A previous XRD study carried out on the powdered gypsum confirmed the presence of bassanite (CaSO₄·½H₂O) as the main mineral phase with a much lower presence of anhydrite or anhydrous calcium sulfate (CaSO₄), calcite (CaCO₃) and dolomite (CaMg(CO₃)₂) (Figure 3). The very low level of gypsum dihydrate detected among its minor components would have been due to the natural hydration of the binder with ambient water vapor.

Gypsum powder was also studied by TG. A first mass loss of 4.88% was obtained corresponding to the transformation of calcium sulfate hemihydrate (CaSO₄·½H₂O) into anhydrite (CaSO₄) at around 130 ºC. Subsequently, at a temperature of approximately 670 ºC, a second mass loss of 2.49% was observed, again due to the decarbonation of calcite (CaCO₃) accompanied by a loss of CO₂ (CaO + CO₂) (Figure 4).

Finally, a study of the material was carried out using XRF, from which the percentage concentration of the elements that make up the gypsum expressed in the form of oxides was obtained (Table 1). As expected, the two major oxides were those associated with the elements Ca and S. The H₂O and CO₂ data were not determined by XRF but obtained from TG since this technique has greater sensitivity and reliability for detecting elements with low atomic number due to their low X-ray absorption.

Once set and cured, the heat transfer of the tested plasters increases with their density (27), as was confirmed by apparent density measurements. The 0.4 w/g and 0.7 w/g mixtures provided apparent densities of 1.472 g/cm³ and 1.034 g/cm³, respectively. Gypsum 0.4 w/g is denser than gypsum 0.7 w/g. The temperatures recorded during the test for the 0.4 w/g ratio were higher than those recorded for the 0.7 w/g ratio, except in the final period from minute 28 onwards, when slightly higher temperatures were recorded for the 0.7 w/g ratio, although the mean temperature was still lower (Figure 5). The greater porosity of the gypsum at 0.7 w/g probably produced a greater long-term alteration (> 28 min) as a result of exposure to direct fire, thus increasing the diffusion of heat in the final portion of the test from the exposed face to the non-exposed.

If the results obtained for the ratios 0.4 w/g and 0.4 w/g + 0.6 g/kg C. A. are compared, it can be seen that upto about minute 7 both the maximum and mean temperatures increased at a similar rate. From that moment onwards, the slope flattens out substantially in boths, corresponding to the evaporation of part of the crystallization water of the gypsum. This plateau lasted about 2 minutes less in the samples containing citric acid, indicating that the generation of CaSO₄·2H₂O by hydration of CaSO₄·½H₂O was hampered in the presence of citric acid, as demonstrated in recent studies (23). At about minute 14 of the test, the maximum temperatures recorded in the specimens containing citric acid were higher than those recorded in the gypsum specimens without additives.

However, the average temperatures obtained throughout the test were very similar. Three tests per dosage were performed, confirming the trends described and the consistency of the measurements (Figure 5).

Figure 6 shows a selection of the images obtained with the thermographic camera after 5, 10, 15, 20, 25 and 30 minutes of exposure to fire. The composition of each specimen is shown in the upper right-hand corner of the image, while the bottom right-hand corner shows the maximum temperature as well as the average temperature recorded in the specimen. These images represent a temperature map of the non-fire exposed face of several 2 cm thick gypsum specimens (both with and without additives) that could be used that might be found as materials used for passive protection of other construction and structural elements.

The lowest degree of heat transmission almost throughout the test was that of the 0.7 w/g gypsum board, although during the final few minutes, the temperatures were slightly higher than for the 0.4 w/g ratio probably because its greater porosity led to a greater alteration due to exposure to fire. Little difference was observed between the gypsum with and without added citric acid for the same w/g ratio. For the 0.4 w/g ratio, the temperature peak was lower than for the 0.4 w/g ratio + 0.6 g/kg C. A., but the heat was distributed over a wider area of the specimen, which meant that the average surface temperatures of the specimens were practically similar.

The slightly different behavior of gypsum with added citric acid from the gypsum without additive could be due to the effect of citric acid on the microstructure of gypsum, which would affect the transmission of heat through the plate.

Once the 30-minute fire test had finished, photographs of the face that had been exposed to fire (EF) were taken. Figure 7 shows the cracking map of the 0.4 w/g dosed specimen and 0.4 w/g + 0.6 g/kg of citric acid dosed specimen. As can be seen, the cracking pattern differs considerably, and agrees with the heat distribution observed by thermal camera on the surface of the specimen. The specimen containing citric acid had fewer but wider cracks than those that appeared in the specimen without citric acid. This might also explain why the temperatures were higher in the central zone in the presence of citric acid (due to the greater width of the fissure).

Figure 8 shows the effect of citric acid on the microstructure of the gypsum samples studied.

As can be seen, citric acid has a strong effect on the gypsum microstructure. The specimens that do not contain citric acid had acicular crystals with twins and crystals behaved differently had a or had crystalline morphologies (Figure 8a). However, in the gypsum with added citric acid, more prismatic shapes and a degree of amorphization of the crystalline framework could be observed (Figure 8b). These results correspond to those already carried out on this subject (22, 23, 29).

After the test, the specimens exposed to fire from both the non-exposed face (NEF) and the exposed face (EF) were observed by scanning electron microscope. As can be seen in Figure 8c and Figure 8d the gypsum crystals of the EF were still perfectly recognizable which agrees with the temperature records obtained by thermal imaging camera since the average temperature did not exceed 110 ºC. However, the microstructure of the samples directly exposed to fire (EF) differed from the above (Figure 8e and Figure 8f). In the first place, different crystalline behaviour corresponding to bassanite and anhydrite were observed, since the temperatures recorded with the thermocouple of the EF being around 800 ºC and therefore much higher than the complete dehydration of the plaster. In addition, small cracks were observed as a result of the thermal stress of the coatings exposed to direct fire, especially in the samples with added citric acid.

