CEMENT and CONCRETERESEARCH. Vol. 2, pp. 663-680, 1972.

Pergamon Press, Inc

Printed in the United States.

THE RELATIONSHIP BETWEENSTRENGTHAND THE

COMPOSITION AND FINENESS OF CEMENT
K.M. Alexander
CSlRO Division of Building Research,
NSW Cement and Concrete Laboratory,
Australia 2006

(Communicated by R. W. Nurse)

ABSTRACT
Data from new experiments and from work published during the past
40 years are subjected to regression analysis to determine the
relationship between strength and the composition and fineness of
cement. The original authors differed in their opinions on the
relative importance of CRS* and CRA. The controversy is examined
against a background of ~xperienc~ with a group of materials that
are representative of Portland cement in general, and in terms of
an analysis based on a model in which, during the f i r s t weeks of
hardening, the strength developed by C3S depends on the
proportion of C3A in the cement.

Die Daten neuer Versuche und in den letzten 40 Jahren

ver~ffentlichter Arbeiten werden der Regressionsanalyse
unterworfen, um die Beziehung zwischem der Festigkeit und
der Zusammensetzung und Feinheit des Zements zu bestimmen.
Die ursprUnglichen Verfasser sind verschiedener Meinung Uber
die relative Bedeutung von CRS und CRA. Die Streitfrage
wird aufgrund der Erfahrung ~ i t eine~ Gruppe von Materialien,
typisch fur Portlandzement, allgemein untersucht und analysiert
an einem Modell, in dem die, in den ersten Wochen des Abbindens
entwickelte Festigkeit von C3S, von dem C3A-Anteil im Zement
abhangt.

In abbreviated formulas: C=CaO, A=A1203, F=Fe203, S=SiO2.

663
664 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

Introduction
Over the years, many laboratories throughout the world have studied
the relationship between strength and the composition and fineness of
Portland cement. Information is available from experiments in which each of
the cement compounds is examined individually, as in the classical work of
Bogue and Lerch ( I ) , which is widely quoted in the standard texts of
reference. However, the method which is most commonly used is to determine
the strength, composition and fineness for a range of cements and to subject
the data to regression analysis. Extensive investigations along these lines
have been carried out in, for example, the United States, where the work of
Gonnerman (2), Gonnerman and Lerch (3), and Blaine, Arni and DeFore (4)
covers the wide variety of cements that have been manufactured, at one time
or another, during the present century. Interest continues to be shown in
this approach, and several publications along these lines have appeared in
recent years (4, 5, 6).

Despite the attention given to the subject, some of the most vital
issues remain unresolved. For example, from the viewpoint of concrete
strength, the importance of C3A relative to that of C3S is s t i l l a matter
for controversy (5). I t is with this issue in particular that the present
investigation is concerned.

From an examination of published evidence i t appears that a great deal

of the uncertainty in this area can be traced to two causes, namely, working
with a range of materials that is not representative of Portland cement in
general, and basing the regression analysis on an unsuitable model. I t is
the purpose here to present evidence from a study in which particular
attention has been given to minimising errors from these sources.

In this investigation, errors attributable to biased sampling are

reduced by working with the combined data from new experiments and from
studies published during the last 40 years. The data are for 228 cements,
and the results of 16,400 strength tests on these materials are used in the
regression analysis. Regarding the second cause of uncertainty, the
approach adopted is to explore the limitations of the model that is
traditionally employed in this field of research, and then to repeat the
analysis using what is believed to be a more realistic model of the way in
which the various cement compounds contribute to the strength of concrete.
Vol. 2, No. 6 665
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

The Cements

The data for the present analysis are taken from a number of
independent studies, each based on a d i f f e r e n t group of cements. The main
characteristics of these cements are as follows:

GROUP A: The 71 cements used by Gonnerman (2) which were produced by the
Portland Cement Association Fellowship at the U.S. Bureau of Standards during
1930-33. The range of compositions was purposely expanded beyond that of
normal Portland cement, but the range of fineness is narrow, since each
cement was ground to the same residue on 200 mesh sieve. Thus the
c o e f f i c i e n t of variation for surface area is only one third of the average
for series B to E. The SO3 content is constant at 1.8 percent.

