Magazine of Concrete Research, 2008, 60, No.

3, April, 165–175
doi: 10.1680/macr.2007.00004

Packing density of cementitious materials:

measurement and modelling
H. H. C. Wong and A. K. H. Kwan

The University of Hong Kong

Packing density has great effect on the performance of a concrete mix. However, little research has been carried
out on the packing density of cementitious materials owing to the lack of an established measurement method.
Herein, a new method, called the wet packing method, is presented. With this method, the packing densities of
blended cementitious materials, consisting of ordinary Portland cement (OPC), pulverised fuel ash (PFA) and
condensed silica fume (CSF), were measured. The results verified the theory that the packing density could be
significantly increased by blending two or even three cementitious materials together. Comparison between the
measured results and the predicted values by three existing packing models, together with some additional tests,
revealed that in the presence of a third-generation superplasticiser, the packing density of CSF is dependent on the
lime content. When the lime-containing OPC and PFA contents are low, the CSF particles would flocculate and
pack rather loosely, but at higher OPC and PFA contents, or with lime added, the CSF would pack to a higher
density. With the effect of lime accounted for, very good agreement between the measured results and the predictions
by the packing models was achieved.

cles and lubricates the concrete mix. Basically, a higher

Introduction
packing density of the aggregate would lead to a re-
The packing density of particles, which is defined as duced paste demand (the amount needed to fill up the
the ratio of the solid volume of the particles to the bulk voids) and a larger amount of excess paste (the amount
volume occupied by the particles, is a fundamental in excess of that needed to fill up the voids) for lubri-
parameter governing the properties of many materials cating and improving the workability of the concrete
made from particles, such as ceramics.1 In the field of mix. Apart from improving the workability, the higher
powder technology, the packing density of particles has packing density of the aggregate may also be utilised to
been a major theme of research.2 As a concrete mix is reduce the paste volume for higher dimensional stabi-
also composed largely of particles, its properties are lity or to reduce the water/cementitious materials
greatly influenced by the packing density of the parti- (w/cm) ratio for higher strength.
cles and hence research on the packing density of the As the overall performance of a concrete could be
constitutive materials can help to improve understand- enhanced by increasing the packing density of the
ing of the behaviour of concrete. aggregate, packing density optimisation of the aggre-
Early in the 1960s, Powers3 studied the effects of the gate has been a key issue in concrete mix design.
packing density of aggregate on the properties of fresh Standard and fairly simple methods for measuring the
concrete. He argued for the excess paste theory, which packing density of aggregate are available in the exist-
states that concrete may be conceived as a mixture of ing codes.4,5 However, the packing density optimisation
aggregate and cement paste and it is the cement paste of aggregate is still largely conducted by means of a
in excess of the amount needed to fill up the voids lengthy trial-and-error process. To reduce the labour
between the aggregate particles that disperses the parti- and time required, some packing models developed in
the field of powder technology have been successfully
applied to predict theoretically the packing density of
Department of Civil Engineering, The University of Hong Kong, aggregate for packing density optimisation.6
Hong Kong, China
Extending the excess paste theory and following the
(MACR-D-07-00004) Paper received 19 December 2006; last revised geometric similarity principle, it may be postulated that
19 June 2007; accepted 21 November 2007 the packing density of the cementitious materials
165

www.concrete-research.com 1751-763X (Online) 0024-9831 (Print) # 2008 Thomas Telford Ltd

