S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

M AT E R I A L S S C I E N C E Copyright © 2023 The

Authors, some
Hot mixing: Mechanistic insights into the durability of rights reserved;
exclusive licensee
ancient Roman concrete American Association
for the Advancement
of Science. No claim to
Linda M. Seymour1, Janille Maragh1, Paolo Sabatini2, Michel Di Tommaso3, James C. Weaver4, original U.S. Government
Admir Masic1* Works. Distributed
under a Creative
Ancient Roman concretes have survived millennia, but mechanistic insights into their durability remain an Commons Attribution
enigma. Here, we use a multiscale correlative elemental and chemical mapping approach to investigating NonCommercial
relict lime clasts, a ubiquitous and conspicuous mineral component associated with ancient Roman mortars. License 4.0 (CC BY-NC).
Together, these analyses provide new insights into mortar preparation methodologies and provide evidence
that the Romans employed hot mixing, using quicklime in conjunction with, or instead of, slaked lime, to
create an environment where high surface area aggregate-scale lime clasts are retained within the mortar
matrix. Inspired by these findings, we propose that these macroscopic inclusions might serve as critical
sources of reactive calcium for long-term pore and crack-filling or post-pozzolanic reactivity within the cemen-
titious constructs. The subsequent development and testing of modern lime clast–containing cementitious mix-
tures demonstrate their self-healing potential, thus paving the way for the development of more durable,
resilient, and sustainable concrete formulations.

Downloaded from https://www.science.org on February 12, 2024

INTRODUCTION materials such as volcanic ash ( pulvis) (6), the detailed formulations
Ordinary Portland cement (OPC) is a key component of concrete, of which were tailored to their specific intended applications (7, 8).
the most ubiquitous construction material in the world, but its pro- Whereas aerial lime mortars relied on the uptake of CO2 from the
duction has serious environmental consequences. The manufacture air to harden, hydraulic mortars combined lime and water with re-
of OPC releases up to 1 metric ton of CO2e emissions per metric ton active silicates and aluminosilicates ( pozzolanic materials) to form
of produced material, and current strategies to reduce this impact cementitious hydrates [e.g., calcium alumino-silicate hydrates (C-
are insufficient as demand continues to rise (1). These emissions A-S-H)] (9–13).
are primarily generated when limestone and clay are calcined to This reaction, also known as the pozzolanic reaction, can be
form clinker (mainly tricalcium and dicalcium silicates, also written as
known as alite and belite, respectively), which is then finely
ground. One method to reduce cement’s carbon footprint (which xCaðOHÞ2 þ yAl2 O3 � zSiO2 þ ðn xÞH2 O
accounts for up to 8% of total global greenhouse gas emissions), ! ðCaOÞx � ðSiO2 Þz � ðAl2 O3 Þy � ðH2 OÞn x ðor C-A-S-H gelÞ
is to improve the longevity of concrete through the incorporation
of self-healing functionalities. The resulting extended use life, com- Any excess unreacted lime slowly carbonates in air via the reac-
bined with a reduction in the need for extensive repair, could thus tion
reduce the environmental impact and improve the economic life CaðOHÞ2 þ CO2 ! CaCO3 þ H2 O
cycle of modern cementitious constructs (2).
In contrast to their modern counterparts, ancient Roman By developing these hydraulic mortars, the Romans were able to
mortars and concretes have remained durable in a variety of cli- create a stronger, more durable material that allowed them to build
mates, seismic zones, and even in direct contact with seawater, as larger, more complex-shaped architectural structures for purposes
in the case for maritime concrete. Because of this proven longevity that were not previously possible, including constructions in the sea
on the order of millennia, these ancient construction materials are (14–16).
attractive model systems for the design of sustainable, durable ce- The production process for Roman mortar began with the cal-
mentitious composites for modern engineering applications (3–5). cining of lime from a source such as limestone, marble, or travertine
For many centuries, and throughout the entire ancient Roman (all predominantly calcite, CaCO3) to form quicklime [calcium
Empire, architectural elements, such as walls and foundations, oxide (CaO)] (17). This lime-based material, which can be hydrated
and infrastructure systems, including aqueducts, roads, and using water (a process known as slaking) or added directly (a
bridges, were created from unreinforced concrete. This concrete process known as hot mixing) (18–21), was then mixed with volca-
was typically composed of volcanic tuff and other coarse aggregates nic ash, ceramic fragments (cocciopesto), or other pozzolana, sand,
(caementa), and bound by a mortar based on lime and pozzolanic and water to form the hydraulic mortar.
Regardless of the application, strict specifications for the raw ma-
terials were detailed by the ancient scholars Vitruvius and Pliny, es-
1
Department of Civil and Environmental Engineering, Massachusetts Institute of pecially for the limestone, which was to be pure white, so as to lack
Technology, Cambridge, MA 02139, USA. 2DMAT srl, Udine, 33100, Italy. 3IMM, Is- impurities (6, 17). Previous studies on Roman architectural mortars
tituto Meccanica dei Materiali SA, via al Molino 55, 6916 Grancia, Switzerland.
4
Wyss Institute for Biologically Inspired Engineering, Harvard University, Cam-
determined that the calcined lime in these samples frequently con-
bridge, MA 02138, USA. tained <5 weight % oxides other than CaO (7, 14). In recent years,
*Corresponding author. Email: [email protected]

