Archaeological and Chemical Investigation on the High Imperial Mosaic Floor Mortars of the Domus Integrated in the Museum of Archaeology D. Diogo de Sousa, Braga, Portugal

: This paper intends to characterize the ﬂoor mortar layers ( nucleus , rudus and statumen ) of the high imperial mosaics of the domus integrated in the Museum of Archeology D. Diogo de Sousa, the oldest roman housing testimonies known in Braga, Portugal. It offers an important archaeological and historical contextualization and ﬁrst chemical characterization attempt on the mortars. The study of 13 mortar samples was carried out at a chemical level through X-ray ﬂuorescence spectroscopy (XRF). All samples presented low lime content when compared to similar studies. A high chemical similarity between nucleus mortars ( opus signinum ) and chemical composition differences between rudus and statumen mortars was determined, conﬁrmed by statistical analyses. Their composition was distinctly related to the stratigraphic position of each ﬂoor mortar layer, following Vitruvius’ model, and to the external conditions and treatments (e.g., capillary rise with soluble salts and application of chemical treatments), to which they were submitted.


Introduction
Roman mortars are noted for their high durability and complex technological knowledge, related with its composition and execution methods. Their study provides fundamental information regarding the knowledge of the archaeological sites. Some studies have been developed on the characterization of roman mortars in Portuguese archaeological sites, such as Beja-Pisões, Tróia, Conimbriga and Marvão-Anmaia [1][2][3][4]. The roman floor mosaics' substrate has several preparatory mortar layers, carefully built (in terms of number, thickness and composition). Specifically, these kind of mortar substrates are under study elsewhere, such as Greece, Italy, Slovenia and Spain [5][6][7][8][9][10][11][12]. However, in Bracara Augusta, the Roman name of Braga (Portugal), the roman mortars from its archaeological sites are poorly analyzed and there is no known bibliographic references concerning their characterization.
The Roman city of Bracara Augusta, and its evolution between the end of the first century BC and Late Antiquity (V-VII centuries) is known based on the numerous remains recovered by urban archaeology over the last 40 years, in the context of the "Campo Arqueológico de Braga" (CAB; Braga archeological Campus) and the Bracara Augusta Project, created in 1976 [13], within the scope of preventive archaeology interventions, conducted by the Archaeology Unit of the University of Minho (UAUM) and the Archaeology Office of the Municipality of Braga (GACMB). In this process, the role of the Museum of Archaeology D. Diogo de Sousa (MDDS), by its contribution in supporting research to the defense and preservation of the archaeological heritage of Braga, must be highlighted. Braga is a city Figure 1. Location of the Stables Archaeological Site in the plan of the roman city, low imperial and medieval walls. Adapted from [19]. This is the largest excavated area in the southern part of the ancient Roman city and currently houses the facilities of the MDDS, open to the public since 2006. In this place, there were two rectangular buildings, used as stables, dating from the beginning of the 20th century, and located at an elevation of approximately 184 m above sea level. These buildings are surrounded in the west by the Bombeiros Voluntários Street and in the east by an unbuilt area [19,20]. The archaeological work was spread over four excavation campaigns that took place between 1986 and 2002. The first intervention occurred in 1986, to assess the archaeological potential of the area (three surveys) [16]. Between 1988 and 1989, there was a second campaign, covering the entire area of the Museum implantation (24 surveys) [20], followed by a third one, in 1996, extending the intervention to the Museum's gardens [19]. Finally, the fourth and last campaign took place in 2002, at a time when the Museum was already built, in the gardens, on the eastern part of the building [19]. The work carried out allowed the identification of the archaeological remains belonging to several insulae of the Roman city, some of which had housing functionalities, chronologically set between the 1st and the 7th century, allowing us to understand the  [19]. This is the largest excavated area in the southern part of the ancient Roman city and currently houses the facilities of the MDDS, open to the public since 2006. In this place, there were two rectangular buildings, used as stables, dating from the beginning of the 20th century, and located at an elevation of approximately 184 m above sea level. These buildings are surrounded in the west by the Bombeiros Voluntários Street and in the east by an unbuilt area [19,20]. The archaeological work was spread over four excavation campaigns that took place between 1986 and 2002. The first intervention occurred in 1986, to assess the archaeological potential of the area (three surveys) [16]. Between 1988 and 1989, there was a second campaign, covering the entire area of the Museum implantation (24 surveys) [20], followed by a third one, in 1996, extending the intervention to the Museum's gardens [19]. Finally, the fourth and last campaign took place in 2002, at a time when the Museum was already built, in the gardens, on the eastern part of the building [19]. The work carried out allowed the identification of the archaeological remains belonging to several insulae of the Roman city, some of which had housing functionalities, chronologically set between the 1st and the 7th century, allowing us to understand the evolution of the city in the blocks located to the south of it. Therefore, it is a large area, which was left out of the medieval city, having certainly been transformed into agricultural spaces, which contributed to the preservation of the existing ruins. Testimonies of the various stages of occupation were found, such as pavements of opus signinum, a pavement of carved stones in the shape of diamonds and rectangles, various mosaics, and walls [16,19,20]. The interpretation of this set allowed the identification of two blocks of the Roman city, designated, respectively, by A and B arranged on each side of the south segment of the maximum kardo. Block A was occupied by a domus, from which most of the central structures were recovered. From block B, the data collected was more dispersed and complex, and did not allow for a reliable interpretation [19]. The essential part of the exhumed archaeological remains is now buried, apart from the mosaic targeted by this study, which formalized the soil of a room in a house, preserved and musealized in the basement of the MDDS service block [16,20].

