New Insights into the NT Build 492 Rapid Chloride Migration Test

Transcription

1 New Insights into the NT Build 492 Rapid Chloride Migration Test P. Spiesz 1, * M. M. Ballari 2 H. J. H. Brouwers 3 ABSTRACT In this study a new theoretical model is developed for the popular Rapid Chloride Migration test (NT Build 492). This model can in a proper way predict the transport of chlorides due to application of the electrical field. In order to improve the traditional chloride transport model, the non-linear inding of chlorides and non-equilirium etween free- and ound- chlorides are introduced. The new system of equations is solved numerically and applied to experimental data otained elsewhere. From these simulations several parameters are otained, namely: the effective chloride diffusion coefficient, the non-linear inding constants and the chloride mass transfer coefficient. This opens the new possiility of otaining the effective chloride diffusion coefficient (usually derived from long-term, laorious techniques) from the short-term migration tests. The otained inding constants are in line with experimental data found in literature. The chloride mass transfer coefficient shows a tendency to decrease in time during the application of electrical field, which can give evidence of changes in the microstructure and/or in the pore solution of concrete. KEYWORDS Concrete; Chloride; Diffusion; Accelerated test. * Corresponding author 1 Department of Architecture Building and Planning, Eindhoven University of Technology, The Netherlands, p.spiesz@tue.nl 2 Department of Architecture, Building and Planning, Eindhoven University of Technology, The Netherlands, j.rouwers@tue.nl 3 Department of Architecture, Building and Planning, Eindhoven University of Technology, The Netherlands, m.allari@tue.nl

2 P. Spiesz, M.M. Ballari and H.J.H. Brouwers 1. INTRODUCTION Deterioration of concrete elements and structures eing in contact with chloride-earing solutions (e.g. seawater or de-icing salts) in majority of cases is caused y the chloride-initiated corrosion of steel rears reinforcing concrete. The rears are passivated as long as the chloride concentration at their level does not reach the critical threshold value. Once the critical concentration of chloride ions is reached, the corrosion process is triggered and leads in time to the deterioration of concrete. The cover layer (the ulk of concrete etween its surface and the reinforcing steel) is the arrier against the chloride source. Nevertheless, in concrete, eing a porous material, the transport of chlorides takes place due to capillary suction and diffusion processes. Thus, the critical chloride concentration at the level of steel reinforcement will e reached, and this is only a matter of time. However, this time, eing related to the service lifetime of concrete, is dependent on the quality (rate of chloride diffusion) and quantity (distance that chlorides have to penetrate) of concrete. Hence, the chloride ingress speed (usually expressed as the chloride diffusion coefficient in concrete) and the thickness of the cover layer of concrete are usually employed in service life design models of concrete structures. In order to determine the diffusion coefficient, a numer of laoratory measurement techniques have een developed over years, among which the NT Build Rapid Chloride Migration test (also termed as the CTH or RMT test) is nowadays ecoming more and more popular. The importance of this technique increased when its output value was included in the European DuraCrete Project for the service life design of concrete. Although the test is commonly accepted and performed, this paper addresses some issues to its theoretical model. In this model, the inding of chlorides y cement hydrates is assumed to e linear and in instantaneous equilirium, however, it is known that the inding is non-linear and certain duration of the exposure to free- chlorides is necessary in order to reach the equilirium. Hence, it is demonstrated that the currently adopted chloride transport model is oversimplified, which explains the origin of ig differences etween the theoretical and measured chloride concentration profiles in concrete after application of the electrical field. This difference shows that the output value of the migration test is incorrect, thus should not e used in service life models. As presented in this paper, the transport model for the chloride migration in concrete is modified y including the non-linear chloride inding isotherm and non-equilirium conditions etween free- and ound- chlorides. As the result of implementation of this new theoretical model, the chloride concentration profiles in concrete can e predicted in a good way. From application of the model to the experimental data, several parameters can e otained (e.g. effective diffusion coefficient of chlorides, inding parameters). 2. RAPID CHLORIDE MIGRATION TEST (NT BUILD 492) The Rapid Chloride Migration (RCM) test developed y Tang [1996], is a non-steady-state test, ased on the ionic migration induced y an external electrical voltage applied across a concrete specimen saturated in Ca(OH) 2 -saturated water solution. Due to the potential difference applied etween the electrodes, chloride ions move from the upstream solution, through the concrete specimen, towards the downstream solution, for a certain time. Afterwards, the specimen is split open and sprayed with AgNO 3 a colorimetric indicator for chlorides. Finally, the chloride penetration depth is measured and ased on its value the non-steady-state chloride diffusion coefficient (D RCM ) is calculated from a model ased on the Nernst-Planck equation. This model reads as follows [Tang 1996]: 2 2 c J0 D 0 c zfe c c zfe c D 2 RCM 2 t x c (1) 1 x RT x x RT x c where: c concentration of chlorides in the pore solution, t time, J 0 - total flux of chlorides through a unit area of solution, x distance, D 0 diffusivity of chlorides in pore solution of concrete, c concentration of chlorides ound in concrete, z ion valence, F Faraday constant, E electrical

