Solid–liquid distribution of selected concrete admixtures in hardened cement pastes

Waste Management 26 (2006) 741–751 www.elsevier.com/locate/wasman

Solid–liquid distribution of selected concrete admixtures in hardened cement pastes Martin A. Glaus *, A. Laube, Luc R. Van Loon Laboratory for Waste Management, OHLD/002, Paul Scherrer Institut, CH-5232 Villigen, Switzerland Accepted 31 January 2006 Available online 15 March 2006

Abstract The distribution between hardened cement paste and cement pore water of selected concrete admixtures (BZMs), i.e., sulfonated naphthalene–formaldehyde condensate (NS), lignosulfonate (LS) and a gluconate-containing plasticiser used at the Paul Scherrer Institute for waste conditioning, was measured. Sorption data were fitted to a single-site Langmuir isotherm with affinity constants K = (19 ± 4) dm3 g1 for NS, K = (2.1 ± 0.6) dm3 g1 for LS and sorption capacities q = (81 ± 16) g kg1 for NS, q = (43 ± 8) g kg1 for LS. In the case of gluconate, a two-site Langmuir sorption model was necessary to fit the data satisfactorily. Sorption parameters for gluconate were K1 = (2 ± 1) · 106 dm3 mol1 and q1 = (0.04 ± 0.02) mol kg1 for the stronger binding site and K2 = (2.6 ± 1.1) · 103 dm3 mol1 and q2 = (0.7 ± 0.3) mol kg1 for the weaker binding site. Desorption of these BZMs from cement pastes and pore water in cement specimens prepared in the presence of the BZMs were then used to test the model. It was found that only minor parts of NS and LS could be mobilised as long as the cement composition was intact, whereas the sorption of gluconate was found to be reversible. The Langmuir model makes valuable predictions in the qualitative sense in that the pore water concentration of the BZMs is reduced by several orders of magnitude as compared to the initial concentrations. In view of the necessity for conservative predictions used in the safety analysis for disposal of radioactive waste, however, the predictions are unsatisfactory in that the measured pore water concentrations of NS and LS were considerably larger than the predicted values. This conclusion does not apply for gluconate, because its concentration in cement pore water was below the detection limit of 50 nM.  2006 Elsevier Ltd. All rights reserved.

1. Introduction Concrete admixtures (abbreviated as BZMs1) are used to improve the workability of cement, to influence physical properties, such as compressive strength, durability or setting time, and to improve mix rheology (Dodson, 1990; Ramachandran, 1995; Spiratos and Jolicoeur, 2000). In the context of the safety of cementitious repositories for lowand intermediate-level or long-lived intermediate-level radioactive waste planned in Switzerland, BZMs are of concern in the same way as other organics, such as cellulose or *

Corresponding author. Tel.: +41 56 310 22 93; fax: +41 56 310 22 05. E-mail address: [email protected] (M.A. Glaus). 1 The abbreviation derives from the German word Betonzusatzmittel, no commonly used abbreviation for concrete admixtures has been found in the literature. 0956-053X/$ - see front matter  2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.wasman.2006.01.019

ion exchange resins. Sorption of radionuclides on the solid repository matrix is a key factor governing their possible later release from the repository (Hadermann, 1997). Organic substances contained in the repository may positively or negatively influence this sorption. Therefore, the role of organics needs to be addressed quantitatively in an overall performance assessment. The assessment of BZMs is complicated because of a number of special circumstances: (i) BZMs comprise a large variety of completely different chemical substances, ranging from simple carbohydrates to complex mixtures of macro-molecules; (ii) a single type of BZM is, in turn, most often composed of a variety of different compounds, the composition being in many cases ill-defined, or proprietary; (iii) it is not known which BZM will be used in the construction of cementitious radioactive waste repositories in Switzerland – the only BZMs that can be identified at the moment are those used for waste conditioning.

742

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

The pore water concentration of BZMs is a key parameter in the assessment of their role in safety analysis of a radioactive waste repository, as it directly affects the extent of possible radionuclide complexation. It is further related to the amount of BZMs bound to the surface, which may influence competitively the sorption on cement of radionuclides or other sorbing ligands. It has been shown unequivocally in the literature that BZMs, such as superplasticisers and set retarders, strongly sorb onto the surface of cement particles (Ernsberger and France, 1945; Blank et al., 1963; Rossington and Runk, 1968; Daimon and Roy, 1978; Popescu et al., 1991; Spanka and Thielen, 1995) and single mineral phases of cement (Blank et al., 1963; Rossington and Runk, 1968; Ramachandran, 1972; Costa and Massazza, 1984; Fukaya and Kato, 1986; Yilmaz and Glasser, 1991). However, these investigations were mostly done to understand the mechanism of action of BZMs and, for this reason, were mostly carried out on early stages of cement hydration. The aim of the present work is to test, whether (i) simple sorption models are suited to predict the pore water concentration of the BZMs and (ii) whether sorption mechanisms are identical for different classes of BZMs. For this purpose, sorption data for a few typical superplasticisers and set retarders on fully hydrated cement pastes were measured and fitted with one- or two-site Langmuir isotherms. Desorption of these BZMs from cement pastes and pore water analysis in cement specimens prepared in the presence of the BZMs were then used to test the predictive capacity of the sorption model. Reversibility is a prerequisite for applying Langmuir type sorption models and is therefore a good means to test the validity of the model. It is a common practice in many works concerned with BZM analysis to calculate concentrations of BZMs from measurements at a single wavelength. This may lead to biased results, because fractionation of heterogeneous BZMs may affect the spectral properties of the compounds. Possible changes in UV–Vis spectra of the BZM fractions in cement pore water have been examined in the present work, and irregularities found were included in the estimation of the overall experimental uncertainty.