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 1 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

detailed investigations of ancient mortars and concretes allowed for including incomplete or over-burning during the calcining of
the identification of some of the key chemical and mineralogical lime (20), carbonation before concrete preparation (30), incomplete
processes associated with interactions between pozzolanic materials dissolution during setting (12), or insufficient mixing of the
and hydrated lime and provided insights into the mechanical per- mortar (14).
formance of these materials (11, 12, 22–26). In maritime structures, these lime clasts can be heterogeneous in
Studies focusing on the durability of Roman concrete construct- composition; may contain calcite, vaterite, brucite, ettringite, hy-
ed in marine environments, for example, evidenced the dissolution drocalumite, C-A-S-H, tobermorite, and Al-tobermorite (5, 12,
of lime and vitric tuff clasts at high pH, followed by the precipitation 27, 30–34); and can be categorized into one of the following three
of C-A-S-H–containing reaction rims and, subsequently, the post- groups: (i) geologic, (ii) partially dissolved, and (iii) fully dissolved
setting crystallization of Al-tobermorite and phillipsite in the (transformed) (12). Geologic inclusions are calcite-bearing aggre-
matrix (11, 27). In both Augustan and Imperial era architectural gates that did not fully calcine during the production of quicklime.
concretes, a similar C-A-S-H precipitation and subsequent crystal- The other two, partially and fully dissolved clasts, have provided
lization and growth of platy strätlingite crystals in the perimeters of insight into the chemical evolution of maritime concretes. For
scoriae and the cementing matrix were observed (24, 28). More re- example, partially dissolved clasts show a C-A-S-H–containing re-
cently, both Al-tobermorite and strätlingite crystals were found in action rim, gradating toward a calcite-rich core (12). Fully dissolved
the mortars of the Augustan period (ca. 30 BCE) tomb of Caecilia clasts, in contrast, exhibit C-A-S-H throughout and, in some cases,
Metella (29). The prolonged reactivity of volcanic aggregates and the formation of Al-tobermorite (12, 27). In both partially and fully
their potential role in the long-term durability of these materials dissolved clasts, the clast exteriors contain hydrocalumite and et-
has thus been an ongoing focus of recent studies on Roman tringite, attributed to hydration with seawater (12). Although
concretes. these clasts are well characterized in maritime Roman concretes,

Downloaded from https://www.science.org on February 12, 2024

In addition to the features described above, aggregate-scale relict less is known regarding the microstructure and chemical composi-
lime clasts, also referred to as remnant lime or lime lumps, are a tion of relict lime clasts in open-air Roman constructions (struc-
ubiquitous and conspicuous feature of both architectural and mar- tures on land) and the role that they might play in processes
itime Roman concretes. The presence of these distinctive bright associated with the durability of these structures (5, 35).
white features has been previously attributed to several scenarios

Fig. 1. Collection location and distinctive features of the ancient Roman concrete samples used in this study. The test samples came from the archaeological site of
Privernum, near Rome, Italy (A), and shown as a photogrammetry-based three-dimensional reconstruction (B). The architectural mortar samples were collected from the
bordering concrete city wall (C). Large-area EDS mapping of a fracture surface (D) reveals the calcium-rich (red), aluminum-rich (blue), silicon-rich (green), and sulfur-rich
(yellow) regions of the mortar. Further imaging of polished cross-sections (E) shows aggregate-scale relict lime clasts within the mortar (the large red features denoted by
asterisks). The colored arrows in (E) denote the pore-bordering rings visible in the EDS data that are rich in calcium (red) or sulfur (yellow), which are enlarged at right to
show additional detail. Photo credits (B and C): Roberto Scalesse and Gianfranco Quaranta, Associazione AREA3, Italy.

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 2 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

To address these yet unresolved questions, we report on the fragments) using this approach, and the results are representative
chemical characterization of relict lime clasts found in 2000-year- of the trends observed.
old Roman concrete samples obtained from the archaeological These elemental maps (Fig. 1, D and E, and fig. S1) also show
site of Privernum, Italy. The investigated samples are composition- sulfur throughout the mortar, which, as proposed previously (38),
ally consistent with other architectural mortars encountered could be due to primary sulfur originating from the components
throughout the Roman Empire (8, 13, 16, 36, 37), and were used to prepare the mortar, secondary sulfur that entered the
sourced from the masonry mortar of the city wall, an open-air struc- binding matrix through the pores of the mortar, seen as rings, or
ture. We characterized the composition of the lime clasts and their a combination of these. While all of these considerations are possi-
surrounding matrix using large-area scanning electron microscopy ble, detailed elemental analysis (fig. S1) shows that sulfur is present
and energy dispersive x-ray spectroscopy (SEM-EDS), powder x-ray primarily in the pore volume, with secondary sulfur likely originat-
diffraction (XRD), and confocal Raman imaging. The results of ing from one or more of a variety of external, environmental sources
these analyses provide compelling evidence for hot mixing of including soil and acid rain. The latter intrusion is a common deg-
Roman mortar using quicklime instead of, or in addition to, radation pathway in which cementitious materials are leached on
slaked lime. From these findings, we propose that persistent, aggre- exposure to acid rain containing sulfur-based ions, as calcium car-
gate-scale, high surface area lime clasts that result from this process bonate is converted to calcium sulfate hydrates (e.g., gypsum) (39,
could serve as a source of reactive calcium for long-term pore and 40). Raman mapping (fig. S2) demonstrates that the sulfur-bearing
crack filling and therefore provide a chemically dominated intrinsic pore walls contain predominantly gypsum, thus precluding the use
self-healing mechanism. Motivated by these discoveries, we devel- of calcium sulfoaluminate fibers as a filling agent in these samples
oped a modern Roman-inspired cementitious mixture that incor- (38). Despite centuries of exposure to environmental attack,
porated a hot mixing preparation method and observed the however, the masonry wall remains well preserved around the pe-

