) approach nearer to point (3). The coordinates of point (1), i.e. the feed mass flow and the crystallizer cooling (operating) temperature, define the production capacity of the system. The recirculation in the crystallizer is adapted to this production capacity, and the supersaturation is kept within the metastable range. The circulation flow, therefore, is a major design criterion. The circulation flow can be calculated by equation 1.

This is a limiting relationship for the production capacity of a crystallizer, and if exceeded, spontaneous
4

nucleation can occur. Relatively high circulation flows are usually necessary even for minor production capacities, because the typical metastable field has a range of a only few g/l: as an example, for a production of 1 t/h and a Delta C of 1 g/l, the circulation flow would be 1000 m3/h. Fig. 6 illustrates the relationship between desupersaturation rate and total crystal surface area. The generation of supersaturation drives crystal growth on the crystal surface area A1 while the suspension is circulated through the crystallizer. Looking at the recirculation as a series of differential elements, a typical saw-tooth function is evident. Because residual supersaturation is added to the fresh supersaturation generated in the element that follows, all supersaturation should preferably be consumed within a single loop. If there is a large amount of crystal surface area available (A2), the desupersaturation (measured in kg mass/m2 of crystal surface area) of the liquid is faster, resulting in lower residual supersaturation at the completion of the loop. Suspension densities (mass of crystals/mass of suspension) between 15 and 25 wt.-%, usually satisfy this requirement.

The mass deposition rate dm/dt varies exponentially with supersaturation Delta C, and linearly with crystal surface A (eqn. 2). The constant kg reflects the influence of temperature. The nucleation rate B0 (eqn. 3) is an exponential function of supersaturation, the specific energy input and the crystal mass mT .
5

As mentioned earlier, two of the main crystallizer design parameters are fixed, to ensure better control of crystallization. These are the C, which, in practice, is set to about half of the metastable range, and the crystal mass mT (or the crystal surface area A), set to val-ues between 15 and 25%wt magma density. The remaining design parameters that can influence product quality are the crystallization temperature T and the energy input E. With increasing temperature, the constant kg of eqn. 2 will usually increase. The energy input E affects inversely the secondary nucleation rate. Influencing the CSD and the average crystal size, therefore, remains possible by reducing E and operating the crystallizer at a higher temperature. Another important factor in determining crystal size is the crystal retention time T, as shown by eqn 4, which describes the linear growth rate G and eqn. 5, which describes the mean crystal size L.

Theoretically, longer retention times are necessary if larger crystals are desired; however, in practice, also the opposite can be observed. The reason is the mechanical attrition rate Ga which reflects the reduction of crystal size due to breakage. The effective linear crystal growth rate Geff , therefore, is the sum of Gk minus Ga. (eq. 6).

. Because attrition is increases with increasing crystal size, there is a critical particle size, where the effective crystal growth rate becomes zero (eq. 8). Therefore, under certain conditions, there exists a maximum achievable crystal size and longer retention times may lead to smaller crystal sizes. Note: see futhermore at US patent 3,303,870 : evaporator apparatus US patent 4,113,552 : elutriation led and method
Types of crystallizers

In designing a crystallizer, all the parameters discussed in Part 3 above need to be com-bined, so as to provide the optimal environment for crystal growth. Crystallizers with longer retention times operate at lower specific energy input, resulting in lower nucleation rates. Crystal-to-impeller impacts are the most effective source for nuclei production, as they usually generate at least a hundred-fold more nuclei than crystal-to-wall and crystal-to-crystal collisions. Types of crystallizers, therefore, differ from one another
6

mainly in the design and location (in clear liquor or in suspension) of the circulation pump (compare fig. 4). In summary, crystals grow larger by:

Maintaining proper supersaturation levels, Decreasing the energy input, Decreasing the mechanical stresses, Decreasing the attrition, and, at the same time, increasing the retention time

Also, conditions for crystallization improve at higher temperatures. The nucleation rate in a crystallizer is influenced not only by the specific energy input, but also by the mixing intensity. The specific by impeller pumps can be defined as:energy input

Here, denotes the specific density of the suspension or solution, g the acceleration due to gravity, H the delivery head, the efficiency, and Vcryst the filled volume of the crystallizer. In designing the circulation pump, the impeller diameter D and the number of revolutions n of the impeller per unit time can be varied at constant power N:

As a result, the tip speed and, hence, the secondary nucleation rate B can be altered at a constant rate of dissipated energy . Industrial-scale experience indicates that:

If crystallization is used as a simple precipitation process only, the production process may require only crystals that can be separated easily. Such crystal size can average about 0.2 mm, and this allows impeller pumps that operate at higher speeds. This is applied to FC-type crystallizers (fig. 4, left). For coarser crystals the need for reduced energy input is achieved in the so-called draft-tube-baffled (DTB) crystallizers (fig. 4, centre), which are equipped with an impeller pump located within a draft tube. These are circulators with large diameters which operate at low tip speeds that lead to low secondary nucleation. These impellers have diameters up to more than 1 m, resulting in far lower tip speed than conventional pumps located in an external circulation loop. Additionally, in the case of evaporative crystallization, the DTB has a further advantage over the FC type. In the DTB crystallizers
7

the heat exchanger which is responsible for most of the pressure loss (and thus the higher energy input) is not part of the slurry circulation. Thus, the central impeller pump consumes only about one third of energy than an equivalent FC unit, and thus provides improved crystal growth environment. The heat exchanger is installed in a second circuit operated only with clear liquor. The operation with clear liquor has an additional beneficial effect on crystal size. The DTB has an integral settling section (the Baffle), from which clear liquor can be with-drawn for the heater circuit. The clarification is such that the withdrawn liquid still contains the smaller crystal particle fractions. These fines are typically destroyed in the heater cir-cuit (heating increases the saturation point of the liquid), and this fines destruction reduces the number of small crystals in the crystallizer. In DTB type crystallizers, therefore, average crystal sizes between 0.5 up to 1.5 mm can be achieved. For production of very coarse crystals the circulation of the suspension is avoided alto-gether (to minimize the crystal attrition due to the pump impeller). In fluidized-bed crystal-lizers only clear liquor is circulated (fig. 4, right) and therefore the secondary nucleation rate is drastically reduced. Such crystallizers can produce crystals of 5-6 mm. Apart from the fundamental arrangements shown in Fig. (4), several other hybrid types of crystallizer designs are used in industry. Fig. 7 shows a collection of the simplest category

Figure 7 FC-type crystallizers The stirred-tank crystallizer (1) is selected for vacuum-cooling crystallization. The horizon-tal crystallizer (3) is also a vacuum-cooled crystallizer, and is characterized by several stages arranged in series within a single outer shell. Compared to the single-stage, vertical, agitated-tank crystallizer, a horizontal multi-stage crystallizer can result in lower investment costs. The draft-tube crystallizer (2) can operate with low-speed circulators, and therefore better control of the process supersaturation, and lower primary nucleation are possible. This type is used for products which have low growth rates and narrow metastable ranges. The forced-circulation crystallizer (4) is comparable in function to the drafttube crystallizer. The controlled recirculation of the suspension is carried out by means of an axial-flow
8

pump through an external heat exchanger. This type of crystallizer can also be used for vacuum-cooling crystallization (5). In FC crystallizers, the suspension density is determined by the mass flux. Higher densities (e. g., for achieving a faster desupersaturation rate) can be achieved in Fluidized Bed or DTB crystallizers if clarified solution is removed from the crystallizer separately (see fig. 4, centre and right). The lowintensity recirculation of the suspension, the removal of the fines, and the separate removal of clear liquor, as well as the classifying effect, contribute strongly to the production of coarser crystals. Thus, products such as ammonium sulphate, potassium chloride or urea are produced in this type of crystallizer, with average particle sizes of about 1.5 mm. Still coarser particles can be produced in fluidized-bed type crystallizers. Best known is the "Oslo"-type (fig. 4, right). At present, two variations of this crystallizer type exist. The original, also known by the name "Krystal has certain operating problems due to the for-mation of incrustations that quickly hinder circulation. The more recent "MESSO" variant was developed especially for crystallization of substances prone to form incrustations, and does not have these problems: By reversing the flow in the evaporation section, the superheated solution from the heat exchanger moves from the walls to the center, from where it is removed. This means that the solution supersaturation increases (as evapora-tion occurs) in a zone away from wall surfaces, and the formation of incrustations in the evaporation section is thus avoided. Operating times of several weeks can be achieved by such Fluidized Bed crystallizers.

Peripheral Components
Usually, a crystallization system is not complete at the crystallization step alone. The suspension in the crystallizer still has to be separated, the crystals have to be dried and packed. Vapors leaving the crystallizer need to be condensed, and the non-condensable gases need to be removed by vacuum pumps. Fig. 8 shows a simplified flow sheet of such a complete, vacuum evaporation crystallization plant.

