5.

6 Phase composition of portland cement paste

The distribution of phases in a cement paste will depend primarily on the composition of
the cement clinker, the total amount of sulfate (including added gypsum), and the amount of
time the cement has been hydrating. Table 5.2 lists some phase data for a typical type I
OPC paste made at a w/c of 0.45. The oxide composition of such a cement was given in
Chapter 3. For the mature paste, it was assumed that the C
3
S and C
2
S reacted to near
completion, the C
3
A was completely consumed, and half of the C
4
AF had reacted. The
phase distribution in the mature paste was estimated based on the reactions given in the
previous section, assuming that the conversion of ettringite into monosulfate was
complete. In this case, the total amount of sulfate was such that some ettringite remained
in the paste. Note the low volume percent of capillary porosity in the mature paste (~11%),
which is a result of the relatively low w/c of 0.45. This indicates that the paste will be
relatively impermeable, and thus resistant to many forms of degradation. At a slightly
higher w/c of 0.5, the capillary porosity would increase to about 17%, making the paste
considerably more vulnerable.
Table 5.1: Phase data for a Type I OPC paste made with a w/c of 0.45.
Volume %
Phase Density (g/cm
3
) At Mixing Mature Paste
C
3
S 3.15 23.40 1.17
C
2
S 3.28 7.35 0.78
C
3
A 3.03 4.42 0.00
C
4
AF 3.73 2.87 1.39
Gypsum (C

H
2
) 2.32 3.47 0.00
C-S-H (solid)
a
2.65 0 29.03
C-S-H (with gel pores)
b
1.90 0 49.99
Portlandite (CH) 2.24 0 13.96
Ettringite (AFt) 1.78 0 6.87
Monosulfoaluminate (AFm) 2.02 0 15.12
Water 1.00 58.49 31.69
Gel porosity -- 0 20.96
Capillary porosity -- 58.49 10.73
a
Formula 1.7C-S-4H.
b
Formula 1.7C-S-1.6H.
Whereas the values in the last column of Table 5.2 were determined from a relatively
complex series of calculations that took into account the specific cement composition and
reactions, it is possible to estimate the overall volumes of hydration product, gel porosity,
and capillary porosity for any portland cement paste using a few simple equations. This
approach and the model described below were developed by T.C. Powers, an influential
cement researcher, and his coworkers in the 1940s [6],[7]. The individual hydration
products are not distinguished, but are considered as a single phase with a characteristic
density and internal gel porosity. This is clearly a simplification, but it is justified on the
basis of the relatively small differences in the mineral composition of ordinary portland
cements. The model also assumes that the initial and final volumes of a cement paste are
equal, which is close enough to true to make no difference.
The first step is to determine the degree of hydration of the cement, , which varies
between 0 initially and 1 for complete hydration of the cement. The degree of hydration is
proportional to the amount of water that has been chemically combined into the hydration
products, which can be determined by measuring the non-evaporable water. A small
sample of hardened paste (not concrete) is crushed into small pieces and then dried (see
Section 7.3.2) to remove the liquid or evaporable water. After weighing, the sample is
ignited by heating it to 1050C and then weighed again. At this temperature the hydration
products decompose and the non-evaporable water is driven off, essentially returning the
specimen to the original cement composition. The degree of hydration is estimated as
5.8
where w
n
is the weight of non-evaporable water divided by the weight of the ignited sample.
As a paste hydrates, the volume of hydration product (including gel porosity) increases
and the volume of capillary porosity decreases. The w/c of the paste determines the
amount of space available for hydration product to form. At higher w/c the cement can
completely hydrate leaving residual capillary porosity. At lower w/c the capillary porosity will
fill up before all of the cement has reacted, stopping hydration prematurely. Separating
these conditions is a critical w/c value that provides just enough space for all of the cement
to react with no leftover capillary porosity. This value was determined experimentally by
Powers and Brownyard to be w/c = 0.38 [6]. At this w/c, the fully hydrated specimen
consists of only hydration product, and so the volume of hydration product is equal to the
initial volume of cement and water. Assuming that the specific volume of portland cement is
0.32 cm
3
/g, this gives a value of v
hp
= 0.70 cm
3
of hydration product per gram of starting
cement. This value of v
hp
is independent of the w/c of the paste, and will scale linearly with
the degree of hydration, giving the following general formulas:
or

