Powder Technology 249 (2013) 297303

Contents lists available at ScienceDirect

Powder Technology
journal homepage: www.elsevier.com/locate/powtec

Importance of particle size and shape on the tensile strength distribution

and de-agglomeration of cohesive powders
Shyamal C. Das a,b,, Srinivas Ravindra Babu Behara a,1, David A.V. Morton a, Ian Larson a, Peter J. Stewart a
a
b

Drug Delivery, Disposition and Dynamics, Monash Institute of Pharmaceutical Sciences, Monash University (Parkville campus), Australia
New Zealand's National School of Pharmacy, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand

a r t i c l e

i n f o

Article history:
Received 18 April 2013
Received in revised form 14 August 2013
Accepted 23 August 2013
Available online 31 August 2013
Keywords:
De-agglomeration
Agglomerate strength
Size
Shape
Intermediate lactose
Dry powder inhaler

a b s t r a c t
Purpose: The purpose of the study was to understand the role of particle size and shape changes in modifying agglomerate strength distribution and de-agglomeration of cohesive lactose powders.
Methods: The relative de-agglomeration of three lactoses of different particle size distributions (Lactohale 201 or
LH201, Lactohale 210 or LH210 and Lactohale 220 or LH220) was determined from laser diffraction particle sizing
of the aerosol plume at different air ow rates. The agglomerate strength distributions were estimated by Monte
Carlo simulation using the primary particle size, work of cohesion and tapped density distributions determined
by laser diffraction, inverse gas chromatography and tapping apparatus, respectively. The morphology and particle shape parameters were determined by scanning electron microscopy and the Morphologi G3.
Results: The estimated agglomerate strength correlated well with the de-agglomeration of all lactose samples at
different air ow rates. While the work of cohesion of the lactose samples was not signicantly different, the
packing fraction was dependent on the proportion and shape of intermediate-sized, cohesive particles between
5.4 and 14 m. For example, while the proportion of particles b5.4 m was similar for all lactose samples, the proportion of intermediate-sized, cohesive particles increased in the order of LH201 b LH210 b LH220. The
intermediate-sized, cohesive particles were more elongated than the b5.4 m fraction and the extent of elongation of the lactose samples increased in the order of LH220 N LH210 N LH201.
Conclusion: The study reinforced the role of agglomerate strength distributions in understanding deagglomeration of cohesive materials. Modication of particle size distributions and shape characteristics contributed to the agglomerate strength changes in the lactose samples. The study enhanced the fundamental understanding of powder de-agglomeration and provided strategic approaches that could be used to improve
inhalation product performance.
2013 Elsevier B.V. All rights reserved.

1. Introduction
Dry powder inhaler formulations often contain micron-sized drug
particles (usually b5 in size) mixed with large, free-owing lactose and
cohesive ne lactose. Due to high adhesion/cohesion of these micron
size particles, the drug is agglomerated alone, with ne carriers, with
large carriers or with both ne and large carriers resulting in complex
multi-particular agglomerates [1]. However, for effective drug delivery
to the deep lung/alveolar region, the agglomerates need to be deagglomerated and dispersed to particles b5 m [2]. De-agglomeration of
powders was found to be inuenced by the agglomerate strength which
was indirectly measured by air shear pressure using an Aerosizer [3,4]
or by the mechanical strength measured using a dual column physical

Corresponding author at: New Zealand's National School of Pharmacy, University of

Otago, P.O. Box 56, Dunedin 9054, New Zealand. Tel.: +64 3 479 4262; fax: +64 3 479 7034.
E-mail address: Shyamal.das@otago.ac.nz (S.C. Das).
1
Current Address: Department of Mechanical and Nuclear Engineering, School of
Engineering, Virginia Commonwealth University, Richmond, Virginia, USA.
0032-5910/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.powtec.2013.08.034

testing machine [5]. Agglomerate strength () can be calculated using

the Eq. (1) [6]:
4

15:6

W
d

!
1

where d is the particle diameter, is the packing fraction (volume of

particles/volume of aggregates) and W is the work of adhesion or cohesion of particles. In fact, real powders are heterogeneous in nature
having a distribution of particle size, a distribution of work of cohesion
and a distribution of packing fraction resulting in a distribution of agglomerate strength [7]. While the de-agglomeration and dispersion of
powders at a particular ow rate may relate to a single, average value
of agglomerate strength [5], the complex de-agglomeration behavior
of cohesive micronized powders (b5 m) at a range of ow rates is better explained by the agglomerate strength distribution [8]. According to
Eq. (1), for the same material where the work of cohesions of different
powders is similar, the agglomerate strength, and therefore, the de-

