M ETODOLOGÍA DE TOMA DE MUESTAS Y ANÁLISIS UNE EN 13725

In the production of silicon and ferrosilicon alloys by reduction of silica with carbon in an electric furnace, some SiO vapour is formed. This by- product oxidises in the vapour phase on contact with air and the resulting fume is condensed to yield a very fine silica powder, typically with a particle size of 100–500 nm and a nitrogen BET surface area around

20 000 m2/kg. Quality varies with source, the best material having silica

contents in the range 90–94%, although Wolsiefer et al. (1995) made concretes giving similar performance test results with silica contents in the range 79–95%. However, even a surface coating of carbon can greatly reduce pozzolanic activity. The term microsilica is widely used and includes very pure silicas such as those manufactured as reinforcing fillers

for plastics. These can have surface areas above 150 000 m2/kg.

Condensed silica fume (CSF, an abbreviation widely used in the literature) has a very low bulk density with values down to one-tenth of

the true particle density of silica glass, 2300–2400 kg/m3. It is frequently

transported in the form of compacted granules, as pellets, or as an aqueous slurry with an added dispersant such as a superplasticiser (Section 10.1). Minor impurities include: alumina, magnesium oxide, lime and alkalis and possibly some iron and iron silicide. Pelletised silica fume is suitable for milling with Portland clinker in a cement mill. Granules may be broken down and dispersed in a concrete mixer by the aggregate. Aged agglomerated aqueous suspensions may need to be redispersed before use. Wolsiefer et al. used sixteen samples from ten North American sources covering this range of possibilities. Portland/silica fume cement is specified in ENV 197 as a blend of 6–10% with clinker (Table 9.1). The silica content of the fume must exceed 85%, the specific surface exceed

15 000 m2/kg and the loss on ignition be less than 4%.

Silica fume (SF) particles possess no long range order and29Si NMR

reveals that it is highly condensed with bulk silicon atoms linked to four neighbours by bridging oxygen atoms (Bijen and Pietersen, 1994). After prolonged soaking in water, surface oxygens in such a silica are present as silanol (Si–OH) groups. It is highly reactive and, in a blended Portland cement paste, the portlandite content may pass through a maximum in

under one day as it is consumed in the pozzolanic reaction which, if the SF is well dispersed, is likely to be largely completed after two weeks. Rates of reaction of individual phases in a silica fume cement paste can be

followed by a combination of29Si NMR and thermal analysis (Justnes et

al., 1993). Taylor (1990) calculated that up to about 35 g of microsilica

can react with 100 g of Portland cement, assuming that the C/S ratio of the C–S–H formed cannot fall below 0.8.

The use of silica fume in concrete is limited by its cost and availability. To offset the increased water demand introduced by this fine material, a superplasticiser is incorporated into a concrete mix and the accelerating effect of silica on alite hydration may make the inclusion of a retarder desirable (Section 10.1). The increase in cohesiveness of a mix it produces can be of value, in underwater concreting for example (Neville, 1995). Efficient dispersion of silica fume in a concrete is not only necessary to ensure maximum benefit from its contribution to properties such as strength and permeability but also to avoid a possible alkali-silica reaction in which an agglomerate of silica particles is big enough to act as an especially reactive aggregate forming an alkali-rich swelling gel (Lagerblad and Utkin, 1994).

The high reactivity of silica fume leads to an increase in early heat liberation in a concrete, although as a replacement for Portland cement SF progressively reduces total heat liberation as levels increase above 10%. Incorporation of silica fume can enhance concrete strength (increases by a factor of around 2 are possible) but the principal practical benefit is reduced permeability and hence improved durability. The latter derives from the acceleration in the early hydration of alite combined with the high pozzolanic reactivity of the silica which accelerates the consumption of portlandite and the formation of C–S–H. Total porosity does not appear to be reduced significantly by the pozzolanic reaction but, as observed with fly ash, the pore size distribution is modified by an increase in the proportion of fine pores in the C–S–H gel. Some larger but unconnected pores are formed where portlandite crystals have dissolved and reacted with the silica. The densification of the transition zone at the paste– aggregate interface can contribute to a reduction in the permeability of a concrete by a factor of around 100.

