The stability of bound chlorides in cement paste with sulfate attackJian Geng a ,b ,⁎,Dave Easterbrook b ,Long-yuan Li b ,Li-wei Mo aa Research Center of Green Building Materials and Waste Resources Reuse,Ningbo Institute of Technology,Zhejiang University,China bSchool of Marine Science and Engineering,University of Plymouth,UKa b s t r a c ta r t i c l e i n f o Article history:Received 10July 2014Accepted 25November 2014Available online 27December 2014Keywords:Sulfate attack (C)Bound chlorides (D)Stability (C)Fly ash (D)Ground granulated blast-furnace slag (D)This paper presents an experimental investigation on the stability of bound chlorides in chloride-contaminated cement pastes with and without FA/GGBS when subjected to Na 2SO 4and MgSO 4attack.It is shown that bound chlorides were released in the chloride-contaminated pastes when exposed to Na 2SO 4or MgSO 4solution.This is mainly attributed to the decomposition of Friedel's salt (FS),where Cl −bound in FS is replaced by SO 42−.How-ever there were fewer released chlorides found in the pastes exposed to MgSO 4solution than in those exposed to Na 2SO 4solution.This is partly due to the low pH in the pore solution and partly due to the blocking effect of brucite on ionic transport caused by MgSO 4.The inclusion of FA/GGBS in concrete can increase the decomposition of FS and thus the release of bound chlorides.However,it also resists the penetration of Na 2SO 4and thus reduces the attack of Na 2SO 4.©2014Elsevier Ltd.All rights reserved.1.IntroductionThe corrosion of reinforcing steel in concrete structures,due to chlo-ride ion contamination,is one of the main reasons for the deterioration of concrete structures.There are two forms of chloride ions in concrete.One is free chlorides and the other is bound chlorides.It is well-known that the corrosion of reinforcing steel is mainly induced by the free chlo-rides,so reducing free chlorides by increasing bound chlorides will be bene ficial to the durability of concrete structures.According to the bind-ing mechanism,chloride ions can be bound through chemical reactions and physical absorption.In the former,chloride ions are mainly bound in Friedel's salt (FS)(3CaO·Al 2O 3·CaCl 2·10H 2O)through hydration reactions between chloride ions,tricalcium aluminate (C 3A)and its hydration products.In the latter,chloride ions are mainly absorbed by calcium silicate hydrate (C –S –H gel).It was reported that the formation of bound chlorides could be affected by a multitude of factors such as the quantity of C 3A in cement,supplementary cementitious materials (SCM),alkalinity of pore solution,Ca/Si and Ca/Al of hydration products,chloride salt type,and service condition of concrete structures [1–5].In summary,the chloride binding capacity of concrete can be improved by using SCM or cement with high C 3A content.However,many researchers have identi fied that the stability of bound chlorides,espe-cially of FS,can be affected by pH,carbonation,and chemical erosion [6–9].Sulfate attack is another problem for the durability of concrete struc-tures.The attack of sodium sulfate (Na 2SO 4)and magnesium sulfate (MgSO 4)on concrete is a common phenomenon.The mechanisms of Na 2SO 4and MgSO 4attack on concrete are different,mainly due to the solubility of phases formed with sodium and magnesium ions [10–12].With regard to Na 2SO 4attack,the deterioration of concrete is attributed to the formation of expansion products such as gypsum (CaSO 4·2H 2O)and secondary ettringite (AFt)(3CaO·Al 2O 3·3CaSO 4·32H 2O)according to the following equations:Ca ðOH Þ2þNa 2SO 4þ2H 2O →CaSO 4·2H 2O þ2NaOHð1Þ3ðCaSO 4·2H 2O Þþ3CaO ·Al 2O 3þ26H 2O →3CaO ·Al 2O 3·3CaSO 4·32H 2Oð2Þ2ðCaSO 4·2H 2O Þþ3CaO ·Al 2O 3·CaSO 4·12H 2O þ16H 2O →3CaO ·Al 2O 3·3CaSO 4·32H 2O :ð3ÞWhereas for MgSO 4attack,the transformation of the cementitious C –S –H gel to the non-cementitious magnesium silicate hydrate mush (M –S –H),which has very little strength,is the main reason for the dete-rioration of concrete,although gypsum and secondary AFt are also formed during the attack.In addition,brucite,i.e.Mg(OH)2,will form when magnesium is present in the pore solution,which has low solubil-ity