Content uploaded by Åsa Ringbom
Author content
All content in this area was uploaded by Åsa Ringbom on Jan 12, 2016
Content may be subject to copyright.
214 Lindroos et al.
Problems in radiocarbon dating of
Roman pozzolana mortars
ALf Lindroos – JAn Heinemeier – ÅsA ringbom – fionA brock –
PiA sonck-kootA – miiA PeHkonen – JuHAni suksi
1. introduction
Mortars, including concrete-like Roman pozzolana mortars, from various archaeological
sites have been radiocarbon dated since the 1960s, but often without taking into account
the distinction between hydraulic or non-hydraulic samples.1 The method of radiocarbon
dating of the binder carbonate of mortar has always been known to be problematic, and
hence it has been used with caution. In the case of non-hydraulic mortars the method
has been studied experimentally and the problems are fairly well understood,2 the main
issue relating to the presence of old, radiocarbon-dead material in incompletely calcined
limestone residues and in calcite grains present in commonly used geological mortar
llers such as sand and gravel. Our protocol for dating non-hydraulic mortars has been
presented in several publications.3 We have monitored the effects of the contaminants
using sequential dissolution of mortar with acid under vacuum for up to 16 hours, until the
samples are almost completely dissolved. Successive fractions, or increments, of carbon
1 J. LAbeyrie - g. deLibriAs, ‘Dating of old mortars by the carbon-14 method’, Nature 201
(1964) 742; g. deLibriAs – J. LAbeyrie, ‘The dating of mortars by the carbon-14 method’, in r.m.
cHAtters - e.A. oLson (ed.), Proceedings from the 6th International Conference on 14C and tritium
dating, Washington D.C., Clearinghouse for Fed. Sci. & Tech. Inf., Natural Bur. Standards, U.S.
Dept. Commerce, 1965, 344-347; m. stuiver – c.s. smitH, ‘Radiocarbon dating of ancient mortar
and plaster’, in r.m. cHAtters - e.A. oLson (ed.), Proceedings from the 6th International Conference
on 14C and tritium dating, Washington D.C., Clearinghouse for Fed. Sci. & Tech. Inf., Natural Bur.
Standards, U.S. Dept. Commerce, 1965, 338-343.
2 r.L. foLk – s.Jr. vALAstro, ‘Successful technique for dating of lime mortars by carbon-14’,
Journal of eld Archaeology 3 (1976) 203-208; C. PAcHiAudi – J. mArecHAL – m. vAn strydonck
– m. duPAs - m. dAucHot-deHon, ‘Isotopic fractionation of carbon during CO2 absorption by
mortar’, in M. stuiver – R.S. krA (ed.), Proceedings from the 12th International 14C Conference,
Radiocarbon 28:2A (1986) 691-697; J. Ambers, ‘Stable carbon isotope ratios and their relevance
to the determination of accurate radiocarbon dates for lime mortars’, Journal of Archaeological
Science 14 (1987) 569-576; M. vAn strydonck – m. duPAs, ‘The classication and dating of lime
mortars by chemical analysis and radiocarbon dating: A review’, in W.H. WALdren - J.A. ensenyAt -
r.c. kennArd (ed.), The 2nd Deya International Conference of prehistory, Vol II (BAR International
Series 574) 1991, 5-43.
3 J. Heinemeier – H. Jungner – A. Lindroos – Å. ringbom – t. von konoW – n. rud, ‘AMS 14C
dating of lime mortar’, Nuclear Instruments and Methods in Physics Research B 123 (1997) 487-495;
A. Lindroos, ‘Carbonate Phases in Historical Lime Mortars and Pozzolana Concrete: Implications
for 14C Dating’, Department of Geology and Mineralogy, Åbo Akademi University, Diss., Åbo 2005;
A. Lindroos – Å. ringbom – J. Heinemeier – m. brAskén – A. sveinbJörnsdottir, ‘Mortar dating
using AMS 14C and sequential dissolution: Examples from Medieval, non-hydraulic lime mortars
from the Åland Islands, SW Finland’, Radiocarbon 49:1 (2007) 47-67; Å. ringbom - J. Heinemeier
- A. Lindroos - f. brock, ‘Mortar dating and Roman pozzolana, results and interpretations’, in this
volume.
Comm. Hum. Litt. Vol. 128 215
dioxide (CO2) are collected from each sample and AMS radiocarbon-dated, and the results
are presented as age proles. The theoretical principles required to interpret the 14C proles
have been presented in other publications.4
The Roman pozzolana mortars are considered to be hydraulic, and have received less
attention than non-hydraulic mortars. This is certainly not because they are deemed to be
any less important, but because they have been considered even more difcult to date. It
is known that hydraulic mortars are less permeable to atmospheric CO2, that they contain
less dateable carbonate, and that they are constantly active chemically and can generate
carbonates whenever they are disturbed. Our research group has analyzed a large number
of Roman pozzolana and other hydraulic mortars since 1988 and we have studied the effect
on the radiocarbon age of the presence of young carbonates, such as re-crystallizations
and eforescent growths, in them. Some of the results have previously been published.5
In this article we present our experimental data for hydraulic mortars and evaluate the
effectiveness of our dating method by comparing our results with well-known dates from
brick stamps and other archaeological criteria.
2. HydrAuLic or non-HydrAuLic mortAr?
A mortar is considered hydraulic when a substantial part of the binder minerals within it are
produced by direct reactions between the lime and the ller material (e.g. calcium silicates
and silica gels), or by hydration of other burned minerals (e.g. burned clay minerals).