GROUP B: This comprises 50 cements of known composition which were part of

the "series 73" studied by Gonnerman and Lerch (3). These cements were
manufactured in the United States during 1916-1950. In this group, SO3
is constant at 1.75 percent, C3A is almost constant at 10.2 percent with a
standard deviation of only 1.1 percent, and there is a strong association
between C3S and surface area (correlation c o e f f i c i e n t (R) of 0.75).

GROUP C: The twenty "series 308" cements used by Gonnerman and Lerch (3),
manufactured in 1940, and representing the five types of cements specified
in ASTM C 150-40T.

GROUP D: The 23 non-air-entraining cements used in the long term study

conducted by the Portland Cement Association (7).

GROUP E: Consists of 64 commercial Portland cements produced in Australian

factories during 1965-70. Alexander, Taplin and Wardlaw (5) have studied 30
of these cements, and the remainder are new materials used in the present
investigation.

Although the results for each of the groups A-E were analysed
separately, most of the analytical work reported here was done on composite
series in which the data were combined, and in which the C3A content of the
cements was limited to a maximum of 14.0 percent. The need for this l i m i t
is i l l u s t r a t e d by the data for 28-day and l-year strengths of group A
concretes (Fig. i ) . Over the range 0 to 14 percent C3A, strength increases
l i n e a r l y with C3A*, and the two variables are strongly associated. At 28
days, for example, the correlation c o e f f i c i e n t is .80. At higher C3A,
strength decreases with increasing C3A, the rate of decrease being greater
at i year.

*SO3 constant in this group. See l a t e r comment on e f f e c t of gypsum content.

666 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

40 44
A B

E
"E
Z 30
c-

t-
Q T ¢-

36
O

20 i

O
"10

OO
i 14% i
C~ 14%
10
I i i I 28

4 12 20 4 12 20
C3A, percent C3A, percent

FIG. 1

Compressive strength vs C~A for concretes made from series A

cements. A: Ate28 days. B: At 1 year.

I t is possible that the change in behaviour at around 14 percent C3A

is caused by changes in the texture of the hardened paste• In this
connection i t is interesting to recall Bogue and Lerch's (1) observation
that, in contrast to the dense structure of C3S paste, pastes formed from
C3S mixed with 15 percent of C3A develop a rather open porous structure.

Series 1 and 2 Cements

Potential compound compositions for series 1 and 2 cements were

calculated by ASTMStandard C150-64, and were corrected for free lime.
Surface areas used in the analysis were either air permeability values, or
Wagner surface areas multiplied by 1.7.

Series 1 consists of the 143 cements in groups A, D and E which have C3A
not greater than 14.0 percent. T h i s is the main series used in the analysis,
since complete sets of data are available at 3, 7 and 28 days and at 1 year.
A histogram of the distribution of C3S contents for series 1 cements is
given in Fig. 2A, from which i t will be apparent that over the range 30 to
70 percent C3S, there are from 20 to 44 cements in each 10 percent C3S class
interval. The corresponding histogram for C3A content (Fig. 2B) shows that
between 0 and 12 percent C3A, there are 16 to 34 cements in each 2 percent
C3A class interval• The wide range of C3S and C3A contents and the low level
of association (R=.06) between C3S and C3A are illustrated by the scatter
diagram (Fig. 3A) for C3S vs C3A.
The upper and lower limits for the surface area of series 1 cements are
Vol. 2, No. 6 667
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

50 A B C
/i
40
//

7"~
E 30 /.4
(J

"5 20

.. F.zC
I///
;// . . . . . . . .

0 20 40 60 0 4 8 12 200 280 360 440 520 600

C3S, percent C3A, percent Surfoce oreo rn~/kg

FIG. 2

Histograms for the distribution of (A)C3S, (B)C3A

and (C) surface area, of series I cements.

580 and 239 m2/kg respectively, and 92 percent of the cements have surface
areas between 240 and 400 m2/kg. Over this l a t t e r range there are from 18 to
45 cements in each 40 m2/kg class interval (Fig. 2C).

Series 2 contains 184 cements and consists of all series 1 cements

supplemented by group C cements and any group B cements for which surface
areas are available. As with series I , the upper l i m i t for C3A is 14 percent.
The range of C3S and C3A contents is shown in the form of a scatter diagram, in
Fig. 3B. A complete set of series 2 data exists only for the age of 28 days.

