INTRODUCTION
The radiocarbon (14C) dating of mortars has been a challenge for many laboratories around the world for many years (Labeyrie and Delibrias Reference Labeyrie and Delibrias1964; Stuiver and Smith Reference Stuiver, Smith, Chatters and Olson1965). The main difficulty in dating this type of material is that the elements likely to contaminate the sample are of the same chemical nature as the sample itself, namely calcium carbonates (CaCO3) (Lindroos et al. Reference Lindroos, Heinemeier, Ringbom, Braskén and Sveinbjörnsdóttir2007; Artioli et al. Reference Artioli, Secco, Addis, Bellotto and Gruyter2017). The LMC14 (Laboratoire de Mesure du Carbone 14), the French national platform for 14C preparation and measurement, has attempted to develop a reliable method for dating mortars (Heinemeier et al. Reference Heinemeier, Jungner, Lindroos, Ringbom, von Konow and Rud1997; Marzaioli et al. Reference Marzaioli, Lubritto, Nonni, Passariello, Capano and Terrasi2011). In this context, we took part in the international intercomparison campaign MODIS2 (MOrtar Dating Intercomparison Study). In this paper, the preparation protocol developed, which includes the extraction of CO2 by thermal decomposition, is first presented. Then, an overview of the preliminary study is described, giving the appropriate extraction temperature range, determined by Thermal Gravimetric Analysis. 14C measurements were performed on the ARTEMIS AMS facility of the LMC14 laboratory in Saclay, France. The results obtained are compared to the consensus values presented at the Radiocarbon conference in Zürich in September 2022. A final discussion lays out the strengths and limitations of the thermal decomposition approach for mortar dating (Barrett et al. Reference Barrett, Donnelly and Reimer2020; Daugbjerg et al. Reference Daugbjerg, Lindroos, Hajdas, Ringbom and Olsen2021).
MATERIALS AND METHODS
Samples
The LMC14 laboratory took part in the activities of the Mortar international group in 2014 in Padua (Italy). After the Mortar Dating International Meeting (MODIM) in 2018 in Bordeaux (France), it was decided to participate in the MODIS2 intercomparison campaign. The method chosen to extract the CO2 from the mortar was thermal decomposition associated to a prior Thermal Gravimetric Analysis (TGA). For each material, the weight loss acquired during the TGA gives the working temperature range to extract CO2 from the mortar sample. This approach is similar to the one used for the extraction of CO2 from lead white for radiocarbon dating, in which the laboratory has expertise (Beck et al. Reference Beck, Messager, Coelho, Caffy, Delqué-Količ, Perron, Mussard, Dumoulin, Moreau, Gonzalez, Foy, Miserque and Bonnot-Diconne2019; Messager et al. Reference Messager, Beck, de Viguerie and Jaber2020, Reference Messager, Beck, Germain, Degrigny, Serneels, Cano and Cardoso2021, Reference Messager, Beck, Blamart, Richard, Germain, Batur, Gonzalez and Foy2022).
For thermal decomposition, sample tests were run to set up the method before applying it to MODIS2 samples. Two “simple” mortars, previously well studied by the community, were selected:
MODIS1-1: The first sample of the MODIS1 intercomparison (Hajdas et al. Reference Hajdas, Lindroos, Heinemeier, Ringbom, Marzaioli, Terrasi, Passariello, Capano, Artioli and Addis2017), a wall bedding mortar from the church of Nagu in the Åboland archipelago (Finland).
NAGU025: A mortar sample from the east gable of the nave of the church of Nagu in the Åboland archipelago (Finland). The sample was taken between a roof truss and a piece of scaffolding, which were both dated independently, samples 023w and 024w (Sjöberg et al. Reference Sjöberg, Lindroos and Ringbom2011).
After these initial tests had been performed, MODIS2 samples were prepared. The sample descriptions are as follows:
MODIS2-1: The church of Saltvik on the Åland Islands, Finland in the central parts of the Baltic sea on the Precambrian granitic basement. The sample is from the AD 14th c. tower (Heinemeier et al. Reference Heinemeier, Ringbom, Lindroos and Sveinbjörnsdóttir2010).
