HISTORIC LIME MORTARS COMPOSITION AND TERMINOLOGY FOR RADIOCARBON DATING—CASE STUDIES BASED ON THIN-SECTION PETROGRAPHY AND CATHODOLUMINESCENCE

Abstract

Since the first developments of anthropogenic lime materials radiocarbon (¹⁴C) dating in the 1960s, numerous studies have been undertaken and developed to investigate the topic further. Historic mortars are complex composite and open system materials that can incorporate a large range of components. Due to the complexity of the historic lime mortars composition, they are not part of a routine protocol in most radiocarbon laboratories and reliable dating is not always achieved. A thorough characterization needs to be performed and different preparation methods can be considered as a function of their compositions. A vast range of terms are employed to qualify the lime mortars components and alterations that can possibly have an influence on the dating result. Here, a detailed description of these components and the various terms used is listed. To illustrate this, images obtained by thin-section petrography and cathodoluminescence are presented in addition to radiocarbon results using stepwise acid hydrolysis on Belgian mortars having different provenance, state, age and composition. Depending on the type of aggregate used, the type of binder and its conservation state, the eventual presence of weathering carbonates and the assumed speed of the carbonation process, the reliability of radiocarbon measurements using the stepwise acid hydrolysis technique is discussed and confronted with presumed historical constraints.

INTRODUCTION

Mortars, or more generally anthropogenic lime carbonates, have been used for many centuries to build elements of architecture. The oldest known lime production dates from the Neolithic period or even to the end of the Palaeolithic (Ronen et al. 1991; Kingery et al. 1988; Karkanas 2007; Friesem et al. 2019). The production process of lime mortars consists of heating geological limestone (CaCO₃) to form quicklime (CaO) which is then blended with water to obtain slaked lime [Ca(OH)₂] and mixed with aggregates (often sand) to create the mortar. The hardening is due to a carbonation reaction: atmospheric CO₂ is incorporated to the mortar and gives rise to the formation of calcium carbonate (CaCO₃), then called anthropogenic lime carbonates (Hale et al. 2003; Ringbom 2011). This carbonation process theoretically allows radiocarbon dating of the mortars since the quantity of ¹⁴C absorbed is relative the ¹⁴C quantity in the atmosphere at a certain time. Radiocarbon dating of lime-based mortars was first developed in the 1960s on the basis that ¹⁴C decay should act similarly to living organisms after their death. These first outcomes were very encouraging and the French researchers Jacques Labeyrie and Georgette Délibrias already draw attention on performing microscopic observations to verify the eventual presence of foraminifers (loose microfossils present in the sand used) susceptible to false the date result, since they are made of very ancient carbonates (Labeyrie and Delibrias 1964). Several types of carbonates can be present in mortars because of the raw materials used, the mortar formation process itself, and possible subsequent weathering. They are listed and described in detail later in the manuscript (see the section “Lime Mortar Components and Terminology”).

Intercomparison between laboratories gives consistent results, whatever the technique used, as some samples give reliable results and other samples show a range of dates making them unsuitable for radiocarbon dating using the current common methods available (Hajdas et al. 2017; Hayen et al. 2017). Reference materials such as charcoal or straw extracted from the mortars and dated with ¹⁴C or roof frames allowing dendrochronology and contextual information are needed to confirm the ¹⁴C dates obtained for a mortar sample. Numerous characterization methods have been employed to describe the composition of mortars and identify possible contaminants for radiocarbon dating. The two most common techniques used in mortar dating studies are thin-section petrography and cathodoluminescence microscopy (Daugbjerg et al. 2021b).

Petrographic analysis on resin-impregnated thin sections gives indications on the mineralogical composition, texture and structure of the binder, and on the different types of carbonates present. It also helps identifying the aggregates, the possible presence of additives, the macroporosity and the state of preservation of the mortar. It is an indispensable first step in the characterization of historic mortars (Hughes and Cuthbert 2000; Nawrocka et al. 2005; Elsen 2006; Weber et al. 2013) and has been extensively used for mortar dating studies (Lindroos et al. 2007; Heinemeier et al. 2010; Ortega et al. 2012; Nonni et al. 2013, 2018; Hayen et al. 2016; Addis et al. 2019; Michalska and Pawlyta 2019; Caroselli et al. 2020, 2020). Cathodoluminescence by optical microscopy on uncovered thin-sections is used to distinguish the different types of carbonates and it can highlight the presence of other minerals such as quartz, K-feldspar and plagioclase (Heinemeier et al. 1997, 2010; Lindroos et al. 2007; Ortega et al. 2012; Al-Bashaireh 2013; Lichtenberger et al. 2015; Prevosti et al. 2016; Addis et al. 2019; Toffolo et al. 2019, 2020; Ponce-Antón et al. 2020). A bright red to orange/yellow color is observed for most limestones whereas anthropogenic lime carbonates exhibit a weak, dull tile-red purple to brown luminescence. The color can vary depending on the instrument used (Pagel et al. 2000) and the limitations of the technique are the lack of luminescence of certain limestones and the limited spatial resolution compared to cathodoluminescence by scanning electron microscopy (Lindroos et al. 2007; Toffolo et al. 2020). Image analysis can be applied to both methods to determine the binder aggregate ratio and the relative porosity in the case of petrography (Pecchioni et al. 2014) and evaluate the quantities of the different types of carbonates for cathodoluminescence (Heinemeier et al. 1997; Lindroos et al. 2007; Murakami et al. 2013). Petrography and cathodoluminescence are complementary and the most suitable to determine possible complications for radiocarbon dating, although additional techniques (such as SEM-EDS, FTIR, XRD, TGA/DSC, etc.) can also be applied in a case-by-case approach, if further characterization is needed to fully understand the composition of the mortar.