80 /
!
• A 80
I B

e•ee • •
i ee ~ • • •

c 60 l "" • • 60 "" • "

(J

o.
!
|

j
• S •
•
• • •
so• •
0 o~
,. . . : . ~ , ' ,
• •
8e •
• e=
e
~

•
0
aJ i
. ." -,'" -
•
.
°
•
::'...,
" •"
,
•
.';';
..s~o- .
.%

|•
[" ..... - ...
o 40 ";. •
"-" ." . "; "s'.
• o 40 L"
~ " -"'-~
' ". .". . . •- ;"" : ~ ' . •

l
: • • •
•e o • • • % % •

20 ~ .', 20 ~ •
"I " I I

0 8 16 0 8 16
C3A, percent CsA, percent
FIG. 3

S c a t t e r diagrams f o r C3S and CRA c o n t e n t s .

A. Series 1 cements. B. Series 2 cements.
668 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

Conditions of Test

Each group of cements was tested in either paste, mortar or concrete, or

in both mortar and concrete. The methods of test for the individual groups
and for the combined series are defined in Table 1. Dependingon the
laboratory, specimens were either moist cured or stored continuously immersed.
Water/cement ratio, cement content, aggregate type, size, shape and grading,
specimen size and shape, and many other factors, varied from laboratory to
laboratory.

Discussion

Normalized Strengths for Series 1 and 2 Cements

The strengths used in the analysis of series 1 and 2 data are normalized
values, which are defined as the compressive strength that would have been
observed i f the cement had been tested by the procedure that Gonnerman (2)
used for the materials in group A. Gonnerman'sconditions were selected as
standard because he examined the widest range of cements and measured concrete
strengths at the 'key' age of 28 days for every cement in the group.
Strengths were normalized by the following procedure:

Cements from groups B to E in turn were paired with similar cements in

group A. The criterion for deciding when cements were similar was
established by regression analysis on data for the individual groups A to E.
From this analysis i t seemed reasonable to accept that cements were similar i f
C3S, C3A and surface area matched to within 5 percent, l percent and 50 m2/kg
respectively. At each age, and for each of the groups B to E, the slope (m)
of the line of best f i t was calculated for cement strength vs strength of the
similar cement from Group A. Normalizedstrength for the cements in the
group were then estimated by multiplying the observed strength by the
appropriate value of m.

Conflictin 9 Evidence on the Relative Importance of C3S and C3A

I t is widely held that during the early weeks of hardening, the strength
of Portland cement is largely determined by C3S content. Evidence in
support of this view comes from studies on the individual cement minerals
and from experiments on Portland cement i t s e l f . Data for the individual
cement minerals are given in Fig. 4A, in which Bogue and Lerch's (1)
measurements on the strengths of pastes made either from C3S or from C3A are
plotted against time. The evidence shows that for ages of up to six months,
C3S paste develops more than ten times the strength of C3A paste.
Vol. 2, No. 6 669
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

TABLE 1
Test Conditions for Group A to E Cements
in Series i and 2

GROUP 3-DAY 7-DAY 28-DAY 1-YEAR

A 1 : 2.75 mortar 6 bag concrete

W/C 0.53 W/C 0.56
2-inch cubes 3x6-inch cylinders

5 bag
concrete
W/C 0.53
6x12-inch
cylinders

5 bag
concrete
W/C 0.49
6x12-inch
cylinders

D I : 2.75 4.5 bag + 1 : 2.75 6 bag f

mortar* concrete" mortar* concrete
2-inch 6-inch 2-inch 6-inch
cubes modified cubes modified
cubes cubes

E Paste, W/C 0.50, O.5-inch cubes

* Made and tested in accordance with ASTMC109-49

f Constant slump, 2 to 3-inch.

Experiments on Portland cement i t s e l f , which confirm the importance of C3S,

are reported by Gonnerman and Lerch (3). These latter experiments formed
part of an investigation into the changes in composition, fineness and
strength of cement manufactured during the f i r s t half of the present century.
Gonnerman and Lerch decided that the large differences in strength between
the cements of 1950 and those produced 35 years earlier were mainly caused by
increases in fineness and in potential C3S.

