Introduction

Clay is one of the oldest waterproofing materials used in construction. Archaeological evidence around the world suggests that unfired clay was being used to seal cisterns, reservoirs, and canals at least as early as the Neolithic (Mays et al. 2013; Kirke 1980). Vernacular usage often suggests a good understanding of different locally available resources existed. In semi-arid climates, buildings were traditionally protected from precipitation by sealing flat roofs with compacted clay (Minke 2012; Warren 1998). This included floors made from Huwwar in the Levant and Arga in Tibet (Rollefson 1996, p. 223; Feiglstorfer 2020). In the German state of Saxony, there is evidence of buildings dating back to the sixth century having clay barriers laid under the foundations or packed into stone casing above the plinth (Ziegert 2003). The use of clay to protect buildings declined with the advent of industrially produced materials, such as bitumen.

Use remained common in hydraulic engineering and landfill construction to seal against pressing water or chemical leachates. Developments in these fields led to the emergence of bentonite panels and specialised clay barrier mixtures in the second half of the twentieth century, the use of which has in turn migrated back across to architectural engineering. These products depend on the properties of highly swelling sodium bentonite clay to peptise when wet and minimise hydraulic conductivity through the layer (Michette et al. 2017). Despite their unprocessed raw materials, these products can have a high embodied energy due to the geographically limited availability of sodium bentonite. A small number of conservation projects in parts of Germany have made use of naturally occurring, locally available, waterproof clay to produce barriers. As well as being reversible, this method, therefore, has a very low embodied energy.

Clay waterproofing techniques are not covered in building codes. With the exception of bentonite panels, they are rarely addressed in practical guidelines, technical literature, or scientific research on earth building. Suitability tests for different specialised mixtures and naturally occurring clays used in landfill engineering are widely available. On the basis of these, Michette et al. (2017) established some general principles for the requirements of suitable clay barrier material for use in architectural conservation.

  • A grain size distribution that forms a low porosity when compacted.

  • The presence of swelling clays to further restrict porosity and bind moisture during wetting.

  • The pressure and volumetric deformation caused by swelling must, however, not result in shrinkage cracks during drying.

  • To reduce the risk of shrinkage, the optimum moisture content to achieve Proctor density should be close to or below the shrinkage limit.

  • The hydraulic conductivity, at close to Proctor density, should be no greater than 10–10 m/s.

  • Besides reversibility, installed material needs to demonstrate a chemical stability and compatibility with historic fabric.

The geographic range of historical precedents suggests that suitable material may be relatively common. Material used in the various vernacular techniques is characteristically calcareous, either in its raw state or through additives. This will reduce shrinkage and improve density if calcite forms pore-clogging cement. There is also research on the use of naturally occurring resources for landfill engineering (e.g., Met and Akgun 2015). Volcanic material has been favourably assessed (Navia et al. 2005). There is no guidance on establishing the general suitability of material for use in architectural conservation or construction. Many places in which historical use is unknown may have suitable, locally available resources, which could be exploited for conservation or newbuild projects.

The objective of this paper is to establish a protocol for assessing locally available geo-materials for use in clay barriers to protect historic masonry. It presents a series of laboratory tests based on standards used in related fields of geo-engineering, such as landfill engineering. It also introduces a series of rapid-tests used to assess material for earthen construction, which are suited for field laboratories. It integrates these methodologies to enable the selection and modification of suitable barrier material. An ongoing conservation research project in Pompeii, Italy provides a useful case study for investigating the viability of a range of different geological deposits and the installation of a barrier in heritage settings (see also: Michette et al. 2018; Breuninger 2018; Breuninger et al. 2019). The aim of this research is to facilitate the wider use of clay barriers, and explore the use of hyper-local materials in general, for conservation and construction projects.

