Cement and Concrete Research 33 (2003) 155 – 164

Porosity, pore size distribution and in situ strength of concrete

Rakesh Kumara, B. Bhattacharjeeb,*
a
Bridges Division, Central Road Research Institute, New Delhi 110 020, India
b
Department of Civil Engineering, Indian Institute of Technology-Delhi, Hauz Khas, New Delhi 110 016, India
Received 23 July 1997; accepted 24 July 2002

Abstract

In this study, in situ strength of concrete was determined through compression test of cores drilled out from laboratory cast beams. The
apparent porosity and pore size distribution of the same concrete were determined through mercury intrusion porosimetry, performed on
small-drilled cores. The normal-strength concrete mixes used in the experimental investigation were designed to exhibit a wide variation in
their strengths. To ensure further variation in porosity, pore size distribution and strength, two modes of compaction, two varieties of coarse
aggregates, different levels of age, curing period and exposure condition of concrete were also introduced in experimental scheme. With the
data so generated, an appraisal of the most frequently referred relationships involving strength, porosity and pore size of cement-based
materials was carried out. Finally, a new empirical model relating the in situ strength of concrete with porosity, pore size characteristics,
cement content, aggregate type, exposure conditions, etc., is presented.
D 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Mercury porosimetry; Pore size distribution; Pore system; Concrete; Cement content

1. Introduction Water –cement ratio also governs the transition zone por-
osity in concrete [4]. Thus, there are a number of well-
Concrete prepared with hydraulic cement binder can be established strength versus water – cement ratio relation-
regarded as a chemically bonded ceramic. The hydration ships, which indirectly relate the strength of concrete with
reaction of cement results in a product consisting of solid its pore system characteristics [4,8,9]. These relationships
and a pore system [1]. Pores are thus inherent to concrete. serve their purpose very well in the design of concrete
Pores in concrete can also result from inadequate compac- mixes. A few of these relationships take into account air
tion. This pore system governs the most important prop- content and degree of hydration of concrete. However, such
erties of concrete, notably its strength [2,3]. Well-compacted indirect relationships do not take into account the pores
concrete prepared with hard low-porosity aggregates may be present in hardened concrete in structure due to inadequate
assumed to be a multiphase material consisting of coarse compaction, etc. Further, the pore system in concrete also
aggregates embedded in mortar matrix. The mortar matrix changes with degree of hydration and chemical changes due
consists of fine aggregates, the solid cement hydrates, to aggressive environments, etc. A direct relationship, on the
unhydrated cement, etc., and the pore system [4]. The pore other hand, can facilitate the strength estimation of in situ
system present in the mortar of concrete, however, is concrete from the knowledge of its pore system character-
markedly different from the pores of well-compacted mortar istics. For the purpose of mix design, however, strength/
prepared independently using identical proportions of the water – cement ratio relationships are more useful. The most
relevant ingredients. The above difference in the two pore important characteristics of pore system are porosity and
systems is due to the transition zone pores present at pore size distribution, which can be determined through
mortar – aggregate interface [5 – 7]. Capillary porosity of mercury intrusion porosimetry (MIP). However, MIP results
hardened cement paste depends on water-to-cement ratio. are affected by a number of factors and the same must be
suitably accounted for in the experimental procedure adop-
ted [10 – 14]. Secondly, the smallest size of pore, in which
*
Corresponding author. Tel.: +91-11-6591193; fax: +91-11-6862037. mercury can intrude, depends upon the maximum intrusion
E-mail address: [email protected] (B. Bhattacharjee). pressure applied. Consequently, extent of porosity that can

0008-8846/03/$ – see front matter D 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 8 - 8 8 4 6 ( 0 2 ) 0 0 9 4 2 - 0
156 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164