The mortars are, however, based on slaked lime (calcium hydroxide, Ca(OH)2) and at
some stage they are in contact with atmospheric or dissolved CO2. Therefore dateable
carbonates do form according to reaction (1):
CO2 + Ca(OH)2→CaCO3 + H2O
Other minerals formed by direct reaction between the lime and other constituents dominate
within the mortar. These minerals are usually complex chains of hydroxides, aluminates
and silicates, as well as some un-reacted calcium hydroxide or portlandite. Common
minerals present in mortar are presented in Table 1.
4 See Lindroos and Lindroos et al., cit. n. 3
5 Lindroos, cit. n. 3; J. HALe – J. Heinemeier – L. LAncAster – A. Lindroos – Å. ringbom, ‘Dating
ancient mortar’, American Scientist 91:2 (2003) 130-137; Å. ringbom – J. HALe – J. Heinemeier – A.
Lindroos – f. brock, ‘Mortar dating in Medieval and Classical archaeology’, Construction History
Society Newsletter 73 (2006) 11-18; Å. ringbom – J. HALe – J. Heinemeier – L. LAncAster – A.
Lindroos, ‘When did the Mortar Harden? A new method for dating buildings and other structures
through AMS radiocarbon analysis’, Poster presented at the XVI International Congress of Classical
Archaeology, The Associazione Internationale di Archeologia Classica, Harvard University, August
24, 2003.
216 Lindroos et al.
Table 1. Common concrete minerals according to Lechtman and Hobbs6.
formuLA HAbitus minerAL
Ca3Si2O7 . 3H2Ogel tobermorite
Ca6Al2(SO4)3(OH)12 . 26H2O rods, bers ettringite
CaO . Al2O3 . 6H2Ohydrogarnet
Ca(OH)2plates portlandite
The concrete minerals and calcite dissolve readily in diluted mineral acids and many
organic acids, and it is therefore possible to classify the mortars to some extent using their
partial dissolution chemistry. Van Strydonck et al.7 describe different ways to characterize
historical mortars, including the hydraulic index (HI) which they dened as the ratio of
the soluble non-carbonate components in the binder, to the soluble carbonate component
in the binder:
SiO2 + Al2O3 + Fe2O3
CaO + MgO
It should be noted, however, that CaO is in fact a major component in all common binder
minerals. Samples with a ratio >0.5 are dened as hydraulic. In principle we have adopted
this denition, but some problems are considered:
i. HI is designed to distinguish between lime mortar and modern concrete. In
modern concrete the hydraulic minerals are mostly formed from burned mineral
products and not from reactions with the ller. In ancient mortars the hydraulic
minerals are formed through reactions between a reactive ller and lime. The
reactivity of the ller in turn is dependent on its´ mineralogy, crystallinity,
devitrication and secondary alterations.8 A reactive ller tends to be soluble,
and thus a contaminant in the HI determination. Therefore the results tend to be
biased towards hydraulic classication.
ii. The Rome area is an alkaline volcanic province and potassium is a major
component in the rocks, which comprise of up to 10% K2O.9 It occurs in
6 H.n. LecHtmAn – L.W. Hobbs, ‘Roman concrete and the roman architectural revolution’, in
W.D. kingery (ed.), Ceramics and Civilization, vol III, Westerville OH 1987, 81-128; L.W. Hobbs
- r. siddALL, Cementitious materials of the ancient world, in this volume.
7 m. vAn strydonck – m. duPAs – m. dAucHot-deHon – c. PAcHiAudi – J. mArecHAL, ‘The
inuence of contaminating (fossil) carbonate and the variation of δ13C in mortar dating’, in M.
stuiver – W.H. WALdren, Proceedings of the 12th International 14C conference, Radiocarbon 28
(1986) 702-710.
8 m. JAckson – f. mArrA, ‘Roman stone masonry: Volcanic foundations of the ancient city’,
American Journal of Archaeology 110 (2006) 403-436.
9 E.g. M. gAetA – c. fredA – J.n. cHristensen – L. dALLAi – f. mArrA – d.b. kArner –
P. scArLAto, ’Time-dependent geochemistry of clinopyroxen from the Alban hills (Central Italy):
Clues to the source and evolution of ultrapotassic magmas’, Lithos 86 (2006) 330-346 and references
therein.
Comm. Hum. Litt. Vol. 128 217
refractory K-feldspar and leucite grains,
but also in soluble micas and volcanic
glasses. Sodium is also abundant, but
is less soluble. In Fig. 1, a widespread
volcanic rock unit is presented as a pie
diagram.10 Elements with concentration
values <1% are omitted. The Loss
On Ignition (LOI) component will
be discussed in the next chapter. We
have included K2O in the pie diagrams
describing mortar compositions, but it is
not included in the HI. We do not know
if hydraulic potassium minerals are
formed as a result of reactions with the volcanic glasses, but if ashes are present
in the mortar then potassium hydroxide (KOH) is also present.
iii. Our sample preparation method is not designed to determine HI as accurately
as possible, but rather to isolate dateable carbonate from the sample. In order
to get a representative HI value we use a coarse grain-size fraction, where
the separation is poor. Old mortars are so heterogeneous that it is not usually
possible to take a big enough sample for crushing, grinding and homogenisation
for proper HI determination. However, grain-size fractions >100 µm usually
yield rather uniform HI values, whereas ner fractions are highly enriched
in carbonate and yield low HI values. Fig. 2 shows an example: the <38 µm
fraction yields the value 0.28 and the 100-150 µm fraction yields the value
0.51. The ground 150-300 µm fraction has an HI of 0.54.
10 Values from c. fredA – m. gAetA – d.m. PALLAdino – r. trigiLA, ‘The Villa Senni eruption
(Alban Hills, Central Italy): the role of H2O and CO2 on the magma chamber evolution and on the
eruptive scenario’, Journal of Volcanology and Geothermal Research 78 (1997) 103-120.