MODIS2-2: The church of Hamra on the Swedish island of Gotland to the south of the Åland Islands in a Mesozoic limestone terrain. The sample is from the attic of the chancel and dates from the (early) AD 14th c. (Roosval Reference Roosval1911; Ranta et al. Reference Ranta, Hansson, Lindroos, Ringbom, Heinemeier, Brock and Hodgins2009).
MODIS2-3: The early Christian Basilica of Santa Eulalia in Mérida, western Spain in a metamorphic schist area. The sample is from the inner corner of the north/northwest wall dating from AD 304–570 (Mateos Cruz Reference Mateos Cruz1999).
Before distributing these 5 samples among the different laboratories, some treatments were performed by the MODIS group. For MODIS1-1 and NAGU025, the samples were available as a fine fraction 46–75 μm and original pieces. For MODIS2 samples, the samples were split into pieces and each laboratory received one piece of each sample for characterization of the mortar. The rest of each sample was crushed with plastic covered pliers and sieved in a mechanically vibrated sieve for 15 min. The <150 micrometer (µm) grain-size fraction was collected and homogenized by shaking. Each laboratory received about 1000 mg of the fraction in a small glass vial. When filling the vial small subsamples from different parts of the master sample were merged.
Thermal Gravimetric Analysis (TGA)
For the TGA, samples were prepared as follows:
The MODIS 2 mortar samples were gently crushed to avoid producing small grains of geological carbonates, which can contaminate binder carbonate (Heinemeier et al. Reference Heinemeier, Ringbom, Lindroos and Sveinbjörnsdóttir2010). Sieves < 100 µm were used to collect the < 100 µm fraction of the powder; this fraction was used to carry out the TGA and then the thermal decomposition.
Prior to thermal decomposition, the samples were characterized by TGA. The analysis uses about 10 mg of the 46–75 μm size fraction. This quantity is placed in an oven equipped with a precision balance, under a controlled nitrogen atmosphere. By increasing the temperature, an increasing fraction of the solid sample thermally decomposes into gas. The TGA measures the weight loss of the sample as a function of temperature. For the present study, the TGA instrument, TA SDTQ600, from the Laboratoire d’Archéologie Moléculaire et Structurale (LAMS) of Sorbonne University (Paris, France) was used.
For the different samples, TGA was performed to determine the range of working temperatures that need to be applied to each mortar in order to collect CO2. From these results, the temperature steps can be adjusted for each sample, during the thermal decomposition phase. The weight loss and the derivative of the weight loss curve for the two test samples and the 3 MODIS2 samples are shown in Figure 1.
Figure 1 TGA plots of MODIS1-1, NAGU025, MODIS2-1, MODIS2-2 and MODIS2-3 samples showing the mass loss (in %; left y-axis) and the derivative of the weight loss curve (right y-axis). The data were acquired for 200 min at a rate of one point every 0.5 s.
The TGA result on the test sample (MODIS1-1) shows small wiggles at low temperatures. Significant weight loss starts to take place after 600°C and the major weight loss occurs between 610 and 740°C. The derivative of the weight loss curve shows a peak at these temperatures, corresponding to the release of CO2 from the sample. The NAGU025 sample shows a similar curve, starting at a slightly lower temperature, around 580°C. The peak ends at 700°C. The spike at about 760°C is an artefact. After these good results, the MODIS2 samples were processed.
The MODIS2-1 and MODIS2-2 samples present a similar behavior to that of the test sample. The MODIS2-2 sample releases CO2 at a lower temperature. The MODIS2-3 sample shows some irregularities in the weight loss curve below 550°C, which does not bode well for the radiocarbon dating results if it comes from other CO2 sources.
The TGA results are summarized in Table 1 and were used in the next step of the method, the thermal decomposition.
Table 1 Working temperatures for the different samples in degrees Celsius given by TGA.