3 New Insights into the NT Build 492 Rapid Chloride Migration Test field, R universal gas constant, T temperature, D RCM non-steady-state chloride diffusion coefficient. Eq. (1) is further simplified and solved for D RCM [Tang 1996]: D RCM RT xd xd (2) zfe t RCM where: x d chloride penetration depth indicated y a colorimetric indicator, α laoratory constant and t RCM duration of the test. 3. A NEW MODEL OF CHLORIDE TRANSPORT DUE TO MIGRATION IN CONCRETE 3.1 Verification of the Traditional Model In Fig. 1 the theoretical chloride concentration profile, representing the solution of Eq. (1) is shown. On the other hand, an example of the experimental chloride concentration profile from Stanish [2002] is given in Fig. 2. Similar experimental profiles for chloride migration tests can e also found e.g. in Yuan [2009], Castellote et al. [1999] or Gruyaert et al. [2009]. Figure 1. Theoretical chloride concentration profile (Eq. 1) [Tang 1996]. Figure 2. Experimental chloride concentration profile [Stanish 2002]. As can e deduced from Figs. 1 and 2, the chloride transport model, governed y Eq. (1), cannot predict in the proper way the ehaviour of chlorides migrating into concrete due to the electrical field. Thus, the value of D RCM calculated from Eq. (2) is incorrect. 3.2 Non-linear Chloride Binding in Non-Equilirium It is worth to emphasize that the chloride inding, introduced in the constant term of c / c (see Eq. 1), is assumed to e linear (c = β c, where β is a constant). Nevertheless, as presented e.g. in Tang [1996] and Ziara [2001], the amount of ound chlorides increases non-linearly with an increase of the free-chlorides amount. The linear chloride inding isotherm does not predict in a proper way the relation etween ound and free-chlorides and it can e applicale only within a limited range of freechlorides concentration [Tang 1996]. Tang [1996] shows experimentally that the chloride inding in concrete oeys the Freundlich isotherm in free-chloride concentrations in the range of [mol/dm 3 ] and the Langmuir isotherm at low Cl - concentrations (< 0.05 [mol/dm 3 ]). The Freundlich isotherm is given as follows: C n K c (3) where: C concentration of ound chlorides, K chloride inding capacity of concrete and n inding intensity parameter.

