Love Eriksson Vt 2018
Examensarbete, 15 hp
Master´s programme in Environmental Archaeology, 120 hp
Hyperspectral imaging on mortars from the inner walls of Carcassonne
A study on the application of near infrared spectroscopy as a non- destructive classification method on historical mortars
Love Eriksson
Love Eriksson Vt 2018
Examensarbete, 15 hp
Master´s programme in Environmental Archaeology, 120 hp
Preface
Photo of inner wall by Kajsa Arnberg. Thank you Kajsa for letting me to use this wonderful photo!
I would like to thank my supervisors, Peter Holmblad, Johan Linderholm and Philip Buckland who have guided me and for their tremendous help through my studies here at the university.
And for this thesis also Paul Geladi who helped me when using the hyperspectral camera and later analysis as he read through and gave good advice on how to improve it. And most importantly had it not been for Claudia Sciuto would this thesis not been possible as she not only provided me with the material and interesting subject to work with but also helped me in the discussion and analysis of the material that she was very knowledgeable about.
I am very grateful for my classmates in the master’s program who have made the studies so much fun and interesting as they brought with them new perspectives on our prehistory.
Abstract
. The aim of this thesis is to study and evaluate the application of hyperspectral image analysis as a non-destructive analysis method for historical mortars. This method was applied on 35 sampled mortars in varying sizes and type from the inner walls of the fortified medieval city Carcassonne. By using near infrared spectroscopy and classifying the complex multivariate data by applying the SIMCA method (Soft Independent Modelling of Class Analogies) it is possible to conduct an in depth analysis of the samples. This can then further our understanding about the construction phases as well as construction techniques used as indicated through the chemometric analysis that can identify and group the mortars in accordance to raw material and transformation process. From this could four distinct groups be found in the PCA models, two Roman periods and two high medieval periods, allowing to study Carcassonne prior to and after its enclosure. A find from the first Roman periodindicates on a bathhouse or public building existing prior to the construction of the defensive wall, leading to the hypothesis that maybe more parts of the inner wall might contain older structures like this. The application of hyperspectral image analysis on historical mortars has proven itself a useful tool and simple method for studying mortars.
Abstract
. Målet med denna uppsats var att studera och evaluera applikationen av hyperspektral bildanalys som en icke-destruktiv analysmetod på historiskt murbruk.Instrumentet applicerades på 35 murbruksprover i varierande storlek och typ tagna från de inre murarna av den befästa medeltida staden Carcassonne. Med nära infraröd spektroskopi och klassificering av den multivariata genom SIMCA metoden (Soft Independent Modelling of Class Analogies) var det möjligt att göra en djupgående analys av proverna. Detta
tillvägagångssätt kan då främja vår förståelse om stadens konstruktionsfaser och
konstruktionstekniker som indikeras genom den chemometriska analysen som kan identifiera murbruket utefter råmaterial samt hur murbruket tillverkats. Från dessa metoder kunde fyra distinkta grupper finnas i PCA modellerna, två romerska perioder och två högmedeltida perioder, vilket öppnade för tolkning både innan och efter stadsmurarna rests. Ett fynd från den första romerska perioden indikerar på förekomsten av ett badhus eller publik byggnad vars väggar sedan återanvänts vid konstruktionen av den inre stadsmuren. Detta fynd leder till hypotesen att potentiellt andra delar av den inre stadsmuren kan innehålla väggar från äldre byggnader som denna. Applikationen av hyperspektral bildanalys på historiskt murbruk har påvisat sig ett användbart verktyg och simpel metod för att studera murbruk.
Keywords
Environmental archaeology, Hyperspectral imaging, NIR-spectroscopy, Non-destructive, PCA-models, Historical mortars, Carcassonne
Language English
Index
1. Introduction ... 1
1.1 A short history about Carcassonne ... 2
1.2 The chemistry of mortars ... 3
1.3 Historical documents on mortar production ... 4
1.4 Different time, different construction technique? ... 4
1.5 Aim and research questions ... 5
2. Theory ... 6
3. Material ... 7
3.1 The different uses of the wall as shown by mortar ... 7
4. Method ... 9
4.1 Hyperspectral imaging (HSI) ... 9
4.2 Preparation of mortars for scanning ... 9
4.3 Statistics ... 10
4.3.1 PCA analysis ... 10
4.3.2 SIMCA method ... 10
5. Results ... 12
5.1 Image analysis of red Roman pigments ... 12
5.2 Problem encountered using HSI ... 15
5.3 PCA analysis ... 18
5.4 SIMCA ... 20
6. Discussion ... 22
6.1 Application of HSI and NIR ... 22
6.2 Mortars and their compositions ... 23
7. Conclusion ... 28
Reference list ... 29
Appendix ... 1
Figure index
Tables
Table 1. Official phases as used by the PCR-programme ordered into groups named in the text
……….5 Table 2. The sampled mortars and their composition. The type of mortars are based on ocular observations made by Claudia Sciuto in MAL (Environmental archaeology laboratory in Umeå). ………14
Figures
FIGURE 1.LOCATION OF CARCASSONNE IN OCCITANIE,AUDE IN FRANCE ... 1 FIGURE 2.RECTIFIED ORTHOPHOTO OF THE CURTAIN WALL BETWEEN TOWERS 25 AND 26.
DETAILS OF SAMPLED MORTARS (SEE SAMPLE LOCATION IN APPENDIX 1). FOTO:CLAUDIA
SCIUTO ... 8 FIGURE 3.LOADING LINE BEFORE (ABOVE) AND AFTER (BELOW) REMOVING WAVELENGTHS
BELOW 1000NM.IT BECOMES CLEAR THAT THOSE BELOW 1000NM ARE SOLELY
DISTURBANCE GATHERED FROM STRAY LIGHT WHEN SCANNING. ... 12 FIGURE 4.IMAGE ANALYSIS BEFORE CLEANING THE IMAGE FROM BACKGROUND AND SHADOWS.
