Early Age Cracking in Cementitious Systems 207

6.2 MEASUREMENT METHODS FOR TESTING OF EARLY AGE

AUTOGENOUS STRAIN

Tor Arne Hammer

SINTEF Civil and Environmental Engineering
Trondheim, Norway
Øyvind Bjøntegaard and Erik J. Sellevold
The Norwegian University of Science and Technology,
Trondheim, Norway

6.2.1 Introduction

Today autogenous shrinkage is a “hot topic” with active research taking place in many countries
and yearly conferences devoted almost exclusively to the subject.

It is a trend that each publication presents new results and discusses and interprets these with
regard to other sets of results. However, as pointed out in a state-of-the-art review [1], autogenous
shrinkage as presented in different publications varies enormously, both in sign, i.e. from
contraction to expansion and magnitude. The influence of temperature varies even more. It is of
major importance to start a discussion of measurement techniques before going too far on
interpretations. This paper is a modest introduction to such a discussion.

6.2.2 Volumetric Measurement of Paste vs. Linear Measurement of Concrete

6.2.2.1 General
The purpose of the measurement of autogenous deformation may vary from one of fundamental
understanding of mechanisms to one of providing basis for practical design. The purpose will also
determine the choice of measuring system. From a fundamental point of view it may be of interest
to measure the volumetric autogenous deformation (AD) of pastes from the earliest possible point
of time, while from a design point of view it is of interest to determine the contribution to the
crack risk. The latter requires a linear measure (the one-dimensional horizontal component) of
concretes. In principle, AD may contribute to settlement and plastic shrinkage (at low water to
binder ratios) [2]. This is not considered here and is beyond the scope of this report. The start of
time of testing for design purposes should correspond to the time were the AD starts to become
stress-inducing (when restrained), i.e. time zero (see section 6.1).

Investigations have revealed that there is no consistent relationship between the volumetric AD of
paste and the measured linear AD of the concrete. This is of course unsatisfactory concerning the
reliability of the test methods. The point is discussed by Hammer [3] and Barcelo et al. [4]. A
summary is presented below.
208 RILEM TC 181-EAS: Final Report - July 2002

6.2.2.2 Volumetric measurements of pastes

The main driving force for AD, the chemical shrinkage, can be measured in an Erlenmeyer flask
as follows: A small amount of paste is filled in the flaskand the rest of the volume is filled with
water. The volume change is recorded as the reduction of the water level in the system [5]. The
flasks are placed in a water bath for temperature control. The method has proven to be reliable,
and it may be started at a short time after mixing (less than 15 minutes). The method is based on
the assumption that the self-desiccation pores fill with water. This is true in most cases, at least in
the first days, because the thickness of the paste sample is small (i.e. maximum 10 mm). Another
set of methods is based on weighing the amount of water needed to re-fill a water filled closed
sample container with a sealed paste-mortar sample. The sealing may be rubber membrane [6],
plastic sheets [7] or oil [8].

The measurement of the volumetric autogenous shrinkage (i.e. the external result of the chemical
shrinkage) is more complicated, mainly because of the role that bleeding plays: The test methods
discussed in [3] are based on measurement of a sealed samples (e.g. in rubber bags), which leave
any bleed water to remain on the top surface. A number of investigations have shown that the
subsequent absorption of the bleed water will be observed as an extra contraction, i.e. extended
time before the autogenous shrinkage curves deviates from the chemical shrinkage curve. In order
to avoid the bleeding, the rubber bag may be placed on rollers to be rotated. The effect of rotation
is discussed by Justnes et al [9].

6.2.2.3 Linear measurements of pastes

Barcelo et al. [4] have compared linear measurements and volumetric shrinkage measurements,
using the rubber bag (condom) method, on pastes and found considerably different results: The
volumetric method gave approximately 3 times as large shrinkage as the linear measurements
(converted to volumetric scale) at 72 days of age. They discussed several plausible reasons for the
difference:

- Friction between the paste and the mould in the linear test
- Pressure of the rubber bag induced by the stretching during filling
- Lower capillary underpressure in the volumetric test due to the sealing
- Non-isotropy; the sample is rotated in the volumetric test, and not in the linear

However, they concluded that the reasons discussed could not explain the magnitude of the
difference, and, that further research is needed.

