Concrete Science and Engineering, Vol. 2, December 2000, pp 170-175 MRS CONFERENCE PAPERS Modeling of chloride penetration into concrete Tracing five years field exposure Tang Luping 1,2 and Lars-Olof Nilsson 1 (1) Chalmers University of Technology, Dept. of Building Materials, Gothenburg, Sweden (2) SP Swedish National Testing and Research Institute, Boraas, Sweden ABSTRACT This paper presents the current development of the numerical model ClinConc for predicting chloride penetration into concrete. In the beginning of the 1990 s, as a part of the work in a Swedish national research project, some 40 types of concrete specimens were exposed to sea water at the field station on the west coast of Sweden. The chloride profiles in concrete were measured after exposure for 0.6, 1, 2 and 5 years. These data of chloride profiles together with the transport properties of concrete measured in the laboratory are greatly useful for modeling of chloride penetration into concrete. In the middle of the 1990 s a numerical model for predicting chloride penetration into concrete, called ClinConc, was developed from our previous work. The model is essentially based on the current knowledge of physical and chemical processes involved in the chloride transport and binding in concrete. In this model most of the factors affecting chloride penetration are considered in a relevant and scientific way. The ambition is to predict chloride penetration profiles by using those parameters as input data that can be measured independently without relying upon any curve-fitting procedure. After tracing five years field exposure, it is proved that the model could predict field chloride profiles fairly well, especially in the penetration depth. It has been found, on the other hand, that the model somewhat underestimates the total chloride content in the surface zone. This discrepancy could be solved by simply including a factor describing the time effect of chloride binding. 1. INTRODUCTION Chloride-induced corrosion of reinforcement in concrete is a major problem for the safety and durability of reinforced concrete structures. Chloride penetrates into concrete in different ways, mostly by diffusion through water saturated pores and convection by capillary suction and permeation. Chloride ingress is generally accompanied by chemical and physical binding. The mechanisms behind chloride ingress are complicated and not fully understood. Some comprehensive reviews could be found elsewhere [1, 2]. In the beginning of the 1990 s, as a part of the work in a Swedish national research project, some 40 types of concrete specimens were exposed to sea water at the field station on the west coast of Sweden [3]. The chloride profiles in concrete were measured after exposure for 0.6, 1, 2 and 5 years. These data of chloride profiles together with the transport properties of concrete measured in the laboratory are greatly useful for modeling of chloride penetration into concrete. In the middle of the 1990 s a numerical model for predicting chloride penetration into concrete, called ClinConc, was developed from our previous work [4-6]. The model is essentially based on the current knowledge of physical and chemical processes involved in the chloride transport and binding in concrete. Not like other empirical models, which are basically based on achieving regression parameters from curve-fitting measured chloride profiles, in the model ClinConc most of the factors affecting chloride penetration are considered in a relevant and scientific way. Thus the model could be classified as a scientific one. The ambition is to predict chloride penetration profiles by using those parameters as input data that can be measured independently without relying upon any curvefitting procedure. In this paper, the predicted results are compared with those measured from the field exposures and the differences in the predicted and measured profiles are discussed. Possible approaches to the improvement of the model are given. 2. FIELD EXPOSURES Concrete slabs of size 1000 700 100 mm were cast in steel molds in the laboratory and moist cured. At an age of two weeks the slabs were mounted on the sides ISSN 1295-2826/00 RILEM Publications S.A.R.L. 170