be determined by porosimetry test depends upon the nature porosity and pore size distribution of concrete. It was also
of the pores, the size of the smallest pore likely to be desired that the strength data so generated should adequately
encountered in the material and the maximum intrusion cover the range of strength usually encountered for normal-
pressure applied. strength concrete. High-strength concrete, produced using
The pore system in cement-based materials consists of water-reducing agent and pozzolanic microfiller, was left
four types of pores. These are: (a) gel pores, which are out of the scope of this work.
micropores of characteristic dimension 0.5 – 10 nm; (b)
capillary pores, which are mesopores with average radius 2.2. Mix proportions and experimental factors
ranging from 5 to 5000 nm; (c) macropores due to deliber-
ately entrained air; and (d) macropores due to inadequate Strength and porosity of concrete depend upon water –
compaction. In concrete, in addition to the above pores, cement ratio. Considering the practical limits of water –
there can be cracks at aggregate – mortar interface due to cement ratios (0.38 – 0.65) for workable concrete prepared
shrinkage. The gel pores, which are mostly of 1.5 –2.0 nm without water-reducing agent, six concrete mixes were
size, do not influence the strength of concrete adversely designed so as to ensure adequate variation in strength. Same
through its porosity, although these pores are directly related ordinary Portland cement was used throughout this invest-
to creep and shrinkage. Capillary pores and other larger igation. Similarly, throughout the investigation, the same
pores, on the other hand, are responsible for reduction in land-quarried local sand confirming the Zone II of British
strength and elasticity, etc. [4,7,15 –17]. Thus, while dealing Standard and potable laboratory tap water were used as fine
with an empirical strength –porosity relationship of con- aggregate and mixing water, respectively. To ensure further
crete, contribution of the gel pores in the overall porosity variation in strength and porosity, two modes of compaction,
and pore size distribution of concrete can be neglected, namely, compaction through mechanical vibration and man-
without introducing any significant error. Hence, to deter- ual compaction through tamping rod, were adopted in the
mine the pore system characteristics influencing the experimental programme. Two types of graded coarse aggre-
strength, the maximum pressure in the porosimetry test gates of 20 mm maximum size were also used: one being the
must be sufficient to cause intrusion of mercury in the crushed quartzite rock while the other was obtained in the
smallest capillary pore. In mercury porosimetry, a major laboratory by crushing common surface clay bricks. The
part of gel pores remains nonintruded. Further, the closed former was hard with negligibly small porosity and the latter
pores also remain nonintruded. One other limitation pointed was soft and porous with a water absorption value of 13.7%.
out as regard to mercury porosimetry is that it measures Degree of hydration, i.e., the age, curing, exposure to
entry sizes rather than true pore size that is related to ink aggressive environment, etc., also affects the strength and
bottle effect [18]. Thus, the porosity determined, as above, pore system of concrete. Thus, further variation in strength,
is apparent porosity. porosity and pore size distribution of concrete was ensured
A number of relationships relating strength of cement- by adopting the age, curing period, exposure to acidic water
based materials with their pore system characteristics are and thermal exposure as experimental factors. Two levels of
available in the literature. In this paper, firstly, the results of age, namely, 28 and 84 days, and two levels of curing
an experimental investigation are presented, whereby data period, i.e., moist curing for 1 and 27 days, were also used.
on in situ cube compressive strength of concrete—estimated Similarly, two levels of exposure conditions, namely, expo-
through compression test of cores drilled out from labor- sure to open air and exposure to acidic environment of pH
atory cast beams—are generated, together with MIP data for 4 –5, and three levels of thermal exposure, i.e., exposure to
the same concrete. This is followed by an appraisal whereby 26 (room temperature), 300 and 600
C were used. As
most frequently referred relationships involving strength factorial experiment design would have resulted in a large
and pore system characteristics of cement-based materials number of samples, the levels of the factors were adopted in
are evaluated for their suitability in strength estimation of in a restricted way. Details of the samples prepared are given
situ concrete. Finally, a new empirical relationship for in situ subsequently.
strength of concrete is proposed, which takes apparent
porosity, pore size characteristics (corresponding to 33,000 2.3. Casting
psi intrusion pressure) and binder content of the concrete
into account. Thirty-two concrete beams of dimensions 1000
200

100 mm as shown in Fig. 1 were cast using six designed
mixes designated as Mix1 –Mix6. Four more beams were
2. Experimental investigation also cast using crushed brick coarse aggregates instead of
quartzite aggregates. The mix proportions used for casting
2.1. Objectives and scope these beams were the same as those of Mix1 and Mix2,
except that additional absorption of water by the aggregates
The main objectives of this experimental investigation was also accounted for in this case. These mixes are
were to generate sufficient data on strength, apparent designated as Mix7 and Mix8, respectively. Half of the total
 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164 157

