Compositional Characterization and Chronology of Roman Mortars from the Archaeological Site of Arroyo De La Dehesa De Velasco (Burgo De Osma- Ciudad De Osma, Soria, Spain)

This study characterises the mortar materials used in the construction of walls and floors at the Arroyo de la Dehesa de Velasco site, located near the Roman city of Uxama Argaela (the modern Burgo de Osma—Ciudad de Osma, Soria, Spain). Multilayer mortars have been characterised by petrographic, mineralogical (X-ray diffraction and scanning electron microscopy with energy dispersive analyses and geochemical analysis (X-ray fluorescence). Additionally, radiocarbon dating of the mortar binder fraction was performed in order to establish the chronology of the building in the absence of other archaeological chronological records. The results showed that similar siliceous aggregates and lime binders were used in the fabrication of multilayer system mortars. Some multilayer wall mortars show ceramic fragments or brick powder to produce hydraulic mortars and improve the resistance to moisture. The raw materials used for the construction of the site were of local origin and the construction was built during the first century BC, according to radiocarbon dating.


Introduction
Mortars are an essential artificial material used since ancient times to support walls and substrates of buildings, and their durability is related to the nature of the components and manufacturing methods. Mortars consist of a mixture of aggregates and binder related both to the geographic location of the building and to the historical context in which the mortars were manufactured [1][2][3][4][5][6][7][8]. The study of mortars offers clues about raw materials and their provenance as well as on the manufacturing technology [9][10][11][12][13][14][15][16][17]. The characteristics of the mortars can be used to determine the chronological sequences of the construction [18][19][20][21] and they can be used for building dating [22][23][24][25][26][27][28]. Mortars also provide keys to restoration and conservation issues [29][30][31][32][33] and for the formulation of new composite materials [34][35][36][37][38][39]. Although the ancient Romans detailed the recipes and technological guidelines for building construction, putting them into practice by the craftsmen was no small feat [40,41]. According to the primary function or the finishing, different technical solutions were applied in historical buildings and constructions and these techniques were extensively used all around the Roman Empire.
The Roman archaeological site of Arroyo de la Dehesa de Velasco was accidentally discovered during the construction of the Duero motorway (A-11) between the towns of Burgo de Osma and

Methods
The entire mortars were extracted parallel to the wall or floor substrate using a water-cooled circular saw. After cleaning, mortar samples were cut perpendicularly to the bedding to observe macroscopic features and the layer profile of the structural mortars, such as colour, thickness, grain shape and size of aggregates.
The mortar component composition, the mineralogy of the binder and the aggregates, grain size and morphological relationships were characterised by petrographic analysis. As thin sections as necessary to complete the total thickness of mortars were obtained. Seventeen thin sections of mortar samples were analysed by Nikon Eclipse LV100pol polarised light microscopy (Nikon, Tokyo, Japan), equipped with a DSF-11 digital camera and a DSL-2 control unit.
A JEOL JSM-6400 Scanning Electron Microscope (SEM) (JEOL, Tokyo, Japan) was used for detailed textural features. Mortar samples were Au-coated to eliminate charging effects. SEM observations were performed in secondary electron (SE) mode.
The mineralogical composition of the coarse binder fraction (<20 m) and the potential neoformed minerals were analysed by X-ray diffraction (XRD). The coarse fraction of the binder was separated by a settling extraction procedure described in Reference [51]. The mortar sample was crumbled up manually and disaggregated in ultrapure water raised up to pH = 8 by adding NH4OH using an ultrasonic bath, and the suspended fraction was extracted from the topmost supernatant to obtain the <20 microns fraction. Mineralogy was determined on a powder sample using a Philips X'Pert diffractometer (Malvern PANalytical, Almelo, The Netherlands) equipped with a graphite monochromator adjusted to Cu-kα1 X-radiation operating at 40 kV and 20 mA. The data collection was performed in a continuous scan ranging from 5° to 70° 2θ at an acquisition rate of 0.02° 2θ per second. The mineral phase identification and semi-quantitative determinations were performed using X'Pert HighScore Plus 3.0 software (Malvern PANalytical).
Major element concentrations of the whole layer samples were determined by wavelength dispersive X-ray fluorescence (WDXRF) using a PANalytical Axios Advanced PW4400 XRF spectrometer (4 kW Rh anode SST-mAX X-ray tube) at the SGIker Facilities. Fused beads were obtained after heating (at ∼1200 °C for 3 min in Pt/Au crucibles) a mix of 0.2 g of sample with 3.8 g of a lithium borate flux (Spectromelt A12, Merck, Kenilworth, NJ, USA) and 2 drops of a lithium bromide solution using a PANalytical Perl'X3 fusion machine. Lower limit of detection for major elements are in the range of 0.01 wt%. The loss on ignition (LOI) was calculated after heating the powder sample at 900 °C.