Figure 1. Pie diagram representing the
composition of a widespread volcanic
rock from the Rome area: The upper
ow unit of the “Villa Senni” eruption.
Figure 2. The value of the hydraulic index is strongly dependent on the grain-size of the fraction
analyzed. Carbonate-rich ne fractions yield low (non-hydraulic) values.
Torre delle Milizie (Rome 007)
1h in 10% HCl at 20°C
HI: 0.28 HI: 0.51
<38µm fraction 100-150µm fraction
Al2O3
SiO2
K2O
FeO
CaO
MgO
218 Lindroos et al.
Grinding the coarse grain size fraction further (a common laboratory procedure) is not
recommended because it may affect the dissolution which strongly depends on mineralogy.
Iron silicates in particular tend to dissolve so slowly that grinding enhances their solubility
signicantly. The hydraulic indices are therefore only reference values, which should be
used with precaution especially if compared with values obtained by other labs. One
should also consider which part of the sample has been analysed and what it represents.
Is it a bulk sample (and therefore, how representative is it for the whole mortar batch) or
is it a grain-size window of an aliquot of a small sample crushed in a certain way? It is
evident that all samples must be prepared and analyzed the same way to be comparable.
Our hydraulic indices have been measured from the unwashed 150-300 µm grain-size
fractions that have been produced by crushing 50-150 g of the sample with plastic covered
pliers and vibrating the splinter in a sieve series for 20-25 min. The 150-300 µm powders
are leached in 10% (by volume) hydrochloric acid (HCl) at 20oC for 1h and the elemental
concentrations are determined using ICP-AES (at the Geological Survey of Finland and/
or at the Dept. of Analytical Chemistry, Åbo Akademi University). The conversion of the
ICP-AES data to concentrations of oxides within the mortar is a simple stochiometric
calculation.11
It should be noted that the above preparation procedure is developed for 14C dating
(where the washed 46-75 µm grain-size fraction is usually used) and not for bulk
chemistry. Fig. 3 shows the HI plotted against the approximate age of the Roman mortars.
It is evident that the division between non-hydraulic and hydraulic is rather arbitrary and
there is an extensive “grey-zone” composed of mortars that are not clearly one or the
other. It is questionable what purpose the distinction between hydraulic and non-hydraulic
serves in a 14C dating context. Therefore non-hydraulic mortars from the Roman empire
11 P. sonck-kootA – A. Lindroos – J-o. LiLL – J. rAJAnder – e-m. viitAnen – m.H. PeHkonen –
J. suksi – f. mArrA – s-J. HeseLius, ’Characterization of volcanic material used in ancient Roman
mortars using external-beam PIXE’, Nuclear Instruments and Methods in Physics Research B 266
(2008) 2367-2370.
Figure 3. The hydraulic index
of Roman mortar samples (our
data) plotted against the known,
expected or determined age of
the samples. Although HI is the
standard way to characterize a
mortar, the ICP-AES analysis
can easily, and at a low extra
cost, be extended to cover many
trace elements as well. This is
of special interest in the Rome
area because the geochemistry
and mineralogy of the volcanoes
around Rome is rather extreme and
some minerals and geochemical
anomalies can be identied also
in the mortar llers. Preliminary
comparisons with geological
material and provenance studies
have previously been published11
and more detailed studies will be
presented elsewhere.
Comm. Hum. Litt. Vol. 128 219
are included in this study. From a historical perspective, in the Rome-Naples area there
seems to be a diffuse trend from non-hydraulic BC mortars to more hydraulic AD mortars.
However, in Iberia all mortars with a geological ller are non-hydraulic.
3. 14c dAting of mortArs; PrinciPLes And ProbLems
An AMS-based procedure of dating non-hydraulic mortars and interpreting the results is
presented in an earlier publication.12 The same procedure is used for hydraulic mortars, as
described below, but the interpretation of the data is different, as we will discuss shortly.
• The mortar samples to be dated are collected from areas that represented the
surface of a mortared construction when it was built. Prior to sampling the
mortar surface is scraped clean with a chisel and ~ 100 g mortar is collected
using a hammer and chisel. The sample pieces are collected into plastic bags
that are then sealed for transport.
• One piece of mortar is tested for alkalinity using phenolphthalein (alkaline
mortars absorb modern CO2 and yield biased 14C ages). In the lab the pH of the
mortar is tested once more by putting the same piece that had been tested using
phenolphthalein in water or by taking some of the water after wet sieving to
check if it is alkaline.
• The mortar is inspected visually and microscopically and some mortar pieces
are studied using cathodoluminescence (CL). This method usually reveals
natural carbonates from limestone and marble.
• The remaining material is crushed with plastic-covered pliers and the splinters
are vibrated in a sieve series for 20-25 min. This selectively shakes off dateable
soft carbonate binder mineral dust and leaves harder geological material in the
coarse fractions.
• Of the fractions collected during sieving, the 150-300 µm fraction is usually
utilized for chemical characterization and the 46-75 µm fraction for dating. The
most ne-grained materials are seldom dated because they are more difcult
to study and their behaviour is difcult to model since they may either react
violently during acid hydrolysis or remain partly oating on the acid.
Prior to dating the fraction is studied using CL (to check if the sieving has removed the
limestone) and the CO2 content is determined using Loss On Ignition (LOI) at 550-1000oC.
The sample is then reacted with 85% phosphoric acid (H3PO4) under vacuum and the CO2
released is collected in 5 successive increments or fractions representing roughly 20% of
the total gas each. In many cases it is, however, advisable to increase the resolution in the
beginning of the dissolution and collect a rst fraction representing <10% of the CO2. The
process is allowed to continue for at least 16h until the dissolution is nearly complete.
