sample preparation

After a macroscopic description the mortar samples were divided into two parts: one part for luminescence dating and the other part for XRD measurements. The difficulty to extract pure coarse grain quartz fraction of sufficient quantity from sand used as inert in lime mortar has led the authors to study the luminescent emission from polymineral fine grain phase “enriched in quartz” through HF etching procedures during preparation phase of the samples (Prasad, 2000; Mauz and Lang, 2004; Stella et al., 2013). The possibility to use this procedure must be validate by several tests (e.g. preheating test, recovery test, thermal transfer test…). After the removal of two external millimetres, mortar samples were pretreated with 10% H₂O₂ for four days to remove organic material, and with 20% hydrochloric acid for 120 min. to dissolve carbonates. Subsequently the samples were treated with 20% HF at RT for 15 minutes, washed in 10% HCl for 25 minutes to remove fluorides, and then the 4–11 μm grain fraction enriched in quartz was selected and deposited onto 9.8 mm diameter discs (Gueli et al., 2010; Stella et al., 2013). The purity of fractions was then evaluated by the ratio between the post-IR OSL/T2 (normalized OSL intensity after IR stimulation) and OSL/T1 (normalized OSL emission) that resulted close to unity (Mauz and Lang, 2004; Prasad, 2000; Shen et al.,2007; Zhang and Zhou, 2007; Gueli et al., 2010; Gerardi et al., 2012; Stella et al., 2013).

For brick samples the polymineral fine grain fraction was obtained by PH3DRA standard procedure (Gueli et al., 2009, 2010; Guibert et al., 2009a). As for mortars, the outer two millimetre layer was removed, then the sample was crushed and sieved to select the fraction below 40 μm. Following, a sequence of etching procedures was made: 10% HCl for 1h to remove the carbonate, 10% H₂O₂ for 48 hours to remove the organic component, 1% HF for 1 hour to remove clay mineral, and 10% HCl for 25 minutes to eliminate fluorosilicates possibly formed. Through a sedimentation procedure, polymineral fine grain fraction in the 4–11 μm range was obtained and then deposited onto 9.8 mm diameter disc. In preparation lab, room lighting is provided by Illford DL10 lamps equipped with Ilford #902 filters.

Equipment

TL for bricks, OSL and IRSL (InfraRed Stimulated Luminescence) for mortars were performed using semi-automated Risø readers (TL-DA-10 and TL-DA-15) with EMI 9235QA photomultipliers (Bøtter-Jensen, 1997; Bøtter-Jensen et al., 2000). TL glow curves were recorded in the TL-DA-10 detection system using Corning 7-59 and Schott BG-12 optical filters. OSL and IRSL signals were obtained using a TL-DA-15 reader equipped, respectively, with 41 blue LEDs (470± 30 nm) and with a laser diode (830 ± 10 nm). The stimulation units delivered about 30 mWcm^–2 for OSL and 240 mWcm^–2 for IRSL at 90% power. Both OSL and IRSL emissions were detected using a Hoya U340 optical filter.

Artificial luminescence signals were induced by two different ⁹⁰Sr-⁹⁰Y calibrated beta sources integrated in the Risø systems delivering, respectively, 6 Gy/min in TL-DA-15 model used for mortars and 1.32 Gy/min in TL-DA-10 model used for bricks. ²⁴¹Am calibrated alpha source delivering 2.7 Gy/min was used to evaluate the luminescence efficiency coefficient k (Tables 3 and 4) necessary to correct the alpha dose contribution to the annual dose (Aitken, 1985; Guibert et al., 2009a).

OSL measurements for mortars

For six aliquots of each sample ED measurements and recovery test as a function of preheating temperature (160–260°C) were made (Choi et al., 2003; Thomas et al., 2003; Kiyak and Canel, 2006; Stella et al., 2013). Each dose value was determined using the modified SAR protocol that provides the measure of the recuperated OSL signal following a zero Gy regenerative dose in a SAR measurement cycle (recuperation test). At the end of each measurements cycle, an optically stimulation for 40 s at 280°C allows hence to check if any thermal transfer of charge was occurred from levels insensitive to light to the OSL traps (Murray and Wintle, 2003).

