particle aerosolisation and break-up in dry powder inhalers: evaluation and modelling of the...

Particle Aerosolisation and Break-up in Dry Powder Inhalers:Evaluation and Modelling of the Influence of Grid Structuresfor Agglomerated Systems

WILLIAM WONG,1 DAVID F. FLETCHER,2 DANIELA TRAINI,1 HAK-KIM CHAN,1 JOHN CRAPPER,3 PAUL M. YOUNG1

1Advanced Drug Delivery Group, Faculty of Pharmacy, University of Sydney, Sydney, New South Wales 2006, Australia

2School of Chemical and Biomolecular Engineering, University of Sydney, Sydney, New South Wales 2006, Australia

3Pharmaxis Ltd., Sydney, New South Wales 2086, Australia

Received 2 February 2011; revised 11 April 2011; accepted 20 May 2011

Published online 21 June 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.22663

ABSTRACT: This study aimed to investigate the influence of grid structures on the break-up and aerosol performance of a model inhalation formulation through the use of standardisedentrainment tubes in combination with computational fluid dynamics (CFD). A series of entrain-ment tubes with grid structures of different aperture size and wire diameters were designedin silico and constructed using three-dimensional printing. The flow characteristics were sim-ulated using CFD, and the deposition and aerosol performance of a model agglomerate system(496.3–789.2:m agglomerates containing 3.91 :m median diameter mannitol particles) wereevaluated by chemical analysis and laser diffraction, respectively. Analysis of the mannitol re-covery from the assembly indicated that mass deposition was primarily on the grid structurewith little before or after the grid. Mass deposition was minimal down to 532:m; however, forsmaller grid apertures, significant blockage was observed at all airflow rates (60–140 L·min−1).Analysis of the particle size distribution exiting the impactor assembly suggested that mannitolaerosolisation was dependent on the void percentage of the grid structure. It is proposed thatinitial particle–grid impaction results in a shearing force causing agglomerate fragmentationfollowed by immediate re-entrainment into the turbulent airstream within the grid apertureswhich causes further dispersion of the fine particles. Such observations have significant impli-cations in the design of dry powder inhaler devices. © 2011 Wiley-Liss, Inc. and the AmericanPharmacists Association J Pharm Sci 100:4710–4721, 2011Keywords: CFD; dry powder inhaler; impaction; agglomerate; deagglomeration; aerosols;in silico modelling; pulmonary drug delivery; simulations; particle size

INTRODUCTION

There are a wide range of dry powder inhalation (DPI)devices available in the market or under develop-ment, which utilise different mechanisms to dispersethe micron-sized active pharmaceutical ingredients(APIs).1–3 For effective delivery of dry powder par-ticulates to the respiratory tract, for the treatmentof local and systemic disease states, primary drugparticles need to have an aerodynamic diameter lessthan approximately 5:m.4 However, API drug parti-cles within this size range have a high-surface area

Correspondence to: Paul M. Young (Telephone: +61-2-9036-7035; Fax: +61-2-9351-4391; E-mail: [email protected])Journal of Pharmaceutical Sciences, Vol. 100, 4710–4721 (2011)© 2011 Wiley-Liss, Inc. and the American Pharmacists Association

to mass ratio which induces high cohesive–adhesiveforces between contiguous surfaces within the formu-lation. This is reflected in the relatively low aerosolefficiencies in conventional DPI formulations, witha large proportion achieving fine particle fractions(FPFs: the percentage dose of API with an aerody-namic diameter suitable for inhalation therapy) ofless than 30%.2 Various approaches have been under-taken to improve the ability to aerosolise these APIdrug particles, the majority of which are focused onmodifying the formulation or the device design.

Although there are many formulation variableswhich can be altered to achieve adequate levels ofdrug delivery to the lung, two formulation meth-ods have been primary utilised: carrier-based andagglomeration-based systems.5 These methods are

4710 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011

EVALUATION AND MODELLING OF THE INFLUENCE OF GRID STRUCTURES FOR AGGLOMERATED SYSTEMS 4711

used to ensure efficient entrainment of the API intothe airstream whilst providing a means of sampledilution (when small microgram range doses are re-quired). Despite this, aersosol performance is still rel-atively poor. Subsequently, significant research hasbeen undertaken at both the fundamental and empir-ical level to understand the complex processes drivingparticle deagglomeration and aerosolisation withindifferent dry powder formulations.

The diverse range of DPI devices reflects themultitude of approaches undertaken to enhanceaerosol performance; however, little research hasbeen conducted to study the fundamental mecha-nisms of break-up. In general, these systems havebeen optimised through performance modificationand empirical study designs. Fundamental studieswhich have been conducted often utilise entrain-ment tubes incorporating dispersing apparatuses;however, these studies generally focussed on theaerosolisation of micron-sized drug particles fromcarrier-based formulations rather than the break-up of agglomerate-based systems, containing micron-sized primary particles.6–8 In some cases, agglomer-ate systems have been studied using this approach9;however, most of these studies have been cross-disciplinary (e.g. in the printing or minerals indus-try); hence such studies have limited relevance due tothe large variation in the materials and particle sizedistributions.

