Construction and Building Materials 55 (2014) 183–193

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier .com/locate /conbui ldmat

A service life based global warming potential for high-volume fly ashconcrete exposed to carbonation

http://dx.doi.org/10.1016/j.conbuildmat.2014.01.0330950-0618/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +32 9 264 55 22; fax: +32 9 264 58 45.E-mail address: nele.debelie@ugent.be (N. De Belie).

Philip Van den Heede, Nele De Belie ⇑Magnel Laboratory for Concrete Research, Ghent University, Technologiepark Zwijnaarde 904, B-9052 Ghent, Belgium

h i g h l i g h t s

� HVFA concrete is less resistant to carbonation than OPC concrete but still is durable.� Colorimetric assessment underestimates the actual (microscopic) carbonation depth.� Service life of HVFA concrete estimated with a simplified model exceeds 100 years.� When curing and weather effects are considered, service life also exceeds 100 years.� The GWP of carbonation exposed HVFA concrete is 18–27% lower than OPC concrete’s GWP.

a r t i c l e i n f o

Article history:Received 2 October 2013Received in revised form 7 January 2014Accepted 11 January 2014Available online 7 February 2014

Keywords:Fly ashSteel depassivationAccelerated carbonation testingThin section analysisService life predictionLife cycle assessmentGlobal warming potential

a b s t r a c t

To evaluate the global warming potential (GWP) of carbonation exposed high-volume fly ash (HVFA) con-crete, its expected service life should be known. In the early stages of product development, this is donewith rudimentary prediction models based on simple colorimetric carbonation testing. More sophisti-cated methods (e.g. thin section analysis) and prediction models that consider concrete curing and mete-orological conditions (cf. Fib Bulletin 34) can be used later on if the former predictions look promising.This paper shows that both rudimentary and advanced prediction models result in significant GWPdecrease (�18% to 27%) for HVFA concrete, regardless the underlying carbonation assessment method.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Within the development process of a new potentially ‘green’concrete type, it is at all time imperative to keep an eye on its ex-pected reduced cement related greenhouse gas (GHG) emissions incomparison with traditional concrete. It is known that the produc-tion of ordinary Portland cement involves considerable CO2 emis-sions. Based on in depth literature study this is around 842 kgper ton on average [1]. Now, the environmental benefit that couldbe achieved by replacing a considerable portion of the cement withsupplementary cementitious materials such as fly ash (FA) will de-pend greatly on the strength and durability performance of thematerial. In case its compressive strength is lower in comparisonwith traditional concrete, structure dimensions will evidently belarger. As a result, more concrete will need to be manufactured

to meet the structural requirements. If the concrete in its expectedfuture natural environment – in this case an environment wherecarbonation-induced corrosion is at risk – will require regularmaintenance, repair or even full replacement, the impact of thenecessary rehabilitation actions will add on to the impact of theoriginal concrete amount needed to build the structure [1].

Evidently, little information on the actual durability perfor-mance of a new concrete type is readily available in the earlystages of product development, and this for two reasons. (i) Thenew material has not been applied yet in any real structure thatcould be regularly inspected for damage. (ii) Accelerated laboratorydurability tests, e.g. experiments done to predict the time to car-bonation-induced depassivation of the reinforcing steel, are timeconsuming. True, exposing the concrete at its proper testing ageto an increased CO2 concentration followed by a time-dependentevaluation of the carbonation depth may very well result in theexperimental determination of a carbonation rate for the concretecomposition under investigation. These carbonation depths could

184 P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193

simply be visualized by using the well-known colour indicatorphenolphthalein [2] or even more accurately with microscopy onthin sections [3,4]. However, without a full understanding of theconcrete’s curing behaviour and the influencing environmental fac-tors of its natural environment even the most accurately deter-mined experimental carbonation rate will not result in a realisticservice life prediction on the short term. This is because the effectof curing can only be evaluated by performing carbonation tests atmultiple testing ages. The expected meteorological conditions (rel-ative humidity, amount of precipitation, etc.) of a specific environ-ment in which the concrete would be used once on the market,should also be assessed more thoroughly. Logically, a quantifica-tion of these influencing factors asks for more effort and time with-in the product development process. Both items will be addressedonly if a first rough service life and sustainability assessment basedon rather easy-to-obtain data already gives promising results.

Another important aspect that deserves attention is the appliedCO2 concentration during an accelerated carbonation test. Thisconcentration can vary considerably with the chosen test method.For instance, NBN B15-100 suggests a CO2 concentration of 1% dur-ing the carbonation test for equivalent performance assessment ofconcrete with type II additions, but gives the freedom to work withother CO2 concentrations as well. According to Fib Bulletin 34 con-crete needs to be stored in a carbonation cabinet at 2% CO2 for28 days to determine its inverse effective carbonation resistance[5]. Lo and Lee applied the same CO2 concentration [6]. NT Build357 on the other hand specifies an atmosphere of 3% CO2 [7] andBorges et al. expose concrete to a CO2 concentration of 5% to accel-erate carbonation [8]. In our research, the applied CO2 concentra-tion was 10% by volume cf. [9,10]. Literature survey also showsthat still higher CO2 concentrations are inherent to some testmethods for accelerated carbonation. For instance, a CO2 concen-tration of 20% was applied in [11]. The AFPC-AFREM procedureon accelerated carbonation testing is even more extreme as it re-quires a carbonation chamber conditioned at 50% CO2 [12]. Thus,it is clear that there is not much uniformity among the availablecarbonation test methods. Nevertheless, some researchers are con-vinced that increasing the CO2 concentration too much will resultin important chemical changes of the concrete’s paste fraction thatwould not occur if the concrete could naturally carbonate at 0.03%.Castellote et al. found that the polymerisation of the calcium sili-cate hydrate (C–S–H) after carbonation of OPC pastes increaseswith the increase of the CO2 concentration applied [13]. When car-bonating at 0.03% CO2 and 3% CO2 there is remaining C–S–H gel,although with a lower Ca/Si ratio than that of an uncarbonatedsample (Ca/Si = 1.87), but still with the characteristics of thisphase. The obtained Ca/Si ratios for the C–S–H gels of samples car-bonated at 0.03% CO2 and 3% are quite similar and amount to 1.23%and 1.18%, respectively. When carbonating at 10% and 100% CO2

the C–S–H gel completely disappears. It is completely transformedinto polymerised Ca-modified silica gel. In a CO2 atmosphere of 3%or less, this transformation is only partial. The presence of remain-ing unhydrated cement and ettringite also highly depends on theimposed CO2 concentration. After carbonation at 0.03% CO2 and3% CO2, unhydrated cement and ettringite are still there. Testingat 10% CO2, makes the ettringite completely disappear and leavesonly a small amount of unhydrated cement. A CO2 concentrationof 100% results in a complete disappearance of both the ettringiteand the unhydrated cement. Based on these findings Castelloteet al. conclude that the maximum allowable CO2 concentrationfor acceleration of the carbonation process is 3% [13]. Only underthese conditions, a dramatical change in microstructure of thepastes is avoided. Nevertheless, they emphasize that these obser-vations only hold true for OPC pastes and they advise that theirvalidity should also be checked for pastes consisting of alternativebinder materials (e.g. FA).

Borges et al. studied the carbonation behaviour of pastes con-taining high amounts of blast-furnace slag (BFS) by exposingthem at 5% CO2. When dealing with OPC + BFS pastes attentionneeds to be paid to both the carbonation of calcium hydroxide(CH) and C–S–H. The carbonation of CH usually results in a den-sification of the microstructure because the volume of the car-bonates (calcite) formed is 11–12% greater than the volume ofCH. Carbonation of C–S–H on the other hand causes a polymer-isation of the silica chains in the C–S–H, which may be respon-sible for a volumetric decrease (shrinkage), a coarsening of thematrix and cracking [8]. There is evidence that CO2 reacts simul-taneously with the CH and the C–S–H [14]. This statement isalso confirmed by Thiery et al. [15]. According to the latter, CHcarbonation is initially more rapid than that of the C–S–H gel,but this situation soon reverses because of the formation of alayer of CaCO3 microcrystals on the surface of CH crystals.Now, in blended cement pastes there is less CH available dueto pozzolanic reaction. As a consequence, rapid decalcificationof the C–S–H is expected when the paste is highly permeableto CO2. This will result in carbonation shrinkage which is accel-erated when the Ca/Si ratio drops below 1.2 [16]. In a blendedpaste characterised by a low permeability, carbonation of thelow Ca/Si ratio C–S–H is believed to be less at risk. Borgeset al. observed from permeability experiments that OPC + BFSpaste is very sensitive to the carbonation shrinkage phenome-non, although carbonates filled some of the pores which in-creased the overall density and reduced the overall porosity[8]. Similar findings could be expected for blended pastes con-sisting of OPC and FA.

