Cement and Concrete Research 35 (2005) 1555 – 1565

Lime pozzolana mortars in Roman catacombs: composition,

structures and restoration
Sergio Sánchez-Morala,*, Luis Luquea, Juan-Carlos Cañaverasb, Vicente Solerc,
Javier Garcia-Guineaa, Alfredo Aparicioa
a
Museo Nacional de Ciencias Naturales (CSIC), C/ José Gutiérrez Abascal 2, E-28006 Madrid, Spain
b
Universidad de Alicante, Departamento de Ciencias Tierra and Medio Ambiente, E-03080 Alicante, Spain
c
Instituto de Productos Naturales y Agrobiologia de Canarias, Avda, Astrofisico Francisco Sanchez 3, E-38206 La Laguna, Tenerife, Spain

Received 21 April 2003; accepted 19 August 2004

Abstract

Analyses of microsamples collected from Roman catacombs and samples of lime–pozzolana mortars hardened in the laboratory display
higher contents in carbonated binder than other subaerial Roman monuments. The measured environmental data inside the Saint Callistus and
Domitilla catacombs show a constant temperature of 15–17 8C, a high CO2 content (1700 to 3500 ppm) and a relative humidity close to
100%. These conditions and particularly the high CO2 concentration speed-up the lime calcitization roughly by 500% and reduce the cationic
diffusion to form hydrous calcium aluminosilicates. The structure of Roman catacomb mortars shows (i) coarser aggregates and thicker beds
on the inside, (ii) thin, smoothed, light and fine-grained external surfaces with low content of aggregates and (iii) paintings and frescoes on
the outside. The observed high porosity of the mortars can be attributed to cracking after drying linked with the high binder content.
Hardened lime lumps inside the binder denote low water/mortar ratios for slaking. The aggregate tephra pyroclasts rich in aluminosilicate
phases with accessorial amounts of Ba, Sr, Rb, Cu and Pb were analysed through X-ray diffraction (XRD), electron microprobe analysis
(EMPA) and also by environmental scanning electron microscopy (ESEM) to identify the size and distribution of porosity. Results support
procedures using local materials, special mortars and classic techniques for restoration purposes in hypogeal backgrounds.
D 2004 Elsevier Ltd. All rights reserved.

Keywords: CaCO3; Ca(OH)2; Mortar; Aggregate; Pozzolan; Microcracking; Crystal size; Curing

1. Introduction rock dated to the middle Pleistocene originating from the

Colli Albani volcanic district. Corridors are often partially
Lime–pozzolana mortars cover large surfaces of Roman covered with bricks making up the structural arches, walls
catacomb walls and are usually painted with frescoes (Fig. and domes. Many cubicles containing family or communal
1a). In the eastern part of Rome, outside the ancient walls, tombs were dug as squared dome shapes and covered with
Saint Callistus and Domitilla are two of the oldest mortar and stuccoes (Fig. 1b). Mortar prevents failure of the
catacombs excavated in volcanic tuff. Saint Callistus is soft substrate and provides a clean, plain, light surface to be
famous for the popes’ tombs of the second century AD. decorated with religious paintings and symbols. The
These important archaeological–architectural heritages catacombs were built from the second to the eighth century
attract thousands of tourists. Many wall surfaces of Saint AD, and the main part of the decorated mortar originates
Callistus and Domitilla catacombs are made of volcanic from the first stages of the catacomb use.
The excellent properties of lime–pozzolana mortars, such
as high strength, insolubility and hardening even under
* Corresponding author. water, were known to the ancient Roman builders since the
E-mail address: [email protected] (S. Sánchez-Moral). third century BC. Significant Roman monuments, such as
0008-8846/$ - see front matter D 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cemconres.2004.08.009
1556 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565

Fig. 1. (a) Aspect of walls in a cubicle covered by mortar and painted frescoes in the Saint Callistus catacomb (Rome). (b) Section showing the volcanic
substrate overlaid by thin layers of mortar and stucco. (c) Section of individual sample showing coarse and fine aggregate inside the mortar beds and pure
calcite in the mortar substrate interface. (d) Thin section of both the volcanic substrate and the mortar with two thin calcite layers overcoating the
substrate.