Recent advancements in the performance and ac-cessibility of high-end computing facilities have re-sulted in the increasing use of computer modellingto investigate particle fluidisation and deagglomera-tion. Computational fluid dynamics (CFD) is a well-established methodology utilised to predict fluid flowand has been applied to predict the flow behaviourin devices used for inhalation,10–15 and flow in therespiratory tract.16,17 Although these studies gave in-sight into DPI aerosolisation, it is difficult to elucidatethe role of specific physical mechanisms taking placedue to the multitude of variables operating within aconventional DPI. Discrete element modelling (DEM)has also been increasingly used to simulate particledeagglomeration18–20 where interactions and veloci-ties of the particles within the agglomerate can bepredicted over a period of time. However, this tech-nique is generally in vacuum, and hence does not ac-count for fluid flow. DEM is also computationally ex-pensive when studying agglomerates which containover 1,000,000 particles in a micron size range. Even-tually, the coupling of CFD with DEM will enableaccurate prediction of the entire aerosolisation pro-cess; however, current limitations in computationalpower have limited studies to small numbers of largeparticles.21,22

In order to study the aerosolisation process inagglomerate-based DPI systems, the authors have

undertaken a series of studies to evaluate how astandard formulation behaves with respect to spe-cific deagglomeration mechanisms (i.e. airflow, turbu-lence, impaction, etc.). Previous studies have utiliseda combination of CFD and experimental entrainmenttubes to study the break-up and aerosolisation of amodel agglomerate system (containing micron-sizedmannitol particles) as a function of airflow, turbu-lence level and impact force against the surface ofan entrainment tube.23,24 The initial study utilised aseries of venturi tubes to induce turbulent flow withcharacteristics equivalent to commercial DPI deviceswhilst minimising other potential break-up mecha-nisms (such as wall or grid impaction), whereas thesubsequent study utilised a series of impactors withdifferent impaction angles to induce agglomerate col-lision upon the impact surface whilst reducing the ef-fect of break-up in turbulent flow. Interestingly, thesestudies found agglomerate impaction to play a moredominant role in the aerosol performance. However,this was more complicated than impact force alonebecause upon impact particle mass was depositedon the surface, where subsequent airflow directly re-entrained the particles within the airstream.

To further investigate the mechanism of break-up and aerosolisation in agglomerate-based DPI sys-tems, the influence of impaction against a grid struc-ture was studied. Grid structures are commonly usedin commercial DPI devices, and unlike the impaction“plate” or surface, particle mass is less likely to bedeposited on the surface of the grid upon agglomer-ate impaction. A series of entrainment tubes contain-ing grids of varying wire diameter and aperture sizeswere designed. The flow behaviour through these en-trainment tubes were evaluated using CFD analysisand subsequently compared with physical aerosol anddeposition measurements to ascertain the influence ofthe grid structure on the aerosolisation mechanism.

MATERIALS AND METHODS

Materials

Spray-dried micron-sized primary mannitol particles(batch number M08-060) were obtained from Phar-maxis Ltd. (Sydney, New South Wales, Australia).Water was purified using reverse osmosis (MilliQ,Molsheim, France), and all organic solvents wereobtained from Sigma (Sydney, New South Wales,Australia) and were of analytical grade.

Preparation of a Model Particulate System

Agglomerates were formed by subjecting primarymannitol particles to Turbula mixing (Bachofen AGMaschinenfabrik, Basel, Switzerland), for 15 min at42 rpm in a custom-built aluminium cylinder 25 mmin diameter and 36 mm in length. A fraction of

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011

4712 WONG ET AL.

agglomerates between 500 and 800 :m was producedfor subsequent testing by passing the agglomeratesthrough a nest of sieves (ISO 3310-1 test sieves,Endecotts Ltd., London, UK). The separated agglom-erates were stored in sealed containers at 45% rela-tive humidity (RH) and 25◦C for a minimum of 48 hprior to their use.

Physical Characterisation of the Primary MannitolParticles and Agglomerated Systems

The methods and results of the physical char-acterisation of the primary mannitol powder andagglomerated systems have been reported in de-tail elsewhere.23 In general, the primary manni-tol powder and agglomerated systems were charac-terised in terms of particle size, morphology, den-sity, mass, and surface area. The primary particlesize distribution was determined using laser diffrac-tion (Malvern Mastersizer 2000; Malvern, Worces-tershire, UK) by suspending the particles in chlo-roform. A refractive index of 1.52 for mannitoland 1.44 for chloroform was utilised for the parti-cle size measurements. Agglomerate particles werevisualised using optical microscopy (CX41 micro-scope with DP12 digital camera; Olympus, Tokyo,Japan). Samples were dispersed on a glass slideand the resulting image was captured using an as-sociated charge-coupled device (CCD) camera andsoftware. Captured images were post-processed us-ing image analysis software (Image-J; NationalInstitute of Mental Health, Bethesda, Maryland)to determine particle area and diameter parameters,from which the agglomerate size distribution was cal-culated. The density of the primary mannitol parti-cles was determined using helium pycnometry at 27◦C(Accupyc 1340 Gas Pycnometer; Micromeritics, Nor-cross, Georgia) and the surface area of the primarymannitol particles was determined by nitrogen ad-sorption at 77 K (Tristar II 3020; Micromeritics). Inaddition, the agglomerate mass was measured usinga seven figure Cahn microbalance, (DVS-1; SurfaceMeasurement Systems Ltd., London, UK), from whichvalues for the primary particle number and agglom-erate density were calculated for use in subsequentcomputational studies and in silico simulation.

