3.1. Characterization of Bricks and Cement Raw Meal
The properties of commercial magnesia–spinel bricks A and B employed in the coating tests are listed in
Table 2.
Brick A presents higher density and, consequently, lower porosity and permeability, what is important to avoid alkali and clinker infiltrations, but may be a disadvantage in terms of physical adherence of coating on brick surface. The difference in density and porosity between the bricks is a result of the distinct process parameters used during the industrial production. Brick B shows a similar alumina level when compared to brick A, suggesting an equivalent spinel amount, besides zirconia addition that reacted partially with calcium oxide forming calcium zirconate (CaZrO3), which is a high refractoriness phase (Tm ~ 2340 °C). In terms of mechanical strength, Brick A exhibits higher resistance but at high temperatures, the bricks present similar values.
Table 3 presents the characterization of the cement raw meal as received (100% cement raw meal) and modified (with 30% potassium sulphate and 5% coal). The pyrometric cone equivalent test (P.C.E.) evaluates the softening temperature of the meals and, as expected, the addition of K
2SO
4 reduced this temperature from 1520 °C to 1395 °C. This reduction probably will influence the coating tests, as the modified cement meal will present more liquid phase than the meal as received. The particle size showed a similar distribution for both meals but the mean size (d50 ~ 18 µm) was coarser than the mean size of ~5 µm tested by Rigaud et al. in their work about coating adherence on magnesia–spinel bricks with good adherence ability [
6].
Regarding chemical analysis, the cement raw meal as received presented a typical content of the main oxides (CaO, SiO
2, Al
2O
3 and Fe
2O
3), on the other hand, the addition of K
2SO
4 in the modified meal changed completely the level of these oxides. The silica ratio calculation (SR = SiO
2/(Al
2O
3 + Fe
2O
3)) indicated higher values for the meals than the ideal value of 2 studied by Rigaud et al. [
6], which will reduce the amount of clinker liquid phase and consequently the adherence strength.
In the XRD, with the patterns illustrated in
Figure 3, the calcite was identified as the main mineral, and then quartz, muscovite and chlorite were detected in lower concentrations. For the modified meal, the arcanite (K
2SO
4) was present as well.
3.2. Qualitative Coating Test
The evaluation of adherence ability of the coating for the bricks A and B by the contact method described by Kosuka et al. [
8] is shown in
Table 4 for different raw meals. The test was performed in duplicate and the refractory samples were taken from the same brick. Brick B showed better coating adherence than brick A when the cement raw meal as received was used in the coating test. Indeed, the higher porosity and permeability of brick B contribute with a physical coating adherence. Moreover, the reaction of CaO from cement clinker with no reacted ZrO
2 from brick B, forms CaZrO
3, which increases the connection between the clinker–brick interface [
10] and contributes with a chemical coating adherence. However, when the modified cement raw meal was used, as proposed by Kosuka et al. [
8], both bricks presented similar behavior with no coating adherence after first cycle, and the coating agent was easily removed after second and third cycles. The increasing of coating adherence with cycles was related with the excessive use of additions to cement raw meal.
The visual aspect of the bricks A and B after 3 cycles of qualitative coating test using cement raw meal as received is illustrated in
Figure 4. The coating agent did not adhere on brick A surface, and an infiltration clinker layer is observed in the hot face of this brick.
For brick B, the coating agent remained adhered after attempts to remove with hand and no reacted area is visually observed in the hot face, which demonstrates the protective action of the coating.
When the modified cement raw meal was used, the bricks A and B presented no difference in coating adherence, as shown in
Figure 5. Both bricks demonstrated a deep infiltration of clinker layer, which is related to the high percentage of liquid phase without forming a connection between clinker and brick. Although more liquid phase in the clinker accelerates the reaction between refractory and clinker, what is essential to construct a protective layer, the excessive amount of liquid infiltrates deeper into the brick and impairs the coating formation.