the Pantheon, Roman Baths or Coliseum in Rome, were studying microsamples collected from the wall catacombs
built using Roman concrete called bOpus CaementiciumQ in by optical microscopy, environmental scanning electron
the first century AD [1]. The structural and physical– microscopy (ESEM), electron microprobe (EM), X-ray
chemical characterization of the lime–pozzolana mortars diffraction (XRD), ICP-AES, atomic absorption spectro-
used in building and the determination of the prevailing scopy (AAS), porosimetry, etc.; (ii) performing original
microenvironmental conditions could provide enough infor- experiments of lime–pozzolana hardening under X-ray
mation to formulate new specific mortars for restoration diffraction with thermal control; and (iii) measuring the
purposes, using original neighbouring raw materials. The environmental data inside catacombs (RH values, CO2
region of Rome is volcanic in nature, and the hills of the concentration, temperature, pH and chemical composition
City consist of a relatively soft and light volcanic material of infiltration waters).
known as tuff. In different places, it has slightly different
textures and colours (grey, brown, and black are typical), but
it is always lighter than lava, very coarse-grained with 2. Method and techniques
numerous small cavities. Unfortunately, recent restorations
using inadequate materials incompatible with the original 2.1. Environmental data
bedding mortar may cause permanent damage [2]. The use
of physically–chemically suitable mortar, aesthetically Microclimatic parameters were measured by installing a
compatible, must help restoration works [2–5]. Detailed monitoring system consisting of microsensor instruments, a
studies on historical mortars sampled in the Roman PC-based data logger system and a signal-conditioning unit
catacombs may contribute to understanding the aspects of inside Saint Callistus catacomb. Microsensors installed on
missing traditional techniques. air/floor/lamps measured temperature (measurement accu-
The aim of this work is to determine the chemical and racy 0.01 8C), air relative humidity (0.1%), air CO2
mineralogical composition, texture, structure and properties concentration (2.8 ppm) and 222Rn concentration (pylon
of lime–pozzolana mortars from Roman catacombs. The AB5 scintillometer with a diffusion detector). Automatic
research concentrated on three aspects of research: (i) recording of data every 2 min was performed during 6
 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565 1557

months. This information will be helpful in understanding 2.4. Experimental reaction by a self-isothermal XRD
the mortar hardening and preservation conditions. Also, an
external weather station was installed outside the catacomb The experimental in situ hardening of different mixtures
to record outdoor conditions. of portlandite–pozzolana mortars was performed in a self-
made chamber for X-ray diffraction during which a
2.2. Samples and microscopy sequential test was performed for 132 h. The procedure
was as follows: (1) a continuous experiment for 13 h,
To reduce damage in the ancient walls, mortar samples recording first 265 isothermal XRD sequential profiles of 3
ranging from 5 to 30 g were carefully collected from Saint min each at 17 8C from 278 to 388 2u; (2) six subsequent
Callistus and Domitilla catacombs, using a scalpel, a small series of 24 XRD profiles each, approximately for 1 h at
chisel and a little hammer. Samples were observed under the 178C, after addition of water to the sample to observe
stereomicroscope to separate different layers for subsequent portlandite carbonation under completely wet conditions;
analysis. Optical observation of small chips of the samples and (3) additional series of 24 XRD profiles in the same wet
cut-polished in thin sections up to 35 Am in thickness was conditions after 1 day and 1 weekend to evaluate the
done, using a polarizing microscope (Nikon Eclipse C600 progress of the carbonation. These XRD profiles, recorded
POL) equipped with a digital Nikon Coolpix 950 camera. In under thermally controlled conditions, were performed
this manner, the mineralogy and type of aggregate, binder, using a Phillips PW1710/00 powder diffractometer with
porosity, cracking, secondary mineral formation and binder– CuKa radiation. Patterns were obtained by step scanning
aggregate ratios were determined. Morphology, textures, from 278 to 388 2u in 0.05 steps with a count of 2 s per step.
relationships, composition, crystal shapes and sizes of lime– Improvements made to the X-ray diffractometer include (a)
pozzolana mortars were also studied by environmental a new original water refrigerator; (b) a new stainless steel
scanning electron microscopy (ESEM). For this technique, door to hold internal cooling fluids; (c) internal heaters
sample metallisation was done using gold vapour in vacuum (resistance and halogen lamp) in new sample holders under
(50 2 of gold cover) in a Bio-Rad SC515 sputter coating PID thermal control; (d) a new electronic circuit and
unit. Samples were observed in a Philips XL20 SEM at software to record differential thermal analyses between
accelerating voltages of 20–30 kV. The EDS analyses were the sample and reference; and (e) a specific subdoor to fit
obtained using a Phillips EDAX PV9900 with a light the relative humidity detector [6,7]. The simplest method
element detector type ECON. The porosity was determined was to modify only the door of the sample chamber; a new
by two methods, microscopy of thin sections and mercury rounded door was constructed to fit newly constructed and
intrusion porosimetry, using a Micromeritics Autopore III S marketable supplies. A self-made MS-DOS program pack-
9400. age, i.e., PLV-SIRDAT by J. D. Martin-Ramos (jdmartin@
goliat.ugr.es), was used for the full-duplex control of the X-
2.3. Chemical and structural analyses ray diffractometer (Philips PW1710/00 with a PW1712
communication card) via a RS232 serial port. PLV is made
The chemical and mineralogical composition of indi- up of several programs: (a) REGISTRO—to control and
vidual mineral grains was determined by electron microp- acquire diffraction data; (b) LECTOR—a diffractogram
robe (EM; Jeol Superprobe JXA-8900M), bulk and evaluation program; (c) X—a program for fully automatic
channel-selected (TAP, PETJ, LIF, PETH) X-ray spectra phase identification in mixtures with sine and cosine 2u
search and identification routines. The standards used were corrections; and (d) R—for refining cell parameters.
natural and synthetic crystals from the collection of the SIRDAT is a specific program that allows sequential data
bServicio de Microscopı́a Electrónica Lluis Bru,Q Complu- records to be processed.
tense University, Madrid. The chemical composition of
mortar components was obtained through several techni-
ques such as inductively coupled plasma-mass spectrom- 3. Results
etry (ICP-MS) for trace elements; atomic absorption
spectroscopy (AAS) and electron microprobe analysis 3.1. Environment data
(EMPA) for major and minor elements. The concentration
of ferrous iron was determined by wet chemistry using a Microenvironmental recordings inside the catacombs
redox titration with ceric sulphate as titrant, measured with show humidity values close to saturation (relative humidity
an ORP electrode. The semiquantitative mineral composi- z97%), high CO2 concentrations (1700 and 3500 ppm in
tion of the lime–pozzolana mortars was determined the Domitilla and Saint Callistus catacombs) and a constant
through X-ray diffraction (XRD) using a Phillips PW- temperature of 15–17 8C during most of the year. Infiltration
1710 powder diffractometer with CuKa radiation. Patterns waters are scarce. Only two dripping points could be
were obtained by step scanning from 38 to 758 2u with a sampled in the Saint Callistus catacomb. Water chemistry
count of 0.2 s per step, exploration speed of 28/min at 40 shows pH values (7.5 and 7.8) close to neutral and locally
kV and 40 mA. high concentrations of nitrates (148 ppm), which could be
1558 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565