Construction of the Grid Assembly

An entrainment tube incorporating a stainless steelgrid prior to the exit was designed with computer-aided design software (ANSYS DesignModeler 12.0;ANSYS, Canonsburg, Pennsylvania) and constructedfrom acrylonitrile butadiene styrene (ABS) using arapid prototyping three-dimensional printer (Dimen-sion Elite, Dimension Inc., Eden Prairie, Minnesota).The tube was constructed of 290 mm long interlock-ing sections with an inner diameter of 19 mm anda 280 mm long final section incorporating the grid

Figure 1. Schematic of full grid assembly (a) and of thegrid insert (b).

structure. This final section comprised upper andlower interlocking components between which in-terchangeable grid inserts of varying aperture sizescould be positioned. When assembled, the total tubelength was 2.20 m with the grid structure located150 mm upstream of the outlet (Fig. 1a). Six wo-ven stainless steel grids (Star Screens Australia,Sydney, New South Wales, Australia) of specificationsgiven in Table 1 were cut to size and sandwiched be-tween two interlocking ABS discs constructed usingthree-dimensional printing to form the grid inserts(Fig. 1b).

CFD Analysis

A commercially available CFD code, ANSYS CFX12(ANSYS), was used to simulate the flow of air at25◦C through the grid assembly to predict turbulenceproperties around the grid structure, and to track

Table 1. Grid Specifications

Aperture Size(:m)

Wire Diameter(:m)

Thickness(:m)

Void percentagea

(%)

1999 550 1010 621028 550 870 42741 320 600 49532 310 550 40310 200 420 37150 100 220 36

aVoid percentage is a measure of the open area within the gridstructure as a percentage of the total cross-sectional area of theentrainment tube.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps


particles through the flow field. To reduce compu-tational expense, the flow field was modelled for asection of the mesh containing four holes, with sym-metry planes set at the boundaries. The Reynolds-averaged Navier Stokes equations were used to de-termine the flow field throughout the grid assemblyand the shear stress transport model with scalablewall functions was used to model turbulence. A finitevolume method based on a hexahedral mesh with in-flation at the walls was used to capture the boundarylayer behaviour correctly.

Mesh independence analysis was conducted to en-sure that the computational results were independentof mesh size. This was achieved by studying the ve-locity profile along a line normal to the grid, passingthrough the centre of the grid aperture, to comparethe spreading of the jet-exiting aperture. The resolu-tion of the CFD mesh was doubled until no change inthe velocity profile was observed.

In-Line Aerosol Particle Size Analysis

The particle size distribution of the mannitol pow-der emitted from the grid assembly was measuredusing laser diffraction to evaluate the break-up ofthe model agglomerate system for different grid aper-tures at different flow rates. The apparatus (Fig. 1a)was mounted vertically on a scaffold with the exit portconnected in-line to a Spraytec particle sizer (MalvernInstruments, Malvern, UK). A Gast Rotary vain pump(Erweka GmbH, Heusenstamm, Germany) was con-nected to the outlet of the particle sizer and the flowwas calibrated using a flow metre (TSI 3063; TSIinstruments Ltd., Buckinghamshire, UK). For eachmeasurement, 50 mg of the agglomerated powder wasintroduced into the centre of the airstream at the topof the grid assembly using a funnel.

The geometric particle diameter distribution at theoutlet was measured in real time using the Spraytecsoftware at a data collection rate of 2500 sweeps persecond over a 0.1–2000:m size range. Three volumet-ric flow rates (60, 100, and 140 L·min−1) and six gridaperture sizes (1999, 1028, 741, 532, 310, and 150:m)were studied in triplicate. After each experiment, thegrid assembly was disassembled, washed with 5 mLof purified water and analysed using a validated high-performance liquid chromatography (HPLC) methodand system described elsewhere to determine themannitol deposition within the assembly.25 In gen-eral terms, a LC20AT pump, SIL20AHT autosam-pler, CBM-Lite system controller with a pc-computerrunning LC solution v1.22 software and a RID-10Arefractive index detector (Shimadzu, Sydney, NewSouth Wales, Australia) were utilised at an operat-ing flow rate of 1 mL·min−1. An 8 mm Resolve C18Radial Pack chromatography Cartridge (Waters AsiaLtd., Singapore) was used for separation. After sam-

ple recovery, the entire grid assembly was washedwith ethanol and air dried prior to re-assembly.

Statistical Analysis

Experimental data were analysed for significancewith one-way analysis of variance (with Tukey’s post-hoc analysis) using SPSS Statistics 17.0 (SPSS Inc,Chicago, Illinois). A difference was considered signif-icant when p < 0.05.