298

S.C. Das et al. / Powder Technology 249 (2013) 297303

agglomeration will vary depending on the differences in particle size

and packing fraction.
The packing density and packing fraction of powders are inuenced
by particle size distribution [9,10] meaning the agglomerate strength
and therefore, the de-agglomeration of powders will be inuenced by
particle size distribution. The inuence of particle size of lactose on
drug dispersion has been studied; however, the focus of these studies
has been on either the free-owing larger carrier particles or on the
very small micronized particles usually less than about 5 m. For example, the dispersion of the drug was either decreased or remained
unchanged with increasing particle size of large carriers [1114]. However, Islam et al. also showed that it is ne lactose (FL), not the carrier
lactose size, which was important in increasing de-agglomeration of
salmeterol xinafoate (SX) [13]. The FL, usually of size b 5 m and either
associated with carrier lactose or additional to the formulation, improves dispersion through either passivation of the active sites
[15,16] or by the formation of mixed agglomerates of drug and ne lactose particles which are less strong than drug alone agglomerates [1,17].
However, the role of intermediate sized cohesive lactose (i.e., mean particle size is generally in the range of 515 m which is larger than the
micronized lactose particles used to enhance aerosolisation) and the
mechanism of changing dispersion performance was less explored. An
increase in the dispersion of salbutamol sulphate (SS) was found
when only 1.5% w/w of lactose of volume mean diameter (VMD)
15.9 m was added to a mixture of SS and coarse sieved lactose
(VMD, 90.8 5.0 m) [15]. However, in that study, the lactose sample
was not completely free from nes (b5 m) raising the question of
whether the result was due to nes or intermediate size lactose. In another study, the dispersion of SX from SX-FL binary mixtures was higher
using FL of VMD 7.9 m than using FL of VMD 3 m [3]. Using particle
size-shear pressure proles of the powders, they argued that the formation of more open packed structures of the mixture containing FL of
VMD 7.9 m enhanced the aerosolisation in comparison with the mixture containing FL of VMD 3 m. In the FL of VMD 7.9 m, 90% of particles were less than 14 m. Conventional carrier lactose powders used in
inhalation formulations have a broad distribution of particle size including ne cohesive fraction (b5 m), intermediate size semi-cohesive
fraction (for example, 514 m) and larger, free-owing particles. It
is, therefore, important to understand the key particle fractions providing the greatest impact on de-agglomeration and to give focus to the
role of intermediate size fractions of carrier lactose and the mechanism
of dispersion change. It is now possible to determine the agglomerate
strength distribution directly from particle size distribution, tapped
density distribution and work of cohesion distribution determined by
laser diffraction, tapping apparatus and inverse gas chromatography,
respectively [7]. Moreover, various particle shape parameters such
elongation and circularity can be analysed capturing 2-dimensional
(2D) images of the 3-dimensional (3D) particles by the Malvern
Morphologi G3 [18]. These shape parameters can give a useful indication about the particle shape which may have effect on packing and deagglomeration [19].
It is hypothesized that the intermediate size cohesive lactose fraction
will create a more open-packed structure of the cohesive powder bed
resulting in more facile de-agglomeration and faster achievement
of the peak de-agglomeration. Therefore, this experiment was designed to study the inuence of intermediate sized lactose fraction
on de-agglomeration and to understand the mechanism. The deagglomeration of the three lactose of different particle size distributions having similar work of cohesions and similar proportion of
nes (b 5.4 m) but differing proportions of intermediate size fractions was examined at different air ow rates. The work of cohesion
distribution, particle size distribution, and packing fraction distribution were determined and the agglomerate strength distribution
was calculated from these parameters using a Monte Carlo simulation. The shape parameters of agglomerates were also examined.
Understanding the roles of these intermediate size semi-cohesive

lactose fractions will help better design powder formulations or to

develop carrier lactose particles that can give improved efciency.
2. Materials and methods
2.1. Materials
Three lactose samples: Lactohale 201 (LH201), Lactohale 210
(LH210) and Lactohale 220 (LH220) supplied by DFE Pharma,
Netherlands were used as received. GC grade hexane, heptane, octane,
nonane, decane, dichloromethane and ethyl acetate (all from SigmaAldrich GmbH, Steinheim, Germany) were used for surface energy
analysis.
2.2. Methods
2.2.1. Pre-conditioning of powders
In order to conrm that powders have been exposed to similar mechanical processing prior to investigation, each powder was preconditioned by a standardised hand mixing experiment that has been
developed and validated in our laboratory [20]. Five grams of each powder in each batch was mixed in a glass jar for ve minutes using three
ceramic balls (10 mm diameter). At the end of every 30 s mixing, the
jar was tapped both horizontally and vertically to release the powder
stuck on jar's cones and wall. The mixing was conducted by gentle shaking so that no particle size reduction occured; this was conrmed as no
statistically signicant difference (P N 0.05) was observed in primary
particle size distributions before and after pre-conditioning.
2.2.2. Work of cohesions calculation
Non polar, polar and total surface energy distributions were determined using a nite dilution experiment with inverse gas chromatography (IGC, Surface Measurement Systems Ltd, London, UK)
according to a literature method [21,22]. In brief, approximately
0.6 g of lactose was packed in pre-silanised glass columns
(300 mm 3 mm internal diameter) by tapping for four minutes
using a tapping apparatus (Surface Measurement Systems Ltd,
London, UK). The powder lled columns were closed at both ends
with silanised glass wool and conditioned for 2 h at 303 K to remove
surface impurities. A series of alkanes such as hexane, heptane, octane, nonane and decane were used to determine non-polar surface
energy (NP), and two polar probes such as dichloromethane and
ethyl acetate were used to determine polar surface energy (P). All
these probes were passed through the column at concentrations of
0.03, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 0.94 p/p0 (where p denotes the
partial pressure and p0 the vapour pressure). Helium owing at
10 sccm (standard cubic centimetres per minute) was used to carry
the probes in to the column. A ame ionization detector was used
to detect retention times. The retention time for methane run at a
concentration of 0.1 p/p0 was regarded as the dead time, where
there was no interaction between probe and sample. Retention volumes were calculated from retention times. The BrunauerEmmet
Teller (BET) surface area was determined from hexane adsorption isotherms, and the surface coverage (n/nm) was calculated from the
adsorbed amount (n) and monolayer capacity (nm, the number of
moles of the probe adsorbed for monolayer coverage). At each surface
coverage for each probe, the net retention volume (VN) was calculated.
The NP was the slope (2 NA NP) of a plot of RTlnVN against a NP of
alkanes [23]. The P was calculated from acidic (+) character determined from interaction with a monopolar basic probe, ethyl acetate,
and basic () character determined from the interaction with a
monopolar acidic probe, dichloromethane as per van Oss concept
[24,25]. The total surface energy (T) was calculated by summing up
the non-polar (NP) and polar surface energies (P) [26]. The work of cohesion (W) was calculated from the non-polar and polar surface energies [27].