Site and setting

Re-exposed archaeological fabric is often at a particularly high risk of rapid decay (Wilhelm et al. 2017). The Via Nucerina necropolis in Pompeii has been a focal point of work carried out within the Pompeii Sustainable Preservation Project (PSPP). The PSPP is an international collaboration co-ordinated by the Fraunhofer Institute for Building Physics which aims to develop case-specific, evidence-based conservation strategies for the archaeological site of Pompeii. The Via Nucerina consists of approximately 10 individual tombs, which were fully excavated in the early 1980s. Their plaster surfaces have since severely deteriorated (Fig. 1). Recent surveys established that decay of the plaster and masonry is ongoing and linked to ground moisture intrusion (Michette et al. 2018). The installation of a waterproof clay barrier was proposed for the foundation of Tomb D–N, which had been selected for long-term environmental monitoring and initial implementation of conservation strategies, such as edge repairs. The barrier will provide an opportunity to determine the suitability of clay barriers in the context of valuable archaeological heritage, develop appropriate installation techniques, and monitor their effect on historic fabric in-situ. Provided suitable material can be found, the barrier is to be made from locally available resources.

Fig. 1
figure 1

South side of Tomb D–N, Via Nucerina, Pompeii during an excavation in 2017. Damage to the plaster surfaces suggests rising damp following ground moisture intrusion as an active decay mechanism

Full size image

Geologically, the region around Pompeii is defined by volcanic material from Vesuvius and Mesozoic sedimentary rocks (Iannace et al. 2009). The Monti Lattari to the south of Pompeii and the Neapolitan Apennines to the east consist of limestone, dolomitic limestone, and marlstone. These protrude through the more recent volcanic deposits. In valleys and basins, the carbonate base rock is layered with clays and pyroclastic material. This mixture of carbonates, clays, and pozzolans gave rise to important ceramic and cement industries in the Roman era (Artioli et al. 2019; Brandon et al. 2005). The manufacturing industries persist to some extent, although the quarrying industry has drastically declined, and the majority of raw materials are imported. Nevertheless, the geological diversity of the region provides a good opportunity to incorporate different materials into the development of assessment protocols for suitable clay barrier material.

Methodology

Materials

A total of ten different geo-materials were collected from within an approx. 30 km radius of Pompeii (Fig. 2; Table 1). The materials were selected with the help of a local expert, on the basis of availability and to reflect the diversity of local resources. Between 10 and 50 kg of each material was collected for transport to the laboratory of the Technical University in Munich, Germany. This included three clayey soils, selected as potential barrier material or base material for a mixture. Three pozzolanic sands, a slaked lime, a rock meal (a by-product of stone cutting), and excavation material from ancient and modern Pompeii were selected as potential additives to a mixture. A naturally occurring, commercially available waterproof clay from Saxony, Germany, previously analysed by Michette et al. (2017), was also tested to provide a comparison.

Fig. 2
figure 2

Geological map of the area around Pompeii showing the location of sampling sites in this study (after Breuninger 2018)

Full size image
Table 1 Purpose and availability of tested samples
Full size table

Overview of methods

The materials were tested according to steps 1–3 in the pipeline shown in Fig. 3 (step 4 is the topic of ongoing research). Grain size distribution and Atterberg limits were determined to enable a general classification of soil type. Alongside an initial visual assessment of composition and considerations on availability, this enabled a selection of material deemed suitable for further testing. X-ray diffraction (XRD) was performed on powdered samples of the most suitable materials to determine mineral composition. Any selected base material was tested for proctor density, hydraulic conductivity, swelling deformation, and shrinkage limits to assess its suitability as barrier material based on protocols used in other fields of engineering (Gartung et al. 1993). Mixtures were then produced, using selected additives, to improve the properties of the base material. The mixtures were re-tested. To reduce the number of laboratory experiments and minimise material expenditure, two rapid tests were performed on all mixtures. The general classification, mineral composition, and barrier suitability tests were then performed on the mixtures deemed most suitable. The commercially available waterproof clay was used for orientation when assessing suitability.

Fig. 3
figure 3

Experimental pipeline for assessing suitability of raw materials and mixtures as clay barrier material

Full size image

General classification

DIN 18196 (2011) enables a classification of soil types for civil engineering and construction purposes according to grain fraction, plasticity, and organic content. This is useful both for identifying potentially suitable barrier material (e.g., no organic content) and for categorising clay barrier material within existing construction standards. The grain size distribution and Atterberg limits are required for this classification.

