Powder Technology 211 (2011) 189–198

Contents lists available at ScienceDirect

Powder Technology

j ourna l homepage: www.e lsev ie r.com/ locate /powtec

Density behavior of cohesive granular materials

Rafael Mendez a, Francis S. Romanski b, M. Silvina Tomassone b,⁎a Department of Chemical Engineering, University of Puerto Rico at Mayaguez, Puerto Ricob Department of Chemical and Biochemical Engineering, Rutgers University, 98 Brett Road, Piscataway, NJ 08854, 98 Brett Rd. C-118 Piscataway NJ, 08854, United States

⁎ Corresponding author. Tel.: +1 732 445 4949.E-mail address: silvina@soemail.rutgers.edu (M.S. To

0032-5910/$ – see front matter. Published by Elsevierdoi:10.1016/j.powtec.2010.11.024

a b s t r a c t

a r t i c l e i n f o

Article history:Received 15 June 2010Received in revised form 22 September 2010Accepted 21 November 2010Available online 3 December 2010

Keywords:Particle technologyPowder flowBulk DensityTap Density

A new experimental methodology for the characterization of density in a powder bed utilizing X-ray micro-computerized tomography (micro-CT) was developed to quantify the density fluctuations in three commonpharmaceutical powders (α-Lactose Monohydrate, Lactose 310, and Avicel 102). The method begins by filling anacrylic cylinderwith powder and subsequently subjecting the system to vibrations using amechanical shakerwhilemonitoring the density at predetermined bed heights. Three key parameters were isolated including frequency,amplitude, and number of strokes. It was found that the three powders exhibited different packing rates and finalstates. It was also found that the density increased in the powder bed as a function of the number of taps, frequency,and amplitude. Additionally, a more uniform density profile was achieved by utilizing higher amplitudes. Thecohesiveproperties of the threepowderswere investigatedusing the FT4powder rheometer and correlatedwith theresults foundwith themicro-CT scanner. It was found that changes in densityweremore significant in less cohesivepowders, such asAvicel. As powders increase in cohesion, itwas found thatmoremechanical energywas required toalter the agglomerated powder bed. Additionally, the density at the top of the powder bed was significantly moredense than that at the bottom for Avicel, however, the results were directly opposite for the other more cohesivepowders. The results have indicated that micro-CT may be used as a more comprehensive and higher resolutiontechnique for analyzing the density of powders and provide a unique insight to packing at different powder bedheights.

massone).

B.V.

Published by Elsevier B.V.

1. Introduction

The mechanical properties of granular systems are of greatimportance to many chemical industries including petrochemical,specialty chemical, food sciences, catalysis, and most abundantly, thepharmaceutical industry [1]. Intrinsic properties, such as cohesion anddensity have a direct impact on the performance, flowability, andblending processes commonly used in powder manufacturing.However, cohesion and density are neither simple, nor easy toquantify. Despite the tremendous relevance and wide applicability,the flow properties of powders are not well understood [1,2].Unfortunately, the behavior of granular systems, in particularcohesive granular systems, is intrinsically more complex than theflow of fluid media. Complexities such as yield stress, failure surfaces,the coexistence of multiple constitutive regimes, and a densityhysteresis all add to the complex nature of granular systems.Furthermore, as a direct result of the complex behavior of granularsystems, the experimental methodologies used to observe thesystems are significantly more rudimentary than those used forfluid media. In addition, the computational work in granular system isalso in relative infancy when compared to fluid mechanics [3].

Therefore, there exists a significant gap in the ability to predict notonly the intrinsic properties, such as density and cohesion, but theoverall flowability and behavior of granular systems.

In general, there are two accepted classifications for characterizingthe flowability of powders. Direct methods characterize the granularsystem in a consolidated state, such as in a shear cell, while indirectmethods characterize powder in a flowing, loosely packed state, suchas the measure of the angle of repose [2,4–6]. Additionally, thepacking behavior of powders is typically represented using densitiesand density ratios and is generally used as an additional classificationfor the behavior of granular systems. For example, both the bulk andtap density of powders are individually useful, however, they becomemuch more useful when combined to form flow indices such as theHausner ratio or the Carr index [7]. An example of packing behavior isexhibited in the Hausner ratio (HR), which simply is the ratio of thetap density to the bulk density of the powder. Generally, if theHausner ratio is high, the granular system is assumed to exhibit poorflowability [8]. While the Hausner ratio and Carr index are widelyused both academically and industrially due to the relative ease ofanalysis, they are not easily validated and are subject to much debate[5].

