effects of binder rheology on melt agglomeration in a high shear mixer

International Journal of Pharmaceutics 176 (1998) 73–83

Effects of binder rheology on melt agglomeration in a highshear mixer

Helle Eliasen, Torben Schæfer *, H. Gjelstrup Kristensen

Department of Pharmaceutics, The Royal Danish School of Pharmacy, 2 Uni6ersitetsparken, DK-2100 Copenhagen, Denmark

Received 27 July 1998; accepted 8 September 1998

Abstract

Lactose monohydrate was melt agglomerated in an 8-l high shear mixer using Gelucire 50/13, Stearate 6000 WL1644, or polyethylene glycol (PEG) 3000 as meltable binder. The impeller speed was varied at two levels, and massingtime was varied at six levels. In order to obtain a similar agglomerate growth, a larger binder volume had to be usedwith Gelucire 50/13 than with Stearate 6000 WL 1644 and PEG 3000. The lower viscosity of Gelucire 50/13 gave riseto agglomerates of a wider size distribution and a higher porosity as well as more adhesion of mass to the bowl. Alower binder viscosity resulted in more spherical agglomerates at the low impeller speed. © 1998 Published by ElsevierScience B.V. All rights reserved.

Keywords: Binder viscosity; Melt agglomeration; High shear mixer; Adhesion; Porosity; Water of crystallization

1. Introduction

In a wet agglomeration process, the agglomer-ate growth is affected by the viscosity of thebinder liquid. Initially, the agglomerate formationand growth depend on the distribution of thebinder, which is facilitated by a low viscosity. Thesubsequent agglomerate growth by coalescence ispromoted by a higher binder viscosity (Ennis etal., 1991; Schæfer and Mathiesen, 1996b). Kening-ley et al. (1997) found that the minimum binder

viscosity required to form granules increased withthe particle size of the solid material.

A prerequisite of a formation of agglomeratesin a mixer granulator is that the agglomerates areable to resist the impact with the rotating im-peller. The strength of the dynamic liquid bridgeswithin an agglomerate is dominated by viscousforces (Mazzone et al., 1987). A lower binderviscosity and a larger particle size will decreasethe agglomerate strength (Schæfer and Mathiesen,1996b; Keningley et al., 1997). When the agglom-erate strength is so low that fracture occurs, noagglomerates will be formed, and an adhesion of* Corresponding author.

0378-5173/98/$ - see front matter © 1998 Published by Elsevier Science B.V. All rights reserved.

PII S0378-5173(98)00306-8

H. Eliasen et al. / International Journal of Pharmaceutics 176 (1998) 73–8374

moist mass to the wall of the mixer bowl will beseen instead (Keningley et al., 1997). This meansthat a lower binder viscosity might augment theadhesion of mass to the wall. It has also beenfound, however, that the adhesion becomes in-creased at a very high binder viscosity (Schæferand Mathiesen, 1996b).

The densification of the agglomerates duringmixing is promoted by a lower binder viscosity(Ennis et al., 1991; Schæfer and Mathiesen,1996b; Iveson, 1997). With low viscosity binders,the densification is primarily affected by fric-tional forces between the particles within the ag-glomerate, whereas viscous forces dominatewhen the viscosity is above 1000 mPa· s (Ken-ingley et al., 1997). The densification increasesthe liquid saturation of the agglomerates, andmore binder liquid is squeezed to the surface atcollisions between agglomerates. A lower binderviscosity will cause a higher surface plasticity,which will promote the rounding of the agglom-erates (Schæfer and Mathiesen, 1996b).

In previous melt agglomeration experiments,no effect of binder viscosity on the agglomera-tion properties of eight hydrophobic binders allhaving viscosities below 50 mPa· s was found(Thomsen et al., 1994), whereas marked effectsof binder viscosity were found in experimentswith the more hydrophilic polyethylene glycolshaving viscosities in the range 100–26500mPa· s (Schæfer and Mathiesen, 1996b). Thehydrophilicity of the meltable binder is assumedto affect the wetting of the material and theadhesion of mass to the mixing bowl (Schæfer etal., 1990; Thomsen et al., 1994; Peerboom andDelattre, 1995). It is difficult to distinguishclearly between the effects of viscosity and hy-drophilicity on the basis of the previous meltagglomeration experiments since the low viscos-ity binders examined were hydrophobic, and thehigh viscosity binders examined werehydrophilic.