THE CLUSTER THAT NEEDS TO BE EXCLUDED CAN CLEARLY BE SEEN IN THE RIGHT OF THE
PCA, HOWEVER ONE MUST BE CAREFUL TO NOT EXCLUDE ALL WHICH SEEMS TO BE
OUTLIERS AS THESE MIGHT ALSO CONTAIN PARTS OF THE AGGREGATE OR OTHERWISE. ... 13 FIGURE 5.AFTER CLEANING UP THE IMAGE FROM SHADOWS AND BACKGROUND THE DIFFERENT
TYPES OF MORTARS AND THE RED PIGMENTATION BECOMES VERY CLEAR IN THE SECOND COMPONENT.AND AS CAN BE SEEN IN THE PCA THE AGGREGATES AND MORTAR (B)
GATHER TO THE LEFT AND THE RED PIGMENTATION (A) FORMS ITS OWN CLUSTER TO THE RIGHT. ... 14 FIGURE 6.STRIPING EFFECT SEEN IN THE SECOND COMPONENT OF THE IMAGE ANALYSIS ON
CAR16_M2.THIS WAS THE FIRST MORTAR SCANNED IN THE SECOND RUN, WHICH MIGHT HAVE RESULTED IN DUST FALLING ON THE WHITE REFERENCE AS THE CAMERA WAS
REFOCUSED AND NOT THOROUGHLY CLEANED. ... 16 FIGURE 7.THE MORTAR SAMPLE CAR16_28 FROM THE FIRST RUN AND RESCANNED FOR
QUALITY ASSURANCE.THE ORIGINAL SCAN TO THE LEFT HAS A SLIGHTLY MORE “POINTY”
SCATTERPLOT THAN THAT OF THE RERUN AS SEEN ON THE RIGHT.THE RERUN CANNOT SHOW THE EXACT SAME RESULT AS THE FIRST RUN, HOWEVER, BOTH DISPLAYED THE STRIKINGLY SIMILAR RESULTS WITHOUT ANY SIGN OF A STRIPING EFFECT IN THE SECOND COMPONENT. ... 17 FIGURE 8.LOADING LINE FOR AVERAGE SPECTRA OF ALL ANALYSED MORTARS PRESENTED IN
THE PCA MODELS. ... 19 FIGURE 9.PCA MODEL OF ALL SAMPLED MORTARS.THESE SHOW A GENERAL TREND FROM
OLDER (LEFT) TO YOUNGEST (RIGHT), WHY THIS IS MIGHT BE BECAUSE OF EROSION,
WEATHERING OR OTHER FACTORS SUCH AS MATERIAL USED TO PRODUCE THE MORTAR.THE SECOND COMPONENT SEEM TO REPRESENT AGGREGATE AS THE DIFFERENT TYPES OF
MORTARS GROUPS TOGETHER... 19 FIGURE 10.SIMCA ANALYSIS OF THE WHOLE DATASET.THERE ARE OUTLIERS FROM THE FIRST
AND SECOND PHASE BUT WHY THEY ARE SPECIAL IS CURRENTLY OUTSIDE THE REACH OF
THIS THESIS, BUT HEY ARE DISTINCT IN THEIR GROUPING IN COMPARISON TO WHAT IS SEEN IN PHASE FOUR. ... 20 FIGURE 11.SIMCA ANALYSIS CONSISTING ONLY OF PHASE ONE AND TWO (ROMAN PERIOD)
EXCLUDING ALL THE OUTLIERS AND OTHER PHASES TO BETTER OBSERVE THE SIMILARITIES OF THE MORTARS.THE TWO GROUPS ARE DISTINCT FROM EACH OTHER BUT THEY BELONG TO THE SAME CLUSTER.SEE PCA IN FIGURE 6 FOR CLUSTERING OF PHASE 1 AND 2. ... 21 FIGURE 12.LOCAL MODEL OF ROMAN SAMPLES FROM PHASE 1-2, EXCLUDING BOTH HYDRAULIC
MORTARS AND THE MEDIEVAL RENOVATION.THIS MUCH LIKE THE SIMCA ANALYSIS SHOWS A CLEAR RELATION BETWEEN THE TWO PERIODS BASED ON THE SIMILARITIES OF THE MATERIAL USED TO PRODUCE THE MORTARS... 25 FIGURE 13.LOCAL MODEL OF ROMAN SAMPLES FROM PHASE 1-2, EXCLUDING BOTH HYDRAULIC
MORTARS AND THE MEDIEVAL RENOVATION.LOOKING AT COMPONENT 2 AND 3 FURTHERS THE NOTION THAT THE TWO PERIODS HAVE USED SIMILAR MATERIALS OR PRODUCTION TECHNIQUES, HOWEVER, THE GROUP IS NOT AS CLOSE AS SHOWN ABOVE BUT IT HAS FEWER OUTLIERS IN THE SPACED CLUSTER. ... 26 FIGURE 14.LOCAL MODEL OF ALL HYDRAULIC MORTARS SAMPLES.THEY ALL DERIVE FROM
PHASE ONE EXCEPT FOR TWO POINT.,THIS IS PROBABLY BECAUSE OF THE SECOND PHASE BEING USED AS RUBBLE/WALL FILLING CONSIDERING THAT IT SHOULD NOT EXIST WITH THE EARLIER PHASE AS WELL AS NO OTHER HYDRAULIC MORTARS BELONGING TO THE SECOND PHASE ... 27
1
1. Introduction
This thesis regards the analysis of mortar sampled from several sections of the inner wall in the fortified medieval city Carcassonne. The sections of the inner walls have previously been studied using near infrared spectroscopy (NIR) with the aim of applying it in the field (Allios et al. 2016). This study was part of a larger project, PCR-programme collectif de recherche, aiming to investigate the town fortification system during which samples of mortars were taken from different sections of the inner wall. This thesis focuses on these samples in order to analyse the material and try to determine different construction and renovation phases.
The analysis have been carried out using hyperspectral imaging (HSI), applying a non-
destructive method to studying mortar. The instrument has the possibility to study and classify the material components of the mortar without dissolving it in wet chemicals or manual disaggregation, which are more common methods (Casadio et al. 2005, Fischer & Kakoulli 2006, Linderholm & Geladi 2014). Using NIR for characterising the material and chemical composition of the mortars, there is a possibility for increasing the understanding of historical construction techniques as well as possibly construction phases in Carcassonne. The scope of this study is however relatively limited both in terms of time and type of analysis, but if providing good results can act as a pilot study for discussion of further applications of HSI on historical-archaeological contexts. One such future application might allow this relatively new and advanced tool to study ancient structures and their materials in a broader perspective than before.
Figure 1. Location of Carcassonne in Occitanie, Aude in France
2
1.1 A short history about Carcassonne
When analysing Carcassonne one must rely on its history and identified phases of
construction, which often can be found in historical documents and books. This is important to know when considering which phases are represented in the material and why (see table 1).
Carcassonne is a city located in south-western France near the river Aude (fig. 1). The older part of the fortified city which is the aim of the study can be found on a cliff by the name Carcaso, the site of the old Roman colony. The area around this site has seemingly been inhabited since the prehistory but is notable for the Iberian people who lived there around 500 BC before the Romans conquered, settled and later fortified the cliff later adding walls and towers. The Romans lost the city to the Visigoths in the 5th century AD, who in turn lost it to the Saracens for some time until the Franks took firm control over it in the 8th century AD.
For each conquest, the city was fortified even further with more towers, walls and
embankments. The most intense of these periods of fortification seem to have been during the 12-13th century AD when the French king ordered the construction of an outer wall and 14 towers as well as the improvement of the inner keep (Encyclopædia Britannica 1947 & 1964, Danmarks nationalencyklopedi 1996). This high medieval expansion and restoration was not as simple as to the others where protection was the goal, this is a complex case where religion and other factors played in as the Albigensian crusade fought the Cathar heresy (Sciuto 2018).
