effect of aggregate shrinkage of concrete

Journal of Advanced Concrete Technology Vol. 6, No. 1, 31-44, February 2008 / Copyright 2008 Japan Concrete Institute 31

Invited paper

Effect of Aggregate on Drying Shrinkage of Concrete Tadashi Fujiwara1

Received 29 September 2007, revised 6 November 2007

Abstract Aggregate accounts for a large volumetric rate within concrete and influences the properties of concrete. In this investi-gation, the effect of aggregate properties on drying shrinkage of concrete was examined. It had been thought before that aggregate plays a role in restraining the shrinkage of cement paste, and that the shrinkage of aggregate itself can be ne-glected. However, according to the results obtained by this experiment, the shrinkage of aggregate is not necessarily negligible and it is strongly related to the shrinkage of concrete. Lightweight aggregate with a lower modulus of elastic-ity offers less restraint on the potential shrinkage of cement paste, then the large shrinkage of concrete is expected, but the actual shrinkage of lightweight concrete is comparatively small. It was demonstrated by this experiment that be-cause of the small shrinkage of lightweight aggregate, shrinkage of lightweight concrete is relatively small. The shrink-age of normal aggregate is generally larger than that of lightweight aggregate, and it is necessary to pay attention to the usage of normal aggregate.

1. Introduction

Structural artificial lightweight aggregate was used in Japan for the first time in 1964. The advantage of struc-tural lightweight concrete using lightweight aggregate is the reduction of dead load, so the application of light-weight concrete increased rapidly after that. Meanwhile, investigations of lightweight concrete were also pursued continually. The results of these investigations contrib-uted to the spread of lightweight concrete, but a number of problems and questions regarding lightweight con-crete also came out from these investigations.

One of the problems was that tensile and bending strengths decrease considerably during drying, and one of the questions was that drying shrinkage is compara-tively small. Drying shrinkage of lightweight concrete is expected to be large because lightweight aggregate with a lower modulus of elasticity offers less restraint on the potential shrinkage of cement paste. However, various experiments done in Japan have confirmed that shrink-age of concrete made with lightweight aggregate is ap-proximately equal to or even smaller than shrinkage of normal concrete.

The author carried out research in order to identify the causes of these problems and questions. It was proved that the reduction of tensile and bending strengths during drying is caused by the steep gradient of water content from the surface toward the interior of lightweight concrete, which causes large shrinkage stress. It was also made clear that drying shrinkage of lightweight concrete is small because of the small shrinkage of lightweight aggregate itself. These results were published in the Proceedings of Japan Society of

Civil Engineers by Goto and Fujiwara (1976, 1979), and the author thought that these problems were resolved completely.

Since then, the authors papers did not arouse public interest because the use of lightweight concrete in Japan did not increase, contrary to expectations. However, the authors results have been quoted recently in order to clarify the cause of a problem that does not concern lightweight concrete directly. The problem is the occur-rence of remarkable construction cracks and the suspi-cion attached to the quality of the used normal aggre-gate (JSCE 2005).

The author also dealt with normal aggregate for comparison with lightweight aggregate during the ex-amination of drying shrinkage, and lightweight aggre-gate showed smaller shrinkage than normal aggregate. In other words, shrinkage of normal aggregate was, generally speaking, larger than that of lightweight ag-gregate, and some normal aggregates showed extremely large shrinkage. In spite of this result, so far it has been thought that normal aggregate does not generally shrink to any significant degree. Then, the problem that the shrinkage of normal concrete appears large beyond ex-pectations emerged, and the authors indication that shrinkage of normal aggregate itself cannot be disre-garded began attracting special interest for the first time (Imamoto and Arai 2008 ; Asamoto et al. 2008).

The authors original papers were published only in Japanese. Professor Mihashi, Editor-in-Chief of ACT, considered a clarification of the influence of aggregate on shrinkage of concrete written up approximately 30 years to be of use and recommended to publication of the papers in question as an invited paper of ACT in order to disseminate this fact to the world. This paper was written based on his recommendation, with the last two past papers united as one and translated into Eng-lish.

1Professor, Dept. of Civil and Environmental Engineering, Iwate University, Morioka, Japan. E-mail:[email protected]

32 T. Fujiwara / Journal of Advanced Concrete Technology Vol. 6, No. 1, 31-44, 2008

2. Change in length of aggregate due to absorption and drying

2.1 Purpose Aggregate accounts for a large volumetric rate within concrete and influences the nature of concrete. Conven-tionally, it has been thought that drying shrinkage of concrete is caused mainly by shrinkage of cement paste, and that aggregate serves as a restraint on the shrinkage of cement paste. Accordingly, lightweight aggregate with a lower modulus of elasticity should offer less re-straint on the potential shrinkage of cement paste and this should lead to higher shrinkage of concrete than normal aggregate. However, not a few experimental results in Japan indicate that shrinkage of concrete made from lightweight aggregate is approximately equal to or even smaller than shrinkage of normal concrete.

These results are contrary to the general concept, and there is the possibility that the length change of aggre-gate itself is related to this unique phenomenon. The purpose of this experiment is to measure the length change of aggregate due to absorption and drying.

2.2 Method of experiment (1) Length change measurement method Aggregate particles are one of the difficult objects to measure the length change of, because they are com-paratively small in size and are irregular in shape and roundish, and have an uneven surface. These character-istics seem to be the reason why this field of research is difficult. Therefore, the establishment of measuring methods is the first step to be solved. In this experiment, three methods were devised as the result of many con-siderations and improvements.

The first method uses an electric wire strain gauge that is stuck on the circumference of the aggregate and travels almost fully around the circumference of the aggregate, as shown in Fig. 1. In the case of this method, dampproofing of the gauge is very important because the specimen is soaked in water. Aggregate is an absor-bent material, so moisture absorption from the back of gauge as well as the surface of gauge must be noted. The specimens were carefully prepared with full atten-tion paid to this point.