It is well known that the definition of a word can vary depending on the field of study. Historic lime mortars can be examined in various contexts including archaeology, geology, art history, conservation, restoration, etc. This article aims to give a comprehensive view of lime mortar components, with petrographic and cathodoluminescence photomicrographs, and their potential influence on radiocarbon dating using stepwise acid hydrolysis. The various terms and definitions used are listed and categories are defined in the case of radiocarbon dating of archaeologic or historic lime mortars. Examples on Belgian mortars from Flanders having various composition help illustrating the different constituents and phenomena.

MATERIALS AND METHODS

Materials

The results about the Roman mortar of Tongeren (Belgium) obtained in previous studies (Hayen et al. 2016, 2017; Hajdas et al. 2017) and for three new samples are described in this paper: two samples come from the St Martin’s church in Rutten (Belgium) and one comes from the Ten Bogaerde site in Koksijde (Belgium). The sampling was performed following as much as possible the advice listed by Daugbjerg et al. (2021b).

Occupation in Rutten is attested as far back as the Roman period and possibly up to the Neolithic. The present St Martin’s church is a neo-classical church. It was built in 1844 against the preserved 13th century Romanesque tower. Saint Martin (ca. 316–397 AD) is the protector of soldiers, travellers, horsemen, and patron of many hospitals, as well as of tailors, weavers and beggars. In 2014, the St Martin’s church in Rutten (Belgium) was excavated before the installation of an underfloor heating system. The underground foundations, where the samples sp32 3A and sp162 7B were sampled, were presumably dated to the Merovingian period (480–750 AD).

The Ten Bogaerde site at Koksijde is a monastic grange of the Cistercian abbey of the Dunes, dating back from before 1183 AD. The 13th century medieval barn, built in brick architecture by abbot Nicholas of Baillieul (1232/3–1253 AD), is one of the most imposing buildings preserved, but the building has lost its impressive wooden roof structure in a fire in 1593 AD. The inner floor of the barn has been excavated in 2014. The lost piers of the medieval roof structure have been discovered during the excavations. Two pier types of the original 13th century roof structure could be distinguished: brick piers and piers of reused stone material. Besides that, two brick piers can be related to a renovation (undated between 1250 and 1600 AD) of the original construction. An investigation of the archaeological structures was possible after cleaning up by use of clear water. Afterwards, several mortars were sampled from the outer part of the structures for radiocarbon dating. To exclude contamination by ground water, sheltered places were selected for sampling which resulted in dry mortar samples. In total, nine samples were selected for radiocarbon dating, but only the results obtained for the sample 7002-3 are presented here. The presumed historical date of the latter is estimated around 1232/3–1253 AD (Lehouck and Termote 2016).

Methods

Standard thin-sections (area of maximum 2 × 3 cm², ± 30 µm thickness) were prepared from epoxy-impregnated samples with a yellow fluorescent dye to facilitate the visualization of macropores. The thin-sections were intentionally left uncovered by a cover slip in order to allow cathodoluminescence observations. The petrographic examination was carried out at RICH with a polarising microscope (ZEISS, Axioplan, transmitted light) at 25× to 400× magnification. The photomicrographs were taken with a high resolution digital camera (DeltaPix Invenio 5DII). A series of essential information to retrieve the composition and the crucial features potentially affecting the radiocarbon dating results was set up with different categories and subcategories. For the binder, the type is described as well as if the texture is micritic or in some rare cases micro-sparitic. The state of the lumps is stated, so if they are completely burned, underburned or overburned with their average and maximum sizes and frequency. For the aggregates, their grain size, mineralogy, and general shape are defined. The global mortar appearance is specified, it comprises the homogeneity, macroporosity and the pore structure. Finally, possible admixtures and alterations are listed.

The cathodoluminescence pictures were acquired with an in-house cold-cathode luminescence microscope available at KU Leuven and called Technosyn model 8200, Mark II. The chamber was maintained at ∼0.05 Torr vacuum and the beam width measured 5 mm. The current was kept between 600 to 800 µA to obtain around 3 kV electron beam. The images were taken at 40× magnification with a ProgRes^® C10^plus camera. The observations and photographs were performed both on uncovered thin-sections and on powdered samples with a particle size < 75 µm that are used for radiocarbon dating.

