Gypsum based mixes for conservation purposes: evaluation of microstructural and mechanical features ; Mezclas basadas en yeso para su aplicación en conservación: evaluación de sus propiedades microestructurales y mecánicas

This paper presents the study of four gypsum mixtures, focusing on the role of both inorganic and organic additives and on the micro-structural features and mechanical properties. Additives have been chosen among those most reported in historical recipes, for example magnesia, lime putty, rabbit glue. The selected mixes refer to gypsum-based materials used in artworks manufacture, such as plasters, mouldings, stuccoworks, pastiglia. Blank reference materials were prepared on purpose according to the specific recipe, in order to verify the final composition and to highlight the hardening mechanisms and the formation of setting compounds. The chemical composition was related to workability properties and final mechanical resistance and the action of additives as retardants was studied with interesting results. For instance, MgO imparts good properties to the mechanical features, especially with regard to the compressive strength characteristics.


INTRODUCTION
Gypsum is a building material that has been used over the centuries since the Egyptian civilization and still presents a major role in current building technologies (1)(2)(3). Gypsum based materials were utilised in plastering and rendering, bedding mortars, stuccoworks and some outstanding decoration techniques of small artworks, such as pastiglia. This Italian term recalls a low-relief decoration present in several early Renaissance caskets, paintings and frames, with gilt or painted finishes. Gypsum had also a special role in prototyping techniques. The use of this material in plaster coatings was mainly due to its fireproofing, thermal and acoustic qualities, its low cost and ease of preparation (4)(5)(6)(7). Choices in stuccoworks making, for example low and high reliefs, mouldings and in the round statues implied the workability properties of the mixes used. For these reasons gypsum was often mixed with both organic and inorganic natural and artificial materials, such as lime putty and organic additives (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20). The blend with animal glue and pigments was widely used in painting ground layers and in gilded artworks, especially in the south of Europe (21)(22)(23). In the techniques called scagliola and pastiglia plasticity and workability of gypsum mixes played an important role in the final restitution of tiny details and shiny surfaces (24,25). Gypsum has been determined in literature (14)(15)(16)26) as a component in the external renderings of Spanish and Italian buildings (27)(28)(29), nevertheless the expectations given by its low solubility in water. Its use was also important as a support and/or binding medium for mural paintings, as in many Egyptian cycles (1)(2)(3)30).
The technology of gypseous materials allowed several working steps since very early stages, to modern industrial processes; in fact, gypsum is formed in nature in sedimentary evaporative depositional environment. To use gypsum as a binder, thermal dehydration is required. This process starts at a temperature of 42°C, but decomposition is slow. So a range of temperature between 100°-140°C is normally used in industrial processes in a wet environment. The product of the dehydration is bassanite (CaSO 4 ½H 2 O), which rehydrates quickly when water is added. If gypsum is heated up to 200°C a soluble phase of anhydrite (γ-CaSO 4 ) is obtained, which rehydrates in presence of water too. Heating bassanite, over 360°C leads to the formation of the insoluble anhydrite phase β-CaSO 4 , while α-CaSO 4 is produced at 1180°C. Above 1375°C, the complete dissociation of anhydrous calcium sulphate into calcium oxide, sulphur dioxide and oxygen occurs. According to the literature there are still doubts about the transition conditions between one phase and the next (31)(32)(33)(34).
Here we link the composition of gypsum based mixes with their final performances, discussing how this type of knowledge could be useful in conservation practice, for example when the conservator has to integrate a missing part with a new paste (35). Scientific literature in the specific field is scarce, especially with regard to gypsum based traditional recipes. Mixtures tested in this paper have been prepared according to the historic recipes given by conservators' suggestions, together with the available scientific bibliography (10,12,13,21,33,(36)(37)(38)(39)(40)(41)