The purpose of the present work was to eluci-date the effects of viscosity of three hydrophilicmeltable binders covering the viscosity rangefrom approximately 30 to 400 mPa· s.

2. Materials and methods

2.1. Materials

Lactose 450 mesh (a-lactose monohydrate,DMV, The Netherlands) was used as startingmaterial.

Polyethylene glycol (PEG) 3000 (Hoechst,Germany), Gelucire 50/13 (esters of polyethyleneglycol and glycerol) (Gattefosse, France), andStearate 6000 WL 1644 (esters of polyethyleneglycol) (Gattefosse, France) were used asmeltable binders. Gelucire 50/13 has a HLBvalue of 13, and Stearate 6000 WL 1644 has aHLB value of 18 (Product information, Gatte-fosse, France). PEG 3000 is indicated to have aHLB value of 20 (Attwood and Florence, 1983).PEG 3000 was used as flakes whilst Gelucire50/13 and Stearate 6000 WL 1644 were providedas blocks, which were rasped prior to use.

The size distribution by volume of the lactosewas determined by a Malvern 2601Lc laser dif-fraction particle sizer (Malvern Instruments,UK). The median particle diameter and thespan were found to be 23 mm and 2.2. The spanis the difference between the diameters at the 90and the 10 percentage points relative to the me-dian diameter.

The BET multipoint surface area of the lac-tose, determined by a Gemini 2375 surface areaanalyzer (Micromeritics, USA), was 0.94 m2/g.

The true densities of the materials were deter-mined by an AccuPyc 1330 gas displacementpycnometer (Micromeritics) using helium purge.The density of the lactose was 1.545 g/cm3. Thedensities of the molten binders were estimated at70, 80, and 90°C by weighing a known volumeof the binder in a volumetric flask with a nomi-nal volume of 50 ml. The exact volume of thevolumetric flask at each of the temperatures wasdetermined on the basis of the weight of ther-mally equilibrated water and the density of thewater.

The water contents on a wet weight basiswere estimated by volumetric titration as previ-ously described (Schæfer and Mathiesen, 1996a).The water content was found to be 5.0% for thelactose.

H. Eliasen et al. / International Journal of Pharmaceutics 176 (1998) 73–83 75

The melting ranges of the meltable binders wereestimated by a Perkin Elmer DSC 7 differentialscanning calorimeter as previously described(Schæfer and Mathiesen, 1996a).

The viscosities of the molten binders were esti-mated at 70, 80, and 90°C by a rotation vis-cosimeter, Rotovisco RV 12 (Haake, Germany),with a MV sensor system and MV II K rotor. Theviscosity values used are the results obtained atthe highest rpm that gave rise to a deflectionwithin the scale. The results did not indicate thatthe viscosity values were dependent on the rota-tion speed of the rotor.

The loss modulus and the storage modulus ofthe molten binders were estimated with a con-trolled stress rheometer, Carri-Med RheometerCSL2100 (TA Instruments, Leatherhead, UK).The measurements were performed at 70°C at afrequency of 1 Hz in oscillatory mode using acone and plate measuring system. The cone was a6-cm 1° stainless steel cone. A sinusoidal oscilla-tory shear stress was applied and the strain re-sponse was monitored. The data were collected bya computer, and the loss modulus and the storagemodulus were calculated by computer software(TA Instruments Rheology Solutions SoftwareCSL V1.1.6) from data within the linear viscoelas-tic region.

2.2. Equipment

The 8-l laboratory scale high shear mixer(Pellmix PL 1/8, Niro A/S, Denmark), describedin a previous paper (Schæfer et al., 1993a), wasemployed in the experiments. The temperature ofthe heating jacket was set to 35°C in all theexperiments.

2.3. Mixing procedure

The load of the mixer was 1 kg of lactose. Theconcentration of the meltable binders was keptconstant at 23% m/m of the amount of lactose.

In all the experiments, the starting material andthe meltable binder were dry mixed at an impellerspeed of 1300 rpm. The impeller speed was eitherlowered to 1000 rpm or raised to 1500 rpm after3 min of mixing for Gelucire 50/13, after 8 min of

mixing for Stearate 6000 WL 1644, and after 10min of mixing for PEG 3000. These differentmixing times correspond to a change in impellerspeed when the product temperature was about1°C below the melting point of the binder. Themelting point was observed as an inflection pointon the recorded product temperature curve. Thisinflection point was defined as the start of mass-ing time.