The Bracara Augusta Mosaics and the Findings of the Stables Archaeological Site
Braga's mosaics were studied by Abraços [21,22], in the coordination of a large team. The first set identified was discovered in the "Campo das Carvalheiras" area, in the context of the construction of the new Orphans Seminary; however, most of the documented elements result from the work carried out since 1976, with the formation of the CAB and the UAUM. The large number of archaeological excavations carried out by the UAUM have allowed the identification, registration, safeguarding and collection of materials and a better understanding of the Bracara Augusta mosaics. The city has been known, throughout its history, by diverse periods of building construction and economic growth [14,15,23], associated with successive changes in its status, which manifested themselves in various programs for the city's beautification, which integrates the mosaic decoration of many of its public and private buildings. Few of the city's mosaics remain in situ, the majority being deposited in MDDS [22]. A set of 69 mosaic coverings, including elements from Dume and Falperra, essentially decorated with geometric motifs, and some sets decorated with aquatic motifs and vases, were documented. Mosaics from the 1st century are known, however most of these panels date from the 3rd to 4th centuries, as a result of the intense construction and urban renewal program that the city experienced by that time [22].
With specific regard to the mosaics of the CVL archaeological site, there are five records. The first identified set was in 1991, composed of two panels dating from the 1st century, which covered the floor of two rooms of a housing complex. The second, later, dating from the Suevo-Visigothic period, was identified in 1986 in the Museum gardens, and matches a pavement in opus sectile and opus tesselatum, both preserved in situ. Three fragments were also discovered in works carried out in 1988, testimonies of the opus tesselatum technique, whose dating was not possible to specify, given the size of the archaeological discovery and context [22].
The mosaic pavement, analyzed in this present study, and the associated structures, such as walls and pipes, were in a reasonable state of conservation [24]. The rarity of this type of mosaic in the region is due to its size and constitution, and the soil acidity that does not favors its conservation, besides the fact that is considered the oldest Roman housing testimonies known until now in Braga [25], dating from the August period.  [26]. Adapted from [27].
Furthermore, several types of Braga's ceramics [28], as well as lime and marble [29] were imported from other places in the Roman Empire, such as Lusitania. The importance of marble, one of the most relevant stone resources in the Roman period, from the Marble Triangle (Estremoz, Borba, Vila Viçosa-Portugal), to Asturica Augusta, was studied in [30], and referenced therein. These being the marble suppliers of Bracara Augusta. However, as documented by a study of the city's architectural elements [31], in a universe of 356 analyzed pieces, only six were in marble. In addition to the above, its absence may also be justified by the costs associated with its acquisition or, possibly, an economy of reuse, very common in the Roman times and in the medieval world, in the context of the construction of new buildings, or even its reduction to obtain lime, an element of great importance in the Middle Ages. Similar to what was documented in the excavations of the Saint Raymond Museum, in Toulouse [32], there would be kilns for reducing marble pieces, including statuary, to obtain lime. Additionally, the assumptions related to the cost of the material and its reuse are in fact admissible, however the importation of marble and lime to Braga cannot be excluded and could have been superior to the scarce archaeological evidence available.

Description and Conservation of the Mosaic's Pavement
The construction techniques and materials used by the Romans were described, in the 1st century AC, by Vitruvius, in the work entitled "De Architectura" [33], according to the following down to top layers ( Figure 3): (a) a first preparatory layer (statumen), consisting of rolled pebbles and stones, measuring about 12 cm (the size of a fist); (b) over, the rudus layer was placed, consisting of 3/4 of sand or gravel and small pebbles and ¼ of lime in a layer with ~22 cm (3/4 of a roman foot), to protect the mosaic against humidity and infiltrations; (c) the nucleus, a opus signinum layer, consisting of a thin lime mortar with fragments of tile, or ground brick, to level the floor to receive the tesserae, in a 1.5 cm layer; finally, (d) the tesserae was placed over the nucleus, in a layer of 2 to 3 mm, consisting of lime and very fine marble dust, that filled the interstices.  [26]. Adapted from [27].
Furthermore, several types of Braga's ceramics [28], as well as lime and marble [29] were imported from other places in the Roman Empire, such as Lusitania. The importance of marble, one of the most relevant stone resources in the Roman period, from the Marble Triangle (Estremoz, Borba, Vila Viçosa-Portugal), to Asturica Augusta, was studied in [30], and referenced therein. These being the marble suppliers of Bracara Augusta. However, as documented by a study of the city's architectural elements [31], in a universe of 356 analyzed pieces, only six were in marble. In addition to the above, its absence may also be justified by the costs associated with its acquisition or, possibly, an economy of reuse, very common in the Roman times and in the medieval world, in the context of the construction of new buildings, or even its reduction to obtain lime, an element of great importance in the Middle Ages. Similar to what was documented in the excavations of the Saint Raymond Museum, in Toulouse [32], there would be kilns for reducing marble pieces, including statuary, to obtain lime. Additionally, the assumptions related to the cost of the material and its reuse are in fact admissible, however the importation of marble and lime to Braga cannot be excluded and could have been superior to the scarce archaeological evidence available.

