Ancient Roman Mortars from Villa del Capo di Sorrento: A Multi-Analytical Approach to Define Microstructural and Compositional Features

This research provides a characterization of ancient Roman mortars from “Villa del Capo di Sorrento” (commonly known as “Villa di Pollio Felice” or “Bagni della Regina Giovanna”). A deepened analysis of cementitious binding matrix and aggregates was conducted with the aims of determining possible sources of raw materials and the mix recipe, and to evaluate the minerogenetic secondary processes. Twenty samples taken from the Villa were investigated by means of a multi-analytical approach, including polarized optical microscopy on thin sections, X-ray powder diffraction, scanning electron microscopy analysis, energy dispersed spectrometry, simultaneous thermal analyses, and mercury intrusion porosimetry. Bedding mortars were made with slaked lime mixed with volcanic materials, whereas coating mortars were made adding to previous recipe as ceramic fragments. All samples were classified as hydraulic mortars. Cementitious binding matrix was characterized by gel-like C-A-S-H, calcite, hydrocalumite, and gypsum, deriving from lime/pozzolanic material. Geomaterials used for mortar production had a local origin. Pozzolanic materials, such as volcanic fragments, scoriae, pumice, and crystal fragments derived from both pyroclastic rocks of the Campi Flegrei district and from rocks of the Somma-Vesuvio complex; porosity test suggest that the products related to minerogenetic secondary processes, make mortars more resistant.


Introduction
The Campania region was renowned during ancient times as Campania felix, thanks to its climate, beautiful landscapes, and fertile land. For these reasons, the region was among the favorite places to live in Roman times, with highly populated cities and otium villae on the coast [1].
Nowadays, the region has plenty of major archeological sites, such as the eternal Vesuvian cities, Pompeii, and Herculaneum, the Campi Flegrei, which were often visited by rich Roman senators because of the natural baths and the stunning villae maritimae, found from Posillipo to Punta Campanella [2].
These sites, like other remarkable examples in the Campania region, are still preserved despite their location in an aggressive environment, such as the seaside, and the impact of waves and weathering, which is why they have been the object of many scientific studies.
The Department of Earth Sciences, Environment, and Resources (DiSTAR), Federico II University of Naples, for more than twenty years has been engaged in the application of mineralogical and petrographic studies of several ancient finds and monuments, especially of Campania region, such as Roman concrete, mortars, and ceramics, e.g., [3][4][5][6][7].
In this study the area of interest is the Sorrento Peninsula, between the village of Aequa (near Vico Equense) and the far side of the Sorrentine peninsula with its adjoining islets (Figure 1a). A total of 24 ruins have been identified as structures related to villae maritimae, commonly dated, based on their building techniques, between Late Republican period (133-21 B.C.) and the start of the 2nd c. A.D. [2]. In particular, we have focused the study on ancient roman mortars of one of the most remarkable villa maritima of the Sorrento Peninsula: Villa del Capo (commonly known as Villa di Pollio Felice or Bagni della Regina Giovanna; Figure 1).
Twenty lime-based mortars were carefully selected from different structures of Villa del Capo (Figure 1c) with the aim of characterizing them, especially regarding the cementitious binding matrix and aggregates, to determine the provenance of raw materials and mix recipe, improve the knowledge of Roman construction techniques, and study the secondary minerogenetic processes affecting the investigated mortars.
The Department of Earth Sciences, Environment, and Resources (DiSTAR), Federico II University of Naples, for more than twenty years has been engaged in the application of mineralogical and petrographic studies of several ancient finds and monuments, especially of Campania region, such as Roman concrete, mortars, and ceramics, e.g., [3][4][5][6][7].
In this study the area of interest is the Sorrento Peninsula, between the village of Aequa (near Vico Equense) and the far side of the Sorrentine peninsula with its adjoining islets (Figure 1a). A total of 24 ruins have been identified as structures related to villae maritimae, commonly dated, based on their building techniques, between Late Republican period (133-21 B.C.) and the start of the 2nd c. A.D. [2]. In particular, we have focused the study on ancient roman mortars of one of the most remarkable villa maritima of the Sorrento Peninsula: Villa del Capo (commonly known as Villa di Pollio Felice or Bagni della Regina Giovanna; Figure 1).
Twenty lime-based mortars were carefully selected from different structures of Villa del Capo (Figure 1c) with the aim of characterizing them, especially regarding the cementitious binding matrix and aggregates, to determine the provenance of raw materials and mix recipe, improve the knowledge of Roman construction techniques, and study the secondary minerogenetic processes affecting the investigated mortars.