Methods
The entire mortars were extracted parallel to the wall or floor substrate using a water-cooled circular saw. After cleaning, mortar samples were cut perpendicularly to the bedding to observe macroscopic features and the layer profile of the structural mortars, such as colour, thickness, grain shape and size of aggregates.
The mortar component composition, the mineralogy of the binder and the aggregates, grain size and morphological relationships were characterised by petrographic analysis. As thin sections as necessary to complete the total thickness of mortars were obtained. Seventeen thin sections of mortar samples were analysed by Nikon Eclipse LV100pol polarised light microscopy (Nikon, Tokyo, Japan), equipped with a DSF-11 digital camera and a DSL-2 control unit.
A JEOL JSM-6400 Scanning Electron Microscope (SEM) (JEOL, Tokyo, Japan) was used for detailed textural features. Mortar samples were Au-coated to eliminate charging effects. SEM observations were performed in secondary electron (SE) mode.
The mineralogical composition of the coarse binder fraction (<20 µm) and the potential neoformed minerals were analysed by X-ray diffraction (XRD). The coarse fraction of the binder was separated by a settling extraction procedure described in Reference [51]. The mortar sample was crumbled up manually and disaggregated in ultrapure water raised up to pH = 8 by adding NH 4 OH using an ultrasonic bath, and the suspended fraction was extracted from the topmost supernatant to obtain the <20 microns fraction. Mineralogy was determined on a powder sample using a Philips X'Pert diffractometer (Malvern PANalytical, Almelo, The Netherlands) equipped with a graphite monochromator adjusted to Cu-kα1 X-radiation operating at 40 kV and 20 mA. The data collection was performed in a continuous scan ranging from 5 • to 70 • 2θ at an acquisition rate of 0.02 • 2θ per second. The mineral phase identification and semi-quantitative determinations were performed using X'Pert HighScore Plus 3.0 software (Malvern PANalytical).
Major element concentrations of the whole layer samples were determined by wavelength dispersive X-ray fluorescence (WDXRF) using a PANalytical Axios Advanced PW4400 XRF spectrometer (4 kW Rh anode SST-mAX X-ray tube) at the SGIker Facilities. Fused beads were obtained after heating (at ∼1200 • C for 3 min in Pt/Au crucibles) a mix of 0.2 g of sample with 3.8 g of a lithium borate flux (Spectromelt A12, Merck, Kenilworth, NJ, USA) and 2 drops of a lithium bromide solution using a PANalytical Perl'X3 fusion machine. Lower limit of detection for major elements are in the range of 0.01 wt%. The loss on ignition (LOI) was calculated after heating the powder sample at 900 • C. The radiocarbon dating of mortar samples was performed on the finer binder fraction corresponding to 0.5-<2 µm fraction of a carefully selected layer of multilayer plaster by the 14 C accelerator mass spectrometry (AMS) at Beta Analytic Inc. (Miami, FL, USA) and 14 C conventional ages were calibrated by OxCal v4.2.3 software [52] and the Intcal13 atmospheric calibration curve [53]. The finest binder fraction was obtained by a settling extraction following the procedure described in References [49,51].

Results and Discussion
The macroscopic characteristics are summarised in Table 1. The macroscopic characteristics of the floor mortars are different since Sample AD-192 shows a multilayer system whereas AD-114 is a single layer mortar. Mortars from the walls all consist of a multilayer system or multi-layered plasters formed by three to six layers. Samples AD-132, AD-134, AD-243 and AD-058 are formed by three layers and AD-124 and AD-242 by more than five layers ( Figure 2). Wall mortars are composed of a fine light-coloured matrix with coarse-grained rock fragments visible to the naked eye as aggregates. Rock fragments are mainly sub-rounded in shape, of variable colour, from milky to yellow-orange-brown, and composed mainly of siliceous detrital sands (quartz and quartzites) and a minor amount of subangular limestones and sandstones. Voids are common both parallel to the layers and randomly oriented, probably formed by the dissolution of grass stalks. The binder is light in colour from creamy to orangish, and white lumps or scarcely mixed lime inclusions [54][55][56] are abundant, ranging from macroscopic to microscopic scale. Vegetable fibre remains, probably straw, are observed in most mortars, particularly in the inner layers. In most cases, Minerals 2020, 10, 393 5 of 17 cohesion and adhesion of layers in the mortars is excellent. Only two wall mortars show reddish-pinkish layers due to the occurrence of ceramic fragments and/or powder, known as cocciopesto.
