CHINA PARTICUOLOGY Vol. 3, No.

3, 191-195, 2005

Mini Review

WET-GRANULATION RESEARCH WITH APPLICATION TO SCALE-UP

Gabriel I. Tardos
Department of Chemical Engineering, The City College of The City University of New York, New York, NY 10031, USA E-mail: tardos@chs2s0.engr.ccny.cuny.edu

Summary Granulation is a unit operation by which larger granules are produced from fine, powdery particles to improve appearance, flow properties and mixedness, reduce dustiness and, in general, produce engineered particles with superior attributes. Agglomeration in wet granulation is achieved by introducing a binder fluid onto a shearing mass of fine powders. This paper gives a general overview of the process with emphasis on a simplified granulation model based on a dimensionless parameter containing inertia and viscous dissipation energies between colliding particles: the so-called Stokes number. The model incorporates most common features of all granulation devices (mixers) used in the pharmaceutical industry. Also described in the paper is a computer simulation that captures the movement of flowing powder in an ideal mixer-granulator with constant shear rate. A fraction of the total number of particles is wet (covered by binder and therefore sticky) while the rest of the particles are dry. The numerical simulation depicts two distinct regimes of agglomeration found in a typical granulator: granule growth and subsequent breakup. During granule growth-simulations, final granule size and shape distributions are obtained by analyzing the size and shape of formed granules using a pattern-recognition technique. A second kind of simulation, also using rapid granular flow modeling, follows the rotation and deformation of an agglomerate held together by a liquid binder. Results from these simulations yield critical values of the Stokes number. Below the critical value, the agglomerates are stable and only rotate in response to shear while above the critical value they break into two or more pieces. At the critical value, they attain a steady elongated shape. Using values of the critical Stokes number, the model predicts the size of formed granules. The existence of the critical state in which granules attain a characteristic elongated shape is used to measure shear forces in a granulator by employing calibrated test particles of known strength. This knowledge is employed in granulation scale-up to determine a kinematic rule that conserves stresses in the small and the large-scale machines. It was found that conserving the magnitude of internal stresses in the moving powder yields granules with similar attributes in granulators of different size. Keywords
agglomeration, granule growth theory, aggregation simulation

Introduction
During a granulation operation fine powders are agglomerated into larger granules to produce engineered particles with superior attributes such as improved appearance, flow properties and mixedness and reduced dustiness. Agglomeration in wet granulation is achieved by introducing a binder fluid onto a shearing mass of fine powders. The process is used extensively in the continuous-mode in the detergent and chemical industries or in batch-mode, mostly in the pharmaceutical industry. During batch-wet granulation the process is enclosed in a vessel, and usually no external controls are applied as granules grow and consolidate inside the equipment. The most widely used control parameter is impeller power draw and the equipment is stopped when either the current or the torque on the impeller reaches a predetermined value along the process. Wet granulation usually contains two main periods, binder introduction and wet massing. During binder introduction, water or another appropriate liquid (binder) is poured, sprayed or dripped into the moving powder mix.

The expectation is that, due to the movement of the powder and the wetting properties of the liquid, the binder is distributed on most powder surfaces rendering them sticky enough to allow particle-particle coalescence upon mutual collisions. Wet massing is the subsequent process by which, due to relative movement of different particles in the mixer and to the presence of sticky particle surfaces, granules grow and consolidate. The two processes can usually not be separated either spatially or temporally unless they are performed in different vessels (Mort et al., 2001), and the net outcome is that growth occurs both during binder introduction and wet massing, resulting in a wide granule size distribution and significant variability from run to run. Binder granulation can be achieved in different types of mixers ranging from rotating drums and pans to high shear mixers and fluidized and spouted beds (Tardos et al., 1997; Talu et al., 2000; Iveson at el., 2001). Fluidized bed granulation is in some ways different from other types of mixer granulation (Turton, et al., 1999; Tardos & Mort, 2003) in that the gas supplied to produce powder agitation through fluidization also causes binder evaporation and

Note: This is a shorter version of a plenary lecture on granulation technologies delivered at the AAPS meeting in Lansdown Va., June 2003.