The second method uses a contact type dial gauge. In this method, several pieces of aggregate are connected in order to ensure a gauge length of 100 mm, as shown in Fig. 2. The measurement points are stuck on four faces, and the length change is the average of the four points. This is a mechanical measuring method that is not affected by moisture absorption and is more appro-priate than the electric method using a wire strain gauge. Warping due to wetting and drying was a concern be-cause the specimen is long and thin, but it was con-firmed by another preparatory experiment that while some warping occurs, it has practically no influence on length change values.

The third method uses an electronic micrometer, as

shown in Fig. 3. In this method, changes in electric re-sistance are used for measuring changes in length like the electric wire strain gauge, but the value of the resis-tance is not affected even if the measured part is ex-posed to water. However, this method is very sensitive to vibration, so special attention was paid to the mini-mization of vibration.

(2) Water content change measurement method In the case of the method using contact type dial gauge, changes in water content can be known directly by measuring changes in the mass of specimen. On the other hand, changes in water content cannot be meas-ured directly in the case of the method using an electric wire strain gauge and the method using an electronic

Fig. 1 Method using electric wire strain gauge.

Fig. 2 Method using contact type dial gauge.

Fig. 3 Method using electronic micrometer.

T. Fujiwara / Journal of Advanced Concrete Technology Vol. 6, No. 1, 31-44, 2008 33

micrometer. For the latter two cases, changes in water content were measured using another specimen that had the same surface area for absorption as the one for measuring the length change. Both were laid under the same environmental conditions. (3) Materials used For normal aggregate, river gravel from Shizukuishi River in Iwate Prefecture and Shiroishi River in Miyagi Prefecture was used. Because river gravel is a mass of stones with various qualities and its length change is thought to be affected by the quality of the stones, the river gravel was classified according to the stone quality. In order to confirm the influence of the stone quality, the length change of a core sample with a diameter of 15 mm consisting of rock from Iwate Prefecture was also measured.

For lightweight aggregate, three kinds of commer-cially available artificial lightweight aggregate were used. They were pelletized type M, coated type M and pelletized type B.

Table 1 lists the physical properties of the used ag-gregate.

In order to make the characteristics of the length change of aggregate clearer, the length change of ce-ment paste made from Ordinary portland cement was also measured, using a specimen consisting of a cylin-der with a diameter of 15 mm.

2.3 Experimental results and discussions (1) Comparison among length change measur-ing methods In this experiment, three methods for measuring length changes of aggregate were used. Figure 4 shows the length changes measured by these three methods for artificial lightweight aggregate pelletized type M. The aggregate under absolute dry condition was left in air with 70% relative humidity, and was then soaked in water for 24 hours. After soaking, it was left again in air with 70% relative humidity. The temperature of both the air and in water was 20C. The figure shows the change in strain from the beginning of soaking.

Each of the specimens swelled in water and shrank in air; this commonality persists regardless of the measur-ing method that is adopted, so the tendency of length

change can be known by any of the measuring methods. Comparing the absolute strain of each method, the change in strain measured by electric wire strain gauge is larger than that of the other two methods, and there is a residual strain after long-term drying in the case of this method (Fig. 4). These changes seem to be caused by insufficient dampproofing of the gauge. As men-tioned above, special attention was paid to dampproof-ing but there is a possibility that the dampproofing was not perfect. Thus the reliability of values measured us-ing the electric wire strain gauge can be judged as somewhat poor.

In the case of the electronic micrometer, the reliabil-ity of measured values is thought to be high because the results obtained with this method are not influenced by moisture absorption. However, measuring various specimens with this method takes a long time because the equipment can measure only one specimen at a time.

These two methods have their respective difficulties, while the method using the contact type dial gauge is not affected by moisture absorption and allows the si-multaneous measurement of many specimens. The val-ues measured by this method are close to the ones ob-

Table 1 Physical properties of used materials. Normal aggregate Lightweight aggregate Stone

Materials Shizukuishi Shiroishi

Pelletizedtype M

Coatedtype M

Pelletizedtype B

Marble Andesite Sandstone

0 2.49 2.60 1.20 1.15 1.19 2.71 2.59 2.52 2.57 2.62 1.29 1.26 1.21 2.72 2.67 2.57 w 3.2 0.8 8.1 9.6 1.3 0.2 2.9 1.9

0: Density under oven dry condition (g/cm3) : Density under saturated surface dry condition (g/cm3) w: Water absorption (%)

0

50

100

150

200

0 50 100 150 200Time (hrs)

Stra

in (

10-6

)

Electric wire strain gauge

Contact type dial gauge

Electronic micrometer

Sample: Lightweight aggregate (Pelletized type M)

in water in air

Fig. 4 Comparison of three methods for measurement of length change.


tained with the electronic micrometer, so the reliability of this method is thought to be high. In this experiment, the contact type dial gauge was mainly used for these reasons. (2) Length change of normal aggregate The variations of strain and water content of normal aggregate with the passage of time were measured under two different experimental conditions. One condition consisted in only soaking the aggregate in water and the other condition consisted in drying the aggregate dried in air with relative humidity of 70% after soaking it in water for 24 hours. Figure 5 shows the variations of granite from Shiroishi River.

The aggregate swelled in water and shrank in air al-most to the pre-soaking length. These changes in strain with the passage of time are similar to the changes in water content. Although this relationship between strain and water content is natural, it differs from the case of lightweight aggregate, as mentioned later.

The river gravels with other stone qualities and sam-ples made from rock showed the same tendency in length change as described in this figure. However, the absolute value of length change was influenced consid-erably by stone quality. Table 2 lists the average swell-ing strain of each aggregate by absorption in water for approximately 1600 hours. The swelling strain of river

gravel and rock remarkably depends on the stone qual-ity.