Wong and Kwan

should have a similar effect as that of the aggregate: it systematic concrete mix design method based on pack-
is the water in excess of the amount needed to fill up ing density optimisation of the cementitious materials.
the voids between the cementitious materials that dis- In the current paper, a new method of measuring the
perses the particles and lubricates the cement paste. packing density of cementitious materials is presented.
Hence, a higher packing density of the cementitious The method has been applied to measure the packing
materials would lead to a reduced water demand (the densities of blended cementitious materials containing
amount needed to fill up the voids) and a larger amount different proportions of OPC, pulverised fuel ash (PFA)
of excess water (the amount in excess of that needed to and condensed silica fume (CSF). The results so ob-
fill up the voids) for lubricating and improving the tained served two purposes: (a) to demonstrate how the
flowability of the cement paste. Apart from improving packing density could be increased by double- or
the flowability, the higher packing density of the ce- triple-blending different size cementitious materials to-
mentitious materials may also be utilised to reduce the gether; and (b) to check and verify the applicability of
paste volume for higher dimensional stability or to the existing packing models to cementitious materials.
reduce the w/cm ratio for higher strength. During the course of the research, it was found unex-
It is generally believed that blending of ordinary pectedly that the packing density of CSF in the pre-
Portland cement (OPC) with finer cementitious materi- sence of a third-generation superplasticiser is
als would increase the packing density. This has been dependent on the lime content—that is, the content of
demonstrated indirectly by the improved performance Ca(OH)2 in the paste.
of the mortar or concrete produced when OPC is
blended with a finer cementitious material. For exam-
ple, Lange et al.7 have shown that by blending OPC Measuring packing density of cementitious
with a finer blastfurnace-slag-based cement, the water
materials
demand of the cementitious materials could be reduced.
Kwan8 has found that at a w/cm ratio lower than 0.28, There is still no standardised method for measuring
the addition of condensed silica fume, which has an the packing density of any fine powders smaller than
average particle size smaller than 1.0 ìm, could im- 100 ìm, especially cementitious materials. Different re-
prove the workability of the concrete mix. Xie et al.9 searchers used different methods yielding widely differ-
have blended OPC with an ultra-fine pulverised fuel ent results. Yu et al.10 measured the packing density of
ash, which is finer than OPC, to produce a higher- fine powders by pouring into a container and tapping
performance concrete than possible with the use of only under dry condition. This dry packing method is af-
OPC. The improved performance so achieved has been flicted by conglomeration of the particles and its results
attributed to the increase in packing density of the are highly dependent on the degree of compaction
cementitious materials arising from the filling effect of applied. In the context of cementitious materials, it also
the finer material. However, the actual packing densi- has the problem of ignoring the effect of the mixing
ties of the cementitious materials have never been water. DeLarrard11 determined the packing density
directly measured and the suggestion that blending of from the water demand of the cementitious materials
OPC with finer materials could improve packing den- taken as the water volume needed to produce a thick
sity has remained a postulation which needs to be paste. This method is based on the assumption that the
verified. water volume needed to produce a thick paste is the
From the above studies, it is evident that the packing same as the minimum voids volume of the cementitious
density optimisation of cementitious materials is also a materials. Dewar12 determined the packing density of
key issue in concrete mix design. However, there are cementitious materials as the solid concentration of the
two major difficulties. First, there has been no estab- paste formed at standard consistence using a standard
lished method for measuring the packing density of consistence test.13 This method is based on the assump-
cementitious materials. The existing methods for meas- tion that the water volume of the paste formed at
uring the packing density of aggregate, which are based standard consistence is the same as the minimum voids
on dry packing of the particles and thus may be called volume of the cementitious materials. Both DeLarrard’s
dry packing methods, are not applicable to any fine method and Dewar’s method ignore the presence of any
powders smaller than 100 ìm such as cementitious air voids in the paste.
materials because the presence of inter-particle forces In this research, a new method, called the wet pack-
at micrometre size level would cause conglomeration of ing method, has been developed. As the name implies,
the particles leading to unreasonably low packing den- the packing density of the cementitious materials is
sity results.2 Second, owing to lack of data from actual measured under wet condition with water and admix-
packing density measurement, it is not known whether tures, if any, added. Unlike the previous methods, the
the existing packing models are applicable also to ce- new method directly measures the bulk densities of the
mentitious materials. Without experimental and theor- paste formed at different w/cm ratios. At a high w/cm
etical methods for estimating the packing density of ratio, the paste is like a slurry with the cementitious
cementitious materials, it is not possible to develop any materials dispersed as a suspension in water, whereas at
166 Magazine of Concrete Research, 2008, 60, No. 3
 Packing density of cementitious materials: measurement and modelling

a low w/cm ratio, the paste contains a fair amount of cles, the larger particles would occupy a solid volume
air voids because the water added is insufficient to fill within the bulk and porous volume of the smaller
up the voids between the cementitious materials. With particles and thereby also increase the packing density
too much water added at high w/cm ratio or some air (occupying effect). Hence, the blending together of dif-
entrapped at low w/cm ratio, the solid concentration of ferent size particles could improve the packing density.
the cementitious materials tends to be low at both high However, if the two different size particles are not of
and low w/cm ratios. In between, there exists an opti- distinct size, there would be particle interactions such
mum w/cm ratio at which the solid concentration at- that the smaller particles would loosen the packing of
tains a maximum. The maximum solid concentration is the larger particles in contact with each smaller particle
taken to be the packing density of the cementitious (loosening effect) or the larger particles would act like
materials. By measuring the bulk densities of the paste, walls and reduce the packing of the smaller particles in
the presence of any entrapped air voids is implicitly contact with each larger particle (wall effect).
taken into account. To cater for multiple mix of more than two different
The bulk density of a paste formed at a known w/cm size particles (also called multi-component mix or
ratio is measured by filling the paste into a cylindrical polydisperse mix), the above binary packing model has
mould of known volume V and measuring the mass of been extended into a variety of packing models. Most
the paste M inside the mould (the mould used is of of them are based on the linear packing theory14 and
62 mm diameter 3 60 mm height but any other mould may thus be classified as linear packing models. The
of similar size may also be used). The solid volume Vc linear packing theory stipulates that among the multiple
of the cementitious materials may be calculated from components of different sizes there is at least one
the following equation component that is fully packed and dominant (dominant
M in the sense that it divides the other components into
Vc ¼ (1) smaller size and larger size components, which fill
rw uw þ rÆ RÆ þ r
R
þ rª Rª
voids and occupy solid volumes respectively) and that
in which rw is the density of water; rÆ , r
and rª are the specific volume (ratio of bulk volume to solid
the densities of the individual cementitious materials volume) of the mix is the maximum of a series of
denoted by Æ,
and ª; uw is the w/cm ratio by volume; linear functions of the volume fractions of the indivi-
and RÆ , R
and Rª are the volumetric ratios of Æ,
and dual components (each linear function corresponds to
ª to the total cementitious materials. From this, the one component assumed to be dominant). In the 1980s,
voids ratio u and the solid concentration  can be the linear packing theory was improved to account for
evaluated as the particle interactions by Stovall et al.15 and Yu and
u ¼ ð V
Vc Þ=Vc (2) Standish.16
Apart from the linear packing models, there are also
 ¼ Vc =V (3) packing models that deal with the multiple components
by a stepwise approach of considering only two compo-
The values of u and  so obtained are dependent on the nents at a time.12 One stepwise approach is to start with
w/cm ratio of the paste. Plotting the values of u and  the finest component, mix it with the next coarser
against the w/cm ratio (as will be illustrated later in component, evaluate the specific volume of the mixture
Fig. 2), it can be seen that there exist a minimum value using a binary packing model, then, treating the mix-
of u and a maximum value of . The minimum value ture as a single component, mix it with another coarser
of u is the minimum voids ratio while the maximum component, and so on until all components have been
value of  is the packing density of the cementitious included. Alternatively, the mixing sequence may also
materials. proceed from the coarsest component to the finest
component. In theory, other mixing sequences may also
be adopted.
In this research, three most advanced and popular
Packing models
packing models are employed to predict theoretically
In general, the packing density of a mix of monosize the packing density of the mixes of cementitious mate-
spherical particles varies from 0.60 for a non-ordered rials tested so as to investigate whether the test results
and unvibrated mix to 0.74 for an ordered and hexago- agree with theoretical results and to find out whether
nal close packed mix. The packing density of a binary the existing packing models are applicable to cementi-
mix of smaller particles and larger particles is depen- tious materials. They are the packing models developed
dent mainly on the volume fractions of the particles. by Yu et al.17 in 1996, by DeLarrard11 in 1999 and by
When a small amount of smaller particles is added to Dewar12 in 1999.
the larger particles, the smaller particles would fill the The model developed by Yu et al. (hereafter referred
voids between the larger particles and thereby increase to as model A) is a linear packing model with the
the packing density (filling effect) while when a small particle interactions allowed for. The interaction func-
amount of larger particles is added to the smaller parti- tions that allow for the loosening and wall effects were
Magazine of Concrete Research, 2008, 60, No. 3 167
Wong and Kwan