explained by the farming activity in the surrounding quartz in a vitreous glassy matrix. This matrix could be the
landscape. Minor values of sulphate (46 and 33 ppm) have main component of the whole rock. The aggregate is
also been observed. CO2 partial pressure was 10
2.06 to composed of fragments of vitreous volcanic rock, usually
10
1.97 bar. The microclimate monitoring system measuring including euhedral phenocrysts with sharp edges and
temperature, relative humidity and CO2 concentration was irregular shapes. Different sources of pyroclastic rocks are
operative in Domitilla and Saint Callistus catacombs. assumed, given that phenocrysts are more abundant in the
Recently, lighting time, air and rock temperature and mortar aggregates than in the surrounding substrate. Like-
222
Rn concentration are being measured in the Saint wise, X-ray diffraction analyses show divergence between
Callistus catacomb. A new sensor detecting temperature the substrate pyroclasts and the aggregates composition
differences between the air and rock surface was also (Table 1). Phyllosilicates are much more abundant in the
installed in the Oceanus cubiculum. This wealth of new data host rock than in the aggregate, in respect to pyroxenes (3:1
will be used in additional studies regarding the impact of in mortar versus 8:1 in host rock). Secondary minerals,
tourists visiting the architectural heritage. such as analcime, are more abundant in the mortars than in
the host rock. Occasionally, no feldspars occur in the
3.2. Mortar characterization mortar, but analcime and secondary products are detected.
Possible secondary gypsum was observed in three samples
Mortar-coating thickness onto the volcanic substrate in percentages less than 5%. Large calcite fragments are
ranges from 2 to 20 mm. Mortars are composed of one, two also found as aggregate (Fig. 1d). Lime masses remaining
or three separate beds with different sizes and shapes of isolated after the slaking can explain them. Both the mortar
aggregates (Fig. 1c). Usually, the mortar shows only two aggregates and the volcanic rocks are rich in iron
layers: the inner ranging from 11 to 13 mm and the outer aluminosilicate phases (approximately 50%; Table 2). The
from 0.3 to 0.6 mm. The external surface of mortars often CaO content is also important, mainly in mortar samples
displays areas with a smooth grey–white–brown colour. A (ca. 17%). The initial ratios of ferrous versus ferric iron in
mixture of volcanic rock aggregate and calcite binder forms the pyroclastic rocks point to the redox conditions of
the common Roman mortar from catacombs. Mortar beds crystallization (e.g., 1:4.9 for the analysed samples of
coating the rock surface or covering former mortars are substrate). In addition, this (Fe2+/Fe3+) ratio is a sensitive
thicker and comprise coarser aggregates. External beds are indicator of the oxidation state (e.g., 1:2.5 in the mortar
usually composed of fine volcanic dust and calcite. One samples analysed). During weathering processes, Fe (III) is
analysed mortar sample shows a thin layer of quartz more conservative than Fe (II). The increment in the
fragments mixed with lime, covering the volcanic aggregate proportion of Fe (III) versus Fe (II) indicates more intense
mortar. Frequently, the rock-mortar interface shows a thin weathering of the rock [8,9] from its formation until its use.
bed of pure calcite binding both surfaces (Fig. 1c, d). The Roman builders would select the volcanic rock less altered
lime mortar displays two or three separated beds including for use as aggregate in the mortars. The trace element
occasional air pockets among them. Thin sections of concentrations of mortars studied are quite homogeneous,
aggregate tuffs show calcite in the fissures. Under the and they are linked with the frequent addition of fresh tuff
polarizing microscope, thin sections of mortars show fragments. Barium and strontium are the most abundant
aggregate/binder ratios ranging from 0.5:1 to 1.1:1 (Fig. trace elements in both rock and mortars. The most
1d), being highly variable even inside a single sample. remarkable feature is the high content in copper, lead and
Aggregate grain sizes are also highly variable. It is possible rubidium of the mortars in comparison with the volcanic
to distinguish two sizes of population in the coarse mortars, rocks. This could be related to the composition of the
one from 0.5 to 2 mm and one from 100 to 300 Am. The recipients in which the mortars were prepared. Mortars
grain size distribution is apparently bimodal. Mortar show high values of loss on ignition ranging from 17% to
aggregates are composed of small fragments of tephra 25%, in connection with its high content in calcium
(Fig. 1d), which is a frequent substrate volcanic rock in the carbonate. Mercury porosimetry measurements (Fig. 2)
Lazio region, almost certainly obtained by digging galleries indicate that volcanic rock and mortar show a large
and tombs. These fragments show pyroxene phenocrysts porosity, averaging ca. 42%, while external mortar beds
(8–14% of augite–aegirine rich in Na and Fe), sanidine, or stucco show a lower porosity, ca. 39%. The pore size
biotite, analcime (20%), phyllosilicate phases and calcite distribution is homogeneous, all being below 100 Am (Fig.
(5–10%). Occasionally, they are mixed with detrital sandy 2). Observations by ESEM and polarizing microscopy of

