Documente Academic
Documente Profesional
Documente Cultură
TECHNICAL REPORTS
chloride diffusion/migration coefficients of concrete
L. Tang1, 2 and H. E. Sørensen3
1) SP Swedish National Testing and Research Institute, Borås, Sweden
2) Chalmers University of Technology, Gothenburg, Sweden
3) FORCE Institute, Brøndby, Denmark
A B S T R A C T R É S U M É
This paper presents the results of the round-robin Cet article présente les résultats expérimentaux réalisés
test carried out in the Nordic countries to evaluate the en Scandinavie pour évaluer la répétabilité et la reproducti-
repeatability and reproducibility of three Nordic test bilité des mesures accélérées du coefficient de diffusion des
methods for measuring the chloride diffusion coefficient ions chlore dans les bétons. Trois procédures ont été testées :
of concrete. The three methods were: NT BUILD NT BUILD 443, NT BUILD 355 et l’essai accéléré
443, NT BUILD 355 and the CTH Rapid Test. A total CTH. 9 laboratoires et 4 pays ont été impliqués dans ce
of nine laboratories from Denmark, Sweden, Norway projet : Danemark, Suède, Norvège et Islande. Ont été
and Iceland participated in the test. Concrete specimens utilisés des échantillons de béton fabriqués avec 3 types de
with three types of binder and two water-cement ratios ciment et 2 rapports E/C différents. Les résultats d’essais
were used in the test, and the test results were evaluated ont été analysés selon la norme ISO 5725-94. Ceux-ci
in accordance with international standard ISO 5725-94 montrent que parmi les 3 méthodes testées, la procédure
for the determination of repeatability and reproducibil- CTH donne la meilleure précision dans la détermination
ity, in spite of small numbers of trials and laboratories. du coefficient de diffusion des ions chlore pour le ciment
The results show that, among the three methods tested Portland ou les bétons avec fumée de silice. La procédure
in this project, the CTH Rapid Test in general gives the NT BUILD 443 fournit aussi des précisions acceptables.
best precision in the determination of chloride diffusion Par contre, si la procédure NT BUILD 355 donne des
coefficient for Portland cement or silica fume concrete. résultats satisfaisants pour les bétons perméables, elle
The NT BUILD 443 method also gives satisfactory pre- n’offre pas une bonne reproductibilité lorsque la perméabi-
cision. The NT BUILD 355 method gives satisfactory lité des matériaux diminue.
precision for permeable concrete, but shows poor repro-
ducibility for dense or low-permeability concrete.
1. INTRODUCTION BUILD 443 [14] and the CTH Rapid Test which was
recently accepted as a Nordic standard NT BUILD 492
The chloride diffusion coefficient is considered as a [15]. The theory behind these three methods has been
very important parameter, dominating chloride transport reviewed and discussed elsewhere [16]. Due to different
in concrete. Different test methods have been developed theoretical bases, the measured values of chloride diffu-
for determining chloride diffusion coefficient, such as sion coefficients obtained by using different test meth-
diffusion cell tests (steady state diffusion) [1-3], immer- ods are not generally directly comparable. The literature
sion tests (non-steady state diffusion) [4-5], migration showed a large scattering of the values from measure-
cell tests (steady state migration) [6-9], and the CTH ments of the chloride diffusion coefficient. In addition
Rapid Test (non-steady state migration) [10-12]. In the to the different theoretical bases, measurement error
Nordic countries, three rapid test methods were devel- might be one of the reasons to this scatter. So far, infor-
oped in the past decade: NT BUILD 355 [13], NT mation about the precision of each test method is very
Editorial Note
Dr. Tang Luping is a RILEM Staff Member and works at the Swedish National Testing and Research Institute, a RILEM Titular Member. He parti-
cipates in the work of RILEM TC 178-TMC: ‘Testing and modelling chloride penetration in concrete’.
limited. In order to determine the uncertainty of the three specimens for each test series were cut from the
measured values for chloride diffusion coefficient, a central portion of two concrete cylinders. The speci-
round-robin test was carried out in the Nordic coun- mens were then prepared for testing according to the
tries. This work was performed as a Nordtest project, respective standards or instructions.