However, in a study of the 7-day to 6-month strengths of normal Portland

cements produced in the 15 Australian plants in 1967-68 (5), i t was found
that the regression coefficient for C3A was two to five times that for C3S
670 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

A B
E 3
t
f"o, x
~ 6O Z I ~'s.
°

C I
ID I
~ 2
.c" 40 I
O
I
C O

e-
O
20
P
c,A
.¢'~'7 -- -- , , IZ
i i

? 28 90 180 7 28 56 90 180
Age, days Age, days

FIG. 4

A. The compressive strength of C3S and C3A; Data

from Bogue and Lerch (1). B. R~gressiofi coefficients
for C3S and C3A in Portland cement paste tested in
compression: Data from Alexander et al. (5).

(Fig. 4B). The highest ratio of the coefficients was at 28 days, and at
this crucial age, the strength differences for the 15 cements in this sample
were mainly attributable to differences in C3A. Equation (1), for cements
of most recent origin (group E), shows that the observation on the high
ratio of the coefficients is s t i l l valid.

YP28 = -2.74 + 0.441C3S + 2.620C3A + 0.0705SW (I)

where YP28 is the estimated compressive strength of paste at 28 days, and SW
is surface area. YP28and SW are in N/mm2 and m2/kg respectively, and C3S
and C3A are in percent.

Thus, in emphasising the role played by C3A, the conclusions from the
regression work (5) are in conflict with the opinion of Gonnerman and Lerch
(3). However, the latter authors used group B cements, for which the C3A
content was almost constant, and they naturally found that variations in
performance were due to C3S and surface area. The present emphasis on C3A
(Fig. 4B) also runs contrary to the evidence (Fig. 4A) from experiments with
pure minerals. In this instance the two opinions can be reconciled by the
suggestion (5) that C3S reacts more rapidly in cements with high C3A content.
In fact, this hypothesis suggests that the linear regression equation is not
the appropriate model. Before this topic is pursued i t will be advantageous
to compare the data for pastes and concretes, and to consider further
evidence on the C3A effect.
Vol. 2, No. 6 671
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

A Comparison Between Paste and Concrete

The question of whether the relationship between paste strength and

composition and fineness is relevant to the behaviour of cement in concrete
has been debated for many years, and is sufficiently important to warrant
further attention here. Considertherefore the evidence in Fig. 5. In
this diagram, calculated 28-day concrete strengths for 97 cements are
plotted against the corresponding values that were determined by experiment.
The 97 cements are the materials whose 28-day concrete strengths form part of
the series 2 d a t a . Thesecements are identified in Table 1. The
calculated strengths were obtained by means of equation (1), which was
derived from results that came entirely from tests on pastes. All strengths,
whether calculated or measured directly, are normalized.

The band within which 95 percent of the sample observations are expected
to fall has been calculated from the standard error of estimate for equation
(1), and is defined by the broken line in Fig. 5. T h i s expectation is
realized since 98 percent of the observed strengths fall within the band.

This test of the relevance of data from paste tests is particularly

severe because the paste data upon which equation (I) is based are from tests
on 64 modern cements, whereas the 97 cements whose concrete strengths are
estimated are a different set of materials, comprising several groups of
cements that were manufactured 20 to 40 years ago. Furthermore, the various
groups of concretes were made years apart, the tests were conducted in

40 //
,/
•
/
i/ • •
E
•// 8 • ~ •
/// eI• I | ~i a• //
// • r / /
=" 30
¢- // • ~ • ///

P /
//
| • • /
,./
/ • • • //
// • • • e//
// • /
~ 20 /
// ee
• el •
• Y
/

• //
• O/,
• o • ~11 I
b_l •
i I
i

t / i i i

10 20 30 40
Observed strength N/ram 2
FIG. 5
Observed vs estimated normalized strength for concrete tested in
compression at 28-days. Strengthsestimated from the regression
of paste strength on composition and fineness.
672 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

d i f f e r e n t laboratories, and there were considerable differences between the

experimental procedures in these laboratories.

I t is concluded therefore that the tests on cement paste provide a

reliable indication of the efficiency with which the various cements w i l l
perform in strength tests on the corresponding concretes. This conclusion
is supported by evidence that has recently been published by Dalziel (8).