Geological Context
Villa del Capo is located along the northwestern sector of the Sorrento Peninsula at the base of the north facing slope of the Corbo Mt. The ancient building is placed on the distal part of a small promontory that developing in NNW-SSE direction forms the so-called Punta del Capo. Being part of the Sorrento Peninsula horst, the Corbo Mt. is formed of Cretaceous limestones, that widely crop out in the Sorrento Peninsula promontory, overlined by a transgressive Miocene succession, locally covered by quaternary rocks [9] (Figure 2). Forming a morphologic boundary between two semi-grabens located to the north and to the south (i.e., Gulf of Naples and Gulf of Salerno, respectively [9,10] (Figure 2), the Sorrento Peninsula is characterized by a structural setting dominated by NNW-SSE normal faults, resulting from the Plio-Quaternary transcurrent and extensional tectonics [10][11][12]. Less significant is the structural expression of the Late Miocene-Pliocene compressive tectonics. Similar to the Corbo Mt., the Lattari Mts. Range, forming the Sorrento Peninsula, is consistently mantled by pyroclastic soils representing the product of the activity of the Somma-Vesuvio and the Campi Flegrei. Such deposits have a thickness ranging from few centimeters to ten meters [12] and were mostly produced by the Holocenic Vesuvius eruption of 79 A.D. Between 18 ka ago and 79 A.D., the Lattari Mts. did not receive any significant fallout deposits, because during this time span, the Vesuvio Plinian eruptions were dispersed in other directions (NE to E; [13]. The pyroclastic cover has been subject to both mass wasting and fluvial denudation processes that completely removed it from steeper slopes.

Geological Context
Villa del Capo is located along the northwestern sector of the Sorrento Peninsula at the base of the north facing slope of the Corbo Mt. The ancient building is placed on the distal part of a small promontory that developing in NNW-SSE direction forms the so-called Punta del Capo. Being part of the Sorrento Peninsula horst, the Corbo Mt. is formed of Cretaceous limestones, that widely crop out in the Sorrento Peninsula promontory, overlined by a transgressive Miocene succession, locally covered by quaternary rocks [9] (Figure 2). Forming a morphologic boundary between two semi-grabens located to the north and to the south (i.e., Gulf of Naples and Gulf of Salerno, respectively [9,10] (Figure 2), the Sorrento Peninsula is characterized by a structural setting dominated by NNW-SSE normal faults, resulting from the Plio-Quaternary transcurrent and extensional tectonics [10][11][12]. Less significant is the structural expression of the Late Miocene-Pliocene compressive tectonics. Similar to the Corbo Mt., the Lattari Mts. Range, forming the Sorrento Peninsula, is consistently mantled by pyroclastic soils representing the product of the activity of the Somma-Vesuvio and the Campi Flegrei. Such deposits have a thickness ranging from few centimeters to ten meters [12] and were mostly produced by the Holocenic Vesuvius eruption of 79 A.D. Between 18 ka ago and 79 A.D., the Lattari Mts. did not receive any significant fallout deposits, because during this time span, the Vesuvio Plinian eruptions were dispersed in other directions (NE to E; [13]. The pyroclastic cover has been subject to both mass wasting and fluvial denudation processes that completely removed it from steeper slopes.

The Roman Villa del Capo
Roman Villa del Capo, which the popular tradition knew as I bagni della regina Giovanna (the baths of Queen Giovanna) or Villa di Pollio Felice, was identified by antiquarian tradition, since the sixteenth century, as well as by part of archaeological studies [15]. Located in a unique position, due to the presence of a circular basin, the villa is part of a dense

The Roman Villa del Capo
Roman Villa del Capo, which the popular tradition knew as I bagni della regina Giovanna (the baths of Queen Giovanna) or Villa di Pollio Felice, was identified by antiquarian tradition, since the sixteenth century, as well as by part of archaeological studies [15]. Located in a unique position, due to the presence of a circular basin, the villa is part of a dense network of maritime residences that studded the entire Gulf of Naples between the first century B.C. and the first century A.D. [16].
The Roman residence, organized with different living areas and services, was located on the top of the promontory, and characterized by several continuous terraces along the slope, about 200 m long, down to the sea where there were dockings [15]. The central  area was characterized by four levels with a compact nucleus of buildings; the presence  of a peristyle with backyard and central basin on the upper terrace has been documented, where open rooms paved in opus sectile and a system of ramps and stairs linked this place with others. The structures of the Villa, facing the Sorrento side, due to the irregularly ascending shape of the area was leveled with the basis of the Villa made by opus caementicium foundations, exhibiting arches on the first floor and a covered porch with residential rooms at the second one [15]. The thermal area is identified by several rooms near the sea, where traces of their decoration survive, although partially destroyed by sea storms [2]. Finally, Villa del Capo continued to be inhabited for several centuries despite the damages caused by Somma-Vesuvio eruption of 79 A.D. [16].