Thermal Decomposition
An amount of 200 mg of mortar powder was placed in a quartz tube with quartz wool to avoid the sample being sprayed in our vacuum line. The quartz wool was combusted in an oven at 900°C overnight to clean it.
All the mortars were first heated at 550°C during 30 min to remove possible organic contaminants and Layered Double Hydroxide (LDH) (Artioli et al. Reference Artioli, Secco, Addis, Bellotto and Gruyter2017; Ricci et al. Reference Ricci, Secco, Marzaioli, Terrasi, Passariello, Addis, Lampugnani and Artioli2020) which can introduce young carbon dioxide. With the TGA, this kind of release was not clearly recorded, except for the MODIS2-3 sample which showed some weight loss between 200 and 500°C. Since the curves below 550 degrees were not completely flat, it was decided to apply this preheating for all the samples.
The heating of each mortar was carried out on the manual line (Figure 2) under vacuum (Dumoulin et al. Reference Dumoulin, Comby-Zerbino, Delqué-Količ, Moreau, Caffy, Hain, Perron, Thellier, Setti, Berthier and Beck2017). A sample was collected approximately every 20–30°C, depending on the amount of gas obtained. The different fractions collected cover the working range highlighted by TGA in Table 1. The different subranges of temperature are noted in the reference of the sample. When the amount of CO2 produced was very high during the first 10 min, a second fraction was collected until the end of the CO2 extraction on this subrange. This second sample is denoted “fraction 2” and corresponds to the higher temperature of the subrange.
Figure 2 LMC14 manual vacuum line with on the left the quartz tubes for the mortar thermal decomposition with temperature monitoring. In the middle, the H2O trap and the N2 trap. On the right, the Pyrex tubes to collect the different CO2 fractions.
The principle of the Thermal Decomposition (TD) method is based on the idea that the different types of carbonates present in the mortar sample decompose at different temperatures. The discussions around the TD of mortars (Toffolo et al. Reference Toffolo, Regev, Mintz, Kaplan-Ashiri, Berna, Dubernet, Yan, Regev and Boaretto2020; Barrett et al Reference Barrett, Keaveney, Lindroos, Donnelly, Daugbjerg, Ringbom, Olsen and Reimer2021; Daugbjerg et al. Reference Daugbjerg, Lindroos, Hajdas, Ringbom and Olsen2021) showed that the CO2 collected in the first phase comes from LDH (around 500°C–550°C). Then comes the CO2 coming from mortar binders (around 500°C–650°C) and lastly the geological carbonates (between 700°C and 800°C). In practice, when collecting the different fractions, these different types of carbonate may be realised at the same temperature and they may overlap. The 14C activity is used to discriminate the appearance of the different phases. The different radiocarbon measurements are used to locate a “flat zone” in fraction modern, corresponding to the pure mortar binder, before a dip generated by the older geological carbonate signature (Lindroos et al. Reference Lindroos, Ringbom, Heinemeier, Hodgins, Sonck-Koota, Sjoberg, Lancaster, Kaisti, Brock, Ranta, Caroselli and Lugli2018, Reference Lindroos, Ringbom, Heinemeier, Hajdas and Olsen2020a, Reference Lindroos, Heinemeier, Ringbom, Daugbjerg and Hajdas2020b). Thus the plateau in modern fraction at the lowest temperatures, before the dip, corresponds to the 14C activity of the binder. This concept is close to that of sequential dissolution, which considers that the binder reacts faster and more readily than the geological carbonates, which is more refractory (Goslar et al. Reference Goslar, Nawrocka and Czernik2009; Daugbjerg et al. Reference Daugbjerg, Lindroos, Hajdas, Ringbom and Olsen2021).