Downloaded from https://www.science.org on February 12, 2024

effective self-healing of induced cracks measuring up to 0.5 mm rimeter of the archaeological site.
in width. The results from our multiscale compositional analysis To identify both the major and minor mineral phases present in
of Roman concrete and the development of their modern counter- the concrete wall samples and map their specific distribution pat-
parts provide a starting point to understand the role of various pro- terns, we used two modes for quantitative analysis of the mortar’s
duction parameters on the durability of ancient mortars and elemental composition: first on freshly fractured surfaces (Fig. 2, A
concretes and demonstrate the applicability and potential value of to C) and second on prepared, polished cross-sections (Fig. 2, D to
these ancient design principles in the development of cementitious F). Quantitative EDS data were plotted pixel by pixel on ternary di-
materials for the modern world. agrams (Fig. 2, B and E) to reveal the relative ratios of the three key
elements of the cementitious phases—calcium, silicon, and alumi-
num (41).
RESULTS Immediately evident from these analyses is the compositional
The Roman mortar samples investigated in this work came from the shift toward the (red) calcium vertex of the ternary diagram for
ancient city of Privernum, near Rome, Italy (Fig. 1A). Privernum the fractured surface (Fig. 2, A and B), suggesting that the sample
was occupied as early as the 2nd century BCE through the 13th likely failed along weaker binding phases, the relict lime clast, and
century CE, and the site was first excavated in the 18th century, the binder immediately surrounding the clast. Quantitative EDS
with the most recent work performed from 1990 to present. The data were then clustered to visualize the distribution of small differ-
samples analyzed in the present study were part of the city wall ences in composition throughout the mortars. Points on the ternary
(Fig. 1, B and C), are compositionally consistent with other diagrams (Fig. 2, B and E) and individual pixels on the original map
Roman concrete samples dating from the same time period (8), area (Fig. 2, C and F) were colored on the basis of the assigned
and correspond with the city’s Roman occupation, which lasted cluster. Within the polished cross-sections (Fig. 2F), the different
through the fall of the empire. investigated relict lime clasts all separated into two clusters—one
To investigate centimeter-, millimeter-, and submillimeter-scale within the clast reaction rim and another in the core. Evident
compositional details of the mortar samples, we used SEM-EDS. here is a transition from the calcium-rich core to a rim containing
Large-area elemental mapping of the freshly fractured surface of a higher concentrations of silicon and aluminum. Further analysis of
representative Roman mortar sample (Fig. 1D) shows its complex the EDS maps also shows that the matrix immediately surrounding
chemical composition (with a dominant presence of calcium, the lime clasts is richer in calcium than other parts of the matrix,
silicon, and aluminum), the aggregate-scale relict lime clasts, and suggesting that some leeching has occurred in the sample. Just
evidence for pore filling by secondary mineral phases (fig. S1). outside the lime clasts and extending to the edge of the sample, a
Further elemental mapping of polished mortar samples (Fig. 1E) matrix phase rich in calcium with aluminum and silicon can also
provided additional fine-scale structural and compositional infor- be seen. As shown in Fig. 2F, these points (red) are distinct from
mation, the results of which were consistent with previous studies the primary binding phase (yellow), which has ratios of silicon, alu-
of Roman mortars and concretes (11, 12, 27). Note that to obtain minum, and calcium more closely associated with C-A-S-H binders
reliable elemental mapping data from the irregularly fractured con- (41). Petrographic analysis (fig. S3) shows that the binder is pre-
crete samples, we used a multidetector EDS system, which both dominantly isotropic, as would be the expected product from the
minimized the potential contributions of shadowing artifacts and pozzolanic reaction ( poorly crystalline C-A-S-H), with calcite oc-
facilitated the high-throughput analysis of a large number of test curring in the matrix near lime clasts and within some pores.
samples. In total, more than 15 different fractured and polished A better understanding of the mineral assemblage and the po-
mortar sample regions were analyzed (from 10 different collected tential identification of crystalline phases within the mortars was
obtained through powder XRD. Two sample types were examined

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 3 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from https://www.science.org on February 12, 2024

Fig. 2. Multimodal compositional analysis of ancient Roman concrete. EDS maps of a freshly fractured surface (A) and polished cross-section (D) are quantified, and
each pixel is replotted as a relative ratio of calcium, silicon, and aluminum on a ternary diagram (B and E), with the most Ca-rich phases shown in red and denoted by the
red circle near the Ca vertex in (B). Pixels in the ternary diagrams (B and E) and phase maps (C and F) were colored on the basis of the cluster of EDS data to which they
belong. Surrounding each of the lime clasts is a clear rim of a compositionally unique phase (the hydration rim), which is denoted by the light blue zones in (F) and is
shown at higher magnification at right with color-coded text for additional clarity. Powder XRD of a relict lime clast (G) and its surrounding matrix (H) confirms the
presence of calcite and quartz and suggests the presence of vaterite among other phases [e.g., halite (*) and volcanic minerals (diopside, leucite, etc.) (+)] in the mortar.

in this way: extracted relict lime clasts from the mortars (Fig. 2G) clast, and one within the clast reaction rim. The reaction rim cluster
and the binding matrix complete with other aggregate material is also seen incorporated into the cementitious matrix and corre-
(Fig. 2H). The remnant lime clasts were predominantly calcite sponds to a composition range akin to calcium-enriched C-A-S-H
and quartz (likely due to imperfect separation of the clast from sur- phases (41). Each pixel of the lime clast region was recolored on the
rounding binder and aggregates), although the diffraction data also basis of the cluster that it belongs to in Fig. 3E.
suggest the presence of small quantities of potential recrystallization High-resolution mapping data of the mineral phases in this same
phases such as vaterite. Overall, the elemental and mineral assem- cross-section were obtained using Raman microspectroscopy
blage in the Privernum mortar was well aligned with the composi- (Fig. 3D). This methodology was chosen because the different
tion of previously studied Roman mortar samples (9, 19, 25, 34, 35), calcium carbonate polymorphs of calcite, aragonite, and vaterite
highlighting the consistency of these samples from archaeological exhibit different Raman spectra due to the different binding envi-
site to site. ronments within the crystal structures (42). Two carbonate-domi-
Large-area SEM-EDS mapping of a representative resin-embed- nated phases were identified with this method. Specifically, one
ded cross-section of the Privernum mortar is shown in Fig. 3A. The such phase (Fig. 3D, yellow) is characterized by a shift of −3
relict lime clasts, identified by their high concentrations of Ca, were cm−1 of the ν1 carbonate peak with respect to the 1085 cm−1 peak
studied in detail using correlative EDS and Raman microspectro- of calcite and a fluorescence signature distinct from the rest of the
scopy (Fig. 3, B to E). The elemental composition of the lime sample (fig. S4). The distribution of the phases identified through
clasts was clustered as described above to identify the distribution Raman spectroscopy is shown in Fig. 3E, and fig. S4 further dem-
of compositional differences within the samples. The clustering onstrates sample heterogeneity. In some of the relict lime clast
results were plotted on the ternary axes (Fig. 3C) to characterize spectra, a shoulder that can be associated with disordered CaCO3
the relative ratios of the primary cementitious components. Six was present on the ν1 peak (see fig. S5). Broadening of the ν1
clusters were visualized corresponding to quartz aggregates, resin- peak compared to a geologic Iceland spar calcite standard was
rich regions, the cementitious matrix, two clusters within the lime also observed (fig. S6).