RESULTS

Physical Properties of the Primary Mannitol Particles

Analysis of the size distribution by laser diffraction(mean of triplicates ±standard deviation) indicateda lognormal distribution with 90% of the particleshaving a volume diameter ≤6.82 ± 0.37:m and 10%≤1.95 ± 0.02:m. The median (d0.5) particle diame-ter of the primary particles was 3.91 ± 0.15 :m. Thisvalue was used for theoretical calculations of the ag-glomerate structure as outlined in a subsequent sec-tion. Such observations are in good agreement withvalues published in previous studies.25 The true den-sity of the primary mannitol particles was 1435.2 ±0.6 kg·m−3, as measured by helium pycnometry. Thisis also in good agreement with values published inprevious studies.26 The surface area of the primarymannitol particles was 2.606 ± 0.026 m2·g−1, as mea-sured by nitrogen adsorption and calculation usingthe BET method.

Physical Properties of the Mannitol Agglomerates

The mannitol agglomerates utilised in this study havepreviously been investigated extensively.23 Using op-tical microscopy in conjunction with image analy-sis, the particle diameter of 300 agglomerates wasmeasured from eight fields of view. The count-basedsize distribution was determined by grouping the sizedata into eight successive size intervals (Fig. 2). Lin-ear regression of the particle size distribution over a5%–95% cumulative percentage indicated a normaldistribution (R2 = 0.993) with 90% of the particleshaving a diameter of less than 759.9 :m and 10% ofthe particles having a diameter of less than 525.6:m. The median (d0.5) agglomerate diameter was642.8 :m. The agglomerates were also weighed ona Cahn microbalance and the mean mass was mea-sured to be 91.1 ± 22.1 :g (n = 25). From the meanmass of the agglomerate, in conjuction with the me-dian primary mannitol particle diameter and density,an agglomerate density of 655 kg·m−3 was calculated.

Mannitol Deposition Within the Grid Assembly

As this study investigates the effect of a grid onthe break-up of agglomerates, it is expected that a


4714 WONG ET AL.

Figure 2. Size distribution of mannitol agglomerates.

substantial amount of mannitol would be depositedon the grid section. To examine this, the grid assem-bly was disassembled into the upper section, lowersection and grid insert (as outlined in Fig. 1a) aftereach experiment and mannitol deposition on each sec-tion was measured using HPLC. For all flow ratesand grid aperture sizes, minimal mannitol depositionwas observed on the upper section, upstream of thegrid, and on the lower section, downstream of the grid(<0.3 and <2.5 mg, respectively). The mannitol de-posited on the grid insert for each grid aperture sizeat 60, 100 and 140 L·min−1 is shown in Figure 3. Theamount of mannitol deposited on the grid insert ap-pears to be independent of the flow rate for grid aper-tures of 1999, 1028 and 741 :m (2–3.5 mg), whereasfor the 532:m grid aperture, the amount of man-nitol deposited decreased with increasing flow rate(19.6 mg at 60 L·min−1, 6.0 mg at 100 L·min−1 and 3.5mg at 140 L·min−1). As described previously, morethan 90% of all mannitol agglomerates have a diam-eter exceeding 525.6 :m. As such, for a grid aper-ture size of 532:m, it is likely that any agglomeratesinsufficiently fragmented upon impact with the gridstructure would remain trapped above the grid. Athigher flow rates, agglomerates impact the grid struc-ture with greater force, resulting in more agglomer-ate break-up and a reduction in mannitol depositedabove the grid. Significant deposition was observedfor grid apertures of 310 and 150:m (>50% of ag-glomerates introduced), as the size of the mannitolagglomerates markedly exceeded the aperture size,resulting in blockage and overloading of the grid. Assuch, grid apertures of 310 and 150 :m were excludedfrom subsequent analysis.

Figure 3. Mannitol deposition on the grid.

CFD Analysis of the Grid Assembly

The turbulence kinetic energy across the centre planeof a grid aperture for the 1999 and 532 :m grids at140 L·min−1 are shown in Figure 4. In both cases, theturbulence kinetic energy was greatest in the regionsdirectly behind the rigid grid structures and aroundthe edge of the potential core, generated from the jetflowing out of the grid structure, due to large veloc-ity gradients present within this region. As the jetfrom the grid aperture mixes with the bulk flow, theturbulence kinetic energy also falls. Minimal turbu-lence kinetic energy was generated upstream fromthe grid and within the grid structure. As expected,greater levels of turbulence kinetic energy were gen-erated within the smaller 532:m grid; however, theturbulence kinetic energy decreases faster due to the



Figure 4. Turbulence kinetic energy across the centre plane of a grid aperture for the 1999:m(a) and 532:m (b) grid apertures at 140 L·min−1. (Note: Different scales for turbulence kineticenergy is used in each case.)

smaller jets generated by the grid and rapid mixingdownstream.

The volume-averaged turbulence kinetic energyand integral shear strain rate from the centre plane ofthe grid to various distances downstream of the grid,for grid apertures of 1999 and 532:m, are shown inFigure 5. The integral shear strain rate is a mea-sure of the potential of a turbulent flow to break-uppowder agglomerates, and was defined and utilisedprevious by Coates et al.13 Turbulent flows of higherintegral shear strain rates will exert a greater aero-dynamic shearing force upon an agglomerate. In gen-eral, both volume-averaged turbulence kinetic energyand integral shear strain rate increased with increas-ing flow rate and a reduction in grid aperture size.Volume-averaged turbulence kinetic energy peaked atapproximately 7 mm downstream from the grid beforegradually reducing further downstream for all cases,whereas the volume-averaged integral shear strainrate decreased rapidly immediately downstream ofthe grid for all cases.

