34 Hobbs - siddall

Cementitious materials of the ancient world

linn W. Hobbs - RutH siddall

1. intRoduction

Among the earliest transformative technologies developed by human beings is the

production of cementitious materials from raw material resources. Deliberately
manufactured lime and gypsum cements appear to have been utilized as early as the 7th
millennium BCE for mortars and plasters in wall and building construction,1 and perhaps
far earlier, probably augmenting mud plasters, mortars and adobe which continued to be
used. Bitumen, and perhaps other harvested organic polymeric substances (such as latex in
Mesoamerica), was used as an ancient water-impermeable mortar in brick constructions.
The irst truly large-scale monumental use of manufactured cements was, of course, the
pozzolanic mortars of Roman concrete.2
Inorganic cements function in many ways as “portable” stone, the raw materials
transformed in some way into a inely-divided substance that, with addition of water,
exhibits plasticity and after a time sets to a rock-hard consistency. They may properly,
therefore, be considered ceramic materials, though pyrotechnologies may be involved
only in their production (if at all) and not in their transformation into a durable monolithic
inorganic solid. The chemical and physical phenomena underlying the production and
setting of these cements are quite disparate, but there exist some commonalities, and the
setting reactions can be usefully assigned to generic categories.
Unfortunately, there is often confusion in archaeological communities about the
identity and processing history of archaeological cements, not least because it is not
always straightforward to identify the cementitious phase or phases unambiguously from
the set product (particularly after centuries of possible environmental and evolutionary
alteration) and to connect these to a cement type and prior processing protocols. At the
very least, however, there needs to be a more informed awareness of the categories of
cementitious materials utilized in ancient times and a general appreciation for the science
of their setting reactions, the resulting product chemistries and microstructures, and
the mechanical properties they exhibit. The purpose of this contribution is to provide a
minimal background that is hoped will prove useful to archaeologists, archaeometrists and
historians. Some connections to modern cements will also be made.

1 W.H. GouRdin - W.D. KinGeRy, ‘The beginnings of pyrotechnology: Neolithic and Egyptian
lime plaster’, Journal of Field Archaeology 2 (1975) 133-149; W.D. KinGeRy - P.B. VandiVeR - M.
PRicKett, ‘The beginnings of pyrotechnology, part II: Production and use of lime and gypsum plaster
in the pre-pottery Neolithic Near East’, Journal of Field Archaeology 15 (1988) 219-244; G.O.
Rollefson, ‘The uses of plaster at Neolithic ’Ain Ghazal, Jordan’, Archaeomaterials 4 (1990) 33-
54.
2 H.N. lecHtman - L.W. Hobbs, ‘Roman Concrete and the Roman Architectural Revolution’, in
W.D. KinGeRy (ed.), Ceramics and Civilization, Vol. 3, (American Ceramic Society) Columbus, OH,
1987, 81-128; J-P. adam, Roman Building: Materials and Techniques, Batsford, London, 1994.
Comm. Hum. Litt. Vol. 128 35

It is important at the outset to make some important distinctions in terminology, in order to

avoid confusion and inject some degree of descriptive precision and uniformity of usage.
We should note that the following deinitions we adopt are not related to the chemistry
of the phases and can be applied to a range of materials that may be based on different
chemistries (gypsum, calcium carbonate, silicate, etc.). These deinitions describe the
physical condition of the material and, by association, the context in which it is used.
A cement is a material that provides internal cohesion, either to components of itself
(the binder or cementitious phase, typically arising from some form of chemical reaction
between precursors occurring during setting) or between other included inert-component
phases that are cemented or “glued” together by the binder phase. The cement may be
deined as including inely-divided inert components with dimensions ranging from the
microscopic features of the cementing phase itself (typically the latter are of order 2-5
µm) up to about 1 mm in size. Such inclusions may represent a relatively minor proportion
of the cement, but occasionally these additions may comprise a substantial or even the
major volume fraction, because their inclusion economizes on the amount of cementitious
phase – which is the more costly in terms of availability or processing – needed in the
cement. The ine-scale inert inclusions also often confer improved mechanical properties,
such as toughness (resistance to fracture) or hardness (elasto-plastic resistance to the
deformation of scratching or indentation) and a reduction in shrinkage (and hence reduced
cracking) during the setting process. Fine quartzite (crystalline silica) and other siliceous
(e.g. diatomite) sands are among those inclusions frequently added. These inert additions
are analogous to the temper inclusions added to (or present as a natural constituent of)
aluminosilicate clays that provide additional bulk and toughness to both green and ired
clay ceramics. In cements, the inclusion inventory also frequently includes incompletely
reacted particles of the cementing phase precursors.
Although the cementitious bond to the inert phase can be chemical, in many cases a self-
cohering binder phase may simply engulf the added inert-phase particles and mechanically
entrap them. Binder self-coherence may arise from establishing a boundary-less continuous
glassy phase or from atomic bonding across grain boundaries in a polycrystalline binder
phase reaction product. The micromorphology of cementitious phases is usually quite
important to understanding the cementing strength and other mechanical properties, such
as toughness and hardness of composite ensembles, as are the binder-inclusion bond
strength and the inclusion distribution also. However, it should be appreciated that such
micromorphological features are often not readily visible using light optical microscope
techniques because they occur on a micrometer or sub-micrometer scale.
A concrete comprises a cementitious composite made up of a cement to which has
been added inert material (aggregate) with macroscopic dimensions, typically deined
as being > 5mm.3 An upper limit is not placed on aggregate size, as this depends on the
scale and methods of the building operation in question. Gravel stones in Portland-cement
concretes are ubiquitous modern examples; so, too, are the carefully graded, selected and

3 J.E. PRentice, Geology of Construction Materials, Chapman and Hall, London, 1990. See
Chapters 3 (p. 68) and 4 (p. 110) therein for discussion of the mechanical and compositional
properties of coarse and ine aggregates.
36 Hobbs - siddall

emplaced caementa aggregates of Roman concrete. A mortar is commonly understood

to be a cementitious material used in-between bricks or building stones (and generally
will contain aggregate with dimensions < 5mm) – a role clearly served by Roman cement
within the brick (or earlier stone) facings of brick-faced Roman concrete. But, in illing the
space between the facing walls and with the addition of caementa, Roman cement there
becomes a concrete. The term plaster is generally reserved for wall coverings, the desired
wall smoothness of which generally dictates only ine-scale inclusion phases (such as
sand with dimensions typically smaller than 1 mm). Cement loorings are often similar to
plasters, though frequently with the additional requirement of water impermeability and
integrity while wet.
Although most cementitious mortars are poured or cast as water slurries, few set by
“drying”. None involves a formal chemical dehydration step during setting, and only one
(clay-base mud mortars) solidiies by removal of (only) excess water. Setting of the rest
involves at least one chemical reaction with the cementitious material phases themselves:
rehydration (gypsum mortars), carbonation (lime mortars), silica polymerization and
calcium silicate and aluminate phase formation (Roman pozzolanic cements), and
hydration-induced silica polymerization and aluminate hydration (Portland cement).

2. mud moRtaRs, PlasteRs and adobes

The most commonly used cementitious material before the late Palaeolithic era was mud
containing substantial amounts of clay, employed perhaps as early as 100,000 years BP in
Africa and soon after in Mesopotamia, Persia, Anatolia and elsewhere. The material was
utilized as a mortar, as a plaster, to form bricks – initially unired (sun-dried mud, or adobe,
brick) but later ired, and even as monolithic mud-aggregate adobe concrete. (The word
adobe comes rather directly from Middle Kingdom Egyptian, through Coptic and Arabic,
and thence into Spanish.) Silica sand was the most common inorganic inclusion; grasses,
straw, dung and other ibrous organic matter were commonly included to impart additional
toughness. The clay content of the mud is quite important, because many clays are highly
cementitious and when wet also confer the requisite plasticity for the manipulation of mud
mortars and plasters and the shaping of monolithic mud-based structures. The etymology
of the English word clay is similarly revealing: the Old English is claeg (akin to modern
German klei), and in Dutch, kleven, means “to stick.” All four versions in fact derive from
the earlier Latin glus, or “glue,” whence also Old French, glu, meaning “bird lime,” which
represents a crossover to yet another cementitious technology.
Clays are hydrated aluminosilicates, in which Al3+ ions are bound ionically to oxygen
in [AlO6] octahedra that share edges with each other and occupy structural alumina-like
sheets, and Si atoms are covalently bonded to oxygen in corner-sharing [SiO4] tetrahedra
occupying adjacent silica-like sheets. These two kinds of sheets are alternately bonded
to each other, by oxygen ion sharing, to form structural layers of two or more sheets.
Hydration contributes (OH)– ions substituting for O2– ions in the octahedral aluminate
sheets, or occluded water molecules in between layers. For example, kaolin Si2Al2O9H4,
which may be written more revealingly in terms of its bonding topologies as Si2O5•Al2(OH)4
Comm. Hum. Litt. Vol. 128 37