and could densify the pore system and thus affect the transport ofCement and Concrete Research 68(2015)211–222⁎Corresponding author.E-mail address:gengjian@ (J.Geng)./10.1016/j.cemconres.2014.11.0100008-8846/©2014Elsevier Ltd.All rightsreserved.Contents lists available at ScienceDirectCement and Concrete Researchj o u rn a l h o m e p a g e :h t tp ://e e s.e l s e v i e r.c o m /C EM C O N /d e f a u l t.a s pions in the cement paste.The mechanism of MgSO4attack occurs according to the following equations:CaðOHÞ2þMgSO4þ2H2O→CaSO4·2H2OþMgðOHÞ2ð4Þx CaO·y SiO2·z H2Oþx MgSO4þð3xþ0:5y−zÞH2O→xðCaSO4·2H2OÞþx MgðOHÞ2þ0:5yð2SiO2·H2OÞð5Þ4MgðOHÞ2þSiO2·nH2O→4MgO·SiO2·8:5H2Oþðn−4:5ÞH2O:ð6ÞIn fact,sulfate attack and chloride contamination are often found to coexist in concrete structures which are exposed to marine and saline environments.The effects of the sulfate and chloride on a concrete structure's durability are multifaceted.On the one hand,the existence of sulfate,especially of Na2SO4,inhibits the formation of FS and reduces the quantity of bound chlorides[13–15].On the other hand,the exis-tence of chloride ions is beneficial for the resistance of concrete to Na2SO4and MgSO4attack[15–18].However,Baghabra argued that the effect of chloride ions on MgSO4attack was slight because the trans-formation of cementitious C–S–H gel to non-cementitious M–S–H was not affected by chloride ions[19].Despite the work on the interaction of sulfate and chloride in con-crete mentioned above,there is very little work on the effect of sulfate attack on the stability of bound chlorides in concrete.Brown and Badger investigated the distributions of bound sulfates and chlorides infield concrete cores exposed to mixed NaCl,Na2SO4and MgSO4attack. They found that there was extensive AFt in the absence of a gypsum zone for some concrete cores[20].Xu et al.obtained similar results, i.e.that sulfate attack could lead to the release of bound chlorides[21]. Both studies suggested the transformation of FS to AFt due to sulfate attack,but the mechanism of FS transform to AFt and the stability of bound chlorides absorbed by C–S–H gel under sulfate attack were not discussed in depth.It is well known that the use offly ash(FA)and ground granulated blast-furnace slag(GGBS)in concrete can not only improve the chloride binding capacity of concrete,but also the resistance of concrete to sulfate attack[22,23].Hence,it would be interesting to know how they affect the stability of bound chlorides when the concrete is under sulfate attack.The purpose of this paper is to report the experimental in-vestigation on the stability of bound chlorides in cement paste under Na2SO4and MgSO4attack,and the corresponding influence of FA and GGBS on the stability of bound chlorides.The stability of bound chlorides in cement paste was examined by analyzing the change of a dimensionless index,R cl,which represents the mass ratio of bound chlo-rides to initial total chlorides in the sample after it was exposed to a5% Na2SO4solution or a5%MgSO4solution for28,56or90days.The mech-anisms of the release of bound chlorides are discussed based on the results of X-ray diffraction(XRD),Fourier transform infrared(FT-IR) and differential thermo-gravimetric analysis(DTG).2.Experiment2.1.MaterialsThe materials used in the experiments were Type42.5Ordinary Portland Cement(OPC),grade II FA and GGBS.The chemical composi-tions of OPC,FA and GGBS are listed in Table1.The potential phase com-positions of OPC,calculated from chemical analysis by Bogue,are given in Table2.All other chemical reagents used in the experiments,but not listed in the tables,are analytically pure.2.2.MethodsIn order to reduce the experimental running time but still able to achieve good and representative results,chloride binding was achieved by using0.5mol/L NaCl solution as the mixing water for the casting of samples.The mass ratio of the mixing water to the binder(cement and SCM)was0.5,which was the same for all samples.The influence of single and combined use of FA and GGBS on the stability of bound chlorides was also investigated.The replacement of cement with SCM was30%by weight,and the proportions of FA to GGBS in the