Elsevier

Chemical Engineering Science

Volume 92, 5 April 2013, Pages 134-145
Chemical Engineering Science

Distribution nucleation: Quantifying liquid distribution on the particle surface using the dimensionless particle coating number

https://doi.org/10.1016/j.ces.2013.01.010Get rights and content

Abstract

Control of nucleation in wet granulation often determines the size and structure of the final granules. After a decade of research, our knowledge of “immersion nucleation”, where a large droplet engulfs fine particles, is relatively well established. However, far less is known about the second mechanism called “distribution nucleation”, where the powder particles are gradually coated by a layer of small droplets. In this paper, the similarities between particle coating and granulation were used to define the five steps of distribution nucleation. A Bernoulli model was developed to describe the fractional surface coating F, and a new dimensionless parameter, the particle coating number Φp, was defined as the ratio of the theoretical area coated by the drops, assuming no overlap, to the total surface area of the particle.

The particle coating number was experimentally validated by adding drops randomly over the surface of a particle and measuring the fractional surface coating using image analysis. Standard ping pong balls (40 mm diameter) plus four sizes of foam balls ranging from 20 to 50 mm diameter were coated with either small or large drops of melted PEG1000. The results demonstrated that the particle coating number Φp can be used to predict the fractional surface coverage F using simple, known parameters (without any fitting coefficients), and was able to account for differences in particle size and drop size. Deviations in the experimental data from the theoretical predictions were observed, and attributed to drop spreading and merging and heat conduction into the particle, which all affect the experimental drop footprint area. When the effective footprint area was used as a fitting parameter all the experimental data collapsed onto a single line, demonstrating that the particle coating number captures the key coating coverage behaviour.

The particle coating number can also be used to predict the effect of changing particle size, surface area, liquid level, or drop size on the coating fraction, which is in turn known to be linked to granulation kinetics. The particle coating number opens up new options for granulation process control, and is expected to be valuable in a variety of particle wetting and coating applications where small drops are distributed over larger particles.

Highlights

► The five steps in distribution nucleation were defined. ► New dimensionless parameter, particle coating number, was derived and experimentally validated. ► Fractional coating can be predicted in advance based on simple parameters. ► Particle coating number is expected to improve modelling studies and to aid process control of fluidised bed granulation.

Introduction

Wet granulation is an important process which converts powders into structured granules by adding a liquid binder, and in so doing improves powder flow, increases bulk density, minimises dust and reduces segregation. Granulation is used in a broad range of industries, including in the chemical, pharmaceutical, food, and, mining industries. Each industry has its own specialised granulation equipment, including pans, drums, mixers, and fluidised beds (Litster and Ennis, 2004). During granulation, there are three general mechanisms that can occur simultaneously or in rapid succession: wetting and nucleation, growth and breakage (Iveson et al., 2001). The balance of these mechanisms depends on the equipment design, formulation, and operating conditions.

Nucleation is the first mechanism where the powder is wetted by the binder fluid and the initial “nuclei” granules are formed. The liquid binder usually is added via an atomised spray, particularly for fluid bed granulation (Boerefijn et al., 2009, Hede et al., 2008, Schaafsma et al., 1999, Schaafsma et al., 2000, Schæfer and Wørts, 1977, Seo et al., 2002).

There are two nucleation mechanisms which are distinguished by the relative ratio of binder droplet diameter ddrop to the particle diameter dparticle. The first mechanism, known as “immersion nucleation,” occurs when the small powder particles get immersed in a relatively large binder droplet (ddrop>dparticle). Over the last decade, research on immersion nucleation has become well established (Abberger et al., 2002, Ax et al., 2008, Chouk et al., 2009, Hapgood et al., 2010, Hapgood et al., 2002, Hapgood et al., 2003, Hapgood et al., 2004, Hapgood et al., 2009, Iveson et al., 2001, Johansen and Schæfer, 2001, Lee and Sojka, 2011, Litster et al., 2002, Litster et al., 2001, Marston et al., 2010, Nguyen et al., 2009, Oullion et al., 2009, Plank et al., 2003, Seo et al., 2002, Wildeboer et al., 2007, Wildeboer et al., 2005). The large droplet contacts the powder particles and produces a highly saturated initial nucleus, where the size is proportional to the size of the drop (Ax et al., 2008, Hapgood et al., 2003, Schaafsma et al., 2000, Waldie, 1991). There are generally five steps in immersion nucleation (see Fig. 1), starting with the (1) formation of the droplets; (2) droplet impact on the powder surface and possible breakage (Agland and Iveson, 1999, Chouk et al., 2009, Lee and Sojka, 2011, Marston et al., 2010); (3) droplet coalescence at the powder surface; and ending with the (4) penetration of the drop into the powder bed and/or the (5) mechanical dispersion of the droplet through the powder (Hapgood et al., 2003).