established by curve-fitting of test results. One major Table 1. Mix proportions of the mixes of cementitious materi-
characteristic of this model is the incorporation of als tested
particle shape (in terms of sphericity) in the formulae
Mix No. Mix proportions: % by volume
for the interaction functions.
The model developed by DeLarrard (hereafter re- OPC PFA CSF
ferred to as model B) is also a linear packing model
with the particle interactions allowed for and the inter- 1 100 0 0
action functions established by curve-fitting of test 2 0 100 0
results. Particle shape is not considered in the formula- 3 0 0 100
4 85 15 0
tion. Instead, the degree of compaction applied is taken 5 70 30 0
into account by incorporating a compaction index K in 6 55 45 0
the formulation. 7 40 60 0
The model developed by Dewar (hereafter referred 8 25 75 0
to as model C) is not a linear packing model. It adopts 9 85 0 15
10 70 0 30
a stepwise approach of considering only two compo- 11 55 0 45
nents at a time, starting with the finest component, 12 40 0 60
mixing this component with the next coarser compo- 13 25 0 75
nent to form a combined component and then mixing 14 0 85 15
with another coarser component and so on. Particle 15 0 70 30
16 0 55 45
interactions have been considered in terms of two inter- 17 0 40 60
action parameters m and Z established partly by geo- 18 0 25 75
metric analysis and partly by calibration with test 19 70 15 15
results. 20 55 30 15
The mathematical details of the above three packing 21 40 45 15
22 25 60 15
models can be found in Refs 17, 11 and 12, respec- 23 55 15 30
tively, and are thus not repeated here. For each packing 24 40 30 30
model, a computer program in spreadsheet format has 25 25 45 30
been written. These computer programs are easy to use 26 40 15 45
(provided the user has basic knowledge of packing 27 25 30 45
28 10 45 45
density modelling) and fast (computer time generally 29 25 15 60
within a few minutes), and should be useful tools for
packing density analysis and optimisation.

the method given in BS 812: Part 2,4 their solid densi-

Testing program
ties had been measured to be 3110, 2329 and 2202 kg/
A comprehensive testing program, involving the m3 , respectively. On the other hand, their particle size
packing density measurement of 29 mixes of OPC + distributions are as shown in Fig. 1. For the OPC and
PFA + CSF at various mix proportions by the proposed PFA, the particle size distributions were obtained by
wet packing method, has been launched. The 29 mixes laser diffraction measurement while for the CSF, the
were designed on a volumetric basis with the mix particle size distribution was obtained from the suppli-
proportions of OPC, PFA and CSF expressed in terms er. Both the PFA and CSF were basically spherical in
of volume fractions (the volume fraction of each con- shape and are thus assumed to have a sphericity of
stitutive material is the ratio of the solid volume of the 1.000. The OPC was angular in shape. Its specific sur-
constitutive material to the total solid volume of all face area had been measured in accordance with BS
constitutive materials in the mix). Table 1 presents the 4359: Part 221 as 317 m2 /kg whereas the specific sur-
details of the mix proportions of the mixes tested. It face area calculated from the particle size distribution
should be noted that each mix is given a mix No. and with the particles assumed to be spherical was found to
the 29 mixes may be divided into three categories: non- be 251 m2 /kg. From these two values, the sphericity of
blended mixes (mix No. 1 to 3 for pure OPC, pure PFA the OPC was determined, as stipulated by Yu et al.,17
and pure CSF), double-blended mixes (mix No. 4 to 8 as 251/317 ¼ 0.792.
for OPC + PFA, mix No. 9 to 13 for OPC + CSF and During mixing of the cementitious materials with
mix No. 14 to 18 for PFA + CSF) and triple-blended water to form a paste for testing, a superplasticiser (SP)
mixes (mix No. 19 to 29 for OPC + PFA + CSF). was added to each mix. The SP used was a third-
The OPC, PFA and CSF employed were all com- generation SP based on a cross-linked polycarboxylate
monly used materials obtained from the market. They ether polymer. It has a solid mass content of 20% and a
had been tested to comply with BS EN 197,18 BS 3892: relative density of 1.03. According to the supplier, the
Part 119 and ASTM C1240-05,20 respectively. Using normal dosage of the SP, measured in terms of liquid
168 Magazine of Concrete Research, 2008, 60, No. 3
 Packing density of cementitious materials: measurement and modelling
100
OPC
(note that all w/cm ratios referred to herein are by
PFA volume and that the w/cm ratio by volume is the same
CSF
80 as the water ratio—that is, the ratio of volume of water
to volume of solid).
Cumulative volume: %