Description and Conservation of the Mosaic's Pavement
The construction techniques and materials used by the Romans were described, in the 1st century AC, by Vitruvius, in the work entitled "De Architectura" [33], according to the following down to top layers ( Figure 3): (a) a first preparatory layer (statumen), consisting of rolled pebbles and stones, measuring about 12 cm (the size of a fist); (b) over, the rudus layer was placed, consisting of 3/4 of sand or gravel and small pebbles and 1 4 of lime in a layer with~22 cm (3/4 of a roman foot), to protect the mosaic against humidity and infiltrations; (c) the nucleus, a opus signinum layer, consisting of a thin lime mortar with fragments of tile, or ground brick, to level the floor to receive the tesserae, in a 1.5 cm layer; finally, (d) the tesserae was placed over the nucleus, in a layer of 2 to 3 mm, consisting of lime and very fine marble dust, that filled the interstices. Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 14 Many of these Vitruvius model mosaics have subsisted until our days, but this structure was not always strictly respected, leading to a faster deterioration of the mosaic pavement.
In Figure 4, a section of the pavement of the mosaic of the MDDS crypt was drawn, in which can be observed the local adaptation of the different stages of the mosaic construction according to the Vitruvian model. In this case, not all the layers that support the mosaic respect the thickness of the Vitruvius' model. In Figure 5 it is shown the layers corresponding to the nucleus, rudus and statumen of the MDDS crypt mosaic floor.  The protection of the area of the mosaic structure was carried out during the building construction for prior conservation actions (Figure 6a), although it was not efficient in controlling successive floods due to the low quota, at groundwater level, in comparison Many of these Vitruvius model mosaics have subsisted until our days, but this structure was not always strictly respected, leading to a faster deterioration of the mosaic pavement.
In Figure 4, a section of the pavement of the mosaic of the MDDS crypt was drawn, in which can be observed the local adaptation of the different stages of the mosaic construction according to the Vitruvian model. In this case, not all the layers that support the mosaic respect the thickness of the Vitruvius' model. In Figure 5 it is shown the layers corresponding to the nucleus, rudus and statumen of the MDDS crypt mosaic floor. Many of these Vitruvius model mosaics have subsisted until our structure was not always strictly respected, leading to a faster deterioratio pavement.
In Figure 4, a section of the pavement of the mosaic of the MDDS cry in which can be observed the local adaptation of the different stages construction according to the Vitruvian model. In this case, not all the laye the mosaic respect the thickness of the Vitruvius' model. In Figure 5 it is sh corresponding to the nucleus, rudus and statumen of the MDDS crypt mosa  Many of these Vitruvius model mosaics have subsisted until our days, but thi structure was not always strictly respected, leading to a faster deterioration of the mosai pavement.
In Figure 4, a section of the pavement of the mosaic of the MDDS crypt was drawn in which can be observed the local adaptation of the different stages of the mosai construction according to the Vitruvian model. In this case, not all the layers that suppor the mosaic respect the thickness of the Vitruvius' model. In Figure 5 it is shown the layer corresponding to the nucleus, rudus and statumen of the MDDS crypt mosaic floor.  The protection of the area of the mosaic structure was carried out during the building construction for prior conservation actions (Figure 6a), although it was not efficient in controlling successive floods due to the low quota, at groundwater level, in comparison  The protection of the area of the mosaic structure was carried out during the building construction for prior conservation actions (Figure 6a), although it was not efficient in controlling successive floods due to the low quota, at groundwater level, in comparison to the top of the hill located to the west-Hill of Cividade. Archeologists decided, to improve the protection of the area, to bury the mosaic floors until 1997, when the construction of the space that houses the mosaics was finished. The protective materials were removed (except the thin layer of sand that covered the mosaic floors) and some uncovered structures were consolidated and restored with a mortar based on hydraulic lime and gravel and a biocide, applied in the areas with biological colonization. A year later, when the sand was removed it was observed that the tesserae as well as other structures (walls, pipes and mortar pavements) were in a reasonable state of conservation. In 2003 and 2004, in order to drain the excess of water that reached the mosaic floors, a gallery was built, although it was not efficient. The floor mosaics were submitted to several treatments to remove the biological colonization [24,33], however it is still present (Figure 6b) due to room unfavorable humidity and temperature conditions. The anomalies in the structure of the crypt's mosaics may be the result of the high accumulation of water due to: (i) capillary rise of groundwater, likely through the more permeable materials of the mosaic structure; and (ii) rising damp due to the accumulation of surface water near the structure surrounding the mosaics. The lack of drainage system around the mosaic floor and lack of ventilation had unfavorable results on the mosaics [35,36]. to the top of the hill located to the west-Hill of Cividade. Archeologists decided, to improve the protection of the area, to bury the mosaic floors until 1997, when the construction of the space that houses the mosaics was finished. The protective materials were removed (except the thin layer of sand that covered the mosaic floors) and some uncovered structures were consolidated and restored with a mortar based on hydraulic lime and gravel and a biocide, applied in the areas with biological colonization. A year later, when the sand was removed it was observed that the tesserae as well as other structures (walls, pipes and mortar pavements) were in a reasonable state of conservation.