After the SAR sequence, the same regeneration dose of the first point of beta irradiation is given again to check whether the sensitivity corrected OSL (L_R/T_R)/(L₁/T₁) is reproducible by the recycling ratio R. This value, moreover, identifies the presence of a possible systematic error in the interpolation of L_n/T_n onto the dose–response curves. On the same aliquots, the recovery test was performed. The cycle of SAR modified protocol was repeated 7 times using increasing regeneration doses from 0.5 to 7.5 Gy. For ED calculation, data are acceptable if the recovery test and recycling ratio are within ±10% of unity and recuperation test near to zero (Murray and Wintle, 2003). Being young samples, whose OSL signal is not dominated by the fast component, the integral signal of the first 0.4 s was used as OSL intensity, after subtraction of the background calculated from the following 0.4 s, for minimize the contribution from the medium and slow OSL signal components (Ballarini et al., 2007; Pawley et al., 2010; Gerardi et al., 2012). The uncertainties were quantified on the basis of statistical counting of luminescence signals and applying error propagation. Following the test dose, a cut heat of 160°C was applied. Table 2 shows the temperature range, where the ED has a constant value, and the preheating temperature value chosen for mortars dating. The individual ED values were entered into bins of 0.1 Gy and so the frequency distributions were obtained. Shapiro-Wilk normality test was used to determine if a data set is well-modeled by a normal distribution and to compute how likely it is for a random variable underlying the data set to be normally. The test rejects the hypothesis of normality when the p-value is less than or equal to 0.05. Failing the normality test allows you to state with 95% confidence the data does not fit the normal distribution.

For each sample passing no significant departure from normality was found. So, the frequency distributions were related with Gaussian fits (with standard deviations) and Mean ED values were obtained (Table 3). The results show a good degree of precision for equivalent dose results but in the case of fine grains deposited on discs, their high number (some thousands) will tend to average all signals proportionally to the inverse of the square root of the number of grains. This also implies the reduction of the dispersion between aliquots, so the deviation from a normal distribution is much less observable. The methodology used is then only indicative but not exhaustive to evaluate the bleaching degree of fine grains. In this case, the association between dates of bricks and dates of mortars coupled with historical and architectural survey represents an useful approach to obtain indications about the accuracy of the dating results.

The ED for age calculation was obtained by radial plots analysis (Galbraith et al., 1999; Stella et al., 2013) using RadialPlotter software (Vermeesch, 2009). Fig. 2 shows the radial plot of the dose distribution used to obtain the ED value for each mortar sample. In these graphs the position on the x-axis is a measure of the precision (t/se) with which ED is known. This axis is also expressed in terms of the relative error (se/t, expressed as a percentage). Table 3 reports for each mortar sample the number of aliquots, the k-value and the ED Central Value obtained by radial plot.

Fig. 2

ED distribution obtained with radial plots for mortar samples.

TL measurements for brick

The ED values for bricks were determined using TL measurements and especially the added dose method on the polymineral fine grain phase (Aitken, 1985; Guibert et al., 2009a). For each sample, 24 aliquots were prepared and grouped by six. The first group was used for measuring the natural TL signal. The remaining groups were serially and incrementally irradiated using a calibrated beta source. TL glow curves were recorded by heating the aliquots up to 500°C at a uniform heating rate of 5°C/s in an ultrapure nitrogen environment. The inter-aliquot intensity variations were corrected by second glow normalisation. Temperature region between 310 and 360°C TL (290–380°C plateau) was used for equivalent dose calculation (Q_β). In order to evaluate the possible nonlinearity behaviour of the sample at low artificial beta doses, q_β correction was determined from the intercept of the “second” growth curve behaviour. The same aliquots were then exposed to small and incrementing beta doses and a growth curve was constructed. The intercept of the curve on the dose axis is the correction value for supra-linear growth (q_β) (Aitken, 1985). ED was calculated adding Q_β and q_β values. From artificial luminescence signals induced by calibrated alpha doses the luminescence efficiency coefficient k was determined (Guibert et al., 2009a). All samples showed a growth linear luminescent behaviour vs. dose. For each sample measurements of luminescence loss over time were performed (Aitken,1985). The results obtained exclude the presence of anomalous fading. Table 4 shows k value, Q_β and q_βvalue obtained for each brick sample.