From Fig. 2 it can be seen that for the case of pure

60
OPC, at a w/cm ratio of 0.8, the voids ratio u was very
close to the w/cm ratio, indicating that the voids were
40 completely filled with water. Upon reduction of the w/
cm ratio to lower than 0.7, the voids ratio became
20
slightly higher than the w/cm ratio, revealing that the
voids were only partly filled with water and there was a
small amount of air in the voids (note that the differ-
0 ence between the voids ratio and the w/cm ratio is
0·1 1 10 100
Particle size: µm
actually the air ratio—that is, the ratio of volume of air
to volume of solid). As the w/cm ratio was reduced to
Fig. 1. Particle size distributions of the OPC, PFA and CSF lower than 0.6, the voids ratio became significantly
higher than the w/cm ratio. Eventually at a w/cm ratio
of 0.525, the voids ratio reached a minimum value of
mass, should be 0.5–3.0% by mass of the cement. As 0.607 and the solid concentration reached a maximum
the different cementitious materials have different den- value of 0.622. At any w/cm ratio lower than 0.525, the
sities and it is the solid volume rather than the mass water added was not enough to produce a paste and
that is more important, the SP dosage is expressed therefore a large amount of air was entrapped in the
herein in terms of the liquid mass of SP per unit solid mixture, leading to an increase in voids ratio and a
volume of the cementitious materials. The dosage of decrease in solid concentration.
SP used in each mix was 93.3 kg/m3 (same as the upper All the other mixes of cementitious materials tested
limit of normal dosage recommended by the supplier). were found to exhibit similar wet packing behaviour, as
depicted in Fig. 2. In other words, for each mix tested,
as the w/cm ratio decreased from a relatively high
value of about 1.0 to a relatively low value, the voids
Results and discussions ratio decreased to a certain minimum value and then
started increasing whereas the solid concentration in-
Wet packing behaviour of cementitious materials creased to a certain maximum value and then started
The wet packing behaviour of the cementitious mate- decreasing. From these test results, the packing density
rials may be depicted by plotting the voids ratio and of each mix was obtained as the maximum solid con-
the solid concentration of the paste formed against the centration of the mix.
w/cm ratio by volume, as illustrated in Fig. 2 for the
case of pure OPC. In the voids ratio/solid concentration Measured packing densities
diagram, an equality line of u ¼ w/cm ratio is also The measured packing densities of the non-blended
drawn to compare the voids ratio with the w/cm ratio cementitious materials are tabulated in Table 2 while
the measured packing densities of the blended cementi-
0·9 0·70 tious materials are tabulated in the second column of
Voids ratio Table 3.
Solid concentration From Table 2, it is evident that the packing density
0·8 of the OPC was slightly lower than that of the PFA.
0·65
Since the OPC and PFA have similar particle size
Solid concentration

distributions, the difference in packing density could

0·7
Voids ratio

not have been caused by any significant difference in

0·60 particle size. On the other hand, as the OPC particles
0·6

0·55 Table 2. Measured packing densities of pure cementitious

0·5 materials
u ⫽ w/cm ratio
Mix No. Cementitious material Measured packing density
0·4 0·50
0·4 0·5 0·6 0·7 0·8 0·9 1 OPC 0.622
w/cm ratio by volume 2 PFA 0.646
3 CSF 0.397
Fig. 2. Voids ratio/solid concentration diagram of pure OPC
Magazine of Concrete Research, 2008, 60, No. 3 169
Wong and Kwan

Table 3. Comparison between measured results and predictions by packing models

Mix No. Measured Model A Model B Model C

packing density
Predicted Difference: % Predicted Difference: % Predicted Difference: %
packing density packing density packing density

4 0.637 0.628
1.4 0.627
1.6 0.626
1.7

5 0.641 0.633
1.2 0.632
1.4 0.631
1.6
6 0.644 0.637
1.1 0.636
1.2 0.635
1.4
7 0.643 0.640
0.5 0.640
0.5 0.639
0.6
8 0.645 0.643
0.3 0.643
0.3 0.643
0.3
9 0.703 0.688
2.1 0.661
6.0 0.663
5.7
10 0.726 0.659
9.2 0.661
9.0 0.635
12.5
11 0.671 0.577
14.0 0.579
13.7 0.578
13.9
12 0.645 0.514
20.3 0.515
20.2 0.519
19.5
13 0.590 0.463
21.5 0.463
21.5 0.465
21.2
14 0.748 0.707
5.5 0.674
9.9 0.673
10.0
15 0.745 0.658
11.7 0.661
11.3 0.635
14.8
16 0.689 0.577
16.3 0.578
16.1 0.577
16.3
17 0.575 0.513
10.8 0.514
10.6 0.518
9.9
18 0.410 0.463 +12.9 0.463 +12.9 0.465 +13.4
19 0.718 0.694
3.3 0.665
7.4 0.649
9.6
20 0.731 0.698
4.5 0.669
8.5 0.658
10.0
21 0.730 0.702
3.8 0.671
8.1 0.669
8.4
22 0.742 0.704
5.1 0.673
9.3 0.675
9.0
23 0.736 0.659
10.5 0.661
10.2 0.623
15.4
24 0.735 0.658
10.5 0.661
10.1 0.625
15.0
25 0.752 0.658
12.5 0.661
12.1 0.634
15.7
26 0.665 0.577
13.2 0.579
12.9 0.574
13.7
27 0.691 0.577
16.5 0.579
16.2 0.575
16.8
28 0.700 0.577
17.6 0.579
17.3 0.576
17.7
29 0.639 0.514
19.6 0.515
19.4 0.516
19.2