Figure 6 depicts three likely scenarios that resultin either the agglomerate or fragments of the agglom-erate passing through the grid. Position a representsan agglomerate passing wholly through the centreof the grid aperture. Positions b and c represent anagglomerate impacting partially on the grid, follow-ing which fragments and/or the agglomerate wouldpass through the grid aperture in close proximity tothe edge of the grid structure. As such, it is impor-

tant to determine the intensity of turbulence withinthese regions, which may contribute to post-impactionbreak-up of the agglomerate and/or its fragments. Fig-ure 7 shows the turbulence kinetic energy and inte-gral shear strain rate measured along lines positionedwithin the afore-mentioned regions for all grid aper-ture sizes at 140 L·min−1. Measurement lines weredrawn from the centre plane of the grid in a down-stream direction normal to the flow. To represent po-sition a, a line was drawn at the centre of the gridaperture (defined as the “middle” in Fig. 7). To rep-resent position b, a line was drawn 10% of the gridaperture width inwards from the midpoint of the gridaperture edge (defined as the “side” in Fig. 7). To rep-resent position C, a line was drawn 10% of the gridaperture width inwards from both grid aperture edgesat the corner (defined as the “corner” in Fig. 7). Tur-bulence kinetic energy and integral shear strain rateswere observed to be highest at the corner region andlowest in the middle region for each grid aperture size.In general, turbulence kinetic energy increased withdecreasing void percentage, which is a measure of theopen area within the grid structure as a percentageof the total cross-sectional area of the entrainmenttube (refer to Table 1). The 741 :m grid aperture ac-tually has a larger void percentage than the 1028 :mgrid aperture. As such, there is a larger proportion ofthe cross-sectional area which is open, and thus, theturbulence kinetic energy is lower. Maximal turbu-lence kinetic energy along the corner region occurred


4716 WONG ET AL.

Figure 5. Volume-averaged turbulence kinetic energy and integral shear strain rate down-stream of the grid for aperture sizes of 1999:m (a and b, respectively) and 532 :m (c and d,respectively).

between 1 and 5 mm downstream of the grid andgradually reduced further downstream. As the gridaperture size was reduced, the turbulence kinetic en-ergy along the side and middle region peaked closerto the grid structure. The integral shear strain rateincreased with decreasing grid aperture size. Mini-mal integral shear strain rate was calculated in themiddle region for all grid aperture sizes. With the ex-ception of the 1999:m grid, the integral shear strainrate peaked rapidly in the side and corner regionswithin the grid structure for all grid aperture sizes,and decays rapidly once outside of the grid structure.The integral shear strain rates measured for the 1999:m grid aperture were substantially lower than theother grid aperture sizes.

Aerosolisation Performance

The aerosolisation efficiency of the different grid aper-tures can be described by the aerodynamic size dis-tribution measured at the outlet of the grid assembly.Figure 8 shows size distributions of mannitol parti-cles emitted after passing through grid apertures of1999 and 532:m at 60 and 140 L·min−1, respectively,in both cases. Analysis of these data suggests a multi-modal distribution for all data sets. The higher value

peak may be attributed to the primary agglomeratesize distribution in the case of the 1999:m grid aper-ture, whereas in the 532:m grid aperture, the higherpeak corresponds to large agglomerate clusters of asize similar to the grid aperture, passing through thegrid after impaction. A central region spanning fromaround 20 to 100:m is present in all cases and islikely to represent particle clusters, fractured fromthe main agglomerate upon impaction. Interestingly,a lower peak spanning from 3 to 20:m is observedat 140 L·min−1 for the 1999 :m grid aperture and atboth 60 and 140 L·min−1 for the 532 :m grid aper-ture. This region, which corresponds to the primarymannitol particles and small multiplets, was observedto increase in magnitude with increasing flow ratesand decreasing grid aperture sizes, whereas the peakat larger sizes reduced in magnitude. This reflectedan improved aerosolisation efficiency and break-upof the primary agglomerate. It is also important tonote here that the data were represented as a conven-tional (pharmacopeia) volume distribution, and thus,a weighting towards the larger particle diameters willalways exist (as one 500 :m sphere will have thesame volume as 1 million 5 :m spheres of equivalentdensity).



Figure 6. Schematic representation of the interaction be-tween agglomerates and the grid structure. (a) Agglomer-ate passes through the centre of the grid aperture with-out impacting it. (b) Agglomerate impacts one edge of thegrid structure and passes through the grid aperture in closeproximity to the edge. (c) Agglomerate impacts two edges ofthe grid structure and passes through the grid aperture inclose proximity to the corner.

