12th International Congress on the Deterioration and Conservation of Stone

Wednesday 24 October 2012

Oral PresentationsMethods and Materials of Cleaning, Conservation, Repair and Maintenance; Logistics and Planning Session IX: 10:30 12:15

CHARACTERIZATION OF HYDRAULIC MORTARS CONTAINING NANOTITANIA FOR RESTORATION APPLICATIONS N. Maravelaki1*, E. Lionakis1, C. Kapridaki1, Z. Agioutantis2, A. Verganelaki1, V. Perdikatsis2 Department of Science Department of Mineral Resources Engineering 1,2 Technical University of Crete, Polytechnioupolis, Akrotiri, 73100 Chania, Crete, Greece
2 1

Abstract In this work nano-titania of anatase form has been added in mortars containing (a): binders of either lime and metakaolin or natural hydraulic lime and, (b): fine aggregates of carbonate or silicate nature. The aim was to study the effect of nano-titania in the hydrolysis and carbonation of the above binders widely used in the design of restoration mortars, as well as the mechanical properties of the derived mortars. The nano-titania proportion was 4.5-6 per cent w/w of binders. The physicochemical and mechanical properties of the nano-titania mortars were studied and compared to the respective ones, without the nano-titania addition, used as reference. DTA-TG, FTIR and XRD analyses indicated the evolution of carbonation, hydration and hydraulic compound formation during a one-year curing. The mechanical characterization indicated that the mortars with the nano-titania addition showed improved mechanical properties over time when compared to the specimens without nano-titania. The results evidenced carbonation and hydration enhancement of the mortar mixtures with nano-titania. The hydrophylicity of nano-titania improves the humidity retained in mortars, thus facilitating the carbonation and hydration process. This property can be exploited in the fabrication of mortars for adhering fragments of porous limestones from monuments, where the presence of humidity controls the mortar setting and adhesion efficiency. A specifically designed mechanical experiment based on the direct tensile strength proved the suitability of these mortars with nano-titania as adhesive materials for restoration applications. Keywords: adhesive mortars, nano-titania, metakaolin-lime, hydraulic lime, hydration, mechanical properties 1. Introduction The adhesion of mortars to fragments of archaeological stone or other building materials is an important intervention, which results in a substantial structural integrity between the adhered materials, leading to the slowing or preventing from further decay. Treatment options include the application of adhesives and grouts, as well mechanical pinning repairs. Commonly used adhesives such as epoxy, acrylic and polyester resins demonstrated excessive strength, high irreversibility and, if improperly applied, their removal may be more damaging to the historic fabric (Amstock 2000).
*

Corresponding author. Tel.: +30 (28210) 37661; fax: +30 (28210) 37841. E-mail address: pmaravelaki@isc.tuc.gr (Noni Maravelaki)

In this research work, two different kinds of stone from Piraeus, namely Aktites and Mounichea stones corresponding to a hard dolomitic limestone and a marly limestone of the area, respectively, were selected due to their common employment as main construction materials of the Athenian Acropolis building during the Archaic period. The preferred treatment strategy for the reassembling and adhesion of these fragments was addressed through designing bonding mortars compatible to these stones. Repair criteria, were as follows: (a) physico-chemical and mechanical compatibility between repair materials and stone, (b) adequate strength to resist tensile and shear forces, (c) retreatability, (d) longevity, (e) affordability and (f) ease of installation. The design of adhesive mortars with binders of either hydrate lime-metakaolin or natural hydraulic lime has been adopted with the aim of formulating a complex system characterized by the highest compatibility. Nowadays, both of hydrate lime-metakaolin and hydraulic lime mortars are widely used in the field of the restoration and conservation of architectural monuments, due to its capability to enhance the chemical, physical, structural and mechanical compatibility with historical building materials (stones, bricks and mortars) (Rosario, Velosa, Magalhaes, 2009). This compatibility is a very critical prerequisite for the optimum performance of conservation mortars considering the damages caused to historic monuments during the past decades, due to the extensive use of cement mixtures. Into this framework, in special designed mortars consisting of binders of either lime and metakaolin or natural hydraulic lime and fine aggregates of carbonate nature, nanotitania of anatase (90 per cent) and rutile (10 per cent) has been added in 4.5-6 per cent w/w of binder. The aim was to study the effect of nano-titania in the hydration and carbonation of the above binders and to compare the physico-chemical properties of the nano-titania mortars with those mortars without nano-titania (used as reference). Thermal analysis (DTA-TG), infrared spectroscopy (FTIR) and X-ray diffraction (XRD) analyses were performed to investigate the evolution of carbonation, hydration and hydraulic compound formation during a six-month curing period. Furthermore, the stonemortar interfaces, the adhesion resistance to external mechanical stress as related to the physico-chemical characteristics of the stone-mortar system and the role of the nano-titania as additive were reported and discussed in this paper. 2. Experimental procedure

2.1 Design of adhesive mortars: binders, fillers and aggregates Binders of either lime (L: by CaO Hellas) with metakaolin (M: Metastar 501 by Imerys), or natural hydraulic lime (NHL: NHL3.5z by Lafarge) as well as nano-titanium dioxide (T: nano-structured nano-titania by NanoPhos), used as filler due to its photocatalytic activity, are employed for the design of the adhesive mortars. The already established photocatalytic activity of nano-titania in anatase form (Hyeon-Cheol, Young-Jun, Myung-Joo et al. 2010) will significantly enhance the hydration and carbonation process, thus affecting the adhesion performance. Moreover, self-cleaning properties of the adhesive mortars can be also attained due to the photocatalytic action of the nano-titania. XRD, FTIR and DTA-TG techniques were used to characterize the raw products. In Table 1, the mortar mixes are presented, where the ratio of water to binder (W/B) ranges from 0.8 to 0.6. The required quantity of lime that will react with metakaolin was

fixed in a weight ratio equal to 1.5, ensuring the pozzolanic reaction. Any unreacted quantity of lime, after its carbonation, provides elasticity to the final mortar and enables the mortar to acquire a pore size distribution similar or compatible to porous stone, thus allowing a homogeneous distribution of water and water vapor in the complex system. Furthermore, the enhanced derived elasticity can function as a tool for the arrangement and absorption of external stresses, which otherwise could lead to the mechanical failure of the mortar. Table 1: Mortar mixes (composition in mass %) Samples Sand Binders Code Cc NHL M L NHLT1 48 49 NHLT2 33 64 ML1 50 20 30 MLT1 47 20 30

Filler Nano-titania 3 3 0 3

B/A 1 2 1 1

W/B 0.7 0.6 0.8 0.8

Preliminary tests on the adhesive capability of the designed mortars with fragments of porous limestones pointed out the inefficient performance of NHL mortars. Thus it was decided to exclude this formulation from further study. TiO2 nano-powder dispersion in a small amount of water was achieved through ultrasonic treatment for 15 min; afterwards the obtained TiO2 colloidal solution was subjected to UV radiation (365 nm) for 30 min to activate the nano-titania. Then the dispersed TiO2 solution was mixed with the other raw material and stirred with a handheld mixer for 5 min. Due to the fact that the fine aggregates can contribute to the avoidance of shrinkage and cracking during the setting process, the addition of sand with fine grains was deemed essential. Consequently, equal proportions of sand passing through the 125 and 63 m sieves were added in the mix, which were previously washed by water to free the harmful soluble salts. 2.2 Assessment of the adhesive mortars 2.2.1 Physico-chemical properties of mortars When binders of powdered pozzolans, such as metakaolin are mixed with lime, or natural hydraulic lime is mixed with water, they produce a new binder that exhibits a hydraulic character due to reactions among the amorphous phase of pozzolans and lime (Aggelakopoulou, Bakolas and Moropoulou 2011), as well as hydration of NHL (Maravelaki-Kalaitzaki Karatasios Bakolas et al. 2005). The pozzolanic and hydration reactions, which take place in room temperature and in conditions of high relative humidity, lead to the formation of a hydrous gel of calcium silicate hydrate (CSH) and calcium aluminate hydrated phases (CAH), which modify the microstructure of the paste and increase both the hydraulic properties and the strength of the mortar (Tziotziou Karakosta Karatasios et al. 2011). Therefore, the study of the hydration is essential in order to evaluate the performance of the mortar, in terms of physical and mechanical properties, which are also interrelated to the longevity of the mortar (Papayianni and Stefanidou 2006).

The above mixtures (Table 1) were molded in prismatic and cubic moulds, with dimensions of 4416 cm and 555 cm, respectively and then placed in a curing chamber for setting, at RH = 65 3 % and T = 20 2 oC, according to the procedure described in the EN 196-1 standard. Pastes of these mixtures with and without nanotitania, with dimensions of 5 mm in diameter and 30 mm in height, were also prepared and sealed into ceramic tubes using Parafilm membrane to avoid moisture loss and drying and were then maintained at the same curing conditions with the studied mortars. The setting process of the paste was interrupted at preset time periods, of 1, 3, 5, 7, 11, 21, 28 and 90 days according to a hydration stop procedure, which involved the immersion of the sample in two stop-bath solutions (acetone and diethyl-ether) for 60 min each, and then drying at 70 oC for 30 h. The development of the hydration and carbonation of powder samples were carried out by XRD, mercury intrusion porosimetry (MIP), FTIR, DTA/TG, EDXRF and scanning electron microscopy (SEM). By identifying CSH and CAH at different ages, not only qualitatively, but also semi-quantitatively, the hydration and carbonation process can be monitored. The mineralogical analysis was investigated by XRD with a Siemens D-500 diffractometer (40 kV/35 mA) and the spectra were collected between 5 and 60o 2 scale, with a step of 0.03/5s. SEM analysis was carried out in fractured surfaces, using a FEI Quanta Inspect scanning electron microscope. DTA/TG was operated with a Setaram thermal analyser; in static air atmosphere up to 1000 oC at a rate of 10 oC/min. The FTIR analysis of KBr pellets was operated in Perkin-Elmer spectrophotometer in the spectral range of 400-4000 cm-1. MIP measurements were recorded using a Quantachrome Autoscan 60 porosimeter, in the range of 2-4000 nm. Physical properties of the stone and mortars were further studied by water absorption by saturation according to EN 13755:2002, as well as capillary water absorption measurements for mortars according to EN 1015-18:1995 and for stones according to EN 1925:1999. 2.2.2 Mechanical estimation of the designed adhesive mortars The Aktites and Mounichea stone samples were cut and shaped into specimens with dimensions of: (a) 4x4x4 cm used to measure the compressive strength of the stone and (b) 4x4x8 cm used to bond them with the designed mortars. The mechanical properties of the designed mortars were characterized by measuring the uniaxial compressive strength (Fc) and the flexural strength (Ff-3pb) according to EN 1015-11:1999. The incising of stone faces by special mechanical tools to provide a rough surface and the wetting of these surfaces, were prerequisite before the application of the mortars. The mortar was partially applied to the stone surface and then the stone specimens were filled with mortar by placing them levelly with the aid of special joint clamps. The joint was kept moist with damp cotton gauze and a polyethylene sheet. The specimens were then placed in a storage chamber for 28 days, according to the conditions described in EN 1015-11:1999. Home designed equipment for both the four point flexural strength (Ff-4pb) (Fig. 1a, b) and the direct tensile strength test (Ft) (Fig 1c), were used for measuring the adhesive performance of the bonded stone-mortar specimens (4x4x17 cm) (Whittaker, Singh and Sun 1992).

Figure 1.

a) Four point bending apparatus with variable support spans b) geometry and calculations for four-point bending test c) direct tensile test apparatus for stone mortar specimens.

3.