192

CHINA PARTICUOLOGY Vol. 3, No. 3, 2005 dissipated during collision and no growth occurs.

cooling (or heating) of the powder. In addition, particle size increase in a fluidized bed is associated with many changes in fluidization characteristics the most important of which are the mixing properties of the bed. These interacting phenomena make fluid bed granulation by far the most complex while at the same time the most versatile allowing drying and cooling transformations to be carried out simultaneously with size increase (agglomeration). On the other hand, simultaneous transformations pose difficult challenges to scale-up of industrial applications. High shear granulation using mechanical mixers that impart high energy to the powder are used extensively in the detergent and pharmaceutical industries with great success. Recent work by the present author showed that if shearing conditions in different but geometrically similar machines are kept constant (Tardos et al., 2004), the only important variable is the total amount of binder present. Under these conditions the process becomes more robust and scale-up to larger size equipment is greatly simplified. Additional information regarding this topic and several papers dedicated to closely related subjects could be found in the special issue of Powder Technology dedicated exclusively to granulation (Tardos, 2001).

Fig. 1

Schematic of two approaching, binder-covered particles in a granulating mix of powders.

To use the above-simplified model for actual predictions, it is necessary to assign values to the relative velocity, U0. A rough estimate is given as U0=a 0, where 0 is the shear rate in the mixing powder. Using the above velocity, one can calculate the limiting (critical) granule size where Stcoal=Stcoal*:
coal (2) acr = 9 Stcoal * 8 p 0 . Below the critical size, acr, granule growth is rapid while above this critical size it slows considerably since coalescence of larger particles ceases at the critical point where Stcoal=Stcoal*. An important observation is related to the characteristic particles size, a, which is well defined for two equal particles but, in a more general case, becomes an equivalent size calculated from, 1/a=1/a1 + 1/a2 where a1 and a2 are the sizes of the colliding particles (granules). It is immediately apparent that during collisions between small and large particles when a1 a2 , the equivalent

12

Theoretical Basis Granule Growth Kinetics

Significant advance in the theoretical approach to treat granulation was achieved in recent years (Ennis et al., 1991; Khan & Tardos, 1997; Iveson et al., 2001). To simplify the process, it is assumed that sufficient binder is present in the granulator following the binder addition phase and that the binder is appropriately spread on enough granular surfaces. It is also assumed that the particles have a characteristic dimension, a, that for a spherical particle can be taken as its diameter. The liquid binder is characterized by its surface tension and viscosity and the relative velocity between particles is taken to be U0. The system of two approaching particles is shown schematically in Fig. 1 where we take the distance between them to be h. The outcome of the collision of the two binder-covered particles is determined by the ratio of the initial kinetic energy of the system and the energy dissipated in the liquid bridge and in the particles. This can be expressed analytically by the definition of a Stokes number (Ennis et al., 1991) related to particle coalescence, Stcoal : Stcoal = (initial energy)/(dissipated energy) 2 (1) = mp(2U0) /(2Fvish). where p is the particle density, mp is the mass of the particle and Fvis= 3U0a2/(4h) (Ennis et al., 1991) is the viscous force due to the liquid bridge. Simple energy considerations show that if the Stokes number is smaller than a critical value, Stcoal<Stcoal*, collisions between particles are effective, coalescence occurs and granule growth takes place. If the opposite is true, Stcoal>Stcoal* particles rebound because the total incoming energy is larger than the one

size is close to that of the smaller particle, aa2. The implication is that for small particles the dimensionless Stokes number is always small, that is, always less than the critical value, Stcoal*, and therefore small particles will be preferentially captured by larger ones in a process called layered growth as shown in Fig. 2.
1/aeq=1/a1+1/a2

a1 a2 For a1 Steq Fig. 2 aeq a2;

a2 ;

Stl* and coalescence will take place. Schematic representation of the layer-growth process.

It would appear from the above analysis that once granules reach the critical size characteristic of the layering (coating) regime, any increase in the critical Stokes number or decrease in shear rate will increase the size of the granules. Since larger granules are exposed to larger shear forces due simply to their size, we present here a simplified model that accounts for the breakage of these granules that limits their growth. To formalize this mathematically, it is useful to define a new dimensionless Stokes number (Tardos et al., 1997) that relates initial kinetic en-

Tardos: Wet-Granulation Research with Application to Scale-up ergy in the shearing mass to the internal energy resisting deformation, in the following form: Stdef = (external kinetic energy)/(energy of deformation)
2 = mpU0 2Vp 0 . (3) The notations in this equation are similar to those in Eq. (2) where in addition, Vp, is the particle (granule) volume, mp=Vpp is the granule mass and ( 0) is some characteristic strength of the granule that can be taken as its yield strength ( 0)=y. The Stokes number defined above increases with increasing particle size, a, and reaches at some point during granulation a critical value, Stdef*, above which granules start to deform and break. At the critical point, when Stdef=Stdef*, a limiting granule size is attained:

193

( )

systems. Our main contribution was to add liquid layers with well-defined characteristics such as viscosity and surface tension to some particles thereby making them sticky and watch them agglomerate. Without going into details, we show in Fig. 4 the size distribution of agglomerates from a calculation with 40 000 particles of which 20% were sticky and were covered by a liquid layer with a thickness of 10% of the particle diameter. Both the surface tension and the viscosity of the layer (defined within a dimensionless parameter, the so-called Capillary number, Ca=U0/ ) influence the size of the agglomerates formed unlike the simple model presented above where only viscosity of the binder had an effect.
Cumulative distribution/% Cumulative Distribution, [%] t=0 t=0

acrdef = ( 2 ySt def * p )

12

0 .

(4)

100 80 60 40 20

The above considerations are summarized in Fig. 3, where both the growth limit characterized by Stcoal=Stcoal* and the deformation (breakage) limit characterized by Stdef=Stdef* are presented. The ratio of Stokes numbers is presented on the ordinate while the critical sizes associated with equilibrium points are presented on the abscissa.

time

t*=0, d(3,2)= 1 t*=0, d(3,2)=1 t*=10, d(3,2)=2.43 t*=10, d(3,2)= 2.43 t*=60, d(3,2)=3.80 t*=60, d(3,2)= 3.80 t*=100, (3,2)=4.18 t*=100, d(3,2)= 4.18 t*=160, d(3,2)=4.22 t*=160, d(3,2)= 4.22 t*=200, (3,2)=4.30 t*=200, d(3,2)= 4.30

0
Fig. 4

10

12

14

16

Dimensionless agglomerate sizer, dagg/dp Dimensionless Agglomerate Size, [dagg/dp] Particle size distribution and Sauter mean diameters [d(3,2)] of agglomerates in the constant shear field as a function of time.

Fig. 3

Summary of the coalescence and granule breakage model.

As seen in the figure, particles will grow by nucleation and coalescence until they reach the growth limit, acrcoal. Since this limit is calculated strictly for equal-size particles, unequal particles can still grow by layering to yield granules somewhat larger than acrcoal. Granules that grow beyond, acrdef, become unstable, deform and eventually break. The process will stabilize at some point between the two limiting sizes and will be characterized by a size distribution that will shift slowly toward larger sized granules and will, after a long enough time, reach acrdef. At that point, this size will be a theoretical steady state condition.

Molecular Dynamic Simulations (MDS) of Granulation

The goal of these computations was to simulate the collision of large numbers of particles moving in a shear field in which some particles are dry and some are wet and therefore sticky (Talu at al., 2000). The calculations are relatively straightforward in the sense that many of these simulations have been performed to date with dry particle

As seen in the figure, all particles are of relative diameter dagg/dp=1 at time t*=0 (mono-disperse particle size distribution). In these calculations, t* is a dimensionless time defined as t*=t 0, where 0 is the shear rate. Here we took the size of the particle, a=dp, and the size of the agglomerate, acr=dagg, to be equal to their diameter. As time progresses, larger agglomerates form and the size distribution widens, as one would expect. The increase in size slows as time progresses and beyond a dimensionless time of approximately t*=100, a steady state distribution is reached. This can also be seen from the Sauter mean diameter (d(3,2)) calculated in the figure that only changes by about 2.8% as time doubles from t*=100 to t*=200. This appears to be a dynamic steady state when agglomerates form, grow to the critical size, break and reform and grow again resulting in an almost constant particle size distribution. Several calculations of the kind shown above were performed with different values of the Stokes (defined by Eq. (1)) and Capillary numbers (Ca=U0/ ). The Stokes number can be viewed as a dimensionless viscosity while the Capillary number can be perceived as a dimensionless surface tension, and a given pair of values completely defines the properties of the sticky layer on the particle (granule) surface. The average granule size resulting from these calculations was correlated with the parameters of the system using a general equation of the form:

194

CHINA PARTICUOLOGY Vol. 3, No. 3, 2005

dagg dp = ( St * St ) ,

(5)

where =0.4 for the case of granules formed by coalescence and =1 for the case of granule deformation and breakup. Essentially, Eq. (5) is a simple recasting of equations (2) and (4) as they result from the simple theoretical model presented above with the only observation that =0.4 instead of 1/2. Values for the critical Stokes number, St* contain all other parameters of the problem that are not directly incorporated into the Stokes number. For growth by coalescence, the numerical simulation gives:
St * = 44r 1.65 Ca 0.32

2.5

for St < 1 ,
2.5

(5a)

D according to the relationship NDn=constant, where n is a scaling index. The simplest scaling rule maintains constant tip speed at all scales using n=1, while n=0.5 is consistent with constant Froude number (the Froude number is the ratio of centrifugal to gravitational forces in the mixer and can be expressed as Fr=N2D/g). The law of equal tip speed was used overwhelmingly in the past. In fact, many mixers are built and hard-wired in the factory to exactly match the equal tip speed rule upon scale-up. Only during the last several years were other considerations taken into account (see for example Landin et al., 1996a and 1996b; Mort & Tardos, 1999; Iveson et al., 2001).

14 St * = 92r 1.65 ( h1 dp ) Ca 0.32 for St > 1 , (5b) and for granule breakup the result can be summarized as: St * = 2.25 Ca for Ca < 10 (5c) Here r is the ratio of binder covered to the total number of particles in the system and h1 is the thickness of the binder layer. The above equations were obtained by curve fitting of the calculated values of granule size using the numerical scheme and recasting of the final result in the form of Eq. (5). The advantage of the Molecular Dynamic Simulation is that it predicts an actual value of the critical Stokes number St* as compared to the simple energy-based model that only yields the result given in Eq. (5). The above equations are of practical value since the size of agglomerates resulting from coalescence can be calculated from equations (5a and 5b) while the largest agglomerate present under given magnitude of shear is give by Eq. (5c).

Fig. 5 Deformed and un-deformed agglomerates from an MDS simulation and test pellets from an actual experiment. Under the condition of equal shear stresses both the agglomerate and the test pellets assume an elongated shape before they break.

Measurement of Stresses in a Granulator: Application to Scale-up

An important result from the molecular dynamic simulation (MDS) was the finding that at and around the critical condition before breakage, agglomerates assume an elongated shape as shown on the LHS of Fig. 5. We used the above finding (Tardos et al., 2004) to measure stresses present in a granulator by introducing into the shearing mass of powder, test pellets of known, constant yield strength. The pellets were made of an almost perfect plastic (Bingham solid of relatively low flow viscosity) that exhibited constant yield strength over a wide range of shear rates. Materials of decreasing yield strength were used in granulations performed at a given binder/solid ratio until deformation was observed as shown on the RHS of Fig. 5. It was considered that at the point of deformation the shear force in the wet-mass equaled the yield strength of the test pellet. We found that in the dry powder and at low binder levels, the shear forces in a high-shear granulator are relatively low. When the binder content reaches a critical value, shear forces intensify significantly and grow almost exponentially as more binder (liquid) is added. Scale-up rules for vertical axis granulators are typically kinematic in nature, with the main impeller rotational speed, N, varied as a power-law function of the impeller diameter

The purpose of the experiments (Tardos et al., 2004) was to determine the scaling index n so as to achieve shear stresses of equal magnitude in different scale granulators using the same powder mix and the same amount of binder. We used granulators of different scales from 2 liters to 25 liters and several different pharmaceutical formulations and found that the index should be n=0.80.85 the spread depending mostly on the shape of the impeller as shown in Fig. 6. This is an important finding since it yields a scaling criterion for vertical-axis high-shear granulators based on solid physical principle.

Fig. 6

Scale-up of high-shear granulators with vertical axis with different impeller geometries and different inserts.

