MICROSILICA CONTANING REFRACTORY CASTABLESINFLUENCE OF ADDITIVES ON PLACING PROPERTIES

Presented at the 47th annual symposium on Additives for monolithics, at St. Louis, USA, March 23-24, 2011.

Microsilica containing refractory castables - the influence of additives on placing properties

Bjrn Myhre, Elkem Silicon Materials, P.O Box 8126 Vaagsbygd, 4675 Kristiansand, Norway. bjorn.myhre@elkem.no

Abstract:
The surface of microsilica is covered by numerous silanol groups that are able to dissociate in an aqueous medium. The dissociation is responsible for interactions in castables with microsilica and cement. When microsilica is used in castables such interactions are normally modified by use of a

suitable additive. The additive disperses, but the amount and type may have side-effects that may not be expected. Both set time as well as ageing of the castable has been found very dependent on type and amount of additive. In this paper some of the interactions in microsilica containing castables are shown and explained.

Surface of Microsilica
Although the overall composition of microsilica is SiO2, the surface of a microsilica particle is not plain siloxane bonds, but is also partially hydroxylated and hydrated. Producers of fumed silica (particularly Evonik Degussa) have been very active in the characterisation and modification of

the silica surface in order to tailor specific properties of precipitated and fumed silica. In general, several connectivities of surface silicon have been identified by means of FT-IR- and NMRspectroscopy, as shown in Figure 1.

Figure 1. Surface Si connectivities idientified on silica.

Based on theoretical and experimental results , it seems that the maximum density of silanol (Si-OH) groups on the surface of silica is approximately 4.6nm-2. The surface of microsilica is covered by silanol groups; although the presence of impurities plays some role, measurements have indicated silanol densities ranging from approximately 2 to 4.5nm-2 for microsilica from silicon production. The good correlation between theoretical values and those found for microsilica may indicate that the impurities of the microsilica are not smothered out on the surface of the microsilica particles but rather as discrete particles or dissolved in the silica. A combination is probable. The presence of the silanol groups makes the microsilica easy to disperse in aqueous systems. Depending on the pH, a fraction of the silanol groups dissociate resulting into a negatively charged surface. The surface charge or zetapotential can be measured and values higher than approximately 25mV (positive or negative) are often taken as an indication of stable suspensions. Figure 2 shows a measurement of such a zeta potential for microsilica 971, and we see that the microsilica has a negative charge over the whole range. Due to dissociation of the silanol groups, the negative surface charge increases with pH up to approximately pH 7. At higher pH than 7, the zeta potential flattens out and at still higher pH the microsilica starts to dissolve.
MS titrated by 1M HCl 0.0 Zeta-potential (mV) -5.0 -10.0 -15.0 -20.0 -25.0 -30.0 -35.0 -40.0 0.00 2.00 4.00 pH 6.00 8.00

Additional to the microsilica, refractory castables often contain reactive alumina in combination with microsilica. Looking at the Zeta-potential of the alumina, being positive over most of the range, Figure 3, one could expect that microsilica and alumina would attach and flocculate due to the opposite charges. In practice this does not happen as the pH of castable mixes normally are on the basic side of neutral where the alumina has very low or even negative surface charge.

CTC50 titrated by 1M HCl 40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 -5.0 0.00

Zeta-potential (mV)

2.00

4.00 pH

6.00

8.00

10.00

Figure 3 Zeta potential of Alumina CTC50 from Almatis as a function of pH. A 10% slurry titrated with HCl and with zeta potential being measured using electro-acoustic techniques (Acoustosizer, Collodial Dynamics)

Microsilica and deflocculants

A refractory castable can be visualised as an aggregate structure with voids filled by aqueous slurry of microsilica, cement and other fines. The negative charge of the microsilica makes it vulnerable for attack by cations which causes gelling or solidification of the system. The concentrations necessary for gelling are depending on type of cation but a quick calculation of the number of sites (silanol groups) per gram microsilica estimates that approximately 0.15mmol per gram of microsilica could be sufficient for a full coverage. This quantity is in many (less pure) microsilicas available as water soluble impurities in the microsilica itself, and an autogenous gelling of such a slurry (e.g. 50%) may occur. Although any cation can be used for gelling microsilica slurry, the

Figure 2 Zeta potential of microsilica 971 as a function of pH. A 10% microsilica slurry titrated with HCl and with zeta potential being measured using electro-acoustic techniques (Acoustosizer, Collodial Dynamics)