Results and discussion

3.1 Stones and raw materials characterization Table 2: Mineralogical composition and properties of the stones and raw materials. Stone Cc Do Cl Qz Kl Il Alb Fc WCC WSC (MPa) (g cm-1 s-1/2) % 6.2 83.0 0.6 3.7 1.8 4.2 0.5 42.3 0.0026 3.7 D1 (0.001) (0.1) 1.5 78.7 2.8 6.1 2.5 7.8 0.5 28.7 0.0141 13.1 D2 (0.004) (2.8) 12.0 80.0 0.9 2.7 2.4 1.9 0.1 107.0 0.0011 2.3 D3 (0.0002) (0.2) 4.2 80.0 1.3 3.3 3.3 7.0 0.9 6.6 0.0007 2.9 D4 (0.0004) (0.9) 11.1 70.9 1.1 5.4 0.9 8.1 2.2 22.7 0.0211 8.3 K (0.005) (2.0)
Cc: Calcite; Do: Dolomite; Cl: Chlorite; Qz: Quartz; Kl: Kaolinite; Il: Illite; CH: Calcium Hydroxide; C2S-beta: Larnite; C3S: Alite; At: Anatase; Rt: Rutile; Fc: Compressive Strength; WCC: Water Capillary Coefficient; WSC: Water Saturation Coefficient

Mineralogical composition and properties, both of raw materials and stones, determined by XRD and thermal analyses are presented in Table 2. According to these results, the Piraeus stone can be classified in: (a) micritic dolomitic limestone hard and compact, with a low to medium porosity, small grain size and high values of mechanical strength (D1, D3); (b) marly limestone/dolomitic limestone with an oolitic texture, brownish or yellowish to light grey-color with a low to medium porosity and low values of mechanical strength (D2, D4, K). Furthermore, Table 2 reports the values of the compressive strength, as well as mean values of water capillary and water saturation coefficient calculated in three stone specimens. The water capillary and water saturation tests indicate that the pore system

of the studied stones differed significantly and therefore the designed mortars should be adapted accordingly. Stones with high compressive strength, such as D1 and D3 absorb a low water amount correspondingly to their low quantity of aluminosilicates. However, in the D2, D4 and K stones no clear relationship exists between compressive strength and hygric properties. Even though these stones contained similar amounts of aluminosilicates, they differed both in the absorbed quantity of water and the values of the compressive strength. This relies on the structural inhomogeneity of the samples and especially, as in the case of D4 stone, on the absence of inter-connected pores, which affect the stone hygric behavior. 3.2 Physico-chemical evaluation of mortars The total bound water (Htb) identified in the studied samples in the temperature range from 100 to 460 oC corresponds to the dehydration of calcium silicate hydrates (CSH), calcium aluminate hydrates (CAH) and the residual bound water (Aggelakopoulou, Bakolas and Moropoulou 2011). The dehydration of Ca(OH)2 (CH) occurred in the temperature range ~480 500 oC, while the decomposition of CaCO3 took place at a temperature higher than 600 oC.

Figure 2.

DTA curves for NHL mortars without nano-titania (dashed line) and with nanotitania (solid line) at 7 days of curing along with the evolution of carbonation illustrated in the inset plot.

In particular, Figure 2 depicts the DTA curves for the NHL mortars with and without nano-titania at 7 days of curing. The inset plot illustrates the evolution of the ratio unreacted-CH (CHun) to formed-Cc(Ccf) for a curing period of 1, 3, 5 and 7 days. As the setting process proceeds mass losses attributed to the release of CO2 from CaCO3 increased, while the mass loss of CH dehydration decreased, due to the transformation of CH into hydraulic components and calcite. The lime consumption and the formation of hydraulic phases are more pronounced for the nano-titania mortars at different curing times. The same observation was also obtained from the thermal analysis of metakaolinlime mortars with and without nano-titania. SEM micrographs of the mortars ML1 without- (Figure 3a) and with nano-titania MLT1 (Figure 3b) after one year of curing corroborate the results of DTA analysis.

SEM micrograph of MLT1 (Figure 3b) shows that a dense network of hydraulic amorphous components was formed providing more elasticity to the mortar matrix. Moreover, characteristic hexagonal crystals of portlandite were located in both SEM micrographs; nevertheless in the case of MLT1 mortar the portlandite crystals are obviously fewer. The FTIR spectra of the above studied mortars are in full accordance with the SEM micrographs, showing after curing of one year traces of portlandite and enhanced hydraulic compound formation (Maravelaki-Kalaitzaki, P. Lionakis, E. Agioutantis, Z. et al. 2012).

Figure 3.

SEM micrographs of (a) ML mortar and (b) MLT mortar showing portlandite crystals (P) and the hydraulic amorphous components.

These variations can be explained by the TiO2 photocatalytic action for the mixtures with nano-titania which leads both to calcite formation and enhanced hydration of CSH and CAH products, similar to what was observed on the early age hydration of Portland cement by adding increased dosage of fine titanium oxide (Jayapalan, Lee and Kurtis 2009). 3.3 Mechanical Evaluation Table 3 reports the physical and mechanical properties of the designed mortars. The physical properties of the designed mortars differed insignificantly indicating that the nano-titania addition neither modified the microstructure nor affected the hygric behaviour of the materials. The lowest Fc values were recorded for the NHL samples, while the Fc values decreased with curing time in the ML1 samples. Even though, the Fc values recorded at four weeks curing for the MLT1 samples are lower than the corresponding values for samples without nano-titania (ML1), nevertheless, the Fc values of the MLT1 samples reached higher values than the ML1 samples after three months and one year curing, thus indicating the beneficial effect of the nano-titania in the compressive strength. The decrease of Fc values over time in the ML1 compositions has been already reported by other authors (Aggelakopoulou, Bakolas and Moropoulou 2011, Velosa Rocha and Veiga 2009) and was most probably attributed to the microcracking formation due to shrinkage during the curing. Figure 4 depicts typical stress-strain curves of hydrated lime-metakaolin with (MLT1) and without nano-titania (ML1), in two time intervals of 4 weeks and 3 months.

Figure 4.

Stress-strain diagram of samples without nano-titania at 4 weeks (ML1 [4w]) and 3 months (ML1 [3m]) curing time and with nano-titania at 4 weeks (MLT1 [4w]) and 3 months (MLT1 [3m]) curing time.

Table 3: Physical and chemical properties of the designed mortars. Code WCC* WSC* P Pr SSA Cur Fc^ g cm-1 % % m m2/g -ing MPa s-1/2 27.80 32.96 0.295 12.0 4w 4.15 (0.56) NHLT1 0.0149 (0.007) (1.09) 3m 5.47 (0.28) 1y 5.51 (0.95) 25.58 37.79 0.092 12.84 4w 5.41 (0.29) NHLT2 0.0080 (0.0004) (0.27) 3m 7.22 (1.61) 1y 0.0040 33.92 31.46 0.031 14.02 4w 14.85 (1.53) ML1 (0.0001) (0.26) 3m 11.57 (1.39 1y 10.62(2.39) 0.0074 29.52 32.54 0.031 16.01 4w 9.08* (0.89) MLT1 (0.003) (1.07) 3m 14.19 (0.70) 1y 15.40(0.70)

Ff3pb* MPa 1.69 1.15 1.21 -

E GPa 0.17

0.37

0.42 0.76 0.56 0.59 1.10 0.91

(*) mean value of three samples; (^) mean value of 6 samples; Fc: compressive strength; Ff-3pb: Flexural strength; E: Elasticity modulus, [4w]: 4 weeks; [3m]: 3 months; [1y]: 1 year; -: n. a.

By combining the results of physical and mechanical properties the mortars NHLT2, ML1 and MLT1 were selected as adhesive means for the stones under consideration (Table 4). The mortar NHLT1 showed similar values to others except for its high water coefficient capillary, which can be considered of secondary importance for the adhesive ability. However, NHLT1 exhibited difficulty in joining the stone specimens and therefore was not included in the finally selected mortars. It seems that nano-titania with its hydrophylicity created an environment, which not only enhanced the hydraulic component formation, but also controlled the shrinkage, thus avoiding microcracking (Karatasios Katsiotis Likodimos, et al. 2010). Further

support to this statement is derived from the dense network of hydraulic components observed in the SEM micrographs (Fig. 3) for the MLT1 samples. Table 4: Mechanical properties of stone-mortar specimens cured for four weeks. Mortar Number of Ff-4pb Ff-4pb Number of Ft Ft Code adhered stone (MPa) adhered stone (MPa) specimens specimens 4 (D1, D2) 2.39 (0.7) 2 (D1, D2) 0.51 (0.26) NHLT2 3 (D3, K) 1.34 (0.47) 2 (K) 0.15 (0.08) MLT1 1 (D4) 1.07 1 (D1) 0.09 ML1 Table 4 reports the results of the 4-point bending test and the direct tensile test for the adhered stone-mortar specimens. In all tests, failure was observed at the interface between stone and mortar. The results revealed that the higher bonding strength (higher flexural and tensile) was measured when applying the NHLT2 mortar as compared to the MLT1 and ML1 mortar.

4.

Conclusions

This research work addressed an important problem in the restoration sector concerning the reassembling of stone fragments from ancient monuments using noncement mortars. The proposed adhesive mortars contain hydraulic lime or metakaolin and lime as binders, carbonate sand with grains smaller than 250 m in a B/A ratio from 1 to 2, as well as nano-titania as additive in a binder replacement of 4.5-6%. The mechanical characterization indicated that the mortars with nano-titania showed increased compressive and flexural strength and modulus of elasticity when compared to the specimens without nano-titania. The results also indicate enhanced carbonation and hydration of mortar mixtures with nano-titania. The hydrophylicity of nano-titania improves the humidity retention into mortars, facilitating thus the carbonation and hydration processes. This property can be exploited in the fabrication of mortars tailored to adhering porous limestones, where humidity controls the mortar setting and adhesion efficiency. Home-designed equipment applied to measure the bonding strength of stonemortar systems revealed that the nano-titania addition in both metakaolin-lime and hydraulic lime mortars improves the adhesive property of the mortar when applied to porous stones. ACKNOWLEGMENTS This work was carried out in collaboration with the The Akropolis Restoration Service (YSMA) of the Hellenic Ministry of Culture and Tourism. The authors would like to thank Emeritus Prof. Vasileia Kasselouri-Rigopoulou, (Committee for the Restoration of the Acropolis Monuments-ESMA) for collaboration and scientific support, the Committee for the Restoration of the Acropolis Monuments and Dr E. Aggelakopoulou Head of the Technical Office for the Acropolis Monuments' Surface Conservation. This research has been co-financed by the European Union (European Social Fund - ESF) and Greek national funds through the Operational Program "Education and Lifelong

Learning" of the National Strategic Reference Framework (NSRF) - Research Funding Program: Heracleitus II, Investing in knowledge society through the European Social Fund. 5. References Aggelakopoulou E, Bakolas A, Moropoulou A. 2011. Properties of limemetakolin mortars for the restoration of historic masonries. Applied Clay Science, 53(1):1519. Amstock JS. 2000. Handbook of adhesives and sealants in construction. New York: McGraw-Hill. Hyeon-Cheol J, Young-Jun L, Myung-Joo K, Ho-Beom K, Jun-Hyun H. 2010. Characterization on titanium surfaces and its effect on photocatalytic bactericidal activity. Applied Surface Science, 257(3):741746. Jayapalan AR, Lee BY, Kurtis KE. 2009. Eect of Nano-sized Titanium Dioxide on Early Age Hydration of Portland Cement. In Nanotechnology in Construction, Proceedings of the NICOM3. Karatasios I, Katsiotis MS, Likodimos V, Kontos AI, Papavassiliou G, Falaras P, Kilikoglou, V. 2010. Photo-induced carbonation of lime-TiO2 mortars. Applied Catalysis B: Environmental, 95(1-2):78-86. Maravelaki-Kalaitzaki, P. Lionakis, E. Agioutantis, Z. Kparidaki, C. Verganelaki, A. Mayrigiannakis, S. Stavroulaki, M. Perdikatsis, V. Kallithrakas-Kontos, N. 2012. Physico-chemical and mechanical characterization of hydraulic mortars containing nano-titania for stone applications, 4th International Symposium on Nanotechnology in Construction, Agios-Nikolaos. Papayianni I, Stefanidou M. 2006. Strengthporosity relationships in limepozzolan mortars. Construction and Building Materials, 20(9):700-705. Rosario Veiga M, Velosa A, Magalhaes A. 2009. Experimental applications of mortars with pozzolanic additions: Characterization and performance evaluation. Construction and Building Materials, 23(1):318-327. Tziotziou M, Karakosta E, Karatasios I, Diamantopoulos G, Sapalidis A, Fardis M, Maravelaki-Kalaitzaki P, Papavassiliou G, Kilikoglou V. 2011. Application of 1H NMR in hy-dration and porosity studies of lime pozzolan mixtures. Microporous and Mesoporous Materials, 139(1-3):16-24. Velosa AL, Rocha F, Veiga R. 2009. Influence of chemical and mineralogical composition of metakaolin on mortars characteristics. Acta Geodynamica et Geomaterialia, 153(6):121-126. Whittaker B., Singh N., Sun G. 1992. Rock Fracture Mechanics: Principles, Design and Applications, Elsevier.