Materials and Methods
In the context of the ongoing collaboration and related authorization of the former Soprintendenza archeologia della Campania and to the assistance of archaeologists, who have carried out several studies on the investigated area, mortar samples were carefully collected in the most representative architectural structures of the archaeological site of Villa del Capo ( Figure 1c, Table 1). The sampling of mortars was preceded by an accurate on-site survey to define the most suitable samples to collect in terms of archaeological and architectural significance. A total of 20 samples was collected comprising ten bedding mortars (Group A), eight coating mortars (Group B) and two floor mortars (Group C).
Samples were removed mechanically with hammers and chisels, ensuring to avoid external and clearly altered portions to study materials that are as close as possible to the original conservation state.
In order to achieve this study's goals, various analytical techniques were used to obtain a mortar's complete petrographic, mineralogical, and chemical characterization and degree of hydraulicity.
The first approach to the study of mortar samples was macroscopic observation, for: (1) identification of the materials and (2) planning of analytical procedures.
Polarized optical microscopy (POM) on polished thin section was performed to observe the textural features and the petrographic composition of the samples with a Leica DFC280 microscope (Leica Camera, Wetzlar, Germany). Percentage of binder and aggregate was measured via modal analysis on four representative thin sections from each different sector of the investigated area, selected on the base of their macro-and microscopical features. 1500 points for each section were counted using a Leica Q Win image analysis software. Maximum uncertainty of percentage for a total amount of 1500 points is about 2.8% [17].
Qualitative mineralogical analysis was performed by X-Ray Powder Diffraction (XRPD) using a Panalytical X'Pert Pro diffractometer (Malvern PANalytical, Almelo, The Netherlands) equipped with a RTMS X'Celerator detector with Cu-Kα radiation, operating at 40 kV and 40 mA. Scans were collected in the range 4-70 • 2θ using a step interval of 0.017 • 2θ, with a step counting time of 120 s. Panalytical Highscore Plus 3.0c software (Malvern PANalytical, Almelo, The Netherlands) and PDF-2/ICSD databases were used for identification.
Micro-textural observations and quantitative micro-chemical analyses were carried out by Scanning Electron Microscopy coupled and Energy Dispersive Spectroscopy (SEM/EDS; JEOL JSM-5310 (Jeol Ltd., Tokyo, Japan) coupled with Oxford Instruments Microanalysis Unit, INCA X-act detector (Oxford Instruments plc, Abingdon, Oxfordshire, UK). Measurements were performed with an INCA X-stream pulse processor (ETAS group, Stuttgart, Germany) using a 15-kV primary beam voltage, 50-100 µA filament current, variable spot size, from 30,000 to 200,000× magnification, 20 mm WD, and 50 s net acquisition real time). The INCA Energy software was employed, using the XPP matrix correction scheme, developed by Pouchou and Pichoir [18], and the Pulse Pile up correction. The quant optimization was carried out using cobalt (FWHM-full width at half maximum peak height-of the strobed zero = 60-65 eV).
Details of the utilized standard and precision and accuracy are provided in [19]. Micro-chemical analyses were performed to determine major chemical composition of binder, lime lumps and aggregates. Hydraulicity index (HI) of binder was calculated according to Boynton using the (SiO 2 + Al 2 O 3 + Fe 2 O 3 )/(CaO + MgO) ratio.
Thermal Analyses (TGA/SDTA) were carried out with a Mettler Toledo TGA/SDTA 851e instrument (Mettler Toledo, Columbus, OH, USA) and Mettler Toledo STARe SW 7.01 software, with the main goal of determining total (binder plus aggregates) hydraulic features of these materials. Samples were previously dried at 40 • C in a drying oven for 48 h. Thermal analysis was performed within the temperature range of 25-1000 • C, heating rate of 10 • C/min in nitrogen atmosphere (flow 60 mL/min).
Porosity was determined on fragments of mortar (binder and aggregate were not separated) and was evaluated using mercury intrusion porosimetry (MIP), according to ASTM D4404-18 [20]. Due to the scarce amount and small dimensions of samples, on the bases of macroscopic and microscopic features, three fragments were selected, approximately 1 cm3 in size. Selected fragments were dried in an oven for 24 h at 105 • C, and then analysed on Thermo Scientific equipment PASCAL 140 (ThermoFischer, Waltham, MA, United States) with a maximum injection pressure of 0.4 MPa and PASCAL 240 with a maximum injection pressure of 200 MPa. Total volume of pores of radius between 3.75 nm and 800 µm (expressed in mm 3 /g) was determined; open porosity (expressed in vol %); bulk density (g/cm 3 ); apparent density (g/cm 3 ), and specific surface (m 2 /g); graphical and numerical representation of the distribution of pore sizes were also determined.

Texture and Optical Microscopy
From a macroscopic point of view, some samples appear to be intact with respect to others that seem to be quite dusty and friable. In addition, they globally show grain size from fine to coarse sand [21]. A brief macroscopic description of mortar Groups (A, B, and C) along with representative sample images is summarized in Table 2. Bedding mortars light grey color Up to 10 mm ++

Texture and Optical Microscopy
From a macroscopic point of view, some samples appear to be intact with respect to others that seem to be quite dusty and friable. In addition, they globally show grain size from fine to coarse sand [21]. A brief macroscopic description of mortar Groups (A, B, and C) along with representative sample images is summarized in Table 2. Optical microscopy shows that all mortars are characterized by presence of lime lumps (mm to cm). Their formation generally is related to lime binder not well mixed in mortars but sometimes they could be also formed due to not adequate slaking processes of lime or to non-homogeneous temperature in the kiln (under/over-burnt fragments of limestone) [22][23][24]. Secondary calcite on pore rims and pumice vesicles also occurs.
Group B (coating mortars) is characterized by a binder that ranges from greyish to brownish color and shows a cryptocrystalline (31.4 Vol.%) and micritic texture (18.9 Vol.%; Figure