Yet, when considering the findings of Castellote et al. [13], atleast part of the shrinkage due to the carbonation of C–S–H ob-served by Borges et al. [8] may also be attributed to the fact thatthe carbonation test was performed in an atmosphere with 5%CO2 instead of the proposed maximum value of 3%. Thus, perform-ing carbonation tests at a CO2 concentration above 3% may overes-timate the carbonation of C–S–H and thus the measuredcarbonation depths and rates that result from it.

However, this effect does not take into account another phe-nomenon inherent to carbonation. The production of calcium car-bonate always coincides with the release of water. Whencarbonating concrete at a high CO2 concentration, the amount ofwater produced could be more than the porous matrix is capableof expelling in the same time interval. The time needed to establisha condition of equilibrium again is believed to slow down the prop-agation of the carbonation depth [17,18]. da Silva et al. also men-tion that CO2 solubility is low when high CO2 concentrations areused. The penetrating CO2 first needs to transform into acid inthe presence of water before the actual carbonation reaction cantake place and the amount of CO2 capable of dissolving in wateris limited [17]. If these mechanisms would turn out more dominantthan the coarsening of the pore structure due to the carbonationshrinkage attributable to the acceleration of the carbonation test,applying a high CO2 concentration would rather underestimatethe carbonation depth and rate under field conditions.

Thus, depending on the prevailing mechanism – either the in-crease in permeability due to carbonation shrinkage [8] or theblocking effect of the water released during carbonation [18] to-gether with the limited solubility of CO2 in water [17] – experi-mentally estimated field carbonation rates from highlyaccelerated tests either over- or underestimate the carbonationresistance and expected service life of a concrete with a high con-tent of supplementary cementitious materials in its natural envi-ronment. A calculation of the global warming potential based onthis service life assessment may attribute a higher or lower envi-ronmental burden to the material than when the service lifeassessment would be based on a more natural carbonation test.

P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193 185

2. Research purpose and methodology

This paper specifically investigates the difference in globalwarming potential for a potential low-carbon concrete composi-tion located in an environment subject to carbonation-inducedcorrosion when based on either a rudimentary or a more sophisti-cated concrete and environment specific service life prediction cf.Fib Bulletin 34 [5]. Since a shift from a simplified to a more com-plex approach is inherent to the process of product development,a critical evaluation of this difference is considered to be relevant.It may certainly help product developers to avoid an inaccurateassessment of the actual environmental benefit of a newly devel-oped concrete composition. The concrete type under investigationis a high-volume fly ash (HVFA) concrete composition cf. Malhotraand Mehta [19].

Within a first research step, the effect of the applied techniqueto determine carbonation depth was investigated. A comparisonwas made between the fast and accessible colorimetric measure-ments with phenolphthalein and the more labour intensive, yetmore accurate microscopic measurements on thin sections. The ef-fect of the applied CO2 concentration (1% or 10%) was also studied.

Within a second research step these experimentally and theo-retically obtained carbonation rates for the studied HVFA composi-tion were implemented in both a rudimentary service lifeprediction model and the more sophisticated model prescribedby Fib Bulletin 34 [5].

In a final research phase, the well-known life cycle assessment(LCA) methodology was adopted to calculate a global warming po-tential (GWP) in kg CO2eq for both the HVFA concrete compositionand the appropriate reference concrete for the applicable exposureclasses for carbonation-induced corrosion. All calculations wereperformed in the LCA software SimaPro.

3. Materials and methods

3.1. Concrete mixtures

In total, eight concrete mixtures were manufactured (Table 1). CompositionsT(0.55) and T(0.50) are Ordinary Portland Cement (OPC) references with a minimumcement content and a maximum water-to-cement ratio (W/C) ratio conforming toNBN B15-001 for exposure classes XC3 and XC4, respectively. The first exposureclass corresponds with a moderately humid environment with exposure to carbon-ation-induced steel corrosion. The second exposure class also represents an envi-ronment with exposure to carbonation-induced steel corrosion, but now incombination with wet/dry cycles.

FA containing concrete compositions F(1)15 and F(2)15 were made in compli-ance with the k-value concept of NBN B15-001. The two compositions are identicalexcept for the fact that fly ashes F(1) and F(2) came from two different sources. Byusing the k-value concept, the maximum fly ash-to-binder (F/B) ratio for a mini-

Table 1Mixture proportions, slump and experimental strength classes of the tested concrete com

T(0.55) T(0.50)

1 Sand 0/4 (kg/m3) 715 7142 Aggregate 2/8 (kg/m3) 516 515

Aggregate 8/16 (kg/m3) 672 6713 CEM I 52.5 N (kg/m3) 300 3204 Fly ash (kg/m3) 0 05 Water (kg/m3) 165 1606 SPa (ml/kg B) 0 0

W/B 0.55 0.50F/B 0 0Slumpb S3 S1Strength classc C30/37 C45/55

a Dosage of superplasticizer (SP) in ml per kg binder.b S1 (10–40 mm), S2 (50–90 mm), S3 (100–150 mm), S4 (160–210), S5 (P220 mm).c Based on the 5% characteristic (compressive strength) value cf. EN 1990 (n = 3, VX: u

mum total binder content equals only 15% which indicates that the environmentalbenefit of this partial cement replacement will be rather limited. To increase thisbenefit, HVFA mixtures F(1)50 and F(2)50 were developed.

Fly ashes F(1) and F(2) meet the requirements of NBN EN 450-1 to qualify foruse in concrete: the loss on ignition was lower than 5% (Class A) and the 45 lm fine-ness was less than 40% (Class N). Thus, the former property turns out similar forboth fly ashes, while the latter indicates that fly ash F(1) is twice as fine as flyash F(2) (Table 2).

Note that the HVFA mixtures are characterised by a higher total binder content(cement + FA) and a lower water-to-binder (W/B) ratio compared to the referencesT(0.55) and T(0.50). This was mainly done to ensure a strength class equal or higherthan the one minimum required for the XC3 and XC4 environment. Although thestrength classes of F(1)50 and F(2)50 were higher than the minimum indicativestrength classes C25/30 and C30/37 for exposure classes XC3 and XC4 (cf. NBNN15-001), they were not always higher than the experimental strength classes ofthe references (Table 1). The experimental strength class of each studied concretecomposition was obtained from a compressive strength test (cf. NBN B15-220) per-formed on three cubes (n = 3) with a 150 mm side after 28 days of optimal curing at20 �C and 95% relative humidity (RH). The 5% characteristic compressive strengthvalue needed to determine the strength class was calculated cf. NBN EN 1990, whileassuming the variation coefficient VX unknown from prior knowledge.

3.2. Curing and sample preparation

Per concrete mixture, cubes with a 100 mm side were cast. After casting, thecubes were kept at a constant temperature and relative humidity (RH) of 20 �Cand 95%, respectively. Demoulding took place the next day whereupon the speci-mens were stored again under the same conditions until the age of 21, 84 or175 days. At that moment, 5 of the 6 cube surfaces were treated with an imperme-able coating to ensure a unidirectional flow of CO2 throughout the samples duringthe experiment. The untreated side was always a cast surface of the cube.

3.3. Carbonation testing

3.3.1. Accelerated carbonation testingPer testing age, the accelerated carbonation test involved exposing three coated

concrete cubes to an atmosphere containing 10% CO2 at 20 �C and 60% RH cf. [9,10].Apart from the CO2 concentration, these test conditions were thought to be repre-sentative for exposure class XC3.

Another three cubes of each concrete mixture apart from OPC reference T(0.55)were alternatingly submerged in water for one week and exposed to the same testconditions (10% CO2, 20 �C and 60% RH) for another week to simulate exposure classXC4 cf. [9].

The last saw cut taken from each test series subjected to the accelerated carbon-ation tests (either after 18 weeks or after 10 wet/dry cycles) were not only testedfor carbonation using the colorimetric method, but also by means of microscopicanalysis on thin sections. This was only done for the samples tested at 28 and91 days. The carbonation resistance after 182 days of optimal curing was only as-sessed colorimetrically.