2.6. MIP

A number of factors affect the MIP results. Most import-

ant among them are the method of sampling, number of
sample, sample conditioning, rate of pressure application,
maximum intrusion pressure applied, values of contact
angle and surface tension of mercury used in Washburn’s
equation, etc. [11]. Again, the dimensions of the penetrom-
eter restrict the maximum size of the sample used in MIP.
Hence, only small samples in the form of core, crushed
Fig. 1. Concrete beam specimen. chunk, tablets, etc., can be used in porosimetry. Therefore,
to obtain representative results, within desirable accuracy,
the sample size, i.e., number of samples to be used for a
given concrete specimen, needs to be statistically ascer-
number of beams cast with each mix was compacted
tained. Therefore, prior to actual investigation, a preliminary
through mechanical mode of compaction (using immer-
experimental investigation was carried out before arriving at
sion-type needle vibrator) and the other half was compacted
a suitable method of sample collection, number of sample to
manually with the help of a 25-mm-diameter tamping rod.
be tested, form of the sample, rate of pressure application,
These modes of compaction are abbreviated as VC and HC,
etc. [10 – 12]. Further, for this study, contact angle and
respectively. All the beams were demoulded after 24 h of
surface tension values of mercury were adopted from
casting and cured as explained in Section 2.4.
available literature. In addition to the above major factors,
certain minor factors—such as expansion of sample cell
2.4. Curing and exposure conditions
under pressure, differential mercury compression, sample
compression and hydrostatic head of mercury, etc.—also
To ensure adequate curing, the beam specimens after
affect the MIP results to a limited extent. The effects of
demoulding were wrapped under wet hessian cloth, wetted
these factors are of minor consequences; hence, their affects
continuously by sprinkling water. Most of the beams were
were neglected [11].
cured for 27 days, but a few beams were cured for 1 day and
left exposed to atmosphere prior to sample extraction. Some
2.6.1. Sample preparation, number of sample and con-
of the beams after 27 days of curing were submerged in
ditioning and tests parameters
acidic water (pH 4– 5) for a period of 84 days. Due to size
The number of samples required in MIP to obtain the
restriction of the furnace, cores drilled out from some of the
average value of porosity and mean distribution radius
beams, instead of beams themselves, were subjected to 300
(defined later) within ± 15% accuracy was obtained sta-
and 600
C temperatures. The details of concrete mix
proportions, cube compressive strength of concrete, number
of beams, etc., are presented in Table 1. Table 1
Mix proportions and details of concrete beam specimens cast
2.5. Test for in situ compressive strength Mix Mix Coarse 28-Day Number of
proportions designation aggregate cube beam cast
In situ strength of concrete in beams was determined (C:S:A:w/c) type compressive using modes
strength of compaction
through core test. From each beam, three cores, 75 mm in (MPa)
diameter and 100 mm in length, were drilled. The cores VC HC
were drilled perpendicular to the direction of the casting as 1:2.5:5.1:0.65 Mix1 Quartzite 31 1 1
shown in Fig. 1. The compression test on dry cores was 1:2.2:4.2:0.56 Mix2 Quartzite 34 7a 7a
1:1.8:3.9:0.51 Mix3 Quartzite 35 1 1
conducted on universal testing machine by ensuring a rate of
1:1.5:3.6:0.46 Mix4 Quartzite 38 1 1
loading of 12 MPa/min as per standard practice [19,20]. The 1:1.3:3.2:0.42 Mix5 Quartzite 43 1 1
representative in situ cube strength of concrete in the beams 1:1.1:2.7:0.38 Mix6 Quartzite 45 5b 5b
was estimated from average failure load of three cores 1:2.5:5.1:0.65 Mix7 Broken brick 14 1 1
according to Eq. (1) [20]: 1:2.2:4.2:0.56 Mix8 Broken brick 16 1 1
a
Includes: 1(27 days cured and tested on 28th day) + 1(1 day cured and
2:5fl tested on 28th day) + 1(27 days cured and tested on 84th day) + 1(1 day
fcu ¼ : ð1Þ
1:5 þ 1=l cured and tested on 84th day) + 1 (27 days cured and subjected to acidic
environment) + 1(27 days cured and subjected to 300
C) + 1(27 days cured
and subjected to 600
C) = 7.
In Eq. (1), fcu is the estimated in situ cube compressive b
Includes: 1(27 days cured and tested on 28th day) + 1(27 days cured
strength; fl is the determined cylinder compressive strength and subjected to acidic environment) + 1(27 days cured and subjected to
of a core with length/diameter = l. For the present study, l 300
C) + 1(27 days cured and subjected to 600
C) + 1(1 day cured and
was 4/3 for all the cores. tested on 84th day) = 5.
158 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164

tistically using Stein’s two-stage formula. This was found to of six samples collected from a beam, average intrusion
be six [10 – 12]. Thus, six numbers of samples were tested curve was obtained for concrete in each beam. Thus, 36
and average results were taken as the representative of a such average intrusion curves were produced [11]. The
concrete specimen. Preliminary investigation carried out apparent porosity of the concrete (corresponding to 33,000
revealed that a small-cored sample of the concrete is the psi intrusion pressure) for each of the six samples taken
most appropriate form for MIP study [10 – 12]. It was also from a beam was calculated using the individual cumulative
observed that rate of pressure application has little effect on intrusion volume and the relevant weight measurements for
the measured porosity and pore size distribution [10 – 12]. a sample, and averaged. Average apparent porosity of
This was also confirmed by other research works [21,22]. concrete in each of the 36 beams was thus obtained. From
Therefore, in this study, six small cores of 25 mm diameter the average intrusion curves, the values of mean distribution
and 15 – 25 mm length were drilled out from each of the radius, rm, was estimated according to the equation given
concrete beams. Oven drying is reported to be the best below [23]:
method for sample conditioning and the corresponding X
i¼n
contact angle to be adopted is 117
[10,13,14]. Therefore, Vi lnri
the samples were dried in an oven at 105 –110
C for 24 h i¼1
lnrm ¼ ð3Þ
and stored in a desiccator prior to testing. X
i¼n
Vi
i¼1
2.6.2. Testing
Testing was performed on Quantachrome Autoscan-33 where, for the continuous intrusion curve divided into n
mercury porosimeter having a pressure range from subam- discrete radii ranges, Vi is the incremental intrusion of
bient to 33,000 psi. The contact angle and the surface mercury corresponding to ith radius range represented by
tension of mercury were assumed to be 117
and 0.484 N/ the mean radius ri. The porosity belonging to the pore size
m, respectively, for the oven-dried samples [10]. Conse- ranges greater than 106, 53– 106 and 10.6 –53 nm, and less
quently assuming the cylindrical pores, the Washburn’s than 10.6 nm is also calculated from these intrusion curves.
equation yields: These results along with in situ strength of concrete in
beams are presented in Table 2.
r ¼ 63; 750=p ð2Þ