Texture and Optical Microscopy
From a macroscopic point of view, some samples appear to be intact with respect to others that seem to be quite dusty and friable. In addition, they globally show grain size from fine to coarse sand [21]. A brief macroscopic description of mortar Groups (A, B, and C) along with representative sample images is summarized in Table 2. Optical microscopy shows that all mortars are characterized by presence of lime lumps (mm to cm). Their formation generally is related to lime binder not well mixed in mortars but sometimes they could be also formed due to not adequate slaking processes of lime or to non-homogeneous temperature in the kiln (under/over-burnt fragments of limestone) [22][23][24]. Secondary calcite on pore rims and pumice vesicles also occurs.
Binder phase in group A (bedding mortars) is characterized by pale brown/beige to grey color and shows various grades of crystallinity from cryptocrystalline (36.8 Vol.%; Figure  Vol.%) formed by clinopyroxene, sanidine, plagioclase, biotite, and, sometimes, calcite, and crystal fragments of sanidine, plagioclase, and clinopyroxene (5.2 Vol.%). Volcanic fragments can be identified as volcanic tuff, characterized by pumice and obsidian fragments, lithics and crystal fragments of sanidine, clinopyroxene, biotite, and plagioclase set in an ashy matrix mainly constituted by volcanic glass shards, affected by secondary mineralization processes [25]. In sample BG12 presence of acicular crystals was also observed, apparently calcite, filling the vugs. (Figure 3c). The binder/aggregate ratio was about 2.7 (Table 3).
Group B (coating mortars) is characterized by a binder that ranges from greyish to brownish color and shows a cryptocrystalline (31.4 Vol.%) and micritic texture (18.9 Vol.%; Figure