At a later stage of the research, another 24 concrete cubes (side: 100 mm) ofHVFA mixture F(1)50 were cast. Half of the cubes were subjected to an acceleratedcarbonation test at 10% CO2, 20 �C and 60% RH, while the other twelve were kept ina carbonation cabinet at 1% CO2, 20 �C and 60% RH for a similar test. According to daSilva et al. a concentration of 1% CO2 develops the same reaction products as anormal atmosphere at 0.03% CO2 [17] and can thus be considered as a more or lessnatural carbonation test. Exposure started 28 days after the cubes were optimally

positions.

F(1)15 F(2)15 F(1)50 F(2)50

698 698 645 645503 503 465 465656 656 606 606300.8 300.8 225 22548 48 225 225160 160 158 1582.0 3.0 5.0 5.00.46 0.46 0.35 0.3515 15 50 50S3 S4 S2 S5C45/55 C45/55 C40/50 C35/45

nknown).

Table 2Physical and chemical characteristics of the cement and fly ash types.

CEM I 52.5 N F(1) F(2)

LOI (%) 1.5 4.8 4.445 lm fineness (%) – 13.2 26.6Density (kg/m2) 3120a 2230 2260Specific surface, Blaine method (m2/kg) 390a – –

CaO (%) 63.37 2.80 7.58SiO2 (%) 18.90 48.54 50.83Al2O3 (%) 5.74 33.34 20.45Fe2O3 (%) 4.32 3.52 7.52SO3 (%) 3.34 0.80 0.68MgO (%) 0.89 0.72 1.77Na2O (%) 0.47 0.34 1.00K2O (%) 0.73 1.54 1.69

a Data provided by the cement manufacturer.

186 P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193

pre-conditioned and coated (as described in Section 3.2). Field carbonation rates ob-tained from both experiments were compared with each other to quantify the effectof the CO2 concentration applied (10% versus 1%).

3.3.2. Colorimetric measurement detailsPer saw cut treated with phenolphthalein, nine different measurements (one

every 10 mm) were done, and this to the nearest millimeter. After spraying the1% phenolphthalein solution onto the concrete slices, the carbonated area will showcolourless, while the non-carbonated area will colour purple. The point of transitionfor phenolphthalein is situated at around pH 8.0–9.8 [20]. Since a pH between 9 and13 is required to maintain the protective passivation layer upon the steel reinforce-ments, the colorimetric method gives a more or less adequate estimation of the CO2

ingress. However, given the pH range for colour transition, there is still a risk for anunderestimation of the actual carbonation depth. To evaluate the magnitude of thisunderestimation, the actual penetration depth was also evaluated on thin sections.

3.3.3. Microscopic measurement detailsAfter 28 and 91 days of optimal curing followed by 18 weeks of exposure or 10

wet-dry cycles, a 45 � 30 � 15 mm3 prism was sawn from one 15 mm thick slice insuch a way that a 45 � 15 mm2 face was the exposed surface. Thin sections wereprepared from the collected prisms in accordance with [21]. Note that for theOPC concrete compositions thin sections were made for only one curing age aftercompletion of the accelerated carbonation test. For OPC reference T(0.55) the28 days curing period was considered, while for mixture T(0.50) thin sections weretaken from the concrete that was optimally cured for 91 days.

All thin sections were examined with a Leica DM LP polarizing microscope un-der crossed polars. The luminous intensity of the polarized light equaled 6.5 on ascale of 10. Images were taken with a Leica DFC295 camera and analysed withthe Leica Application Suite v.3.7 software. The carbonated zone was characterisedby a lighter brownish stain (Fig. 1).

Fig. 1. The carbonated zone with a lighter stain as observed under the microscopeunder crossed polars. (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)

3.3.4. Carbonation rate and resistanceFor each concrete mixture tested, the measured carbonation depths xc (in mm)

with the phenolphthalein colour indictor were plotted as function of the squareroot of the exposure time t (in weeks) to determine an experimental (accelerated)carbonation coefficient Aacce (in mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

). This approach is in agreement withFick’s law of diffusion for the steady state condition. The carbonation rate obtainedcannot be considered as a realistic one, because a CO2 concentration of 10% exceedsthe natural CO2 concentration in air (0.03%) by far. To obtain a first estimation of thecorresponding carbonation rate under field conditions from accelerated carbon-ation tests performed at 10% CO2, Audenaert [9] used a conversion formula that ex-presses the ratio of the accelerated and field carbonation coefficients (Aacce andAfield) in terms of their corresponding CO2 concentrations cacce and cfield. The sameformula (Eq. (1)) was also used in this research.

Aacce

Afield¼

ffiffiffiffiffiffiffiffifficaccepffiffiffiffiffiffiffiffiffifficfieldp ð1Þ

Note that Sisomphon and Franke [22] used a very similar conversion formula.Only the CO2 concentration of their accelerated carbonation test was different(3% instead of 10%). Since literature indicates that 3% CO2 could be the maximumallowable CO2 concentration for an accelerated carbonation test (see Section 1),one could conclude that Eq. (1) may not be applicable for all values of cacce. Wetherefore calculated Afield of one HVFA mixture (F(1)50) from both the outcome ofthe carbonation test performed at 10% CO2 and 1% CO2 (see Section 4.3.2).

3.4. Service life prediction

3.4.1. Simplified approachPlotting the measured carbonation depths as a function of the square root of

time usually results in a fairly linear relation. For long, this square-root-time rela-tion for the carbonation rate based on Fick’s second law has been considered asan effective way to estimate roughly when the carbonation front is expected toreach the reinforcing steel and trigger its depassivation. As such, a very basic limitstate function can be defined for carbonation-induced steel depassivation (Eq. (2)):

gða; xcðtÞÞ ¼ a� xcðtÞ ¼ a� Afield

ffiffitp

ð2Þ

with a, the concrete cover (mm), xc(t), the carbonation depth at time t (mm), Afield,the carbonation rate under field conditions (mm/

ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

) and t, the time (years).Nevertheless, it must be said that Eq. (2) provides a rather simplified approach. Asmentioned in Section 3.3.4 the natural carbonation rate Afield is obtained from Eq.(1) which is nothing more than a simple conversion formula to go from the higherCO2 concentration under laboratory conditions to the expected CO2 concentrationon site (usually around 0.03% CO2 by volume). For instance, it disregards the contin-uously changing natural environment of the concrete. Moreover, the beneficial effectof any measures taken to prevent premature desiccation near the concrete surface –also known as curing – on its effective carbonation resistance is not taken into ac-count that way.

3.4.2. Advanced approachFib Bulletin 34 provides a more sophisticated concrete and environment specific

limit state function (Eq. (3)) [5].

gða; xcðtÞÞ ¼ a� xcðtÞ ¼ a�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 � ke � kc � R�1

NAC;0 � CS

q�ffiffitp�WðtÞ ð3Þ

with a, the concrete cover (mm), xc(t), the carbonation depth at time t (mm), ke, theenvironmental function that accounts for the relative humidity in practice (–), kc, theexecution transfer parameter that deals with concrete curing (–), R�1

NAC;0, the inverseeffective carbonation resistance under natural carbonation conditions ((mm2/years)/(kg/m3)), CS, the expected increased atmospheric CO2 concentration with time with-out additional emissions attributed to motorized traffic (kg/m3) and W(t), the weath-er function (–).

Environmental function ke (Eq. (4)) takes into account that the humidity level(RHreal) of the actual concrete environment may differ from the reference relativehumidity (RHref: 60%) imposed during the accelerated carbonation test.

ke ¼1� RHreal

100

� �fe

1� RHref100

� �fe

0B@

1CA

ge

ð4Þ

where fe and ge represent two constant parameters which are independent of theexposure conditions and management phases. Data for RHreal are normally collectedfrom a weather station close to the location of the concrete structure. Since the mainpurpose of this research is an environmental evaluation of a newly developed HVFAconcrete composition which has not been applied yet in practice, a fictional concretestructure assumed to be located near the weather station of Zaventem, Belgium wasconsidered. The daily mean values recorded at this weather station between 1999and 2008 were provided by the Royal Meteorological Institute (KMI). They were usedto estimate the (Beta) distribution for RHreal.

A value for the execution transfer parameter kc (–) that accounts for the curingeffect is obtained from Eq. (5).