S.C. Das et al. / Powder Technology 249 (2013) 297303

2.2.3. Primary particle size distributions determined by laser diffraction

Approximately 200 mg of pre-conditioned lactose was sonicated in
5 ml isopropyl alcohol for 5 min. The primary particle size distributions
of the three lactose samples (LH201, LH210 and LH220) were determined by laser diffraction (Mastersizer-S, Malvern Instruments,
Worcestershire, UK) using the 300RF lens and the 150 ml dispersion
unit. The mean particle size distribution of six replicates was characterized by the derived parameters (d10, d50 and d90) using an imaginary refractive index of 0.01.
2.2.4. Packing fraction distributions
By pouring 5 g of pre-conditioned powders slowly into a 10 ml
measuring cylinder from a xed height, the tapped volume distributions for each powder in four replicates were determined over 1024
taps of an automatic tapper (AUTOTAP, Quantachrome Instruments, Boynton Beach, FL, USA). Normalising the weight by true density of lactose (1.545 g cc 1) [28], the true volume or volume of
particles was calculated. The packing fraction distributions were calculated dividing the true volume by tapped volume at each tap as
shown in Eq. (2):

Packing fraction

True volume or volume of particles

Tapped volume or volume of aggregates

299

sprinkled on double-sided carbon black tape mounted on a sample

holder. The powders were gold coated for 2 min by a sputter coater
(SCD005, BAL-TEC AG, Balzers, Germany) using an electrical potential
of 2.0 kV at 25 mA under a partial vacuum of 0.1 mbar for an expected
coating thinness of 15 nm.
2.2.7.2. Particle shape parameters. Particle shape parameters such as
elongation, aspect ratio, convexity mean and HS circularity mean of
pre-conditioned lactose powders were analysed using a Malvern
Morphologi G3 [18]. The specications of operations were as follows:
sample volume: 3 mm3, injection pressure: 1.0 bar, injection time:
30 ms, setting time: 300 s, light intensity: 80% and foil type: 6 microns.
A post analysis lter was applied to sort images of particular size range.
2.2.8. Statistical modelling and analysis
The statistical signicance was carried out using one-way analysis of
variance with Tuckeys post-hoc analysis and between the groups was
carried out using independent sample t-test using an alpha of 0.05
(SPSS version 17.0, SPSS, Inc., IL, USA).
3. Results
3.1. Work of cohesion of lactose powders

2.2.5. Powder strength distributions using Monte Carlo simulations

The distributions of three parameters, particle diameter, packing
fraction and work of cohesion were determined experimentally. Cumulative probability distributions were constructed from these experimental values by linear interpolation between frequent measurements.
Assuming that there was no correlation between the distributions of
work of cohesion, tapped density, and particle diameter, Monte Carlo
simulations were employed to determine powder strength distributions
of 1,000,000 random samples using the Eq. (3) [7]:
4

i 15:6

t W i
di

!
3

where, i denotes powder strength of ith random value determined

from the ith random value for particle diameter (di), true density (t)
and work of cohesion (Wi). The Math random Perl package (version
0.71) was used to draw random samples. The descriptive statistics
such as the expectation value (arithmetic mean), standard deviation,
and representative percentiles of powder strength were calculated
from these random values for i. Perl scripts were developed to implement these calculations and data processing.

The work of cohesion (WC) of all three lactose samples (LH201,

LH210 and LH220) was similar at the surface coverage determined by nite dilution experiment using IGC (Fig. 1). The maximum surface coverage that produced acceptable data in this study was low; for
example, 4% for LH201 and only around 1.5% for LH210. For acceptable
data, the retention volumes at different concentrations for different
probes were considered. In principle, the retention volume should increase with the increase in probe concentration if there were no
intermolecular interactions between probe molecules. Therefore, the
concentration at which no further increase in retention volume was observed was not included in surface coverage calculation. The surface
coverage was calculated by normalising the amount of the probe
adsorbed to the material by the amount of probe required for monolayer coverage (known as monolayer capacity of the probe). In the past,
2025% surface coverage for coarse lactose was reported compared to
around 5% for ne lactose [22]. The nonpolar surface energy was calculated using Schultz approach [23] from the slopes of ve alkane probes.
In general, as the number of alkanes used to determine the surface energy is increased, the outcome is more reliable. During the nonpolar surface energy calculation, the surface energy values with R2 0.999 were

2.2.6. Aerosolisation of lactose by particle sizing of aerosol plume

Powders (20 1 mg) were dispersed horizontally from gelatine
size 3 capsules (Capsugel, NSW, Australia) using a Rotahaler (GSK,
Middlesex, UK) through an inhalation cell of a laser diffraction instrument
(Spraytec, Malvern Instruments, Worcestershire, UK) for ve seconds at
30, 45, 60, 90 and 120 l min1. The air ow rates were chosen as they are
possible for patients with respiratory disorders (50400 l min1)
[2931]. The measurements were conducted on ve replicates by capturing 100 measurements per second over a ve second period. A standard
format of average scatter data along with concentration weighted average
for the particles less than 5.4 m size was employed [32,33].
2.2.7. Morphology
2.2.7.1. Scanning electron micrographs. The morphology of each lactose
powder was examined by a scanning electron microscope (Phenom,
FEI Company, Hillsboro, Oregon, USA). Each powder (~20 mg) was

Fig. 1. The work of cohesion distributions of Lactohale 201 (LH201), Lactohale 210
(LH210) and Lactohale 220 (LH220) determined by inverse gas chromatography at nite
dilution (n = 2).