combined samples were either1:1or7:3.The detailed mix proportions of the samples tested are listed in Table3.A total of106samples were tested.All samples were of a size of 40mm×40mm×160mm.There were three groups of samples.The first group(2×5×7samples)were cured at a standard curing condi-tion(20±2°C and95%RH)for periods of1,3,7,14,28,56and90days for the investigation of the effect of curing time and SCM on the evolu-tion of bound chlorides in the cement paste.The second group(2×5×3 samples)were examined for the effect of Na2SO4attack on the stability of bound chlorides.In this group,all samples,after the56days standard curing,were dried at a room temperature(20±2°C and60%RH)for 1day.Then,for each sample itsfive surfaces were sealed by paraffin wax and one40mm×40mm surface was left untouched.After then, all samples were immersed in a covered plastic container(575mm ×400mm×275mm)of5%Na2SO4solution for28,56and90days at the standard curing condition(20±2°C and95%RH).The third group(2×1×3samples)were for the samples only with OPC,which were cured as the same as those done in the second group.The only dif-ference is that they were immersed in a similar covered container of5% MgSO4solution for28,56and90days at the standard curing condition (20±2°C and95%RH)for the examination of the effect of MgSO4at-tack on the stability of bound chlorides.The volume of the sulfate solu-tions used in the immersion tests was25L and the storage solutions were not renewed during the immersed tests.In the second and third groups,when the attack time reached28,56, and90days,the samples were dried at room temperature for1day,and then were sliced into four pieces parallel to the exposed surface (starting from the exposed surface)and each piece is one cm thick. Afterwards,each piece was broken into small blocks,which were then immersed in anhydrous ethanol for7days to terminate hydration. These small blocks were ground intofine powder by passing through a sieve of0.15mm mesh aperture size,which was then stored in a des-iccator with silica gel and soda lime at11%RH to minimize carbonation before it was used in the tests for chloride content titration and other material characterization analyses.The initial total chloride content(C t)of the sample cured at the stan-dard curing condition can be calculated based on the mixing water of Table1Chemical composition of main materials(data presented by mass%).SiO2CaO MgO Fe2O3Al2O3SO3Ignition loss OPC19.6760.43 4.56 4.20 5.70 2.30 2.54FA43.10 6.300.247.2638.200.70 2.04GGBS23.5052.80 6.500.7011.80 1.650.78Table2Potential phase composition of OPC(data presented by mass%).Potential phase composition OPCC3S51.58C2S17.77C3A8.01C4AF12.773.91212J.Geng et al./Cement and Concrete Research68(2015)211–2220.5mol/L NaCl solution,which is 8.863mg ·g −1.The free chloride content (C f )was measured using the traditional leaching method according to the standard of Test Code for Hydraulic Concrete (SL352-2006)and the total chloride content (C t )was measured using the acid-soluble method (SL352-2006).In order to analyze the stability of bound chlorides in concrete,the dimensionless index (R cl )was exam-ined,which is de fined as follows,R cl ¼C t −C f %ð7Þwhere 8.863mg.g −1is the initial total chloride content in the sample.X-ray diffraction (XRD)/reference intensity ratio (RIR)analysis and DTG can be used to approximately determine the quantity of FS,AFt and calcium hydroxide (CH)in the samples.XRD/RIR can determine the relative mass relations among different minerals in a sample,which is calculated according to the following equations [24,25]:W i ¼I i =RIR iX i ¼1I i=RIR i ðÞð8ÞW 1þW 2þW 3þ⋯þW l ¼1ð9Þwhere W i is the relative mass of mineral i ,RIR i is the reference intensityratio of mineral i ,which can be collected from the PDF card of the Inter-national Centre for Diffraction Data (ICDD),I i is the integral intensity of the highest peak of mineral i ,which is calculated using X'Pert HighScore Plus ™software,and N is the number of minerals in the sample.XRD/RIR is usually used to determine the quantity of substances in metals because of simple compositions [25].For cement based materials,it is rather complicated to