Chronological classification of ancient mortars employing spectroscopy and spectrometry techniques: Sagunto (Valencia, Spain) case.

Forty-two mortar samples, from two archaeological excavations located in Sagunto (Valencian Community, Spain), were analysed by both portable energy dispersive X-ray fluorescence spectroscopy (pED-XRF) and inductively coupled plasma mass spectrometry (ICP-MS) to determine major and minor elements, and traces including rare earth elements (REE). Collected data were crossed with those previously obtained from Sagunto Castle mortars and principal component analysis (PCA) was applied to discriminate the construction phases of the unearthed buildings. REE permitted to ascribe most of the masonries to the Roman Imperial period. Moreover, a statistical model was built by employing partial least squares discriminant analysis (PLS-DA) in order to classify the mortars from Roman Imperial period and from Islamic period, due to the problematic overlapping between these two phases. Results confirmed the effectiveness of the developed indirect chronology method, based on REE data, to discriminate among historic mortars from different construction periods on a wide scale including different Sagunto archaeological sites.


Introduction
Mortar is a building material composed essentially by binder and aggregate fractions and, in some cases, by additives of different types [1][2][3]. In particular, lime mortar had a key function in Roman architecture: Roman people made their walls and structures with this material adding reactive materials, like pozzolan materials (i.e. pulvis puteolanus, cocciopesto), to give a hydraulic character to the mortars [4][5][6][7][8]. The use of mortar in architecture is documented during the Middle Ages and in the following historical periods [6][7][8][9][10][11][12][13].
The chemical analysis by statistical approach, together with mineralogical and petrographic characterization of ancient mortars and polished stones has shown to be a useful tool in the interpretation of the construction phases of several archaeological sites and historical complexes [14][15][16][17][18][19][20]. Moreover, ancient mortars are subjected to decay phenomena which must be also detected and evaluated for conservation issues [21][22][23][24]. This paper shows the results of the analyses carried out on ancient mortars collected from buildings discovered during two recent archaeological excavations at Sagunto, a town located in the Eastern Spain, ca. 30 km north of Valencia, close to the Costa del Azahr on the Mediterranean Sea.
Sagunto is well known in the world for its complex history and its area has been occupied since the Iberian Age. During the Roman period, Sagunto was interested by the construction of important buildings such as the Circus and the Theatre. Hereafter, Sagunto was occupied by Islamic people and during the Modern Ages the city was involved by the Napoleonic Wars [25].
The collected mortar samples were analysed by both portable energy dispersive X-ray fluorescence spectroscopy (pED-XRF) and by an inductively coupled plasma mass spectrometer (ICP-MS) to determine their major elements, minor elements and trace elements. The mineral element concentration and, in particular, the measured contents of rare earth elements (REE) of these samples were compared with those previously obtained from Sagunto Castle [26] applying principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) to determine the construction periods.

Sampling
The studied mortars (Table 1) were sampled during two archaeological excavations located in the city of Sagunto (Figure 1). Twenty-nine samples were collected during the archaeological excavation of Los Huertos St., an area characterized by the presence of buildings and materials from the Roman Imperial phase and from the Islamic occupation in the Middle Ages. Twenty-three samples of lime mortars were collected from these buildings: nine samples come from the Circus (C), a building dating back to the Roman Imperial period, one from the cloaca (CLC), one from a jamb (JMN), two from two different walls (Eastern Wall, EW, and Northern Wall, NW), one from a noria (NR), four from a pilaster (P). Six samples of earth mortar come from two rooms that the archaeologists interpreted as may belonging to the Islamic occupation (Room 1 the samples R1, and Room 2 the samples R2) and in this study were analysed as control samples. Two sample of lime mortar were collected from the Room 3 (R3), two from an unidentified building (UN) and one from a well (WLL). Thirteen samples of lime mortars were collected from different masonries during the emergency archaeological excavation of an area close to Sagunto's Railroad Station (RS named samples). Data of twenty-three lime mortar samples from different buildings of the Castle of Sagunto, studied by Gallello et al. 2017 [26], were taken into account as calibration set to perform data analysis and to date the structures of the above quoted excavations. Part of the calibration samples come from masonries dating back to the Roman Imperial Period: five samples from the Theatre (TR) and one from the Curia (CUR); the other samples come from masonries dated to the Islamic occupation phase: four samples of the first part of the Islamic Wall (MI), one of a wall which was considered modern before the study (MM), one from an Hermitage (ERM), two from two Islamic reworks of the Torre Central Estudiantes (TCE) and two from the tabernae of the Imperial Forum. Moreover, two samples from the Curia (CUR) and five samples from the Basilica of the Imperial Forum (FBI), whose dating was uncertain between the above quoted two phases, were added to the data set.

