A Chronology of Ancient Earthquake Damage in the Modena Cathedral (Italy): Integrated Dating of Mortars (14C, Pollen Record) and Bricks (TL)

ABSTRACT The 15th century cross-vaults of the medieval Modena Cathedral (UNESCO site) consist of intricate patches of different masonry portions bound by three types of lime mortars and at least two types of gypsum mortars. Such anomalous structure suggests multiple repair works over time after damaging earthquakes. The absolute dating of lime mortars (14C) and bricks (TL) integrated with the pollen record of mortars allowed to clarify the construction and restoration history of the vaults and to link the repairs to the earthquake chronology for the area. The results reveal that the original construction of the vaults (1404–1454) was carried out using lime mortar binding reused Roman and medieval older bricks. Lime mortar was used also for later repairs caused by earthquakes in the 16th and 17th centuries. Gypsum mortars were then used to entirely rebuild some vaults and to repair others in the 18th and 19th centuries. The study indicates that unexpected damage could be revealed by the detailed chronology of masonry binders. These data represent fundamental steps to implement earthquake risk assessments and strengthening projects of ancient buildings.


Introduction
In 2012 a seismic crisis struck the central Emilia Romagna region (northern Italy), strongly damaging the monumental heritage (Parisi and Augenti 2013). The medieval cathedral in Modena suffered significant damage, as brick fragments fell to the ground from the cross vaults (Baraccani et al. 2017). The vaults were a 15th century addition to the original 12th century structure of the cathedral. A complex network of fractures cut the vaults, witness of previous earthquake damage, but their internal structure was hidden by a thick plaster on both sides. Our fallen fragments inspection revealed that some part of the masonry of the vaults were bound with the traditional lime binder, but others were bound with a gypsum mortar. As a strengthening project was planned to increase the safety level of the Cathedral, the plaster cover was removed and we performed the detailed survey of the 23 vaults. An unexpected scenario was revealed, most of the vaults showed a complex patchwork of juxtaposed lime and gypsum masonry, whereas others consisted onl.y of gypsum (Tirelli et al. 2020). This complex situation prompted two main questions: 1) Was the use of different binders possibly a clue for much larger ancient damage than that shown by the fracture survey alone? And 2) was lime or gypsum the original binder used for the construction of 15th century vaults?
This paper reports the results of an integrated study based on independent dating techniques to define the repair phases of the cathedral. The aim was to explore the possible correlation of the main restoration works to the earthquake chronology. Radiocarbon dating was carried out to date lime mortars and thermoluminescence (TL) to date bricks. As gypsum mortar is not directly datable using radiocarbon, we performed also pollen analyses to obtain an indirect indication of the construction timing suggested by the local vegetation record, which saw the introduction of new ornamental species in the late history of the city.

The Modena cathedral: construction, earthquakes and repairs
The Modena Medieval UNESCO World Heritage Site consists of the Cathedral and the Civic tower Ghirlandina, celebrated examples of Romanesque art (Figure 1a). The building of the monumental complex started in 1099, date of the Cathedral's foundation (Lubritto et al. 2015;Silvestri 2013), to 1319 when the construction of the Ghirlandina bell tower was completed (Piccinini 2009). The church and the tower are covered by an impressive variety of ornamental stones coming from the despoliation of Roman monuments (Lugli et al. 2009) and also a large part of the bricks forming the core structures were taken from ancient Roman buildings (Panzeri et al. 2019). The cathedral roof was modified in the 15th century with the addition of 23 cross-vaults (Figure 1b). Indirect and fragmentary clues from documents describing building contracts, accounts and correspondence indicate that the vaults were constructed in different phases, from 1404 to 1454 (Baracchi 1993). The first vaults to be built were the easternmost in the central nave (C1, from 1404 to 1415) and in the southern nave (S9, from 1404 to 1433) The last one was the westernmost in the central nave, next to the façade (1447?-1453; see Table 3 for the complete chronology).
Ancient documents indicate that the cathedral suffered multiple damaging events, but the location and extent of the repairs is virtually unknown. The catalogue of Italian earthquakes (Rovida et al. 2019) reports a number of historical earthquakes that struck the area after the vaults construction. According to the estimated values of Peak Ground Acceleration (PGA) for the cathedral (Baraccani et al. 2014

Vaults masonry survey
We performed a detailed survey of the extrados of the 23 vaults (Figures 1c, 2) to map the presence of different mortars after the preliminary map reported in Tirelli  (2020). The vaults have been numbered starting from the apses according to the nave orientation (north, central and south). Vaults N1 and S1 were explored only through two small openings in the mortar cover.