The architect Viollet-le-Duc between 1850-1880 in a great effort to restore the city and its monuments has renovated these fortifications and the cathedral within. It is during this restauration that the fortifications somewhat changed character into a display of how people during the 19th century viewed the medieval ages, c.f. national romanticism throughout Europe (Coldstream 2002).
Table 1 Official phases as used by the PCR-programme ordered into groups named in the text
Phase Chronology (AD) Group
Phase 1 I-III century Roman Phase 2 IV-VIII century Late antiquity Phase 3 IX-X century Early medieval Phase 4 XI-XII century High medieval Phase 5 XIII century High medieval Phase 5a Louis IX High medieval Phase 5b Philippe III-Philipe IV Late medieval Phase 6 XIV-XVIII century Historical Phase 7a XIX century Indeterminate
Phase 7b XIX century Viollet-le-Duc (1850-1914) Phase 8 XX-XXI century 1920-1996
3
1.2 The chemistry of mortars
Before the aim of this study is presented, it is important to know how people in the past prepared and produced mortars. This needs understanding of both historical records as well as chemistry and geology which all play part in understanding the binder-matrix of the mortar.
Previous studies using chemical methods to study this will be presented in order to highlight some possibilities that might be detectable and observed when using HSI.
First the chemical composition and related processes must be known. Most commonly found in mortars are calcium hydroxide [Ca(OH)2] which occurs when slaking the calcium
monoxide with water turning it into usable lime which when combined with water and sand or clay will produce common lime mortar (Hägg 1989, Rapp & Hill 1998, Hobbs & Siddall 2011, Pavía & Caro 2008).
Slaking: CaO(s) + H2O(l) → Ca(OH)2(s)
The process following slaking is applying it in between the building material allowing it to harden when exposed to air (Hägg 1989, Hobbs & Siddall 2011),
Hardening: Ca(OH)2 + CO2 → CaCO3(s) + H2O
Another compound sometimes used when producing mortar derives from gypsum, a binder which is unlikely to occur in the sampled material. Gypsum occurs naturally as a combination of calcium sulphate dehydrate [CaSO4∙2H2O]. The treatment of such a material must reach a temperature between 120-200°C removing the water from the compound turning it into a powder like state. When later again combined with water it turns into a paste which can be used for either plaster or construction as gypsum mortar; this kind of mortar is sensitive to water (Hägg 1989, Hobbs & Siddall 2011).
The cement used in the restauration during the 1800s will be presumed to be Portland cement as it is not entirely known what kind of material they used. Cement has been known since the antiquity when lime was combined with volcanic pozzolan to produce hydraulic mortars, although manufactured differently (Hägg 1989, D’Ambrosio 2015). When producing the Portland cement clay is added, instead of pozzolan, which is treated in a rotary kiln until it reaches temperatures of around 1400°C after which some gypsum is added [Ca3OSiO4]. The cement can include many types of silicates, foremost calcium silicate but may also include di- or trioxides of Ca, Al and FeIII depending on where the clay is collected; these kinds of
silicates which naturally exists in pozzolans can also be found in hydraulic mortars (Hägg 1989, Miriello et al. 2010).
4 However, if the hydraulic mortar actually contains pozzolans is questioned by Pavío and Caro (2008) and Siddall (2011) who suggests that hydraulicity might also derive from the addition of ceramics to the mortar. They base this assumption on petrographic analysis, known
structures and historical records such as Vitruvius and Cato. This assumption might lead to the conjecture that the silicones could come from the sand or clay as well as ceramics themselves, which had the same mechanical properties as pozzolan in the binder.
1.3 Historical documents on mortar production
“After reaching clarity in choosing which sand to use one must be very careful with what matters lime. It should be done with white rocks or basalt. Lime which is of hard, heavy stone
is good for masonry, lime of porous, and light stone is good for plaster.”
– (Vitruvius De Architectura translated by author from Dalgren 2009, p. 40) Vitruvius recommends choosing limestone for mortar and plaster according to these criteria, but in reality this might only have been followed as far as local deposits allowed. The prominence of what lime or sand used is of secondary importance in this thesis as it aims to see the differences and not search for their respective deposits. These deposits might also be hard to find considering the complex geological situation in and around Carcassonne.
However, Vitruvius amongst others provide a great source when it comes to perceiving historical production of mortar and to some extent ways of construction. Vitruvius 10 books on architecture will be used in this thesis to understand the process of producing historical mortars along with the article from Pavía and Caro (2008) who analyses mortars and evaluating Roman records on the matter.
1.4 Different time, different construction technique?
The production of hydraulic mortars did not cease after the fall of the Roman Empire and lived on in medieval Europe as well as the Byzantine Empire, although changing names throughout the centuries. Moreover, it can be considered that because of the use of crushed ceramics instead of pozzolanic ashes that the structure is of an older descent (Pavía & Caro 2008, Siddall 2011). This does not mean that such hydraulic mortar is exclusive to that era, because something common when building during the Roman and medieval periods was the repurposing of older buildings. This repurposing might come in the forms of taking building material from existing structures or simply building over them as the structure was either
5 repurposed or expanded upon (Wickham 2010, Bork et al. 2011). This means that a blend of Roman mortars and medieval mortar could have existed as building material, mixed or layered as structures were repaired or renovated.
Construction techniques in Western Europe remained relatively the same as during the Roman Empire because of this factor, in many previous provinces masons passed on the knowledge of construction. The social conditions and structures of society however changed at a faster pace after the fall of the Roman Empire, the socioeconomical situation and changes in society later showed itself in the architecture as well.
1.5 Aim and research questions
With hyperspectral imaging and near infrared spectroscopy one can study the chemical and material structures of the mortars. By doing so and combining it with known construction phases and spatial locations, one can begin to try and formulate a relative chronology of the sampled material. However, this study will mainly focus on the application of a non-
destructive method of studying the material and chemical components of the mortars.
1. How can the application of NIR-hyperspectral imaging contribute to the interpretation and classification of historical mortars?
2. What kind of mortars can be identified and do they belong to different construction phases? If they belong to different phases what are the characteristic differences?
6
2. Theory
It is the author’s own views on the subject of theory mainly lies with what is known as
classical or cultural-historical archaeology and processual archaeology. The former because of the objective and uniformist view of categorizing and interpreting ancient tools and the latter because of its directions and applications of natural sciences in archaeology providing a theory for the interpretation of sites, past lifestyles and paleolandscapes. Much of this thesis is based on the theoretical basis of processual archaeology and its way of applying natural science for analysing and categorizing the archaeological and in this case historical source material (Olsen 1997, Johnson 2010). Moreover, it is on this scientific basis that newer methods such as NIR can better be implemented into the field of archaeology shortening the bridge between proxies.
The processualist approach and its relation to methods, source material and interpretations is based on empirical observations and tend to be more objective in its approach. It also allows for a more thorough analysis and possibility for other scientists to verify the raw data, making the process more transparent and allow for others to make their own interpretations using the same raw data. This thesis follows these ideas and will present the interpretations in an objective light focusing more on the applied method and its results only reserving some hypothesis about the wall itself.