The procedures for cooling of the agglomeratesand for estimation of the adhesion to the bowlwere the same as described in a previous paper(Schæfer, 1996).

2.4. Granule characterization

2.4.1. Size distributionThe amount of lumps larger than 4 mm was

determined as previously described (Schæfer,1996).

The granule size distribution was estimated by asieve analysis of a sample of about 100 g drawnby scooping from the cooled fraction finer than 4mm, and the geometric mean diameter (dgw) andthe geometric standard deviation (sg) were calcu-lated. A series of 12 ASTM standard sieves wasvibrated for 10 min by a Fritsch analysette 3vibrator (Fritsch, Germany). Sieves in the rangeof 180–2000 mm were used for all theexperiments.

Size fractions including the three sieve fractionsclosest to the mean diameter were prepared byvibrating for 5 min, and these size fractions wereapplied for the measurements in Sections 2.4.2,2.4.3 and 2.4.4.

2.4.2. Porosity and liquid saturationThe intragranular porosity, e, was calculated

from the equation:

e=6g−6t6g

=1−rg

rt

(1)

where 6g is the volume and rg the density of thegranule sample including all intragranular poresand voids, and 6t is the true volume, i.e. thevolume exclusive of intragranular pores andvoids, and rt the true density of the granulesample. 6g is estimated by a mercury immersion


method (Schæfer et al., 1992a), and rg is calcu-lated from rg=M/6g, where M is the mass ofthe granule sample. rt is either determined by agas displacement pycnometer on a ground gran-ule sample or calculated from the mixing ratioand the true densities of the constituents.

In order to estimate the porosity in the ag-glomeration phase of the process in which themolten binder acts like a liquid, the porosityvalues are corrected for the volume of solidifiedbinder within the intragranular pores and voids.The corrected intragranular porosity, ecorr, isgiven by:

ecorr=e+6bs

6g(2)

where

6bs=Mwb

(1+wb) rbs

(3)

and

6g=6t

1−e=6s+6bs

1−e(4)

where

6s=M

(1+wb) rs

(5)

6bs is the true volume of binder in the solidifiedgranule sample, wb is the ratio of the mass ofbinder to the mass of solid particles, rbs is thetrue density of the solid binder, 6s is the truevolume of the solid particles, and rs is the truedensity of the solid particles.

By inserting Eqs. (3) and (5) into Eq. (4) andthen Eqs. (3) and (4) into Eq. (2), the latterbecomes:

ecorr=e+1−e

1+rbs

rswb

(6)

The calculation of the corrected porosity isbased on the assumption that the total volumeof binder is deposited within intragranular poresand voids.

The liquid saturation of the agglomerates isgiven by:

S=6bm

ecorr6g(7)

where

6bm=Mwb

(1+wb)rbm

(8)

and

6g=6s

1−ecorr

(9)

and where 6bm is the volume of molten binder inthe sample and rbm the density of the moltenbinder at the actual product temperature.

By inserting Eq. (5) into Eq. (9) and thenEqs. (8) and (9) into Eq. (7), the latter becomes:

S=(1−ecorr) wbrs

ecorrrbm

(10)

The calculation of the liquid saturation is basedon the assumption that the packing of the solidparticles within the agglomerates is unaffectedby the solidification of the molten binder duringcooling.

2.4.3. Water contentThe water content on a wet weight basis, i.e.

the water content corresponding to the loss ondrying, was estimated by volumetric titration aspreviously described (Schæfer and Mathiesen,1996a).

2.4.4. Image analysisThe agglomerates were placed on an illumi-

nated desk, and measurements were carried outusing a video camera (MTI CCD72EX, DAGE-MTI, USA) connected to a 55-mm lens (Micro-Nikkor, Nikon, Japan), video digitiser software(Media Pro+/HiRes v2.01+ , Rombo Produc-tions, Scotland, UK), and image analysis soft-ware (Global Lab, Image Version 3.1, DataTranslation Inc., USA); 200 agglomerates wereused for one analysis.