Another point is that it is known that the latex used in the condoms is not 100 % water tight. In
order to provide some information about the significance of this, a simple test was performed at
NTNU/SINTEF: A condom was filled with pure water and placed on a balance in order to
measure water loss to the surrounding air (approximately 30 % RH) over time. The result showed
a weight decrease, i.e. a shrinkage of approximately 0.5 % by volume in 24 hours, which is close
to a magnitude higher than the measured linear shrinkage (converted to volumetric scale). This
demonstrates that water may penetrate the rubber if there is a sufficient driving force, and thus,
result in increased mass of the sample in the condom, which on the balance gives the same result
as reduced volume. At the time of writing tests are being carried out at “Lafarge” in Lyon, in order
to confirm if the capillary under-pressure of the paste is sufficient to suck water through the
condom wall.
 Early Age Cracking in Cementitious Systems 209

6.2.2.4 Linear measurements of mortars and concretes

Bjøntegaard has tested the AD at 20°C of a concrete and of the equivalent paste (w/b = 0.40 with
5 % silica fume), Figure 6.2.1 [1]. The paste did not contain any plasticisers, while the concrete
contained both plasticiser with a retarding effect, and superplasticer. As can be seen, the
development the first week is significantly different: The expansion of the concrete at the
beginning is due mainly to re-absorption of bleed water, (see 4.2). The paste did not show any
significant bleeding. The slower autogenous shrinkage development of the concrete in the first
week is probably partly due to swelling of the paste matrix as it absorbs water from the aggregates
(absorbed water in the aggregates constitutes approximately 15 liter/m3), and partly due to a
restraining effect of the aggregates. However, from approximately two weeks and on, there is no
significant difference in the developments. This is somewhat surprising since one should expect a
slower rate in the concrete due to a restraining effect of the aggregates.

The results demonstrate that early age AD of concrete can not be predicted from paste
measurements. Indeed the relationship between the two as shown in Figure 6.2.1, can not be
explained in a consistent way today. This is of course highly unsatisfactory and, until resolved,
possess a question masking the reliability of the experimental results.

400

350
Autogenous shrinkage (10 )

Cement paste x 0.28

-6

300

250

200

150

100
Concrete
50

-50
0 168 336 504 672 840
Time (hours)

Figure 6.2.1: Autogenous shrinkage of a concrete with w/b = 0.40 and an equivalent paste (but
without admixtures) converted to the paste content of the concrete (28 %)

6.2.3 Measuring Systems for Mortars and Concretes

Several types of measuring systems have been used to measure the length change of beams, slabs
or cylinders. They may be divided into six groups:

1 “Cast in nails” through a hole in the middle of the end plates with the nail heads
embedded in a concrete beam, e.g. Figure 6.2.2 [1]
2 Moveable endplates with “plugs” in a concrete beam, e.g. Figure 6.2.3 [10]
3 Horizontal transverse cast-in-bars through a concrete beam, e.g. Figure 6.2.4 [11]
4 Vertical cast in bars in a concrete slab, e.g. Figure 6.2.5 [12]
210 RILEM TC 181-EAS: Final Report - July 2002

5 Cast in strain gage, e.g. [13]

6 Metal plates placed on top of cylinders, e.g. Figure 6.2.6 [14]

The movement has been measured using inductive displacement transducers (IDT) or linear
variable differential transformers (LVDT). But also “non-contact” transducers like reflection of
electronic pulses or laser against a metal chip on the concrete have been used [10] and [12].