The grain size distribution was determined on all suitable material according to DIN 17892-4 2017. All grains < 0.063 mm were separated by wet sieving and the coarse fraction was then dry sieved. The dry sieving was done by hand using a set of 0.063 mm, 0.2 mm, 0.63 mm, 2.0 mm and 63 mm sieves. The fine grain fraction (silt and clay, < 0.063 mm) was analysed using the sedimentation hydrometer method. Fresh material was used for the sedimentation and 25 ml sodium pyrophosphate was added to the slurry to prevent coagulation. The grain size distributions of mixtures prepared from several raw materials were calculated from the grain size distributions of the individual components.

The liquid limit was determined on material with a silt and clay content of more than 40% according to DIN 18122 (1998) using the Casagrande method. 30–50 g of soil was softened into a paste and placed in the bowl of the Casagrande device. A groove was cut through the sample using the standard tool. The bowl was repeatedly lifted to a defined height and dropped. The experiment ended when the groove had closed to a length of 1 cm. The number of drops was recorded and the moisture content of the soil was determined. The experiment was performed 4 times with different moisture content. The water content associated with 25 drops, calculated by regression from the 4 tests, is the liquid limit.

To determine the plastic limit, 5–10 g samples were wetted to a kneadable putty and rolled out by hand on an absorbent surface into rolls of approximately 3 mm thickness. The sample was then kneaded together again. Rolling out and kneading were repeated until the rolls began to crumble. The moisture content was determined on the crumbling rolls. This process was repeated on 3 samples and the plastic limit determined from the mean moisture content.

X-ray diffraction

XRD was performed on all suitable raw materials and mixtures to analyse composition and to identify potential chemical reactions occurring in the mixtures. Powder and texture samples were prepared. Powder sample preparation can lead to strong preferential orientation, especially of clay minerals, which can lead to large errors in XRD measurement. Analysing the difference between powder samples and texture samples can provide a good indication of mineralogical composition.

The powder samples were made from 2 g of homogenised material, which was ground to a fine powder in a micron mill and pressed into tablets. For texture samples, 3 g of material was soaked in distilled water, shaken, and left for 16 h. A drop of the leftover suspension was centrifuged onto a glass slide and left to dry. This process improves the identification of individual minerals in a sample. A third sample was produced from material with suspected smectite content by treatment with ethylene glycol, to aid the distinction of smectite and chlorite.

The samples were analysed in the XRD device at the Technical University of Munich (Phillips PW 1800 Cu Kα), diffractograms were generated using the program X'Pert HighscorePlus, these were evaluated using the BGNMwin software and existing mineral databases to identify and semi-quantitively determine distributions of individual minerals (Moore and Reynolds 1997).

Proctor density

The Proctor density describes the highest possible dry density of a compacted, loose grained material. To achieve Proctor density, a material needs to be compressed at the optimum moisture content. The Proctor density and the optimum moisture content were determined on all suitable base material and mixtures according to DIN 18127 (2012). Loose material was compressed into a steel cylinder (dimensions) in three layers, using a weight (mass 2.5 kg) dropped from a height of 30 cm 25 times per layer. The test was repeated multiple times at different moisture contents, starting just below the plastic limit and increased in increments of 2–3%. The resulting dry densities of the compressed material were calculated and the relationship of dry density to moisture content was plotted to establish the compaction curve. The Proctor density is obtained from the peak of the compaction curve.

Hydraulic conductivity

The water permeability coefficient kf was determined in accordance with DIN 18130-1 (1998). All samples were tested with variable hydraulic gradient using a standpipe device. For some samples, additional pressurization of up to 1 kbar was used because of their low grain size and, therefore, expected low conductivity.

The samples were installed with proctor density and the optimum water content. The test fluid (tap water) had a temperature of 24 °C; therefore, the kf value was converted to the reference temperature (10 °C). Before recording, the samples were flowed through for at least 1 h.