12 Lindroos et al., cit. n. 3, see also ringbom et al., cit. n. 3, in this volume.
220 Lindroos et al.
• In the Aarhus AMS lab each CO2 fraction is divided into two aliquots, one for
dating and one for determination of the δ13C and δ18O stable isotope values
(note that the δ18O values are only used for reference because they are produced
using 85% acid). In the Oxford lab each CO2 fraction is passed through an
elemental analyser and isotope ratio mass spectrometer prior to graphitising the
sample for AMS dating.
• The stable isotopic values measured rarely represent single carbonate phases,
but are instead the weighted average of CO2 from all soluble carbonate phases.
They may, however, still yield important qualitative information.
4. cArbonAte content
Hydraulic mortars are known for their
ability to harden under water without
access to atmospheric CO2. However,
while the mortar is not submerged it does
absorb carbon from the atmosphere in
sufcient amounts to be reected in a 14C
date. We have analyzed a large number of
ancient mortars with a range of hydraulic
properties. Fig. 4 shows some statistics on
the hydraulic mortars and their soluble
carbonate content. The values are the
carbonate-equivalents calculated from
the carbon dioxide yields during the acid
dissolution for dating.
The values presented do not cover the total carbonate inventory of the samples because
the viscous phosphoric acid is not an efcient solvent and it is relatively difcult to
reproduce the same carbon yield in replicate dissolution experiments. In order to measure
the carbonate content properly we used the standard “loss on ignition” (LOI) method
whereby the sample is rst heated to 550oC to remove organic material and water in the
hydroxides before the temperature is raised to 950oC to remove the carbon dioxide. The
weight loss is registered and the carbonate content is calculated assuming that calcium
carbonate is composed of 56% CaO and 44% CO2. Fig. 5 shows a LOI vs acid hydrolysis
plot. In general there are undissolved carbonate residues in the samples and the samples
plot under the 100% dissolution equiline. Especially carbonate rich, non-hydraulic mortars
and pure lime-lumps show irregular dissolution behaviour. Some points plot slightly above
the 100% line, reecting the poor precision of the CO2 yield measurements in H3PO4
hydrolysis.
Figure 4. Calcium carbonate content in Roman
mortars. The calculated carbonate values are based
on the CO2 yield in 85% H3PO4 hydrolysis in 16h.
The lled dot to the left denotes an exceptional
BC mortar from the Jupiter Anxur temple of
Terracina.
Comm. Hum. Litt. Vol. 128 221
5. PArticuLAr ProbLems in tHe dAting of PozzoLAnA mortArs
The general problems in 14C dating of mortars have been discussed to some extent in the
literature.13 The most common problems are the degree of calcination of the burned lime
and the presence of limestone grains in the sand and gravel material used as ller. Our
contribution to this issue has previously been published 14 and these problems will not be
discussed further here. The pozzolana mortars have some specic problems, which are
discussed below.
5.1 Carbonate in the ller
The term pozzolana originates from Vitruvius, who described a volcanic soil, “pulvis
puteolanus”, from the village Puteoli, nowadays Pozzuoli, at the active volcanic Campi
Flegrei district on the northern shore of the Bay of Naples. According to Vitruvius, mixing
this soil with lime would make the mortar hard and durable. In geological terms the deposit
is a loose pyroclastic surge deposit (ne-grained, hot, vitreous material deposited from
horizontal blasts during a violent volcanic eruption). Post-depositional, hydrothermal
alteration (by hot acidic vapours) makes these deposits more reactive as mortar llers.15
Similar volcanic materials are also common in the Rome area, originating from the
Monti Sabatini volcano NW of Rome and the Colli Albani volcano SE of Rome. The
city is founded on the deposits of these volcanoes, many of which have been used as
pozzolana. It is of special signicance for radiocarbon dating of these mortars that both
volcanoes have grown on a Mesozoic limestone basement. This means that ascending
magmas interact with the limestone foundation and calcinate and react with the limestone
13 b. WiLLAime – r. coPPens – r. JAegy, ‘Datation des mortiers du chateau de Chatel-sur Moselle
par le carbone 14’, PACT Journal 8 (1983) 345-9; m. vAn strydonck – m. duPAs – m. dAucHot-
deHon, ‘Radiocarbon Dating of Old Mortars’, PACT Journal 8 (1983) 337-43; Lindroos et al., cit. n.
3; Heinemeier et al., cit. n. 3 ; foLk – vALAstro, cit. n. 2; m.s. bAxter – A. WALton, ‘Radiocarbon
Dating of Mortars’, Nature 225 (1970) 937-938.
14 Lindroos et al., cit. n. 3.
15 JAckson - mArrA, cit. n. 8.
Figure 5. CO2 yield in phosphoric
acid hydrolysis for 14C dating
relative the total CO2 inventory.
Only samples that have been
dissolved for at least 16h are
plotted. The relatively low
carbonate content of the Roman
mortars compared with normally
60-70% for younger non-
hydraulic mortars is a drawback
in radiocarbon dating because
contaminants are omnipresent
and the ratio of binder calcite
to contaminants determines the
suitability of the sample for dating.
The different kind of contaminants
will be described below.
222 Lindroos et al.
before16reaching the surface.17 The processes result in an unusual geochemistry and
mineralogy of the magmas as well as extensive CO2 release during the volcanic eruptions.