shown in Fig. 1. Owing to the lack of knowledge of the exact composition of the BZMs used in this work, their concentrations specified here refer to the dry weight of BZM per volume of solution. Only in the case of gluconate are concentrations are given on a molar basis. Portland cement (CPA 55 HTS, Lafarge, France) was used for preparing cement pastes. Hardened cement paste devoid of BZM (abbreviated to HCP) was the same material described by Van Loon et al. (1997). It was crushed and sieved to <70 lm. Artificial cement pore water (abbreviated to ACW-I), prepared by the procedure described in Van Loon et al. (1997), had the following composition: 114 mmol l1 of Na, 180 mmol l1 of K, 1.6 mmol1 of Ca and a pH of 13.4. Hardened cement pastes prepared in the presence of BZM (abbreviated to BZM-HCP) were prepared according to Table 1. Portland cement was mixed in a Variac mixer (Moulinex) with half of the total water added as liquid at room temperature; remaining water was added thereafter as small-grained ice. BZM solutions were immediately added together with cement except for LS, which was added only after 90 min of cement hydration. The suspensions were first mixed manually and, after obtaining some degree of homogeneity, a mixer was used. Mixing speed was chosen such that the temperature did not exceed 75 C. Aliquots of the pastes were cured in 250 cm3 cylindrical polystyrene vessels (Semadeni, Switzerland; 65 mm diameter) after submerging the paste with 10 cm3 of water. Part of this water was successively taken up by the hydration processes and therefore replaced. After an observation time of 30 days, the samples were stored in a glove box under controlled N2 atmosphere (CO2, O2 < 5 ppm). Pore water from BZM-HCP was expressed at the Institut fu¨r Bauforschung (Rheinisch-Westfa¨lische Technische

R

n

H

O

C H SO3M

OH n

MeO Me = CH3 M = metal ion

2. Experimental section

Reagents of highest purity obtained from Fluka (Buchs, Switzerland) or Merck (Dietikon, Switzerland) and MilliQ water were used throughout. Solutions of sulfonated naphthalene–formaldehyde condensate (abbreviated to NS) and lignosulfonate (desugarised, abbreviated to LS) were obtained from MBT (Switzerland). These solutions contain only the major components of the commercial products. So-called PSI plasticiser (abbreviated to PP) is added to cements used for the conditioning of radioactive waste in Switzerland. PP contains gluconate in significant amounts. The main structures of the BZMs used here are

H

C H2

SO3M

2.1. Reagents and samples

CH2OH

R

SO3M

R = H, CH3, C2H5 M = metal ion

(1) Sulfonated naphthalene-formaldehyde (2) Lignosulfonate (LS) condensate (NS) COOH

COOH H

C

OH

HO

C

H

H

C

OH

H

C

OH

HOH2C

C

OH

CH2 H

C

OH

CH2OH

CH2OH

(3) Gluconic acid (gluc)

(4) α-Isosaccharinic acid (α-ISA)

Fig. 1. Structural formulae of the BZM used in this work. The structure shown for LS is a gross simplification. Actually, LS is composed of a variety of structurally related monomers, irregularly aggregated to a macromolecule. a-ISA is a product of alkaline cellulose degradation used as a structural analogue to gluconate with weaker sorption properties.

M.A. Glaus et al. / Waste Management 26 (2006) 741–751 Table 1 Composition of BZM-HCP cores

Cement (g) Water (g) BZM (g) B/Cd (%) W/Ce

NS-HCP

LS-HCP

PP-HCP

Blank-HCP

1049 715 35.0a 1.5 0.7

1051 716 31.5b 1.2 0.7

1006 736 12.6c 1.25 0.7

1058 742 0 0 0.7

743

L. UV–Vis measurements of BZM need to be conducted with strict control of pH. The reason for this precaution is the fact that hydroxide ions strongly absorb in the far UV region, and consequently, differences in pH may lead to errors when subtracting blank spectra. Further, extinction coefficients may be strongly pH dependent for certain BZMs (Spanka and Thielen, 1995).

a

45% (weight per weight) solution. 40% (weight per weight) solution. c Undiluted solution. d Ratio of BZM to dry weight of cement (expressed as percentage of weight per weight). e Ratio of water to dry weight of cement (on a weight basis). b

Hochschule, Aachen, Germany) using a 500 N mm2 press. Three replicate samples were used per sampling time. Before use, the outer part of the core (a few millimetres) that could have been chemically altered during preparation was removed. During the expression process, the pore water was protected against CO2 contamination by an argon blanket. Only a few millilitres of pore fluid could be obtained from one BZM-HCP core. 2.2. UV–Vis spectrophotometry UV–Vis spectra were recorded in steps of 1 nm on a Camspec M330 single-beam photometer (Biolabo, Lausanne) using 1 cm quartz cuvettes. UV–Vis spectrophotometry has often been applied to measure concentrations of NS and LS. The fact that most organic and inorganic substances absorb in the wavelength region near 200 nm necessitates a careful choice of reference spectra. Fig. 2 shows spectra of supernatants of HCP equilibrated with ACW-I at different solid to liquid ratios (S/L). The spectra suggest that cement–borne substances are exuded from HCP, and they underline the necessity to correct spectra of BZM in ACW-I equilibrated with HCP, even at relatively low S/

0.8

S/L = 17.1 g dm-3

0.7

Optical density

0.6 0.5

10.2

0.4 6.7 0.3 0.2

3.5

0.1

1.2

0 230

240

250

260

270

280

290

300

Wavelength (nm) Fig. 2. Spectra of unknown substances exuded by HCP at different S/L into ACW-I.