Another key theory which will be utilized in this thesis is the relatively new concept of materiality. This theory combines elements from many other theories but mainly ideas and concepts from processualism and post-processualism, allowing for a broader and more
complex interpretation than one or the other. The concept of materiality is not to solely look at the object and its physical components but also the surrounding environment of raw materials and social practises of the people who made it (Jones 2004). This is important when observing a monumental construction like Carcassonne, because of its history and many periods of construction. To understand the different social organization of the nations which has had control over the city and the city surroundings and available raw materials and their ability to endure, is all embodied in the mortars and their material composition. Therefore, this
theoretical point of view is appropriate to apply in this case.
7
3. Material
The samples of mortar were collected between two stones meaning that only one side (if any) is exposed to weathering. In total thirty-five samples of mortar were collected in varying sizes from different sections of the inner walls. Most of the samples were in good shape and what weathering that had occurred could be overlooked when scanning the mortar. However, green lichens had partially covered some mortars; lichens grow well on limestone which is a
problem in terms of conservation of the fortifications and other structures (Friberg & Sundér 1996;103). The majority of the known samples were recognized as Roman hydraulic mortars (see Table 2 & Appendix 1). In the field notes, some considerations were made on sections that might have been restored or suspicion on what might be modern concrete covering historical mortars.
3.1 The different uses of the wall as shown by mortar
As seen from the mortars sampled from the inner walls there are two distinct functions of one segment of the wall (fig. 2). The prime use and what is found in the majority of the material is for defensive purposes and later restauration. However, some Roman materials tell of other uses such as the hydraulic mortar and two samples which display red pigments from what could be fragments of a painting (see below) and another sample which is perceived as preparation for a mosaic floor.
The mortar (Car16_2) which has the red pigments on it is likely a fragment of a painted surface and can be seen through the plaster which forms the basis of the painting and covers the coarser mortar, something which has been observed in other bathhouses in Roman Gaul (Izzo et al.2015, Coutelas 2011). The mortar itself resembles that of other hydraulic mortars found around the inner walls, however, it is suggested by Coutelas (2011) that mortar with crushed brick or pottery were rare and always limited to small masonries such as pools. This is not the case for the samples analysed as they come from more than one spatial location from the inner walls.
Car16_21 is the likely preparation for mosaic floors, and it contained a mosaic bit on it when sampled but that was not available when the material was obtained for this thesis. Such flooring has similarities with other buildings in the Roman empire (Cardoso et al. 2014, Izzo et al. 2015), and because it was found close to Car16_2 and has finely crushed pottery/brick in it, it probably belongs to the same structure which later was repurposed during the second
8 Roman phase. These two together seem to indicate a possible bathhouse previously standing where there is now a wall connecting to a semi-circular tower (fig. 2).
Figure 2. Rectified orthophoto of the curtain wall between towers 25 and 26. Details of sampled mortars (see sample location in Appendix 1). Foto: Claudia Sciuto
9
4. Method
4.1 Hyperspectral imaging (HSI)
Hyperspectral imaging has been used previously when considering conservation of mortar and stones of ancient structures (Fischer & Kakoulli 2006), but it has yet to be used to be used for analysis of the material itself. Because of the method proving itself useful when analysing soil in an archaeological context (Linderholm & Geladi 2012), it might also provide a useful tool for differentiating and analysing mortars which has previously only been done in
conservational purposes.
The camera used was the SWIR 3 Spectral Sisuchema which has a spectral range of 900- 2500nm in a 381-pixel array. It was placed over a moving plate that could scan samples of a length of 20cm and a width of 10cm.
4.2 Preparation of mortars for scanning
The material in this study was divided into two groups depending on size and shape. One group consisted of relatively flat and small pieces of mortar that will be referred to as the first run. The other group consisted of irregularly shaped mortars and lager pieces of mortar (both in width and in length), this will be referred to as the second run. This division allowed for better images to be taken with the hyperspectral camera as it allowed for a relatively good focus on all mortars.
Before scanning, all samples of mortar were cleaned with a soft brush, allowing a clearer picture to be taken by the hyperspectral camera. Moreover, by cleaning the mortars from the dust it lessened the probability that dust would fall on the white reference or onto other vital parts of the camera. After each sample of mortar had been scanned the moving plate was cleaned off as well from dust and bits that might have come loose from the mortar, these traces were often collected under the mortar. This cleaning was important as it ensured that no parts from other samples would be mixed up, which also when later cleaning the image from shadows and dust would save some time. It thus created a clean environment and good conditions for scanning the mortars.
The mortar was first scanned on the exposed side i.e. that which faced the outside of the wall and could suffer from weathering and/or lichens. The mortar was then flipped around to scan the inside, capturing the binder matrix and aggregate of the mortars. Because of the variance of the samples it was at times difficult to differentiate or not possible to detect such surfaces
10 on the sampled mortar, these samples where therefore labelled side 1 and side 2 when
scanned.
4.3 Statistics 4.3.1 PCA analysis
Because of the use of hyperspectral imaging this is a multivariate problem it is hard to understand and compare the complex nature of the different mortars but using principal component analysis (PCA) and chemometrics the variables can be visualized in a multivariate space. The different components represented in the model aid in the interpretation of the complex material in a two-dimensional plot using an x- and y-axis; although the model itself considers x, y and z dimensions representing the different components. It can therefore represent several variables (as many variables as there are bands) in a two-dimensional space although their relation between the variables themselves might e.g. exist in a 3-dimensional space if only three components are being used (Shennan 1988).
The PCA models created were based on the average spectra of the mortars to represent each individual object and sample (see 5.1). By doing so the objects would spread in the model creating patterns depending on their chemical and material compositions the HSI captured.
4.3.2 SIMCA method
When classifying the mortars and their binder-matrix based on the NIR spectra in the PCA models the supervised multivariate classification technique (SIMCA) was used (Wold et al.
1983, Branden & Hubert 2004). The SIMCA method observes and divides the samples based on chemometric observations, which in this case is divided into general groups of Roman, Medieval and Renovation mortars. The groups are then tested against each other as well as the uncertain samples in order to classify them; there is also the possibility of samples not fitting into any known group at all. This is likely because of their average spectra is not applicable to any of the known groups, but the SIMCA model has a certain level of misclassification which might be caused by outliers within the known groups that must be acknowledged (Branden &
Hubert 2004).
11 Table 2 The sampled mortars and their composition. The type of mortars are based on ocular observations made
by Claudia Sciuto in MAL (Environmental archaeology laboratory in Umeå).
Lab ref Type of mortar Binder composition Pores Aggregates type Binder color Car16_m1 hydraulic mortar lime 1-3% Mostly pottery/brick and some pebbles 10YR 8/1
Car16_m2 hydraulic mortar lime 1-3% Mostly pottery/bricks 10YR 8/2
Car16_m3 lime? <1% Pebbles silex? 10YR 6/1
Car16_m4 lime?