The XRD patterns for unused bricks A and B and for the hot face of the bricks after the contact coating test using both cement raw meals are illustrated in
Figure 6 and
Figure 7. When the cement raw meal as received was applied as coating agent, brick A presented C
4AF and Q phase (C
20A
13M
3Si
3 or Ca
20Al
26Mg
3Si
3O
68 [
4]), besides MgO and MgO.Al
2O
3, which are the original phases for this brick.
The tetracalcium alumina ferrate (C
4AF) is one of the main mineralogical phases of the clinker (C
3S, C
2S, C
3A and C
4AF) thus it indicates that a clinker infiltration has occurred. On the other hand, the Q phase is a result of the reaction between silicates phases from the clinker (C
3S and C
2S) with spinel from the brick. In addition to spinel corrosion, the Q phase also contributes to decrease the refractoriness of the brick due to its low melting point (1300–1400 °C) [
4]. The Q phase was also identified on the hot face of brick B, although no reacted area was visually observed on the brick surface after the coating test.
When the modified cement raw meal was used, both bricks showed similar amounts of K2SO4, Q phase and mayenite (C12A7, an intermediate phase in the formation of Q phase), which makes a comparison between the bricks difficult.
Therefore, the cement raw meal as received (100% cement raw meal) is the most suitable meal to be used in the contact coating test as the mixture with potassium sulphate and coal is more aggressive in terms of liquid phase generated.
3.3. Quantitative Coating Test
The results of the sandwich test proposed by Rigaud et al. [
5,
6,
7], but modified in the present work, are shown in
Table 5. For this test, the modified cement raw meal was not used due to the unsatisfactory results obtained in the qualitative coating test. The test was performed in triplicate and the refractory samples were taken from the same brick.
In accordance with previous evaluation of contact test using cement raw meal as received, brick A did not present coating adherence as well, thus the value for CMOR was considered as zero. In contrast, brick B showed a mean value of 1.2 MPa for CMOR conducted after the sandwich test, confirming its superior coating adherence regarding brick A. This value agrees with the results obtained by Rigaud et al. [
6] for magnesia–spinel bricks.
Figure 8 illustrates the difference in the coating ability of the bricks A and B after the sandwich test but before performing CMOR.
After the sandwich coating test, the clinker–brick interface of the bricks A and B was observed using the optical microscope. According to microstructural change, three zones can be distinguished: coating, changed brick and unaltered brick.
Figure 9 displays the changed area for both bricks.
It is noticed the presence of coating adhered on brick B, in addition to a greater microstructure preservation when compared to brick A, which presented more corroded spinel grains due to no coating being formed.
The microstructure analysis points out that the liquid phase of the clinker deeply penetrated into the bricks matrix, reacting with spinel grains and generating phases with low refractoriness, as mayenite (C
12A
7) and Q phase (C
20A
13M
3Si
3 or Ca
20Al
26Mg
3Si
3O
68) [
4], which contributes with formation of more liquid phase in the system and creates the only possible connection between brick and cement clinker [
6]. Although this liquid phase is essential for adhering coating on magnesia–spinel bricks, the formation of a large quantity of liquid is prejudicial for coating retention. Therefore, the preservation of the original structure is important to achieve adherence strength and to build sustainable coating on brick surface, which explains the better coatability of brick B. In fact, the coating action will protect brick B from further infiltrations.
Finally, the comparison of qualitative and quantitative coating tests has showed that both methodologies, as used in the present work, are able to differentiate coatability of magnesia–spinel refractory bricks with similar properties to the bricks A and B. Furthermore, a similarity between the test results can be established: the “bad” classification in contact test is related to CMOR close to 0 MPa in sandwich test, whereas the “very good” is related to CMOR > 1 MPa in sandwich test. So, it is expected that a refractory brick that presents “good” classification in contact test would present CMOR between 0 and 1 MPa in sandwich test. Despite that relation, the quantitative sandwich coating test is preferable because it presents numerical results, in addition to require only one firing cycle.