chloride binding capacity of mortars made with various

Journal of Advanced Concrete Technology Vol. 6, No. 2, 287-301, June 2008 / Copyright © 2008 Japan Concrete Institute 287

Scientific paper

Chloride Binding Capacity of Mortars Made with Various Portland Cements and Mineral Admixtures Tetsuya Ishida1, Shigeyoshi Miyahara2 and Tsuyoshi Maruya2

Received 9 May 2007, accepted 18 February 2008

Abstract The authors experimentally studied the chloride binding capacity of mortar specimens made with various combinations of Portland cement, blast furnace slag, and pozzolans. In the experiment, a pore liquid extraction method, chloride titration test, a quantitative analysis of Friedel's salt based on the XRD method, and a mercury intrusion porosimetry test were conducted in order to measure chloride ions, adsorbed chlorides on the pore wall, and solid-phase chlorides (Friedel's salt), separately. It was clearly shown that the amount of Friedel's salt strongly depends on the type of binder used, whereas adsorbed chlorides is controlled by the micropore structure and the characteristics of the hydrated products.

1. Introduction

The framework for the prediction of service life, dura-bility design and maintenance of concrete structures has recently been under improvement. For example, the Standard Specifications for Concrete Structures pub-lished by the Japan Society of Civil Engineers (2002) specifies the verification for reinforcing bars corrosion caused by the ingress of chloride ions in the design of concrete structures. The method of testing the diffusion coefficient of chloride ions is also standardized (JSCE 2003). Challenges for the future are to improve per-formance evaluation techniques and durability design by quantifying and generalizing the movement and equilib-rium of chlorides based on the latest framework.

Chlorides present in hardened cement pastes are gen-erally classified into chloride ions and bound chlorides (Maruya et al. 1998) (Fig. 1). Chloride ions present in concrete are dissolved in the pore liquid and are in a freely-mobile form. Bound chlorides, on the other hand, apparently do not move at ordinary concentration gra-dients or in ordinary advective environments. Bound chlorides can be further classified into constituents of hydrates, as Friedel's salt (C3A•CaCl2•10H2O), and chlorides adsorbed into the pore walls. The total quantity of chlorides in a hardened cement paste is the sum of the amount of chloride ions and bound chlorides. Hardened cement pastes with good binding capacity are expected to be relatively low in chloride ion content and highly re-sistant to steel corrosion. Accordingly, it is important to have a grasp of the relationship between chloride ions and bound chlorides in concrete under various conditions so as to improve the accuracy of method used to predict

the chloride-induced corrosion of reinforcing bars in concrete.

Chlorides bound as Friedel's salt (solid-phase chlo-rides) are thought to be dependent on the Al2O3 content in the binder and the quantity of monosulfate formed (Glass et al. 1997). Further, it is expected that the quan-tity of chlorides adsorbed into hydrates (adsorbed chlo-rides) will vary depending on the geometric characteris-tics of pore structure, electric charge of pore surface, etc. The objective of this research is to quantify the equilib-rium between chloride ions and bound chlorides in hardened cement pastes. The specimens consist of ce-ment and varying mineral compositions and gypsum contents as well as some made with cement and ground granulated blast furnace slag and pozzolans to examine the effects of mineral admixtures.

2. Test methods

(1) Materials and mix proportions The physical properties and chemical composition of the ordinary Portland cement (OPC), high early-strength Portland cement (HPC), low-heat Portland cement (LPC), ground granulated blast furnace slag (BFS), fly ash (FS) and silica fume (SF) used in the test are listed in Table 1. In this test, the used blast furnace slag did not contain gypsum. The specific surface area of each material ex-cept for the silica fume corresponds to the Blaine fine-

1Associate Professor, Department of Civil Engineering, University of Tokyo, Japan. E-mail:[email protected] 2Civil Engineering Laboratory, Taisei Technology CenterTaisei Corporation, Japan.

Cement hydrate

Cl- Cl-Cl Cl

Cl ClCl- Cl-

Cl- Cl-Pore liquid Cl- Adsorbed

Chloride

Cl Solid-phase Chloride

Cl- Chloride ion

Bound Chloride

Fig. 1 Classification and definition of chlorides in hardened cement pastes3).

288 T. Ishida, S. Miyahara and T. Maruya / Journal of Advanced Concrete Technology Vol. 6, No. 2, 287-301, 2008

ness index, whereas that of the silica fume in the table is not an actual measurement of the lot used in the test but the average (published data measured by the BET method) of products on the market.

The series of test specimens and their materials and mix proportions are listed in Table 2. For the series using Portland cement without mineral admixtures, specimens made with ordinary Portland cement (NC), high early-strength Portland cement (HC) and low-heat Port-land cement (LC) were prepared. To examine the effects of gypsum content on the amount of bound chlorides, two test series (NGS and NGL) with varying gypsum contents were prepared. The total quantity of cement and

gypsum is in this research defined as the quantity of binder. The purity of gypsum is more than 98%. From a practical point of view, the quantity of gypsum specified in the mix seems quite large. In this research, intention-ally large quantity of gypsum was used in order to see the effect of gypsum on chloride binding capacity of hard-ened cement paste matrix. Prefixes B, F and S added to the specimens in mineral admixture series represent ground granulated blast furnace slag, fly ash and silica fume, respectively. The number suffix on these test series indicates the admixture replacement ratio as a percentage by mass of total binder. Mortar specimens having a wa-ter-to-binder ratio of 50% were prepared using JIS

Table 1 Physical properties and chemical compositions of materials used in the tests.

Chemical composition (% by mass) Materials Density

(g/cm3)

Specific surface

area (cm2/g)

LOI SiO2 Al2O3 Fe2O3 CaO MgO SO3 Cl

OPC 3.18 3110 0.49 21.5 5.65 2.93 64.9 1.07 1.75 0.005

HPC 3.13 4320 1.39 20.0 5.01 2.79 65.9 0.80 2.93 0.008

LPC 3.24 3410 0.41 25.8 3.41 3.51 62.4 0.82 2.40 0.005

BFS 2.91 4150 0.01 34.31 14.33 0.25 43.0 5.66 － <0.001

FA 2.16 3480 3.20 58.24 25.58 5.02 2.42 1.34 － <0.001

SF 2.25 200000 1.13 95.8 0.62 0.68 0.18 0.46 0.10 －

Table 2 Series of test specimens, materials and mix proportions used in the tests.