5 μm

(a) (b)
Figure 1. Layer structure of the clay kaolin, 2Si2O5•Al(OH)4. a) Representation of atomic structure,
showing two-dimensional sheet polymerization of [AlO6] and [SiO4] structural units, OH–
substitution for O2–, and capacity for hydrogen bonding across adjacent layers (after R.c. eVans,
Crystal Chemistry, 2nd ed., Cambridge Univ. Press, 1964). b) Tabular micromorphology of stacked
clay platelets in hardened kaolin clay (scanning electron micrograph, reproduced with permission
from W.d. KinGeRy - P. VandiVeR, Ceramic Masterpieces: Art, Structure and Technology, Free
Press, New York, 1986).
or its intrinsic stoichiometries as 2SiO2•Al2O3•2H2O, comprises two-dimensional Si4O10
hexagonal network sheets of corner-sharing [SiO4] tetrahedra bonded to an adjacent two-
dimensional hexagonal network sheets of edge-sharing partially-hydrated [AlO4(OH)2]
octahedra (Fig. 1a) in paired layers. Each paired layer is bonded to the next paired layers
by weak hydrogen bonding. Other clays are characterized by different numbers of sheets
and stacking orders per layer.
Clays therefore exhibit strongly two-dimensional tabular morphologies, in common
with other sheet aluminosilicates like mica K(Si3Al)O10Al2(OH)2 and talc Si4O10Mg3(OH)2.
Clay tabular grains are typically a few micrometers across and tens of nanometers thick
(Fig. 1b). H2O molecules can readily intercalate between tabular platelets, allowing
adjacent platelets to slide across each other readily; the easy shear deformation results in
plasticity, while the remanent hydrogen bonding and induced surface electrostatic charges
contribute stickiness. In excess water, the tabular plates disaggregate and form colloidal
suspensions that settle slowly. Soaking and settling is a common clay processing route for
production of plastic clay material from initially dry clay deposits.
Drying wet clays removes water and plasticity, but the resulting material exhibits limited
strength and toughness because of the weak hydrogen interlayer bonding. For this reason,
sand and ibrous organic matter are commonly added to clays to improve strength and
fracture toughness of unired brick and adobe. Firing removes chemically incorporated
water and substitutes hydrogen bonding by much stronger ionic Al-O or covalent Si-O
bonding across layers and between tabular platelets to form grain boundaries that
improve strength and decrease water permeability. Unired brick, adobe and clay mortars
nevertheless exhibit suficient compressive strength (~1 MPa) to have supported the
weight of the tall ziggurats of the second-millennium BCE Sumerians. (At least the lower
courses of these structures were commonly faced with ired brick and laid with bitumen
mortar to protect the inner unired brick from water erosion.) Sun-drying does not confer
38 Hobbs - siddall

improved mechanical properties to unired clays over those achieved by dehydration at

more ambient temperatures and serves simply to hasten the dehydration process.

3. GyPsum moRtaRs and PlasteRs

Gypsum plasters are known to have been in use in Egypt from the Predynastic period
onwards,4 though gypsum cements appear to have evolved much earlier.5 The material
is variously recorded as being used for making good interior surfaces: walls, loors and
ceilings, as well as a cementitious adhesive for repairing stone tools and for sealing
vessels. It was also used as a mortar for securing stone masonry. The Cypriote Aceramic
Neolithic settlement at Kalavasos-Tenta in southern Cyprus, for example, dated to the
7th millennium BCE, appears to have utilized gypsum mortars in stone-built domestic
structures,6 and until even the early 20th century CE gypsum use survived in Cyprus as
a traditional water-prooing stucco over mud-brick domestic structures. 7 Early analyses
of Egyptian pyramid materials8 and more recent pyramid explorations9 have revealed that
gypsum mortar was extensively used in grouting rougher-cut internal pyramid blocks and
even to cement rubble in the upper reaches of the pyramids. It is also the likely cementing
component of rubble ramps that appear to have been used for at least some pyramid
material transport, and wet gypsum plaster is suggested to have been used additionally as
a lubricant for manoeuvering large masonry objects.10
Much later, gypsum was widely used as the principal mortar in the building of domed
mosques during the 13th-15th century CE Mamluk sultanate in Egypt.11 It likewise appears
to have been used as the mortar in a rapid method of constructing thin-tiled vaulting,
irst documented in late 14th-century Spain, capitalizing on its substantial compressive
strength and rapid setting.12 The latter novel approach to construction of lightweight
vaults, domes and staircases had spread to France by the 18th century and was notably
revived by the early 20th-century Valencian architect and builder Rafael Guastavino.
After emigrating to the United States, Guastavino contributed this unique (and largely

4 B. aston - J.A. HaRRell - I. sHaW, ‘Stone’, in P. nicHolson - i. sHaW (ed.), Ancient Egyptian
Materials and Technology, Cambridge University Press, 2000, Chapter 2, 5-77; A. lucas, Ancient
Egyptian Materials and Industries, 4th ed., Dover Publications, Mineola, NY, 1989.
5 GouRdin-KinGeRy, cit. n. 1.
6 Dr. maRia PHiloKyPRou, PhD Thesis, University of Cyprus.
7 The longstanding use of gypsum from the Neolithic to modern times on Cyprus (as well as
early use of pozzolanic mortars) was kindly pointed out by Dr. George Constantinou, former director
of the Cyprus Geological Survey.
8 lucas, cit. n. 4
9 M. leHneR, The Complete Pyramids: Solving the Ancient Mysteries, Libri, New York, 2008.
10 lucas, cit. n. 4
11 B. ciPRiani, Development of Construction Techniques in the Mamluk Domes of Cairo, SM
Thesis, Massachusetts Institute of Technology, 2005; B. ciPRiani - J. ocHsendoRf, ‘Construction
techniques in medieval Cairo: The domes of the Mamluk mausolea’, in G. mocHi (ed.), Proc. Conf.
on Theory and Practice of Construction: Knowledge, Means and Models, Ravenna, Italy, 2006,
575-587.
12 J.A. ocHsendoRf, Guastavino Vaulting: The Art of Structural Tile, Princeton Architectural
Press, Princeton, NJ, 2009.
Comm. Hum. Litt. Vol. 128 39

unappreciated) technology to scores of America’s most iconic buildings, including Grand

Central Terminal, Pennsylvania Station and Cathedral of St. John the Devine in New York,
Boston Public Library, and the Biltmore Estate in North Carolina.13 Other longstanding
uses of gypsum include decorative plaster mouldings and, importantly, as a ground (gesso)
for panel paintings and gold leaf.
The attractions of gypsum cements and plasters are very low processing temperatures,
considerable strength, and ready availability (for example, in Egypt, from the Nile Delta
and evaporite deposits elsewhere). Gypsum rock is hydrated calcium sulfate CaSO4•2H2O,
crystallizing as an evaporite out of fresh and sea waters and also from hot springs where
it is found in combination with free sulfur. The crystal structure of gypsum is a layer
structure (Fig. 2a), comprising sheets of covalently-bonded [SO4] tetrahedra bound by
oxygen-sharing to adjacent sheets of ionically-bonded [CaO6] octahedra in which the
oxygen has been partially substituted by OH– ions. Bonding across sheets occurs through
weak hydrogen bonding. This structural anisotropy is relected in a preferred crystal growth
mode, which results in elongated lat laths (often twinned) and ibrous morphologies.
When the stable hydrate gypsum is heated to modest temperatures, it breaks down via
a series of hydration states, eventually to form the most stable (anhydrous) form anhydrite
b-CaSO4. A subhydrate phase, bassanite CaSO4•xH2O, with x = 0.8-0.5, has been identiied
as an intermediate stage of dehydration;14 the name hemihydrate is generally applied to
the composition CaSO4•0.5H2O. The hydrate, hemihydrate and anhydrite compositions
exhibit slightly different crystal structures, distinguishable by X-ray diffraction (Fig. 3).
The subhydrates are formed during heating to temperatures from 70-90 °C. Gypsum-
plasters may be formed by calcination at a minimum temperature of 90 °C, but they are

5 μm

(a) (b)
Figure 2. Structure of gypsum cement a) Layer crystal structure of fully-hydrated gypsum,
CaSO4•2H2O, showing [SO4] tetrahedra, [CaO6] octahedra, OH– substitution for O2–, and capacity for
hydrogen bonding across adjacent layers (after R.c. eVans, Crystal Chemistry, 2nd ed., Cambridge
Univ. Press, 1964). b) Lath crystallites of CaSO4•2H2O hydrate in a set gypsum mortar one day
after hydration (examined wet in an environmental scanning electron microscope, source: Linn W.
Hobbs).
13 ocHsendoRf, cit. n. 12
14 C. bezou - A. nonat - J-C. mutin - A. nøRlund-cHRistensen - N.S. leHman, ‘Investigation of
the crystal structure of gamma-CaSO4, CaSO4•0.5H2O and CaSO4•0.6H2O by powder diffraction,’
Journal of Solid State Chemistry 117 (1995) 165-176.
40 Hobbs - siddall

(a) Gypsum rock

(b) Hemihydrate gypsum

Figure 3. X-ray diffraction spectra (Cu Ka radiation) from a) hydrated gypsum rock (peaks labeled
“1” correspond to CaSO4•2H2O), and b) hemihydrate from gypsum rock dehydrated 6 h at 160 ˚C
(peaks labeled “2” correspond to CaSO4•0.5H2O).