9.5 Blastfurnace slag and blastfurnace slag cements

9.5.1 Composition

In the production of iron in a blast furnace, a flux such as limestone or dolomite (and sometimes bauxite) is added to the charge of iron ore and coke. This produces a liquid slag at 1400–1550º which contains the siliceous and aluminosilicate impurities originating in the iron ore and the

coke and which is immiscible with the liquid iron. The principal oxide components of the slag are thus lime, silica and alumina and compositions are often represented in a ternary diagram such as Fig. 9.2, although significant amounts of magnesia (up to ca. 20%) may be present when dolomite is used. Slow cooling of the slag after it is tapped from the furnace allows crystallisation to occur and the product has little or no cementitious value. The principal crystalline phases usually formed are

merwinite (C3MS2) and melilite, which is a solid solution of gehlenite

(C2AS) and a˚kermanite (C2MS2). However, rapid cooling of the liquid

slag, involving mechanically converting it to small droplets (granulation) or pellets, inhibits crystallisation (Fig. 9.6). Granulation produces a particularly high proportion of glass (usually >90%) and, after grinding, a product which resembles sand and has latent hydraulicity is obtained (ground granulated blastfurnace slag: ggbs). As the term ‘latent’ implies,

Merwinite Melilite Melilite Calcite Quartz Quartz (a) (b) 38 34 30 26 22 20 Degrees 2θ Cu Kα

Fig. 9.6. X-ray diffraction patterns of two slags: (a) glass with some crystallisation; (b) glass halo (Moranville–Regourd, 1998)

an activator is necessary for this property to be developed and an alkali (Section 10.4), lime, gypsum and Portland cement are effective, although major usage involves only the latter. In its application, ggbs may be blended with Portland cement or interground with clinker to an optimum

SO3level.

Slag cements have been produced since before 1900 in continental Europe and ENV 197 covers the wide range of compositions commercially available (Table 9.1). In the UK, the great majority of ggbs is used by direct introduction into concrete in the ‘ready mixed’ industry. The required chemical and physical properties of the slag for this application are specified in BS 6699: 1992. The composition of slag from a particular plant depends on the purity (source) of the iron ore (usually an iron oxide — hematite or magnetite), the coke and flux employed and, unless sources of these are changed, variability is limited. However, compositions vary from one plant to another and the range encountered worldwide is considerable. Data for some European plants are given in Table 9.4. The reducing conditions in the blastfurnace, produced by the carbon monoxide formed by the partial combustion of the coke in the charge, result in any iron and manganese in the resulting glass being in the divalent state and some metallic iron may also be present, although usually these are limited to tenths of one percent and slag is almost white in colour. The magnesium oxide content (limited to 14% in BS 6699) does not cause a soundness problem because the magnesium ion is distributed in the glass and not present as the mineral periclase (Section 6.6). Much of

the sulfur is present as the reduced form S2 and this is contained in the

glass. Hydration of the slag in the presence of cement produces a blue- green colour which subsequently fades as oxidation takes place.

ENV 197 specifies that a blastfurnace slag for blended cement manu- facture must have a glass content of at least 66.6% and that the sum of the Table 9.4. Composition ranges for 16 granulated blastfurnace slag sources in seven European countries (Livesey, 1993)

Oxide Range: % Oxide Range: %

SiO2 33.9–38.1 K2O 0.31–0.72 CaO 36.6–42.8 Na2O 0.20–0.45 Al2O3 8.8–13.3 SO3 0.0–2.4y MgO 6.7–12.8 S2 _0.8–1.3 Glass content 89–99* Ratios (C + M)/S 1.27–1.47 (C + M + A)/S 1.53–1.85

* 15 samples — one sample had a glass content of only 75%. y 15 samples: 9  0.1%; 6 > 0.5%.

oxide mass percentage contents (CaO + MgO + SiO2) be at least 66.6%. In

addition, the ratio (CaO + MgO)/SiO2must exceed 1.0. The requirement

for the remainder is simply that it be Al2O3with only small amounts of

other oxides. A number of attempts, with limited success, have been made in the past to correlate the oxide composition of slags with their potential hydraulicity (Smolczyk, 1980). The basicity ratio or modulus, usually written as (C + M + A)/S, where the symbols represent the mass percentages of these oxides in the slag, has been the most commonly employed and hydraulicity rated as good, moderate or poor when the value lies above 1.0, from 0.5 to 1.0, and below 0.5, respectively. Minimum values of 1.0 or 1.4 have been specified in some national standards.