Tardos: Wet-Granulation Research with Application to Scale-up

195

Conclusions
A review of the more recent advances in the theoretical and experimental treatment of granulation of fine powders was presented. A simple theoretical model based on energy conservation principles during collision of bindercovered particles was described. The model provides insight into a very complex process and reduces it to simple concepts on the particle-particle and particle-binder micro-level. Based on this fundamental understanding, a simulation model was developed that captures the micro-level approach but also makes it more general and incorporates other important interactions such as capillary, collision and viscous forces between particles. The simulation generates particle and granule size distributions, as would a real granulation process. To further improve the model and make it equipment specific, one needs to incorporate into the simulation a realistic powder flow model: efforts to achieve this are under way. An important finding from the simulation, i.e., the existence of the critical state where granules deform before breaking, was successfully used to measure stresses in an actual high-shear granulator. It was shown that using pellets with well-defined yield strength, one could indirectly measure shear forces in a granulating mass of wet powder. It was found that shear forces in a High-Shear mixer-granulator of the vertical type (Fielder) are actually not always high. Shear forces in the dry mix appear to be very low. The mixer has the potential to produce high shear forces, but these forces are transmitted to the powder mass only if it becomes cohesive due to binder addition. Shear forces in the granulator are strongly wet-mass dependent, and they increase rapidly as soon as a critical cohesivity is achieved; this appears to coincide with the onset of granule growth. Based on the above finding, we proposed an improved granulation process in which dry powders are pre-wetted before introduction into the main granulating device. This scheme has the potential to produce larger shear forces during subsequent wetting and binder introduction and thereby improve homogeneity and consequently final granule properties. It was also shown that geometrically similar granulators could be scaled-up to exhibit constant shear forces. A scale-up index for the scaling law NDn=constant, of 0.8<n<0.85 was found to uniformly apply to scaling from a 2-L machine to a 7.5-L device and further to a 25-L mixer. The shape of the impeller was found to have a non-trivial effect on scale-up in that a tangential impeller imparted

more shear to the powder then the axial impeller under otherwise identical conditions. These are important findings since it is the first scale-up model to be based on scientific criteria and achieve similar shear forces and granule attributes in different size equipment.

References
Ennis, B. J., Tardos, G. I. & Pfeffer, R. (1991). A microlevel-based characterization of granulation phenomena. Powder Technol., 65, 257-272. Iveson, S. M., Litster, J. D., Hapgood, K. & Ennis, B. J. (2001). Nucleation, growth and breakage phenomena in agitated wet granulation processes: a review. Powder Technol., 117, 3-39. Khan, M. I. & Tardos, G. I. (1997). Stability of wet agglomerates in granular shear flows. J. Fluid Mech., 347, 347-368. Landin, M., York, P., Cliff, M. J., Rowe, R. C. & Wigmore, A. J. (1996a). Scale-up of pharmaceutical granulation in a fixed bowl mixer-granulator. Int. J. Pharm., 133, 127-131. Landin, M., York, P., Cliff, M. J., Rowe, R. C. & Wigmore, A. J. (1996b). The effect of batch size on scale-up of a pharmaceutical granulation in a fixed bowl mixer-granulator. Int. J. Pharm., 134, 243-246. Mort, P. R. & Tardos, G. I. (1999). Scale-up of agglomeration processes using transformations. KONA, 17, 64-76. Mort, P. R., Capeci, S. W. & Holder, J. W. (2001). Control of agglomerate attributes in a continuous binder-agglomeration process. Powder Technol., 117, 173-176. Talu, I., Tardos, G. I. & Khan, M. I. (2000). Computer simulation of wet granulation. Powder Technol., 110, 59-75. Talu, I., Tardos, G. I. & van Ommen, J. R. (2001). Use of stress fluctuations to monitor wet granulation of powders. Powder Technol., 117, 149-162. Tardos, G. I., Khan, I. M. & Mort, P. R. (1997). Critical parameters and limiting conditions in binder granulation of fine powders. Powder Technol., 94, 245-258. Tardos, G. I. (2001). Granulation and coating of particles. Powder Technol., Vol. 117 (pp.1-2), Special issue dedicated to the topic of granulation. Tardos, G. I. & Mort, P. J. (2003). Fluidization: applications for coating and granulation. In Yang, W. C. (Ed.), Handbook of Fluidization and Fluid-Particle Systems (pp.445-468). New York: Marcel Dekker. Tardos, G. I., Hapgood, K. P., Ipadeola, O. O. & Michaels, J. N. (2004). Stress measurements in high-shear granulators using calibrated test particles: applications to scale-up. Powder Technol., 140, 217-227. Turton, R., Tardos, G. I. & Ennis, B. J. (1999). Fluidized bed coating and granulation. In Yang, W. C. (Ed.), Selected Topics on Fluidization, Solids Handling and Processing (Chapter 6, pp.331-429). Noyes Publications.
Manuscript received December 29, 2004 and accepted January 18, 2005.