Table 1
Average mineral composition of volcanic rock (15 samples), coarse mortar (18 samples) and fine mortar (12 samples)
Material Phyllosilicate Augite Feldspar Quartz Analcime Calcite Gypsum
Volcanic rock 68 10 5 b2 10 5 –
Coarse mortar 30 10 2 b3 14 42 1
Fine mortar 13 2 2 3 6 74 –
 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565 1559

Table 2
Average bulk material elemental analyses—volcanic rock (15 samples), coarse mortar (18 samples) and fine mortar (12 samples)
Material SiO2 TiO2 Al2O3 Fe2O3 FeO MnO MgO CaO Na2O K2 O P2O5 LOI
Volcanic rock 45.28 0.68 18.55 6.88 1.21 0.22 2.22 7.16 2.18 2.09 0.27 13.04
Coarse mortar 35.32 0.43 12.31 3.73 1.45 0.15 1.95 19.39 1.95 1.25 0.29 21.44
Fine mortar 27.55 0.33 10.11 3.53 0.69 0.10 1.99 28.98 1.23 1.22 0.15 24.07

Material Ba Sr Rb Zr Pb Ce La U Th Li V Cu
Volcanic rock 2068 1101 266 363 119 255 149 10 67 61 191 46
Coarse mortar 1538 923 332 250 170 193 114 9 50 46 147 116
Fine mortar 2440 1183 334 256 269 201 117 9 50 45 150 68
Major oxides and loss on ignition (LOI) in wt.%, trace elements in ppm.

mortar thin sections (Fig. 3a) link this large porosity with first step in the lime putty preparation is to dissolve the
the main fissures and orthogonal cracking. A network of calcium hydroxide in water:
fissures reaches up to the external surface, helping mineral
dissolution. Fissures occur while the sample hardens by fast CaðOHÞ2 X Ca2þ þ ðOHÞ

2
reduction during cooling in new atmospheric conditions. A saturated solution with pH 12.8 is produced. Before
Occasionally, calcite masses fill these cracks (Fig. 3b). total slaked lime dissolution, carbonate ions come into the
process, originating from atmospheric carbon dioxide. The
3.3. Mortar hardening calculations rate of CO2 uptake by water is relatively fast. The halftime
of the reaction CO2(g)=CO2(aq) is only minutes and
The mortar carbonation or hardening is a complex decreases with increasing air CO2 pressure [10]. The
process that includes several consecutive reactions. The subsequent formation of carbonic acid is very fast (halftime

Fig. 2. Mercury porosimetry results of both samples: (a) mortar and (b) volcanic substrate. Note the large differences.
1560 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565

Fig. 3. ESEM photomicrographs of ancient Roman and experimental mortar samples: (a) layers of mortar and stucco and surface alterations; (b) orthogonal
fissure filled by neoformed calcitic fibbers; (c) experimental mortar section showing: (I) voids left by loss of aggregates, (II) orthogonal cracks and fissures,
(III) aggregate inside calcitic binder; (d) experimental mortar interface between volcanic aggregate and calcite binder showing remaining portlandite crystals.

~0.1 s), being close to 10
6 times the halftime of the where R is in mmol/cm2 s, and brackets denote activities.
homogeneous reaction: The value of k 4 is a function of temperature and Pco2. For
Pco2b10
1.5 bar, the function
H2 CO03 X Hþ þ HCO

3

For this reason, the complete reaction–formation of log k4 ¼
7:56 þ 0:016T
0:64 log Pco2
calcium and carbonate species is fast enough to achieve (T in Kelvin, Pco2 in bars) allows the rough calculation of
equilibrium in few minutes. Calcite precipitation–dissolu- the rate constant k 4 [10]. Temperature functions and rate
tion is described by the overall reaction [11]: constants have been suggested in previous publications
Ca2þ þ 2HCO
[11,12].
3 X CaCO3 ðcalciteÞ þ CO2 ðgÞ þ H2 O
Through this equation, using the PHRQPITZ software
In which the essential reactions with their respective rate [13], the theoretical rate of calcite precipitation from a
constants (k) are as follows: calcium hydroxide saturated solution under different air
Ca2þ þ HCO