Additional Evidence on the S e n s i t i v i t y of Strength to C3A

In this section, the examination of additional evidence on the C3A

e f f e c t i s , for two reasons, mainly concerned with the situation at 28 days.
F i r s t l y , there is the widespread commercial interest in 28-day strength, and
secondly, this is the age at which the strongest association is observed
between strength and C3A (5).

The dependence of strength on composition and fineness w i l l be discussed

in terms of a regression in the form of equation (2).

Y = a + bC3S + cC3A + dSW (2)

where Y is the estimated strength, and a, b, c and d are constants. This
form is preferred because i t has been shown (5) that when the regression of
strength on composition and fineness is based on the t r a d i t i o n a l model, the
s i l i c a t e s and aluminates should each be represented by only a single variable.
Further details are given l a t e r in this discussion.

The evidence (5) on the s e n s i t i v i t y of strength to C3A came from

experiments on only 15 cements, which were all normal cements manufactured at
about the same time. The question that arises is whether t h e i r behaviour is
typical of Portland cement in general. Consider therefore the data for the
concretes, mortars and pastes made from the 143 cements in series I. In
Fig. 6 the coefficients for the correlation of strength with C3S, C3A and
surface area in this series are plotted against time. As with the 15 cements
used in the p i l o t project, the strongest association of strength with C3A
occurs at 28 days. The correlation of strength with C3S passes through a
maximum at 7 days, while the correlation of strength with surface area
decreases with time throughout the period of 3 days to i year.

Further information on the C3A effect for series I cements was

published by Alexander (9) and is reproduced here, in s l i g h t l y modified form,
in Fig. 7. This evidence shows that of the 39 cements with C3A less than
4.0 percent, 36 develop 28-day strengths that are below the mean value (AB)
Vol. 2, No. 6 673
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

0.8

.-= 0.6
O

FIG. 6 03
O
U

Plot of correlation coefficients for CRS, c 0"4

o
C3A and Sw against age. Data are for -
q
normalized strengths of series l cements•
uP> 0.2

i i i

3 7 28 365
Age in doys, Log scole

for the series• Conversely,at the high C3A end of the scale, 33 of the 35
cements with C3A greater than 10.0 percent develop above-average strength•
This behaviour is not attributable to strong associations between C3A and the
other main variables that control strength since C3A is correlated only .07
with C3S (Fig. 3A) and only .01 with surface area.

At each age, series I cements can be divided into two subgroups

depending on whether the strength is above or below average for the series•
The ratio between the average C3A for stronger and weaker cements is given in
Fig. 8, together with the corresponding ratios for C3S and surface area. The
data show that as the age of 28 days is approached, stronger cements are
increasingly distinguished by their high C3A contents• At 28 days the C3A
50

4o
@
o° •
i
0

• %• •

.~ 30
I
'.:.:.."
•• | • • "•

I • i , ,
c A • • ! ,$ •

$ ;•, I@
:':. • , • ,.
g

20 I • •h I ."• :

10 ~ ------,----
0 4 8 12 16
C3A, percent

FIG• 7•

Relationship between 28-day normalized strength and C3A content for

pastes, mortars and concretes made from series 1 cements• (R = .65)
674 VoI. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

1-8

1"6 FIG. 8
Comparison between stronger and weaker cements
U3 on the basis of composition and fineness.
1"4 S = average C3S, C3A or SW for series 1 cem-
ents that are of above-average strength at the
indicated age. W = corresponding value for
the remaining cements.
1-0 i | i

3 ? 28
Age, days

r a t i o between the two subgroups has risen to 2.1, with the average C3A for
the stronger cements running at 8.9 percent compared with 4.2 percent for the
weaker materials. The corresponding ratios for C3S and surface area decrease
with time, and by 28 days have fallen to 1.19 and 1.07 respectively.