2.3. Measurements of total organic carbon Total organic carbon (TOC) was analysed by a Dohrmann DC-180 Carbon Analyser (Schmidlin, Neuheim, Switzerland) using a UV-promoted persulfate wet oxidation and non-dispersive infrared detection of the CO2 evolved. The accuracy of the results obtained by this technique depends on the completeness of oxidation of the substrate. This prerequisite may not be achieved for all types of organic compounds. In the case of polymeric organics, TOC results may underestimate the true content by a factor up to 3 (Kaplan, 1992). No information on the completeness of oxidation of BZMs can be found in the literature. For this reason, a few tests with variably diluted stock solutions of the NS, LS and PP have been performed in this study. A linear relationship between measured TOC and the amount of BZM solution has been found at concentration ranges between 2 and 60 mg dm3 (R2 > 0.999). This is a first indication for complete oxidation of the BZM carbon to CO2. Another indication comes from a comparison of measured TOC values with the specification of the manufacturer, which were in acceptable agreement (<5%). 2.4. Analysis of gluconate by high performance anion exchange chromatography (HPAEC) Measurements were carried out on a Dionex DX-500 system (Dionex, Switzerland) consisting of a metal-free GP 40 quaternary gradient pump, an ED 40 electrochemical detector and an AS3500 SpectraSYSTEM autosampler (Thermo Separation Products), equipped with a 9010 motor-driven Rheodyne injection valve, a 100 ll PEEK injection loop and a 250 ll sample syringe. A 4 mm · 250 mm Carbopac PA-100 separation column equipped with a 4 mm · 50 mm Carbopac PA-100 guard column (Dionex, Switzerland) was used for separation. 100 ll samples were injected in full loop mode. For pure solutions of gluconate, the eluent was 0.25 M NaOH. If strongly retained components were present in the sample, i.e., if the sample matrix was PP in ACW-I, this eluent was – after elution of gluconate – superimposed by a step gradient of sodium acetate reaching a maximum acetate concentration of 0.75 M, while the NaOH concentration remained 0.25 M. Peaks were detected by pulsed amperometry using a gold working electrode in combination with an Ag|AgCl reference electrode. The waveform applied to detect the analytes and to clean and regenerate the gold surface, was taken from Rocklin et al. (1998). By repetitive injection of standard solutions of gluconate across the

744

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

whole sample schedule, the variability of detector response with time could be monitored. A corrected peak height of the sample ðy 0p Þ was used to calculate analyte concentration (xp), according to: xp ¼ F 

y 0p  a b

ð1Þ

where F is the dilution factor of the sample, and a and b are the parameters of the linear regression of the calibration points. y 0p is calculated from the measured peak height of the sample, yp, by applying a correction function (fk), usually a polynomial expression between first and third degree: yp y 0p ¼ ð2Þ fk fk is obtained from the dependence of the peak height of the repetitive control measurements on time. Best results for samples in an ACW-I matrix or cement pore water were obtained by sample pre-treatment with OnGuard-H cartridges in the H+-form (Dionex, Switzerland), resulting in an exchange of alkaline earth and transition metal ions against protons, which leads to a neutralisation of the sample. With best performance of the detector, gluconate concentrations of 108 M could be detected. 2.5. Sorption on and desorption from HCP of BZMs All experimental steps were carried out in a glove box under controlled N2 atmosphere (CO2, O2 < 5 ppm). Crushed and sieved particles with sizes <70 lm of a HCP, hydrated for several years at high water to cement ratio, were used as the solid. Polysulfone centrifuge tubes were filled with a suitable amount of cement particles (0.5 g in the case of NS and gluconate, 0.5–1 g in the case of LS) and between 10 and 30 cm3 of BZM solutions in ACW-I resulting in S/L of the order of 25 g dm3. Total concentrations of BZMs (further denoted as ‘‘input concentrations’’) added ranged between 0.02 and 5 g dm3 (NS), between 0.4 and 8 g dm3 (LS) and between 0.1 and 100 lM (gluconate). The suspensions were shaken end-over-end for a desired time and then centrifuged for 15 min at 27,000g (max.). In some cases a layer of cement particles remained at the surface of the supernatant. The supernatant withdrawn was therefore additionally filtered with a 0.45 lm membrane filter discarding the first few millilitres before analysis. Each data point, i.e., at specific S/L or BZM input concentrations, was measured in duplicate or triplicate. Spectrophotometric analysis was used to measure the input concentrations (g dm3) of NS and LS ([BZM]in) and equilibrium concentrations in the supernatants obtained after centrifugation ([BZM]eq). Wavelength ranges between 280 and 310 nm were used to calculate NS concentrations and between 270 and 390 nm to calculate LS concentrations according to the following equation, where the index k denotes the wavelength of evaluation:

½BZMeq;k ¼

Dsup k 

Dcem k F

bk

 ak

F

ð3Þ

where Dsup is the optical density of the supernatant and k Dcem the optical density of the solution phase of a suspenk sion of HCP in ACW-I at comparable S/L, but without BZM. F is the dilution factor, and ak and bk are the calibration parameters. Dcem was determined according to Fig. 2. k [BZM]in,k and [BZM]eq,k used to calculate the amount of BZM sorbed onto cement, [BZM]sorb,k, given as g BZM per kg of HCP:   V tot ½BZMsorb;k ¼ ½BZMin;k  ½BZMeq;k  ð4Þ mc where Vtot is the total volume of the liquid phase (dm3), and mc is amount of HCP (kg). [BZM]eq,k and [BZM]sorb,k were averaged across the range of wavelength evaluation in order to give single values for [BZM]eq and [BZM]sorb. Combined standard uncertainties were calculated according to the recommendations given by the EURACHEM Working Group (Williams et al., 1995). For desorption of BZMs crushed HCP was first loaded to saturation with 20 cm3 of BZM containing ACW-I. After centrifugation (27,000g (max.), 15 min), 18 cm3 of supernatant were removed, analysed for BZM and replaced by the same volume of fresh ACW-I. In the case of NS and LS this suspension was further shaken end-over-end, and filtered aliquots were periodically analysed by spectrophotometric measurements for BZM. After 16 days, the step of centrifuging and replacing the supernatant by fresh ACW-I was repeated and changes in BZM concentration in the solution phase were monitored for another 15 days. In the case of gluconate, the equilibration time for one desorption step was between 1 and 3 days; approximately 40 exchange steps were performed. The kinetics were not followed upon exchange of ACW-I. In all cases the concentration of BZM measured in the supernatants of a given exchange step were corrected for the amount of BZM remaining in solution from the previous extraction step due to incomplete exchange of the liquid phase. 2.6. Extraction of BZM from BZM-HCP Extraction of cement paste with 1 M Na2CO3 is commonly used to determine the contents of NS, LS and many other BZMs (Connolly et al., 1980; Jeknavorian et al., 1998). Preliminary tests, however, revealed that the short extractions times recommended in Wexler and Brako (1963) were not sufficient to quantitatively extract NS and LS. This is indirectly confirmed by Sanders et al. (1995), who applied three successive extractions, each of 20 min, to extract LS from hardened concrete. Moreover, it was not possible to fully extract gluconate from PPHCP. It is assumed that, at the alkaline pH of a carbonate solution, the sorption of gluconate is too strong for a successful extraction. For these reasons, the following modifications in the procedure of Wexler and Brako (1963) were

