Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences

Kraków, 2008

Hydration of natural cements

Hydratace naturálních cement

Renata Tišlová

A PhD thesis prepared under supervision of: Roman Kozowski PhD, DSc

Acknowledgements

Many people and institutions helped me to complete this research. I would like to thank them all. I am especially grateful to my supervisor Roman Kozowski PhD, DSc for his professional support, sharing knowledge and setting an excellent example of enthusiasm for the research. I owe also thanks to my colleagues at the Instittute of Catalysis and

Surface Chemistry, Polish Academy of Sciences in Krakow, Anna Klisiska-Kopacz,

Antonina Kozowska, Grzegorz Adamski, ukasz Bratasz, Dariusz Mucha for their unfailing support as well as to the Faculty of Restoration, the University of Pardubice, for making my work on the thesis possible.

The work described in this thesis has been in part carried out within the EU research project: ROCEM – Roman Cement to restore built heritage effectively (Contract No.

EVK4-CT-2002-00084) within the 5th Framework Programme, Thematic Priority:

Environment and Sustainable Development, Key Action 4: City of Tomorrow and Cultural

Heritage and the national research project „Wdroenie technologii cementu romaskiego do praktycznej konserwacji zabytków“, project of the Sectorial Operational Programme –

Growth of the Competitiveness of Enterprises, No WKP_1/1.4.1/1/2005/8/8/222/2005/U.

Thanks are due to all colleagues working in the ROCEM project particularly from the ICSC PAS, Krakow and the University of Bradford, UK for supplying samples of

Roman cements and data on their composition and properties. Thanks are also due to the

University of Applied Arts, Vienna, Austria for cooperation on the SEM analysis, as well as to Wolfgang Schwarz for his valuable contribution to the X-ray diffraction analysis. I´m also grateful to Edison Coatings Inc., USA and Vicat, France, for supplying samples of natural cements they produce for this study.

2 Finally, it would have been not possible accomplishing this research without support of my family, especially my husband Petr.

3 Table of Contents

Acknowledgements 2 Table of Contents 4 Abbreviations 7 1. Roman cements key materials of the built heritage of the nineteenth/early twentieth centuries 8 1.1 History 9 1.2 Definition 10 1.3 Raw materials and production 13 1.4 Historic Roman cement mortars 18 1.5 Conservation problems 27 2. Calcination of marls to produce Roman cements 30 3. Hydration of Portland cement 30 3.1 Early hydration of OPC 32 3.2 The late period of OPC hydration 36 3.2.1 Composition and structure of C-S-H 37 3.2.2 Morphology of C-S-H 38 3.3 Microstructure of the OPC pastes and mortars 39 4. Study Aims 41 5. Materials investigated 43 5.1 Cements 43 5.2 Cement pastes - design, setting and strength development 48 5.2.1 Setting 49 5.2.2 Compressive strength 50 5.3 Mortars – design and curing conditions 53 5.4 Historic Roman cement mortars 55 6. Experimental methods used 57 6.1 X-ray diffraction of cement materials 57 6.1.1 X-Ray diffraction of cement powders 57 6.1.2 In-situ X-ray diffraction of pastes 58

4 6.2 Mercury intrusion porosimetry 59 6.2.1 Physical basis of the method, its strength and limitations 59 6.2.2 Cement paste drying 60 6.3 Specific surface area 61 6.3.1 The BET equation 62 6.3.2 The t-method 63 6.3.3 The water vapour surface area 64 6.4 Thermal analysis 64 6.5 Scanning electron microscopy (SEM) 64 6.6 Adhesion 66 7. Results and discussion 67 7.1 Hydration during wet-air curing 67 7.1.1 Growth of crystalline hydrates and consumption of the components of original cements in the hydration process as measured by the in-situ XRD 67 7.1.1.1 Initial stage of hydration 67 7.1.1.2 Late stage of hydration 81 7.1.1.3 Hydration of Roman cement mortars 87 7.1.2 The development of specific surface area in Roman cement pastes 90 7.1.2.1 Experimental approach 90 7.1.2.2 Interpretation of the measurement data 93 7.1.2.3 Specific surface area of Roman cement pastes and mortars 95 7.1.3 Thermal analysis 101 7.1.3.1 Identification of hydrated products in Roman cement pastes 101 7.1.3.2 Quantification of the hydrated product content 106 7.1.3.3 The degree of hydration 109 7.1.4 The microstructure of Roman cement pastes by means of SEM-EDX analysis 116 7.1.5 Pore structure of hydrated Roman cements as measured by mercury intrusion porosimetry (MIP) 121 7.1.5.1 Pore structure of Roman cement pastes 121 7.1.5.2 The influence of water content on the pore structure of the pastes 126 7.1.5.3 Porosity structure of Roman cement mortars 128 7.2 Hydration on exposure to real-world environments 129 7.2.1 Hydration of mortars in different curing conditions 129 7.2.2 Porosity of historic mortars 132 7.2.3 The influence of different porous substrates on the hydration of mortars 137

5 7.2.4 The influence of water repellent treatment on hydration 138 7.3 The influence of hydration on the adhesion of Roman cement repair mortars 140 7.3.1 The adhesion of Roman cement repairs and the effect of curing conditions 141 7.3.2 The influence of mortar composition on the adhesion 142 7.3.3 Adhesion of subsequent layers of fresh mortar 143 Conclusions 146 List of Figures 150 List of Tables 156 List of Equations 158 References 159

6 Abbreviations

Cement Chemical Nomenclature:

S = SiO2 A = Al2O3 F = Fe2O3 S = SO3

C = CaO M = MgO H = H2O

C = CO2 K = K2O N = Na20

C2S dicalcium silicate (belite)

C3S tricalcium silicate (alite)

C4AF tetracalciumferroaluminate (ferrite, also denoted as brownmillerite)

C3A tricalciumaluminate (aluminate)

AFm hexagonal calcium ferro-aluminate-hydrates, written in chemical

- - nomenclature [Ca2(Al,Fe)(OH)6]·X·xH2O, where x=(0; 1), X= OH or Cl , ½

2- 2- SO4 or CO3

AFt trigonal calcium ferro-aluminate-hydrates, most commonly ettringite

Ca6Al2(SO4)3(OH)12.26H2O

CH calcium hydroxide (portlandite)

A1 calcium aluminum oxide carbonate hydroxide hydrate

2[Ca2Al(OH)6]·1/2CO3·OH·5.5H2O (C4A C 0.5H12)

A2 calcium aluminum hydroxide hydrate 2[Ca2(Al,Fe)(OH)6]·OH·H2O (C4AH13)

A3 calcium aluminum oxide carbonate hydrate 2[Ca2 (Al,Fe)(OH)6]· CO3 .5 H2O

(C4A C H11)

7 Chapter 1: Roman cements key materials of the built heritage of the nineteenth/early twentieth centuries

The nineteenth and early twentieth centuries were an era of rapid urban expansion and extraordinary building activity in all European cities. The wealth and power of the new social elites were expressed in sumptuous building façades, decorated with rich architectural and ornamental forms imitating the grand styles of past epochs (Figure 1).

While the stylistic costume of the buildings drew upon the past, the technology used in manufacturing the decorative elements was entirely contemporary. The key material was a natural, highly hydraulic binder, known as Roman cement, which was an alternative to the traditional decorative materials such as stone, brick or terracotta. Its main advantages were that it was fast-setting, and featured high early strength, excellent durability, and a beautiful warm yellow-to-brown colour. The above features combined with low cost of production enabled the easy and economic manufacturing of renders and decorative elements on the external facades of buildings. Roman cement was often referred to as the exterior equivalent of gypsum plaster as it offered the same speed of setting and manipulation yet could withstand exterior conditions very effectively. Its unique properties enabled craftsmen to develop a stylistic language of architectural decoration which has determined the aesthetic appearance of central areas in most European cities today.

The advent of functionalism in the twentieth century, favouring the architectural language of simplicity and absence of any decorations, brought a quick decline in the production and use of Roman cement with newer Portland cement dominating the construction market. The renaissance of interest in Roman cement and its use spans the last

8 twenty years, and is a result of the growing interest in European art of the late 19th/early

20th centuries. Attempts have been undertaken to investigate historic Roman cement stuccoes and develop strategies and adequate measures for their conservation. The use of inadequate restoration materials that do not fit the original, and the lack of knowledge on

Roman cement technology were recognized as key conservation problems, which, in turn, have led to extensive research aimed at re-establishing this historic material and technology to conservation practice.

Figure 1: Example of a façade decorated with Roman cement stucco, Trade Academy in Krakow, Poland, 1904-1905, Jan Zawiejski.

The present thesis is an element of this research effort. In its introductory part, information is provided about the historic cements and stuccoes based on historic literature but also on the outcome of the recent plethora of research work on this subject.

1.1 History

The birthplace of Roman cement production was England. The advent of Roman cement era dates back to 1796 when James Parker patented a novel binder which became

9 known as Parker’s or Roman cement (Parker, 1796). In early production, Parker used

Septarian nodules of clayey limestone from costal clay beds in Sheppey and Harwich in the south of England. In mainland Europe, the manufacture of Roman cement was developed after 1840. Natural cements were first produced in the French Alps in 1842 under the name

“Ciment Prompt” (Eckel, 1905, Cornier, 1904 reported in Cailleux et al., 2007). Later, the manufacture of Roman cement expanded to other sub-Alpine regions – Austria, southern

Germany, northern Italy (Eckel, 1905, Callebaut et al., 2001), Switzerland, as well as

Spain (Varas et al., 2005). An important historic centre of Roman cement production was the Austro-Hungarian Empire; many cement factories were established in the Tyrol, in the area west and south of Vienna, in Bohemia and Galicia, today´s southern Poland

(Tarnawski, 1887).

The production of natural cements in the United States started in 1818-1819 with the construction of the Erie Canal in Madison County. American natural cement was patented two years later, in 1821, by Canvass White. Before patenting it, White visited

England and studied the information concerning the materials and methods related to

Parker´s “Roman cement” (Eckel, 1905). In the following years, the cement industry rapidly developed reaching its maximum production between 1870-1880. One of the greatest natural-cement districts was the Rosendale region of eastern New York. Cement rock was discovered there in 1825 and cement production in Rosendale commenced the same year.

1.2 Definition

Roman cements are natural, highly hydraulic binders, produced from clayey limestones containing 15-40% silica, alumina and iron oxides without preliminary mixing and grinding (Eckel, 1905). The simple production process involved crushing the stone to

10 fragments and burning them at comparatively low temperatures (800-1200°C), below the sintering point, which is the same or a little higher than that used for producing pure lime.

In case the geological source had an admixture of magnesium carbonate, the cements contained magnesium compounds after burning. The burnt stones were ground to a fine powder and stored in wooden barrels.

Two terms are used to describe products manufactured by the above procedure:

“natural” and “Roman” cements and the convention of their usage in this paper is as follows: the term “natural cement” is used for a large class of low-energy cementing products obtained by calcining naturally occurring single rock sources below the sintering point and then grinding the burnt material to a fine powder. The natural cements obtained can differ both in performance requirements and applications. The term ‘Roman cement’ is applied to a narrower category of European rapid-setting natural cements as this feature was an important performance requirement, especially for the easy manufacture of stuccoes for building exteriors. This requirement was explicitly specified in the Austrian

Standard of 1878. Therefore, American natural cements cannot be termed ‘Roman cements’ as they are used in masonry mortars and cast-in-place concrete and the ASTM C-

10 standard of 1974 specifies 30 minutes as a minimum setting time. The differences in performance between various historic natural cements can result from differences in geological sources exploited, different chemistry of the burnt materials and different hydration mechanisms and strength development. It should be stressed that, though the term “Roman cement” was introduced historically in England to describe cements produced from calcitic septarian nodules, its later translation into various national languages was used for materials produced from both beach and marl sources.

Natural cements can be placed in the family of hydraulic binders between hydraulic limes and Portland cements. However, they differ from hydraulic limes in that they are not

11 high in free lime and therefore do not slake when water is added. Compared to Portland cement, they have lower burning temperatures, which yield a different chemistry and influence their setting and hardening behaviour. The aesthetic feature is also important: natural cements possess warm yellow-to-brown colour compared to the grey colour of

Portland cement. The Cementation Index (CI) was introduced by Eckel in 1905 to evaluate the hydraulic behaviour of different natural cements. CI is calculated according to the following formula (Eq.1):

CI = (2.8*SiO2 + 1.1*Al2O3 + 0. 7*Fe2O3 ) / (CaO + 1.4* MgO) (Eq.1)

Eckel (1905) states that natural cements are located within the broad range 1.00-

2.00 as given by the Index. He distinguishes three groups of cements, with the best falling between the narrower limits of 1.15-1.60. They are characterized by the incomplete decomposition of calcite; optimal calcination temperatures from the medium temperature region yield cements with rapid setting and high strength. Cements with CI falling into

1.00-1.15 are generally rich in calcium and magnesium oxides. If lower calcination temperatures are applied, a high content of free lime and magnesia is produced. Thus, comparatively higher temperatures should be applied, which yield inferior cements. The third group of cements, characterized by CI 1.60-2.00, is high in clayey and low in lime and magnesia content. Therefore, rather lower calcination temperatures should be recommended, otherwise puzzolana cements are produced.

According to the Austrian standard from 1880, which was modified in 1890,

Roman cements are distinguished according to their setting times (Tarnawski, 1887). Fast- setting cements are those which set within seven minutes from the moment water is added.

A slow variety of Roman cement starts to set after 15 minutes. The setting time and

12 strength for each group of Roman cement mortars specified in the standard is given in

Table 1.

Several specifications introduced between 1896 and 1904 provided characteristics of American natural cements: fineness, specific gravity, setting times as well as strength of the pastes and mortars (Table 2). The specification of 1904 introduced the definition of a natural cement as a material which is produced by “calcination of an argillaceous limestone at a temperature only sufficient to drive off the carbonic-acid gas”. The specification also stipulated that an increase of the cement fineness resulted in a significant improvement of the paste and mortar properties - setting and tensile strength.

Table 1: Setting time and strength for Roman and Portland cement mortars as specified by the Austrian standards of 1880 and 1890.

Age Tensile strength Compressive strength [days] 2 2 [N/mm ] [N/mm ]

Roman cement Portland Roman cement Portland cement cement

fast slow fast slow

setting 15 15 15 15 (min)

7 d 0.4 0.5 1 not specified

28 d 0.8 1 1.5 6 8 15

1.3 Raw materials and production

Suitable rock sources, which were exploited for the production of Roman cements, can be found in different geologic formations: the best known English Roman cements were produced by calcining Septarian nodules from Eocene London clays, or from Jurassic and

Cretaceous formations along the coastlines (Figure 2).

13 Table 2: Fineness, setting time and strength for American natural cements as given by several historic specifications.

Specification Fineness Setting time Tensile strength [N/mm2]

7 d 28 d

1896 90 per cent - 0.5 0.9 (New York State (d<360 m) Canals)

80 per cent (d<254 m)

1901 95 per cent 20 min – 4 hours* 0.6 1.4 (Engineer Corps, (d<300 m) * (determined on the U.S. Army) paste, water-cement ratio (0.3)

85 per cent (d<150 m)

1904 90 per cent 10 min – 0.7-1.4 1.4-2.1 (American Society (d<150m) (30 min - 3 hours) for Testing

Materials) 70 per cent (d<75 m)

In continental Europe, deposits of stratified marls were mined in France, especially in the Jurassic areas of Burgundy and the Cretaceous region near Grenoble. The marls quarried in the Eastern Alps were Jurassic, Cretaceous or Eocene, such as in the Bergamo area in northern Italy, in the Tyrol, the area near Salzburg and in the area west and south of

Vienna (Figure 3). Other important sites of production were situated mainly in the foothills of the Swiss Alps, in Southern Germany, Bohemia and Galicia, today’s southern Poland.

Analysis of several feedstock materials historically used for Roman cement manufacture are given in Table 3 (Hughes et al., 2007A, 2008).

14

Figure 2: Sheppey septaria, UK. Nodules of clayey limestone in clay beds used for the original production of Roman cement.

Figure 3: The quarry of marls, Lilienfeld, Austria. Typical quarry of raw material used for the production of Roman cement in mainland Europe.

The raw materials used for natural cement manufacture in the United States were argillaceous limestones with similar composition to that of European rock sources. One difference was the high percentage of magnesium typical of most American natural-cement rocks; e.g. the cement rock of Rosendale district, New York contains about 25% magnesium carbonate (Eckel, 1905).

The manufacture of Roman cement comprises three processes – the crushing of marlstone, firing and grinding. The calcination was historically carried in various types of

15 kilns; the most common was vertical continuous mixed-feed types. With growing industrialization during the nineteenth century, an increasing number of big factories were running batteries of kilns for the production of Roman cements (Figure 4).

Table 3: Mineralogy of selected source material used for Roman cement production.

Quartz Feldspar Calcite Dolomite Pyrites Clay minerals (%) (%) (%) (%) (%) Lilienfled 9 2 63 3 - (16% illite + (Austria) 7% chlorite) Folwark 4 - 71 1 - (24% (Poland) smectite) Sheppey 19 1 61 - 2 (16% illite + (England) 1% kaolinite)

Figure 4: Batteries of vertical continuous kilns used for the production of American natural cement. Round Top Hydraulic cement company, Washington County, United States.

During calcination, calcium carbonate and clays were decomposed and new phases containing silica, alumina and lime were formed. Opinion on the optimum calcination temperatures for Roman cement production varies in the historic literature. Significantly,

Eckel (1905) states that higher calcination temperatures between 1100-1300oC produce better cements with a higher content of calcium silicate. Lower temperatures result in cements with strength behaviour more controlled by aluminates and ferrites. Mitchell

16 (undated) states that if the heat is too high it produces a complete decarbonation which yields inferior cements. It was also recognized early on that optimal calcination conditions depend on the composition of the rock source. Eckel concluded that more intense calcination must be applied on cements with CI falling below 1.1, when compared to lower temperatures of around 900oC being sufficient for materials with CI of 2.0. The calcination time has been variously reported as being between 30-72 hours (Smeaton, 1837 reported in

Hughes et al. 2007B; Mitchell, undated; Thurston, 1939 reported in Hughes et al., 2007B,

2008); this probably included the heating, residence and cooling times.

Hughes et al. (2007A, 2007B, 2008), who did an extensive programme on calcination of several Roman cement source materials, state, however, that the CI is an insufficient parameter for marl classification and that the optimal calcination conditions are variable depending on many factors especially the source and properties of marls comprising microstructure and clay mineralogy. However, the historic descriptions of quality cements have been confirmed by current research – optimal cements are indeed produced at low temperatures and contain evidence of incomplete calcination through the presence of quartz and calcite. Further details may be found in Weber et al. (2007B).

After burning, the cement clinker was ground to a fine powder. Historically, millstones were used as they were the only available fine grinders. Later, modern grinding machinery was introduced, such as tube or ball mills. The final fineness was an important property of the cements. Therefore, it was defined by historic standards and specifications.

For American natural cements, the grinding was originally required to produce cements fine enough that 95% passed a 50-mesh sieve, later even a greater fineness was required (Table

3). Roman cements produced within the ROCEM project met the requirements of another historic specification - the Austrian norms (1880, 1890) (Tarnawski, 1887) – stipulating that

17 more than 80% of cement particles must pass through the 50-mesh sieve. The historic cements were packed usually into 250 kg barrels or 60 kg sacks and shipped by rail or river.

1.4 Historic Roman cement mortars

With the growing interest in the science-based conservation of the built heritage of the late nineteenth and early twentieth centuries, investigations of composition and properties of the historic Roman cement mortars were initiated. Project RENDEC (1997-

1999) provided information on decorative renders used in Central Europe in the period around 1900 (Rendec, 1999). Callebaut et al. (2000) describe the application of hydraulic mortars as restoration materials on the facade of St. Michael´s Church in Leuven, Belgium during the nineteenth century; Cailleux et al. (2007) characterize the composition and deterioration mechanism of historic mortars based on natural cements used to produce renders and stucco decorations on the facades of buildings in the Rhône-Alpes region.

Pecconi et al. (2005) provide information on petrographical, mineralogical and chemical characteristics of the “artificial stones” used to decorate palaces in Florence in the nineteenth and twentieth century. Varas et al. (2005) report on the use of hydraulic mortars in Spain.

Within the ROCEM project (2003-2006), historic Roman cement mortars collected from buildings dated between 1804-1904 situated across Europe were extensively studied

(Vyskocilova et al., 2007, Weber et al., 2007A, B, Hughes et al., 2007A, B, C, D, 2008,

Tislova et al., 2008, Kisiska et al., 2008). The samples collected were representative of different modes of application – cast ornaments, renders, hand-run elements and thin surface coatings. The investigations revealed the following characteristic features of the materials (Weber et al., 2007B, www.heritage.xtd.pl):

18 Historic mortar composition

Aggregate: The ratio of aggregate to cement varied depending on mortar application; for cast and hand-run mortars the aggregate content was low, typically 20-

25%, for renders and joint mortars it was generally higher, about 40-50%. The aggregate used was usually quarzitic sand; however limestone or other sources of locally available materials were also used. The aggregate grain-size distribution varied significantly; typically, fine-grained aggregate was used for thin surface applications such as finish layers of stucco and hand-run elements or cement-wash coatings. Coarse grained aggregate was used for bulky cast elements or core mortars of hand-made mouldings and renders.

Binder: the microstructure of hydrated Roman cements shows a very fine-grained matrix composed mainly of amorphous components which encapsulate unhydrated remnants of original cement phases. The phases identified in historic mortars by X-Ray diffraction are belite (C2S), the principal crystalline phase of original Roman cements, portlandite (CH), wollastonite (CS), gehlenite (C2AS) or rankinite (C3S2). Calcite is present as a mineral relict or a secondary product. Other minor phases such as anorthite

(CaAl2Si2O8) and leucite (KAlSi2O6) can be detected, which are formed as a result of the reaction within the complex system SiO2-CaO-Al2O3-Fe2O3; gypsum (CaSO4.2H2O), ettringite or monosulphate (C3A.CaSO4.12H2O) are also occasionally present. Gypsum is either formed as a result of the reaction of mortar components with sulphur dioxide present in the polluted air, or is added as a retarder of the setting. Ettringite and monosulphate are formed on the hydration of sulphur-containing cements.

Weber et al. (2007B) found that up to 27% of Roman cement mortars consist of distinct lumps of original cement grains, called “phenograins”, which are embedded in the groundmass. From SEM/EDX studies performed by this author and Varas et al. (2005) the

19 phenograins can be classified into three groups: overfired, well fired and underfired, which is a result of a considerable temperature differences within the kiln. The overfired phenograins comprise coarse belite crystals, wollastonite, gehlenite, brownmillerite (C4AF) and several minor phases such as rankinite, kilchoanite (C3S2), leucite and anorthite.

Exceptionally, as a result of local higher temperatures (“hot spots”) in the kiln, alite (C3S) can be detected (Gödicke-Dettmering and Strübel, 1996, Varas et al., 2005). These overfired grains are mildly reactive and principally act as fillers, which play a positive role of absorbing shrinkage strains.

The second group of phenograins is classified as well-fired. They are generally highly reactive and show a characteristic rim of hydrated products around an unhydrated core. Undecomposed relicts of the raw material – clay remnants, quartz grains and occasional fossil structures can be classified as the “underfired” grains. In historic natural cement mortars, all three groups of grains were identified showing the good adherence within a dense and brittle matrix composed of clusters of needle-shaped C-S-H structures

(Callebaut et al., 2001).

Mortar microstructure: The natural cement mortars form well-hydrated, highly compact and dense structures. On the other hand, they are highly porous materials reaching values of accessible water porosity between 30-40% by volume (Weber et al., 2007B).

Cailleux et al. (2007) found smaller values for historic natural cement concretes, between

15-17%. The porosity values are similar or even higher than those measured for either aerial lime-based mortars or natural hydraulic limes (NHL) and cement mortars respectively. Aerial lime-based mortars show porosities between 16-22% which are approximately similar to those measured for ordinary Portland cement concretes or lime- cement mortars (Mosquera et al., 2006, Lanas et al., 2006). Porosity values of 18-30%

20 were measured by Lanas et al. (2004) for well-matured NHL mortars. However, there is a general agreement that porosity is strongly dependent upon several factors such as mortar composition, especially binder-to-aggregate ratio, water content and grain-size distribution of the aggregate. For hydraulic binders, curing time plays an important role in mortar performance.

Physical and mechanical parameters: Historic Roman cement mortars are characterized by high porosity, water uptake and water vapour permeability. At the same time they show high strength and moduli of elasticity (Table 4) (Weber et al., 2007B, www.heritage.xtd.pl). This characteristic combination of properties makes these materials strong but very brittle and porous. The maximum compressive strength values of 50

N/mm2 were measured for castings which generally have higher binder content and are produced at comparatively lower water-to-cement ratios than the render applications. Less binder and more water, typical in the formulations of renders and joint mortars, reduce strength. The addition of lime also reduces strength but, on the other hand, improves elasticity of the mortars. The fresh mortars prepared with the addition of lime have better workability and prolonged setting.

Similar values for all mechanical and physical properties were reported by

Sommain (2007) for Prompt natural cement mortars. Only limited information could be found for American natural cement mortars. The historic specification from 1904 (Eckel,

1905) provided the tensile strength values for 7 and 28-day-old mortars (Table 5).

Currently, Wathe (2005) reports 13 MPa as an approximate value of compressive strength for historic Rosendale cement mortars.

21 Table 4: Physical and mechanical properties of historic Roman cement mortars.

Historic Compressive Tensile Modulus of Bulk Water- Water Water Water Roman strength strength elasticity density accessible absorption absorption vapour cement [N.mm-2] [N.mm-2] [kN.mm-2] [g.cm-2] porosity by coefficient permeability [g.m-2.h-1/2] -10 mortar [% v/v] immersion 10 [% w/w] [kg.m-2. s-1.Pa-1]

Render 11 0.6 5.4 1.4 ± 0.01 39 27 23 11 (with lime addition)

Render 38 ± 19 1.6 ± 0.8 21 ± 10 1.7 ± 0.04 28 ± 9 18 ± 7 9 ± 4 4

Cast 44 ± 7 2.1 ± .1.5 17 ± 1 1.64 ± 0.02 31 ± 1 19 ± 1 7 ± 0.5 3 element

Table 5: Tensile strength of mortars from American natural cements (Eckel, 1905).