Major and minor elements determination
All the analyses were carried out on each entire mortar sample previously pulverized (Dmax < 63 m) and homogenized through agatha mortar and pestle. Major and minor element concentrations were obtained by using a S1 Titan energy dispersive portable X-ray fluorescence spectrometer (pED-XRF) from Bruker (Kennewick, Washington DC, USA) equipped with a Rh X-ray tube and X-Flash®SDD detector. Geochem-trace calibration was used to perform the quantitative analyses and S1Sync software from Bruker was employed to measure Al 2 O 3 , SiO 2 , CaO, Ti and Fe. The accuracy of the measurements was verified by using the following certified reference materials: soil NIM GBW07408 and limestone NCS DC60108a ( Table 2). All the reading's standard errors range between 1 and 5 wt%, except for Al 2 O 3 measured on NCS DC60108a, whose standard deviation increase up to 22 wt% for concentration less than 0.5 wt% probably due to the low sensitivity of the instrument for the determination of this element.    solutions for ICP analysis in HNO 3 5%, containing the above-quoted elements at a concentration of 1000 mg/L, were used as stock standards for calibration. 5 mL volumetric flasks were used adding the corresponding volume of standard solutions 0.15 mL of HNO 3 , 0.45 mL of HCl and the water necessary to reach the final volume. The concentration of trace elements ranges from 1 g/L to 600 g/L except for REE, Y and Sc that ranges from 1 g/L to 100 g/L. The measurement accuracy was verified by using the certified reference materials soil NIM GBW07408 and limestone NCS DC60108a. As internal standard, 50 L of a 1000 mg/L Rh solution were added to each sample and to each calibration standard. The analyses were performed though an Agilent 7900 inductively coupled plasma mass spectrometer. The measurement conditions are shown in Table 3 shows the ICP-MS parameters employed for the analyses and Table 4 shows the main analytical features obtained for the measured mass of each considered isotope, including the instrumental detection and quantification limits (LOD and LOQ), and the coefficient of determination (R 2 ) of the corresponding calibration lines.

Statistical data processing
The PCA models were built by using data obtained from a total of forty-two analysed samples and the set of data obtained from samples dated to the Roman Imperial period and the Islamic occupation phase obtained previously by Gallello et al. [26]. Major and minor elements, trace elements including REE were employed as variables. This technique was used to explore large geochemical datasets by reducing the number of variables and providing a deep insight into the structure of the variance of the dataset. Data were processed through mean center and autoscale prior to modelling and the obtained model was cross validated through leave one out method.

Geochemical results
The analytical results of major and minor elements as well as trace elements and REE data are reported in the supplementary materials (Annex 1 and Annex 2). Railroad Station: 20 ± 2 g/g) and Sc concentrations (Los Huertos Street: 1.1 ± 0.3 g/g; Railroad Station: 1.1 ± 0.1 g/g), while samples from Los Huertos Street have a slightly higher concentration of Y than the ones from the Railroad Station excavation (4.1 ± 0.8 g/g and 3.5 ± 0.4 g/g respectively). Concerning the earth mortars from Room 1 and Room 2, they have comparable contents of REE, Y and Sc. REE total amounts go from 37 g/g to 53 g/g, and Y and Sc range from 5 g/g to 10 g/g, and from 2 g/g to 3 g/g respectively.
To give a deep insight in the mechanisms that influence the amount of REE in the mortar samples, it is interesting to look at the correlation among the REE and analysed elements for each group of mortars. Table 5  suggesting that the major contribution in lanthanides comes probably from clay fraction, other silicate minerals and silicate rock clasts.