Figure 6 shows the relation between absorption and swelling strain of each specimen made from river gravel and rock due to soaking in water for 24 hours. The line connecting the average of each stone quality of river gravel is almost straight, so it is recognized that absorp-tion and swelling strain are roughly proportional to each other. On the other hand, the deviation of each specimen from this straight line is large, so swelling strain cannot be interpreted by absorption only.

(3) Length change of lightweight aggregate An example of the results about lightweight aggregate, measured under the same experimental condition as normal aggregate, is shown in Fig. 7. Swelling strain of lightweight aggregate after long term soaking is also shown in Table 2.

The tendency of aggregate to swell in water and shrink in air almost to the original length was observed alike in all three kinds of lightweight aggregate. Light-weight aggregate shows a characteristic change at the initial stage of drying, in which the rate of shrinkage of the aggregate is relatively slow, while the water content decreases promptly.

Figure 8 shows the relation between absorption and swelling strain of lightweight aggregate, like Fig. 6. While each measured value of pelletized type B is close

0

100

200

300

400

0 50 100 150 200Time (hrs)

Stra

in (

10-6

) ,

0.0

0.5

1.0

Wat

er c

onte

nt (%

) , w

, in water

w, in water

, in air

w, in air

0

500

1000

1500

2000

0 2 4 6 8Absorption (%)

Swel

ling

strai

n (

10-6

) Average ofeach group

Tuff

Andesite

Rhyolite

Shale

Granite

Fig. 5 Typical example for length change of normal aggregate (Granite from Shiroishi river).

Table 2 Swelling strain by absorption for long time(10-6).

1) Normal aggregate 2) Stone 3) Lightweight aggregate Granite Andesite Rhyolite Tuff Shale

Shizukuishi 450 690 1070 990 520 Marble Andesite Sandstone

Pelletized Type M

Coated Type M

PelletizedType B

Shiroishi 450 960 460 1200 - 20 370 320 130 350 150

Fig. 6 Relation between absorption and swelling strain for normal aggregate and rock.


to one another and so the quality of this aggregate is supposed to be uniform, the qualities of pelletized type M and coated type M vary widely. Compared in terms of average, the strain of coated type M is largest. When the average of each lightweight aggregate is correlated linearly, the inclination of the line is considerably dif-ferent from that of normal aggregate. This means that the magnitude of length change cannot be interpreted solely by absorption for both normal and lightweight aggregate.

(4) Length change of cement paste In order to clarify the role of aggregate in the shrinkage of concrete, similar measurements were carried out for cement paste. After curing in water for 7 days, the specimen of cement paste was dried in air with relative humidity of 70% until the shrinkage became almost constant, and was then soaked in water again. The length changes were measured under two conditions. One condition was soaking of the paste only in water for a long time and the other condition was drying of the paste in air after soaking in water for 24 hours. Figure 9 shows the changes in length and water content of the cement paste with a 50% water-cement ratio as an ex-ample of the measured results.

The tendency that the paste swells in water and shrinks in air is same as the case of aggregate.

(5) Comparison of each material Figure 10 shows typical examples of the change in length of each material due to absorption and drying. The length change of cement paste is considerably lar-ger compared with aggregate, so it can be reconfirmed from this figure that the dominant cause of length change of concrete is the length change of cement paste.

The length change of normal aggregate shows a value several times larger than that of lightweight aggregate. This experimental result is completely contrary to the expectation that the length change of lightweight ag-gregate must be larger because aggregate is porous and has a small elastic modulus. This comparison between aggregates gives an important key for solving the ques-tion of why drying shrinkage of lightweight concrete is unexpectedly small. It is highly probable that drying shrinkage of lightweight concrete is small because dry-ing shrinkage of lightweight aggregate itself is small.

Drying shrinkage of concrete causes many problems such as cracking. The small shrinkage of lightweight aggregate is desirable for lightweight concrete since it makes the drying shrinkage of concrete small. On the other hand, the length change of normal aggregate in almost all cases is larger than that of lightweight aggre-gate although it depends on the stone quality. As shown in Fig. 6, there exists normal aggregate whose strain is approximately 50010-6 in spite of its absorption of un-der 1%. This example indicates that the estimation of length change of normal aggregate done in the past was underrated.

0

50

100

150

0 50 100 150 200Time (hrs)

Stra

in (

10-6

) ,

0

5

10

Wat

er c

onte

nt (%

) , w

, in water

w, in water

, in air

w, in air

Fig. 7 Typical example for length change of lightweight aggregate (Pelletized type M).

0

100

200

300

400

500

0 2 4 6 8Absorption (%)

Swel

ling

strai

n (

10-6

)

Average of each group

Coated type M

Pelletized type M

Pelletized type B

Normalaggregate

Fig. 8 Relation between absorption and swelling strain for lightweight aggregate.

0

1000

2000

3000

0 50 100 150 200Time (hrs)

Stra

in(

10-6

),

0

10

20

30

40

50

60

Wat

er c

onte

nt (%

), w

, in water

w, in water

, in air

w, in air

Fig. 9 Typical example for length change of cement paste (W/C = 50%).


(6) Mechanism of length change of aggregate The drying shrinkage of concrete is dominated by the shrinkage of hardened cement paste. As for the mecha-nism of shrinkage of the hardened cement paste, much research has been done thus far and several theories have been presented. Regarding aggregate on the other hand, there has been a tendency to ignore the length change of aggregate, so research that discusses the mechanism of length change has been almost non-existent. In response, Kondo et al (1966) measured the length change of lightweight aggregate made ex-perimentally by themselves due to absorption and dry-ing, and they found that the aggregate swells remarka-bly in the late stage of drying. They thought that the swelling is caused by the disappearance of the negative pressure of the meniscus.