For the radiocarbon dating, the CO₂ extraction was performed with the stepwise acid hydrolysis technique (Van Strydonck et al. 2015; Wojcieszak et al. 2020). For this, the softest pieces of the sampled mortar specimens were carefully selected. If necessary, a metallic brush or a scalpel was used to remove any soil particles, paint layer or other residues possibly present on the pieces surface. The mortar pieces were then gently crushed in a ceramic mortar making successive impacts with a metallic pestle. During the process, care was taken to take hard fragments or fragments identified as aggregates out and to apply a relatively soft pression with the pestle. Next, the crushed sample was separated with a series of 500, 250, 125, 100, and 75 µm sieves. The fraction of particles with a size < 75 µm is used for the CO₂ extraction because the finer the fraction is, the less sand grains are present (including calcareous ones that could influence the dating results) and it provides sufficient material for extracting 4000 mbar of CO₂. The CO₂ extraction process consisted of gradually adding a known quantity of acid (a tube was designed with a 5 ml burette attached on it) to the fraction of particles lower than 75 µm and gather 7 to 9 CO₂ fractions for dating. In most cases, the binder reacts very quickly in acid since it is porous and fine grained while the dead carbon inclusions are much harder, less porous and react more slowly (Sonninen and Jungner 2001). Beforehand, to evaluate the quantity of sample needed for CO₂ extraction, the carbon percentage was measured using around 50 mg of the fraction < 75 µm under vacuum with 5 ml of orthophosphoric acid (H₃PO₄) in excess for 30 min. H₃PO₄ is used to preserve the vacuum pump because it contains less water. The volume of CO₂ released allowed determining the quantity of carbon present and then the quantity of sample needed to obtain 40 mbar for the first fraction. For the CO₂ extraction, the amount of sample was calculated to capture the first 1% of carbon reacting with hydrochloric acid (HCl), followed by approximately 2, 3, 4, 10, 30, 50, and 100%. During the process, the powdered mortar sample was kept in suspension (water is added first) under vacuum while 2 ml of 0.1 M HCl were added for the first fraction (1% of CO₂) and also for the individual next fractions (2, 3, and 4% of CO₂), then different amounts of 2.2 M HCl were added for the remaining fractions (10, 30, 50, and 100% of CO₂). The choice of HCl was made for practical reason, it is a stronger acid and allows faster reaction kinetics (a whole working day is needed for the full mortar CO₂ extraction). The different fractions were dated individually, and the CO₂ percentage was plotted as a function of the radiocarbon determination (Figures 2f, 3f, and 4f). The extrapolated date was obtained using the best mathematical fit and it was compared to the average value (combined radiocarbon dates with the Oxcal 4.4 program) of the first fractions falling in the same range, often the first three or four fractions (first 3 or 4 individual percent of CO₂ extracted). Both values should be relatively close to each other since the binder often reacts first (Lindroos et al. 2018).

When mortar samples display soft white inclusions measuring several millimetres or weighing at least 20 mg, the pure lime lump technique (Pesce et al. 2009; Lindroos et al. 2018) is also performed. For this, the inclusions are extracted from the mortar and powdered with ceramic mortar and pestle. For the CO₂ extraction, the powder is kept under vacuum and reacts with orthophosphoric acid at 14.8 M. Only one fraction is dated. The lime lumps of the samples presented in this paper were too little in size to perform the pure lime lump technique.

The graphitization was performed on a manual vacuum line (Van Strydonck and Van der Borg 1990) at 680°C with iron as catalyst in the presence of hydrogen. Typically, the graphite targets contain ∼1 mg of graphite (for 40 mbar of CO₂). The ¹⁴C dating was performed thanks to accelerator mass spectrometry with a MIni CArbon DAting System (MICADAS) (Synal et al. 2007; Boudin et al. 2015). The calibrated dates were determined with the OxCal 4.4 program (Bronk Ramsey 2009) using the IntCal20 northern hemisphere radiocarbon age calibration curve (Reimer et al. 2020).

When charcoal fragments were spotted in the mortar, they were extracted with a scalpel and submerged in HCl for several hours to eliminate any carbonated contaminants. If necessary the solution was changed and the sample was rinsed thoroughly with ultrapure Milli-Q™ deionized water when no signs of reaction was observed in an acid pH. The graphitization and radiocarbon determination were performed as described for charcoal sample in Wojcieszak et al. (2020).

LIME MORTARS COMPONENTS AND TERMINOLOGY

Apart of the binder, many other mortar components can have an influence on the radiocarbon content, mainly the ones containing carbonates. These carbonate contaminations are underburned pieces of the limestone used for lime production, fossil carbonates such as shell fragments or rounded limestone particles naturally occurring in calcareous sands, carbonate-rich aggregates such as crushed limestone/marble/calcite or even crushed (reused) mortars and specific minerals containing carbonates such as layered double hydroxides (LDHs) possibly formed in hydraulic mortars (Artioli et al. 2017). The presence of geogenic/fossil carbonates results in a lower ¹⁴C/¹²C ratio which may generates an older date, as they are radiocarbon free. In the case of reused mortar fragments an older date will also be obtained, but to a much lesser extent, since they are not radiocarbon free. Another category of carbonates results from weathering and can cause an increase or a decrease of the ¹⁴C/¹²C ratio depending on the incorporation of more recent carbon or older carbon respectively. A final factor to be taken into account is the hardening speed of the mortars. This varies according to different parameters, in particular according to the distance from the surface (Daugbjerg et al. 2021a). Some mortars may take several decades or even hundreds of years to fully carbonate (Hajdas et al. 2017). Figure 1 describes the different mortar constituents with their eventual impact on the ¹⁴C/¹²C ratio. This will depend on the final concentration of these components in the isolated mortar fraction used for CO₂ extraction before graphitization and dating. The following results will focus on stepwise acid hydrolysis on particles < 75 µm compared, when possible, to a charcoal piece found in the mortar samples. Other preparation methods have been developed using different particle separation techniques, sizes, and reaction processes (Folk and Valastro 1976; Lindroos et al. 2007; Heinemeier et al. 2010; Marzaioli et al. 2011; Ortega et al. 2012; Michalska and Czernik 2015; Michalska et al. 2017; Nonni et al. 2018; Michalska 2019; Sironić et al. 2023), as well as thermal decomposition (Labeyrie and Delibrias 1964; Toffolo et al. 2020; Barrett et al. 2021, 2023; Daugbjerg et al. 2021a); but they will not be discussed in details here. Another important aspect for radiocarbon dating lime mortars is the sampling methodology which was meticulously studied by Daugbjerg et al. (2021b).