300

S.C. Das et al. / Powder Technology 249 (2013) 297303

accepted, which also limited the acceptable maximum surface coverage.

There were other limitations associated with IGC experiment. For example, in this study, dichloromethane and ethyl acetate were chosen as
monopolar acidic and monopolar basic probe as used by others [24]
and as recommended by the instrument manufacturer even though
they are not completely monopolar; a more detailed discussion about
these limitations have been described [21].
As there was little difference between the work of cohesion distributions of the three lactose samples (Fig. 1), an average of the work of cohesion distributions was calculated. It is not possible to determine real
surface energy/work of cohesion up to 100% surface coverage. Thus, the
average work of cohesion was extrapolated to 100% surface coverage
(Fig. 2). This extrapolation is often problematic as the extrapolated
values may not represent the actual surface energy/work of cohesion.
However, this prole was consistent with previous published results
[22,34] and with the fact that work of cohesion of most of the lactose surface should be similar and the difference in work of cohesion, if any,
should be for a small part of surface of high energy state. The orientation
and spacing of different chemical groups could be different at different
parts of the surface. Since surface energy determination in IGC is based
on the interaction between probes and surface groups, it is logical that
surface energy of a material will vary at different parts of the surface.
The purpose of the extrapolation was to determine a WC cumulative
distribution that could be used to calculate the tensile strength of the cohesive powder bed. The assumption used in this calculation was that
there was no difference in WC among the three lactose samples used
in this study. As the particle size and packing fraction distribution can
have a signicant role on de-agglomeration behavior, these parameters
were determined.
3.2. Particle size distributions
There was no difference among the lactose samples in the low particle size range, but the difference was observed in the presence of larger
particles (Fig. 3). For example, the concentrations of particles less than
5.4 m were similar (~1011%) in three lactose samples (LH201,
LH210 and LH220) whereas 99% particles were less than 63 m for
LH220 compared to 100 m for LH201 (Table 1).
The concentration of intermediate-size semi-cohesive particles
(5.414 m) was the highest in LH220 and the lowest in LH201. For example, the concentration of particles between 5.4 m and 14 m was
around 17% in LH201 as opposed to 24% in LH220. Particles less than
14 m was chosen in this study because a cohesive particle fraction
(VMD 7.9 m) that showed better de-agglomeration previously had
90% of particles less than 14 m [3].

Fig. 3. The cumulative primary particle size distribution (PSD) of Lactohale 201 (LH201),
Lactohale 210 (LH210) and Lactohale 220 (LH220) determined by laser diffraction in liquid medium (n = 5, error bars indicate standard deviation of ve replicates).

The presence of different particle size distributions is likely to lead to

different packing arrangement and packing fraction distributions.
Therefore, packing fraction distributions of all three lactose samples
were determined and reported.
3.3. Packing fraction distributions
It is difcult to determine packing fractions in various locations of a
powder bed and then construct a packing fraction distribution. Due to
the difference in powder structure, the powder packed differently during consolidation process. For example, a powder bed with a higher
packing fraction will be more consolidated than a powder bed with a
lower packing fraction. In this study, tapped density was determined
after different numbers of taps and was used to calculate the tapped
density distribution. By converting the tapped density to packing fraction, the packing fraction distribution was constructed. The bulk density
and the packing fractions of three lactose samples prior to tapping were
not different. For example, the packing fractions of LH220, LH210 and
LH201 after 64 taps (~3% taps) were 0.32 0.03, 0.34 0.02 and
0.34 0.00, respectively. The packing fraction of LH220 was less than
LH201 and LH210 with increased tapping (Fig. 4). The difference between LH201 and LH210 appeared after 5% of the total tapping required
to consolidate the powder bed. For example, the packing fractions of
LH220, LH210 and LH201 after 512 taps (~ 50% taps) were 0.41
0.01, 0.45 0.01 and 0.48 0.01, respectively. The lower packing
fraction of LH220 could result in agglomerates of lower strength
which can be revealed by calculating agglomerate strength distribution.
3.4. Agglomerate strength distribution
The agglomerate strength distribution was determined from particle
size distribution, work of cohesion distribution and packing fraction distribution by Monte Carlo simulation [7]. It was assumed that all the
Table 1
The primary particle size distribution parameters of three lactose samples of Lactohale 201
(LH201), Lactohale 210 (LH210) and Lactohale 220 (LH220) determined by laser
diffraction in liquid medium (n = 5, standard deviations of ve replicates are shown
after ).

Fig. 2. The work of cohesion versus percent surface coverage under calculated as average
of the three lactose samples by best t regression equation.