Graphitization and AMS Dating
Samples were then graphitized on the routine bench of the LMC14 laboratory (Moreau et al. Reference Moreau, Messager, Berthier, Hain, Thellier, Dumoulin, Caffy, Sieudat, Delqué-Količ, Mussard, Perron, Setti and Beck2020). As usual in radiocarbon dating, blank samples were produced and measured in order to subtract the contamination induced by the preparation processes. C1 IAEA marble was burned at 850°C in quartz tubes with the same quartz wool as that used for the samples and the CO2 was collected on the same vacuum line. The measurements of different CO2 fractions were performed on the ARTEMIS AMS facility (Moreau et al. Reference Moreau, Messager, Berthier, Hain, Thellier, Dumoulin, Caffy, Sieudat, Delqué-Količ, Mussard, Perron, Setti and Beck2020). This is a 9SDH-2 NEC tandem at 2.6 MV terminal voltage (Moreau et al. Reference Moreau, Caffy, Comby, Delqué-Količ, Dumoulin, Hain, Quiles, Setti, Souprayen, Thellier and Vincent2013). The large SNICS Cesium sputter source (134 positions) was used for this study. Blank subtraction (C1 IAEA marble) was performed as in the routine measurement (Donahue et al. Reference Donahue, Linick and Jull1990). The fraction modern of mean blank value subtracted from the raw data was F m = 0.00187 ± 0.00056 for the two test samples and F m = 0.00242 ± 0.00080 for the MODIS2 samples. These blank levels are slightly higher than the mean C1 blank value obtained with H2PO4 hydrolysis.
RESULTS
The preliminary tests on MODIS1-1 and NAGU025 gave good results in preparation and in radiocarbon dating. MODIS2 samples were also dated.
The 30 minutes’ preheating at 550°C removes possible contaminants such as LDH, which can contaminate the mortar with young carbon. As explained above, the method consists in locating a flat zone in fraction modern at the lowest temperatures corresponding to the 14C activity of the binders before a dip generated by the old geological carbonates. Radiocarbon analyses for MODIS1-1, NAGU025 and MODIS2 are displayed in Table 2. The age and Fm values are given at one sigma, but the values are not rounded, unlike the usual practice in reporting radiocarbon results (Stuiver and Polach Reference Stuiver and Polach1977; Mook and van der Plicht Reference Mook and van der Plicht1999).
Table 2 AMS 14C results after blank subtraction, giving the Sample code, the reference with the temperature subrange (or fraction 2 with the final temperature), the graphitized mass of carbon, the Fraction modern with its error bar (one sigma), the radiocarbon age (in BP) with its error bar (one sigma).
For each sample, identification of the flat zone was performed by looking at the curve given by the plots. A conservative approach was adopted, consisting in avoiding taking into account any fraction that may have been contaminated by other carbonates than the binder. The flat zone in fraction modern at the lowest temperatures was spotted and the values which dip at higher temperatures were excluded as they correspond to the release of the older limestone. Then combined calibrations of the selected data were performed with Oxcal v4.4.4 (Bronk Ramsey Reference Bronk Ramsey2009), using the IntCal20 calibration curve (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey, Butzin, Cheng, Edwards and Friedrich2020). Results are shown on Figure 3.
Figure 3 Flat zone identification in Fm versus temperature for each sample (black dashed double arrow) and combined age and combined calibration of the selected data at 95.4% probability (Oxcal v4.4.4).
The comparison with existing values is given in calibrated ages when an independent dating exists for this sample (for example, a dating on wood). BP ages are also provided for ease of comprehension.
The MODIS 1-1 sample presents a large flat zone from the starting temperature range (600°C) up to a drop at high temperature (above 700°C), caused by geological carbonates. The second fraction at 700°C is a little lower than the first fraction extracted in the temperature interval [680,700°C] and may be slightly contaminated by geological carbonates. In line with our conservative approach, this second fraction was excluded from the calculation of the combined age. The combined calibration on the first 7 points out of the 9 gives the calibrated age interval (with probability) of [1406 calAD–1433 calAD] (95.4%). The consensus value for this sample, resulting from the round-robin exercise named MODIS1 (Hajdas et al. Reference Hajdas, Lindroos, Heinemeier, Ringbom, Marzaioli, Terrasi, Passariello, Capano, Artioli and Addis2017; Hayen et al. Reference Hayen, Van Strydonck, Boaretto, Lindroos, Heinemeier, Ringbom, Hueglin, Michalska, Hajdas and Marzaoili2016, Reference Hayen, Van Strydonck, Fontaine, Boudin, Lindroos, Heinemeier, Ringbom, Michalska, Hajdas and Hueglin2017) gives the following two calibrated intervals: [1329 calAD–1337 calAD] (2.1%) and [1396 calAD–1443 calAD] (93.4%). The present results are in agreement with the consensus value and are also fully compatible with the wood dated at [1410 calAD–1432 calAD] (95.4%).