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 4 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from https://www.science.org on February 12, 2024

Fig. 3. Detailed analysis of lime clast elemental (EDS) and mineral (Raman) data. Large-area (5 mm image width) SEM-EDS elemental map of a polished Privernum
wall section (A). The chemical composition of a representative lime clast (denoted with an orange box) is explored in detail (B). Elemental composition from quantified
EDS shows six clusters, presented on a ternary diagram of calcium, silicon, and aluminum (C), which are colored on the basis of the cluster of EDS data to which they
belong. Raman spectroscopy (D) reveals calcite (red), additional calcium carbonate phases (yellow), quartz (green), and resin (blue), with the characteristic peaks indi-
cated. The distribution maps of the identified phases (E) are shown on the basis of their corresponding EDS and Raman scan data: quartz aggregates (blue), resin-rich
regions (orange), the cementitious matrix ( purple), two clusters within the lime clast (green and red), and a clast reaction rim (yellow).

The recent discovery of calcite-filled cracks in Roman concrete modern), the bright white color of the lime clasts and their
(29) suggested a potential long-term healing process that requires a porous microstructure were visible. Further examination of the
Ca-rich source. Considering the ubiquity of relict lime clasts in lime clasts from the polished samples of the Roman-inspired con-
Roman concrete (5, 12, 24, 43) and their high surface area due to crete formulations using polarized optical microscopy revealed a
their particulate microstructure (Fig. 4, A to D), these inclusions similar visibly distinctive periphery that was seen in the native
might provide the requisite Ca reservoirs for these processes. In- Roman material (cf. Figs. 2F and 3E and figs. S7 to S9).
spired by these observations, we set out to create a modern analog The potential for the high–surface area lime clasts to act as a re-
of this material and explore its properties. Considering the relatively active calcium source for self-healing properties was further inves-
even distribution of the lime clasts and their well-defined morphol- tigated using our Roman-inspired concrete formulations. The cured
ogies, these observations raise the intriguing proposition that they samples were fractured by splitting the sample longitudinally
may have been added to the concrete mix in their intact form (a (Fig. 5A), and the opposite walls of the cracked sample were ad-
practice known as hot mixing). Considering the documented use joined at a distance of ca. 0.5 mm.
of hot mixing in both ancient and modern mortar preparations After an initial preconditioning step (45), the re-mated sides of
(19, 44), we explored this approach in our synthetic formulations. the test samples (which also included lime clast–free controls) were
In our modern cementitious formulations, we combined OPC, then introduced into a flow circuit (Fig. 5B) and subjected to a cons-
pulverized fly ash, sand, and water at a mass mixing ratio of ca. tant water flow. The flow rate over 30 days was then recorded using
1:0.2:2:1. To this mixture (known herein as Mix 1), quicklime was electronic flow meters with a precision of ±5 liters/hour. In these
added to different samples, spanning a concentration varying from studies, a measured decrease in flow rate was predicted to corre-
7.5 to 15 mass % (which was chosen on the basis of the encountered spond to potential self-healing (crack-filling) activity. Typical
frequency range of relict lime clasts in different Roman concrete initial values of flow rates through open cracks ranged between 10
samples) to investigate the effects of quicklime content on the con- and 30 liters/hour and reduced over the course of 1 to 3 weeks de-
crete’s performance. The resulting mixtures were cast into cylindri- pending on the crack geometry, to almost zero when the crack was
cal molds measuring 10 cm in diameter and 10 cm in height and eventually sealed (Fig. 5C). When the flow of water eventually
cured under water for 28 days before use. stopped (and the flow rate was zero or at a negligible flow), the
Once the production of our modern concrete samples was com- test was stopped, and the nature and distribution of secondary prod-
plete, we explored the microarchitecture of the lime clasts and com- ucts formed in the crack (Fig. 5, D to F) during the process were
pared their features to those found in their ancient Roman evaluated using optical microscopy (Fig. 5, D and E) and Raman
equivalents (Fig. 4). In both types of samples (ancient and spectroscopy (Fig. 5F).

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 5 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from https://www.science.org on February 12, 2024

Fig. 4. Compositional and morphological characterization of ancient and modern lime clasts. (A and E) Optical micrographs showing the conspicuous bright white
color of the lime clasts, which can easily be identified from large-area elemental mapping via SEM-EDS (B, C, F, and G), are morphologically distinct from other calcareous
aggregate material (denoted by a yellow asterisk) and exhibit a distinctive particulate microstructure (D and H). The dotted boxes in (B) and (F) denote the locations of the
magnified views in (C) and (G), respectively. (I) A higher magnified view of a single fractured lime clast from our hot-mixed Roman-inspired concrete formulation. The
visibly distinguishable periphery of these lime clasts [denoted by the “outer shell” labels in (J) and (L)] can be easily seen via both bright-field and polarized illumination of
polished samples, while polarized illumination specifically emphasizes the high density of internal cracking in the clasts (K and L). The dotted box in (K) denotes the
location of the magnified views in (L).