more typically produced by heating in the temperature range 120-165 °C to guarantee

formation of the hemihydrate,
CaSO4•2H2O = CaSO4•0.5H2O + 1.5H2O (1)
The loss of water removes the hydrogen bonding across the structural layers, and the
monolithic rock then fractures easily along the layer interfaces and is easily ground to a
ine powder. In fact, fully-hydrated gypsum rock is much harder to crush mechanically
than limestone, so the dehydration effect is quite dramatic.
Mixed with sand and water, the ground subhydrate becomes a strong, rapid-setting
mortar or plaster, setting within ifteen minutes or less. Because only extremely modest
pyrotechnologies need be employed – dung ires, or perhaps only solar heating – gypsum
cements were an attractive alternative to lime cements for fuel-poor Egypt. Heating gypsum
to temperatures well in excess of 200 ˚C results in loss of all water content and the formation
of the stable anhydrite b-CaSO4, but the only cementitious consequence is inconveniently
more rapid setting, and the extra heat energy expended confers no advantage. Non-industrial
manufacture of gypsum plasters using highly variable pyrotechnologies often results in a
range of subhydrates present, indicating incomplete calcination or slaking reactions. The
setting reaction is one of simple rehydration to CaSO4•2H2O, reversing reaction (1) and
rendered particularly rapid because of the high internal surface area of the ground-up layer
structure and easy diffusion paths. In rehydrating, the ine needle-like lath crystallites of
the ground-up hemihydrate re-cement to each other and interlock mechanically (Fig. 2b),
entrapping sand (or other) temper inclusions. The resulting iber composite morphology
is particularly fracture-resistant in comparison to the equiaxed lime cement morphology
(§4) and contributes not only toughness but surprising compressive strength (> 70 MPa).
Comm. Hum. Litt. Vol. 128 41

Hemihydrate gypsum cement forms the basis for modern wall plasters and plasterboard,
whose persistence to today still derives from their rapid setting, surface inish and strength.
The dehydration reaction is endothermic, so the plasters additionally confer a degree of ire
resistance before they lose strength from dehydration. A large gypsum deposit mined from
the hill of Montmartre, at the edge of Paris, has given its name to the modern commercial
product, plaster of Paris. The primary drawback of gypsum cements and plasters is that
they are water-permeable and degrade rapidly in wet conditions (facts well known to most
homeowners). Hence, while they are adequate for indoor applications or dry climates like
Egypt’s, they cannot be used long-term outdoors in wetter climates. Gypsum plasters have
long been used as grounds for wall paintings in many societies and, as mentioned earlier,
are well known as gessos in Renaissance panel paintings. Gesso undercoats (gesso grosso)
have been shown to contain predominantly ground anhydrite, whereas the surface coats
(gesso sottile) are shown to be typical hemihydrate plasters of Paris.15

4. lime moRtaRs and PlasteRs

Limestone is a generic term applied to a wide range of sedimentary rock types
predominantly composed of the minerals calcite and its polymorph aragonite (both
CaCO3) and dolomite (Mg,Ca)(CO3)2. Many of these carbonate rocks are biogenic and
form in marine conditions at depths no greater than 4500-5000 m. Freshwater limestones
are well known as speleothems (cave deposits) and are also associated with warm-water
springs. Many limestones are relatively pure, although they can frequently contain
varying proportions of detrital material, predominantly quartz. Iniltration of other ions
into the porous sediment results in crystallization of additional minerals (e.g. clays and
feldspathoids), under pressure within the sediment, that cement together the calcite grains
(a process known geologically as diagenesis16) to form a compact composite rock. Marbles
are common as metamorphosed (recrystallized) limestones, although their dense, low
porosity fabric renders them relatively poor lime sources unless they are inely crushed
prior to calcination. Igneous carbonate rocks (carbonatites) are extremely rare, though
locally abundant in the central East African Rift. In addition, calcium carbonate forms
from non-geological accumulations of coral and shells (of both marine and freshwater
molluscs and avian eggshells).
Calcium carbonate crystallizes in three polymorphs: calcite (trigonal), aragonite
(orthorhombic), and the geologically rarer vaterite (hexagonal). The atomic structure
of calcite may be compared to a distorted rocksalt (NaCl) structure, with Ca2+ cations
occupying Na+ sites and planar CO32– molecular ions the Cl– sites (Fig. 4a). Anhydrous
CaCO3 is always crystalline, but biogenic calcium carbonate incorporating hydrogen can

15 N. eastauGH - V. WalsH - T. cHaPlin - R. siddall, Pigment Compendium: A Dictionary of

Historical Pigments, Elsevier Butterworths-Heinemann, Oxford, 2004.
16 m.e. tucKee - V.P. WRiGHt - J.A.D. dicKson, Carbonate Sedimentology, Blackwell Science,
Oxford, 1996.
42 Hobbs - siddall

Ca O
(a) (b) (c)
Figure 4. Crystal structures for a) calcite CaCO3, b) lime calcine CaO, and c) slaked lime Ca(OH)2
(after R.c. eVans, Crystal Chemistry, 2nd ed., Cambridge Univ. Press, 1964).

initially nucleate as an amorphous solid17 because of relaxed topological constraints.18

A large range of sea animals, from microscopic plankton and crustacea to macroscopic
mollusks, form exoskeletons from mineralization of extracellular matrix by combining
Ca2+ ions from dissolved calcium salts with CO32– ions from dissolved CO2 to form highly
insoluble calcium carbonate mineral, mostly in the calcite and aragonite polymorphs.
Lime cements are formed from the decomposition of calcium carbonate, usually
from limestone rock but occasionally from collected mollusk shells. Calcium carbonate
dissociates into CaO (lime) and evolved CO2 when heated to temperatures above 800 ˚C,
in a process known as calcination,
CaCO3 = CaO + CO2h (2)
Centuries of lime burning from Neolithic through to Mediaeval times were responsible
for considerable deforestation, owing to the large quantities of wood fuel used in lime
kilns. Dolomitic limestones can be calcined at much lower temperatures, as low as 400 ˚C
in some cases. The resulting lime (CaO) and periclase (MgO) exhibit only limited solid
solubility19 and therefore evolve separately in lime cements.
The Ca2+ and O2– ions in lime (i.e. calcined limestone) arrange themselves in the
cubic rocksalt structure (Fig. 4b). Lime reacts highly exothermically with water to form
Ca(OH)2, or slaked lime, which crystallizes in a very different layer structure characterized
by hydrogen bonding across layers (Fig. 4c),
CaO + H2O = Ca(OH)2 (3a)

17 L. addadi - S. Raz - S. WeineR, ‘Taking advantage of disorder: Amorphous calcium carbonate

and its roles in biomineralization’, Advanced Materials15 (2003) 959-970.
18 L.W. Hobbs, ‘Topological approaches to the structure of crystalline and amorphous atom
assemblies’, in J.J. noVoa - D. bRaGa - L. addadi (ed.), Engineering of Crystalline Materials
Properties: State of the Art in Modeling, Design and Applications, Springer, Dordrecht & London,
2008, 193-230.
19 G. cedeR - P.D. tePescH - A.F. KoHan - A. Van deR Ven, ‘A model to predict ionic disorder
and phase diagrams: Application to CaO-MgO, Gd2O3-ZrO2, and to sodium b’’-alumina’, Journal of
Electroceramics 1 (1997) 15-26.
Comm. Hum. Litt. Vol. 128 43

a) Limestone

b) Calcined limestone

Figure 5. X-ray diffraction spectra (Cu Ka radiation) from a) limestone (peaks labeled “2” correspond
to calcite CaCO3) and b) limestone calcined 24 hours at 1000 ˚C (labeled peaks correspond to lime
CaO).

The slaking of MgO from calcined dolomitic limestones to form Mg(OH)2 (brucite) is
analogous. Sand and water are added to slaked lime to form a plastic lime mortar that sets
by reabsorbing atmospheric CO2 over weeks or months to reform CaCO3,
Ca(OH)2 + CO2 = CaCO3 + H2Oh (3b)
Lime mortars and plasters therefore function as “portable” limestone. The phase changes
are easily followed by X-ray diffraction (Fig. 5). The looser layer structure of slaked lime
is crucial in assisting the diffusion of gaseous CO2 into the solid. The evolution of water in
the recarbonation reaction (3b) is particularly notable, because it indicates that lime will
not set under water (i.e. it is not a hydraulic mortar): the presence of excess water during
setting would drive reaction (3b) in the
reverse direction.
The slow setting of lime cements
by recarbonation is inconvenient for
building technologies, and they are
generally reserved for application as
mortars between stones or brick courses,
where they remain under compressive
10 μm
loading during and after setting. The
set microstructure is not very strong
(~1 MPa), comprising loosely adherent
polycrystalline grains of micrometer Figure 6. Micromorphology of set lime mortar
dimension (Fig. 6). The nearly cubic four weeks after slaking (examined wet in an
environmental scanning electron microscope,
structure of calcite does not contribute source: Linn W. Hobbs).
44 Hobbs - siddall