Tensile strength [N/mm2]

Age (days) 7 d 28 d

Mortar 0.7-1.4 1.4-2.1 (cement- standard sand - 1:3, by volume)

Historical applications of Roman cement mortars

Due to their unique properties described above, Roman cement mortars were used in many domains of application: casting decorative elements, rendering, run work and surface applications such as thin finishing layers and coatings.

Decorative cast elements: were massively produced at low cost due to quick setting and satisfactory initial strength of Roman cements. They were more durable in exterior conditions than decorations made of gypsum. They were manufactured by pouring mortar into moulds made of animal glue; their quick setting allowed manipulation within several minutes; the elements could be de-moulded and immediately used or stored in moist conditions. The final strengths of Roman cement mortars developed with time; therefore storage of castings at humid conditions, favouring the hydration progress was essential. On the other hand, the exposure of fresh castings to rain on the façade was sufficient to ensure a slow hydration over a prolonged period of time. Complex cast elements could consist of several parts. Bulky ornaments were reinforced with internal iron armatures and were fixed to the masonry using wrought iron hooks or nails. Most bulky elements were hollow to make them lighter (Figure 5).

As previously described, the mortars used for casting were poorer in aggregate in comparison with renders. The volume ratios of cement to aggregate were 2:1 or 3:1 whereas water-to-cement ratio was about 0.65. A higher amount of water would lower the strength and facilitate the formation of bubbles on the casting interface. However, more water might have been added to produce exact copies of elements of complicated shape or comprising fine details. The cast or hand-run elements were usually produced in two or more layers: the inner part of a casting or core layer of mortars was coarse-grained whereas the final thin coat contained fine aggregate. The stuccoes were frequently

23 finished by the application of a Roman cement wash. Sometimes, several fine-grained surface layers were applied.

Figure 5: Stucco reinforcement with iron wires; the inner part of castings was usually hollow.

The extremely fast setting of Roman cements could make the proper handling of mortars difficult during the casting process. Therefore, small amounts of retarders were added, usually about 0.2% related to the weight of the cement. Historically, sucrose, citric acid or gypsum was used. However, the use of gypsum is not recommended due to its later reaction to ettringite. Another possibility of retarding the setting was exposure of the cement to air for several hours as Roman cements are extremely sensitive to moist air.

Plain renders and hand-run applications: Renders, plain, rusticated or lined out with false joints, were applied to imitate large stone surfaces and details. Hand-run architectural elements such as cornices, window or door framings were important in composing the structure of a façade. The renders were applied as a single render coat on

24 the brick masonry or in several layers where the render coat was followed by the second coat providing a final level structure. The layer thickness varied between 2-50 mm. Due to their minor shrinkage, it was possible to apply Roman cement mortars much thicker than lime mortars. Also hand-run elements had a sandwich structure consisting of a coarse interior core finished with several outer layers. This was a result of repeatedly applying the mortar and passing the profile over it. In France, mortars were typically used for making large building structural components (Cailleux et al., 2007). Big concrete blocks were produced and used as stone blocks to construct buildings. They were mostly prepared by casting in a factory; only rarely was mortar applied directly onto the façade.

American natural cements were originally used in render and joint mortars due to their comparatively slower setting and lower early strength (www.rosendale.org) (Figure 6).

Figure 6: Joint mortars and renders, typical applications of American natural cements.

The mortar design used for renders and works run in-situ was different from that for castings given above. The mortars were characterized by a larger proportion of aggregate; the optimum cement-to-aggregate ratio was 1:1.5 by volume, for thin coats this was even lower, 1:1. The aggregate particle distribution varied depending on the application; coarse aggregate was used in core mortars, finer ones for finish layers. The

25 optimal water-to-cement ratio was the same as for the cast elements or higher to attain a good consistency of the mortar. A higher content of water was required especially for surface or wash layers. To extend the workable time the addition of a retarder was recommended, especially in cases when large areas of renders or in-situ elements were prepared. To extend the workable time to at least 30 minutes a much higher concentration of a retarder had to be used, varying between 0.3-0.5% (related to cement).

Surface applications – cement-wash coatings: In most cases, Roman cement castings and renders were originally left unpainted. Their typical ochre colour, ranging from pinkish-brown to dark-brown, resembled the colour of other building materials like stone, brick or terracotta.

a) b)

Figure 7: a) Cross section of Roman cement stucco covered with several later coatings of paint, b) surface layer on Roman cement stucco altered to 3 mm depth.

A surface layer can often be seen on Roman cement stuccoes (Figure 7b), which can be the result of a surface treatment with drying oils – a common practice in the nineteenth century. The treatment improved the aesthetic appearance of the façade and its

26 resistance against water. With time, due to the gradual soiling of the surfaces, Roman cement stucco and facades were usually painted over (Figure 7a).

1.5 Conservation problems

Generally, most historic Roman cement mortars have survived more than a century in a very good state of preservation. An irregular network of fine shrinkage cracks is a distinct feature of Roman cement stuccoes. The cracks were formed during drying of renders and architectural castings exposed to external conditions (Figure 8a). Only rarely do they widen if the stucco surfaces are exposed to the severe impact of heavy rain, especially on the top of buildings. The fine surface cracking usually does not lead to damage.

Exposure to excessive dampness is a frequent cause of the destruction of the stuccoes. In the upper parts of the facades the source of dampness is the infiltration of rain water. In the area above ground, the infiltration of water is due to capillary rise. Both can lead to damage of the stuccoes due to water freezing or the transmission and crystallization of salts. Another problem can be the corrosion of iron elements embedded in the Roman cement castings (Figure 8b).

Surfaces of the renders or decorative castings exposed to a polluted urban environment are frequently covered with soiling crusts containing gypsum or other salts.

Cailleux et al., (2007) noted the presence of syngenite (K2Ca(SO4)2.2H2O) on the external surface of the natural cement concrete exposed to the urban environment and thenardite

(Na2SO4) in the subsurface zones. Such salts are generated by the reaction between sulphur dioxide contained in the polluted air and alkali ions present in the cement material. The crystallization of ettringite (3CaO.Al2O3.3CaSO4.32H2O), which is formed during the hydration of cements containing sulphur, can be extremely damaging. Sulphur

27 can be present in raw materials from which cements were originally manufactured or originates from gypsum added to retard the setting of mortars. The crystallization of salts can cause the gradual degradation of the surface and subsurface layers and loss of the original material. In extreme cases, the external compact surface is completely lost and the aggregate grains can be seen (Figure 8c).

a)

c) b)

Figure 8: Deterioration phenomena for Roman cement mortars: a) fine shrinkage cracks characteristic for Roman cement mortars, b) corrosion of iron elements causing the disintegration of the host mortar, c) erosion of Roman cement mortar surface.

Serious conservation problems are usually connected with later repairs and renovation interventions which irreversibly alter the original stucco surface. Years of neglect, the accumulation of repairs, cement coatings and paint layers adversely affect the state of preservation and appearance of Roman cement stuccoes. The use of Portland cement mortars or synthetic resins as binders in the repair materials makes them

28 incompatible and enhances the damage. Original renders and decorative castings have often been removed when in poor condition rather then conserved. Once removed or replaced with improper materials, information on the original technology, skills of craftsmen and past aesthetic concepts was lost. These misplaced interventions have caused the aesthetic degradation of a substantial part of the built heritage of the nineteenth and early twentieth centuries (Figures 9a, b). Therefore, unaltered facades preserving their original colour and architectural surface are very rare, in spite of the fact that Roman cement technology was widely used during the period of urban growth in the nineteenth and early twentieth centuries across the whole of Europe.

b)

Figure 9: Incompatible repairs of Roman cement mortars: a) the improper use of Portland cement, which causes the damage of Roman cement stuccoes, b) Roman cement casting disfigured by a thick coating of paint layers. a)

29 Chapter 2: Calcination of marls to produce Roman cements

The historic literature concerning the Roman cement manufacture was reviewed in

Chapter 1.3. The extensive calcination programme for several kind of raw materials from

Austria, Poland and the United Kingdom was carried out by Hughes (2007A, B, 2008). As the cements produced were investigated in this study, their characteristic mineralogy, which is the product of several solid-state reactions taking place during the burning process, is presented in detail in Chapter 5.

Chapter 3: Hydration of Portland cement

The main aim of this thesis has been to gain insight into a detailed mechanism of the hydration processes that occur in natural cements, especially Roman cements. The general approach is to set off this mechanism by showing the similarities and differences with the extensively studied hydration process of ordinary Portland cement (OPC). The main aspects of setting, strength gain and development of microstructure of OPC are briefly reviewed in this chapter.

Many studies have been conducted on the composition and hydration of ordinary

Portland cement (e.g. Bensted, 1983, Taylor, 1990, Beaudoin and Ramachandran, 1992,

Nonat and Muttin, 1992A, B, Bonen and Diamond, 1994, Schwarz et al., 1994, Kuzel,

1996, Diamond, 1998, Bensted, 2002, Odler, 2000, Famy et al., 2003, Lothenbach and

30 Winnefeld, 2006) and the microstructure of hardened pastes and mortars (e.g. Diamond,

1998, Richardson, 1999, Jennings, 2000, Head and Buenfeld, 2006).

The main differences between OPC and other hydraulic binders are well known.

OPC differs in the mineral composition resulting from its different manufacture. The OPC clinkers are produced at temperatures above 1300°C, at which sintering of the clinker is observed, thus higher than those commonly used for Roman cement or hydraulic lime production. The principal constituents of OPC are calcium silicates, aluminates and ferrites. In contrast to Roman cements, tricalcium silicate (alite) - abbreviated as C3S, is the major calcium silicate phase and makes up 50-70% of the OPC clinkers. Dicalcium silicate C2S (belite) usually constitutes 15-30% of the cement clinker; it can occur in several polymorphic modifications, however, C2S is predominantly formed. Calcium aluminate phases are tricalcium aluminate (C3A) and ferrite (C4AF), also called brownmillerite. Typical amounts of both vary between 5-15%. Minor phases are free lime

(CaO), periclase (MgO) and alkali sulphates. MgO is usually formed from MgCO3 of the original limestone. The presence of glassy components can also be detected.

The hydration of Portland cement comprises reactions of its components with water, in which solid hydrated products are formed. The principal reaction is the formation of calcium silicate hydrate (C-S-H), which accounts for the hardening of

Portland cement pastes and mortars. The minor reactions observed are the hydration of aluminate phases, CaO and MgO, respectively.

The hydration of OPC proceeds in three main phases – from the initial state of a dispersed suspension of cement paste, via setting, to the final state of strengthened system

(Nonat and Mutin, 1992B). The initial state of the paste, which is observed approximately within the first two hours of hydration, is characterized by an increase in the paste viscosity. The increase is related to the coagulation of the newly precipitated hydrated

31 cement particles. After this period, the setting of the cement paste can be observed, accompanied by the evolution of the mechanical properties of the paste. This phase of hydration occurs within 2-8 hours and is accompanied by an accelerated hydrate precipitation and formation of the contacts between the hydrated grains. The paste strength is proportional to the quantity of precipitated hydrates and depends on the w/c ratio – the less water there is in the paste or mortar, the more rigid the paste. The compressive strength of the set paste of w/c=0.4 is about 0.2 MPa.

3.1 Early hydration of OPC

Early hydration reactions comprise processes which occur in the initial stages of hydration up to several hours. The main reactive phases in the initial period are aluminates such as C3A and C4AF, alite and belite are less reactive (Taylor, 1990). Therefore, the dominant hydrated products formed are calcium-aluminate hydrates - AFm (Al2O3-Fe2O3- mono) and AFt (Al2O3-Fe2O3-tri), whereas portlandite (CH) and C-S-H are formed in smaller amounts. A small amount of brucite (magnesium hydroxide) is also detected in cements containing MgO (Taylor, 1990, Odler, 2000, Lothenbach and Winnefeld, 2006).

The main AFm phases have a hexagonal structure and are described by the general formula [Ca2(Al,Fe)(OH)6]·X·xH2O, where x=(0; 1) and X denotes one formula unit of a single charged anion like OH- or Cl-, or half a formula unit of a doubly charged anion like

2- 2- SO4 or CO3 . The AFm phases have a layered structure derived from that of CH by the ordered replacement of one Ca2+ in three by Al3+ or Fe3+ (Taylor, 1990). The principal layers alternate with interlayers containing the X anions, which balance the charge, and

2+ 3+ 3+ H2O molecules. The replacement of Ca by smaller Al or Fe ions distorts the structure of the principal layer, so that alternate Ca2+ ions move in opposite directions from its central plane. This allows each to coordinate the oxygen atom of an interlayer H2O

32 - molecule in addition to its six OH ions. The principal layer, together with the H2O

2+ + molecules thus bonded to the Ca ions, has the composition [Ca2(Al,Fe)(OH)6.2H2O] . In the simpler AFm structures, these units are stacked in such a way as to produce octahedral cavities surrounded by three H2O molecules from each of the adjacent layers. These cavities may contain X anions, H2O molecules, or both. The main AFt phase, initially occurring in the cement pastes and mortars, is ettringite Ca6Al2(SO4)3(OH)12.26H2O, which is formed by the reaction of the aluminate phases and calcium sulphate (Taylor,

1990, Kuzel and Pöllmann, 1991, Kuzel, 1996, Christensen et al., 2004). Other phases, which may be formed by the hydration of aluminate phases, are CAH10 , C2AH8, and

C4AHx (x=7, 11, 13 and 19) (Taylor, 1990, Bensted, 2002, Gartner et al., 2002, Jensen et al., 2005). When carbonate ions are present in the reaction environment, hemicarbonate

C4AC 0.5H12 and monocarbonate C4A C H11 can be formed (Kuzel and Pöllmann, 1991,

Kuzel and Meyer, 1992, Kuzel, 1996). These phases are rarely common in CO2 free pastes

(Kuzel and Meyer, 1992, Jensen et al., 2005). For blended cements containing metakaolin or other highly silicious materials such as fly ash, other phases can develop on hydration; especially gehlenite hydrate (C2ASH8) and hydrogarnet (C3ASH6) are many times reported

(Serry et al., 1984, Silva and Glaser, 1993, Rojas and Sánchez de Rojas, 2003).

Carbonate ions enhance the AFm formation at the expense of ettringite. In the presence of citrates and absence of carbonates, mainly ettringite and only minor amounts of AFm phases are formed. With time, ettringite is reported to convert to monosulfate

Ca4Al2(SO4)(OH)12.6H2O (Kuzel and Meyer, 1992, Kuzel, 1996, Gartner et al., 2002).

However, if sufficient carbonate hydrates are present, this reaction is inhibited (Kuzel,

1996). The list of several calcium aluminate and carbonate hydrate phases formed on hydration of OPC is given in Table 6.

33 Many authors point to a metastable character of the AFm phases (Silva and Glaser,

1990, Silva and Glaser, 1993, Cambrera and Rojas, 2000, Rojas and Sanchez de Rojas,

2003, Jensen et al., 2005). The formation of the metastable hydrated phases, such as

C4AHx (x=7, 11, 13, 19), CAH10, C2AH8, depends upon the reaction temperature and solution composition. The appearance of C4AH13 in the cement pastes is generally caused by the supersaturation of the aqueous phase with respect to calcium hydroxide. High

2+ - concentration of Ca and OH in the pore solution, created by Ca(OH)2 dissolution, enables C4AH13 to precipitate. Silva and Glaser (1993), Christensen et al. (2004) and

Jensen et al. (2005) describe the C4AH13 decomposition. The first authors propose its transformation into hydrogarnet (C3ASH6) under long curing. Another possibility is its transformation into another stable phase, C3AH6, which is reported to be formed also on the transformation of C4AH19 (Jensen et al., 2005). However, Kuzel and Pöllmann (1991) report results which are in contrast with the previous finding, that C4AH19 should be stable during long curing. The same author reports the two-step decomposition of CAH10; the first stage of transformation proceeds via the formation of an amorphous phase consisting presumably of Al(OH)3, C2AH10 and C4AH13 which undergo further decomposition to the ultimately stable C3AH6. An almost identical decomposition reaction sequence is observed for C2AH8. Simultaneously, with the proposed decomposition reactions, the amorphous

Al(OH)3 or very poorly crystalline gibbsite should be formed together with the main hydration products.

The second group of transformation reactions has already been described; it comprises processes which occur in the presence of CO2 (Kuzel and Pöllmann, 1991,

Kuzel and Meyer, 1992). AFm phases are converted into carbonate hydrates; the authors describe the transformation of C4AH19 first into hemicarbonate C4A C 0.5H12, and with increasing CO2 content, into monocarbonate C4A C H11.

34 Table 6: Main hydrated phases detected on OPC hydration.

Formula Composition Interlayer contents Ref. (per formula unit of + Ca2Al(OH)6

- 2- OH CO3 H2O 0 6 C4AH19 2[Ca2Al(OH)6]· 2OH·12H2O 1 Taylor, (1990) C AH 2[Ca Al(OH) ]·2OH·6H O 1 0 3 4 13 2 6 2

C4AH7 2[Ca2Al(OH)6]·2OH 1 0 0

C4A C 0.5H12 2[Ca2Al(OH)6]·0.5CO3 ·OH. 1/2 1/4 11/4 5.5H2O

C4A C 0.5H11.25 2[Ca2Al(OH)6]·0.5CO3 ·OH. 1/2 1/4 19/8 4.75H2O

C4A C 0.5H10.5 2[Ca2Al(OH)6]·0.5CO3 ·OH. 1/2 1/4 2 4.H2O

C4A C 0.5H6.5 2[Ca2Al(OH)6]·0.5CO3 ·OH 1/2 1/4 0

C4A C H11 2[Ca2Al (OH)6]·CO3 ·11H2O 0 1/2 5/2 Fischer and Kuzel, (1982) C4A C H8 2[Ca2Al(OH)6]·CO3 ·3H2O 0 1/2 1

C4A C H6 2[Ca2Al(OH)6]·CO3 0 1/2 0

The authors propose that the transformations proceed probably via the

- - 2- 2- topochemical interlayer exchange 2OH OH .1/2CO3 CO3 . Similar transformation products are formed on hydration of C4AH13 (Rojas and Sáchéz de Rojas, 2003). Pérez

Méndez and Trivio Vázquez (1984) propose the total decomposition of AFm phases into

CaCO3 (aragonite) and aluminum oxide. Similar reactions are observed for pastes containing gypsum. In the pastes without CaCO3, ettringite is formed followed by monosulfate or hemisulfate (Kuzel and Pöllmann, 1991). In the presence of CaCO3, its

35 formation is however suppressed, and hemicarbonate and monocarbonate are formed instead.

3.2 The late period of OPC hydration

The main hydration reaction during the late period is the formation of C-S-H on the dissolution of calcium silicates, especially C3S. The side product of this reaction is calcium hydroxide (CH). -C2S behaves similarly, only less CH is formed and the reaction is slower (Taylor, 1990, Beaudoin and Ramachandran, 1992). About 70% of C3S undergoes hydration in the course of 28 days, compared to 30% of C2S. The C-S-H gel in the cement paste is also formed in the initial period of hydration, however, in negligible amounts only.

The rate of the reaction is controlled by several factors. First, the dissolution of

C3S may be rapid if the dissolved ions are quickly consumed. This occurs in the presence of complex forming components (chelates, citrates) or cations leading to very stable insoluble hydroxides. The dissolution of alite can be retarded by the formation of a protective layer – a diffusion barrier - on the grain surfaces. The layer is assumed to consist of C-S-H precipitated in the initial period (Gartner and Gaidis 1989, Taylor 1990,

Thomas, 2007). A second possibility is that the protective layer is formed by ettringite in the pastes containing gypsum or other AFm phases (Cottin, 1992). The presence of the protective layer on the alite grain surface induces a dormant period during which further

C-S-H formation is retarded. The rate of C-S-H precipitation increases when the suspension of cement pastes is more diluted, e. g. when higher water-to-cement ratio is used (Nonat and Mutin, 1992A, B, Schwarz et al., 1994). With time, C-S-H reacts with

2- atmospheric CO2 or CO3 to form CaCO3. The carbonation brings about the destruction of

36 C-S-H and hydrated silica is formed as the final product (Taylor, 1990, Richardson,

2004).

As already described above, hydrated aluminates containing carbonates, especially hemicarbonate and monocarbonate phases, are formed in the later stages of hydration, as the consequence of a slow dissolution of calcite present in the pastes. At the same time, the amount of ettringite decreases. If MgO is present, it initially precipitates as brucite

(Mg(OH)2) and later converts to hydrotalcite (Lothenbach and Winnefeld, 2006) or Mg is incorporated into the C-S-H phase. Hydrotalcite, due to its variable composition, can be

n- expressed by the general formula Mg1-x(Al, Fe)x(OH)2.[A ]x/n.mH2O, where the structure is composed of positively charged brucite-like layers intercalated with anions [An-] such

- - 2- 2- as OH , Cl , SO4 and CO3 and water molecules. For example, OH-hydrotalcite

Mg4(Al)4(OH)14.3H2O was reported by Wang et al. (2001).

3.2.1 Composition and structure of C-S-H

The composition and structure of calcium silicate hydrate has long been studied

(e.g. Taylor, 1990, Chen et al., 2004, Nonat, 2004, Richardson, 2004). In general, the term

“C-S-H” was implemented to indicate the equilibrium structure in the CaO-SiO2-H2O system, which cannot be expressed by a specific composition. Considerable variations exist in the C-S-H composition, especially in the Ca:Si ratio, for example in different areas of the paste. Lower Ca/Si ratios were observed for C-S-H gel of the “outer products” when compared to higher Ca/Si ratios of the “inner products”. This observation was introduced by Lachowski and Diamond (1983). Other authors (e.g. Taylor, 1950,

Gard and Taylor, 1976, Chen et al., 2004, Richardson, 2004) similarly distinguish two types of C-S-H, described as I and II. Chen et al. (2004) report that C-S-H (I) is normally produced on hydration of C3S and -C2S, whereas a prolonged reaction in an excess of

37 water yields mostly C-S-H (II). The first C-S-H type is characterized by a low Ca/Si ratio and layers elongated in one direction, which result in the fibrous structure. The structure was found to be similar to tobermorite and was first reported by Megaw and Kelsley

(1956). Tobermorite is described by an approximate formula Ca4(Si6O18H2).Ca.4H2O with

Ca/Si ratio of 0.83, but, as suggested by Taylor and Howison (1956), the ratio could be higher. The C-S-H (II) can be imperfectly characterized by a jennite structure with much higher Ca/Si ratio varying between 1.5 to 2.0 (Gard and Taylor, 1976, Taylor, 1986). The tobermorite structure is composed by “dreirketten” - silicate chains, which are formed by the repeated intervals of three silicate tetrahydra with interatomic spacing of the Ca-O sheets. Two of the tetrahydra are “paired” and share two oxygen atoms with central Ca-O sheets, while the third shares only one, forming the “bridging” tetrahedra. The interlayer spacing additionally contains molecules of water and additional Ca2+ ions which balance the negatively charged composite layer. The structure of jennite is also based on the dreierkette silicate chains with Ca-O interlayer sheets. However, every other dreierkette is replaced by OH groups, which are balanced by Ca2+ ions yielding the Ca-OH bonds.

However, C-S-H appears to generally have a disordered structure and varies in the degree of polymerization, and the extent of crystallinity and composition when mixed with other hydrated phases such as CH, AFm or other. Also, when cement paste undergoes carbonation with time, C-S-H is progressively decalcified and the space between C-S-H is filled with microcrystals of calcium carbonate (Famy et al., 2003, Richardson, 2004).

3.2.2 Morphology of C-S-H

C-S-H (I) has a fibrillar structure of a directional morphology controlled by space constraints. If precipitated in large pores, the fibrils formed are coarse, compared to fine

38 fibrils with the smallest dimension of about 3 nm and variable length between a few nanometers to several tens of nanometers, formed in smaller pores (Richardson, 2004).

The second type, C-S-H (II), shows a quite different morphology (Williamson, 1972,

Groves, 1987). It appears like an aggregate of globular particles with particle dimension of about 4-6 nm. The pores within this type of gel are less than 10 nm.

A poorly crystalline or nearly amorphous character of C-S-H phases makes their structural characterization difficult. Techniques that are usually used for studying the microstructure of other materials, e.g. microscopy, surface area and pore-size distribution measurements, or X-ray diffraction, provide only fragmentary information on the C-S-H.

However, new techniques have been successfully used for the study of C-S-H microstructure: nuclear magnetic resonance (NMR), small-angle X-ray scattering (SAXS) and small-angle neutron scattering (SANS). Imaging techniques such as scanning electron microscopy (SEM) and transmission electron microscopy (TEM) combined with energy- dispersive X-ray analysis (EDXA) also provide useful information on the C-S-H composition, other phase distributions and the porosity of the system (e.g. Lachowski et al., 1980, Richardson and Groves, 1993, Richardson, 1999, Jennings, 2000, Richardson,

2004, 2007).

3.3 Microstructure of the OPC pastes and mortars

The structure of Portland cement composites has been extensively studied (Taylor,

1990, Diamond, 1998, Richardson, 1999, Jennings, 2000, Olson and Jennings, 2001,

Odler, 2003, Diamond, 2004, Nonat, 2004). In the cement pastes, two distinctive regions can be seen: the dense hydration products (previously referred to as “outer” and “inner”

C-S-H) and C-S-H groundmass precipitated in space between the cement grains originally filled with water. In young paste, there are residua of only partially hydrated cement

39 particles, between which hydrated groundmass can be observed. Hydrated products formed on the cement grains are seen as thin rims of dense C-S-H gel. As hydration proceeds, the residual grains get significantly smaller or undergo full hydration. With time, a more homogenous, and very compact hydrated cement structure can be observed.

The pore structure of the paste and mortars vary significantly depending on the water-to-cement ratio and degree of hydration (Diamond, 1998). The porosity of concretes is additionally dependent upon the aggregate. The largest pores in the paste matrix, with diameters ranging from 10 to 100 m, originate from air trapped in the concrete. These are well-defined by their characteristic circular shape. The groundmass typically comprises capillary pores, with diameters up to 10 m. With hydration time, the pore sizes diminish slowly due to the formation of the C-S-H gel (Cook and Hover, 1999) and the visual distinction between the hydrated grains and the groundmass is difficult. The porosity of the C-S-H product itself is very fine with pores somewhat under 10 nm in diameter (Richardson, 1999, Jennings, 2000, Jennings et al., 2007).

40 Chapter 4: Study Aims

The principal aim of this study has been to gain better insight into the detailed mechanism of the hydration process that occurs in Roman cements, as better understanding of the chemistry and performance of these materials is a vital condition of their proper application in the conservation of architecture of the late nineteenth and early twentieth centuries. The key concept is compatibility between historic and repair mortars, which is broadly defined as the capacity of the repair mortar to interact with the original historic material without causing technical or aesthetical deteriorations (Arioglu and Acun,

2005, Moropoulou et al., 2005, Lanas et al., 2006, Delgado Rodrigues and Grossi, 2007).