On the basis of these initial findings and to assess the real-world DISCUSSION
applicability of a self-healing Roman-inspired quicklime-contain- Previous evidence suggests that in ancient Roman cementitious
ing concretes in the modern built environment, we developed a structures, the relict lime clasts can react over time with other
second formulation (known herein as Mix 2) using an industrially mortar components, forming both amorphous (C-A-S-H) and crys-
relevant concrete recipe that included coarse aggregates and super- talline (e.g., Al-tobermorite and strätlingite) phases (11). These
plasticizer. These concrete samples (along with quicklime-free con- maturation pathways, however, are not the only ones observed in
trols) were cast as rectangular prisms, and their drying shrinkage Roman cementitious materials and, as shown here, can also
was measured 90 and 365 days after casting. After 90 days, the follow different trajectories depending on the local environmental
quicklime-containing Mix 2 concrete samples had shrunk 9% less conditions (Figs. 2 and 3). While the polished cross-sections of the
than that of the quicklime-free controls, demonstrating the role of ancient Roman concrete samples described in the present study
quicklime as an effective shrinkage compensator during the early (Fig. 2, D to F) do show clear inclusion of silicon and aluminum
stages of the curing process. At 365 days after casting, the drying in the lime clast, suggesting that some conversion to C-A-S-H
shrinkage of the two concretes had converged (<1% difference), and/or its crystalline homologues occurred, calcium carbonate, pre-
thus confirming the marginal effects of expansive quicklime in dominantly as calcite, remains within their cores (Fig. 3D).
this industrially relevant concrete mix. It has already been suggested that calcium carbonates originating
from different sources (e.g., carbonated slaked lime, carbonated free
lime, or geologic calcite) show differences in their morphology and

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 6 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from https://www.science.org on February 12, 2024

Fig. 5. Modern mortar self-healing experiments. After casting, the Roman-inspired hot-mixed concrete samples were mechanically fractured and then re-mated (with
a gap of 0.5 ± 0.1 mm) and preconditioned for our crack-healing studies (A). Using an integrated flow circuit (B), water flow through the sample over the course of 30 days
was documented with a flow meter. Compared to the lime clast–free control (orange line), after 30 days, water flow through the lime clast–containing sample (blue line)
ceased (C), and examination of the cracked surface revealed that it had been completely filled with a newly precipitated mineral phase (D and E), which was identified as
calcite from Raman spectroscopy measurements (F).

crystal structure (46, 47). For example, slaked lime [Ca(OH)2] trans- those made using slaked lime putty or powder (18, 19). The charac-
forms into calcite (CaCO3) when exposed to atmospheric CO2; teristic features of the lime inclusions described in the present study
however, it is relatively more disordered than geologic calcite could also be explained by the temperature increase associated with
samples and, for this reason, can, in some cases, be differentiated hot mixing, which facilitates both the reactions responsible for
as such (46). hardening the mortar and the formation of a gradient in hydration
During the production of concrete, water availability plays a crit- around the lime clasts, preventing them from further dissolution in
ical role in dictating the reaction kinetics and the evolution of dif- the mortar matrix.
ferent mineral phases. For example, previous Raman spectroscopy Regarding the production processes that may have favored the
and XRD studies investigating the hydration and carbonation of lime clasts remaining intact instead of dissolving and reacting in
calcium oxide samples exposed to different ambient humidity the binding matrix, previous research on other ancient Roman
levels demonstrated that at relative humidities below 20%, mortars has suggested a series of hypotheses. These possibilities
calcium oxide converts to amorphous calcium carbonate (48). include (i) hydration and subsequent partial carbonation of the
However, at relative humidity levels of 20 to 40%, conversion to lime before incorporation in the mortar; (ii) overburned lime,
calcite, aragonite, and vaterite was observed, while samples wherein the CaO is passivated and less reactive; (iii) underburned
exposed to humidity levels above 60% showed only calcite. These lime, wherein the lime is not completely calcined; or (iv) the use of
results suggest that the relative humidity of the calcium oxide mat- coarse quicklime instead of a fine powder or paste (18, 30). The gra-
uration environment can be inferred through the measurement of dient in composition seen in the EDS maps of the lime clasts in the
calcium carbonate polymorph ratios within the lime clasts. Privernum samples suggests that some hydration-driven reactions
Using these criteria and on the basis of the lime clast composi- have taken place between the lime and aluminosilicates in the
tion from ancient Roman concretes reported here, a potentially binder, thus likely resulting from one of the latter two pathways.
valid explanation of these water-depleted conditions could be hot The use of coarse particles of quicklime (alone or in conjunction
mixing, referred to as such because of the exothermic nature of with slaked lime) in the mortar is another possibility, which
the CaO hydration reaction (19). In these preparations, quicklime could allow hydration reactions to take place on the clast exterior
is simultaneously mixed with the pozzolans, aggregates, and water and allow the development of the chemical gradient seen in the
instead of, or in addition to, slaked lime as the calcium source in the lime clasts of the Privernum mortar and elsewhere (5, 12, 32, 49).
mortar. Hot mixing thus has the potential to change the conditions The heat resulting from the hydration of quicklime could reduce
under which the lime clasts are transformed within the mortar. It is the availability of water, potentially resulting in the low-humidity
known that hot mixing of other ancient lime-based mortars is one conditions that could explain the Raman spectra reported here. Fur-
method that leads to lime inclusions and that modern reproduc- thermore, these lime clast hydration gradients are also observed in
tions of these mortars have been shown to set faster compared to our modern quicklime-containing mixes that were developed and