3 X CaCO3 þ H
þ Pco2 values and a constant temperature (15 8C) has been
calculated. An open carbonate system was assumed where
with a rate constant k 1 the CO2 is continuously restocked by gaseous exchange.
Ca2þ þ HCO

3 X CaCO3 þ H2 CO3 The value used for the equilibrium constant for calcite at 25
8C was (
log K eq=8.48; [14]). Two values have been
with a rate constant k 2
chosen to simulate a hypothetical hardening process of lime
Ca2þ þ HCO

3 þ OH
X CaCO3 þ H2 O putty: a Pco2 of 10
3.7 bar that could correspond to the
with a rate constant k 3. external atmosphere over the catacomb and 10
2.6 bar, a
Considering constant Pco2 and temperature and assuming similar value to that measured inside by the microenvir-
no inhibition for absorbed particles, the rate of solution or onmental monitoring.
calcite precipitation is given by the sum of the rates of its Fig. 4 shows how the calcite precipitation rate increases
elementary reactions and may be written as follows: together with the CO2 partial pressure, although the total
precipitated volume is slightly smaller for the highest CO2
R¼ k4 ½Ca2þ ½HCO
þ
3 
k1 ½H 
k2 ½H2 CO2 
k3 ½H2 O values. Calcite precipitation immediately begins when the
 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565 1561

Fig. 4. Calcite precipitation rate under normal (10
3.7 bar) and high (10
2.6 bar) CO2 pressure environment.

saturated solution is kept in touch with CO2. This process

could explain the observed large calcite fragments in the
mortars. The initial high rates decrease very fast in respect to
the calcite precipitation. In open-air environments, with
Pco2 values similar to hypogeal backgrounds, lime solution
achieves saturation with respect to calcite in 37 min, four
times quicker that in the case of atmospheric CO2 pressures.

3.4. Mortar hardening measurements by XRD at 17 8C

Previous calculations allow a test on in situ hardening of

Roman mortars composed of pure slaked lime and grinded
volcanic aggregate from the catacombs. The result is a fine-
grained mortar with microstructural features close to the
original fine-grained Roman mortar (Fig. 3c). Observation
by ESEM of the experimental mortars show fissures linked
with sample hardening, which resembles those of the
original mortars. The experimental mortars display a smaller
cohesion between aggregates and binder (Figs. 3–6). This
smaller cohesion could be associated with the incomplete
carbonation demonstrated by the presence of small por-
tlandite crystals in the innermost part of the experimental
mortar (Fig. 3d).
The experiment was performed in a self-made chamber
for X-ray diffraction, where 847 sequential XRD profiles of
calcite and portlandite phases were recorded for 132 h.
Measurements were limited to the angular region from 278
to 388 2u to observe the gradual growth of XRD peaks of
calcite (3.03–2.84–2.49 2) and the decrease of portlandite
XRD peaks (3.11–2.63–2.44 2), as shown in Fig. 5. In good
agreement with the predicted portlandite–calcite phase
transition, the main exchanges of peak sizes were observed
during the first step of the test, in the initial 13 h. However,
in the subsequent XRD profile series, after water addition,
only slow changes are observed (Fig. 5b). The parameter-
isation of the XRD measurements was carried out analysing Fig. 5. Mineralogical evolution of experimental mortar in the XRD
sequential profiles: (a) fast portlandite substitution by calcite in the first
the full width, half maximum, using the PLV-SIRDAT hours of hardening, (b) complete mineralogical evolution of mortar during
software of Dr. Martin-Ramos, for the main XRD peaks of experiment. Note the progressive and partial removal of portlandite linked
calcite and portlandite phases. The FWHM data were with the calcite neoformation during the carbonation process.
1562 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565

Fig. 6. Theoretical evolution of calcitization rate in different pCO2 (10
3.7 and 10
2.6 bar) conditions and real data obtained through XRD under standard
conditions.

plotted versus time to be compared to theoretical calcu- and later producing a thin dust of volcanic aggregate, sized
lations. Both results, experimental and theoretical harden- below 200 Am. External thinner bed and stuccoes provide a
ing, display similar paths. The experimental carbonation of smooth surface to be decorated with frescoes. Unfortu-
the portlandite phase, however, is slower than those nately, these are often fissured by shrinking, as could be
forecasted by the calculations (Fig. 6). observed under the microscope. The volcanic aggregates
contain a considerable amount of vitreous matrix and
unstable minerals, such as sanidine feldspar, favouring
4. Discussion chemical reactions with the lime in the interfaces. This
reinforces the binding. The irregular fragments’ shapes also
Traditional Roman mortar handling techniques nowadays increase the aggregate–binder reactive surface improving
are accepted to produce restoration lime mortar. [1,2,4,5]. the bonding and decreasing the shrinking of the mortar,
Romans carried lime putty in baskets and mixed aggregate resulting in a higher cohesion and strength but a lower
from local sources just before its application [15]. Lime– plasticity of the mixture [18,19]. Differences in mineralog-
pozzolana mortar is a mixture of aggregate and lime. Lime ical and chemical composition and alteration degree imply
(CaO) is produced by heating limestone (CaCO3) to ca. 900 that the aggregates do not originate from around the
8C. The resultant quicklime (CaO) is mixed with water to catacombs at the same point of the final building. It is
obtain slaked lime (Ca(OH)2). The final concrete results suggested that the high content in detritic and secondary
from the addition of surrounding aggregate materials, such clays (sometimes up to 80%) of the local material led the
as quartz sand, grog and rocks [1]. From the third century Roman builders to employ altered volcanic materials from
BC onwards, in the Lazio region, the use of volcanic rocks different sources. The content in pozzolanic components,
was widespread, thanks to their fast hardening, easy such as pyroxene, sanidine and volcanic glass, could be a
handling and high strength even under water and under reason for this. Vitruvius, in the first century, reported a
fire. The characteristics and strength properties of a mortar preference for sands with low clay content [1]. While it is
arise from the composition of aggregate and binder and the clear that Roman builders often used local materials,
specific evolution during hardening [16]; this last factor nevertheless, they selected well-known substances with
depends on placing techniques and environment. the best behaviour.