Dose rate measurements

U, Th and K contents (Table 5) were determinate from high resolution gamma spectrometry measurements (HPGe) (Goedicke, 2011). The comparison between the activity of ²³⁸U (deduced from the ²³⁵U and ²³⁴Th gamma emissions) and that of ²²⁶Ra (deduced from ²¹⁴Pb and ²¹⁴Bi γ emissions in equilibrium with ²²²Rn) shows an equilibrium of the U-series (Table 5) (Guibert et al.,2009b). Annual dose components due to radioelements in both bricks and mortars were calculated from radioactive contents by conversion factors of Guérin et al. (2011).

The dose contributions of the sample were corrected on the basis of porosity factor (W) experimentally measured and of saturation factor (F) estimated on the basis of sampling point (height, inside or outside) (Aitken, 1985). In this particular case an F value of 0.3 ± 0.2 was considered. The annual environmental dose rate was measured using TL dosimeters (GR200A) enclosed in capsules placed in situ at the sampling points (Gueli et al., 2009; Stella et al., 2013) adding cosmic radiation calculated according to Prescott and Hutton (1988).

Table 6 shows the contributions to the annual dose for each sample: internal components D_αint and D_βint, corrected for water content, and the environmental contribution D_env.

Age Calculation

ED values obtained on the two FG brick and mortar samples were entered in the following age equation:

where ED is the equivalent dose, k is the alpha efficiency coefficient, D_{α int}, D_{β int} and D_env are, respectively, the annual dose contributions, corrected with humidity and porosity factors (W, F), from the sample itself and from environment.

Table 7 shows, for each sample, the technique and the method used for ED determination, the annual dose rate, the ages and the corresponding AD dates obtained.

Samples preparation and equipment

Mineralogical characterization was based on X-ray diffraction (XRD) performed with a Philips X’Pert Pro PW3710 diffractometer equipped with a CuKα radiation, operating at 40 kV and 20 mA. Non-oriented aggregates of pulverized bulk material were measured between 2 and 60° using a step of 0.02° 2θ and 1.0 s scanning time in each step.

The macroscopic description of the mortars was performed in the laboratory after sampling. The mineralogical characterization of the mortars by XRD was performed later.

Contact spectrophotometry was used for colorimetric characterization of the mortars. The measurements were made with a portable Konica-Minolta CM2600D instrument, equipped with an integrating sphere in the geometry d/8° after the usual procedures for black and white adjustment, selecting an area of 11 mm diameter. Such measures are carried out directly in situ and are non-destructive. For this reason, no sample preparation was required.

Semi-quantitative analysis

After the qualitative interpretation of all the mineral constituents and taking into account the limitations imposed by the method, a semi-quantitative estimation was carried out in order to facilitate the mineralogical interpretation of the results. However it must be kept in mind that the percentages obtained should be only relative indicators of the concentration of the minerals and not absolute values since the associated error is high.

Semi-quantitative analyses were performed by measuring the areas of the peaks corresponding to the typical diffraction maxima spacings of each mineral divided by the respective reflector powers (Schultz, 1964; Moore and Reynolds, 1997; Thorez, 1976; Dias, 1998). The whole occurring phyllosilicates (total percentage) were determined by considering the main peak area at Bragg’s law interatomic spacing d=4.48 Å peak.