Swelling deformation

The swelling deformation was determined using the powder swelling test in accordance with Recommendation No. 11 of the DGED Working Group 19—Experimental Rock Technology (DGEG 1986) according to Thuro (1993).

Samples measured for swelling deformation were compressed into 7.2 cm diameter, 2 cm high steel rings at proctor density and placed in an Oedometer. A load of approximately 0.5 kN/m2 was placed on the saturated samples and volumetric expansion was recorded on a dial gauge. The total swelling deformation was determined when no further volumetric expansion occurred.

Shrinkage limit

The shrinkage limit was determined according to DIN 18122-2 (1997). Dry density and dry volume were determined by dip weighing after coating with paraffin.

At the shrinkage limit, the material changes from semi-solid to solid state. Thus, from this limit on, no volume change is to be expected with further drying. According to Richwien & Lesny (2000), the distance between the plastic and shrinkage limits is an indicator of the susceptibility of the material to settlement. The greater the difference between the plastic and shrinkage limits, the greater the settlement.

Rapid tests

The rapid tests were based on the ball dropping test and the swelling–shrinking test described in Minke (2000). These tests were chosen to bring assessment protocols in line with common testing protocols used in earth construction and inform the development of appropriate field-testing protocols.

Ball dropping tests were performed on balls of approximately 4 cm diameter formed from material close to plastic limit and dropped from height of 1.5 m. Material which deformed significantly or collapsed upon impact was not considered for ongoing tests.

Swelling–shrinkage tests were performed on discs of approximately 7 cm diameter and 2 cm height formed from material close to liquid limit and oven dried at 40 °C. Material which displayed volumetric shrinkage or crack formation was not considered for ongoing tests.

Results and discussion

Overview of results

The results are presented according to the pipeline shown in Fig. 3, beginning with the general assessment of all materials, the identification of potential barrier material and additives, and the initial barrier assessment of any materials deemed potentially suitable in their raw state. Following this are the results of the rapid tests performed on a wide range of mixtures to narrow down potentially suitable barrier mixtures. Finally, the second round of barrier assessment tests and the mineralogical analysis of the potentially suitable mixtures are presented. For better comparison, figures and tables showing results of the first (i.e., raw state) barrier assessments and second (i.e., mixture) barrier assessments are combined.

Initial assessment

The first round of tests eliminated the possibility of using the clayey soils in their raw state (Table 2). Despite having a very similar grain size distribution to the Saxonian clay in one case, both clays from the Monti Lattari region had a significant amount of organic matter, making them unsuitable for use as barrier material. After the subsequent barrier assessment, the clay from the Salerno basin had a suitable hydraulic conductivity (Table 4), but swelling deformation was high due to the high clay content. Shrinkage cracks could form during drying. Four possible additives were selected after the first round of tests on the basis of suitable grain size distributions, variation of expected reaction with the selected base material, and availability.

Table 2 Results of classification according to DIN 18196 (by grain size distribution and Atterberg limits) and suitability of sampled materials for barrier
Full size table

Rapid tests

The four additives were combined with the Salerno basin clay in different proportions to produce 18 mixtures (Table 3). The rapid tests were used to narrow this down to a selection of four, which were deemed suitable for further barrier suitability testing (Table 4). Figure 4 shows the ball dropping test. The samples with a high content of the highly plastic Salerno basin clay (samples on the left side) tend to deform, while the samples with the lowest content of the plastic material (right side) are more likely to collapse at impact. The swelling–shrinkage test (Fig. 5, Table 3) showed an increasing amount of shrinking with higher content of Salerno basin clay.