To visualise the relationship between grid aperturesize and agglomerate break-up, the particle d0.1 wasplotted as a function of grid aperture size for all flowrates (Fig. 9a). In general, a reduction in grid aperturesize leads to increased agglomerate break-up, whichis reflected in a reduction of d0.1. As expected, increas-ing flow rate also resulted in increased agglomeratebreak-up. It is important to note that the 741:m gridaperture size was an exception to this trend, wherebythe next larger grid aperture (1028 :m) displayedmore efficient agglomerate break-up at all flow rates,reflected by the lower d0.1 values. However, as shownin Table 1, the 741:m grid aperture actually has alarger void percentage than the 1028:m grid aper-ture. As previously described, the void percentage isa measure of the open area within the grid structureas a percentage of the total cross-sectional area ofthe entrainment tube, and represents the likelihoodof agglomerate impacting the grid structure. As the741:m grid aperture has a larger void percentagethan the 1028:m grid aperture, there is a large pro-portion of the cross-sectional area which is open, andthus, it is less likely for an agglomerate to impactupon the grid despite having a smaller aperture sizethan the 1028:m grid aperture.

When particle d0.1 was plotted as a function of thevoid percentage for all grid aperture sizes at all flowrates (Fig. 9b), it can be seen that as the void per-centage is reduced, the particle d0.1 decreases, reflect-

ing an improved efficiency in agglomerate break-up.In addition, agglomerate break-up was more effectiveat higher flow rates. At both 60 and 100 L·min−1,the reduction of d0.1 resulting from decreasing voidpercentages was found to be statistically significant.However, no statistically significant difference wasobserved between the d0.1 for all void percentages at140 L·min−1.

DISCUSSION

This study demonstrated that effective aerosolisationcan be achieved by utilising grid structures in a DPI.As the grid voidage or void percentage was reduced,an increase in aerosol performance and agglomeratebreak-up was observed. However, the mechanism bywhich a grid structure enhances agglomerate break-up involves a complex interaction between impactionupon the grid structure and turbulent shear flow. Ifthe grid apertures are too large, such as in the caseof the 1999:m grid aperture size, agglomerates canpass through the centre of the aperture without im-pacting upon the grid structure. Such agglomerateswill encounter minimal forces acting to break themup, as the turbulence kinetic energy and integralshear strain rates in the centre of the grid apertureare small (Figs. 4 and 7). Agglomerates which im-pact upon the grid are likely to fracture and producefragments or multiplets which will subsequently bere-entrained in close proximity to the edges of thegrid structure, entering into regions of high integralshear and turbulence kinetic energy, which act to fur-ther break-up these fragments. Increasing the flowrate results in greater overall energy within the sys-tem, whereby agglomerates would impact upon thegrid structure with greater force, be re-entrained intohigher velocity flow fields and encounter stronger tur-bulent shear.

Reducing the void percentage increases the likeli-hood of agglomerate impaction with the grid struc-ture; however, this involves a consideration of aper-ture size and grid wire diameter. As demonstratedin the case of the 741:m grid aperture size, whichwhilst having a smaller grid aperture than the 1028:m case also had a smaller wire diameter (320:mcompared with 550:m). This resulted in a void per-centage which was larger than that for the 1028:mgrid aperture size, and as such, a lower likelihood ofagglomerates impacting upon the grid structure. As-suming the wire diameter is kept constant, the voidpercentage can be reduced by reducing grid aperturesize. However, as demonstrated by the cases of the310 and 150:m grid aperture sizes, there is a pointwhere the grid aperture becomes too small for theefficient passage of aerosolised material. In fact, at60 L·min−1, there is significant blockage in the caseof the 532 :m grid aperture size; however at higher


4718 WONG ET AL.

Figure 7. Turbulence kinetic energy and integral shear strain rate measured along linespositioned through the middle, side and corner of a grid aperture for all grid aperture sizes at140 L·min−1. (Note: The 741:m grid aperture size has a larger void percentage than the 1028:m grid aperture size. Refer to Table 1).

flow rates, agglomerates impact with sufficient forceto pass through the grid structure (Fig. 3).

As shown in Figure 9b, reducing the void percent-age at a flow of 140 L·min−1 made no statistical dif-

ference to the particle d0.1, suggesting that the ag-glomerate break-up was independent of grid void per-centage at 140 L·min−1. However, when the percent-age of particles less than 5:m was plotted against



Figure 8. Size distributions of mannitol particles emitted through grid apertures of 1999 and532 :m at 60 and 140 L·min−1. (The mean air velocity at each flow rate is shown in brackets)

the grid void percentage (Fig. 10), it can be seen thateven at 140 L·min−1, reducing the grid void percent-age resulted in an increase in agglomerate break-up,reflected by an increased percentage of particles lessthan 5:m. However, due to the large variation of thedata, the trend of an increased percentage of particlesless than 5:m with decreasing void percentage at 140L·min−1 narrowly failed to achieve statistical signif-icance (p = 0.06). This large variation is likely dueto the fact that the agglomerates impacted with sub-stantially greater kinetic energy at 140 L·min−1. Thisresulted in a marked increase in number of “loose”fragments produced upon impact with the grid, re-flected by the broader peaks between 10 and 100:min the particle size distribution, shown in Figure 8.Consequently, a greater, but more variable, amountof fine particles was released. It is noteworthy thatcomplete disintegration of the agglomerate was notachieved in this study as the percentage of particlesless than 5:m did not exceed 12%.