The characteristics of the hydrated cements are expected to vary with exposure of the specimens to different external environments. The lack of moisture in the external conditions may result in restricted hydration and affect the microstructure and strength of the mortars. Quantifying such influence would be of great importance for the application of Roman cement in the conservation of historic mortars as ideal wet-air curing of mortars is rarely possible in the course of practical work on the facades of buildings. Therefore, this research on hydration of Roman cement was divided into two phases. First, the general mechanism of the hydration was established for ideal wet-air curing conditions.

Then, the effect of real-world, drier environments on hydration was investigated.

A further aim of this study was to clarify to which extent the hydration process varies with different application techniques; historically, Roman cements were universal binders used to produce a range of decorative elements on the facades from architectural castings to plain renders, thin finish layers and coatings. The outer thin layers are especially vulnerable to external impacts. Furthermore, different applications require

41 variable mortar composition; in particular, the water content in mortars can vary significantly. Mortars used for castings and thin surface applications require more water than renders and hand-run elements. The impact of the water-to-cement ratio on the progress of hydration was therefore also investigated.

Another practical aspect investigated was the effect of additives which are commonly used in conservation to improve workability, adhesion and water repellent properties of the restoration mortars. Restorers usually tend to apply protective treatments on freshly cleaned and repaired facades. This is done by applying solutions of water- repellent agents in organic solvents or their emulsions in water. There is nearly always pressure to reduce time on site due to the economics of conservation contracts and the minimum curing times for repair mortars, before any further treatment, should be identified with precision. Therefore, the effect of water repellent treatments on the hydration process was also considered.

Finally, historic mortars collected from buildings across Europe were investigated, as their structure and characteristics set criteria for the optimum formulation and application of the materials used in the restoration process which would produce compatible repairs with good adhesion and durability.

42 Chapter 5: Materials investigated

5.1 Cements

Roman cements were produced by burning naturally occurring deposits of limestone rich in clay minerals ground after burning to the required fineness.

Within this study, the hydration mechanism of several Roman cements produced from typical raw materials originating from different sources was studied. Two European marls were studied – one was sourced from Lilienfeld in Austria, which had been the site of historic cement production in the nineteenth century and the other originated from an active quarry in Folwark, Poland. Sepatarian nodules were collected from beach sources at

Sheppey and Harwich in the United Kingdom which, similarly to Lilienfeld, had been used for the production of historic English Roman cements.

Two commercially available products - Prompt (supplied by Vicat Company from

France) and Rosendale cement (supplied by Edison Coatings Inc. in the United States) - were also included in this study. Both are produced from local sources of marls. Prompt natural cement is a product of burning marl material, which comes from deposits located in the Chartreuse Massif in Isere, France (Sommain, 2007). Rosendale cement is produced from “argillaceous limestone” in the New York’s Hudson Valley, which is rich in dolomite, as described by Edison Coatings Inc. (www.rosendalecement.net).

All Roman cements were produced within the EC project ROCEM – “ROman

CEMent to restore built heritage effectively” (www.heritage.xtd.pl) on the laboratory scale at the School of Engineering, Design and Technology, University of Bradford, UK and the

Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences, Krakow. The

43 details of the firing programme, grinding and sieving is given in Hughes et al. (2007A,

2008). The Roman cement samples studied are described by the combination of the temperature and residence time.

The mineralogical composition of Folwark and Lilienfeld marls and that of the septaria from Sheppey and Harwich, as well as their microstructure and morphology, are discussed in detail in Weber et al. (2007B), Hughes et al. (2007A, 2008). The oxide composition and Cementation Index (CI) of the cements investigated are presented in

Table 7. The CI calculated according to Eckel (1905) (Eq.1, Chapter 1.2) was originally introduced to asses the suitability of marls or other rocks for the production of cements and to compare cements of different oxide composition.

It can be seen that the chemical composition of the cements studied conformed with

Eckel´s requirement of Cementation Index for Roman cements which should vary within

1.00-2.00. All cements are similar in their high contents of calcium, silicon and aluminium oxides giving the materials expected hydraulic properties. Higher contents of iron are typical of both English cements and contribute to their brown colour. The Roman cements are all low in sulphur except for Sheppey and Prompt; where the sulphur is incorporated in ye’elimite Ca4(AlO2)6SO4 . The sulphur content is an important feature of the cement as it can lead to the formation of ettringite in the mortars.

The phase composition of the cements studied is given in Table 8. The main crystalline phases are silicates, especially belite (C2S), which exists as two polymorphs -

’ and . The first modification is more stable in samples calcined at lower temperatures when compared to -belite which is produced more abundantly in cements calcined at high calcination temperatures. A detailed discussion of the formation and stabilization of belite is given by Lea (1970), Hong and Young (1999). Aluminates are generally incorporated in the “amorphous” phase which is believed to comprise poorly crystalline

44 aluminates or alumosilicates or de-hydrated clay remnants (Vyskocilova et al., 2007). The crystallization of gehlenite (C2AS) or brownmillerite (C4AF) at the expense of the amorphous phase is also observed under more intense calcination conditions. The brownmillerite formation characteristic of both English cements and the Prompt is a consequence of their higher iron content. In spite of high sulphur content in both Sheppey and Prompt clinkers, the crystalline phase containing sulphur - ye’elimite Ca4(AlO2)6SO4

- was detected only in the Prompt.

The presence of gehlenite, brownmillerite and higher -belite content in the two commercially produced cements indicate their relatively high calcination temperatures.

This observation conforms to data referred to in historic literature (Chapter 1.3) where higher calcination temperatures around 1100-1300°C are recommended for the production of American natural cements (Eckel, 1905). There is a marked difference between

European natural cements and Rosendale cement. The latter generally contains high levels of magnesium oxide (periclase) and high levels of quartz. The magnesium oxide content is similar in both chemical and mineralogical analyses suggesting that the conversion of magnesium on burning to other phases is minimal.

It is evident from the presented results that the degree of calcination and phase composition of cements are more influenced by temperature than residence time. Hughes et al. (2007A, B, 2008) distinguish three categories of Roman cements: optimal which coincides with the highest four-week strength and the most rapid setting, sub-optimal at lower temperatures and super-optimal at higher temperatures. Based on that classification,

Roman cements investigated in this study can be classified as optimal and super-optimal.

Folwark 830°C/650min, Lilienfeld 860°C/300min and 920°C/300min belongs to the optimal cements.

45 Table 7: Chemical composition of cements investigated

Cements investigated (marl sources and calcination conditions) Folwark Lilienfeld Harwich Sheppey Lilienfeld Folwark Prompt Rosendale Temperature [°C] 830 860 890 890 920 960 unknown unknown Time [min] 650 300 500 500 300 300 - -

SiO2 25.8 29.2 24 19.2 29.1 29.9 19.1 31.3

Al2O3 6.6 9.2 7.8 6.3 9.2 7.3 7.1 5.0

Fe2O3 2.0 3.2 8.0 5.5 3.6 2.8 3.3 2.4 CaO 49.4 46.6 43.2 50.7 46.8 51.3 52.8 32.3 MgO 1.7 1.6 2.5 2.3 1.5 1.9 3.5 17.6

SO3 0.8 0.1 1.1 2.7 traces 0.8 3.7 0.9

Na2O+ K2O 1.7 2.9 2.2 1.7 2.8 1.9 2.2 2.5

TiO2 0.2 0.3 1.2 0.4 0.4 0.4 0.4 0.3 LOI 11.4 6.8 6.3 9.0 6.5 3.7 7.7 7.8 Free CaO 6.3 2.4 3.2 4.7 0,2 6.5 5.4 3.8 Insoluble in 7.3 11.6 12.6 7.2 6.5 6.4 4.0 24.3

HCl+Na2CO3 CI (Cem. Index) 1.56 1.93 1.74 1.20 1.92 1.73 1.10 1.67

Table 8: Mineralogical composition of natural cements investigated.

Cements investigated (marl sources and calcination conditions)

Folwark Lilienfeld Harwich Sheppey Lilienfeld Folwark Prompt Rosendale Temperature [°C] 830 860 890 890 920 960 unknown unknown Time [min] 650 300 500 500 300 300 - - Quartz 275244 114 Calcite 20 15 0 0 6 0 14 10 Brownmillerite 00131100 50 Gehlenite 021041545 -belite 8 25 17 21 20 50 20 10 Lime 400000 00 Portlandite 005706 33 '-belite 30 22 33 27 30 9 13 9 Periclase 000300 319 Spurrite 000000 00 Ye'elimite 000000 20 Total belite 38 47 50 48 50 59 23 19 '-belite/ total belite 0.79 0.47 0.66 0.56 0.6 0.15 0.56 0.47 ratio

Amorphous 36 29 26 29 36 16 35 30

Abbreviations: quartz – SiO2, calcite – CaCO3, brownmillerite – Ca2(AlxFe1-x)2O5 (C4AF), gehlenite – Ca2Al2SiO7 (C2AS), belite – Ca2SiO4 (C2S), lime – CaO, portlandite – Ca(OH)2, periclase – MgO, spurrite – Ca5(SiO4)2CO3 (C5S2 C ), ye’elimite – Ca4(AlO2)6SO4 (C4A6S).

As all are produced at relatively low calcination temperatures, they yield a significant content of unreacted calcite, though the formation of belite, especially more reactive ’- belite, is well advanced (Table 8). Minor components are quartz and free lime observed in

Folwark cement.

Sheppey and Harwich 890°C/500min and Folwark 960°C/300min can be described as super-optimal cements. The increased calcination temperature produced a high degree of decarbonation of calcite and the lime formed was subsequently combined into the belites, and also partially into gehlenite or brownmillerite.

5.2 Cement pastes - design, setting and strength development

To study the hydration mechanism of Roman cements, pastes from cements and water were prepared at the water-cement ratio of 0.65. Prismatic specimens of 20x2x2 cm were cast in steel moulds and after setting were immediately de-moulded and cured under specified conditions for periods of up to two years. Samples of the hardened cement pastes were taken after the predetermined curing period and tested. To stop hydration, several methods of drying were used as described in Chapters 6.2.2, 6.3 and 6.4. After drying, the samples were stored in a dessicator until tested.

The water-cement ratio of 0.65 was determined practically as being that which yielded paste of sufficient fluidity to allow the casting of the most rapidly setting pastes into the moulds before the onset of the setting process. This value was also commonly used by conservators preparing mortars. The same water-cement ratio was used also for the Prompt and Rosendale cements to compare their hydration behaviour under the same conditions. However, it should be noticed that due to the slow setting and lower initial strength, the Rosendale cement is normally used at much lower water-cement ratios

(ASTM standard, 1974).

48 5.2.1 Setting

The setting times for the Roman cement pastes of the same composition and curing conditions as used in this study were determined at the School of Engineering, Design and

Technology, University of Bradford, UK. They are shown in Table 9. It is apparent that, in general, the Roman cements set very fast with setting times of approximately 1.5-3 mins without any differences between optimal and super-optimal cements. These values agree well with the information given in historic literature (Chapter 1.2). The only exception is

Harwich 890°C/500min, which has a setting time of 9 minutes. The low setting times do not permit proper handling of these materials. Therefore, the use of retarders is necessary; their addition prolongs the setting significantly as can be seen for Folwark 830°C/650min with 0.4% of citric acid added as a retarder. This dosage level extended the workable time to at least 10 minutes.

Table 9: Setting time of Roman cement pastes investigated, w/c 0.65, CA – citric acid.

Cements investigated (marl sources and calcination conditions)

Folwark Folwark Lilienfeld Harwich Sheppey Lilienfeld Folwark

0.4 CA Temperature [°C] 830 830 860 890 890 920 960 Time [min] 650 650 300 500 500 300 300 Setting time [min] 3 10 3 9 2.5 1.5 1.5

Table 10: Setting time of Prompt and Rosendale natural cement pastes investigated, w/c 0.65.

Rosendale Prompt

USA France Temperature [°C] 890 920 Time [min] 500 300 Setting time [min] > 20 min 2-3

49 According to Sommain (2007), the Prompt cement sets within 2-3 minutes (Table

10), though the materials used within this study set slower at around 10 minutes. Similar values were reported for materials supplied to the market with the addition of a retarder.

The Rosendale natural cements were historically specified as materials with a minimum setting time of 30 minutes (ASTM standard, 1974). The commercial products supplied presently are characterized by a setting time varying between 20 and 60 minutes (Table

10).

5.2.2 Compressive strength

The compressive strength measurements were done at the School of Engineering,

Design and Technology, University of Bradford, UK for the pastes at ages of 6 hours to 6 months. The details of the experimental procedure and the detailed description of the results are provided in Hughes et al. (2007A, 2008). Characteristic profiles of compressive strength development for several pastes studied within this work are shown in Figure 10.

They show the following distinctive features:

Initial strength development occurs immediately after setting – both

Lilienfeld pastes are characterized by early strength of 3-4 MPa making

these cements highly suitable for practical use. The early strength is

especially useful in making decorative castings since it permits an early

removal of the objects from their moulds. In contrast, both Folwark,

Harwich, Sheppey, Prompt and Rosendale cement pastes show lower values

of the initial strength not exceeding 1 MPa.

Dormant period of variable length – characterized by a negligible increase

in compressive strength values. A dormant period varies significantly across

50 Figure 10: Compressive strength development profiles of Roman cement pastes, w/c 0.65.

Table 11: Compressive strength of Prompt cement mortars (Vicat, France).

Mortar (cement/aggregate=1:2 by

volume, w/c=0.65) 7 days 28 days 6 months Compressive 6.5 914 strength [MPa]

the samples; it may last from 1 day for Folwark 830°C/650min to several

weeks; 14 days dormant period was observed for Sheppey and Harwich

890°C/500min, 8 weeks for Lilienfeld 920°C/300min and Folwark

960°C/300min. Prompt cement displays a dormant period of 7 days (Sommain,

2007).

Late-stage strength gain – an increase in the compressive strength can be seen

after the dormant period in all Roman cements reaching the final values of 13-

20 MPa after 6 months of hydration. Lilienfeld pastes are characterized by

higher final strength values when compared to generally weaker Folwark

cements. Additionally, a small decrease in strength is observed in the later

stages of hydration for all Roman cement pastes. This phenomenon has been

described in Justnes et al. (1996) and Pérez Méndez and Trivio Vázquez

(1984), though its explanation requires further studies.

The strength profile of the Prompt cement paste was not determined within this study. However, Sommain (2007) showed the strength profiles for mortars produced at w/c=0.65 and cement-to-aggregate ratio of 1:2 (by volume); the compressive strength reached 14 MPa after 6 months of hydration (Table 11). For historic concretes, the compressive strength values of about 20 MPa were reported by Cailleux (2007). The strength development of the Rosendale cement is initially rather slow but finally reaches values of around 13 MPa, which is comparable with other historic Roman cement mortars

(Wathe, 2005).

The strength profiles for pastes retarded with 0.4% of citric acid are also presented in Figure 10. As one can see, the compressive strength values obtained for the retarded

52 pastes in the initial period up to an age of some 4 weeks are comparable with those of the non-retarded samples. At subsequent ages, substantial enhancement of the strength can be observed.

5.3 Mortars – design and curing conditions

Several mortars were prepared from Folwark 830°C/650min Roman cement by mixing the cement with aggregate and water. Due to the quick setting of this cement citric acid was added as a retarder at 0.3% (relative to dry cement).

The composition of the first set of mortars replicated the formulation of historic mortars used in rendering and casting. The aggregate-cement ratio was 0.2; quartz sand

(d=0.25-0.5 mm) was used as the aggregate. The water-cement ratio was varied between

0.65 and 1.0 as higher water contents could historically be used to produce architectural castings and hand-run work when compared to plain rendering. Mortars were cast into silicone moulds or laid on different substrates (historic Roman cement mortars and fired- clay bricks), de-moulded after setting and stored at 20°C under three different curing conditions for up to 6 months: wet-air curing (the samples were kept over distilled water under near 100% relative humidity), water-curing (the samples were immersed in water) and air-curing (the samples were stored in a climatic chamber at 75% relative humidity).

Two other set of mortars was treated with the water-repellent silicone emulsion Funcosil®

WS and the air-entraining and water-repellent Aida® Porenmörtel-Konzentrat (Remmers, www.remmers.com) to see the effect of such treatment on the hydration process and to define an optimal time required for the mortar curing before application of a water repellent. The emulsion was applied by brush onto the mortar surface (0.5 l/m2) and a good water-repellent effect was obtained after drying as shown in Figure 11. Aida®

Porenmörtel-Konzentrat was dissolved in water (1:20 by volume) prior to application.

53

Figure 11: The effect of water-repellent treatment.

The composition of the second set of mortars replicated the formulation of historic fine finish layers and surface coatings. Very fine quartz sand (d=0.25 mm) was used as the aggregate and the mortars were formulated for two water-cement ratios – 0.7 and 1.0. The mortars were applied by a spatula on different porous substrates with a thickness not exceeding 5 mm (Figure 12a).

a) b)

Figure 12: Roman cement surface applications: a) finish layer, b) and wash.

Finally, a Roman cement wash was prepared by mixing pure cement with water in two water-cement ratios of 0.7 and 1.0 respectively, which allowed the proper application

54 of such material with a brush on the porous substrate. Both mixtures were applied on the surface of the old Roman cement mortars as a 2-3 mm thick layer and cured at the conditions given previously (Figure 12b). Simultaneously, one set of wash samples were deposited on the quartz glass plate to record X-ray diffraction patterns in situ.

Two different substrates were tested similar as for original Roman cement mortars: fired-clay brick and historic well-matured Roman cement stucco. The substrate was pre- wetted in two ways: either it was immersed in water or water was applied with a brush until saturated.

5.4 Historic Roman cement mortars

Several samples of historic Roman cement mortars were included in this study to obtain their pore structure characteristics and compare them to the repair materials. The samples were collected from historic buildings of the 19th and 20th centuries across Europe

(Table 12).

Table 12: Selected historic Roman cement mortars.

Building sampled Element

Trade Academy, (A) cast architectural detail Krakow, Poland (B) plain pre-cast profile (C) run cornice (D) run window framing (E) plain render White Lion House, hollow casting of a lion Eye, Suffolk, UK outer surface inner surface Office building at Liebiggasse 5, Vienna, hollow casting of a console Austria outer layer inner core

55 The samples represent a wide spectrum of Roman cements produced from different local sources of raw materials, which yielded products with a range of composition and properties. The selection covered various modes of Roman cement applications such as cast elements, run in-situ stucco, mortars and surface-coating applications (Chapter 1.4).

56 Chapter 6: Experimental methods used

Wet chemical analyses to determine the oxide composition of the cements studied followed the procedures of the European Standard EN 196-2: 2005. X-ray diffraction

(XRD) was used to determine their mineralogy (Hughes et al., 2007A, 2008 and Weber et al., 2007B).

The in-situ X-ray diffraction of pastes, mercury intrusion porosimetry (MIP), specific surface area measured by nitrogen and water vapour adsorption, and thermal analysis (DTG/TG) were used to characterize the progress of hydration in pastes and mortars. The microstructure of the pastes at different stages of hydration was studied using scanning electron microscopy (SEM).

6.1 X-ray diffraction of cement materials

X-Ray diffraction is a powerful technique for the study of crystalline materials.

This method was used for a quantitative determination of the phase composition of cements and the identification of hydrated phases in pastes. The method provided also valuable information on the polymorphic modifications of the phases and their degree of crystallinity.

6.1.1 X-Ray diffraction of cement powders

The phase composition of cements was determined using a Siemens D5005 X-ray diffractometer, CuK radiation (40kV, 40 mA). The spectra were recorded between 2-70°

2 with a step of 0.04° 2, 4 s/step. The quantitative phase analysis was performed using

57 the Rietvield method, which made possible the calculation of the weight fractions of crystalline phases (De La Torre et al., 2001) from the entire powder diffraction diagram.

The crystal structure of each phase and the equipment characteristics were input parameters in the analysis. Corundum from Merck was added as an internal standard to quantify the amorphous phase content (Mumme et al., 1995, Scrivener et al., 2004,

Gaultieri et al., 2006). TOPAS V2.1 software, BrukerAXS, Karlsruhe (2003) was used to conduct the refinements.

6.1.2 In-situ X-ray diffraction of pastes

In-situ X-ray diffraction of cement pastes was used to follow the consumption of the crystalline constituents of the cements as well as the formation of crystalline products during the hydration process. Immediately after mixing, the paste was poured into a plastic sample holder and covered with a polyethylene foil to prevent the evaporation of water and reaction with carbon dioxide. The X-ray diffraction patterns were recorded between 8-55° 2with a step of 0.04° 2, 3 s/step using the same instrument as for the powder cement samples. The measurements were taken after several predetermined curing periods until six months of hydration. Between measurements, the pastes were stored over distilled water under near 100% relative humidity.

This method was developed by Schwarz et al. (1994) and Schwarz (1995) to study the kinetics of Portland cement hydration. The in-situ measurements did not require the drying of the pastes prior to the XRD measurements and hence allowed the undisturbed tracing of the products of hydration. As already reported (Taylor, 1990, Kuzel and Meyer,

1992) such products can be very sensitive to drying during sample preparation, which can cause uncontrolled dehydration and the disordering of the crystal structures, especially of the AFm and AFt phases. In the in-situ experiments, the peak intensities of each phase in

58 the X-ray diffraction pattern are proportional to its volume in the paste and the course of hydration can be semi-quantitatively evaluated by following changes in these peak intensities with the curing time.

6.2 Mercury intrusion porosimetry

Mercury intrusion porosimetry (MIP) has been widely used in the study of porosity and pore structure characteristics of the cement-based materials: pastes, mortars and concretes (Taylor, 1990, Laskar, 1997, Diamond, 1998, Cook and Hover, 1999, Diamond,

2000, Friás and Cambrera, 2000, Rojas and Sánchez de Rojas, 2005). Knowledge of porosity and pore size distribution allows better understanding of many physical and mechanical properties such as strength, permeability or durability of the cement-based materials (Poon et al., 1997, Mosquera et al., 2006). Therefore, porosity and pore structure can be considered one of the major criterion of compatibility between mortars.

In this study, the pore structure of the paste and mortar samples was determined using a Poremaster mercury intrusion porosimeter from Quantachrome, allowing the study of pore sizes in the range 440-0.0035 m. The measurements were carried out on the fractured samples taken from the specimens after the predetermined hydration period.

Samples were dried according to a procedure described in Chapter 6.2.2. After drying, the samples were stored in a dessicator until tested.

6.2.1 Physical basis of the method, its strength and limitations

The method is based on the well-known Washburn equation (Eq.2) providing the means of calculating the pressure P required to intrude the pore of a given diameter d by mercury with surface tension  and contact angle  

59 d = -4  cos P (Eq.2)

The Washburn equation directly indicates that the method cannot measure very fine pores of diameters below 0.0035 m as they would require high pressures beyond the capacity of the instrument.

The model assumes the cylindrical shape of the pores and their connection to the outer surface of the specimen, i.e. their accessibility to mercury. However, neither assumption is fulfilled in the real cement materials. Diamond (1998, 2000) and Cook and

Hover (1999) conclude that only a small part of the pores open directly to the outer surface. Most of the pores are inside the paste and are reachable only by penetration through a long chain of pore spaces of varying size and shape. Therefore, the intrusion of mercury into the pore system is severely restricted by the initial pressure which causes the breakthrough access to the interior pore space. Due to this so called “ink bottle effect”, the

MIP method generally does not reflect the real pore size distribution in the pastes and the total porosity values determined differ from those determined by immersion of pastes in water (Cook and Hover, 1999).

The mercury porosimetry is capable, however, of precisely determining the minimum throat dimension of the pore network connected to the specimen surface – the

‘critical’ or ‘threshold’ diameter according to Diamond (2000). This parameter allows following the progress of hydration with high reliability as shown by the literature data

(Cook and Hover, 1999) and the findings reported in this work.

6.2.2 Cement paste drying

The MIP testing requires a complete removal of the free water from the pore system. Several methods can be applied – oven drying between 50 and 105°C, freeze-

60 drying, D-drying or solvent replacement drying (Diamond, 1971, Parrot, 1981, Feldman and Beaudoin, 1991, Konecny and Naqvi, 1993, Gallé, 2001). However, the removal of water can lead to desorption and de-saturation which can cause changes in the paste microstructure such as the generation of micro-cracks or the collapse of fine pores. The further effect is the dehydration of the hydrous phases like C-S-H gel, or the AFm and AFt phases (Gallé, 2001).

Acetone drying was selected for this study as one of the “mildest” methods which does not significantly affect the pore structure of cement pastes and mortars. Fractured paste specimens were soaked in acetone for 24 hours immediately after sampling. In order to remove the acetone, the specimens were then dried in a rotary vacuum flask at 20oC for

4 hours. After drying, they were stored in an air-tight dessicator until testing.

6.3 Specific surface area

The specific surface area has been widely accepted as an important parameter correlating with the development of the microstructure of cement pastes during hydration

(Litvan, 1976, Rarick et al. 1996, Juenger and Jennings, 2001, Odler, 2003, Jennings and

Thomas, 2004, Odler, 2004, Thomas et al., 2004). Measuring the parameter complements the mercury intrusion porosimetry, as it gives direct information on the evolution of the entire pore space of the pastes. In this study the specific surface area was determined by the measurements of water vapour and nitrogen adsorption which are known to give different values of the parameter.

The measurement of water vapour adsorption was conducted gravimetrically for the relative pressure range 0-0.98 with the use of a Sartorius vacuum microbalance at the constant temperature 24 ± 0.2oC. For each experiment, about 0.1 g of a dried sample was used (Chapter 6.2.2). The sample was evacuated under vacuum for 3.5 hours at the

61 temperature of the measurement to remove physically adsorbed water. Then portions of water vapour were added until the required relative pressure was attained and the corresponding mass increase of the sample was recorded. After the saturation water vapour pressure had been reached, the desorption branch was measured. On average, 15 points were measured on the adsorption branch of the isotherm and 10 points on the desorption branch, and approximately one week was necessary for the complete measurement.

Nitrogen adsorption was carried out at 77K using a vacuum volumetric equipment

Autosorb-1 from Quantachrome at 77K. The samples were outgassed prior to the measurements.

The water vapour and nitrogen accessible specific surface areas were calculated using the BET equation (Chapter 6.3.1). The water vapour accessible specific surface area was also determined using the t method (Chapter 6.3.2).

