2.1 Ground granulated blast furnace slag

GGBS is the rapid cooling floating part of iron mix at the exit of metallurgic furnaces. It can be cooled by water or ambient air. It is a glassy/crystalline composite in which the glassy part can reach 95% of the entire material mass. The slag used in this study is a fine white powder, with a 4500 cm² g^–1 specific area. Its composition is given in Table 1.

2.2 Chemical activation

To understand the activator mechanism in the hydration process of slag, four activators have been studied: Na₂CO₃, natural gypsum (CaSO₄$\dot{}$2H₂O), Ca(OH)₂ and Portland cement [6]. These activators were chosen for their slag activation ability and because they do not present large hazards to be handled. The different activators performances are compared to those of Portland cement, which is used in ordinary manufactured slag cement.

2.3 Sample preparation

The entire samples are prepared in the same way, using a water/total binder mass ratio of 0.5. The slag and the activator at the chosen ratio were mixed with water. The samples were then mould in cylindrical test tubes of 40-mm diameter and preserved safe in dark room in a pool of water until the date of the compressive tests. After compression, a central part of the sample was taken and then the hydration process stopped by using acetone/ether mix. The composition of the different samples is shown in Table 2.

2.4 Solid state NMR

NMR study is based on quantitative ²⁹Si (spin I = 1/2) high-resolution magic angle spinning experiences [7]. High spinning rate decreases the dipolar interaction and the anisotropic part of the chemical shift, and accumulations (up to 10 000 spectra) are sufficient to obtain a low signal/noise ratio spectrum. Spectra are obtained with a Bruker ASX solid-state NMR spectrometer operating at 11.7-T magnetic field. Spinning rate was about 7 kHz in 7-mm ZrO₂ rotors. All experiences use single π/2 pulses, recycling delay of 5.75 s, without ¹H decoupling. The relaxation times are respected, and all the spectra are weight-normalised. The ²⁹Si isotropic chemical shifts are referenced relative to tetramethylsilane Si(CH₃)₄ (TMS) at 0 ppm; Si[(CH₃)₃]₈Si₈O₂₀ (Q8M8) is used as a secondary reference.

In order to reduce quadrupolar effects of ²⁷Al nuclei (spin I = 5/2), NMR experiences may be carried out at high magnetic field and high spinning rate. Hence, ²⁷Al NMR was realised in high-field magnets (11.7 T and 19.6 T), with spinning rates up to 30 kHz in 2.5-mm ZrO₂ rotors. All experiences employed single π/12 pulses sequences. Chemicals shift are referenced relative to a 1.0 M AlCl₃–6 H₂O solution.

The assignment of ²⁹Si and ²⁷Al resonance peaks is based on chemical shift range established for minerals and cements [8,9].

2.5 Compressive tests

Compressive tests were made using a hydraulic press with a maximum force of 1500 kN. Each formulation was tested under compression for different hydration times: 7, 14, 28, 90 and 500 days. These times are conventional in French civil engineering [18]. For 90 days of hydration in Portland cement, the hydration ratio is considered to be equal to 95%.

3.1 Hydration products

Basic activator destroys the glassy structure of the slag and releases in the water/slag solution the different elements necessary for the constitution of the hydration products (C–S–H and secondary products of aluminates hydration). Typically, in classical use of GGBS as a substitution of Portland cement, cement alkali plays the role of the slag activator. The hydroxyls ions OH^– cut the Al–O and Si–O bonds [5]. Their combination with Ca²⁺ in solution leads to the formation of calcium silicate hydrate (C–S–H) [(CaO)_x(SiO₂)_y(H₂O)_z], hydrogarnet (Ca₃[Al(OH)₆]₂), sulphoaluminate hydrate phases like ettringite [(Ca₃Al(OH)₆.12H₂O$\dot{}$(SO₄)₃$\dot{}$2 H₂O)], and Afm phases [(Al₂O₃–Fe₂O₃-mono)] [10].

C–S–H can be represented as short chains of silicates with a microstructural order at nanometre scale. The microstructural model of Tobermorite can be used to approach the C–S–H structure. The nature and the structure of C–S–H in Portland cement were discussed by several authors [11,12]. A schematic representation is shown in Fig. 1. Due to their ill-crystallised structure, it is very difficult to study the microstructure by XRD or SEM. Solid-state NMR does not matter with the regularity of the crystalline network. The stoichiometry of such hydrated species is not established. It depends on many parameters and some are not yet identified. Nevertheless, for Ca/Si ratios below 1.0, between 1.0 and 1.5 and upper 1.5, Nonat has identified α, β, and γ types and structures for C–S–H [13]. In other Ca/Si ratios cases, tendencies of long or short chains formation can be evidenced, but no precise chemical formula.

3.2 Mechanical performances

Experimental curves are shown in Fig. 2. For all the activators, without special curative treatments or special additives, we can observe that the mechanical performances are equal or better than those obtained with a classical Portland cement, except with Ca(OH)₂ activation. The reference samples using Portland cement as the activator, after 7 days of hydration, reach 10 MPa, and a maximum of 30 MPa after 90 days for the M3 sample. These results agree with the classical performances of a slag cement grout formulation.