P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193 187

kc ¼tc

7

� �bc

ð5Þ

with bc, the exponent of regression (–) and tc, the period of curing (days). Accordingto Fib Bulletin 34 [5], any measures taken to prevent premature desiccation of theconcrete surface (water curing, air curing while covering the concrete surface withsheets, etc.) are seen as ways that guarantee proper curing. In this research, the per-iod of curing corresponded with the period during which the samples were precon-ditioned at 20 �C and 95% RH. Note that factor 1/7 in Eq. (5) may not be valid forevery concrete type. The use of this factor implies that 7 days of optimal curingwould be enough to achieve the highest carbonation resistance possible. The ratherslow pozzolanic FA reaction that takes place in HVFA concrete may require longercuring. This issue will be addressed more thoroughly in Section 4.2.

The value for R�1NAC;0 is normally calculated from the effective inverse carbon-

ation resistance of dry concrete R�1ACC;0, determined at a certain point in time t0 using

the accelerated carbonation test prescribed by Fib Bulletin 34 (Eq. (6)) [5].

R�1NAC;0 ¼ kt � R�1

ACC;0 þ et ð6Þ

with kt, a regression parameter which considers the influence of the test method (–)and et, an error term which takes into account inaccuracies that occur conditionallywhen using the accelerated test method ((mm2/years)/(kg/m3)). Since the acceler-ated carbonation test method used in this research is different from the one pro-

posed by Fib Bulletin 34 [5], R�1NAC;0 cannot be calculated for the tested concrete

compositions by means of Eq. (6). Therefore, we decided to replaceffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

R�1NAC;0 � CS

� �r

in Eq. (3) with Afield, the same load variable that was used for the simplified limitstate function described in Section 3.4.1. To still incorporate the expected increasedatmospheric CO2 concentration with time (cf. Fib Bulletin 34 [5] around 0.0082 kg/m3 or 0.05% CO2 by volume), Afield was calculated from Eq. (1) while assuming0.05% CO2 by volume for cfield.

By means of weather function W(t) (Eq. (7)) the effect of occurring wettingevents such as (driving) rain is included in the limit state function.

W ¼ t0

t

� �w

¼ t0

t

� �ðpSR �ToWÞbw

2

ð7Þ

with t0, the time of reference (years), w, the weather exponent (–), pSR, the probabil-ity of driving rain (–), ToW, the time of wetness (–) and bw, the exponent of regres-sion (–). The probability pSR corresponds with the average distribution of the winddirection during raining events. Its value for vertical elements was estimated fromKMI weather station data measured in Zaventem, Belgium between 1999 and2008. For the same location and time period, the average number of rainy daysper year was used to calculate ToW. According to Fib Bulletin 34, the minimumamount of precipitation for a rainy day is 2.5 mm [5].

Thus, combining Eqs. (4), (5), and (7) into Eq. (3) where Afield replacesffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiR�1

NAC;0 � CS

� �rresults in a limit state function that considers the effect of concrete

curing and the varying relative humidity and weather conditions (Eq. (8)).

gða; xcðtÞÞ ¼ a� xcðtÞ ¼ a�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 � ke � kc

p� Afield �

ffiffitp�WðtÞ ð8Þ

3.4.3. Model input parametersAn overview of the applied distributions and its characterising parameters

(mean, standard deviation, minimum and maximum) for each of the input param-eters to either the simplified and advanced limit state functions are shown inTable 3. The reliability indices (b) and probabilities of failure (Pf) associated withlimit state functions (2) and (8) were calculated using the First Order ReliabilityMethod (FORM) available in the probabilistic Comrel software. In compliance with

Table 3Quantification of the input parameters to the simplified and advanced limit state function

Distribution Mean

aXC3 (mm) Lognormal 35aXC4 (mm) Lognormal 40Afield (mm/

ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

) Normal Table 5RHreal (%) Beta 79RHref (%) Constant 60fe (–) Constant 2.5ge (–) Constant 5.0bc (–) Normal �0.567tc (d) Constant 28/91t0 (years) Constant 0.0767pSR (–) Constant 0.16ToW (–) Constant 0.31bw (–) Normal 0.446

the Fib Bulletin 34 [5], these parameters need to meet the requirements for thedepassivation limit state (b P 1.3 and Pf 6 0.10) to qualify for use in a XC3 orXC4 environment.

3.5. Life cycle assessment

In compliance with ISO 14040, the LCA consisted of four major steps: the defi-nition of goal and scope, the inventory analysis, the impact analysis and theinterpretation.

3.5.1. Definition of goal and scopeThis LCA was conducted to quantify the theoretical reduction in greenhouse gas

emissions that could result from using the proposed HVFA concrete composition in-stead of traditional OPC concrete or FA concrete conforming to the k-value concept.To do this correctly, the LCA study should take into account differences in strengthand durability with the latter concrete types. Therefore, the amount of concreteneeded to construct and maintain an axially loaded column (height: 2.5 m, cross-section: rectangular) carrying a design load of 1500 kN for 100 years was chosenas the functional unit. The experimental strength classes given in Table 1 were usedfor the column design. All calculations regarding concrete and steel reinforcementdimensioning were done in accordance with NBN EN 1992-1-1. In case service life isbelieved to be less than 100 years, additional concrete manufacturing necessary torepair the column is included as well. In this research, it was assumed that a full col-umn repair consisted of a replacement of the entire 35 mm (aXC3) or 40 mm (aXC4)concrete cover. The proposed concrete covers are in agreement with the guidelinesof NBN EN 1992-1-1 for a construction class upgrade to achieve a service life of100 years.

3.5.2. Inventory analysisPer concrete constituent, the necessary inventory data were collected from the

Ecoinvent database [23]. Their proper short descriptions as mentioned in the data-base together with the amounts used to manufacture 1 m3 of each concrete mix-ture, are shown in Table 4.

The required amounts of sand, aggregates, cement, FA, water and superplasti-cizer (SP) for concrete manufacture were assumed to be accurately weighed andtherefore considered as constants. For the allocation of impacts attributed to theindustrial by-product FA, the economic allocation coefficients as proposed by Chenet al. were applied, which is 1.0% of the impact of the coal fired electricity produc-tion corresponding with the production of 1 kg FA [24].

SP inventory data were obtained from an environmental declaration publishedby the European Federation of Concrete Admixture Associations [25]. The transportof each constituent to the concrete plant was not incorporated in the LCA since itsenvironmental impact is always very case specific. The impact of the reinforcingsteel was also not included because each column contained more or less the sameamount of steel.

For all Ecoinvent data, unit processes (U) were used in the modelling of eachconcrete mixture. This was done to enable a full probabilistic uncertainty analysisof the calculated environmental scores using Monte Carlo simulation. This approachaccounts for the small variations in emission values associated with the productionof each concrete constituent and enables the calculation of a standard deviation onthe concrete’s GWP.

3.5.3. Impact analysis and interpretationThe IPCC 2007 GWP 100a impact method was used to calculate the Global

Warming Potential (GWP) expressed in CO2 equivalents for a timeframe of100 years. All calculations were done in the LCA software SimaPro 7.3.3.

s for service life prediction.

Stdv. Lower boundary Upper boundary

8 – –8 – –0.1 – –9 40 100– – –– – –– – –0.024 – –– – –– – –– – –– – –0.163 – –

Table 4Overview of the life cycle inventory data used per m3 of each studied concrete mixture.

Material data (kg) T(0.55) T(0.50) F(1)15 F(2)15 F(1)50 F(2)50

1. Sand, at mine/CH U [23] 715 714 698 698 645 6452. Gravel, round, at mine/CH U [23] 1188 1186 1159 1159 1071 10713. Portland cement, strength class Z 52.5, at plant/CH U [23] 300 320 300.8 300.8 225 2254. Fly asha 0 0 48 48 225 2255. Tap water, at user/CH U [23] 165 160 160 157.5 157.5 157.56.Superplasticizer [25] 0 0 0.8 1.2 2.5 2.5

Processing data (kW h)Electricity, low voltage, production BE, at grid/BE U [23] 3.83 3.83 3.83 3.83 3.83 3.83

a Partially contains Ecoinvent data: ‘Electricity, hard coal, at power plant/BE U [23]’, through economic allocation cf. Chen et al. [24].

188 P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193

4. Results and discussion

4.1. Colorimetrically versus microscopically assessed carbonationdepths

Microscopic measurements usually result in higher carbonationdepths than colorimetric measurements with the phenolphthaleincolour indicator (Fig. 2). This observation holds true for practicallyevery studied concrete mixture after 28 and 91 days of optimalcuring followed by either ±18 weeks of continuous exposure to10% of CO2 at 20 �C and 60% RH (simulated exposure class: XC3)or 10 wet/dry cycles (simulated exposure class: XC4). MixtureF(1)50, optimally cured for 28 days and exposed to the simulatedXC3 environment, was the only exception. Statistical analysis(Independent Sample T tests, Significance level = 0.05) showed thatthe recorded differences in carbonation depth were only non-sig-nificant for the F(1)50 mixtures destined for exposure class XC4,and this for two curing ages (28 and 91 days).