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 7 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

investigated in the present study (Fig. 4, J to L, and fig. S9). Further frescoes and wall plaster, for example, the ancient scholars would
evidence supporting the hypothesis of quicklime use in the primary often suggest the aging of lime in water before use and ensure
mix is the cracks observed in petrographic photomicrographs that that it was as finely ground as possible (17), because incompletely
form during the hydration of quicklime in both ancient samples hydrated lime particles, known as bottaccioli in these applications,
(12) and modern reproductions (Fig. 4L), as well as the highly could absorb water over time and expand, damaging the paint
porous nature of lime clasts observed in the fractured surfaces of (fresco) layer. For this reason, both Vitruvius (6) and Pliny (17) de-
both ancient and modern lime inclusions (Fig. 4, D and H). scribe the preparation of lime for plasterwork to involve a thorough
Both of these phenomena (the formation of internal microcracks soaking or softening process (macerata). When referring to lime for
and a porous architecture) can be attributed to volume changes as- structural use, however, Vitruvius uses the word extincta (II.5.1)
sociated with lime cycle transformations. The CaO reaction with instead of macerata. While extincta and macerata are both frequent-
water to produce Ca(OH)2 is accompanied by a density-induced ly interpreted as referring to slaking, Vitruvius’ change in diction
volume expansion of ca. 33% (assuming the densities of CaO and points to a potentially different process. On the basis of the
Ca(OH)2 at 3.34 and 2.2 g/cm3, respectively) and accounts for clast results of our chemical characterization of the Privernum
crack formation and the observed porosity. While these volumetric mortars, it is thus possible that in contrast to the use of macerata
changes in the concrete constituents might be expected to negative- (which specifically refers to the slaking process), extincta could
ly affect concrete performance, our results with modern lime clast– refer to lime hydrated simultaneously with the other mortar com-
containing industrially relevant concrete formulations have demon- ponents, supporting the hot mixing hypothesis proposed here. Fur-
strated that, even after 1 year, lime clast–containing concretes thermore, it has also been previously proposed that some Roman
exhibit negligible drying shrinkage compared to their lime clast– mortars were prepared through the sequential addition of both
free controls. slaked lime and quicklime in a two-step process (51).

Downloaded from https://www.science.org on February 12, 2024

Despite extensive literature related to the composition and appli- Hydration of quicklime during hot mixing produces an exother-
cations of ancient Roman concretes, the exact order of operations mic reaction as CaO hydrates to form Ca(OH)2. The temperature
for Roman mortar production based on historical evidence increase in the mortar is approximately 55° to 60°C over ambient
remains ambiguous. There is even debate as to whether preparation (52), with a presence of hot spots characterized by temperatures ex-
techniques differed between the production of marine and terrestri- ceeding 200°C (19). Previous thermodynamic modeling of the poz-
al cementitious structures (12). Imperial age mortar (according to zolanic (hydration) reaction in Roman marine concrete has
Vitruvius) was prepared by mixing lime with volcanic sand (mate- suggested that temperatures up to 97°C are possible within thick
ries ex calce et harena mixta). During the Republican period, Cato, concrete structures from the pozzolanic reaction alone (23). The
in his De Agri Cultura (50), describes the mortar mix as calx hare- heat from these two reactions, the hydration of CaO and the pozzo-
natus (“lime with sand”). The wet mortar mix would then be mixed lanic reaction, could facilitate the formation of C-A-S-H and even
with tuff and brick caementa to form a concrete. In general, for crystalline phases such as Al-tobermorite and strätlingite, in the

Fig. 6. Schematic of the proposed mechanism for self-healing within ancient Roman mortars. Through the process of hot mixing, the calcium-rich lime clasts are
encapsulated by the cementitious matrix, which, following the formation of a hydration rim, ultimately undergo carbonation (top row). Upon cracking (bottom row),
water can infiltrate, transporting a calcium-enriched solution into the pore network to heal the damage ( process 1) or serve as reactive calcium for post-pozzolanic
reactions (such as C-A-S-H formation) at the interface between the volcanic tephra and the surrounding matrix ( process 2).

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 8 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

cementing matrix and along the reaction rim of the lime clasts in the healing tests carried out on our modern samples described in the
early stages of the mortar hydration. Evidence for these reactions is present study (Figs. 4 and 5, C to F) further support this hypothesis.
further supported by the inclusion of silicon and aluminum in the An important question to be asked at this point is how the
EDS clusters identified within the reaction rims of the lime clasts gradual carbonation of lime clasts affects the self-healing capacity
studied herein and several pieces of evidence reported elsewhere of the final mixes. It is clear that the freshly hot-mixed concrete
(34). The pozzolanic reaction is sensitive to changes in temperature that incorporates lime clasts has autogenous self-healing potential.
(39), and, hence, hot mixing provides a plausible explanation for the If a crack propagates through the concrete, then the high–surface
gradient in composition seen within the lime clasts of the Priver- area Ca(OH)2-rich clasts release Ca2+ into a highly alkaline solution
num mortar samples described in the current work and those that readily carbonates and seals the crack (Fig. 5). A long-term car-
from other Roman concretes (12, 32). bonation of clasts would intuitively compete with this process con-
Whereas previous evidence supports the formation of Al-tober- sidering the buffering effect of carbonates and subsequent pH
morite minerals in lime clasts of maritime structures due to the heat reduction of the fluid that is oversaturated in chemical constituents.
of the pozzolanic reaction (23), the results presented here suggest Nevertheless, in the case of lime clast carbonation [a process that is
that, by contrast, when quicklime is introduced via hot mixing in distinct from the carbonation of portlandite in modern concrete
terrestrial structures, the reaction is limited to the outer rim of formulations (59)], a high surface area (evident in SEM images)
the lime clast, encapsulating calcium-rich core structures within further improves solubility (60), thus benefiting the mechanism
mortar matrix. This particular configuration can be observed in proposed herein.
the comparison of Roman and modern samples (Figs. 2 and 4), The proposed post-pozzolanic process introduced here can be
both showing the same type of lime clast inclusions, hydration- compared to the autogenous healing in modern cementitious ma-
driven cracking, and distinctive reaction rims. The initially terials wherein cracks are filled by carbonation products or via the