Parameters

LH201

LH210

LH220

d50
b5.4 m
5.414 m
b99% particles

24.2 0.6
10.3 0.3
17.3 0.5
100 m

20.4 0.6
10.8 0.4
21.7 0.5
85 m

18.2 0.7
11.1 0.7
24.3 0.8
63 m

S.C. Das et al. / Powder Technology 249 (2013) 297303

301

Table 2
Representative agglomerate strength values of Lactohale 201 (LH201), Lactohale 210
(LH210) and Lactohale 220 (LH220) calculated by Monte Carlo simulation (n =
1,000,000).
Lactose

Average agglomerate strength (mPa)

Variance

1%

80%

99%

LH201
LH210
LH220

2.81
2.41
1.42

195.71
164.13
85.46

0.14
0.15
0.06

8.65
8.66
6.16

19.16
18.09
12.12

no signicant (P N 0.05) difference in percent relative de-agglomeration

was observed between LH220 and LH210 at any of the ow rates.
All the data of percent relative de-agglomeration proles have been
empirically modelled using sigmoid 3-parameter model (shown in
Eq. (4)) and nonlinear least squares regression [35].

Fig. 4. Cumulative packing fraction distributions of Lactohale 201 (LH201), Lactohale 210
(LH210) and Lactohale 220 (LH220) (n = 4, data represent mean standard deviation).

three variables are independent. It is seen that the agglomerate strength

of LH220 was lower than LH201 and LH210 (Fig. 5).
The agglomerate strength for up to 99% particles of LH220 was lower
than that of LH201 and LH210. Moreover, the average agglomerate
strength of LH220 was lower than LH210 and LH201 (Table 2). Lower
the agglomerate strength means easier de-agglomeration.

3.5. Relative de-agglomeration determined by laser diffraction

The percent relative de-agglomeration versus air ow rate proles
for all three lactose of different particle size distributions (LH201,
LH210 and LH220) are shown in Fig. 6. The air ow rates required to
reach maximum percent relative de-agglomeration [33] of the three lactose samples were different. For example, LH220 reached the maximum
percent relative de-agglomeration at 45 l min1 which was followed
by LH210 at 60 l min1 and by LH201 at 90 l min1 (Fig. 6). The maximum percent relative de-agglomerations were approximately 90% for
LH220 and LH210, and reached close to 100% for LH201.
Although there was no signicant (P N 0.05) difference in percent relative de-agglomeration of the three lactose samples at 30 l min1, the
percent relative de-agglomeration of LH220 was signicantly (P b 0.05)
higher than LH201 at ow rates of 45 l min1 and 60 l min1. Moreover,

Fig. 5. Powder strength distributions of Lactohale 201 (LH201), Lactohale 210 (LH210) and
Lactohale 220 (LH220) determined from their cumulative particle size distributions, cumulative work of cohesion distributions and cumulative packing fraction distributions
by Monte Carlo simulation.

a
1 e

xx0
b

where, a, b and x0 represent the maximum percent relative deagglomeration, the change in percent relative de-agglomeration with
air ow rate, and the air ow rate required to achieve 50% relative deagglomeration, respectively. This approach has been used to characterize the aerosolisation behavior of the powders [32,33]), and the requirements for determining the best t have been described earlier [36]. The
estimated parameters of the lactose samples are shown in Table 3.
The estimated maximum percent relative de-agglomeration for
LH201 was higher than that for LH220 or LH210 (Table 3), which was
consistent with the experimental de-agglomeration versus air ow
rate proles for the three lactose samples (Fig. 6). The change in percent
relative de-agglomeration (b) with air ow rate, and the air ow rate
required to achieve 50% relative de-agglomeration (x0) were different
among the lactose samples.
4. Discussion
The correlation between the average agglomerate strength and the
sigmoidal modelling parameters, estimated from the de-agglomeration
versus air ow rate proles, of (1) ow rate required to achieve 50% relative de-agglomeration (x0) and (2) the change in percent relative deagglomeration with air ow (b) was shown in Fig. 7. Although no significant change was observed in either b or x0 when the agglomerate
strength was increased from 1.4 mPa to 2.4 mPa, both b and x0 rapidly

Fig. 6. The percent relative de-agglomeration proles of percent of particles less than
5.4 m versus air ow rate for the aerosolised plume of Lactohale 201 (LH201), Lactohale
210 (LH210) and Lactohale 220 (LH220) dispersed from Rotahaler and determined by
laser diffraction at different ow rates (30 l min1 to 120 l min1) (n = 5, error bars indicate standard deviations of ve replicates).

302

S.C. Das et al. / Powder Technology 249 (2013) 297303

Table 3
Nonlinear least squares estimated parameters of a (the maximum percent relative deagglomeration), b (the change in percent relative de-agglomeration with air ow rate)
and x0 (the air ow rate required to achieve 50% relative de-agglomeration) for the 3parameter tting of the percent relative de-agglomeration versus air ow rate proles
for Lactohale 201 (LH201), Lactohale 210 (LH210) and Lactohale 220 (LH220).
Parameter