5.9
In the original model, the amount of gel water was defined as the amount of evaporable
water in a fully hydrated specimen of w/c = 0.38, which was measured to be v
gel
= 0.21 g/g
starting cement. Assuming that the gel pore water has a specific volume of 1 cm
3
/g, this
gives the formulas
or

5.10
It should be noted that vacuum or oven drying a paste removes not only liquid water but
also some of the water held in the structure of the C-S-H gel and aluminate phases. Thus
eq. 5.10 will give gel pore volumes that are somewhat larger than the amount of liquid water
trapped inside the C-S-H gel phase. The volume of unreacted cement in a paste is:
or
5.11
By definition, the volume of capillary porosity is v
cap
= v
tot
v
hp
v
cem
. Using eqs. 5.9 5.11
gives:
or
5.12
Using a value of = 0.95, we can compare the results of the Powers model with the
calculated phase distributions for a specific cement composition shown in Table 5.2. Using
the versions of eqs. 5.9 5.12 on the right (normalized to paste volume) gives:
v
hp
= 86.4% v
gel
= 25.9% v
cem
= 2.1% v
cap
= 11.6%
The corresponding values from Table 5.2 are:
v
hp
= 85.9% v
gel
= 21.0% v
cem
= 3.3% v
cap
= 10.7%
Agreement is reasonable, with the exception of the gel porosity. In Table 5.1 the gel
porosity includes only liquid water trapped in the C-S-H gel, so this is expected.

Chemical and autogeneous shrinkage
There are two types of volume changes that are associated with the hydration process:
chemical shrinkage and autogeneous shrinkage. As the terms imply, in both cases the
volume of the paste decreases. These two phenomena often occur together, and both scale
with the degree of hydration, so they are frequently confused.
Chemical shrinkage represents the change in the total saturated volume of the paste as
the cement is converted to hydration products. This volume change is negative, becuase
the the solid hydration products are denser than the cement and water from which they
form. The main reason for this is that the effective density of water that is bound into phases
such as C-S-H, CH, and ettringite is signifcantly higher than the value of 1 g/cc associated
with liquid water. This may be confusing because we have previously stressed the point that
the hydration products have a greater volume than the cement (only) from which they form.
The difference is whether the original liquid water is included in the calculation.
Autogeneous shrinkage occurs when the capillary pores start to empty of water in order
to provide water for continued hydration. This generates capillary stresses which cause the
paste to shrink. This is exacly the same mechanism that causes drying shrinkage; the only
difference is where the water that leaves the capillary pores is going. Autogeneous
shrinkage will only occur if there is no external supply of water to refill the capillary pores.
Such a condition is called sealed curing, because in a laboratory setting it is what occurs
when fresh paste is sealed into a plastic container so that no water can enter or leave.
Autogeneous shrinkage will occur in pastes of any w/c subjected to sealed curing, but in
practice it only becomes noticeable for w/c below about 0.5. This is because water is
drained first from the largest pores, and emptying of large pores causes little capillary
stress. Autogenous shrinkage will necessarily occur in the interior of large concrete or
cement paste volumes, because additional water cannot be supplied from the surface. (One
clever attempt to circumvent this problem is to use porous, water-saturated aggregate to
provide an internal source of water, but this is impractical in most circumstances).
In order for a cement paste to hydrate fully to = 1, there must be enough water to react all
of the cement, and there must be enough space for the hydration product to form. Both of
these parameters depend on the w/c of the paste, and the Powers model described above
can be used to estimate the minimum w/c value for complete hydration based on each of
the two criteria.