4 P. Spiesz, M.M. Ballari and H.J.H. Brouwers The chloride inding data presented in Ziara [2001] shows that even for higher chloride concentration (up to 3 [mol/dm 3 ]), the Freundlich equation descries the inding correctly. During the RCM test, the concentration of chlorides in the ulk solution yields 2 [mol/dm 3 ], thus the Freundlich isotherm can adequately represent the chloride inding for this test. Considering the fact that the inding of chloride is non-linear, the assumption of constant c / c in Eq. (1) is incorrect, thus the value of D RCM calculated from Eq. (2) should e treated with scepticism. When chloride ions are transferred from one phase (liquid) to another (solid) across an interface that separates the two, the resistance to mass transfer causes a concentration gradient in each phase. Due to these limitations in the mass transfer, usually a certain time is required in order to achieve the equilirium etween the concentrations in liquid and solid. From literature it is known that the equilirium for chloride sorption in concrete can e achieved after 7 days [Tang 1996] or days [Theissing et al. 1978] of exposure. For long-term diffusion tests, the assumption of equilirium is acceptale since the chloride exposure period is sufficiently long. However, the duration of the RCM test usually amounts to 24 hours, and only sometimes varies from 6 hours (for poor quality concrete) up to 4 days (for very good quality concrete). Thus, during the migration process, equilirium etween free- and ound- chlorides cannot e achieved, which implies a necessity of the application of the Cl - mass transfer rate. The mass transfer rate of chlorides from the pore solution, through the liquid-solid interface and towards the solid, is considered to e induced y a concentration gradient in these phases. 3.3 New Chloride Transport Model for the RCM Test The main idea of the new model proposed in this paper is ased on the introduction of the non-linear chloride inding isotherm and non-equilirium etween free- and ound- chlorides concentrations in the system ased on the Nernst-Planck equation. The simplified Nernst-Planck equation is commonly used to descrie ionic transport in porous medium due to comined actions of migration and diffusion. However, as presented in Narsilio et al. [2007], when a sufficiently large electrical voltage (U) is applied across the concrete specimen, the flux of ions due to electrical migration dominates over the diffusion flux due to concentration gradients. The fact that during the RCM test U = [V] allows neglecting the flux of chloride ions due to the gradient of concentrations. The inding of chlorides takes place instantaneously at the surface of the active solid, ut there is a resistance to the mass transfer through the liquid-solid interface. This limitation in the mass transfer rate of chlorides is responsile for the non-equilirium conditions in the system and is governed y the mass transfer coefficient k. Therefore, r, eing the mass transfer rate, reads: r k c c s (4) where: c s chloride concentration in liquid at liquid-solid interface. Calculating c s from Eq. (3) and inserting into Eq. (4) gives [Brouwers 1999]: 1 C n r k c (5) K When considering the total volume of concrete (solid state and pores completely saturated with liquid), the reaction term in the total chloride mass alance in concrete yields 0, ecause the rate of the disappearance of chlorides in the liquid phase is equal to the rate of chlorides increment in the solid phase. However, when considering liquid and solid phases separately, there will e non-zero reaction term as shown in Eq. (5). Thus, assuming that the chloride transport due to diffusion is neglected and applying Eq. (5) and the Nernst-Planck equation, the chloride mass alance, for liquid and solid respectively, reads:

5 New Insights into the NT Build 492 Rapid Chloride Migration Test 1 c c C n u k c t x K 1 n C C (1 ) s kc t K (6) (7) where: φ porosity of concrete, ρ s density of the solid state of concrete and u ionic migration velocity, u = DzF/RTE, D diffusion coefficient. During the migration of chlorides a penetration front can e localized. The position of the front ψ(t) is governed y the following equation: ut () t (8) The oundary and initial conditions pertaining to Eqs. (6) and (7) read: c( x 0, t) c C ( x, t) C 0 i (9) where: C i - the initial ound chloride concentration prior to migration test. The D RCM coefficient (see Eq. 1) is the apparent diffusion coefficient, ecause its value is depending on the inding term. As can e seen in Eqs. (6) and (7), the diffusion coefficient D, applied in the new model, is an independent constant, thus its value is related only to the diffusivity of chlorides in free liquid and the pore system in concrete (considering constrictivity and turtuosity). Hence, the D is representing the effective diffusion coefficient in concrete D eff. 3.4 Numerical Solution of the New Chloride Transport Model In order to otain the numerical solution of Eqs. (6) and (7) a forward discretization is performed, and reads as follows: 1/ n ( i 1, j) ( i, j) ( i 1, j 1) ( i 1, j) C c c c c ( i1, j) u k c( i 1, j) (10) t x K 1/ n C( i1, j) C ( i, j) C( i, j) (1 ) s kc( i, j) (11) t K where: Δt interval of time, i position in time, i = 1.. t/ t, Δx interval of distance, j distance in concrete and j = 1.. L / x. Solving Eqs. (10) and (11) for c (i+1,j+1) and C (i+1,j) respectively, gives: c c k c 1/ n x c( i 1, j) c( i, j) C ( i 1, j ) ( i 1, j 1) ( i 1, j) ( i 1, j) u t K C C k c 1/ n t C( i, j) ( ) ( i1, j) ( i, j) i, j s(1 ) K (12) (13)