Petrographic analysis
A total of 32 mortar samples (lime and gypsum) were collected for petrographic analyses. Thin sections were prepared by vacuum impregnation with epoxy resin. The petrographic parameters of mortars (aggregate/ binder, aggregate size and sorting and porosity) were calculated from digital image analysis of thin sections taken under the optical microscope in transmitted light according to the method described by Middendorf et al 2019. Three images for sample were photographed form 4.5 × 3.0 mm thin sections and evaluated using the program ImageJ. Only three samples (C2-3AM, N2-2DM and S1-2M) were not processed by image analysis due to high sample fragmentation. The aggregate sand supply sources have been identified by petrographic composition according to Lugli, Marchetti Dori, and Fontana (2007).

Radiocarbon dating of lime mortars: sequential dissolution method
Six lime mortars from the vaults showing multiple repairs (C1, C2, C3, C5, N9) were prepared for radiocarbon dating following the sequential dissolution methodology (Heinemeier et al. 2010;Lindroos, Heinemeier, and Ringbom 2019;Lindroos et al. 2007). The mortar samples were crushed and dry sieved using increasingly fine mesh widths ranging between 75-500 μm in order to separate the soft carbonate binder mineral dust from the harder limestone particles. The grain-size fraction <100 μm was subsequently wet sieved and the 46-75 μm grain-size fraction was isolated and dried for 14 C dating. The alkalinity test with phenolphthalein dissolved in alcohol was performed on the coarser grain size fraction (301-500 μm) to assess the presence of calcium hydroxide, which would react with the modern CO 2 in the atmosphere, producing secondary calcite. For this reason, alkaline samples are rejected for AMS dating. The 46-75 μm grain-size fraction was observed with cathodoluminescence (CL) to identify possible contaminants sources (underburned limestone fragment and carbonate aggregate).   Dissolution was done in a vacuum system with 85% phosphoric acid at 0° poured on the sample powders. Several consecutive CO 2 fractions were liberated and collected at different stages of the dissolution process. To minimize the amount of CO 2 from slowly dissolving contaminant carbonates, the first fraction was collected within 10 s. The following fractions were isolated after reaction times listed in Table 1. The mortar samples C1-1M, C2-3M, C3-1M, C5-4M, C5-5M, N9-3M, were analyzed in three CO 2 fractions. The vials containing CO 2 were analyzed at the AMS 14 C Dating Centre of Aarhus University and at the AMS facility Laboratory of Ion Beam Physics ETH Zurich. The conventional 14 C dates were calibrated using the OxCal 4.3 program (Bronk Ramsey 2017). All calibrated results are reported at the 95% confidence level (2 sigma).

TL dating of bricks
The thermoluminescence dating was performed on the same vaults selected for 14 C dating, analyzing ten brick samples (C1-1L, C3-1L, C5-1L, C5-4L, C5-5L, C5-6L, C5-7L, N3-3AL, N9-1L, N9-3L). Only in one case (C1-1L) we could not sample and date the binding mortar due to restoration works. We used the fine-grain dating technique (Zimmerman 1971), which requires a relatively small amount of sample. The samples were prepared under dim red light, and the polymineral fine grain fraction (1-8 μm) was deposited on stainless steel discs. The fine-grain fraction was obtained from a suspension of the sample powder in acetone, an ultrasonic bath being used to disperse coagulation. Grains in the size range of 1 to 8 μm were separated by settling time of 20 and 2 minutes, respectively, using a 60 mm column (Aitken 1985). For the evaluation of the palaeodose, the TL Multiple Aliquot Additive Dose protocol (MAAD, Aitken 1985) was applied. TL measurements were performed using a home-made system based on the photon counting technique with a photomultiplier tube (EMI 9235QB) coupled to a blue filter (Corning BG12). The samples were heated from room temperature to 480°C at 15°C s −1 . Artificial irradiations were carried out by a 1.85 GBq 90Sr-90Y beta source (doserate: 3.57 Gy min −1 ) anda 37 MBq 241Am alpha source (dose-rate: 14.8 Gy min −1 ).
The annual dose rate was directly evaluated from the measurement of the radioactivity of the samples and their surroundings. The 238 U and 232 Th concentrations were obtained by total alpha counting using ZnS (Ag) scintillator discs assuming a Th/U concentration ratio equal to 3.16 (Aitken 1985).
The contribution due to 40 K content was obtained from the total concentration of K measured by flame photometry. For the annual dose rate calculations, an amount of water corresponding to 20 ± 5% of saturation was assumed The external dose rate was evaluated on site with an ionization chamber for environmental dosimetry, to avoid the problem of inhomogeneity of the building materials around the samples.