2.4.5. Scanning electron microscopyPhotographs were taken by a scanning elec-

tron microscope (SEM) (Jeol JSM 5200, Japan).


Table 1Physical properties of the meltable binders

Water content Peak temperatureBinder Density of molten binderTrue density Melting range(°C) (°C)(g/cm3)(g/cm3) (%)

90°C70°C 80°C

0.98 0.2 35–44 42Gelucire 50/13 1.120 0.99 0.985546–580.2Stearate 6000 WL 1.05 1.041.191 1.06

164449–57 55PEG 3000 1.225 1.08 1.07 1.07 0.2

2.5. Experimental design

A series of 66 factorially designed experimentswas carried out using either Gelucire 50/13,Stearate 6000 WL 1644 or PEG 3000 as themeltable binder. For each binder, the massingtimes were varied at 1, 3, 5, 10, 15, and 20 min atan impeller speed of 1000 rpm and 1, 3, 5, 10, and15 min at an impeller speed of 1500 rpm. Allexperiments were carried out in duplicate.

The results shown in this paper are mean valuesof two experiments unless otherwise stated. Thedata were analysed by analysis of variance.

3. Results and discussion

3.1. Binder properties

The binder properties are presented in Tables 1and 2. It appears from Table 2 that Gelucire50/13 has the lowest viscosity, Stearate WL 1644the highest viscosity, and PEG 3000 an intermedi-ate viscosity. The loss modulus indicates the dy-namic viscous behavior connected with the abilityto dissipate the energy associated with irreversibledeformation, whereas the storage modulus indi-cates the amount of energy stored elastically upondeformation (Radebaugh et al., 1989). It is seenfrom Table 2 that Stearate 6000 WL 1644 hasdistinct viscoelastic properties since the value ofthe storage modulus is high and larger than thevalue of the loss modulus. Gelucire 50/13 andPEG 3000 show no elastic deformation at all.Although Gelucire 50/13 is less hydrophilic thanStearate 6000 WL 1644 and PEG 3000, all three

binders are primarily hydrophilic. The compari-son of the effects of their rheological propertieson agglomeration is assumed, therefore, not to bemarkedly affected by differences in hydrophilicity.

3.2. Mean granule size and granule sizedistribution

The appropriate binder concentration wasfound by trial and error in preliminary experi-ments. A binder concentration of 23% m/m waschosen independently of the type of binder sincethis gave rise to a similar agglomerate growth.This indicates that different binder volumes haveto be used for different binders in accordance withprevious findings (Thomsen et al., 1994) since thecorresponding binder concentrations expressed as% v/m are 23.5 for Gelucire 50/13, 21.9 forStearate 6000 WL 1644, and 21.5 for PEG 3000 at80°C.

The effects of type of binder and impeller speedon the mean granule size and the granule sizedistribution are presented in Fig. 1. From 5 minof massing, there is a significant effect (pB0.001)of the impeller speed on mean granule size (Fig.1a) but no significant effect of the type of binderin accordance with the fact that different bindervolumes were chosen in order to obtain a similaragglomerate growth. A higher impeller speed in-creases the mean granule size because of a higherdeformability of the agglomerates at collisions(Adetayo et al., 1993; Schæfer and Mathiesen,1996a). Further, the deformability is augmenteddue to a lower viscosity at a higher impeller speed,because a higher impeller speed increases theproduct temperature (Fig. 2a).


Table 2Rheological properties of meltable binders

Gelucire 50/13 PEG 3000Stearate 6000 WL 1644

Viscosity (mPa · s)21741470°C 4915831680°C 3911821790°C 29

460.33Loss modulus (Pa) 1.3

0 166Storage modulus (Pa) 0

The initial agglomerate growth, however, is af-fected by the type of binder. With Stearate 6000WL 1644 and PEG 3000, a higher impeller speedresults in a larger mean granule size at 1 min ofmassing, whereas no effect of impeller speed isseen with Gelucire 50/13. This is reflected in asignificant interaction (pB0.05). Early in an ag-

glomeration process, the risk of comminution ofagglomerates will be higher, because agglomeratesof a rather loose structure are formed. A com-minution is promoted by a higher impeller speedand a lower agglomerate strength. The loosestructure combined with the lower binder viscos-ity of Gelucire 50/13 gives rise to weaker agglom-erates. The higher growth rate to be expected atthe higher impeller speed is counteracted, there-fore, by an increased comminution. Consequently,no effect of impeller speed is seen initially withGelucire 50/13.