6.3.1 The BET equation

The two-parameter BET sorption equation (Brunauer et al., 1940) was used to interpret the adsorption data by expressing the equilibrium gas adsorption as a function of relative pressure (Eq.3):

p 1 C1 p 4 (Eq.3)  4 4 n(p0 p) nm C nm C p0

where: p is the experimental gas partial pressure and p0 its value at saturation, n the amount of gas adsorbed by a gram of the adsorbent at gas relative pressure p/p0, nm the monolayer capacity in the same units as n, C - energy constant related to the difference of free enthalpy of water molecules in the liquid state and in the monolayer.

62 The BET constants nm and C were determined by a least-squares regression of the above relation over a relative pressure range from 0.05-0.3, identified as the applicability interval. The specific surface area was calculated from nm using a cross-sectional area of

0.114 and 0.162 nm2 for water and nitrogen molecules, respectively.

6.3.2 The t-method

The t-method was first proposed by Lippens and deBoer (1965). In this method, the values of relative pressure p/po on the ordinate axis are replaced by those of t, the statistical thickness of the gas layer adsorbed on a suitable non-porous material of known specific surface area (Eq. 4):

V ads tp(/ po ) (Eq.4) SBET

where Vads is the volume of gas adsorbed at gas relative pressure p/p0 and SBET the specific surface area.

The first universal t (p/po) curves for water vapour were published by Hagymassy et al. (1969). Later, t-curves were proposed to be a function of the chemical nature of the adsorbent and therefore many t-curves were measured to serve as standards to materials of similar chemistry. Badmann et al. (1981) used non-porous di- and tri-calcium silicates to determine the statistical thickness of adsorbed water layer between 0.01 – 0.95 p/po and proved their suitability in interpreting the water sorption isotherms of hardened cement pastes. Their t curve was expressed by a simple two-parameter equation (Eq.5):

  tp( / po ) 3.85 1.89ln( ln( p / po )) (Eq.5)

63 6.3.3 The water vapour surface area

The application of water vapour for the measurement of surface area of cement- based materials has recently been reviewed (Thomas et al., 1996, Rarick et al., 1996,

Juenger and Jennings, 2001, Jennings and Thomas, 2004, Odler, 2003). Together with nitrogen it is commonly used for the evaluation of cement paste hydration (Odler, 2003,

Thomas et al., 2004). However, quite different results are obtained between the two adsorbates when applied on cement pastes.

Differences between SH2O and SN2 can be attributed to the ability of the water molecules to penetrate more of the gel porosity than the larger nitrogen molecules

(Powers, 1967, Tislova et al., 2008). Odler (2003) noticed that a significant fraction of pores in materials containing C-S-H are inaccessible even at the maximum partial pressure p/p0=1.0. This is very similar to the “bottle ink” effect already described for mercury intrusion porosimetry; within cement pastes, some pores are connected to the surrounding only with very narrow entrances, which are inaccessible to nitrogen at commonly applied conditions. On the other hand, surface area measurements using water can yield invalid results while water is a fundamental component of CSH and when absorbed, a chemical reaction may occur (Feldman and Sereda, 1970). BETN2 increases when the ability of nitrogen to diffuse is increased; one possibility is via an increase in a measurement temperature (Odler, 2003).

6.4 Thermal analysis

Thermal analysis comprising differential thermal analysis (DTA), thermogravimetry (TG) and differential thermogravimetry (DTG) has been used for the study of cement-based systems for a number of years, mostly to complete the XRD analysis (Odler and Abdul-Maula, 1984, Ubbríaco and Calabrese, 1998, Rojas and

64 Cambrera, 2002). The method is particularly useful in the detection of gel products and tracing the kinetics of the hydration reaction (Mouret et al., 1997, Poon et al., 1999, Lam et al., 2000, Escalante-Garcia, 2003, Rojas and Sánchez de Rojas, 2003).

The TG method is based on the detection of weight loss due to the decomposition of phases present in the hydrated cement systems when fired up to 1000°C. DTA locates the ranges of temperature corresponding to the thermal decomposition of different phases in the cement paste.

In this study, the measurements were carried out using a Simultaneous Thermal

Analyzer STA 409 PC Luxx from Netsch. All experiments were conducted on powdered paste samples by heating them from 30°C up to 1000°C at a heating rate of 10-20 K/min under static air atmosphere conditions.

A special method of freeze-drying the samples was developed for this technique to ensure that free water was eliminated from the pastes without influencing water chemically bound in the hydrous phases (Copeland and Hayes, 1953, Gallé, 2001, Mitchell and

Margeson, 2006). After a pre-determined period of hydration, a fractured paste sample was immersed in liquid nitrogen and then vacuum-dried at the temperature of dry ice (5 x 10-4

Torr).

6.5 Scanning electron microscopy (SEM)

The microstructure of the pastes at different stages of hydration was studied using the back-scattered electron imaging (BSEI) mode of environmental scanning electron microscopy (ESEM). The Philips XL30 apparatus was used (high vacuum, accelerated voltage of 20kV). Fractured specimens of the Roman cement pastes were spattered with gold prior to the measurement.

65 6.6 Adhesion

The adhesion of the repair mortars to the original substrate was measured according to ASTM D4541 using PosiTest® AT (Figure 13). The steel disks of 20 mm in diameter were bonded on the top of the repair mortar with epoxy glue. The adhesion between repair material and Roman cement substrate was characterized by the results of pull-off strength.

Figure 13: PosiTest® AT for the adhesion measurement.

66 Chapter 7: Results and discussion

7.1 Hydration during wet-air curing

7.1.1 Growth of crystalline hydrates and consumption of the components of original cements in the hydration process as measured by the in-situ XRD

The development of the in-situ XRD patterns of selected cement pastes is shown in

Figures 14-27 as a function of their curing time. For comparison, the diffraction patterns of the anhydrous cements are also shown. The patterns of the pastes contain all reflections of crystalline phases present in the original cements with the exception of lime and portlandite, which dissolve in water. Additionally, the formation of hydrated phases and its kinetics can be followed.

In several characteristic cases, the integrated intensities of peaks related to newly- formed hydrated phases are plotted as a function of the hydration time to better show the kinetics of the hydration reaction (Figures 15, 17, 19, 23, 25).

A few examples of the hydration of fast-setting Roman cements with the addition of citric acid used as a retarder are also discussed (Figures 22-25).

7.1.1.1 Initial stage of hydration

The initial stage of hydration comprises processes which occur within the first six

hours, during which the pastes attain their initial strength values. The main hydrated

phases formed during the initial stage are the AFm and AFt phases.

AFm phases

67 The major crystalline phases formed immediately after the mixing of the cement clinker with water are calcium aluminate hydrates – AFm. Their formation can be accounted for by the reaction within the amorphous phase, which contains poorly crystalline calcium aluminates or decomposed clay remnants. The AFm phases are formed in varying amounts depending on the content of their precursors in the cements. The effect can be clearly seen when the hydration of Folwark (Figures 14, 15) and Lilienfeld cements are compared (Figures 18, 19). In the Lilienfeld paste, AFm phases are formed in large quantities whereas the same process is less efficient in Folwark cement. A high content of aluminum oxide in the former combined with an efficient reaction to cementitious phases on firing (Weber et al., 2007B) can account for this difference. The reduction in AFm content can be clearly seen in cements calcined at more intense calcination conditions

(Harwich 890°C/500min, Sheppey 890°C/500min, Lilienfeld 920°C/300min, Folwark

960°C/300min) (Figures 26, 27, 18, 16), which is a consequence of the recrystallization of the reactive calcium aluminates into the non-hydrating phases - gehlenite and brownmillerite (Table 8) .

The AFm phases detected are described by the general formula

- - [Ca2(Al,Fe)(OH)6]·X·xH2O, where X denotes one single-charged anion like OH or Cl or

2- 2- half a formula unit of a doubly-charged anion like SO4 or CO3 (Taylor, 1990). They are characterized by a basal XRD reflection at 0.80 – 0.82 nm. In general, three AFm phases are formed on the hydration of Roman cement in a characteristic sequence. The main AFm phase, formed immediately after mixing the cements with water, is a poorly crystalline calcium aluminum oxide carbonate hydroxide hydrate 2[Ca2Al(OH)6]·1/2CO3·OH·5.5H2O

- 2- in which OH and CO3 serve as X anions in the previously given formula. This phase, denoted in this work as A1, is written in the cement chemical nomenclature as C4A C 0.5H12.

o It shows the strongest diffraction maxima at around 10.78 2 (d006 = 0.82 nm) and 21.66°

68 2 (d0012 = 0.41 nm). Its structure is described by Fischer and Kuzel (1982); as for all AFm phases, the unit cell is based on a hexagonal structure where a=0.577 nm; c= 4.92 nm.

With increasing reaction time two new hydrates appear in the pastes - C4AH13 and

C4A C H11. The first one, denoted as A2 in this study, is calcium aluminum hydroxide hydrate 2[Ca2(Al,Fe)(OH)6]·OH·H2O, showing two characteristic diffraction maxima at

o around 11.26 (d002 = 0.79 nm) and 22.63 2 (d004 = 0.39 nm) respectively. The structure was described first by Butler et al. (1959 reported in Taylor, 1990) as monoclinic with the unit cell parameters a=0.99 nm, b=1.14 nm, c=1.684 nm and ß = 111o. It differs from the

- previous C4A C 0.5H12 phase in the content of OH groups, which are the only X anions in the general formula [Ca2(Al,Fe)(OH)6]·X·xH2O. Due to the close proximity of its diffraction maxima to those of C4A C 0.5H12, the two phases partially overlap and their separation in the XRD pattern is often complicated. C4AH13 has been frequently identified as the main aluminate hydrate phase formed during the hydration of Portland cement

(Plowman and Cambrera, 1984), lime/pozzolana mixtures (Moropoulou et al., 2004) or pozzolana blended Portland cement systems (Saad Morsy et al., 1997, Rojas and

Cambrera, 2002, Rojas and Sánchez de Rojas, 2003). The phase has a metastable character and tends to quickly convert to the carbonate hydrate, as in the presence of CO2 and/or

CaCO3 in the pastes, or to decompose (Silva and Glaser, 1990, Rojas and Sánchez de

Rojas, 2003).

The other phase is calcium aluminum oxide carbonate hydrate 2[Ca2 (Al, Fe)

2- (OH)6]· CO3 .5 H2O, which contains only CO3 anions in the X position. The formation of

- 2- this phase is most likely related to the interlayer exchange of OH by CO3 in the two other

AFm phases formed initially (Taylor, 1990, Kuzel and Pöllmann, 1991, Kuzel and Meyer,

1992, Kuzel, 1996, Péra et al., 1999). This exchange is accompanied by a shift in the position of the diffraction maxima to 11.68 (d001 = 0.76 nm) and 23.51° 2 (d002 = 0.38

69 nm), respectively. The compound of the general formula C4A C H11, is described by

Fischer and Kuzel (1982). The unit cell is based on the triclinic structure with the unit cell parameters a=0.578 nm, b=0.574 nm, c=0.786 nm, =92.60o, ß=101.98o and =120.05o.

Within this study, it is denoted as A3. Table 13 summarizes the information on the three

AFm phases identified in the investigated Roman cement pastes.

Table 13: AFm phases identified in the Roman cement pastes studied.

Formula Composition Interlayer contents (per formula unit of + Ca2Al(OH)6

- 2- OH CO3 H2O

C4A C 0.5H12 2[Ca2Al(OH)6]·0.5CO3 1/2 1/4 11/4 ·OH. 5.5H2O

C4AH13 2[Ca2Al(OH)6]·2OH·6H2O 1 0 3

C4A C H11 2[Ca2Al (OH)6]·CO3 0 1/2 5/2 ·11H2O

The presence of carbonate hydrates is due to the presence of calcite in the pastes or due to the secondary carbonation taking place with increasing hydration time, as evidenced by a rapid increase in the intensity of the main diffraction maximum of calcite at

29.48o 2 he results of the present hydration experiments are supported by previously

2- published studies pointing to the control of the hydration reaction by the CO3 available in the cement paste (Kuzel and Meyer, 1992, Kuzel, 1996, Péra et al., 1999). They describe the same sequence of the reaction – the first phase is hemicarbonate, which slowly converts to monocarbonate with an increasing reaction time due to the interlayer exchange

- 2- of OH groups by CO3 .

70

Figure 14: X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65) at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A2 =

C4AH13, A3 = C4A C H11, B = belite, Q = quartz, C = calcite. Reflection F comes from foil covering the hydrating sample.

Figure 15: Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Folwark Roman cement 830oC/650min (w/c=0.65).

Abbreviations: A1 = C4A C 0.5H12, A2 = C4AH13, A3 = C4A C H11.

71

Figure 16: X-ray diffraction patterns of Folwark Roman cement 960oC/300min and its paste (w/c=0.65) at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A3 =

C4A C H11, P = portlandite, B = belite, Q = quartz, C = calcite, G = gehlenite. Reflection F comes from foil covering the hydrating sample.

Figure 17: Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Folwark Roman cement 960oC/300min (w/c=0.65).

Abbreviations: A1 = C4A C 0.5H12, A3 = C4A C H11.

72

Figure 18: X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65) at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A2 =

C4AH13, A3 = C4A C H11, B = belite, Q = quartz, C = calcite, G = gehlenite, P = portlandite. Reflection F comes from foil covering the hydrating sample.

Figure 19: Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Lilienfeld Roman cement 920oC/300min (w/c=0.65).

Abbreviations: A1 = C4A C 0.5H12, A2 = C4AH13, A3 = C4A C H11.

73 AFt phases - ettringite

Ettringite - calcium aluminum sulfate hydroxide hydrate C6AS 3H32 – is formed in the pastes of cements that originally contain SO3. It manifests itself by its most prominent

o o o maxima at 9.1 2 (d100 = 1.08 nm), 15.8 2 (d110 = 0.62 nm) and 22.9 2 (d-1-14 = 0.43 nm) (Taylor, 1990). A particularly high amount of ettringite is formed on hydration of

Prompt cement (Figure 20) due to its comparatively high content of SO3 incorporated in ye´elimite (Ca4(AlO2)6SO4) (Table 8). Ettringite formed in the initial stages of hydration converts to the AFm phases described above. Similarly, on hydration of Rosendale cement, ettringite is formed immediately after setting (Figure 21). It is the main hydrated phase in the paste and its concentration declines slowly with time. A small amount of ettringite is present in the system even after six months of hydration.

Figure 20: X-ray diffraction pattern of Prompt cement and its paste (w/c=0.65) at increasing curing time. Abbreviations: E = ettringite, A1 = C4A C 0.5H12, A2 = C4AH13, Y = ye´limite, B = belite, Br = brownmillerite, Q = quartz, C = calcite, L = lime. Reflection F comes from foil covering the hydrating sample.

74

Figure 21: X-ray diffraction pattern of Rosendale cement and its paste (w/c=0.65) at increasing curing time. Abbreviations: E = ettringite, A2 = C4AH13, P = portlandite, B = belite, Q = quartz, C = calcite, L = lime, Pe = periclase. Reflection F comes from foil covering the hydrating sample.

Inert phases

Two inert phases that did not take part in the hydration were detected in the Roman cement pastes - brownmillerite (C4AF) and gehlenite (C2AS). Both phases were present in the cements fired at more intense calcination conditions. Brownmillerite was present in a high amount in two English cements and in a lesser amount in Prompt and its concentration remained unchanged during six months of hydration. The same was observed for gehlenite, present in Lilienfeld 920°C/300min and Harwich and Sheppey

890°C/500min. Periclase (MgO), present in Rosendale cements burned from dolomitic sources, was another example of the inert phase.

75 Effect of citric acid on the early stage of hydration

Admixture of citric acid as a retarder in the pastes of fast-setting Roman cements proves to be very important for their practical use in conservation practice. The retarders extend workable time significantly, which enables a proper handling and application of the pastes and mortars. In the case of the retarded Folwark 830°C/650min paste, the addition of 0.4% of citric acid by weight of dry cement prolonged the setting to 10 mins when compared to the non-retarded paste (Table 9). At the same time, the use of retarder has only negligible effect on the strength within the first 4 weeks of hydration followed by subsequent enhancement of the strength (Figure 10).

The effect of retardation is due to a significant blocking of the precipitation of the

AFm phases. Though the AFm phases could be detected in the XRD patterns after just 15 minutes of hydration of the retarded samples (Figure 22-25), they were formed at a much lower rate for the entire hydration period. The observation was interpreted by assuming that citrate ions acted as a chelating agent due to their preferential coordination on the

Ca2+, Al3+ or Fe3+ centres. By this reaction, very stable chelate complexes of iron (II or III) hydroxo-citrates with aluminum are formed, which affect the nucleation and further crystal growth of AFm or AFt (Schwarz et al., 1994, Ozkul, 2000, Cody et al., 2004, Rai et al.,

2004). It was further observed in this study that the addition of citric acid does not affect the general pattern of the formation of the AFm phases as the same hydrated products are formed in retarded and non-retarded systems.

Surprisingly, the retarder addition stabilizes the ettringite formed when compared to non-retarded samples. It is noted in the literature that the nucleation of ettringite may be affected in the presence of citrate anions, and a more stable mineral such as calcium aluminate monosulfate may be formed instead (Schwarz et al., 1994). However, no diffraction maxima related to monosulphate were detected.

76

Figure 22: X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65), modified by the addition of 0.4% of citric acid, at increasing curing time.

Abbreviations: E = ettringite, A1 = C4A C 0.5H12, A3 = C4A C H11, P = portlandite, B = belite, Q = quartz, C = calcite. Reflection F comes from foil covering the hydrating sample.

Figure 23: Integrated intensity of the main diffraction maxima of the hydrated phases as a function of the curing time – paste of Folwark Roman cement 830oC/650min (w/c=0.65), modified by the addition of 0.4% of citric acid. Abbreviations: A1 = C4A C 0.5H12, A3 =

C4A C H11, E = ettringite.

77

Figure 24: X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65), modified by the addition of 0.4% of citric acid, at increasing curing time.

Abbreviations: A1 = C4A C 0.5H12, A3 = C4A C H11, E = ettringite, P = portlandite, B = belite, Q = quartz, C = calcite. Reflection F comes from foil covering the hydrating sample.

Figure 25: Integrated intensity of the main diffraction maxima of hydrated phases as a function of the curing time – paste of Lilienfeld Roman cement 920oC/300min (w/c=0.65), modified by the addition of 0.4% of citric acid. Abbreviations: A1 =

C4A C 0.5H12, A3 = C4A C H11, E = ettringite.

78 Correlation between the formation of AFm phases and early strength of pastes

The analysis of the growth patterns of AFm phases allows the association of them with the development of the paste’s early strength only; they do not seem to affect further changes in strength. Of all the cements studied, the A1 phase is formed immediately and most abundantly in Lilienfeld cement pastes (Figure 18), which is in agreement with a higher aluminum content in the amorphous phase (Table 7, 8) of these cements, especially the 920°C/300min sample. This observation correlates with a high early strength of

Lilienfeld cement paste attaining 4 MPa (Figure 10). Comparatively less of the A1 phase is initially formed in the paste of the optimally fired Folwark 830°C/650min (Figure 14, 15) and in the cement pastes of both English cements Harwich and Sheppey 890°C/500min

(Figure 26, 27) correlating with their lower initial strength not exceeding 1 MPa.

Figure 26: X-ray diffraction patterns of Harwich cement 890oC/500min and its paste

(w/c=0.65) at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A3 = C4A C H11, E = ettringite, P = portlandite, B = belite, Q = quartz, C = calcite. Reflection F comes from foil covering the hydrating sample.

79

Figure 27: X-ray diffraction patterns of Sheppey cement 890oC/500min and its paste

(w/c=0.65) at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A3 = C4A C H11, P = portlandite, B = belite, Q = quartz, C = calcite. Reflection F comes from foil covering the hydrating sample.

Very little AFm phases are also initially formed in the intensely calcined Folwark

960°C/300min (Figure 16) and their increase with hydration time is lower reflecting a lack of active amorphous aluminates in the cement which could contribute to the increase in strength.

Little AFm phases are also formed on the hydration of the Prompt and Rosendale natural cements (Figure 20, 21). In Rosendale cement, a very low amount of AFm can be correlated with its different chemical composition and nature of the clinker phases.

Finally, it should be stated that the structural transformation and gradual amorphisation of the AFm phases initially formed do not affect the late-stage strength of the pastes (Figure 10). It is evident that the strength of the pastes remains unchanged

80 though the dominating diffraction maxima of these phases slowly diminish with increasing reaction time. This is very well illustrated by the transformation of the A1 hydrate formed in large quantities in the Lilienfeld 920°C/300min paste.

7.1.1.2 Late stage of hydration

Reactions in the late stage of hydration account for properties that are characteristic of Roman cements such as high strength and excellent durability. It is these properties which account for the good preservation of Roman cement mortars over hundreds of years.

The dominant reaction is hydration of belite (C2S) resulting in the formation of amorphous calcium silicate hydrates – the C-S-H gel. Two further processes are observed – the disappearance of crystalline calcium aluminate hydrates and the carbonation of the paste matrix, progressing form the surface of the specimens.

Formation of C-S-H gel due to the hydration of belites

It is generally known that, in cements, C-S-H is produced by the reaction of calcium silicate phases with water. In Roman cements, belite (C2S) is the only phase which can react to C-S-H in contrast to Portland cements, where C-S-H is predominantly produced from alite (C3S) (Chapter 3). Due to the amorphous nature of the C-S-H gel, its formation can only be detected indirectly, by following the consumption of C2S. Belite is present in Roman cements as two polymorphs - ’ and  and their contents depend on the calcination conditions as shown in Chapter 5.1. However, it is impossible to distinguish the two varieties in the in-situ XRD of pastes as their strongest diffraction maxima are similarly located at around 32-34° 2and they can also overlap with other crystalline phases such as brownmillerite or calcite. -C2S has only one separate strong diffraction maximum at around 41.2° 2. Therefore, it is difficult to follow the consumption of each

81 crystalline modification during hydration, i.e. to determine which variety is more active in the hydration process. any studies which have compared the reactivity of the two phases have yielded conflicting results (Skalny and Young, 1980, Fukuda et al., 2001). It is apparent in the present study that ´-C2S predominantly contained in the Folwark

830°C/650min (Figure 28), already starts to react after 24 hours of hydration when compared to -C2S predominating in the Folwark 960°C/300min cement (Figure 29), which has a dormant period of at least four weeks. However, this is not uniformly consistent across all the cements produced from different sources as other factors influencing belite reactivity must be taken into account. Ishida et al. (1993) note that high activity is more influenced by the specific surface area and crystal size than the nature of the polymorphs. Cottin (1992) explains the slow reaction to the C-S-H gel by the formation of a protective layer on the surface of anhydrous calcium silicate particles inhibiting further hydration reactions, which eventually leads to dormant periods of varying length. This protective layer can be formed by the AFm or AFt phases as it is described for the C3S hydration in Portland cements. The length of the dormant period is expected to depend on the stability of this protective layer, which is controlled by the physical and chemical properties of the silicate substrate particles. Both arguments can explain the high reactivity of cements fired under ‘mild’ conditions as being due to the presence of poorly-crystalline, highly-dispersed belite by coincidence present mainly as

/  -C2S. In contrast, optimal and super optimal cements, show longer dormant periods which increase with increasing calcination temperatures. Lilienfeld 920°C/300min is a good illustration of this tendency (Figure 30). Though containing quite a high proportion

/ of  -C2S, on hydration, displays a long dormant period lasting 4 weeks.

82

Figure 28: X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65) at increased curing time. Relative intensities of the diffraction maxima of calcite (C) and belites (B).

Figure 29: X-ray diffraction patterns of Folwark Roman cement 960oC/300min and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite (C), gehlenite (G) and belites (B).

83 Prompt cement has a dormant period of seven days (Figure 31). A very slow formation of C-S-H was observed for Rosendale cement paste (Figure 32) with a significant consumption of belites starting only after three months of hydration.

However, it is not the result of a long dormant period but rather the generally low reactivity of the cement. This observation was confirmed by the low strength of Rosendale cement mortars for prolonged periods of curing (Chapter 5.2.2).

Further late-stage processes occurring in cement pastes

Two further processes simultaneously take place with C-S-H formation, in the late stage of hydration - the amorphisation of the crystalline AFm and AFt phases initially formed and the carbonation of the pastes due to the reaction of calcium hydroxide with

CO2 from the air.

The gradual decline of the diffraction maxima of the AFm phases with increasing hydration time was observed for all the pastes studied; for some they disappeared completely after six months. Significant depletion starts after 28 days of hydration for most pastes studied and is accompanied by the formation of an amorphous phase, which can be detected as an increase in the background of the recorded XRD patterns. Many authors point to a metastable character of the AFm phases (Silva and Glaser, 1990, 1993,

Cambrera and Rojas, 2000, Rojas and Sanchéz de Rojas, 2003); a transformation of

C4AH13 into hydrogarnet after long curing was mentioned many times. Jensen et al. (2005) propose a complete decomposition of the AFm phases into amorphous Al(OH)3 or very poorly crystalline gibbsite. Another explanation is offered by Pérez Méndez and Trivio

Vázquez (1984) who report that, in the presence of CO2 in the paste system, some of the

AFm phases convert into carbonated phases (aragonite) and alumina. However, further work is required to improve the understanding of this process.

84

Figure 30: X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite (C), gehlenite (G) and belites (B).

Figure 31: X-ray diffraction patterns of Prompt cement and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite (C), gehlenite (G) and belites (B).

85

Figure 32: X-ray diffraction patterns of Rosendale cement and its paste (w/c=0.65) at increasing curing time. Relative intensities of diffraction maxima of calcite (C), gehlenite (G), belites (B) and periclase (Pe).

Correlation between the formation of C-S-H gel and the late strength of the pastes

The formation of C-S-H gel clearly correlates with the late increase in the strength attaining levels of 12-20 MPa (Figure 10). The final strength levels are characteristic of a given cement source, i.e. they do not vary within the selected field of the calcination conditions controlling the nature and crystallinity of belites. Hence, pastes prepared from

Folwark and Lilienfeld cements attain maximum strength values of 12 and 20 MPa, respectively. English cements are situated in between, attaining a maximum strength of about 15 MPa.

The addition of 0.4% of citric acid as a retarder does not reduce the strength; conversely, higher values of compressive strength were measured for these samples

(Figure 10). However, the retarder had a negative effect on strength when its concentration was increased.