In contrast to traditional methods to estimate the tap density,which rely on measuring the volume and mass of a sample, a novelmethodology utilizing an X-ray CT scanner is proposed. In an X-raymicro-CT scanner, X-rays are used to create a cross section image by

Fig. 1. Particle size distribution.

Table 1Key properties of pharmaceutical excipients.

Material Conditioned bulk densitya

(g/cm3)Cohesion parameter(kPa)

Mean particle size(μm)

Avicel 102 0.381 0.110 132Regularlactose

0.393 0.384 81

Alphalactose

0.629 0.774 9.1

a The conditioned bulk density (CBD) was measured by the FT-4 Rheometer after thepre-treatment. In the pre-treatment, a 45° angle impeller blade was introduced into thepowder column in clockwise direction and moving out in the opposite direction threetimes. The measure cylinder is a split vessel assembly designed to allow a precisevolume of powder to be sampled, by rotating the upper part of the vessel away fromthe lower part. The powder remaining in the lower vessel can be weighed using the FT4in-built balance and the conditioned bulk density can be calculated.

190 R. Mendez et al. / Powder Technology 211 (2011) 189–198

the projection of a thin-beam X-ray through an object, in this case, thepowder bed. The X-rays are then absorbed, scattered, and transmitted.Those that are transmitted through the object at each predeterminedangle are subsequently measured and retained as attenuation data.Specifically, this measured attenuation data is ameasure of the reductionin X-ray intensity that results from the absorption and scattering causedby the granular system. Subsequently, the collected data is summed overthe various angles and reconstructed computationally, thus building a CTimage that represents a cross section of the granular system. Differencesin density are seen as a grey-scale contrast. Since the local rate ofattenuation is directly proportional to the mass density along the givenpath, the output results are a representation of the density of the granularsystem. Experimentally, this system can now be subjected to variousforces, i.e. “tapping” and the density recalculated, eventually giving arepresentative profile of the density as a function of the number of taps,the amplitude, and the frequency of taps. This method of using micro-CThas been used previously to establish density profiles for many powdersystems, most notably, glass spheres, metals, salts, and ceramics [9–12].Unfortunately, thismethodhas seen limited use in the characterization ofpharmaceutical powders [13], of which many are extremely cohesive; infact, themost use of X-ray tomography has been in the analysis of tabletsand pharmaceutical ribbons [13–15].

This work focuses on the development of an experimental methodfor the characterization and understanding of density and densityfluctuations in granular beds as a function of cohesive powderproperties. In particular, the use of X-ray micro-CT scanning wasimplemented and used to study three widely used pharmaceuticalpowders with varied, well established, cohesive properties (α-lactosemonohydrate, lactose monohydrate 310, and Avicel 102) [3,16]. Theeffect of various forms of mechanical energy through the variation offrequency, amplitude, and number of impacts on the density profilesof the powders were established and compared with those obtainedthrough conventional methodologies. The remainder of this paper isorganized as follows: Section 2 describes in detail the methodologyused to incorporate X-ray CT-scanning to study the density of apharmaceutical granular system. Section 3 exhibits the results andincludes a detailed discussion. Section 4 contains the conclusions ofthe study and areas where the technique can be further explored.

2. Materials and methods

2.1. Materials

α-Lactose Monohydrate (Lacto Chem., later referred to as alphalactose), Lactose Monohydrate 310 N.F. (regular lactose, ForemostFarm), and microcrystalline cellulose (Avicel 102; FMC) were used inthis study. Eachwas obtained from Fisher Scientific (USA), and used asreceived. Powders were kept in the same location and at a constanttemperature and relative humidity. The particle size distributions ofthe three powders, as measured using the Beckman-Coulter LS 13-320Laser Diffraction powdermodule are shown in Fig. 1. Selected relevantphysical properties of the three excipients are shown in Table 1; thecohesion parameter and bulk density data were established using theFT4 Powder Rheometer.