The second mechanism, known as “distribution nucleation”, occurs when a large powder particle is wetted by small binder droplets distributed over the particle surface (ddrop<dparticle). Distribution nucleation is poorly understood (Boerefijn and Hounslow, 2005, Štěpánek and Rajniak, 2006), although there is a wealth of literature on fluid bed coating of particles, where there are clear similarities (Hemati et al., 2003, Karlsson et al., 2011, Nienow, 1995, Panda et al., 2001, Teunou and Poncelet, 2002, Turton and Cheng, 2005), particularly since small atomised drops are used to form a coating layer around the exterior of larger particles. However, granulation (or agglomeration) of the coated particles is considered an undesirable side effect of coating processes (Boerefijn et al., 2009).

Coating begins with the droplet deposition on the particle surface, which includes the impact, wetting and spreading of the fluid over the surface (Karlsson et al., 2011, Link and Schlunder, 1997, Panda et al., 2001). Gradually, more and more drops cover the particle surface, gradually forming a wet coating layer. Simultaneous drying within the fluid bed causes the coating to dry and minimises agglomeration of the particles. However, if the flux of fluid is too high (Boerefijn et al., 2009, Hede et al., 2008), the bed will become too wet, allowing liquid bridges to form between colliding particles, which may results in agglomeration of the coated particles.

Building on the existing work on coating, we define the five stages of distribution nucleation, as shown in Fig. 2. Initially, the binder fluid must be atomised into droplets which are significantly smaller than the particles to be granulated. Step 2 indicates a single droplet impacting on the particle and creating a solid–liquid interface, which will be affected by a range of processes including surface chemistry, fluid rheology, drop dynamic spreading and contraction, particle roughness, interaction with adjacent droplets, the overall drying rate etc. until the drop reaches a pseudo-equilibrium position on the particle surface.

Step 3 in Fig. 2 shows the particle coating stage, where the particle is gradually covered by more and more droplets. Each droplet successfully deposited increases the fraction of the particle surface area which is coated with droplets. The random distribution of the droplets landing of the particle surface will not be uniform, and as a result the particle will not be uniformly wetted (Karlsson et al., 2011, Panda et al., 2001, Rajniak et al., 2007, Štěpánek and Rajniak, 2006). Steps 1–3 of distribution nucleation have been investigated in numerical volume of fluid (VOF) and Monte-Carlo simulations by Štěpánek and Rajniak (2006). In a VOF simulation, droplets were deposited (stage 1) randomly onto particles of varying roughness. The drops were allowed to spread over the undulating particle surface and reach an equilibrium position before another drop was added. These simulations produced elegant 3D representations of droplets coating single particles as a function of droplet size, particle roughness, and the equilibrium contact angle of the fluid on the solid. They tracked the increase in the liquid coverage of the particle surface, called the “surface coverage”, as more drops were deposited on the particle. The surface coverage fraction was initially directly proportional to the number of drops deposited on the particle, but as the drop density increased, the probability of each new drop landing on the dry section of the particle decreased. As a result, the rate of increase of the particle surface coverage levelled off and asymptotically approached 100% (Štěpánek and Rajniak, 2006). The fraction of the particle surface covered ψ as a function of the liquid to solid ratio xLS was fitted to empirical curves of the form:Ψ=1exp(kxLS)where k is a fitting parameter, fitted for each set of experimental simulation condition (e.g., same roughness and contact angle). The simulations also showed that liquid is not uniformly distributed over the particle surface and that the surface coverage is affected by the particle roughness. These simulations were later used to model real pharmaceutical formulations and explain differences in the experimental growth curves (Rajniak et al., 2007, Štěpánek et al., 2009).