86 The strength development profiles of Roman cements described are of considerable practical significance. The fast initial hydration reaction and early strength development make possible the swift manufacturing of architectural castings and easy execution of on- site rendering and run work. The second phase of the hydration reaction – the precipitation of C-S-H gel - gives Roman cement mortars characteristic hardness and compact microstructure resulting in excellent weather-resistance and long-term durability to outdoor exposure. As C-S-H is poorly crystallised, the kinetics of its formation could not be directly traced using the X-ray diffraction. Therefore, the growth of the silicate gel into the pore space of hardened cement pastes was studied in this work with the use of mercury intrusion porosimetry. Further, adsorption of water vapour and nitrogen and thermal analysis were used to determine the amounts of hydrated products formed as a function of the curing time. The obtained results, which give a deeper insight into the progress of hydration in the pastes of natural cements, are presented in subsequent chapters.

7.1.1.3 Hydration of Roman cement mortars

In addition to the extensive study of bulk specimens of pastes two sets of Roman cement washes (2-3 mm thin) were investigated. They were prepared as described in

Chapter 5.3 and their in-situ XRD patterns were recorded for hydration period up to 3 months (Figures 33, 34).

Both samples show the same pattern of hydration as observed for the bulk pastes

(Figures 14-27). An immediate formation of AFm proceeds, however, in less amounts compared to bulk samples. This might be due to an extreme sensitivity of such thin applications to curing conditions; even a slight deviation from wet-air curing might yield the retardation of hydration.

87

Figure 33: X-ray diffraction patterns of the Folwark 830°C/650min wash layer (w/c=1.0), at increasing curing time. Abbreviations: A1 = C4A C 0.5H12, A3 = C4A C H11, E = ettringite, P = portlandite, B = belite, Q = quartz, G = gehlenite, C = calcite. Reflection F comes from foil covering the hydrating sample.

Figure 34: X-ray diffraction patterns of the Folwark 830°C/650min wash layer (w/c=0.7), with addition of citric acid (0.3%) at increasing curing time. Abbreviations: A1 =

C4A C 0.5H12, A3 = C4A C H11, E = ettringite, P = portlandite, B = belite, Q = quartz, G = gehlenite, C = calcite. Reflection F comes from foil covering the hydrating sample.

88 The AFm formation is further suppressed in wash containing citric acid. Moreover, the surface applications are extremely vulnerable to early carbonation. The carbonation accounts for a characteristic whitening of such surface layers, which contributes to the natural appearance of the Roman cement facades.

89 7.1.2 The development of specific surface area in Roman cement pastes

Measurements of specific surface area (SSA) by gas adsorption have been widely used for the evaluation of hydrated cements. The two adsorbates used almost exclusively were nitrogen and water vapour. Factors that may affect the SSA measurements have been reviewed in Chapter 6.3. SSA is also obtained in the mercury intrusion porosimetry; however, this method yields values corresponding only to pores larger than 3.5 nm in diameter (Chapter 6.2).

In this study, the SSA measurements were applied to follow the microstructure development during the hydration of Roman cement pastes and mortars; it complements information obtained by other methods described earlier.

7.1.2.1 Experimental approach

As an SSA measurement is based on gas adsorption, it is fundamental to remove air as well as physisorbed species from the paste specimen to enable the migration of the adsorbate gas into its pore system. The experimental difficulty consists, however, in the limited possibility of distinguishing the water physisorbed on the surface of the existing pores from water strongly bound in the hydrates or as interlayer water within their structure. By way of example, Figure 35 shows a mass decrease as a function of the outgassing time for the paste of Lilienfeld cement 920oC/300min cured for 28 days. As one can see, the plot follows an exponential type function and the final mass decrease is estimated at 2.8%. The release of water is continuous but is especially intense within the first three hours of the outgassing, during which the weakly bound water is removed. The subsequent measurements of the sorption isotherms for samples, outgassed for two time

90 periods of 3.5 and 42 hours, shown in Figure 36, proves that strongly bound and physically adsorbed water can be separated in the adsorption data.

Figure 35: Mass change with the outgassing time.

Figure 36: Adsorption-desorption isotherms of water vapour on Lilienfeld 920°C/300min Roman cement paste cured for 28 days, outgassed for 3.5 and 42 h. EMC - equilibrium moisture content, p/po - relative pressure of water vapour.

The sorption consists of two components. Firstly, there is an irreversible initial uptake of the initial portions of water vapour dosed to the system until the re-hydration of

91 the sample is reached. Then, a physical adsorption of water vapour occurs, described by a sigmoid shape of the type II isotherm in the IUPAC 1985 classification (Sing et al., 1985).

It indicates a progressively thickening adsorbed water layer as the vapour pressure is increased up to the saturation pressure, at which point the adsorbed layer becomes a bulk liquid. It can be seen in Figure 36 that the physical adsorption part is the same for varying outgassing periods but it is displaced vertically along the y-axis as levels of the initial rehydration of the surface increase with the increasing removal of the strongly bound water by the outgassing. To better interpret water isotherms of the pastes studied, p/po values on the ordinate axis were replaced by those of t, the statistical thickness of the water layer absorbed on the non-porous material, calculated according to Eq.4 in Chapter 6.3.2 and

Eq.3 in the Chapter 6.3.1 was then used to calculate the BET values.

The same EMC - p/po adsorption isotherms as in Figure 36 are shown in Figure 37 as EMC-t plots where EMC stands for the equilibrium moisture content expressed in per cent.

Figure 37: EMC – t plots of the adsorption data shown in Figure 36. EMC – equilibrium moisture content, t – statistical thickness of the adsorbed water layer.

92 Since the volume adsorbed is a product of surface area and water layer thickness, a straight line graph is obtained as long as the water layer can grow freely on the entire surface of the sample. The straight line should go through the origin and its slope represents a water vapour accessible surface area. The slope is identical for the two plots and hence the surface area values are the same irrespective of the outgassing time.

However, upward displacements of the EMC-t plots are observed due to the initial uptake of water vapour during re-hydration.

The important conclusion of the above measurements is that the amount of water adsorbed physically does not depend on the outgassing schedule. Three hours under a vacuum of a residual pressure of less than 10-3 mbar were selected somewhat arbitrarily for further measurements.

7.1.2.2 Interpretation of the measurement data

Adsorption-desorption isotherms of water vapour are shown in Figure 38 for several Lilienfeld cement 920oC/300min pastes, cured for up to two years. The isotherm for 28-day paste is practically identical to that of 1-day paste and was not included in the figure. There is a characteristic evolution of the isotherms reflecting the consecutive formation of hydration products: initially, AFm phases and, when curing time exceeds the dormant period of four weeks, predominantly the C-S-H. The adsorption-desorption isotherms of the young pastes, containing mainly AFm phases, show relatively small water vapour uptake without any hysteresis loop, which indicates an open porous structure with no narrow entrances or ‘ink bottle pores’ (Chapter 6.3.3). No evolution of the pore structure formed at the outset of hydration is observed until the end of the dormant period is reached and the C-S-H gel starts to grow more significantly. The matured pastes which contain significant amounts of C-S-H gel show increased water vapour uptake and a

93 characteristic hysteresis loop closing at about p/po=0.35 indicating a dense microstructure with ‘ink bottle pores’ (Hagymassy et al., 1972).

Figure 38: Adsorption-desorption isotherms of water vapour for Lilienfeld 920°C/300min Roman cement paste cured at various ages. Definitions of EMC and p/po as in Figure 36.

Figure 39: EMC – t plot of the adsorption data shown in Figure 38. Definitions of EMC and t as in Figure 37.

94 The same EMC - p/po adsorption isotherms as in Figure 38 are shown in Figure 39 as EMC - t plots. The straight line graphs are obtained as long as the water layer grows freely on the entire surface of the sample. Their slopes are water vapour accessible surface areas that include the walls of pores provided that the distance between the walls is large compared with molecular dimensions. However, when the water layer formation is accompanied by excess capillary condensation, an upward deviation in EMC-t plots is observed, due the presence of an extra amount of water when compared to the standard.

7.1.2.3 Specific surface area of Roman cement pastes and mortars

Table 14 shows specific surface areas of the pastes determined with the use of several methods: St H2O calculated from the EMC-t plots for water vapour adsorption as described above, BETH2O and BETN2 calculated using water and nitrogen adsorption data in a low pressure range and a least-square fit for the linearized two-parameter BET equation, and SHg - cumulative pore surface area of pores accessible to mercury (up to 3.5nm in diameter) calculated under the assumption of a cylindrical pore geometry. The adsorption data used to calculate the BETH20 values were those of the ‘pure’ physical adsorption of water vapour, which were obtained by subtracting from the raw measurements the irreversible initial uptake of the water vapour due to re-hydration, determined from the

EMC-t plots as shown in Figure 39.

General conclusions can be drawn from the results shown in Table 14. First, an excellent agreement between BETH20 and St H20 clearly points to the reliability of both the calculation of the BET areas and the t-curve of Badmann et al. (1981) relating the statistical thickness of the adsorbed water layer to the relative pressure, used for the interpretation of the sorption isotherms. As a result, the obtained values of the water vapour accessible surface areas can be taken as reliable and useful indicators of the growth

95 of the hydration products, at least for a comparative evaluation of several cements which is the focus of the present investigation. The water vapour accessible surface areas are believed to represent the total surface of the pastes at all stages of hydration (Odler, 2003).

Table 14: Specific surface area of Roman cement pastes cured at various times as determined by different experimental methods.

Cement Hydration St H2O BETH20 BETN2 SHg 2 2 2 2 time [m /g] [m /g] [m /g] [m /g] 15 min 10.0 10.0 10.7 - Folwark 830°C/650min 7 days 37.9 38.0 51.3 63.3

28 days 61.5 58.8 41.7 82.0

3 months 67.5 65.5 54.2 73.0

6 months 70.9 69.3 37.9 92.8

2 years 123.8 111.4 54.7 -

Folwark 15 min 13.7 13.7 7.0 - 960°C/300min 28 days 27.8 26.8 22.0 25.4

3 months 85.3 84.7 38.6 63.5

6 months 93.3 93.4 36.3 80.0

2 years 97.7 98.7 38.7 -

Lilienfeld 15 min 13.7 13.7 11.7 - 920°C/300min 1 day 19.7 19.9 11.5 18.7

28 days 21.7 23.2 16.4 27.5

3 months 73.6 73.6 26.5 72.9

6 months 109.1 105.0 46.1 102.6

2 years 124.3 122.3 51.7 115.5

96 In the case of Roman cements, the initial stage of hydration involves the formation of both AFm and C-S-H phases, however the crystalline AFm phases are dominant

(Chapters 7.1.1.1). Their presence leads to the surface area not exceeding 20 m2/g. The formation of C-S-H gel brings about an increase in the surface area up to 120 m2/g, the value being approximately similar for all pastes studied and proportional to the degree of hydration and thus to the amount of the C-S-H gel formed. The development of the surface area (Figure 40) correlates well with the strengths of the pastes on the one hand (Chapter

5.2.2) and the porosity structure evolution on the other (Chapter 7.1.5.1).

The BET specific surface areas determined with nitrogen as the adsorbate are close to those found by water vapour adsorption during the early stage of hydration, and corresponds to the formation of the open structure by the AFm phases. When only the growth of the C-S-H gel is initiated, the BETN2 values become consistently lower than those found by water vapour adsorption. The BETN2 values are similar to those reported by

Guerrero and Goi (2005) for belite cement pastes reaching 40 m2/g after 180 days.

Figure 40: Evolution of the BETH2O surface area of three studied Roman cement pastes with hydration time.

97 The various interpretations of the differences between nitrogen and water accessible

SSAs have been recently reviewed by Odler (2003) who points to the formation of denser and more interwoven porous spaces that are totally or partially impermeable to N2 molecules at the liquid nitrogen temperature as the principal explanation.

The present study has further revealed a generally close agreement of the SSAs determined by mercury porosimetry with those found by water vapour adsorption, irrespective of the degree of paste hydration. As mercury porosimetry allows determining only the surface of those pores into which mercury has penetrated, the observation points to the fact that multilayer physical adsorption of water is located in pores with diameter above about 3.5 nm.

As can be seen from the experimental evidence presented, the determination of the water vapour adsorption opens the possibility of following the progress of the hydration in the Roman cement pastes and mortars. This method, in particular, could be useful in determining the amount of C-S-H formed, as suggested earlier by Olson and Jennings

(2001).

The second part of this study focused on the hydration mechanism of the conservation mortars. The experiments described earlier point to a considerable sensitivity of the hydration progress in the Roman cement mortars to even slight deviations from the ideal moist conditions. This is especially pronounced in thin layers, like finish layers or washes of thickness not exceeding 3 mm that are vulnerable to swift drying and carbonation (Chapters 7.1.1.3, 7.1.3.2).

Also the SSA measurements show restricted hydration of the sample of the surface coatings, which is illustrated in Table 15 and Figure 41. The significantly lower BETH2O values for these samples, especially at late ages, point to reduced hydration irrespective of the curing time. The same restricted hydration is described by Thomas et al. (1996) and

98 Rarick et al. (1996) who point to carbonation of the samples as a further process blocking or even reducing SSA development. Carbonation gives rise to the growth of calcite crystals within C-S-H fibrils (Chapter 3.2.1) which leads to partial destruction of the gel porosity (Richardson, 2004). Carbonation is particularly marked for samples with a higher surface/volume ratio. Extensive carbonation of the wash coatings was also found by X-ray diffraction (Chapter 7.1.1.3) and confirmed by thermal analysis (Chapter 7.1.3.1).

Table 15: Specific surface area of Roman cement mortars cured at various times as determined by water vapour adsorption.

Mortar Hydration St H2O BETH20 composition time [m2/g] [m2/g] 7 days 26.7 22.4 Folwark 830°C/650min 28 days 40.8 42.4

3 months 61.5 63.6

finish layer 6 months - - w/c = 0.7 4 hours 21.4 21.3 Folwark 830°C/650min 24 hours 27.8 24.3

7 days 41.2 45.8 wash layer w/c = 1.0 28 days 48.5 48.3

3 months 51.9 51.9

6 months 85.4 81.4 4 hours 23.3 24.5 Folwark 830°C/650min 4 days 44.6 45.4

7 days 38.8 39.4 wash layer w/c = 0.7 28 days 41.4 42.4

3 months 43 43.9

6 months 60.9 60.3

99

Figure 41: BETH2O surface area of different Roman cement applications prepared from Folwark 830°C/650min (w/c=0.65), 0.3% of citric acid as a retarder was used.

100 7.1.3 Thermal analysis

The combined thermogravimetric (TG) and differential thermal analyses (DTA) were used to gain further insight into the nature of hydrated phases, as well as to determine quantitatively the degree of hydration of pastes and mortars at different curing stages and conditions. As in other parts of the study, a particular attention was paid to thin coatings which are very vulnerable to swift drying in the real world conditions.

Pastes for the investigations were produced from two cements – Folwark

830°C/650min and Lilienfeld 860°C/300min. They represent two typical varieties of

Roman cements; on hydration, the first yields low initial strength and a short dormant period whereas the other high initial strength and a long dormant period (Chapters 5.2.2,

7.1.1.1, 7.1.1.2). It can be recalled at this point that the initial strength depends on the quantity of AFm phases formed during the early hydration.

The TG/DTA curves of the pastes are shown in Figures 42, 43 (Folwark

830°C/650min) and 44, 45 (Lilienfeld 860°C/300min). The curves of dry cements are also included for comparison. Curing times at which the analyses were carried out were the same as selected for the X-ray diffraction investigations. Folwark 830°C/650min paste cured for 1.5 year was also analysed representing a sample of well-matured paste.

7.1.3.1 Identification of hydrated products in Roman cement pastes

Generally, two important regions can be identified during the thermal decomposition of pastes or mortars. The weight loss until 85°C is related to physically bound water, e.g. free or evaporable water present in the pore system of the paste, which does not take part in the hydration reaction. It can be very easily removed from the pastes on heating or using other drying methods. Non-evaporable water, also called chemically bound water or strongly bound, is contained in the products of hydration. Generally, its

101 content can be calculated as the weight loss between 85 and 1000°C (Serry et al., 1984,

Vedalakshmi et al., 2003, Pane and Hansen, 2005). When samples are properly dried, only negligible amount of free water is present in the pastes and the weight loss is predominantly related to water bound in the hydrated phases.

Several phases can be identified in the Roman cement pastes from endothermic peaks observed in the DTA thermograms and the corresponding weight losses. A major endothermic effect and mass loss in the temperature range 440-520°C is attributed to the dehydration of Ca(OH)2 (CH) which proceeds according to Eq.6. The effects within the temperature range 600-780°C are associated with the decomposition of amorphous and crystalline calcium carbonate which proceeds according to Eq.7.

Ca(OH)2 CaO + H2O (Eq.6)

CaCO3 CaO + CO2 (Eq.7)

Portlandite is present in significant amounts only in the pastes of Folwark

830°C/650min, in which initially it is a product of slaking of reactive free lime contained in the cement (Chapter 5.1). Portlandite prevails in the paste even after long curing times without a significant transformation to calcium carbonate. Calcium carbonate is present both in the cements and the pastes. In later stages of hydration reaction, a shift in the position of the calcium carbonate minimum is observed for both pastes. It can be interpreted as being due to the precipitation of poorly crystalline phase which decomposes at lower temperatures. Such observation was described by Dweck et al. (2000). Another possibility is that the carbonation of paste produces calcium carbonate of such thermal characteristics.

102 Three further endothermic effects are observed in the ranges 85-95, 105-160 and

200-245°C respectively. The first two are attributed to calcium silicate hydrate C-S-H

(Serry et al., 1984, Taylor, 1990, Saad Morsy et al., 1997, Ubbríaco and Tasselli, 1998,

Rojas and Cambrera, 2002). In both pastes analysed, the DTA spectrum is dominated by the second broad peak with minimum around 135°C.

The third minimum at about 220°C, can be related to the major AFm phase formed on the Roman cement hydration which was identified by the X-ray diffraction as

C4A C 0.5H12, the A1 hemicarbonate phase in Chapter 7.1.1.1. As described earlier,

C4A C 0.5H12 is metastable and converts within initial 24 hours to C4AH13 (A2) and

o C4A C H11 (A3). This is confirmed by a gradual disappearance of the minimum at 220 C on the thermograms. According to Ubbríaco and Tasseli (1998) and Rojas and Sánchez de

Rojas (2003), the decomposition temperature for C4AH13 is 160-210°C, i.e. lower than that for the hemicarbonate. It would partially overlap with the effect related to CSH and A1, which is confirmed by the significant broadening of the peak in this temperature range.

Therefore, it is impossible to determine quantitatively contents of the individual hydration products from the DTA data.

The TG thermograms reveal a different progress of the hydration reaction in each paste, as it was earlier observed by the in-situ X-ray diffraction (Chapter 7.1.1.1). Folwark

830°C/650min shows a continuous increase in the hydrated phase content (Figure 42) whereas Lilienfeld 860°C/300min shows the dormant period of 14 days during which a negligible increase in the hydrated phase content is observed (Figure 44). Only after the dormant period, a significant evolution of C-S-H occurs as indicated by increasing water loss in the temperature region below 400°C and also by an increasing area of the endothermic minimum at 135°C on the DTA curves. In the Folwark 830°C/650min paste

103 cured for 1.5 year, the two minima related to C-S-H merge to one broad feature shifted to around 100oC.

Figure 42: Thermogravimetric curves of Folwark Roman cement 830°C/650min and its paste (w/c=0.65) at increasing curing time.

Figure 43: DTA thermograms of Folwark Roman cement 830°C/650min and its paste (w/c=0.65) at increasing curing time.

104

Figure 44: Thermogravimetric curves of Lilienfeld Roman cement 860°C/300min and its paste (w/c=0.65) at increasing curing time.

Figure 45: DTA thermograms of Lilienfeld Roman cement 860°C/300min and its paste (w/c=0.65) at increasing curing time.

105 The results obtained from the thermogravimetric study of the pastes have confirmed the general pattern of hydration reaction revealed by the XRD investigations. The latter were however able to follow precisely only the formation and transformation of the crystalline AFm phases and not the initial stages of the belite hydration to C-S-H. Thermal analysis did neither help in determining quantitatively the various products formed on hydration due to the overlapping of the effects corresponding to the decomposition of CSH and AFm.

7.1.3.2 Quantification of the hydrated product content

Table 16 and 17 show weight losses related to the decomposition of phases which are formed on hydration of Roman cements: C-S-H, AFm, Ca(OH)2 and CaCO3. The total weight loss at 1000°C represents the loss on ignition (LOI). As discussed in the previous chapter, it is impossible to separate the effects due to the dehydration of CSH and AFm.

Table 16: Weight losses measured by TG analysis for Folwark 830°C/650min paste.

Curing time CSH and Ca(OH)2 CaCO3 CaCO3 (%)/ LOI AFm (wt.%) (wt.%) Ca(OH)2 (%) (wt.%) (wt.%) 85-400°C 440-520°C 600-780°C 85-1000°C

cement 0.25 0 8.81 5.54 11.40 15 min 3.51 1.12 8.03 7.17 13.03 30 min 3.94 1.07 9.16 8.56 12.99 1 h 3.87 1.15 8.09 7.03 13.85 24 h 5.41 1.29 7.90 6.12 15.62 7 d 7.46 1.25 7.91 6.33 18.12 14 d 8.16 1.14 7.46 6.54 18.49 21 d 8.76 1.35 7.27 5.39 18.79 3 m 10.36 1.06 6.22 5.88 19.79 1.5 y 12.26 0.74 6.11 8.26 24.68

Therefore, the process is quantified as total weight loss between 85 and 400°C, at which portlandite starts to decompose. Contents of Ca(OH)2 and CaCO3 in the pastes were

106 calculated from the weight losses related to their dehydration or decarbonation according to Eq.6 and 7, respectively, and are plotted in Figures 46 and 47 as a function of curing time. Contents of these phases in the original cements are used as starting points of the curves. Additionally, weight loss due to the water release from CSH and AFm is plotted as it indirectly reflects the evolution of the hydrate content with curing.

The contents of portlandite and calcite in both cements agree well with their mineralogical compositions established with the use of the X-ray diffraction and presented in Chapter 5.1; 20% and 15% of calcite in Folwark 830°C/650min and Lilienfeld

860°C/300min, respectively, as well as insignificant amounts of portlandite in both cements were found.

Table 17: Weight losses measured by TG analysis for Lilienfeld 860°C/300min paste.

curing time CSH and CaCO3 LOI AFm (wt.%) (wt.%) (wt.%) 85-400°C 600-780°C 85-1000°C

cement 0.85 1.26 7.24 15 min 5.83 1.89 12.13 30 min 6.28 1.09 12.15 1 h 6.51 1.27 12.73 24 h 6.19 1.14 10.96 7 d 8.55 1.29 14.64 14 d 10.20 1.16 16.32 21 d 10.26 0.95 15.21 3 m 14.16 1.13 21.59

During hydration, the calcium carbonate content in both pastes decreases - rapidly in the initial stages until 14 days and slowly thereafter. The decline in calcium carbonate is paralleled by the increase in the AFm content. Both observations can be accounted for by the consumption of calcium carbonate when the two main carbonate AFm phases - hemicarbonate (C4A C0.5H12) and monocarbonate (C4A CH11) are formed, predominantly

107 within the first 14 days as documented by the X-ray diffraction studies. CH content initially increases only in Folwark 830°C/650min paste due to slaking of reactive free lime present in the original cement (Chapter 5.1) and remains unchanged without significant transformation into calcium carbonate via carbonation within the first 3 months of hydration.

Figure 46: Composition of Folwark Roman cement 830°C/650min paste as a function of curing time.

Figure 47: Composition of Lilienfeld Roman cement 860°C/300min paste as a function of curing time.

108 CaCO3 (%)/Ca(OH)2 (%) ratios presented in Table 16 were used to evaluate the carbonation degree of the individual pastes. In Folwark 830°C/650min paste some relative increase in the CaCO3 content was observed only for 1.5 year old paste, which indicates a slow carbonation of CH (Lanas et al., 2004). In Lilienfeld pastes, no portlandite was detected.

7.1.3.3 The degree of hydration

The weight loss data allow calculating the degree of hydration of pastes  defined as a ratio of the chemically bound water wb,t after a given hydration time t, to the

chemically bound water in fully hydrated material (wb , ) (Eq.8).

w  b,t (Eq.8) wb,

wb,t was calculated according to Eq.9:

 wwb, t b total(1 ) w b cement (Eq.9)

in which

w (%) LOI  m  m (Eq.10) b,total paste CaCO3 Ca(OH )2

w (%) LOI  m  m (Eq.11) b,cement cement CaCO3 Ca(OH )2

The total weight loss at 1000°C (LOI) and weight losses due to the decomposition of

CaCO3 and Ca(OH)2 are expressed as per cent of the ignited weight of the sample.

Substituting Eq. 9 in Eq. 8 one finally obtains:

109 ww  b total b cement (Eq.12)  wwbb  cement

wb,cement for Folwark 830°C/650min and Lilienfeld 860°C/300min was determined as 0.3% and 0.6%, respectively. The temperature range used in the determination of LOI was between 85 and 1000°C, slightly different from the common practice in which 100 or

105°C is used instead of 85°C (Lam et al., 2000, Pane and Hansen, 2005). The change was introduced because the C-S-H decomposition was found to start at the latter temperature

(Chapter 7.1.3.1).

Thewb, values range between 0.23 and 0.25 in the pastes of Portland cements;

0.23 is generally used in the calculations of the degree of hydration (Copeland et al., 1960,

Taylor, 1990, Mouret et al., 1997, Pane and Hansen, 2005). For Roman cement pastes studied, the maximum wb values were determined for two long-cured pastes as 0.23 for 1.5 year Folwark 830°C/650min and 0.26 for 7 months Lilienfeld 860°C/300min (Tables 18 and 19). The latter value was finally used for the calculation of  for the both pastes according the Eq.12.

The calculated values of chemically bound water wb, t and degree of hydration  after each hydration period are presented in Tables 18 and 19 and are plotted in Figures 48 as a function of curing time.

It can be seen that, during initial 24 hours of curing, values of wt,b and for Folwark

830°C/650min are lower than those for Lilienfeld by around 50%. The finding confirms the results obtained by the X-ray analysis that Lilienfeld cement is initially more reactive compared to Folwark. Also the subsequent hydration progress of the two cements agree well with previous findings: the Folwark paste shows a further fluent increase of and wb with hydration time, whereas the hydration degree of the Lilienfeld paste increases sharply within first 24 hours but only a negligible increase can be seen until 21 days, which

110 correlates well with the dormant period. A further sharp increase can be observed for the late-age samples, starting at 3 months. The final values of wt,b and obtained for the matured pastes of both cements are, however, very similar, confirming that Roman cement pastes and mortars can attain similar final levels of hydration, not significantly influenced by the reactivity of a given cement.