where p is in pounds per square inch and r is in nanometer. 3. Appraisal of existing models
With this pressure, the smallest size of pore into which
mercury can be intruded is 2 nm. Thus, the pressure is Quite a few relationships involving strength and porosity
sufficient to ensure intrusion of mercury in all the capillary of cement-based materials have been reported in literature.
pores, as the reported radius of the smallest size capillary Notable among them are linear relationship of the form s=
pore is 5 nm. However, majority of the gel pores would s0  Kp, power exponent relationship of form s = s0(1  p)m,
remain nonintruded. The largest radius (pore size) that can exponential relationship of the form s = s0e  Kp and s =
be accounted for in the pore size distribution is 0.2 mm with Kln( p0s/p). In all these relationships, s stands for compress-
subambient pressure filling apparatus. The sample cell fitted ive strength at porosity p, s0 stands for compressive strength
with the base cell of capacity 17.7 cm3 was used throughout at zero porosity, p0s stands for porosity at zero strength, m and
the experiment. Six numbers of samples were tested for K are empirical constants. On plotting the compressive
given concrete to ensure adequate accuracy of the MIP strength of concrete given in Table 2 against their respective
results representing particular concrete. All tests were apparent porosity, it was observed that there are two distinct
performed at a constant moderate scanning rate indicated clusters of points. One cluster corresponds to concrete made
by Point 5 of the machine knob on its 0 –10 scale [10]. To with quartzite aggregates and the other corresponds to brick
obtain representative pore size distribution curve for the aggregates. The best-fit linear curve of the form s = s0  Kp
concrete in a particular beam, the results of six porosimetry between the strength and apparent porosity, for both the
data were averaged. For this purpose, the intruded volumes clusters of points taken together, yields s = 34.25  0.615p,
of mercury for all the six samples at a particular radius of the with a coefficient of correlation Cr =53%. Similarly, when the
pore were averaged to obtain average intruded volume of strength porosity data for the concrete made with quartzite
mercury at that pore radius. This procedure was repeated at aggregates were used alone, the resulting equation becomes
a large number of radii to generate the resulting average s = 53.45  2.301p, with a coefficient of correlation Cr =
pore size distribution curve. 52%. Poor correlation is exhibited in both cases; thus, linear
relationship of the form s = s0  Kp, suggested by Hassel-
2.7. Results man [24], seems to be oversimplified. Two values of s0
estimated as above differ quite significantly from each other
In situ strength of concrete in all 36 beams was estimated and apparently depend on aggregate type. Results of regres-
according to the procedure stated. From the intrusion curves sion for other forms of relationship are given in Table 3 and
 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164 159

Table 2
In situ strength, porosity, mean distribution radius, etc., of concrete
Beam Mix Age, Exposure In situ Apparent rm Porosity in pore size range (%)
number curing condition strength porosity (nm)
>106 nm 53 – 106 nm 10.6 – 53 nm < 10.6 nm
(day) (MPa) (%)
1 Mix1 28, 27 Atm. 18.3 12.96 34.3 3.931 1.084 4.907 3.038
2 Mix2 28, 27 Atm. 28.4 11.93 38.7 4.388 0.946 3.728 2.868
3 Mix2 28, 1 Atm. 26.8 10.87 58.7 5.166 0.898 2.639 2.163
4 Mix2 28, 27 300
C 22.7 11.10 41.3 4.388 0.816 3.840 2.065
5 Mix2 28, 27 600
C 21.5 13.53 42.3 5.006 1.001 4.376 3.147
6 Mix2 28, 27 Acidic 27.5 12.75 26.6 3.036 1.258 5.060 3.391
7 Mix2 84, 27 Atm. 29.7 10.80 39.3 3.965 0.816 3.515 2.504
8 Mix2 84, 1 Atm. 26.8 10.83 52.9 4.867 0.895 2.741 2.322
9 Mix3 28, 27 Atm. 30.3 11.80 45.8 4.085 1.344 4.569 1.802
10 Mix4 28, 27 Atm. 35.3 11.22 31.2 3.252 0.813 4.444 2.711
11 Mix5 28, 27 Atm. 40.3 11.50 30.4 3.175 0.907 4.536 2.882
12 Mix6 28, 27 Atm. 43.2 9.26 28.1 2.592 0.830 3.573 2.265
13 Mix6 28, 27 300
C 38.7 10.38 41.9 4.018 0.746 3.586 2.030
14 Mix6 28, 27 600
C 28.3 16.55 34.2 5.125 1.121 5.979 4.325
15 Mix6 28, 27 Acidic 42.5 9.50 23.0 1.894 0.839 4.032 2.735
16 Mix6 84, 1 Atm. 39.3 9.63 30.3 2.497 0.849 3.643 2.639
17 Mix7 28, 27 Atm. 14.2 33.70 146.9 22.06 3.014 4.764 3.860
18 Mix8 28, 27 Atm. 16.4 33.14 126.7 20.18 3.490 5.400 4.067
19 Mix1 28, 27 Atm. 15.5 11.22 41.6 4.163 0.717 3.830 2.510
20 Mix2 28, 27 Atm. 24.0 12.04 35.4 3.656 0.996 4.539 2.849
21 Mix2 28, 1 Atm. 23.2 11.39 71.3 5.719 0.870 3.083 1.718
22 Mix2 28, 27 300
C 14.9 12.23 31.2 3.938 0.902 3.743 3.642
23 Mix2 28, 27 600
C 13.6 15.37 49.6 6.287 1.369 4.443 3.271
24 Mix2 28, 27 Acidic 23.7 12.01 30.5 3.408 1.118 4.474 3.010
25 Mix2 84, 27 Atm. 25.7 10.38 47.5 4.140 0.739 3.371 2.130
26 Mix2 84, 1 Atm. 23.9 10.40 68.3 4.999 0.802 3.021 1.578
27 Mix3 28, 27 Atm. 30.7 11.30 43.0 3.849 0.997 4.404 2.050
28 Mix4 28, 27 Atm. 33.8 13.55 45.0 5.159 0.936 4.999 2.460
29 Mix5 28, 27 Atm. 37.7 11.85 29.3 3.091 0.840 5.152 2.767
30 Mix6 28, 27 Atm. 35.4 9.90 36.9 3.432 0.709 3.546 2.213
31 Mix6 28, 27 300
C 28.8 9.92 43.6 3.672 0.630 3.699 1.919
32 Mix6 28, 27 600
C 24.2 13.31 36.9 4.108 0.986 5.229 2.983
33 Mix6 28, 27 Acidic 36.2 9.28 35.0 2.622 0.761 4.088 1.805
34 Mix6 84, 1 Atm. 36.3 9.54 35.9 3.097 0.552 3.618 2.270
35 Mix7 28, 27 Atm. 17.7 33.60 122.1 20.05 3.897 5.327 4.330
36 Mix8 28, 27 Atm. 19.6 31.70 109.7 18.08 3.291 5.740 4.579