In 2003 and 2004, in order to drain the excess of water that reached the mosaic floors, a gallery was built, although it was not efficient. The floor mosaics were submitted to several treatments to remove the biological colonization [24,33], however it is still present (Figure 6b) due to room unfavorable humidity and temperature conditions. The anomalies in the structure of the crypt's mosaics may be the result of the high accumulation of water due to: (i) capillary rise of groundwater, likely through the more permeable materials of the mosaic structure; and (ii) rising damp due to the accumulation of surface water near the structure surrounding the mosaics. The lack of drainage system around the mosaic floor and lack of ventilation had unfavorable results on the mosaics [35,36]. The mosaic 1 feature squares were made of granite tesserae alternated with squares made of white limestone tesserae (Figure 7). The tray included, to this end, a drainage opening also covered with mosaic but with small squares with about 8 to 9 cm on the side. Mosaic 2 was decorated with squares of hourglass lines, equally made with granite and limestone tesserae (Figure 7). The mosaic covered a room whose maximum dimensions are 2.23 m in width and 3.22 m in length, with squares of approximately 20 cm on each side. In both cases, the limestone tesserae is poorly preserved [22] and only the granitic features remain preserved. The characterization of the tesserae composition determined that the white ones are composed of limestone, and the dark ones by granite, possibly pink granite from the Conde area, in Braga, assuming that this mosaic was made with local raw material [24].
The durability of the floor mortars that supports those mosaic pavements is remarkable, although some conservation actions may be needed, as some cohesion loss was observed in those renders. No prior studies considering chemical characterization of those mortars in Braga were found. To achieve this, a study on those mortars was handle in this paper. The mosaic 1 feature squares were made of granite tesserae alternated with squares made of white limestone tesserae (Figure 7). The tray included, to this end, a drainage opening also covered with mosaic but with small squares with about 8 to 9 cm on the side. Mosaic 2 was decorated with squares of hourglass lines, equally made with granite and limestone tesserae (Figure 7). The mosaic covered a room whose maximum dimensions are 2.23 m in width and 3.22 m in length, with squares of approximately 20 cm on each side. In both cases, the limestone tesserae is poorly preserved [22] and only the granitic features remain preserved. The characterization of the tesserae composition determined that the white ones are composed of limestone, and the dark ones by granite, possibly pink granite from the Conde area, in Braga, assuming that this mosaic was made with local raw material [24]. MD) of the mosaic floors ( Figure 7). The samples were collected from original and wellpreserved floor mortars (although some cohesion loss was observed during sampling, mainly in nucleus layer) and were macroscopically grouped, considering the color and the removal depth, in five different layers: layer 1-five samples from nucleus upper zone (MA2, MA3, MB1, MC1, MD1), layer 2-five samples from nucleus lower zone (MB2, MC2, MC3, MC4, MD2), layer 3-one sample between nucleus and rudus (MA4), layer 4-one sample from rudus (MD3) and layer 5-one sample from statumen (MA1) ( Table 1).   The durability of the floor mortars that supports those mosaic pavements is remarkable, although some conservation actions may be needed, as some cohesion loss was observed in those renders. No prior studies considering chemical characterization of those mortars in Braga were found. To achieve this, a study on those mortars was handle in this paper.

Materials
The study was conducted in the four different areas of the mosaic floor integrated in the crypt of the MDDS building. Micro-sampling was performed and carefully removed to prevent contamination and following the protocol of the technicians of MDDS.
This investigation was based on the characterization of 13 samples collected from the different mortar layers of the mosaic floor considering four different areas (MA, MB, MC, MD) of the mosaic floors ( Figure 7). The samples were collected from original and well-preserved floor mortars (although some cohesion loss was observed during sampling, mainly in nucleus layer) and were macroscopically grouped, considering the color and the removal depth, in five different layers: layer 1-five samples from nucleus upper zone (MA2, MA3, MB1, MC1, MD1), layer 2-five samples from nucleus lower zone (MB2, MC2, MC3, MC4, MD2), layer 3-one sample between nucleus and rudus (MA4), layer 4-one sample from rudus (MD3) and layer 5-one sample from statumen (MA1) ( Table 1).
The bedding mortar, a very thin layer consisting of lime and very fine marble (Figure 2), was not visible and, if it existed, may be below the original tesselatum; it was not possible to collect samples from this layer. Samples from the nucleus layer were removed from the nucleus upper layer (layer 1, close to the outer surface that previously was in contact with bedding mortar) and from the nucleus lower layer (layer 2, close to the rudus layer) to investigate compositional differences. Additionally, it was only possible to collect one sample from the interface between nucleus and rudus (MA4), and from rudus (MD3) and from statumen (MA1) layers; these samples need to be removed from deep and only one zone at the mosaics' floor mortars was defined by the museum technicians, to avoid as much damage as possible. The thickness of floor mortar layers varies. During nucleus sampling it was difficult to identify which group the samples belonged to in some cases.                     The bedding mortar, a very thin layer consisting of lime and very fine marble ( Figure  2), was not visible and, if it existed, may be below the original tesselatum; it was not possible to collect samples from this layer. Samples from the nucleus layer were removed from the nucleus upper layer (layer 1, close to the outer surface that previously was in contact with bedding mortar) and from the nucleus lower layer (layer 2, close to the rudus layer) to investigate compositional differences. Additionally, it was only possible to collect one sample from the interface between nucleus and rudus (MA4), and from rudus (MD3) and from statumen (MA1) layers; these samples need to be removed from deep and only one zone at the mosaics' floor mortars was defined by the museum technicians, to avoid as much damage as possible. The thickness of floor mortar layers varies. During nucleus sampling it was difficult to identify which group the samples belonged to in some cases.