Figure 1 Mortar constitution and possible influence on lime binder radiocarbon dating using stepwise acid hydrolysis.

Carbonates present in lime mortars can be classified in three categories in the context of lime mortar dating: i) binder-related particles originating from completely burned particles that fully reacted just after the erection of the building or in the following few weeks/months and did not undergo weathering processes—they are the carbonates which are targeted for radiocarbon dating. They have been sometimes described as “datable fraction,” “pristine carbonate fraction,” “original fraction,” “anthropogenic carbonate” (Hajdas et al. 2017; Urbanová et al. 2020) or “pyrogenic calcium carbonates” (Toffolo 2020), but some of these terms can also include compounds formed in the following categories; ii) carbonates formed subsequently because of weathering processes, delayed hardening or reactive compounds; and iii) geogenic/fossil carbonates having a geological or taphonomic origin and containing dead carbon (Nawrocka et al. 2005; Hajdas et al. 2017; Lindroos et al. 2018). All three categories can be present as lumps/inclusions and/or can be dispersed as smaller particles within the matrix and can have an influence on the radiocarbon dating. The following paragraphs describe in details the mortar constitution and their possible influence on the date obtained with some practical examples on Belgian mortars.

Inorganic Aggregates

Aggregates found in historic mortars can be of natural or anthropogenic origin and either be unaltered or crushed (Pecchioni et al. 2014). Standard petrographic methods are very useful for the identification of the type of aggregates (Elsen 2006; Ortega et al. 2012; Nonni et al. 2013; Pecchioni et al. 2014; Michalska and Pawlyta 2019; Caroselli et al. 2020). Lime mortars can be classified as hydraulic or non-hydraulic depending on the types of calcined lime and aggregate added. “Hydraulic” refers to the ability of hardening under water, but these types of mortars can also harden in ambient air (Forster 2004).

Non-hydraulic mortars, also called air hardening lime mortars, can contain different types of inert aggregates. Inert aggregates, such as siliceous sands, non-reactive ceramics fragments or slag, do not have any impact on radiocarbon dating. Here, “inert” refers to the fact that they are not part of the chemical reaction taking place during the hardening/carbonation process.

Aggregates of air hardening lime mortar can also include other inert materials containing carbonated inclusions such as calcareous sand, crushed white marble, shell fragments, crushed limestone or crushed (reused) mortar fragments (Miriello et al. 2018). All these types of inorganic aggregates incorporate carbon within the mortar which can have an influence on the ¹⁴C/¹²C ratio.

An initially non-hydraulic lime binder can incorporate reactive (in opposition to inert) aggregates inducing pozzolanic reactions forming compounds with cementitious properties. The resulting binders can be qualified as hydraulic because, as natural hydraulic binders (see section Binder), they possess the ability to harden both in air or under water. The hardening of these mortars happens both through carbonation of calcium hydroxide and through reaction of calcium hydroxide with hydraulicizing components mainly forming calcium silicate hydrates (C-S-H) (Pecchioni et al. 2014; Seymour et al. 2023). The porosity decreases around these reactive aggregates, often giving a darker shade to the mortar. They are called pozzolanic or latent hydraulic materials and can be composed of volcanic pyroclastic rocks (pozzolana, tuffs), diatomaceous earths, crushed/powdered ceramics (called “cocciopesto” or “terrazzetto” in Italy, “Homra” in Arabic countries, “Horasan” in Turkey or “Surkhi” in India) or calcined clay (metakaolin) (Böke et al. 2006; Elsen 2006; Shi and Day 1993). Moreover, magnesium-silicate-hydrates (M-S-H) formed in some modern cementitious systems (Zhang et al. 2011) have also been found in render employed in the Nora cisterns during the Roman period (Secco et al. 2020). Their formation seems to be promoted by the use of mixed Mg-rich pozzolanic aggregates (e.g., animal ashes) and crushed ceramics. Other products of the hydraulic/pozzolanic reaction in modern and ancient binders are layered double hydroxides (LDH) (Artioli et al. 2017; Secco et al. 2020; Toffolo 2020). These latter compounds contain carbonates that can compromise the radiocarbon dating when using acid hydrolysis (this is further described in the section Delayed hardening). Radiocarbon dating of hydraulic mortars is often more complex compared to non-hydraulic moratrs (Heinemeier et al. 2010). Although, recently, FTIR spectroscopy was used to help pre-screening of hydraulic mortars that could yield correct radiocarbon results (Asscher et al. 2020).

Radiocarbon dating of non-hydraulic mortars showed more reliable results compared to hydraulic mortars (Hajdas et al. 2017; Hayen et al. 2017). As it is the case for the sample sp32 3A from the St Martin’s church in Rutten (Belgium) illustrated in Figure 2. The petrographic observations (Figure 2a) indicate a mortar in good condition and the presence of a siliceous sand (slightly glauconitic and feldspathic quartz sand without fossil carbonates). Glauconite, an iron potassium phyllosilicate (mica group) mineral, is recognised as rounded green particles in plain polarised light (PPL, Figure 4c), when partially oxidised the color is orange (Figure S3a), and when the oxidation is full, it appears black (Figures 2a, 3a, and 4a). Glauconite is a common occurrence in Belgian mortars (Fontaine et al. 2014). Sandy-clay agglomerates (Figure S1a and b) and mica flakes (Figure S1e and f) are also observed in the thin-sections of the sample sp32 3A. The lime lumps are mainly completely burned (Figure 2a), but some of them can be underburned (Figure 2b, 2c, 2e, S1c, and S1d). The cathodoluminescence image of the powder < 75 µm (Figure 2d) used to extract the CO₂ for radiocarbon dating shows very few amounts of limestone observable in bright red. The results obtained for the different CO₂ fractions of the mortar are shown in Figure 3f (see Table S1 for the numerical results). The extrapolation curve (with an order 3 polynomial trendline) gave an estimation of 1244 ± 33 BP (676–880 cal AD [2σ]) and the combination of the three first fractions 1219 ± 20 BP (770–890 cal AD [2σ]). Both results matched with the presumed historical date (480–750 AD) and the date obtained for the charcoal piece extracted from the mortar (1293 ± 32 BP–657–797 cal AD [2σ]).