Car16_m5 Car16_m6 Car16_m7
Car16_m8 lime? 5-10% 10YR 8/1
Car16_m9 lime 10-20% Quartz 7.5YR 8/2
Car16_m10 hydraulic mortar lime 1-3% Pottery with lumps of lime 5YR 8/3
Car16_m11 lime 5-10% Various, river sand? 5YR 8/1
Car16_m12 hydraulic mortar lime Various
Car16_m13 hydraulic mortar lime 1-3% Mostly pottery/bricks white
Car16_m14 lime 1-3% Quartz 10YR 8/1
Car16_m15 lime 1-3% Various, river sand? 10YR 8/1
Car16_m16 lime 1-3% Quartz 10YR 8/1
Car16_m17 lime 5-10% Various, river sand? 10YR 8/1
Car16_m18 hydraulic mortar lime 1-3% Mostly pottery/bricks 10YR 8/2
Car16_m19 lime 5-10% 10YR 8/2
Car16_m20 lime 5-10% Various, river sand? 10YR 8/1
Car16_m21 lime 1-3% Various white
Car16_m22 lime 1-3% Various, river sand? 10YR 8/1
Car16_m23 lime 1-3% Quartz 10YR 8/1
Car16_m24 lime 1-3% Various 10YR 8/1
Car16_m25 lime 1-3% Various 10YR 8/1
Car16_m26 lime 1-3% Various 10YR 8/1
Car16_m27 lime 1-3% Various 10YR 8/1
Car16_m28 lime Various
Car16_m29 lime 1-3% Various 10YR 8/1
Car16_m30 hydraulic mortar lime 1-3% Mostly pottery/bricks 10YR 8/2
Car16_m31 lime 1-3% Various, river sand? 10YR 8/1
Car16_m32 lime 1-3% Various 10YR 8/1
Car16_m33 hydraulic mortar lime 1-3% Mostly pottery/bricks 7.5YR 8/2 - 8/3
Car16_m34 hydraulic mortar lime 1-3% Various 10YR 8/1
Car16_m35 hydraulic mortar lime 1-3% white
12
5. Results
5.1 Image analysis of red Roman pigments
The process and workflow of the image analysis that was carried out on all 35 samples will be explained by analysing the sample Car16_2 (see table 2). This sample is 15 cm long and at its widest point 12 cm and the sample consists of two types of mortar, Roman hydraulic mortar and medieval filling covering the first mortar. The sample also has a flat surface, which has been prepared likely for a wall painting as seen from the red pigmentation that covers it. The red pigmentation could possibly derive from red ochre [Fe2O3] or burnt yellow ochre that was commonly used to produce red paint (Béarat 1996, Hradil et al. 2003).
Before the image analysis began all wavelengths below 1000nm were excluded because they only acted as disturbance to the rest of the spectra (see fig. 3).
Figure 3. Loading line before (above) and after (below) removing wavelengths below 1000nm. It becomes clear that those below 1000nm are solely disturbance gathered from stray light when scanning.
13 The following process was to remove the shadow and background from the image in question, this could often easily be recognized as its own cluster beside what was the mortar (fig. 4).
After excluding this the image needed a second clean-up from stray pixels and shadows which formed under the mortar that were not contained in the main shadow/background cluster. The result is a clear picture with distinct features of both the red pigments and mortars (fig. 5).
Figure 4. Image analysis before cleaning the image from background and shadows. The cluster that needs to be excluded can clearly be seen in the right of the PCA, however one must be careful to not exclude all which seems
to be outliers as these might also contain parts of the aggregate or otherwise.
14 Figure 5. After cleaning up the image from shadows and background the different types of mortars and the red pigmentation becomes very clear in the second component. And as can be seen in the PCA the aggregates and
mortar (B) gather to the left and the red pigmentation (A) forms its own cluster to the right.
When comparing images 4 and 5 in the second component, the two clusters become very clear in both the PCA model and the hyperspectral image. The pigmentation which otherwise did not show at all is now highlighted in a bright red. And the two materials can now easily be separated into two distinct objects where an average spectra then can made to use for a PCA analysis putting all mortars into direct relation with each other and their components. This process was repeated for all 35 samples of mortars.
15
5.2 Problem encountered using HSI
Tools such as NIR and HSI that are based on near infrared optical information are susceptible to scanning stray light and producing strange results if the white reference is dirty. Therefore, when using such machines it is of pivotal importance to keep a clean working environment as well as keeping a constant lighting in the room.
Mortars can be and are relatively dusty as particles fall off the sampled material, which as explained in the methods section caused for a systematic cleaning procedure. However, as often when conducting an experimental study, problems arise in the form of strange readings, which on the spectral image resulted in striping (fig. 6). This result required explanation as to try and explain what happened and how to avoid it.
All samples had to be reviewed to control that the hyperspectral images was of good quality.
After reviewing all the samples in the image analysis software Prediktera evince, it became obvious that only two samples were affected by this distortion. A subset was then selected and scanned again in order to verify the presence of eventual technical problems (compare with fig. 7); however, the new scans on the subset showed the same image noise, therefore further troubleshooting had to be made.
Notes taken throughout the scanning process in combination with the planned method of scanning the mortars were reviewed; this allowed two possible factors to be detected. The first factor is dust on the white reference, which might have been caused by less thorough cleaning during the first sample in each series. The second factor could be the irregular size and shape of the mortars that might have caused for scattering effects when the camera was refocused.
The second factor was the reason why the samples were divided into two groups; because of the need to refocus the camera to accommodate for the difference between small and large mortars. The camera therefore needed to be focused twice. However, as seen from the scans and the rescanned subsamples this was not the cause for the technical problem considering that only the first sample scanned in each set were effected.
The conclusion of this troubleshooting seems to point towards the first factor being the cause for the strange readings, meaning that a more thorough cleaning process must be engaged even from the first sample of each series.
16 Figure 6. Striping effect seen in the second component of the image analysis on Car16_m2. This was the first mortar scanned in the second run, which might have resulted in dust falling on the white reference as the camera
was refocused and not thoroughly cleaned.
17 Figure 7. The mortar sample Car16_28 from the first run and rescanned for quality assurance. The original scan to the left has a slightly more “pointy” scatterplot than that of the rerun as seen on the right. The rerun cannot show the exact same result as the first run, however, both displayed the strikingly similar results without
any sign of a striping effect in the second component.
18
5.3 PCA analysis
Initial observation were made in Prediktera breeze (version 2018.17.0) to ensure good quality images for later image analysis and PCA models made in Prediktera evince (version 2.7.9).
The loading line as seen in figure 5 represents the whole set of samples. Several peaks can be found at 1450nm and 1950nm with an inverted water peak in the third component at 2340nm.
These peaks are distinctive, albeit hard to distinguish between which of the two CH- and OH- groups the peak represents (Socrates 2001); it is likely a mix of both. The first component consists of 90% of the variation in the PCA model and the second component 8,8% (fig. 5).