According to the capillary tension theory about dry-ing shrinkage of hardened cement paste, shrinkage due to drying is caused by the negative pressure in the cap-illary cavities that form the meniscus during drying. As drying progresses, the number of capillary cavities that form the meniscus increases, which increases shrinkage. When water in capillary cavities has been removed al-most perfectly by further drying, the negative pressure disappears so that the hardened cement paste swells to the pre-drying length. Kondo et al. thought that the cause of swelling of lightweight aggregate at the late stage of drying is the disappearance of the negative pressure of the meniscus according to the capillary ten-sion theory. If lightweight aggregate swells in concrete, it reduces the shrinkage of concrete, so the cause of the phenomenon of the shrinkage of lightweight concrete being unexpectedly small can be interpreted rationally.

In authors experiment, commercially available lightweight aggregate was used. Figure 7 shows the length change of the commercially available lightweight

aggregate due to absorption and drying. Even if the ag-gregate is dried up until the water content becomes ap-proximately zero, swelling does not occur. The figure shows only one measured example but all lightweight aggregates measured here merely shrank. Therefore, the swelling at the late stage of drying seems to be a phe-nomenon peculiar to the lightweight aggregate made experimentally, so the cause of the small shrinkage of lightweight concrete used actually cannot be explained by this swelling due to drying. It is also impossible to interpret the length change of commercially available lightweight aggregate by the capillary tension theory.

As for the mechanism of drying shrinkage of hard-ened cement paste, the surface adsorption theory, swell-ing pressure theory, interlayer water theory, as well as the capillary tension theory, are known as influential theories. According to any of these three theories, the paste will swell by absorption and shrink by drying. The length changes of both normal and lightweight aggre-gate due to absorption and drying measured in this ex-periment coincide with the length changes derived from the three theories. The interlayer water theory, however, is thought to be applicable only to hardened cement paste, while the other two theories may be able to ex-plain the length change of aggregate.

According to the surface adsorption theory, the sur-face energy of solids changes with adsorption of water to the surface of the solid or withdrawal of water from the surface, and the length of the solid changes as a re-sult. In the case of adsorption, the surface energy de-creases and the solid swells. According to the swelling pressure theory, the pressure that separates solid particle occurs due to seepage water in the case of absorption, causing the solid to swell.

Figure 11 shows changes in mass and length when aggregates under absolute dry condition were left in a fog room with relative humidity of 100%. Although the increases in mass of both normal and lightweight ag-gregate were slight, swelling strain was large. Especially in the case of the normal aggregate, the swelling strain was almost equal to the aggregate soaked in water, as shown in Fig. 5. This result means that the adsorption of water to the solid surface within aggregate causes the swelling, and that the decrease in surface energy is the dominant cause of the swelling compared with the swelling pressure.

The length change of porous material caused by the change in surface energy is expressed as follows ac-cording to Bangham et al. (1930).

/ = S/E (1) where, : specific gravity, S: surface area, : change in surface energy, E: modulus of elasticity.

Although many factors are related to the length change, the surface area seems to be the most important factor. Assuming that other factors are constant, the length change can be expressed as follows.

0

500

1000

1500

2000

2500

0 50 100 150 200Time (hrs)

Stra

in (

10-6

)

Cement paste(W/C = 50%)

Normal aggregate(Shiroishi river)

Lightweight aggregate(Pelletized type M)

in water in air

Fig. 10 Length change of each material.


/ = KS (2) The surface area, here, is the internal surface area,

which is the surface area of the substance surrounding pores within aggregate. This internal surface area was measured by the nitrogen adsorption method. Figure 12 shows the relation between this internal surface area and the swelling strain after soaking for approximately 1600

hours. It was difficult, as mentioned above, to find a definite

relation between absorption and the swelling strain for both normal and lightweight aggregates. On the other hand, there exists an obvious proportional relation be-tween the internal surface area and the swelling strain for both normal and lightweight aggregates, although there is some scattering of the data. This means that the most important factor related to length change of ag-gregate is the internal surface area, and that equation (2) based on the surface adsorption theory is basically cor-rect.

(7) Characteristic of lightweight aggregate In spite of the fact that lightweight aggregate is porous, its length changes due to absorption and drying are comparatively small. It can be seen from Fig. 12 that because of the small internal surface area, the length change of lightweight aggregate is small. The small internal surface area of lightweight aggregate, in spite of having a great many pores within itself, seems to come from the particular pore structure.

Figure 13 shows the pore size distribution of aggre-gates measured by the mercury penetration method. The lightweight aggregate has many pores whose total vol-ume is 3.5 times of the pores in the normal aggregate. The lightweight aggregate exhibits a great number of large size pores in particular, while having fewer small size pores compared to the normal aggregate. The in-ternal surface area is determined mainly by the amount of small size pores. The lightweight aggregate has a few small size pores, so the internal surface area seems to be small.

In manufacturing lightweight aggregate, raw material such as expanded shale is sintered at high temperature and the raw material solidifies into glass. At this time,

0

0.5

1

0 50 100 150 200Time (hrs)

Incr

ease

in m

ass (

%)

0

100

200

300

400

Swel

ling

strai

n (

10-6

)

Normal aggregate(Granite from Shiroishi river)Lightweight aggregate(Coated type M)

Fig. 11 Length change of aggregate due to adsorption (in air with R.H.100%).

0

500

1000

0 2 4 6 8Surface area (m2/g)

Swel

ling

strai

n

10-6

)

Normal aggregate (Shiroishi river)Normal aggregate (Shizukuishi river)Lightweight aggregate

Fig. 12 Relation between internal surface area and swelling strain by absorption for long time.

0.00

0.02

0.04

0.06

0.08

0.10

0.12

Diameter

Pore

vol

ume

(cm

3 /g)

Normal aggregate(Tuff)Lightweight aggregate(Pelletized type M)

nm3.5 17.7 177 1.77 17.7 177

m

Fig. 13 Pore size distribution measured by mercury penetration method.


gas occurs within the material and is confined by the glassy material, so large size pores form there. As the result of recrystallization to glass, the raw material looses its micro pores, resulting in a small number of micro pores of lightweight aggregate. The pore structure within lightweight aggregate characterized by an im-portant number of large size pores and a small number of micro pores comes from the high-temperature sinter-ing during the manufacturing process.