accurately determine the kinds of hydration products,which increases the dif ficulty of the quantitative analysis.However,if the quantity of one of the minerals can be determinedusing other methods,the calculation process of XRD/RIR becomes pared with the FS and AFt,the quantity of CH can be accurately determined using DTG.Therefore,the quantities of the FS and AFt can be calculated by solving the following algebraic equations,m FS :m AFt ¼T 1ð10Þm FSFS þm AFt þm CH ¼T 2ð11Þm AFtm FS þm AFt þm CH ¼T 3ð12Þm CHm FS þm AFt þm CH¼T 4ð13Þwhere m FS ,m AFt and m CH are masses of FS,AFt and CH,respectively,T 1,T 2,T 3and T 4are the mass ratios,which can be calculated from Eqs.(8)and (9).Note that,m CH can be determined by DTG and thus m FS and m AFt can be determined by Eq.(10)plus any one taken from Eqs.(11)–(13).XRD was carried out using the D8Advance instrument of Bruker AXS with a Cu K αradiation generated with 40kV and 30mA.The diffraction spectra were collected in the range of 5–60°(2θ)scale,with a step sizeTable 3Mix proportions (data presented by mass %).Samples OPC FA GGBS w/b a NoteCN 100000.5Exposed to 5%Na 2SO 4solutionCF 703000.5CG 700300.5CF1G17015150.5CF7G3702190.5CM1000.5Exposed to 5%MgSO 4solutionaw/b represents the mass ratio of mixing water (i.e.0.5mol/L NaCl solution)to binder (cement +SCM).Fig.1.Variation of R cl with standard curing time in samples of differentmixes.Fig.2.Values of R cl in the surface layer of the sample at various different sulfate attack times (CM was exposed to MgSO 4,while all the others were exposed to Na 2SO 4).Fig.3.Values of R cl in different layers of the sample after 90days sulfate attack (1st layer is next to the surface and 4th layer is away from the surface.CM was exposed to MgSO 4,while all the others were exposed to Na 2SO 4).213J.Geng et al./Cement and Concrete Research 68(2015)211–222of 0.02°/s.FT-IR was performed for the samples on a Nicolet Nexus 470spectrometer using the KBr pellet technique in the range of 400–4000cm −1.DTG was carried out in a Netzsch TG-209F1thermal an-alyzer,using a heating rate of 20°C/min at the range of 25–1000°C,in N 2atmosphere.3.Stability of bound chlorides 3.1.Standard curing conditionThe variation of R cl during the standard curing time is shown in Fig.1.It can be seen from the figure that R cl in the samples with SCM is higher than that in the sample only with OPC when they have the same curing time,which is more obvious after the curing time exceeds 14days.Up to 28days,the combined use of FA and GGBS results in higher values of R cl in CF1G1and CF7G3than in the samples with only either FA (CF)or GGBS (CG).However,after the 28days standard curing,the R cl value of the samples has an order of CF ≈CF7G3N CF1G1N CG,which increases with the increased proportion of FA to GGBS.The latereffect of FA on chloride binding is mainly due to its slow pozzolanic re-action.The results shown in Fig.1indicate that the inclusion of SCM in concrete can increase the chloride binding capacity and the effect of FA on chloride binding is more signi ficant than that of GGBS.Furthermore,they also show that the R cl values of all samples increase very obviously before 28days but after that there is less change,suggesting that the equilibrium between free and bound chlorides has been reached.3.2.Sulfate attack conditionFig.2shows the expected decrease in R cl of the surface layer of all samples with the sulfate attack,but the rate of the decrease is higher than that was reported [21].The R cl value in the surface layer of sample CN exposed to Na 2SO 4solution,for example,decreases from 59.8%to 4.3%after only 28days.After that,R cl continuously decreases with the attack time but with a slow reduction rate,from 4.3%at 28days to 1.9%at 90days.The results for locations other than the surface layer at 90days are shown in Fig.3.It can be seen from the figure that,although the 4th layer of sample CN is far away from theexposedFig.4.XRD patterns of samples CN(CM),CF and CG at standard curing condition for (A)28and (B)56days (E:ettringite (AFt),F:Friedel's salt (FS),CH:calcium hydroxide,M:mono-sulfoaluminate,V:Vaterite,CSH:C –S –H gel,C:calcite).214J.Geng et al./Cement and