Figure 2 Mortar sample sp32 3A from the St Martin’s church in Rutten (Belgium). (a) characteristic aspect in plain polarised light (PPL) of the thin-section showing three completely burned lime lumps (L) as the larger subrounded inclusions in brownish gray, the smaller white subangular/subrounded grains are quartz (Qz) and K-feldspars (Fsp) grains, and the rounded darker ones on top left are oxidised glauconite (oGlt); all inclusions or grains are surrounded by the intergranular binder in brownish gray (color comparable to the L) and some macropores (P) are visible in light yellow; (b) PPL image of a large underburned limestone inclusion (uL); (c) zoom in on a mainly underburned limestone inclusion showing reminiscence of the limestone structure (LS) in gray; (d) cathodoluminescence image of the powdered fraction < 75 µm with a few bright red limestone inclusions, bright blue K-feldspars and bright green plagioclases; (e) cathodoluminescence image of a thin-section showing an underburned lime lump in lighter and diffuse red on the top left, smaller bright red limestone fragments that possibly correspond to underburned inclusions, quartz grains in dark purple and bright blue K-feldspars; and (f) radiocarbon results for the charcoal extracted from the mortar and for the powdered mortar with particles < 75 µm as a function of the CO₂ fraction.

However, when aggregates containing high amounts of fossil carbonates are present in a mortar, air hardening mortars can also be problematic for radiocarbon dating. Another sample (sp162 7B) from the St Martin’s church in Rutten presented in Figure 3 and S2 illustrates this scenario. The thin-section examination (Figure 3a, 3b, 3d, 3e, S2a and S2b) highlighted a high quantity of limestone in the aggregate used. It was recognised thanks to thin-section petrography (Figure 3a, 3b, 3e, S2a and S2b) because of the presence of sandy biopelmicrite and fossil inclusions and thanks to the high quantity of bright red in the cathodoluminescence images of the thin-section (Figure 3d and S2c). Cathodoluminescence on the powdered sample < 75 µm also revealed a high amount of limestone represented by numerous bright red particles (Figure 3c). Charcoal (Figure S2d, S2e and S2f) and sandy-clay agglomerates (Figure S2g and S2h) were also identified in the thin-sections. In addition, the micritic binder shows a partial dissolution (Figure S2g). The radiocarbon content of the three first fractions combined was 1492 ± 20 BP (565–640 cal AD [2σ]) and the extrapolation curve of all fractions (Figure 3f, and Table S1 for the numerical results) using a third order polynomial trendline was 1471 ± 34 BP (552–648 cal AD [2σ]). The last fraction of the graph (3709 ± 34 BP—2202–1981 calBC) is much older than the one obtained for the sample sp32 3A (1560 ± 32 BP—425–575 calAD). In addition, the charcoal result was estimated at 1297 ± 32 BP (657–776 cal AD [2σ]) which is younger than the age obtained for the mortar. This result is impossible since the charcoal can only be older or of the same age as the mortar. The high amount of fossil carbonates leads to a lower ¹⁴C/¹²C ratio using the stepwise acid hydrolysis method on particles < 75 µm.

Figure 3 Mortar sample sp162 7B from the St Martin’s church in Rutten (Belgium). (a) characteristic aspect in PPL of the thin-section showing the micritic binder in gray, limestone (LS) inclusions in darker gray, quartz (Qz) and K-feldspar (Fsp) grains as white subangular/subrounded inclusions, glauconite (Glt) occurs as rounded and greenish/brownish/black grains as a function of the oxidation degree, and the irregular porosity (P) in light yellow; (b) limestone inclusion (LS, biopelmicrite); (c) cathodoluminescence image of the powdered fraction < 75 µm with numerous bright red limestone inclusions, a few bright blue K-feldspars and bright green plagioclases; (d) cathodoluminescence image of a thin-section showing numerous limestone (biopelmicrite) inclusions in bright red, quartz grains in dark purple and bright blue K-feldspars; (e) same image in PPL; and (f) radiocarbon results for the charcoal extracted from the mortar and for the powdered mortar with particles < 75 µm as a function of the CO₂ fraction.

Furthermore, inorganic aggregates can be subjected to alteration caused by weathering processes and also cause variation of the ¹⁴C/¹²C ratio (see “Weathering Carbonates” section).

Binder

The binder derives from the calcination of natural limestone, which is then mixed with water, and it carbonates by incorporating atmospheric CO₂. It can be called anthropogenic or pyrogenic carbonate since it was produced by heat treatment. The texture of the binder under PPL is most commonly micritic (the individual grains cannot be distinguished from each other, < 4 µm) as shown in Figures 2 and 3, and in rare cases micro-sparitic (distinguishable individual grains, 4–10 µm). The crystal sizes limits between “micrite” and “microspar” are those according to Tucker (1981). Two types of binder are considered in the case of lime mortars: hydraulic and non-hydraulic.