Pollen analysis
A total of 6 pollen samples (C5-6M, C5-7M, N3-3AM, N8-2M, S9-1M, S9-4M) were extracted from the masonry after carefully chiseling away the exposed surface to avoid contamination by preset-day pollen rain. The external rim of the sample fragments were eliminated in the laboratory and the sample cores were treated for pollen extraction by sieving and heavy liquid separation according to Van der Kaars et al. (2001) and Florenzano et al. (2012). Lycopodium tablets were added to calculate concentrations (p/g = pollen/gram). Pollen identification was performed under the optical microscope at 1000x magnification on permanent slides by means of reference pollen collection, keys and atlases (e.g. Moore, Webb, and Collins 1991;Reille 1992).

Map of the vaults mortars
The map of the 15th century vaults shows a patchwork of lime and gypsum mortars ( Figure 2). The complex juxtaposition of different mortars does not allow to distinguish the original construction portions from the later repairs just using cross-cutting relationships. The repair sequence is particularly intricate in the northern nave (vaults N7, N8 and N9) and for the western part of the central nave (C4 and C5), where at least two different gypsum mortars were identified (G1 and G2, Figure 1c). All the mortars in the southern vaults consist of gypsum, with the only exception of in the westernmost vault (S9), where a small portion of lime mortar is present.
It was not possible to map all the different types of lime mortars by visual inspection alone, because a discrimination can be done only by petrography and radiocarbon dating, as illustrated below.

Petrography of lime mortars
The lime-based mortar show frequent rounded underburned limestone fragments up to 5 mm, with common micritic reaction edges, which fade into the surrounding binder, and scarce lime lumps, typically a few millimeters in diameter or less. The aggregate consists of fluvial sand with the following composition: micritic limestone fragments, sparry calcite, monocrystalline and polycrystalline quartz, k-feldspar, plagioclase, siltstone fragments, fossil fragments, serpentinite fragments, biotite, iron oxide and/or hydroxide grains. Six samples also present abundant bricks fragments (cocciopesto). No calcite neomorphism was observed in any of the samples. The mortars were classified in four groups based on the aggregate grain size, the aggregate binder ratio and sand provenance (Table 2).
Group 3 is characterized by lime mortars with an aggregate binder ratio from 2.1 to 2.2 and contains frequent underburned fragments (N5-1M, N8-1M, N9-3M). The aggregate grain size is from fine to medium sand, moderately sorted. The sand provenance is from the Panaro River ( Figure 3c, Table 2).
Group 4 (samples C1-4AM, C4-1AM and N3-2AM) consists of the lime plaster coating the upper side of the vaults. The aggregate binder ratio is from 2 to 2.1, the aggregate is well sorted and no underburned limestone fragments are present. The aggregate grain size is from fine to medium sand and the provenance is from the Secchia River ( Figure 3d, Table 2).

Petrography of gypsum mortars
Two different types gypsum mortar (G1 and G2) were identified by cross-cutting relationships and masonry texture (Figure 3e, sample C5-6M, gypsum mortar G1, and sample C5-7M, gypsum mortar G2, Figure 3f, Table  2). They are characterized by a binder of microcrystalline gypsum containing a few deformed selenite crystal fragments, Other components are fragments of carbonate rock and shale. The two mortars show different grain size and porosity: mortar G1 includes larger selenite crystal fragments, up to 3,5 mm across, and higher porosity with large rounded pores, whereas mortar G2 contains abundant fragment of micritic carbonate rocks. Anhydrite crystals are also present as result of the high temperature firing of gypsum in the kiln areas adjacent to the burning fuel. Mortar G2 was identified only in the central vaults C4 and C5. The attribution to the G1 mortar group of the vaults in the southern nave was done by visual inspection for the presence of large selenite fragments.