6 P. Spiesz, M.M. Ballari and H.J.H. Brouwers With the initial and oundary conditions: c C i, j 1 ( ) ( i1, j) c 0 C i (14) 4. APPLICATION OF THE NEW MODEL TO EXPERIMENTAL DATA A wide dataase of total chloride concentration profiles otained for concrete tested using the RCM test is presented in Stanish [2002], therefore it has een used for further investigations presented in this paper. The experimental data of Stanish [2002] is summarized in Tale 1. The measured profiles are representing the total chloride concentration in concrete, which can e represented as follows: c (1 ) s C C t (15) c where: C t total concentration of chlorides in concrete and ρ c total density of concrete. Tale 1. Testing conditions and properties of concrete [Stanish 2002]. Concrete OPC 0.35 c 0 t RCM 6, 9 and 18 [h] T ρ c 2557 [g/dm 3 ] L ρ s 2601 [g/dm 3 ] φ U 60 [V] C i Concrete OPC 0.45 c 0 t RCM 6, 9 and 18 [h] T ρ c 2553 [g/dm 3 ] L ρ s 2606 [g/dm 3 ] φ U 60 [V] C i 70.9 [g/dm 3 ] [K] 0.05 [m] [g/g] 70.9 [g/dm 3 ] [K] 0.05 [m] [g/g] w/c cement water coarse aggregates fine aggregates w/c cement water coarse aggregates fine aggregates [kg/m 3 ] 146 [kg/m 3 ] 1025 [kg/m 3 ] 695 [kg/m 3 ] [kg/m 3 ] 163 [kg/m 3 ] 1025 [kg/m 3 ] 698 [kg/m 3 ] Based on the data shown in Tale 1and the experimental total chloride concentration profiles, the values of k, K, n and D eff are optimized y using Microsoft Excel Solver tool and applying Eqs. (12), (13) and (15). The optimized values of these parameters and the corresponding simulated chloride profiles are shown respectively in Tale 2 and Figs. 4 and 5. Figure 4. Optimized total chloride concentration profiles, Concrete OPC 0.35 (Tale 1). a) 60V, 6h; ) 60V, 9h; c) 60V, 18h.

7 New Insights into the NT Build 492 Rapid Chloride Migration Test Figure 5. Optimized total chloride concentration profiles, Concrete OPC 0.45 (Tale 1), a) 60V, 6h; ) 60V, 9h; c) 60V, 18h. Tale 2. Optimized model parameters. OPC 0.35 OPC 0.45 t RCM k K n D eff t RCM k K n D eff [h] [ 10-6 s -1 ] [ 10-3 dm 3n /g n ] - [ m 2 /s] [h] [ 10-6 s -1 ] [ 10-3 dm 3n /g n ] - [ m 2 /s] When analyzing the inding capacity K, it can e noticed that the values otained from the numerical model are in line with experimental values presented in Tang [1996] and Ziara [2001], taking into consideration that these experiments were performed on cement pastes, thus, in concrete, due to lower paste content, the inding capacity must e respectively lower. The inding capacity K, which determines the upper limit of ound chlorides, is dependent on the w/c ratio, hence is related to the degree of hydration and connectivity to the cement hydration products. Hence, in the case of concrete with a higher w/c ratio the value of K should also e increased, and this trend can indeed e oserved in Tale 2. The optimum value of n, descriing the intensity of inding and the maximal amount of ound chlorides, is found to e in the range of , which is also in line with the experimental data presented in Tang [1996] and Ziara [2001]. The effective chloride diffusion coefficient in concrete - D eff, is retrieved from the simulations. This coefficient, which is independent from the inding of chlorides and refers only to the pore system in concrete, so far has een otained only from time consuming and laorious steady-state tests. These values of D eff presented in Tale 2 are in line with the values presented in literature [Tang 1996], for concrete with similar w/c ratios. In none of the analyzed experimental chloride concentration profiles the maximal total chloride concentration was achieved except for the surface layer of concrete. This confirms validity of the assumption of non-equilirium in the new chloride transport model. In order to account for this phenomenon, k, eing the mass transfer coefficient, was introduced. The values of k are showing a tendency to decrease during the migration process, which can e oserved in Tale 2. One can find in Figs. 4 and 5 that the value of k is playing deciding role on the shape of each profile. When the chloride mass transfer coefficient is larger, equilirium etween c and C can e reached faster, which results in a sharp chloride profile (e.g. see in Fig. 4a). However, when the value of k ecomes smaller, the mass transfer rate of chlorides ecomes the limiting step and the profile is characterized y a more flattened shape such as in Figs. 4c and 5c. The diminishing chloride mass transfer rate in concrete needs further investigations, since it plays an important role in the case of chloride migration in concrete, and has not een studied yet.