Zeta-potential (mV)

characteristics of the resultant slurry are dependent on the charge and size of the cation that attaches. In the case of monovalent cations, the gelling action has been attributed to a mechanism where its the hydrated cations that link adjacent silica particles. Such a gel normally feels quite soft and easy to liquify by agitation, while that from polyvalent cations tend to have a more brittle and harder appearance. With polyvalent cations, it is easy to imagine that the excess positive charge of the cation links with another negative charge on an adjacent microsilica particle. In this way threedimensional networks may form and this is what we normally regard as gelled microsilica. In castables, cations are generally available from the cement and aggregates, and it is likely that the initial setting of most castables is closely related to gelling of the microsilica containing slurry phase. In the case of ultralow and no-cement systems with microsilica, it is probable that most of the setting is dependent of this solidification although it is often triggered by small amounts of additives that promote the gelling. Such additives can e.g. be small quantities of cement. The absorption of polyvalent cations on the silanol sites is most likely a dynamic process and requires a certain concentration in the water phase before absorption gets significant. The amounts of soluble cations necessary for gelation are therefore probably higher than just to attach to the silanol sites, but it seems reasonable to expect that the order of magnitude may be similar. The tendency to react with cations to produce a more or less rigid gel, creates problems if used in densely packed castable systems where cement is present. The calcium and aluminate cations from the cement dissolve and attack the otherwise fluid bond phase. To prevent this coagulation, the calcium ions have to be prevented from absorbing on the microsilica. This may either be done by having a pH below 5, but as this may affect setting; addition of a proper amount of a polyelectrolyte or a surface active agent is a common stratagem. The surface active agent, normally termed deflocculant (in Portland cement based systems termed plasticiser or superplasticiser), is believed to absorb to the surfaces thus preventing the absorption of calcium but it also creates an equal, often negative charge of the particles of the bond system. As the particles then will repel each other, casting can take place at reduced water additions. Deflocculants commonly used are polyphosphates, e.g. sodium

hexametaphosphate (Calgon) at approximately 0.2 wt% addition level, and polyacrylates (Darvan 811D, 0.05 wt %). The effect of Calgon on microsilica and reactive alumina is illustrated in Figure 4.
0.0 -10.0 -20.0 -30.0 -40.0 -50.0 -60.0 -70.0 0 0.5 1 1.5 2 2.5 Surfactant content (mg/m2)
10% MS Slurry 5% Calgon

10% CTC50 slurry 5% Calgon

Figure 4: The effect of addition of Calgon. Zeta potential of 10% aqueous slurry of microsilica 971 or CTC50 as a function of Calgon addition expressed as mg/m2 surface.

The deflocculating mechanisms involved falls normally into two main groups or combination thereof: 1) Steric stabilisation, the deflocculant attaches to the fines and the size of it prevents agglomeration of particles 2) Electrostatic stabilisation, the deflocculant creates an equal electrical charge on the surface of the fines. Electrostatic repulsion prevents particles from agglomeration. 3) Combination of 1) and 2):

Additional to the above mechanisms, use of deflocculants makes it possible to maintain flow of refractory castables for some time, sufficient to place the castable before setting (gelling) commences. Sometimes setting may get adversely affected though, giving unacceptably long or short set times. This has been explained with several mechanisms. For long set time with polycarboxylate ethers, one explanation is that the surface of the cement get shielded by the side

chains and that longer side chains consequently 2 retard more than shorter .
Self-flow (%)