Materials
Materials used for the mixes and the laboratory model samples are summarized below:  (11,16,33,41,42) The amount of water was considered crucial to achieve the final performances, so its weight in each mixture was measured.
MgO was selected because it is reported by crafts handbooks as an additive that improves the mechanical performances of gypsum based pastes (33), although this effect was not yet been verified by scientific studies. It is known from the literature that magnesium could be present in lime mortars, when magnesian lime is fired in the kiln (43,44) starting from rock raw materials with dolomite impurities (4). Lime putty was selected because, when mixed with gypsum, it forms a well-known and diffused blend utilised in stuccoes practice (11,13,33,38,41). Finally, the choice of the rabbit glue is due to the well-documented use in scagliola and pastiglia techniques, for example in the creation of altar frontals (11,15,41). It is worthwhile to note that magnesium oxide and calcium carbonate have been detected in recent characterisation studies (9,11,13,28,45), as well as animal glue or other protein based materials (11,41).
The study of the final properties of gypsum based specimen has been particularly neglected in recent literature, which in fact reports only on the different type of aggregate (11,42), the addition of fibrous materials (46), or on the blend with organic compounds (47).
The raw materials were characterized by X-ray diffraction (XRD), infrared (FTIR) and Raman spectroscopy to verify the presence of undesirable components. Bassanite is the main component of "Gesso Alabastrino", with traces of anhydrite and dolomite, probably coming from the raw materials. The purity of MgO was confirmed, while the lime putty is designated as CL 90-Q according to EN 459-1 (48). Rabbit glue showed the typical FTIR absorption peaks of collagen (49).
Mixtures were prepared at controlled environmental parameters (65% RH and 20 °C) and the same mixing conditions were ensured for each specimen, using the same mixer and same velocity. Steady curing conditions were maintained for 28 days (50% RH and 20°C) (50). The composition of the 4 mixtures are reported in Table 1. With regard to mixtures A, B and C, they are considered to be among the most widespread ones in the art technique manuals. Instead, the mixture D was chosen in order to study the effects due to the addition of MgO. A systematic survey of historical records leads to the ratios reported in Table 1. A set of preliminary tests suggested the appropriate amount of water, considering the one already present in lime putty, or in the rabbit glue. Anyway, the proper consistency was evaluated by means of a flow table.
A series of specimens were prepared, sized 160×40×40mm as required by current standards (51), with the aim to study physical and mechanical properties, together with the microstructure. The setting time and the hardening were evaluated with Vicat needle (51), the workability was estimated through the flow table method, the flexural and compressive resistance, and the linear shrinkage were measured according to current standard (51).
Flexural strength measurements were performed after 28 days, averaging the final value on a basis of three replicates. Compressive strength tests were performed after 28 days, averaging on the 6 replicates produced by the fracture of specimens used in flexural test.

Analytical methods
The raw materials and the mixtures were characterized by XRD, FTIR, and Raman Spectroscopy. The mineralogical composition after the specimen setting was analysed too, in order to understand the effects induced by the added compounds. For each specimen three replicates were carried out. X ray diffraction was carried out on powdered samples by PANalytical X'Pert PRO X-ray diffractometer, with geometry goniometer θ-θ. The diffractograms were recorded between 3° and 75° 2θ with a scanning speed of 0.21 θ/sec, using a Cu Kα radiation, a PW generator 3040/60 in the conditions of 40kV and 40mA, and a solid-state multi-detector X'Celerator PW3015/20, with Ni filter. The powdered samples were deposited on a ground glass support. Results were interpreted by the use of the X'Pert HighScore software.
A Micro-Raman spectrometer Horiba Jobin-Yvon LabRam was used, coupled to a confocal microscope Olympus BH-4 equipped with motorized XY stage. The instrument works with two laser lines: 632.8 nm (He-Ne, 20 mW) and 473.1 nm (Nd: YAG laser, 100mW). The Raman signal is dispersed by a holographic grating with 1800 lines/mm on a Peltier cooled CCD detector (256 × 1024 pixels). The Rayleigh radiation is blocked by notch filters BraggGrateTM (OptiGrate), which allow collecting signals from 10 cm −1 from laser line. The laser power reaching the sample can be limited through the use of neutral density filters. The entrance slit was set at 150 microns. Spectral resolution is ~1.5 cm −1 using the 632.8 nm. Spectra were acquired in the range from 100 to 4000 cm −1 using the line at 473.1 nm with the objective 50x ULWD. Laser power of the order of tens of mW was used, averaging 3-4 analysis with acquisition times of 20-160 seconds each, using the filters D1 and D0.6.
Fourier Transform Infrared spectra (FTIR) were recorded with a Thermo Scientific Nicolet iS10 instrument in the 4000-600 cm −1 range, equipped with an ATR accessory with diamond crystal, with a resolution of 4 cm −1 , collecting 32 scans. Previously to each sample a background analysis was recorded, and automatically used to correct the measurement.
In depth morphological and microstructural observations were carried out using a stereomicroscope Leitz Wild M420 at different magnifications in order to observe the surface morphology in detail and a Scanning electron microscope JEOL 5910 LV, source tungsten filament, coupled with X-ray spectrometer (EDS) in dispersion of IXRF-2000 energy. Analyses were conducted in low vacuum. The compositional nature and distribution of elements on sample was investigated by the EDS spectra and maps from 0 to 20 keV. The measurements were carried out on calibrated images to study the dimensions of gypsum crystals in the samples by image analysis. The software ImageJ was used for this purpose.