Compatibility of Roman cement mortars with gypsum stones and anhydrite mortars: The example of Valre Castle (Sion, Switzerland)
C. Gosselin1*, F. Girardet2, S.B. Feldman3
Laboratory of Construction Materials, Ecole Polytechnique Fdrale de Lausanne, CH-1015 Lausanne, Switzerland, christ.gosselin@gmail.com 2 Rino Sarl, CH-1807 Blonay, Switzerland, fred.girardet@bluewin.ch 3 Active Minerals International LLC, Cockeysville, USA, s.feldman@activeminerals.com
1

1. Abstract
Durable and reversible restoration of stones and historical mortars is a major concern to those interested in conservation of historical structures and contemporary practices of restoration are continuously revisited with the help of the technical and scientific researches. However, the historical feedback on repairing techniques recently showed that Roman cements (RC), developed and widely used through the XIXth Century, were particularly well-adapted to repair historical masonries. The current article presents a case study of RC mortars applied on gypsum stones and historical anhydrite mortars, both soluble and known to be sensitive to the chemical compatibility with hydraulic binders. Mineralogical analysis of samples from the basilica Notre Dame de Valre (Sion, Switzerland) shows that late XIXth C. RC joints and renders have perfectly lasted in contact with the structural gypsum stones and anhydrite mortars from the XIIIth C. Results from XRD and SEM work suggest that the present RC was produced at a temperature high enough to form significant amounts of -C2S and C2AS, remaining unreacted after very long term hydration. The extent of C2S hydration is notably reduced due the precipitation of silica gel, a carbonation product, at the boundary of the cement grains. The high capillary porosity developed during hydration is homogeneously distributed, enhancing the transport properties. These conclusions were supported by complementary observations. First, elemental mapping through the strong RC /anhydrite mortars interface does not indicate any accumulation of sulfate salts at the boundary. Additionally, in contrast to the RC mortars, the rapid expansion and degradation of grey Portland cement mortars was observed, confirming the limitations of the latter applied on gypsum stones.

Keywords: Roman cement, gypsum stone, anhydrite mortars, microstructure, compatibility

* Now at Geotest, Le Mont-sur-Lausanne, Switzerland, christophe.gosselin@geotest.ch

2. Introduction
The basilica Notre Dame de Valre (also called Castle of Valre) is part of a fortified village built in the XIIth C., on the top of a hill overlooking the city of Sion. In 1877, the cantonal Great Council called upon the Government to report on the ownership of the feudal castles of the canton and the measures for their conservation. A major restoration campaign of the Valre Castle was conducted between 1892 and 1902 under the supervision of the Swiss Society for the Conservation of Historic Buildings. The work was led by the Office Kalbermatten under the direction of the architect Theophile van Muyden [1].

The masonry is composed of different types of local stones (calcareous, tuff and gypsum stones). Two varieties of local gypsum stones were used (white gypsum stones, i.e. alabaster, and yellowish gypsum stones containing sand) for both ashlar masonry and sculpted ornamentations in areas unexposed to rain. The protection of the gypsum stones from the rain led to their relatively good state of conservation despite their high solubility (2 g/l). However, the dissolution of gypsum stones in areas exposed to rain, and the subsequent release of sulphate ions can lead to major conservation concerns related to their compatibility with the different hydraulic binders used for the joints, the structural and decorative elements [2]. This article discusses the specific case of restoration joints made of XIXth C. Roman cement based mortars. Several types of mortars are present on the faades, according to the period of construction and restoration of the building. Six types of mortars were identified as shown in Table 1. The present study focuses on the compatibility of the XIXth C. Roman cement mortars (types 3 and 4) versus that of the Portland cement mortars (type 5) with the gypsum stones and the anhydrite mortars (type 2).
Type of mortar Type 1 Type 2 Type 3 Type 4 Type 5 Type 6 Description Original pietra rasa lime mortar (XIIIth C.), white Original anhydrite mortars (XIIIth C.), pinkish Restoration Roman cement mortars (1896), thin pointing mortar, concave (spatuled), covering the original lime mortar, beige to greyish Restoration Roman cement mortars (1898), thick and extruded repointing mortars, beige to greyish Restoration Portland cement mortars (end of the XIXth C., as Type 3 and 4), cast mortar to repair elements made of gypsum stone, dark grey Restoration hybrid mortar (hydraulic lime and white cement) (1997-2003), repointing deep joint, white Table 1 Types of original and restoration mortars identified on the faades

The use of anhydrite (CaSO4) or gypsum (CaSO4.2H2O), commonly with lime, to produce joint, precast or render mortars is reported in several historical buildings from the Ancient Egypt period (e.g. Cheops and Unas pyramids [3]) to more recent periods of decorative architecture [4,5], through the Medieval period in Europe (e.g. North German [6] and French cathedrals [7]). The choice of anhydrite or gypsum was usually motivated by the appearance (colour, texture) of hardened calcium sulphate based mortars close to that of lime mortar. Mixtures of lime and gypsum (so-called estrich gyps) are also reported in German monuments. Roman cement is a hydraulic binder developed in Europe in the early XIXth C. and widely used both for civil engineering and architectural restoration applications. Several studies in the field of stone and mortars conservation have revisited and highlighted the unique and lasting properties of this material to repair historical masonries [8-10]. Roman cements are sometimes called Natural cement, in contrast to artificial Portland cements (i.e. co-ground with added gypsum) in some countries such as France [11] or USA [12]. This family of hydraulic binder results from the calcination of naturally occurring limestones rich in clay minerals below the sintering point (800-1000C) and the grinding of the burnt material to a fine cement. A typical shaft kiln was used during the early production of these cements. Roman cement has been produced from the 1830s in the North of Switzerland (Solothurn and Aarau) and increasingly used throughout the XIXth C. The development of the Swiss railway network in the 1850s required rapid materials for the lines construction and contributed to the promotion of the Swiss Roman cements. The annual production of cement reached 60 t in

1851. However, despite of the quality of the local raw materials, the Swiss production could not compete with the other European producers and the importation of Roman cement from France increased from 1857 (with an import of 35000 tons recorded in 1876). The two main production sites were Vicat (Grenoble) and Pavin de Lafarge (Virieu le Grand, Isre) [13].

3. Samples and methods

Figure 1 View of the Northern faade of the Valre castle and samples location (after Amsler and Gagliardi)

Figure 1 gives the location of the mortars samples from the Northern faade of the church. The mortar 2 was sampled from the Eastern faade (annex buildings in ruins not illustrated in Figure 1). Figure 2 gives detailed views (cross sections) of the four samples (three Roman cement mortars and one Portland cement mortar) used for the microstructural characterisation (SEM and XRD). Scanning Electron Microscopy (SEM, Philips Quanta 200) was used to study the microstructure of the mortar samples. The samples were impregnated with epoxy resin and polished to obtain cross sections. The microanalysis of phases and elemental mappings were done with Energy Dispersive Spectroscopy (EDS, Bruker AXS Quantax). XRD analysis was done on sieved mortar samples using an XPert Pro PANAlytical diffractometer (Cu tube, =1.54 ). The Rietveld method was applied for the crystalline phase quantification.

4. Results
1. Macroscopic observations of the mortars samples
The Portland cement mortar (sample 1) was applied to replicate a column of a window frame originally made of white gypsum stone. From Figure 2, the bulk mortar sample looks beige (most probably due to the superficial atmospheric carbonation), but the cross section prepared for microscopy reveals a grey matrix. The bulk mortar is dense and no specific degradation pattern is observed. The outer subsurface (exposed to the environment) is notably distinguished by a colour slightly switching from grey to light beige. The dissolution of the gypsum stone substrate, appearing yellow after epoxy resin impregnation, is well advanced and a high amount of matter was lost during the sample preparation. A thick reaction interface is marked at the mortar/stone interface (top right image of Figure 2).

Figure 2 Photographs and description of the mortars samples

The bulk Roman cement samples have different colours according to their location and application. Sample 2 was applied as a thin render on the original XIIIth C. anhydrite mortar. While the matrix of the cement mortar is dark grey (outer surface and bulk), that of the anhydrite mortar is pinkish and contains coarse white inclusions. The two mortars are strongly bound by a very thin interface. Sample 3 is a thick joint partly covering and profiling an adjacent a gypsum stone (some fragments of stone remains visible in Figure 2). This sample was removed from the faade and collected on the first roof of the Northern faade. The outer surface of the sample looks beige but the cross section shows a dark grey matrix, comparable to the Portland cement sample 1. The cross section also reveals the formation of a thin (500 micron) and yellowish subsurface that is exposed to the atmosphere. This specific layer was already reported in French [9] and Austrian [8] samples but its origin (atmospheric oxidation, footprint of organic product, such as wax, originally used for technical or aesthetical purposes,) is not fully

understood. The mortar is composed of coarse aggregates. Micropores from mixing are randomly distributed through the matrix. Sample 4 was collected from a core in a thick joint. The light colour (beige) of the bulk mortar remains even after the cross section preparation and this mortar is notably lighter than the previous ones. Figure 2 shows a dense matrix despite a high macroporosity formed during the mixing.

2. XRD analysis on the mortars

The crystalline composition of the RC mortars, suggested by the best-fit values of the Rietveld analysis, is given in Table 2. The nomenclature of cement chemistry (C=CaO, S= SiO2, A= Al2O3, F= Fe2O3, C=CO3, H=H2O) is used for given phases in this table.
Origin of phase Phase Microcline KAlSi3O8 Albite NaAlSi3O8 Clinochlore (Mg,Fe)[(OH)8AlSi3O8 Muscovite K2Al4[(OH,Fe)AlSi3O10 Actinolite Ca2(Mg,Fe)5Si8O2.2(OH)2 Quartz SiO2 Periclase MgO Tileyite Ca5(Si2O7)(CO3)2 Belite -2CaO.SiO2 or -C2S Alite 3CaO.SiO2 or C3S Gehlenite 2CaO.Al2O3.SiO2 or C2AS Ferrite Ca2(Al,Fe)2O5 or C4AF Calcite CaCO3 or CC Vaterite CaCO3 Aragonite CaCO3 Portlandite Ca(OH)2 or CH Sample 1 1.2 14.3 3.7 4.0 2.2 17.7 10.7 7.4 3.0 0.7 9.0 7.5 Sample 2 4.2 15.0 5.6 3.6 2.5 13.9 2.0 9.5 26.2 1.8 6.2 Sample 3 2.0 5.7 3.5 5.0 1.6 16.8 1.4 1.4 26.3 2.9 2.6 15.3 0.8 1.7 Sample 4 4.2 16.4 4.7 3.0 1.6 13.0 12.4 35.4 1.0 6.2 -

(1)

(2)

(3)

(4) (5) (6) (7)

Gypsum CaSO4.(H2O)2 3.6 8.4 3.4 2.0 4.4 8.4 Ettringite Ca6(Al(OH)6)2(SO4)3(H2O)26 Table 2 Quantitative XRD of the mortar samples : (1) from aggregates, (2) from aggregates or anhydrous cement, (3) from anhydrous cement, (4) from aggregates, anhydrous cement or carbonation product, (5) carbonation products, (6) hydration products, (7) hydration products (Ett.) or reaction products with external sulfate