The NAGU025 sample shows a flat curve, with a very extensive temperature range. The first point seems to be younger and is probably contaminated by modern Carbon. This point should not be taken into account in the flat zone. It gives a combined radiocarbon age of 533 ± 11 BP and a combined calibrated age of [1401 calAD–1425 calAD] (95.4%). This has to be compared to the two independent dates of woods found above and under this location (Sjöberg et al. Reference Sjöberg, Lindroos and Ringbom2011). The calibrated dates for the roof truss of the nave and the east gable of the nave are respectively 526 ± 25 BP and 540 ± 30 BP. The calibrated date intervals are respectively {[1325 calAD–1345 calAD] (10.0%) and [1393 calAD–1440 calAD] (85.4%)} and {[1316 calAD–1355 calAD] (29.6%) and [1388 calAD–1437 calAD] (65.8%)}. These three results are in perfect agreement.
The MODIS2-1 LMC14 data plot presents a flat zone. It is quite difficult, however, to decide how many points should be included in the selection (the first two, three or four points). Conservative selection is once again the best way to go. Fraction 2 at 670°C is older than the previous three points and appears to be contaminated by geological carbonate. The 650–670°C fraction finishes extracting CO2 at the same temperature as fraction 2 (670°C), so it is preferable to discard this point from the flat curve. As for the first two extractions corresponding to the temperature between 600°C and 650°C, the corresponding combined age is 693 ± 22 BP, which gives the calibrated age intervals (with probability) of [1275 calAD–1305 calAD] (75.3%) and [1365 calAD–1384 calAD] (20.2%). This MODIS2 sample has to be compared to the preliminary results shown during the Radiocarbon 24 conference in Zürich by M. Scott (to be published): the median age was 634 BP and the first and third quartiles of the age distribution of the laboratories participating in the MODIS2 international intercomparison were respectively 484 BP and 767 BP. The LMC14 results are within this range. A dendrochronological date was established at 1381 calAD. This date is within the calibrated age intervals measured at the LMC14, thus showing good agreement on this mortar sample.
The gas production on MODIS2-2 started at a lower temperature (given by the TGA). The plateau is clear in Figure 3 and it corresponds to the first two collections of CO2 between 550 and 625°C. The combined age is 689 ± 22 BP and the combined calibrated age intervals are [1276 calAD–1307 calAD] (71.3%) and [1363 calAD–1385 calAD] (24.2%). Preliminary results show a median age at 682 BP and [Q1,Q3] interquartile intervals of [602 BP, 899 BP]. It is suggested that this mortar is from the 14th century. An early architectural interpretation claims that the chancel where the dated material comes from is from AD 1300. The LMC14 result is in agreement with all of these assumptions and with the preliminary consensus value.
The age curve as a function of temperature for the MODIS2-3 sample exhibits a strange behavior, with no flat zone. So, to give a result, only the first extraction (560 to 595°C) was considered indicating that the age of the shrine could be 1874 ± 30 BP. The preliminary results on this sample are widely scattered over 5 centuries. The preliminary consensus value (median age of the distribution of the laboratories’ measurements) is 1802 BP and the interquartile interval is [1680 BP, 2024 BP]. However, the shrine is assumed to be in the age interval [AD 304–AD 570]. The results of the laboratories are mostly too old, highlighting a problem in the preparation step, caused by contamination in geological carbon due to the proven presence of dolomite in the mortar.