Concrete Research 68(2015)211–222surface,there is still a notable decrease in the R cl value from59.8%at the beginning of the Na2SO4attack to16.6%after90days of attack.This demonstrates that the stability of bound chlorides in concrete is very susceptible to Na2SO4attack.Note that the data plotted in Fig.2show that there is also a decrease in the R cl values of the samples with SCM after Na2SO4attack for28days, but the R cl values are still higher than that of the sample CN only with OPC.This suggests that the use of SCM can alleviate the effect of Na2SO4attack on the stability of bound chlorides.This is partly because the effect of SCM on the diffusion of ions,since the ionic diffusion coef-ficient in cement paste with SCM is normally lower than that in OPC paste,and partly because the cement paste with SCM has more bound chlorides[26].Additionally,in contrast with the results obtained under the standard curing condition,the R cl values of the samples with SCM increase with the decreased proportion of FA to GGBS,and also the R cl value of the surface layer of sample CF is the lowest of all samples containing SCM,following the Na2SO4attack.This suggests that Na2SO4attack can also alter the effect of SCM on the stability of bound chlorides.This appears to be consistent with what is reported in literature[21].The stability of bound chlorides in concrete under MgSO4attack is also shown in Figs.2and3.When the MgSO4attack time extends from0to28days,the R cl value of the surface layer of sample CM decreases from59.8%to26.3%,which is slower than that of sample CN exposed to Na2SO4solution.When the attack time reaches90days, the R cl value of sample CM's surface layer decreases to7.5%,which is still almost four times as high as that of sample CN.This indicates that the stability of bound chlorides is less susceptible to MgSO4attack when compared with Na2SO4attack.Again,thisfinding is consistent with what is reported in other experiments[21,27].The different reductions of R cl in samples CM and CN reflect the different effects of MgSO4and Na2SO4on bound chlorides.During the immersion process free chloride ions will diffuse out and sulfate ions will diffuse into the specimen.The former may decrease the bound chlo-ride level in the sample owing to the equilibrium between the free and bound chlorides.The latter can transform FS into AFt,which not only can reduce the bound chlorides but also can change the pore system and thus affect the diffusion rate of ions.In addition,when magnesium is present,brucite will be formed,which can also change the pore sys-tem and thus affect the transport of ions and the R cl value.The slower reduction of R cl found in sample CM shown in Figs.2and3indicates that the magnesium ions must have some influence on the sulfate attack to the bound chlorides.This influence could be physical and/or chemi-cal.The former is mainly due to the forming of brucite in the surface layer,which reduces the inward diffusion of sulfate ions and the out-ward diffusion of chloride ions.Indeed,the measured free chloride con-centration after the90days immersion was found to be higher in sample CM than in sample CN and have the ratios of about1:0.72for the surface layer and1:0.81for the4th layer.An accurate analysis for the diffusion effect on the bound chlorides requires having more data on thinner layers and knowing the binding isotherms.Nevertheless, the above results did indicate that the diffusion of chloride ions was affected by magnesium ions.The chemical effect of magnesium ions on bound chlorides will be discussed in the next section.Note that the ionic diffusion coefficient in concrete with SCM is nor-mally smaller than that in concrete only with OPC.Thus,the inclusion of SCM in cement paste can provide additional resistance to the ingress of sulfate ions,which in turn can affect the stability of bound chlorides. More discussion on this will be provided in the next section.4.Material characterization analyses4.1.X-ray diffractionThe XRD patterns of samples CN,CF and CG cured at the standard curing condition for28and56days are shown in Fig.4.From the XRD patterns one can identify the FS with a very obvious diffraction