Table 2 discriminates different sources of crystalline phases. Some common minerals such as microcline, albite, clinochlore, muscovite and actinolite originate from the local sand and were identified in various amounts in all samples. Quartz and calcite could also be attributed to the aggregate fraction in the Portland cement mortar (sample 1) but part of these phases can be also attributed to residual remnants of Roman cements [14]. Despite the long time of contact with moisture on the faades, the roman cement samples (2 to 4) contain unreacted phases. The polymorph of belite is identified in different amounts. In the typical range of calcination temperature of roman cement, -2CaO.SiO2 dominates and hydrates afters few weeks of contact with mixing water and moisture [15]. When the calcination temperature becomes higher, -2CaO.SiO2 can form but is less reactive than the polymorph. By achieving higher temperature in the kiln, more alumina from the raw clay minerals becomes available to form gehlenite (2CaO.Al2O3.SiO2), a low reactivity phase. Note that -belite and gehlenite are also identified in the Portland cement sample. Indeed -

belite is a secondary reactant in Portland cements much less reactive than tri-calcium silicate 3CaO.SiO2, the main reactive phase responsible for the strength development. Gehlenite, which is sparingly present in contemporary Portland cements and has been identified in early PC (XIXth C.) [16], is said to be suggestive of underburnt Portland cement (usually calcined above 1400C) [17]. As abovementioned, the presence of calcite can be attributed to the mortar sand but also to the carbonation of cement hydration products. Both Portland and Roman cement, and tri- and dicalcium silicate, respectively, hydrate to form calcium silicate hydrate (C-S-H) and calcium hydroxide (CH). Remaining calcium hydroxide was identified in samples 1 and 3 (Table 2) but its carbonation under atmospheric conditions (e.g., CO2, wetting/drying cycles) led to the precipitation of calcite and other metastable calcium carbonate polymorphs (aragonite and vaterite), particularly in the Roman cement samples. The metastable CaCO3 polymorphs transform to stable calcite by dissolution-precipitation reaction but can coexist in dry environment [18], which is the case of the Northern faade of the present structure. Note that the C-S-H phase could also be subject to carbonation and comparable carbonation products could be attributed to this reaction. Gypsum and ettringite were identified in the different mortars samples. Ettringite is a primary hydration product in Portland cement (reaction between the calcium aluminate and the sulfate form added gypsum) but rapidly dissolve after the sulfate depletion (first days of the cement hydration). This phase can precipitate again when external or internal sources of sulfate are available. In the present case, this mortar was applied on a column originally made of gypsum stone, continuously providing sulfate allowing ettringite to precipitate. Some Roman cements (American Rosendale RC, French Vicat and Vassy RC) are reported to contain sulfate phases [9,19] leading to the early formation of ettringite. However ettringite was identified only in the sample 3 (reprofiling mortar applied on the structural gypsum stone) while gypsum seems to be the common reaction products present within this series of samples.

3. Microstructure of the mortars by SEM

The microstructure of the mortar samples is illustrated in Figure 3 to Figure 5. The main phases (typical to Portland cement) comprising a cement grain from sample 1 (Figure 3.a) were identified as C3S (with Ca/Si= 3.15 0.3), C3A (with Al substituted by Fe, Ca/(Al+Fe) = 1.62 0.02) and C4AF (with Ca/Al = 2.09 0.05 and Ca/(Al+Fe) = 1.16 0.02). Note that C3S was not identified by the XRD technique. The observation of this sample at lower magnification showed a many unreacted cement grains, suggesting a low degree of hydration. The nature of the hydration products (around the cement grains) is discussed below. In sample 2 (Figure 3.b), three main phases were identified: C2S (rich in Si with Ca/Si= 1.77 0.2), C2AS rich in Si, Ca/Si = 1.7 0.16 and Ca/(Al+Si) = 0.76 0.05) and the silica gel appearing in dark in the BSE images, at the boundary of the cement grain. While C2S and C2AS are common phases formed in Roman cement at relatively high temperature of calcination [14], the presence of the silica gel at the boundary can be considered as a weathering product of the cement exposed to atmospheric carbonation. This product results from the reaction between C2S and CO2, leading to the decalcification of C2S and precipitation of silica gel, which would be a residual product of Ca depletion. This specific reaction is scarcely reported but the mechanism and the nature of the silicate polymer were recently studied on synthesized C2S and Portland cement [20].

Figure 3-c shows another example of Roman cement grain in which only rounded C2S grains were detected. Figure 3-d illustrates a Roman cement grain composed of C2S surrounded by significant amounts of silica gel (as carbonation product of parent C2S).

Silica gel C2S

C4AF C3A C3S

C2AS

MgO

a: Sample 1

b: Sample 2

C2S

C2 S
Silica gel

c: Sample 3 d: Sample 4 Figure 3 Main features of the cement grains

Figure 4 illustrates the microstructure of the samples 1 and 2 and gives the elemental composition (Ca, Al, Si, S, expressed in the atomic ratio graphs) of the hydration products. In the Portland cement mortar (sample 1), we distinguish typically two types of calcium silica hydrate, the inner C-S-H immediately surrounding the cement grains, and the outer C-S-H precipitated further in the microstructure. The latter are less dense and can incorporate other phases such as ettringite or monosulfoaluminate. The EDS dots distribution in the atomic ratio graph suggests that outer C-S-H is intermixed with both ettringite (Ett.) and/or monosulfoaluminate (Ms), while the latter was not detected by XRD, possibly because of their inherent poor crystallinity. The most important feature here lies in the comparison with the composition of the hydration products of the roman cement mortar (sample 2). First, no distinction between outer and inner products was observed in the hydrated binders. In addition, the atomic ratio graphs show a different phase assemblage than previously described. The distribution of Al, Ca and Si suggests that a C-A-S-H type phase intermixed with calcium carbonate dominates in the microstructure, with no clear evidence of a pure endmember C-S-H phase. The degree of carbonation of this sample is very high (e.g., the micrograph of Figure 4 show a cement grain with fully carbonated C2S) which probably makes the analysis of outer hydration products difficult. However, recent studies on modern Roman cements [21] support the fact that C-S-H and CH, as hydration products of C2S, are

not homogeneously distributed in the microstructure even after 90 days of hydration, in contrast to Portland cements. The S/Ca vs. Al/Ca graph of the sample 2 shows that no sulphur bearing phase was detected, in the hydration products of the mortar.
inner CSH outer products +CC

cement grain cement grain

outer CSH +Ett. carbonated C2S

0.5 0.45 0.4 0.35 Al/Ca 0.3 0.25 0.2 0.15 0.1 0.05 0 0 0.2 0.4 Si/Ca 0.6 0.8 1 innerCSH outerCSH

Ms

0.5 0.45 0.4 0.35 Al/Ca 0.3 0.25 0.2 0.15 0.1 0.05 0

Ms

Ett

Sample 1 Al/Ca vs. Si/Ca

Ett

Sample 2 Al/Ca vs. Si/Ca

outer products

C1.7SH

C1.7SH
0 0.2 0.4 Si/Ca 0.6 0.8 1

CHorCC
0.5 0.45 0.4 0.35 S/Ca 0.3 0.25 0.2 0.15 0.1 0.05 0 0 0.1 0.2

CHorCC Ett Sample 1 S/Ca vs. Al/Ca

S/Ca 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05
0.3 0.4 0.5

Ett Sample 2 S/Ca vs. Al/Ca Ms

outer products

Ms
innerCSH outerCSH

0 0 0.1 0.2 Al/Ca 0.3 0.4 0.5

CHorCC

Al/Ca

CHorCC

Figure 4 Microstructure and composition of hydration products, samples 1 and 2

The distribution of sulphur at the interface between the XIIth C. anhydrite mortar and the XIXth Roman cement mortar (sample 2) was studied by elemental mapping. Figure 5 shows a clear interface with no concentration gradient of sulphur at the boundary of the Roman cement mortar. Nonetheless sulphur may diffuse and few dot are visible in the Figure 5-b, which were attributed to the reaction between sulphur and CH to precipitate gypsum in the porosity.

Romancement mortar

Anhydrite mortar

a: BSE image b: Elemental distribution of sulfur Figure 5 Elemental mapping of the interface between anhydrite mortar and RC mortar (sample 2)

Figure 6 shows BSE images and the respective segmented grey level images that illustrate the distribution of the porosity (appearing in black) in the samples 2 and 4. In Figure 6- b and d, the pores resulting from mortar mixing are distinguished from the capillary porosity developed during the hydration of Roman cement. The network of the capillary pores is randomly distributed through the sample, allowing soluble sulphur to migrate and calcium suphate (or ettringite) to precipitate in available space. This provides to Roman cement mortars some specific transfer properties adapted to the substrates (here, anhydrite mortar or gypsum stone).

a: BSE Sample 2

b: Segmented BSE (porosity) Sample 2

c: BSE Sample 4 d: Segmented BSE (porosity) Sample 4 Figure 6 BSE images of samples 2 and 4 (a and c) and segmentation of grey level: distribution of the porosity in the samples 2 and 4 (b and d)

5. Conclusions
The use of Roman cement mortars in the late XIXth C. on the faades of the church of Valere (Sion, CH) was presented. The specific application as joints of structural gypsum stones and render of earlier mortar made of anhydrite was discussed. Roman cement belongs to a specific family of hydraulic binders developed during the XVIIIth C and which shows interesting compatibility properties adapted to stone conservation. The samples show a grade of cement containing significant amounts of unreacted gehlenite and -C2S, suggesting a calcination process using a temperature range above 1000C, likely that of the French Vicat cement and higher than typically reported for Roman cements (8001000C). Recent studies showed that the microstructure development of hydrated modern Roman cements is controlled by dissolution of amorphous CaO-Al2O3 and fine calcium

carbonate (both from the raw cement) and precipitation of carbonated phases in the CaOAl2O3-CO3-H2O system. If sulphur is present in the raw cement, ettringite and monocarboalumiante precipitate, as in the case of the French Vicat cement containing gypsum [19,21]. These elongated crystals (plate of carbonated phases or ettringite needles) precipitate very rapidly to form a poorly packed microstructure in the first minutes of hydration. However, this phase assemblage develops minimal strength, allowing the rapid application of mortars for cast elements or renders. This microstructure becomes progressively and locally denser after the later reaction of calcium silicate (mainly -C2S but also -C2S for the cements produced at higher temperature) forming, in theory, C-S-H and CH. However the presence and location of these two phases in hydrated Roman cement remains difficult to identify in contrast to the case of C3S hydration in Portland cements. This identification becomes more challenging in highly carbonated samples, where different calcium carbonate polymorphs are uniformly distributed in the primary hydration products. Carbonation acts also on the non reacted C2S, during which decalcification leads to the formation of silica gel at the cement grain boundary. This reaction may contribute to the reduction of the extent of C2S dissolution and cement hydration. The excellent state of conservation of Roman cement mortar could be explained by its microstructure and the high capillary porosity developed during hydration [22,23]. This allows the sulphur, dissolved from the gypsum stone, to migrate through the Roman cement mortar and to potentially react to form gypsum (or ettringite in sample 3) without generating internal stress and subsequent cracks formation. This specific feature is one of the most important characteristics that make Roman cements highly suitable as a restoration mortar in contrast to the Portland cement mortar that was inappropriate for replicating a window column originally made of gypsum stone.

6. Acknowledgements
The European Project ROCARE is acknowledged for the financial support of this case study. The architect C. Amlser is also gratefully acknowledged for his interest in this material and the current opportunity to use the Church faade as a demonstration site for modern Roman cement mortar applications. CG would like to dedicate this paper to the memory of Pr. Michele Coutant, her knowledge and her passion of the conservation of cultural heritage.

7. References
[1] [2] [3] [4] [5] Raemy-Berthod, C., 2003. Inventaire suisse darchitecture, 1850-1920: Sion. Socit dHistoire de lArt en Suisse 9 92. Winkler, E.M., 1997. Stone in Architecture: Properties, Durability. Springer Regourd, M., Kerisel, J., Deletie, P. and Haguenauer, B., 1988. Microstructure of mortars from three Egyptian pyramids. Cement and Concrete Research 18 (1) 81-90. Cotrim, H.l., Veiga, M.d.R.r. and de Brito, J., 2008. Freixo palace: Rehabilitation of decorative gypsum plasters. Construction and Building Materials 22 (1) 41-49. Luxan, M.P., Dorrego, F. and Laborde, A., 1995. Ancient gypsum mortars from St. Engracia (Zaragoza, Spain): Characterization. Identification of additives and treatments. Cement and Concrete Research 25 (8) 1755-1765. Livingston, R.A., A. Wolde-Tinsae and Chaturbahai, A., 1991. The Use of Gypsum Mortar in Historic Buildings. In: C. Brebbia, Structural Repair and Maintenance of Historic Buildings, Southampton UK, Adams, J., Kneller, W. and Dollimore, D., 1992. Thermal analysis (TA) of lime- and gypsum-based medieval mortars. Thermochimica Acta 211 (0) 93-106.