under which the repeatability and reproducibility of the
three Nordic test methods were evaluated. A total of
nine laboratories from Denmark, Sweden, Norway and 2.2 Brief description of the test methods
Iceland participated in the test. Due to the high cost of
the round-robin test, only three types of concrete were 2.2.1 NT BUILD 355 – Steady state migration test
used in the test. This paper presents some of the results The Nordic standard NT BUILD 355 [13] was
from the project. A detailed report has been published developed in the beginning of the 1980’s and revised in
within the project [17]. the middle of the 1990’s. This is a steady state migration
test. The test procedure according to the revised version
involves:
2. EXPERIMENTAL – Coating the curved surface of the specimen with, for
example, epoxy resin;
2.1 Sample preparation – Saturating the specimen by immersion in saturated
lime water until the weight changes by not more than
A total of three types of concrete with different types 0.1% per day;
of binder, and two water-binder ratios were manufac- – Mounting the specimen between the migration cells
tured at SP. The mix proportions and properties of con- and filling the upstream cell with 5% NaCl solution and
crete are listed in Table 1. the downstream cell with 0.3 N NaOH solution;
– Applying an external potential of 12 V DC between the
two cells and measuring the actual
Table 1 – Mix proportions and properties of concrete (kg/m3) potential drop across the specimen by
Mix Binder type Water- Binder Aggregate Admixture Compr. using two reference electrodes;
binder content content wt% of Strength* – Qualitatively checking the down-
ratio binder MPa stream cell for chlorides by using
92%(wt) SRPC + 8%(wt) 0.8 slightly acidified 1 M AgNO3 solu-
A silica fume (slurry) 0.4 420 1860 (Cementa 82.6 tion until a white precipitate can be
corresp. to CEM II/A-D 92M)
observed;
B
100% SRPC
0.5 380 1860 0 63.2 – Quantitatively determining chlo-
corresp. to CEM I ride contents in the downstream cell
Dutch slag cement at least once a day over at least seven
C containing ~70%(wt) slag, 0.5 390 1860 0 45.1 days by using a standardised method;
corresp. to CEM III/B
– Performing linear regression analy-
* At the age 28 days according to the Swedish Standard SS 13 72 10. sis of at least five points of the linear
part of the c-t (concentration-time)
Each type of concrete was mixed in a 40 litre paddle curve until a linear correlation coefficient of at least 0.9
mixer and cast into 200 mm long × 100 mm diameter is obtained;
plastic cylindrical moulds. Immediately after casting the – Calculating the chloride f lux J from the slope of the
moulds were covered with thick plastic films to prevent linear regression.
evaporation from the concrete surface. One day after The chloride diffusion coefficient, Dssm, is then cal-
casting the concrete cylinders were numbered according culated using the following equation:
to the casting order and stored (still in the moulds) in RTL RTL V2 ∆c2
water at about 20°C. At an age of 7 days, the moulds Dssm = ⋅J = ⋅ ⋅ m 2 / s (1)
were stripped and the cylinders were continuously cured zF∆Ec1 zF∆Ec1 A ∆t
in water at about 20°C. At 21 days, the concrete cylin- where
ders were sorted for the nine laboratories plus extra R: gas constant, R = 8.314 J/(K·mol);
specimens for reservation by arranging similar distribu- T: average value of the initial and final temperatures
tions for each test method to minimise the effect of cast- in the anolyte solution, K;
ing order. The sorted concrete cylinders were wrapped L: thickness of the specimen, m;
with wet textile materials and sealed in thick plastic films z: absolute value of ion valence, for chloride, z = 1;
before being sent to the respective laboratories. On F: Faraday constant, F = 9.648 × 104 J/(V·mol);
arrival at each laboratory, the cylinders were again kept ∆E: absolute value of the potential dif ference
in water at about 20°C until a specified age. The extra between the upstream solution and the down-
reserved specimens were kept at SP by storage in water stream solution, measured by using two refer-
at about 20°C. ence electrodes, V;
At the specified age (about five weeks for the immer- c1, c2: chloride concentration in the upstream and
sion test, about two months for the migration tests), downstream cell, respectively, kg/m3;
480
Tang, Sørensen
( c)
β
involves: t −0.6
481
Materials and Structures/Matériaux et Constructions, Vol. 34, October 2001
3. THEORETICAL ANALYSIS OF
MEASUREMENT ERRORS
If we exclude the influence of material heterogeneity,
the errors caused by different measurement parameters
can be estimated by differentiating the relevant parame-
ters involved in the calculation equations. The detailed Fig. 1 – Coefficient of variation (COV) caused by different error
evaluation procedures were described elsewhere [16]. sources.