LH201

LH210

LH220

a
b
x0

106.57
29.42
44.91

87.87
10.55
30.07

87.55
6.83
29.24

increased when agglomerate strength was increased from 2.4 mPa to

2.8 mPa. Although data were limited, the correlation shown in Fig. 7 indicated that a critical agglomerate strength was required before change in
the parameters occurred. In relation to the third estimated parameter of
percent relative de-agglomeration, the agglomerate strength distribution
of LH220 was lower than LH210 and LH201 (Fig. 5) which was consistent
with faster approach to maximum percent relative de-agglomeration by
LH220 than by LH210 or LH201.
This study considered the reasons for the change in agglomerate
strength for the three lactose samples of different particle size distributions. Based on Eq. (1), the change in the agglomerate strength could be
related to differences in work of cohesion, particle size, or packing fraction [6]. The work of cohesion of the three lactose samples was similar
(Fig. 1). Therefore, the work of cohesion was likely to have little signicance on the difference in agglomerate strengths and de-agglomeration
behavior of these three lactose samples shown in Fig. 6.
There are several aspects of particle size that should be considered.
4.1. Presence of nes
The presence of nes in dry powder formulations can inuence deagglomeration [17,37]. However, the similar extents of nes in all the
three lactose samples means that any difference in de-agglomeration
could not be attributed to nes less than 5.4 m.
4.2. Presence of large particles
In contrast, the presence of large, free owing lactose at different
concentrations may produce a different ball-milling effect on the comminution of agglomerates in the three lactose samples during
aerosolisation. However, this mechanism, where the force magnitude
and the extent of ball milling were dependent on the size of the particles, does not seem to be the major contributor to de-agglomeration.
LH201 (d90 of 58.8 m) showed less de-agglomeration than LH220
(d90 of 38.8 m), in spite of having the larger size distribution with

potentially greater contribution to ball milling of agglomerates. Thus,

the presence of large lactose particles may not be the reason for the different de-agglomeration patterns of the three lactose samples.
4.3. Presence of intermediate size cohesive particles
Aerosolisation of particles less than 5.4 m will require the materials
in the cohesive matrix to de-agglomerate. The extent of de-agglomeration
has been shown to depend on the particle size distribution of the materials in the cohesive matrix; agglomerates containing the smaller particles
are less readily aerosolised than agglomerates consisting of larger cohesive particles because of the inuence on the strength of the agglomerate
[3]. The presence of higher amounts of cohesive intermediate-size lactose
seen in LH220 (in contrast to LH201 or LH210 (Table 1)) may lead to the
formation of more loosely-packed agglomerates and thus aerosolisable
powder beds demonstrated in Fig. 6 [1,3].
The extent of powder bed consolidation of the three lactose samples
with tapping, shown by the packing fraction distributions, was reasonably well correlated with the percentage of intermediate-sized lactose
in the particle size distributions. The higher the concentration of
intermediate-sized lactose, the lower is the packing fraction. The packing fraction was lowest for LH220 and highest for LH201.
In addition to the packing fraction correlations with percent of intermediate size lactose, the shape characteristics of the intermediate size
lactose particles (5.414 m) were different to the small lactose particles (b 5.4 m). Four shape parameters (aspect ratio, elongation, convexity and high sensitivity (HS) circularity) of the three lactose
samples were determined (Table 4). Aspect ratio is dened as the
width/length ratio and the lower the aspect ratio, the longer the particle
[38]. Elongation is dened as 1 aspect ratio[39]. The values for elongation can vary from 0 for a circle or square which is symmetrical in all
axes to almost 1 for an elongated particle. Convexity is a measure of
surface roughness having the value of 1 for a completely smooth surface
and the value of 0 is for a highly spiky and irregular surface. The circularity or HS circularity is the measure of perfectness of circle shape
[40]. For a perfect circle the value is 1 while for a spiky or irregular object
circularity will be close to 0 [18].
For each of the three lactose samples (LH220, LH210 and LH201), the
5.414 m particles (individual particles or agglomerates) were signicantly more elongated than b5.4 m units as evident by the higher elongation mean and lower aspect ratio (Table 4; P b 0.001). Moreover, HS
circularity mean of 5.414 m particles were also signicantly lower
than corresponding values of b5.4 m units, indicating that 514 m
units are less circular than the corresponding b5.4 m units
(P b 0.001). The particle shape parameters for 5.414 m particles of
different lactose samples were signicantly different (P b 0.001). The
Table 4
The shape parameters (elongation mean, aspect ratio mean, convexity mean and high
sensitivity circulatory mean) of Lactohale 201 (LH201), Lactohale 210 (LH210) and
Lactohale 220 (LH220).
Shape
parameters

LH220

514 m
b5 m

LH210

514 m
b 5 m

LH201
Fig. 7. The relationship between the change in percent relative de-agglomeration (b)
with air ow rate and the air ow rate required to achieve 50% relative deagglomeration (x0) with average agglomerate strength.