Na₂CO₃ activation for short hydration times (7 and 14 days) is the most efficient (more than 20-MPa compressive strength at 7 days). At medium and long time of hydration (28 and 90 days), the mechanical performances reach respectively 20-MPa and 24-MPa compressive strength for M4 and M5. Na₂CO₃ seems to be the more efficient of this study (up to 40 MPa after 90 days of hydration, M5 formulation) but for high activator ratio, some rheological problems can occur, due to a quick rigidification of the grout.

Gypsum activator does not present after 7 days of hydration enough mechanical properties to be measured, whatever the activator concentration is. Nevertheless, this activation presents interesting rheological properties for civil engineering works, considering the great rigidification delay of the grout. After 14 days of hydration, the compressive strength is more than 12 MPa for M7 and 25 MPa for the two others. After 90 days of hydration, M7 reaches 27 MPa, whereas medium and higher concentrations reach 30MPa compressive strength. With gypsum, the mechanical performances are nearly the same, whatever the activator concentration is.

One also obtains very tight results for Ca(OH)₂ whatever the concentration is. After 90 days of hydration, the compression strength is only 15 MPa, which is not enough for an industrial use.

In the case of cement, slag or other hydraulic binder, many other parameters like granular stacking, porosity or thermal shrinkage have a great influence on the mechanical performance.

3.3 ²⁹Si NMR

The various hydration products of hydrated slag can be observed by the solid-state NMR of several nuclei. The most relevant nuclei are ²⁹Si (spin ¹/₂), ²⁷Al (spin ⁵/₂), ²³Na (spin ³/₂), ²⁵Mg (spin ^–5/₂), ¹H (spin ¹/₂ ) and ⁴³Ca (spin ^–7/₂).

First of all, the C–S–H slag study is done by the evaluation of the different electronically environments of the tetrahedron SiO₄, elementary unit of the chain structure. Second, the NMR study of ²⁷Al, which can exist in tetrahedral, pentahedral and octahedral coordination, gives information about the hydrated aluminates, the sulphoaluminates products (octahedral coordinence), the possible substitution of aluminium in the SiO₄ chain (tetrahedral coordinance) or in the CaO plane (pentahedral and octahedral coordinence).

Fig. 3 shows the different ²⁹Si NMR spectra obtained after 28 days of hydration for the higher activator concentration. For all the activators, spectra can be decomposed into six elementary peaks. According to the chemical shift table of ²⁹Si [9], the attribution of these peaks may be: Q⁰ (–74ppm), Q^1(1Al) (around –75ppm) Q¹(–77ppm), Q^2B (–82ppm), Q^2(1Al) (–80ppm) and Q² (–84ppm). Qⁿ attributions correspond to the connectivities of SiO₄ species, Q^2B (where ‘B’ means ‘bridging’) to a middle tetrahedron of the dreierkette C–S–H chain structure. Q⁰ species corresponds to the anhydrous slag. Nevertheless this first decomposition into Qⁿ unit does not allow us to determine the geometry of the hydrates: the superposition of Q^1(1Al) with Q⁰ peaks and Q^2B with Q^2(1Al) peaks impedes the calculation of an elementary-chain length. The growth or decrease of the different species when varying the hydration times can be evidenced. The most important decrease of Q⁰ species occurs for Na₂CO₃ activation, which corresponds to the high mechanical strength measured. As shown by the ²⁹Si NMR spectra, the different activators have not the same effect on the hydration process. The role of the alkali cation seems to be important in the slag activation/hydration process. A comparison between the different spectra of a same activator family for different activation ratio is represented in Fig. 4. The decreasing kinetics of the Q⁰ unit increases with the activator ratio.

3.4 ²⁷Al NMR

²⁷Al NMR for classical cement composition is very interesting to evaluate the occurrence of the different hydration products containing aluminium, which can cause serious damages to a construction.

²⁷Al NMR spectra for all the activators are shown in Fig. 5. In octahedral site, ²⁷Al NMR shows three different peaks with Portland and gypsum activation [14]: ettringite (a special form of Aft phases) at 12 ppm (position of the maximum peak resonance intensity), monosulphoaluminate at 9 ppm and a non-assigned peak at 3.5 ppm are formed [17]. With Na₂CO₃ and Ca(OH)₂, Afm phases are present, but ettringite and monosulphoaluminate have disappeared (no SO₄^2– in water/slag solution).

The aluminium inside the anhydrous slag is on a tetrahedral site. By using high-field MAS NMR spectrum (19.6 T), an improvement of the spectrum resolution is obtained (Fig. 6): a great peak appears at 71 ppm, which could be the signal of the silicon substitution by aluminium inside the C–S–H chain. By coupling this result with the substitution hypothesis, it is possible to introduce a hydrated product for hydrated blast furnace slag: C–Al–S–H (alumina-silicate of hydrated calcium). Taking into account this substitution phenomenon in ²⁹Si spectra, we can assign the Q^2B peak as a double peak Q^2B and Q^2(1Al) [15,16].