were obtained using the data of the concrete made with where s and s0 have the same meaning as defined earlier,
quartzite aggregate only. and a, b, c and d are the constant coefficients in the
Poor correlation is observed for all the above forms of equation. p>106 nm is the porosity with radius r>106 nm,
curve and it appears that simple strength – porosity relation- p106 – 53 nm is the porosity between pore radius 53 and 106
ships are not applicable in this case. In Fig. 2, these curves nm, p53 – 10.6 nm is the porosity between pore radius 10.6 and
are shown together with the data used for curve fitting. 53 nm and p < 10.6 nm is the porosity with radius r < 10.6 nm.
Regression for the above forms of curve using data for both Considering all the data presented in Table 2, inclusive of
aggregates taken together results in even poorer correlation. both types of aggregates, the multiple linear regression
Thus, simple relationships involving only porosity are yielded the values of the various coefficients as: s0 =
inadequate in explaining the observed variation of in situ 32.09, a = 1.135, b =  4.343, c =  1.993 and d = 4.942. Si-
strength of concrete with measured apparent porosity. milar to the observations made earlier for cement pastes by
A completely different form of strength – porosity rela- Atzeni et al. [23], two coefficients b and c are negative, in-
tionship was originally proposed by Older and Rößler [25]
and was further modified by Atzeni et al. [23]. A slightly
Table 3
modified form of this relationship is:
Results of regression for simple relationships
Form of equation Equation Cr (%)
s ¼ s0  ap>106 nm  bp10653 nm  cp5310:6 nm s = Kln( p0s/p) s = 37.1ln(0.284/p) 55
s = s0(1  p)m s = 68.74(1  p)8.15 54
 dp<10:6 nm ð4Þ s = s0e  Kp s = 74.4e  8.96p 51
160 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164

3.2. Strength – porosity relationship of Luping [26]

In this model, pores are divided into different size groups

and the fracture process is simulated through a computer
model. The basic theory underlying the model is available in
the literature [26]. For a given portion of material containing
the ith pore size group, the limiting condition of failure is
given by the following inequality:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Km Ami
scri  ð6Þ
ri

where Ami represents the fraction of solid (material) and can

be represented as Ami = 1  4riNi(1  riNi), Ni represents the
number of pores of ith size group per unit side length of the
overall material, ri is the mean pore radius of the group and
Fig. 2. Strength – porosity relationships. Km is an empirical constant that includes the modulus of
elasticity of material at zero porosity and the surface energy,
etc. The model assumes a specific regular geometrical
dicating that increased pore volume in certain size ranges of
arrangement of pores in all three spatial directions. The
pores increases the strength of concrete. Difficulty in
material is assumed to fracture progressively with applied
explaining the negativity of coefficients led to rejection of
stress, starting from the fraction containing the largest pores
this model by Atzeni et al. [23] in the past. Thus, this type of
to that containing the smallest pores. The stress it withstands
relationship is also unsuitable for correlating in situ strength
at the final stage of fracture is assumed as the strength of the
of concrete with its measured pore size characteristics.
material.
To evaluate the strength of concrete from its pore size
3.1. Relationship of Atzeni et al. [23]
distribution data obtained through MIP, the knowledge of
the constant Km for concrete in Eq. (6) is essential. In the
The relationship of Atzeni et al. relating strength of
absence of information on Km evaluation of this model for
cement pastes with its porosity, strength at zero porosity,
its applicability to the relationship among strength, porosity
etc., is given as:
and pore size distribution for concrete is rather difficult.
s0 ð1  pÞ However, values of Km can be evaluated if strength and
s¼K pffiffiffiffiffi : ð5Þ porosimetry data are available. Starting from the finest pore
rm
size group and the strength of the material, values of Km can
In this equation, s, s0, p and K have the same meaning
as defined earlier. To consider the effect of pore size distri-
bution on strength, the authors used a parameter rm in the
relationship and termed it as mean distribution radius, as
defined earlier in Eq. (3). Atzeni et al. [23] advocated the
estimation of s0 from linear strength –porosity relationship;
however, the analysis presented in Section 2 demonstrates
that this estimation is likely to be grossly erroneous; thus, to
pffiffiffiffiffi
start with, the in situ strength is plotted against ð1  pÞ= rm ,
as shown in Fig. 3. Using the data for quartzite aggregate
alone, the best-fit linear curve between in situ strength and
pffiffiffiffiffi pffiffiffiffiffi
ð1  pÞ= rm , is s ¼ 189:3ð1  pÞ= rm þ 2:45. In this
equation, p is in fraction, rm is in nanometer and s is
in megapascal. However, the intercept 2.45 is relatively
small and from physical consideration, when p ! 1 or
(1  p) ! 0, rm tends to be large (rm ! 1), resulting in
pffiffiffiffiffi
½ð1  pÞ= rm ! 0; therefore, for p ! 1, s must be neg-
ligibly small [s ! 0]. Thus, the curve between s and
pffiffiffiffiffi
ð1  pÞ= rm can be fitted through the origin. The coef-
ficient of correlation of this linear plot is 70%. The
coefficient of correlation, estimated for nonlinear relation-  
1p
ffiffiffi
ffi
ships between above two variables, was lower than above. Fig. 3. Relationship between in situ strength and p
rm .
 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164 161