Since the Leica Application Suite v.3.7 software provides neces-sary tools for measuring distances, the microscopic technique for

(a)

(c)

Fig. 2. Difference between colorimetric and microscopic assessed carbonati

measuring the carbonation depth is certainly more precise (accu-racy: ±0.2 mm). This is in contrast with the determination of a phe-nolphthalein colour change boundary to the nearest millimeter bymeans of a simple ruler. Moreover, microscopic measurements al-low for the detection of incipient carbonation and partially carbon-ated areas in the vicinity of larger pores or the more porousinterfacial transition zone of aggregates. As a consequence, onewould expect that a microscopically assessed carbonation depthwould always be characterised by a larger standard deviation. Astatistical homogeneity of variance check (Levene’s test, Signifi-cance level = 0.01) for the microscopic and colorimetric measure-ments would then turn out significant for all concrete mixtures.However, this statement only holds true for mixtures T(0.50) at91 days (XC4), F(1)15 at 28 days and 91 days (XC3, XC4), F(2)15at 28 days and 91 days (XC3, XC4) and F(2)50 at 28 days and91 days (XC4).

An overall comparison between the different concrete mixturesfor exposure class XC3 shows that HVFA compositions F(1)50 andF(2)50 are much more susceptible to carbonation than the concretemixtures F(1)15 and F(2)15 with only 15% FA and OPC reference

(b)

(d)

on depths in simulated XC3 (a and b) and XC4 environments (c and d).

P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193 189

T(0.55) (Fig. 2a and b). This observation follows from both the col-orimetric and microscopic carbonation depth measurements, andthis for the two curing ages. When looking at the results for expo-sure class XC4, the same conclusions can be drawn. Evidently, thecyclic immersion of the concrete in water to simulate a XC4 envi-ronment resulted in considerably lower carbonation depths. Aftereach immersion period the concrete needed to dry out before theCO2 could steadily penetrate the concrete again.

4.2. Colorimetrically measured accelerated and estimated fieldcarbonation coefficients

The accelerated carbonation coefficients Aacce (Table 5) are inline with the individual colorimetric carbonation depth measure-ments at the end of each carbonation test (Fig. 2). For exposureclass XC3, the Aacce values of the HVFA mixtures F(1)50(4.3 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

) and F(2)50 (3.6 ± 0.1 mm/ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

) arearound three times the corresponding carbonation coefficient ofOPC reference T(0.55) (1.3 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

) after 28 days of opti-mal curing prior to exposure. Prolonged optimal preconditioningfor 91 days resulted in an important decrease of the carbonationcoefficient for HVFA mixtures F(1)50 (2.4 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

, –44%) and F(2)50 (2.3 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

, – 36%). This time depen-dent decrease was less pronounced for the OPC reference(1.0 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

, – 23%), which indicates that a considerablepart of OPC hydration had already taken place by the age of28 days. On the other hand, the hydration process of HVFA con-crete takes longer, because a sufficient amount of Ca(OH)2, a hydra-

Table 5Accelerated carbonation coefficients (Aacce) and estimated field carbonation coefficients (A

Class Mix Aacce (mm/ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

)

28 d R2 91 d

Colorimetric measurementsXC3 T(0.55) 1.3 ± 0.1 0.70 1.0 ± 0.1

F(1)15 0.0 – 0.0F(2)15 0.0 – 0.0F(1)50 4.3 ± 0.1 0.99 2.4 ± 0.1F(2)50 3.6 ± 0.1 0.98 2.3 ± 0.1

XC4 T(0.50)a,b 0.0 – 0.0F(1)15a,b 0.0 – 0.0F(2)15a,b 0.0 – 0.0F(1)50a 1.0 ± 0.1 0.97 0.9 ± 0.1F(1)50b 1.4 ± 0.1 0.98 1.2 ± 0.1F(2)50a 0.9 ± 0.1 0.95 0.7 ± 0.1F(2)50b 1.4 ± 0.1 0.94 0.9 ± 0.1

Class Mix Aacce (mm/ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

)

28 d 91 d

Microscopic measurementsXC3 T(0.55) 1.6 ± 0.5 –

F(1)15 0.7 ± 0.2 0.5 ± 0.3F(2)15 0.2 ± 0.1 0.3 ± 0.2F(1)50 3.8 ± 0.4 3.4 ± 0.5F(2)50 3.9 ± 0.7 3.0 ± 0.5

XC4 T(0.50)a – 0.6 ± 0.4T(0.50)b – 0.8 ± 0.6F(1)15a 0.2 ± 0.2 0.3 ± 0.2F(1)15b 0.2 ± 0.1 0.4 ± 0.2F(2)15a 0.2 ± 0.1 0.3 ± 0.2F(2)15b 0.3 ± 0.1 0.4 ± 0.2F(1)50a 1.0 ± 0.3 0.9 ± 0.4F(1)50b 1.5 ± 0.4 1.2 ± 0.5F(2)50a 1.0 ± 0.4 1.0 ± 0.5F(2)50b 1.5 ± 0.6 1.4 ± 0.6

a Aacce values were calculated with inclusion of the immersion period in water.b Aacce values were calculated with exclusion of the immersion period in water.

tion product of the remaining OPC, needs to be present first beforethe FA can start to react. The further densification of the micro-structure due to the pozzolanic FA reaction is therefore expectedat later age. The significantly improved carbonation resistanceafter 91 days as reported in Table 5 seems to demonstrate this phe-nomenon. However, when looking at the Aacce values obtained after182 days of optimal preconditioning, a slightly increased carbon-ation coefficient was recorded for mixtures F(1)50 (2.7 ± 0.1 mm/ffiffiffiffiffiffiffiffiffiffiffiffiffiffi

weeksp

) and F(2)50 (2.6 ± 0.1 mm/ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

). This is in contrastwith the corresponding carbonation coefficient of OPC referenceT(0.55) which had decreased further on to 0.6 ± 0.1 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

.Gruyaert et al. also observed either a slight increase or decreasein carbonation rate between 3 months (�91 days) and 6 months(�182 days) of optimal sample preconditioning for concrete inwhich 50%, 70% or 85% of the cement was replaced with BFS[10]. An additional carbonation test performed by Gruyaert et al.after 18 months of curing resulted again in a carbonation coeffi-cient similar to the one measured after 91 days. For BFS concrete91 days is therefore seen as the optimal curing period beyondwhich the concrete’s carbonation resistance will not improve any-more [10]. A similar time span is expected for the studied HVFAconcrete compositions and OPC reference T(0.55), although anadditional carbonation test at later age would also be advised tofind further confirmation for this statement. Under the assumptionthat 91 days is indeed the optimal curing period, the expressionused in the advanced limit state function for service life prediction(Section 3.4.2) to account for the concrete’s curing behaviour (Eq.(5)) was modified: Factor 1/7 (which assumes that 7 days of opti-

field) as measured colorimetrically and microscopically.

Afield (mm/ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

)

R2 182 d R2 28 d 91 d

0.71 0.6 ± 0.1 0.97 0.7 ± 0.1 0.5 ± 0.1– 0.0 – 0.0 0.0– 0.0 – 0.0 0.00.92 2.7 ± 0.1 0.98 2.2 ± 0.1 1.3 ± 0.10.93 2.6 ± 0.1 0.94 1.8 ± 0.1 1.2 ± 0.1

– 0.0 – 0.0 0.0– 0.0 – 0.0 0.0– 0.0 – 0.0 0.00.95 1.0 ± 0.1 0.90 0.5 ± 0.1 0.5 ± 0.10.96 1.4 ± 0.1 0.90 0.7 ± 0.1 0.6 ± 0.10.85 0.7 ± 0.1 0.97 0.5 ± 0.1 0.4 ± 0.10.86 1.0 ± 0.1 0.97 0.7 ± 0.1 0.5 ± 0.1

Afield (mm/ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

)

28 d 91 d

0.9 ± 0.3 –0.4 ± 0.1 0.3 ± 0.20.1 ± 0.1 0.2 ± 0.11.9 ± 0.2 1.7 ± 0.32.0 ± 0.4 1.6 ± 0.3

– 0.3 ± 0.3– 0.4 ± 0.30.1 ± 0.1 0.2 ± 0.10.1 ± 0.1 0.2 ± 0.10.1 ± 0.1 0.2 ± 0.10.2 ± 0.1 0.2 ± 0.10.6 ± 0.2 0.5 ± 0.20.8 ± 0.2 0.6 ± 0.30.6 ± 0.2 0.5 ± 0.30.8 ± 0.3 0.8 ± 0.3

190 P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193

mal curing would be sufficient) was replaced with factor 1/91. Themodified formula was applied for all studied concrete mixtures.