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

introduced for the extraction of gluconate from PP-HCP: (i) the extraction times were prolonged to several days; (ii) 1 M NaHCO3 was used with PP-HCP. 3. Results and discussion 3.1. Sorption on and desorption from HCP of NS and LS Sorption isotherms of NS and LS on HCP are shown in Fig. 3. Previous experiments have shown that the distribution of NS between HCP and ACW-I does not depend significantly on the S/L in the range tested (1–20 g dm3, data not shown). The evaluation of the data in terms of a sorption model is complicated by the fact that the spectra of the BZM in the supernatants are not identical with those in the input solutions. As is illustrated in Fig. 4, a wavelength shift is observed in the case of NS indicating that the different components con-

tained in the BZM sorb differently. In order to account for these additional uncertainties, the amounts of the BZM in solution and on HCP were calculated from absorbancies in the supernatants measured at different wavelengths. As shown in Fig. 3, the distribution of the BZMs between HCP and the liquid phase calculated for different wavelength gives quite a consistent picture; the sorption data may be fitted by a one-site Langmuir type sorption equation: ½BZMsorb ¼

K  q  ½BZMeq 1 þ K  ½BZMeq

ð5Þ

where K is the sorption-affinity constant (dm3 g1 of BZM) and q the sorption capacity (g BZM per kg of HCP) of cement for BZM. The optimum fit parameters obtained are K = (19 ± 4) dm3 g1 for NS, K = (2.1 ± 0.6) dm3 g1 for LS and q = (81 ± 16) g kg1 for NS, q = (43 ± 8) g kg1 for LS. 50

100

A

B 40

[BZM]sorb (g kg-1)

80

[BZM]sorb (g kg-1)

745

60

40

30

20

10

20

0

0 0

0.5

1

1.5

2

2.5

0

3

1

2

[BZM]eq (g dm-3)

3

4

5

[BZM]eq (g dm-3)

Fig. 3. Sorption isotherms of NS (A) and LS (B) on HCP measured at S/L 17 g dm3 (NS) and 50 or 100 g dm3 (LS) after a contact time of 20 h. Data points are average values of evaluation at different wavelengths, the variability of which is comprised in the error bars. The solid lines denote fit curves using the parameters given in the text. The dashed lines denote the ranges of uncertainty introduced by the uncertainties of the fit parameters.

1.4

25

20

15 [BZM]in -3

(g dm ): 10

5

1.5 – 5

Input solution

1

Supernatants

0.5 0.1 0.02

0 250 260 270 280 290 300 310 320 330

Wavelength (nm)

B

Optical density (arbitrary scale)

Optical density (arbitrary scale)

A 1.2 1 0.8 0.6

[BZM]in:

0.4

(g dm-3)

Input solution

0.2

0.5 – 10

Supernatants 0 260

280

300

320

340

360

380

400

Wavelength (nm)

Fig. 4. Comparison of normalised spectra of input solutions and supernatants of NS (A) and LS (B) after contact with HCP. The spectra were normalised such that the optical densities of the peak maxima became identical.

746

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

Desorption of NS and LS was measured in two steps after loading HCP with these BZMs up to saturation. For each step, desorption kinetics were followed for 2 weeks. The spectra of aliquot samples of the desorption equilibrium solutions give a consistent picture for both NS and LS. They are identical with the spectra observed in the sorption experiments (data not shown). For this reason data were evaluated the same way as in the sorption experiments using a few wavelengths that cover the main absorption bands of NS and LS. A representation of concentrations of NS and LS desorbed is given in Fig. 5. The error bars comprise uncertainties due to evaluation of the results at various wavelengths, to baseline shifts of the spectrophotometric measurements and to averaging of triplicate sets of experiments. Within the time of observation, desorption of both NS and LS do not reach the equilibrium states that would be expected based on the Langmuir model. Based on the inventory of BZM and

the sorption parameters evaluated, a maximum NS concentration of 117 mg dm3 is expected for the first desorption step, and 111 mg dm3 for the second. This second exchange step, however, shows that almost the same concentrations are reached as in the first step, suggesting that part of the NS and LS is reversibly sorbed and that the desorption kinetics are fast. Most of the NS and LS is, however, either not available to desorption, or desorption kinetics are too slow to be detected in the time window of the experiments. This is in agreement with observations made in the literature for the sorption of LS on hydrated cement phases (Ramachandran, 1972). 3.2. Sorption on and desorption from HCP of gluconate Fig. 6A shows the sorption isotherm measured for gluconate on HCP and the results of desorption experiments. Most of the sorption data shown in Fig. 6A were measured 200

120 First step Second step

A

First step Second step

B

96

[LS]'eq (mg dm-3)

[NS]'eq (mg dm-3)

150 72

48

100

50 24

0

0 0

100

200

300

400

0

500

100

200

300

400

500

Time (h)