Figure 6.2.2: Measuring system with “cast in nails” 100·100·500 mm beam [1]

Figure 6.2.3: Measuring system with moveable end plates in a 40·40·160 mm beam [10]
 Early Age Cracking in Cementitious Systems 211

Figure 6.2.4: Measuring system with horizontal “cast in bars” in a 150·150·500 mm beam [11]

Figure 6.2.5: Measuring system

with vertical cast in bars in a
270·270·100 mm slab (designed
for plastic shrinkage mainly)
[12]

Figure 6.2.6: Measuring system with

concrete in 100 · 375 mm flexible tubes
and vertical length measurement
(displacement transducers on top) [14].
212 RILEM TC 181-EAS: Final Report - July 2002

6.2.4 Sources of Errors

6.2.4.1 Friction
The fact that all systems involve measurement on a specimen in a mould and that the measuring
points have to be fixed before casting (except the method presented in (14)) causes some points of
concern that require attention, as outlined below. This is mainly related to the influence of friction
between the concrete and the mould, and settlement which may give undesired movement of the
measuring points.

At the beginning, when the concrete stiffness is low, the friction may restrain the free movement
(the body creeps). However, the stiffness develops rapidly, and thus, the frictional effect becomes
insignificant probably some hours after the time of setting.

The settlement may cause a rotation of the measuring points, resulting in a horizontal movement
which will appear as shrinkage [3] in the time before setting. Another issue is that the settlement
may cause cavities around the measuring points that may give a “slack” which, in turn, may
influence the measurements at the beginning. Also, when fixed to the transducers the settlement
action may give a bending in the connecting rod that introduces friction in the transducer or in any
support, e.g. the hole in the end plate.

6.2.4.2 Bleeding
As already mentioned and discussed in [3], the measurements of AD can be influenced by bleed
water collecting on the concrete surface during the fresh phase (i.e. before setting). After setting,
this bleed water may then be reabsorbed by the concrete as self-desiccation occurs, resulting in
reduced autogenous shrinkage or even expansion. In addition, other water sources, such as water
available in the aggregates and internal bleeding may also play a role, since very little water is
required to “refill” the self-desiccateded pores and thereby eliminate autogenous shrinkage.

Bjøntegaard [1] has investigated the effect of (surface) bleeding on a concrete with w/b ratio of
0.40 (Figure 4.2.7), where the specimen dimensions were 100 x 100 x 500 mm. The following
trends were observed:

- The concrete shows expansion for few hours after setting as shown by the three nominal
identical tests marked “With bleeding” in Figure 6.2.7. As can be seen, the re-
absorption of bleed water (totally 1 l/m3 concrete) gives an expansion of 30 - 40 · 10-6
in the period between 7 and 12 hours.
- In the test marked “Bleeding removed” the bleed water was removed from the concrete
surface just before setting, which led to reduction in the expansion, to values around 10
· 10-6. This small expansion may be due to the fact that there may be internal water
available (i.e. from aggregate and internal bleeding). In that case, the removal of
external bleed water will not be enough to eliminate the entire expansion in this
concrete.
- The test denoted “Addition of extra bleed water” is somewhat special since extra water
was added to the concrete surface just before setting (giving totally 4 l/m3 concrete on
the surface). As can be seen, the expansion now lasted up to 16 hours and was as high
as 110 · 10-6.
 Early Age Cracking in Cementitious Systems 213

Hence, considering the tests described above it is quite evident that there is a marked influence of
bleed water on the subsequent measurement of AD. It is also notable that bleeding causes both a
small reduction of the w/b-ratio as well as a permanent increase of the degree of capillary
saturation. These two effects have probably the opposite effect on long term AD and the net result
of bleeding is therefore not obvious. Results from testing of AD at realistic temperature histories
showed an increase of the long term AD and tensile stresses when restrained, when the bleed
water was removed in one case, but no influence in another [1]. In any case, bleeding will disturb
a meaningful comparison between different concretes having different bleeding characteristics.
Also, the effect of bleeding will depend on specimen size - a point to keep in mind when
comparing results from different laboratories.

Furthermore, in most practical cases, only the concrete near the surface will be affected by
bleeding. And, as practical cases often involve quite massive cross sections, a more realistic test
procedure would therefore be to remove the bleed water before setting. On the other hand, for
more special practical situations, such as floors and thin concrete decks, the whole thickness is
influenced by bleeding and the bleed water should not be removed in AD measurements.