Spectroscopic and X-Ray diffraction analysis
The characterization of the raw materials and of the mixtures was carried out combining mineralogical (X ray diffraction) and spectroscopic analyses (Raman Spectroscopy and FTIR). Each analysis was performed on three replicates. The mixtures B, C and D were analysed after the curing time (1month) too in order to verify the effect of the addition of MgO, rabbit glue and lime putty on hardening: gypsum, anhydrite, and brucite were identified in mixture D, while gypsum, calcite, anhydrite and portlandite were found in the mixture B. The presence of brucite Mg(OH) 2 is due to the hydration of MgO during the preparation of the mixture D. A high content of bassanite was put in evidence, in mixture C, as well as the presence of gypsum and anhydrite. According to Elert et al. (54), the presence of bassanite after hardening is due to the action of animal glue as a barrier that inhibits the formation of gypsum.
The presence of calcite and portlandite in mixture B is linked to the process of lime putty hardening. The determination of brucite (mix D) and portlandite (mix B) suggests the incomplete carbonation of lime putty in the samples. This phenomenon could also be present in building materials, for the very low kinetics of carbonation reaction depending on relative humidity, size and geometry of the artefact. In fact, in the case of bulky architectural elements, it is hard to ensure the effective CO 2 diffusion from the atmosphere (55)(56)(57)(58). For these reasons, specimens will be allowed to set completely and the same survey will be the object of a next paper. The specimen B and D have been however studied, considering their interest in building material topic and in view of a comparison with completely carbonated specimens.

Thermogravimetric analysis and Differential Scanning Calorimetry (TG-DSC)
Thermogravimetry analyses (Table 2) confirmed the above-mentioned results highlighting the presence of brucite in the mixture D. The amount of brucite was lower than expected from the amount of MgO in the mixture; probably some of the unreacted MgO worked as an aggregate. Moreover, a small amount of calcite is present in all specimens, due to impurities in the raw materials used to obtain gypsum. A greater amount of calcite was identified in the mixture B, which also contains a certain amount of portlandite due to carbonation not yet complete (59).

Scanning Electron Microscopy observations (SEM)
SEM observations (Figure 1) show the variability of the morphology and size of gypsum crystals, due to the presence and variety of additives. In Figure 1a (Mixture A) the image of typical acicular gypsum crystals is shown, while in Figure 1b (Mixture B) both acicular and rounded crystals are visible, probably due to the presence of both gypsum and calcite, that were also found by XRD analysis. The acicular gypseous crystals were not detected in mixture C (Figure 1c), and only tabular crystals are observable. Finally, the image of the mixture D (Figure 1d) clearly shows the lowering of the dimension of acicular crystals and the appearance of an amorphous matrix embedding the crystals. The elemental distribution on the samples surface is given in Figures 2-4.
Mixture A was not reported because its composition is obvious. On the contrary, in the rest of the SEM-EDS maps some features are evident and they deserve comments. For instance small lumps were observed, possibly due to the non-uniform blending or to the presence of magnesium impurities, even in the case where magnesia was not intentionally introduced (MixC - Figure 3). In particular, in Figure 2 (mixture B) it is possible to observe an accumulation of calcium in correspondence of the flat surface of the specimen.
Thank to SEM images the crystal size distribution was evaluated. In mixtures A, B and D ( Figure 5) crystals length is much than the width (Table 3). On the other hand, mixture C presents a smaller difference between length and width of the crystals.