Downloaded from https://www.science.org on February 12, 2024

matrix-incorporated clasts would then undergo a slow transforma- further hydration of unreacted pozzolanic particles (61, 62). It
tion into calcium-rich, highly porous phases of various polymorphs should be noted that cementitious compounds contained in the
of calcium carbonate. concrete can always induce some degree of natural self-healing.
Inspired by these discoveries, it is thus likely that the high abun- For example, it is generally assumed that in these types of self-
dance of aggregate-scale lime clasts in ancient Roman mortars could healing studies, where water is pumped through a damaged struc-
thus serve as a source of calcium for post-pozzolanic processes in a ture, at relatively low–water pressure gradients and up to 10% rela-
pore- and crack-filling “self-healing” mechanism that combats the tive movement of the crack, any crack with a width of up to 0.20 mm
progressive degradation of these cementitious materials (Fig. 6). can naturally self-heal, while for relative movements between 10 and
Over time, as cracks and pores form, the intrusion of water causes 20%, the crack width for natural self-healing is between 0.15 and
the dissolution of calcium-based phases in the relict lime clasts car- 0.20 mm (63). However, for cracks wider than the above values
rying them into the pore network. As the calcium-rich fluids leach (such as those used for the self-healing experiments in Fig. 5),
into the cracks or the connected pore network, many pathways exist any measured reduction in water flow through the structure
for potential post-pozzolanic reactions. For example, excess pozzo- induced by self-healing would imply that additional mechanisms
lanic material, such as volcanic ash that did not react during the have occurred.
initial setting and curing, or aggregates of volcanic origin, can The effectiveness of the pore- and crack-filling mechanism de-
now dissolve and react with Ca-rich fluids originating from the scribed here is predicated on the frequency of cracks intersecting the
lime clasts to form C-A-S-H phases, thus reinforcing the interfacial abundant highly porous and intrinsically reactive lime clasts. In
zones between volcanic aggregates/ash and the binding matrix (29). support of this requisite damage mode, the optical and SEM
This strengthening is associated with the consolidation of the inter- imaging of freshly fractured ancient Roman concrete surfaces
facial zones and increased mechanical performance of the C-A-S-H reveals that fractures predominantly propagate through the lime
compared to its precursors (53). Another possible pathway is the clasts in the architectural mortar, which are visible on both sides
recrystallization of CaCO3 phases within the pore/crack space. of the crack interface. It should be noted, however, that lime produc-
This pathway, in which secondary calcium carbonate is precipitated tion in antiquity could lead to underburned/overburned lime that
through a mechanism similar to the one occurring in the formation could potentially influence their surface area, subsequent reactivity,
of calthemites (54, 55), relies on the wetting and drying cycles ex- and mechanical properties. While our analyses suggest that lime
perienced through normal weathering conditions. These processes clast composition in Roman mortars is relatively consistent from
have been observed previously in both modern and ancient carbon- sample to sample, in some cases, higher-density lime clasts in
ate-based mortars (56–58). In contrast to these previous studies, ancient mortars can act as aggregate, with cracks instead propagat-
however, we suggest here that the hot mixing–transformed lime ing through the interfacial transition zone (5).
clasts act as a calcium source for these processes and, furthermore, The entire self-healing mechanism suggested here occurs upon
the precipitation of calcium carbonate as a crack-filling mechanism stimulation from external forces that would, otherwise, cumulative-
is already a known pathway for autogenous healing in other ancient ly result in material failure if left unchecked. The calcium-rich
lime-containing mortars (57). In the present study, we demonstrate phases within the lime clasts remain stored until they are needed,
the calcium enrichment of matrix phases adjacent to the lime clasts thus potentially allowing the healing properties of these materials
(shown in red in Fig. 2F), supporting the hypothesis that the lime to persist over millennia. To this end, microcracks and pores
clasts are a source of calcium for leaching and recrystallization induced by environmental changes (e.g., freeze-thaw cycling and
within the pore space of the mortars (Fig. 2, D to F). Microcrack acid rain) or mechanical damage (e.g., settling and seismic loads)
filling by calcite has been recently observed in ancient Roman can be filled, with water activation, once the cracks reach the lime
mortars from the tomb of Caecilia Metella (29), and the self- clast particles.

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 9 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

In this work, we successfully demonstrate that these ancient con- Scanning electron microscopy and energy dispersive x-ray
crete design practices could be applied for the development of spectroscopy
modern OPC-based formulations, thus showing far-reaching impli- Elemental analysis to complement the microstructural data was ob-
cations for extended concrete design life, long-term durability, and tained using quantitative SEM-EDS. Element mapping was used to
sustainability. Whether the damage occurs within years of construc- study the composition and the spatial distribution of phases within
tion or centuries thereafter, so long as the lime clasts remain, these the mortar, specifically the lime clasts and the surrounding matrix,
self-healing functionalities can persist. Inspired by the demonstrat- and its chemistry was investigated on both polished and freshly frac-
ed potential of this technology, future studies will explore how such tured surfaces. The latter was included to avoid potential artifacts
an ad hoc self-healing mechanism can be implemented in modern from the sample preparation process, such as chemical alteration
infrastructure, both for reinforced concrete and for unreinforced of the lime clasts (e.g., carbonation) during the polishing process.
applications as are commonly used, for example, in concrete The scanned areas were selected on the basis of the presence of
three-dimensional printing efforts (64). lime clasts within the section. Polished samples were embedded
and prepared by a commercial third party, and the fractured surfac-
es were exposed immediately before sample mounting and imaging.
MATERIALS AND METHODS All samples were imaged (uncoated) in low vacuum (15 Pa) with
Experimental design an acceleration voltage of 20 keV using a Tescan Vega GMU scan-
The following analytical methods were carried out to (i) investigate ning electron microscope. The fracture surface measurements were
the multiscale compositional heterogeneity of ancient Roman acquired using a Bruker dual-detector EDS system to minimize
mortars and (ii) to extract the design principles for the development shadowing artifacts. The pixel size for the EDS data was adjusted
and testing of modern Roman-inspired concrete formulations. to correspond to the estimated interaction volume, in this case ca.