50

Bleeding removed
Autogenous deformation (10-6)

With bleeding (3 tests)

-50

-100

Addition of extra bleed water

-150
0 6 12 18 24 30
Time (hours)
Figure 6.2.7: Effect of bleed water on the autogenous deformation [1]

6.2.4.3 Temperature control

The AD has proven to be more temperature dependent than most of other concrete properties [1].
The hydration generated temperature rise in a specimen with a cross section of 100 – 150 mm in a
mould surrounded by air of constant temperature, may be up to 10°C, depending on the material
composition, specimen thickness and the insulation given by mould walls. Therefore, temperature
control is necessary in order to maintain the target isothermal temperature of the specimen through
the whole test. This has been done by using hollow moulds in which a fluid flows [1, 11] or by
submerging the whole system in a fluid bath at a constant temperature [14].

Temperature control is necessary when testing is done at given temperature histories (e.g.
according to a simulation of a given part of a structure). Then, control of the temperature
movement of the measuring points, transducers, base, etc, is required.
214 RILEM TC 181-EAS: Final Report - July 2002

6.2.4.4 Moisture protection

Usually, plastic or aluminium foils are used to seal the specimen. As the body remains in the
mould during testing (leaving the top surface “uncovered”, only) and the testing time is relatively
short, such foils will normally give sufficient sealing. However, weight measurement of the
specimen (and mould) at the start and the end of testing should be done in order to confirm it.

6.2.5 Recommended Test Method

A principle sketch of a test rig presently being used at NTNU/SINTEF is shown in Figure 6.2.2. It
is constructed with the purpose of minimizing the sources of error discussed above. The rig allows
recording of free length change vs. time for hardening mortar and concrete specimens of 500 mm
length and a 100 mm x 100 mm cross section [1]. The horizontally oriented specimen is
surrounded by a fully temperature controlled mould. Temperature control is provided by water
circulation in copper tubes that are fixed to the four copper plates which form the inner walls of
the mould. A large number of 20 oC isothermal tests (i.e. where the circulating water keeps 20oC
constant temperature) have shown that the temperature increase of the specimen is kept below
1oC.

At each end of the specimen an inductive displacement transducer (not shown) measures the
length change ΔL. The transducers are connected by an invar steel bar (not shown) to minimize
the sensitivity to variations in the ambient air temperature. The signals are recorded separately and
added to obtain the total length change. The transducers are fixed to a measuring rod made of
invar steel, with a disc at the end which extends 10 - 15 mm into the concrete sample; hence the
“active” length of the specimen is 470 - 480 mm. The measuring rods are placed through holes
that are made in the end plates. The end plates of the form are held in place during the early period
when the concrete applies hydrostatic pressure. After setting, the end plates may be gently moved
a few mm away from the specimen to allow it to expand freely.

The mould is lined with two layers of plastic foil with talcum powder in between, in order to
minimise the friction between the concrete and the mould. A thermocouple is placed at the centre
of the specimen. The top surface of the specimen is covered with a layer of aluminium foil that
provides a tight seal against diffusion of water. This aluminium foil is wide enough to be bend
down on both sides of the rig where it is fixed with a tape. The 5 mm copper plate cover is then
placed on top and weight is applied to provide a seal against moisture loss.

The length change measurements should be started as early as possibly (before setting is a
minimal requirement). A standard test procedure should consider how to handle the effect of any
bleeding (see section 6.4.4.2).

At each test the signals from the two inductive displacement transducers and the thermocouple are
recorded continuously.

Very good reproducibility for nominally identical mixes was demonstrated with this rig. A Round
Robin test, carried out within the Brite EuRam-project “IPACS”, involving six European
laboratories (NTNU/SINTEF included), testing nominally identical mixes, showed however a
large scatter. The tests were performed at a target isothermal temperature of 20ºC and also at a
“realistic temperature history” with a max. temperature of 40ºC. It was concluded that a part of the
 Early Age Cracking in Cementitious Systems 215

scatter may be explained from the fact that some of the laboratories used mixes which deviated
from the nominal mix. Nevertheless, the conclusion is also that differences in measuring
techniques are responsible for a part of the scatter. Further work is planned within “IPACS” in
order to investigate this.