2.2. Equipment summary

Particle size analysis was performed using the Beckman-Coulter LS13 320 Laser diffraction particle size analyzer (Beckman-Coulter, Brea,CA). Bulk density and cohesion data were obtained using the FT4Powder Rheometer (Freeman Technology, Worcestershire, UK). Themechanical shaker was constructed using the following components:Function generator fromGlobal Specialties 105-2001 (Wallingford, CT),a Techron5515amplifier (AETechron, Elkhart, IN), a 110Volt, 1/14 Amppower supply amplifier, a PCB 48712 10–10–100 mV/G accelerometer(PCB Group, Inc, Depew, NY), and a Vibration Test Systems VG-100-6

shaker (Vibration Test Systems, Aurora, OH). The transparent acryliccylinder containing the granular material to be tested was constructedwith a 3.8 cm internal diameter and a 3 cm fill height. An image of thesetup is shown in Fig. 2. For operation, a signal of step inputwas given tothe shaker,whichwas amplified by the amplifier. A step signalwas usedto resemble a “tap” as close as possible. A mounted piezoelectric deviceon the base of the shaker was connected to an accelerometer recordingacceleration, displacement, and velocity movement in the shaker. TheSkyScan 1172 high-resolution micro-CT system was used for X-raytomography (SkyScan, Kontich, Belgium).

2.3. Experimental procedure

Initially, the acrylic cylinder was filled to roughly 80% of its totalvolume. It is important to note that the powder was always pouredfrom an identical height to minimize pouring effects. After filling, thesamplewasweighed and recorded. The initial state was scanned usingthe SkyScan 1172 micro-CT. Following the initial scan, the powderwas shaken using the mechanical shaker varied by number of strokes(or “taps”), frequency, and amplitude. Three frequencies were used:2 Hz, 4 Hz, and 8 Hz, as well as, two distinct amplitudes 4 a.u. (3 mm)and 5 a.u. (6 mm). The samples were carefully removed from theshaker and analyzed in the micro-CT after 30, 120, 250, 500, 750,1000, 2000, 4300, and 8000 taps.

The SkyScan 1172 used for scanning was optimized for theappropriate resolution by using a pixel size of 28″ in a 1000×500 pixelmodewith a close proximity, filter-less camera. The reconstruction of theimages was conducted at a grey-scale threshold from 0 to 0.6; the beam

Fig. 2. Mechanical powder shaker setup.

Fig. 4. Cylinder with powder bed. The two solid black lines are the limits of themeasured scanned area.

191R. Mendez et al. / Powder Technology 211 (2011) 189–198

hardening, ringartifacts andpost alignmentswereadjusted toachieve thebest possible images.Abasic schematic of howthe systemworks is shownin Fig. 3. The powder was analyzed from 5 to 23 mm in the micro-CT, animage of the cross section tested is shown in Fig. 4. For all experiments,524 reconstructed micro-CT images were used to measure the powderdensity and powder density variation with respect to height, and thepowder porosity. In addition to the tomography measurements, theheight of thebedwasmeasuredvisually after each set of “taps” in aneffortto simply validate themicro-CT results. An example of the reconstructionof the powder bed produced with the micro-CT and analyzed with theCTAn (CT analyzer) software is shown in Fig. 5. This image is an exampleof a lactose powder bed at a height of 21.065 mm from the bottom of theacrylic cylinder. The image is 2 mm less in radius than the actual crosssection of the cylinder tominimize anypotential electrostatic interactionswith the acrylic walls of the cylinder. The volume of interest (VOI) wasrecorded and calibrated for Hounsfield units (HU), a quantitative unit fordescribing radio-density by calibrating a scale between water and air,with water having a density of 0 HU and air−1000 HU, deionized waterwas scanned initially for calibration. Each cross section was taken andconverted into Hounsfield units, the data points were averaged for theentiremid-section of the bed and used to establish the average density ofthe bed for eachdatapoint for the entire powder bed. The averagedensityof the cross sections at different, predetermined heightswasmeasured tocompute the variation of density with respect to height.