Vicat needle
The initial setting time and final setting time were measured by Vicat needle using standard UNI EN 13279-2 (51). As described in Figure 6 and in  Table 4, the comparison between pure gypsum (A) and the mixes B (lime putty) and D (MgO) was interesting. The curves of Figure 6 show the penetration measurements performed by Vicat needle on the mixtures. The time value where the curve starts represents the initial setting time, while the final setting time is reached at the highest value of the curve (PEN. [mm]). Lime putty and MgO acted as a retardant in gypsum initial setting time. It was also verified that the mixture containing rabbit glue did not provide useful values in Vicat measures because it had not hardened in the measuring range (15h). Anyway, the measure confirmed the role of rabbit glue as strong retardant. This topic will deserve in depth study in the future, with an on purpose procedure overlooking the standard used. It is important to remark that currently no reference values are available for these mixtures, except for pure gypsum used as binder in plastering. In order to comply with the standard EN 13279-1,2, the initial setting time of more than 20 minutes was established (51,60).
The mixture A showed ( Figure 6) the initial setting time at 9.57 min and the final setting time at 14.27 min, corresponding to a penetration value of 37.65 mm. The standard value for the ending point of the final setting time resulted as 40mm, corresponding to the thickness of the sample holder (51). The initial setting times of mixture B (gypsum and lime putty) and mixture D (gypsum and MgO) are around 17.06 minutes and 13 minute, respectively, while the final setting time is around 24 minutes for both of them. The mixture B seems to set quickly, but after the final setting is completed in 10 minutes. This could be considered as a strange performance, surely connected to the presence of lime putty and to its slow hardening due to the carbonation reaction.

Flow table
The results of workability tests, carried out according to the procedure described in UNI EN 13279-2 2004 (51), showed higher values than those indicated for pure gypsum standards ((165 ± 5) mm) (51). Moreover, gypsum added with lime putty or rabbit glue seemed to have less workability (Table 5)  These mixtures could be used in material integration, re-pointing and other restoration practices. Hence, the workability test proved to provide fruitful results, although they should be  performed specifically on a case-by-case basis, in order to better contextualize the needs of each circumstance.

Mercury Intrusion Porosimetry (MIP)
The porosity values and the pore size distribution are reported in Table 6. The total porosity ranges between 32.7% (mixture C) and 46.9% (mixture B). Mixtures A, C and D show similar values with regard to the average pore radius and in mixture B, the most representative pore class shows a diameter around 1.7 μm. In Figures 7 and 8 the total open porosity and total cumulative volume are plotted versus the pore radius, clearly showing that mixture B has a different microstructure than the other mixes. This behaviour is also shown in Figure 9, where the pore size distribution is plotted versus the relative volume and the pore radius (μm). The results of mixture C are quite interesting, as it appears to be the only one to have a bimodal pore size distribution ( Figure 9). As expected, the organic additive filled the inorganic matrix reducing the total open porosity to the lowest value of 32.7%. Moreover, the bulk density resulted to be the highest of the ones measured. The literature corroborates the findings, even if it refers to lime mortars with organic additives (61, 62).

Flexural and compressive strength
Mechanical tests have been carried out on the hardened mortars after 28 days of curing time, as suggested by standards (51). The measured values are reported in Tables 7 and 8.
The results of flexural strength measurements are presented in Figure 10. The mixture made with gypsum and lime putty (B) presents the lowest value, due to the incomplete carbonation at the time of the measurement. The mixture C displays values slightly lower than mixture A, while the mixture with MgO (D) is characterised by greater variability, with values both higher and lower than the mixture made of pure gypsum. Figure 11 presents the results of compressive strength (mean values of 6 replicates). It is evident that the addition of MgO highly increases the values of compressive strength, whereas mixture    B shows the lowest result, due to the incomplete carbonation at this stage (63, 64).

Linear shrinkage
The measures were carried out after 3, 7, 14, 28 days (Table 9). Three specimens for each mixture were analysed as suggested by UNI EN 13454-2 (65). Each mixture expand, generally after an initial shrinking of about 5-10 days, then the dimensions do not vary significantly for the rest of the test. Mixture C shows the maximum difference in size from the starting point to the end of the measurement (-2.41mm). Mixture C is the only one that expands, but remaining below the initial dimensions, with a total effect of shrinkage.

Contact angle
The contact angle is given by the encounter of the tangent of liquid-vapour interface with a solid liquid interface. By convention, a surface material having a contact angle with water greater than 90° is defined as hydrophobic, while a surface with angles minor than 90° is hydrophilic. The measurement were performed according to UNI EN 15802:2010 (53). The only mixture on which it was possible to measure the contact angle was mixture C (gypsum and rabbit glue): it resulted as 41° (std. dev. 11). This value does not correspond to a hydrophobic surface, even if it allowed the comparison with absorbing properties of the other mixtures. Mixtures A, B and D in fact did not allow any contact angle measurement, due to the very fast kinetic of water drop absorption. We can conclude that rabbit glue induces the variation of the properties of the surface addressing to lower absorption and lower hydrophilicity.