Note that the R2 values that correspond with the carbonationcoefficients of the HVFA mixtures are closer to unity (R2: 0.92–0.99) than the ones reported for mixture T(0.55) after 28 and91 days (R2: 0.70–0.71). This can be explained by the fact that amuch more well-defined carbonation front existed for the HVFAmixtures after applying the phenolphthalein colour indicator. Thecolorimetric determination of the carbonation front for the OPCreference T(0.55) was more difficult because the colour changeboundary on the sample saw-cuts was less clear. In other words,the samples of the latter OPC mixture were characterised by a par-tially carbonated zone. The presence of this zone made it difficultto determine the exact location of colour transition. As a conse-quence, plotting the carbonation depths as a function of the squareroot of time resulted in a less pronounced linear relation.

With respect to the FA mixtures conforming to the k-value con-cept of NBN B15-001, no carbonation was recorded using the col-orimetric measuring technique. As a result, the acceleratedcarbonation coefficient of mixtures F(1)15 and F(2)15 is seemingly0.00 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiweeksp

. However, the microscopic measuring techniquerevealed that these two concrete mixtures indeed showed signs of– be it limited – carbonation (Fig. 2). The microscopic carbonationmeasurements obtained after around 18 weeks of exposure to 10%CO2 were also used for the calculation of an accelerated carbon-ation coefficient. Since only one exposure period (±18 weeks)was considered, linear regression could not be used for this pur-pose. A rudimentary carbonation coefficient was obtained by sim-ply dividing the measured carbonation depth (Fig. 2) byffiffiffiffiffiffi

18p

weeks. Logically, the corresponding single measurement stan-dard deviations are considerably higher than the standard devia-tions on the colorimetric carbonation coefficients obtained bymeans of linear regression. Nevertheless, the same technique wasapplied to come up with a microscopic carbonation coefficientfor the OPC reference and the HVFA mixtures. Comparison of thetwo measuring techniques shows that apart from the F(1)50 mix-ture tested at 28 days, all mixtures were characterised by slightlyhigher microscopic carbonation coefficients.

With respect to the results for exposure class XC4, more or lessthe same conclusions are valid as for exposure class XC3. Logically,the carbonation coefficients obtained for the former exposure classare much lower relative to the ones obtained for the latter. This isnot only the case when the immersion period in water is included(a), but also when it is excluded (b) from the exposure time. Appar-ently, the amount of water absorbed by keeping the specimenssubmerged every two weeks is not sufficiently released again fromthe concrete during each subsequent week of exposure to 10%. As aresult, penetration of CO2 was hindered substantially. This explainswhy the Ab

acce values of HVFA mixtures F(1)50 and F(2)50 for expo-sure class XC4 still differ significantly from their correspondingAacce values measured in exposure class XC3. When based uponcolorimetric measurements, the former coefficients are about 32–49% lower than the latter. The microscopic measuring techniqueresulted in very similar percentages (34–47%). However, the latterpercentages are not valid for mixtures F(1)15 and F(2)15. Some-times, their low carbonation coefficients recorded for exposureclass XC4 were even a little higher than the ones recorded for expo-sure class XC3 (e.g. F(2)15 at 91 days).

4.3. Service life prediction

4.3.1. Simplified approachWhen looking at exposure class XC3, HVFA concrete mixtures

F(1)50 and F(2)50 would be characterised by a much shorter timeto carbonation-induced steel depassivation (as estimated from Eq.(2)) than OPC reference T(0.55) and the FA containing concrete

mixtures F(1)15 and F(2)15 conforming to the k-value concept (Ta-ble 6). However, for the colorimetric and microscopic carbonationassessment methods that have been experimentally applied in thisresearch, the time to steel depassivation would still easily exceedthe predefined service life of 100 years, and this after both 28 days(F(1)50: 132–176 years, F(2)50: 196–159 years) and 91 days(F(1)50: 371–219 years, F(2)50: 433–247 years) of optimal curing.

The very low carbonation rates for all mixtures and both curingperiods, as obtained under the laboratory conditions that shouldrepresent exposure class XC4, were responsible for the generallyapplicable time to depassivation of at least 1000 years (the upperlimit of the timeframe considered in Comrel). Thus, a predefinedservice life of 100 years would not be a problem in this simulatedXC4 environment. However, this statement is not necessarily validfor exposure class XC4 in general. As already pointed out in Sec-tion 4.2, regular immersion of the concrete in water every 2 weeksconsiderably hindered further penetration of CO2. Concrete car-bonation rates in environments characterised by more irregularwet/dry cycles and longer drying periods, could easily be muchhigher and result in faster depassivation of the embedded reinforc-ing steel.

4.3.2. Advanced approachApplying the more sophisticated limit state function (Eq. (8))

gives a totally different outcome of the estimated service lives.Regardless the exposure class, the concrete composition, the ap-plied curing period and the carbonation assessment method usedfor determining the carbonation coefficient, the estimated timeto steel depassivation would always be more than 1000 years (Ta-ble 6). To see which model input parameter postpones steeldepassivation the most, only one parameter at a time (ke, kc orW(t)) was considered in Eq. (8), while the other two were tempo-rarily removed from it. The resulting series of service life predic-tions – three in total – was performed for HVFA mixture F(1)50,cured for 28 days and then assumed to be located in a XC3 environ-ment (aXC3 = 35 mm, Afield colorimetric = 2.2 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

). When onlydepending on model input parameter ke, kc or W(t), the estimatedtime to steel depassivation would be 876 years, 33 years and310 years, respectively. Thus, the low ke value due to the high rel-ative humidity (79 ± 9%) inherent to the Belgian climate (Zaven-tem, 1999-2008, source: KMI) is mainly causing the long steeldepassivation period when estimated with the advanced limitstate function (Eq. (8)). The low W(t) value that accounts for thetime of wetness (ToW = 0.31) and the probability of driving rain(pSR = 0.16) for the same location and time period, also substan-tially postpones steel depassivation. The curing factor kc on theother hand accelerates the mechanism substantially, and thismainly due to the fact that HVFA mixture F(1)50 requires longeroptimal curing (91 days). However, when all three parameters(ke, kc and W(t)) are combined in Eq. (8), the latter effect is com-pletely compensated by the values of ke and kc and the resultingtime to depassivation exceeds 1000 years. Based on this combinedFib Bulletin 34 [5] approach, one would expect that repair of struc-tures made of either of the studied concrete compositions wouldnot need to be feared within the first 100 years after construction.This can mainly be attributed to the fact that the expected meteo-rological conditions of the environment in which the concretewould be used do not favour carbonation-induced steeldepassivation.

Apart from the above mentioned parameters, there may be oneother cause for the rather long depassivation periods obtained,namely the experimental origin of the field carbonation coefficientAfield. This value was calculated by means of Eq. (1) from the accel-erated carbonation coefficient Aacce obtained after exposing theconcrete to 10% CO2. As already mentioned in Section 1, Castelloteet al. believe that the maximum CO2 concentration during an accel-

Table 6Influence of the underlying carbonation assessment method and limit state function for service life prediction on the estimated times to steel depassivation for the studiedconcrete compositions.

Simplified

XC3

28 days T(0.55) F(1)15 F(2)15 F(1)50 F(2)50

Colorimetric >1000 – – 132 196Microscopic 750 >1000 >1000 176 159

91 days T(0.55) F(1)15 F(2)15 F(1)50 F(2)50

Colorimetric >1000 – – 371 433Microscopic – >1000 >1000 219 247

XC4

28 days T(0.50)a F(1)15a F(2)15a F(1)50a F(2)50a

Colorimetric – – – >1000 >1000Microscopic – >1000 >1000 >1000 >1000

91 days T(0.50)a F(1)15a F(2)15a F(1)50a F(2)50a

Colorimetric – – – >1000 >1000Microscopic >1000 >1000 >1000 >1000 >1000

Advanced

XC3

28 days T(0.55) F(1)15 F(2)15 F(1)50 F(2)50

Colorimetric >1000 – – >1000 >1000Microscopic >1000 >1000 >1000 >1000 >1000

91 days T(0.55) F(1)15 F(2)15 F(1)50 F(2)50

Colorimetric >1000 – – >1000 >1000Microscopic – >1000 >1000 >1000 >1000

XC4

28 days T(0.50)a F(1)15a F(2)15a F(1)50a F(2)50a

Colorimetric – – – >1000 >1000Microscopic – >1000 >1000 >1000 >1000

91 days T(0.50)a F(1)15a F(2)15a F(1)50a F(2)50a

Colorimetric – – – >1000 >1000Microscopic >1000 >1000 >1000 >1000 >1000

a Times to depassivation were calculated with inclusion of the immersion period in water.