Texture and Optical Microscopy
From a macroscopic point of view, some samples appear to be intact with respect to others that seem to be quite dusty and friable. In addition, they globally show grain size from fine to coarse sand [21]. A brief macroscopic description of mortar Groups (A, B, and C) along with representative sample images is summarized in Table 2. Optical microscopy shows that all mortars are characterized by presence of lime lumps (mm to cm). Their formation generally is related to lime binder not well mixed in mortars but sometimes they could be also formed due to not adequate slaking processes of lime or to non-homogeneous temperature in the kiln (under/over-burnt fragments of limestone) [22][23][24]. Secondary calcite on pore rims and pumice vesicles also occurs.
Binder phase in group A (bedding mortars) is characterized by pale brown/beige to grey color and shows various grades of crystallinity from cryptocrystalline (36.8 Vol.%; Figure  Vol.%) formed by clinopyroxene, sanidine, plagioclase, biotite, and, sometimes, calcite, and crystal fragments of sanidine, plagioclase, and clinopyroxene (5.2 Vol.%). Volcanic fragments can be identified as volcanic tuff, characterized by pumice and obsidian fragments, lithics and crystal fragments of sanidine, clinopyroxene, biotite, and plagioclase set in an ashy matrix mainly constituted by volcanic glass shards, affected by secondary mineralization processes [25]. In sample BG12 presence of acicular crystals was also observed, apparently calcite, filling the vugs. (Figure 3c). The binder/aggregate ratio was about 2.7 (Table 3).
Group B (coating mortars) is characterized by a binder that ranges from greyish to brownish color and shows a cryptocrystalline (31.4 Vol.%) and micritic texture (18.9 Vol.%; Figure 3c), small and fractured lime lumps, with not well-defined edges (6. Optical microscopy shows that all mortars are characterized by presence of lime lumps (mm to cm). Their formation generally is related to lime binder not well mixed in mortars but sometimes they could be also formed due to not adequate slaking processes of lime or to non-homogeneous temperature in the kiln (under/over-burnt fragments of limestone) [22][23][24]. Secondary calcite on pore rims and pumice vesicles also occurs.
Binder phase in group A (bedding mortars) is characterized by pale brown/beige to grey color and shows various grades of crystallinity from cryptocrystalline (36. formed by clinopyroxene, sanidine, plagioclase, biotite, and, sometimes, calcite, and crystal fragments of sanidine, plagioclase, and clinopyroxene (5.2 Vol.%). Volcanic fragments can be identified as volcanic tuff, characterized by pumice and obsidian fragments, lithics and crystal fragments of sanidine, clinopyroxene, biotite, and plagioclase set in an ashy matrix mainly constituted by volcanic glass shards, affected by secondary mineralization processes [25]. In sample BG12 presence of acicular crystals was also observed, apparently calcite, filling the vugs. (Figure 3c). The binder/aggregate ratio was about 2.7 (Table 3).
Group B (coating mortars) is characterized by a binder that ranges from greyish to brownish color and shows a cryptocrystalline (31.4 Vol.%) and micritic texture (18.9 Vol.%; Mortars from group B can be identified as cocciopesto, a typical building technique used in ancient Rome for waterproof structures, such as cisterns and floors [3,26]. The binder/aggregate ratio is of about 1.5 (Table 3). optical activity. The other secondary aggregates that characterize this Group (B) are volcanic and carbonate fragments (2.8 Vol.%-0.5 Vol.%), pumice, and scoriae with reaction rim (4.8 Vol.%; 2.2 Vol.%; Figure 3g) and crystal fragments of plagioclase, sanidine, and clinopyroxene (5.8 Vol.%). Particularly relevant is the presence of leucite-bearing scoriae in BG8 sample (Figure 3h) and the presence of crystal fragments of garnet in BG7 sample (Figure 3i). Mortars from group B can be identified as cocciopesto, a typical building technique used in ancient Rome for waterproof structures, such as cisterns and floors [3,26]. The binder/aggregate ratio is of about 1.5 (Table 3).
Transition between layers 1 and 0 is well defined and evident (Figure 4g). Modal analysis, performed on layer 1, shows binder/aggregate ratio equal to 1.8.
Transition between layers 1 and 0 is well defined and evident (Figure 4g). Modal analysis, performed on layer 1, shows binder/aggregate ratio equal to 1.8.
Layer 0 (L0) is characterized by white color of the binder, with mainly micritic aspect with no aggregates (Figure 4h).
XRPD results confirmed occurrence of lime-based mortar with volcanic aggregate, as shown by semi-quantitative analyses reported in Table 4.
Regarding binder phases ( Figure 5), calcite is the most abundant phase with subordinate gypsum [CaSO 4 ·2(H 2 O)] and hydrocalumite Ca 2 Al(OH) 6 [Cl 1 -x(OH)x]·3(H 2 O). Gypsum, since it is only in a few samples, could be ascribed to sulphation processes of calcite as a consequence of the decrease in pH value, caused by dissolution of atmospheric SO 2 [29].
Hydrocalumite, also known as Friedel's salt or AFm phase in cement science, belongs to layered double hydroxides (LDHs) family [8,30,31]. LDHs are among the few oxidebased materials with permanent anion exchange capacity, developed through isomorphous substitution [30]. Hydrocalumite has not only an ordered Ca-Al distribution in the hydroxide layer, but well-ordered Cland water in the interlayer space. The interlayer order is due to coordination of the water molecules to Ca in the hydroxide layer, which results in an unusual 7-coordinate Ca environment. This phase occurs naturally and generally forms by reaction of Cl-containing de-icing salts with the calcium aluminates of Portland cement [32]. In Villa del Capo mortars hydrocalumite is ascribable to the reaction between the Ca(OH) 2 , sea-water, and hydroxyaluminate derived from pozzolanic materials [31].
As far as volcanic aggregates fraction is concerned ( Figure 5), XRPD analyses suggest that volcanic tuff fragments can be associated to Campi Flegrei ignimbrite products (Neapolitan Yellow Tuff and/or Campanian Ignimbrite), due to the presence of typical zeolitic association of this material, i.e.,: phillipsite, chabazite, and analcime [33,34]. In addition, presence of sanidine, clinopyroxene, and mica could be also associated to the Campi Flegrei ignimbrite formations.
All samples show the presence of halite, associated with marine aerosol. XRPD analysis allowed to detect the presence of amorphous fraction, recognized by the rising of the pattern background, related to volcanic glass components and C-A-S-H (calcium, aluminum, silicate, hydrate) phases. association of this material, i.e.,: phillipsite, chabazite, and analcime [33,34]. In addition, presence of sanidine, clinopyroxene, and mica could be also associated to the Campi Flegrei ignimbrite formations. All samples show the presence of halite, associated with marine aerosol. XRPD analysis allowed to detect the presence of amorphous fraction, recognized by the rising of the pattern background, related to volcanic glass components and C-A-S-H (calcium, aluminum, silicate, hydrate) phases.

Binder and Lime Lumps
SEM-EDS analysis was carried out to obtain information about the type of lime used to produce the mortars, defining the Hydraulic Index (HI) of the binder and lime lumps.
SEM-EDS analysis of binder showed presence of newly formed hydraulic phases (C-A-S-H), and confirmed, as XRPD suggested, presence of halite and hydrocalumite ( Figure 6). Gel C-A-S-H derived from reactions between lime and pozzolanic material (volcanic and ceramic materials). Pozzolanic materials consist predominantly of silica and alumina that are able to combine with slaked lime in the presence of water to produce new reaction products exhibiting a binding character, the so-called CASH phases [35]. Occurrence of hydrocalumite recognized in relict pores of Villa del Capo mortars, is associated to the migration of Clanions from the sea-water saturated in Ca(OH) 2 to aluminum-rich sites along the edges of the relict lime clasts or in the voids of mortars [36].
EDS analyses on binder were realized averaging out four measurements for each investigated mortar sample and the results were considered as representative of chemical composition. Based on the mean values of the detected major elements (