4.1. Mortar volcanic aggregates 4.2. Carbonated binder and environmental CO2

All aggregate components studied, with a few excep- Two factors help hardening during the mortar carbona-
tions, are volcanic rocks. Mortar observation shows two or tion: (i) high CO2 concentrations in the catacombs’ environ-
three stratified beds thinning outward. The use of several ment and (ii) high porosity through fissures that allow an
coating beds prevents the cracking of walls [1]. Aggregate extra interchange surface assimilating carbon dioxide [19].
sizes change depending on the bed depth, as has been The high humidity and high CO2 concentration in cata-
observed in other Roman monuments [3]. The inner first combs’ microenvironment are favourable conditions for
layers are less aesthetic and display minor porosity as a calcitization. [16]. Human breathing by builders, inhabitants
result of mixed aggregates sizes. Mixed sizes of coarse or visitors produces extra CO2. The low thickness of the
aggregates with a slight amount of pozzolanic dust mortar bed also helps a fast carbonation, inasmuch as
aggregate in the walls increase the reactive surface [17] surface hardening is fast, but at depth, hardening can take
and increase the calcitization speed [16]. The bimodal grain several years depending on environmental conditions,
size distribution probably indicates a separate fragmentation porosity and thickness [19]. Carbonation is a long-term
of the volcanic rocks, first obtaining grains of ca. 1–2 mm process taking many months; extremely fast carbonation
 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565 1563

could inhibit slower pozzolanic reactions, reducing the to avoid detachment of rock fragments and to decorate
cationic diffusion. tombs and crypts.
The main difference between the theoretical kinetic mode The unexpectedly high porosity of the analysed mortars
and the experimental simulation carried out is the initial does not agree with the traditional water resistance and high
assumption of an open carbonate system in the former. The strength of Roman mortars having smaller porosity. The
gradual calcification of the mortar causes a progressive microscopy study of the pore size distribution support as the
closure of the system. Subsequently, the CO2 in the mixture porosity of volcanic fragments is very different to that of the
is not totally restocked, hindering sometimes a complete mortar. The substrate porosity is mainly formed by
carbonation. Mortars, under real conditions, evolve in open, intergranular spaces, while the mortar porosity is caused
closed and mixed situations. The high porosity of the by parallel fissures produced by volume losses during a
Roman mortars and the high CO2 concentration inside the faster hardening of the surface. Larger pores are empirically
catacombs help CO2 absorption by the mixture. However, linked to coarser grains and imply higher CO2 diffusion and
carbonation is faster in a completely open system. The slow faster calcitization [16]. Calcite minerals have partially
hardening of the flowing mixture produced by the fast filled the pores; several dissolution–precipitation processes
carbonation in the high CO2 environment of catacombs of calcium carbonate have also been noted in the fissures
could have stimulated the Roman builders to increase the (Fig. 3b).
binder amount in respect to the aggregate. An important
factor to be considered in the chemical reactions at the 4.3. Ancient Roman technique to coat walls in hypogeal
binder–aggregate boundaries is the pH inasmuch as a high environments
pH more easily induces reactions of amorphous silicates,
such as volcanic glass. During a common calcitization Vitruvius explained, about pozzolanic mortars, that the
process, the pH decreases from 12.5 to 7–8 [16], and the convenient aggregate–lime ratio is 3:1 [1]. Plinius, in the
aforementioned chemical reactivity is reduced. first century AD, proposed a 4:1 ratio [20]. Different authors
The Pco2 average values of the catacombs’ environment have observed similar ratios in ancient mortars of Roman
lead to the lime solutions achieving saturation with respect monuments but usually with aggregate ratios closer to 2:1
to calcite in a short time (ca. 37 min), involving a very fast [3,4]. Petrographical study of the Roman catacombs’
decrease in pH (Fig. 7), which could inhibit chemical mortars by polarised light microscope shows that the
reactions between aggregate and lime. Fast calcitization of aggregate–binder ratio is highly variable, ranging from
the binder under higher concentrations of atmospheric CO2 0.5:1 to 1.1:1. This proportion shows larger quantities of
means a pH reduction and low reactivity of the mortars. For lime but still far from those employed in the building of
this reason, after CO2 exposure, thin layers of lime mortar harbour infrastructures and breakwaters (2:1) [1] and those
are formed. This differs from the classic pattern of observed in other monuments [4,21]. The increase on the
pozzolanic mortars. The physical properties of mortars in lime amount could be a useful technique for special
the catacombs are supported more in the binder carbonation environmental conditions, such as the catacombs, where
than in the pozzolanic activity. In this sense, it is important calcitization is extremely fast. It could help setting onto
to stress that mortars in catacombs were not used as vertical surfaces in high humidity. The mechanical conse-
structural framework but for covering ceilings and walls quence of mortars with high lime contents has higher
strengths but also a higher risk of cracking through volume
changes during hardening [19].
Limestone lumps inside mortar are linked with lime
fragments not slaked due to water shortage during the
slaking process. Most of the observed micritic fragments in
the catacomb walls are slightly rounded or as pyramidal
shapes formed in the mixture, hardened on the surface and
were later included, still soft, in the remaining binder. This
could be in good agreement with dry-slaking techniques
using low water–binder ratios [2,16]. Low water–lime ratios
mean low diffusion rate, and secondary formation in the
interfacial transition zone is denser [22]. In some cases,
radial crystal growth around voids indicates a secondary
origin after volcanic glass dissolution. The pronounced
mechanical and chemical differences between lime and
volcanic rock may cause wall flaking in many areas of the
catacombs. Voids and small discontinuities between beds
Fig. 7. Theoretical rate of the pH decrease during lime putty calcitization suggest that builders waited for the first layer to set before
under different pCO2 environmental conditions. placing the second layer and the following coarse mortar.
1564 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565