Luping, Nilsson Fig. 2 Schematic of the model ClinConc for prediction of chloride penetration. Fig. 1 Annual temperature in the sea water at Träslövsläge harbour on the west coast of Sweden. of pontoons at Träslövsläge harbor on the west coast of Sweden. The annual chloride concentration in the sea water is in a range of 10 to 18 g/l and the annual temperature in the sea water varies from +2 C to +20 C following a sine curve, as shown in Fig. 1. At specified exposure times, that is, 0.6 or 0.8, 1, 2 and 5 years, cores with diameters of 75 to 100 mm were taken from the concrete slabs at three different zones, i.e. submerged, splashing, and atmospheric zones, and the chloride penetration profile in each core was determined. The detailed sampling and chloride analyzing procedures were described elsewhere [8]. Since the current model is for the submerged zone only, the field data from this zone will be used for comparison. 3. MODEL CLINCONC The detailed description of the model ClinConc can be found in our previous publications [4-6]. The model consists of two main procedures: 1) simulating free chloride transport through the pore solution in concrete by using a chloride diffusion equation with the free chloride concentration as the driving potential, and 2) calculating the distribution of the total chloride content in concrete by considering non-linear chloride binding isotherms including the significant effects of both alkalinity and temperature. It is essentially important that the various dimensions in the calculation should be clearly specified. The model is schematically shown in Fig. 2. It can be seen that in this model the main input data include concrete mix design, workmanship and exposure conditions. The chloride binding data, which have a strong effect on chloride penetration profiles, are stored in a database for numerical calculation. The only parameter that needs to be measured is the chloride diffusion coefficient by using the CTH rapid method, which has recently been accepted as a Nordic standard test method [9]. The effects of temperature, curing age and concrete skin on diffusivity have also been taken into account. The computer program is in progress. The version 2.2b is presently available. 4. COMPARISON WITH THE FIELD DATA The mixture proportions of concrete [3] used for comparison are listed in Table 1 where the chloride diffusivity [8] measured by using the CTH Rapid Test is also given. The field data of chloride profiles were reported elsewhere [8, Table 1 Mixture proportions of concrete used for comparison in this study Concrete Compositions Diffusivity Cement* Silica fume Water Air Aggregate Mix ID w/b kg/m 3 kg/m 3 kg/m 3 % kg/m 3 D CTH -12 m 2 /s #1-40 (OE) 0.40 420 0 168 6 1692 8.1 #1-50 0.50 370 0 185 6.4 1684 19.9 #H4 0.40 399 21** 168 5.9 1685 2.7 #2-40 0.40 420 0 168 6.2 1675 7.1 * Swedish Slite OPC for #2-40 and Swedish Degerhamn SRPC for the others. ** In slurry. 171
Concrete Science and Engineering, Vol. 2, December 2000 Fig. 3 Chloride profiles of Concrete #1-40(OE) after 0.6 to 5 years exposure. Fig. 5 Chloride profiles of Concrete #H4 after 0.6 to 5 years exposure. Fig. 4 Chloride profiles of Concrete #1-50 after 0.8 to 5 years exposure. Fig. 6 Chloride profiles of Concrete #2-40 after 1 to 5 years exposure. 172
Luping, Nilsson 10]. The calculated and measured chloride profiles in four different types of concrete are shown in Figs. 3 to 6. Figs. 3 and 4 show examples of SRPC concrete with different waterbinder ratios, Fig. 5 shows an example of silica fume concrete, and Fig. 6 shows an example of OPC concrete. From these figures it can be seen that, no matter which type of concrete, the predicted chloride profiles correspond very well with the measured profiles up to two years field exposure. For the profiles after five years exposure, it seems as if ClinConc underestimated the chloride contents at the depths of 0-20 mm. However, the predicted penetration depths appear in a good agreement with the measured profiles, except for Concrete #1-40(OE), which will be discussed later. It can be found from the measured profiles that the surface chloride content tends to increase with exposure time. This tendency was not taken into account in the ClinConc. The chloride binding isotherms used in the ClinConc were those obtained in the laboratory after about two weeks equilibrium [11]. On the other hand, the effect of alkalinity on chloride binding was also based on a limited investigation [12]. In reality, the pore solution compositions may change due to leaching and penetration of different substances, resulting in different characteristics of chloride