As already indicated in Fig. 7, changes with the pas-sage of time in water content and length due to absorp-tion and drying of lightweight aggregate differ from each other. Figure 14 shows the relation between water content and strain of lightweight aggregate, pelletized type M, which was soaked in water for a long time and then was dried in air.

There exists a proportional relation between water content and strain in the case of normal aggregate, and the locus of the drying curve almost coincides with that of the absorption curve. On the other hand, in case of lightweight aggregate, strain is not linearly proportional to water content, and the loci of the absorption and dry-ing curves do not coincide with each other. The charac-teristic of the pore structure of lightweight aggregate is considered to contribute to the complexity of the rela-tion between water content and strain. Considering the pore structure, the relation can be explained as follows.

The penetration of water into micro pores takes place first when lightweight aggregate is soaked in water. The aggregate swells rapidly because the surface area with adsorbed water is large. When water penetrates into large pores by further soaking, swelling slows down while the water content increases considerably. Water evaporation from large pores takes place first during drying, so the water content decreases promptly but shrinkage does not occur rapidly because water remains in the micro pores. When drying progresses until water

evaporates from the micro pores, shrinkage becomes rapid.

3. Effect of aggregate on drying shrinkage of concrete

3.1 Purpose According to the measurement of length change of ag-gregate due to absorption and drying, the length change of lightweight aggregate tends to be smaller than that of normal aggregate. From this result, drying shrinkage of lightweight concrete is presumed to be small mainly because shrinkage of aggregate itself is small, but it is necessary to confirm this presumption through the measurement of drying shrinkage of concrete. The pur-pose of this experiment is to clarify the effect of aggre-gate on drying shrinkage of concrete including the con-firmation of the presumption. 3.2 Method of experiment (1) Preparation of aggregate sample For convenience of analysis, it is desirable to use ag-gregate with uniform stone quality. The quality of lightweight aggregate particles produced from a par-ticular material seems to be approximately uniform, but river gravel is composed of stones with various qualities so crushed stones were used in this experiment as nor-mal aggregate. The crushed stones are not commercial available and those used were made from rocks gathered from Iwate Prefecture for this experiment. Classifying rock stones according to the origin of the rock, they are 19 kinds of igneous rocks, 4 kinds of metamorphic rocks and 15 kinds of sedimentary rocks, as shown in Table 3. Adding 5 artificial lightweight aggregates to 38 normal aggregates, a total of 43 aggregates were ex-perimentally investigated.

Normal aggregates crushed using a jaw crusher have angularities. Because the degree of angularity of each aggregate is different, the angularities may influence drying shrinkage of concrete. In order to reduce this influence as much possible, the angularities of every normal aggregate were removed to some extent by using a Los Angeles abrasion test machine in which normal aggregate was worn via 10,000 rotations. Next, normal aggregate was classified into coarse aggregate with sizes of 15-5 mm and fine aggregate with sizes of 5 mm under. The grading of coarse aggregate was adjusted such as 15-10 mm and 10-5 mm become equivalent. The fine aggregate was divided into 5 sizes using sieves with sizes of 5, 2.5, 1.2, 0.6 and 0.15 mm, and then each size of aggregate was mixed in equivalent amounts. Granules under 0.15 mm were excluded from fine ag-gregate. The grading of lightweight aggregate was also adjusted in the same manner as for normal aggregate. Thanks to these adjustments, it is not necessary to con-sider the effects of shape, maximum size and grading of aggregate on drying shrinkage of concrete.

Every aggregate was used under saturated surface-dry

0

50

100

150

0 5 10 15 20Water content (%)

Stra

in (

10-6

)

Drying

Absorption

Fig. 14 Relation between water content and strain of lightweight aggregate (Pelletized type M).


condition. Lightweight aggregate was boiled to achieve a fully saturated condition.

(2) Properties of aggregate The density under absolute dry condition and initial water content of used aggregates are also listed in Table 3. The initial water content of aggregate means the wa-ter content which the aggregate exhibited prior to the casting of concrete. The static modulus of elasticity and Poissons ratio of normal aggregate were measured us-ing a cylindrical specimen with a diameter of 30 mm and a height of 60 mm that was cored from parent rock. Lightweight aggregate was arranged in a rectangular form for the measurements. The length change of ag-gregate was measured by the method using a contact type dial gauge like in the previous experiment. A slen-der specimen with a diameter of 20 mm cored from rock was used for measuring the length change of normal aggregate.

(3) Properties of cement paste The water-cement ratio of concrete was 0.59 as de-scribed later. Cement paste with this water-cement ratio showed remarkable segregation and therefore it was very difficult to make specimens. For this reason, the properties of hardened cement paste were estimated from the measured results of mortar with various con-tents of fine aggregate, although the details of the ex-periment are not described here. The properties esti-mated were the static modulus of elasticity, Poissons ratio and drying shrinkage of hardened cement paste. (4) Properties of concrete In order to facilitate comparison among concrete speci-mens, mix proportions by volume were kept constant throughout, as shown in Table 4.

The concrete specimens were prisms of 4040160 mm and were dried in air at 20C, with relative humid-ity of 60% after water curing for 28 days. The length change of concrete due to drying was measured by a comparator. The changes in length and mass of every

Table 3 Properties of used aggregate.