Content per unit volume of concrete (kg/m3) Test series Water Cement Gypsum Fine aggregate HC 416.5 832.9 － 838.0 NC 419.0 838.0 － 838.0

LC 422.0 844.1 － 838.0

NGS 416.2 792.7 39.7 838.0

NGL 413.4 747.9 78.8 838.0

Test series Water Cement (OPC)

Ground granulated blast furnace slag Fine aggregate

B20 416.0 665.7 166.4 838.0 B40 413.1 495.7 330.5 838.0 B60 410.2 328.2 492.2 838.0 B80 407.3 162.9 651.8 838.0

Test series Water Cement (OPC) Fly ash Fine aggregate

F20 404.3 646.8 161.7 838.0 F40 390.5 468.6 312.4 838.0

Test series Water Cement (OPC) Silica fume Fine aggregate

S20 406.1 649.7 162.4 838.0 S40 393.9 472.7 315.1 838.0

T. Ishida, S. Miyahara and T. Maruya / Journal of Advanced Concrete Technology Vol. 6, No. 2, 287-301, 2008 289

standard sand (percentage of water adsorption: 0.42%; density: 2.64 g/cm3) for the fine aggregate and ion-exchanged water for mixing the cement pastes.

The properties of the fresh mortar specimens were almost same except for the test specimen with a 40% silica fume replacement ratio. Accordingly, 0.4% of an air-entraining and high-range water-reducing agent (as a percentage by mass of binder) was added to this test specimen to adjust its properties.

(2) Method of preparing test specimens and curing conditions The method used to mix the cement pastes was in ac-cordance with JIS R5201 10.4.3 and the mortar were mixed and placed in forms at 20°C. The dimensions of the prepared test specimens were φ50 x 100 mm. Forms were removed on the day following placement and the specimens were then cured under water for 28 days. For series F and S specimens, which included Pozzolans, the specimens were cured under water for 91 days so that the pozzolanic reactions sufficiently proceed. After comple-tion of curing, test specimens were cut into slices and cubes for the various types of test, as described later. To be more precise, sections 10 mm from the ends of a test specimen were removed to eliminate the effects of seg-regation. From the remaining 80 mm long section of the cylindrical test specimens, six φ50 x 10 mm slice speci-mens and about 250 cubic specimens 7 mm on a side were taken using a wet-type diamond cutter. The test specimens were then immersed in the NaCl solution of the prescribed concentration described later. (3) Test items and methods a) Definition and determination of the quantities of chloride ions and bound chlorides Basically the JCI method (JCI-SC4) was used to analyze the quantity of chlorides in the hardened cement pastes. Strictly speaking, however, the quantity of soluble chlo-rides obtained by this means is not equal to the total quantity of chloride ions present in the pore liquid. This is because soluble chlorides are extracted in warm water at 50°C, so they can be expected to include some chlo-rides that would be adsorbed at a normal temperature of

20°C and some of the solid-phase chlorides that take the form of Friedel's salt. For this reason, a comparison was made between the total quantity of chloride ions meas-ured with a pore liquid extractor (Barneyback and Dia-mond 1981) and the quantity of soluble chlorides in a test specimen obtained under the same conditions to formu-late the relationship between the two quantities for the respective test series (Maruya et al. 1998). This allows for conversion of measurements obtained using the rela-tively simple and easy JCI method to the true quantity of chloride ions. The total quantity of chloride ions in a test specimen was determined from the product of the con-centration of chloride ions in the pore liquid and the volume of the pore liquid. The volume of pore liquid was defined as the volume of pores (3.2 nm to 320 μm in radius) measured using mercury intrusion porosimetry in the test specimen after immersion in the NaCl solution. More specifically, it was assumed that the volume of spaces in which chloride ions were capable of being present was equal to the volume of pores measured by mercury intrusion porosimetry.

After the 10 mm test slices were immersed in NaCl solution (using three levels of concentration as listed in Table 3) for 14 and 28 days, the concentration of chloride ions in the pore liquid was measured, the quantity of total soluble chlorides was measured by the JCI method and the quantity of Friedel's salt was determined by powder X-ray diffractometry (XRD). Further, in order to accel-erate the penetration of chloride ions from the exterior in a short period of time, cubic specimens with a large ratio of immersed area to volume were used in addition to the slice specimens for the measurement of the total quantity of chlorides and soluble chlorides. The concentrations of chloride ions in the solutions used for immersion of test specimens and immersion periods are listed in Table 3.

b) Determination of quantity of Friedel's salt The quantity of Friedel's salt was determined by the XRD internal standard method (referred to here as the XRD method) using synthesized Friedel's salt and its calibra-tion curve. Methods of determining the quantity of cal-cium aluminate hydrates by the XRD method aiming at pastes of Portland cement and synthetic clinker minerals

Table 3 Test items.

Test items Shape of test specimens Concentration of chloride ions in

immersion solution

Period of immersion

Measurement of chloride ion concentration Measurement of quantity of chlorides by the JCI method Determination of quantity of Friedel’s salt

Sliced φ50 x 10 mm

3 levels*1 (1, 3, 10%)

2 periods 14, 28 days

1% 56 days 3% 28, 56 days Measurement of quantity of

chlorides by the JCI method Cubic

7 mm on a side 10% 14, 28, 56 days *1 5 levels (1, 3, 6, 10, 15%) for test series NC and B20


have been reported in several papers (Sakai et al. 1998; Inoue et al. 2002). However, there have been few ex-amples of the XRD method being used for mortar specimens. Accordingly, the procedure used to determine the quantity of Friedel's salt by the XRD method is de-scribed below (Fig. 2).