The CO2 activity continues long after the actual eruptions and hot CO2 and sulphur dioxide
rich vapours alter the volcanic deposits to suitable mortar llers. The vapours and hot
springs also produce carbonate minerals in the volcanic rocks and even travertine deposits
composed of mainly calcite. Many of the rocks are around half a million years old and
they may contain marine carbonates and carbonates originating from submarine volcanic
activity. The risk of carbonate contamination from the pozzolana ller is therefore high
when dating Roman mortars. We have sampled the pozzolanas in order to study their
geochemistry, mineralogy and in this context their hydroxide and carbonate content. Fig.
6 shows the results of the LOI analyses at 550o and 950o. The volcanics with a high LOI
at 550o (where hydroxide water is removed) are potentially good, reactive pozzolanas
whereas the ones with a high LOI at 950o are potential sources of contamination in 14C
dating of mortars.
In our pozzolana samples we have found the common carbonates calcite, aragonite and
dolomite within voids and cracks in the volcanic rocks. Marble splinters and dust are also
common in the mortars. According to our experience so far all these contaminants dissolve
16 JAckson - mArrA, cit. n. 8; M.E. bLAke, Ancient Roman Construction in Italy from the
Prehistoric to Augustus, Part 1, Washington D.C., Carnegie Institution of Washington 1947, 308-
352; L.C. LAncAster, Concrete Vaulted Construction in Imperial Rome, Cambridge University
Press 2005.
17 E.g. A. rittmAnn, ‘Die geologische bedingte evolution und differentiation des Somma-
Vesuvius magmas’, Zeitschrift für Vulkanologie 15 (1933) 1–2.
Figure 6. LOI (loss on ignition) values for volcanic rocks from Rome (1-5, 7-14, 17-19 and 21),
Pompeii (16, 20), Herculaneum (6) and Pozzuoli (15), showing their content of hydroxides (550°)
and carbonates (950°). The rocks have been used as building material in ancient Rome.16
Comm. Hum. Litt. Vol. 128 223
in acid more slowly than the bulk of the binder carbonates. This means that it is possible to
date mortars that include these geological contaminants, if the CO2 for dating is extracted
at an early stage of the dissolution progress. Two samples from the Colosseum (Fig. 7)
show that if less than 10% (F<0.1) of the sample is dissolved the CO2 evolved will date
the time of construction. One of the three proles from the same sample is quite different,
which is due to experimenting with different dissolution temperatures and pressures and
emptying of the vacuum system between the collection of the individual CO2 fractions.
The experiments show that it is possible to affect the solubility of the contaminants relative
to the solubility of the binder carbonate. However, the rst CO2 evolved still comes from
the binder carbonate.
The mortar from the temple of Jupiter
Anxur, Terracina, from the republican
time shows a similar 14C prole with
contaminants that release CO2 relatively
early during dissolution but it still appears
possible to date the samples if only small
CO2 fractions are collected and dated. The
age presented in the Fig. 8 is preliminary;
an actual dating requires several samples
and a better control of possible carbonate
re-crystallization.
5.2 Carbonates that are younger than the construction
The pozzolana mortars are famous for their ability to heal themselves after rupturing by
growing new minerals in the fractures. The most common of these new minerals is calcite
that can grow when alkaline inclusions come into contact with the ambient atmospheric
CO2 or dissolved CO2 in percolating waters. The alkaline inclusions are common because
most pozzolana mortars are in fact alkaline. Fig. 9 shows the alkalinity of some mortars
when put in water and measured 3 times within 4 hours and then the next day (not plotted,
but taken into account in the plot).
Figure 7. 14C proles from the
Colosseum. Sample 001 contains
only little contaminants and up to
50% (F<0.5) of it can be dissolved
before the 14C age is signicantly
affected by them. Sample 002 has
more contaminants and, depending
on dissolution technique, only
5-20% can be dissolved before the
14C age is affected.
1850
1900
1950
2000
2050
2100
2150
2200
2250
2300
0 0.2 0.4 0.6 0.8 1
Samples 001 and 002
Profiles AAR-7347-1.1-3
AAR-7348 -2.1- 3
A AR -73 48- 1.1- 8
AAR-7348 -2.1- 8
The Colosseum
14C age BP
F
001
002 002
002
1937±21 BP
=AD 25-85 (68.2%)
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900
3000
0 0.2 0.4 0.6 0.8 1
Sample 001
Grain-size 46-75 µm
Analyses A AR-9780.1-5
Terracina
14C age BP
F
2126±31 BP (210-90 BC, 68.2%)
Figure 8. 14C prole of the Jupiter Anxur temple,
Terracina, with a preliminary dating result.
224 Lindroos et al.
Re-activated carbonate growth seems to
be very common in Roman mortars and
the presence of eforescent carbonates
usually also affects the 14C dating. Fig.
9 may suggest that the sampling and
grinding and especially exposing the
sample powders to water may trigger
carbonate growth. A low initial pH and a
rapid drop in pH seems to correlate with
accurate dating. The young carbonates
seem, however, to dissolve rapidly in the
chemical preparation and affect only the
initial CO2 fractions. We have studied one
sample from the Colosseum and a series
of well-preserved Trajan and Hadrian
mortars from Trajans Market and Ostia in some detail. In Fig. 10 we plotted the age of the
rst CO2 fraction extracted versus the pH of the mortar. The size of the CO2 fraction is
reected by the size of the dot. Samples with low initial pH yield relatively accurate dates
whereas the more alkaline samples with pH values >8.5 yield younger ages for the initial
CO2 fractions than anticipated. The smaller the rst CO2 fraction is, the younger the age.
Is it then possible to date alkaline samples? If the only problem is alkalinity then the
later CO2 fractions should be devoid of rapidly dissolving young carbonates and yield the
correct age. If there is also geological carbonate present, it might be difcult to read the
right age from a 14C prole because there is the possibility that the interference of young
carbonates and geological carbonates will result in a plausible date that is biased. It should,
however, be considered that the binder carbonate is by far the most abundant carbonate
phase in a well-preserved mortar and if a series of samples consistently reproduce the same
14C age for the mid-CO2 fractions of the samples then they reect the age of the binder
because the likelihood of getting several equal and plausible 14C ages by uncontrolled
combinations of young carbonates and geological carbonate is negligible.