An Alternative to the Traditional Model for the

Regression of Strength on Composition and Fineness

The t r a d i t i o n a l method (2,6) of analysing strength in terms of potential

mineral composition and fineness is to f i t an equation of the form

Y = a + bC3S + cC2S + dC3A + eC4AF + fS W (3)

This form of equation can be successfully used for strength prediction,

but i t does not provide a s u f f i c i e n t l y r e a l i s t i c basis for evaluating the
contributions made by the individual minerals. I t has been pointed out (5)
that there are good reasons for discarding the model on which equation (3) is
based, both from the viewpoint of cement chemistry, and for the reason that
there are highly s i g n i f i c a n t associations between the independent variables.
I t was shown (5) that at ages of up to two months, a better description of
strength is obtained by a regression in which the s i l i c a t e s are represented
by C3S alone, and the aluminates by C3A. Equation (1) for group E cements
provides a convenient example from the present investigation. In this
equation the highest correlation between independent variables is 0.16. The
standard errors for the regression coefficients for C3S, C3A and surface area
are O.llON/mm2, O.315N/mm2 and O.0351m2/kg respectively, and all coefficients
are s i g n i f i c a n t at the 5 percent level. The multiple correlation coefficient
for strength is .80. The corresponding t r a d i t i o n a l regression equation
for these data is:
Vol. 2, No. 6 675
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

Y'P28 = -5.14 + 0.386C3S - 0.087C2S + 2.980C3A + 0.641C4AF + 0.0617S W (4)

There are strong associations between the independent variables in
equation (4): C3S is correlated .97 with C2S, and C3A is correlated .72 with
C4AF. The standard errors of the regression coefficients for C3S, C2S, C3A,
C4AF and surface area are 0.608, 0.614, 0.734 and O.782N/mm2, and O.0372m2/kg
respectively. Only the coefficient for C3A is s i g n i f i c a n t at the 5 percent
level. Despite the introduction of terms for a further two of the major
cement minerals, the multiple correlation c o e f f i c i e n t for strength remains at
the value (.80) observed for equation ( i ) .

Modification of the regression relationship to the form shown in

equation (I) resolves some of the d i f f i c u l t i e s of i n t e r p r e t a t i o n , but there
is a serious lack of agreement (cf. Fig. 4) between the results obtained by
using the modified equation, and those obtained when the strengths of the
individual minerals are measured d i r e c t l y . In the present investigation
this aspect was therefore examined in greater d e t a i l , and i t was found that
the regression coef f i ci ent for C3S increased with increasing C3A. This
suggested that i t would be more appropriate to use a model in which C3A
affects the strength of Portland cement by increasing the rate of strength
gain of C3S. Various forms of regression that complied with this requirement
were then tested. The best f i t for the data at 3, 7 and 28 days was
obtained with a relationship of the form shown in equation (5).

Y = a + b'C3S + cSW (5)

where
b' = b(1 + dC3A) (5a)
and where a, b, c and d are constants.

The following interpretation of the behaviour of cement at early ages

is in keeping with the implications of equation (5). At a given fineness
and at zero C3A, Portland cement develops a basic or background strength
which, for the present purpose, is attributed e n t i r e l y to C3S. I f C3A is
present in the c l i n k e r , the strength of cement increases beyond the level
that is basic for that age. I t was deduced e a r l i e r (5) that the increase
occurs because C3S hydrates more rapidly in cements which have high C3A
contents*. In other words~ not only the basic strength~ but the strength

*Cf. the unpublished work of Kantro and Weise, quoted by Verbeck (10),
which shows that the rate of reaction and of strength development of C3S is
increased by introducing AI203 and MgO into C3S during clinkering.
676 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

in excess of the basic level also can be attributed to C3S. The strength
increase is d i r e c t l y proportional to C3A and depends, in addition, on the
magnitude of d (equation 5). I t w i l l be shown la te r that d varies with age.
The regression of strength on composition and fineness for series i
cements at 3, 7 and 28 days is given, in the t r a d i t i o n a l form, in equations
6, 8 and I0, and in the alternative form, in equations 7, 9 and 11. There
is a I : I correspondence between the estimates from the two forms of equation,
and the two sets of estimates correlate .99, .99 and .98 at 3, 7 and 28 days
respectively.