The evaluation results are summarised as follows. Test conditions: U = 30 ± 0.2 V, L = 0.05 ± 0.0002 m, T = 298 ±3 K,
By differentiating Equation (1) one can obtain the t = 24 ± 0.1 h, and c0 = 2 ±0.1 mol/l; assuming ∆xd = ± 0.5 mm.
measurement errors involved in the NT Build 355 test:
all it increases the uncertainty of the method. According
∆c to Sørensen [24], an error of 15% is estimated from a
∆ 2
∆Dssm
= +
( )
∆T ∆ ∆E ∆L ∆c1 ∆V2 ∆A
+ + + + +
∆t
(6)
round-robin test with three laboratories.
Dssm T ( )
∆E L c1 V2 A ∆c2
The measurement errors involved in the CTH Rapid
Test can be shown in Fig. 1 [16], where the thick line is
∆t the sum of the errors caused by various sources, and can
Under the test conditions of T = 298 K, ∆E = 10 V, be called the “pessimistic” maximum error in the mea-
L = 0.05 m, c1 = 0.5 mol/l, V2 = 236 ml or 0.000236 m3, surement. It can be seen that the measurement of pene-
and A = π × 0.12/4 = 0.00785 m2, for a normal educated tration depth results in a large error, but when a penetra-
operator with normal laboratory equipment, the mea- tion depth is larger than 10 mm, the maximum error can
surement errors except the last one in the above equa- be minimised into a range of a few per cent.
tion, which will be discussed later, might be limited to
∆T = ± 3°C, ∆(∆E) = ± 0.2 V, ∆L = ± 0.2 mm or
± 0.0002 m, ∆c 1 = ± 0.01 mol/l, ∆V 2 = ± 1 ml or 4. TEST RESULTS AND PRECISION ANALYSIS
0.000001 m3, and ∆A = ± 0.00005 m3. The accumula-
tive error will be: The measured chlor ide dif fusion coef f icients
reported from different laboratories are summarised in
∆c Table 2. Since the NT BUILD 355 test needs expensive
∆ 2
∆Dssm ∆t equipment and is time-consuming, only 2 to 3 laborato-
= 0.06 + (7) ries participated the test. Owing to the laborious and
Dssm ∆c2
expensive fact of the NT BUILD 443 test, only Mix B
∆t was tested in several laboratories while two laboratories
The last term in the above equation is the error caused tested Mixes A and C. The ISO 5725-2 standard proce-
by the determination of the slope from a concentration- dure [25] was employed for the precision analysis. From
time plot, which is dependent on the chloride analysis (val- the preliminary precision analysis, it was found that the
ues of c2) and the judgement of linearity or “steady state”. NT BUILD 443 test results reported by Laboratory 4 for
This might be the largest error source involved in the Mix B and Mix A exceeded the criteria for outlier and
steady state test. An example of more than 100% error in straggler respectively. These data were rejected in the
the determination of slope has been given in Ref. [16]. calculation of repeatability and reproducibility for this
It is difficult to make a mathematical analysis of the method. After rejection of the outlier and straggler data,
measurement error involved in the NT BUILD 443 test, only one laboratory (Laboratory 1) returned valid results
because the final calculation for chloride diffusion coeffi- for Mix A from the NT BUILD 443 test. Thus the data
cient is by curve-fitting. The error from a curve-fitting reported by Sørensen [24] from another round-robin
technique is trivial if a correct computer program is test with three participating laboratories and similar con-
employed. The main error sources may be in the determi- crete were employed to calculate the repeatability and
nation of chloride profiles, which involves sampling tech- reproducibility for Mix A from this test.