The calcitization rate of pure lime depends on the the hypogeal level of CO2 and the porosity provide a
atmospheric CO2 content. Using the chemical composition ceiling to the cationic diffusion forming hydrous calcium
and the stoichiometric reactions proposed by Plummer et al. aluminosilicates.
[23], it is observed that wet lime under 3000 ppm CO2 – This type of study is essential in the description of useful
calcitizes five times faster than in normal conditions parameters and details on Roman mortars, thereby
(standard 270 ppm; Fig. 6). This example is applied to an facilitating restoration. The final objective must be the
open system assuming a thin layer in touch with air. In the use of materials and recipes compatible to the original
case of thicker layers, the calcitization rate decreases mortar. We conclude that special restoration mortars must
inasmuch as the total bed changes to an almost closed be used in those cases, where the original material is
system as a consequence of the external fast calcitization located in hypogeal environments.
that blocks the contact with the atmospheric CO2.

Acknowledgements
5. Conclusions
We are very grateful to Rafael González, Maribel Ruiz
– Lime–pozzolana mortars have been employed to cover and Marisa Vallejo of the Department of Geology, Museo
the most representative surfaces of the Roman cata-
Nacional de Ciencias Naturales (CSIC), Madrid, and to
combs’ walls. The observed mineralogical, chemical and
Olga Cazalla of the Universidad de Granada. Special thanks
textural properties of these mortars could provide an
also to Ugo and Giusepe in the Roman catacombs. This
insight into the pasting techniques used by the ancient
research has been supported by the European Union
Roman builders inside the catacombs, where a specific
Research Project CATS EVK4-CT200-00028 on biogenic
microenvironment plays an important role in the chem-
decay prevention in Roman catacombs.
ical evolution of the mortar pastes.
– The measured environmental data inside the Saint
Callistus and Domitilla catacombs show a constant
temperature of 15–17 8C, a high CO2 content (1700 to References
3500 ppm) and a relative humidity close to 100%. These
conditions and particularly the high CO2 concentration [1] M.L. Vitruvius, Los Diez Libros de Arquitectura. Ed. Iberia, Madrid,
2000.
speed-up the lime calcitization roughly by 500%. [2] P. Degryse, J. Elsen, M. Waelkens, Study of ancient mortars from
– Roman builders made one to three mortar beds, using Sagalassos (Turkey) in view of their conservation, Cem. Concr. Res.
local aggregates. However, the in situ volcanic aggre- 32 (2002) 1457 – 1463.
gate was not used. The high proportion in clay of the in [3] A. Gqleç, T. Tulun, Physico-chemical and petrographical studies of
situ tephra probably leads them to choose different old mortars and plasters of Anatolia, Cem. Concr. Res. 27 (1997)
227 – 234.
sources of less altered aggregates. Hardened lime lumps [4] M.P. Riccardi, P. Duminuco, C. Tomasi, P. Ferloni, Thermal, micro-
inside the binder denote low water–mortar ratios for scopic and X-ray diffraction studies on some ancient mortars, Therm.
slaking. Acta 321 (1998) 207 – 214.
– Frequently, a thin layer of pure lime was smeared onto [5] O. Cazalla, E. Sebastian, G. Cultrone, M. Nechar, M.G. Bagur, Three-
the soft and crumbly natural surface of the host rock. way ANOVA interaction analysis and ultrasonic testing to evaluate air
lime mortars used in cultural heritage conservation projects, Cem.
Builders waited for the desiccation of this first bed to Concr. Res. 29 (1999) 1749 – 1752.
plaster a second bed of pure lime mortar. Two [6] J. Garcia-Guinea, R. Abella, S. Sánchez-Moral, R. Benito, J.D. Martı́n-
explanations are suggested: (i) providing hardening of Ramos, Examining hydrated minerals using optically stimulated
this soft substrate and (ii) possible experiments to see X-Ray diffraction, an inexpensive modification of traditional diffrac-
how fast the lime cures in situ, given the specific tometers, J. Sediment. Res. 70 (2000) 964 – 967.
[7] J. Garcia-Guinea, R. Ortiz, V. Correcher, A. LaIglesia, J.D. Martı́n-
atmosphere inside the catacomb. Ramos, Improvements to X-ray diffractometer: heaters, hygrometry,
– The aggregate–lime mixture shows a higher content of thermo-differential and spectra analyses, Rev. Sci. Instrum. 72 (2001)
binder in comparison to contemporary written reports 4005 – 4007.
and other Roman monuments. The structure of these [8] H.W. Nesbitt, G.M. Young, Formation and diagénesis of weathering
profiles, J. Geol. 97 (1989) 129 – 147.
mortars shows (i) coarser aggregates and thicker beds on
[9] H.W. Nesbitt, R.E. Wilson, Recent chemical weathering of basalts,
the inside; (ii) thin, smoothed, light and fine-grained Am. J. Sci. 292 (1992) 740 – 777.
external surfaces with low content of aggregates; and (iii) [10] D. Langmuir, Aqueous Environmental Geochemistry, Prentice-Hall,
paintings and frescoes on the outside. The observed high New Jersey, 1997, 600 pp.
porosity of the mortars can be attributed to cracking after [11] L.N. Plummer, D.L. Parkhurst, T.M.L. Wigley, Critical review of the
drying linked with the high binder content. kinetics of calcite dissolution and precipitation, in: E.A. Jenne (Ed.),
Chemical Modelling of Aqueous Systems, Am. Chem. Soc. Symp.
– The high content in volcanic glass and unstable minerals Ser., vol. 93, Am. Chem. Soc, Washington, DC, 1979, pp. 537 – 573.
in the aggregate helps the chemical reactions around the [12] E. Busenberg, L.N. Plummer, A comparative study of the dissolution
aggregate grains. Conversely, the fast calcitization under and crystal growth kinetics of calcite and aragonite, in: F.A. Mumpton
 S. Sánchez-Moral et al. / Cement and Concrete Research 35 (2005) 1555–1565 1565