were angular in shape whereas the PFA particles were packing of the CSF particles and thereby substantially
spherical in shape, the difference in packing density reduced the packing density.
was more likely attributable to the difference in particle The results in the second column of Table 3 indicate
shape (angular particles generally pack to a lower den- very clearly that double or triple blending of the ce-
sity than spherical particles22 ). mentitious materials could improve the packing density.
For the CSF, the measured packing density was For the double-blended mixes, the influence of blending
found to be surprisingly low. One possible cause was is shown in Fig. 3 by plotting the packing density
the surface interaction between the SP and the CSF against the volume fraction of the finer material. For
particles. The SP used was a third-generation SP. Un- the double-blended OPC + CSF and PFA + CSF mixes,
like first- and second-generation SPs, a third-generation when a small amount of CSF was added, the packing
SP is a comb-type SP capable of dispersing OPC parti- density of the mix was significantly increased. For
cles not only by electrostatic repulsion but also by instance, when 30% CSF was added to the OPC, the
steric repulsion derived from the grafted polyethylene packing density was increased by 17% from 0.622 to
oxide (PEO) side chains of the SP molecules.23 How- 0.726 and when 15% CSF was added to the PFA, the
ever, it may not have the same effectiveness in the packing density was increased by 16% from 0.646 to
dispersion of CSF particles. When CSF is mixed with 0.748. Such significant increases in packing density
water, the amorphous silica (SiO2 ) of the CSF under- may be attributed to the filling effect of the ultra-fine
goes hydrolysis, resulting in the formation of silanol CSF particles. However, when more CSF was added,
(
Si
OH) groups on the surfaces, and in the presence the packing density of the mix started to decrease
of a third-generation SP, the surface hydroxyl of the because the CSF, which has a lower packing density,
silanol groups would form hydrogen bonds with the had become dominant. For the double-blended OPC +
ether oxygen in the PEO side chains of the SP mole- PFA mixes, the packing density of the mix increased
cules, leading to PEO adsorption on the CSF surfaces slightly as more and more PFA was added. Since the
and flocculation of the CSF particles.24 The PEO ad- OPC and the PFA have similar size distributions, the
sorption and flocculation might have hindered dense PFA had little filling effect when added to the OPC.
170 Magazine of Concrete Research, 2008, 60, No. 3
 Packing density of cementitious materials: measurement and modelling
0·8
increased to higher than possible with only double
OPC ⫹ PFA
OPC ⫹ CSF
blending of OPC with either PFA or CSF. These results
0·7 PFA ⫹ CSF provide solid experimental evidence to verify the
long-held postulation that blending different size
cementitious materials together can improve the pack-
ing density.
Packing density

0·6

Predictions by packing models

0·5 The predicted packing densities of the blended ce-
mentitious materials (mix No. 4 to 29) by the three
packing models (models A, B and C) are tabulated in
0·4 the third to eighth columns of Table 3 for comparison
with the respective measured packing densities. When
model A was applied to evaluate the predicted packing
0·3 densities, in order to account for the effect of particle
0 20 40 60 80 100 shape, the OPC, PFA and CSF were assigned sphericity
Volume fraction of finer material: %
values of 0.792, 1.000 and 1.000, respectively. When
Fig. 3. Packing densities of double-blended mixes model B was applied to evaluate the predicted packing
densities, in order to incorporate the effect of compac-
tion, different typical values of the compaction index K
were tried. It was found that a K-value of infinity (the
The slight increase in packing density was attributable value for the case of perfect compaction) would yield
to the gradual replacement of the angular and less the closest agreement between the predicted and meas-
densely packed OPC particles with the spherical and ured packing densities (this is in fact an indication of
more densely packed PFA particles. the high effectiveness of the proposed wet packing
For the triple-blended mixes, the influence of blend- method in the mixing and packing of the cementitious
ing is shown in Fig. 4 by plotting the packing densities materials). A K-value of infinity was therefore adopted
of the OPC + PFA + CSF mixes in the form of a when applying model B. Despite the apparent differ-
ternary packing density diagram. From this diagram, it ences between the three packing models, the results
can be seen that by triple blending 25% OPC, 45% presented in Table 3 clearly show that the three packing
PFA and 30% CSF, the packing density could be in- models actually yielded very similar predicted packing
creased to 0.752 (21% higher than the packing density densities differing by not more than 6.5% in all the
of pure OPC). In fact, there is a small region in the mixes studied.
diagram near the lower right-hand corner with the Comparing the predicted packing densities to the
packing density consistently higher than 0.750. Hence, measured packing densities, it can be seen that for the
by triple blending of OPC with both PFA and CSF to double-blended OPC + PFA mixes (mix No. 4 to 8),
utilise the better particle shape of PFA and the filling the predicted packing densities by all the three packing
effect of CSF, the packing density of the mix could be models agreed almost exactly with the corresponding
measured packing densities, with maximum errors of
only 1.4%, 1.6% and 1.7% when models A, B and C
100% OPC were applied, respectively. For the double-blended OPC
+ CSF mixes (mix No. 9 to 13), the differences be-
tween the predicted packing densities and the corre-
0·6

sponding measured packing densities were somehow

50

much larger; at a CSF content of 15%, the differences

were about 6% but as the CSF content increased to
75%, the differences gradually increased to around
22%. For the double-blended PFA + CSF mixes (mix
No. 14 to 18), the differences between the predicted
packing densities and the corresponding measured
packing densities were also quite large, ranging from
0·55 about 10% at a CSF content of 15% to slightly higher
0·7