The representation of the whole dataset was one of the reasons for why these components were used, the other reason was because they provided clear and distinct clusters in the PCA models.
The third component in the loading line for all samples has a peak at 1056nm, which could indicate red colour. Such peaks occur even when observing subsets of the data making it likely that the third component often represents red pigmentation. However, it must be considered that even if it in the whole set is represented in the third component it does not always apply to the individual mortars. This peak representing red pigmentation can also be found represented in the second component (compare fig. 3 with fig. 8).
The PCA models were made using the first and second component, as they not only explained most of the variation in the sampled material but also because they provided clear trends. In figure 9, three clusters emerge, one to the left being the Roman hydraulic mortar which almost groups with the second cluster of early-late antiquity cluster in the centre and a high medieval renovation in the lower right corner of the score plot. These categories are based on the ocular observation of the mortars in addition to observational notes made during the sampling
process (see Table 1 & 2 and Appendix 1). The application of NIR can be used to confirm ocular observations made on the mortars, which in this case seem to be correct as the groups correspond with the assigned observations made prior to the analysis.
The groupings are however not entirely clear, and this might be because of some level of misclassification or limitations in ocular observations but may also be because of outliers. But there are more or less distinct groups which can be made allowing for a closer inspection on those samples which are unclear or seemingly misplaced.
19 Figure 8. Loading line for average spectra of all analysed mortars presented in the PCA models.
Figure 9. PCA model of all sampled mortars. These show a general trend from older (left) to youngest (right), why this is might be because of erosion, weathering or other factors such as material used to produce the mortar. The second component seem to represent aggregate as the different types of mortars groups together.
A general tendency is found in the grouping of the mortars that is based on the type of mortar that is observed (fig. 9). Some exceptions apply as outliers that mainly come from the first Roman phase grouping with the high medieval phases. However, the hydraulic mortars form two distinct groups, one found to the left in the score plot and another closer to the top which is the red pigmentation.
20
5.4 SIMCA
The mortars that are analysed produce a very complex multivariate data often hard to place into one cluster or category. To aid with such a categorization a SIMCA analysis can compare and classify these complex datasets into groups based on how similar and different the known sample groups are to each other (Branden & Hubert 2004). The SIMCA approach to
categorizing the samples in this case provided a somewhat uncertain result. As seen in the SIMCA model that there are differences where three main observations can be made.
The first observation is that the first and second phase cluster together but distinctively belong to their respective half of the plot. The second observation is that the fifth phase is clustered near the centre of the SIMCA classification scheme and should be considered its own cluster albeit close to that of the first two phases. The third observation is about the fourth phase, which are outliers and they do not belonging to any distinctive group (fig. 10). These observations means that the Roman material albeit belonging to two different phases are comparably similar (fig. 11).
Figure 10. SIMCA analysis of the whole dataset. There are outliers from the first and second phase but why they are special is currently outside the reach of this thesis, but hey are distinct in their grouping in comparison to
what is seen in phase four.
21 Figure 11. SIMCA analysis consisting only of phase one and two (Roman period) excluding all the outliers and
other phases to better observe the similarities of the mortars. The two groups are distinct from each other but they belong to the same cluster. See PCA in figure 6 for clustering of phase 1 and 2.
22
6. Discussion
6.1 Application of HSI and NIR
One of the main points for applying this method to study historical mortars is not only because of the methods development in the field of archaeology, but also for testing the application of hyperspectral imaging (HSI) on mortars from an interesting geological and historical context. Acting as a pilot study, this thesis applies and evaluates the application if HSI to discuss its future applications as well as its ability to characterise the materials in the mortars. The latter of which previously only has been available through destructive methods dissolving the mortars (e.g. Casadio et al. 2005). The HSI is limited in this aspect as it likely only can evaluate the sand/lime ratio in comparison to the more precise chemical methods and should be used in conjunction with the conventional methods for the best result. A chemical analysis of the sampled mortars will be done after the completion of this thesis to further the discussion and evaluation of the application of HSI on mortars.
The method itself is relatively easy and quick to apply to seemingly any material as it is versatile and only needs some knowledge in how to operate the camera. Anyone could conduct the first step of the process as long as they are in an evenly lit room. However, as seen in the chapter 5.1 where the process of the image analysis is explained there is some need for understanding the samples themselves as well as how the near infrared spectra and principal components analysis (PCA) works for a proper analysis to be done. An example of the early processing of the HSI is the cleaning of an image from shadows and background. It is not always as clear or simple as shown in figure 4 and 5 considering that some shadows and scattering effect remained on the outline of the mortar. Identifying and removing such parts can be difficult and at times harmful as it could actually be part of the mortar which is being removed if one is not careful; although this is a problem which can be undone with relative ease. This processing using hyperspectral imaging allows for the removal of unnecessary and inadequate data to create more meaningful results, so even with the complexity of the mortars it can be presented in a more comprehensible way (Grahn & Geladi 2007).
Using the hyperspectral imaging for analysing mortars has its advantages and can confirm or contradict what ocular observations had been made in the field and laboratory. This can be seen in both simpler image analysis and PCA score plots but can be much more advanced such as the analysis done on the red pigmentation. Therefore, the tool can be used for both overviews of the dataset as well as going in depth to analyse the different parts if the mortars.
23 The analysis and application undertaken in this thesis is relatively simple and superficial whereas the tool can be used for much more. However, due to time constraints, no chemical analysis or otherwise could be made. This might have been for the better, allowing for a basic approach to the application to be discussed laying ground for future discussions and
development. Nevertheless, it is clear that the application of hyperspectral imaging for the analysis of mortars in terms of classification is likely to provide good results.
6.2 Mortars and their compositions
Regarding the chronology of the mortar, a subject that should be discussed first, one can find a general trend from the oldest to the youngest mortar sample in figure 9 (see p. 19). It is interesting to note that the trend from oldest to the newest mortar sample in the model as represented by the first component is going from left to right. This distinction is not exact but a general indication of something being represented in the first component, which might be erosion (e.g. weathering, oxidation…), colour or choice of binder. The Roman periods are very similar to each other and probably represent similar construction strategies or at least a similar way of producing mortars. Most likely, these two periods collected their raw materials from the same sources, but this is hard to detect considering the complex nature of the
geology around Carcassonne. Similarly, we can find that the high medieval mortar samples form two distinct groups along with some samples from the late antiquity; this might be because of the mortar being used as wall filling and this have been mixed with the samples.
In terms of chronology the mortars seem to correlate somewhat with the first PCA component. Whether this is due to the material composition or colour of the mortar is
uncertain. The scattering along the y-axis which is dictated by the second component seem to be because of the mortars material composition and maybe also to some extent transformation processes. The second component has two distinct CH-groups (1451 & 1956nm) and a less distinct OH-group (1732nm) (Socrates 2001) (fig. 8). The grouping seen in the PCA might therefore be because of the binder and/or the aggregates, which would differ between construction phases because of the masons likely using different sources to gather their raw materials and produce the mortar. This in combination with the first component that
represents the light and absorbance gathered from the mortars may be the reason for the relative chronology.