The final two steps of distribution nucleation shown in Fig. 2 are inter-particle collisions (step 4) and the formation liquid bridge (step 5) to create a nucleus granule. Nucleation is defined here as the beginning of agglomeration and therefore involves more than one particle held together by a liquid bridge. In the case of immersion, nucleation is initiated the instant that the droplet meets the powder as it covers more than one particle at collision. In the case of distribution nucleation, nucleation is not instantaneous; it requires two or more partially wetted particles colliding at their wetted region. This collision results in liquid bridges being formed between the particles, which form stronger and more rigid bonds as they dry. To increase the chances of two partially wetted particles forming a liquid bridge and initiating nucleation, adequate surface coverage is required. Particles with low surface coverage which collide may not form a liquid bridge after collision, either because they collided at a dry patch, or because the volume of liquid in the collision zone was insufficient to form a liquid bridge. However, adding excess liquid to create a high surface coating of the particles increases the cost, the granule growth rate and risk of uncontrolled granulation, and drying time and energy.

Whether a nucleus granule forms depends on several factors—the collision velocities, the fluid properties, the particle roughness and the amount of fluid present at the contact point (Ennis et al., 1990, Ennis et al., 1991). Specifically, the probability of a collision resulting in a stable liquid bridge between two particles i and j after a collision Cij can be written as (Štěpánek et al., 2009):Cijsucc=ψijphysψijgeomCijwhere ψij are the physical and geometric success factors. The physical success factors can be calculated by the Stokes’ coalescence models (Ennis et al., 1991, Liu et al., 2000) and are relatively well described. A major contributor to the geometric success factor is the presence or absence of liquid at the collision point (Štěpánek et al., 2009). Determining whether there is liquid at the contact point has so far been impossible, let alone whether there is sufficient liquid at the contact point to form a liquid bridge. Whilst the sophisticated model of Štěpánek and Rajniak (2006) could, in theory, be applied during fluid bed simulations, the model would be extremely complex and this approach is currently impractical.

In addition, there will also be a distribution of the number of drops deposited per particle, which adds a further level of variation in the liquid coating of the fluidised particles. Understanding both the variation in liquid coating between particles as well as over the surface of a single particle has, to date, been an extremely complex undertaking. In this paper, we are focussing on the liquid distribution over the surface of a single particle—separate work on understanding inter-particle coating variability is underway.

Although there are several dimensionless design groups for immersion nucleation, including dimensionless spray flux (Hapgood et al., 2004, Litster et al., 2001) and drop penetration time (Hapgood et al., 2002, Nguyen et al., 2009) which have been applied to a variety of processes (Ax et al., 2008, Cavinato et al., 2011, Chouk et al., 2009, Hapgood et al., 2010, Plank et al., 2003, Tan et al., 2009), there are currently no dimensionless groups which describe distribution nucleation. This is a serious gap in knowledge because nucleation in fluid beds tends to occur via the distribution mechanism (Boerefijn and Hounslow, 2005). Further study of distribution nucleation is required to develop a suite of design rules equivalent to the design rules developed for immersion nucleation.

There is currently no method to quantify the fractional surface coverage of a particle or to accurately predict whether two colliding particles will have sufficient local liquid to form a liquid bridge between the particles. This paper investigates distribution nucleation by deriving and validating a new dimensionless particle coating number Φp to predict the fractional coverage of a particle with liquid as a function of particle size, particle surface area, number of drops, and drop footprint area. The ability to accurately predict particle coating and surface coverage will have immediate applications in granulation modelling, process design and troubleshooting of fluid beds. In addition, it will advance our understanding of processes which involve either particle coating or distribution nucleation.

Section snippets

Bernoulli model for coating coverage F

The model system consists of a single particle being bombarded sequentially with droplets from uniformly random directions. This system is similar in concept to that of Štěpánek and Rajniak (2006); however, once deposited (ignoring overlap) all droplets cover the same footprint area on the surface of the particle. Theoretical model development then follows similarly to Freireich and Wassgren (2010) and Freireich et al. (2011); except the fractional area coated per coating event is smaller,

Materials and methods

To validate the dimensionless particle coating number Φp (Eq. (12)) and its relationship with the actual particle coating coverage fraction F (predicted using Eq. (13), (14)), experiments were conducted by depositing drops randomly on particles of different sizes and measuring the fraction of the surface coated with drops using image analysis. The particles used in this study were ping pong balls (PP) and polystyrene foam balls (FB) as summarised in Table 1.