100 90 80 70 60 50 40 30 20 10 0 0 0.2 0.4 FS20 dosage (mg/m2) 0.6 0.8

24 m /g

18 m /g

Deflocculants and flow: To check quality of microsilica or cement or other components of low cement castables, Elkem has since the early 1990ies, used a model LCC based on white fused alumina with 6% CAC and 8% microsilica. The castable is composed to follow an Adreassen type PSD with a q-value close to 0.25 giving flow properties close to self-flowing at 4.15wt% water (=13 volume %). The castable composition is given in Table 1. Table 1 Standard LCC castable composition. [wt%] Microsilica (18 or 24 m2/g) 8 CA-14 [wt%] 6 White Fused Alumina. 3-5mm 10 White Fused Alumina. 0,5-3 mm 32 White Fused Alumina. 0-0,5mm 16 White Fused Alumina. -74 micron 12 Calcined alumina: CT 9FG 16 Deflocculant: Castament FS20 0,01-0,1* Water [wt%] 4.15 * Standard dosage, 0.05% To check the effect of the Castament FS20 on flow of a castable, the standard LCC mix (Table 1) was used with the amount of deflocculant as the variable. Two different microsilica qualities with surface area of 18 and 24m2/g were used. In Table 1, the composition of this mix is given. In Figure 7, the resultant self-flow values are shown as a function of deflocculant addition per microsilica surface area (mg/m2). The two microsilica qualities gives very similar results when expressed as a function of surface area, only the maximum self flow differs somewhat. Although lower self flow was obtained with the microsilica with the higher surface area, and conclusions in the direction of lower flow at higher surface area would be tempting, we do not have data proving that this actually is the case.

Figure 5. Self flow of white fused alumina based LCC as a function of deflocculant addition. Effect of dispersant Above, the good correlation between microsilica surface area, flow and amount of additive was shown. Those findings were based on flow measurements of LCC based on fused alumina, but the same correlations exist for bauxite based systems. Table 2: Bauxite based LCC compositions. (%) 1 2 Chinese bauxite 130 30 4mm Chinese bauxite 030 30 1mm Fused Al. -74 mic 18 18 Secar 71 6 6 Calcined alumina 8 8 Microsilica (971U) 8 8 FS20 0.01-0.15 Calgon 0.01-0.40 Water 5.50 5.50 In Figure 6, self-flow of castable 1 and 2 (Table 2) is presented as a function of addition of dispersant in mg per square meter microsilica surface. In practical terms, an addition of 1 mg/m2 equals to an addition level of approximately 0.2%. Earlier results on white fused alumina systems(Table 1) have shown similar dependencies for FS20 and Calgon, as shown in Figure 7.

120 100 Self-flow (%) 80 60 40 20 0 0.00

Self-flow (%)

120

FS20 Calgon

FS20

100 80 60 40 20 0 0 0.5 Dosage (mg/m )

2

Calgon Darvan 811D

0.50

1.00

1.50

2.00
2

2.50

3.00

1.5

Dosage (mg/m )

Figure 6: Effect of dosage of dispersants on selfflow of bauxite-based LCC (number 1&2 in Table 2). If Figure 6 and 7 are compared, it is seen that for FS20 the appearance is quite similar, while for Calgon it is in principle similar, but somewhat skewed. There can be several reasons for the slightly distorted shape of the curve for Calgon, but the first that come into mind is that the phosphate reacts with impurities of the bauxite, thus necessitating higher dosages than in the purer white fused alumina system. Further, as also Darvan 811D is included, it should be commented. Darvan being a poly-acrylate as opposed to FS20 being a carboxylate-ether differs by having a pronounced optimum addition level as opposed to the more flat plateau level that is seen for FS20. This is probably connected to the mode of stabilisation of colloidal slurries of the two different additives. Darvan is by electrostatic repulsion, FS20 by a combined steric and electrostatic mechanism.

Figure 7: Effect of dispersant addition on self flow of white fused alumina based castables with 6% cement and 8% microsilica, 4.15% water. If we look at the Zeta-potential for the combination of microsilica and the additives in question the following is found, Figure 8.
0 -10 Zeta-potential (mV) -20 -30 -40 -50 -60 -70 -80 -90 0 0.5 1 1.5 2 Surfactant content (mg/m2)
FS20 Darvan 811D Calgon

Figure8: Zeta potential of 10% aqueous slurry of microsilica 971 as a function of dispersant addition expressed as mg/m2 microsilica surface. Figure 8 shows the zeta-potential of a microsilica slurry as a function of dispersant addition. Here, Darvan 811D, Calgon and FS20 have been tested. Two phenomena may be noticed, firstly the lowering (or lack of it) of zeta-potential caused by FS20 indicates that the stabilisation mechanism is primarily of steric (uncharged) nature. The second observation is that there is a good correlation between the zeta-potential and the flow. This is such that the maximum flow appears at, or close to, the point of the relevant zeta-potential curve where it either attains minimum or, in the case of Darvan, the slope changes abruptly.