peak at around11°2θ.Fig.5shows the relative masses of AFt,FS and CH in samples CN,CF and CG after they were cured in the standard condition for56days.It can be seen from thefigure that the use of FA and GGBS is beneficial to forming more FS.This result can be attributed to two rea-sons.First,the forming process of FS in concrete has been associated with the quantity of aluminate in cementious materials.The higher the quantity of aluminate,the more FS is formed.According to the chemical composition shown in Table1,there is a larger quantity of alu-minate in GGBS and FA than in OPC,which can be released due to the latent hydraulic property of GGBS and the pozzolanic property of FA, which is beneficial to the formation of FS.Secondly,the formation of FS would be hindered because SO42−can react with aluminate prior to Cl−to form mono-sulfoaluminate(AFm)and AFt[13–15].In addition, C–A–H and C–S–H gel,formed due to the hydration reactions induced by FA and GGBS,are also beneficial to chloride binding.As shown in Fig.5,although the quantity of aluminate in FA is higher than that in GGBS,the quantity of FS in sample CF is still lower than that in sample CG after standard curing for56days.It was believed that only reactive alumina Al2O3r−in SCM could react with Cl−to form FS[5].The quantity of CaO in FA used in this study is6.3%,which is low calciumfly ash according to Chinese specification GB/T15696-2005,and where Mullite is the main form of Al2O3,so it is adverse to the formation of FS.Nevertheless,a notable decrease in the intensity of diffraction peak (IDP)of CH can be found in the XRD patterns of sample CF over the curing time from28to56days,which is induced due to the pozzolanic reaction between CH and FA.As a result,more C–S–H gel and C–A–H are formed,which could increase the bound chlorides in sample CF.It should be noticed that the IDP change at around30°(2θ)shown in Fig.4correlates with both C–S–H gel and calcite(CaCO3),because of the overlap of the two strongest diffraction peaks at29.25°(2θ)and 29.40°2θ,respectively[8,28].The XRD patterns of sample CN under Na2SO4attack are shown in Fig.6.It can be observed from Fig.6A that the IDP of FS in the surface layer of sample CN becomes very weak after Na2SO4attack for28 days,which indicates that FS has been decomposed due to the Na2SO4 attack.A quantitative analysis of FS,AFt and CH of sample CN after the Na2SO4attack for28and90days is shown in Fig.7.It can be seen from thefigure that the relative mass of FS in the sample decreases very quickly from2.04to0.45after the28days attack.This suggests that the stability of FS is very susceptible to Na2SO4attack,which may also explain why the decrease of R cl is quick as is shown in Fig.2.How-ever,when the attack time is extended from28to90days,the change in the quantity of FS is slight,which indicates that a large quantity of FShasFig.5.Analysis of ettringite(AFt),Friedel's salt(FS)and calcium hydroxide(CH)in sam-ples CN/CM,CF and CG after they had56days standard curing(wt.%represents the mass percentage of AFt/FS/CH in sample).215J.Geng et al./Cement and Concrete Research68(2015)211–222been decomposed following 28days of the Na 2SO 4attack.Moreover,it can be seen from Fig.7that the quantity of FS gradually decreases from the inside to the surface,which correlates with the change of the R cl value shown in Fig.3.In addition,one can see from Fig.6B that AFt with a diffraction peak at around 9°(2θ)can be detected in every layer of sample CN after the Na 2SO 4attack for 90days.The data shown in Fig.7for AFt indicate that the quantity of AFt in the fourth layer of sam-ple CN is higher than its initial value,which con firms that the attack of Na 2SO 4has reached the fourth layer of the sample.Fig.7also shows the expected opposite changes of FS and AFt with time.The XRD patterns of samples CF and CG after the Na 2SO 4attack for 90days are shown in Fig.8.Similar to the sample CN,the diffraction peaks of FS in the samples with SCM,especially in sample CF,become very weak.Similar to the analysis of the sample CN,Fig.9shows the relative mass of FS,AFt and CH of samples CF and CG after the Na 2SO 4attack for 90days.It seems that the