Radiocarbon dating of lime mortars
The results of the alkalinity test show a low alkalinity for all samples, indicating that they are suitable to be dated. The cathodoluminescence inspection confirmed the presence of older carbonate contaminants seen under the optical microscope.
The 14 C measurements with the hydrolysis data are presented in Table 1. The dating results are evaluated from the 14 C profiles of Figure 5 and are interpreted in Table 3. The majority of the profiles displays a concave curvature, or increasing slope (Lichtenberger et al. 2015), due to the rapid dissolution of the mortar binder and the slow onset of dissolution of the contaminants (Figure 4a,c,e,g,i), see the discussion paragraph.
For sample C2-3M the first two CO 2 fractions agree within the given error margins and contamination can be ruled out (criterion I, Heinemeier et al. 2010;Lindroos et al. 2018;Ringbom et al. 2011). The combined calibration of fraction 1 and 2 yields a 14 C age of 492 ± 24 BP corresponding to a calibrated calendar age of 1409-1445 (Figure 4d).
Sample C3-1M provided a low CO 2 pressure value for fraction 1 and the hydrolysis was performed a second time in order to obtain a larger CO 2 volume (sample C3-1M.2). This fraction yielded 14 C ages of 506 ± 26 BP corresponding to a calibrated calendar age of 1399-1445 ( Figure 4f).
The results of sample C5-4M dating were rejected because the 14 C profile shows a decreasing slope (Figure 4m). In this case, it was not possible to read the binder age from the profile (Figure 4n).

Pollen in lime and gypsum mortars
The state of pollen preservation is good, concentration is low, from about 1000 to just over 2000 np/g, as reported for other mortars (Langgut et al. 2013). A total of 108 taxa were identified in the mortar samples and their association and environmental significance will be given elsewhere. Here we report the taxa relevant ( Figure 5) for dating purpose as chronological markers, such as "exotic" ornamental trees of recent introduction and plants that were extensively cultivated in particular historic periods in the area. These taxa are: 1) Ginkgo biloba in samples N3-3AM and N8-2M; 2) Ailanthus altissima in sample N8-2M; 3) Cedrus in sample N3-3AM and 3) Cannabis sativa (hemp) discovered in five samples (C5-6M, C5-7M, N3-3AM, S9-1M, S9-4M). Pollen concentration of taxa is always <1%, except for hemp, which has a range from 0.4 to 6.2%.

Mortar composition
In the 50 years' time span of the vault construction inferred from documents (1404-1454), group 2 mortar appears to have been used to build the older vaults, whereas the younger vaults were built using group 1 and 3 mortars ( Table 3). The only lime repair that we have discovered was manufactured using the mortar of group 3. It follows that the identified mortar groups have no chronological significance, but probably reflect the various contractors that worked on the vault construction and later repairs using their own preparation techniques and raw materials in various phases. The variable aggregate grain size may be due to collection of sand in the Panaro river from distinct sedimentation areas, such as meander bars, longitudinal bars or terraces and in different time periods of the year, during the flood (spring and fall) or the dry seasons (summer and winter). Sands from the Panaro river were also used as aggregate for the construction of the apses (mortar A and B of Caroselli et al. in press).
The lack of underburned fragments in group 4 suggests a relatively recent, semi-industrial, lime preparation technique, possibly dating to a 19th century intervention designed to protected the fractured vaults with a thick lime mortar cover. Aggregate sand supplied from the Secchia river started to be used in the cathedral only since the 13th century modifications (Caroselli et al. in press).
The use of gypsum mortars in the rebuilding and in the repairs of the cathedral vaults represents an unexpected, unique case. In the area the use of gypsum as binder started only from the second half of the 17th century (Lugli 2019) when the famous scagliola artworks from Carpi (Modena) and the stucco decorations typical of the baroque art became very popular. Gypsum mortar in brick masonries was normally used for very limited repairs and not for structural elements.
The reason why the ecclesiastical authority decided to switch from lime to gypsum mortar is unknown. Gypsum mortar would have been less expensive in terms of: 1) fuel consumption, as lower temperature are required for firing gypsum (105°C) than lime (in excess of 800°C) and 2) preparation technique, as gypsum does not need to be mixed with a sand aggregate because of it expansion properties upon hydration. Gypsum mortars are also lighter and provide faster setting time than lime mortars. The carbonate and shale fragments observed within the gypsum groundmass are not added elements, but derive from the selenite raw rock used to prepare the mortars (Figure 3f, Lugli et al. 2010).
Unfortunately, gypsum mortars cannot be directly dated, and in some cases the reuse of bricks older than the vaults construction time does not allow to pinpoint the timing of the repairs (see below). The cross-cutting relationships indicate that gypsum mortar G1 predates G2 (Figure 1).