binding. In addition, an increased saturation degree of the air voids near the surface might also be a reason for the increase in the chloride content at the vicinity of the surface. The concrete mixtures used in this comparison were all air entrained with a volume of about 6%. The saturation degree of the air voids will increase after such a long period of immersion, especially in contact with a salt solution. Nevertheless, changes in binding characteristics and saturation degree might be the reasons why the predicted chloride contents in the surface zone differ from the measured ones. These changes in the surface zone may not necessarily influence the transport properties in the deep zone of concrete where the pore solution compositions may not significantly vary. This can explain why the predicted penetration depths are in a good agreement with the measured ones. The chloride profile in Concrete #1-40(OE) is an exception which cannot be explained in the above ways. Further measurement and more observations are needed before any conclusion could be drawn. 5. POSSIBLE APPROACHES TO THE IMPROVEMENT OF THE MODEL The general limitations of the model ClinConc have been discussed elsewhere [13]. From the above comparison it has been found that the chloride contents closer to the surface are somewhat underestimated after five years exposure. Several reasons for these differences are possible. An obvious effect that is not included in the prediction, but in the measurements, is the high binder content close to the exposed surfaces because of the wall effect at the cast surface. A higher cement content than in the predictions would result in a larger amount of bound chloride and consequently more chloride in total. Another possible reason could be a time effect on chloride binding, giving a larger amount of bound chloride with time. Such an effect is indicated by some data [14], but not yet fully proved. Some field data indicates a time-dependent surface chloride content. This is, however, in the splash zone where drying and wetting eventually rise the chloride content. A time-dependent chloride binding in constantly submerged concrete is something quite different. The cement distribution profile could be estimated by determining calcium content parallel to the determination of chloride content if the aggregate does not contain acid soluble calcium. This has, in fact, been done in our study and the measured chloride profiles were reported in mass percent of binder. However, the measurement uncertainty in calcium content is large, especially when concrete is contaminated by chlorides. Better test methods are needed for determining cement profiles. In the above prediction as shown in Figs. 3 to 6, the time effect on chloride binding was not included. Today, the knowledge and quantitative information about the time effect of chloride binding under submerged zone are very limited. Further study is needed to improve the understanding of this time effect. At present, a simple modification could be included in the model, that is, to multiply the present binding function by a factor f t, which can be expressed as: f t = a ln(t Cl + b) + 1 (1) where t Cl is the local chloride contamination time in years, and a and b are time-dependent coefficients. It should be noticed that, in the above equation, the time t Cl does not mean the exposure duration, but means the duration of chloride contamination. No matter how long the exposure time is, at the depth where chlorides have not penetrated yet, the time t Cl is always zero. Attention should be paid to this in computing. According to the measured chloride profiles [10], the values of a and b could be assumed as 0.36 and 0.5, respectively. An example of f t - t Cl curve is illustrated in Fig. 7. By considering this factor in the model ClinConc, much better prediction could be achieved, as shown in Fig. 8. 173