DISCUSSION
The results presented in this paper are in agreement with the consensus values and the independent dating carried out mainly on pieces of contemporary wood in mortar except for one sample MODIS2-3.
The two test samples from the church of Nagu in Finland made it possible to validate our procedure on samples already characterized and chemically and mechanically treated by the MODIS group. The LMC14 approach, i.e. treatment protocol (see above) + TGA + thermal decomposition + dating + “flat zone” identification on the age versus heating temperature curve, gives radiocarbon results in very good agreement with the known values.
These mortars are “pure” binder mortars, with small amounts of geological limestone. Their composition has been studied (Lindroos, personal communication; Hadjas et al. Reference Hajdas, Lindroos, Heinemeier, Ringbom, Marzaioli, Terrasi, Passariello, Capano, Artioli and Addis2017) and shows that it is possible to obtain a correct and credible dating quite easily on this type of mortar.
The results obtained on the three mortars selected for the MODIS2 intercomparison are more mixed and differ from the expected values depending on the nature of the mortar. MODIS2-1 was judged as a straightforward sample, as this kind of mortar performed well at 95% success (Heinemeier et al. Reference Heinemeier, Ringbom, Lindroos and Sveinbjörnsdóttir2010). The result is correct and therefore the dating of this kind of mortar with thermal decomposition is advisable.
The MODIS2-2 mortar presents an abundant limestone contamination of the aggregate. The result of the thermal decomposition shows very good agreement with the expected value. The protocol based on thermal decomposition gives reliable results because the limestone contamination is clearly evidenced after 625°C.
The case of MODIS2-3 is much less convincing. The results of most laboratories, using different methods (thermal decomposition, sequential dissolution…) show dates that are very often much older than the expected age. With TD we cannot get any “flat zone” and the 14C activities dip quickly. This is attributed to the presence of dolomite (geological carbonate) in the dated samples which begin to react around 600°C with the binder.
CONCLUSION AND PERSPECTIVE
The LMC14 team took part in the MODIS2 intercomparison campaign on mortar by testing the dating of mortar after thermal decomposition. Four out of five samples gave results in agreement with the expected values, showing that this method is reliable. These four samples are similar in characteristics, with a slight contamination of the binder during the manufacturing stage or during the stage of addition of the aggregate. On the other hand, the method used at the LMC14 on the 5th sample (MODIS2-3) did not make it possible to obtain an accurate date. Geological carbonate contaminations were clearly visible in the results.
The LMC14 dating method combining TGA with thermal decomposition is thus to be recommended for mortars that have been well characterized beforehand (Toffolo Reference Toffolo, Regev, Mintz, Poduska, Shahack-Gross, Berthold, Miller and Boaretto2017, Reference Toffolo, Regev, Mintz, Kaplan-Ashiri, Berna, Dubernet, Yan, Regev and Boaretto2020; Urbanova Reference Urbanova, Boaretto and Artioli2020) which will be called “pure” binder mortars and which do not contain contaminants that are too complicated to separate thermally. To go further, putting ourselves in the shoes of the users of 14C dating for mortars, it may be economically unsuitable to request multiple dating on a mortar to obtain the age curve as a function of temperature (fractions). What could be advised, in view of the results obtained, is a simplified thermal decomposition approach, reducing the number of fractions: this would involve heating the properly prepared sample as usual for 30 minutes up to 600°C to eliminate organic contaminants and LDH phase, then collecting the heating fraction between 600 and 620°C and sending it for dating.
This simplified operating mode nevertheless implies a rigorous characterization beforehand, to determine whether or not the mortar is sufficiently “pure” to be dated. Mortars with high geological carbon contamination require further methodological investigations to be certain of being able to obtain a reliable date. This could lead to a third MODIS international intercomparison campaign, once convincing developments have been found.
ACKNOWLEDGMENTS
The authors thank Alf Lindroos from the Department of Geology and Mineralogy, Åbo Akademi University, Finland for providing test samples and fruitful discussions of the different methods used for mortar dating. The LMC14 is funded by five French organizations: CEA, CNRS, IRD, IRSN, and MC.