Table 3 Volumetric shrinkage of mixture samples following swelling–shrinkage test
Full size table
Table 4 Comparison of key variables for barrier suitability across the raw Salerno clay, the selected mixture, and the Saxonian waterproof clay used for orientation
Full size table
Fig. 4
figure 4

Results of ball dropping test using different mixing proportions of Salerno Clay and (top row) excavation soil 2; (second row) black pozzolanic sand; (third row) rock meal; (fourth row) slaked lime; (bottom row) slaked lime and black pozzolanic sand. (Image corrected to enhance contrast and sharpen cracks) (after Breuninger 2018)

Full size image
Fig. 5
figure 5

Results of swelling–shrinkage test using different proportions of Salerno Clay and (top row) excavation material; (2nd row) black pozzolanic sand; (3rd row) rock meal; (4th row) slaked lime; (bottom row) slaked lime and black pozzolanic sand (after Breuninger 2018)

Full size image

The mixtures created with the excavation soil performed variably, with the 90% clay/10% soil mixture deforming heavily following ball drop and shrinking by more than 10% in volume, the 80/20 mixture showing deep cracks following the shrinkage test, and the 60/40 mixture collapsing following the ball dropping test. The 70/30 mixture was deemed to have passed both tests. In the black pozzolanic sand mixtures, the 90% clay/10% pozzolana and 80/20 mixtures deformed heavily following ball drop. Although a small piece of the 60/40 mixture broke off, the ball did not collapse and the sample was deemed to have passed. All mixtures performed well in the shrinkage test, with only 90/10 shrinking by more than 10% in volume and no visible cracks forming, so the 70/30 and 60/40 mixtures were selected for further testing. In the rock meal mixtures, the 90% clay/10% rock meal and 80/20 mixtures deformed heavily, and the 60/40 mixture collapsed entirely following ball drop. Although shrinkage was limited, fine cracks were visible in the 70/30 and 60/40 mixtures. None of the mixtures were deemed suitable. During the oven drying of the slaked lime mixtures, salt crystals formed on the surface. Mixtures using the slaked lime were deemed inappropriate for barrier material in contact with the historic masonry because of potential salt contamination; however, the most suitable mixture using slaked lime and pozzolanic sand was further investigated in this study to explore the potential effect of a chemical reaction between the two components on barrier performance.

Barrier assessment

The four mixtures selected for ongoing assessment were tested for grain size distribution (Fig. 6) and Atterberg limits (Fig. 7). While still noticeably different in composition, the mixture of 60% Salerno clay and 40% black pozzolanic sand is closest in character to the Saxonian clay. The barrier assessment tests further indicated that the 60/40 mix is the most suitable barrier material (Table 4). This mixture performed well according to all the tested criteria. Only two samples have an acceptable hydraulic conductivity (60% clay/40% pozzolana and 70% clay/30% pozzolana). All other samples are too permeable to act as a barrier. Out of those two samples, the 60/40 mix has an optimum moisture content closer to the shrinkage limit, the lower deformation potential, the lower hydraulic conductivity, and is, therefore, more suitable.

Fig. 6
figure 6

Grain size distribution of clay barrier material from Saxony, alongside that of clay from Salerno basin in raw state and in four mixtures selected for barrier assessment

Full size image
Fig. 7
figure 7

Plasticity Index (Ip), liquid limit (wL), and classification according to DIN 18196 of clay barrier material from Saxony, and clay from Salerno basin in raw state and in four mixtures selected for barrier assessment

Full size image

Mineralogical composition

XRD analysis was performed on the raw Salerno basin clay, the four mixtures subjected to the barrier assessment, the individual components used in the mixtures, and the Saxonian barrier clay for comparison. The results of the XRD analysis are shown in Table 5. The analysis identified smectite and calcite in the Salerno basin clay, the presence of which was also confirmed in the Saxonian clay.

Table 5 XRD results of materials selected for barrier assessment in raw state and in mixtures
Full size table

Discrepancies were noted when comparing the composition of the mixtures quantified by XRD analysis with the expected composition calculated from the constituent raw materials. In most cases, these discrepancies are likely to be due to the small sample sizes and the semi-quantitative nature of the evaluation software. The small sample sizes could explain why olivine is not identified in the mixtures, despite being present in small quantities in the black pozzolana and the excavation soil. It can also explain discrepancies in minerals which are present as larger crystals, such as augite. Misrepresentations by the evaluation software when analysing the diffractogrammes can explain the large discrepancies in illite/muscovite and smectite contents. These clay minerals have a similar crystal structure and produce similar impulses that complicate their precise identification. Comparing the overall clay mineral content of the analysed mixtures with the expected result based on the individual components produces much closer results.