Later in the process, and especially at the highimpeller speed, the agglomerate growth becomesmore difficult to control. This is reflected in apoor reproducibility of the mean granule size.From 5 to 15 min of massing, the standard devia-tion estimated by the analysis of variance wasfound to be 102 mm at the low impeller speed and210 mm at the high impeller speed. This explainswhy the differences in agglomerate growth be-tween binders seen in Fig. 1a were not found tobe significant. With the PEG 3000, one of thereplicates at 20 min of massing at the low impellerspeed had to be interrupted owing to formation oflarge balls caused by an uncontrollable agglomer-ate growth. This illustrates the poor reproducibil-ity of the process. Consequently, all the resultswith PEG 3000 at 20 min and 1000 rpm areresults of single experiments.

In accordance with previous findings (Schæferand Mathiesen, 1996b), the initial amount oflumps was found to be relatively high (10–30%),but a marked comminution of lumps occurred byfurther massing. The effect of the impeller speedon the initial amount of lumps was similar to the

Fig. 1. Effects of type of binder and impeller speed on themean granule size (a) and the geometric standard deviation (b)during massing. ( ) Gelucire 50/13, 1000 rpm; (�) Gelucire50/13, 1500 rpm; (× ) Stearate 6000 WL 1644, 1000 rpm; (�)Stearate 6000 WL 1644, 1500 rpm; (�) PEG 3000, 1000 rpm;(2) PEG 3000, 1500 rpm.


effect on the mean granule size since no effect wasseen with Gelucire 50/13, whereas a higher impellerspeed resulted in more lumps with Stearate WL1644 and PEG 3000.

As can be seen from Fig. 1b, the granule sizedistribution becomes significantly narrower (pB0.001) at prolonged massing, because loose lumpsare broken down. Gelucire 50/13 gives rise to asignificantly wider size distribution (pB0.001) thando the other binders. The lower agglomeratestrength makes the agglomerates produced withGelucire 50/13 more susceptible to comminution.This causes breakdown to occur simultaneously tothe formation of agglomerates resulting in a largervariation of the size of the agglomerates. At pro-longed massing, the agglomerates gain strength bydensification. Consequently, they will be more ableto resist the comminution. Therefore, no differencein size distribution between binders is seen from 5min of massing at the high impeller speed and from15 min of massing at the low impeller speed. Thisindicates that the densification occurs more rapidlyat a higher impeller speed.

3.3. Porosity and liquid saturation

The product temperature increases during mass-ing because of heat of friction (Fig. 2a), a higherimpeller speed giving rise to a significantly higher(pB0.001) product temperature. Previous experi-ments (Schæfer and Mathiesen, 1996a) have shownthat an escape of water of crystallization com-mences at a product temperature between 80 and90°C, and that this will delay the agglomerategrowth because of a formation of small pores in thesurface of the lactose particles. Accordingly, a clearfall in the water content of the agglomerates is seenbetween 3 and 5 min of massing at the high impellerspeed and between 10 and 15 min of massing at thelow impeller speed (Fig. 2b). From 10 min ofmassing, the water content is significantly lower(pB0.001) at the high speed. The evaporation ofwater of crystallization is seen to be unaffected bythe type of binder.

Fig. 3a shows the effects of the type of binder andimpeller speed on the corrected intragranularporosity. Normally, a fall in the intragranularporosity will be seen at prolonged massing, because

Fig. 2. Effects of type of binder and impeller speed on theproduct temperature (a) and the water content (b) duringmassing. ( ) Gelucire 50/13, 1000 rpm; (�) Gelucire 50/13,1500 rpm; (× ) Stearate 6000 WL 1644, 1000 rpm; (�)Stearate 6000 WL 1644, 1500 rpm; (�) PEG 3000, 1000 rpm;(2) PEG 3000, 1500 rpm.

the agglomerates become densified. This might,however, be counteracted by the pore formationcaused by the evaporation of water of crystalliza-tion, resulting in an increase in the porosity(Schæfer and Mathiesen, 1996a). In the presentexperiments, a significant increase (pB0.001) in theporosity is seen at prolonged massing, except forGelucire 50/13 at the low impeller speed (Fig. 3a).Gelucire 50/13 results in a significantly higher(pB0.001) porosity than do Stearate 6000 WL1644 and PEG 3000, which have similar porosities.Since a higher binder concentration promotes adensification of agglomerates due to a larger lubri-cation effect (Schæfer et al., 1990), the higherporosity obtained with Gelucire 50/13 cannot beexplained by the larger binder volume used for thisbinder. Generally, a significantly higher porosity(pB0.001) is obtained at the high impeller speed.