From the analysis of the results obtained, it is clear that, once cured and before being tested by fire, the gypsum is rich in calcium sulfate dihydrate, with small amounts of bassanite and anhydrite, as expected in this type of material (Figure 9a and Figure 9b).

Additional XRD tests were carried out on both faces of the specimens tested by fire, to verify the effect of temperature on the mineral phases described above. On the non-exposed face (NEF), the greater presence of the most hydrated mineral in the series (gypsum) was clearly observed for the sample without citric acid (Figure 10a) compared with the same sample containing citric acid (Figure 10b). For example, the major peak at low 2-theta angles was substantially reduced by incorporation of citric acid. This may confirm the effect of citric acid on morphological changes (crystal habits) of gypsum (23, 30) leading to more heat-conductive plasters. Finally, it is worth noting that XRD is a semi-quantitative technique and, although the peaks intensities may reveal some differences, this should be confirmed by quantitative methods like TG. As expected from the temperature records on the unexposed face, there were no differences in thermally stable minerals such as quartz or in other minerals that are much more stable than gypsum, such as calcite and dolomite.

However, the results obtained on the exposed face (EF) indicate that after that time of exposure to fire practically all of the gypsum had become anhydrite (Figure 10c and Figure 10d). Neither of the figures show much sign of plaster, which suggests that practically all of the material has been dehydrated as indicated by the SEM images described above. Even thermally stable minerals such as calcite and dolomite, which decompose above 750-800 ºC, were almost completely descomposed. Therefore, the XRD studies corroborated that a direct fire scenario produces a thermal stress that considerably alters the microstructure and composition of the coating.

A thermogravimetric study allows the degree of hydration of gypsum to be based on its resistance and degree of exposure to fire (31). The results of the behavior of specimens not subjected to the fire exposure tests, fire-tested specimens from the non-exposed face (NEF) and fire-tested specimens from the exposed face (EF) are presented below.

In specimens not subjected to fire (Figure 11), the results obtained indicate that in the initial state, the first mass loss occurred at around 150 ºC, corresponding to the transformation of gypsum dihydrate (CaSO₄·2H₂O) to bassanite (CaSO₄·½H₂O) and finally anhydrite (CaSO₄). Both processes occurred at temperatures that were very close and were therefore seen as a single water loss process in the thermogram. The first step of the TG curve is due to water crystallization release from CaSO₄·2H₂O. As can be seen, the weight loss for 0.4 w/g ratio + 0.6 g/kg C.A is lower than that of 0.4 w/g. Taking into account the high sensitivity of TG technique the observed difference, although minor, may indicate lower formation of CaSO₄·2H₂O in the former sample. Finally, the TG results are consistent with the maximum temperature profile found in Figure 5. After 30 min, the maximum temperature recorded on the non-exposed face of samples containing citric acid was higher than that recorded for the remaining ones.

The second and third mass loss in all the thermograms corresponded to the thermal decomposition of carbonates (calcite) into calcium oxide and carbon dioxide at around 650-700 ºC and the decomposition of anhydrite sulfate (CaSO₄) in calcium oxide and carbon sulfur dioxide at 1300-1350 ºC.

In the thermograms made on specimens from the face not exposed to fire (NEF), after 30 minutes of testing, the three above described steps are also evident (Figure 11). The first loss of mass at about 150 ºC corresponding to the loss of water produced by the sequential change from gypsum to bassanite and anhydrite, was considerably lower in the dosage containing citric acid, possibly due to the fact that the additive inhibits the hydration of the bassanite forming less product with as much crystallization water as possible (gypsum).

This result corroborates the analyses obtained for the NEF in the XRD, where more intense diffraction signals of the gypsum mineral were observed for the 0.4 w/g dosage with respect to the same sample with citric acid. There was also a small difference in mass loss between the 0.4 and 0.7 w/g coating which was slightly greater in the latter.

Again, there a second mass loss occurred between 650 and 700 ºC due to the thermal decomposition of carbonates (calcite) and a third loss between 1300 and 1350 ºC due to the decomposition of anhydrous calcium sulfate (anhydrite).

These results are very different in EF (Figure 11). After 30 minutes of testing, all the gypsum originally present in the specimen had been decomposed as the loss of crystallization water from gypsum (CaSO₄·2H₂O) and bassanite (CaSO₄·½H₂O) disappears. Therefore, the XRD and TG results confirm a practically quantitative transformation of the totally or partially hydrated phases of calcium sulfate into anhydrite (CaSO₄).

Small differences were found in the CO₂ mass loss associated with carbonates. The specimen least affected by exposure to direct fire was 0.4 w/g since the loss of CO₂ was 1.46% which was higher than that detected in the 0.7 w/g and 0.4 w/g samples with added citric acid (0.64% and 0.34%, respectively). It is important to emphasize that on the face exposed to the fire, the sensors recorded temperatures at around 800 ºC. Therefore, the duration of the experiment was prolonged by exposing the coating for 30 minutes to temperatures high enough to partially decarbonate the carbonates and even eliminate them in their entirety. Differences in the compactness and microstructure of the coatings may explain the differences observed in the thermograms of Figure 11. However, the 0.4 w/g originally contained the highest amount of CaSO₄·2H₂O, which is the mineral that offered the greatest resistance to increased temperatures. This fact may be the key to delaying the destruction of carbonates and explain the higher concentration of calcite in the 0.4 w/g specimen.