Figure 9. pH evolution of dry-sieved mortar
powders when put in water.
6.5
7
7.5
8
8.5
9
9.5
10
012345678
Alkalinity of Roman samples
pH
Time in water (h)
Trajans market (018A)
Trajans market (018B)
Tivoli (005)
Trajans market (019A)
Insula de Casa e Giardino, Ostia (027)
Colosseum (002)
Basilica Ulpia (004)
1000
1200
1400
1600
1800
2000
7 7.5 8 8.5 9 9.5 10
004
Basilica
Ulpia 019A
027
018A
018B
5%
10%
20%
026
025A
pH, Roman mortars
14C age
pH
002
Colos-
seum
14C BP time-span for Trajan-Hadrian
CO2 fraction
percentage
Figure 10. The 14C ages of
the rst CO2 fractions for
a series of well-preserved
samples from central
Rome. The samples are
sorted by initial pH value
when the 150-300 µm
grain-size fraction was put
in water (100 mg in 5 ml).
18A, 18B, 19A and 25A
are from Trajan’s market
and 26 and 27 are from
Ostia, and from the time of
the Hadrian reign.
Comm. Hum. Litt. Vol. 128 225
Unfortunately, 14C dating is rather expensive and it is quite common that the few available
14C proles are ambiguous. To ensure that we actually get the age of the binder carbonate,
several successive CO2 fractions from each sample, preferable in the 20-60% dissolution
interval, should yield the same age.18 Fig. 11 shows two problematic proles from alkaline
samples from Ostia (see Fig. 10). The samples were taken next to each other from “Insula
de Case e Giardino”, which was built in the time of Hadrian.19 This age would require
a radiocarbon age of around 1900 BP, which we got from the second CO2 fractions in
the rst dating attempt where only three successive CO2 fractions were collected (open
symbols). The multi-fraction proles do not reveal such an age. Instead they show that
the samples have at least three types of contaminants: 1) a very rapidly dissolving young
contaminant affecting the rst CO2 fraction of the lower prole, 2) a slower-released
young contaminant affecting the rst half of both proles and 3) the common slowly-
dissolving geological contaminants slowly raising the age at the ends of the proles. No
age plateaux are formed because the second young contaminant dissolves too slowly and
it will be exhausted so late during dissolution that it tends to overlap with the dissolution
of the geological contaminants. Based on this information we can reason that the mid-
sections of the age proles are less affected by contamination.
5.3 Diagenetic alterations
When mortared constructions are buried under soil they will slowly become a part of the
geological cycle and the processes that turn soft soil into sedimentary rocks will affect
them. These processes are referred to as diagenesis and they include all the chemical,
physical and biological changes undergone by the sediment after its initial deposition
(excluding surface weathering). The most relevant of these processes to radiocarbon dating
of mortars is the formation of a carbonate matrix between grains and carbonate crystals
18 See also ringbom et al., cit. n. 3, in this volume.
19 Personal communication with Janet DeLaine.
Figure 11. 14C proles of two Hadrian
samples from Ostia, each analyzed
twice. The samples were rst analyzed
in 3 successive CO2 fractions (open
symbols) and the second fractions
yielded the right age. When we
increased the number of CO2 fractions
(proles with lled symbols) and
thus the resolution of what actually
happened during the dissolution
we can see that there is no interval
covering several CO2 fractions with a
similar age in either of the proles and
no age plateaux are formed.
1000
1200
1400
1600
1800
2000
2200
2400
2600
0 0.2 0.4 0.6 0.8 1
F
Samples Rome 26 and 27
Grain-size 46-75µm
Analyses:
AAR-6287.1-3
AAR-6288.1-3; 6288.2,1-8
OxA X-2147-06-09
OxA X-2146-46
026
027
Insula de Case e Giardino, Ostia
14C age BP
Known age in 14C years
226 Lindroos et al.
inside voids and hollows. The new carbonates comprise both calcite and its polymorph
aragonite as well as other less common carbonate minerals.20
We undertook some sampling in Pompeii and Herculaneum to explore these problems.
Samples were taken at the Helsinki University excavations in Pompeii and at the British
School in Rome excavations in Pompeii and Herculaneum. Seven mortar samples, a
lime lump and a limestone inclusion were analyzed (Table 2). Sample Herculaneum 001
from cast moulds was covered in volcanic ash in AD 79, which prevented the access of
contemporaneous atmospheric CO2. Instead it yielded a modern age corresponding to the
excavation.
The general impression from the analyses (Fig. 12) is that it is very difcult to date
excavated mortars of this kind. The rst problem to consider is the CO2 emmissions
from the active Somma-Vesuvius volcano and the whole, large Campi Flegrei volcanic
system at the Bay of Naples. If the atmospheric CO2 was locally contaminated with excess
volcanic CO2 it is impossible for us to assess to what extent the 14C age is inuenced
by radiocarbon dead material from the volcanoes. Any 14C prole would be expected to
look perfectly normal but it could be biased towards an older age. However, many of the
proles we have, do not look normal and they are severely biased either to the older or to
the younger side. In 3 proles the rst CO2 fractions are older than the second fractions,
which is very unusual (Fig. 12). We have recently encountered these kinds of proles in
samples from basement rock areas with marble (i.e. metamorphic limestone).21
The lime lump EPUH 002L from Casa di Marcus Lucretius displayed a rather puzzling
14C prole (Fig. 13). The original carbonate within the lump has been entirely replaced by
carbonate with a 13th century 14C age and some even more recent material that dissolves
rst. The mineralogy of the samples is generally not very different from other ancient
mortars. CL analysis inidicated that many of them looked suitable for dating or they
displayed a lot of luminescent carbonate that is not limestone, but looks more like re-
20 m. tucker, Techniques in Sedimentology, Blackwell, Oxford 1988.