Fig. 3. Effects of type of binder and impeller speed on thecorrected intragranular porosity (a) and the liquid saturation(b) during massing. ( ) Gelucire 50/13, 1000 rpm; (�) Gelu-cire 50/13, 1500 rpm; (× ) Stearate 6000 WL 1644, 1000 rpm;(�) Stearate 6000 WL 1644, 1500 rpm; (�) PEG 3000, 1000rpm; (2) PEG 3000, 1500 rpm.

lower strength promotes comminution. A simulta-neous communition and formation of agglomer-ates delay the densification. Therefore, Gelucire50/13 gives rise to a higher porosity than the moreviscous binders.

Previous melt agglomeration experiments inhigh shear mixers (Schæfer et al., 1992b, 1993b;Knight, 1993) have shown that the liquid satura-tion of the agglomerates has to be about 100% inorder to obtain a marked agglomerate growth bycoalescence. As can be seen in Fig. 3b, the differ-ences in the intragranular porosity are reflected inthe liquid saturation, a higher porosity giving riseto a lower liquid saturation. At the high impellerspeed, a higher porosity was obtained. Conse-quently, a higher impeller speed gives rise to alower liquid saturation. The fall in the liquidsaturation that has to be expected at an increasingporosity is partly counteracted by a thermal ex-pansion of the binder liquid caused by the increas-ing product temperature.

Fig. 4 illustrates that the surface of the agglom-erates becomes smoother at the high impellerspeed. At 1000 rpm, lactose particles can be iden-tified. This indicates that more binder liquid issqueezed to the surface at the high impeller speedin spite of the lower value of the liquid saturation.This documents that the higher porosity and thelower liquid saturation found at the high speedare not due to less densification, but are caused bythe above-mentioned formation of small pores inthe surface.

The higher porosity caused by Gelucire 50/13gives rise to a lower liquid saturation initially inthe process only. This is because different % v/mof the binders are used. Since the liquid saturationhas to be similar in order to obtain a similaragglomerate growth with different binders, thehigher porosity caused by Gelucire 50/13 had tobe compensated for by using a larger volume ofbinder.

3.4. Adhesion

It appears from Fig. 5 that Gelucire 50/13 givesrise to a significantly larger (pB0.001) adhesionof mass to the bowl than do Stearate 6000 WL1644 and PEG 3000. The lower viscosity of Gelu-

Initially, the formation of agglomerates of aloose structure is promoted by a high impellerspeed, because it is more likely for comminutionto occur simultaneously with the formation ofagglomerates at high shear forces. Subsequently,the agglomerates become densified, and a higherimpeller speed causes a higher porosity, becausemore water of crystallization evaporates (Fig. 2b).The pores formed by evaporation of water ofcrystallization are not assumed to affect the ag-glomerate strength since the intragranular packingof the particles is unaffected by these small poresin the particle surface. This means that the ag-glomerates in the present experiments becomestronger due to a denser particle packing, al-though an increasing porosity is observed.

As mentioned above, the agglomerates pre-pared with Gelucire 50/13 have a lower strengthdue to the lower viscosity of that binder. The


cire 50/13 makes the agglomerates more de-formable when hitting the wall. This will cause anincreased contact area between agglomerate andwall resulting in more adhesion. The agglomeratesare deformed to a larger extent at a higher im-peller speed, and a higher speed, therefore, givesrise to a larger adhesion in the case of Gelucire50/13. The higher viscosity of Stearate 6000 WL1644 and PEG 3000 makes the agglomerates lessdeformable. Consequently, these binders give riseto an adhesion that is rather low and independentof the impeller speed.