be estimated from Eq. (6) through an inverse computation 4. Proposed model

process suggested by the authors.
Km includes the effect of modulus of elasticity and 4.1. Basic model
surface energy of concrete at zero porosity, and thus
strongly depends on the nature of pore free solid. The nature Compared to other models, the model of Atzeni et al.
of solid again is dependent on cement content of the [23] exhibited a somewhat better degree of correlation
concrete. Thus, a plot of Km values against cement content between the strength and independent variable defined. This
expressed as fraction is shown in Fig. 4 for concrete can be partially explained through Griffith’s theory. Accord-
exposed to ambient laboratory temperature and cured for ing to Griffith’s theory, the critical stress resulting in rapid
27 days in standard manner. The resultant fitted linear growth of crack and fracture causing failure under tension is
equation is Km = 54,402.1C  4255.5, with a coefficient of given by:
correlation equal to 90%. It can be observed that at the rffiffiffiffiffiffiffiffiffi
cement content (expressed as fraction of total weight) of 2ET
st ¼ ð7Þ
0.078, the value of Km is zero and, consequently, the pcl
strength of concrete is also zero. Thus, this is the minimum
possible cement content for concrete. Within the range of where E, T and cl stand for modulus of elasticity, fracture
cement content and w/c ratio used in this work, the solid surface energy of the material and half-crack length,
matrix strength of concrete is independent of porosity and, respectively. For a porous material like concrete containing
hence, is also independent of w/c ratio. The matrix strength pores of different sizes, E and T are effective modulus of
increases with increase in cement content. Thus, Km can be elasticity and effective fracture surface energy, respectively,
estimated through the equation obtained for the concrete for the uncracked overall material consisting of solid matrix
prepared with the given aggregate and cement used. An and the pore [16,26,27]. For porous materials, the effective
elaborate discussion on the subject, however, is available modulus of elasticity is a function of porosity, as pores do
elsewhere [11]. not contribute to elastic modulus. Numerous models relating
the porosity and elastic modulus are available [4,15 – 17,26].
3.3. Discussions on the applicability of existing models The simplest one is the linear form, E = E0(1  p). The
fracture surface energy is the energy required for creating a
Analysis presented in the previous subsections dem- unit area of interface between the solid and air [4,16,17,27].
onstrate that both simple strength – porosity relationship For porous material, the interface between the solid and air
and Older and Rößler [25] types of models are inadequate already exists at pores; however, for pore free solid portion,
in explaining the variation of in situ strength of concrete the same needs to be created. The effective fracture surface
with apparent porosity and pore size distributions. The energy required for fracture of gross unit area of porous
model of Atzeni et al. [23], on the other hand, is able to material, thus, will be lower with increase in porosity. This
explain the above behaviour better. Luping’s model is the has been recognized in the past by Brandt [16] and
most complicated among all the models discussed. In the Wittmann [27] and relationships between effective fracture
absence of the value of the empirical constant Km, it is surface energy and porosity have been proposed for
difficult to make even relative assessment of strength concrete and other cement-based materials. The simplest
through this model. form of this relationship is again linear, namely, T =
T0(1  p). The effective modulus of elasticity (E) and the
effective fracture surface energy (T) of concrete are thus
dependent on the porosity, since only solid contributes to
them. Assuming simple linear models and assuming the
effective half-crack length as the average radius of the pores
estimated through the mean distribution radius [23,26], the
resulting equation for the tensile strength at the onset of
fracture can be expressed as:
rffiffiffiffiffiffiffiffiffiffiffiffiffi
2E0 T0 ð1  pÞ
st ¼ pffiffiffiffiffi ð8Þ
p rm

where E0 and T0 represent the modulus of elasticity and

specific surface energy of pore free solid, respectively. This
model is similar to that suggested by Wittmann [27] except
that the effective modulus of elasticity is also assumed to be
proportional to (1  p). It may be mentioned here that other
Fig. 4. Km versus fraction of cement contents of concretes. complex forms of relationship involving E, T and p reported
162 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164