Y N3 = 10.19 + 0.170C3S + 0.279C3A + 0.0485S w (R=.82) (6)

Y'N3 = 8.61 + 0.154C3S(I + 0.030C3A ) + 0.0468S w (R:.82) (7)

Y N7 = -8.59 + 0.284C3S + 0.544C3A + 0.0355S w (R=.84) (8)

Y'N7 = -5.33 + 0.236C3S(I + 0.043C3A ) + 0.0329S w (R=.84) (9)

Y N28= 1.97 + 0.188C3S + 1.014C3A + 0.0385S w (R=.83) (I0)

Y'N28 = 7.79 + 0.I06C3S(I + 0181C3A) + 0.0325S w (R=.81) (11)

where the subscripts N3, N7, N28 indicate that the estimated strengths are
normalized values for the ages of 3, 7 and 28 days respectively.

In equation 5 (a), d is the proportion by which each 1 percent of C3A

raises the cementitious a c t i v i t y of C3S above the basic level for C3S at zero
C3A. I t w i l l be convenient to refer to d as the s e n s i t i v i t y of C3S to C3A.
Series i cements show a sixfold increase in d between 3 and 28 days, and i t
w i l l be seen from Fig. 9 (open circles) that over this interval of time, the
relationship between d and age is linear.
The picture that emerges from this evidence is as follows. During the
first one or two days of hardening, when cementitious a c t i v i t y is mainly
confined to hydration of the f i n e r particles and the outer layers of the
coarser p a r t i c l e s , the contribution that C3S makes to strength is not greatly
dependent upon the C3A content of the cement. However, dependency increases
rapidly with time, and a high level is soon attained. Maximumdependency
is observed at about 28 days, and persists for some weeks thereafter. C3A
probably operates by accelerating hydration rather than by changing the
cementitious properties of the hydration products. At l a t e r ages, hydration
approaches completion even in the absence of C3A, and thus, with the passage
of time, the a b i l i t y of C3A to increase strength diminishes.
Vol. 2, No. 6 677
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

e Series 1 cements
20 o Group E cements, high SOz
• Group E cements. Low SOz
e

FIG. 9
L~ 10
The s e n s i t i v i t y of C3S to C3A, vs age.
=-8
=<
bC
oi "

3 ? 28
Age, clays

Gypsum, and the S e n s i t i v i t y of C3S to C3A

In view of the influence of gypsum on the behaviour of C3A the question

arises as to whether or not the revised model should be refined to take
this factor into account. Therefore, the s e n s i t i v i t y of C3S to C3A was
calculated for groups of cements that were selected on the basis of t h e i r
gypsum contents. Many procedures were used, starting with the simplest, in
which cements were selected from series I and 2 materials i f t h e i r gypsum
contents or t h e i r C3A/gypsum ratios f e l l between certain a r b i t r a r y l i m i t s .
Other methods that were t r i e d included, for example, the use of an equation
which related the l i m i t i n g gypsum contents to fineness and chemical
composition. This equation was based on work in which Lerch ( I I )
determined, by d i r e c t experiment, the optimum gypsum levels for a range of
Portland cements.
This phase of the analysis led to the conclusion that with the f u l l
range of cements in this investigation, there was no consistent evidence that
the s e n s i t i v i t y of C3S to C3A varied with gypsum content. However, i t is
doubtful i f any of the selection procedures would be valid for the wide range
of p a r t i c l e size d i s t r i b u t i o n s and the great differences in chemical
composition that exist in series I and 2. Commenton the quantitative
effects of gypsum content w i l l therefore be confined to observations on a
more homogeneous sample, namely, group E cements. Group E is the largest
individual group in the investigation and, since the cements are a l l of
recent manufacture, t h e i r performance is relevant to contemporary concrete
practice.

Group E cements were c l a s s i f i e d , in terms of gypsum content, by the

678 Vol. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

O
o
o

o
oa ooo o
o

OoO o° o ~ --
Q.I
0
IlJ FIG. 10
~2 o ao o °° o° oo °

o
oo
o o o
c~
The division of group E cements into
o3
O O
two subgroups on the basis of SO3 content

5 10 15
C3A,percent

procedures developed for series i and 2 cements. Analysis showed that i t

was satisfactory to divide the 64 cements into two equal subgroups as shown
in Fig. 10. The distribution of C3A in each subgroup wassimilar, but
gypsum contents in the one subgroup were about one third higher than in the
other. The sensitivity of C3S to C3A for high gypsum and low gypsum
cements at 3, 7 and 28 days is given in Fig. 9. I t will be seen that within
the accuracy of the experiment, d was not affected by the change of gypsum
content, and that d for high gypsum cements and for low gypsum cements did
not differ from the value for series 1 cements in general. There are
indications that the situation may be more complex than this (12,13) but
the present data do not enable this possibility to be checked by analysis.