Fig. 5. Effects of type of binder and impeller speed on theamount of adhesion to the bowl during massing. ( ) Gelucire50/13, 1000 rpm; (�) Gelucire 50/13, 1500 rpm; (× ) Stearate6000 WL 1644, 1000 rpm; (�) Stearate 6000 WL 1644, 1500rpm; (�) PEG 3000, 1000 rpm; (2) PEG 3000, 1500 rpm.

Fig. 4. SEM photographs of agglomerates produced withGelucire 50/13 at 1000 rpm (a) and 1500 rpm (b). Massingtime: 15 min.

3.5. Agglomerate shape

The shape of the agglomerates is expressed asthe aspect ratio, which is the ratio of length towidth (Fig. 6). The type of binder, the impellerspeed, and the massing time were found to havesignificant effect (pB0.001) on the aspect ratio. Alonger massing time and a higher impeller speedresult in more spherical agglomerates. A signifi-cant interaction (pB0.001) between the type ofbinder and the impeller speed was found, because

Fig. 6. Effects of type of binder and impeller speed on theaspect ratio. ( ) Gelucire 50/13, 1000 rpm; (�) Gelucire50/13, 1500 rpm; (× ) Stearate 6000 WL 1644, 1000 rpm; (�)Stearate 6000 WL 1644, 1500 rpm; (�) PEG 3000, 1000 rpm;(2) PEG 3000, 1500 rpm.


the effect of the type of binder is significant onlyat the low impeller speed. At low speed, the mostspherical agglomerates are formed with Gelucire50/13. This is because a lower binder viscositycauses a higher surface plasticity of the agglomer-ates making the rounding of the agglomerateseasier, even at liquid saturations being lower thanthose of the other binders. Since a higher impellerspeed promotes the rounding, the effect of thebinder viscosity on the rounding disappears at thehigh level of impeller speed.

The effect of binder viscosity on the surfaceplasticity cannot explain why Stearate 6000 WL1644 at the low impeller speed gives rise to morespherical agglomerates than does PEG 3000, al-though Stearate 6000 WL 1644 has the highestviscosity. The viscoelastic properties of the Stearate6000 WL 1644 might explain these findings. Aformation of spherical agglomerates is likely to bea result of shear forces generated because of thesliding of the agglomerates across each other. Thisrounding of the agglomerates will be counteractedby a flattening of the agglomerates caused by directimpacts between the agglomerates, between theagglomerates and the rotating impeller, and be-tween the agglomerates and the wall of the bowl.Agglomerates prepared with Stearate 6000 WL1644 might become less flattened at direct impacts,because the binder liquid is able to deform elasti-cally, and because the higher binder viscosity makesthe agglomerates less deformable. Thus, agglomer-ates prepared with Stearate 6000 WL 1644 mightbecome more spherical, because they are more ableto withstand the flattening at direct impacts.

The formation of spherical agglomerates is as-sumed to be a result of complex interactionsbetween flattening and rounding. This might ex-plain why PEG 3000 gives rise to more sphericalagglomerates than Stearate 6000 WL 1644 at 20min of massing, where the effect of surface plastic-ity on the rounding might be more important,because the agglomerates have become moredensified.

4. Conclusions

In a high shear mixer, agglomeration is a result

of the balance between coalescence of agglomeratesand breakdown of agglomerates. A lower binderviscosity will decrease the agglomerate strength andthus promote a comminution of the agglomerates.The present experiments indicate that the agglom-erate growth becomes clearly influenced by com-minution at a binder viscosity being belowapproximately 50 mPa· s, but this viscosity limitwill depend on other variables such as the impellerspeed and the particle size of the solid.

The agglomerate properties might be affected bya low binder viscosity, because a formation ofagglomerates simultaneous to a breakdown ofagglomerates will result in porous agglomerates ofa rather wide size distribution. The higher intra-granular porosity caused by a low binder viscositymight increase the required amount of binder liquidin order to obtain a liquid saturation that issufficiently high for agglomerate growth by coales-cence. The densification of the agglomerates willbecome delayed if a marked comminution occurs,but a densification will gradually increase theagglomerate strength and might finally make theagglomerates able to resist the comminution.

A low binder viscosity seems to cause an in-creased adhesion of mass to the bowl due to ahigher deformability of the mass. More studies offactors influencing the adhesion might be useful,however, since the adhesion is also supposed to beinfluenced by electrostatic charging and by theaffinity of the binder to the bowl.