0
0·7

0·
0·7
0·

60
than 16% at higher CSF contents. For the triple-blended
00

0·500
65

50

0
00
0

0·450
OPC + PFA + CSF mixes (mix No. 19 to 29), the
predicted packing densities were all lower than the
100% CSF 100% PFA
corresponding measured packing densities by signifi-
Fig. 4. Ternary packing density diagram of OPC + PFA cant differences, which were generally larger at higher
+ CSF CSF contents. The relatively small differences between
Magazine of Concrete Research, 2008, 60, No. 3 171
Wong and Kwan

the predicted and measured packing densities at low presence of lime and the increase in pH might have
CSF content and the substantially larger differences reduced the PEO adsorption and CSF flocculation, and
between the predicted and measured packing densities consequently increased the packing density of the CSF.
at high CSF content indicated that something might To study the effect of lime on the packing density of
have happened at high CSF content that had not been pure CSF (mix No. 3), different amounts of lime (in
fully accounted for. the form of calcium hydroxide powders) were added to
the water before mixing with the CSF and the resulting
Effect of lime on packing density of CSF packing densities of the CSF measured. Four different
The three mixes with the highest CSF contents are amounts of lime, namely 0.74, 7.41, 37.05 and 74.10
mix No. 3 containing 100% CSF, mix No. 13 contain- g/l (corresponding to 0.01, 0.1, 0.5 and 1.0 mole/l,
ing 25% OPC + 75% CSF, and mix No. 18 containing respectively), had been added to four separate samples
25% PFA + 75% CSF. They have measured packing of the CSF, each with the same amount of SP (93.3 kg/
densities of 0.397, 0.590 and 0.410, respectively. While m3 ) added. The resulting packing densities of the CSF
pure CSF has a rather low packing density of 0.397, were measured to be 0.461, 0.479, 0.520 and 0.522,
the addition of 25% OPC dramatically increased the respectively, as listed in Table 4. These results showed
packing density by 49% to 0.590. Such increase in that the presence of lime could improve the packing
packing density was much larger than attributable to density of CSF by as much as 31%. Basically, the
the occupying effect of the OPC particles. On the other packing density of CSF would increase steadily with
hand, the addition of 25% PFA only increased the the lime content and eventually reach a constant and
packing density by 3% to 0.410. As the OPC and PFA maximum value at a sufficiently high lime content.
have similar size distributions and thus should have Herein, the maximum packing density so achieved un-
more or less the same occupying effects, the large der the lime-sufficient condition of having a suffi-
difference in packing density between the mix with ciently high lime content is called lime sufficient
25% OPC and the mix with 25% PFA revealed that the packing density.
addition of OPC and the addition of PFA have different The packing densities of mix No. 13 containing 25%
effects on the packing density, which could not be OPC + 75% CSF, mix No. 16 containing 55% PFA +
explained just from the physical packing point of view. 45% CSF, mix No. 17 containing 40% PFA + 60% CSF
Chemistry might have played a certain role causing the and mix No. 18 containing 25% PFA + 75 % CSF were
packing density of CSF to be changed when OPC or also measured again with different amounts of lime
PFA was added. added. As before, different amounts of lime had been
One major chemical difference between OPC and added to separate samples of each mix and the same
PFA is that when OPC is mixed with water, it under- amount of SP (93.3 kg/m3 ) was added to each sample.
goes hydration and releases a substantial amount of The measured packing densities at different lime con-
lime as a by-product of hydration into the water tents are presented in Table 4. It is seen that for mix
whereas when PFA is mixed with water, only the small No. 13, the packing density was rather insensitive to
lime content in the PFA (usually a few per cent) is the addition of lime, revealing that the lime liberated
dissolved into the water. Hence, OPC and PFA would by the 25% OPC in the mix was already sufficient to
liberate different amounts of lime into the water. Since avoid reduction in the packing density of CSF owing to
lime is alkaline, the pH of the water increases as lime lime deficiency. Among the other mixes, mix No. 16
goes into the water and the increase in pH should be was also insensitive to the addition of lime but both
larger with OPC added than with PFA added. Accord- mix No. 17 and 18 had their packing densities signifi-
ing to Somasundaran and Zhang,24 pH has a strong cantly increased when lime was added. This revealed
effect on the hydrolysis of silica and because of gradual that without OPC, at least 55% PFA would be needed
dissociation of the silanol groups, both the adsorption to produce a lime-sufficient condition for avoiding re-
density of PEO on CSF surfaces and the flocculation of duction in the packing density of CSF. Nevertheless, in
CSF particles would decrease as the pH increases. The practice, since the OPC content in the mix is unlikely

Table 4. Measured packing densities of mixes containing CSF in the presence of lime

Mix No. Mix proportions: % by volume Amount of lime added to the mixing water: g/l

OPC PFA CSF 0 0.74 7.41 37.05 74.10

3 0 0 100 0.397 0.461 0.479 0.520 0.522

13 25 0 75 0.590 0.594 0.593 — —
16 0 55 45 0.689 0.681 0.683 — —
17 0 40 60 0.575 0.589 0.625 0.631 —
18 0 25 75 0.410 0.457 0.576 0.580 —