Methods
This present study aims to characterize the chemical composition of the mosaic floor mortars. The chemical analyses were performed on finely crushed mortar samples by Xray fluorescence spectroscopy (XRF) through a Panalytical Axios spectrometer PW4400/40 X-ray (Marvel Panalytical, Almelo, The Netherlands) equipped with Rh tube under argon/methane, at University of Aveiro (UA), using Omnian37 and Pro-Trace2021 software for major and minor elements analyses, respectively. The loss of ignition (LOI) was determined by heating the mortars samples at 1000 °C, using an electric furnace for 3 h. The major elements analyzed were Al2O3, CaO, Cl, Fe2O3, K2O, MgO, MnO, Na2O, P2O5, SiO2, SO3, and TiO2 with detection limit of 1%. The detection limits of the trace elements analyzed were: As = 4.06 mg/kg, Ba = 6.90 mg/kg, Br = 0.78 mg/kg, Co = 4.54 mg/kg, Cr = 1.96 mg/kg, Cs = 4.78 mg/kg, Cu = 2.84 mg/kg, Ga = 0.94 mg/kg, Nb = 0.84 mg/kg, Ni = 2.00 mg/kg, Rb = 0.64 mg/kg, Sn = 3.02 mg/kg, Sr = 0.72 mg/kg, Th = 2.52 mg/kg, V = 2.78 mg/kg, Zn = 1.28 mg/kg and Zr = 0.80 mg/kg. Precision and accuracy of analyses and procedures were monitored using UA internal standards, certified reference material and quality control blanks. Results were within the 95% confidence limits. The relative standard The bedding mortar, a very thin layer consisting of lime and very fine marble ( Figure  2), was not visible and, if it existed, may be below the original tesselatum; it was not possible to collect samples from this layer. Samples from the nucleus layer were removed from the nucleus upper layer (layer 1, close to the outer surface that previously was in contact with bedding mortar) and from the nucleus lower layer (layer 2, close to the rudus layer) to investigate compositional differences. Additionally, it was only possible to collect one sample from the interface between nucleus and rudus (MA4), and from rudus (MD3) and from statumen (MA1) layers; these samples need to be removed from deep and only one zone at the mosaics' floor mortars was defined by the museum technicians, to avoid as much damage as possible. The thickness of floor mortar layers varies. During nucleus sampling it was difficult to identify which group the samples belonged to in some cases.

Methods
This present study aims to characterize the chemical composition of the mosaic floor mortars. The chemical analyses were performed on finely crushed mortar samples by Xray fluorescence spectroscopy (XRF) through a Panalytical Axios spectrometer PW4400/40 X-ray (Marvel Panalytical, Almelo, The Netherlands) equipped with Rh tube under argon/methane, at University of Aveiro (UA), using Omnian37 and Pro-Trace2021 software for major and minor elements analyses, respectively. The loss of ignition (LOI) was determined by heating the mortars samples at 1000 °C, using an electric furnace for 3 h. The major elements analyzed were Al2O3, CaO, Cl, Fe2O3, K2O, MgO, MnO, Na2O, P2O5, SiO2, SO3, and TiO2 with detection limit of 1%. The detection limits of the trace elements analyzed were: As = 4.06 mg/kg, Ba = 6.90 mg/kg, Br = 0.78 mg/kg, Co = 4.54 mg/kg, Cr = 1.96 mg/kg, Cs = 4.78 mg/kg, Cu = 2.84 mg/kg, Ga = 0.94 mg/kg, Nb = 0.84 mg/kg, Ni = 2.00 mg/kg, Rb = 0.64 mg/kg, Sn = 3.02 mg/kg, Sr = 0.72 mg/kg, Th = 2.52 mg/kg, V = 2.78 mg/kg, Zn = 1.28 mg/kg and Zr = 0.80 mg/kg. Precision and accuracy of analyses and procedures were monitored using UA internal standards, certified reference material and quality control blanks. Results were within the 95% confidence limits. The relative standard The bedding mortar, a very thin layer consisting of lime and very fine marble ( Figure  2), was not visible and, if it existed, may be below the original tesselatum; it was not possible to collect samples from this layer. Samples from the nucleus layer were removed from the nucleus upper layer (layer 1, close to the outer surface that previously was in contact with bedding mortar) and from the nucleus lower layer (layer 2, close to the rudus layer) to investigate compositional differences. Additionally, it was only possible to collect one sample from the interface between nucleus and rudus (MA4), and from rudus (MD3) and from statumen (MA1) layers; these samples need to be removed from deep and only one zone at the mosaics' floor mortars was defined by the museum technicians, to avoid as much damage as possible. The thickness of floor mortar layers varies. During nucleus sampling it was difficult to identify which group the samples belonged to in some cases.