strongly anisotropic growth directions, like those found in gypsum, so the grains in the set
mortar retain isotropic morphologies, with little mechanical interlock potential. Set lime
mortars are resistant to neutral water, but erode rapidly in acidic conditions, for example
those present in acid rain – an important consideration when evaluating whether lime
mortars were used in ancient stone constructions.
Limes have been in use since the early Neolithic period, with earliest records from
pre-pottery Neolithic B (PPNB) Levant archaeological sites, where the material was used
for a range of applications including constructing vessels, as decorative mouldings on
human skulls, for making-good loors, hearths and walls in domestic interiors, and also
as adhesives.20 Limes were used extensively in civic and domestic Greek architecture,
especially as linings for cisterns, baths and fountains, for setting mosaic loors and, again,
for making-good loors and walls Lime cements and their related materials are unknown
in Egypt until the Ptolemaic period. During the Roman domination of the Mediterranean,
non-hydraulic limes continued to be used in many small-scale projects, notably for setting
mosaics and for supports and grounds for wall paintings.21 By the Mediaeval period in
Europe, non-hydraulic limes were widely employed in many architectural and decorative
projects, and a well-organized industry is recorded22 and frequently shown in illustrations
of the time. Outside the Mediterranean region, limes were known to many other societies,
although they have not been the subject of major research. Exceptions include studies of
Mayan limes in Mesoamerica.23 The manufacture of lime cements continues on a small
scale in many developing countries,24 where future ethnographic studies have potential to
shed much light on pre-industrial lime production and use. Lime mortars have also been
used extensively as grounds and supports in fresco and secco painting techniques, from the
Roman period to the present day. The extensive use of agricultural limes from many periods
and societies should also not be overlooked in studies of lime manufacturing technologies.
An advantage of lime mortars (and, by extension, cements —such as Roman pozzolanic
cements—containing a lime constituent) for the archaeologist is the exciting prospect of
reliable radiocarbon-dating, using accelerator mass spectrometry (AMS) methods applied
to the 14C content of the atmospheric CO2 uptake to form CaCO3 during setting. While
signiicant caveats still attend this application to mortars, extensive comparisons of
mortar dates with dendrochronologies and 14C dates for accompanying wood architectural

20 GouRdin-KinGeRy, cit. n. 1; O. baR-yosef, ‘The Natuian in the Southern Levant’, in T.

cuyleR-younG - P.E.L. smitH - P. moRtensen (ed.), The Hilly Flanks and Beyond: Studies in Ancient
Oriental Civilisations, vol. 36, University of Chicago Press, Chicago, IL, 1983, 11-42.
21 adam, cit. n. 2
22 H.m. colVin, Building Accounts of King Henry III, Oxford University Press, 1971; n. daVey,
A History of Building Materials, Phoenix House, London, 1961.
23 I. VillaseñoR - C.A. PRice, ‘Technology and decay of magnesian lime plasters: the sculptures
of the funerary crypt of Palenque, Mexico’, Journal of Archaeological Science 35 (2008) 1030-
1039; M.I. VillaseñoR-alonso, Lowland Maya Lime Plaster Technology: A Diachronic Approach,
PhD Thesis, Institute of Archaeology, Unversity College London, University of London, 2008.
24 M. WinGate, Small-scale Lime Burning: A Practical Introduction, Intermediate Technology
Publications, London, 1985.
Comm. Hum. Litt. Vol. 128 45

components and wood fragments in mediaeval Scandinavian churches have resulted in

> 90% correlations and dating certainties to better than 200 years in that period.25

5. silica-based Pozzolanic cements

Oxygen and silicon are the two most abundant elements in the Earth’s crust, and the
overwhelming majority of mineral species are silicates. Pure silica (SiO2) is encountered
geologically most frequently as the mineral quartz, an easily accessible resource widely
distributed in beach and desert sands. Quartz – in the form of sand – is a common, even
principal ingredient of almost all cementitious mortars because of its universal abundance
and availability. However, quartzite sand features merely as bystander in cementitious
reactions, serving as an inert temper to reduce volume fraction of the more costly (in terms
of pyroprocessing) cementitious agents and to improve mechanical properties.
Almost all silica forms (except for a super-dense phase found in meteoritic impact
zones), together with a majority of silicate minerals, are based on atomic arrangements
of covalently-bonded [SiO4] tetrahedra that bond to each other by sharing their vertex
oxygens. This bonding motif generates a half-dozen compact polymorph topologies for
crystalline and amorphous silicas26 and a larger number of less compact framework cage
structures redolent of geodesic dome connectivities. The tetrahedron-tetrahedron vertex
sharing represents a kind of inorganic polymerization reaction. Polymerization in one,
two or three dimensions is respectively responsible for chain, sheet and network silicate
structures. If all four oxygen atoms are shared with other tetrahedra, the connectivity is
designated Q4; if only three of the four vertex oxygens are connected in common, the result
is Q3 (as in planar clay mineral structures). Q2 connectivities lead to chain polymerization,
and Q1 to pairs of linked tetrahedra. Q0 units (effectively [SiO4]4– molecular ions)
characterize orthosilicates (such as ZrSiO4) in which isolated [SiO4] tetrahedra do not
share oxygens with other [SiO4] tetrahedra and are bound more ionically to constituent
cations. Pure silicas are comprised of fully-polymerized Q4 units in three-dimensional
networks.
Silicas are only vanishingly soluble in water solutions over a wide range of pH, which
is why they are used by some organisms (notably, plants and diatoms) as a mineralizing
agent for internal or external skeletal or defensive functions and for several millennia as
manufactured glass containers in the domestic consumer market and chemical industry.
SiO2 is not, however, completely unreactive – especially in inely divided form with large

25 J.R. Hale, ‘Beyond the Eternal City: Roman Concrete in Global Perspective’, in these
proceedings; Å. RinGbom - J. HeinemeieR - A. LindRoos - F. BRocK, ‘Mortar dating and Roman
pozzolana – results and interpretations’, in these proceedings; A. LindRoos - J. HeinemeieR - Å.
RinGbom - F. BRocK - P. SoncK-Koota - M. PeHKonen - J. SuKsi, ‘Problems in radiocarbon dating of
Roman pozzolana mortars’, in these proceedings.
26 L.W. Hobbs - C.E. JesuRum - V. Pulim - B. beRGeR, ‘Local topology of silica networks’,
Philosophical Magazine A 78 (1998) 679-711; L.W. Hobbs - C.E. JesuRum - B. beRGeR, ‘The
topology of silica networks’, Chapter 1, in J.-P. duRaud - R.A.B. deVine - E. dooRyHee (ed.),
Structure and Imperfections in Amorphous and Crystalline Silica, John Wiley & Sons, London,
2000, 1-47.
46 Hobbs - siddall

surface area and in its non-crystalline (“amorphous”) polymorphs. Silicas are especially
reactive in highly alkaline environments (which is why one doesn’t put ine “crystal”
glasses into dishwashers with basic dishwashing detergents) and combine with alkali (Na,
K) and alkaline earth (Ca, Mg) elements to form a range of crystalline alkali silicates and
low-melting glasses. The reactions are the origin of the well-known “silicate garden”27
experiments of childhood, where addition of a soluble Ca salt to a solution of soluble silica
(such as “water glass,” a sodium silicate solution) results in precipitation of an insoluble
hydrated calcium silicate polymerized gel that hardens. The product, dubbed “C-S-H”
because of its still-uncertain structure and composition, is the principal cementing phase in
modern Portland cement (§6)28 and probably Roman concrete (§5.2).29 (Cement chemists
traditionally use the shorthand notation C = CaO, S = SiO2, H = H2O, A = Al2O3, etc.
which conveniently tracks chemistries and stochiometries in cement reactions, C-S-H thus
representing some indeterminate form of a hydrated calcium silicate.)
Pozzolanic cements are, by the narrowest deinition of the term, those based on inely-
divided highly-porous aluminosilicate ash powders derived from pyroclastic volcanic
lows.30 The name originates from the Roman village of Puteoli (modern Pozzuoli) –
located across the Bay of Naples from the frequenly erupting Mt. Vesuvius volcano, but
suficiently distant to be unaffected by Vesuvian pyroclastic lows, such as the eruption
of 79 CE (documented by Pliny the
Younger) that destroyed Pompei and
Herculaneum. Pozzuoli lies instead
on the Campi Flegrei volcanic ield,
and the Pozzuoli tuffs (Fig. 7) used
extensively as the pozzolana in early
Roman concrete derive from much
earlier eruptions of other volcanos in
the early Holocene. A broader reference
for the term includes any inely divided
highly siliceous aluminosilicate
material with large surface area and
consequential chemical reactivity – Figure 7. Campanium yellow tuff deposits, exposed
in an old quarry face near Pozzuoli and typical of
which may include siliceous silts and those described by Vitruvius as pozzolanas (source:
sludges from river beds and estuaries, Ruth Siddall).