3. Results and discussion

In this section, a discussion of the results obtained for the powderdensity as a function of frequency of taps, amplitude of taps, and total

Fig. 3. Schematic of X-ray computerized microtomography method.

number of taps is presented. In addition, the relative densities of thepowder bed were taken at various heights in the experimentalcylinder to observe not only the overall and average density of thesystem, but also the difference in densities between the top andbottom of the sample itself. As previously described, samples weretaken using the SkyScan 1172 micro-CT scanner. Representativeimages may provide a qualitative understanding of the system. As anexample, Fig. 6 displays representative images from alpha lactosesubjected to zero, 1000, and 8000 taps at 2 Hz and amplitude of 6 mm.It is important to note that the images on the left represent the bottomof the bed (Fig. 6a, c, and e), and the images at the right (Fig. 6b, d, andf), the top of the bed. In the zero-tap, or initial condition sample, thereare significant amounts of agglomerates apparent in the both the topand bottom of the sample, however, the image of the top of the bedhas a clear overall trend exhibiting more void space (dark areas), andthus, a lower overall density than the bottom. After 1000 taps, therewas a noticeable increase in the homogeneity of the samples, and also,

Fig. 5. A reconstructed cross sectional image of regular lactose bed at a height of21.065 mm from the bottom of the acrylic cylinder. The image is 2 mm radius less theactual cross section.

Fig. 6. Example of micro-CT Scan images forα-lactose monohydrate as function of the numbers of taps for frequency 2 Hz and amplitude 5 a.u a) zero taps, top, b) zero taps, bottom,c) 1000 taps, top, d) 1000 taps, bottom, e) 8000 taps, top, f) 8000 taps, bottom.

192 R. Mendez et al. / Powder Technology 211 (2011) 189–198

a closer resemblance between the top and bottom. Finally, after 8000taps, the top and bottom are relatively homogeneous, with a notedhigher packing at the bottom; additionally, all traces of agglomeratesare clearly absent.

3.1. Effect of frequency and amplitude on powder density

The shaking frequency and amplitude are directly related to theenergy applied to the column of powder, and as a result, the more

Fig. 7. Frequency and amplitude effect on powder density for a)αmonohydrate lactose,b) Avicel 102, and c) regular lactose.

193R. Mendez et al. / Powder Technology 211 (2011) 189–198

energy applied, the more compacted the resulting powder bed. It wasobserved, as shown in Fig. 7, that there is a marked increase in densitywith an increase in the number of taps, as well as, shaking frequency.These results are more pronounced at higher amplitudes. It isimportant to note that at higher amplitudes, intuitively, thecompaction occurs with a fewer number of taps. For all threepowders, it is clear that the density increases dramatically after thefirst few hundred taps, and afterwards, the difference in tap densitydecreases very little. However, the regular lactose does increase indensity at a comparatively slower rate than Avicel and alphamonohydrate lactose. The results also show small changes in densityfor alpha lactose at all frequencies at the lower of the two amplitudes,while the more pronounced increase is marked with the higheramplitude of 6 mm. It is hypothesized that alpha lactose, the mostcohesive of the three materials, requires more energy for particlerearrangement and the breakage of any agglomerates. In order todisplay the statistical difference between the independent variablesthe results were tabulated into a series of analysis of variance(ANOVA) tables in order to statistically determine the importance ofeach effect on the individual powders. Tables 2, 3, and 4 are theresulting ANOVA tables measuring the statistical differences of theeffect of amplitude and frequency on the density of the powder.Avicel, a free flowing powder, had a statistically significant effect ofboth amplitude and frequency on the resulting density on the powderat the 95% confidence level; these results are tabulated in Table 2.Similarly, alpha lactose, also had a statistically significant effect ofboth amplitude and frequency on the final density of the powder bed,these results are summarized in Table 3. In contrast, the results fromthe regular lactose indicate that neither the frequency, nor theamplitude were statistically significant at the 95% confidence level;the results are tabulated in Table 4. It is important to note, that theparticle size of regular lactose and that of Avicel are on the same orderof magnitude, as shown by Fig. 1 (81 vs. 132 μm), therefore, it isassumed that the particle size was not the reason for the discrepancyin the packing data.

As previously described, the Hausner Ratio (HR) is the ratiobetween the tap density and the bulk density and is instrumental indetermining the flowability of powders. Fig. 8 shows the HausnerRatio as a function of frequency and amplitude. It can be observed thatfor all recorded frequencies, the HR was noticeably higher when theamplitude was increased to 6 mm from 3 mm, this effect is alsonoticeably lower when the frequency is higher. Therefore, it wasconcluded that the higher the energy input, the higher the resultingHR. Additionally, it was apparent that in this particular group ofexperiments, the amplitude and frequency were qualitatively equal tothe type of excipient used.