Table 18: Chemically bound water content wb,t and degree of hydration  for the Folwark 830°C/650 min paste.

Curing time LOIpaste Chemically Degree of (wt.%) bound water hydration wb,t 

cement 11.40 0.003 - 15 min 13.03 0.041 0.159 30 min 12.99 0.029 0.110 1 h 13.85 0.050 0.193 24 h 15.62 0.073 0.281 7 d 18.12 0.106 0.409 14 d 18.49 0.118 0.454 21 d 18.79 0.122 0.469 3 m 19.79 0.153 0.587 1.5 y 24.68 0.234 0.898

Table 19: Chemically bound water content wb,t and degree of hydration  for the Lilienfeld 860°C/300min paste.

Curing time LOIpaste Chemically Degree of (wt.%) bound water hydration wb,t 

cement 7.05 0.006 - 15 min 12.13 0.071 0.277 30 min 12.51 0.077 0.300 1 h 12.73 0.080 0.312 24 h 10.96 0.067 0.256 7 d 14.64 0.107 0.420 14 d 16.32 0.131 0.511 21 d 15.21 0.121 0.472 3 m 21.89 0.206 0.807 7 m 24.45 0.256 1.000

111

Figure 48: Chemically bound water wb,t versus curing time for pastes of Folwark 830°C/650min and Lilienfeld 860°C/300min cements.

Figure 49: Degree of hydration  versus curing time for pastes of Folwark 830°C/650min and Lilienfeld 860°C/300min cements.

Three selected hydration times were marked in Figure 49 to indicate the minimum requirements for the adequate curing of Roman cement pastes. It can be seen that at least

14 days of moist curing is required to obtain materials in which hydration degree attains

50%, which confirms generally a slow progress of hydration in Roman cement mortars and the need for comparatively long curing times.

112 Thermal analysis can be potentially a particularly promising technique for determining the progress of hydration in practical applications as it is fast and requires minute specimens which can be easily withdrawn from historic objects especially from thin finishing layers and coatings. In Figure 50 and 51, the DTA/TG curves for washes prepared from Folwark 830°C/650 min cement mixed with water in ratios 0.7 and 1.0, cured for 6 months, are compared. To retard setting, the first mixture was modified with

0.3% of citric acid.

Figure 50: DTA and TG thermograms of the Folwark 860°C/300min wash layer (w/c=0.70), with addition of citric acid (0.3%).

The results clearly show that the general hydration pattern does not change when compared to the bulk paste samples as evidenced by the same sequence of minima on the

DTA curves, without significant differences between both varieties of wash samples.

However, no evidence of the single endothermic peak related to hemicarbonate phase can

113 be found, substituted by similar effect around 190°C related to the C4AH13 (Ubbríaco and

Tasseli, 1998, Rojas and Sánchez de Rojas, 2003). The absence of the carbonate phases is consistent with results obtained by X-ray diffraction performed on the washes with the same composition (Chapter 7.1.1.3).

Figure 51: DTA and TG thermograms of the Folwark 860°C/300min wash layer (w/c=1.0).

In the wash modified with citric acid, the retardation effect is observed in the initial stages of hydration as a significant lowering of the degree of hydration (Table 20) when compared with not modified samples. Calcite content in wash with higher w/c=1.0 was found to increase strongly with curing time (Figures 50, 51), probably as a consequence of better diffusion of the CO2 through a more porous structure of the matrix. The effect of the water content on the porosity structure is discussed further (Chapter 7.1.5.2).

114 Table 20: Degree of hydration  of Folwark 830°C/650min wash layers (w/c=0.7 and 1.0).

Curing time Degree of hydration 

w/c=0.7 w/c=1.0 (0.3% citric acid) 1 d 0.123 0.267 7 d 0.376 0.489 28 d 0.546 0.506 3 m 0.551 0.601

115 7.1.4 The microstructure of Roman cement pastes by means of SEM-EDX analysis

The Scanning Electron Microscopy (SEM) with Energy Dispersive X-ray

Spectrometry (EDX) was widely used to characterize the hydraulic mortars (Maravelaki-

Kalaitzaki et al., 2005, Varas et al., 2005), concretes (Hu and Li, 1999) or cement pastes

(Hanehara et al., 2001, Guerrero and Goi, 2005). In this study, SEM-EDX was used to characterize the microstructure and morphological features of Roman cement pastes, which complemented the results obtained by other methods: X-ray diffraction (Chapter 7.1.1),

BET analysis (7.1.2), DTA/TG (7.1.3) and the MIP (7.1.5). The investigations were carried out at the Insitute of Art and Technology, University of Applied Arts in Vienna, Austria.

The SEM analysis was carried out on fractured samples of Lilienfeld

920°C/300min paste. The selected cement is a good example of typical Roman cement as it possesses characteristic features of the material – a high initial strength, a long dormant period of two months followed by an increase in final strength (Chapter 5.2.2). The micrographs of the unhydrated cement clinker are also shown in Figures 52a, b for comparison.

The clinker generally consists of irregular aggregates of belite phases; the main constituents of the Roman cements clinkers. They form leaves or tissues consisting of very fine plates. The morphology of alumina phases is not clearly visible as they are dispersed within the matrix (Figure 52a). However, some decomposed clay remnants can be seen as layered structures within the matrix of the clinker (Figure 52b); the predominant elements detected being aluminum and potassium. Furthermore, calcium-rich zones are also observed, pointing to the presence of unreacted calcium carbonate, oxide or hydroxide.

116 a) Tissue-like agglomerates of C2S (belite) with dispersed alumina phases.

Ca/Si = 2.03 Ca/Al = 3.70

b) Amorphous relicts of clay minerals rich in alumina and alkali (predominantly potassium) with a layered morphology.

Al/Si=0.71 (Al+K)/Si=1.47

Figure 52: SEM/EDX analysis of Lilienfled 920°C/650min Roman cement clinker.

Figures 53a-d show SEM images of Roman cement pastes at various stages of hydration; the curing times were similar to those in other studies. Part a represents fresh paste studied within 15 minutes after mixing the cement with water. It clearly shows morphology which is quite different from that of the original cement. It is known from the in-situ X-ray diffraction that crystalline AFm phases are predominantly formed at this early stage of hydration. The tissue-like structures which typically adhere to the original cement particles are clearly related to AFm (Figure 53b) as their chemical composition -

Ca/(Al+Fe)=4.0 – corresponds to the stoichiometry of the AFm species (Chapter 7.1.1.1).

Furthermore, novel globular structures precipitate on the original cement particles and have been identified as nuclei of the C-S-H phase (Figure 53c). The same structures of calcium

117 silicate hydrates are described by Varas et al. (2005) within historic natural cement mortars and Montgomery et al. (1981) in fly ash concrete. The morphology of the pastes cured for three days shows a progressive decrease in the porosity due to an increasing content of hydrated phases.

a) Morphology of the paste after 15 minutes of hydration. Needle-like growths of AFm and globular particles of CSH.

b) Morphology of the paste after three days of hydration.

Tissue-like structures: Ca/Si=8.45 Ca/(Al+Fe)=4.02

c) Globular particles within the paste after three days of hydration.

Globular structures: Ca/Si=2.39 Ca/(Al+Fe)=3.54

118 d) Morphology of the two-month-old paste – characteristic interlinking the matrix reduces porosity.

Compact plate-like structures: Ca/Si=2.19 Ca/(Al+Fe)=5.46

Figure 53: SEM/EDX analysis of hardened Lilienfeld 920°C/300min Roman cement paste at different curing periods (w/c=0.65).

The observations confirm the general conclusion that AFm phases are mainly formed during the initial stages of hydration and their formation leads to an open porous structure. The SEM-EDX study provides additional information that small quantities of C-

S-H are also formed in the young paste matrix. The C-S-H growth, however, is slow for at least 28 days.

Morphology of one-year-old Roman cement paste – compact structure rich in CSH.

Compact plate-like structures: Ca/Si=1.85 Ca/(Al+Fe)=4.19

Figure 54: SEM/EDX analysis of hardened Lilienfeld 920°C/650min Roman cement paste cured for one year (w/c=0.65).

119 The SEM photomicrographs of late-age pastes (Figures 53d and 54) show distinctly different morphology. The interlinked structure is predominantly composed of C-

S-H rather than AFm. It displays significant compactness and reduced porosity. The tissue- like structures of AFm are overgrown by the C-S-H gel.

The EDX analysis performed at different spots of the sample area mainly identified high contents of Ca, with the Ca/Si ratios of 2.2, and only relics of AFm. No evidence of calcite crystals formation was found, probably as a consequence of an intimate mixture of calcite or vaterite crystals in the fine-textured gel and a general little carbonation in the bulk. The morphology of the one-year old Lilienfeld paste also shows a very compact structure of a homogenous morphology in which one cannot distinguish individual crystal structures characteristic of the earlier stages of hydration.

The described observations agree well with the results obtained by other methods as they point to C-S-H growth at later stages of hydration, which reduces the amount and size of pores. The process is closely related to the increase in final strength.

120 7.1.5 Pore structure of hydrated Roman cements as measured by mercury intrusion porosimetry (MIP)

7.1.5.1 Pore structure of Roman cement pastes

Figures 55 a-f compare the differential mercury intrusion curves, i.e. incremental pore volume intruded as a function of pore diameter for several Roman cement pastes studied at ages ranging from 15 minutes to 1 year. For comparison, the hydration of

Ordinary Portland Cement (OPC) paste was also carried out. The OPC pastes were produced at the same water-cement ratio as for other Roman cements, which allowed comparisons to be made.

Early stage of hydration

At an early age, pastes exhibit a single, sharply defined peak at a pore diameter of

0.2 to 0.8 m. The presence of the peak indicates a one-step intrusion of mercury into a capillary network connected to the specimen surface. It corresponds to the minimum throat dimension of this network – the ‘critical’ or ‘threshold’ diameter according to

Diamond (2000). Cook and Hover (1999) systematically determined the threshold pore widths for Portland cement pastes as a function of the w/c ratio and the curing time. They varied between approximately 2 and 0.02 μm and, as expected, smaller values were obtained with increased curing time and decreased w/c ratio. The values for one-day pastes clustered between 2 and 0.4 μm for w/c ratios between 0.7 and 0.4; only the 0.3 w/c paste did not exhibit the initial mercury intrusion peak, corresponding to a large threshold pore width, at every curing period tested.

The pore structure of young Roman cement pastes correlates with the strength development profiles in the pastes; the initial single peak recorded in the pastes cured for the least amount of time remains unchanged during the dormant period of the paste, which

121 can last up to several weeks. Only then does the peak shift to smaller pore sizes, which is

assumed to be as a result of the filling of larger pores by the C-S-H gel.

a) Lilienfeld 920°C/300min b) Folwark 960°C/300min

c) Harwich 890°C/500min d) Sheppey 890°C/500min

Figure 55: Differential volume of intruded mercury versus pore diameter for Roman cement pastes (w/c=0.65) at increasing curing time: a) Lilienfeld 920°C/300min, b) Folwark 960°C/300min, c) Harwich 890°C/500min, d) Sheppey 890°C/500min.

Although all Roman cement pastes exhibit a unimodal distribution of the throat

pore dimensions at early curing times, the actual threshold pore width varies between

cements depending on their mineral composition, especially on the reactive alumina

components, which was shown to vary across Roman cement samples (Chapter 5.1).

Thus, the smallest values were observed for Liliefeld 920°C/300min cement (0.2 m)

(Figure 55a) whereas the largest values were observed for the intensely calcined Folwark

122 cement 960oC/300min (0.8 m) (Figure 55b) and Harwich 890°C/500 min (0.95 m)

(Figure 55c), which contain high amounts of inactive aluminate phases – gehlenite or

brownmillerite. The observation provides further evidence that, on hydration, cements that

are capable of formation of larger amounts of the AFm phases, produce a denser initial

microstructure and higher early strength. Pastes of rapid-hydrating calcium sulfoaluminate

cement, capable of the very fast development of hydration products – ettringite and

aluminum hydroxide, show a similar threshold diameter at early ages (Bernardo et al.,

2006).

a) Prompt cement b) Rosendale cement

Figure 56: Differential volume of intruded mercury versus pore diameter for Prompt and Rosendale cement pastes (w/c=0.65) at increasing curing time.

The microstructure of pastes of the two commercial cements follows the general

pattern described above (Figures 56a, b). On hydration, the Prompt cement paste shows a

pore structure development similar to that of Roman cements. The paste of the mildly

reactive Rosendale cement shows a very high initial threshold pore width of 2.5 m.

Moreover, the larger pores do not disappear even after long curing times, which leads to a

clear bimodal pore size distribution in the matured paste. Both observations can be

explained by the comparatively low content of active aluminate phase in the cement and

123 thus its lower capacity to bind water during hydration. The high content of the unbound

excess water produces the same effect as high water-cement ratio. The influence of the

water content in the pastes will be discussed further in Chapter 7.1.5.2.

Figure 57 shows the differential mercury intrusion curves for pastes of Folwark

830°C/650min that are pure or retarded with citric acid. They point to a similar pore

structure development. Both threshold pore values are located very close to each other at

around 0.3-0.55 m after the first 24 hours of hydration.

a) Folwark 830°C/650min b) Folwark 830°C/650min with 0.4% citric acid

Figure 57: Differential volume of intruded mercury versus pore diameter for the Folwark 830°C/650min paste (w/c=0.65) pure and with the addition of 0.4% of citric acid as a retarder, at increasing curing time.

Late stages of hydration

As was previously described in Chapters 5.2.2, 7.1.1.1, 7.1.1.2 all cement pastes

display a dormant period of varying length. The phenomenon is also reflected in the

porosimetric data. When the curing time of any given cement exceeds the dormant period,

the initial peak significantly shifts to smaller pore sizes between 0.05 and 0.01 m. When

the effect is compared with the strength measurement (Figure 10) or the development of

hydration of the pastes as studied by the in-situ XRD, TG/DTA, BET and SEM (Chapters

124 7.1.1, 7.1.2, 7.1.3, 7.1.4), the change in pore size can been clearly related to the gradual disappearance of belite and the formation of C-S-H gel which fills larger pores. The pore sizes in the matured paste agree well with the data given in the literature for matured pastes. In particular, Cook and Hover (1999) have shown that the threshold pore widths for Portland cement pastes cluster between 0.2 and 0.02 mat curing times of 56 days.

Rosendale cement paste also displays a shift in pore size. The results agree well with the outcome of the XRD study, that strength development in the Rosendale pastes is almost exclusively due to C-S-H gel formation. Its precipitation starts after seven days of hydration and stays practically unchanged during further hydration (Figure 56b).

The development of the pore structure of the OPC paste shown in Figure 58 is somewhat different. The initial threshold pore width is 1 m. However, a bimodal distribution of pore sizes is achieved quickly, pointing to the swift formation of the regions of lower porosity due to the C-S-H growth.

Ordinary Portland Cement (OPC)

Figure 58: Differential volume of intruded mercury versus pore diameter for OPC (w/c=0.65) at increasing curing time.

In Roman cements, C-S-H gel is generated more slowly so it refines only gradually the pore structure, which is reflected in a continuous shift of the pore network to lower

125 threshold pore widths, rather than the appearance of two porosity regions (the bimodal distribution). The threshold pore width of matured, six-month-old OPC pasties 0.01 m, similar to the same data for matured pastes of optimal Roman cements.

7.1.5.2 The influence of water content on the pore structure of the pastes

Figure 59 compares the differential mercury intrusion curves for pastes prepared from Folwark Roman cement 830°C/650min at three w/c ratios of 0.6, 0.8 and 1 respectively, studied at three ages of hydration - 1 day, 14 days and 3 months. One-day curves make possible the comparison of the pore structure of fresh pastes, 14-day plots illustrate the progress of hydration due to C-S-H gel precipitation and the data at three months represents the matured pore structure of fully hydrated pastes.

The mercury intrusion curves of all three pastes hydrated at early ages exhibit a single peak at a pore diameter depending on the w/c ratio used: 0.7 m for w/c=0.6, 1.2

m for w/c=0.8 and 1.9 m for w/c=1.0 (Figure 59a). As was described in the previous chapter, the threshold pore diameter 0.7 fits the typical threshold pore diameter range of

0.2-0.8 m observed for the early-age pastes with a w/c ratio of 0.6. However, the threshold pore diameter of the samples prepared at an increased w/c ratio of 0.8 and 1 has been considerably shifted to higher values. The formation of larger pores is the result of the evaporation of the excess water which stays unbound in the pores during hydration.

With increasing curing time, smaller pores of 0.01-0.03 m in diameter develop, due to filling the larger pores by the C-S-H gel, as described earlier. Characteristically, at a low w/c ratio of 0.6, the paste exhibits a final unimodal distribution of the throat pore dimensions centred at 0.02 m, the bimodal distribution being seen only in the transition phase of hydration (Figure 59b). In contrast, a bimodal distribution prevails in the pastes prepared at an increased w/c ratio pointing to a coexistence of two porosity regions. This

126 observation can be interpreted in terms of the progress of the hydration reaction. While water content in the paste with water-cement ratio of 0.6 is sufficient to be combined into the hydrated phases, the evaporation of the excess amount of water leaves large pores in the paste structure.

As was discussed earlier, similar bimodal pore distribution was observed for super- optimal cements incapable of forming large amounts of hydrated phases.

a) 1 day of hydration b) 14 days of hydration

c) 3 months of hydration

Figure 59: Differential volume of intruded mercury versus pore diameter for Folwark 830°C/650min cement pastes of various w/c ratios cured for: a) 1 day, b) 14 days, c) 3 months.

127 7.1.5.3 Pore structure of Roman cement mortars

After the detailed study of Roman cement pastes, an extensive study on the pore structure of the repair mortars was carried out with regard to their potential use in the restoration of historic mortars. The mortars tested and their different curing conditions were described in Chapter 5.3, the results are presented in following Chapter 7.2.

128 7.2 Hydration on exposure to real-world environments

7.2.1 Hydration of mortars in different curing conditions

The first set of samples was cured at ideal humid conditions - over distilled water near 100% relative humidity (wet-air curing) or under water (wet curing). The second set was cured at the relative humidity of 75% (air curing) reflecting real-world environmental conditions to which repair materials are exposed on building facades. Figures 60a-c compare the differential mercury intrusion curves for the wet, wet-air and air-cured

Roman cement mortars prepared at a w/c ratio of 0.6, over a three-month study period.

a) wet-air curing b) water-curing

c) air-curing

129 Figure 60: Differential volume of intruded mercury versus pore diameter for Roman cement mortars cured at various conditions: a) wet-air curing, b) water-curing, c) air curing.

During early hydration, all curves exhibit a single peak pointing to a unimodal pore size distribution. However, whereas wet and wet-air curing lead to a dense well- hydrated microstructure after 28 days of hydration with a threshold pore diameter of 0.02

μm, the air-cured specimen does not change its microstructure with curing time as a consequence of restricted hydration due to loss of water at an early stage (Figure 60c). A similar influence of external environmental exposure on porosity and pore-size distribution was reported by Khatib and Mangat (2003). The dominant pore diameter of the top surface of a cube cast of Portland cement paste at a w/c ratio of 0.45, exposed to a dry environment with RH of 25%, was 1.5 μm when compared to 0.2 μm obtained for the bottom surface where the availability of water had allowed more hydration to take place.

However, when the air-cured sample, even after a prolonged storage at dry conditions, was re-wetted by soaking in water and then moist-cured, the hydration continued as illustrated in Figure 61.

Figure 61: Rehydration of air-cured mortar.

130 The change in pore size distribution, which reflects hydration progress, can also be observed for Roman cement mortars collected from historic buildings. Figure 62 shows differential mercury intrusion curves for a sample of historic Roman cement mortar from

Neue Hofburg in Vienna, re-wetted and kept under water for a period of up to four months. A clear shift in the diameter of the initial large pores to smaller values is observed. The experiments point to a considerable sensitivity of the hydration progress in the Roman cement stuccoes to even a slight departure from the ideal moist curing. On the other hand, the re-wetting of facades with rain water starts hydration again allowing the stuccoes to gain sufficient strength with time.

Figure 62: Differential volume of intruded mercury versus pore diameter for a sample of a cast element from Neue Hofburg in Vienna, Austria. Rehydration by re-wetting.

131 Figure 63: Differential volume of intruded mercury versus pore diameter for a 5 mm thin mortar layer at increasing time of moist curing. No significant change was observed in the progress of hydration in a 5 mm thin finish layer (Figure 63). The hydration was not restricted or delayed as had been expected for a relatively thin layer more liable to drying or carbonation. Clearly, in the practical application of finish layers or surface coatings particular attention should be paid to providing moist-curing conditions to prevent drying.

7.2.2 Porosity of historic mortars

All historic mortars studied (Table 21) were found to possess high porosity. Values between 20% and 45% were obtained, which are similar or higher than values found for aerial or hydraulic lime mortars (Lanas et al., 2006, Marques et al., 2006).

Table 21: Characteristics of selected historic Roman cement mortars.

Building sampled Element Open porosity (%)

Trade Academy, (A) cast architectural detail 35 Krakow, Poland (B) plain pre-cast profile 23 (C) run cornice 44 (D) run window framing 43 (E) plain render 23 White Lion House, hollow casting of a lion Eye, Suffolk, UK outer surface 26 inner surface 23 Office building at Liebiggasse 5, hollow casting of a console Vienna, Austria outer layer 34 inner core 44

Differential MIP curves for historic Roman cement mortars mostly exhibited bimodal distribution of the throat radius size. This is illustrated in Figure 65 by the curves obtained for samples of Roman cement mortars collected from the facade of the Trade

132 Academy, Krakow, Poland, built 1904-1905. The various decorative elements sampled

(architectural castings and run in-situ stucco), which are part of the stylistic language of the facade’s decoration, are indicated in Figure 64. The comparison is done on purpose for samples collected from one building so that a potential variability in the original materials used, mortar formulation and execution techniques are kept to a minimum.

Figure 64: Trade Academy of Krakow, Poland, 1904-1905, Jan Zawiejski. Sampling spots from decorative elements produced with various application techniques.

Several observations are apparent from the comparison. The bimodal distribution of pore sizes point to a coexistence of fine pores with a threshold diameter below 0.1 μm and of larger pores with diameters between 0.8-2 μm typically observed. For one sample, collected from a cast ornament, only one peak is observed, however, one can hardly speak about a unimodal pore size distribution as the peak is very broad and is a result of the superposition of a range of pore sizes between 0.02 and more than 1 μm. The prevailing bimodal or multi-modal distributions of pore sizes points to the coexistence of well-reacted cement matrix and large pores resulting from the evaporation of excess water added to the mortar prepared for run in-situ application.

133

Figure 65: The differential volume of intruded mercury versus pore diameter for samples collected from various decorative elements of the Trade Academy, Krakow, Poland.

Favourable wet-curing conditions, limiting water evaporation and favouring a good progress of the hydration reaction could have occurred, especially on the hydration of large castings formed by applying Roman cement mortars to moulds; historic textbooks always recommend storage of the castings at humid conditions after their removal from the mould.

It is not accidental that a considerable shift of the threshold pore diameter, indicating an extensive formation of C-S-H gel filling the larger pores, was observed just for a cast architectural detail. Further evidence is provided by the analyses of samples collected from the free-standing large casting of a lion, sited over the entrance to White Lion House, in

Eye, Suffolk, UK (Figure 66). The casting was hollow formed with a cast thickness that varies from about 25 mm to 100 mm in the thicker areas around the head and the mane.

Samples were collected from the outer surface, exposed to the external environment after removal from the mould, as well as from the inner surface of the hollow interior of the casting. The mercury intrusion curves of all samples (Figure 67) exhibit a single peak or a narrow cluster of peaks at a small pore diameter of around 0.1 m or below. The inner

134 surface, when the wet-curing conditions must have prevailed for the longest time, showed the densest well-hydrated microstructure.

Figure 66: Casting of a lion, White Lion House, Eye, Suffolk, UK (mid-nineteenth century).

Figure 67: Differential volume of intruded mercury versus pore diameter of mortars for samples collected in the cross section of a large hollow casting of a lion.

It was usual to produce Roman cement stuccoes in two or more coats, the inner coat being a coarse-grained ‘core’ on which a fine-grained thinner finish layer was applied.

Figure 68 shows a large architectural casting – a console – from an office building in

Vienna, dated 1882-3. The pore structures in the inner and outer zones of the casting are shown in Figure 69. The outer part, strongly exposed to air, exhibits large pores with a

135 threshold diameter of around 3 m, a result of incomplete hydration interrupted by the evaporation of water. In contrast, the threshold diameter of the pores in the core of the same element has shifted to lower values, which indicates a more fully hydrated material.

Figure 68: Office building, Vienna, 1882- 1883, Emanuel Trojan von Bylanow. A console.

Figure 69: The differential volume of intruded mercury versus pore diameter for samples of coarse-grained core and fine surface layer collected from a cast console. Office building in Vienna, 1882-1883.

The threshold pore diameter of the ‘air’ pores, resulting from the evaporation of the excess water, generally did not exceed 3 m. The larger pores were only observed in severely deteriorated samples of Roman cement mortars and they clearly were a result of

136 secondary weathering. A sample from a collapsed concrete floor of Castle House in

Bridgewater, Somerset, UK, serves as an example of considerable secondary porosity due to severe deterioration (Figure 70).

Figure 70: Differential volume of intruded mercury versus pore diameter collected from a deteriorated concrete floor - Castle House, Bridgewater, Somerset, UK, 1851.

7.2.3 The influence of different porous substrates on the hydration of mortars

When bricks or old mortars absorb water from repair mortar, its hydration will be reduced in the same way as on exposure to dry external conditions. Therefore, pre- wetting masonry or old Roman stucco before repair mortar is placed is strongly recommended, similarly to the application of other types of mortars. Figure 71 illustrates the effect of different substrates and their pre-wetting on the pore structure of the hardened repair mortars. Two different substrates were taken on which repair mortar can be applied: brick and historic Roman cement mortar. Both are very porous, attaining porosity values of about 30-40% (Table 21). A significant absorption of water by the highly absorbent dry porous substrate is observed when mortar is applied. The threshold pore size shifts to a low value of 0.1 m as the w/c ratio effectively decreases due to the

137 absorption of water not consumed in the initial hydration reaction by the dry substrate

(Figure 71).

In contrast, the water saturated substrate does not absorb water from the repair mortar and thus the threshold pore size of 0.5 is observed, likewise for the wet-air cured mortar.