Binder and Lime Lumps
SEM-EDS analysis was carried out to obtain information about the type of lime used to produce the mortars, defining the Hydraulic Index (HI) of the binder and lime lumps.
SEM-EDS analysis of binder showed presence of newly formed hydraulic phases (C-A-S-H), and confirmed, as XRPD suggested, presence of halite and hydrocalumite ( Figure  6). Gel C-A-S-H derived from reactions between lime and pozzolanic material (volcanic and ceramic materials). Pozzolanic materials consist predominantly of silica and alumina that are able to combine with slaked lime in the presence of water to produce new reaction products exhibiting a binding character, the so-called CASH phases [35]. Occurrence of hydrocalumite recognized in relict pores of Villa del Capo mortars, is associated to the migration of Cl --anions from the sea-water saturated in Ca(OH)2 to aluminum-rich sites along the edges of the relict lime clasts or in the voids of mortars [36]. SEM-EDS analysis was carried out to obtain information about the type of lime used to produce the mortars, defining the Hydraulic Index (HI) of the binder and lime lumps.
EDS analyses on binder were realized averaging out four measurements for each investigated mortar sample and the results were considered as representative of chemical composition. Based on the mean values of the detected major elements (  The Hydraulicity Index (HI), calculated according to Boynton's formula [37], showed values ranging between 0.12 and 0.25 wt%; in particular, mortar samples of Group A and C should be considered as weakly hydraulic lime (0.10< HI > 0.15 wt%; Figure 7); samples of Group B are moderately hydraulic limes (0.16 < HI < 0.31 wt%; Figure 7). Regarding the investigation of lime lumps (Table 6), three measurements were performed for each detected lump and performed in their central portion, to reduce the level of contamination. Lime lumps were found to be composed mainly of CaO, with very high values of CaO + MgO ranging from 90.56 wt% (BG9) to a maximum of 96.56 wt% (BG6) and low values of SiO 2 + Al 2 O 3 + Fe 2 O 3 all less than 5 wt%. The Hydraulicity Index related to lime lumps shows values from 0.01 and 0.05 wt% (0.03 < HI < 0.05 wt%; Figure 7), as a result of lumps with aerial properties (H.I. < 0.10). Considering these results, it is possible to infer that mortars became hydraulic by addition of aggregates (volcanic materials and ceramic fragments) with peculiar features. These aggregates, as said previously, produced a "pozzolanic reaction", due to their silica and alumina content, that reacted with calcium hydroxide leading to the formation of calcium aluminum silicate hydrates, C-A-S-H phases [8,38,39], furtherly testified and confirmed by reaction rims around pumice, scoriae, and ceramic fragment (Figure 3).   mortars became hydraulic by addition of aggregates (volcanic materials and ceramic fragments) with peculiar features. These aggregates, as said previously, produced a "pozzolanic reaction", due to their silica and alumina content, that reacted with calcium hydroxide leading to the formation of calcium aluminum silicate hydrates, C-A-S-H phases [8,38,39], furtherly testified and confirmed by reaction rims around pumice, scoriae, and ceramic fragment (Figure 3).  Table 5. Average values of major oxides (wt%, recalculated to 100%, EDS) of lime lumps (L). SiO2 + Al2O3 + Fe2O3, CaO + MgO, HI (hydraulic index) is also shown.

Volcanic Aggregates
SEM-EDS analysis of volcanic aggregates were performed to obtain additional information about the provenance of raw materials. Results confirmed the above-mentioned hypothesis of the employment of Campi Flegrei tuff aggregates, due to the presence of phillipsite with well-defined prismatic crystal habit and pseudo cubic crystals of chabazite [33].
EDS microanalysis was carried out on pumice from mortar samples (see Supplementary Materials Table S1). Using Total Alkali versus Silica diagram (TAS) for the effusive volcanic rocks [40], it is possible to observe that pumice fragments show trachytic compositions; their classification follows the compositional trend of Campi Flegrei products (Figure 8a). Results also suggested the use of geomaterials from the Somma-Vesuvio complex, due to the presence of volcanic scoriae containing analcime (typical product of leucite alteration; Figure 8b) and garnet fragments [13] typical of these materials. In support of this hypothesis, chemical composition of analyzed garnet (calculated following [41]; (see Supplementary Materials Table S2) shows similarity with garnets of Somma-Vesuvio (andradite 46-70 mol% and grossularia 16-45 mol%; taken by [42]; unpublished data on garnets from intrusive Somma-Vesuvio rocks. L. Melluso, personal communication).