All wall plasters display an outermost, very-thin whitish pictorial layer (1-2 mm) known as the marmorino layer (L0) [40,[57][58][59]. Towards the inside, two layers (L1, L2) can be distinguished composed of sandy siliceous aggregates of whitish-greyish colour in a calcitic binder. Each layer 2 cm in thickness contains homogeneous grain-size aggregates, and both layers have similar aggregate compositions differing in aggregate sorting. The outermost layer (L1) displays a fine grain size, whereas the innermost layer (L2) shows a coarse grain size. Samples AD-124 and AD-242 consist of six and five layers respectively, and are characterised by reddish-pinkish layers enriched in crushed ceramic materials intentionally added. Besides, these mortars show a clayey brownish scratch coat layer in contact with the wall substrate or a joining coat composed of subangular quartz and feldspars with scarce lime lumps.  All wall plasters display an outermost, very-thin whitish pictorial layer (1-2 mm) known as the marmorino layer (L0) [40,[57][58][59]. Towards the inside, two layers (L1, L2) can be distinguished composed of sandy siliceous aggregates of whitish-greyish colour in a calcitic binder. Each layer 2 cm in thickness contains homogeneous grain-size aggregates, and both layers have similar aggregate compositions differing in aggregate sorting. The outermost layer (L1) displays a fine grain size, whereas the innermost layer (L2) shows a coarse grain size. Samples AD-124 and AD-242 consist of six and five layers respectively, and are characterised by reddish-pinkish layers enriched in crushed ceramic materials intentionally added. Besides, these mortars show a clayey brownish scratch coat layer in contact with the wall substrate or a joining coat composed of subangular quartz and feldspars with scarce lime lumps.
The two floor mortar samples are quite different: Sample AD-192 corresponds to a multilayer system consisting of three layers with a light-coloured matrix. The outermost layer (L0) is a marmorino-like layer. The upper layer (L1) consist of aggregates of crushed limestone, homogeneous in grain size, whereas the inner layer (L2) shows rounded siliceous pebbles of more heterometric grain size ( Figure 2). Layer L2 additionally contains moulds of the large pebbles probably used in the statumen or a preparatory bottom layer formed of pebbles in order to stabilize the ground [40,[57][58][59]. Sample AD-114 is characterised by only one sandy-clayey brownish layer with rounded aggregates of siliceous nature and sand-grain size.
The petrographic analysis is able to characterise the aggregate mineral composition and the binder features (Table 1) [60]. Wall plaster aggregates possess a grain size ranging from fine to coarse sand, according the classification in Reference [61], and subangular to subrounded shape [62]. Their nature corresponds to quartz, with a minor amount of potassium feldspar, and the rock fragments are mainly quartzite, limestones and arenites. Quartz and quartzites generally possess a high sphericity index and subrounded shapes, whereas limestone and sandstone fragments are subangular with low sphericity. Three-layer wall plasters show grain size grading of aggregates from coarse grain size in the inner layer, L2, to fine grain size in the outer layer, L1. Besides, each layer shows a distinctive sorting leading to layering. Vegetable fibre remains are usually observed in some layers, mainly in the inner layers. Multilayer wall mortars show mainly similar features related to grain size distribution ( Figure 3). In contrast, distinctive aggregate sorting patterns are observed in each layer, varying between poorly to moderately sorted [63,64]. The ceramic fragments in the cocciopesto layer are poorly sorted showing angular shapes and low sphericity and sometimes showing reaction rims. The outermost marmorino layer shows angular carbonate aggregates embedded in a micritic matrix with an aggregate/binder ratio of <1 or nearly 1 (Figure 3). The innermost clayey mortars corresponding to the scratch coats or joining coats differ from the other wall mortar layers in the aggregate nature and grain size distribution. This layer is characterised by a high density of low-roundness quartz and potassium feldspar aggregates of homogeneous grain size showing an aggregate/binder ratio of >1.