Methods
This present study aims to characterize the chemical composition of the mosaic floor mortars. The chemical analyses were performed on finely crushed mortar samples by Xray fluorescence spectroscopy (XRF) through a Panalytical Axios spectrometer PW4400/40 X-ray (Marvel Panalytical, Almelo, The Netherlands) equipped with Rh tube under argon/methane, at University of Aveiro (UA), using Omnian37 and Pro-Trace2021 software for major and minor elements analyses, respectively. The loss of ignition (LOI) was determined by heating the mortars samples at 1000 °C, using an electric furnace for 3 h. The major elements analyzed were Al2O3, CaO, Cl, Fe2O3, K2O, MgO, MnO, Na2O, P2O5, SiO2, SO3, and TiO2 with detection limit of 1%. The detection limits of the trace elements analyzed were: As = 4.06 mg/kg, Ba = 6.90 mg/kg, Br = 0.78 mg/kg, Co = 4.54 mg/kg, Cr = 1.96 mg/kg, Cs = 4.78 mg/kg, Cu = 2.84 mg/kg, Ga = 0.94 mg/kg, Nb = 0.84 mg/kg, Ni = 2.00 mg/kg, Rb = 0.64 mg/kg, Sn = 3.02 mg/kg, Sr = 0.72 mg/kg, Th = 2.52 mg/kg, V = 2.78 mg/kg, Zn = 1.28 mg/kg and Zr = 0.80 mg/kg. Precision and accuracy of analyses and procedures were monitored using UA internal standards, certified reference material and quality control blanks. Results were within the 95% confidence limits. The relative standard

Methods
This present study aims to characterize the chemical composition of the mosaic floor mortars. The chemical analyses were performed on finely crushed mortar samples by X-ray fluorescence spectroscopy (XRF) through a Panalytical Axios spectrometer PW4400/40 X-ray (Marvel Panalytical, Almelo, The Netherlands) equipped with Rh tube under argon/methane, at University of Aveiro (UA), using Omnian37 and Pro-Trace2021 software for major and minor elements analyses, respectively. The loss of ignition (LOI) was determined by heating the mortars samples at 1000 • C, using an electric furnace for 3 h. The major elements analyzed were Al 2 O 3 , CaO, Cl, Fe 2 O 3 , K 2 O, MgO, MnO, Na 2 O, P 2 O 5 , SiO 2 , SO 3 , and TiO 2 with detection limit of 1%. The detection limits of the trace elements analyzed were: As = 4.06 mg/kg, Ba = 6.90 mg/kg, Br = 0.78 mg/kg, Co = 4.54 mg/kg, Cr = 1.96 mg/kg, Cs = 4.78 mg/kg, Cu = 2.84 mg/kg, Ga = 0.94 mg/kg, Nb = 0.84 mg/kg, Ni = 2.00 mg/kg, Rb = 0.64 mg/kg, Sn = 3.02 mg/kg, Sr = 0.72 mg/kg, Th = 2.52 mg/kg, V = 2.78 mg/kg, Zn = 1.28 mg/kg and Zr = 0.80 mg/kg. Precision and accuracy of analyses and procedures were monitored using UA internal standards, certified reference material and quality control blanks. Results were within the 95% confidence limits. The relative standard deviation was between 5% and 10%.
All statistical analyses were performed using IBM SPSS ® statistics v25. The normality of the data was verified (Shapiro-Wilk's test, p > 0.05). ANOVA, Tukey's test, t-student, K-means, cluster and discriminant analysis were used to determine groups and statistically significant differences (p < 0.05).

Results and Discussion
The chemical compositions, related to the binder and aggregates of the mosaic floor mortars samples, are detailed in Tables 2 and 3. The K-Means Cluster analysis of all samples identified two sets of samples in both major and trace elements, with MA1, MA4 and MD3 forming a cluster clearly separated from the remaining samples. The ANOVA 1-way analysis showed that the variables with significant statistical influence to define these two groups were Al 2 O 3 , K 2 O, SiO 2 , Ba, Cs, Ga, Nb, Rb, Sn (p = 0.00), MgO, Zr (p = 0.001), P 2 O 5 (p = 0.002), Ni (p = 0.003), Br, Sr (p = 0.006), Na 2 O (p = 0.007), and Fe 2 O 3 (p = 0.015). All samples were characterized by low CaO content (1.1-5.5%) and high concentration in F 2 O 3 (6.2-11.1%) and Al 2 O 3 (20.7-35.9%), when compared with the same elements from other studies on roman mortars [8,10,37]. A study on Pompeii mortars [38] found low percentages of CaO (3.81%) and attributed this low content to the high content of volcanic rock fragments as well as to the to the chemical composition of the binder composed by a mixture of lime and clay. Another possible explanation can be related to the degradation of those present study mortars (some cohesion loss was observed) as a result of the unfavorable humidity conditions to which the mosaic floor has been submitted that may lead to the leaching of the lime, as cohesion loss was observed during sampling [39]. The SiO 2 content was higher than Al 2 O 3 in all samples, although in samples MA4 (between nucleus and rudus), MD3 (rudus) and MA1 (statumen), this difference was much more evident, with over 2X's. The highest SiO 2 content was found in samples MA4, MD3 and MA1, with 47.3%, 54.1% and 47.6%, respectively. These samples showed lower Al 2 O 3 and higher K 2 O content than nucleus ones (layers 1 and 2). MD3 showed the lowest Al 2 O 3 (20.7%) and LOI (2.7%) concentrations, among all samples. The low LOI (2.7-11.2%) in all samples, when compared to other studies, e.g., [8,40], can be related to the high percentage of clay minerals and its release of OHand calcite release of CO 2 [41].