Ceramic Aggregates
Regarding ceramic fragments, SEM-EDS analysis allowed to confirm extreme differences occurring between them, even ones from the same mortar sample (Figure 9). These differences are testified by the different mineralogical composition, different texture, different porosity, and by the different chemical composition of matrix.
Chemical analysis on matrix of representative ceramic fragments (Table 7) showed different concentration in CaO ranging from 2.87 wt% for sample BG12 to a maximum of 19.08 wt% for BG4b.
The concentration of this oxide is a strong discriminant of ceramics in relation to their final use. When the percentage of this oxide is less than 6% [44] the clayey raw material used to produce ceramics is defined non calcareous, otherwise it is said calcareous (CaO > 6%). Non calcareous clays are best suited for making cookware, due to the better thermal shock resistance of the end products, whereas calcareous clays are generally employed to produce pottery used as containers [45].
Differences between ceramic fragments does not allow for defining provenance but represent a testimony of the re-use of ceramic waste materials in preparations of mortars. The relevant role played by ceramic fragments was to provide hydraulicity to the mortars, as pointed out by HI evaluation (Figure 7). Coating and floor mortars that contain both pozzolana and ceramics materials are characterized by highest HI values.

Ceramic Aggregates
Regarding ceramic fragments, SEM-EDS analysis allowed to confirm extreme differences occurring between them, even ones from the same mortar sample (Figure 9). These differences are testified by the different mineralogical composition, different texture, different porosity, and by the different chemical composition of matrix.
Chemical analysis on matrix of representative ceramic fragments (Table 7) showed different concentration in CaO ranging from 2.87 wt% for sample BG12 to a maximum of 19.08 wt% for BG4b.
The concentration of this oxide is a strong discriminant of ceramics in relation to their final use. When the percentage of this oxide is less than 6% [44] the clayey raw material used to produce ceramics is defined non calcareous, otherwise it is said calcareous (CaO > 6%). Non calcareous clays are best suited for making cookware, due to the better thermal shock resistance of the end products, whereas calcareous clays are generally employed to produce pottery used as containers [45].
Differences between ceramic fragments does not allow for defining provenance but represent a testimony of the re-use of ceramic waste materials in preparations of mortars. The relevant role played by ceramic fragments was to provide hydraulicity to the mortars, as pointed out by HI evaluation (Figure 7). Coating and floor mortars that contain both pozzolana and ceramics materials are characterized by highest HI values.

Differential Thermal and Thermogravimetric Analysis
Simultaneous thermal analyses combined with mineralogical composition of samples allowed to obtain further information about the hydraulic characteristics of m tar samples. Table 8 reports the percentage of weight loss estimated from the TG-D curves within the selected temperature ranges. In the temperature range from 25 to

Differential Thermal and Thermogravimetric Analysis
Simultaneous thermal analyses combined with mineralogical composition of the samples allowed to obtain further information about the hydraulic characteristics of mortar samples. Table 8 reports the percentage of weight loss estimated from the TG-DTG curves within the selected temperature ranges. In the temperature range from 25 to 120 • C the weight loss is due to dehydration of hygroscopic or adsorbed water (i.e., phyllosilicates), from 120 to 200 • C the weight loss of water from hydrated salts occurs, between 200 and 600 • C the weight loss is due to structurally bound water (SBW) from the hydraulic compounds and, finally, the loss of CO 2 as a consequence of the decomposition of calcium carbonate (CaCO 3 ) takes place at temperature range between 600 and 900 • C [24,46,47]. Generally, lime mortars (or non-hydraulic mortars) are characterized by less than 3% of structural bound water (SBW) of the hydraulic components and high CO 2 amount (>32%), whereas mortars with higher amounts of water bound and proportionally small quantities of CO2 are considered hydraulic [48].
The CO 2 to structurally bound water ratio in relation to CO 2 percentage (% weight loss in the temperature range of 600-800 • C) is shown in Figure 10. The inverse trend of hydraulicity of mortars is being augmented exponentially with CO 2 . This representation allows a good classification of the mortar nature [24,46,48,49]. From the observation of Figure 10 and the values in Table 8  °C the weight loss is due to dehydration of hygroscopic or adsorbed water (i.e., phyllosilicates), from 120 to 200 °C the weight loss of water from hydrated salts occurs, between 200 and 600 °C the weight loss is due to structurally bound water (SBW) from the hydraulic compounds and, finally, the loss of CO2 as a consequence of the decomposition of calcium carbonate (CaCO3) takes place at temperature range between 600 and 900 °C [24,46,47]. Generally, lime mortars (or non-hydraulic mortars) are characterized by less than 3% of structural bound water (SBW) of the hydraulic components and high CO2 amount (>32%), whereas mortars with higher amounts of water bound and proportionally small quantities of CO2 are considered hydraulic [48].
The CO2 to structurally bound water ratio in relation to CO2 percentage (% weight loss in the temperature range of 600-800 °C) is shown in Figure 10. The inverse trend of hydraulicity of mortars is being augmented exponentially with CO2. This representation allows a good classification of the mortar nature [24,46,48,49]. From the observation of Figure 10 and the values in Table 8 samples from Villa del Capo are highly hydraulic mortars and can be classified as natural pozzolanic mortars.