Some Limitations of the Present Model

Although use of the present model leads to improved agreement between

the conclusions from regression analysis, and those from tests on the
individual minerals, there are a number of limitations that should be borne
in mind. For example, when C3A exceeds 14 percent there is a complete
departure from linearity in the strength-C3A relationship (Fig. 1). Also,
i f the analysis is to be extended to later ages, allowance must be made for
the contribution that C2S makes to strength, and for any interaction between
C2S and C3A. Thereis also the possibility that at all ages, in addition
to modifying the behaviour of the silicates, C3A may make a direct
contribution to strength. T h i s contribution is likely to be small and, in
view of a strong interaction between the variables C3A and (C3S)x(C3A)
(equation 5), evaluation would be d i f f i c u l t . Finally, there is the need to
consider the influence of C4AF. At this stage, i t does not seem that the
Vol. 2, No. 6 679
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

model could be greatly improvedby taking this factor into account. The
data of Bogue and Lerch (1) show that as a cement, C4AF ranks well below
even C3A. In addition, with the Portland cements that have been examined
here, there is only a weak association between C4AF and strength. This
latter aspect has been considered in greater detail in an earlier
investigation (5).

Summary

In this investigation of the relationship between strength and the

composition and fineness of cement, results published during the past 40 years
are pooled, combined with new data, and subjected to regression analysis.
The earlier studies are also analysed individually, and experiments in which
the cement compounds were mixed separately with water are examined. On the
basis of these studies i t is concluded that:

1. Thereare considerable differences of opinion on the importance of

C3A relative to that of C3S.
2. The controversy has arisen because the cements used by some
investigators were not representative of Portland cement in general, and
because regression analyses were based on an unsuitable model.
3. An alternative model is described, in which, during the f i r s t few
weeks of hardening, the strength developed by C3S depends on the proportion
of C3A in the cement.
4. Analysis of the data in terms of the alternative model shows that
the sensitivity of C3S to C3A increases with age, and passes through a
maximum at about 28 days. At this age, the differences between the strengths
of the present cements are largely attributable to differences in C3A.
Within the limits of the experiment, the sensitivity of C3S to C3A is not
affected by the proportion of gypsum in the cement.

5. The regression of paste strength on the composition and fineness of

cement can be used for the prediction of concrete strength.

References

l • R.H. Bogue and W. Lerch, Ind. Eng. Chem., 26, 837 (1934).
2. H.F. Gonnerman, ASTMProc. 34, 244 (1934).
3. H.F. Gonnerman and W. Lerch, ASTMSpecial Tech. Pub. No. 127
(Portland Cement Assoc. RD Lab. Bull.39) (1951).
. R.L. Blaine, H.T. Arni and M.R. DeFore, NBS Build. Sci. Series 8,
Part 3, Section 7 (1968).
680 VoI. 2, No. 6
STRENGTH, RELATIONSHIP, COMPOSITION, FINENESS, CEMENT

. K.M. Alexander, J.H. Taplin and J. Wardlaw, Proc. 5th Int. Symp. Chem.
Cement, Tokyo, 111,152 (1968).
. P. Gourdin, Rev. Mat. Constr., No.651, 411 (1969).
7. P. Klieger, J. Am. Concr. Inst. (Proc.) 54, 481 (1957).
8. J.A. Dalziel,Cement Technol., 2, 105 (1971).
9. K.M. Alexander, Proc. Civil Eng. Mats. Conf. Southampton (1969),
(ed. M. Te'eni), Part 2, 1187, Wiley-lnterscience (1971).
I0. G. Verbeck, J. Portland Cement Assoc. RD Lab., ~, 57 (1965).
II. W. Lerch, ASTMProc. 46, 1251 (1946).
12. L.E. Copeland, E. Bodor, T.M. Chang and C.H. Weise, J. Portland Cement
Assoc. RD Lab., ~, 61 (1967).
13. G.L. Kalousek and J.E. Kopanda, Cement and Concrete Research, ~,
63 (1971).