The standard substance of Friedel's salt was synthe-sized by hydrating the mixtute of C3A and CaCl2 with stoichiometric ratio at a water-to-powder ratio of 2, and cured for 7 days (Japan Cement Association 2001). The C3A was prepared by firing the mixture of reagent grade calcium carbonate and aluminum oxide. Ideally, the density and chemical composition of a sample for pre-paring the XRD method calibration curve should be close to those of the test specimen to be measured (The Japan Society for Analytical Chemistry 2005). For this test, standard sand at 50% as specified in the mix proportion was added so that the sample was under conditions as close as possible to the test specimen after immersion in the NaCl solution. Calcium carbonate was used as a diluent to supply calcium; it was mixed with the synthe-sized Friedel's salt in prescribed quantities. For example, the quantity of Friedel's salt required to give 1% on a mass percentage basis on the calibration curve was ob-tained by mixing standard sand, calcium carbonate and Friedel's salt to make a 50/49/1 mix of standard sand/calcium carbonate/Friedel's salt on a mass per-centage basis. As an internal standard, 10% α-Al2O3 (by mass of sample) was added to the above sample. The standard sand, diluent, standard substance (Friedel's salt) and internal standard (α-Al2O3) were dried at reduced pressure using a water aspirator for 7 days and pulverized before use.

The diffraction peak used for the determination of quantity was at 2θ = 11.3° [002] for Friedel's salt and at 2θ = 52.5° [024] for α-Al2O3 (CuKα = 1.5405 Å). X-ray diffraction measurement conditions were as follows: tube voltage = 40 kV, tube current = 250 mA, scan rate = 0.2 deg/min and step size = 0.02 deg. The area of dif-fraction peak was determined by calculating an inte-grated intensity of the function after fitting of the peak propfile using the least-squares method. The calibration curve was prepared from the ratio of diffraction peak area (of the standard substance to internal standard). In this case, if a diffraction peak of Kuzel's salt was present adjacent to the that of Friedel's salt, the integrated inten-sities of the Friedel's salt and Kuzel's salt were separated by the addition of a function corresponding to the dif-fraction peak of Kuzel's salt and by function-fitting the diffraction peak. By this means, only the integrated in-tensity of the Friedel's salt was used. As can be seen from Fig. 3, the calibration curve shows good linearity.

Using the calibration curve, the quantity of Friedel's salt in test specimen after immersion in the NaCl solution was determined as follows. A mortar test specimen im-mersed in the solution for the prescribed period was pulverized with acetone to stop hydration, and after suc-tion filtration the pulverized specimen was dried in a water aspirator for 7 days, which corresponds to the same drying condition as in the case where the calibration curve was obtained. The test specimen was further pul-verized before measurement and 10% internal standard (α-Al2O3) by mass of specimen was added to the pul-verized material. The mass of the test specimen used for determination of the quantity of Friedel's salt was stan-dardized at 2.0 g. The ratio of diffraction peak area of the

1. Synthesis of a standard substance (Friedel’s salt)

2. Preparation of calibration curve

3. Determination of the quantity of a sample to be examined

• Synthesis of C3A by firing• Synthesis of Friedel’s salt by hydration: C3A + CaCl2 = 1:1 (mole ratio) Hydration for 7days

• Drying: aspirator for 7days• Preparation of sample for determining calibration curve: Mixing of standard sand, CaCO3, Friedel’s salt, and internal standard substance (α-Al2O3) at prescribed mix proportions• X-ray determination of quantity

• Drying: aspirator for 7days• Preparation of sample to be examined: mixing of sample to be examined + internal standard substance (α-Al2O3, 10% by mass of cement)• X-ray determination of quantity

Fig. 2 Procedures for determining quantity of F salt.


Friedel's salt to that of the internal standard, as obtained from X-ray diffraction, was calculated and the quantity of Friedel's salt in the mortar was then determined from the calibration curve.

c) Method of calculating quantity of chlorides nor-malized by mass of binder The total quantity of chlorides and soluble chlorides determined by the JCI method and the quantity of Friedel's salt determined by the XRD method are given as a ratio to the mass of the sample (% by mass of sam-ple). Although this measure is effective for evaluating the total quantity of chlorides in the concrete, it does not provide a generalized unit suitable for various mixing conditions when evaluating differences in chloride binding capacity due to the type of cement and admix-tures. For example, taking into account the fact that chlorides in hardened cement pastes are bound by the cement hydrates, then in mixes with a large cement content per unit volume of concrete, the quantity of chlorides bound per mass of sample is high because the quantity of hydrates contributing to binding chlorides is large. For this reason, the measured chloride amounts were rearranged as a ratio to the mass of binder (% by mass of binder).

To convert the measured quantity of chlorides per mass of sample to the quantity of chlorides per mass of binder, the simplest and easiest method is to assume that the unit mass of binder, the unit mass of water and the unit mass of aggregate in the test specimen are the same as in the original mix. However, as shown in Fig. 4, the quantity of binder may vary even for an equivalent sample mass, since the quantity of water remaining in a sample after D-dry or drying in a water aspirator varies with drying conditions and the degree of hydration. To eliminate factors causing such errors wherever possible, corrections are made by the loss on ignition method in converting from the quantity of chlorides by mass of sample to that by mass of binder.

First, the ratio of a component in a test specimen is determined after D-dry or drying in a water aspirator using the Equation (1) below, using the quantity of water determined by the loss on ignition method.

( )100S B

Wn

P P IP

I+ ⋅

=−

(1)

where PWn = ratio of mass of water (mainly bound water) contained in the sample determined by the loss on igni-tion; PS = ratio of mass of fine aggregate; PB = ratio of mass of binder; I = measured loss on ignition (% by mass of sample). Because the quantity of fine aggregate is the same as the quantity of binder, PS and PB in Equation (1) are both set at 1.0.

From the above equation, the quantity of chlorides per mass of sample obtained by the JCI method and XRD method can be converted into a quantity per mass of binder using Equation (2).

B

WnBSsamplebinder P

PPPClCl ++⋅= (2)

where Clbinder = quantity of chlorides per mass of binder (% by mass of binder) and Clsample = quantity of chlorides per mass of sample (% by mass of sample).