1) Igneous rock 2) Metamorphic rock 4) Artifical ligftweight aggregate No. Stony kinds 0 w0 No. Stony kinds 0 w0 No. Kinds 0 w0 1 granite 2.62 0.76 20 hornfels 2.66 1.14 39 pelletized type M 1.27 19.332 2.27 6.72 21 3.03 0.57 40 coated type M 1.24 24.373 diorite 2.95 0.47 22 2.66 0.60 41 pelletized type B 1.25 17.374 gabbro 2.82 1.08 23 2.95 0.47 42 pelletized type J 1.34 9.655 diabase 2.74 0.96 3) Sedimentary rock 43 pelletized type S 1.36 5.376 porphyry 2.65 0.53 24 limestone 2.68 0.55 0: Density under oven dry condition (g/cm3)7 rhyorite 2.08 9.08 25 2.66 0.67 w0: Initial water content (%) 8 1.83 17.21 26 2.67 0.57 9 andesite 2.49 4.06 27 schalstein 3.00 0.63

10 2.46 3.46 28 2.89 0.81 11 2.56 2.30 29 2.39 7.76 12 2.08 10.81 30 leparetic tuff 2.15 8.49 13 2.59 2.41 31 andesite tuff 2.39 5.53 14 2.45 3.47 32 2.95 0.24 15 2.74 1.54 33 dacite tuff 1.74 14.83 16 2.51 3.54 34 tuff 1.61 21.41 17 dacite 2.47 3.85 35 sandstone 2.26 6.17 18 basalt 2.54 2.05 36 2.17 7.60 19 2.47 5.86 37 conglomerate 2.74 0.71

38 2.27 6.39

Table 4 Mix proportions.

Quantity of material per unit volume Fine Coarse aggregate

Water Cement aggregate G

510mm 1015mm

Maximum size of coarse

aggregate (mm)

Water cement

ratio W/C (%)

Fine aggregate percentage

s/a (%)

W (kg/m3)

C (kg/m3)

S (/m3) (/m3) (/m3)

15 59 48.9 214 362 313 164 164


concrete specimen were judged to reach approximately their equilibrium state after drying for 280 days. There-after, all specimens were dried in an electric oven at 105C until mass changes no longer occurred, and then the strain under absolute drying was measured.

3.3 Experimental results and discussions (1) Relation between properties of aggregate and drying shrinkage of concrete The extents of shrinkage of concrete made with each kind of aggregate are shown in Fig. 15. As is evident from the experimental condition, the differences among the various concrete specimens lie only in the properties of aggregate, and thus the extents of shrinkage of con-crete shown in the figure are brought about only from the difference in physical properties of aggregates. The extent is wide on the whole, and the ratio of maximum shrinkage to minimum shrinkage can reach 14. It is true that some of the aggregates used in this experiment are

of no practical use, but there is no doubt that the shrinkage of concrete depends strongly on the properties of aggregate.

Figure 16 shows the relation between the basic prop-erties of aggregate and drying shrinkage of concrete under oven dry condition. Some concretes showing ex-tremely large shrinkage are excluded from these figures, because they seem to require special consideration.

In case of normal aggregate, the basic properties of aggregate strongly relate to shrinkage of concrete, so it is necessary to check basic properties when selecting normal aggregate. Assuming that the region drawn by longitudinal lines in the figures indicates the conditions for normal aggregate to be of good quality, the extent of drying shrinkage of normal concrete whose aggregate meets these conditions is wider than that of lightweight concrete. This means that it is possible for drying shrinkage of normal concrete to be large, even if good-quality normal aggregate is used. The data of lightweight concrete are considerably apart from the tendency of normal concrete so that the drying shrink-age of both types of concrete cannot be comprehen-sively interpreted by the basic properties of aggregate.

It has been thought that aggregate restrains shrinkage of hardened cement paste and reduces the whole shrinkage of concrete as the result. Considered from such a viewpoint, the modulus of elasticity seems to be the most important property of aggregate. Figure 17 shows the relation between the modulus of elasticity of aggregate and drying shrinkage of concrete. Insofar as normal concrete is concerned, the larger the modulus of elasticity of aggregate, the smaller the shrinkage of con-crete, so the effect of restraint due to aggregate is ap-parent. On the other hand, lightweight concrete shows a different shrinkage behavior from normal concrete. This means that it seems not to be sufficient to take up only the restraining effect of aggregate as the role of aggre-gate with regard to drying shrinkage of concrete.

As mentioned above, length change of aggregate it-

0

1000

2000

3000

0 1 2 3

Density, 0 (g/cm3)

Dry

ing

shrin

kage

, c

(10

-6)

0 10 20Initial water content, wo (%)

0 20 40

Void percentage, p (%)

Normal concreteLightweight concrete

Fig. 16 Effect of physical properties of aggregate on drying shrinkage of concrete.

0

2000

4000

6000

Dry

ing

shrin

kage

, c (

10-

6 )

Igne

ous r

ock

Met

amor

phic

rock

Sedi

men

tary

rock

Ligh

twei

ght a

ggre

gate

Ave

rage

Oven dryingNatural drying

Fig. 15 Extent of drying shrinkage of concrete.


self due to drying is not necessarily negligible. Figure 18 shows the relation between length change of aggre-gate and drying shrinkage of concrete. The length change of aggregate is the strain from fully saturated condition to absolute dry condition.

There is an obvious proportional relation between these quantities, and especially both data of normal and lightweight concrete place approximately on the same straight line. This means that the length change of ag-gregate particle largely affects the shrinkage of concrete, and that because of the small length change of light-weight aggregate, the shrinkage of lightweight concrete can be small.

(2) Effect of length change of aggregate on drying shrinkage of concrete The above experiments demonstrated that the length change of aggregate is caused by a change in surface energy of the substance, and that the value of length change depends on the internal surface area. As shown in Fig. 12, there is a proportional relationship between the internal surface area and swelling strain after soak-ing for approximately 1600 hours. The aggregates used in the experiment here differ from those used in the pre-vious experiment. Figure 19 shows the relation between the internal surface area and drying strain of aggregate used here.