[6]

[7]

[8]

[9]

[10] [11] [12]

[13] [14] [15] [16] [17] [18]

[19]

[20]

[21]

[22]

[23]

Weber, J., Gadermayr, N., Bayer, K., Hughes, D., Kozlowski, R., Stillhammerova, M., Ullrich, D. and Vyskocilova, R., 2008. Roman cement mortars in Europe's architectural heritage of the 19th century. ASTM Special Technical Publication, 69-83. Gosselin, C., Verges-Belmin, V., Royer, A. and Martinet, G., 2009. Natural cement and monumental restoration. Materials and Structures/Materiaux et Constructions 42 (6) 749-763. Hughes, D.C., Swann, S. and Gardner, A., 2007. Roman cement: Part one: Its origins and properties. Journal of Architectural Conservation 13 (1) 21-36. Avenier, C., Rosier, B. and Sommain, D., 2007. Ciment naturel. Glnat Werner, D. and Burmeister, K., 2007. An Overview of the History and Economic Geology of the Natural Cement Industry at Rosendale, Ulster County, New York. Journal of ASTM International 4 (6) 14. Dariz, P., 2009. Romanzement in der Schweiz Geschichte des natrlich hydraulischen Bindemittels in der Eidgenossenschaft. Restauro 8 8. Hughes, D.C., Jaglin, D., Kozlowski, R. and Mucha, D., 2009. Roman cements - Belite cements calcined at low temperature. Cement and Concrete Research 39 (2) 77-89. Hughes, D.C., Jaglin, D., Kozlowski, R., Mayr, N., Mucha, D. and Weber, J., 2007. Calcination of marls to produce Roman cement. Journal of ASTM International 4 (1) Pinter, F., personnal communication: Vienna. Campbell, D.H., 1999. Microscopical examination and interpretation of portland cement and clinker. Portland Cement Association Thiery, M., Villain, G., Dangla, P. and Platret, G., 2007. Investigation of the carbonation front shape on cementitious materials: Effects of the chemical kinetics. Cement and Concrete Research 37 (7) 1047-1058. Vyskocilova, R., Schwarz, W., Mucha, D., Hughes, D., Kozlowski, R. and Weber, J., 2008. Hydration processes in pastes of Roman and American natural cements. ASTM Special Technical Publication, 96-104. Shtepenko, O., Hills, C., Brough, A. and Thomas, M., 2006. The effect of carbon dioxide on -dicalcium silicate and Portland cement. Chemical Engineering Journal 118 (1-2) 107-118. Gosselin, C., Scrivener, K.L. and Feldman, S.B., 2011. Microstructure of roman cements used for architectural restoration. International Conference on Cement Chemistry, Madrid, Klisinska-Kopacz, A., Tislova, R., Adamski, G. and Kozlowski, R., 2010. Pore structure of historic and repair Roman cement mortars to establish their compatibility. Journal of Cultural Heritage 11 (4) 404-410. Bayer, K., Gosselin, C., Hilbert, G. and Weber, J., 2011. Microstructure of historic and modern Roman cements to understand their specific properties. In: T.K. Alenka Mauko, Tinkara Kopar, Nina Gartner, 13th Euroseminar On Microscopy Applied To Building Materials, Ljubljana, 2-3.

IDENTIFICATION OF 19th CENTURY ROMAN CEMENTS BY THE PHASE COMPOSITION OF CLINKER RESIDUES IN HISTORIC MORTARS Nina Gadermayr1, Farkas Pintr2, Johannes Weber1
1

University of Applied Arts Vienna, Institute of Arts and Technology, Section of Conservation Sciences A-1013 Vienna, Salzgries 14/1, Austria 2 Federal Office for the Protection of Monuments, Scientific Laboratory, A-1030 Vienna, Arsenal 15/4, Austria

Abstract The term cement in its historic context stands for several groups of highly hydraulic binders which may differ in the temperatures of calcination and hence yield mortars of significantly differing properties. A key to identify the exact type of binder used in a historic cement mortar consists in the textural and mineralogical features of unhydrated residual clinker particles. Employing various techniques of light and scanning electron microscopy allows define fingerprints for given cements, a task hardly addressed in earlier studies on natural cements calcined at low temperature. The present contribution focuses on the most characteristic residual clinker phases found in historic Roman cement mortars, a group of materials widely used in the 19th century urban architecture and civil engineering. These binders were produced from natural marlstones through shaft kiln calcination at temperatures virtually below sintering. High amounts of non-crystalline reactive compounds form in this low temperature regime. Along with fine-grained impure C2S, they have developed into a relatively homogeneous hydrated matrix. In parallel, however, non- or low reactive compounds have assembled together in binder-related nodules characteristic for Roman cement mortars. Most of these compounds are of non-crystalline nature, they comprise solid solution silicate phases as well as crystalline CS (wollastonite), coarse C2S and C2AS (gehlenite). By means of thin section and reflected light microscopy, combined with scanning electron microscopy and X-ray microanalysis, the phase assemblages can be observed and identified in virtually all of the historic mortars. Various classes of residues were distinguished within the above general frames. According to their texture and mineral content, they are classified into overred or underred in respect to the optimum temperature of calcination. Optimum clinker assemblages with high amounts of reactive phases may form nodules as well, characterised by their specifically dense hydrate structure. Following a general description of the above mentioned classes, the contribution presents a group of samples from historic faades in the city of Budapest as an example of the usefulness of the classification in the practice of mortar analysis. The data suggest that different brands of Roman cements characterised by specific setting times may have existed and used on purpose for specific mortar applications. The paper aims at providing information useful for the identification and characterisation of historic natural mortars. Keywords: Roman cement, natural cement, phenograins, clinker microscopy, SEM

Introduction Natural cements manufactured by the low temperature calcination of marlstones fine-grained carbonate rocks with significant amounts of clay and other silicates were amongst the most important building materials of the 19th century in Europe. Almost a century after their recession in the first half of the 20th century and following a period of neglect, these binders, frequently referred to as Roman cements (RCs), are now attracting new interest by the restoration community. This development was accompanied by a number of scientific and technical studies on the composition and properties of the cements, pastes and mortars. Amongst other activities, the EU-funded projects ROCEM (2003-06) and ROCARE (2009-12) have provided detailed information which can be viewed e.g. in the project website www.rocare.eu. A number of handbooks and other texts edited in the 19th and early 20th centuries deal with the technical aspects of the Roman cement technology, e.g. Pasley (1830), Tarnawski (1887), Tetmajer (1893), Schoch (1904), Eckel (1905), Bohnagen (1914), Khl & Knothe (1915). Analyses of the chemical composition of RCs, descriptions of the process of production, and definitions of their key properties can be found in these sources. Historic standards took account of the typically short setting time of Roman cements, which were less than 15 minutes for many brands. Evolvement of the final strength is generally delayed in comparison to Portland cements (PCs), so that the 28days strength of the former is of not too much significance for its quality. Concerning the temperature of calcination, this was established by trial calcinations yielding binders which match the historic ones by all means (Hughes et al., 2007, 2009). Thus, the temperature needed to produce optimum Roman cements from an appropriate raw feed averages at or even below about 900 C. At this temperature, the mineralogical composition of a RC analysed by X-ray diffraction is characterised by a specific phase assemblage. According to Hughes et al. (2010, quoted verbatim in the following), optimal cements are characterised by maximum -belite content, a high content of an amorphous phase and residual calcite and quartz indicating incomplete calcination; Carbonated belite, spurrite, was observed in some cements at low temperatures. As the calcination temperature is increased, the -belite converts to the -belite form observed in natural hydraulic limes and Portland cement; spurrite was also reduced. Additionally, aluminosilicate as gehlenite is observed. These developments are accompanied by a reduction in the amorphous phase and residual calcite and quartz. Brownmillerite was observed in cements with a high iron content. (End of quotation). When dealing with the identification of historic RCs, it must be kept in mind that, being natural cements from different sources, they show a relatively wide spread of their chemical composition, sometimes falling within the much narrower range of PCs. This fact, as well as the well-known problems related in general to the analysis of binders in historical mortar samples, makes it difficult to identify a Roman cement mortar by just the chemical composition of the binder. Also X-ray diffraction can be of just a limited use, since fingerprint products like -belite have reacted away by hydration to C-SH; similar holds for the reactive portion of the amorphous phase, which is anyway not definable by this method. It is thus due to a petrographic analysis based on methods of microscopy to assess the characteristic features of a historic Roman cement mortar, at least in cases where the experienced eye of an expert fails to take an unambiguous decision on the nature of the mortar.

1.

Studying the compounds in the way outlined above means to observe and analyse the residual cement clinker in an otherwise hydrated and frequently carbonated mortar sample, by all their petrographic and mineralogical properties accessible to the light and electron microscope. This approach forms the methodical base of the present study which thus aims to contribute to a better knowledge of characteristic fingerprints for the identification of Roman cement in historic mortars. Beyond this, it will be shown that a more precise classification of the type of RC can be achieved. 2. Methods Following vacuum-impregnation in epoxy resin, the mortar samples were cut perpendicular to the surface and processed to petrographic thin sections. They were then studied under a polarising microscope (PL) in transmitted light, occasionally also in reflective light in the case of polished thin sections. The scanning electron (SEM) studies were performed on polished thin sections, or on their counterpart sections, respectively, employing a back-scattered electron detector (BSE). Usually, the highvacuum working mode was selected with prior carbon coating of the surfaces, though a number of analyses were also made on uncoated samples in the low-vacuum mode. Chemical analyses were performed by an energy-dispersive X-ray analyser (EDS) linked to the SEM. 3. Phenomenology of residual clinker in historic Roman cement mortars The most characteristic common features of all historic RC mortars are the binderrelated nodules, phenograins according to the nomenclature by Diamond & Bonen (1993), which are formed by various types of incompletely hydrated cement particles or densely hydrated agglomerates, respectively. They can be observed in any microscope even at low resolution, though determination of their specific nature needs more precise methods of analysis such as polarising microscopy on thin sections combined with electron microscopy. The abundant presence of such nodules is believed to be due to the particularities of the historic process of production, see e.g. Weber, Gadermayr, Bayer et al. (2007): The raw feed, a rock material with all its natural impurities and inhomogeneities, enters the vertical shaft kiln unground in fist-sized fragments. The temperature gradients in such a kiln are paralleled by gradients within every single lump of stone, so that the resulting clinker is likely to cover a wide range of different grades of calcination. The low temperatures below effective sintering even favour uneven conditions of reaction. Notwithstanding the occasionally reported fact that obviously over- or underburned material was removed manually before the clinker was ground (Khl and Knothe 1915), the final product is a blend of differently reacted cement grains which, as a consequence, vary in respect to their reactivity with water to form hydrates. Given the generally coarse particle size of a historic RC frequently in the range of several hundreds of micrometers up to one millimetre, those clinker particles with non or incomplete hydration remain well visible to the eye of the observer. They bear significant information on the historic process of production as well as on the nature of the raw feed. In earlier studies performed on Roman cements calcined at defined temperatures in the laboratory, a classification of these phenograins was attempted in respect to the strength development of the corresponding pastes (Weber, Gadermayr, Kozowski et al., 2007). Reflecting their respective grades of calcination, the residual

clinker can be classified roughly as either underfired(1), optimally fired(2) or overfired(3), even if the boundaries between the groups are fluid to some extent. Residual cement nodules can be observed even in carbonated Roman cement mortars, as noted by St.John, Poole and Sims (1998) - and classified by their structure. 3.1 Type 1 nodules underfired Underfired nodules calcined to a sub-optimal degree, herewith referred to as type 1 nodules, still show some textural features inherited from the marl (Figures 1a+b): Unreacted silicates such as predominantly quartz, mica and feldspar can be observed embedded in a flaky to fibrous matrix of Ca-Al-silicates. Partial diffusion of Ca and/or K into silicate minerals can be traced by very thin marginal zoning visible by SEM-BSE, which is in general better developed in quartz grains than in feldspars, but can be also observed in coarse micas. It is especially the quartz which serves as indicator for underfired conditions, as it appears angular and with a still intact crystal lattice, showing the typical birefringence visible under crossed polars in PL, with no or only very thin diffusion rims of Ca or K to be observed by SEM. Calcite and the matrix clay minerals appear to be the first components to react at elevated temperatures. Calcite is incompletely decomposed, however, and often forms pseudomorphs after fossil shells. These features related to calcite are another significant criterion for sub-optimal calcination. The texture of the clayey matrix, especially in respect to the orientation and porosity of the constituents, forms a third sensitive indicator of the conditions of calcination. Compared to optimally fired state, the matrix of underfired nodules principally differs by its relatively dense structure near the limit of resolution of the SEM.