Next, a method of converting the concentration of chloride ions obtained with the pore liquid extractor is described below. The concentration of chloride ions per mass of binder Cfree (% by mass of binder) is given by Equation (3) below.

y = 0.5131x + 0.0213R2 = 0.9929

0.0

1.0

2.0

3.0

4.0

5.0

6.0

0.0 5.0 10.0

Area ratio of Friedel’s salt to Al2O3

Quantity of Friedel’s salt added (% by mass) Fig. 3 Calibration curve prepared by XRD internal stan-dard method.

C

ompo

nent

ratio

(% b

y m

ass)

Initial mix proportion

Component ratio in test specimen after drying

0.0

0.5

1.0

1.5

2.0

2.5

3.0 WaterBinderFine aggregate

PS

PB

PWn

Fig. 4 Determination of component ratio of a component by loss on ignition method.


100⋅⋅⋅=B

VMCC ionClClfree (3)

where CCl = concentration of chloride ions in the pore liquid [mol/m3]; MCl = atomic mass of chlorine [mol/m3]; Vion = volume in which chloride ions are present [m3/m3]; and B = mass of binder per unit volume of concrete [kg/m3]. As noted above, because the volume of pores (of radius 3.2 nm to 320 μm) measured by mercury intrusion porosimetry is defined as the volume in which chloride ions are capable of existing, the Equation (4) holds as follows.

n

ion

WSBVV++

= (4)

where V = volume of pores (of radius 3.2 nm to 320 μm) per unit mass measured by mercury intrusion po-rosimetry [m3/kg]; S = mass of fine aggregate per unit volume of concrete [kg/m3]; and Wn = ratio of mass of water (mainly bound water) contained in a sample de-termined by the loss on ignition [kg/m3]. Finally, Equa-tion (5) is obtained from Equations (3) and (4).

( ) 100⋅++⋅⋅

= nClCl

free WSBB

VMCC (5)

3. Chloride binding capacity of mortar using Portland cement without admixtures

(1) Relationship between chloride ions and soluble chlorides The concentration of chloride ions in the pore liquid is measured by potentiometric titration of a solution ob-tained with a pore liquid extractor using a silver nitrate solution. Figure 5 shows actual examples of measure-ments. The figure shows the concentration of chloride ions in the pore liquid extracted from test specimens HC, NC and LC immersed in NaCl solutions with three levels of concentration for 28 days. The 1, 3 and 10% NaCl solutions are equal to 173, 529 and 1,900 mol/l in chlo-ride ion concentration. As this figure shows, the con-centration of chloride ions present in hardened cement pastes is higher than the concentration in the immersion solution and varies with the type of cement. Similar tendencies have been reported in past research (Someya et al. 1989; Maruya et al. 1998). In addition, the lower the concentration of the NaCl immersion solution, the relatively greater the increment in the concentration of chloride ions in the pore liquid. This phenomenon is interpreted as follows: positively charged pore surfaces attract negative chloride ions, resulting in the increase in the concentration of chloride ions in the pore liquid (Maruya et al. 1998). Accordingly, this result indicates

0

100

200

300

400

HC NC LC

1%NaCl solutionimmersed for 28days

0

200

400

600

800

HC NC LC0

500

1,000

1,500

2,000

2,500

HC NC LC

Concentration of Cl- in extracted pore liquid [mmol/l]

173

529

1,900





Fig. 5 Concentration of chloride ions in pore liquid extracted from test samples.


that the same value cannot be assumed for the concen-trations of chloride ions in hardened cement pastes and immersion soluition even if the chloride ions are in equilibrium.

Figure 6 shows the relationship between the quantity of chloride ions and that of soluble chlorides obtained from the test. These values are expressed as percentage by mass of powder. The quantity of soluble chlorides is determined from Equation (2) and the quantity of chlo-ride ions in a test specimen is determined from Equation (5). For the test series with gypsum added, the quantity of powder is the sum of the cement and gypsum quantities. As shown in the figure, there is a greater quantity of soluble chlorides than the chloride ions in the test specimens under the same conditions of immersion pe-riod and concentration. This relationship between chlo-ride ions and soluble chlorides is formulated as given by Equation (6) by a power approximation of the test results within the range of the test conditions, i.e. for chloride ions up to 2.5% by mass of binder.

HC Csol=1.56・Cfree0.64 (R2=0.996)

NC Csol=2.07・Cfree0.55 (R2=0.983)

NGS Csol=1.70・Cfree0.65 (R2=0.954)

NGL Csol=1.37・Cfree0.70 (R2=0.860)

LC Csol=1.67・Cfree0.59 (R2=0.972)

(6)

where Cfree = quantity of chloride ions [% by mass of binder] and Csol = quantity of soluble chlorides [% by mass of binder]. From a physical viewpoint, the power function means that the quantity of soluble chlorides decreases as the quantity of chloride ions increases. To be more specific, the quantity of partially adsorbed and solid-phase constituents extracted from test specimens using warm water decreases relatively and the quantity of chloride ions and the quantity of soluble chlorides gradually equalizes in the region where the concentration of chloride ions is high. (2) Effects of quantity of gypsum Figure 7 shows the relationship between the quantities of chloride ions and bound chlorides in test series NC, NGS and NGL. The test results in the figure include those obtained by both the pore liquid extraction and JCI methods. The quantity of soluble chlorides determined by the JCI method is converted into the quantity of chloride ions using Equation (6). Calculation results exceeding the applicable range of Equation (6), i.e. chloride ions exceeding 2.5% by mass of binder, are removed from the plot to avoid loss of reliability. The quantity of bound chlorides is determined by deducting the quantity of chloride ions from the measured total quantity of chlorides. As can be seen from the figure, the quantity of bound chlorides shifts greatly with the addi-

Chloride ions [% by mass of binder]0.0 0.5 1.0 1.5 2.0 2.5

0.0

1.0

2.0

3.0

4.0

5.0HC (high early-strength Portland cement)

NC (ordinary Portland cement)

LC (low-heat Portland cement)

Soluble chlorides [% by mass of binder]

(a)

0.0 0.5 1.0 1.5 2.0 2.50.0

1.0

2.0

3.0

4.0

5.0

NGL (with gypsum added)

NC (ordinary Portland cement)NGS (with gypsum added)

Chloride ions [% by mass of binder]


(b)

Fig. 6 Relationship between quantities of chloride ions and soluble chlorides in mortar using Portland cement without admixtures.