multilayer system of three layers. The floor AD-114 shows a homogeneous medium-grain size of subangular quartz aggregates in a matrix composed of a mixture of lime and clay with an aggregate/binder ratio of >1. Sample AD-192 shows a clear grading upwards from medium-pebble grain size in the innermost layer to silt in the marmorino layer. The layers contain aggregates of different nature. The innermost layer, L2, is composed of centimetric very rounded quartz and quartzite rock fragments, whereas layer L1 is formed by centimetric angular limestone fragments of heterogeneous origin, such as nodulous limestones, miliolid limestones, bioclastic mudstones and recrystallised limestones [60] The outermost marmorino layer shows angular carbonate aggregates embedded in a micritic matrix. The multilayer system in the walls and in the floors was created according to the building methods documented by Vitruvius [40] and used in numerous Roman sites. The mortar technology usually consisted of two or three layers under the pictorial layer promoting a suitable support for wall paintings. The wall mortars were composed of a mixture of lime and sands, with an aggregate/binder ratio of 2:1 or 3:1, with a slight increase in binder content towards the outside. The cocciopesto layers were probably manufactured to confer hydraulicity to mortars, improving the strength and moisture resistance [65][66][67]. Similar technology was used for floor substrates, which were manufactured according to recipes described by ancient authors [40,41]. Floor mortars consist of an inner layer formed by a mixture of natural aggregates and binder with an aggregate/binder ratio of >1 and an upper layer consisting of a mixture of crushed coarse limestone aggregates and binder, with an aggregate/binder ratio of <1. No crushed bricks were used in the floor layering. Both wall and floor structures were finished with a thin lime, with a euhedral calcite or limestone fragments layer or a marmorino layer (Figure 3). The microtextural feature of the marmorino layer can be observed in detail by SEM, revealing a microcrystalline calcite (Figure 4a). SEM analyses also Floor mortars are different since Sample AD-114 has a single layer and Sample AD-192 has a multilayer system of three layers. The floor AD-114 shows a homogeneous medium-grain size of subangular quartz aggregates in a matrix composed of a mixture of lime and clay with an aggregate/binder ratio of >1. Sample AD-192 shows a clear grading upwards from medium-pebble grain size in the innermost layer to silt in the marmorino layer. The layers contain aggregates of different nature. The innermost layer, L2, is composed of centimetric very rounded quartz and quartzite rock fragments, whereas layer L1 is formed by centimetric angular limestone fragments of heterogeneous origin, such as nodulous limestones, miliolid limestones, bioclastic mudstones and recrystallised limestones [60] The outermost marmorino layer shows angular carbonate aggregates embedded in a micritic matrix.
The multilayer system in the walls and in the floors was created according to the building methods documented by Vitruvius [40] and used in numerous Roman sites. The mortar technology usually consisted of two or three layers under the pictorial layer promoting a suitable support for wall paintings. The wall mortars were composed of a mixture of lime and sands, with an aggregate/binder ratio of 2:1 or 3:1, with a slight increase in binder content towards the outside. The cocciopesto layers were probably manufactured to confer hydraulicity to mortars, improving the strength and moisture resistance [65][66][67]. Similar technology was used for floor substrates, which were manufactured according to recipes described by ancient authors [40,41]. Floor mortars consist of an inner layer formed by a mixture of natural aggregates and binder with an aggregate/binder ratio of >1 and an upper layer consisting of a mixture of crushed coarse limestone aggregates and binder, with an aggregate/binder ratio of <1. No crushed bricks were used in the floor layering. Both wall and floor structures were finished with a thin lime, with a euhedral calcite or limestone fragments layer or a marmorino layer (Figure 3). The microtextural feature of the marmorino layer can be observed in detail by SEM, revealing a microcrystalline calcite ( Figure 4A). SEM analyses also allow the observation of some vegetal ghosts since druse habit calcium carbonate is formed, covering the pre-existing vegetal remains ( Figure 4B). allow the observation of some vegetal ghosts since druse habit calcium carbonate is formed, covering the pre-existing vegetal remains (Figure 4b). Multilayer systems are common in important buildings and were a common practice in ancient Roman territories [59,[68][69][70], applied under pictorial-bearing walls and on floor substrates to promote their durability and conservation [71,72]. Although very little remains of the building have been found, the use of these constructive techniques indicates the social and/or economic significance of the Arroyo de la Dehesa de Velasco site. Table 2 summarises the XRD mineralogical composition of the coarse fraction (<20 m) of the mortar binder. The binder fraction is composed mainly of calcite and, in minor amounts, phyllosilicates and quartz. Clay minerals are represented by kaolinite and illite. SEM images show a planar-habit slightly-folded kaolinite-like mineral (Figures 4c and 5). The lack of clay minerals in some samples does not mean that they are totally absent but that they may be under the detection limit.