514 m
b 5 m

Elongation
mean

0.366
0.007
0.240
0.016
0.343
0.008
0.203
0.020
0.290
0.009
0.167
0.008

Aspect
ratio
mean

Convexity
mean

0.634
0.012
0.760
0.021
0.657
0.013
0.797
0.018
0.710
0.015
0.833
0.018

0.945
0.014
0.988
0.009
0.961
0.020
0.973
0.012
0.969
0.013
0.967
0.011

High
sensitivity
circularity
mean

0.680
0.012
0.878
0.011
0.741
0.014
0.861
0.014
0.798
0.013
0.856
0.016

S.C. Das et al. / Powder Technology 249 (2013) 297303

elongation shown by the mean aspect ratio and the elongation

mean for the three lactose samples changed in the order of
LH220 N LH210 N LH201 indicating that the particles of LH220 were
more elongated than LH210 and LH201; the HS circulatory changed in
the order of LH220 b LH210 b LH201 indicating that the particles of
LH220 was less circular than LH210 and LH201 (Table 4). All differences
were signicant (P b 0.001). However, there was no signicant difference in the convexity mean for all fractions (P = 0.251 for the 5.4
14 m fraction and P = 0.102 for the b 5 m fraction); the value of
the convexity mean indicated that the particles were relatively smooth.
There were also signicant differences in the elongation of the particle
fraction b 5.4 m indicated by the differences in elongation and aspect
ratio means indicating that the particles of LH220 in the fraction
b5.4 m were more elongated than LH210 and LH201 (P = 0.004 and
P = 0.010, respectively).
Not only did the LH220 contain a greater concentration of intermediate sized particles in the range of 5.414 m, but also LH220 possessed
intermediate sized particles that were more elongated. In addition, the
particles in the b 5.4 m fraction were more elongated although the concentration in each lactose was about the same. Thus, the composition of
the cohesive matrix of the three powders constituted by all particles less
than 14 m was likely to be different. It is not possible to say whether
the concentration or the shape of the particles in the cohesive matrix
produced the packing fraction change; however, the more loosely
packed structure of LH220 was responsible for its higher aerosolisability
at low ow rates.
5. Conclusion
The study reinforced the role of agglomerate strength distributions
in understanding de-agglomeration of cohesive materials. The importance of this study is that the reasons for change in agglomerate
strength have been established and related to the physical characteristics (size and shape) of a cohesive fraction in the range 5.4 to 14 m.
Strategies to reduce agglomerate strength relate to reducing packing
fraction and work of cohesion. In this study, work of cohesion remained
relatively constant between the powder samples; however, packing
fraction was manipulation of size and shape and this contributed to
the agglomerate strength changes in the lactose samples. The study enhanced the fundamental understanding of powder de-agglomeration
and provided strategic approaches that could be used to improve inhalation product performance.
Acknowledgements
The authors would like to thank DFE Pharma and Capsugel for supplying lactose and gelatine capsules respectively. SD acknowledges Dr
Jurgen Bulitta for the help about Monte Carlo simulation.
References
[1] H. Adi, I. Larson, P.J. Stewart, Adhesion and redistribution of salmeterol xinafoate
particles in sugar-based mixtures for inhalation, Int. J. Pharm. 337 (2007) 229238.
[2] Y. Qui, A.L. Adjei, P.K. Gupta, Absorption and bioavailability of inhaled peptides and
proteins, in: A.L. Adjei, P.K. Gupta (Eds.), Inhalation delivery of Therapeutic Peptides
and Proteins, Mercel Dekker, Inc., New york, 1997.
[3] H. Adi, I. Larson, H. Chiou, P. Young, D. Traini, P. Stewart, Agglomerate strength and
dispersion of salmeterol xinafoate from powder mixtures for inhalation, Pharm. Res.
23 (2006) 25562565.
[4] S. Das, I. Larson, P. Young, P. Stewart, Inuence of storage relative humidity on the
dispersion of salmeterol xinafoate powders for inhalation, J. Pharm. Sci. 98 (2009)
10151027.
[5] S. Adi, H. Adi, H.-K. Chan, W.H. Finlay, Z. Tong, R. Yang, A. Yu, Agglomerate strength
and dispersion of pharmaceutical powders, J. Aerosol Sci. 42 (2011) 285294.
[6] K. Kendall, C. Stainton, Adhesion and aggregation of ne particles, Powder Technol.
121 (2001) 223229.
[7] S. Das, S. Behara, J. Bulitta, D. Morton, I. Larson, P. Stewart, Powder strength distributions for understanding de-agglomeration of lactose powders, Pharm. Res. 29
(2012) 29262935.