Set of microsilica containing castables

Up to the first half of the 1980ies, the common way of determining set time of refractory castables was by assessing set with the use of a Vicat-needle set up, or less sophisticated methods like assessment by knocking, squeezing etc. All these methods had a number of difficulties, of which the labor and lack of precision when used on castable systems was the most important. 3 In 1984, Fentiman, George and Montgomery , correlated results from Vicat measurements with temperature development of castables. Instead of using a Vicat needle (penetration of a needle into fresh mortar) the temperature of the castable (or mortar) is recorded. Adiabatic calorimetric studies of aluminous cement in contact with water have shown that there are two exotherms. The first occurs immediately and is associated with heat of wetting and rapid dissolution. A dormant period then follows during which hydrate nuclei form and develop. Once critical nuclei have been formed, bulk precipitation of hydrate occurs giving rise to a second exotherm4. This second exothermic reaction may be used to determine when the setting starts by simply embedding a thermocouple into a proper amount of fresh castable or mortar. If the thermocouple is connected to a strip-chart recorder, the exotherm manifests as a temperature increase. The time to initial set(TTS) may thus be measured (Fig. 9). The time to temperature increase correlates well with data obtained by the Vicat needle. Figure 10 shows a typical exotherm and the correlation between Vicat and the exothermic test. The exothermic test can, as opposed to the Vicat needle, also give reliable results when applied on castables.
Time TTM

Figure 10.Correlation between Vicat initial set and results from exothermic test for calcium aluminate mortar. For castables with both microsilica and CA-cement, the castable normally shows an additional exotherm, which has been associated with the formation of a layer of CASH compounds on the surface of the cement particles that dissolves and enables setting to proceed, giving the final exotherm5. Usually, the appearance of the first exotherm (TTS) is associated with the end of workability of the castable, and the maximum temperature of the second exotherm (TTM) is regarded as the time to final set. The temperature method has been found easy and reliable, but is not suited for castables with very low cement content because of the small amount of heat liberated. The minimum cement content for reliable measurements may be around 3-5% depending on sample size and apparatus. At UNITECR 2005 in Orlando, Myhre presented a study in which some interesting correlations between microsilica and set time were shown. The system investigated was based on the standard castable composition given in Table 1, with the microsilica quality as the variable. All microsilica samples came from the same source, but taken at different occasions during a period of a few months for two successive years. In this context it must be stressed that the microsilica in question is an uncontrolled grade, collected as a by-product from silicon production. To further emphasise this point, the original figure from Orlando has been manipulated to also include results from mixes using the controlled microsilica grade 971 sampled and tested during the same months but in 2007.
6

TTS

START

Figure 9. Typical Strip-chart print out showing temperature in a microsilica containing CA-cement based refractory castable.

12:00 36:00 30:00 6:00 24:00 0:00 18:00 12:00 6:00 0:00 24/1

2004 2005 971-2007

28:48:00
FS20

24:00:00 19:12:00 14:24:00 9:36:00 4:48:00 0:00:00 0.00

Calgon Darvan 811D

TTS [h:min]

3/2

13/2

23/2 4/3 Sampling date

14/3

24/3

TTS (h:min)

0.30

0.60

0.90
2

1.20

1.50

Dosage (mg/m )

Figure 11. Initial set as a function of production date for microsilica from one silicon furnace during winter/spring 2004 and 2005. Values obtained with microsilica grade 971 is added for reference. The set time as defined by the time to the first exotherm of the castable, TTS, shows a very pronounced dependency on the addition level, with Calgon being the prominent example. In Figure 12 and 13 the set time is shown for the bauxite and the white fused alumina system, respectively. In both cases the set time of the Calgon dispersed castables has a very long set time at addition levels around 0.35mg/m2. This coincides in the white fused alumina example unfortunately also with the maximum flow. In the bauxite example, as the maximum flow was obtained at higher Calgon additions this is no longer a problem. FS20 shows a flatter response to the addition level, although by very high addition level there is a tendency towards prolonged set.
24:00 20:24 16:48 TTS (h:min) 13:12 9:36 6:00 2:24 FS20 Calgon