Porosity
Mercury intrusion porosimetry were performed on three selected samples: BG13 for Group A and BG13 and BG17 for Group B, due to the scarce amount of available material. Table 9 reports cumulative volume, bulk density, apparent density, open porosity, and specific surface, whereas Figure 11 shows representative pore size distribution. Open porosity of BG13 sample (Group A) was slightly lower (38.25%) than BG5 and BG17 (B group; 40.03%-42.27%), all analyzed mortars show unimodal and broadened shape of cumulative pore size distribution.
Bulk densities of samples were within the range of 1.38 g/cm 3 to 1.57 g/cm 3 , whereas apparent densities were within the range of 2.57 g/cm 3 to 2.89 g/cm 3 . Relative volume curves are positively skewed and highlighted that pore radii mainly range between 5 and 100 nm. Pore sizes of samples fall within the characteristic field of hydration product porosity, usually considered below 100 nm [50,51].
Comparing these results with those obtained for modern hydraulic mortars, reported in [3], it is clearly evident that Villa del Capo mortars, as well as other ancient roman pozzolanic mortars of previous study [3,19], possess very small pore size radii (5-100 nm), about one order of magnitude lower than modern hydraulic mortars (100-1000 nm). These differences are probably due to the vesicular structure of pozzolanic materials (i.e., pumice) that represents a fundamental feature of the complex pore structure of the cementitious matrix of ancient mortars. Secondary minerogenetic products, in this case hydrocalumite and C-A-S-H gels, fill pores, thus enhancing bonding of pumice clasts and making mortars less permeable [36,52]. These findings (1) decrease the possibility of alteration of the mortars; (2) increase mechanical resistance, and consequently, (3) also increase durability of manufacts.

Conclusions
This work presents an extensive multi-analytical characterization of ancient mortars from a very important Roman archaeological site: Villa del Capo in the Sorrento Peninsula.
A petrographical study of the investigated samples highlighted some important as- Figure 11. Cumulative (continuous lines) and relative (dashed lines) pore size distribution Villa del Capo mortars (BG5; BG13; BG17).

Conclusions
This work presents an extensive multi-analytical characterization of ancient mortars from a very important Roman archaeological site: Villa del Capo in the Sorrento Peninsula.
A petrographical study of the investigated samples highlighted some important aspects regarding production technology of materials, suggesting different recipes for different mortar types.
Bedding mortars result from: (1) a mixture of slaked lime, water, fine grained volcanic materials, and aggregates of volcanic rocks, whereas, coating and floor mortars may be considered as (2) a mixture of slaked lime, water, fine grained volcanic, and ceramic materials with volcanic and ceramic aggregates.
The "recipe" of coating mortars is identified as Cocciopesto or Opus signinum, a typical building technique used in ancient Rome for lining of tanks, terraces, thermal environments, and flooring [26].
Geomaterials used for mortars production had local provenance and are very well consistent with the surrounding geological setting. In fact, volcanic fragments, scoriae, pumice, and crystal fragments (i.e., clinopyroxene, feldspar, and garnet) derived from both pyroclastic rocks of the Campi Flegrei district and from rocks of the Somma-Vesuvio complex, as inferred by optical microscopy, and mineralogical and chemical composition. In coating and floor mortars (group B and C) there is an addition of ceramic fragments that improve pozzolanic aptitude of the mortar. It was not possible to define their provenance, due to strong differences among samples, which likely suggest a recycling of materials.
Provenance of carbonate rocks used to produce lime for the investigated materials is still unknown, even if it is highly reasonable to assess that they were produced "on site" from carbonate deposits of Mesozoic age that border the Campanian plain ( Figure 2). SEM-EDS analysis and TGA investigation revealed that studied mortars can be classified as hydraulic mortars, which is also confirmed by reaction rims observed around pozzolanic materials.
XRPD and SEM-EDS analysis highlighted that binder fractions are characterized by different secondary products (reaction products), including amorphous C-A-S-H gel, calcite, gypsum, and hydrocalumite.
Results of porosity tests together with microstructural observations suggest the products related to minerogenetic secondary processes, especially gel C-A-S-H and hydrocalumite settling in the porous of pozzolanic materials, make mortars more resistant.
Studies like this, that entail meticulous characterization of all the components for the realization of the artificial geomaterials, paying special attention to secondary minerogenetic processes, are crucial in the perspective of future restoration works, and especially for the preservation of archaeological sites that cannot be independent from thorough knowledge of materials used for the construction and of their state of conservation, along with that of the building.
Author Contributions: Sampling, C.R. and R.E.; archaeological information, R.E.; formal analysis and data curation, C.R., L.G. and P.C.; writing-original draft preparation, C.R.; writing-review & editing, C.R., L.G. and P.C.; supervision, P.C. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: All data derived from this research are presented in the enclosed figures and tables. additional samples was supported by the following AIM projects [AIM18352321-1; AIM1835232-1, CUP: E66C18001140005]. The authors would like to thank Roberto de Gennaro for the invaluable assistance during EDS microanalyses, Sergio Bravi for his technical ability in thin sections preparation and Tommasina Budetta, former Sovrintendenza archeologia della Campania that provided authorization and useful support on sampling. Thanks, are also due to CTG Italcementi Heidelberg Group for supporting this research activity. The authors would like to thank the journal editors and the anonymous reviewers for their useful comments that helped us in the revision of this manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.