Time (h)

Fig. 5. Kinetics of desorption of NS (A) and LS (B) from HCP. ½NS0eq and ½LS0eq denote concentrations corrected for the quantities remaining from the previous equilibration step due to incomplete exchange of the liquid phase. The ordinates are chosen to represent maximum possible concentrations based on equilibration of the systems according to the Langmuir model (cf. Eq. (5)). 100

1 00

B

1 0-1

1 0-2 Sorption

[α-ISA]sorb (mol kg -1)

[gluc]sorb (mol kg-1)

A

10-1

10-2

Sorption

Desorption

Desorption

1 0-3 1 0-8 1 0-7 1 0-6 1 0-5 1 0-4 1 0-3 1 0-2 1 0-1 1 00

[gluc]eq (mol dm-3)

10-3 10-5

10-4

10-3

[α-ISA]eq

10-2

10-1

100

-1

(mol dm )

Fig. 6. Sorption on and desorption from HCP of gluconate (A) and a-ISA (B), respectively. The solid lines denote fit curves using the parameters given in the text. The dashed lines denote the ranges of uncertainty introduced by the uncertainties of the fit parameters. Error bars denote typical combined standard uncertainties of a single data point.

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

after a contact time of 1 day. Preliminary experiments, in which the reaction time was varied between 1 day and 3 mo (also included in Fig. 6A), have shown a slight dependence on reaction time, but hardly exceeding the experimental uncertainty. A similar conclusion can also be drawn for the variation of S/L. Fig. 6A contains data measured at S/L between 5 and 25 g dm3. No significant pattern showing a dependence on S/L emerges. The fit curve has been obtained using a two-site Langmuir isotherm of the following form (the abbreviation ‘gluc’ stands for gluconate, for other symbols: cf. Eq. (5)): ½glucsorb ¼

K 1  q1  ½gluceq K 2  q2  ½gluceq þ 1 þ K 1  ½gluceq 1 þ K 2  ½gluceq

ð6Þ

Eq. (6) does not give mechanistic information about the molecular processes of sorption; it is only a mathematical formula, by which the experimental data can be fitted. As for NS and LS, the uncertainties in the fit parameters have been estimated rather subjectively based on the combined standard uncertainties of individual data points and the scatter in sorption data obtained for the parameter variations tested (ageing time, S/L, in the case of gluconate). The optimum sorption parameters are K1 = (2 ± 1) · 106 dm3 mol1 and q1 = (0.04 ± 0.02) mol kg1 for the stronger binding site and K2 = (2.6 ± 1.1) · 103 dm3 mol1 and q2 = (0.7 ± 0.3) mol kg1 for the weaker binding site. Desorption of gluconate from HCP has been investigated using three independent replicate experiments. The results suggest that gluconate molecules bound to the weaker sorption site were available for desorption, because they follow in some degree the sorption isotherm. The question of reversibility of gluconate binding to the stronger site cannot be unequivocally clarified by this experiment, because sorption to this binding site is too strong for gluconate to be effectively removed. Non-statistical uncertainties may propagate from one ACW-I exchange step to the next, because the amount of gluconate sorbed for a given step is calculated based on the total content of the system in gluconate from the previous step. In view of the inherent problems associated with the measurements of gluconate desorption, similar experiments were conducted with a-isosaccharinic acid (a-ISA). a-ISA is a structurally related polyhydroxycarboxylic acid, the sorption of which on HCP has been measured in previous work (Van Loon et al., 1997). The site capacities of HCP for a-ISA are of a similar order of magnitude as for gluconate (q1 = 0.1 ± 0.01 mol kg1, q2 = 0.17 ± 0.02 mol kg1); sorption affinity constants are, however, lower (K1 = 1730 ± 385 dm3, K2 = 12 ± 4 dm3). The results of sorption and desorption experiments using a-ISA are shown in Fig. 6B. Owing to the weaker sorption of aISA on HCP, compared to gluconate, it is possible to quantitatively recover a-ISA also from the stronger binding site. In view of the chemical analogy between a-ISA and gluconate it is thus reasonable to assume that desorption of the latter from HCP is also reversible.

747

3.3. Content in and desorption from crushed BZM-HCP of BZM The results of BZM extraction from crushed BZM-HCP material harvested after a curing time of 20 mo are given in Table 2; gluconate has been recovered quantitatively. Reproducible results were obtained for the extraction of all BZM-HCP samples investigated across the time scale investigated. Note that the spectra of NS are rather identical to those of fresh solutions of NS. The results suggest that the BZMs did not degrade in 20 mo. This conclusion can safely be drawn for gluconate due to the specificity of the analytical method, but in the case of NS and LS it is suggested by unchanged absorption in the wavelength range between 250 and 320 nm, that the main structural components are still conserved. It is especially remarkable that the spectra of the extracts of LS-BZM do not show the signature of transformation to vanillin, which has been observed for the pore waters (cf. the following section). Thus, it can be hypothesised that the LS molecules present in the cement pore water have different reactivity than those bound to HCP. However, a decrease in degree of polymerisation may have taken place during ageing. Desorption of NS, LS and gluconate from crushed NSHCP, LS-HCP and PP-HCP was measured by four consecutive extractions with fresh ACW-I. Desorption was monitored by UV–Vis and TOC measurements in the supernatants in the case of NS and LS and by HPAEC measurements in the case of PP-HCP. Within the range of uncertainty, the results of UV–Vis spectrophotometry were in fair, but acceptable agreement with those from the TOC measurements for the sampling after 4 mo and in good agreement for the sampling after 20 mo. PP-HCP samples were not further investigated after showing that the concentration of gluconate remained below the Table 2 Extraction by Na2CO3 (NS- and LS-HCP) and by NaHCO3 (PP-HCP) of BZM from crushed BZM-HCP material hydrated for 20 mo BZM-HCP (g)

Contact time (d)