The mineralogy of floor mortar binder agrees with the petrological study. Single layer sample AD-114 shows a large amount of phyllosilicates composed of a mixture of illite and kaolinite minerals according to the sandy-clayey nature of the mortar ( Table 2). In contrast, the binder of Sample AD-192 is formed by calcite. For the three-layer wall mortars, only one whole sample has been analysed by XRD because layering is attributable to grain size distribution, as the petrographic study has revealed. Among the three-layer samples, the binder of AD-058 stands out because of its high phyllosilicate content, mainly kaolinite, whereas the binders of the other samples are only formed by calcite. The multilayer wall mortars AD-124 and AD-242 also show a large amount of phyllosilicates. In most layers, kaolinite is more abundant than illite, and illite is more abundant than kaolinite only in cocciopesto layers. The large amount of illite in cocciopesto layers is because of the intentionally added ceramic fragments. The occurrence of kaolinite in multilayer mortars with cocciopesto layers can suggest intentionality in mortar manufacture. Multilayer systems are common in important buildings and were a common practice in ancient Roman territories [59,[68][69][70], applied under pictorial-bearing walls and on floor substrates to promote their durability and conservation [71,72]. Although very little remains of the building have been found, the use of these constructive techniques indicates the social and/or economic significance of the Arroyo de la Dehesa de Velasco site. Table 2 summarises the XRD mineralogical composition of the coarse fraction (<20 µm) of the mortar binder. The binder fraction is composed mainly of calcite and, in minor amounts, phyllosilicates and quartz. Clay minerals are represented by kaolinite and illite. SEM images show a planar-habit slightly-folded kaolinite-like mineral ( Figures 4C and 5). The lack of clay minerals in some samples does not mean that they are totally absent but that they may be under the detection limit. The mineralogy of floor mortar binder agrees with the petrological study. Single layer sample AD-114 shows a large amount of phyllosilicates composed of a mixture of illite and kaolinite minerals according to the sandy-clayey nature of the mortar ( Table 2). In contrast, the binder of Sample AD-192 is formed by calcite. For the three-layer wall mortars, only one whole sample has been analysed by XRD because layering is attributable to grain size distribution, as the petrographic study has revealed. Among the three-layer samples, the binder of AD-058 stands out because of its high phyllosilicate content, mainly kaolinite, whereas the binders of the other samples are only formed by calcite. The multilayer wall mortars AD-124 and AD-242 also show a large amount of phyllosilicates. In most layers, kaolinite is more abundant than illite, and illite is more abundant than kaolinite only in cocciopesto Crushed ceramics were added to confer hydraulic properties to non-hydraulic mortars. The reaction rim surrounding the crushed ceramics suggests the occurrence of aluminosilicate hydrated products formed by the reaction aluminosilicate phases in the presence of water with the calcium hydroxide. Although, hydraulic newformed phases were not easy to identify by XRD due to their low crystallinity and low concentration in the binder [73][74][75]. However, in the binder of cocciopesto layers, reflections at 7.03 Å and 3.51 Å can be attributed to amesite (Mg2 Al2 SiO5 (OH)4), which is a magnesium aluminosilicate hydrate phase (M-A-S-H) (Figure 5a). On the contrary, these reflections are not observed in the binder of layers without ceramic fragments, where only calcite and quartz were identified (Figure 5b).  Crushed ceramics were added to confer hydraulic properties to non-hydraulic mortars. The reaction rim surrounding the crushed ceramics suggests the occurrence of aluminosilicate hydrated products formed by the reaction aluminosilicate phases in the presence of water with the calcium hydroxide. Although, hydraulic newformed phases were not easy to identify by XRD due to their low crystallinity and low concentration in the binder [73][74][75]. However, in the binder of cocciopesto layers, reflections at 7.03 Å and 3.51 Å can be attributed to amesite (Mg 2 Al 2 SiO 5 (OH) 4 ), which is a magnesium aluminosilicate hydrate phase (M-A-S-H) (Figure 5a). On the contrary, these reflections are not observed in the binder of layers without ceramic fragments, where only calcite and quartz were identified (Figure 5b).
Cocciopesto mortars have often been used for waterproofing cisterns and waterways [6,66,75,76]. The composite mortar made by packing cocciopesto-bearing and cocciopesto-free layers not only confer resistance to the moisture but also improve the mechanical properties and stability of mortars [67,77].
The chemical composition of all the mortar samples is detailed in Table 3. The chemical composition is related to the composition of aggregates and binder. Mortars show a high content of silica, ranging from 43 wt% to 64 wt%, and of lime, ranging from 16 wt% to 27 wt%, according to the siliceous aggregate and binder nature, respectively. Al 2 O 3 ranges between 0.8 wt% and 8.8 wt%, Fe 2 O 3 t between 0.2 wt% and 2.8 wt% and K 2 O between 0.3 wt% and 1.6 wt%.