0.0 1.0 2.0 3.0 4.00.0

1.0

2.0

3.0

4.0

Increase in quantity of gypsum added

NC

NGS

NGL


Bound chlorides [% by mass of binder]

Fig. 7 Effects of quantity of added gypsum on chloride binding.


tion of gypsum and chloride binding capacity markedly decreases as more gypsum is added. To study the mechanism of this behavior in detail, the authors at-tempted to separate the bound chlorides into solid-phase chlorides and adsorbed chlorides.

The study was carried out using the case of test specimens immersed in NaCl solution for 14 days. The study results obtained by the XRD method are shown in Fig. 8. Only the test results at diffraction angles 2θ = 10-12° are extracted and plotted in the figure. As is clear from test series NC and NGS, the quantity of Friedel's salt formed tended to increase as the concentration of chloride ions increased. In addition, the diffraction peak intensity corresponding to Friedel's salt decreased as the quantity of gypsum increased. No Friedel's salt was ob-served in test series NGL, in which the greatest quantity of gypsum was added. Figure 9 shows the study results obtained by the XRD method (at 2θ = 8.5-10.5°) imme-diately before the immersion of test specimens in the NaCl solution. As is apparent from the figure, the addi-tion of a large quantity of gypsum in test series NGL increased the formation of ettringite (the peak at 2θ = 9.1°) but did not cause a shift to monosulfate (the peak at 2θ = 9.9°). This is why the formation of Friedel’s salt was not observed in test series NGL. A similar tendency has been reported in hardened cement pastes containing gypsum mixed with ground granulated blast furnace slag (Kato et al. 2003).

A diffraction peak at 2θ = 10.6° distinct from the dif-fraction peak corresponding to Friedel’s salt was ob-served in test series NC and NGS at low NaCl concen-trations (Fig. 8). The same tendency was observed in the test series described later where ground granulated blast

furnace slag was used (Fig. 15). This substance identified at low NaCl concentrations is likely to be Kuzel's salt (C3A•(0.5CaSO4•0.5CaCl2)•10H2O), as pointed out in past research (Glasser et al. 1999). However, this dif-fraction peak is not included with the solid-phase chlo-rides in this study because no method of determining the quantity of Kuzel's salt by the XRD method has yet been established (Glasser et al. 1999).

Figure 10 shows the relationships between the quan-tities of chloride ions, solid-phase chlorides and adsorbed

NC, immersed in 1% NaCl solution

2θ [degree]10 11 12

Friedel’s salt

Diffraction intensity

10 11 1210 11 12

Friedel’s salt



NGS, immersed in 1% NaCl solution



NGL, immersed in 1% NaCl solution



Fig. 8 XRD peak (after immersion).

8.5 9.0 9.5 10.0 10.5

Ettringite

2θ [degree]

Monosulfate

Test series NGS


Test series NGL

Fig. 9 XRD peak (before immersion).


chlorides for respective test series. The quantity of ad-sorbed chlorides is determined by deducting the quantity of solid-phase chlorides obtained by the XRD method from the total quantity of bound chlorides. As described earlier, the solid-phase chlorides referred to in the paper indicate only Friedel's salt (C3A•CaCl2•10H2O). Ac-cordingly, compounds that bind chlorides in a different form, such as Kuzel's salt, may be included in the ad-sorbed chlorides.

The quantity of solid-phase chlorides varies greatly depending on the test series and decreases with the amount of gypsum added. On the other hand, no clear difference is observed in the quantity of adsorbed chlo-rides among the various test series. Adsorption behavior was expected to vary with the specific surface area or charge state of the pore walls, but from a macroscopic viewpoint, no particular changes in adsorption behavior were observed with the addition of gypsum. As noted above, the difference in the chloride binding capacity among the test series NC, NGS and NGL is chiefly de-pendent on the quantity of solid-phase chlorides.

(3) Effects of mineral composition Next, the effects of mineral composition are discussed. Figure 11 shows the equilibrium between chloride ions and bound chlorides for test series NC (ordinary mortar), HC (high early-strength mortar) and LC (low-heat mor-tar). Quantities are calculated in the same way as de-scribed in the previous subsection. From the viewpoint of chloride binding capacity, the ranking of mortar types from lowest to highest is NC, HC and LC. This is thought to reflect the Al2O3 content in the cement (Table 1).

As in the previous subsection, bound chlorides were separated into solid-phase chlorides and adsorbed chlo-rides to study in detail the contributions to chloride binding of solid-phase and adsorbed chlorides and the results of the study are shown in Fig. 12. As can be seen from the figure, the quantity of solid-phase chlorides varies greatly with difference in mineral composition. In particular, the quantity of Friedel's salt formed is small in the test series made with low-heat cement. Mineral compositions estimated by Bogue’s formula based on the chemical compositions in Table 1 are shown in Table 4. The estimated proportions of calcium aluminate, C3A, in ordinary cement and low-heat cement are 10.0% and 3.1%, respectively. The quantities of solid-phase chlo-rides in the test series made with ordinary and low-heat cement were expected to reflect the proportion of C3A, but the discrepancy from actual measurements is larger than originally thought.

The chemical proportion of C3A, based not on absolute quantity but in the form of the mole ratio of gypsum to C3A (CaSO4/C3A), is also listed in Table 4. The esti-mated mole ratio for test series LC made with low-heat cement is 2.61, which is nearly the same as that for the test series NGL (2.68) in which a large quantity of gyp-sum is added to ordinary cement. More specifically, the quantities of gypsum contained in both test series are

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0


Solid-phase chlorides (Friedel’s salt) [% by mass of binder]

NC

NGS

NGL

(a)

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0


Adsorbed chlorides [% by mass of binder]

NC

NGS

NGL

(b)

Fig. 10 Relationships between quantities of solid-phase chlorides and chloride ions (effects of addition of gypsum).