NS-HCP

1.001 0.994 0.998 1.014

3 5 7 10

8.1 8.4 8.3 8.3

10.9 10.9 10.9 10.9

74 ± 30 77 ± 30 76 ± 30 76 ± 30

LS-HCP

0.996 1.003 1.010 1.004

3 5 7 10

9.7 10.0 9.3 10.3

8.7 8.7 8.7 8.7

111 ± 32 114 ± 32 107 ± 32 117 ± 32

PP-HCP

1.012 0.998 0.994 0.997

3 5 7 10

4.4 4.7 4.5 4.5

5.6 5.6 5.6 5.6

79 ± 34 84 ± 34 81 ± 34 81 ± 34

a

BZM extracteda

Expectedb

Sample

Recovery (%)

Total amount of BZM extracted (mg for NS and LS, lmol for gluconate). b Total amount of BZM based on the assumption of 50% water loss during storing and crushing of the samples (mg for NS and LS, lmol for gluconate).

748

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

detection limit in HPAEC; UV–Vis spectra did not show any absorbing species in the typical wavelength range investigated (data not shown). Fig. 7 shows a comparison between measured concentrations of NS and LS in the equilibrium solutions and those calculated using the parameters for the Langmuir sorption isotherm (cf. Section 3.1). The Langmuir model is again not suited to describe the experimental data both for NS and LS, even when considering that the initial concentration has an estimated relative uncertainty of 30% owing to the uncertainties in water loss during storage of the samples. The NS data in particular suggest that major parts of this BZM are not available to desorption. 3.4. BZM in pore water of BZM-HCP Fig. 8 show typical spectra measured in the pore water samples expressed from NS- and LS-HCP cores. The spectrophotometric technique for determining the BZM con-

centration is noticeably impaired. The spectra of the pore waters from NS-HCP have different absorption maxima as compared to solutions of pure NS in ACW-I. In the case of pore waters from LS-HCP, a new absorption band is observed. Given this shift of absorption characteristics, exact determination of the concentration of NS and LS in the pore water based on UV–Vis measurements will be biased. The spectra of NS in the pore waters are, however, very similar to those measured after sorption of NS on HCP (cf. Fig. 4). Therefore, it can reasonably be assumed that the results of the UV–Vis measurements are at least order-of-magnitude correct for NS. The reference to the measurements of BZM sorption on cement is relevant in that those sorption measurements were carried out over short time scales, whereas in the present analyses the concentration of BZM in the pore water could possibly have been altered, due to chemical transformation or degradation reactions occurring over the relatively long time span of these experiments. This has to be assumed for LS-

40

12

A

B

35

10

[BZM] eq (mg dm-3 )

[BZM] eq (mg dm-3)

Experimental values Langmuir model

8

6

4

30

Experimental values Langmuir model

25 20 15 10

2

5 0

0 1

2

3

1

4

Desorption step

2

3

4

Desorption step

Fig. 7. Comparison between experimental and modelled equilibrium concentration in desorption of BZM from NS-HCP (A) and LS-HCP (B).

1.5

2.0

A

B

1.2

1.6 Reference solution (40 mg dm-3)

0.9

0.6

0.3

Optical density

Optical density

Pore water solutions

Pore water solutions Reference solution (42 mg dm-3)

1.2

0.8

0.4

0 240

260

280

300

Wavelength (nm)

320

0.0

240

280

320

360

400

Wavelength (nm)

Fig. 8. Corrected UV–Vis spectra of pore waters squeezed from each three specimens of NS-HCP (A) and LS-HCP (B) hydrated for 4 mo.

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

HCP. It is hypothesised in the literature that, with relatively short contact times (a few days), vanillin is produced from hydrolysis of LS (Swenson and Thorvaldson, 1960). Linear correction of the spectra shown in Fig. 8B by subtracting the spectrum of a standard solution of vanillin in 1 M K2CO3 (Sanders et al., 1995) indeed leads to a rather good agreement with the spectrum of the reference solution of LS in ACW-I. BZM concentrations were evaluated from the range of wavelengths, in which the spectra of the original compounds and those measured for the expressed pore waters are as similar as possible. For NS a wavelength range between 260 and 300 nm was chosen and for LS a wavelength range between 270 and 290 nm. It is assumed that the standard uncertainty of the results obtained at different wavelength reflects in a reasonable way the true uncertainty in concentration measurements attributed to the spectral changes. Note that such a procedure for the evaluation of BZM concentration is rather unusual in the literature, where BZM concentrations are most often calculated from a single wavelength. Unfortunately, most authors do not explicitly indicate, whether and how their results are corrected for background absorption due to cement–borne substances. The concentrations of NS and LS in the expressed pore waters were also measured by TOC. A summary of the

749

concentrations, as calculated from UV–Vis spectroscopy and TOC for samples collected after 4 and 20 mo, is given in Table 3. The uncertainties of the results from the UV–Vis measurements are fairly large due to the changes between the spectra of the expressed pore waters and those of the initial solutions. With regard to the relatively small sample-to-sample variations shown in Fig. 8, one can conclude that the uncertainties due to spectral changes are the dominant source of uncertainty in UV– Vis measurements. The standard uncertainties of TOC measurements, on the other hand, are dominated by sample-to-sample variations, while instrumental uncertainties are mostly negligible. As also shown in Table 3, the concentrations of organic carbon as determined from the UV–Vis measurements and from the TOC measurements are in fair agreement for NS but not for LS. From the two samplings it cannot be unequivocally decided whether this discrepancy can be interpreted as a chemical transformation reaction of LS. Within the range of uncertainty the ratios of concentrations measured in the 20 mo samples to those in the 4 mo samples are 1.5 ± 0.15 throughout. This is an indication that the difference in concentration may rather be explained by a decrease of available pore water due to the proceeding stage of hydration of the cement paste than by changes in the properties of the solid with respect to

Table 3 Pore water concentrations of BZM from analysis by UV–Vis spectrophotometry and TOC measurements. The results are averaged over the three samples for each type of BZM-HCP. The standard uncertainties given include sample-to-sample variation and instrumental uncertainties BZM

Sampling after 4 mo

Sampling after 20 mo

From UV–Vis

NS LS PP a

From TOC

From UV–Vis

(mg BZM dm3)

(mg C dm3)

(mg C dm3)

(mg BZM dm3)

(mg C dm3)

From TOC (mg C dm3)

224 ± 56 410 ± 42 n.m.a

121 ± 30 193 ± 20 n.m.a

192 ± 15 617 ± 6 4286 ± 175

324 ± 71 636 ± 46 n.m.a

175 ± 38 299 ± 22 n.m.a

258 ± 26 737 ± 14 4244 ± 451

Not measurable; gluconate was found to be below the detection limit (for details, see the text).