Some differences between the XRD analysis of mixtures and that of the individual components may also indicate chemical reactions have taken place, for example, in the mixture 60% clay/20% black pozzolana/20% slaked lime. Calcium oxide identified in the powder sample of the slaked lime was not identified in the mixture, while kaolinite was identified in the mixture despite not being present in the individual constituents. This could be due to small sample sizes, alternatively it could indicate a reaction between the slaked lime and the clay minerals had taken place. The mixture did not improve hydraulic conductivity but did improve volumetric stability (Table 4). Ongoing work should include further physico-chemical analysis to investigate potential cementing reactions.

Discussion

Finding suitable base material was not as easy as initially hoped. Access to potential additives was also restricted due to the declining quarrying industry. This affected the extent to which this case study could inform general requirements and establish a breadth of suitable material characteristics. The suitable mixture shows similarities to the commercially available waterproof clay, most significantly the presence of smectite and calcite and a good volumetric stability. Although the grain size distribution is the most closely matched of the tested mixtures, it does have a notably higher proportion of fine grains than the Saxonian clay. The waterproofing mechanism seems to occur from the raw state of the ingredients in the mixture. The effect of adding pozzolanic sand was to improve the grain size distribution, rather than cause a cementing reaction. While the selected material will need to be tested long term in-situ to establish full suitability, these findings indicate that the general requirements of clay barriers could conform to a wide range of commonly available resources.

The most suitable material fits the model prescribed by historic precedents of vernacular clay waterproofing material. Those which have been studied in detail, such as Arga (Feiglstorfer 2020), and the commercially available Saxonian clay investigated here (Michette et al. 2017), which is likely to be similar to historic material used in the region (Ziegert 2003), also contain proportions of smectite, calcite and sand which achieve a good volumetric stability when compacted. Other traditional techniques which have fallen out of use, such as Huwwar, a mixture of chalk and clay used in the Levant as early as 5000 BCE to form waterproof layers (Rollefson 1996, p. 223), or the flat earth-roofs of semi-arid regions in eastern and central Asia, have been less well-studied. The associated built heritage of these regions is deteriorating (Jäger 2012). While historic buildings are also at risk in Tibet, the ‘living’ tradition of Arga plays an important role in cultural heritage conservation. Given what we do know, it is likely that the ‘lost’ materials of other flat earth-roof techniques depend on a similar mineralogical composition of smectite and calcite in a robust grain matrix. The testing protocol presented here can be used to help evaluate clay waterproofing materials in these and other regions for use in conservation projects. In the context of the PSPP, and archaeological and architectural conservation more generally, rediscovering techniques which make innovative use of locally available materials can create new opportunities for engaging with and protecting built heritage. Subject to further tests and approvals, a trial barrier using the most suitable mixture will be installed at Tomb D–N on the Via Nucerina during the PSPP Summer school in 2023.

Conclusion

This paper describes the methodology for selecting local raw materials suitable for a clay barrier at the base of a funerary monument in the Porta Nocera area of Pompeii. A mixture consisting of 60% calcareous, clayey soil and 40% pozzolanic sand was selected from a variety of local geo-resources. The selection process included rapid tests based on the ball dropping test and the swelling–shrinkage test, and laboratory tests commonly used in the assessment of barrier material for landfill engineering. The materials were also analysed using XRD. The selected mixture has a hydraulic conductivity and volume stability that is comparable to other clay barrier material used in architectural conservation. More broadly, it has the characteristics of calcareous, clayey material traditionally used in vernacular waterproofing techniques. This study supports work elsewhere on techniques, such as Arga, and establishes commonalities in the mineralogical composition and mechanical function of clay waterproofing materials. The described methodology provides a starting point for investigating material in other regions. The application of the standardised experiments and simple field tests presented here will enable the cataloguing and comparison of suitable material. This will identify general requirements and inform more robust protocols tailored to the specificities of clay for mitigating ground moisture intrusion. This will help develop clay barriers as a sustainable option for protecting buildings from ground moisture, and can facilitate the reactivation of hyper-local, low energy material industries.