24 In the PCA score plot in figure 9 three distinct clusters emerge, one to the left being the Roman hydraulic mortar, a mixed Roman cluster in the centre and a medieval restauration in the bottom right. The medieval restauration was for some time believed to be part of the 19th century renovation, however, looking at the different analysis this seemed unlikely and that the mortars based on their material and similarity to the roman mortars in fact were medieval.
In the upper right corner of the score plot two groups of outliers can be found, these are the mosaic floor and Roman red pigmentation. From this two groups can clearly be defined which is a Roman period and a high medieval period. The latter can be subdivided into two groups, one more similar to Roman mortars and the other more modern. However, a more in-depth analysis of the materials used is needed to confirm that notion.
The Roman hydraulic mortar acts as a nearly homogenous group in terms of how they cluster and which phase they belong to, only having a couple outliers in the PCA score plots (fig. 9 &
fig. 14). Two are classified as phase 2 but this is likely a misclassification as no other such occurrences exist and by this time the wall would have changed a function and would
therefore not have been in need of hydraulic mortars (Encyclopædia Britannica 1947 & 1964, Danmarks nationalencyklopedi 1996, Vitruvius De Architectura Dahlgren 2009, Coutelas 2011, Izzo et al. 2015). The hydraulic mortar with the red pigments is most likely its own group because of the colour, as well as having a surface prepared differently; maybe with a slightly different material composition similar to that of the mosaic floor (Izzo et al. 2016).
Because of their different purpose their material composition would be different from other mortars, and this shows well in the score plot. Other mortar groups are likewise dependent on their material composition, but this might be because of their construction phase meaning that they were made by gathering raw materials from different locations during different phases.
As mentioned earlier, the wall during antiquity has two distinct functions, prior to and after the enclosure of Carcassonne, where different mortars are used for different functions. And if compared with other mortars and studies in the area it is likely that the wall which existed prior to the enclosure of the city belonged to a bathhouse.
Because of the red pigmentation seen on the layer of plaster covering the mortar probably forming a layer of a painting there is a possibility of a decorated wall which is comparable with other bathhouses found in Roman Gaul (Béarat 1996, Hradil et al. 2003, Coutelas 2011, Izzo et al.2015). Similarly, a foundation for a mosaic floor which uses slightly different raw materials and is prepared distinctly different from other mortars used for wall building can be seen both in figure 9. Much like the painted wall similar occurrences and comparisons can be
25 found elsewhere in the Roman Empire (Cardoso et al. 2014, Izzo et al. 2015). Much point towards the possibility of a bathhouse existing prior to the enclosure of the city that later was repurposed into the wall which stands today as seen in figure 2.
The cluster in the middle which consists of both the Roman phases represent a break between construction phases, although how far between is unknown. The two according to both the PCA score plots as well as the SIMCA analysis indicates that the materials are very similar to each other meaning that the two probably gathered their natural resources from the same locations and may have followed the same construction techniques (fig. 12-13). However, for the time being this hypothesis is based on vague assumptions and the analysis is too simple to define such notions. A continuation to prove that this is true would be to chemically dissolve the mortars to detect its contents as well as conducting a more advanced image analysis differentiating the sand/lime ratio (e.g. Casadio et al. 2005). Another cause for this grouping might because of a certain level of “repurposing” of the existing Roman structures as
indicating a possible renovation and expansion of the walls, which would then be a sampling bias (Davis 2010, Bork et al. 2011).
Figure 12. Local model of Roman samples from phase 1-2, excluding both hydraulic mortars and the medieval renovation. This much like the SIMCA analysis shows a clear relation between the two periods based on the
similarities of the material or technique used to produce the mortars.
26 Figure 13. Local model of Roman samples from phase 1-2, excluding both hydraulic mortars and the medieval renovation. Looking at component 2 and 3 furthers the notion that the two periods have used similar materials or production techniques, however, the group is not as close as shown above but it has fewer outliers in the
spaced cluster.
Why this repurposing might have been is because there are clear indications of a bathhouse or other public building existing prior to where the wall and a tower now stands (see chapter 3.1) (Cardoso et al. 2014, Izzo et al.2015, Coutelas 2011). The mortars and their composition, which all derive from the first Roman phase, are in general only hydraulic mortar. The change of mortar type comes in relation to the first wall that encloses Carcassonne. Although, without analysing more of the wall this statement only regards a very small segment of it, but a clear pause in construction and change of function occurred between the two phases.
Because of the relative similarities of the Roman mortars it is difficult to detect any definitive chronological continuation, a possibility to further this chronology is to try and date the mortars using 14C-analysis. When carbon dating mortars it is not considering fragments of charcoal or otherwise but is dating the concentration of CO2 that accumulated in the mortars during the hardening process. This effectively means that it is dating the binder as well as the time since the hardening of the mortar (Heinemeier et al. 1997).
Because the binder is the part being dated it is debatable if it is as good approach in
comparison to more conventional carbon dating of organic materials. Studies conducted on the subject mention the many possibilities of contamination during the process of extracting the carbon from the mortars. Most notable one being the contamination of atmospheric carbon
27 contamination, sand aggregates and under burnt lime (lime lumps in sampled mortar), which would cause the sampled to appear older than they actually are. Moreover, the small size of the mortars dictate which dating methods can be used, the most common being mass
spectrometry (Heinemeier et al. 1997, Nawrocka et al. 2005, Lindroos et al. 2007, Hodgins et al. 2011). It is likely as suggested by Langely et al. (2011) that using this method for dating buildings might best be done in combination with other relative and absolute dating methods.
In this case it would not be possible to cross-date the sampled mortars and the method itself seem prone to contamination and is still disputed for how accurate it actually is. Considering the arguments from the above sources a successful outcome seems doubtful. As for now, the PCA score plot provides a relative chronology even without any precise dates. However, considering the similarities between the first and second Roman phase it would be interesting to date the enclosure of the city in order to discuss the development of the Carcassonne.
Figure 14. Local model of all hydraulic mortars samples. They all derive from phase one except for two point., This is probably because of the second phase being used as rubble/wall filling considering that it should not
exist with the earlier phase as well as no other hydraulic mortars belonging to the second phase
28
7. Conclusion
By applying NIR-hyperspectral imaging to study historical mortars it is clear what potential the tool has for studying such material, and that it even on a basic level provides much
information and means of interpreting the mortars. Studying the results from the PCA models one can determine the mortars based on their type and construction phase as indicated by the different aggregates and binder used as well as technique for producing the mortars
themselves. This is most notable when observing what seems to be a Roman bathhouse repurposed into a defensive wall. Therefore, has the NIR-hyperspectral imaging showed itself as a helpful tool for furthering the analysis of ancient structures and construction techniques using non-destructive means. This study is, however, limited in its scope, but it acts as a platform for developing the procedure of applying HSI to study historical mortars as well as discussing the methods application. Moreover, with this study it is possible to see the many prospects this tool might have to further the interpretation of ancient structures and
construction techniques.