Drops of polyethylene glycol

Selection of footprint area estimation method

The single droplet footprint areas measured and given in Table 2 may be compared to the models derived in Section 2.2.1 (Fig. 4). This figure shows that the projected footprint area given by Eq. (17) always under-predicts the measured droplet footprint. As expected the ping pong ball and foam balls seem to follow similar but separate trends. By adjusting the contact angle until the experimentally measured single footprint area agreed with the calculated footprint areas, the ping pong ball has

Deviations from predicted coating coverage

There are several causes of experimental variation which should be considered. First, the manual rotation of the ball between drop additions is unlikely to be truly random and will therefore result in some deviations from the theoretical coverage. We have now designed and built a robot manipulator which will be able to rotate the ball to a new random position after each drop has been added. In addition, the coating process is inherently a discrete, stochastic, Monte-Carlo type process. Many

Conclusions

The five steps of distribution nucleation indicated that the particle coating step was a critical precursor to particle coalescence. The particle coating number Φp derived here is able to predict the fractional surface coating F, including variations in particle size and drop size, and represents the first quantitative design approach to distribution nucleation and coating. Whilst some experimental variations in drop footprint area resulted in some bias and scatter, using the footprint size as

Acknowledgements

This project was supported financially by an Australian Research Council Discovery Project DP0877961.

References (47)

  • K.P. Hapgood et al.

    Improving liquid distribution by reducing dimensionless spray flux in wet granulation—a pharmaceutical manufacturing case study

    Chem. Eng. J.

    (2010)
  • K.P. Hapgood et al.

    Drop penetration into porous powder beds

    J. Colloid Interface Sci.

    (2002)
  • K.P. Hapgood et al.

    Dimensionless spray flux in wet granulation: Monte-Carlo simulations and experimental validation

    Powder Technol.

    (2004)
  • P.D. Hede et al.

    Validation of the flux number as scaling parameter for top-spray fluidised bed systems

    Chem. Eng. Sci.

    (2008)
  • M. Hemati et al.

    Fluidized bed coating and granulation: influence of process-related variables and physicochemical properties on the growth kinetics

    Powder Technol.

    (2003)
  • S.M. Iveson et al.

    Nucleation, growth and breakage phenomena in agitated wet granulation processes: a review

    Powder Technol.

    (2001)
  • A. Johansen et al.

    Effects of interactions between powder particle size and binder viscosity on agglomerate growth mechanisms in a high shear mixer

    Eur. J. Pharm. Sci.

    (2001)
  • S. Karlsson et al.

    Characterization and mathematical modelling of single fluidised particle coating

    Powder Technol.

    (2011)
  • J.D. Litster et al.

    Scale-up of mixer granulators for effective liquid distribution

    Powder Technol.

    (2002)
  • J.D. Litster et al.

    Liquid distribution in wet granulation: dimensionless spray flux

    Powder Technol.

    (2001)
  • J.O. Marston et al.

    Experimental study of liquid drop impact onto a powder surface

    Powder Technol.

    (2010)
  • T. Nguyen et al.

    Drop penetration time in heterogeneous powder beds

    Chem. Eng. Sci.

    (2009)
  • M. Oullion et al.

    Simulating the early stage of high-shear granulation using a two-dimensional Monte-Carlo approach

    Chem. Eng. Sci.

    (2009)
  • Cited by (37)

    • Experimental characterization of the effect of liquid viscosity on collisions between a multi-component droplet and a heated particle

      2022, Chemical Engineering Science
      Citation Excerpt :

      Coating of particles in a fluidized bed or a rotation drum, which is a repeated process of droplets spraying and particles drying, is widely applied in industry. The micro phenomenon of a droplet colliding with a particle is a critical process (Iveson et al., 2001; Kariuki et al., 2013; Schmidt et al., 2015). The atomized droplets, containing solute and solvent, spread and solidify repeatedly on seed particles, resulting in layer growth or agglomeration of the seed particles (Terrazas-Velarde et al., 2011; Sutkar et al., 2016; Diez et al., 2018).

    View all citing articles on Scopus
    1

    Now at The Dow Chemical Company, Midland, MI, USA

    2

    Now at Department of Chemical and Biological Engineering, Sheffield University, Sheffield, UK

    View full text