in literature, when tried, were unable to explain the 4.2. Effects of age, exposure condition and aggregate type
empirically observed behavior (Fig. 3) better than the model
of Atzeni et al. Exposure to temperature or acidic environment results in
The uniaxial compressive strength of the material is irreversible changes in cement hydrates, resulting in altera-
dependent on the tensile fracture strength of the material; tion of effective binder content of the concrete. Taking into
therefore, the same can be written as: account the effective binder content present after exposure
to various environments, different multiplying factors fe, fa
ð1  pÞ and fT can be introduced in the above equation. The factors
s ¼ K1 pffiffiffiffiffi : ð9Þ fe, fa and fT take into consideration the effects of exposure to
rm
acidic environment, age and temperature, respectively.
When soft porous aggregates are used in the concrete
Empirical constant K1 takes care of E0, T0 and other instead of hard rocks (quartzite), as in this work, the mode
unaccounted factors. E0, modulus of elasticity of solid pore of failure of concrete under compressive load differs con-
free matrix, depends on the binder content of the concrete. siderably. For concrete prepared with strong hard aggre-
T0, the energy required to create unit surface area of solid gates, the observed fracture surface passed through the
matrix of concrete, is dependent on the bond strength of the mortar – aggregate interface and crushing of aggregate was
concrete– particulate system. Estimated modulus of elasti- absent. While for concrete prepared with Mix7 and Mix8,
city of unhydrated cement particles is comparable to that of the crushing of aggregate was always observed during
natural rock. Hydration of cement, on the other hand, results compression tests. Besides, higher porosity of concrete
in an increase in Van der Waals forces, causing an increase was observed for soft aggregate prepared with identical
in the modulus of elasticity of solid matrix [17]. As a first mix proportions due to the contribution from porous ag-
approximation, therefore, both E0 and T0 can be assumed to gregate itself. Therefore, type of aggregate also influences
be directly proportional to the cement content of concrete the void free matrix property, and hence, a multiplying
for the range of mix proportions considered; hence, Eq. (9) factor fca can be also introduced for aggregate type in Eq.
can be rewritten as: (10). Thus, on further modification, Eq. (10) can be rewrit-
ten as:
ð1  pÞ
s ¼ K2 C pffiffiffiffiffi ð10Þ ð1  pÞ
rm s ¼ kfca fe fa fT C pffiffiffiffiffi : ð11Þ
rm
where C is the cement content of mix, expressed as fraction. In the above equation, k is the new empirical constant.
The fitted linear equation between in situ strength and To evaluate fT, the following procedure was adopted. The
pffiffiffiffiffi
Cð1  pÞ= rm passing through the origin for all the data, as concrete samples prepared from only Mix2 and Mix6 were
shown in Fig. 5, results in a value of 1294.3 for K2. The subjected to elevated temperatures (300 and 600
C). Thus,
coefficient of correlation for this equation is estimated to be the value of K2 defined in Eq. (10) for all the 27-day-cured
90%, which shows a considerable improvement over and 28-day-old Mix2 and Mix6 concrete samples exposed
original relationship of Atzeni et al. discussed earlier. to room temperature, 300 and 600
C were estimated and
averaged for two modes of compaction. The average value
so obtained was then divided by the K2 value corresponding
to 26
C to obtain a ratio corresponding to each temperature.
This ratio for both mixes at 26
C, therefore, is unity. The
average of the ratio for two mixes is the fT value corres-
ponding to a given temperature. The best-fit linear curve
between fT and temperature results in a simple equation:
fT = 1  3.5
10  4(T  26), where T is the exposure tem-
perature (
C) of concrete. The values of fe and fa were also
deduced in the same manner for 84 days of exposure to
acidic and atmospheric environments, respectively. The
values obtained are 0.91 and 1.1 for fe and fa, respectively.
The values of factors fe, fT and fa are unity for moist-cured
28-day-old concrete exposed to 26
C temperature only.
Values of fe for other environments can be evaluated by
adopting a similar methodology when data are available. As
per standard practice, IS: 456 [28], etc., the age factor for 3-
month-old concrete nearly coincides with the fa value
  estimated and thus validates the estimation of the same.
1p
ffiffiffi
ffi
Fig. 5. In situ strength versus C p
rm . Similar factor for inadequate curing can also be introduced
 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164 163

3. A new model has been proposed, which takes into account

the apparent porosity, pore size characteristics and cement
content of concrete, etc., in addition to aggregate type,
exposure condition and age of concrete.

Acknowledgements

The authors acknowledge CSIR, HRD Group of the

Government of India, for providing the financial support for
carrying out the research.

References

[1] T.S. Nagraj, B. Zahida, Generalization of Abrams’ laws, Cem. Concr.