Figure 13: Effect of additive on set time of the white fused alumina based LCC. If the two different systems are compared, at the same addition level, the two systems react similarly, yielding similar set times and flow values. The bauxite however needs more water (5.5 as compared to 4.15). Shelf life Shelf life of castable 1 and 2( Table 2) with 0,05% FS20 and 0,20% Calgon respectively, was measured on dry-mixed samples (25kg, 6minutes in Hobart mixer) placed in closed, plastic lined paper bags. Two conditions were chosen, one was an unheated warehouse at the Elkem site in Kristiansand Norway, the other was in a temperature and humidity controlled room, >50%RH at 20C. The differences between the two conditions were thus that the warehouse fluctuated with the weather and seasons while the climate room had close to constant conditions with relatively high humidity. This controlled condition has earlier been found to accelerate ageing significantly and is used in accelerated ageing tests in Elkem. In Figure 14 the self-flow of the castables are shown as a function of storage time. Although not having a very strong effect on flow, some lowering in self-flow is seen particularly for the Calgon samples after extended storage in the climate room. In Figure 15, the set time as expressed by the time to the first exotherm (TTS) is shown as a function of storage time. As opposed to the flow, the storage has a strong effect on set time. Already after 30 days of storage, the set time increases significantly for both dispersants at both storage conditions. While the

0.00

0.50

1.00

1.50

2.00
2

2.50

3.00

Dosage (mg/m )

Figure 12 The effect of dispersant addition on set time of bauxite based LCC (1 & 2 Table 1)

FS20 then becomes relatively stable, the samples with Calgon continues the trend towards longer and longer set time. The set time of the samples in the climate controlled room seems the most affected, extending the set time up to 4 days in the worst case.

phosphates could be an option. The humidity would assist these processes if not take active part. Such assumptions must be thoroughly investigated in separate tests though.

100 FS20 Calgon

96:00:00 76:48:00 TTS (h:min) 57:36:00 38:24:00 19:12:00 0:00:00 FS20 Calgon

80 Self-flow (%) 60 40 20 0
Start point 30 90

Start point
180

30

90

180

Storage time (days)

Storage time (days)

(a) Stored in warehouse (a) Stored in warehouse

100 80 Self-flow (%) 60 40 FS20 Calgon

96:00:00 76:48:00 TTS (h:min) 57:36:00 38:24:00 19:12:00 0:00:00 FS20 Calgon

20 0
Start point 30 90 180

Start point

30

90

180

Storage time (days)

(b) >50% RH at 20oC Figure 15: Set-time of bauxite LCC after storage in a) warehouse in Kristiansand and b) >50% RH at 20oC, respectivey.

Storage time (days)

(b) >50% RH at 20oC Figure 14: Self-flow of bauxite LCC after storage: a) warehouse in Kristiansand and b) the climate controlled room at >50% RH and 20oC, respectively. According to Fryda et al.7, humidity is a key factor in ageing, and this is also shown here as the high humidity in the climate controlled room accelerates decay as compared to the warehouse condition. Very interesting is the difference in behaviour between Calgon and FS20. To explain this would imply some speculations, but immunisation or sealing of the cement surface by calcium

Conclusion.
Because of interactions between cement and microsilica, additives has to be used in order to ensure good flow of a castable. The flow of the castable is then dependent on the proper amount of additive, and the set is dependent on it as well. The optimum addition level is not always easy to predict as sometimes there are opposite effects, like the example with Calgon that has long set time at optimum flow. It may therefore be concluded that concerning flow and set, one will often face a tradeoff between the desired properties. Also regarding shelf life of castables, it is not

insignificant what choice is made. Choosing one additive over another may in some cases be on the expense of the longevity of the castable.

References:
1 R.K. Iler, The chemistry of silica, John Wiley & Sons, New York 1979. 2 H. Hommer and K. Wutz: Recent Developments in Deflocculants for Castables in proc UNITECR05, Am. Ceram Soc. p.186-190 3 C.H. Fentiman et. al in " Advances in Ceramics vol. 13, New Developments in Monolithic Refractories" ed. by Robert E. Fisher, The Am. Ceram. Soc. Inc. Columbus, Ohio 1985 p. 131135 4 L. S. Wells and E. T. Carlson; J. Res. Natl. Bur. Stand., RP2723, 57, 335 (1956) 5 Steffen Mhmel: in Die Reaktionen von Calciumaluminaten bei Hydratation und thermischer Belastung Beitrge zur Chemie der Hochtonerdezemente dr. ing thesis Tech. Univ. Freiberg Germany 2003. 6 B.Myhre, Microsilica in Refractory CastablesHow does Microsilica Quality influence performance? in proc. UNITECR 2005 Orlando USA Nov. 9, 2005 7 Fryda et al., A New Insight into the Characterisation of Ageing and Mechanisms That Determine the Shelf Life of Low Cement Castables in Proc UNITECR 2009, Salvador Brasil Oct 13-16 2009.