Assuming the grid aperture is kept constant, thevoid percentage can be reduced by increasing thewire diameter. However, should the wire diametersubstantially exceed the diameter of the agglomer-ate, impaction onto the middle of the grid structurewould be similar to an agglomerate impacting a plate.Previous studies examining the impaction of an iden-

tical agglomerate system at similar impact velocitiesagainst an impaction plate have shown that uponimpaction, particle mass is deposited upon the im-paction plate independently of flow rate.24 Aerosolperformance was determined by the subsequent re-entrainment of this particle mass by the airflow abovethe impaction plate. DEM of agglomerates of 5 :mmono-disperse particles of similar density impactingon a plate by Tong et al.20 have also reported ductiledeformation and the formation of particles clusterswhich are dispersed radially from the point of impact.In the case of this study, wire diameters were of theorder of the smallest particle diameter or lower (allwire diameters ≤550 :m). As such, it is expected thatan agglomerate would only impact partially againstthe grid structure (either on the side or the middleof the agglomerate) which is dramatically different toimpaction against a plate. As the grid structure actson a small portion of an agglomerate, large pressureswould be exerted upon this area of contact, result-ing in a shearing force. Fragments from this initialimpaction would fall directly through the grid aper-tures into the flow stream, allowing for more efficientre-entrainment in comparison with that for a plate.This is reflected in the lower amount of mannitol de-posited within the grid structure in comparison withthe impactors used in the previous study.24 Reducing


4720 WONG ET AL.

Figure 9. Effect of grid aperture size (a) and void percent-age (b) on the particle d0.1 at all flow rates. (The mean airvelocity at each flow rate is shown in brackets.)

Figure 10. Effect of void percentage on the percentage ofparticle with a size less than 5:m at all flow rates. (Themean air velocity at each flow rate is shown in brackets.)

the wire diameter would exert a larger shear forceon the agglomerate as the area of contact is reduced(similar to a knife blade cutting through an object),allowing for more efficient break-up; however, a re-duction in wire diameter also leads to an increasein void percentage, resulting in a reduced likelihoodof impaction. Further studies are required to find theoptimal balance between aperture size, wire diameterand grid void percentage.

Separating the role of turbulence from that of gridimpactions in the aerolisation and break-up of ag-glomerates proved difficult in this study. However,it is likely that impaction would play the dominantrole in the initial fragmentation of the agglomer-ate. As mentioned previously, the agglomerate mustencounter the grid structure before being carriedthrough the grid aperture into regions of high inte-gral shear and turbulence kinetic energy, as there isminimal turbulence through the centre of the gridaperture. Initial impaction with the grid structurewould cause the agglomerate to fragment, whereasfurther dispersion of the fragments and multipletswith the regions of high turbulence kinetic energyand high shear is likely, however the time a parti-cle would spend in these regions is relatively short.Although large integral shear strain rates are gener-ated by the airflow passing through the grid structure,these are mostly confined to the small region near theedge of the grid within the grid aperture. Below thegrid aperture, the integral shear strain rate rapidlydecreases (Figs. 5 and 7). Likewise, high levels of tur-bulence kinetic energy are generated; however, theseare confined to small regions downstream of the gridstructure and dissipate within 20 mm downstream ofthe grid (Figs. 4, 5, and 7). A previous investigationof the effects of grid structure on the performance ofan Aerolizer R© DPI (Plastiape S.p.A, Osnago, Italy) indispersing 5:m spray-dried mannitol primary parti-cles reported that when grid voidage was decreased,the integral shear strain rates generated downstreamof the grid was reduced,13 which were initially of asimilar magnitude to the maximum levels observedin the corner region of the 1999:m grid aperture at140 L·min−1 (∼13,000 s−1) in this study. Particle gridimpactions were also reduced, and there was an in-crease in device retention, however the FPF of theemitted dose remained unchanged. This would sug-gest that the turbulence generated by the grid hadlittle effect on the performance of the Aerolizer R© DPI;however, it is important to note that the Aerolizer R©

DPI utilises tangential inlets, as well as a capsulefor the dispersion of mannitol, resulting in a cyclonicflow which is vastly different to the flow within theentrainment tube utilised in this study. As such, fur-ther investigation utilising DEM in conjunction withCFD should be considered to capture the break-up



behaviour of the agglomerate upon impaction as wellas to distinguish the specific contribution of gridimpactions and turbulence in the aersolisation andbreak-up of agglomerates.

CONCLUSIONS

This study demonstrated that grid structures canbe effective in promoting break-up and aerosolisa-tion of pharmaceutical agglomerates. Agglomerateimpaction against the grid structure appeared to bethe more dominant break-up mechanism when com-pared with turbulence generated by the grid, withthe initial impact inducing a mechanical shear uponthe agglomerate causing it to fragment. However, tur-bulence generated by the grid may potentially in-duce further dispersion of these fragments into fineparticles.

ACKNOWLEDGMENTS

This research was supported under the AustralianResearch Council’s Linkage Projects funding scheme(project LP0776892). The views expressed herein arethose of the authors and are not necessarily those ofthe Australian Research Council.