Radiocarbon dating
The Modena cathedral lime mortars represent one of the most challenging case for radiocarbon dating. This is because the mortars are rich in incompletely burned limestone fragments and the sand aggregate consists of 33.7-43.3% of carbonate grains (Panaro river, Lugli, Marchetti Dori, and Fontana 2007). These components represent potential contaminants, possibly leading to older ages than the construction time. But the advantage of the sequential dissolution method is that the dead carbon influence is revealed by the shape of the 14 C profile, as contaminants dissolve at a slower pace than the binder and their effect is negligible in the first collected CO 2 fraction (fraction 1, Lindroos et al. 2018).
The cocciopesto-rich mortars (group 1) could also provide biased results because they contain less dateable carbonate, are less permeable to atmospheric CO 2 and may harden very slowly. They may also contain uncarbonated calcium hydroxide, which could capture modern 14 C providing younger ages than the actual construction time Michalska, Czernik, and Goslar 2017;Ringbom et al. 2011Ringbom et al. , 2014. In our case, only one sample was affected by the problem, as revealed by the decreasing slope of the 14 C profile demonstrating delayed hardening (C5-4M, Figure 4m).
All the tested samples are in the range of the vaults construction time, although a slight influence of dead carbon contamination may have caused steep 14 C profiles and a minor shift towards older ages for the initial CO 2 fractions, (Figure 4). These results highlight the good potential for the sequential dissolution method to recognize and reduce to the minimum the contamination effect, even for the worst-case scenarios given by high dead carbon from underburned fragments and carbonate aggregate, and the delayed carbonation effect induced by cocciopesto.

TL dating
The results of the thermoluminescence dating of brick indicate several repairs carried out using bricks produced in the 17th, 18th and 19th centuries ( Table 3). Some of the restoration works also extensively reused bricks manufactured before the construction of the vaults, in the 12th and 14th centuries, and even ancient Roman bricks. The reuse of older bricks was documented for the medieval core structure of the tower and the cathedral (Panzeri et al. 2019), but was not known for the late additions, such as the 15th century vaults. Ancient documents attest that the original construction works were often halted because of the lack of building materials (Baracchi 1987). Ornamental stones and bricks were available only when buried roman monuments were discovered and exploited (see discussion in Lugli et al. in press). As a general rule, the reuse of older bricks could have been a common practice in the earlier repair phases (16th to 17th centuries), although we cannot exclude that bricks coming from demolition of older (medieval) buildings, were also used also for the 18th and 19th century repairs.

Pollen record
The pollen spectra record plants from to the areas nearby the city center trapped during the laying and setting of the mortars. Ginko tree (Ginkgo biloba L.) arrived in Italy (Padua, Milan and Pisa) in the second half of the 18th century (Maniero 2000;Saccardo 1909; Targioni Tozzetti 1853) and its fortune as ornamental plant was rather quick, becoming widely cultivated throughout Europe in the mid-19th century (Crane et al. 2013).
Tree of Heaven (Ailanthus altissima (Mill.) Swingle also arrived in Italy in the mid-18th century and by the mid-19th century it appears widespread and perfectly naturalized in some areas, becoming a highly invasive species (Maniero 2000;Pignatti, Guarino, andLa Rosa 2017-2019;Saccardo 1909).
These two marker plants, Ginkgo and Ailanthus, are listed in the oldest Index Plantarum of the Modena Botanical Garden (Fabriani 1811), which is located not far from the cathedral, and therefore the taxa are witnesses of the urban area at least from the beginning of the 19th century.
Hemp (Cannabis sativa) cultivation started in the area since the Bronze Age (Mercuri et al. 2006) and grew continuously through the Middle Age to the beginning of the 19th and 20th century (Casalgrandi 2018;Venturi and Amaducci 1999), but does not appear in the present-day pollen rain in town (Mercuri, Massamba N'siala, and Barbieri 2001). The relatively high concentration seen in samples S9-1M and C5-7M (up to 6.2%) requires rather extensive field cultivation, setting the age of these two mortars to repairs conducted in the second half of the19th century (Table 3 and Figure 6).