Nucleus renders (layer 1-nucleus upper zone, and layer 2-nucleus lower zone) also displayed higher content of Fe 2 O 3 (8.1-11.2%, except in sample MC1 with 7.6%), Al 2 O 3 (29.3-36.0%) and lower SiO 2 content (31.8-38.5%), when compared to rudus (MD3) and statumen (MA1) render samples, which may be the result of ceramic powder presence in nucleus render composition. The higher content of Fe 2 O 3 is responsible for the rose ochre color, that can be associated with the brick fragments or powder mixed in the mortar [38], and may indicate that iron is in its oxide form. Additionally, layer 1 showed higher CaO content (3.6-5.5%, except in sample MB1 with 2.1%) than layer 2 (1.5-2.4%). The higher CaO content in layer 1 may result from the influence of the composition of the bedding mortar that previously existed above the nucleus layer (although was not visible during sampling) and could be mixed in the nucleus upper zone. According to Vitruvius [33], it was composed by lime and very fine marble dust. In layer 1, MA2 and MC1 samples showed the highest CaO content, with 5.1% and 5.5%, respectively. The highest content in Al 2 O 3 was observed in layer 2 (34.1-36.0%). The higher Cl content was found on the nucleus upper layer (layer 1), which may be the result of capillary rise with chloride soluble salts through the mosaic floor and evaporation occurring at the outer surface of the outer layer, and as a result higher Cl concentration was observed in this layer [42].
Previous studies about the chemical characterization of archaeological and ceramic materials used the hierarchical cluster analysis on chemical data to obtain a more robust interpretation [43,44]. The cluster analyses (Ward linkage) of layers 1 and 2, separated sample MC1 from the others, and defining two other subgroups, one with MA3, MB1, and MA2 samples, and a second one with MB2, MC4, MC3, MC2, MD2, and MD1 samples (Figure 8a). The variables cluster analysis also revealed that Al 2 O 3 and SiO 2 present a distinct concentration pattern from the other major elements, in agreement with previous analysis (Figure 8b). An ANOVA 1-way analysis of these two groups revealed significant statistical differences in Al 2 O 3 and CaO (p < 0.05). The groups descriptive statistics showed that Al 2 O 3 concentration mean for layer 2 (36.62%) was higher than for layer 1 (31.68%), and that CaO mean for layer 1 was higher than for layer 2, with 4.04% and 1.90%, respectively. Despite differences identified, all the other variables did not present significant differences. Additionally, discriminant analysis based on the chemical composition of each nucleus layer (1 and 2) revealed that samples MC2 and MC4 from layer 2 showed more affinity with layer 1, and samples MA2 and MD1 from layer 1 revealed more affinity with layer 2. This may result from sampling contamination due to the small thickness of this layer. As this roman pavement is an archaeological heritage integrated in a museum, it is not possible to collect more samples.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 10 of 14 nucleus render composition. The higher content of Fe2O3 is responsible for the rose ochre color, that can be associated with the brick fragments or powder mixed in the mortar [38], and may indicate that iron is in its oxide form. Additionally, layer 1 showed higher CaO content (3.6-5.5%, except in sample MB1 with 2.1%) than layer 2 (1.5-2.4%). The higher CaO content in layer 1 may result from the influence of the composition of the bedding mortar that previously existed above the nucleus layer (although was not visible during sampling) and could be mixed in the nucleus upper zone. According to Vitruvius [33], it was composed by lime and very fine marble dust. In layer 1, MA2 and MC1 samples showed the highest CaO content, with 5.1% and 5.5%, respectively. The highest content in Al2O3 was observed in layer 2 (34.1-36.0%). The higher Cl content was found on the nucleus upper layer (layer 1), which may be the result of capillary rise with chloride soluble salts through the mosaic floor and evaporation occurring at the outer surface of the outer layer, and as a result higher Cl concentration was observed in this layer [42]. Previous studies about the chemical characterization of archaeological and ceramic materials used the hierarchical cluster analysis on chemical data to obtain a more robust interpretation [43,44]. The cluster analyses (Ward linkage) of layers 1 and 2, separated sample MC1 from the others, and defining two other subgroups, one with MA3, MB1, and MA2 samples, and a second one with MB2, MC4, MC3, MC2, MD2, and MD1 samples (Figure 8a). The variables cluster analysis also revealed that Al2O3 and SiO2 present a distinct concentration pattern from the other major elements, in agreement with previous analysis (Figure 8b). An ANOVA 1-way analysis of these two groups revealed significant statistical differences in Al2O3 and CaO (p < 0.05). The groups descriptive statistics showed that Al2O3 concentration mean for layer 2 (36.62%) was higher than for layer 1 (31.68%), and that CaO mean for layer 1 was higher than for layer 2, with 4.04% and 1.90%, respectively. Despite differences identified, all the other variables did not present significant differences. Additionally, discriminant analysis based on the chemical composition of each nucleus layer (1 and 2) revealed that samples MC2 and MC4 from layer 2 showed more affinity with layer 1, and samples MA2 and MD1 from layer 1 revealed more affinity with layer 2. This may result from sampling contamination due to the small thickness of this layer. As this roman pavement is an archaeological heritage integrated in a museum, it is not possible to collect more samples. Relatively high trace elements As, Pb and Cu contents (Table 3), when compared to other studies [38,40], were found. In most of the nucleus samples, the considerably higher As concentration might be attributed to the application of biocides to remove the biological colonization of the mosaic floor (e.g., [40]). The relatively high content in Cl, Pb Relatively high trace elements As, Pb and Cu contents (Table 3), when compared to other studies [38,40], were found. In most of the nucleus samples, the considerably higher As concentration might be attributed to the application of biocides to remove the biological colonization of the mosaic floor (e.g., [40]). The relatively high content in Cl, Pb and Cu (Table 2) can be due to the considerable degree of exposure to modern construction (concrete and Portland cement), considering that the mosaic floor is integrated in the new cement and concrete building of MDDS (e.g., [40]).