27 R.D. coatman - N.L. tHomas - D.D. double, ‘Studies of the growth of “silicate gardens and
related phenomena’, Journal of Materials Science 15 (1980) 2017-26.
28 J.F. younG - S. mindess - D. daRWin, Concrete, Prentice-Hall, New York, 2002.
29 C.A. lanGton - D.M. Roy, ‘Longevity of borehole and shaft sealing materials: Characterization
of ancient cement-based building materials’, Materials Research Society Symposium Proceedings
26 (1984) 543-549; C.A. lanGton - D.M. Roy, ‘Longevity of borehole and shaft sealing materials:
Characterization of cement-based ancient building materials’, Technical Report ONWI-202,
Distributions Category UC-70, National Technical Information Service, U.S. Department of
Commerce, September 1982.
30 R. siddall, ‘The use of volcaniclastic material in Roman hydraulic concretes: A brief review’,
in B. mcGuiRe - D. GRiffitHs - I. steWaRt (ed.), The Archaeology of Geological Catastrophes
(Special Publications 171), Geological Society, London 2000, 339-344.
Comm. Hum. Litt. Vol. 128 47

Table 1. Representative compositions (wt%) of various pozzolanic materials and cement formulations
Cation as Baia Ash Rome Pit Santorine Rhenish Fired Class F 3 Baia Roman 3 Rome Roman Class C Portland
Oxide Sand Earth Trass Clay Flyash Ash: 1 Cement† Pit: 1 Cement‡ Flyash Cement
Lime* Lime* (I)
SiO2 59.5 48.2 63.2 54.6 58.2 47.6 39.8 37.5 36 25-35 36.8 20.5
CaO 2.1 7.5 4 3.8 3.3 4.4 31.5 6.3§ 29.5 15-25 27.8 63.9
Al2O3 19.3 21.9 13.2 16.4 18.4 25.7 12.9 9.8 16.4 9-15 18.1 5.4
Fe2O3 3.3 9.6 4.9 3.8 9.3 17.3 2.2 2 7.2 4-6 6.2 2.6
MgO 0.2 3.2 2.1 1.9 3.9 1 0.2 21.1§ 2.4 2-3 4.7 2.1
Na2O 3.9 5.1 0.7 0.7-2.5 1.2
11.3 4.1 3.9 7.6 3.1 0.6
K2O 2.7 3.9 1.6 1.1-2.2 0.4
SO3 0.2 0.3 0.7 0.4 1.1 1.7 0.1 0.2 0.2-6 1.9 3

*
Ratio by volume: equal bulk densities assumed for both powdered constituents in lieu of any contemporary information. †Cement
matrix composition of Caesarea Maritima breakwater (oleson et al., cit. n. 36). ‡Cement matrix composition from Forum
Romanum and Ostia apartment buildings (lanGton - Roy, cit. n. 29 (1984); lanGton - Roy, cit. n. 29 (1982). §Anomalously
high Mg and low Ca contents reported in Caesarea Maritima cement suggest possible calcination of dolomite as the lime source.

diatomaceous earth (§5.1), ly ash from modern coal-ired power plants, and silica fume.
Table 1 lists comparatively the chemical compositions of a variety of these reactive
pozzolanas.
Volcanic ashes may occur naturally as inely powdered, friable deposits, but others
comprise extremely tough welded glass-rich rocks. The majority of pyroclastic rocks
derive from explosive eruptions of high-viscosity, volatile-rich magmas. Chemistries
vary from highly siliceous granitic magmas to low-silica alkaline magmas. Lavas are also
suitable – and frequently encountered – pozzolanic additives in Roman constructions;
coarse grinding of the fractions of such material makes them readily observable, both
micro- and macroscopically, in the majority of cases. The rapid cooling of these magmas
results in a signiicant amorphous fraction of alkali silicate and silicoaluminate glasses,
with crystalline fractions of quartz SiO2 and alkali and alkaline-earth aluminosilicate
crystalline phases, such as anorthite CaAl2Si2O8 and laboradite NaCa1-3Al3-7Si5-9O16-36 (Fig.
8b). Roman volcanic ield pit-sand pozzolanas are notable for their higher iron contents
(Table 1). Explosive release of dissolved gases while still molten results in a highly porous
pozzolan microstructure (Fig. 8a), with large surface area (~500 m2/g) and consequent
enhanced reactivity.
Flyash from modern coal-ired electric power plants provides a modern source of
pozzolanic powders. In this case, microscopic aluminosilicate inclusions (like clays) in
burning coal melt into small molten droplets, 1-30 mm in size, that are carried by lue
gases up the furnace chimney, where they rapidly solidify into glassy microspheres. The
lyash is collected (usually by electrostatic precipitation) and bagged as a waste product
used as a low-cost pozzolanic addition to Portland cements. The small sphere diameters
(Fig. 9b) provide the equivalent requisite high surface area for chemical reactivity. Flyash
chemical compositions closely match those of Roman pozzolanas (see below and Table
1), and in some classes (Class C lyash, which includes substantial calcium content) to
the chemical compositions of typical 3:1 pozzolan-lime admixtures in Roman cement,
providing an anhydrous shortcut to Roman cement formulations, to which need be added
only water to provide by hydration the same cementing reaction products (§5.2). There is
48 Hobbs - siddall

(a) (b)
Figure 8. Characteristics of Santorine earth volcanic ash-derived pozzolans (source: Herakles Cement
Corporation, Athens, Greece). a) Highly porous microstructure (scanning electron micrograph,).
b) X-ray (Cu Ka radiation) powder-diffraction spectrum of volcanic ash, showing peaks from
principal crystalline constituents and evidence of glassy phases. L = Labradorite, composition
range NaCaAl3Si5O16 to NaCa3Al7Si9O36, A = Anorthite CaAl2Si2O8, Q = Quartz SiO2, Glassy phase
composition largely Na-Ca silicate.

no evidence that lue dust, or even furnace ash (e.g. from metal smelting, pottery iring, or
lime burning), were collected in ancient times to provide a source of pozzolanic material,
however. The furnace temperatures in low-draft furnaces would, in any event, have been
too low to melt inorganic inclusions using only wood or charcoal fuels, except perhaps in
high-draft Chinese cast-iron smelting furnaces.

5.1 Silica-lime cement reactions

Slaked lime (§4) is a convenient source of alkaline environment for silica and results in
the aqueous reaction
Ca2Si2O6•3H2O (crystalline C2SH3)
2Ca(OH)2 + 2SiO2 = (4)
xCaO•ySiO2•zH2O (amorphous C-S-H gel)

Crystalline C2S2H3 is certainly one possible reaction product, but C-S-H gel with a similar
composition (x~2, y~2, z~3) is found in hydration of the anhydrous C2S component in
Portland cement (§6). The Si2O64– “ion” formular representation in (4) actually represents
one-dimensional chain polymerization of [SiO4] tetrahedron units, vertex-sharing two of
their four oxygen atoms (Fig. 10).
Like the calcite product of lime mortar, this polymerized calcium silicate phase
acts as a solid cementitious space iller between temper inclusions and between the
reacting particulates (Figs. 9a, 9b and 11b); but the hydrated calcium silicate gel phase
may importantly undergo an additional morphological transformation to confer the sort
of mechanical strength deriving from the intertwined laths of gypsum hydrate. In this
process, osmotic pressure of the reactants within the gel envelope extrudes hollow ibrils
(Fig. 11a, also evident at an earlier stage in Figs. 9a and 9b) through which water and the
reactants are transported to continue reactions (4). These ibrils intertwine to confer shear
Comm. Hum. Litt. Vol. 128 49

(a) (b)

Figure 9. Microstructures of two pozzolanic mortars (examined wet in an environmental scanning

electron microscope, source: Linn W. Hobbs). a) Santorine-earth pozzolana-lime mortar at four
weeks, showing early development of gel-ibril matrix cementing pozzolana and lime particulates.
b) Early stage of ly-ash mortar setting, showing wide distribution of ash sphere diameters and the
beginning of gel-ibril matrix after one week of setting.

(a) (b)
Figure 10. One-dimensional chain polymerization of [SiO4] tetrahedral units in Ca2Si2O6. The
(Si2O64–) monomer is indicated in (a). A partial sheet of polymerized chains is illustrated in (b). The
Ca2+ ions reside in the interstitial spaces between adjacent chains.

(a) (b)
Figure 11. Hydrated calcium silicate phase morphologies. a) C-S-H gel ibril development at 330
days after hydration of the tricalcium silicate (C3S) phase common to pozzolanic, lime-silica and
Portland cements (scanning electron micrograph reproduced with permission from: H.m. JenninGs -
b.J. dalGleisH - P.l. PRatt, ‘Morphological development of hydrating tricalcium silicate as examined
by electron microscopy technique’, Journal of the American Ceramic Society 64[10] (1981) 567-72,
b) Schematic representation of the replacement of progressively diminishing reacting particulates by
the hydrated calcium silicate cementing phase in pozzolanic cements (source: adapted from H.f.W.
tayloR, ‘Portland cement: hydration products’, in d.m. Roy (ed.), Instructional Modules in Cement
Science, Materials Educational Council, Pennsylvania State University, University Park, PA, 1985, p. 52).
50 Hobbs - siddall