3.2. Variation of density with height as a function of frequency,amplitude, and cohesion

Figs. 9 through 14 represent the change in density in the respectivepowder bed as a function of height, frequency, and amplitude. Figs. 9and 10 represent Avicel, Figs. 11 and 12 represent regular lactose, andfinally Figs. 13 and 14 represent alpha lactose. The results read left toright in order of increasing frequency, 2, 4, and 8 Hz, respectively.Additionally, the first set of graphs represents the lower amplitude of4 a.u. or 3 mm, and the latter, 5 a.u. or 6 mm, for a given excipient.Each series represents a different number of individual taps forfollowing values: 0, indicating the initial state, 30, 100, 250, 500, 1000,and finally 2000 taps. Very little difference was observed after 2000taps for Avicel, while noticeable differences were apparent for regularand alphamonohydrate lactose leading to the decision to only includedata up to 2000 taps for Avicel while 8000 taps for the others. Aspreviously mentioned, the particle sizes of the powders in questionwere shown in Fig. 1, where the mean particle diameters were 9.1 μmfor alpha lactose, 81 μm for regular lactose, and 132 μm for Avicel. The

cohesion parameters as determined using the FT4 powder rheometerwere 0.110 for Avicel, 0.384 for regular lactose, and 0.774 for alphamonohydrate lactose. The cohesion parameter was determined by thevalue of the yield locus when the normal stress is equal to zero [17].

Table 2Analysis of variance for Avicel 102.

Source DF SS MS F P

Total 5 0.022216444 0.0044432887Amplitude 1 0.0063374879 0.0063374879 27.123257 0.03495Frequency 2 0.015411646 0.0077058228 32.979473 0.02943Error 2 0.00046731024 0.00023365512

Table 4Analysis of variance for regular lactose.

Source DF SS MS F P

Total 5 0.0079309592 0.0015861918Amplitude 1 0.0039783631 0.0039783631 16.777479 0.05475Frequency 2 0.0034783457 0.0017391728 7.3344076 0.11998Error 2 0.00047425039 0.0002371252

194 R. Mendez et al. / Powder Technology 211 (2011) 189–198

For Avicel, Figs. 9 and 10, the changes in density with respect toheight change significantly as the number of taps increases. Theseresults are observably more pronounced at the lower of the twoamplitudes (4 a.u., 3 mm). Additionally, it was quite apparent that atlower frequencies, specifically 2 and 4 Hz, the density increased morenoticeably the center of the cylinder. In addition, the density appearsmuch more uniform at the upper half of the cylinder, while a loweraverage density is clearly visible in the lower half of the measuredareas. However, at the higher frequency, 8 Hz, there was noticeablyless variability in the density with respect to height. The non-intuitiveresults show that packing occurs from the top to the bottom. It wastherefore hypothesized that the reason for the observed top to bottompacking was that the powder at the upper part of the measured areahad significantly more movement as a function of frequency and thus,it may be rearranged faster. It should also be noted that the effects ofthe higher amplitude (5 a.u., 6 mm) on Avicel, shown in Fig. 10,resulted in a noticeably lower change in the effects of frequency withrespect to Fig. 9. Furthermore, the difference between the curves ismuch less apparent, therefore, for Avicel, the higher amplitudespromote higher and more uniform packing, and in turn, decreasedensity oscillations.

The experiments performed on Avicel were repeated on regularlactose. As shown in Fig. 1, the particle size of the regular lactose isvery similar to that of the avicel (mean size 81 μm, compared to132 μm), however, regular lactose is approximately 3.5 times morecohesive than Avicel (0.384 cohesion parameter compared to 0.110for Avicel). Interestingly, the density with respect to height for regularlactose is the exact opposite of the trend exhibited by Avicel, wherethe bottom half of the cylinder becomes more compacted faster thanthe top of the cylinder. The results remain even after all the tapshave been completed. For all frequencies tested, the results arequalitatively similar, again with themaximumobserved density at thebottom of the cylinder. In addition, the effect of amplitude andfrequency appeared qualitatively similar for regular lactose whencompared to Avicel, indicating that the amount of energy to disruptthe powder bed is evident at both amplitudes. This can further beconcluded by the large difference in the density profiles from theinitial state, through the set of taps concluding after 8000.