Cocciopesto layers in the multilayer AD-124 and AD-242 mortars display the highest values of aluminum, iron, potassium and titanium, owing to the occurrence of ceramic and tile fragments. However, other layers in these samples also display relatively high contents of these elements, suggesting the presence of ceramic powder in the mortar. Kaolinite-rich sample AD-058 does not show high aluminum values, suggesting that the mineralogical contents determined by XRD are significantly over-quantified. The two multilayer mortars are quite different in their chemical composition, and the higher calcium content of the AD-124-1 layer is because of the lump abundance. The compositional variation of mortars is plotted in binary diagrams ( Figure 6). Diagrams of CaO/Al 2 O 3 ratio versus oxides allow to distinguish two groups of samples. The group with low CaO/Al 2 O 3 ratio shows high contents of iron and potassium and low silica content corresponding to clay-rich mortars and cocciopesto mortars, and the group with high CaO/Al 2 O 3 ratio displays low iron and potassium contents, and a high content in silica corresponding to siliceous sandy mortars. The CaO content can be related with the variable amount of lumps in some layers. Nevertheless, the lime content is similar in both groups and is not a distinctive chemical feature. Thus, geochemical differences accord with the mineralogical features of the mortars. Hierarchical clustering analysis on the basis of chemical data discriminates two main clusters of mortars (Figure 7). The first cluster groups AD-124-2, AD-242-2, AD-242-4, AD-242-3 and AD-124-5 samples enriched in Al2O3, TiO2, Fe2O3, MgO and K2O corresponding to layers with ceramic fragments and to clay-rich scratch layers. Two sub-clusters can be distinguished, one including layers AD-124-2 and AD-242-2 related to ceramic-rich aggregate layers, and the other including AD-242-4, AD-124-5 clay- Hierarchical clustering analysis on the basis of chemical data discriminates two main clusters of mortars (Figure 7). The first cluster groups AD-124-2, AD-242-2, AD-242-4, AD-242-3 and AD-124-5 samples enriched in Al 2 O 3 , TiO 2 , Fe 2 O 3, MgO and K 2 O corresponding to layers with ceramic fragments and to clay-rich scratch layers. Two sub-clusters can be distinguished, one including layers AD-124-2 and AD-242-2 related to ceramic-rich aggregate layers, and the other including AD-242-4, AD-124-5 clay-rich layers and the AD-242-3 layer with a ceramic powder-rich layer. The second cluster includes Samples AD-058, AD-134, AD-114, AD-243-1, AD-243-2, AD-132, AD-124-1, AD-124-3, AD-124-4 and AD-242-1 corresponding to three-layer siliceous mortars, to the innermost siliceous layer of the floor mortar and clay-free layers of multilayer mortars. Additionally, two sub-clusters can be distinguished, one group displays high SiO 2 and low CaO contents (AD-058, AD-134, AD-114, AD-243-1 and AD-243-2 layers), and the second group shows higher CaO and lower silica content related to the large amount of carbonate-nature aggregates and lime lumps (AD-124-1, AD-124-3, AD-124-4, AD-132 and AD-242-1 layers). According to the mineralogical and petrographic results, raw material typology and provenance issues can be resolved. Siliceous sands, limestones and ceramic fragments were mainly used for the mortar. Various Cretaceous limestones outcropping near Uxama Argaela (Burgo de Osma-Ciudad de Osma) were probably the source for the production of lime. Moreover, limestone aggregates used for the floor mortars, corresponding to bioclastic limestone with miliolid fragments, indicates that Hortezuelos and Hontoria del Pinar limestones located to the south of Burgo de Osma could be the provenance area (Figure 8). Siliceous nature aggregates are the other important materials in the mixture. These correspond to sands composed mainly of rounded quartz and quartzites. The source of such sands can be ascribed to Cenozoic detrital materials, which are very abundant around the site. However, the exact raw material provenance is difficult to establish because of the monotony of facies, scarcity of fossil record and the lack of outcrop continuity. Besides, Quaternary fluvial materials from the Abión River, which are mostly composed of quartz and quartzites, could also be used as aggregates in the mortar manufacture. The lack of particular markers makes it difficult to fix the According to the mineralogical and petrographic results, raw material typology and provenance issues can be resolved. Siliceous sands, limestones and ceramic fragments were mainly used for the mortar. Various Cretaceous limestones outcropping near Uxama Argaela (Burgo de Osma-Ciudad de Osma) were probably the source for the production of lime. Moreover, limestone aggregates used for the floor mortars, corresponding to bioclastic limestone with miliolid fragments, indicates that Hortezuelos and Hontoria del Pinar limestones located to the south of Burgo de Osma could be the provenance area (Figure 8). Siliceous nature aggregates are the other important materials in the mixture. These correspond to sands composed mainly of rounded quartz and quartzites. The source of such sands can be ascribed to Cenozoic detrital materials, which are very abundant around the site. However, the exact raw material provenance is difficult to establish because of the monotony of facies, scarcity of fossil record and the lack of outcrop continuity. Besides, Quaternary fluvial materials from the Abión River, which are mostly composed of quartz and quartzites, could also be used as aggregates in the mortar manufacture. The lack of particular markers makes it difficult to fix the provenance area because lithologies of different ages are mainly composed of similar