DISCUSSION
The results have clarified the characteristics of the gypsum mixtures prepared in this work. The compositional analyses (after 28 days) allowed identifying the chemical and mineralogical composition     (64,67). As for the microstructure, the mixture gypsum/ lime putty showed the highest average values of total porosity and of pore radius. These results could be related to the low mechanical properties measured,  even if further analyses are necessary. The presence of portlandite in this mixture is due to the incomplete carbonation of the raw materials, which probably led to an overall weakness, as resulted from mechanical tests. Regarding the pore size distribution, mixture B presents pore size distribution % and porosity values similar to mixture A (gypsum), even if the average radius is different. However, the results confirmed the retardant role of lime in setting and hardening mechanisms.
Moreover, the average pore radius in mixtures C and D is closer to the one measured for gypsum in mixture A, although mixture C seems to have a bimodal pore radius distribution ( Figure 9). These data should be taken into account when studying the transport phenomena, which are considered crucial when applying fluid products, such as cleaning formulates or consolidants, or in planning filling voids or missing parts with compatible integration materials.
SEM allowed an in depth microstructural survey, which put in evidence the changes induced by the additives, for example those regarding the dimensions and the shape of the gypsum crystals in the mixtures.
The setting and hardening measures clarified that all the additives here considered acted as retardants. An exception to this is the rabbit glue mix, which did not harden at the time of the measurements. Coming to mechanical tests, the mixture including MgO gave an important result: it showed a resistance to flexural strength similar to the gypsum mixture and compression strength higher than each of the other mixtures. As for this mixture, the compressive strength values are higher than the values measured by Sing and Garg in gypsum mortars retardants     (54). The mechanical features of mixes containing MgO suggest the use of these materials when a general strengthening is required. The mixture of MgO and gypsum showed excellent results in the shrinking test too, obtaining a dimensional variation similar to that of pure gypsum. This should be considered a very important result, if we bear in mind the possible application of this material in integration and repointing operations during the restoration practice, because of possible detachment problems. A variation of linear shrinkage slightly greater than that recorded for gypsum was presented by mixture B, while the worst results were obtained in mixture C.
Actually, mixture C is not suitable for injection operations because of its high shrinkage: if used on fragile ancient materials, it could in fact damage or cause cracks in the plaster during the setting and the hardening, especially where the integration is called at filling bulky volumes. Currently it is not possible to compare our results to (42), because no aggregate was added to our mixtures in order to comprehend first of all the features of the binder fraction alone. In any case, we can observe that our results of compressive strength are higher respect to the values reported in (33,42), even if our samples are more porous. The workability tests provided fruitful results: in fact, as to conservation purposes, the different mixes could be proposed for different final applications and the workability values distributed in a wide range could allow the operator the most suitable choice.
A significant difference in water transport phenomena can be assumed between mixture C and the rest of the inorganic mixes here studied.

CONCLUSIONS
When the conservator is using gypsum mixes in integration, embedding or re-plastering operations, very often the ready made products are chosen among those offered by the market. Unfortunately, their composition is not known in most cases. For example, more than one synthetic polymeric addition could be present. Nevertheless, the habit of preparing on site "on purpose" mixes according to the traditional recipes was lost over the XX century. Hence, empiric knowledge of the properties of these mixes is not very often part of the skills of many conservators. Moreover, even if the general empirical properties are esteemed, normally they have not been measured. In this paper a first approach is attempted and proposed.
This work investigates the effect of three very common additives added in gypsum-based mixtures that acted as retardants. The results obtained from the experimental studies can be summarized as follows: • The addition of MgO, lime putty and rabbit glue acted as retardant; any of these additives gives time to the conservator of plastering and modelling in relief integration works, proving rabbit glue to be the most retardant compound; • The mixture made with MgO could provide promising results in restoration practice when a general strengthening is required; anyway, some in-depth studies are needed in order to highlight the possible transformation of Mg compounds in the mix, taking into account that some Mg sulphates may form very dangerous efflorescences. • The rabbit glue changes the microstructure and the pore network. Most probably, the water transport phenomena could be reduced and a small increase in water repellency could be expected as well, even if some condensation problems on the surface should be considered. In fact, mixture C showed a contact angle of 41° and the material should be considered hydrophilic, but the results interestingly proved the increase of water repellency caused by the addition of rabbit glue. However, more in depth studies are in progress to better understand this aspect. • A study of the properties of the mixtures after a longer period of curing time could give more information on the characteristic of these materials and their possible applications.