Figure 1a, b. Typical underfired residual grain (type 1). (a): SEM-BSE image, (b): the same particle in the polarising microscope under crossed polars; a characteristic quartz grain can be seen right from the centre of the micrograph.

3.2 Type 2 nodules optimally fired The common feature of these optimally calcined residues is their relatively high amount of reactive clinker material. Thus, apart from the composition and texture of the unreacted portion, a significant amount of hydration products is visible either as a compact rim of varying thickness up to 150 m around the nodules, or even in the form of nearly completely hydrated nodules which, however, include unreacted phase residues. Such densely hydrated areas may have uncarbonated core areas even in otherwise fully carbonated mortars. According to their characteristic constituents, the

optimally fired nodules can be subdivided into two types; they contain either noncrystalline silicates with clear zoning (2-A), or- belite clusters (2-B), respectively. Residues of the type 2-A (Figures 2a+b, 3a+b, 4a+b) are characterised by nonequilibrium features such as solid solution systems, and by significant zoning due to a partial diffusion of Ca predominantly from carbonates, and K probably from micas into silicates. The most common phases are non-crystalline silica with no birefringence, up to a size of 30 m. Varying amounts of K and Ca are always present, by their diffusion into the silicate they have supposedly contributed to the breakdown of their crystalline lattice. Since K-ions have diffused more easily, this element appears to be concentrated rather in core areas of zoned silicate grains, where eventually it may have caused partial melting visibly by small spherical cavities. Ca, on the other hand, tends to be concentrated rather in the marginal zones, where it may form calcium silicates of stoichiometric composition. Thus, the front of Ca-diffusion into silicates can be observed by a sequence of Ca-silicate minerals with increasing Ca:Si ratios, such as wollastonite CS, rankinite C3S2 and belite C2S. Where no full hydration of the clinker matrix has occurred, it consists of very fine Ca-Al-Si-phases at the limits of resolution. These non-crystalline products of calcination of the clayey portion of the raw feed play a dominant role in the early age hydration of a RC (Tilov 2008). Depending on the specific grade of calcination, the residual matrix appears either as densely fibrous, or as porous and grainy. New formations within the matrix include gehlenite C2AS along with the Ca-silicate minerals mentioned above.

Figure 2a, b. Characteristic densely hydrated nodule of the 2-A type. (a): SEM-BSE image, note the shrinkage cracks, (b): the same particle in the polarising microscope under crossed polars.

Figure 3a, b. Type 2-A typical optimally fired phenograin; (a) SEM-BSE of the whole nodule, (b) detail of (a) with wollastonite CS and belite C2S grown around silicate grains.

Figure 4a, b. SEM-BSE images of a optimally fired clinker nodule, (b): detail of the same clinker residue: crystallisation of C2S at the costs of zoned silicate grains this grain could be classified as transition between types 2-A and 2-B: Belite C2S is growing around silicates and in form of smaller particles in a more flaky matrix.

Type 2-B-residues (Figure 5a+b) are characterised by distinctive clusters of relatively coarse belite C2S, associated with minor amounts of phases like C3S2, CS, C2AS, and occasionally Ca-Fe-Ti-phases. These belite crystals of a size up to 8 m show lamellar structures and equilibrium features such as triple borders. They may contain small amounts of Mg and were obviously formed by a full solid state reaction with the silicates. The crystals are frequently arranged around open spaces probably formed by local melting. As compared to similar voids found in type 2-A clinker nodules, the voids of 2-B tend to be bigger, i.e. up to 80 m in diameter, and their margins are defined by euhedral clinker crystals. The coarse nature of belites prevents them from efficient hydration. Thus, usually just a dense rim of hydrates can be found around such clusters.

Figure 5a, b. SEM-BSE micrographs of a remnant of type 2-B: clusters of belite. (a): note the dense rim of hydration; (b): detail of (a), with euhedral belites (light grey) and rankinite, C3S2 (dark grey).

3.3 Type 3 nodules overfired Overfired super-optimal particles (Figures 6a-d) show no or just weak signs of hydraulic activity. They are easily recognised by their very angular and often shard-like structure. Marginal hydration of type 3 nodules can never be observed, though in the case of small particle size their hydration can be assumed.

Due to different cooling rates in the kiln during the manufacturing process, two forms of overfired particles can be observed in RC: Angular shaped and coarse crystalline (type 3-A) or shard-like and glassy relicts (type 3-B) with air voids. Abundant compounds of the coarse crystalline nodules (type 3-A, Figure 6a+b) are gehlenite C2AS, belite C2S, wollastonite CS, leucite S2AK, and various phases of the CaO-SiO2-Al2O3-K2O-FeO-TiO2 system. The resulting compounds are not only welldefined stoichiometrically, but also in terms of their euhedral crystal habits. Thus, it can be assumed that they were formed under equilibrium conditions. Glassy nodules (type 3-B, Figures 6c+d) show various stoichiometries in the Si-AlK-Ca system, though in general they contain significant amounts of aluminium. In the isotropic glass, the presence of small microlithes indicates devitrification. Overfired nodules probably formed in local areas of higher temperatures in the traditional kilns seem to play an important role as inert filler, also controlling the rapid setting of RC to some extent (Hughes et al., 2010). Generally, this mixture of various residues is consistent with results of laboratory calcinations of marls, where optimal, sub- and super-optimal calcined cements show similar microstructures (Hughes et al., 2010).

Figure 6a-d. SEM-BSE micrographs of overfired phenograins. (a, b): Type 3-A is characterised by its coarse-crystalline structure. (b): detail of the same nodule: C2S (medium grey), wollastonite CS, a mineral of the composition C2S2(A,M) - probably melilite (medium to dark grey), kalsilite KAS (dark), Ca-Ti-Fe-mineral (bright). (c, d): Type 3-B: Si-Al-K-rich glass with small microlithes and bubbles; details of the nodule in (d) the darker parts are richer in Si.

4.

Roman cement clinker residuals in historic mortars from Budapest

Samples collected from seven 19th century apartment houses in Budapest (Figure 6) were selected to check the above described typology of RC clinkers in the course of full mortar analyses. It was hoped that some kind of correlation between the predominant type of clinker nodules and the way of architectural application could be established. A list of samples is given in Table 1. Approximately 100 phenograins were analysed and statistically evaluated for each of the samples. They were classified into one of the three major phenograin types 1, 2 or 3 described in Section 3 of this study. In all of the samples, the optimally fired cement grains revealed to dominate (55 to 66%). The amount of under- and overfired nodules varies from 8 to 28% and from 15 to 32%, respectively (Table 1). The samples can be divided into two groups based on the ratio of under- to overfired particles in a mortar (Figure 7): one group, comprising all render samples, contains higher amounts of overfired nodules; the other one, representing all pre-cast elements, shows higher amounts of underfired phenograins. Based on the statistical evaluation of the observations, the following assumptions can be made. .The high amount of optimally fired cement grains in all samples is likely to indicate a pre-selection of optimally calcined clinker material before the grinding procedure. Furthermore, as suggested by Weber, Gadermayr, Kozowski et al. (2007), the overfired portion of a Roman cement clinker would hydrate more slowly and thus contribute to a prolonged setting time of a mortar. Given the generally short times of workability of Roman cements, retardation is a prerequisite for on-site render works. Therefore, the higher amount of overfired cement grains in the three in-situ run samples suggest slower setting times of theseRoman cement mortars. In contrast to the above, in the other four samples the predominance of optimally and underfired clinkers suggests a faster speed of set of the mortars (Weber, Gadermayr, Kozowski et al., 2007). Since these mortars were also used for render works where prolonged setting times were required, another method of retardation must have been used. A simple method recorded by historic sources was e.g. the storage of Roman cement clinker in the open air for some days. Under these conditions, a certain portion of highly reactive and amorphous Ca-aluminate phases of the clinker would react with the moisture of the air, thus withdrawn from the early hydration responsible for the mortar set. Consequently, a slower speed of set and better workability could have been achieved.
Table 1. Mortar samples taken from 19th century building faades in Budapest with quantitative classification of the phenograins.

Sample ROC-1 ROC-2 ROC-3 ROC-4 ROC-5 ROC-6 ROC-7

Type of sampled faade element In-situ run RC mortar, door framing RC mortar from diamond-shaped element RC mortar from diamond-shaped projection In-situ run RC mortar, window framing Mortar, quoin element, In-situ run RC mortar profile Mortar, RC-based artificial stone column

Figure 6. Characteristic apartment house from 1896 in downtown Budapest.

Figure 7. Quantitative evaluation of phenograins in historic RC mortar samples from Budapest

Conclusions and discussion The selected approach to study the clinker residues in historic mortars arose from the fact that neither chemical nor X-ray diffraction analyses could yield sufficient information to identify natural cements of the Roman cement type versus e.g. Portland cements or highly hydraulic limes. Another objective was related to the search for optimum conditions of calcination in order to reproduce Roman cements matching the historic binders in their properties and composition. To that end, the residual clinkers in historic mortars needed to be studied in comparison with the new products. In general, the variety of phases and textures present in Roman cement phenograins within even one sample is striking. The basic phenomenology and classification of the most frequent types of nodules found in Roman cement mortars, as presented in this contribution, is based on observations made for a considerable number of samples from many European sources. One should therefore keep in mind that the described phenomena are just indicative in a way that significant deviations can occur for certain Roman cements from specific regions. To give just two examples, clinkers found in RC brands produced from dolomitic marl, e.g., could not be considered in this contribution, for the simple fact that so far the authors have not come across them frequently enough. Some historic RC brands from England, manufactured from septaria by using coal as a fuel rather than timber, show distinct features related to higher temperatures and sometimes higher sulphate contents inherited from the stone. The related clinker assemblages were not included in this study. The classification into different types of phenograins reflecting various grades of calcination forms a promising approach to gain more insight into mechanisms of hydration on the one hand, on the other hand it will enable improved differentiation between RC brands once a more systematic assessment will be made. It is hoped that increasing numbers of scientist will use the approach presented to publish comparable observations and data on RC mortars from all over Europe, so that our knowledge on the range of composition within this family of binders could grow.

5.

Acknowledgements The present study is largely based on research activities funded by the Commission of the European Union within the EU research projects ROCEM (2003-06) and ROCARE (2009-12).

6.