Finally, the experiments were repeated using alpha lactose, thesmallest (9.1 μm), and most cohesive of the powders (0.774 cohesionparameter). Figs. 13 and 14 show the results of density with respect toheight as a function of frequency for the two amplitudes 4 a.u. (3 mm),and 5 a.u. (6 mm), respectively. While the density trend, from top tobottom, is in line with the results shown for regular lactose where the

Table 3Analysis of variance for α monohydrate lactose.

Source DF SS MS F P

Total 5 0.06935995 0.0138720Amplitude 1 0.05966048 0.0596605 1084.7024 0.00092Frequency 2 0.00958947 0.0047947 87.17427 0.01134Error 2 0.00011000 5.5002e−05

bottomof the bed has an increase in density before the top, the results atthe lower of the two amplitudes indicate a significant variability inrelationship with the height of the measurement. It was thereforehypothesized, that in the lower amplitude case, it may be possible thatthe amount of energy applied during the shaking was not enough tomove, rearrange, or break particle agglomerates that are a directconsequence of the increased cohesion in the powder. The density at 0taps in both figures showed a different trend, in some cases the highestdensityposition as functionof height changed. Thiswasmorenoticeablefor alphamonohydrate lactose, since thismaterial has a higher cohesion.In addition, it was more difficult to reproduce exactly the same initialdensity behavior, as cohesion was significantly higher, and when thecylinder was filled with highly cohesive material, the initial packingdepends on the air permeability. Additionally, there is very littledifference between the density measurements as a function of thenumber of taps, however, at the higher amplitude, the powder bed wasnoticeably more uniform, and the changes with relation to number oftaps was also more apparent. This further indicates that the effect ofamplitude is significant to the rearrangement and packing of thepowder bed due to the increase amount ofmechanical energy impartedinto the bed.

While cohesion was not the only factor resulting in the vastlydifferent density profiles observed in the aforementioned results, itwas clearly a dominant factor. The other effects that were not takeninto account, yet may have a significant effect on the results are asfollows. First, the particle size, most notably since the alpha lactosewas an order of magnitude smaller than both regular lactose andAvicel, may have played a significant role, particularly in the ability forthe powder to pack tighter. Furthermore, the shape of the particlescould also have an effect on the packing behavior of the particles.Avicel, a noted needle-shaped particle may pack at a different rate, orto a different maximum value than the more uniformly shaped

Fig. 8. Hausner ratio as function of the frequency of shaking.

Fig. 9. Density changes as function of height for Avicel 102 at different frequencies atamplitude 4 a.u.

Fig. 10. Density changes as function of height for Avicel 102 at different frequencies atamplitude 5 a.u.

195R. Mendez et al. / Powder Technology 211 (2011) 189–198

particles of the two lactose powders. Finally, the electrostaticinteractions were assumed to be negligible by viewing a windowinside where wall effects would be significant, however, this would

not stop the charges from propagating to the rest of the bed, andtherefore, may have had an effect on the data. While it was clearly notpossible to control every variable in this system, the measurement

Fig. 11. Density changes as function of height for regular lactose at different frequenciesat amplitude 4 a.u. Fig. 12. Density changes as function of height for regular lactose at different frequencies

at amplitude 5 a.u.

196 R. Mendez et al. / Powder Technology 211 (2011) 189–198

technique clearly established an interesting and consistent set ofresults that are unique, and of much greater resolution than the moretraditional and rudimentary methods commonly used to establish

bulk and tap density. Furthermore, the development of this techniquehas allowed for insight into the different packing behavior of verydifferent pharmaceutical powders.

Fig. 13. Density changes as function of height for α monohydrate lactose at differentfrequencies at amplitude 4 a.u.

Fig. 14. Density changes as function of height for α monohydrate lactose at differentfrequencies at amplitude 5 a.u.