The dendrogram obtained by cluster analysis on the trace elements chemical data for all samples is presented in Figure 9. T-student test revealed significant differences between all variables in layers 1 and 2. An ANOVA 1-way analyses showed significant statistical differences between the 2 layers in V (p < 0.05) content. The descriptive statistics showed that the V mean in layer 2 (70.88 mg/kg) was higher than in group 1 (58.46 mg/kg). All the other variables did not present significant differences between groups. Moreover, discriminant analysis of layers 1 and 2 trace elements content suggested that samples MC2, MC3 and MC4 from layer 2 revealed more affinity with layer 1, and samples MC1 and MD1 from layer 1 showed more affinity with layer 2.
showed the highest CaO content, with 5.1% and 5.5%, respectively. The highest content in Al2O3 was observed in layer 2 (34.1-36.0%). The higher Cl content was found on the nucleus upper layer (layer 1), which may be the result of capillary rise with chloride soluble salts through the mosaic floor and evaporation occurring at the outer surface of the outer layer, and as a result higher Cl concentration was observed in this layer [42].
Previous studies about the chemical characterization of archaeological and ceramic materials used the hierarchical cluster analysis on chemical data to obtain a more robust interpretation [43,44]. The cluster analyses (Ward linkage) of layers 1 and 2, separated sample MC1 from the others, and defining two other subgroups, one with MA3, MB1, and MA2 samples, and a second one with MB2, MC4, MC3, MC2, MD2, and MD1 samples (Figure 8a). The variables cluster analysis also revealed that Al2O3 and SiO2 present a distinct concentration pattern from the other major elements, in agreement with previous analysis (Figure 8b). An ANOVA 1-way analysis of these two groups revealed significant statistical differences in Al2O3 and CaO (p < 0.05). The groups descriptive statistics showed that Al2O3 concentration mean for layer 2 (36.62%) was higher than for layer 1 (31.68%), and that CaO mean for layer 1 was higher than for layer 2, with 4.04% and 1.90%, respectively. Despite differences identified, all the other variables did not present significant differences. Additionally, discriminant analysis based on the chemical composition of each nucleus layer (1 and 2) revealed that samples MC2 and MC4 from layer 2 showed more affinity with layer 1, and samples MA2 and MD1 from layer 1 revealed more affinity with layer 2. This may result from sampling contamination due to the small thickness of this layer. As this roman pavement is an archaeological heritage integrated in a museum, it is not possible to collect more samples. Relatively high trace elements As, Pb and Cu contents (Table 3), when compared to other studies [38,40], were found. In most of the nucleus samples, the considerably higher As concentration might be attributed to the application of biocides to remove the biological colonization of the mosaic floor (e.g., [40]). The relatively high content in Cl, Pb and Cu (Table 2) can be due to the considerable degree of exposure to modern These mortars were characterized by low content in CaO and high content of Fe 2 O 3 and Al 2 O 3 . The nucleus (layers 1 and 2) was characterized by higher Fe 2 O 3 and Al 2 O 3 than rudus and statumen. Rudus and statumen showed higher SiO 2 content than nucleus. Comparing layers 1 with 2: layer 1 showed higher CaO content and layer 2 higher Al 2 O 3 content. These layers showed relatively high As, Pb and Cu contents.
The statistical analyses enabled the differentiation of the two main nucleus layers (layers 1 and 2) from the three outlier samples: between nucleus and rudus, rudus and statumen. The descriptive statistics revealed differences among layers 1 and 2 concerning mainly Al 2 O 3 , CaO and V contents. Considering layers 1 and 2 major elements statistical analyses, MC2 and MC4 from layer 2 shows more affinity with layer 1 and MA2 and MD1 from group 1 shows more affinity with layer 2; analyses on trace elements revealed that samples MC1 and MD1 from layer 1 presented more affinity with layer 2 and MC2, MC3 and MC4 from layer 2 more affinity with layer 1.

Conclusions
The results obtained on the 13 mortars samples from different floor layers of the high imperial mosaics of the domus integrated in the crypt space of Museum of Archaeology D. Diogo de Sousa showed that there are differences on the chemical elements analyzed from the different layers: nucleus upper layer (layer 1), nucleus lower layer (layer 2), between nucleus (layer 3) and rudus (layer 4), rudus and statumen (layer 5) layers.
The chemical characterization of the studied floor mortar layers indicated that their composition was distinctly related to their stratigraphic position in the substrate, according to Vitruvius' model, and to the external conditions and treatments to which each layer was submitted (e.g., capillary rise with soluble salts and application of chemical treatments). Mosaic floor mortars showed low content of CaO and high F 2 O 3 , Al 2 O 3 , Cl, As, Pb and Cu contents. The statistical analyses using the chemical data on major and trace elements was robust and confirmed the clear separation between the nucleus layer (layers 1 and 2) and the other layers.
This study offers an important archaeological and historical contextualization and a first chemical characterization on the roman mortars of Bracara Augusta. Additionally, it is focused on mortar investigation of the oldest Roman housing testimonies of Bracara Augusta musealized in Museum of Archaeology D. Diogo de Sousa, Portugal.
Future studies will focus on the mineralogical analyses of the mosaic floor layers, and correlate them with the results of the chemical analyses discussed in this present study.