strength and signiicant tensile strength and contribute

increased toughness (resistance to fracture) to the
result cementitious matrix inill (Fig. 11b).
Suitably fast kinetics for reaction (4) obtain only
if there is suficient aqueous solubility of either
reactant and appreciable surface area from which the
components may dissolve or on which the reaction
may proceed. The surface area provided by quartzite
sand is not suficient, so reaction (4) does not
contribute signiicant additional cementing function
to normal lime mortars prepared with sand temper. Figure 12. An example of a diatom
Suficiently large surface areas can be obtained with silica frustule (Azpeitia tabularis,
scanning electron micrograph,
silica fume or (as a geological source) diatomaceous source: University College London
earth. Diatoms are unicellular marine and freshwater Microfossil Image Recovery and
phytoplankton (a major group of eukaryotic Circulation for Learning and
Education, www.ucl.ack.uk/GeolSci/
algae) that construct elaborately lacy exoskeletons micropal/diatom.html, accessed
(frustules, Fig. 12) of hydrated amorphous silica. 2009).
Accumulations are readily available from shoreline
deposits and, more frequently, dry lake beds. Diatomaceous earths require heating (to
remove organic matter) and grinding before being used in cement mixtures. Over
geological time, the amorphous silica frustules recrystallize to form crystalline quartz,
and reactivity decreases. Because of the very ine particle size in diatomaceous earth and
the fact that the frustules are generally entirely consumed in the reactions outlined in (4),
recognition of these additives in historical and archaeological cements is problematic. A
recent study31 of mortars used in buildings and monuments in Pisa, with construction dates
ranging from the 12th to 18th centuries CE (including the “Leaning Tower”), has identiied
a hydraulic component exhibiting C-S-H gel formation in some of the mortars, but without
other remanent evidence for the use of aluminosilicate pozzolans (a very low Al content
was measured in the cementing phase) and has postulated the use of diatomaceous earths
as the reactive component. Diatomaceous earth-dolomitic lime cement has been shown to
set relatively quickly (1 day), and synthetic limestone blocks cast from 93 wt% pulverized
limestone temper (70 mm particle size) and only 7 wt% cementing phase acquire signiicant
compressive strength (5 MPa) within a week.32

5.2 Roman Mortar and Roman Concrete

The extraordinary lowering of Roman architectural forms that deined Roman imperial
cityscapes from the irst century BCE to the ifth century CE is largely attributable to the
application of Roman pozzolanic mortars based on the aqueous reaction of lime with

31 M. fRanzini - L. leoni - M. lezzeRini - F. saRtoRi, ‘On the binder of some ancient mortars’,
Mineralogy and Petrology 67 (1999) 59-69.
32 On-going unpublished studies (2009) by M. baRsoum, Drexel University, and by L.W. Hobbs,
MIT of possible ancient use of castable lime-silica mortars.
Comm. Hum. Litt. Vol. 128 51

volcanic ash and volcanic sands. These volcanic materials are widely distributed around the
Mediterranean region, notably including the Bay of Naples (Baiae), the Roman volcanic
ield, and the Santorini archipelago in the Aegean which exploded c.1625 BCE and whose
islands (the largest of which is Thera, Θήρα) today present massive cliffs of ash now mined
for pozzolanic additions to Portland cement. The Roman architectural historian Pollio
Vitruvius, writing in De Architectura published in ~25 BCE, recommended a mixture of
three parts Pozzuoli tuff (“pulvis Puteolanus”) to one part lime to produce a hydraulic
cement “that hardens as well under water as in ordinary buildings.” Analyses of Roman
cement and Baian tuff (Table 1) suggest that the constituent ratio was variable.
Aqueous reaction of slaked lime with pozzolan chemistries, generically represented
as x(Na,K)2O•y(Al,Fe)2O3•zSiO2 or (Na,K)2x(Al,Fe)2ySizOx+3y+2z, results in the hydrated
silicate reaction (4) and additional hydrated crystalline alkali silicate and Ca aluminate/
ferroaluminate phases, e.g.
(3y+2z-2x)Ca(OH)2 + (Na,K)2x(Al,Fe)2ySizOx+3y+2z + (2x+3y-z/2)H2O =
(z/2-x) 2CaO•2SiO2•3H2O (or a-Ca2Si2O6•3H2O C-S-H gel) (5a)
+ x (Na,K)2O•2CaO•2SiO2•3H2O (or a-Na2Ca2Si2O7•3H2O) (5b)
+ y 3CaO•(Al,Fe)2O3•6H2O (5c)

for three anticipated product phases (5a,b and c). Any excess unreacted lime carbonates in
air slowly as well, following reaction (3). The hydrated calcium silicate product phase (5a)
in gel form is likely the most important cementing constituent.
The setting of pozzolanic cements proceeds rather slowly (weeks, typically) because
the reactions (5) are necessarily rate-limited by the arrival of solubilized ions diffusing in
an aqueous medium at ambient temperature. These kinetics were largely determinative
in the evolution of Roman construction methods, particularly the use of (irst) stone
and (later) brick facings as initial space-deining containment forms and the use and
design of wood centering for casting arches, vaults and domes.33 The degree of silica
polymerization in 1800-year old Roman concrete has been investigated,34 together with
some of the crystalline phases evolving over these long time periods, including CaCO3
from carbonation of unreacted slaked lime. The conclusions are that substantial silica
polymerization has occurred, as well as – importantly – signiicant carbonation of
unreacted lime. The much higher strength of fully-set pozzolanic mortars (at least 5 MPa,
ive times that of lime mortars) clearly encouraged their early adoption (by at least the
late 2nd century BCE) and later near-universal application (particularly after the ire in
Neronian Rome in 64 CE), in preference to using lime alone.
Pozzolans notably yield hydraulic mortars that are capable of setting in water, a property
not shared by lime mortars (which do not set by hydration) nor by gypsum mortars (which
do). The reason appears to be linked to the production of the C-S-H gel phase and its
cementing role, both as space-illing adhesive and as mechanical interlock agent because
of the evolving ibrillar morphology extruded by osmotic pressure (Fig. 11). Pozzolanic

33 lecHtman - Hobbs, cit. n. 2; adam, cit. n. 2; L. lancasteR, Concrete Vaulted Construction in

Imperial Rome, Cambridge University Press, 2005, 58-59 (esp.).
34 lanGton - Roy, cit. n. 29 (1984); lanGton - Roy, cit. n. 29 (1982).
52 Hobbs - siddall

mortars must, in fact, be kept at least damp during the setting process to develop maximum
strength. Their hydraulic property appears responsible for their earlier use by the Greeks,
who used them in facing (and perhaps re-facing) ore-washing basins.35 The property was
especially useful to Roman engineers who used pozzolanic mortars in underwater or
riparian applications in dock and harbor construction and other aquatic environments,
such as aqueduct linings. The best documented of these is the massive Caesarea Maritima
breakwater built by Herod the Great off the Palestine seacoast around 25 BCE,36 which
entailed the pouring of Vesuvian pozzolanic mortar into submerged wooden molds or
sunk coffer dams. Anomalously high Mg content of the cement in the breakwater block
concrete suggests possible use of dolomitic limestones or dolomite as the lime source,
with an attractive lower calcination temperature requirement.

5.3 Cocciopesto
Roman construction methods involved use of pozzolanic mortars as concrete, which is a
composite of aggregate (caementa) and mortar. The aggregate served a space-illing and
mechanical function but remained an inert ingredient in the cement, with a major exception
of crushed and ground volcanic rock used as aggregate. Roman mortars employed in
environments where waterprooing was a prerequisite are found to contain sherds of
ground-up pottery or bricks, especially where the site becomes far removed from obvious
volcanic ash sources (for example, in Roman Spain and Britain) and routinely in small-
scale and domestic engineering projects – even in areas, such as at Pompeii, were volcanic
pozzolanas are readily available. Cocciopesto is the modern term for pozzolanic mortar in
which a major constituent is crushed ired pottery. The Romans called this material opus
signinum. There is evidence37 that crushed pottery was used to induce hydraulic properties
in Roman mortars and plasters by at least the 2nd century BCE. This technology continued
in the Mediterranean region following the decline of the (western) Roman Empire and is
well known from Byzantine and Ottoman period architecture,38 where the material became
known as khorasan. The word homra is used in Arabic and surkhi in India, revealing
widespread traditions in the use of these materials. The sherds additionally perform the
role of aggregate in the composite, as well as contributing their own pozzolanic activity.
Pottery and brick are made of ired clays, which are layer-structure hydrated
aluminosilicates (§2) comprised of thin (~10 nm) micrometer-size platelets that are easily
separated by crushing or grinding when the clays are dried or lightly-ired (but not high-
ired). The small platelet size confers large surface area to inely crushed pottery, just as in
diatomaceous silica or volcanic ash, with a chemical composition not far from that of volcanic