The constitution of the glass in blastfurnace slag has been examined by TMS (Section 7.2) and Raman and NMR spectroscopy. These techniques

have shown that it contains isolated SiO44 and dimeric Si2O76 anions and

Ca2+, Mg2+, Al3+and Al–O+as cations. A review of the structural aspects of

this glass (cation oxygen coordination numbers, microheterogeneities, and structural and surface defects) and the physical methods employed in their examination has been given by Regourd (1998). Crystalline material, when present, is identified by optical microscopy and X-ray diffraction (Fig. 9.6). Merwinite and melilite are often found as inclusions in the glass, in sizes ranging from dendrites less than 100 nm wide to crystals visible in the optical microscope. Metallic iron may also be present as inclusions.

9.5.2 Hydration

The reactivity of a slag depends on its glass content and the nature and concentration of the defects present in this phase, both of which are affected by the rate of cooling to below about 800º which it has experienced. As in all phase-interface reactions, the particle size distribution (fineness) of a slag is also a primary variable. Spectroscopic techniques such as X-ray fluorescence have been used to establish that magnesium and aluminium ions in slags and synthetic glasses in the C–M–A–S system are located in both four coordinated (tetrahedral) and six coordinated (octahedral) sites and hydraulicity has been linked to the content of the latter. Regourd (1986) considered that the hydraulic reactivity of a slag is determined by the proportion of octahedral aluminium in the glass which dissolves first in an alkaline medium, the calcium silicate network being hydrolysed subsequently. A pH above that which calcium hydroxide can provide is not required to initiate slag reaction as it is for fly ash. Together with the role of structural defects, including incipient crystallisation, the several factors influencing slag hydration make it hardly surprising that simple bulk composition- reactivity relationships have proved of limited value.

The early hydration of Portland cement clinker is accelerated in the presence of blastfurnace slag and the rate of hydration of the latter is dependent on the efficiency of its dispersion (Lee et al., 1987). Lumley et

al. (1996) used a solvent extraction method (Section 4.2) to determine the

unreacted slag in pastes of several slag cements hydrated at 20º at water/ binder ratios of 0.4 and 0.6. They found that the percentage of the slag hydrated at 28 days was 30–55% and at 2 years 45–75%. These values are

similar to that for belite (C2S) in Portland cement. The slags they

examined differed significantly in reaction rate which, for their samples, was related to differences in hydraulic moduli and fineness. Reaction rate of a slag decreased with decreasing water/binder ratio and with an increasing proportion of slag in the blended cement.

The principal products of hydration of a slag cement are C–S–H (C/S ratios in the range 0.9–1.6 have been found), AFt and/or AFm phases, but with the portlandite expected from the Portland cement component reduced in quantity by interaction with the slag. Microanalytical determination of Al/Ca ratios in the electron microscope suggested that C–S–H and AFm phases are intermixed on a submicron scale (Harrisson

et al., 1987). With high slag proportions, the amount of portlandite

formed may pass through a maximum as hydration proceeds. Stra¨tlingite

(C₂ASH₈), siliceous hydrogarnets (solid solution series C₃AS₃–C₃AH₆)

and a hydrotalcite type phase are formed. Hydrotalcite phases vary in

composition and are related to brucite (Mg(OH)2), having a layer

structure in which some of the Mg2+ions are replaced by Al3+ and some

of the OH ions by CO2

3 .

Hydration products form separately at the surface of both clinker and slag particles and extend inwards with time. The level of heterogeneity found in an examination of the microstructure depends on the mobility of the ions involved in an alkaline medium. Bijen and Pietersen (1994) found that a net movement of calcium, silicon and aluminium out of the slag occurred while magnesium acted as an immobile marker. This would be expected from the low solubility of the magnesium ion in an alkaline medium. They also observed that the reaction of the slag was initiated at the onset of cement hydration, at a pH around 12, and that the subsequent rise observed with a neat cement paste did not occur. Glasser et al. (1987) noted that pH rises little above 12 and that a slag/cement paste also differs from that of neat Portland cement paste in producing a chemically

reducing environment in the pore solution, involving the S2 _{ion, when}

the slag content is as high as 70%. On hydration, the sulfide is considered to be located in the AFm phase and is eventually oxidised to sulfate. Lachowski et al. (1997) found the development of the microstructure in a slag/Portland cement blend to be continuing after six years’ curing at 98– 100% rh, with unreacted slag still present.