detrital mineral phases. Nevertheless, the occurrence of kaolinite allows the provenance of these aggregates to be determined. Kaolinite-bearing detrital materials of Oligocene age were described near the site [78], pointing to these materials as the probable provenance for the aggregates. Thus, of the materials used to elaborate the wall plasters and floor mortars, both lime and aggregates came from the surrounding geological materials. The use of local materials and the multi-layered manufacture follow the guidelines described by Vitruvius. The lack of written sources and archaeological evidence at Arroyo de La Dehesa de Velasco mean that the date of the building complex is unknown. According to archaeological criteria, two samples were selected for radiocarbon dating, Sample AD-132 from Room A, corresponding to the original constructive mortar, and Sample AD-242 from Room C, corresponding to an improvement of the building. Sample AD-242 presents layers with cocciopesto, but these layers have been avoided since dating mortars with cocciopesto can be problematic, especially due to some neoformation minerals that can interfere [48,49].
Binder fraction between 0.5 and <2 µ m of two mortar samples was extracted for radiocarbon dating following the procedures described in References [49,51]. The siliceous L2 layer of Sample AD-132 and the L1 layer of Sample AD-242 were selected to obtain the suitable binder fraction. Prior to dating, the binder fraction XRD was performed to assure the suitability of the selected fraction ( Figure 9). Traces of The lack of written sources and archaeological evidence at Arroyo de La Dehesa de Velasco mean that the date of the building complex is unknown. According to archaeological criteria, two samples were selected for radiocarbon dating, Sample AD-132 from Room A, corresponding to the original constructive mortar, and Sample AD-242 from Room C, corresponding to an improvement of the building. Sample AD-242 presents layers with cocciopesto, but these layers have been avoided since dating mortars with cocciopesto can be problematic, especially due to some neoformation minerals that can interfere [48,49].
Binder fraction between 0.5 and <2 µm of two mortar samples was extracted for radiocarbon dating following the procedures described in References [49,51]. The siliceous L2 layer of Sample AD-132 and the L1 layer of Sample AD-242 were selected to obtain the suitable binder fraction. Prior to dating, the binder fraction XRD was performed to assure the suitability of the selected fraction ( Figure 9). Traces of phyllosilicates leading to contamination are concentrated in the >2 µm and <0.5 µm grain-size fractions (Figure 9a,c). The 0.5-2 µm fraction does not show these potential contaminants; therefore, this binder fraction was selected for dating ( Figure 9b). The result of 14 C dating of the AD-242 sample is 2120 ± 30 BP corresponding to the calendar age of 206-50 BC (Table 4, Figure 10a) and for the AD-132 sample is 2130 ± 30 BP corresponding to the calendar age of 210-52 BC (Figure 10b). Archaeological remains indicating occupation date of the site are very scarce, but several historical milestones allow to limit the range of building age. According to the historical sources, the Romanisation of the region could not begin until after the conquest of Numantia in 133 BC. Besides, Uxama Argaela was burned down in 72 BC by Pompey for supporting Sertorius during the first republican civil war. Both historical milestones limit the oldest age of building age to ca 70 BC. Among the scarce archaeological remains, some fragments of terra sigillata dated between 40-80 AD corresponding to abandonment of the building indicate that the construction of the building must indubitably be earlier. In summary, according to historical data, the age of the building must be before 40 AD (age of abandonment) and later to 133 BC, corresponding to the beginning of the Romanisation on the region. Considering a bayesian approach, the most probable age of construction would be before 50 BC. Both the original construction (sample AD-132) and improvement (sample AD-242) dates of the building are close, suggesting that the bath complex of Arroyo de La Dehesa de Velasco was built during the first century BC. The result of 14 C dating of the AD-242 sample is 2120 ± 30 BP corresponding to the calendar age of 206-50 BC (Table 4, Figure 10A) and for the AD-132 sample is 2130 ± 30 BP corresponding to the calendar age of 210-52 BC ( Figure 10B). Archaeological remains indicating occupation date of the site are very scarce, but several historical milestones allow to limit the range of building age. According to the historical sources, the Romanisation of the region could not begin until after the conquest of Numantia in 133 BC. Besides, Uxama Argaela was burned down in 72 BC by Pompey for supporting Sertorius during the first republican civil war. Both historical milestones limit the oldest age of building age to ca 70 BC. Among the scarce archaeological remains, some fragments of terra sigillata dated between 40-80 AD corresponding to abandonment of the building indicate that the construction of the building must indubitably be earlier. In summary, according to historical data, the age of the building must be before 40 AD (age of abandonment) and later to 133 BC, corresponding to the beginning of the Romanisation on the region. Considering a bayesian approach, the most probable age of construction would be before 50 BC. Both the original construction (sample AD-132) and improvement (sample AD-242) dates of the building are close, suggesting that the bath complex of Arroyo de La Dehesa de Velasco was built during the first century BC.   [52] and IntCal09 atmospheric data [53].