quantities of FS in samples CF and CG are as high as that in sample CN after the Na 2SO 4attack.However,considering the higher quantity of FS in samples CF and CG before the Na 2SO 4attack as shown in Fig.5,the decrease of the quantity of FS in them is quicker than that in sample CN.Therefore,it can be concluded that the stability of FS in the samples with FA or GGBS is susceptible to Na 2SO 4attack when compared to the sample CN.The XRD patterns of samples CN and CM attacked by Na 2SO 4and MgSO 4for 90days are shown in Fig.10.An interesting finding is that there is still an obvious diffraction peak of FS in the sample CM,which is different from the sample CN attacked by Na 2SO 4.The analysis results shown in Fig.11demonstrate that there is more FS in sample CM than in sample CN.Therefore,it can be concluded that the Na 2SO 4attack has more effect on the decomposition of FS in hardened cement paste than the MgSO 4attack.In addition,the IDP of AFt in sample CM is lower than that in sample CN due to the different erosion mechanisms.However,there is still an obvious increase in AFt for sample CM from 0to 90days as demonstrated in Figs.5and 11,which indicates that MgSO 4attack can also lead to the formation of secondary AFt.NoteFig.6.XRD patterns of samples CN with Na 2SO 4attack.(A)1st layer at different days and (B)different layers at 90days (E:ettringite (AFt),F:Friedel's salt (FS),CH:calcium hydroxide,M:mono-sulfoaluminate,V:Vaterite,CSH:C –S –H gel,C:calcite).216J.Geng et al./Cement and Concrete Research 68(2015)211–222that,when magnesium is included in the exposure solution,brucite is formed at the expense of calcium hydroxide,which can affect not only the leaching of chloride from the specimen but also the inward trans-port of sulfate from the exposed solution and thus provide the in fluence on the decomposition of FS and the formation of AFt.However,our XRD result did not reveal a signi ficant amount of brucite and/or gypsum in the surface layer.This is probably due to the specimen layer used in the tests being too thick.Both Skaropoulou and Sotiriadis reported their test results in which brucite was detected in XRD patterns,but the IDP of it was very weak when compared to other phases [11,17].However,in other similar experiments brucite was not detected in XRD patterns [27,29,30].This is probably attributed to the consumption of brucite due to the formation of M –S –H as shown in Eqs.(4)–(6)[19].4.2.Fourier transform infrared (FT-IR)Fig.12shows the FT-IR spectra of sample CN after the Na 2SO 4attack for 28and 90days,respectively.The band at around 3640cm −1is due to the stretching vibration of \OH in Ca(OH)2[30],which is very weak in all samples due to Na 2SO 4attack.The presence of carbonate bands at around 1430and 870cm −1indicates that the samples have already absorbed CO 2molecules from the air before they were immersed into sulfate solution [31].The band at around 1110cm −1comes from asym-metric stretching vibration of S –O in SO 42−,which is identi fied as the fingerprint peak of AFt [32,33].As is shown in Fig.12,owing to more secondary AFt being formed,this band becomes stronger from the in-side to the surface over the attack time.The changes in the bands at around 3440and 1650cm −1are due to the stretching vibration of \OH in structural water of hydration products and the bending vibra-tion of \OH in the interlayer water of hydration products [30].The two bands are also related to the formation of secondary AFt,which be-come strong with the increased quantity of secondary AFt.In addition,the band at around 970cm −1comes from asymmetric stretching vibra-tion of Si –O in C –S –H gel [31,34].It can be observed from Fig.12that there is no obvious change in this band over the attack time,which sug-gests that the stability of C –S –H gel is independent of Na 2SO 4attack.With regard to FS,because chloride ions are not absorbed in the range 400–4000cm −1,the bands at around 730,530and 460cm −1,which are due to Al –O vibrations of [Al(OH)6]3−,can be identi fied as the fin-gerprint peaks of FS [35,36].Owing to the decomposition of FS under Na 2SO 4attack,the strength of these bands appears very weak.Fig.13shows