There exists an obvious proportional relation between these quantities for both normal and lightweight aggre-gates, reconfirming that the most important factor re-lated to the length change of aggregate is the internal surface area. This means that the length change of ag-gregate is caused by the change in surface energy, but there are exceptional aggregates that are not included in the figure. One of the exceptions is normal aggregate made from limestone. In Fig. 18, there are two meas-urement points located on the left side of the vertical axis. This means that these aggregates swelled due to complete drying compared with the initial stage under the fully saturated condition. This swelling of aggregate due to drying is a unique phenomenon and these two aggregates showing the swelling are both limestone. Three aggregates made from limestone were used in this experiment. Figure 20 shows their length change re-lated to change in water content when they were soaked in water and then dried in air.

One of the limestones swells due to absorption and shrinks due to drying so that length change of this ag-gregate seems to be caused by the change in surface energy like in many other aggregates. On the other hand, the mechanism of length change of the other two lime-stones cannot be explained by the change in surface energy because they shrink in water and swell just be-fore the end of drying almost to the pre-soaking length. The driving force causing such unique length change seems to be capillary tension. When the aggregates are soaked in water, a meniscus is formed in capillary cavi-ties and the aggregates shrink by capillary tension. At

0

1000

2000

3000

0 20 40 60 80 100 120Elastic modulus, Ea (kN/mm2)

Dry

ing

shrin

kage

, c

(10

-6)


Fig. 17 Effect of elastic modulus of aggregate on drying shrinkage of concrete.

0

1000

2000

3000

-200 0 200 400 600 800 1000 1200 1400Drying shrinkage, a (10-6)

Dry

ing

shrin

kage

, c (

10-

6 )


Swelling strain

Fig. 18 Effect of length change of aggregate on drying shrinkage of concrete.

0

500

1000

1500

0 2 4 6 8 10Surface area, Sa (m2/g)

Dry

ing

shrin

kage

, a (

10-

6 )

Normal aggregateLightweight aggregate

Fig. 19 Relation between internal surface area and drying shrinkage of aggregate.


the final stage of drying, capillary tension disappears due to the disappearance of the meniscus, so the aggre-gates swell to the pre-soaking length.

Figure 21 shows the relation between the decrease in mass and drying shrinkage of concrete due to drying. Most concretes including lightweight concrete merely shrink with decreases in mass, but two concretes using aggregate made from limestone swell at the last stage of drying, namely oven drying. It is evident that the swell-ing of these concretes is caused by the swelling of the aggregate itself.

As mentioned above, there is an opinion that the dry-ing shrinkage of lightweight concrete is small because lightweight aggregate swells due to drying. The result obtained by the authors experiment, which measured the length change of lightweight aggregate, was the di-rect opposite of the expected result. However, swelling of aggregate due to drying can occur depending on the stone quality of the aggregate, because swelling of ag-gregate made from limestone was also observed during the authors experiment.

The second exception not included in Fig. 19 is ag-gregates showing extremely large shrinkage. Three ag-gregates correspond to this category, and the shrinkages of these aggregates and concretes made with these ag-gregates are shown in Table 5. Because of the ex-tremely large shrinkage of aggregate itself, drying shrinkage of concrete is also extremely large.

This excessive shrinkage seems to be caused by the clay included within aggregate. During mixing and placing of concrete containing these aggregates, the workability of the concrete changed rapidly and fluidity was lost to a considerable extent. These phenomena were assumed to be caused by adsorption of mixing water to clay within the aggregate. It is well known that

clay causes excessive concrete shrinkage, so it is neces-sary to pay special attention to the existence of clay within aggregate.

(3) Effect of aggregate from standpoint of mul-tiphase aspect Concrete is a composite material that is made up of ce-ment paste and aggregate. Considering the properties of cement paste and aggregate, several theoretical formu-lae for shrinkage of concrete have been proposed. Among these theoretical formulae, Picketts equation (1956), below, has often been quoted.

c/p = (1-Va) (3. 1) = 3(1-c)/{1+c+2(1-a)Ec/Ea} (3. 2)

where, c: shrinkage of concrete, p: shrinkage of cement paste, Va: volume concentration of aggregate, Ec: modulus of elasticity of concrete, Ea: modulus of elas-ticity of aggregate, c: Poissons ratio of concrete, a: Poissons ratio of aggregate.

This equation was derived based on the theory about thick spherical shells, and the analytical model used seems to approximate the actual composition of con-crete. Substituting measured values of shrinkage of ce-ment paste and the modulus of elasticity and Poissons ratio of aggregate and concrete into Picketts equation, drying shrinkage of concrete for drying of 280 days was calculated. Figure 22(a) shows the relation between measured and calculated shrinkage of concrete. The predicted shrinkage from this equation does not satis-factorily explain the experimental results, and in par-

0

500

1000

1500

2000

2500

0 5 10 15 20Mass loss (%)

Dry

ing

shrin

kage

, c (

10-

6 )

No.18

24

25

38

42

39

Natural dryingOven drying

Fig. 21 Drying shrinkage of concrete.

Table 5 Drying shrinkage(10-6).

No. 29 33 34 a 6130 7690 4740 c 4440 6010 5460

-250

-200

-150

-100

-50

0

50

100

150

0 0.2 0.4 0.6

Shrin

kage

atra

in (

10-6

)

Water content (%)

AbsorptionDrying No. 24

No. 25 No. 26

Fig. 20 Length change of limestone due to absorption and drying.

Stra

in


ticular it is a characteristic defect of this equation that the extent of the calculated values is very narrow com-pared with the extent of the experimental values. Picketts equation does not include shrinkage of aggre-gate itself, so the adaptability of this equation seems to be poor.