Concrete Science and Engineering, Vol. 2, December 2000 6. CONCLUDING REMARKS Fig. 7 Example of correction factor for time-dependent chloride binding. The model ClinConc for chloride penetration is very promising for predicting actual chloride profiles in submerged concrete structures. The required input is very limited and a short term test method, for instance, the standard non-steady state migration test, can be used to determine it. The predicted results are meant to be used for direct comparisons with measured profiles without involving curve-fitting. Any discrepancies found will immediately indicate where the most important lack of knowledge is to be looked for. After tracing five years field exposure, it is proved that the model could predict field chloride profiles fairly well, especially in the penetration depth. It has been found, on the other hand, that the model somewhat underestimates the total chloride content in the surface zone. This discrepancy could be solved by simply including a factor describing the time effect of chloride binding. Much more research should be directed towards understanding and quantifying these time effects. ACKNOWLEDGEMENT The authors thank the reviewers of this paper for their valuable suggestions. REFERENCES Fig. 8 Predicted chloride profiles of Concrete #2-40 by considering the time-dependent chloride binding in the model ClinConc. Other possible improvements of the model may include: the considerations of concentration dependent chloride diffusivity, diffusion in an unsaturated pore system, and combination of diffusion and convection. More data measured from the well-controlled field exposures are needed for the above improvements. [1] Marchand, J., Gérard, B. and Delagrave, A., Ion Transport Mechanisms in Cement-based Materials, Report GCS-95-07, Dept. of Civil Engg., University of Laval, Québec, Canada, 1995. [2] Nilsson, L.-O., Poulsen, E., Sandberg, P. and Sørensen, H. E., Chloride Penetration into Concrete - State-of-the-art, HETEK Report No. 53, Danish Road Directorate, 1996. [3] Sandberg, P., Systematic collection of field data for service life prediction of concrete structures, in Durability of Concrete in Saline Environment, 7-22, Cementa, Danderyd, Sweden, 1996. [4] Tang, L. and Nilsson, L.-O., A numerical method for prediction of chloride penetration into concrete structures, presented at the NATO/RILEM Workshop, July 10-13, 1994, St. Rémy-lès- Chevreuse, and to be published in The Modeling of Microstructure and its Potential for Studying Transport Properties and Durability, ed. H. Jennings et al. [5] Tang, L., Chloride Transport in Concrete - Measurement and prediction, Doctoral thesis, Publication P-96:6, Dept. of Building Materials, Chalmers University of Technology, Gothenburg, Sweden, 1996. [6] Tang, L. and Nilsson, L.-O., Prediction of chloride penetration into concrete by using the computer program CLINCONC, Proceedings of the 2nd International Conference on Concrete 174
Luping, Nilsson under Severe Conditions, June 21-24, 1998, Tromsø, Norway. [7] Ewertson, C., Climate data from the field station at Träslövsläge harbour. Internal report, Swedish National Testing and Research Institute, Borås, Sweden, 1995. [8] Tang, L., Chloride penetration profiles and diffusivity in concrete under different exposure conditions, Publication P-97:3, Department of Building Materials, Chalmers University of Technology, Gothenburg, Sweden, 1997. [9] Nordtest, Concrete, Mortar and Cement Based Repair Materials: Chloride Migration Coefficient from Non-steady State Migration Experiments, NT BUILD 492, Esbo, Finland, 1999. [10] Andersen, A., Hjelm, S., Janz, M., Johannesson, B., Pettersson, K., Sandberg, P., Sørensen, H., Tang, L. and Woltze, K., Total chloride profiles in uncracked concrete exposed at Träslövsläge marine field station - Raw data from 1992 to 1997, Report TVBM-7126, Division of Building Materials, Lund Institute of Technology, Lund, Sweden, 1998. [11] Tang, L. and Nilsson, L.-O., Chloride binding capacity and binding isotherms of OPC pastes and mortars, Cement and Concrete Research 23 (2) (1993) 347-353. [12] Sandberg, P. and Larsson, J., Chloride binding in cement pastes in equilibrium with synthetic pore solutions as a function of [Cl] and [OH], in Chloride Penetration into Concrete Structures - Nordic Miniseminar, ed. by L.-O. Nilsson, Publication P-93:1, Division of Building Materials, Chalmers University of Technology, 98-107, Gothenburg, Sweden, 1993. [13] Nilsson, L.-O. and Tang, L., Present limitations in scientifically based prediction models for chloride ingress into submerged concrete, Proceedings of the 1st International Meeting on Material Science and Concrete Properties. March 1998, Toulouse, France, 1998. [14] Nilsson, L.-O., Sandberg, P., Poulsen, E., Tang, L., Andersen, A. and Frederiksen, J. M., A system for estimation of chloride ingress into concrete, Theoretical background, HETEK Report No. 83, The Danish Road Directorate, 1997. 175