35 M. Koui - cH. ftiKos, ‘The ancient Kamirian water storage tank: A proof of concrete technology
and durability for three millenniums’, Materials and Structures 31 (1998) 623-627.
36 J.P. oleson - R.L. HoHlfeldeR - a. Raban - R.L. Vann, ‘The Caesarea ancient harbor
excavation project (C.A.H.E.P.): Preliminary report on the 1980-1983 seasons’, Journal of Field
Archaeology 11 (1984) 281-305.
37 lancasteR, cit. n. 33
38 H. böKe - S. aKKuRt - B. iPeKoGlu - E. uGuRlu, ‘Characteristic of brick used as aggregate in
historic brick-lime mortars and plasters’, Cement and Concrete Research 36 (2006) 1115-1122.
Comm. Hum. Litt. Vol. 128 53

ash. Fired kaolin clay, for example, has the anhydrous composition 2SiO2•Al2O3 (§2), like
a high-silica pozzolan (Table 1); when low-ired, it retains considerable porosity (hence
surface area) and exhibits pozzolanic activity when crushed. The fully dehydroxylated
material is referred to as metakaolinite39 and is highly reactive with alkalis, forming the
basis for the synthesis of zeolites (three-dimensional framework aluminosilicates) and a
class of castable ceramic cements with feldspathoid-like framework structures known as
geopolymers.40 Recent nuclear magnetic resonance (NMR) studies using 29Si show that
heating phyllosilicate (clay) minerals generally to 500-700 ˚C (as in low iring) produces
an amorphous fraction with Q3 [SiO4]-connectivity, and that this a-Q3 fraction reacts
readily with hydrated lime to convert a-Q3 units into crystalline Q2 inosilicate (single- and
double-chain) and Q1 sorosilicate (Si2O7 double tetrahedra unit) connectivities.41
A study of 5th-century CE opus signinum from the church of San Lorenzo in Milan42
has conirmed experimentally that the clay materials used, ired between 600-900 °C, are
strongly pozzolanic. The strength of the reaction increases signiicantly with reduction in
particle size; full reaction may therefore be expected only when material is ground into
particles smaller than 5 µm that are likely to be entirely consumed by the reaction and
thus not identiiable during petrographic analysis. Clays ired to temperatures in excess of
900 °C are unlikely to demonstrate pozzolanic reactivity, because new unreactive phases
such as mullite (3Al2O3•2SiO2) form. Extensive studies of the reactions between lime
and metakaolinite43 have shown that the primary clay mineral kaolinite fully converts
to metakaolinite via dehydration reactions after calcination to 730 °C. In this state it
reacts well with lime and water to produce the crystalline cementitious phase tobermorite
(~Ca5Si6O6(OH)2•4H2O) and a series of calcium aluminate hydrates and calcium aluminate
silicate hydrates, which vary chemically depending on the batching of metakaolinite
to lime. All mixes show a hydraulic reaction, forming mortars with high compressive
strengths after 28 days’ curing. Firing of other common clays, such as montmorillonite
and illite, and even micas, also induces some pozzolanic activity.44 Fired to 750 °C prior
to batching with lime and water, montmorillonite and illite both produce signiicantly
lower cement compressive strengths after 28 days curing than does kaolinite, while

39 G. VaRGa, ‘The structure of kaolinite and metakaolinite’, Épitöanyag 59 (2007) 6-9; A.M.
dunsteR - J.R. PaRsonaGe - M.J.K. tHomas, ‘The pozzolanic reaction of metakaolinite and its
effects on Portland cement hydration’, Journal of Materials Science 28 (1993) 1345-1359.
40 J. daVidoVits, Geopolymer Chemistry and Applications, Institut Géopolymère, San Quentin,
France, 2008.
41 E. zendRi - V. luccHini - G. biscontin - Z.M. moRabito, ‘Interaction between clay and lime in
‘cocciopesto’ mortars: a study by 29Si MAS’, Applied Clay Science 25 (2004) 1-7.
42 G. baRonio - L. binda, ‘Study of the pozzolanicity of some bricks and clays’, Construction
and Building Materials 11 (1997) 41-46.
43 M. muRat, ‘Hydration reaction and hardening of calcined clays and related materials: I.
Preliminary investigation on metakaolinite’, Cement and Concrete Research 13 (1983) 259-266;
M. muRat, ‘Hydration reaction and hardening of calcined clays and related materials: II. Inluence
of mineralogical properties of the raw-kaolinite on the reactivity of metakaolinite’, Cement and
Concrete Research 13 (1983) 511-518.
44 J. ambRoise - M. muRat - J. PéRa, ‘Hydration reaction and hardening of calcined clays and
related materials: III. Extension of the research and general conclusions’, Cement and Concrete
Research 15 (1985) 261-268.
54 Hobbs - siddall

mica-lime-water mixtures produce few cementitious phases. Illite ired to 800 °C yields
higher cementing strength (though not comparable to kaolinite). It is important to note,
within the context of this review, that the pozzolanic additives used in these studies were
mechanically ground and therefore presented a homogeneous and ine grain size, unlike
the very poorly sorted materials encountered in Roman period materials.
The best ceramics to be used as pozzolanic additives to form hydraulic lime cements
therefore appear to be metakaolinite-rich bodies that have been ired between temperatures
of 600-900 °C and then added to slaked lime as inely ground particles. Poorer hydraulic
properties result from additions of ired montmorillonite-rich, illite-rich or mixed clays,
but variations in iring temperature may also have a profound effect on their pozzolanic
properties. In summary, both clay chemistry and iring temperature have direct inluences
on the pozzolanicity of ired clay additions.

6. PoRtland cement
The domed Haghia Sophia temple, built in Constantinople between 532 and 537 CE, is
the last great Roman building based on monolithic construction with pozzolanic concrete.
There is little evidence of traditional Roman pozzolanic mortar use from that date until
the 17th century CE, although cocciopesto continued to be used in Italy well into the
Renaissance, and addition of other mortar components (such as diatomaceous earth) could
have contributed hydraulic properties. Lime mortars were the mortars of the great 12th-
15th century European cathedrals and of 16th- and 17th-century European brickwork, while
gypsum mortars were preferentially employed in 13th-15th century Mamluk Egypt45 and
in southern European thin-tile vaulting.46 Nevertheless, some institutional memory of the
hydraulic properties of pozzolans in lime mortars must have remained, or been preserved
in monastic copies of the manuscripts of classical antiquity, that was revisited in the late
Renaissance. Manuscripts of Vitruvius were rediscovered in 1414 by Florentine humanist
and historian Poggio Bracciolini, and their description of Roman building method could
conceivably have led to deployment of pozzolanic cement in construction of the bridge
piers for the Pont de Notre Dame in Paris, laid in 1489 – the irst suspected use of
hydraulic mortars since Roman times. In the 17th Century, the Dutch – in lining their
canals and dykes used to reclaim land from the sea – had experimented with the hydraulic
properties of yet another material source exhibiting pozzolana-like activity: the silts of the
Rhine river. These inely-divided silts, variously labeled “trass,” “tarras” or “terras,” have
much the same aluminosilicate chemistry as pozzolans (Table 1) and present signiicant
surface area for enhanced reactivity with lime. Their use was superseded in late 18th-
century Europe by hydraulic co-calcined limestone-clay mixtures that led the way to
modern Portland cements in the mid-19th century.
Portland cement is produced by co-calcining a mixture of limestone and a clay or shale
rich in alumina and silica. The processing requires high temperatures of ~1500 °C, which

45 ciPRiani, cit. n. 11; ciPRiani - ocHsendoRf, cit. n. 11

46 ocHsendoRf, cit. n. 12
Comm. Hum. Litt. Vol. 128 55

are realistically attained and maintained only with the use of fossil fuels.47 The sintered
product of the resulting solid-state reaction is called Portland cement clinker, which is then
ground to a ine powder that is slaked before use. Powdered gypsum may be added, which
has the property of retarding the set.
This second revolution in hydraulic cement production was initiated by the British
engineer John Smeaton FRS, who was responsible for the construction of the Eddystone
Lighthouse, off the Cornish coast, in the second half of the 18th Century.48 Smeaton
developed hydraulic sets in calcined argillaceous (Al-containing) limestones and then
mixed these with imported pozzolanas, but his results were not published until after his
death in 1792. John Gilbert (1724-95), surveyor and engineer to Francis Egerton, 3rd Duke
of Bridgewater, subsequently utilized co-“burnt” lias lime and clay in pioneering canal
and bridge constructions throughout northern England. The clay was typically dredged
from estuaries, perhaps based on the earlier experiences of the Dutch, though it is not
known when deliberate co-calcination of limestone-clay or lime-clay mixtures was irst
employed. In some cases, certain clayey limestone rock formations (known as “cement
rock”) were discovered in both Europe and North America to produce such cements
when calcined.49 Naturally processed clay-lime cements even exist from geological high-
temperature/low-pressure metamorphism of co-located marls (clay biomicrites, containing
about 20% clay) and limestones.50 In 1796, James Parker patented a hydraulic cement51
(then known as Parker’s cement or “Roman cement” – a clear allusion to a contemporary
appreciation for the Roman hydraulic mortars of two millennia earlier), made by burning
nodules of clay or “septuaria” found on the Kentish coast and later dredged from estuaries.
That was extensively used in riparian construction in the irst third of the 19th century, e.g.
reconstruction of the West India docks in London in 1800. A lightly-calcined clay-lime
mixture cement powder was patented in 1811 (by John Frost of Swanscombe, Kent), but
it was the French chemist Louis Vicat52 who between 1818 and 1828 irst isolated (to the
silica component) the origin of the cementitious properties of these co-calcined clay-lime
mixtures.
“Portland” cement was patented by Joseph Aspdin, a Leeds bricklayer, in 182453 and
was made by co-calcining certain mixtures of limestone and clay that, when rehydrated
and set, resembled the appearance of the most expensive limestone building stone of the