0.0 1.0 2.0 3.0 4.00.0

1.0

2.0

3.0

4.0

HC

NC

LC



Fig. 11 Effects of differences in mineral composition on chloride binding.


relatively high in relation to C3A and it is thought that, as a result, the quantity of Friedel's salt formed from

monosulfate is lower. This hypothesis is considered ap-plicable not only to test series LC and NGL but also to all other test series. In other words, the quantity of Friedel's salt formed correlates with the mole ratio of CaSO4/C3A.

On the other hand, no noticeable difference was ob-served in the quantity of adsorbed chlorides with dif-ferent mineral compositions. This is similar to the result discussed in the previous subsection. Although adsorp-tion behavior was expected to vary with the pore struc-ture of the hardened cement paste and the C/S ratio, it was clear that no particular change in the quantity of adsorbed chlorides took place within the range of test conditions where only Portland cement was used as the binder.

4. Chloride binding capacity of mortar including mineral admixtures

(1) Effects of ground granulated blast furnace slag as cement replacement In line with the method described in the previous section, the first step was to experimentally obtain the relation-ships between quantities of chloride ions and soluble chlorides for the mortars using ground granulated blast furnace slag (Fig. 13). The relationships are formulated as follows.

B20 Csol=2.23・Cfree0.53 (R2=0.977)

B40 Csol=2.17・Cfree0.55 (R2=0.946)

B60 Csol=1.86・Cfree0.56 (R2=0.971)

B80 Csol=1.60・Cfree0.65 (R2=0.985)

(7)

Figure 14 shows the relationship between chloride ion and bound chloride quantities for test series B using ground granulated blast furnace slag (with a cement replacement ratio of 0-80%). As shown in the figure, test specimens with low replacement ratios, such as B20 and B40, have a relatively high chloride binding capacity, whereas the chloride binding capacity of the test speci-

Table 4 Mineral compositions estimated by Bogue’s formula.

Mineral composition estimated by Bogue’s formula (%) CaSO4

*1

Converted quantity of

total CaSO4

*2

CaSO4/C3A Mole ratio Test

series Type of cement

C3S C2S C3A C4AF NC OPC 53.5 21.3 10.0 8.9 3.0 3.0 0.59 HC HPC 70.7 3.9 8.6 8.5 5.0 5.0 1.16 LC LPC 23.1 56.4 3.1 10.7 4.1 4.1 2.61

NGS OPC 51.0 23.0 9.5 8.5 2.8 6.6 1.58 NGL OPC 48.4 19.2 9.1 8.1 2.7 10.2 2.68

*1 SO3 in cement is counted as part of the total quantity of CaSO4 (%). *2 Total quantity of CaSO4 in the binder, including the quantity of CaSO4 added, as a percentage by mass of aggregate

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0

HC

NC

LC


Solid-phase chlorides (Friedel’s salt) [% by mass of binder]

(a)

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0

HC

NC

LC


Adsorbed chlorides [% by mass of binder]

(b)

Fig. 12 Relationships between quantities of chloride ions and solid-phase chlorides and adsorbed chlorides (effects of mineral composition).


men with a 60% replacement ratio is comparable to that of the test series NC without replacement. The test specimen with an 80% replacement ratio is lower than that of all test specimens using Portland cement without mineral admixtures. A similar tendency in the relation-ship between the replacement ratio of ground granulated blast furnace slag and chloride binding capacity has been reported in past research (Kato et al 2003). To study the mechanism of this behavior in detail, bound chlorides were separated into solid-phase chlorides and adsorbed chlorides as in the previous chapter.

The study was carried out using the case where test specimens were immersed in NaCl solution for 14 days. Before the immersion of test specimens in the NaCl

solution, only monosulfate was found since the used BFS did not contain any gypsum. The results for immersed cases obtained by the XRD method are shown in Fig. 15. Results at a diffraction angle 2θ = 10-12° only are ex-tracted. The common tendency among this test series is that the quantity of Friedel's salt formed increases with increasing salt concentration in the solution, but the diffraction peak of Friedel's salt decreases as the re-placement ratio exceeds 40%. To quantitatively analyze these phenomena, the relationships between the quanti-ties of chloride ions and solid-phase and adsorbed chlo-rides are plotted (Fig. 16). As this figure makes clear, the quantities of solid-phase chlorides in test series B20 and B40 are markedly higher. The inferred reason for this that,

0.0 0.5 1.0 1.5 2.0 2.50.0

1.0

2.0

3.0

4.0

B20

B40

B60

B80




Chloride ions [% by mass of binder]0.0 1.0 2.0 3.0 4.0

0.0

1.0

2.0

3.0

4.0

NC

B20

B40

B60

B80

Fig. 13 Relationship between the quantities of chloride ions Fig. 14 Chloride binding capacity of mortar with varying ratio and soluble chlorides (in case where ground granulated of ground granulated blast furnace replacement. blast furnace slag is used).

2θ [degree]

10 11 12 10 11 12 10 11 12 10 11 12 10 11 12

Immersed in 1% NaCl

solution

NC B80B60B40B20

Friedel’s salt


Friedel’ssalt Friedel’s salt Friedel’s salt

Friedel’ssalt Immersed

in 3% NaClsolution

Immersed in 10% NaCl

solution

Fig. 15 XRD peak (test series using ground granulated blast furnace slag).


although the quantity of Al2O3 originating from Portland cement decreases as the replacement ratio is increased, the quantity of Friedel's salt formed increases through monosulfate formed from the reaction between ground granulated blast furnace slag and gypsum or the reaction between ground granulated blast furnace slag itself and chloride ions (Wanibushi et al. 1999). The other phe-nomenon is that the quantity of solid-phase chlorides decreases at replacement ratios of 60% or more. The detailed mechanism behind this is unknown, but possible factors are a change in the chemical composition of the pore liquid and a change in the calcium aluminate hy-drates as the replacement ratio increases.

Next, attention is focused on the quantity of adsorbed chlorides. For test series with high replacement ratios, such as B60 and B80, the quantity of adsorbed chlorides

was markedly higher. Although a simple comparison is not possible because the compositions of the formed C-S-H (particularly the C/S ratio) vary with the re-placement ratio, a possible explanation is that the use of ground granulated blast furnace slag increases specific surface area of the pore structure, resulting in a greater adsorbable area.