As explained in the previous section, hydraulic binders can result from the interaction with the aggregate used. Another way to obtain hydraulic binders is by burning impure limestone which naturally contains latent hydraulic materials. The latter are non-carbonate minerals or siliciclastic materials containing silica and alumina such as clay or microcrystalline silica. Like hydraulic mortars resulting from the use of reactive aggregates, natural hydraulic mortars can take longer to carbonate. When they hardened under water, hydraulic mortars did not interact with atmospheric CO₂ making them unsuitable for radiocarbon dating (Daugbjerg et al. 2021b). If the hardening of hydraulic mortars happened in the presence of air, the same complications as for air hardening mortars are possible in addition to a more frequent delayed hardening due to their lower air permeability (Ringbom et al. 2014; Nonni et al. 2018; Daugbjerg et al. 2021b).

Lime binders, also called carbonate or carbonaceous binder, can be qualified as calcic if the limestone used is relatively pure calcium carbonate (CaCO₃) not containing hydraulicizing components (the most common case), or magnesian/dolomitic if magnesian/dolomitic limestone, i.e., containing different percentages by weight of calcium-magnesium carbonate [CaMg(CO₃)₂], is burned. Magnesian/dolomitic lime binders are heterogeneous due to the presence of hydromagnesite[Mg₅(CO₃)₄(OH)₂•4(H₂O)] areas embedded within the calcitic matrix. Another characteristic feature of magnesian/dolomitic lime binders are the presence of hydromagnesite ghosts with spherical shape (Pecchioni et al. 2014). Dolomitic mortars have been successfully dated but different methods need to be implemented depending on the type of aggregate used (Caroselli et al. 2020).

Binder Lime Inclusions

A lump is a piece or mass of indefinite size and shape. In the context of historic lime mortars, they exhibit a white color and are often irregular spherical particles visible to the naked eye (Hughes et al. 2001). They can be considered as aggregate because even though they derive from the quicklime, they do not act as binder (Leslie and Hughes 2002). Hughes et al. (2001) state that their presence in a mortar indicates that the mortar was mixed “hot” with unslaked quicklime added directly to sand and water or with poorly slaked lime putty containing relict material (Hughes et al. 2001). Other hypothesis for the lime lump formation is that they derived from the hard carbonate crust which can form on top and around the edges of lime putty as it matures in pits or other containers, or they result from a poor mixing of mortars, causing a poor combination of aggregate and lime binder (Bakolas et al. 1995; Bruni et al. 1997). Pesce et al. (2009) call them “unmelted lumps.” Different textures are observed for lime inclusions but most often, they are rounded grains of calcium carbonate showing internal shrinkage cracks (Hughes et al. 2001) as can be observed in Figure 2a. Lindroos et al. (2018) wrote that the most suitable lumps for ¹⁴C dating are soft, white and have a flour-like surface. It is often considered that perfectly formed lime lumps are free from dead carbon contamination (Nonni et al. 2018). Numerous studies show that lime lumps are suitable for radiocarbon dating (Pesce et al. 2009, 2012; Barrett et al. 2021; Lindroos et al. 2018) although they can be subjected to contamination from unburned limestone (Lindroos et al. 2018) or secondary calcite deposits (Caroselli et al. 2020).

In Figure 1, three types of lime lumps resulting from the calcined limestone are considered. First, the lime lumps sensu stricto that are completely burned or well-fired carbonated binder inclusions (see Figure 2a). If they were not subjected to weathering process or delayed hardening, and also large enough, they are good candidates for radiocarbon dating. A second subcategory is made of underburned/incompletely burned carbonated binder inclusions also called “underfired” nodules (Weber et al. 2013); these can induce a decrease of the ¹⁴C/¹²C ratio since they still contain fossil carbon. On thin sections under PPL, they present reminiscence of the limestone texture (Figure 2b, 2c, S1c, and S1d). Pecchioni et al. (2014) show a variety of examples depending on the limestone source. In Figure 2b and 2c, relics of the micro-sparitic limestone texture are clearly seen in the thin-section observed in PPL. In cathodoluminescence by optical microscopy, such lumps appear red but not as bright as pure limestone fragments and some parts of the lump do not luminesce (Figure 2d). The last type of lumps considered are overburned binder inclusions. They may be unreactive inclusions that slake poorly (Leslie and Hughes 2002) and then would have little or even no influence on radiocarbon dating.

Organic Matter

The presence of organic matter in a lime mortar can originate from intentional or unintentional processes. If intentional, it can be incorporated during the mortar production process as additive or as admixture forming organic aggregates (Pecchioni et al. 2014). Additives are liquids intentionally added in relatively small amounts to affect a desired change in properties. Additives can be made of egg yolk or white, casein, animal glue, milk fig, oils, blood, resins, etc. For instance, a recent study by Kang and Kang showed that the addition of perilla oil increases the resistance to freeze-thaw (Kang and Kang 2022). Admixtures are solid substances (also called organic aggregates) sometimes found in mortars, for instance charcoal, coal, straw, wood, hair, grains, seeds, dung, bottom ash, etc. Some of these organic materials can radiocarbon dated. The radiocarbon dating of organic materials, if present, should be preferred to date a mortar because it is often more reliable and more efficient in term of pre-treatment compared to dating the binder or even the lime lumps although it can also sometimes yield aberrant results (Van Strydonck et al. 1992). For wood and charcoal, the old wood effect should be considered but they can at least give a terminus post quem. A very recent study however pointed out the necessity to date multiple organic samples for an improve statistic of the results (Brabcová et al. 2023).