47 younG et al., cit. n. 28

48 a.J. fRancis, The Cement Industry 1796-1914: A History, David & Charles, London, 1977.
49 fRancis, cit. n. 48
50 H.N. KHouRy - E. salameH - I.D. claRK - P. fRitz - W. baJJali - A.E. milodoWsKi - M.R. caVe
- W.R. alexandeR, ‘A natural analogue of high pH cement pore waters from the Maqarin area of
northern Jordan I: Introduction to the site’, Journal of Geochemical Exploration 46 (1992) 117-13;
W.R. alexandeR - R. dayal - K. eaGleson - J. eiKenbeRG - E. Hamilton - C.M. linKlateR - I.G.
mcKinley - C.J. tWeed, ‘A natural analogue of high pH cement pore waters from the Maqarin
area of northern Jordan II: Results of predictive geochemical calculations’, Journal of Geochemical
Exploration 46 (1992) 133-146.
51 Rev. Dr. James PaRKeR, Northleet, Kent, A Certain Cement or Terras to be Used in Aquatic
and Other Buildings and Stucco Work, English Patent No. 2120, London, 27 July 1796.
52 L. Vicat, Mortier et Ciments Calcaires, Paris, 1828.
53 J. asPdin, Leeds, Yorkshire, An Improvement in the Mode of Producing an Artiicial Stone,
English Patent No. 5022, London, 21 October 1824.
56 Hobbs - siddall

era, Portland stone. This stone was quarried from the isle of Portland (hence the name),
lying off the Dorset coast, and used extensively by Sir Christopher Wren and successor
architects to build much of 17th-and 18th-century London. Portland cement found early
application in stucco (to emulate Portland stone) and in aquatic applications such as sills
and cisterns. Nevertheless, the material took some time to really take a hold within the
engineering and construction industry. A noteworthy early structural use of these new
hydraulic cements was as the brick mortar chosen by Marc Isambard Brunel (son of
renowned bridge-builder Isambard Kingdom Brunel) in his construction of the irst tunnel
under the Thames, which was begun in 1825 but not completed until 1843 after several
structural failures.
The principal reaction products of these co-calcinations – viz dicalcium silicate
2CaO•SiO2 (C2S), tricalcium silicate 3CaO•SiO2 and tricalcium aluminate 3CaO•Al2O3
(C3A) or aluminoferrites (C3AF) – are similar to those of set Roman mortars, except that the
products are anhydrous instead of hydrated, the result of solid-state reactions at elevated
temperature mediated by solid-state diffusion, instead of room-temperature ion transport
in a water medium. Upon addition of water, the subsequent setting reactions therefore yield
similar inal products as pozzolanic cements in (5) – viz hydrated calcium silicates and
calcium aluminoferrites – but comprise only simple hydration reactions of the anhydrous
precursors. The solid-state reaction products of co-calcining are best appreciated by
reference to the CaO-SiO2 and CaO-Al2O3 phase diagrams, the high-temperature portion
of the irst of which is reproduced in Fig. 13. The low co-calcining temperatures of early
Portland cements (likely well below 1200 ˚C) would have resulted in 2CaO•SiO2 (C2S)
silicate product and probably little 3CaO•Al2O3 (C3A) aluminate product. C2S hydrates to
the C-S-H gel much more slowly than does
the C3S product that results from modern
high processing temperatures above 1400
˚C; but C2S is, today even, preferred for
large concrete structures, such as dams,
where too rapid heat evolution and poor
thermal conductivity could introduce
unacceptable differential thermal stresses
during setting, leading to cracking. The
calcium aluminate/aluminoferrite phases
(5c) evolved in the hydration of modern
high-ired Portland cements exhibit rod- Figure 13. High-temperature portion of the CaO-
like morphologies (Fig. 14a) that appear SiO2 pseudobinary phase diagram, showing
soon (within hours) after initial hydration formation of 3CaO•SiO2 (C3S) only above 1300
˚C. Lower temperatures used in 18th- and 19th-
and contribute an early initial set that century co-calcining of lime-clay admixtures
makes Portland cements so useful in would have resulted only in 2CaO•SiO2 (C2S)
reaction product, which sets more slowly. (Source:
contemporary building. To retard this d.m. Roy, ‘Portland Cement: Constitution and
initial setting reaction, gypsum (CaSO4) Processing. Part II: Cement Constitution and Kiln
Reactions’, Journal of Educational Modules in
is usually added, resulting in formation Materials Science and Engineering 3[4] (1981)
of calcium aluminosulfates, such as 649).
Comm. Hum. Litt. Vol. 128 57

Figure 14. Microstructures of Portland cement (scanning electron micrographs). a) Ettringite rods
from hydration of the calcium aluminate C3A phase of Portland cement (source: Ruth Siddall). b)
C-S-H gel ibril clusters and ettringite rods in one-day old Portland cement paste (reproduced with
permission from K.L. Scrivener, ‘The microstructure of concrete’, in J.P. Skalny (ed.), Materials
Science of Concrete, American Ceramic Society, Westerville, OH, 1989, 127-161, Fig. 9).

ettringite (6CaO•Al2O3•3SO3•32H2O, or C6AS3H32) that, like the calcium aluminates,

exhibit similar rod-like morphologies. The main set, which occurs over days to weeks,
derives from the morphologically ibrous hydrated calcium silicate gel phase already
described for pozzolanic cements (Fig. 14b).
A wide range of phases can be identiied in both classical 19th century and more modern
Portland cement formulations. Modern Portland cements, which are very inely ground (to
> 300 m2/g surface area), demonstrate an almost complete reaction in the cement paste, and
the ground clinker particles are largely consumed. Petrographic analyses of earlier fully-
set cements54 reveal kernels of unreacted clinker – clinker remnants of 10% remaining for
particle sizes > 600 µm and 45%, with particle sizes between 150-600 µm being typical of
materials made circa 1879. By 1910, considerable improvements in grinding machinery
(principally mechanized jaw crushers)55 reduced clinker particle size, and it is typically
recorded that set ordinary Portland cements post-dating 1910 contain only 10% of clinker
remnants exceeding 90 µm.

7. imPoRtance of analysis
The traditional approach to scientiic examination of archaeological clay ceramic
materials is preparation of thin sections for transmission light microscopy using polarized
light (polarizing light microscopy, or PLM).56 This approach provides diagnostic

54 D.A. st JoHn - A.W. Poole - I. sims, Concrete Petrography: A Handbook of Investigative

Techniques, Arnold, London, 1998.
55 fRancis, cit. n. 48
56 I.K. WHitbRead, ‘A proposal for the systematic description of thin sections towards the study
of ancient ceramic technology’, Archaeometry 31 (1989) 127-138.
58 Hobbs - siddall

information about the mineralogy57 of both intentional and adventitious inclusions that
can be used to source constituent materials; it also provides limited information about
the clay matrix. Thin-section methods can be readily applied, too, to cements, plasters,
mortars and concretes, providing information about both inclusions and, to some extent,
the cement matrix.58 For lime cements, for example, information may be gleaned about
the degree of carbonation, areas of strong hydraulic reactions, the formation of gels and
also, importantly, presence of particles of underburned and unslaked lime that can provide
important information concerning the calcium carbonate source. All of these features can
contribute to our knowledge of the processing and application of cement manufacture.
Individual particles comprising the cement matrix often remain below the resolution of
light microscopes, unfortunately, but a surprising amount of information may be gleaned
nevertheless.59 Ideally, the crystalline phases present should additionally be identiied by
X-ray diffraction (mindful that cementitious product phases are in many instances non-
crystalline).
Scanning (SEM) and transmission (TEM) electron microscopies are usually required
for a more deinitive characterization of matrix reactions. Cement microstructures are
rather easily examined by SEM of fracture sections, with minimal specimen preparation
compared to thin-section production for PLM, though reconstruction simulations involving
hydration may necessitate an environmental SEM that tolerates still-damp specimens.
A more complete identiication would include microchemical analysis using energy-
dispersive or wavelength-dispersive X-ray spectroscopy in SEM or electron microprobe
instruments, chromatography and spectroscopic (e.g. NMR) methods to identify degree
of inorganic polymerization, and analytical TEM of setting-reaction microstructures on
the nanometer scale. While many of these more sophisticated analytical methods may
lie beyond the resources of the ield archaeologist, they are nonetheless indispensable in
properly identifying the cementitious mortar types employed and their source materials,
and particularly in understanding the evolution of cementitious technologies through the
continual processes of experimentation, substitution and discovery.

Acknowledgements
The authors are grateful to Professor Heather N. Lechtman for her careful reading of and
comments on the manuscript; and to the proceedings editors, Professors Åsa Ringbom
and Robert Hohlfelder, for their considerable patience in awaiting its delivery. Dr.
George Constantinou, former director of the Cyprus Geological Survey, and Dr. Maria
Philokyprou, University of Cyprus, contributed illuminating discussions about neolithic
and bronze age mortars usage on Cyprus.

57 A. GlazneR - K. RataJesKi, Atlas of Igneous and Metamorphic Rocks, Minerals and Textures,
University of North Carolina, Chapel Hill, NC, 2002.
58 S. PaVia - S. caRo, ‘An investigation of Roman mortar technology through petrographic
analysis of archeological materials’, Construction and Building Methods 22 (2008) 1807-11.
59 st JoHn et al., cit. n. 54