the FT-IR spectra of samples CF and CG after the Na 2SO 4attack.There is no obvious band at around 3640cm −1in thespectra due to the consumption of CH induced by hydration reactions of FA and GGBS and sulfate attack.It can be observed from Fig.13that there is an increase in the strength of the band of C –S –H gel at 976cm −1in sample CF over the attack time from 56to 90days.Guerre-ro et al.attributed this to the further activating action on FA due to the increase in alkalinity induced by Na 2SO 4attack [15].Moreover,this re-sult also indicates that the stability of C –S –H gel is independent of Na 2SO 4attack.The difference of the bands at 714,535and 458cm −1be-tween samples CF and CG is slight.Fig.14shows the FT-IR spectra of samples CN and CM after Na 2SO 4and MgSO 4attack for 90days,respectively.It is observed from Fig.14that the strength of the band at around 710cm −1in sample CM is much stronger than that in sample CN.Also there is more FS in sample CM than in sample CN,which agrees with the results shown in Figs.10and 11.Moreover,it can be seen clearly from Fig.14that the strength of the band at around 970cm −1in sample CM is lower than that in sample CN.This is likely attributed to the decomposition of C –S –H gel induced by MgSO 4attack.As a result of that,the bound chlorides absorbed by C –S –H gel are released.A weak band at around 1110cm −1in sample CM due to the attack of MgSO 4can induce the formation of secondary AFt.4.3.Derivative thermo-gravimetric analysis (DTG)The DTG curves of sample CN attacked by Na 2SO 4are shown in Fig.15.There are some notable endothermic peaks in the DTG curves.The peak near 100°C is mainly attributed to the dehydration of C –S –H gel and AFt,which are dif ficult to distinguish because of the overlap of dehydration temperature from 85to 130°C [23].The peak near 160°C is attributed to AFm [23].Besides these,the peaks near 340,450and 710°C are attributed to the dehydration of FS,CH and the decomposi-tion of calcite.The absence of the peak for FS in the DTG curve after the Na 2SO 4attack for 28days shown in Fig.15further demonstrates that the stability of FS is susceptible to Na 2SO 4attack.The change in the peak of AFm,which plays an important role in the formation of sec-ondary AFt during the Na 2SO 4attack,is also consistent with the change of FS.Fig.16shows the DTG curves of samples CF and CG after the Na 2SO 4attack for pared to sample CG,sample CF has a weak strength of the peak for FS,which is consistent with the analysis result shown in Fig.9and the R cl data shown in Fig.2.Fig.17shows similar DTG results of samples CN and CM after Na 2SO 4and MgSO 4attack for 90days.It is noticed from the figure that the strength of the peak for C –S –H gel and AFt in sample CM is far lower than that in sample CN.Ac-cording to the FT-IR results shown in Fig.14,this result further indicates that MgSO 4attack will lead to the decomposition of C –S –H gel,resulting in the release of bound chlorides.5.Discussion5.1.Stability of Friedel's saltSuryavanshi and Swamy reported that a drop in alkalinity of pore so-lution due to carbonation could induce the decomposition of FS [8].Con-versely,Na 2SO 4attack can increase the alkalinity of the pore solution,which has a negative effect on chloride binding [23,27,37].The question now is how Na 2SO 4attack affects the stability of FS.The exchange be-tween Cl −and SO 42−is the main mechanism in the formation of FS,which can be explained by the following reaction [27]:3CaO ·Al 2O 3·CaSO 4·12H 2O ðAFm Þþ2Cl −→3CaO ·Al 2O 3·CaCl 2·10H 2O ðFS ÞþSO 2−4þ2H 2O :ð14ÞEssentially,FS belongs to a phase of the AFm family,which has a complex chemical and structural constitution.A general formula for AFm phase is [Ca 2(Al,Fe)(OH)6]+X·m H 2O,where the bracketsindicateFig.7.Analysis of ettringite (AFt),Friedel's salt (FS)and calcium hydroxide (CH)in sample CN after Na 2SO 4attack for 0,28and 90days (wt.%represents the mass percentage of AFt/FS/CH in sample).217J.Geng et al./Cement and Concrete Research 68(2015)211–222。
![2015_-_Jian_Geng_-_Thestabilityofboundchloridesincementpastewithsulfa[retrieved_2015-09-23]](https://img.taocdn.com/s1/m/c873140459eef8c75fbfb32d.png)