All the mineral associations of the analyzed mortar samples present, as essential minerals, a large proportion of quartz followed by calcite and K-feldspars, sometimes with similar values for both these minerals (Table 8, Fig. 3). For all samples, phyllosilicates, in particular clay minerals, as well as dolomite and Na-rich plagioclases are accessory minerals, with the exception of CSFL12 sample in which Na-rich plagioclases reached the 10%.

Fig. 3

Semi-quantitative mineral composition of mortar samples obtained by XRD analysis.

An initial macroscopic observation showed that mortar samples consist mainly of quartz as aggregate and carbonates as cement (binder). In all the analyzed samples the sum of quartz and K-feldspars reaches a mean above 75%. Since some carbonate aggregates (dolostone, dolomitic limestone and limestone fragments) were identified by macroscopic observation, the carbonate composition of the bulk mortar evaluated through XRD corresponds both to the aggregate and the lime paste from the binder, without discrimination. However those carbonate compounds should not influence the binder/aggregate ratio, since they are in a very low occurrence. Furthermore, in what concerns to the binder, this technique only provides the detection of crystalline phases resulting from carbonation and hardening of the original burned limestone that have produced the lime paste (Sanjurjo-Sánchez et al., 2010). Nevertheless, the proportions obtained by XRD semi-quantification confirm the assumption that the sum of Quartz and K-feldspars corresponds to the essential of the aggregate and calcite mostly to the binder (Table 8).

The quartz/K-feldspars ratio is always higher than three. Sample CSFL7 is different by having quartz reaching thirteen times the proportion of K-feldspars, higher proportions of carbonates and noteworthy quantities of phyllosilicates.

Morphometric and textural observations during the initial macroscopic analysis showed a large proportion of mostly subangular to angular silicate sand in all samples. All samples contain also aggregates of variable dimensions below the class of fine gravel, essentially less than 8 mm. Coupled with the XRD data, these characteristics points to a provenance of the aggregates from raw alluvium materials, probably those occurring in the left bank of the Mondego River, where the Convento de S. Francisco is located. The mineralogical composition of the CSFL7 sample is compatible with the same provenance, but with minor amount of carbonate clasts and detrital mica. Since the convent is also at the base of a dolomitic limestone hill, we can assume that a less selected alluvial/colluvial aggregate wasused.

The very scarce presence of phyllosilicates, in particular clay minerals, is remarkable and probably should be explained by the aggregate origin and the technology for the production of mortars. We can assume that raw materials sources were mainly nearby Mondego’s channel deposits comprising a low rate of clay particles, with the above commented exception of sample CSFL7.

The aggregate/binder ratio establishes two groups of samples (Table 8, Fig. 4). CSFL3 and CSFL11 have a quotient higher than six (group II) and for the other samples the quotient is lower than four (group I).

Fig. 4

Sample grouping after the plot of binder vs aggregate proportions.

Colorimetric measurements

Spectral Reflectance Factor of the mortar samples was measured and the CIELAB1976 space coordinates were used for colour specification. The data acquired were relative to the standard observer 10° and D65 illuminant.

Table 9 shows the values of L^∗ (lightness), a^∗ (green-red axis) and b^∗ (blue-yellow axis) obtained for each sample of mortar and cylindrical coordinates C^∗ (Chroma) and h (hue angle) derived (Wyszecki and Stiles, 2000).

The Fig. 5 shows the numerical brightness values (L^∗) associated to hue angle (h) confirming the grouping obtained by XRD analysis. The colour of surface is related to the microscopic structural characteristics and therefore, in cases where it is not possible to extract sufficient material for more accurate characterizations, colorimetry technique could be useful as a tool for mortar classification.

Fig. 5

Level of brightness (L^∗) associated to hue angle (h) for mortar samples.

In this specific case this evidence is validated by the good agreement found between colorimetric and XRD results.