To further the discussion and evaluation of the application of this tool a comparable analysis should be conducted using conventional chemical methods to dissolve the mortars, comparing those results to that of the image analysis. A comparable study could thereby conclude if the tool can provide accurate lime/sand ratios as well as the possibilities and limitations of the HSI image analysis and classification. After doing so the tool can be evaluated as to how well it can study historical-archaeological mortars. Although further application of the tool on a larger material will demand a more rigid protocol to be formulated; the comparable study is needed so that the limitations of the tool can be identified and a method of application be formulated accordingly.
29
Reference list
Allios, D., Guemeur, N., Cocoual, A., Linderholm, J., Sciuto, C., Geladi, P., Gobrecht, A., Bendoula, R., Moura, D., Jay, S. and Gardel, M.-E. 2016. Near infrared spectra and hyperspectral imaging of medieval fortress walls in Carcassonne: a comprehensive interdisciplinary field study. NIR News Vol. 27, No. 3, pp. 16-20
Béarat, H. 1996. Chemical and mineralogical analyses of Gallo-Roman wall painting from Dietikon, Switzerland. Archaeometry Vol . 38, No. 1, pp. 81-95
Bork, R.O., Clark, W. and McGehee, A. 2011. New approaches to medieval architecture.
Ashgate Publishing Ltd., Farnham
Branden, K.V. and Hubert, M. 2004. Robust classification of high dimensional data.
Proceedings in Computational Statistics: 16th Symposium Held in Prague, Czech Republic.
1926-1932. https://wis.kuleuven.be/stat/robust/papers/2004/rsimcaCOMPSTAT2004.pdf (Collected 2018-04-17).
Carcassonne, 1947 and 1964. Encyclopædia Britannica Volume 4. University of Chicago, William Benton
Carcassone, 1996. Den Store Danske Encyclopædi Bind 4. Nordisk Bogproduktion A/S, Haslev
Cardoso, I., Macedo, M.F., Vermeulen, F., Corsi, C., Santos Silva, A. and Mirao, J 2014. A multidisciplinary approach to the study of archaeological mortars from the town of Ammia in the Roman province of Lusitania (Portugal). Archeometry Vol 56, No. 1, pp. 1-24
Casadio, F., Chiari, G. and Simon S. 2005. Evaluation of binder/aggregate ratios in archaeological lime mortars with carbonate aggregate: a comparative assessment of chemical, mechanical and microscopic approaches. Archaeometry Vol 47, No. 4, pp. 671–
689
Coldstream, N. 2002. Medieval Architecture. Oxford university press
Coutelas, A. 2011. The selection and use of lime mortars on the building sites of Roman Gaul.
Ringbom, Å., Hohlfelder, R.L., Sjöberg, P. and Sonck-Koota, P (Editors), Building Roma Aeterna – current research on roman mortar and concrete proceedings of the conference march 27-29 2008, Societas Scientiarum Fennica, Ekenäs tryckeri AB
Dalgren, B. 2009. Vitruvius – Om arkitektur tio böcker. Dymlings, Stockholm
Davis, R.H.C. 1988. A history of medieval Europe – From Constantine to Saint louis.
Longham, London
D’Ambrosio, E., Marra, F., Cavallo, A., Gaeta, M. and Venture, G. 2015. Provenance materials for Vitruvius'harenae fossiciae and pulvis puteolanis: Geochemical signature and historical–archaeological implications. Journal of Archaeological Science, reports 2, pp. 186- 203
30 Fischer, C. and Kakoulli, I. 2006. Multispectral and hyperspectral imaging technologies in conservation: current research and potential applications. Studies in Conservation Vol. 51, No 7, Supplement 1, pp. 3-16
Friberg, G. and Sundér, B. 1996. Stenen i tiden – från 1000-talet till 1940.
Riksantikvarieämbetet. Wallin & Dahlholm AB, Lund
Heinemeier, J., Junger, H., Lindroos, A., Ringbom, Å., Von Konow, T. and Rud, N. 1997.
AMS 14C dating of lime mortar. Nucelar Instruments and Methods in Physics Research B Vol.
123, pp. 487-495
Hobbs, L.W. and Siddall, R. 2011. Cementitious materials of the ancient world. Ringbom, Å., Hohlfelder, R.L., Sjöberg, P. and Sonck-Koota, P (Editors), Building Roma Aeterna – current research on Roman mortar and concrete proceedings of the conference march 27-29 2008, Societas Scientiarum Fennica, Ekenäs tryckeri AB
Hodgins, G., Lindroos, A., Ringbom, Å., Heinemeier, J. and Brock. F. 2011. 14C dating of Roman mortars – preliminary tests using diluted hydrochloric acid injected in batches.
Ringbom, Å., Hohlfelder, R.L., Sjöberg, P. and Sonck-Koota, P (Editors), Building Roma Aeterna – current research on Roman mortar and concrete proceedings of the conference march 27-29 2008, Societas Scientiarum Fennica, Ekenäs tryckeri AB
Hradil, D., Grygar, T., Hradilová, J and Bezdicka, P. 2003. Clay and iron oxide pigments in the history of painting. Applied Clay Science 22, pp. 223-236
Hägg, G. 1989. Allmän och oorganisk kemi. Vannerberg, N.-G. (Rev.), Ninth edition, Almqvist & Wiksell, Uppsala
Izzo, F., Arizzi, A., Cappellettim P., Cultrone, G., De Bonis, A., Germinario, C., Graziano, S.F., Grifa, C., Guarino, V., Mercurio, M., Morra, V. and Langella, A. 2016. The art of building in the Roman period (89 B.B-79 A.D.): Mortars, plasters and mosaic floors from ancient Stabiae (Naples, Italy). Construction and building materials 117, pp. 129-143 Johnson, M. 2010. Archaeological theory – an introduction. Wiley-Blackwell, Chichester Langely, M.M., Maloney, S.J., Ringbom, Å., Heinemier, J. and Lindroos, A. 2011. A comparison of dating techniques at Torre de Palma Portugal: Mortars and Ceramics.
Ringbom, Å., Hohlfelder, R.L., Sjöberg, P. and Sonck-Koota, P (Editors), Building Roma Aeterna – current research on Roman mortar and concrete proceedings of the conference march 27-29 2008, Societas Scientiarum Fennica, Ekenäs tryckeri AB
Linderholm, J. and Geladi P. 2014. Portable near infrared spectroscopy and imaging on archaeological sites and samples in the MOBIMA project. part 1. field spectroscopy. NIR news Vol 25, No. 5, pp. 11-15
Lindroos, A., Heinemeier, J., Ringbom, Å., Braskén, M. and Sveinbjörnsdóttir, Á. 2007.
Mortar dating using AMS 14C and sequential dissolution: examples from medieval, non- hydraulic lime mortars from the Åland islands, SW Finland. Radiocarbon Vol. 49, No. 1, pp.
47-67