In the illustration of Fig (8), one might envision any type of crystallizer both as a single unit or a multiple-effect. Heating methods alternative to live steam, such as mechanical or thermal vapor recompression can be utilized. In this example the vapors from the (last) crystallizer are condensed in a surface condenser. A direct-contact (barometric) con-denser could be used instead. The suspension (in the example) is not thick enough to feed the centrifuge: typically, FC units operate at 15 to 25%wt crystals, whereas centrifuges require feed suspension in the 40-60%wt range. Therefore, the suspension is pre-concentrated in a gravity thickener (shown) or with hydrocyclones. The clear liquor over-flow is recycled. Purge liquor to remove system impurities may be discharged from the system at this point. The thickener underflow is fed to the centrifuge. Depending on the CSD there is the choice among 4-5 types of separation devices: a filter for very small par-ticles, the decanter or peeler centrifuges for the small crystals, and the sieve-worm or the pusher centrifuge for larger crystals. The final drying of the product in most of the cases is achieved in rotary or vibrating fluidized-bed dryers. Cooling of the dry crystals can be done in separate cooler units or in an integral part of the dryer.

Process detail
In addition to the criteria mentioned earlier, the design of specific crystallization processes is influenced by several other factors. Some of the most important are mentioned below.

SURFACE-COOLING CRYSTALLIZATION
The surface-cooling process produces supersaturation directly on the heat exchanger surface. The supersaturation in the heat exchanger is the highest in the entire crystallizer. Incrustations on the heat transfer surface and eventual plugging of the tubes are the nor-mal consequences. This can be an acceptable situation for discontinuous operation, be-cause with each next batch the incrustations may be dissolved again. For continuous processes, however, the surface cooling is only an option if the low operating temperature required in the crystallizer makes vacuum cooling crystallization impractical. If a conti-nuous crystallizer must employ surface-cooling, especially large heat exchanger surface area is supplied, in an effort to increase the operating cycle.

VACUUM-COOLING CRYSTALLIZATION
Vacuum-cooled crystallization is the preferred cooling crystallization method for continuous operation. Because cooling is generated by adiabatic expansion of the solvent, and the condensing of the vaporized solvent is done in a separate heat exchanger, scaling of cooling surfaces is not experienced. Vacuum cooling becomes uneconomical (or imprac-tical) only if operation at very low temperatures is required.

EVAPORATION CRYSTALLIZATION
The evaporative crystallization is generally a vacuum process, much like vacuum-cooled crystallization. The difference is that this process is independent of the concentration and temperature of the feed solution. External heat can be added to the system and the con-centration of mother liquor can be adjusted by evaporation. Like vacuum-cooled crystalli-zation, there are no special encrustation problems in evaporative crystallization. Operating difficulties may arise in the case of concentration of inversely soluble substances, like some sulfates and carbonates. In such cases the same encrustation model exists as in surface-cooled crystallization. High suspension velocities in the heater tubes and high suspension density (to increase the desupersaturation rate) and can improve the operating cycle. Multiple-effect
10

evaporative crystallization plants are supplied in cases where low energy consumption is especially important.

Example - Crystallization of Sodium Chloride

In this example, we examine a case where solar evaporation is combined with a modern salt crystallization system to improve product salt quality. Fig. 9 shows the flowsheet of such a crystallization unit, which is operated with feed liquor concentrated in a solar pond. Along with standard FC type crystallizers, there is installed one fluidized-bed crystallizer for producing a coarse product as well. To improve its energy efficiency, the plant is operated as a multiple-effect evaporation unit. The Messo/Oslo crystallizer is the first effect, ensuring optimum conditions for crystal growth (high temperature). The plant is fed with concentrated liquor from the solar pond into a tank from which the Messo/Oslo crystallizer is fed. This tank has unsaturated liquor, ensuring crystal-free feed (feeding a crystallizer with suspended fines usually reduces its product size). Solution not used for the Mes-so/Oslo unit overflows into a second tank from where the FC crystallizers are fed. The slurry in the FC crystallizer train is advanced from effect to effect and from the last effect to a slurry tank.

Figure 9 Multiple-effect evaporation crystallization for NaCl From there the slurry is fed to the solids separation section, which consists of a hydro cyclone, a washing thickener and a centrifuge. In the washing thickener, the impure mother liquor is exchanged with fresh feed liquor, which has lower impurity levels. The slurry from the washing thickener is fed to the centrifuge where the salt is dewatered and washed with clean water. The wet centrifuge cake is then dried and the crystals sent to the packaging section. The dewatering of the granular product of the Messo/Oslo unit is made separately in a special centrifuge, ensuring lower breakage of the crystals.
11

A plant of this type was erected for the treatment of 60 t/h concentrated liquor from a solar pond on the Mediterranean Sea coast. It is producing 2.5 t/h of granular salt (average size of about 2 mm) and 6.5 t/h of standard vacuum salt. It consumes 11 t/h of steam to evaporate 34 t/h of water. By storing extra concentrated solar pond liquor over the summer period, the plant is able to be operated at full rate for the entire year.

12