303

[8] S.C. Das, S.R. Behara, D.A.V. Morton, I. Larson, P. Stewart, Superiority of powder
strength distribution for understanding de-agglomeration of lactose powders, in:
R. Dalby, P. Byron, J. Peart, J. Suman, S. Farr, P. Young (Eds.), Respiratory Drug Delivery, Davis Healthcare International publishing, LLC, Arizona, 2012, pp. 585590.
[9] H.J.H. Brouwers, Particle-size distribution and packing fraction of geometric random
packings, Phys. Rev. E 74 (2006).
[10] H.Y. Sohn, C. Moreland, Effect of particle size distribution on packing density, Can. J.
Chem. Eng. 46 (1968) 162167.
[11] H. Steckel, B.W. Muller, In vitro evaluation of dry powder inhalers.2. Inuence of carrier particle size and concentration on in vitro deposition, Int. J. Pharm. 154 (1997)
3137.
[12] M.J. Donovan, H.D.C. Smyth, Inuence of size and surface roughness of large lactose
carrier particles in dry powder inhaler formulations, Int. J. Pharm. 402 (2010) 19.
[13] N. Islam, P. Stewart, I. Larson, P. Hartley, Effect of carrier size on the dispersion of
salmeterol xinafoate from interactive mixtures, J. Pharm. Sci. 93 (2004) 10301038.
[14] J. Ooi, D. Traini, S. Hoe, W. Wong, P.M. Young, Does carrier size matter? A fundamental study of drug aerosolisation from carrier based dry powder inhalation systems,
Int. J. Pharm. 413 (2011) 19.
[15] X.M. Zeng, G.P. Martin, S.K. Tee, C. Marriott, The role of ne particle lactose on the
dispersion and deaggregation of salbutamol sulphate in an air stream in vitro, Int.
J. Pharm. 176 (1998) 99110.
[16] P. Lucas, K. Anderson, J.N. Staniforth, Protein deposition from dry powder inhalers:
ne particle multiplets as performance modiers, Pharm. Res. 15 (1998) 562569.
[17] M.D. Louey, P.J. Stewart, Particle interactions involved in aerosol dispersion of ternary interactive mixtures, Pharm. Res. 19 (2002) 15241531.
[18] U. Ulusoy, I. Kursun, Comparison of different 2D image analysis measurement techniques for the shape of talc particles produced by different media milling, Miner.
Eng. 24 (2011) 9197.
[19] X.M. Zeng, G.P. Martin, C. Marriott, J. Pritchard, The inuence of carrier morphology
on drug delivery by dry powder inhalers, Int. J. Pharm. 200 (2000) 93106.
[20] B. Alway, R. Sangchantra, P.J. Stewart, Modelling the dissolution of diazepam in lactose interactive mixtures, Int. J. Pharm. 130 (1996) 213224.
[21] S.C. Das, I. Larson, D.A.V. Morton, P.J. Stewart, Determination of the polar and total
surface energy distributions of particulates by inverse gas chromatography, Langmuir 27 (2011) 521523.
[22] S.C. Das, Q. Zhou, D.A.V. Morton, I. Larson, P.J. Stewart, Use of surface energy distributions by inverse gas chromatography to understand mechanofusion processing
and functionality of lactose coated with magnesium stearate, Eur. J. Pharm. Sci. 43
(2011) 325333.
[23] J. Schultz, L. Lavielle, C. Martin, The role of the interface in carbon-ber epoxy composites, J. Adhes. 23 (1987) 4560.
[24] F. Thielmann, M. Naderi, D. Burnett, H. Jervis, Investigation of the acidbase properties of an MCM-supported ruthenium oxide catalyst by inverse gas chromatography
and dynamic vapour sorption, in: S. Jackson, J. Hargreaves, D. Lennon (Eds.), Catalysis in application, Royal Soc. Chem. Great Britain, 2003, p. 237.
[25] D. Traini, P.M. Young, F. Thielmann, M. Acharya, The inuence of lactose
pseudopolymorphic form on salbutamol sulfate-lactose interactions in DPI formulations, Drug Dev. Ind. Pharm. 34 (2008) 9921001.
[26] I.M. Grimsey, J.C. Feeley, P. York, Analysis of the surface energy of pharmaceutical
powders by inverse gas chromatography, J. Pharm. Sci. 91 (2002) 571583.
[27] C.J. Vanoss, R.J. Good, M.K. Chaudhury, Additive and nonadditive surface tension
components and the interpretation of contact angles, Langmuir 4 (1988) 884891.
[28] R. Rowe, P. Sheskey, M. Quinn, Handbook of Pharmaceutical Excipients, 6th ed.
Pharmaceutical Press, London, 2009.
[29] T.J. Coady, H.J. Davies, P. Barnes, Evaluation of a breath actuated pressurized aerosol,
Clin. Allergy 6 (1976) 16.
[30] P.S.A. Sarinas, T.E. Robinson, A.R. Clark, J. Caneld, R.K. Chitkara, R.B. Fick, Inspiratory
ow rate and dynamic lung function in cystic brosis and chronic obstructive lung
diseases, Chest 114 (1998) 988992.
[31] P.J. Wijkstra, T.W. Vandermark, M. Boezen, R. Vanaltena, D.S. Postma, G.H. Koeter,
Peak inspiratory mouth pressure in healthy-subjects and in patients with COPD,
Chest 107 (1995) 652656.
[32] S.R.B. Behara, P. Kippax, M.P. McIntosh, D.A.V. Morton, I. Larson, P. Stewart, Structural inuence of cohesive mixtures of salbutamol sulphate and lactose on
aerosolisation and de-agglomeration behaviour under dynamic conditions, Eur. J.
Pharm. Sci. 42 (2011) 210219.
[33] S.R.B. Behara, I. Larson, P. Kippax, D.A.V. Morton, P. Stewart, An approach to
characterising the cohesive behaviour of powders using a ow titration
aerosolisation based methodology, Chem. Eng. Sci. 66 (2011) 16401648.
[34] R. Ho, S.J. Hinder, J.F. Watts, S.E. Dilworth, D.R. Williams, J.Y.Y. Heng, Determination
of surface heterogeneity of D-mannitol by sessile drop contact angle and nite concentration inverse gas chromatography, Int. J. Pharm. 387 (2010) 7986.
[35] D. Marquardt, An algorithm for least-squares estimation of nonlinear parameters, J.
Soc. Ind. Appl. Math. 11 (1963) 431441.
[36] N. Draper, H. Smith, Applied Regression Analysis, John wiley and Sons, New York, 1981.
[37] M.D. Jones, R. Price, The inuence of ne excipient particles on the performance of
carrier-based dry powder inhalation formulations, Pharm. Res. 23 (2006)
16651674.
[38] Hiylmaz, U. Ulusoy, M. Yekeler, Effect of shape properties of talc and quartz particles on the wettability based separation processes, Appl. Surf. Sci. 233 (2004)
204212.
[39] M. Stojmenovi, J. uni, Measuring elongation from shape boundary, J. Math. Imaging Vis. 30 (2008) 7385.
[40] M. Yekeler, U. Ulusoy, C. Hiylmaz, Effect of particle shape and roughness of talc
mineral ground by different mills on the wettability and oatability, Powder
Technol. (2004) 6878.