Figure 71: Differential volume of intruded mercury versus pore diameter of repair mortar laid on substrate; the effect of its pre-wetting on pore size distribution.

7.2.4 The influence of water repellent treatment on hydration

The following experimental procedure was adopted to simulate the situation during restoration work on the facades rendered with Roman cement mortars: Roman cement mortar was applied on brick. After one day of hydration under wet-air curing conditions, the sandwich was dried to stop hydration and to remove water from the mortar pore system. The specimen’s edges were covered with a sealant, the compact surface layer of the mortar was removed by scratching and water-repellent emulsion was applied. Immediately after being thoroughly soaked, the specimen was stored under

100% relative humidity, i.e. wet-air curing conditions.

The development of the pore structure in such mortar is shown in Figure 72. The restriction of hydration can be clearly seen as the mortar shows the same structure as on

138 exposure to a dry external environment, even after 28 days of wet-air curing. Only a small shift in the intrusion peak position is observed, possibly a consequence of the small amount of water introduced with the water repellent emulsion itself.

Figure 72: The differential volume of intruded mercury versus pore diameter of repair mortar after water-repellent treatment and wet-air curing for 28 days.

139 7.3 The influence of hydration on the adhesion of Roman cement repair mortars

One of the principal aims of this study has been to identify scientifically-sound criteria for the optimum formulation and application of Roman cement mortars which could produce compatible and durable repairs of the original substrates (Delgado

Rodrigues and Grossi, 2007). Good adhesion between the original and the repair material is another vital condition necessary for the proper execution of conservation work because any factor reducing the adhesive strength of the repair can make otherwise good repair material incompatible.

Pull-off strength measurements are recommended for the assessment of the adhesive strength between concrete substrates and repair materials (Jacquerod, et al.,

1992, Czarnecki et al., 2006). In this study, the adhesion of the repair materials at different stages of hydration was determined for mortars of different composition and curing process. The Roman cement repair tested was 1.5 cm thick render layer. The mortar was cured under ideal moist-air conditions as well as under 75% relative humidity to simulate a drier, real-world environment. As different repair works require mortars of varying consistency, the effect of water-to-cement ratio on adhesion was also investigated.

The effect of polymer additives to mortars on adhesion was also investigated as they are commonly used in the contemporary conservation of buildings to improve not only adhesion but also the workability and durability of the repair materials (Stoch et al.,

1999, Aggarwal et al., 2007, Park et al., 2008). The modification of mortars by polymers is especially important when the repairs are laid on surfaces that are soiled or impregnated with different agents coming from paint layers with which they were coated.

Finally, the adhesion between subsequent layers of fresh mortar was investigated, as many large repairs or reconstructions require a stepwise application of the mortar. The

140 main concern of the craftsmen is good adhesion between all layers forming a repair and time limit between the applications of the subsequent layers which ensures the proper binding of all mortar-layers.

7.3.1 The adhesion of Roman cement repairs and the effect of curing conditions

The adhesion of repair mortars has been studied on composite samples (Figure 76) which were prepared by applying a repair mortar onto the historic Roman cement substrates coming from Neue Hofburg in Vienna according to the procedure described earlier in Chapter 5.3. The repair was produced using mortar prepared from Folwark

Roman cement 830°C/650min of the composition given in Chapters 5.3.

The pull-off strength profiles of repairs are shown in Figure 73 for two curing conditions. In all tests, the substrates were pre-wetted by soaking in water. A clear correlation between adhesion and curing time, equivalent to the degree of hydration, is observed; the adhesion increases significantly during the first 14 days of hydration when a pull-off strength level of 1.3 MPa is attained for the moist-cured specimen. The parameter further increases during the following six months, attaining approximately 2.0 MPa.

The influence of curing conditions on the pull-off strength values can be clearly seen; the proper moist-air curing (Chapter 5.3) produces significantly better adhesion when compared to curing under 75% RH. However, the profiles of the pull-off strength development are the same for both curing conditions. If one assumes that adhesion is closely related to the degree of hydration, this observation is at odds with results presented in Chapter 7.2.1 where the mortars did not show any progress of hydration when exposed to dry-air conditions. The explanation of this discrepancy can be pre-wetting of the substrate which provides water for hydration at the substrate-repair interface for a long time even during dry-air curing. The effect of the substrate pre-wetting on adhesion is

141 illustrated in Figure 74; the pull-off strength significantly increases when the substrate is adequately pre-wetted.

Figure 73: The pull-off strength of Roman cement repair on historic Roman cement substrate from Neue Hofburg, Vienna, cured at various conditions. The substrate was pre-wetted by soaking in water.

Figure 74: The effect of pre-wetting on the pull-off strength of repair on historic Roman cement substrate.

7.3.2 The influence of mortar composition on the adhesion

Increasing the w/c ratio in the repair mortars lowers their adhesion in the same way as curing in dry environments. Figure 75 compares the adhesion development profiles for mortars prepared at two w/c ratios of 0.6 and 1.0, respectively, studied for three months of hydration. Mortars with w/c=1.0 exhibit comparatively lower values. The observation can

142 be explained by the lower strength of more porous mortars formed from water-rich formulations as discussed in detail in Chapter 7.1.5.2.

Figure 75: Adhesion of repair mortars of varying w/c ratio on water-soaked historic Roman cement substrate.

7.3.3 Adhesion of subsequent layers of fresh mortar

When large cavities need to be filled or large elements reconstructed using Roman cement mortars, several layers of mortars are usually applied stepwise. The old wisdom of practical conservation requires that the layers should be laid ‘fresh-in-fresh’ as applying the subsequent layer on the substrate already hardened may compromise adhesion and, thus, adversely affect the quality of the whole repair. The problem was systematically investigated by applying a layer of Roman cement mortar on a substrate made of the same mortar which was stored at moist conditions until setting. The adhesion of the repair layer was then measured as the pull-off strength after increasing time intervals - between 5 and

300 minutes (Figure 76). The results obtained are shown in Figure 77.

It is clearly seen that the pull-off strength starts to decrease shortly after the setting of the substrate and the adhesion of one layer to another is continuously reduced with increasing time between the applications.

143 a) 5 min b) 20 min c) 300 min

Figure 76: Specimens prepared by applying a layer of a Roman cement mortar on the substrate made of the same mortar, moist-air cured at increasing times a) 5 minutes, b) 20 minutes c) 300 minutes. After 300 minutes, a visible borderline between the two layers is observed.

Figure 77: The pull-off strength of the Roman cement mortar layer laid on the substrate made of the same mortar, moist-air cured at increasing time.

The maximum drop from 1.8 MPa to about 1.1 MPa occurs in the first 30 minutes. The observation can be explained by the formation of a smooth, compact, “glassy” surface on the top of the hardened substrate, which acts as a separation layer reducing the adhesion of the subsequent coat. After a long curing time of the substrate, the separation layer is visible as boundary lines in the cross-section of the two-layer sandwich.

These tests confirm the recommendation of the restoration practice that subsequent layers of a mortar should be applied within the shortest possible time intervals (‘fresh-in-

144 fresh’ application). If the requirement is not met and the mortar has set, its surface should be roughened before the next layer is laid or a reinforcement of the repair with metal armature should be considered.

145 Conclusions

This study has focused on the mechanism of the hydration process that occurs in a range of natural cements - belite cements calcined at low temperature - which match the characteristics of the historic binders used in the nineteenth and early twentieth centuries, predominantly to decorate building facades. The hydration process of the natural cements studied involves two principle reactions:

- formation of crystalline AFm and AFt phases, which dominates the initial stage of

hydration during which the pastes attain their initial strength values, and which may

continue for eight weeks,

- hydration of belite resulting in the formation of amorphous calcium silicate hydrates –

the C-S-H gel – which dominates the late stage of hydration and accounts for the

materials final high strength and excellent durability.

Three AFm phases - calcium aluminum oxide carbonate hydroxide hydrates and calcium aluminate hydroxide hydrates with general formulas C4A C 0.5H12, C4AH13 and

C4A C H11 - are formed. One AFt phase – ettringite (calcium aluminum sulfate hydroxide hydrate, C6AS 3H32) - is formed in smaller amounts on hydration of cements containing

SO3. As no crystalline aluminate phase was observed in the cements, the source of aluminates is believed to be within the amorphous phase. The amounts of the AFm phases formed decrease with the decrease in alumina content of the amorphous phase, in particular the effect of the reduction in amorphous alumina due to the crystallisation of gehlenite at the more intense calcination conditions was observed. The formation of AFm phases is also suppressed by the addition of retarders, which extends the workable time and enables a proper handling and application of the pastes and mortars in the conservation practice.

146 Two polymorphs of belite exist in natural cements:  polymorph readily formed on calcination at higher temperatures and ´ which is more stable in cements calcined at lower temperatures. It is impossible to determine which variety is more active in the hydration process as other parameters such as crystal size and surface area of the polymorphs play a crucial role in the belite reactivity. However, high rate of hydration correlates with ‘mild’ firing conditions due to the presence of poorly-crystalline, highly-

/ dispersed belite by coincidence present mainly as  -C2S. In contrast, the cements show longer dormant periods with increasing calcination temperatures.

The above hydration mechanism is reflected in a characteristic development of the pore structure in the natural cement pastes. At early ages, relatively open pore structure is produced by the quick growth of the AFm phases in the pastes with the threshold pore diameter between 0.2-0.8 m. The initial open structure remains unchanged during the dormant period of pastes which can extend up to several weeks. Only then does the threshold pore width shift to smaller values concentrated around 0.02 m, which is the result of filling larger pores by the formation of the C-S-H gel.

The evolution of the pore structure correlates with the development of the water vapour accessible surface area. In the case of the Roman cements, surface area of the early structure produced by the quick growth of the C-A-H phases does not exceed approximately 20 m2/g. Formation of the C-S-H gel brings about an increase of the surface area up to 120 m2/g.

The above hydration mechanism determines the main features of the natural cements which made them historically favoured materials for the easy and economic manufacturing of renders and decorative elements on the external facades of buildings – quick setting, combined with high initial and long-term strength and open pore structure.

147 Indeed, an intensive study of pore structure of historic Roman cement mortars collected from historic buildings across Europe has revealed their high porosity varying between 20 and 45%, which is similar to values found for aerial or hydraulic lime mortars.

Furthermore, the historic Roman cement mortars have rarely been found to develop a dense microstructure characteristic of the abundant formation of the C-S-H gel under ideal condition of moist-curing; massive castings being the only exception identified. The presence of much larger ‘air’ pores, resulting from the evaporation of excess water not consumed during the process of hydration, was in turn almost universally observed.

Further investigations have revealed that the restricted hydration could be due to a number of processes:

- exposure of the freshly laid surface to dry real-world external environments,

- high water-to-cement ratio in the original mortars,

- drawing of water from the stucco mass due to insufficient pre-wetting of the porous

masonry or old mortars.

A further important aim of this study was to gain better insight into the compatibility between the historic and repair mortars where compatibility is broadly defined as a capacity of the repair mortar to interact with the original historic material without causing any damage nor introducing an artificial interface between the original and the repair.

Freshly prepared Roman cement repair materials were found to develop pore structures varying from dense, fine-porous to more open, depending on their w/c ratios and curing conditions, similarly to historic mortars. In general, to attain at least 50% degree of hydration, 14 days of proper wet curing of the repairs is recommended. Otherwise, very weak highly porous mortars with low adhesion to the substrate are produced. Attention should be also paid to a sufficient pre-wetting of the substrate; a significant absorption of

148 water by highly porous substrates like brick or historic stuccoes may remove water needed for mortar hydration. The use of water-repellent solutions or emulsions to protect the facades against rain-water intrusion also hinders the hydration. Thin surface applications such as finishing layers, cement washes or paints are particularly vulnerable to external impacts and require attention in ensuring the proper moist-curing conditions.

Good adhesion between the original and the repair material is another vital condition of producing compatible and durable repairs of the original substrates. The adhesion is predominantly dependent upon the degree of hydration; minimum 14 days of proper curing is required for the repair to gain sufficient adhesive strength. When large cavities need to be filled or large elements reconstructed using Roman cement mortars, the subsequent layers of a mortar should be applied within the shortest possible time intervals (‘fresh-in- fresh’ application), as a smooth, compact, “glassy” surface on the top of the hardened substrate forms otherwise, which acts as a separation layer reducing the adhesion of the subsequent coat.

149 List of Figures

Figure 1 Example of a façade decorated with Roman cement stucco, Trade Academy in Krakow, Poland, 1904-1905, Jan Zawiejski.

Figure 2 Sheppey septaria, UK. Nodules of clayey limestone in clay beds used for the original production of Roman cement.

Figure 3 The quarry of marls, Lilienfeld, Austria. Typical quarry of raw material used for the production of Roman cement in mainland Europe.

Figure 4 Batteries of vertical continuous kilns used for the production of American natural cement. Round Top Hydraulic cement company, Washington County, United States.

Figure 5 Stucco reinforcement with iron wires; the inner part of castings was usually hollow.

Figure 6 Joint mortars and renders, typical applications of American natural cements.

Figure 7 a) Cross section of Roman cement stucco covered with several later coatings of paint, b) surface layer on Roman cement stucco altered to 3 mm depth.

Figure 8 Deterioration phenomena for Roman cement mortars: a) fine shrinkage cracks characteristic of Roman cement mortars, b) corrosion of iron elements causing the disintegration of the host mortar, c) erosion of Roman cement mortar surface.

Figure 9 Incompatible repairs of Roman cement mortars: a) the improper use of Portland cement, which causes the damage of Roman cement stuccoes, b) Roman cement casting disfigured by a thick coating of paint layers.

Figure 10 Compressive strength development profiles of Roman cement pastes, w/c 0.65.

Figure 11 The effect of water-repellent treatment.

Figure 12 Roman cement surface applications: a) finish layer, b) and wash.

Figure 13 PosiTest® AT for the adhesion measurement.

Figure 14 X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65) at increasing curing time.

150 Figure 15 Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Folwark Roman cement 830oC/650min (w/c=0.65).

Figure 16 X-ray diffraction patterns of Folwark Roman cement 960oC/300min and its paste (w/c=0.65) at increasing curing time.

Figure 17 Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Folwark Roman cement 960oC/300min (w/c=0.65).

Figure 18 X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65) at increasing curing time.

Figure 19 Integrated intensity of the main diffraction maxima of AFm phases as a function of the curing time – paste of Lilienfeld Roman cement 920oC/300min (w/c=0.65).

Figure 20 X-ray diffraction pattern of Prompt cement and its paste (w/c=0.65) at increasing curing time.

Figure 21 X-ray diffraction pattern of Rosendale cement and its paste (w/c=0.65) at increasing curing time.

Figure 22 X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65), modified by the addition of 0.4% of citric acid, at increasing curing time.

Figure 23 Integrated intensity of the main diffraction maxima of the hydrated phases as a function of the curing time – paste of Folwark Roman cement 830oC/650min (w/c=0.65), modified by the addition of 0.4% of citric acid.

Figure 24 X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65), modified by the addition of 0.4% of citric acid, at increasing curing time.

Figure 25 Integrated intensity of the main diffraction maxima of hydrated phases as a function of the curing time – paste of Lilienfeld Roman cement 920oC/300min (w/c=0.65), modified by the addition of 0.4% of citric acid.

Figure 26 X-ray diffraction patterns of Harwich cement 890oC/500min and its paste (w/c=0.65) at increasing curing time.

Figure 27 X-ray diffraction patterns of Sheppey cement 890oC/500min and its paste (w/c=0.65) at increasing curing time.

Figure 28 X-ray diffraction patterns of Folwark Roman cement 830oC/650min and its paste (w/c=0.65) at increased curing time.

151 Relative intensities of the diffraction maxima of calcite (C) and belites (B).

Figure 29 X-ray diffraction patterns of Folwark Roman cement 960oC/300min and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite (C), gehlenite (G) and belites (B).

Figure 30 X-ray diffraction patterns of Lilienfeld Roman cement 920oC/300min and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite C), gehlenite (G) and belites (B).

Figure 31 X-ray diffraction patterns of Prompt cement and its paste (w/c=0.65) at increasing curing time. Relative intensities of the diffraction maxima of calcite (C), gehlenite (G) and belites (B).

Figure 32 X-ray diffraction patterns of Rosendale cement and its paste (w/c=0.65) at increasing curing time. Relative intensities of diffraction maxima of calcite (C), gehlenite (G), belites (B) and periclase (Pe).

Figure 33 X-ray diffraction patterns of the Folwark 830°C/650min wash layer (w/c=1.0), at increasing curing time.

Figure 34 X-ray diffraction patterns of the Folwark 830°C/650min wash layer (w/c=0.7), with addition of citric acid (0.3%) at increasing curing time.

Figure 35 Mass change with the outgassing time.

Figure 36 Adsorption-desorption isotherms of water vapour on Lilienfeld 920°C/300min Roman cement paste cured for 28 days, outgassed for 3.5 and 42 h. EMC - equilibrium moisture content, p/po - relative pressure of water vapour. Figure 37 EMC – t plots of the adsorption data shown in Figure 36. EMC – equilibrium moisture content, t – statistical thickness of the adsorbed water layer.

Figure 38 Adsorption-desorption isotherms of water vapour for Lilienfeld 920°C/300min Roman cement paste cured at various ages. Definitions of EMC and p/po as in Figure 36.

Figure 39 EMC – t plot of the adsorption data shown in Figure 38. Definitions of EMC and t as in Figure 37.

Figure 40 Evolution of the BETH2O surface area of three studied Roman cement pastes with hydration time.

Figure 41 BETH2O surface area of different Roman cement applications

152 prepared from Folwark 830°C/650min (w/c=0.65), 0.3% of citric acid as a retarder was used.

Figure 42 Thermogravimetric curves of Folwark Roman cement 830°C/650min and its paste (w/c=0.65) at increasing curing time.

Figure 43 DTA thermograms of Folwark Roman cement 830°C/650min and its paste (w/c=0.65) at increasing curing time.

Figure 44 Thermogravimetric curves of Lilienfeld Roman cement 860°C/300min and its paste (w/c=0.65) at increasing curing time.

Figure 45 DTA thermograms of Lilienfeld Roman cement 860°C/300min and its paste (w/c=0.65) at increasing curing time.

Figure 46 Composition of Folwark Roman cement 830°C/650min paste as a function of curing time.

Figure 47 Composition of Lilienfeld Roman cement 860°C/300min paste as a function of curing time.

Figure 48 Chemically bound water wb,,t versus curing time for pastes of Folwark 830°C/650min and Lilienfeld 860°C/300min cements.

Figure 49 Degree of hydration  versus curing time for pastes of Folwark 830°C/650min and Lilienfeld 860°C/300min cements.

Figure 50 DTA thermograms of the Folwark 860°C/300min wash layer (w/c=0.70), with addition of citric acid (0.3%).

Figure 51 DTA thermograms of the Folwark 860°C/300min wash layer (w/c=1.0).

Figure 52 SEM/EDX analysis of Lilienfled 920°C/650min Roman cement clinker.

Figure 53 SEM/EDX analysis of hardened Lilienfeld 920°C/300min Roman cement paste at different curing periods (w/c=0.65).

Figure 54 SEM/EDX analysis of hardened Lilienfeld 920°C/650min Roman cement paste cured for one year (w/c=0.65).

Figure 55 Differential volume of intruded mercury versus pore diameter for Roman cement pastes (w/c=0.65) at increasing curing time: a) Lilienfeld 920°C/300min, b) Folwark 960°C/300min, c) Harwich 890°C/500min, d) Sheppey 890°C/500min.

Figure 56 Differential volume of intruded mercury versus pore diameter for Prompt and Rosendale cement pastes (w/c=0.65) at increasing curing time.

153 Figure 57 Differential volume of intruded mercury versus pore diameter for the Folwark 830°C/650min paste (w/c=0.65) pure and with the addition of 0.4% of citric acid as a retarder, at increasing curing time.

Figure 58 Differential volume of intruded mercury versus pore diameter for OPC (w/c=0.65) at increasing curing time.

Figure 59 Differential volume of intruded mercury versus pore diameter for Folwark 830°C/650min cement pastes of various w/c ratios cured for: a) one day, b) 14 days, c) three months.

Figure 60 Differential volume of intruded mercury versus pore diameter for Roman cement mortars cured at various conditions: a) wet-air curing, b) water-curing, c) air curing.

Figure 61 Rehydration of air-cured mortar.

Figure 62 Differential volume of intruded mercury versus pore diameter for a sample of a cast element from Neue Hofburg in Vienna, Austria. Rehydration by re-wetting.

Figure 63 Differential volume of intruded mercury versus pore diameter for a 5 mm thin mortar layer at increasing time of moist curing.

Figure 64 Trade Academy of Krakow, Poland, 1904-1905, Jan Zawiejski. Sampling spots from decorative elements produced with various application techniques.

Figure 65 The differential volume of intruded mercury versus pore diameter for samples collected from various decorative elements of the Trade Academy, Krakow, Poland.

Figure 66 Casting of a lion, White Lion House, Eye, Suffolk, UK (mid- nineteenth century).

Figure 67 Differential volume of intruded mercury versus pore diameter of mortars for samples collected in the cross section of a large hollow casting of a lion.

Figure 68 Office building, Vienna, 1882-1883, Emanuel Trojan von Bylanow. A console.

Figure 69 The differential volume of intruded mercury versus pore diameter for samples of coarse-grained core and fine surface layer collected from a cast console. Office building in Vienna, 1882-1883. Figure 70 Differential volume of intruded mercury versus pore diameter collected form a deteriorated concrete floor – Castle House, Bridgewater, Somerset, UK, 1851.

Figure 71 Differential volume of intruded mercury versus pore diameter of

154 repair mortar laid on substrate; the effect of its pre-wetting on pore size distribution.

Figure 72 The differential volume of intruded mercury versus pore diameter of repair mortar after water-repellent treatment and wet-air curing for 28 days.

Figure 73 The pull-off strength of Roman cement repair on historic Roman cement substrate from Neue Hofburg, Vienna, cured at various conditions. The substrate was pre-wetted by soaking in water.

Figure 74 The effect of pre-wetting on the pull-off strength of repair on historic Roman cement substrate.

Figure 75 Adhesion of repair mortars of varying w/c ratio on water-soaked historic Roman cement substrate.

Figure 76 Specimens prepared by applying a layer of a Roman cement mortar on the substrate made of the same mortar, moist-air cured at increasing times a) 5 minutes, b) 20 minutes c) 300 minutes. After 300 minutes, a visible borderline between the two layers is observed.

Figure 77 The pull-off strength of the Roman cement mortar layer laid on the substrate made of the same mortar, moist-air cured at increasing time.

155 List of Tables

Table 1 Setting time and strength for Roman and Portland cement mortars as specified by the Austrian standards of 1880 and 1890.

Table 2 Fineness, setting time and strength for American natural cements as given by several historic specifications.

Table 3 Mineralogy of selected source material used for Roman cement production.

Table 4 Physical and mechanical properties of historic Roman cement mortars.

Table 5 Tensile strength of mortars from American natural cements (Eckel, 1905).

Table 6 Main hydrated phases detected on OPC hydration.

Table 7 Chemical composition of cements investigated.

Table 8 Mineralogical composition of natural cements investigated.

Table 9 Setting time of Roman cement pastes investigated, w/c 0.65, CA – citric acid.

Table 10 Setting time of Prompt and Rosendale natural cement pastes investigated, w/c 0.65.

Table 11 Compressive strength of Prompt cement mortars (Vicat, France).

Table 12 Selected historic Roman cement mortars.

Table 13 AFm phases identified in the Roman cement pastes studied.

Table 14 Specific surface area of Roman cement pastes cured at various times as determined by different experimental methods.

Table 15 Specific surface area of Roman cement mortars cured at various times as determined by water vapour adsorption.

Table 16 Weight losses measured by TG analysis for Folwark 830°C/650min paste.

Table 17 Weight losses measured by TG analysis for Lilienfeld 860°C/300min paste.

156

Table 18 Chemically bound water content wb,t and degree of hydration  for the Folwark 830°C/650 min paste.

Table 19 Chemically bound water content wb,t and degree of hydration  for the Lilienfeld 860°C/300min paste.

Table 20 Degree of hydration  of Folwark 830°C/650min wash layers, w/c=0.7 and 1.0.

Table 21 Characteristics of selected historic Roman cement mortars.

157 List of Equations

Eq.1 CI = (2.8*SiO2 + 1.1*Al2O3 + 0.7*Fe2O3)/(CaO + 1.4*MgO)

Eq.2 d = -4  cos P

p 1 C 1 p Eq.3 4  4 4 n(p0 p) nm C nm C p0

Eq.4 V ads tp(/ po ) SBET

  Eq.5 tp( / poo ) 3.85 1.89ln( ln( p / p ))

Eq.6 Ca(OH)2 CaO + H2O

Eq.7 CaCO3 CaO + CO2

Eq.8 w  b,t wb,

 Eq.9 wwb, t b total(1 ) w b cement

Eq.10 w (%) LOI  m  m b,total paste CaCO3 Ca(OH )2

Eq.11 w (%) LOI  m  m b,cement cement CaCO3 Ca(OH )2

Eq.12 ww  b total b cement  wwbb  cement

158 References

Aggarwal, L.K., Thapliyal, P.C., Karade, S.R. (2007). Properties of polymer- modified mortars using epoxy and acrylic emulsions. Con Build Mat; 21: 379-383.

Arioglu, N., Acun, S. (2005). A research about a method for restoration of traditional lime mortars and plasters: A staging system approach. Build Env; 41: 1233-1230.

ASTM Standard C 10 (1974). Standard Specification for Natural Cement. Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA.

Badmann, R., Stockhausen, N., Setzer, M.J. (1981). The statistical thickness and the chemical potential of adsorbed water films. J Colloid and Interface Sci; 82: 534-543.

Beaudoin, J.J., Ramachandran, V.S. (1992). A new perspective on the hydration characteristics of cement phases. Cem Con Res; 22: 689-694.

Bensted, J. (1983). Early hydration of Portland cement – effects of water/cement ratio. Cem Con Res; 13: 493-498.

Bensted, J. (2002). Calcium aluminate cements. In: J. Bensted, P. Barnes (Eds.), Structure and Performance of Cements, Spon Press, London. 114-139.

Bernardo, G., Telesca, A., Valenti, G.L. (2006). A porosimetric study of calcium sulfoaluminate cement pastes cured at early ages. Cem Con Res; 36: 1042-1047.

Bonen, D., Diamond, S. (1994). Interpretation of Compositional Patterns Found by Quantitative Energy Dispersive X-ray Analysis for Cement Paste Constituents. J Am Ceram Soc; 77, 7: 1857-1882.