References

Bohnagen, A. 1914. Der Stukkateur und Gipser. Reprint Verlag Leipzig Diamond, S. and Bonen, D. 1993. Microstructure of Hardened Cement Paste A new Interpretation, Journal of the America Ceramic Society, 76(12): 2993-99. Eckel, E. C. (1905): Cements, Limes and Plasters 1st ed. New York: John Wiley & Sons Inc. Hughes, D., Jaglin, D., Kozlowski, R. et al. 2007. Calcination of marl to produce Roman cement. Journal of ASTM International, 4(1), Paper ID JAI100661. Hughes, D.C., Jaglin, D., Kozowski R., et al. 2009. Roman cements - Belite cements calcined at low temperature. Cement and Concrete Research 39, 7789. Hughes, D., Weber, J., Kozlowski, R. 2010. Roman Cement for the Production of Conservation Mortars. Preprints 2nd Historic Mortars Conference & Rilem TC 203RHM Repair Mortars for Historic Masonry Final Workshop, Prague, 22-24.September 2010. Khl, H. and Knothe, W. 1915. Die Chemie der hydraulischen Bindemittel Wesen und Herstellung der hydraulischen Bindemittel. Leipzig: Verlag v. S. Hirzel. St. John, D. A., Poole, A. W., Sims, I. 1998. Concrete petrography: A handbook of investigative techniques London etc.: Arnold. Pasley, C. W. 1830. Observations, deduced from experiment, upon the natural water cements of England, and on the artificial cements, that may be used as substitutes for them. Printed by authority, at the Establishment for Field Instruction. Schoch, C. 1904. Die moderne Aufbereitung der Mrtel-Materialien. 2. Aufl., Berlin: Verlag der Thonindustrie-Zeitung. Tarnawski, A. 1887. Kalk, Gyps, Cementkalk und Portland-Cement in sterreichUngarn. Wien: Selbstverlag. Tetmajer, L. 1893. Methoden und Resultate der Prfung der Hydraulischen Bindemittel. Mitteilungen der Anstalt zur Prfung von Baumaterialien am eidgen. Polytechnikum Zrich. 6. Heft, Zrich: Selbst-Verlag der eidgen. Festigkeits-Anstalt. Tilov, R. 2008. Hydration of natural cements. Ph.D. dissertation. Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences. Weber, J., Gadermayr, N., Bayer, K. et al. 2007. Roman cement mortars in Europes architectural heritage of the 19th century. Journal of ASTM International, 4(8), Paper ID JAI100667. Weber, J., Gadermayr, N., Kozowski, R. et al. 2007. Microstructure and mineral composition of Roman cements produced at defined calcination conditions. Material Characterization 58, 1217-1228.

DOLOMITIC LIME MORTAR AND THE IMPACT OF SO2-POLLUTION Anja Diekamp1, Jrgen Konzett2, Patricia Tartarotti3, Peter W. Mirwald2 1 Material Technology Innsbruck (MTI), University of Innsbruck, Technikerstrae 13, 6020 Innsbruck, Austria; anja.diekamp@uibk.ac.at 2 Institute of Mineralogy and Petrography, University of Innsbruck, Innrain 52, 6020 Innsbruck, Austria; 3 Bauforschung Tirol, Murstrae 7a, 6020 Rum bei Innsbruck, Austria

Abstract The impact of SO2-pollution on medieval dolomitic lime mortars from a church near Bolzano, Italy (St. Martin/Kampill), two Castles in the Lower Inn Valley in Tyrol, Austria (Aschach Castle and Matzen Castle) and a stately home in an urban environment in Innsbruck (Angerzell mansion), Austria, was studied. The samples investigated are Romanesque to Baroque mortars and renders, sgraffito-plaster from a Renaissance faade and intonacoplaster from Gothic interior frescos. Their binders essentially consist of calcite [CaCO3], magnesite [MgCO3] and/or hydromagnesite [4MgCO3 Mg(OH)2 4H2O]. These phases are typical for dolomitic lime mortar which was often used in the Eastern Alpine Region during the medieval period. Although dolomitic lime mortars are extremely durable under adverse climatic conditions, they are very susceptible to the damaging effects of Mg-sulfate formation. In fact the investigated samples are often loaded with gypsum [CaSO4 2H2O] and Mg-Sulfates [hexahydrite MgSO4 6H2O, epsomite MgSO4 7H2O]. In case of St. Martin, and Angerzell Mansion modern time (i.e. past !150 years) SO2-pollution is thought to be the only significant source of sulfur. For the Lower Inn Valley, however, SO2-pollution from smelting of copper-silver ores during the 15th and 16th century is an important additional sulfur source. Contrary to what is commonly assumed petrographic and microchemical analysis of selected mortar samples shows that the Mg-phases of the binder (most likely magnesite) are less prone to decomposition and sulfate-formation than is calcite. This finding challenges current concepts for damage assessment and the development of restoration strategies for sulfate-contaminated dolomitic lime mortars and magnesian lime stones as well. Keywords: dolomitic lime, magnesian lime, Mg-Sulfates. 1. Introduction

The effect of magnesium sulfate formation due to direct reaction of SO2 with magnesian limestone or dolomitic limestone is well known. (Balboni et al. 2011; Doehne & Pinchin 2008; Lpez-Arce et al. 2008, 2009, 2011; Odgers et al. 2008; Ruiz-Agudo et al. 2011). In a more indirect way even Mg-carbonate free building stones may be damaged by Mg-sulfate migration from sulfur-loaded dolomitc lime joint mortar. This indirect mechanism may cause alveolization (Siedel 2008, 2010). By contrast, the mechanisms of sulfate formation in the mortars themselves have not been investigated so far in any detail. All that is known at present is that under exposure of SO2 calcite will react to form gypsum and the Mg-Phases (e. g. magnesite) will form Mg-sulfates, in most instances found as epsomite or hexahydrate (Cultrone et al. 2008; Siedel 2000 and 2003).

There are three principle effects that cause damage during Mg-sulfate formation, recrystallization and dissolution: (1) the volume-increase associated with the formation of Mg-sulfate from a precursor Mg-phase of the binder (e. g. magnesite) (Steiger et al. 2008); (2) the volume change associated with the epsomite-hexahydrite phase transition. (Steiger et al. 2011 and references). Both effects can create high crystallization pressures with attendant fracturing and disaggregation of the binder. (3) After their crystallization, all Mg-sulfates are easily leached due to their very high water solubility (710 g/l for epsomite and 660 g/l for hexahydrite at 20C) which causes additional loss of substance from the mortar surfaces (Siedel, 2003). In a dolomitic lime mortar both calcite and the Mg-phases compete for the available sulfur and, hence, their combined effects are crucial for any ensuing damage pattern. In this paper we present the results of a microchemical study to evaluate the effects of Ca- and Mg-sulfate formation on decay patterns of dolomitic lime mortar and -plaster surfaces. 2. Materials and Methods

From each investigated building plaster and mortar samples have been collected representing the main construction periods (Romanesque to Baroque), in most cases accurately dated by art historians. In this paper we will focus on the construction phases in which magnesian/dolomitic lime mortar was used and show significant patters of decay. 2.1 Objects, environmental situation and affected construction phases Church St. Martin in Kampill near Bolzano, South Tyrol, Italia Environment: In the mid thirties an industrial zone was founded in Bolzano causing SO2 emissions and in 1960 a freeway was opened passing directly next to the church. The church is free-standing without adjacent buildings. Mortars/Plasters: Romanesque joint mortar of the tower (dated ~ 1303 A.D.): only in lower areas, which are exposed to rain water a clearly visibly sanding of the joint border/margin can be observed (Fig. 1). On the higher parts of the tower there are also areas less exposed to weathering, where lime painting is preserved on the Mortar surface. Baroque plaster of the registry (constructed ~ 1610 A.D.): according to the orientation of the faades a sanding of the plaster results in different stages of surface-loss. In areas protected from the rain under the eaves salt efflorescence is visible and results to scaling of the plaster. Intonaco-plaster from Gothic interior frescos (~ 1403 A.D.): salt crystallization on frescosurfaces and damages are only found in areas which where exposed to humidity from the ground or outer shale (constructional crack) during a longer time period. Aschach Castle and Matzen Castle in the Lower Inn Valley, Tyrol, Austria Environment: Both Castles are located on exposed positions on the southern side of the Lower Inn Valley. In the Lower Inn Valley SO2-pollution from smelting of copper-silver ores during the 15th and 16th century must have been extreme. E. g. different historical gypsum layers are documented on architectural surfaces of a Baroque Church investigated by Paschinger (2004). In the 70ies of the last century a freeway was constructed in the Inn Valley with a high traffic load (feeder road to on of the most frequented crossings of the Alps).

Plaster: At Aschach Castle a remarkable Renaissance faade with sgraffito-plaster suffered from SO2pollution. The elaborate faade was created around 1580 and 100 years later reworked with a lime paint. In Figures 3 and 4 patterns of decay (sanding) at the surfaces of a cant-bay show a significant correlation with the orientation of the parts of the faade. Gothique surfaces of Matzen Castle show in areas sheltered from trees and vegetation salt efflorescence with an attendant blistering and scaling (Fig. 6). Angerzell Mansion in Innsbruck, Tyrol, Austria Environment: The stately home is located in very narrow built up old tow structures in an urban environment with recent pollution. Although located western of the medieval mining districts in the Lower Inn Valley (towards the main wind direction) a medieval SO2-Pollution can not been excluded. Plaster: Whereas Gothic (natural hydraulic lime binder) and Baroque (lime binder) architectural surfaces are in an excellent state of preservation (only with a minor load of gypsum on form from crusts), Renaissance Plasters (~ 1600 A.D.) show patterns of decay from slightly sanding (Fig. 2) up to alveolization (Fig. 5) in correlation with the orientation of the parts of the faade. 2.2 Analytical methods X-ray diffraction analysis (XRD): was performed on crushed and sieved samples (sieve fraction < 0.063 mm with aggregates removed) in an attempt to characterize the mineralogical composition of the binder, using a SIEMENS D500 powder diffractometer with BraggBrentano optics. Thermogravimetric analysis (DTA/TG): thermogravimetric and calorimetric analyses were performed on powder samples with a simultaneous DTA/TG instrument (SETARAM SETSYS evolution 2400) using powdered sample material placed in a corundum crucible. A heating rate of 10 C min1 in a helium atmosphere was applied. In order to correlate the weight loss observed during heating of the sample due to release of H2O and/or CO2 as result of carbonate and hydroxyl phase breakdown, a Quadrupol mass spectrometer was coupled with the DTA/TG device. In addition to XRD and DTA/TG the mineralogy and textures of a representative selection of the samples were investigated using polished thin sections and a petrographic microscope (Olympus BX40). Identification and quantitative analysis of the mineral phases were then carried out by electron microprobe analysis (EMPA, JEOL JXA 8100) in both energy and wavelength dispersive analytical modes. Analytical conditions of 15 kV accelerating voltage and 10 or 5 nA beam current were used with spot sizes between 3 and 50 m to minimize beam damage of the analyzed phases. To study element distributions wavelength dispersive X-ray mapping was performed. In addition to quantitative and mapping analysis, the backscattered electron (BSE) imaging mode was used to study textures and assemblages on a micrometer scale.

Figure 1: Church St. Martin, tower, N-facade: gypsum crust on (darker areas) and slight sanding in more exposed regions (brighter areas) of Romanesque pointing mortar

Figure 2: Angerzell mansion, Innsbruck, S-facade: destruction of Renaissance plaster with fresco painting by sanding

Figure 3: Aschach Castle, edge of NW-facing cant bay with direction dependent decay pattern of Renaissance sgratto Plaster.

Figure 4: Detail of Fig. 3 (white frame): sanding of Renaissance sgratto Plaster with loss of the original surface on mm-scale.

Figure 5: Angerzell mansion, Innsbruck, S-facade: alveolization of Renaissance Plaster

Figure 6: Matzen Castle, E-facade: Blistering with attended scaling of surface caused by dynamic crystallization of magnesium sulfate on Gothique plaster

Figure 7: Romanesque mortar, St. Martin, Kampill, (thin section photomicrograph, parallel polarizers, eld of view 2.8 mm) with a ~0.30.5 mm thick layer of gypsum at the surface (to the right of dashed line).

Figure 8: Baroque plaster, St. Martin, Kampill, (thin section photomicrograph, parallel polarizers, eld of view 2.8 mm) with a ~1.25mm thick deteriorated zone at the surface, cutting through a lime lump (marked with dashed lines).

Figure 9: Detail of the marked lime lump in Fig. 8 (thin section photomicrograph, parallel polarizers, eld of view 1.13 mm); deteriorated zone to the right of dashed line.

Figure 10: same area as in Fig. 9 (BSE-image).

Mg

Ca

S
0 max

Figure 11: same area as in Figs. 9 and 10, distribution of magnesium (Mg), calcium (Ca) and sulfur (S). Note that Ca and S are correlated which indicates the presence of gypsum. There is no correlation between Mg and S: existing Mg-sulfates are leached out during sample preparation.

3.