Over time microorganisms can develop and form a microbial community. Fungi, bacteria, archaea, cyanobacteria and lichens can lead to biodeterioration processes (for instance dissolution/precipitation—see Weathering carbonates section) but some species can also protect against other destructive ones (Dyda et al. 2019).

Organic matter will not interfere for dating using acid reaction but may have an impact when thermal decomposition is used to extract the CO₂.

Weathering Carbonates

In the literature about radiocarbon dating of historic lime mortars, many terms and concepts are used to describe weathering carbonates. They often include terms like recrystallization, diagenesis, redeposit, dissolution, bleaching, precipitation, secondary and neo-genesis carbonates (Carmine et al. 2015; Toffolo 2020; Daugbjerg et al. 2021b). Here, we restrict recrystallization to the transformation, essentially in the solid state, of metastable polymorphs of carbonates, for example aragonite or vaterite, which were already present within the binder matrix that spontaneously recrystallise into a more stable form (calcite) or to the possible transformation of originally micritic binder matrix (calcite crystals < 4 µm) in larger uniform in size and equant calcite crystals up to 10 µm (microsparite). Recrystallization in this sense would not have an impact on the radiocarbon content since no exogenous carbon is introduced within the crystallographic structure. In opposition, weathering or secondary carbonates deposits would form because of variations of the initial conditions causing dissolution/precipitation and can affect the amount of radiocarbon by either incorporating younger or older carbon (Michalska and Mrozek-Wysocka 2020). It can impact all the carbonates present within the mortar, i.e., calcareous aggregates, binder and binder lime inclusions. They may appear through leaching/redeposition cycles occurring because of relative humidity and pH variations or the presence of percolating water, rainfall and underground water. This phenomenon would also induce the disappearance of less stable forms of carbonates because after their dissolution, the precipitation would most probably lead to the precipitation of calcite at normal pressure (Toffolo 2021). In limestone-rich habitats, bacteria can contribute to precipitation of calcite (Koner et al. 2021) which could also occur in lime mortars.

The mortar sample from Ten Bogaerde in Koksijde (Belgium) showed clear signs of deterioration when observing the thin-section under the microscope: it has been dissolved (Figure 4a and 4c) and secondary carbonates were observed at the border of the macropores (Figure 4b). Moreover, the aggregate contained calcareous grains observed in cathodoluminescence (Figure 4d), but these grains end up in low quantity in the fraction < 75 µm (Figure 4e). The extrapolated estimated age using a linear regression curve was 610 ± 22 BP (1300–1400 cal AD [2σ]) and the radiocarbon content of the three first fractions combined was 625 ± 13 BP (1295–1395 cal AD [2σ]) which are both younger (Figure 4f and Table S1) than the presumed historical date (1232/3–1253 AD). The weathering of the mortar inducing dissolution and precipitation of carbonates introduced younger carbon resulting in an inaccurate radiocarbon dating of the mortar sample.

Figure 4 Mortar sample S7002-3 from Ten Bogaerde in Koksijde (Belgium). (a) characteristic aspect in PPL of the thin-section showing the dissolved binder in light gray surrounding calcareous grains in darker gray (LS), quartz (Qz) and K-feldspar (Fsp) grains in white, rounded oxidised glauconite (oGlt) grains in black, and irregular macropores (P) in light yellow; (b) cross polarised light (XPL) image of a large pore (P) in black showing secondary carbonates (sC) deposit on its edge displaying interference colors of the 2nd to 3rd order; (c) PPL image of a partially dissolved lime lump (dL), irregular macropores (P) in light yellow and a glauconite (Glt) grain in green; (d) cathodoluminescence image of a thin-section showing two calcareous grains in bright red, quartz grains in dark purple, bright blue K-feldspars and green plagioclases; (e) cathodoluminescence image of the powdered mortar fraction < 75 µm showing a few bright red limestone particles and a few bright blue K-feldspars; and (f) radiocarbon results for the powdered mortar with particles < 75 µm as a function of the CO₂ fraction.

Delayed Hardening

Mortar carbonation follows a diffusion process (Van Balen 2005). Carbonates can form up to 1000 years after the construction (Hajdas et al. 2017). Carbonates formed with a delay can be considered as pyrogenic or anthropogenic carbonates but they increase the ¹⁴C/¹²C ratio since their formation is not synchronous with the initial hardening of the mortar. Delayed hardening can happen in the inner parts of a thick wall but also depends on the porosity and the composition of the mortar. In case dolomitic limestone has been used as a raw material, the magnesium components carbonate very slowly (Pecchioni et al. 2014; Michalska et al. 2017; Daugbjerg et al. 2021a). When a mortar did not fully carbonate, modern carbon can be incorporated during the sampling procedure since the mortar is exposed to CO₂ from the atmosphere. Mortars presenting hydraulic properties due to reactive aggregates can take longer to carbonate because of the formation of C-S-H reducing air permeability which slows down the carbonation speed. For natural hydraulic lime, Ca(OH)₂ is also produced when C-S-H are formed. This Ca(OH)₂ could further carbonate in the presence of CO₂ but this reaction takes place at a very slow rate. The higher density caused by the presence of C-S-H reduces the CO₂ diffusion through the mortar (Forster 2004). Layered double hydroxides (LDH) can be found in hydraulic mortars as a product of the hydraulic/pozzolanic reaction (Artioli et al. 2017). LDH contain carbonate groups (CO₃ ^2-) that can exchange CO₂ with the atmosphere and introduce younger carbon to the mortar (Ishihara et al. 2013; Artioli et al. 2017; Toffolo 2020). To summarise, this phenomenon can be linked to the distance from the surface of the wall, the presence of reactive aggregates, the type of binder and also to the carbonate binder lime inclusions (see Figure 1).