Brunauer, S., Emmet, P.H., Teller, E. (1940). Adsorption of gases in multimolecular layers. J Am Chem Soc; 62: 723.

Cailleux, E., Marie-Victoire, E., Sommain, D. (2007). A study of natural cement from the French Rhone-Alps region, In: R.Fort, M. Alvarez de Buergo, M. Gomez-Heras, C. Vazquez-Calmo, Proceedings of the International Heritage, Weathering and Conservation Conference, Vol 1, Taylor and Francis, London, 77-84.

Callebaut, K., Elsen, J., Van Balen, K., Viaene, W. (2001). Nineteenth century hydraulic restoration mortars in the Saint Michael´s Church (Leuven, Belgium). Natural hydraulic lime or cement. Cem Con Res; 31: 397-403.

Cambrera, J., Rojas, M.F. (2000). Mechanism of hydration of the metakaolin-lime- water system. Cem Con Res; 31: 177-182.

Chen, J.J., Thomas, J.J., Taylor, H.F.W., Jennings, H.M. (2004). Solubility and structure of calcium silicate hydrate. Cem Con Res; 34: 1499-1519.

Christensen, A.N., Jensen T.R., Hanson, J.C. (2004). Formation of ettringite Ca6Al2(SO4)3(OH)12.26H2O, AFt and monosulfate, Ca4Al2O6(SO4).14H2O, AFm-14, in hydrothermal hydration of Portland cement and of calcium aluminum oxide-

159 calcium sulfate dihydrate mixtures studied by in situ synchrotron X-ray powder diffraction. J Sol State Chem; 177: 1944-1951.

Cody, A.M., Lee, H., Cody, R.D., Spry, P.G. (2004). The effects of chemical environment on the nucleation, growth, and stability of ettringite [Ca3Al(OH)6]2(SO4)3.26H2O. Cem Con Res; 34: 869-881.

Cook, R.A., Hover, K.C. (1999). Mercury porosimetry of hardened cement pastes. Cem Con Res; 29: 933-943.

Copeland, L.E., Hayes, J.C. (1953). Determination of non-evaporable water in hardened Portland-cement paste. ASTM Bull; 194: 70–74.

Copeland, L.E., Kantro, D.L., Verbeck, G.J. (1960). Chemistry of hydration of Portland cement. In: Proc. Symp. Cem. Chemistry; Vol. 1. Cement and Concrete Association, Washington: 429-468.

Cottin, B. (1992). The first reactions in cement hydration. In: Hydration and Setting of Cements. Nonat, A., Muttin, J.C., Eds. RILEM E.& F.N. Spon, London. 89-100.

Czarnecki, L., Garbacz, A., Krystosiak, M. (2006). On the ultrasonic assessment of adhesion between polymer coating and concrete substrate. Cem Con Comp; 28: 360- 369.

De La Torre, A.G., Bruque, S, Aranda, M.A.G. (2001). Rietvield quantitative amorphous content analysis. J Appl Cryst; 34: 196-202.

Delgado Rodrigues, J., Grossi, A. (2007). Indicators and ratings for the compatibility assessment of conservation actions. J Cult Heritage; 8: 32-43.

Diamond, S. (1971). A critical comparison of mercury porosimetry and capillary condensation pore size distributions of Portland cement pastes. Cem Con Res; 1: 531-445.

Diamond, S. (1998). Concrete porosity revisited. In: Materials Science of Concrete: The Sidney Diamond Symposium, John Wiley & Sons.

Diamond, S. (2000). Mercury intrusion. An inappropriate method for the measurement of pore size distributions in cement-based materials. Cem Con Res; 30: 1517-1525.

Diamond, S. (2004). The microstructure of cement paste and concrete – a visual primer. Cem Con Comp; 26: 919-923.

Dweck, J., Buchler, P.M., Coelho, A.C.V., Cartledge, F.K. (2000). Hydration of a Portland cement blended with calcium carbonate. Therm Acta; 346: 105-113.

Eckel, E.C. (1905). Cements, limes and plasters, 2nd ed., John Wiley & Sons, New York.

Escalante-Garcia, J.I. (2003). Nonevaporable water from the neat OPC and replacement materials in composite cements hydrated at different temperatures. Cem Con Res; 33: 1883-1888.

160 Famy, C., Brough, A.R., Taylor, H.F.W. (2003). The C-S-H gel of Portland cement mortars: Part I. The interpretation of energy-dispersive X-ray microanalyses from scanning electron microscopy, with some observations on C-S-H, AFm and AFt phase compositions. Cem Con Res; 33: 1389-1398.

Feldman, R.F., Sereda, P.J. (1970). A new model for hydrated Portland cement and its practical implications. Eng J Can; 53: 53–59.

Feldman, R.F., Beaudoin, J.J. (1991). Pretreatment of hardened hydrated cement pastes for mercury intrusion measurement. Cem Con Res; 21: 297-308.

Fischer, R., Kuzel, H.J. (1982). Reinvestigation of the System C4A·nH2O– C4A·CO2·nH2O. Cem Con Res; 12: 517–526.

Friás, M., Cambrera, J. (2000). Pore size distribution and degree of hydration of metakaolin-cement pastes. Cem Con Res; 30: 561-569.

Fukuda, K., Wakamatsu, N., Ito, S. (2001). Improvement in hydration reactivity of -phase belite by remelting reaction. J Am Ceram Soc; 84, 3: 639-641. Gallé, C. (2001). Effect of drying on cement-based materials pore-structure as identified by mercury intrusion porosimetry. Comparative study between oven-, vacuum-, and freeze drying. Cem Con Res; 31: 1467-1477.

Gard, J.A., Taylor, H.F.W. (1976). Calcium silicate hydrate (II) (“C-S-H(II)”). Cem Con Res; 6: 667-678.

Gartner, E.M., Gaidis, J.M. (1989). Hydration mechanism I. In: Materials Science of Concrete, The American Ceramic Society, Westerville, OH.

Gartner, E.M., Young, J.F., Damidot, D.A., Jawed, I. (2002). Hydration of Portland cement. In: J. Bensted, P. Barnes (Eds.), Structure and Performance of Cements, Spon Press, London. 57-113.

Gaultieri, A.F., Viani, A., Montanami, C. (2006). Quantitative phase analysis of hydraulic limes using the Rietvield method. Cem Con Res; 36: 401-406.

Gödicke-Dettmering, T., Strübel, G. (1996). Mineralogische und technologische Eigenschaften von hydraulischen Kalken als Bindemittel für Restaurierungsmörtel in der Denkmalpflege. Giessener Geologische Schriften; 56: 131-154.

Groves, G.W. (1987). TEM studies of cement hydration. Mater Res Soc Symp Proc; 85: 3-12.

Guerrero, A., Goi, S. (2005). Microstructure and Mechanical Performance of Belite Cements from High Calcium Coal Fly Ash. J Am Ceram Soc.; 88, (7): 1845-1853.

Hagymassy, J., Brunauer, S., Mikhail, R.S. (1969). Pore structure analysis by water vapor adsorption, I. t-curves for water vapor. J. Colloid Interface Sci. 29: 485-491.

Hagymassy, J. Jr, Odler, I. , Yudenfreund, M., Skalny, J., Brunauer, S. (1972). Pore Structure Analysis by Water Vapor Adsorption, III. Analysis of Hydrated Calcium Silicates and Portland Cements. J. Colloid Interface Sci; 38: 20-34.

161 Head, M.K., Buenfeld, N.R. (2006). Confocal imaging of porosity in hardened concrete. Cem Con Res; 36: 896-911.

Hong, S-H., Young, J.F. (1999). Hydration kinetics and phase stability of dicalcium silicate synthesized by the Pechini process. J Am Ceram Soc 82, 7, 1681-1686.

Hu, S., Li, Y. (1999). Research on the hydration, hardening mechanism, and microstructure of high performance expansive concrete. Cem Con Res; 29: 1013- 1017.

Hughes, D.C., Jaglin, D., Kozlowski, R., Mayr, N., Mucha, D., and Weber, J. (2007A). Calcinations of marls to produce Roman cement. J ASTM Int, 4, 1. JAI100661. Hughes, D.C., Sugden, D.B., Jaglin, D., Mucha, D. (2007B). Calcination of Roman cement: A pilot study using cement-stones from Whitby. Con Build Mater; 22: 1446- 1455.

Hughes, D.C., Swann, S., Gardner, A. (2007C). Roman cements 1: Its Origins and Properties. J Arch Con; 13, 1: 21-37.

Hughes, D.C., Swann, S., Gardner, A. (2007D). Roman cements 2: A conservation strategy. J Arch Con, 13, 3.

Hughes, D.C., Jaglin, D., Kozlowski, Mucha, D. (2008). Roman cements – belite cements calcined at low temperatures. Cem Conc Res; in press.

Ishida, H., Yamazaki, S., Sasaki, K., Okada, Y., Mitsuda, T. (1993). -dicalcium silicate hydrate: Preparation, decomposed phase, and its hydration. J Am Ceram Soc; 76, 7: 1702-1712.

Jaquerod, C., Chippis, Ch., Alou, F., Houst, Y.F. (1992). Nondestructive testing of repair mortars for concrete. In: Proceedings of the 3rd International Colloquium Materials science and restoration, Eslingen, Germany: 872.

Jennings, H.M. (2000). A model for the microstructure of calcium silicate hydrate in cement paste. Cem Con Res; 30: 101-116.

Jennings, H.M., Thomas, J.J. (2004). A discussion of the paper “The BET-specfic surface area of hydrated Portland cement and related materials” by Ivan Odler. Cem Con Res; 34: 1959-1960.

Jennings, H.M., Thomas, J.J., Gevrenov, J.S., Constantinides, G., Ulm, F.J. (2007). A multi-technique investigation of the nanoporosity of cement paste. Cem Con Res; 37: 329-336.

Jensen, T.R., Christensen, A.N., Ilanson, J.C. (2005). Hydrothermal transformation of the calcium aluminium oxide hydrates CaAl2O4.10H2O and Ca2Al2O5.8H2O to Ca3Al2(OH)12 investigated by in situ synchrotron X-ray powder diffraction. Cem Con Res; 35: 2300-2309.

Juenger, M.C.G., Jennings, H.M. (2001). The use of nitrogen adsorption to assess the microstructure of cement paste. Cem Con Res; 31: 883-892.

162 Justnes, H., Van Gemert, A., Verboven, A., Sellevold, E.J. (1996). Total and External Chemical Shrinkage of low w/c-ratio cement pastes. Adv Cem Res; 8, 31: 121-126.

Khatib., J.M., Mangat, P.S. (2003). Porosity of cement paste cured at 45°C as a function of location relative to casting position. Cem Con Comp; 25: 97-108.

Klisinska, A., Tislova, R., Adamski, G., Kozlowski, R. (2008). Pore structure of historic and repair Roman cement mortars to establish their compatibility. J Cult Her; submitted for publication.

Konecny, L., Naqvi, S.J. (1993). The effect of different drying techniques on the pore size distribution of blended cement mortars. Cem Con Res; 23: 1223-1228.

Kuzel, H.J., Pöllmann, H. (1991). Hydration of C3A in the presence of Ca(OH)2, CaSO4.2H2O and CaCO3. Cem Con Res; 21: 885-895.

Kuzel, H.J., Meyer, H.W. (1992). Effect of carbon dioxide on the hydration of pastes. In: Hydration and Setting of Cements, RILEM, E&FN Spon, London. 137-149.

Kuzel, H.J. (1996). Initial hydration reactions and mechanism of delayed ettringite formation in Portland cements. Cem Con Com; 18: 195-203.

Lachowski, E.E., Diamond, S. (1983). Investigation on the composition and morphology of individual particles of Portland cement paste: 1. C-S-H gel and calcium hydroxide particles. Cem Con Res; 12: 177-185.

Lachowski, Mohan, K., Taylor, H.F.W., Moore, A.E. (1980). Analytical electron microscopy of cement pastes: II. Pastes of Portland cements and clinkers. J Am Ceram Soc; 63, 7-8: 447-452.

Lam, L., Wong, Y.L., Poon, C.S. (2000). Degree of hydration and gel/space ratio of high volume fly ash/cement systems. Cem Con Res; 30: 747-756.

Lanas, J., Pérez Bernal, J.L., Bello, M.A., Alvarez, J.I. (2006). Mechanical properties of masonry repair dolomitic lime-based mortars. Cem Con Res; 36: 951-960.

Lanas, J., Pérez Bernal, J.L., Bello, M.A., Alvarez Galindo, J.I. (2004). Mechanical properties of natural hydraulic lime-based mortars. Cem Con Res; 34: 2191-2201.

Laskar, Md. A.I. (1997). Some aspects of evaluation of concrete through mercury intrusion porosimetry. Cem Con Res; 27: 93-105.

Lea, F.M. (1970). The Chemistry of Cement and Concrete, 3rd edition, Edward Arnold, London.

Lippens, R.C., de Boer, J.H. (1965). Studies on pore systems in catalysts: V. The t method. J Catal; 4: 319-323.

Litvan, G.G. (1976). Variability of the nitrogen surface area of hydrated cement

163 paste. Cem Con Res; 6: 139-143.

Lothenbach, B., Winnefeld, F. (2006). Thermodynamic modeling of the hydration of Portland cement. Cem Con Res; 36: 209-226.

Maravelaki-Kalaitzaki, P., Bakolas, A., Karatsios, L., Kilikoglou, V. (2005). Hydraulic lime mortars for the restoration of historic masonry in Crete. Cem Con Res; 35: 1577-1586.

Marques, S.F., Ribero, R.A., Silva, L.M., Ferreira, V.M., Labrincha, J.A. (2006). Study of rehabilitation mortars: Construction of a knowledge correlation matrix. Cem Con Res; 36: 1894-1902.

Megaw, H.D., Kelsley, C.H. (1956). Crystal structure of tobermorite. Nature; 177: 390-391.

Mitchell, J. (undated). Geological researches around London, Vol 4, Manuscript of the Geological Society Library, London, 32-45.

Mitchell, J. (undated). Geological researches around London. Vol 1. Manuscript of the Geol Soc Lib London. 254 – 258.

Mitchell, L.D., Margeson, J.C. (2006). The effect of solvents on C-S-H as determined by thermal analysis. J Therm Anal Cal; 86: 591-594.

Montgomery, D.G., Hughes, D.C., Williams, R.I.T. (1981). Fly ash in concrete – a microstructure study. Cem Con Res; 11: 591-603.

Moropoulou, A., Bakolas, A., Moundoulas, P., Aggelakopoolou, E., Anagnostopoulou, S. (2005). Strength development and lime reaction in mortars for repairing historic masonries. Cem Con Com; 27: 289-294.

Mosquera, M.J., Silva, B., Prieto, B., Ruiz-Ilerera (2006). Addition of cement to lime- based mortars: Effect on pore structure and vapour transport. Cem Con Res; 36, 9: 1635-1642.

Mouret, M., Bascoul, A., Escadeillas, G. (1997). Study of the degree of hydration of concrete by means of image analysis and chemically bound water. Adv Cem Bas Mat; 6: 109-115.

Mumme, W.G., Hill, R.J., Bushnell-Wye, G., Segnit, E.R. (1995). Rietvield crystal structure refinements, crystal chemistry and calculated powder diffraction data for the polymorphs of dicalcium silicate and related phases. N Jb Miner Abh; 169: 35-68.

Nonat, A., Mutin, J.C. (1992A). Introduction. In: Hydration and Setting of Cements. Nonat, A., Muttin, J.C., Eds. RILEM E.& F.N. Spon, London: 21-25.

Nonat, A., Mutin, J.C. (1992B). From hydration to setting. In: Hydration and Setting of Cements. Nonat, A., Muttin, J.C., Eds. RILEM E.& F.N. Spon, London: 171-190.

Nonat, A. (2004). The structure and stoichiometry of C-S-H. Cem Con Res; 34: 1521- 1528.

164 Odler, I., Abdul-Maula, S. (1984). Possibilities of quantitative determination of the AFt-(ettringite) and AFm-(monosulphate) phases in hydrated cement pastes. Cem Con Res; 14: 133-141.

Odler, I. (2000). Special Inorganic Cements. E&FN Spon, London.

Odler, I. (2003). The BET-specific surface area of hydrated Portland cement and related materials. Cem Con Res; 33: 2049-2056.

Odler, I. (2004). Reply to the discussion by H.M. Jennings and J.J. Thomas of the paper “The BET-specific surface area of hydrated Portland cement and related materials”. Cem Con Res; 34: 1961.

Olson, R.A., Jennings, H.M. (2001). Estimation of C-S-H content in a blended cement paste using water adsorption. Cem Con Res; 31: 351-356.

Ozkul, M.H. (2000). Utilization of citro- and desulphogypsum as set retarders in Portland cement. Cem Con Res; 30: 1755-1758.

Pane, I., Hansen, W. (2005). Investigation of blended cement hydration by isothermal calorimetry and thermal analysis. Cem Con Res; 35: 1155-1164.

Park, D., Ahn, J., Oh, S., Song, H., Noguchi, T. (2008). Drying effect of polymer- modified cement for patch-repaired mortar on constraint stress. Const Build Mater; doi: 10.1016/j.conbuildmat.2007.11.003.

Parker, J. (1796). A certain Cement or Terras to be used in Aquatic and other Buildings, and Stucco Work, British Patent 2120, dated July 1796 to James Parker of Northfleet.

Parrott, L. J. (1981). Effect of drying history upon the exchange of pore water with methanol and upon subsequent methanol sorption behaviour in hydrated alite phase. Cem Con Res; 11: 651-658.

Pecconi, E., Malesani, P., Bellucci, B., Frantini, F. (2005). Artificial stones utilised in Florence historical palaces between the XIX and XX centuries. J Cult Heritage; 6: 227-233.

Péra J., Husson, S., Guilhot, B. (1999). Influence of finely ground limestone on cement hydration. Cem Con Comp; 21: 99-105.

Pérez Méndez, M., Trivio Vázquez, F. (1984). Study of the strength developed by stable carbonated phases in high alumina cement. Cem Coc Res; 14: 161-169.

Plowman, C., Cambrera, J.G. (1984). Mechanism and kinetics of hydration of C3A and C4AF, extracted from cement. Cem Con Res; 14: 238-248.

Poon, C.S., Wong, Y.L., Lam, L (1997). The influence of different curing conditions on the pores structure and related properties of fly-ash cement pastes and mortars. Con Build Mat; 11, 7-8: 383-393.

Poon, C.S., Lam, L., Kou, S.C., Lin, Z.S. (1999). A study on the hydration rate of

165 natural zeolite blended cement pastes. Con Build Mat; 13: 427-432.

Rai, S., Chaturvedi, S., Singh, N.B. (2004). Examination of Portland cement paste hydrated in the presence of malic acid. Cem Con Res; 34: 455-462.

Rarick, R.L., Thomas, J.J., Christensen, B.J., Jennings, H.M. (1996). Deterioration of the nitrogen BET surface area of dried cement paste with storage time. Adv Cem Bas Mat; 3: 72-75.

RENDEC - Decorated Renders around 1900 in Europe (1999). Technological Studies and Principles of Conservation and Restoration. EU-Rendec project. Eds. Bundesdenkmalamt, Wien.

Richardson, I.G., Groves, G.W. (1993). Microstructural and Microanalysis of Hardened Portland Cement Pastes. J Mater Sci; 28: 265-277.

Richardson, I.G. (1999). The nature of C-S-H in hardened cements. Cem Con Res; 29: 1131-1147.

Richardson, I.G. (2004). Tobermorite/jennite- and tobermorite/calcium hydroxide- based models for the structure of C-S-H: applicability to hardened pastes of tricalcium silicate, -dicalcium silicate, Portland cement, and blends of Portland cement with blast-furnace slag, metakaolin, or silica fume. Cem Con Res; 34: 1733- 1777.

Richardson, I.G. (2007). The calcium silicate hydrates. Cem Con Res; 38: 137-158.

Rojas, M. F, Cambrera, J. (2000). Pore size distribution and degree of hydration of metakaolin-cement pastes. Cem Con Res; 30: 561-569.

Rojas, M.F., Cambrera, J. (2002). The effect of temperature on the hydration rate and stability of the hydration phases of metakaolin-lime-water systems. Cem Con Res; 32: 133-138.

Rojas, M.F., Sánchez de Rojas, M.I. (2003). The effect of high curing temperature on the reaction kinetics in MK/lime and MK-blended cement matrices at 60°C. Cem Con Res; 33: 643-649.

Rojas, M.F., Sánchez de Rojas, M.I. (2005). Influence of metastable hydrated phases on the pore size distribution and degree of hydration of MK-blended cements cured at 60°C. Cem Con Res; 32: 1292-1298.

Saad Morsy, M., Abo El-Enein, S.A., Hanna, G.B. (1997). Microstructure and hydration characteristics of artificial pozzolana cement pastes containing burnt kaolinite clay. Cem Con Res; 27: 1307-1312.

Schwarz, W., Sujata, K., Jennings, H.M., Gerdes, A., Sadouki, H., and Wittmann, F.H. (1994). Chemically modified hydration of Portland Cement. Building Materials Reports; 6: 5-48. Aedification Publishers, Freiburg, Germany.

Schwarz, W. (1995). Novel cement matrices by accelerated hydration of the ferrite phase in Portland cement via chemical activation: kinetics and cementitious

166 properties. Adv Cem Bas Mat; 2: 189-200.

Scrivener, K.L., Füllmann, T., Gallucci, E., Walenta, G., Bermejo, E. (2004). Quantitative study of Portland cement hydration by X-ray diffraction/Rietvield analysis and independent methods. Cem Con Res; 34: 1541-1547.

Serry, M.A., Taha, A.S., El-Hemaly, S.A.S., El-Didamony, H. (1984). Metakaolin- lime hydration products. Therm Acta; 79: 103-110.

Silva, P.S., Glaser, F.P. (1990). Hydration of cements based on metakaolin: thermochemistry. Adv Cem Res; 3: 167-177.

Silva, P.S., Glaser, F.P. (1993). Phase relations in the system CaO-Al2O3-SiO2-H2O relevant to metakaolin – calcium hydroxide hydration. Cem Con Res; 23: 627-639.

Sing, K.S.W., Everett, D.H., Haul, R.H.W., Moscou, L., Pierotti, R.A., Rouquerol, J., Siemieniewska, T. (1985). Reporting physisorption data for gas solid systems with special reference to the determination of surface area and porosity. Pure Appl Chem; 57: 603-619.

Skalny, J., Young, J.F. (1980). Mechanism of Portland cement hydration. In: Proceedings of the 7th International Congress on the Chemistry of Cement. Vol.1 Imprimerie Centrale de L´Artois, Arras, France: II-1/3-II/1/45.

Sommain, D. (2007). Prompt Natural cement - The Roman cement of Grenoble. Technical specification. The Louis Vicat Technical Center. Material and Microstructure Laboratory.

Stoch, A., Zdaniewicz, M., Paluszkiewicz, Cz. (1999). The effect of polymethylsiloxanes on hydration of clinker phases. J Mol Structure; 511-512: 319- 325.

Tarnawski, A. (1887). Kalk, Gyps, Cementkalk und Portland-Cement in Oesterreich- Ungarn, Wien, Selbstverlag.

Taylor, H.F.W. (1950). Hydrated calcium silicates. Part I. Compound formation at ordinary temperatures. J Chem Soc; 3682-3690.

Taylor, H.F.W., Howison, J.W. (1956). Relationships between calcium silicates and clay minerals. Clay Miner Bull; 3: 98-111.

Taylor, H.F.W. (1986). Proposed structure for calcium silicate hydrate gel. J Am Ceram Soc; 69: 464-467.

Taylor, H.F.W. (1990). Cement chemistry, Academic Press, New York.

Tislova, R., Kozlowska, A., Kozlowski, R., Hughes, D. (2008). Porosity and specific surface area of Roman cement pastes. Cem Con Res; in press.

Thomas, J.J., Hsieh, J., Jennings, H.M. (1996). Effect of Carbonation on the Nitrogen BET Surface area of hardened Portland cement paste. Adv Cem Bas Mat; 3: 76-80.

167 Thomas, J.J., Chen, J.J., Allen, A.J., Jennings, H.M. (2004). Effects of decalcification on the microstructure and surface area of cement and tricalcium silicate pastes. Cem Con Res; 34: 2297-2307.

Thomas, J.J. (2007). A new approach to modeling the nucleation and growth kinetics of tricalcium silicate hydration. J Am Ceram Soc; 90, (10): 3282-3288.

Ubbríaco, P., Calabrese, D. (1998). Solidification and stabilization of cement paste containing fly ash from municipal solid waste. Therm Acta; 321: 143-150.

Ubbríaco, P., Tasselli, F. (1998). A study of the hydration of lime-pozzolan binders. J Therm Anal; 52: 1047-1054.

Varas, M.J., Alvarez de Buergo, M., Fort, R. (2005). Natural cement as the precursor of Portland cement: Methodology for its identification. Cem Con Res; 35: 2055-2065.

Vedalakshmi, R., Sundara Raj, A., Srinivasan, S., Ganesh Babu, K. (2003). Quantification of hydrated cement products of blended cements in low and medium strength concrete using TG and DTA techniques. Therm Acta; 407: 49-60.

Vyskocilova, R., Schwarz, W., Mucha, D., Hughes, D.C., Kozlowski, R., Weber, J., (2007). Hydration Processes in Pastes of Roman and American Natural Cements, J

ASTM Int; 4, (2), ID JAI100669.

Wang, J., Kalinichev, A.G., Kirkpatrick, R.J., Hou, X. (2001). Molecular modelling of the structure and energeties fo hydrotalcite hydration. Chem Mater; 13: 145-150.

Wathe, J. M. (2005). Structural Performance of Natural Cements. Preprints of 1st American Natural Cement Conference, Rosendale, NY, March 31–April 2.

Weber, J., Gadermayr, N., Kozlowski, R., Mucha, D., Hughes, D.C., Jaglin, D., Schwarz, W. (2007A). Microstructure and mineral composition of Roman cements produced at defined calcination conditions. Mat Char; 58: 1217-1228.

Weber, J., Mayr, N., Bayer, K., Hughes, D.C., Kozlowski, R., Stillhammerova, M., Ullrich, D., Vyskocilova, R. (2007B). Roman Cement Mortars in Europe´s Architectural Heritage of the 19th Century.J ASTM Int;4, (8), JAI100667.

Williamson, R.B. (1972). Solidification of Portland cement. Proc Mater Sci; 15: 189- 286.

168

169