(2) Effects of pozzolans as a cement replace-ment Figure 17 shows the relationship between the quantities of chloride ions and soluble chlorides in the test series using fly ash. Regression curves for the relationships are given by the following equations:

)985.0( 84.1 F40

)999.0( 95.1 F20252.0

246.0

=⋅=

=⋅=

RCC

RCC

freesol

freesol (8)

It has been well known that characteristics of fly ash may vary considerably depending on the source and burning process of coal, which may lead to different chlorides behavior in micro-pore structures. Thus, the authors understand that the above equation (8) may be applied to similar types of fly ash used in this test, and further study is necessary in order to derive a generalized equation.

The relationship between the quantities of chloride ions and bound chlorides is shown in Fig. 18. As com-pared with the test series using cement without admix-tures, the quantity of bound chlorides is slightly lower in the high concentration region.

The bound chlorides were separated into solid-phase chlorides and adsorbed chlorides. The resulting rela-tionships between quantities of chloride ions and solid-phase chlorides and adsorbed chlorides are shown in Fig. 19. It has to be noted here that, in some cases, the quantity of solid phase chlorides obtained by the XRD

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0


Solid-phase chloride (Friedel’s salt) [% by mass of binder]

NC

B20

B40

B60

B80

(a)

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0Adsorbed chlorides [% by mass of binder]


NC

B20

B40

B60

B80

(b)

Fig. 16 Relationships between quantities of chloride ions and solid-phase and adsorbed chlorides (effects of re-placement of cement with ground granulated blast fur-nace slag).

0.0 0.5 1.0 1.5 2.0 2.50.0

1.0

2.0

3.0

4.0

F20

F40



Fig. 17 Relationship between quantities of chloride ions and soluble chlorides (in case where fly ash is used).


method exceeded the total quantity of bound chlorides, although this trend was found only in this test series. As already explained, since the quantity of adsorbed chlo-rides is determined by deducting the quantity of solid-phase chlorides from the total quantity of bound chlorides, the quantity of adsorbed chlorides takes a negative value in such a case. In Fig. 19(b), obtained negative values are not plotted, and this is also a reason why some of the solid-phase chlorides shown in Fig. 19(a) become larger than corresponding quantity of bound chlorides.

The quantity of solid-phase chlorides tends to be higher than in test series NC at a replacement ratio of 40%. This indicates that Al2O3 in the fly ash contributes greatly to the formation of Friedel's salt. Focusing on the quantity of adsorbed chlorides, the adsorption capacity of test series F20 is comparable to that of NC but is sub-stantially lower at a replacement ratio of 40%. The rea-son for this decrease is that, although the pore structures become finer and the specific surface area increases with the inclusion of fly ash, the composition of the formed C-S-H, particularly the C/S ratio, decreases substantially, resulting in a decrease in adsorption capacity (Beaudoin et al. 1990).

Next, the mortars with part of the cement replaced with silica fume are studied. The relationships between quantities of chloride ions and soluble chlorides are shown in Fig. 20 and formulated as given by Equation (9).

S20 Csol=1.78・Cfree0.61 (R2=0.958)

S40 Csol=1.04・Cfree1.18 (R2=0.943)

(9)

Equation (9) is applicable up to a concentration of 1.5% of chloride ions (by mass of binder) because the quantity of chloride ions permeating the interior is low.

Figure 21 shows the equilibrium between chloride ions and bound chlorides when part of the cement is replaced with silica fume. Commonly, the silica fume replacement ratio is 5%-15% of the cement and the wa-

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0


NC

F20

F40


(a)

0.0 1.0 2.0 3.00.0

1.0

2.0



NC

F20

F40

(b)

Fig. 19 Relationships between chloride ions and solid-phase chlorides and adsorbed chlorides (effects of replacement of cement with fly ash).

0.0 0.5 1.0 1.50.0

1.0

2.0

3.0

4.0S20

S40



Fig. 20 Relationship between quantities of chloride ions and soluble chlorides (in case where silica fume is used).

0.0 1.0 2.0 3.0 4.00.0

1.0

2.0

3.0

4.0

C hloride ions [% by m ass of b inder]

Bound chlorides [% by m ass of b inder]

N C

F20

F40

Fig. 18 Chloride binding capacity of mortar using fly ash.


ter-to-binder ratio is often kept low for the purpose of increasing strength and durability. In contrast to this standard practice, a high water-to-binder ratio and inten-tionally higher replacement ratio was used in the ex-perimental mix proportions here because it proved very difficult to extract pore liquid from test specimens with a low water-to-binder ratio. This should not affect the most important objective of this research, which is to make clear the role of silica fume as a pozzolan in chloride binding. As is clear from Fig. 21, the quantity of bound chlorides falls as the replacement ratio is increased. As before, the bound chlorides were separated into solid-phase chlorides and adsorbed chlorides. The rela-tionships between quantities of chloride ions and solid-phase chlorides and adsorbed chlorides are shown in Fig. 22. The detected amount of solid-phase chlorides was infinitesimal. The reason for this is considered to be that silica fume itself contains little Al2O3 and the poz-zolanic reaction substantially reduces the concentration of OH- ions; as a result, Friedel's salt disappears. The quantity of adsorbed chlorides in test series NC is similar in tendency to that of S40, while the in test series S20 it is higher. From the viewpoint of chloride binding capacity, the observed macroscopic change in the quantity of ad-sorbed chlorides is thought to occur as a result of a combination of the positive effect of the finer micro-pore structure (an increase in surface area) and the negative effect of a decrease in the C/S ratio of hydrates caused by the addition of silica fume.

As discussed in the previous section, changes in the mineral composition and quantity of gypsum added do not greatly affect the quantity of adsorbed chlorides in mortar using Portland cement without mineral admix-tures. The primary factor that controls chloride binding capacity in such cases is the quantity of Friedel's salt formed. For mortar including mineral admixtures, the

quantity of Friedel's salt formed is dependent on the chemical composition of a binder and, further, the prop-erties of the formed hydrates vary, which leads to a large difference in the macroscopic quantity of adsorbed chlorides. This is one of the characteristic phenomena associated with the use of mineral admixtures.