nique and chloride analysis. As Gran [23] reported, about The mean diffusion coefficients are shown in Figs. 2
10% error for high chloride contests and up to 40% error and 3, and the results from the precision analysis are
for low chloride contests in concrete samples were found summarised in Figs. 4 and 5. The data reported by
when using the standard recommended Valhard titration. Frederiksen et al. [26] were also employed in Fig. 3 for
The large error in chloride content measurement may not comparison. Since the effects of potential drop and con-
mean the same error in the NT BUILD 443 test, but after crete age were not considered in reference [26], the data
482
Tang, Sørensen
Fig. 2 – Mean diffusion coefficients from different test methods Fig. 3 – Comparison between different chloride diffusion coeffi-
(logarithmic scale). cients (normal scale).
for the CTH Rapid Test from reference [26] have, range of 8~14%, and its reproducibility COV is in a
therefore, been re-calculated by using the absolute range of 16~23%. The latter is comparable with the
potential drop ∆E = (U – 2) in Equation (3) instead of CTH Rapid Test.
the applied potential U and corrected to an age of 50 The NT BUILD 355 test reveals good precision only
days (the average age of the specimens for the NT for concrete Mix B (w/c 0.5, moderate permeability)
BUILD 443 test in that study), using Equation (5), but with a repeatability COV of 8.2% and a reproducibility
with 50 instead of 60. COV of 16.6%. The method gives a very poor repro-
483
Materials and Structures/Matériaux et Constructions, Vol. 34, October 2001
REFERENCES
[1] Ushiyama, H. and Goto, S., ‘Diffusion of various ions in hard-
ened portland cement paste’, 6th Intl. Congr. Chem. Cem.,
Moscow, Vol. II-1 (1974) 331-337.
Fig. 6 – Accumulated chlorides passed through concrete Mix A. [2] Page, C. L., Short, N. R. and Tarras, A. El., ‘Diffusion of chlo-
Data of A11-T, A11 and A12 were measured by Laboratory 7, while ride ions in hardened cement pastes’, Cement and Concrete Research
those of A31-T, A31 and A32 were measured by Laboratory 4. 11 (3) (1981) 395-406.
[3] Roy, D. M., Kumar, A. and Rhodes, J. P., ‘Diffusion of chloride
and cesium ions in Portland cement pastes and mortars contain-
ducibility (COV 55~98%) for low permeability con- ing blast furnace slag and fly ash’, Proceedings of the 2nd
crete, such as Mix A or Mix C. The testing time for low International Conference on the Use of Fly Ash, Silica Fume,
permeability concrete is usually very long (1~2 months Slag and Natural Pozzolans in Concrete, ACI SP-91, Madrid,
(1986) 1423-1444.
in this study). Some unintended events, e.g. leakage, [4] Collepardi, M., Marcialis, A. and Turriziani, R., ‘The kinetics of
leaching of ions from the pore solution, etc., might penetration of chloride ions into the concrete’, Il Cemento (4)
occur during this long-time test, resulting in completely (1970) 157-164.
different chloride f low properties, as shown in Fig. 6, [5] Frederiksen, J. M., ‘APM 302 - Danish test method for the chlo-
where different laboratories measured rather different ride ingress into concrete’ (in Danish), Dansk Beton (2) (1992)
22-27.
slopes. As pointed out in the theoretical analysis, the [6] Whiting, D., ‘Rapid determination of the chloride permeability
determination of slope is the largest error source in a of concrete’, Report No. FHWA/RD-81/119 (1981).
steady state test. It seems that this method is more suit- [7] Dhir, R. K., Jones, M. R., Ahmed, H. E. H. and Seneviratne, A.
able to permeable or moderately permeable concrete. M. G., ‘Rapid estimation of chloride diffusion coefficient in con-
484
Tang, Sørensen
crete’, Magazine of Concrete Research 42 (152) (1990) 177-185. Officials, Washington D.C., (1983).