8 P. Spiesz, M.M. Ballari and H.J.H. Brouwers 5. CONCLUSIONS As shown is this paper, the currently adopted mathematical model for the Rapid Chloride Migration test is oversimplified, so the theory cannot satisfactorily predict the chloride ehaviour in concrete during the migration process. For this reason, the chloride diffusion coefficient (D RCM ) eing the output value of the test, has a doutful meaning and should not e used for service life models of concrete. In order to improve the migration model of chlorides in concrete, non-linear chloride inding and non-equilirium etween free- and ound- chlorides concentrations are proposed in this paper. The new chloride transport model is solved numerically and applied to experimental chloride concentration profiles. The values of the non-linear inding parameters (K and n) extracted from the new model are in good agreement with experimental data presented in literature. Due to the limitations in chloride mass transfer through the liquid-solid interface, the equilirium etween free- and ound- chlorides concentration cannot e achieved during the relatively short-term migration test. It is shown that the mass transfer coefficient, represented y k, plays a deciding role on the chloride transport process, hence also on the shape of the chloride concentration profile in concrete. It is found that the value of k tends to decrease in time during the migration process, thus the shape of each analyzed chloride profile changes from sharp in the eginning to flatten at the end of the migration test. This phenomenon has to e further analyzed, since no research has een carried out in this field yet. The effective diffusivity of chlorides can e derived y using the new theoretical model. This reveals a new possiility of application of the short-term non-steady-state migration test in order to otain the effective chloride diffusivity (so far eing only otained y laorious and long-term techniques), which is related only to the pore structure in concrete and is not influenced y inding. REFERENCES Castellote, M., Andrade, C., Alonso, C. 1999, Chloride-inding isotherms in concrete sumitted to non-steady-state migration experiments, Cement and Concrete Research, 29, Brouwers, H. J. H. 1999, Transport model for desorption from natural soils packed in flushed columns, Water Resources Research, 35, Gruyaert, E., Van den Heede, Ph., De Beile, N. 2009, Chloride ingress for concrete containing lastfurnace slag, related to microstructural parameters, Proceedings of the 2 nd International RILEM Workshop on Concrete Duraility and Service Life Planning, , Haifa, Israel. Narsilio, G. A., Li, R., Pivonka, P., Smith, D.W. 2007, Comparative study of methods used to estimate ionic diffusion coefficients using migration tests, Cement and Concrete Research, 37, Stanish, K.D. 2002, The migration of chloride ions in concrete, PhD Thesis, University of Toronto, Toronto, Canada. Tang, L. 1996, Chloride transport in concrete measurement and prediction, PhD Thesis, Chalmers University of Technology, Gothenurg, Sweden. Theissing E.M., Hest-Wardenier, P.v., de Wind, G. 1978, The comining of sodium chloride and calcium chloride y a numer of different hardened cement pastes, Cement and Concrete Research, 8, Yuan, Q. 2009, Fundamental studies on test methods for the transport of chloride ions in cementitious materials, PhD Thesis, University of Ghent, Ghent, Belgium. Ziara, H. 2001, Binding of external chlorides y cement pastes, PhD Thesis, University of Toronto, Toronto, Canada.