A lower binder viscosity gives rise to a highersurface plasticity of the agglomerates and conse-quently to more spherical agglomerates. Thepresent results indicate that a binder having vis-coelastic properties might improve the rounding ofthe agglomerates, but there is a need for furtherinvestigations in order to clarify if viscoelasticityplays an important role in agglomerate growth.

Acknowledgements

The authors wish to thank Gattefosse, France,for supplying Gelucire 50/13 and Stearate 6000WL 1644, and Helle Eliasen wishes to thankSøren Hvidt, Roskilde University, Denmark, foruseful discussions.


References

Adetayo, A.A., Litster, J.D., Desai, M., 1993. The effect ofprocess parameters on drum granulation of fertilizers withbroad size distributions. Chem. Eng. Sci. 48, 3951–3961.

Attwood, D., Florence, A.T., 1983. Surfactant Systems. TheirChemistry, Pharmacy and Biology, Chapman and Hall,London, p. 472.

Ennis, B.J., Tardos, G., Pfeffer, R., 1991. A microlevel-basedcharacterization of granulation phenomena. Powder Tech-nol. 65, 257–272.

Iveson, S.M., 1997. Fundamentals of granule consolidationand deformation. PhD thesis, University of Queensland,Australia.

Keningley, S.T., Knight, P.C., Marson, A.D., 1997. An inves-tigation into the effects of binder viscosity on the agglom-eration behaviour. Powder Technol. 91, 95–103.

Knight, P.C., 1993. An investigation of the kinetics of granula-tion using a high shear mixer. Powder Technol. 77, 159–169.

Mazzone, D.N., Tardos, G.I., Pfeffer, R., 1987. The behaviorof liquid bridges between two relatively moving particles.Powder Technol. 51, 71–83.

Peerboom, C., Delattre, L., 1995. Influence of the nature ofthe excipients on the feasibility and the release characteris-tics of granules made by a thermoplastic pelletizationmethod. Proc. 1st World Meet. Pharm. Biopharm. Pharm.Technol., Budapest, pp. 361–362.

Radebaugh, G.W., Babu, S.R., Bondi, J.N., 1989. Characteri-zation of the viscoelastic properties of compacted pharma-

ceutical powders by a novel nondestructive technique. Int.J. Pharm. 57, 95–105.

Schæfer, T., 1996. Melt pelletization in a high shear mixer. VI.Agglomeration of a cohesive powder. Int. J. Pharm. 132,221–230.

Schæfer, T., Mathiesen, C., 1996a. Melt pelletization in a highshear mixer. VII. Effects of product temperature. Int. J.Pharm. 134, 105–117.

Schæfer, T., Mathiesen, C., 1996b. Melt pelletization in a highshear mixer. VIII. Effects of binder viscosity. Int. J.Pharm. 139, 125–138.

Schæfer, T., Holm, P., Kristensen, H.G., 1990. Melt granula-tion in a laboratory scale high shear mixer. Drug Dev. Ind.Pharm. 16, 1249–1277.

Schæfer, T., Holm, P., Kristensen, H.G., 1992a. Melt pelletiza-tion in a high shear mixer. I. Effects of process variablesand binder. Acta Pharm. Nord. 4, 133–140.

Schæfer, T., Holm, P., Kristensen, H.G., 1992b. Melt pelletiza-tion in a high shear mixer. II. Power consumption andgranule growth. Acta Pharm. Nord. 4, 141–148.

Schæfer, T., Taagegaard, B., Thomsen, L.J., Kristensen, H.G.,1993a. Melt pelletization in a high shear mixer. IV. Effectsof process variables in a laboratory scale mixer. Eur. J.Pharm. Sci. 1, 125–131.

Schæfer, T., Taagegaard, B., Thomsen, L.J., Kristensen, H.G.,1993b. Melt pelletization in a high shear mixer. V. Effectsof apparatus variables. Eur. J. Pharm. Sci. 1, 133–141.

Thomsen, L.J., Schaefer, T., Kristensen, H.G., 1994. Pro-longed release matrix pellets prepared by melt pelletization.II. Hydrophobic substances as meltable binders. DrugDev. Ind. Pharm. 20, 1179–1197.

.

effects of binder rheology on melt agglomeration in a high shear mixer

Documents