The chronology of vault construction and repairs
The concomitant presence of lime and gypsum binder in the mortars of some vaults and the exclusive presence of gypsum in others clearly points to multiple repair phases, which led also to the complete reconstruction of some vaults using at least two different gypsum binders. Such extensive works could only have been triggered by severe damage caused by earthquakes. The possible damaging earthquakes may be identified comparing the time lines given by the 14 C and TL chronological data (Table 3) coupled with the pollen spectra interpretation, as illustrated in Figure 6.
Radiocarbon dating demonstrates that lime mortar was the original vault construction binder from 1404 to 1454. This is also the case for the southern nave, where only a small portion of original lime masonry (vault S9) survived the complete reconstruction of unknown timing with gypsum mortar.
Vault C1 was originally built using lime mortar, and was subsequently partially reconstructed using lime mortar in the 17th century, possibly as a result of the 1660, 1671 earthquakes, as deduced by the brick dates.
Vaults C3 is the only one originally built with bricks produced at the time of the vault construction (15th century). The overlapping 14 C of the mortar and TL date of the bounded bricks allows to refine the construction time from 1440 to 1449, which matches the timing inferred by documents (1446-1447?).
Vault N8 represents the only case where lime mortar provided radiocarbon calendar ages that are younger (16th-17th century) than the original construction time. The vault was modified at least two times. A first restoration by lime mortar and re-used bricks was carried out immediately after the original construction, possibly as result of the 1501, 1505, 1586, 1660 or 1671 earthquakes. A second intervention was then probably triggered by the 1811 earthquake, as suggested by pollen content and brick dating record, or is the result of multiple repairs in the 18th and 19th century. New bricks were used for this second repair, kept together by a gypsum mortar.
Vault C5 was originally built using lime mortar and re-used bricks. At least two different restoration phases followed. A first partial reconstruction was carried out using gypsum mortar G1 and older bricks. A second intervention was performed using gypsum mortar G2 in the 19th century, probably to mend the damage caused by the 1811, 1832 or 1850 earthquakes.
Vault N3 was modified in the 19th century using gypsum mortar, possibly following the 1811, 1832 or 1850 earthquakes, as suggested by the brick dates and pollen record.
Vaults N9 and S9 were originally built using lime mortar and older bricks, whereas gypsum mortar was used for later repair with re-used bricks. Unfortunately, as for the previous vaults modified with gypsum mortar and re-used bricks, the timing of the later intervention could not be defined, but pollen markers found in vault S9 suggest possible late repairs after the 1850 or 1873 earthquakes.
These considerations indicate that lime mortar was rarely used for later repairs in favor of gypsum. Mortar gypsum G1 in the central and northern vaults dates to the 18th-19th centuries, whereas G2 was used in the 19th century, as revealed by brick dates. In the southern nave, mortar G1 could have been deployed even before the 18th century, but the large reuse of ancient brick prevented us to recognize when gypsum started to be preferred instead of lime for the repairs.

Conclusions
The complex structure of the 15th century vaults of the Modena cathedral consists of irregular patches of lime and gypsum mortars suggesting multiple repairs due to damaging earthquakes. The integrated chronology obtained by three independent dating methods has allowed to clarify the vaults construction and restoration history, assessing the extent of the damaged areas.
The sequential dissolution method provided reliable radiocarbon results for lime mortars matching the historic documents time span, despite the potential contamination by dead carbon due to the large presence of underburned limestone fragments and carbonate grains in the aggregate. The thermoluminescence of bricks demonstrated that the cathedral represents an unprecedented example of building material re-use even for the 15th century structural modifications.
The petrographic and absolute dating results revealed that different types of lime mortars were used both for the original construction of the vaults from 1404 to 1454 and for later repairs in the 16th and 17th centuries. At least two types of gypsum binder were subsequently used to repair some vaults in the 18th and 19th centuries. Other vaults were entirely rebuilt using gypsum mortar. The reasons and age of this unprecedented large use of gypsum binder for structural elements are unknown, but were possibly driven by construction cost. Several damaging seismic events were possibly responsible for the reparation works, which are not documented in the historical records.
The results of this multidisciplinary research highlight the possibility that unexpected and severe damage could be revealed by detailed binder chronology, which may add further information to implement earthquake risk assessment and structural strengthening projects. This is because crack and fracture survey alone may reveal only the relatively recent damage pattern suffered by ancient buildings