197R. Mendez et al. / Powder Technology 211 (2011) 189–198

198 R. Mendez et al. / Powder Technology 211 (2011) 189–198

4. Conclusions

An X-raymicro-CT scanner, commonly used for biological imaging,was successfully used to calculate bulk and tap densities on threedistinct powders of varying cohesion. Due to non-destructive andwidely customizable setup, the scanner was able to achieve newinsights into the packing and density behaviors of powders bothbefore, and after a series of “taps” characterized by number,frequency, and amplitude. The established bulk densities werecomparable to the more rudimentary techniques of measuring thevolume and weight of a cylinder of powder. The results demonstrateda clear increase in density as a function of increased taps, frequency,and amplitude. It was determined statistically, that for Avicel andalpha lactose that the amplitude and frequency were significant at the95% confidence level, while they were not for regular lactose.Furthermore, it was shown, that for Avicel, a free flowing powder,that the density of the top of the cylinder was non-intuitively higherthan that of the bottom of the bed; the opposite was true for regularand alpha lactose. It was also found that the tap density, anduniformity of the powder increases with increasing amplitude,indicating that a threshold of mechanical energy input is required tomove, rearrange, and break particle agglomerates; leading to a higherand more uniformly dense powder bed.

In summary, the use of micro-CT scanning for observing thepacking behavior of powders offers new insight and new tools for theever-present mystery of powder and granular behavior. While morerudimentary techniques may be cheap and effective, they do little toimprove powder handling techniques. In the future, this methodologywill help increase the knowledge of powder behavior as not just afunction of cohesion shown here, but also of particle size, shape,electrostatic properties, and various others.

Acknowledgements

In addition, the authors would like to acknowledge the support of theNSF—Grant number 0553819 “Spatial and Temporal Behavior of FlowingCohesive Powders” and also the partial support from NSF-ERC-SOPSCenter.

References

[1] F.J. Muzzio, T. Shinbrot, B.J. Glasser, Powder technology in the pharmaceuticalindustry: the need to catch up fast, Powder Technology 124 (1–2) (2002) 1–7.

[2] PRESCOTT, et al., On powder flowability, 24, MN, ETATS-UNIS: AdvanstarCommunications, Duluth, 2000.

[3] A. Faqih, et al., An experimental/computational approach for examiningunconfined cohesive powder flow, International Journal of Pharmaceutics 324(2) (2006) 116–127.

[4] D. Geldart, E.C. Abdullah, A. Verlinden, Characterisation of dry powders, PowderTechnology 190 (1–2) (2009) 70–74.

[5] A. Santomaso, P. Lazzaro, P. Canu, Powder flowability and density ratios: theimpact of granules packing, Chemical Engineering Science 58 (13) (2003)2857–2874.

[6] D. Geldart, et al., Characterization of powder flowability using measurement ofangle of repose, China Particuology 4 (3–4) (2006) 104–107.

[7] E.C. Abdullah, D. Geldart, The use of bulk density measurements as flowabilityindicators, Powder Technology 102 (2) (1999) 151–165.

[8] K. Thalberg, D. Lindholm, A. Axelsson, Comparison of different flowability tests forpowders for inhalation, Powder Technology 146 (3) (2004) 206–213.

[9] X-ray computerized tomography for density gradients in powder compacts: M.J.P.Wang et al (University of Toledo, Ohio, USA), Metal Powder Report 48 (11) (1993)40-40.

[10] E.R. Nowak, et al., Density fluctuations in vibrated granular materials, PhysicalReview E 57 (2) (1998) 1971–1982.

[11] D.H. Phillips, J.J. Lannutti, Measuring physical density with X-ray computedtomography, NDT & E International 30 (6) (1997) 339–350.

[12] R. Heldele, et al., X-ray tomography of powder injection moulded micro partsusing synchrotron radiation, Nuclear Instruments and Methods in PhysicsResearch Section B: Beam Interactions with Materials and Atoms 246 (1)(2006) 211–216.

[13] M.S. Kadiri, A. Michrafy, J.A. Dodds, Pharmaceutical powders compaction:experimental and numerical analysis of the density distribution, PowderTechnology 157 (1–3) (2005) 176–182.

[14] I.C. Sinka, et al., Measurement of density variations in tablets using X-raycomputed tomography, International Journal of Pharmaceutics 271 (1–2) (2004)215–224.

[15] A.M. Miguélez-Morán, et al., Characterisation of density distributions in roller-compacted ribbons using micro-indentation and X-ray micro-computed tomography,European Journal of Pharmaceutics and Biopharmaceutics 72 (1) (2009) 173–182.

[16] A.W. Alexander, et al., Avalanching flow of cohesive powders, Powder Technology164 (1) (2006) 13–21.

[17] R. Freeman, Measuring the flow properties of consolidated, conditioned andaerated powders—a comparative study using a powder rheometer and arotational shear cell, Powder Technology 174 (1–2) (2007) 25–33.