Res. 26 (6) (1996) 933 – 942.
[2] G.M. Därr, U. Ludwing, Determination of permeable porosity, Mater.
Struct. 6 (2) (1973) 185 – 190.
[3] F.S. Rostasy, R. Weib, G. Wiedemann, Changes of pore structure of
  cement mortar due to temperature, Cem. Concr. Res. 10 (2) (1980)
Fig. 6. In situ strength versus fca fe fa fT C 1p
p ffiffiffi
ffi . 157 – 164.
rm
[4] I. Soroka, Portland Cement Paste and Concrete, Macmillan, London,
UK, 1979.
[5] D. Winslow, D. Liu, Pore structure of paste in concrete, Cem. Concr.
in Eq. (10); however, due to lack of sufficient data, this Res. 20 (2) (1990) 227 – 235.
exercise was not undertaken in the present work. It may be [6] H.W. Reinhardt, in: A. Aguado, R. Gettu, S.P. Shah (Eds.), Concrete
noted here that expressions for fT and fe are obtained through Technology: New Trends Industrial Applications, E & FN Spon, UK,
relatively small set of data; thus, more reliable expressions 1995, pp. 19 – 32.
for fT and fe can be arrived at using the same methodology [7] P.K. Mehta, Concrete: Structure, Properties and Materials, Prentice-
Hall, New York, 1986.
when a larger data set becomes available. [8] S. Popovics, New formulas for the prediction of the effects of porosity
To evaluate the value of factor fca for soft, crushed brick on concrete strength, ACI Mater. J. 78 (2) (1981) 127 – 129.
aggregates, the ratios of actual in situ strength to strength of [9] S. Popovics, Generalization of Abrams’ law, ACI Mater. J. 82 (2)
concrete estimated through Eq. (10), in beams prepared with (1985) 136 – 146.
Mix7 and Mix8, are calculated. These ratios are 1.72 and [10] A.I. Laskar, R. Kumar, B. Bhattacharjee, Some aspects of evaluation
of concrete through mercury intrusion porosimetry, Cem. Concr. Res.
1.96 for mechanically and manually compacted concrete, 27 (1) (1997) 93 – 105.
respectively, for Mix7. Similarly, the ratios for Mix8 con- [11] R. Kumar, Strength and permeation quality of concrete through mer-
crete are 1.58 and 1.71, respectively. The average of the cury intrusion porosimetry, PhD Thesis, Department of Civil Engi-
above four values (1.74) represents fca for crushed brick neering, Indian Institute of Technology Delhi, New Delhi, India, 1997.
aggregate. The value of fca for quartzite aggregate is unity. [12] R. Kumar, B. Bhattacharjee, Study on some factors affecting the
results in the use of MIP method in concrete research, Cem. Concr.
A final plot of in situ strength against the factor,
pffiffiffiffiffi Res., USA (submitted for publication).
{fca fe fa fT Cð1  pÞ= rm }, is given in Fig. 6. The resulting [13] D.N. Winslow, S. Diamond, A mercury porosimetry study of the
best-fit line yields a value of 1.39
103 for k and the evaluation of porosity in Portland cement, ACI J. Mater. 5 (3) (1970)
coefficient of correlation is 85%. It is interesting to note that 564 – 585.
the coefficient of correlation for the above best-fit line, [14] S. Diamond, A critical comparison of mercury intrusion porosimetry
and capillary condensation pore size distribution of Portland cement
considering the data for concrete made with quartzite aggreg- pastes, Cem. Concr. Res. 1 (5) (1971) 531 – 545.
ate alone, i.e., with fca = 1, is 92%. [15] A.M. Neville, J.J. Brooks, Concrete Technology, Longman, Singa-
pore, 1990.
[16] A.M. Brandt, Cement Based Composites: Materials, Mechanical
5. Conclusions Properties and Performance, E & FN Spon, UK, 1995.
[17] K. Newman, Concrete systems, in: L. Holliday (Ed.), Composite Ma-
terials, Elsevier, USA, 1966, pp. 336 – 452.
1. A large set of data on apparent porosity, pore size [18] S. Diamond, Mercury porosimetry; an inappropriate method for the
distribution and in situ strength of concrete has been measurement of pore size distributions in cement-based materials,
presented for concrete of varying mix proportions. Cem. Concr. Res. 30 (10) (2000) 1517 – 1525.
2. It has been demonstrated through an appraisal that most [19] IS: 516, Methods of Tests for Strength of Concrete, BIS (Bureau of
Indian Standards), India, 1956.
of the existing models relating the strength with pore size [20] BS 1881: Part 120, Method for Determination of the Compressive
characteristics of cement-based material are inadequate Strength of Concrete Cores, British Standard Institution (BSI), UK,
in the context of concrete. 1983.
164 R. Kumar, B. Bhattacharjee / Cement and Concrete Research 33 (2003) 155–164

[21] E.J. Sellevold, Mercury porosimetry of hardened cement paste cured [25] I. Older, M. Rößler, Investigations on the relationship between poros-
or stored at 97
C, Cem. Concr. Res. 4 (3) (1974) 399 – 404. ity, structure and strength of hydrated Portland cement pastes: ii.
[22] D.H. Bager, E.J. Sellevold, Mercury porosimetry of hardened cement Effects of pore structure and degree of hydration, Cem. Concr. Res.
pastes: the influence of particle size, Cem. Concr. Res. 5 (2) (1975) 15 (4) (1985) 401 – 410.
171 – 175. [26] T. Luping, A study of the quantitative relationship between strength
[23] C. Atzeni, L. Massidda, U. Sanna, Effect of pore size distribution on and pore size distribution of the porous materials, Cem. Concr. Res.
strength of hardened cement pastes, Proceedings of the First Interna- 16 (1) (1986) 87 – 96.
tional RILEM Congress on Pore Structure and Material Properties, [27] F.H. Wittmann, Fracture Mechanics of Concrete, Elsevier, Amster-
Paris, 1987, pp. 195 – 202. dam, 1983, pp. 56 – 62.
[24] D.P.H. Hasselman, Relation between effects of porosity on strength [28] IS: 456, Indian Standard Code of Practice for Plain and Reinforced
and Young’s modulus of elasticity of polycrystalline materials, J. Am. Concrete, 3rd ed., BIS (Bureau of Indian Standards), India, 1978.
Ceram. Soc. 46 (11) (1963) 564 – 565.