21 Lindroos, cit. n. 3, p. 83, a case from Medieval Barcelona.
Table 2. Samples from Pompeii and Herculaneum
sAmPLe mAteriAL site exPected Age
From the Helsinki University excavations in Pompeii
1. EPUH 002L Lime lump Casa di Marcus Lucretius AD 62-79
2. EPUH 006 Mortar Casa di Marcus Lucretius 2nd C BC
3. EPUH 007 Mortar Casa di Marcus Lucretius 2nd C BC
4. EPUH 008L Sintered lime lump Casa di Marcus Lucretius Geological
From the British School in Rome excavations in Pompeii
5. BSR I:9,3 001 Mortar Casa di Successus 80-25 BC
6. BSR I:9,3 003 Mortar Casa di Successus 25 BC-35AD
7. BSR I:9,11 011 Mortar Insula I:9, Room 1 AD 62-79
From the British School in Rome excavations in Herculaneum
8. Herculaneum 001 Mortar The Suburban baths AD 79
9. Herculaneum 002 Mortar The Palaestra ?
Comm. Hum. Litt. Vol. 128 227
crystallizations, which is common among ancient mortars. Some peculiarities were,
however, observed. For example, even the lime lumps contain carbonates that have a
slightly brighter luminescence than the bulk of the material. All the samples dissolved
rather rapidly and one sample (BSR I:9,11 011) dissolved extremely rapidly and displayed
heavy contamination with luminescent calcite and aragonite. The stable carbon and oxygen
isotopic values do not deviate signicantly from those of other Roman samples and thus
they provide no clue to what kind of alterations have taken place. Only sample BSR I:9,11
011 has rapidly dissolving carbonate with a rather negative δ13C signature (-25.7‰ vs
PDB) and an old 14C age.
Figure 12. 14C proles of mortars from Pompeii and Herculaneum.
2000
2500
3000
3500
4000
4500
5000
5500
6000
0 0.2 0.4 0.6 0.8
1
EPUH 006
EPUH 007
BSR I 9,3 001
BSR I 9,3 003
BSR I 9,11 011
Herc 002
Pompeii & Herculaneum
Mortar samples, grain-size 46-75 µm
F
14
C age BP
Figure 13. 14C prole of a lime lump from a rst century construction in Pompeii. The 46-75 and
76-100 µm grain-size fractions were combined to get enough material from the small lump for 5
CO2 fractions. The carbon in the lump has a 13th century 14C signature and some even more recent
rapidly dissolving material.
228 Lindroos et al.
It22is well known that carbonates co-precipitate trace elements, including U.23 In order
to identify the presence of carbonate precipitations from groundwater we analyzed the
uranium (U) isotopes in the carbonates of the samples. Uranium offers a powerful tool
to study carbonate precipitation from groundwater because the U isotopic composition in
groundwater can be very different from that in rocks. Uranium has 3 naturally occurring
isotopes 234U, 235U and 238U. Their relative abundances in nature are 0.005, 0.720, and
99.275% respectively. 235U and 238U have been around since the beginning of earths´
history and they have a constant ratio (0.00725). 234U, however, is a short-lived daughter
of 238U (with a half-life of 245 000 years) and their ratio can vary considerably in
geological samples being far from their relative abundance in nature, i.e. 234U and 238U are
said to be in radioactive disequilibrium (in activity units 234U/238U≠1). Because 234U is a
decay product and can therefore be more mobile than 238U, it is more prone to end up in
aqueous solution than 238U when minerals are weathered.24 Any samples, archaeological or
geological, exposed to weathering can be expected to have a 234U deciency (234U/238U<1)
and the samples that have been buried and exposed to groundwater precipitations tend to
be enriched in 234U (234U/238U>1). The deviation from unity can be measured accurately
enough with current techniques (a-spectrometry <5% and MC-ICP-MS <1%). Our
working hypothesis is that U coming from limestone is in equilibrium (234U/238U=1) and
insignicant because the concentrations are very low. Uranium in the pozzolana samples
22 d.b. kArner – f. mArrA – P.r. renne, ’The History of the Monti Sabatini and Alban Hills
volcanoes: groundwork for assessing volcanic-tectonic hazards for Rome’, Journal of Volcanology
and Geothermal Research 107 (2001) 185-219.
23 R.J. reeder – m. nugent – g.m. LAmbLe – c.d. tAit – d.e. morris, ‘Uranyl incorporation
into calcite and aragonite: XAFS and luminescence studies’, Environmental Science & Technology
34 (2000) 638–644 and references therein.
24 See e.g. D. PorceLLi – P.W. sWArzenski, ‘The behavior of U- and Th-series nuclides in
groundwater’, in b. bourdon – g.m. Henderson – c.c. Lundstrom – s.P. turner (ed.), Uranium-
Series Geochemistry 52 (2003) 317–361 and references therein.