Table 5 -Pearson correlation coefficient ( ) between ∑REE and each measured element for
the three groups of mortar.

Chemometrics for construction phases discrimination
The identification of the construction phases of the buildings was conducted by comparing the mortars from the Railroad Station and Los Huertos Street to the ones of Sagunto Castle published by Gallello et al. [26] and whose construction periods were retraced in the same work. In particular, this previous study was focused on REE concentrations, due to their proved effectiveness in archaeometric studies about provenance and raw materials of lithic and lithoid archaeological artifacts [28][29][30].
To compare the samples from the excavations to the Imperial and Islamic mortars from Sagunto Castle [26], principal component analysis (PCA) was carried using all variables (i.e. elements) ( Figure 2). Figure 2b shows the contribution of the variables in PC1 (55.00% of samples' variance) and PC2 (13.37% of samples' variance).  Finally, a separation between mortars from Imperial Period (PC1 negative direction), and mortars from the Islamic Period, located in the left lower area of PC can be appreciated. Since the discrimination of lime mortars from Roman Imperial and Islamic buildings seems to be the most problematic issue, principal least squares discriminant analysis (PLS-DA) was employed to build a statistical model to classify uncertain samples. The calibration set and validation set were established by using samples from buildings and masonries that were previously classified by Gallello et al. [26] and by archaeological data as Roman Imperial (C and TR) or as Islamic (MII, TFI and TCE) (Figure 4a). Figure 4b shows the contribution of each variable to the model. As it can be seen, the model clearly separates between mortars of the two different construction phases. Subsequently, the model was applied to a test set composed by uncertain Roman Imperial period or Islamic mortars (Figure. 4a)

Conclusion
The study of the historic mortars from two archaeological excavations (i.e., Los Huertos Street and Railroad Station) located in the city of Sagunto permitted to solve issues that had remained unanswered in the previous studies on these building materials.
The use of multivariate statistics employing REE as variables allowed to classify the lime mortars belonging to the Imperial Roman period, in line with the archaeological data, however the case of the Railroad Station structure needs deeper studies to confirm the chronology as Imperial Roman or Late Roman Age. Also, it was possible to establish the chronological phase of some masonries of Sagunto Castle whose period of construction was uncertain and that can be attributed to the phase of the Islamic occupation or to the Roman Imperial period. PLS-DA model was helpful to discriminate among mortars of uncertain attribution and that can be useful for possible future studies in Sagunto area.
The correlation among REE and major/minor and trace elements are showing that the mechanisms that allowed REE to be discriminating parameters between Sagunto mortars belonging to different periods are related to chemical weathering processes involving limestone clay impurities, aluminosilicates and silicate rock clasts present in the aggregate fraction. Therefore, further works need to be developed focusing in the understanding of the chemical processes that are influencing the REE distributions.
Summarizing, the present study has confirmed the effectiveness of REE data in building materials as markers of different construction phases and has proved the usefulness of the application of ancient mortar analysis on a geographical scale wider than a single archaeological excavation or monuments.

Supplementary materials
Annex I. Concentrations of the measured major and minor elements obtained by pED-XRF.
Annex II. Trace elements and REE results obtained by ICP-MS.

Conflicts of Interest
The authors declare that there is no conflict of interest with any institution or funding body.

Data Availability
The data used to support the findings of this study are included within the article the supplementary information file.

Funding Statement
Authors