Fig. 9. Comparison of pore water concentrations of NS-HCP (A) and LS-HCP (B) based on different assumptions: (i) worst case assuming no sorption of BZM taking place, (ii) measurements and (iii) model calculations based on the Langmuir sorption model.

750

M.A. Glaus et al. / Waste Management 26 (2006) 741–751

sorption. This conclusion is further corroborated by the uniform change in optical density of the spectra. In the case of PP-HPC, gluconate was solely analysed among the numerous chemical components contained in PP. The concentration of gluconate was always below the detection limit of the HPAEC method, about 50 nM for the samples taken after 19 mo. In order to assess the predictive capability of the Langmuir sorption model, the results of the pore water analyses may be compared with calculated equilibrium concentrations in pore water. A tentative calculation of the concentrations of BZM dissolved in the pore water has been carried out by solving Eq. (5) (for NS and LS) or Eq. (6) (for gluconate) using the relevant parameters, viz., the inventories given in Table 1 and the sorption parameters for NS, LS and gluconate. It is assumed that 50% of the water initially added had been lost due to evaporation during storage or consumed by the hydration of cement, but the assumption is not very critical with respect to the results. As illustrated in Fig. 9, the predictions for the concentrations of NS and LS in pore water squeezed from BZM-HCP are lower than the experimentally determined values. The Langmuir sorption model makes a good qualitative prediction for BZM concentrations (decreasing from 40 g dm3 BZM to 0.012 g dm3 NS and 0.18 g dm3 LS, respectively). However, from a quantitative point of view, the model predictions are not conservative, because the pore water concentrations of NS and LS are underestimated (cf. Table 3). 4. Conclusions and implications The experiments show that different classes of compounds behave differently concerning sorption on and desorption from cement pastes. Macromolecular BZM, such as NS and LS are only susceptible to desorption to a minor extent, whereas polyhydroxy carboxylates, such as gluconate, are reversibly sorbed. Eq. (5), an empirical relationship, has been deduced to describe the reversible adsorption of a single species on a single type of sorption site resulting in a monolayer on the surface (Stumm and Morgan, 1981). In view of the heterogeneous mineralogical composition of HCP, it can hardly be expected that only one type of sorption site exists for BZMs. Although batch sorption data of BZM on HCP have often been described using a Langmuir type relation (Blank et al., 1963; Popescu et al., 1991; Fukaya and Kato, 1986; Massazza et al., 1981; Diamond, 1971; Miyake et al., 1985; Jolicoeur et al., 1994), it has also been proposed, based on surface analytical techniques, that e.g., NS forms multilayers on the surface (Uchikawa et al., 1992). For a safe prediction of pore water concentrations of BZMs, such as NS or LS, these need to be measured in pore water expressed from BZM-HCP and cannot be calculated conservatively using a sorption model. The fact that the type and strength of interaction between BZM and cement is dependent on the hydration state of the

cement paste (Jolicoeur and Simard, 1998) further aggravates the situation. However, for a fully hydrated cement paste, a fairly constant mineral composition, and thus a constant concentration of BZM, can be expected. A comparison between BZM inventories and sorption capacities of typical concretes indicate that the surface of BZM-HCP is undersaturated with respect to sorption of BZM. A large part of sorption sites is available for sorption of other ligands or radionuclides. This is an important conclusion with respect to the use of BZMs for the preparation of concretes used for the disposal of radioactive waste. From this point of view it can well be understood that the sorption of radionuclides, such as 63Ni, 152Eu, or 234 Th, and of a-ISA, which was used as a model ligand, has been found to be unaffected by the presence of NS, LS and PP (Glaus et al., 2004). Acknowledgements This work was partially financed by the National Cooperative for the Disposal of Radioactive Waste (Nagra). We gratefully acknowledge M. Brianza (MBT, Zu¨rich), for supplying samples of the BZMs tested in the framework of this study, M. Egloff (PSI), for the preparation of BZM-HCP, F. Huppertz and R. Rankers (ibac, Aachen), for the pore water expressions from BZM-HCP. References Blank, B., Rossington, D.R., Weinland, L.A., 1963. Adsorption of admixtures on portland cement. J. Am. Ceram. Soc. 46, 395–399. Connolly, J.D., Hime, W.G., Erlin, B., 1980. Analysis for admixtures in hardened concrete. In: Proceedings of the International Congress on Admixtures. The Concrete Society, The Construction Press, London, pp. 114–129. Costa, U., Massazza, F., 1984. Adsorbimento di superfluidificanti sul beta-C2S – modifiche del potenziale zeta delle particelle e della reologia delle paste (Adsorption of superplasticizers on beta-C2S – changes in zeta potential of particles and the rheology of pastes). Il Cemento, 127– 140. Daimon, M., Roy, D.M., 1978. Rheological properties of cement mixes: I. Methods, preliminary experiments, and adsorption studies. Cem. Concr. Res. 8, 753–764. Diamond, S., 1971. Interactions between cement minerals and hydroxycarboxylic-acid retarders: I. Apparent adsorption of salicylic acid on cement and hydrated cement compounds. J. Am. Ceram. Soc. 54, 273–276. Dodson, V.H., 1990. Concrete Admixtures. Van Nostrand Reinhold, New York. Ernsberger, F.M., France, W.G., 1945. Portland cement dispersion by adsorption of calcium lignosulfonate. Indus. Eng. Chem. 37, 598–600. Fukaya, Y., Kato, K., 1986. Adsorption of superplasticizers on CSH(I) and Ettringite. In: Proceedings of the Eight International Congress on Chemistry of Cements, Rio de Janeiro, vol. 3, pp. 142–147. Glaus, M.A., Laube, A., Van Loon, L.R., 2004. A generic procedure for the assessment of the effect of concrete admixtures on the sorption of radionuclides an cement: concept and selected results. Mater. Res. Soc. Symp. Proc. 807, 365–370. Hadermann, J., 1997. The pillars of safety. In: Grenthe, I., Puigdomenech, I. (Eds.), Modelling in Aquatic Chemistry. OECD Nuclear Energy Agency, Paris. Jeknavorian, A.A., Mabud, Md.A., Barry, E.F., Litzau, J.J., 1998. Novel pyrolysis gas chromatography/mass spectrometric techniques for the