172 Magazine of Concrete Research, 2008, 60, No. 3

 Packing density of cementitious materials: measurement and modelling

to be lower than 25% by volume, there should be no packing densities by the three packing models agreed
lime deficiency problem that would reduce the packing very closely with the measured packing densities by the
density of CSF. proposed wet packing method. In general, the differ-
ences between the predicted and measured packing
Predictions by packing models with effect of lime densities were well within several per cent and on a par
accounted for with the expected experimental errors in the packing
density measurements. More importantly, the differ-
From the above, the lime-sufficient packing densities
ences no longer increased with the CSF content. For
of mix No. 3, 13, 16, 17 and 18 (taken as the respective
evaluating the accuracies of the three packing models,
maximum packing densities within the ranges of lime
the average differences between the predicted and
contents tested) were obtained as 0.522, 0.594, 0.689,
measured packing densities have been calculated as
0.631 and 0.580, respectively. To take into account the

0.6%,
1.2% and
2.5%, and the average absolute
effect of lime, for pure CSF, the original packing den-
differences between the predicted and measured pack-
sity of 0.397 was replaced by the lime-sufficient pack-
ing densities have been calculated as 1.6%, 2.3% and
ing density of 0.522 as input parameter entered into the
2.9%, for models A, B and C, respectively. Although
packing models for more realistic predictions of the
model A appeared to be the most accurate, all the
packing densities of the various mixes which were
packing models should be acceptable for practical en-
mostly under lime-sufficient condition. Furthermore,
gineering applications. The good agreement between
for mix No. 13, 16, 17 and 18, which might have their
the predicted and measured results also verified the
packing densities afflicted by lime deficiency, their
validity and accuracy of the proposed wet packing
measured packing densities were replaced by their re-
method.
spective lime-sufficient packing densities so that direct
comparison with the predictions by the packing models
based on lime-sufficient condition could be made. Ta-
ble 5 presents the measured and predicted packing
Conclusions
densities of the various mixes after the above changes.
From the comparison in Table 5, it can be seen that A new wet packing method for measuring the pack-
with the effect of lime accounted for, the predicted ing density of cementitious materials in a paste has

Table 5. Comparison between measured results and predictions by packing models with effect of lime accounted for

Mix No. Measured Model A Model B Model C

packing density
Predicted Difference: % Predicted Difference: % Predicted Difference: %
packing density packing density packing density

4 0.637 0.628
1.4 0.627
1.6 0.626
1.7

5 0.641 0.633
1.2 0.632
1.4 0.631
1.6
6 0.644 0.637
1.1 0.636
1.2 0.635
1.4
7 0.643 0.640
0.5 0.640
0.5 0.639
0.6
8 0.645 0.643
0.3 0.643
0.3 0.643
0.3
9 0.703 0.699
0.6 0.678
3.6 0.680
3.3
10 0.726 0.749 +3.2 0.743 +2.3 0.703
3.2
11 0.671 0.685 +2.1 0.688 +2.5 0.683 +1.8
12 0.645 0.632
2.0 0.634
1.7 0.635
1.6
13 0.594 0.586
1.3 0.587
1.2 0.589
0.8
14 0.748 0.719
3.9 0.695
7.1 0.692
7.5
15 0.745 0.747 +0.3 0.748 +0.4 0.704
5.5
16 0.689 0.684
0.7 0.687
0.3 0.681
1.2
17 0.631 0.631 +0.0 0.633 +0.3 0.634 +0.5
18 0.580 0.586 +1.0 0.587 +1.2 0.589 +1.6
19 0.718 0.705
1.8 0.683
4.9 0.681
5.2
20 0.731 0.710
2.9 0.687
6.0 0.690
5.6
21 0.730 0.714
2.2 0.691
5.3 0.703
3.7
22 0.742 0.716
3.5 0.693
6.6 0.704
5.1
23 0.736 0.748 +1.6 0.746 +1.4 0.697
5.3
24 0.735 0.748 +1.8 0.748 +1.8 0.701
4.6
25 0.752 0.748
0.5 0.748
0.5 0.704
6.4
26 0.665 0.685 +3.0 0.688 +3.5 0.676 +1.7
27 0.691 0.685
0.9 0.688
0.4 0.678
1.9
28 0.700 0.685
2.1 0.687
1.9 0.680
2.9
29 0.639 0.632
1.1 0.634
0.8 0.633
0.9