[8] Andrade, C., ‘Calculation of chloride diffusion coefficients in [19] McGrath, P. F. and Hooton, R. D., ‘Influences of voltage on
concrete from ionic migration measurement’, Cement and chloride diffusion coefficients from chloride migration tests’,
Concrete Research 23 (3) (1993) 724-742. Cement and Concrete Research 26 (8) (1996) 1239-1244.
[9] Zhang, T. and Gjørv, O. E., ‘An electrical method for accelerated [20] Justnes, H. and Rodum, E., ‘Chloride ion diffusion coefficients
testing of chloride diffusivity in concrete’, Cement and Concrete for concrete - A review of experimental methods’, Proceedings
Research 24 (8) (1994) 1534-1548. of the 10th International Congress on the Chemistry of Cement,
[10] Tang, L. and Nilsson, L.-O., ‘Rapid determination of chloride June 3-6, 1997, Gothenburg, Sweden, Vol. IV, p. 4iv081.
diffusivity of concrete by applying an electric field’, ACI Materials [21] Zhang, T., ‘Chloride diffusivity in concrete and its measurement
Journal 49 (1) (1992) 49-53. from steady state migration testing’, Doctoral thesis 1997:132,
[11] Tang, L., ‘Electrically accelerated methods for determining chlo- Dept. of Building Materials, University of Science and
ride diffusivity in concrete’, Magazine of Concrete Research 48 (176) Technology (NTNU), Trondheim, Norway, (1997).
(1996) 173-179. [22] Tang, L. and Nilsson, L.-O., ‘Chloride diffusivity in
[12] Tang, L., ‘Chloride transport in concrete - Measurement and pre- high-strength concrete at different ages’, Nordic Concrete Research
diction’, PhD thesis, P-96:6, Dept. of Building Materials, Chalmers 11 (1992) 162-171.
University of Technology, Gothenburg, Sweden, (1996). [23] Gran, H. C., ‘Measurement of chlorides in concrete - An evalu-
[13] NT BUILD 355, ‘Concrete, mortar and cement-based repair ation of three different analysis techniques’, in ‘Chloride
materials: Chloride diffusion coefficient from migration cell Penetration into Concrete Structures - Nordic Miniseminar’, ed.
experiments’, Ed. 2, Nordtest, Espoo, Finland, (1997). by L.-O. Nilsson, Chalmers University of Technology, P-93:1
[14] NT BUILD 443, ‘Concrete, hardened: Accelerated chloride (1993) 71-80.
penetration’, Nordtest, Esbo, Finland, (1995). [24] Sørensen, H. E., ‘Nordtest project 1154-94, Determination of
[15] NT BUILD 492, ‘Concrete, mortar and cement-based repair chloride penetration parameters for concrete’, AEC Laboratory
materials: Chloride migration coefficient from non-steady state Report No. 94-019, AEC Consulting Engineers Ltd., Vedbæk,
migration experiments’, Nordtest, Espoo, Finland, (1999). Denmark, (1995).
[16] Tang, L., ‘Chloride diffusion coefficient of concrete and relevant [25] ISO, ‘Accuracy (trueness and precision) of measurement meth-
test methods - The state of the art and suggestions for future ods and results - Part 2: Basic method for the determination of
work’, SP Report 1997:23, SP Swedish National Testing and repeatability and reproducibility for a standard measurement
Research Institute, Borås, Sweden, (1997). method’, ISO 5725-2:1994, International Standardization
[17] Tang, L. and Sørensen, H. E., ‘Evaluation of the rapid test meth- Organization.
ods for measuring the chloride diffusion coefficients of concrete’, [26] Frederiksen, J. M., Sørensen, H. E., Andersen, A. and
SP Report 1998:42, SP Swedish National Testing and Research Klinghoffer, O., ‘HETEK, The effect of the w/c ratio on chlo-
Institute, Borås, Sweden, (1998). ride transport into concrete - Immersion, migration and resistiv-
[18] AASHTO, ‘Standard method of test for rapid determination of ity tests’, HETEK Report, ed. by J. M. Frederiksen, published by
the chloride permeability of concrete’, AASHTO T 277, the Danish Road Directorate, (1997).
American Association of State Highway and Transportation
485