M.A. Glaus et al. / Waste Management 26 (2006) 741–751 characterization of chemical additives in portland cement and concrete. J. Anal. Appl. Pyroly. 46, 85–100. Jolicoeur, C., Simard, M.A., 1998. Chemical admixture–cement interactions: phenomenology and physico-chemical concepts. Cem. Concr. Compos. 20, 87–101. Jolicoeur, C., Nkinamubanzi, P.C., Simard, M.A., Piotte, M., 1994. Progress in understanding the functional properties of superplasticizers in fresh concrete. In: Malhotra, V.M. (Ed.), Fourth CANMET/ACI International Conference on Superplasticisers and Other Chemical Admixtures in Concrete. American Concrete Institute, Detroit, MI. Kaplan, L.A., 1992. Comparison of high temperature and persulfate oxidation methods for determination of dissolved organic carbon in freshwaters. Limnol. Oceanogr. 37, 1119–1125. Massazza, F., Costa, U., Barrila, A., 1981. Adsorption of superplasticizers on calcium aluminate monosulfate hydrate. In: Malhotra, V.M. (Ed.), Developments in the Use of Superplasticisers. American Concrete Institute, Detroit, MI. Miyake, N., Ando, T., Sakai, E., 1985. Superplasticized concrete using refined lignosulfonate and its action mechanism. Cem. Concr. Res. 15, 295–302. Popescu, G., Olteanu, M., Peretz, S., 1991. Influence of calcium gluconate based additives upon the oil-well cement pastes. Il Cemento, 37–43. Ramachandran, V.S., 1972. Interaction of calcium lignosulfonate with tricalcium silicate, hydrated tricalcium silicate and calcium hydroxide. Cem. Concr. Res. 2, 179–194. Ramachandran, V.S., 1995. Concrete Admixtures Handbook, second ed. Noyes Publications, Park Ridge, NJ. Rocklin, R.D., Clarke, A.P., Weitzhandler, M., 1998. Improved long-term reproducibility for pulsed amperometric detection of carbohydrates via a new quadruple-potential waveform. Anal. Chem. 70, 1496–1501. Rossington, D.R., Runk, E.J., 1968. Adsorption of admixtures on portland cement hydration products. J. Am. Ceram. Soc. 51, 46–50. Sanders Jr., C.I., Sadeghi, S.S., Nelson, N.L., 1995. Limitations of the carbonate extraction/UV spectrophotometric method for determining

751

lignosulfonate-based admixtures in hardened concrete. Cem. Concr. Aggr. 17, 37–47. Spanka, G., Thielen, G., 1995. Untersuchungen zum Nachweis von verflu¨ssigenden Betonzusatzmitteln und zu deren Sorptions- und Elutionsverhalten. Beton 5, 320–327. Spiratos, N., Jolicoeur, C., 2000. Trends in concrete chemical admixtures for the 21st century. In: Malhotra, V.M. (Ed.), Proceedings of the Sixth CANMET/ACI International Conference on Superplasticizers and Other Chemical Admixtures in Concrete. American Concrete Institute, Detroit, MI. Stumm, W., Morgan, J.J., 1981. Aquatic Chemistry, second ed. Wiley, New York. Swenson, E.G., Thorvaldson, T., 1960. Detection of lignosulphonate retarder in cement suspensions and pastes. ASTM Special Technical Publication, vol. 266, pp. 159–169. Uchikawa, H., Hanehara, S., Shirasaka, T., Sawaki, D., 1992. Effect of admixture on hydration of cement, adsorptive behavior of admixture and fluidity and settig of fresh cement paste. Cem. Concr. Res. 22, 1115–1129. Van Loon, L.R., Glaus, M.A., Stallone, S., Laube, A., 1997. Sorption of isosaccharinic acid, a cellulose degradation product, on cement. Environ. Sci. Technol. 31, 1243–1245. Wexler, A.S., Brako, F.D., 1963. Use of the ultraviolet spectrophotometry in determining lignosulfonate additions in cement. Mater. Res. Stand. 3, 364–368. Williams, A., Diamondstone, B., Ellison, B., Haesselbarth, S., Kaarls, W., Mansson, R., Mo¨ller, M., Taylor, H., Thomas, P., Wegscheider, B., Wood, W., De Halleux, R., 1995. Quantifying uncertainty in analytical measurement. In: EURACHEM Working Group on Uncertainty in Chemical Measurement. Crown Copyright, London, UK. Yilmaz, V.T., Glasser, F.P., 1991. Early hydration of tricalcium aluminate–gypsum mixtures in the presence of sulphonated melamine formaldehyde superplasticizers. Cem. Concr. Res. 21, 765– 776.