Magazine of Concrete Research, 2008, 60, No. 3 173

Wong and Kwan

been developed and successfully applied to non-

Acknowledgement
blended, double-blended and triple-blended mixes con-
taining different proportions of OPC, PFA and CSF. The work described in the current paper was fully
From the packing density results of the non-blended supported by a grant from the Research Grants Council
mixes, it was found that, although the OPC and PFA of the Hong Kong Special Administrative Region,
have similar particle size distributions, owing to the China (Project No. HKU 7139/05E).
better spherical shape of the PFA particles, the PFA has
a higher packing density than the OPC. However,
although the CSF particles are also spherical in shape, References
the CSF turned out to have the lowest packing density.
1. Reed J. S. Principles of Ceramics Processing, 2nd edn. Wiley,
This may be attributed to PEO adsorption on the CSF New York, USA, 1995.
surfaces and flocculation of the CSF particles arising 2. Fayed M. E. and Otten L. (eds) Handbook of Powder Science
from hydrogen bonding between the silanol groups on and Technology, 2nd edn. Chapman & Hall, New York, USA,
the CSF surfaces and the PEO side chains of the third- 1997.
generation SP added. On the other hand, the packing 3. Powers T. C. The Properties of Fresh Concrete. Wiley, New
York, USA, 1968.
density results of the blended mixes provide solid evi- 4. British Standards Institution. Testing of Aggregates, Part
dence to verify the long-held postulation that double 2: 1995 Method of Determination of Density. BSI, UK, 1995,
and even triple blending of different size cementitious BS 812.
materials together can significantly improve the pack- 5. American Society for Testing and Materials. Standard Test
ing density. A maximum packing density of 0.752 has Method for Bulk Density (‘Unit Weight’) and Voids in Aggregate.
ASTM, West Conshohocken, PA, 2003, ASTM C 29/C 29M-97.
been achieved by blending 25% OPC + 45% PFA + 6. Goltermann P., Johansen V. and Palbøl L. Packing of ag-
30% CSF. gregates: an alternative tool to determine the optimal aggregate
The three packing models developed by Yu et al.,17 mix. ACI Materials Journal, 1997, 94, No. 5, 435–443.
DeLarrard11 and Dewar12 have been employed to pre- 7. Lange F., Mörtel H. and Rudert V. Dense packing of cement
dict the packing densities of the mixes tested and the pastes and resulting consequences on mortar properties. Cement
and Concrete Research, 1997, 27, No. 10, 1481–1488.
predicted packing densities so obtained were compared 8. Kwan A. K. H. Use of condensed silica fume for making high-
to the corresponding measured packing densities. It was strength, self-consolidating concrete. Canadian Journal of Civil
found that although acceptable agreement between the Engineering, 2000, 27, No. 4, 620–627.
predicted and measured packing densities has been 9. Xie Y. J., Liu B. J., Yin J. and Zhou S. Q. Optimum mix
achieved at low CSF content, there were substantial parameters of high-strength self-compacting concrete with ultra-
pulverized fly ash. Cement and Concrete Research, 2002, 32,
errors at high CSF content. In order to resolve this No. 3, 477–480.
problem, additional tests were carried out to measure 10. Yu A. B., Bridgwater J. and Burbidge A. On the modelling
the packing densities of pure CSF and several other of the packing of fine particles. Powder Technology, 1997, 92,
mixes with different amounts of lime added. It was No. 3, 185–194.
discovered that the packing density of CSF was highly 11. DeLarrard F. Concrete Mixture Proportioning: A Scientific
Approach. E&FN Spon, London, UK, 1999.
dependent on the lime content and that the low packing 12. Dewar J. D. Computer Modelling of Concrete Mixtures. E&FN
densities of pure CSF and other mixes with high CSF Spon, London, UK, 1999.
contents were attributable to lime deficiency (not having 13. British Standards Institution. Methods of Testing Cement
sufficient lime to avoid reduction in the packing density Part 3: 1995 Determination of Setting Time and Soundness.
of CSF owing to PEO adsorption and CSF flocculation). BSI, London, 1995, BS EN 196.
14. Westman A. E. R. and Hugill H. R. The packing of particles.
Nevertheless, with at least 25% OPC or 55% PFA Journal of American Ceramic Society, 1930, 13, No. 10, 767–
added, the OPC or PFA would liberate sufficient lime 779.
into the water to alleviate such lime deficiency. 15. Stovall T., Delarrard F. and Buil M. Linear packing density
To account for the effect of lime, the lime-sufficient model of grain mixtures. Powder Technology, 1986, 48, No. 1,
packing density of CSF was input into the packing 1–12.
16. Yu A. B. and Standish N. Porosity calculations of multi-
models to evaluate the predicted packing densities of component mixtures of spherical particles. Powder Technology,
the various mixes tested and the measured packing 1987, 52, No. 3, 233–241.
densities of the mixes afflicted by lime deficiency were 17. Yu A. B., Zou R. P. and Standish N. Modifying the linear
replaced by their respective lime-sufficient packing packing model for predicting the porosity of nonspherical parti-
densities. After these changes, the predicted packing cle mixtures. Industrial and Engineering Chemistry Research,
1996, 35, No. 10, 3730–3741.
densities by the three packing models agreed with the 18. British Standards Institution. Cement Part 1: 2000 Compo-
measured packing densities by the proposed wet pack- sition, Specifications and Conformity Criteria for Common Ce-
ing method to within an average absolute difference of ments. BSI, London, 2000, BS EN 197.
only 3%. Such good agreement verified the applicabil- 19. British Standards Institution. Pulverized-Fuel Ash Part 1:
ity of all the three packing models to cementitious 1997 Specification for Pulverized-Fuel Ash for Use with Port-
land Cement. BSI, London, 1997, BS 3892.
materials (provided of course the effect of lime has 20. American Society for Testing and Materials. Standard
been accounted for) and the validity and accuracy of Specification for Silica Fume Used in Cementitious Mixtures.
the proposed wet packing method. ASTM, West Conshohocken, PA, 2005, ASTM C 1240-05.

174 Magazine of Concrete Research, 2008, 60, No. 3

 Packing density of cementitious materials: measurement and modelling

21. British Standards Institution. Determination of the Specific 24. Somasundaran P. and Zhang L. Modification of silica–water
Surface Area of Powders Part 2: 1982 Recommended Air Per- interfacial behavior by adsorption of surfactants, polymers, and
meability Methods. BSI, London, 1982, BS 4359. their mixtures. In Surfactant Science Series Volume 90: Adsorp-
22. Kwan A. K. H. and Mora C. F. Effects of various shape tion on Silica Surfaces (Papirer E. (ed.)). Marcel Dekker, New
parameters on packing of aggregate particles. Magazine of Con- York, USA, 2000, ch. 14, pp. 441–462.
crete Research, 2001, 53, No. 2, 91–100.
23. Sakai E., Yamada K. and Ohta A. Molecular structure and
dispersion-adsorption mechanisms of comb-type superplastici-
zers used in Japan. Journal of Advanced Concrete Technology, Discussion contributions on this paper should reach the editor by
2003, 1, No. 1, 16–25. 1 October 2008

Magazine of Concrete Research, 2008, 60, No. 3 175

Copyright of Magazine of Concrete Research is the property of Thomas Telford Ltd and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.