Table 3. Uranium isotopes in pozzolana rocks from Rome and Puteoli determined by α-spectrometry.
sAmPLe k-Ar Age22
(ka)
LeAcHing
Solvent/t (min)
U
(ppm)
234U/238U
(activity ratio)
Pozzolana Nera 407 ± 4 HF near total 18.2 0.93±0.04
Pozzolana Nera 407 ± 4 Aqua Regia 16.3 0.97±0.07
Pozzolana Nera 407 ± 4 1M HCl 10.7 0.83±0.06
Pozzolana Rossa 457 ± 4 1M HCl (10 min) 0.87±0.05
Pozzolana Rossa 457 ± 4 1M HCl (285 min) 0.84±0.05
Pozzolana Rossa 457 ± 4 1M HCl (1410 min) 0.89±0.04
Pozzolana Rossa 457 ± 4 1M HCl (6750 min) 0.88±0.07
Pozzolana Rossa 457 ± 4 HF near total 0.92±0.02
Pozzolana Rossa 457 ± 4 HF near total 9.6 0.91±0.05
Pozzolana Rossa 457 ± 4 HF near total 9.9 0.89±0.03
Pozzolanella 357 ± 2 HF near total 51.0 0.90±0.09
Pozzolanella 357 ± 2 10% HCl 27.9 0.93±0.07
Comm. Hum. Litt. Vol. 128 229
would yield values <1 because they are weathered. Consequently, U in the mortars would
also have values <1 because most of the U in the mortar originates from the pozzolana
ller. Carbonate precipitated from groundwater, however, should contain excess 234U to
compensate for the deciency in the weathered rocks. Values >1 are therefore expected.
We measured the U isotopes in some of the pozzolana rocks (Table 3) and in the calcite
of some mortars (Table 4). The calcite was separated by partial acid leaching and using
ammonium acetate extraction, which selectively dissolves calcite.
Table 3 shows that the pozzolanas are generally signicantly depleted in 234U as
expected. The Roman mortars have a similar 234U deciency, but the mortars from Pompeii
that have been buried under volcanic ashes have gained some 234U and display values
near 1 or signicantly >1 (Table 4). The excess 234U is believed to be present in rapidly
dissolving carbonates. Earlier we have seen that these carbonates have very old 14C of
inorganic origin. Sample BSR I:9,11 011 may have carbon of organic origin because some
30% of the sample dissolves extremely rapidly and displays the δ13C value -25.72‰ vs
PDB. In the lime lump the carbon is of inorganic origin (δ13C = -2.71‰ vs PDB) and
young. These variable isotopic data can best be explained as repeated precipitations from
groundwater carrying old carbon of variable, but mostly geological, origin.
6. discussion
Pozzolana mortars are usually more problematic to date than contemporaneous or younger
lime mortars. The distinction between hydraulic and non-hydraulic mortars seems
somewhat unnecessary to apply to samples prior to dating because the original binder
mineral texture may be lost and a distinction based on the chemistry of partial leaching is
Table 4. Uranium isotopes in mortar carbonates from Pompeii, Herculaneum and Rome determined
by α spectrometry.
sAmPLe LocALity estimAted Age u
(ppm)
234u/238u
(activity ratio)
Rome 007 Torre delle Milizie AD 8th c. 0.62 0.76±0.08
Rome 017 St Costanza AD 4th c. 1.23 0.89±0.05
Ostia 13 Insula di Giove e Ganymede AD 2nd c. 2.60 0.89±0.19
Ostia 13 Insula di Giove e Ganymede AD 2nd c. 2.40 0.86±0.17
Ostia 13 Insula di Giove e Ganymede AD 2nd c. 5.38 0.92±0.38
Ostia 027 Insula di Case e Giardino AD 123-124 0.96 0.95±0.05
EPUH 006 Casa di Marcus Lucretius 2nd c. BC 1.41
EPUH 006 Casa di Marcus Lucretius 2nd c. BC 0.95
EPUH 007 Casa di Marcus Lucretius 2nd c. BC 1.08 1.00±0.05
BSR I:9,3 001 Casa di Successus 80-25 BC 1.07
BSR I:9,3 001 Casa di Successus 80-25 BC 1.13
BSR I:9,3 003 Casa di Successus AD 25-35 1.02
BSR I:9,3 003 Casa di Successus AD 25-35 0.97
230 Lindroos et al.
attached with many factors that are difcult to control. The main problems in 14C dating
of pozzolana mortars are:
1. There is less binder carbonate available than in lime mortars and therefore
numerous other carbonates, that must be considered contaminants, can have a
relatively large impact on the measurements.
2. Young volcanic systems that produce pozzolanic rocks may produce a variety
of carbonate minerals as well, which occur in and grow on the pozzolana. This
is especially the case in mainland Italy where the volcanoes are underlain by
thick limestone sequences.
3. The hard and durable pozzolana mortars are relatively impermeable to
atmospheric carbon dioxide and they commonly have alkaline inclusions that
react with contemporaneous atmospheric CO2 whenever the mortar is ruptured
and the inclusions are exposed.
4. Constructions that have been buried in soil or volcanic ashes may have severe
contamination originating from dissolved volcanic CO2 or dissolved carbonates
carried by groundwater.
The problems are tackled in a similar way as for dating lime mortars, but we strongly
recommend that the alkalinity of the sample should not only be checked with a pH
indicator, but it should be measured quantitatively. Samples with pH values >9 have
usually given too young ages. Mortars excavated from deep in the ground will probably
be so problematic that 14C dating will not give a reliable age for the construction. The
234U/238U ratio for the most rapidly dissolving carbonates in the sample can be used to
indicate the presence or absence of carbonate contamination from groundwater.
Acknowledgements
The authors would like to thank Eeva-Maria Viitanen, Fabrizio Marra, Lynne Lancaster,
Agneta Freccero and Andrew Wallace-Hadrill for guiding us during the sampling, Talvikki
Savolainen for the pH measurements, and the Academy of Finland, the Åbo Akademi
Foundation, the Rector of Åbo Akademi University, and the Magnus Ehrnrooth Foundation
for nancial support.