6.2.6 References
1. Bjøntegaard, Ø., ‘Thermal dilation and autogenous deformation as driving forces to self-induced stresses
in high performance concrete’, Dr.ing. theses 1, The Norwegian University of Science and Technology,
Trondheim (1999) 121 pp.
2. Hammer, T.A., ‘The relation between settlement and plastic shrinkage of high strength concrete’, in
‘Early Age Cracking in Cementitious Systems’, Proc. International RILEM Conference, K.Kovler and
A.bentur, editors, Haifa, March (2001) 89-102.
3. Hammer, T.A., ‘Test methods for linear measurements of autogenous shrinkage before setting’, in
Autogenous Shrinkage of Concrete, Proceedings Int. Workshop 1998, Ei-ichi Tazawa, editor, E & FN
Spon, London (1999) 141-152.
4. Barcelo, L., Boivin, S., Rigaud, S., Acker, P. and Clauvaud, B., ‘Linear vs. volumetric autogenous
shrinkage measurements: Material behaviour or experimental artefact’, in Proceedings of the second Int.
Research Seminar on Self-Desiccation and its Importance in Concrete Technology, B.Persson and G.
Fagerlund, editors, Lund, Sweden (1999)109-126.
5. Justnes, H., Reyniers, B. and Sellevold, E.J., ‘An evaluation of methods for measuring chemical
shrinkage of cementitious pastes’, Nordic Concrete Research, 14 (1/94) (1994) 44 – 61.
6. Setter, N. and Roy, D.M., ‘Mechanical features of chemical shrinkage of cement paste’, Cement and
Concrete Research, 8 (1978) 623 – 634.
7. Ziegeldorf, S. and Hilsdorf, H.K., ‘Early autogenous shrinkage of cement pastes’, in 7th International
Conference on the Chemistry of Cement, Vol. 4, Paris (1980) 333 – 338.
8. Slate, F.O. and Matheus, R.E., ‘Volume changes on setting and curing of cement paste and Concrete
from zero to seven days’, ACI Journal, 1 (1967) 34-39.
9. Justnes, H., van Gemert, A., Verboven, F. and Sellevold, E.J., ‘Total and external chemicla shrinkage of
low w/c ratio cement pastes’, Advances in Cement Research, 8, (31) (1996) 121 – 126.
10. Morioka, M., Hori, A., Hagiwara, H., Sakai, E. and Daimon, M., ‘Measurement of autogenous length
changes by laser sensors equipped with digital computer systems’ in in Autogenous Shrinkage of
Concrete, Proceedings Int. Workshop 1998, Ei-ichi Tazawa, editor, E & FN Spon, London (1999) 191-
200.
11. Takada, K., van Breugel, K., Koenders, A.A.B and Kaptjin, N., ‘Experimental evaluation of autogenous
shrinkage of lightweight aggregate concrete’, in Autogenous Shrinkage of Concrete, Proceedings Int.
Workshop 1998, Ei-ichi Tazawa, editor, E & FN Spon, London (1999) 221-232.
12. Holt, E.E. and Leivo, M.T., ‘Autogenous shrinkage at very early ages’, in Autogenous Shrinkage of
Concrete, Proceedings Int. Workshop 1998, Ei-ichi Tazawa, editor, E & FN Spon, London (1999) 133-
140.
13. Hanehara, S., Hirao, H. and Uchikawa, H., ‘Relationships between autogenous shrinkage, the
microstructure and humidity changes at inner part of hardened cement paste at early ages’, in
Autogenous Shrinkage of Concrete, Proceedings Int. Workshop 1998, Ei-ichi Tazawa, editor, E & FN
Spon, London (1999) 89-100.
14. Jensen, O.M., Christensen, S.L., Dela, B.F., Hansen, J.H., Hansen, P.F. and Nielsen, A., ‘HETEK –
control of early age cracking in concrete – Phase 2: Shrinkage of mortar and concrete”. Report No. 110,
Road Directorate, Denmark Ministry of Transport (1997).