P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193 191

erated carbonation test should be no more than 3% [13]. For thisreason, the applicability of Eq. (1) for carbonation tests that use aCO2 concentration of 10% was experimentally assessed on an addi-tional batch of HVFA concrete mixture F(1)50. Exposure to either10% or 1% CO2 after 28 days of curing obviously resulted in signif-icantly different accelerated carbonation coefficients Aacce. Usingeither Aacce_1% or Aacce_10% as input to Eq. (1) did not result in similarfield carbonation coefficients corresponding with a CO2 concentra-tion of 0.05%. Afield_10% to 0.05% was found to be only 55% of Afield_1% to

0.05%. If exposure to only 1% CO2 indeed does not alter the naturalcarbonation process in an unrealistic way, then this means thatexposing HVFA concrete to 10% CO2 seriously underestimates theactual Afield. Therefore, the colorimetric Afield value of F(1)50 at28 days (2.2 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

) was divided by 55% to have the highesttheoretical field carbonation coefficient possible for this concretemixture (4.0 mm/

ffiffiffiffiffiffiffiffiffiffiffiffiyearsp

) with the colorimetric test method. Afterimplementation of this value in the advanced limit state function(Eq. (8)), the estimated time to steel depassivation is still morethan 1000 years. Apparently, the role of the unfavourable Belgianmeteorological conditions for carbonation-induced steel depassi-vation remains dominant.

It must be recognized that the service life predictions were con-ducted under the assumption that the applied concrete cover wasalways in agreement with limiting values imposed by NBN EN1992-1-1 for a service life of 100 years (aXC3: 35 mm, aXC4:40 mm) and that the concrete was entirely free of defects (cracks,

etc.). Evidently, these optimal conditions cannot always be guaran-teed. Moreover, 28 days of optimal curing or even more is seldomlyused in practice. As a consequence, the studied HVFA mixtures,which are obviously more susceptible to carbonation than OPCconcrete and FA concrete conforming to the k-value concept, mayindeed be characterised by a time to carbonation-induced steeldepassivation <100 years due to the presence of flaws. Carefulinterpretation of theoretical service life predictions is thereforestill advised.

4.4. Environmental consequences in terms of global warming potential

The dimensions and volumes associated with initial construc-tion of the proposed functional unit (column with design load:1500 kN, height: 2.5 m, cross-section: rectangular) for each studiedconcrete composition, together with the corresponding globalwarming potentials (GWPs) are shown in Fig. 3. In case the fieldcarbonation rate of the HVFA mixtures for exposure class XC3would be estimated from colorimetric or microscopic measure-ments and then be used as input in a simplified service life predic-tion model (Eq. (2)), no repair would be needed within thepredefined service life of 100 years. This statement remains trueeven after correcting the (underestimated) field carbonation rateassociated with accelerated carbonation testing in an atmosphereconsisting of 10% CO2 (see Section 4.3.2). As a result, the GWP ofF(1)50 (40.7 ± 4.4 kg CO2eq) and F(2)50 (45.5 ± 4.5 kg CO2eq) would

Column dimensions (m²) 300 × 295 300 × 197 300 × 197 300 × 197 300 × 222 300 × 253 Column volume (m³) 0.221 0.148 0.148 0.148 0.167 0.190 Repair volume (m³) 0.037 0.033 0.030 0.030 0.032 0.034

Fig. 3. Global warming potentials (GWPs) associated with the amount of concrete needed to construct and maintain an axially loaded column for 100 years.

192 P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193

be around 27% and 18% less when compared with the GWP of theOPC reference T(0.55) (55.6 ± 7.7 kg CO2eq) for exposure class XC3.The difference in environmental benefit between the two mixturescan mainly be attributed to the fact that a column made of mixtureF(2)50 would be characterised by larger dimensions due to its low-er strength class (C35/45 instead of C40/50). However, there is stillan environmental benefit of 18% because this strength class is stillhigher than the one of OPC reference T(0.55) (C30/37). On the otherhand, if the FA containing mixtures conforming to the k-value con-cept (F(1)15 and F(2)15) would be seen as the proper compositionof reference, then there would be no environmental benefit at allfor HVFA mixtures F(1)50 and F(2)50. On the contrary, their GWPwould even increase with around 3% and 15%, respectively. Thiscan be explained by the still considerable cement content(225 kg/m3) of the HVFA mixture as well as by their lower strengthclass in comparison with experimental strength class of F(1)15 andF(2)15 (both C45/55).

In case service life would be less than 100 years due to defects,inadequate curing or an insufficient concrete cover on top of therebars, the GWP of the additional concrete manufacturing requiredfor column repair would need to be considered as well. One repairwithin the proposed 100 year time period would result in a totalGWP of 48.3 ± 5.1 kg CO2eq for composition F(1)50 which is only13% less than the one recorded for the repair-free OPC referenceT(0.55) and 22% more than the FA mixtures conforming to the k-va-lue concept. For composition F(2)50, the GWP values would be 4%less and 35% more than the ones of mixtures T(0.55) and F(2)15,respectively. When including two repairs, the resulting totalGWP value for mixture F(1)50 would amount to 56.0 ± 5.7 kg CO2eq

which is around 1% and 41% higher in value in comparison withT(0.55) and F(1)15, respectively. With respect to mixture F(2)50,these percentages would amount to 11% and 55%, respectively.Thus, in case the carbonation resistance in the field would turnout unsatisfactory for some reason and one or two repairs wouldbe imperative, the studied HVFA composition cannot be consideredas beneficial from an environmental viewpoint. For the repair-freeHVFA concrete, both colorimetric and microscopic carbonationassessment results in the important environmental benefit of 18–27%, but only when compared with the OPC reference. This envi-ronmental benefit also holds true if the more advanced limit statefunction (Eq. (8)) would be used for service life prediction pur-poses. When taking into account the concrete’s curing behaviourand the expected meteorological conditions of the environmentin which it would be used, no repair would ever be needed, evenif the field carbonation rate would be estimated with the most crit-

ical carbonation assessment technique and after correction of itsconsiderable underestimation associated with exposing the con-crete to 10% CO2 (see Section 4.3.2).

Although the recorded carbonation rates in the environmentthat simulated exposure class XC4 conditions were always verylow, even for F(1)50 and F(2)50, there would be no environmentalbenefit of using the studied HVFA mixtures in practice. Both thesimplified and more advanced service life prediction indicated thatno repair would be necessary within the 100 year time span,regardless the underlying (colorimetric or microscopic) carbon-ation assessment method. However, the resulting GWPs of F(1)50and F(2)50 need to be compared with the GWP of the appropriateOPC reference for exposure class XC4. Since the experimentalstrength class of reference T(0.50) (C45/55) exceeded the strengthclasses of both F(1)50 (C40/50) and F(2)50 (C35/45), the dimen-sions of a column made of T(0.50) would naturally be smaller. Asa result, the GWPs of F(1)50 and F(2)50 are around 1% and 13%higher than the 40.4 kg CO2eq that can be assigned to the T(0.50)column. If mixtures F(1)15 and F(2)15 would be seen as the com-positions of reference, then the GWPs of F(1)50 and F(2)50 wouldbe around 3% and 15% higher. Thus, the concrete’s strength is akey governing factor with respect to the environmental perfor-mance of a concrete, especially when the durability performanceof the material in the studied environment is not so much a criticalissue. One should definitely try to achieve a strength class for apotentially ‘green’ concrete composition, such as HVFA concrete,which is similar to the one of the composition of reference, and thiswith the lowest cement content possible. Otherwise the decreasein GWP will be less substantial, and in the worst case even non-existing. Note that this finding does not necessarily hold true forevery concrete environment. When exposed to chlorides, servicelife of the same HVFA concrete compositions easily exceeds thepredefined timespan of 100 years, which is much higher than the30–60 years time range for the corresponding OPC reference[26,27]. In that case, the GWP of the HVFA concrete is much moregoverned by its better durability performance than by its lowerstrength class.