The following Hansen-Nielsens equation (1965) was derived by developing Picketts equation, considering the shrinkage of aggregate.

c/p = (1-m){n+1+(n-1)Va2-2nVa}/(n+1)+m

for n1 (4)

c/p = (1-m){n+1- (n-1)Va}/{n+1+(n-1)Va}+m

for n 1 (5)

where, n: Ea/Ep, m: a/p, Ep: modulus of elasticity of cement paste, a: shrinkage of aggregate.

As shown in Fig. 22(b), the extent of the values cal-culated with this equation is wider than that of the val-ues calculated with Picketts equation. This seems to be the result of the fact that the shrinkage of aggregate was reflected into the calculated values, and it is recognized that the shrinkage of aggregate plays an important role for the evaluation of shrinkage of concrete. However, the correlation of the calculated values by this equation with experimental values is not still good, so another factor may be related to the shrinkage of concrete.

Shrinkage of hardened cement paste is restrained by aggregate so that hardened cement paste is subject to tensile stress. Because the stress holds on during drying, plastic deformation of the paste occurs. According to the test of tensile creep of hardened cement paste, creep coefficient of = 2.40 was obtained after loading of 280 days. In order to consider the effect of the creep, effective modulus of elasticity, Eeff = Ep /(1+), was substituted into Hansen-Nielsens equation. As shown in Fig. 22(c), the predicted shrinkage shows a reasonable agreement with the experimental shrinkage taking into consideration tensile creep of paste. Therefore, the plas-tic deformation of paste seems not to be negligible but rather an important factor related to shrinkage of con-crete.

Examining Fig. 22(c) in detail, whereas the correla-tion between measured and calculated shrinkage of normal concrete is fairly good, the correlation of light-weight concrete is relatively bad, that is, the calculated values are somewhat larger than the experimental values. This means that some factors, except for factors in-volved in Hansen-Nielsens equation, can be related to the shrinkage of lightweight concrete. The change in structure of paste due to high absorption of lightweight aggregate can be given as a factor. If water is supplied from the aggregate to the paste during drying, in addi-tion to the fact that shrinkage of the paste would be de-layed, the structure of the paste itself would change due to continuous hydration, resulting in small final shrink-age.

4. Conclusions

In this investigation, the length change of aggregate due to absorption and drying was measured by an original method and the effect of aggregate on drying shrinkage of concrete was examined. The results obtained can be summarized as follows. (1) Length change of aggregate is much smaller than

that of hardened cement paste, so it was confirmed by this experiment that the dominant cause of dry-

0

0.1

0.2

0.3

0.4

0.5

0.6(a) Pickett

0

0.1

0.2

0.3

0.4

0.5

0.6

Calc

ulat

ed v

alue

, c/

p

(b) Hansen-Nielsen

0

0.1

0.2

0.3

0.4

0.5

0.6

0 0.1 0.2 0.3 0.4 0.5 0.6Experimental value, c/p


(c) Hansen-Nielsen

Fig. 22 Examination of theoretical equation on drying shrinkage of concrete.


ing shrinkage of concrete is that of cement paste. However, shrinkage of aggregate is not necessarily negligible.

(2) Length change of lightweight aggregate tends to be smaller than that of normal aggregate. The length change with the passage of time of lightweight ag-gregate differs from change in water content. Lightweight aggregate exhibits a characteristic pore structure, such as a great number of large size pores but a small number of micro pores. The complexity of length change with the passage of time can be explained based on the pore structure and the proc-ess of water moving within aggregate.

(3) Length change of most aggregates seems to be caused by a change in surface energy due to ab-sorption and drying. In this case, the internal sur-face area of aggregate is strongly related to the value of length change. Because of the small inter-nal surface area, length change of lightweight ag-gregate is small.

(4) Because of the small shrinkage of lightweight ag-gregate, shrinkage of lightweight concrete is rela-tively small. Change in structure of paste due to high absorption of lightweight aggregate can also decrease shrinkage of lightweight concrete.

(5) Normal aggregate has been generally considered not to be subject to change in length due to drying. However, there are many normal aggregates show-ing large shrinkage because of a large internal sur-face area, even if their density and absorption meet the standard requirements. Drying shrinkage of concrete with such normal aggregate can be large, so special attention is required.

(6) Some normal aggregates made from limestone shrink due to absorption and swell due to drying. The driving force carrying such unique length

change seems to be capillary tension. Concrete with such aggregate swells at the last stage of drying. When using aggregate that contains clay, special attention is required because shrinkage of concrete can be extremely large.

References Asamoto, S., Ishida, T. and Maekawa, K. (2008).

Volumetric stability of aggregates and shrinkage of concrete as composites. Jour. Advanced Concrete Technology, 6(1), 77-90.

Bangham, D. H. and Fakhoury, N. (1930). The swelling of charcoal. Proc. Roy. Soc. A 130, 81-89.

Goto, Y. and Fujiwara, T. (1976). Volumetric change of aggregate by absorption and drying. Proc. JSCE 247, 97-108. (in Japanese)

Goto, Y. and Fujiwara, T. (1979). Effect of aggregate on drying shrinkage of concrete. Proc. JSCE 286, 125-137. (in Japanese)

Hansen, T. C. and Nielsen, K. E. C. (1965). Influence of aggregate properties on concrete shrinkage. Journal of ACI 62(7), 783-794.

Imamoto, K. and Arai, M. (2008). Simplified evaluation of shrinking aggregate based on BET surface area using water vapor. Jour. Advanced Concrete Technology, 6(1), 69-75.

JSCE concrete special committee for the performance of Tarui bridge, (2005). Midterm report. (in Japanese)

Kondo, R., Sekiguchi, A. and Minagawa, T. (1966). Relation between pore size distribution and length change of artificial light-weight aggregates. CAJ Review of the 28th general meeting, 257-262.

Pickett, G. (1956). Effect of aggregate on shrinkage of concrete and hypothesis concerning shrinkage. Journal of ACI 52, 581-590.

effect of aggregate shrinkage of concrete

Documents