REFERENCES

1. Newman SP, Busse WW. 2002. Evolution of dry powderinhaler design, formulation, and performance. Resp Med96(5):293–304.

2. Smith IJ, Parry-Billings M. 2003. The inhalers of the future?A review of dry powder devices on the market today. PulmPharmacol Ther 16(2):79–95.

3. Islam N, Gladki E. 2008. Dry powder inhalers (DPIs)—Areview of device reliability and innovation. Int J Pharm360(1–2):1–11.

4. Pritchard JN. 2001. The influence of lung deposition on clinicalresponse. J Aerosol Med 14:S19–S26.

5. Young PM, Traini D, Edge S. 2007. Advances in pulmonarytherapy. In Advanced drug formulation design to optimizetherapeutic outcomes; Williams RO, Taft DT, McConville JT,Eds. London:Informa Healthcare, pp 1–51.

6. Louey MD, Van Oort M, Hickey AJ. 2006. Standardized en-trainment tubes for the evaluation of pharmaceutical dry pow-der dispersion. J Aerosol Sci 37(11):1520–1531.

7. Voss A, Finlay WH. 2002. Deagglomeration of dry powder phar-maceutical aerosols. Int J Pharm 248(1–2):39–50.

8. Wang ZL, Lange CF, Finlay WH. 2004. Use of an impinging jetfor dispersion of dry powder inhalation aerosols. Int J Pharm275(1–2):123–131.

9. Calvert G, Ghadiri M, Tweedie R. 2009. Aerodynamic disper-sion of cohesive powders: A review of understanding and tech-nology. Adv Powder Technol 20(1):4–16.

10. Coates MS, Chan HK, Fletcher DF, Chiou H. 2007. Influenceof mouthpiece geometry on the aerosol delivery performanceof a dry powder inhaler. Pharm Res 24(8):1450–1456.

11. Coates MS, Chan HK, Fletcher DF, Raper JA. 2005. Influ-ence of air flow on the performance of a dry powder inhalerusing computational and experimental analyses. Pharm Res22(9):1445–1453.

12. Coates MS, Chan HK, Fletcher DF, Raper JA. 2006. Effectof design on the performance of a dry powder inhaler usingcomputational fluid dynamics. Part 2: Air inlet size. J PharmSci 95(6):1382–1392.

13. Coates MS, Fletcher DF, Chan HK, Raper JA. 2004. Effect ofdesign on the performance of a dry powder inhaler using com-putational fluid dynamics. Part 1: Grid structure and mouth-piece length. J Pharm Sci 93(11):2863–2876.

14. Coates MS, Fletcher DF, Chan HK, Raper JA. 2005. Therole of capsule on the performance of a dry powder inhalerusing computational and experimental analyses. Pharm Res22(6):923–932.

15. Longest PW, Hindle M. 2009. Evaluation of the respimat softmist inhaler using a concurrent CFD and in vitro approach. JAerosol Med Pulm D 22(2):99–112.

16. Longest PW, Vinchurkar S. 2007. Validating CFD predictionsof respiratory aerosol deposition: Effects of upstream transi-tion and turbulence. J Biomech 40(2):305–316.

17. Ma B, Lutchen K. 2009. CFD simulation of aerosol depositionin an anatomically based human large–medium airway model.Ann Biomed Eng 37(2):271–285.

18. Moreno R, Ghadiri M, Antony SJ. 2003. Effect of the impactangle on the breakage of agglomerates: A numerical studyusing DEM. Powder Technol 130(1–3):132–137.

19. Thornton C, Liu LF. 2004. How do agglomerates break? Pow-der Technol 143–144:110–116.

20. Tong ZB, Yang RY, Yu AB, Adi S, Chan HK. 2009. Numer-ical modelling of the breakage of loose agglomerates of fineparticles. Powder Technol 196(2):213–221.

21. Guo Y, Kafui KD, Wu CY, Thornton C, Seville JPK. 2009. Acoupled DEM/CFD analysis of the effect of air on powder flowduring die filling. Aiche J 55(1):49–62.

22. Kloss C, Kahrimanovic D, Pirker S. DANSIS meeting on dis-crete element methods. Kopenhagen, 2009.

23. Wong W, Fletcher DF, Traini D, Chan HK, Crapper J, YoungPM. 2010. Particle aerosolisation and break-up in dry pow-der inhalers 1: Evaluation and modelling of venturi effects foragglomerated systems. Pharm Res 27(7):1367–1376.

24. Wong W, Fletcher DF, Traini D, Chan HK, Crapper J, YoungPM. Particle aerosolisation and break-up in dry powder in-halers: Evaluation and modelling of impaction effects for ag-glomerated systems. J Pharm Sci 100(7): 2744–2754.

25. Wong W, Crapper J, Chan HK, Traini D, Young PM. 2010.Pharmacopeial methodologies for determining aerodynamicmass distributions of ultra-high dose inhaler medicines. JPharm Biomed Anal 51:853–857.

26. Chew NY, Chan HK. 1999. Influence of particle size, air flow,and inhaler device on the dispersion of mannitol powders asaerosols. Pharm Res 16(7):1098–1103.


particle aerosolisation and break-up in dry powder inhalers: evaluation and modelling of the...

Documents