5. Conclusions

To account for the expected durability performance of a poten-tially ‘green’ HVFA concrete composition when calculating its GWPin an environment with exposure to carbonation-induced steeldepassivation, a reliable notion of its service life is imperative. A

P. Van den Heede, N. De Belie / Construction and Building Materials 55 (2014) 183–193 193

general shift from a simplified to a more advanced service lifeassessment can have the following consequences.

(i) The more accurate microscopically assessed carbonationdepth is usually higher than the corresponding colorimetriccarbonation depth. The resulting difference in field carbon-ation rate certainly affects the time to steel depassivationwhen using the simplified limit state function for service lifeprediction. In this case study it can decrease with 37–186 years. Nevertheless, the estimated service life remainshigher than the predefined 100 years.

(ii) The more advanced service life prediction cf. Fib Bulletin 34that takes into account the concrete’s curing behaviour andexpected meteorological conditions, indicated that the timeto steel depassivation for the studied HVFA concrete compo-sition would always exceed 1000 years regardless the under-lying carbonation assessment method. This is mainlyattributed to the unfavourable weather conditions (e.g. rela-tive humidity: 79 ± 9%, time of wetness: 0.31) for carbonationinherent to the Belgian climate. The curing behaviour of HVFAconcrete certainly has an important adverse effect on this, butnot to an extent that it can compensate for the meteorologicaleffects.

(iii) The field carbonation rates that resulted from exposing theHVFA concrete to 10% CO2 at 20 �C and 60% RH are only55% of the field carbonation rates that were obtained afterits exposure to 1% CO2 at 20 �C and 60% RH. Thus, the formertest procedure results in an important underestimation ofthe actual carbonation rate in the field. However, when thisphenomenon is taken into account when performing anadvanced service life prediction, the estimated time to steeldepassivation of the HVFA concrete still easily exceeds100 years.

(iv) The global warming potential of the considered strength anddurability related functional unit for life cycle assessment(axially loaded column, design load: 1500 kN, service life100 years) of the studied HVFA composition is 18–27% lowerthan the GWP of the OPC reference for exposure class XC3when no repairs are needed. If located in an XC4 environ-ment, there would be no environmental benefit at all, mainlybecause the studied HVFA concrete is characterised by alower experimental strength class for column design in com-parison with the applicable OPC reference. The very low fieldcarbonation rates in exposure class XC4 cannot compensatefor the strength effect. In case a FA containing concrete com-position conforming to the k-value concept would be seen asthe proper composition of reference, there would also be noenvironmental benefit because of this strength effect.

Acknowledgements

The authors would like to thank Ghent University for the finan-cial support. Acknowledgements also go the Belgian Building Re-search Institute (BBRI) for the opportunity to use theirequipment, as well as to the Royal Meteorological Institute (KMI)for providing the weather data.

References

[1] Van den Heede P, De Belie N. Environmental impact and life cycle assessment(LCA) of traditional and ‘green’ concretes: literature review and theoreticalcalculations. Cem Concr Compos 2012;34(4):431–42.

[2] RILEM CPC-18. Measurement of hardened concrete carbonation depth. MaterStruct 1988;21(6):453–5.

[3] Curtil L, Gielly J, Murat M. The polarizing microscope: a tool of interest forinvestigation on concrete; application to carbonation. Cem Concr Res1993;23(2):329–34.

[4] Jakobsen NN. A discussion of the paper ‘‘The polarizing microscope: A tool ofinterest for investigation on concrete; application to carbonation’’ by L. Curtil,J. Gielly and M. Murat. Cem Concr Res 1994;24(3):593–4.

[5] CEB-FIP Fib Bulletin 34. Model code for service life design. The InternationalFederation for structural concrete. Task Group 5.6. Lausanne (Switzerland);2006.

[6] Lo Y, Lee HM. Curing effects on carbonation of concrete using aphenolphthalein indicator and Fourier-transform infrared spectroscopy.Build Environ 2002;37(5):507–14.

[7] NT Build 357. Nordtest method, concrete, repairing materials and protectivecoating: carbonation resistance. Espoo (Finland); 1989.

[8] Borges PHR, Costa JO, Milestone NB, Lynsdale CJ, Streatfield RE. Carbonation ofCH and C–S–H in composite cement pastes containing high amounts of BFS.Cem Concr Res 2010;40(2):284–92.

[9] Audenaert K. Transport mechanisms of self-compacting concrete related tocarbonatation and chloride penetration. PhD thesis, Ghent, Ghent University;2006 [in Dutch].

[10] Gruyaert E, Van den Heede P, De Belie N. Carbonation of slag concrete: effect ofthe cement replacement level and curing on the carbonation coefficient –effect of carbonation on the pore structure. Cem Concr Compos2013;35(1):39–48.

[11] Chang CF, Chen JW. The experimental investigation of concrete carbonationdepth. Cem Concr Res 2006;36(9):1760–7.

[12] Rozière E, Loukili A, Cussigh F. A performance based approach for durability ofconcrete exposed to carbonation. Constr Build Mater 2007;23(1):190–9.

[13] Castellote M, Fernandez L, Andrade C, Alonso C. Chemical changes and phaseanalysis of OPC pastes carbonated at different CO2 concentrations. MaterStruct 2009;42(4):515–25.

[14] Groves GW, Brough A, Richardson IG, et al. Progressive changes in thestructure of hardened C3S cement pastes due to carbonation. J Am Ceram Soc1991;74(11):2891–6.

[15] Thiery M, Villain G, Dangla P, Platret G. Investigation of the carbonation frontshape on cementitious materials: effects of the chemical kinetics. Cem ConcrRes 2007;37(7):1047–58.

[16] Chen JJ, Thomas JJ, Jennings HM. Decalcification shrinkage of cement paste.Cem Concr Res 2006;36(5):801–9.

[17] da Silva FG, Helene P, Castro-Borges P, Liborio JBL. Sources of variations whencomparing concrete carbonation results. J Mater Civ Eng 2009;21(7):333–42.

[18] Saetta AV, Vitaliani RV. Experimental investigation and numerical modeling ofcarbonation process in reinforced concrete structures, Part I: Theoreticalformulation. Cem Concr Res 2004;34(4):571–9.

[19] Malhorta VM, Mehta PK. High performance, high-volume fly ash concrete:materials, mixture proportioning, properties, construction practice, and casehistories. 2nd edition. Ottawa: Supplementary Cementing Materials forSustainable Development Inc.; 2005.

[20] General Chemistry. Senese F. Frostburg: Frostburg State University, Dept. ofChemistry; c1997–2010. <http://antoine.frostburg.edu/chem/senese/101/acidbase/indicators.shtml> [updated 2010 February 15; cited 2014 January 2].

[21] Van den Heede P, Furniere F, De Belie N. Influence of air entraining agents ondeicing salt scaling resistance and transport properties of high-volume fly ashconcrete. Cem Concr Compos 2013;37(3):293–303.

[22] Sisomphon K, Franke L. Carbonation rates of concretes containing high volumeof pozzolanic materials. Cem Concr Res 2007;37(12):1647–53.

[23] Frischknecht R, Jungbluth N, editors. Overview and methodology, final reportEcoinvent v2.0 No. 1. St-Gallen: Swiss Centre for Life Cycle Inventories; 2007.

[24] Chen C, Habert G, Bouzidi Y, Jullien A, Ventura A. LCA allocation procedureused as an incitative method for waste recycling: an application to mineraladditions in concrete. Resour Conserv Recy 2010;54(12):1231–40.

[25] EFCA.info. EFCA: European Federation of Concrete Admixture Associations;c2010. Environmental declaration superplasticizing admixtures. <http://www.efca.info/publications.html> [updated April 2006; cited 2013 May 26].

[26] Van den Heede P, Maes M, Gruyaert E, De Belie N. Full probabilistic service lifeprediction and life cycle assessment of concrete with fly ash and blast-furnaceslag in a submerged marine environment: a parameter study. Int J Environ SustDev 2012;11(1):32–49.

[27] Van den Heede P, Maes M, Caspeele R, De Belie N. Chloride diffusion tests asexperimental basis for full probabilistic service life prediction and life-cycleassessment of concrete with fly ash in a submerged marine environment. In:Strauss A, Frangopol DM, Bergmeister K, editors. IALCCE 2012: proceedings ofthe third international symposium on life-cycle civil engineering. Leiden: CRCPress; 2012. p. 913–20.