5. Conclusions

The authors prepared test specimens using Portland ce-ment with various chemical compositions and also with some cement replaced with various mineral admixtures. The total quantity of chlorides in the test specimens was separated into chloride ions, solid-phase chlorides and adsorbed chlorides to clarify and quantify the mechanism

0.0 1.0 2.0 3.0 4.00.0

1.0

2.0

3.0

4.0



NC

S20

S40

Fig. 21 Chloride binding capacity of mortar using silica fume.

0.0 1.0 2.0 3.00.0

1.0

2.0

3.0NC

S20

S40



(a)

0.0 1.0 2.0 3.00.0

1.0

2.0



NC

S20

S40

(b)

Fig. 22 Relationships between quantities of chloride ions and solid-phase chlorides and adsorbed chlorides (effects of replacement of cement with silica fume).


of chloride binding in the hardened cement pastes. The following conclusions are made clear from this research: (1) The quantity of bound chlorides varies greatly with

the quantity of gypsum added. In particular, when gypsum is added in large quantities, the transfor-mation from ettringite to monosulfate does not occur and it is observed that Friedel's salt formation ceases completely. Further, when the mineral composition is varied, the quantity of Friedel's salt formed depends on the Al2O3 content of the cement. These results can be explained quantitatively in terms of the mole ratio of calcium aluminate to gypsum in the binder.

(2) No noticeable difference in the quantity of adsorbed chlorides is observed among test specimens made using Portland cement without mineral admixtures within the range of test conditions. Adsorption be-havior was expected to vary with the specific sur-face area and surface charge of the pore walls as the available area for adsorption, but no particular dif-ference was seen from the macroscopic viewpoint of chloride binding even when the mineral compo-sition and quantity of gypsum were varied.

(3) For the mortar using mineral admixtures, the quan-tities of solid-phase chlorides and adsorbed chlo-rides vary greatly depending on the type of mineral admixture and the admixture replacement ratio. When cement is replaced with ground granulated blast furnace slag at replacement ratios up to 40% and when fly ash is used, the quantity of solid-phase chlorides rises as the quantity of the admixture is increased. The Al2O3 content of ground granulated blast furnace slag and fly ash is considered as con-tributing to the formation of Friedel's salt. When the replacement ratio of ground granulated blast fur-nace slag is increased up to 60% and 80%, however, the quantity of solid-phase chlorides begins to de-crease. Moreover, when silica fume is used, the formation of solid-phase chlorides is hardly ob-served at all.

(4) A macroscopic change in the quantity of adsorbed chlorides is brought about by the combined effect of differences in the specific surface area of the pore structure and the properties (in particular the CaO/SiO2 ratio) of the hydrates formed when min-eral admixture is added.

(5) A method was proposed for determining the quan-tity of Friedel's salt present in mortar test specimens using the XRD method. Further, the authors dem-onstrated that it is possible to determine the con-centration of total chloride ions in the pore liquid and to separate bound chlorides into solid-phase chlorides and adsorbed chlorides and thereby de-termine their respective quantities. The concentra-tion of chloride ions was measured using a pore liquid extractor, the volume of pore liquid using mercury intrusion porosimetry and the quantity of total soluble chlorides by the JCI method.

References Barneyback, R. S. and Diamond, S. (1981). “Expression

and analysis of pore fluids from hardened cement pastes and mortars.” Cement and Concrete Research, 11, 279-285.

Beaudoin, J. J., Ramachandran, V. S. and Feldman R. F. (1990). “Interaction of chloride and C-S-H.” Cement and Concrete Research, 20, 875-883.

Glass, G. K., Hassanein, N. M. and Buenfeld, N. R. (1997). “Neural network modeling of chloride binding.” Magazine of Concrete Research, 49(181), 323-335.

Glasser, F. P., Kindness, A. and Stronach, S. A. (1999). “Stability and solubility relationships in AFm phases Part I. Chloride, sulfate and hydroxide.” Cement and Concrete Research, 29, 861-866.

Inoue, H., Sakai, E. and Daimon, M. (2002). “Analysis of hydration of calcium filler cement.” Cement Science and Concrete Technology, 56, 42-49. (in Japanese)

Japan Cement Association (2001). “Hardened cement paste research committee report.” 292-301. (in Japanese)

The Japan Society for Analytical Chemistry, Study Group on X-ray Analyses (2005). “Practical side of powder X-ray analysis—Introduction to Rietveld analysis.” 54-58. (in Japanese)

JSCE, (2002). “Standard specifications for concrete structures -Materials and construction.” Tokyo: Japan Society of Civil Engineers. (in Japanese)

JSCE, (2003). “Establishment of method of testing chloride ion diffusion coefficient in concrete and trends of test methods to be standardized.” Concrete Engineering Series, (55), Japan Society of Civil Engineers. (in Japanese)

Kato, H., Miyagawa, T., Nakamura, A. and Doi, H. (2003). “Behaviors of chloride ions and effects of gypsum on corrosion of steel reinforcement in mortar using ground granulated blast furnace slag.” Jornal of Materials, Concrete Structures and Pavement, V-61(746), 1-12. (in Japanese)

Koibuchi, K., Nakaumura, A., Sakai, E., Osawa, S. and Daimon, M. (1999). “Effects of potassium chloride on hydration of ground granulated blast furnace slag.” Journal of the Society of Inorganic Materials, 6, 207-212. (in Japanese)

Maruya, T., Tangtermsirikul, S. and Matsuoka, Y. (1998). “Modeling of movement of chloride ions in concrete surface layer.” Jornal of Materials, Concrete Structures and Pavement, V-38(585), 79-95. (in Japanese)

Sakai, E., Sakai, M., Asaga, K. and Daimon, M. (1998). “Phase composition model of cement hydration.” Proceedings of the Japan Concrete Institute, 20(1), 101-106. (in Japanese)

Someya, K., Daisoku, N., Wee, T., and Nagataki, S. (1989). “Characteristics of binding of chloride ions in hardened cement pastes.” Proceedings of the Japan Concrete Institute, 11(1), 603-608. (in Japanese)

chloride binding capacity of mortars made with various

Documents