Results

3.1 Binder-composition of mortars and plasters All damaged mortars and plasters contain magnesian or dolomitic lime binders. They essentially consist of calcite [CaCO3], with varying amounts of magnesite [MgCO3] and/or hydromagnesite [4MgCO3 Mg(OH)2 4H2O]. These phases are typical for dolomitic lime mortar which was often used in the Eastern Alpine Region during the medieval period (Diekamp et al. 2008, 2009). The Romanesque mortar from the Church St. Martin shows in DTA/TG investigations additional a considerably hydraulic character and hydraulic acting relicts in the binder matrix in thin sections. Natural hydraulic components are also often found in investigations of medieval mortars from this region (Diekamp et al. 2012). Although dolomitic lime mortars are extremely durable under the adverse Alpine climatic conditions (Diekamp et al op. cit.), in regions with historical or more recent SO2-Pollution a load of sulfur leads to the formation of gypsum and Mg-Sulfates (XRD), resulting in the following patterns of decay. 3.2 Macroscopic and microscopic patterns of decay Four different types of macroscopic decay patterns can be distinguished which lead to different degrees of deterioration; (1) formation of gypsum crusts (Fig. 1); (2) sanding due to removal of massive gypsum crusts (Fig. 1) and/or Mg-sulfate formation (Figs. 2-4); (3) alveolization/honeycomb weathering (Fig. 5) and (4) blistering with attendant scaling due to Mg-sulfate crystallization and/or hexahydrite-epsomite transition. Using the petrographic microscope and the electron microprobe, the following microscopic decay patterns can be observed: (1) gypsum crusts (Fig. 7); (2) Sulfate formation fronts progressing into the mortars on a mm-scale. The distribution of Mg, Ca and S (Fig. 11) shows that gypsum fronts have progressed deeper into the mortars than the Mg-sulfate fronts. Along these fronts the mineralogy of the binder is changed into a mixture of gypsum and Mgsulfates (Fig. 8). 4. Discussion

4.1 Formation of Ca- and Mg-sulfate: solubility versus dissolution rate The amount of sulfates formed in a mortar through reaction between SO2 and precursor carbonates per time unit is dependent upon the amount of reactants and the time necessary to make available these reactants. The amount is a function of the solubility which is either diffusion-controlled in case of highly soluble compounds (e. g. sulfates) or surface controlled for compounds with a low solubility (e. g. Carbonates). The time required to make available the reactants, on the other hand, is dependent upon the dissolution rate (Morse & Arvidson 2002, Berner 1978). The solubility of magnesite in water in the range 2025C is significantly higher than that of calcite (0.106 g/l vs. 0.014 g/l) (Siedel 2003 and references). The dissolution rate of magnesite, on the other hand, is three orders of magnitude lower than that of calcite at 25C in a wide range of pH-values (Chou et al. 1989, Prokovsky & Schott 1999). For several mortar samples elemental Ca-Mg-S distribution maps show that gypsum formation fronts have penetrated deeper into the mortars than the corresponding Mg-sulfate reaction fronts. This

strongly indicates that dissolution rate is the crucial factor controlling the advancement of sulfate formation fronts within ! 2 mm from the mortar surface. We propose that sulfateinvolved weathering of dolomitic lime mortars starts with gypsum formation and that Mgsulfates form in a more advanced stage of exposure due to the low dissolution rate of magnesite. In any case this requires long periods of high humidity. By comparison wet/dry cycles of short duration favor gypsum formation. Under favorable circumstances this initial gypsum formation can seal the mortar surface and suppress any further Mg-sulfate crystallization. Hence in addition to the solubility-dissolution rate interplay, climatic variations on the scale of a single building will also influence the intensity of sulfate formation and ensuing patterns of decay. 4.2 Surface damaging gypsum versus Mg-sulfate It has to be emphasized that the presence of gypsum as thin layers is not necessarily damaging for reasons outlined in the preceding chapter. It is only if massive layers of gypsum have accumulated over time that surface sanding/scaling may occur due to the different thermomechanical properties of gypsum and mortar (cf. Fig. 1). Unlike gypsum, Mg-sulfates are always damaging due to their high solubility and the volume change associated with phase transitions (see introduction). As is the case for sulfate formation, the intensity of sulfate induced damage (scaling/blistering/sanding/honeycomb weathering) varies according to the building-scale climatic variations. The maximum intensity of Mg-sulfate induced surface damage results from long periods of continuous sulfate growth combined with frequent phase changes between epsomite and hexahydrite. The result of this situation is shown in Figs. 5 and 6. If the Mg-sulfates are frequently removed by dissolution in rain water, the damaging effect is much less pronounced (cf. Figs. 2-4) 5. Summary and Conclusion

If dolomitic lime mortars are used, both gypsum and Mg-sulfates may form by interaction with atmospheric SO2. Gypsum is not necessarily harmful to a plaster surface due to its sealing effect. Mg-sulfates always damage plaster surfaces due to the following effects that may or may not act in concert (1) high crystallization pressure induced by their initial formation; (2) the temperaturehumidity controlled volume increase associated with the hydration of hexahydrite to epsomite; (3) high solubility of Mg-sulfates. Contrary to what has been assumed so far the formation of gypsum and Mg-sulfate is not only controlled by the solubility of calcite and magnesite but also by the dissolution rate of the carbonates. These effects combined favor gypsum formation in an early state of weathering and tend to delay Mg-sulfate formation. The latter invariably requires long periods of humidity. The way the damaging effects associated with Mg-sulfate formation act together is dependent upon the micro climatic conditions on the scale of a single building. Variation of micro climatic conditions may result in a wide range in damaging patterns even on a single faade.

6.

Acknowledgement

This study was carried out within the framework of two linked border crossing projects entitled Historic building materials in Tyrol/Austria and Evaluation of decay at St. Martin in Kampill/Italy. The funding for these projects, provided by the Austrian Federal Ministry for Education, Arts and Culture, the state of Tyrol and the Federal Monument Service of the Autonomous Province of Bolzano South Tyrol is gratefully acknowledged. 7. References

Balboni, E., Espinosa-Marzal, R., Doehne, E. & Scherer, G. (2011): Can drying and rewetting of magnesium sulphate salts lead to damage of stone? Environmental Earth Sciences, 63 (78), 14631473. Berner, R.A. (1978): Equilibrium, kinetics, and the precipitation of magnesian calcite from seawater. American Journal of Science, 278, 14751477. Chou, L., Garrels, R.M. & Wollast, R. (1989): Comparative study of the kinetics and mechanisms of dissolution of carbonate minerals. Chemical Geology, 78, 269282. Cultrone, G., Arizzi, A., Sebastian, E. & Rodriguez-Navarro, C. (2008): Sulfation of calcitic and dolomitic lime mortars in the presence of diesel particulate matter. Environmental Geology 56 (34), 741752. Doehne, E. & Pinchin, S. (2008): Time-lapse macro-imaging in the field: Monitoring rapid flaking of magnesian limestone. In: Lukaszewicz, J., Niemcewicz, P. (eds): In: J.W. Lukaszewicz & P. Niemcewicz (eds.), Proceedings of the 11th International Congress on Deterioration and Conservation of Stone, 1520 September 2008, Torun, Poland, Volume I, 365372. Diekamp, A., Konzett, J., Wertl, W., Tessadri, R. & Mirwald, P.W. (2008): Dolomitic lime mortar a commonly used building material for medieval buildings in Western Austria and Northern Italy. In: J.W. Lukaszewicz & P. Niemcewicz (eds.), Proceedings of the 11th International Congress on Deterioration and Conservation of Stone, 1520 September 2008, Torun, Poland, Volume I, 597604. Diekamp, A., Konzett, J., Mirwald, P.W. (2009): Magnesian lime mortars - identification of magnesium-phases in medieval mortars and plasters with imaging techniques. In: Bernhard Middendorf, Armin Just, Deborah Klein, Andr Glaubitt, Jana Simon (eds.): Proceedings 12th Euroseminar on Microscopy Applied to Building Materials, 15.19.09.2009, Dortmund, Germany, 309317. Diekamp, A., Stalder, R., Konzett, J. & Mirwald, P.W. (2012): Lime Mortar with Natural Hydraulic Components: Characterisation of Reaction Rims with FTIR Imaging in ATRMode. In: Vlek, Jan, Hughes, John J., Groot, Caspar J. W. P. (Eds.): Historic Mortars, Characterisation, Assessment and Repair. RILEM Bookseries, Volume 7, Part 1, 105-113. Lpez-Arce, P., Doehne, E., Martin, W. & Pinchin, S. (2008): Magnesium sulfate salts and historic building materials : experimental simulation of limestone flaking by relative humidity cycling and crystallization of salts. Construction and Building Materials, 58 (289-290), 125 142.

Lpez-Arce, P., Garcia-Guinea, J., Benavente, D., Tormo, L. & Doehne, E. (2009): Deterioration of dolostone by magnesium sulphate salt: An example of incompatible building materials at Bonaval Monastery, Spain. Construction and Building Materials, 23 (2), 846 855. Lpez-Arce, P., Gmez-Villalba, L.S., Martnez-Ramrez, S., lvarez de Buergo, M. & Fort, R. (2011): Preservation strategies for avoidance of salt crystallisation in El Paular Monastery cloister, Madrid, Spain. Environmental Earth Sciences, 63 (7-8), 1487-1509. Morse, J.W. & Arvidson, R.S. (2002): The dissolution kinetics of major sedimentary carbonate minerals. Earth- Science Reviews, 58, 5184. Odgers, D., Pinchin, S., Martin, B., Wood, C., Curteis, T., Doehne, E., Chiari, G., Teutonico, J.M. & Burges. A. (2008): Investigations into Decay Mechanisms of Magnesian Limestone at Chapter House, Howden Minster. In: J.W. Lukaszewicz & P. Niemcewicz (eds.), Proceedings of the 11th International Congress on Deterioration and Conservation of Stone, 1520 September 2008, Torun, Poland, Volume I, 211221. Paschinger, H. (2004): Kalkmrtel und Kalkfarbe in sterreich. In: Kalkmrtel und Kalkfarbe. IFS-Tagung 2004. Institut fr Steinkonservierung, Bericht Nr. 19, 2123. Pokrovsky, O.S. & Schott, J. (1999): Processes at the magnesium-bearing carbonates solution interface: II. Kinetics and mechanism of magnesite dissolution. Geochimica and Cosmochimica Acta, 63 (6), 881897. Ruiz-Agudo, E., Lubelli, B., Sawdy, A., van Hees, R., Price, C. & Rodriguez-Navarro, C. (2011): An integrated methodology for salt damage assessment and remediation: the case of San Jernimo Monastery (Granada, Spain). Environmental Earth Sciences, 63, 7-8, 14751486. Siedel, H. (2000): Effects of salts on wall paintings and rendering in Augustusburg Castle (Saxony). In: Proc. 6th International Congress on Applied Mineralogy, Gttingen Juli 2000, Balkema/Rotterdam, 10351038. Siedel, H. (2003): Dolomitkalkmrtel und Salzbildung an historischer Bausubstanz. In: Heinz Leitner, Steffen Laue, Heiner Siedel (Hrsg.), Mauersalze und Architekturoberflchen. Hochschule fr Bildende Knste, Dresden, 5764. Siedel, H. (2008): Salt-induced alveolar weathering of rhyolite tuff on a building: causes and processes. Proceedings International Conference Salt Weathering on Buildings and Stone Sculptures, Copenhagen, Technical University of Denmark, October 2008, 7988. Siedel, H. (2010): Alveolar weathering of Cretaceous building sandstones on monuments in Saxony, Germany. Geological Society, London, Special Publications 333, 1123. Steiger, M., Linnow, K, Juling, H., Glker, G., El Jarad, A., Brggerhoff, S. & Kirchner, D. (2008): Hydration of MgSO4 H2O and Generation of Stress in Porous Materials. Crystal Growth & Design, 8 (1), 336343. Steiger, M, Linnow, K., Erhardt, D. & Rohde, M. (2011): Decomposition reactions of magnesium sulfate hydrates and phase equilibria in the MgSO4H2O and Na+Mg2+Cl SO42H2O systems with implications for Mars. Geochimica et Cosmochimica Acta, 75 (12), 36003620.