Conclusions
The results of this study have provided information about the technological and chronological aspects of the mortars at the Arroyo de la Dehesa de Velasco site.
The mortars were manufactured in a multilayer system formed by several layers with different thicknesses. Ceramic fragments or powder were added to improve resistance to moisture and the durability of the walls. This multilayer approach agrees with Roman technology recipes written by classical authors and applied by skilled artisans.
Petrographic and mineralogical studies allow to determine the raw materials' provenance area. The only outcrop of Mesozoic limestones in the surroundings of the site, close to Ciudad de Osma-Burgo de Osma, were used to elaborate both lime and aggregates in the floor substrates. Both aggregates Cenozoic and Quaternary siliceous sands from areas close to the site were used. Moreover, the occurrence of kaolinite allows to set the sand provenance to the Oligocene materials.
The petrographic study reveals that the craftsmen followed the theoretical Roman technical guidelines with minor or no variations.
The binder 0.5-2 µ m fraction of layers without ceramic fragments were selected to perform radiocarbon dating, indicating that the building construction occurred during the first century BC.
Author Contributions: J.J.E. and C.A.F. executed the archaeological study, the sampling strategy and collected the samples; A.A.-O. performed the petrographic analysis and wrote the paper; M.C.Z. carried out the XRD mineralogical study and contributed to writing and reviewing the paper; L.A.O. arranged the funding, the project administration, the geochemical study and contributed to writing and reviewing the paper; G.P.-A. performed the sample preparation and mineralogical study for radiocarbon dating. All authors contributed to discussion of results of this study. All authors have read and agreed to the published version of the manuscript.
Funding: This study was partially supported by the T1193-13 project of the Basque Country Government.

Acknowledgments:
We thank anonymous reviewers for constructive comments and suggestions which improved the manuscript. The authors would also like to thank Peter Smith for language assistance.

Conflicts of Interest:
The authors declare no conflict of interest regarding this manuscript.  [52] and IntCal09 atmospheric data [53].

Conclusions
The results of this study have provided information about the technological and chronological aspects of the mortars at the Arroyo de la Dehesa de Velasco site.
The mortars were manufactured in a multilayer system formed by several layers with different thicknesses. Ceramic fragments or powder were added to improve resistance to moisture and the durability of the walls. This multilayer approach agrees with Roman technology recipes written by classical authors and applied by skilled artisans.
Petrographic and mineralogical studies allow to determine the raw materials' provenance area. The only outcrop of Mesozoic limestones in the surroundings of the site, close to Ciudad de Osma-Burgo de Osma, were used to elaborate both lime and aggregates in the floor substrates. Both aggregates Cenozoic and Quaternary siliceous sands from areas close to the site were used. Moreover, the occurrence of kaolinite allows to set the sand provenance to the Oligocene materials.
The petrographic study reveals that the craftsmen followed the theoretical Roman technical guidelines with minor or no variations.
The binder 0.5-2 µm fraction of layers without ceramic fragments were selected to perform radiocarbon dating, indicating that the building construction occurred during the first century BC.
Author Contributions: J.J.E. and C.A.F. executed the archaeological study, the sampling strategy and collected the samples; A.A.-O. performed the petrographic analysis and wrote the paper; M.C.Z. carried out the XRD mineralogical study and contributed to writing and reviewing the paper; L.A.O. arranged the funding, the project administration, the geochemical study and contributed to writing and reviewing the paper; G.P.-A. performed the sample preparation and mineralogical study for radiocarbon dating. All authors contributed to discussion of results of this study. All authors have read and agreed to the published version of the manuscript.
Funding: This study was partially supported by the T1193-13 project of the Basque Country Government.