The Roman mortar of Tongeren (Belgium) dated by seven radiocarbon laboratories for the MOrtar Dating Inter-comparison Study (MODIS) showed an example of this phenomenon (Hajdas et al. 2017; Hayen et al. 2016, 2017). The expected date for the cocciopesto mortar, based on the occupation history of the site, on archaeological findings in the immediate surroundings and on the ¹⁴C dating of charcoal, was the 4th century AD. While the results from the laboratories are in the same range, the age obtained for the hardening process was between the end of the 9th and the 14th centuries AD, which does not correspond to the expected age. Petrographic analysis showed the use of dolomitic lime due to the identification of hydromagnesite ghosts that have been filled up by calcite (Figure S4a to d) and roughly crushed terracotta with fragments of several centimetres in size (Figure S4e and f). A younger age was obtained because of the leaching of the lime binder in combination with delayed hardening and possible secondary carbonate deposits.

DISCUSSION

The combination of thin-section petrography and cathodoluminescence on the fraction of particles with a size < 75 µm intended for the CO₂ extraction provides a very good insight into the binder conservation state and the presence of contaminants or disruptive processes, as the results of the four Belgian mortars characterised by these two techniques clearly show. For the sample sp32 3A from the St Martin’s church in Rutten, the following factors explain why a plausible date that matches the presumed historical date and the date obtained for the charcoal piece extracted from the mortar was obtained: i) the good preservation state of the binder, ii) the use of a non-hydraulic lime combined with the fact that most of the lime binder inclusions appear completely burned, iii) the use of an aggregate devoid of fossil carbonates, iv) the absence of weathering carbonates, and v) the very few amounts of calcareous particles in the powder < 75 µm as revealed by cathodoluminescence. On the other hand, the dating of the sample sp162 7B, coming from the same archaeological context as the previous sample, gives an incoherent date. The charcoal found in the mortar yielded a younger age compared to the surrounding mortar, which is impossible. In this case, despite the relatively good conservation state of the binder (partial dissolution may locally occur), the high content of fossil carbonates in the aggregate leads to a totally unreliable date. The aggregate consists of soft limestone pieces and fossil grains showing a hardness comparable to the lime binder. Hence, they end up in the fraction < 75 µm, as clearly revealed by cathodoluminescence observations.

Clear signs of deterioration for the mortar sample from Ten Bogaerde in Koksijde were highlighted by thin-section petrography with a significant dissolution of the lime binder and the presence of secondary carbonates lining the macropores. These secondary carbonates end up in the fraction < 75 µm introduce younger carbon resulting in an inaccurate dating of the mortar sample. Interestingly, the calcareous grains contained in the aggregate seems not to end up in the fraction < 75 µm. This indicates that it would have been possible to date this mortar correctly if the binder was still in good condition, despite the use of a partly calcareous aggregate.

For the Roman mortar of Tongeren, petrographic analysis showed the use of dolomitic lime with roughly crushed terracotta fragments. In this case, the leaching of the lime binder in combination with delayed hardening and possible secondary carbonate deposits lead to the obtention of a younger age.

CONCLUSION

A summary (Table 1) of the existing mortar terminology currently used in international publications about mortar ¹⁴C dating was elaborated and discussed in order to improve the exchange about concepts and components involved in the process.

Table 1 Summary of the constituents, their definitions and the different terms used in international publications.

It consists of five categories of components (inorganic aggregates, binder, binder lime inclusions, organic matter, and weathering carbonates) and a phenomenon that can possibly influence the dating of the mortar carbonation (delayed hardening). Some of these categories are subdivided and each subcategory or category are discussed depending on their possible impact regarding radiocarbon content using stepwise acid hydrolysis. Aggregates are classified as inert or reactive, both can influence the radiocarbon content except if the inert aggregates are exempt from calcareous grains. For the binder, the hydraulicity plays a role in the carbonation process and can influence the radiocarbon amount in the mortar. Binder lime inclusions can be lime lumps (completely burned), underburned (which still contain dead carbon) or overburned. Organic matter might be present in the form of additives, admixtures, and microorganisms. Pieces of contemporary and datable organic matter present in the mortar, such as wood, charcoal, straw and hair should be preferred, if present, for radiocarbon dating lime mortars. When mortars contain weathering carbonates formed by variations of the ambient conditions or by the presence of microorganisms, exogenous carbon may possibly be introduced and compromise radiocarbon dating. On the other hand, recrystallization of the initially micritic binder into microsparite and spontaneous transformation of a metastable polymorphs of carbonate should not influence the carbon isotope ratio. Delayed hardening, if present, involves an increase in radiocarbon content and can even results in the introduction of modern carbon when sampling.

Often, multiple phenomena and/or inclusions can have an influence on the radiocarbon content present in a lime mortar. A careful characterization is essential to evaluate the possible impact and the reliability of the radiocarbon dating results obtained. Petrography and cathodoluminescence helped to identify components that influence the radiocarbon dating. A high amount of limestone within the mortar, the presence of secondary carbonates and a dissolution of the binder are clear signs that will most probably lead to an unreliable ¹⁴C date using stepwise acid hydrolysis with a particle size < 75 µm.