(Ed.), Studies in Diagenesis, U.S. Geological Survey Bulletin, vol. [17] G. Baronio, L. Binda, N. Lombardini, The role of brick pebbles and
1578, 1986, pp. 139 – 168. dust in conglomerates based on hydrated lime and crushed bricks,
[13] L.N. Plummer, D.L. Parkhurst, G.W. Fleming, S.A. Dunkle, Constr. Build. Mater. 11 (1997) 33 – 40.
PHRQPITZ a computer program incorporating Pitzer’s equations [18] G. Giaccio, R. Zerbino, Failure mechanism of concrete. Combined
for calculation of geochemical reactions in brines, U.S. Geological effects of coarse aggregates and strength level, Adv. Cem. Based
Survey Water Resources Investigation Report, vol. 88, U.S. Govern- Mater. 7 (1998) 41 – 416.
ment Printing Office, Washington, DC, 1988, p. 4153. [19] A.H.P. Maurenbrecher, M.Z. Rousseau, Review of factors affecting
[14] D.K. Nordstrom, L.N. Plummer, D. Langmuir, E. Busenberg, the durability of repointing mortars for older masonry, 9th Canadian
H.M. May, Revised chemical equilibrium data for major water– Masonry Symposium, Publ. Inst. For Research in Construction, Univ.
mineral reactions and their limations, in: E.A. Jenne (Ed.), Chemical New Brunswick, Canada, 2001, 12 pp.
Modeling of Aqueous Systems, Am. Chem. Soc. Symp. Ser., vol. [20] M.A. Plinius, Historia Natural, Facsimile, ITGE, 1998, (1629) 2 vols.
93, Am. Chem. Soc, Washington, DC, 1990, pp. 857 – 892. [21] S. Lodola, Studio archeometrico di laterizi e malte dallo scavo
[15] J.P. Adam, in: Batsford (Ed.), Roman Building: Materials and archeologico di Villa maria (Lomello, pavia). Thesis for the Degree in
Techniques, 1994, 360 pp. Geology, (1993) University of Pavia: 118.
[16] R. Hayen, K. Van Balen, D. Van Gemert, The influence of [22] J.P. Ollivier, J.C. Maso, B. Bourdette, Interfacial transition zone in
production processes and mortar compositions on the properties of concrete, Adv. Cem. Based Mater. 2 (2) (1995) 30 – 38.
historical mortars, 9th Canadian Masonry Symposium, Publ. Inst. [23] L.N. Plummer, T.M.L. Wigley, D.L. Parkhurst, The kinetics of calcite
For Research in Construction, Univ. New Brunswick, Canada, 2001, dissolution in CO2–water systems at 5 to 6 8C and 0.0 to 1.0 atm CO2,
12 pp. Am. J. Sci. 278 (1978) 179 – 216.