Downloaded from https://www.science.org on February 12, 2024

3 μm, so that overlap of data did not interfere with the interpretation
Sample collection of phases present. Each pixel in the elemental map contained an
Samples were collected during the summer of 2016 from bedding EDS spectrum corresponding to the 3-μm pixel area. The quantifi-
mortar in the masonry of the city wall at the Privernum archaeolog- cation was performed using 2 × 2 binning that aggregates the counts
ical site, near Rome, Italy, and stored under ambient conditions. in a 2 pixel–by–2 pixel area before quantifying the elemental com-
The analyses described here were performed on 10 different position. Quantification of the binned EDS data was performed
samples from the Privernum archaeological site, and the results using Bruker’s Esprit 2.1 software with the built-in Linemarker
are representative of the trends observed. PB-ZAF correction. This correction accounts for background radi-
ation, atomic number effects, absorption of x-rays by the specimen,
Roman-inspired concrete samples and fluorescence. Individual EDS maps showing the atomic percent
In this study, we investigated two Roman-inspired concrete formu- of each element (e.g., calcium and silicon) present at each 3 μm–by–
lations (Mix 1 and Mix 2). Mix 1 (a coarse aggregate–free mortar 3 μm pixel were then extracted. The quantified EDS maps were pro-
that was used in our self-healing experiments) consisted of OPC, cessed using custom MATLAB scripts (41) to convert the pixels into
pulverized fly ash, sand, and water, at a ratio of (1:0.2:2:1). To this matrices where each entry represents the atomic percent of calcium,
mixture, quicklime was incorporated, spanning concentrations silicon, magnesium, aluminum, oxygen, or other elements present
varying from 7.5 to 15 mass %. To evaluate the effects of quicklime in each pixel. The relative ratios of these elements of interest were
addition on a representative industrially relevant concrete formula- plotted as ternary diagrams.
tion, we prepared Mix 2, which consisted of OPC, pulverized fly ash, Each EDS dataset was visualized using fuzzy c-means clustering
sand, water, superplasticizer, and coarse aggregate at a ratio of (Fuzzy Logic Toolbox, MATLAB R2018a). Fuzzy clustering differs
(1:0.4:2.5:0.7:0.01:3). To this mixture, quicklime was incorporated from hard clustering in that each data point may be a part of mul-
as a partial substitution of aggregates at a concentration of 4.3 tiple clusters. This assumption holds for this purpose, because a
mass %. given 3 μm–by–3 μm pixel of the measured area may encompass
The compositional details of each of these components are as more than one phase. Data points were ultimately divided into clus-
follows: OPC was CEM I 42.5 and conformed to EN 197-1 (OPC ters using the highest membership value for each data point. Using
type I of ASTM 150); fly ash conformed to EN 450-1 (class F of this technique, each dataset was divided into clusters, and the
ASTM C680); sand conformed to EN 12620 (or ASTM C33); number of clusters was increased until the map was appropriately
coarse aggregate conformed to EN 12620 (or ASTM C33); super- segmented. Clustering was performed on an n-by-5 array of EDS
plasticizer was Masterglenium Sky 799; water conformed to EN data, where n is equal to the number of pixels in the EDS map of
1008; and quicklime (with particle sizes of <4 mm) exhibited a interest, and each column contained the quantity of each of the
CaO content above 96.9%. The self-healing experiments and the five elements of interest (oxygen, silicon, calcium, magnesium,
mortar formulations (the Mix 1 samples and their quicklime-free and aluminum) in atomic percent.
controls) were cast into cylindrical molds measuring 10 cm in diam-
eter and 10 cm in height and cured under water for 28 days before Powder x-ray diffraction
use. The industrially relevant concrete formulations (the Mix 2 XRD samples were prepared by extracting visible lime clasts from
samples and their quicklime-free controls) were cast as rectangular mortar samples. After lime clast extraction, the remaining matrix
prisms measuring 360 mm in length, 120 mm in width, and 120 mm was manually scraped from the sample to reduce the potential for
in height. To determine the volumetric stability of the Mix 2 con- cross-contaimation. Samples were then ground to a fine powder
crete sample, drying shrinkage was measured in accordance with using an agate mortar and pestle no more than 24 hours before
the Swiss norm SIA 2621/F protocol, after 90 and 365 days. data acquisition.

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 10 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Powder XRD was performed in the X-Ray Diffraction Facility at of each test sample were rejoined with an interior crack spacing of
the Center for Materials Science and Engineering at the Massachu- 0.5 ± 0.1 mm and introduced into a metered flow circuit.
setts Institute of Technology. Diffraction data were collected with a 2) Crack validation: To verify the intact nature of the cracks for
PANalytical X’Pert Pro diffractometer using copper Kα radiation in the mounted samples, water was pumped through the flow circuit
a Bragg-Brentano geometry. The scans were performed from 10° to equipped with electronic flow meters with a precision of ±0.5 liters/
90° 2θ at increments of 0.008. Each increment was acquired for hour, before any preconditioning or crack healing measurements
50.165 s. Powder XRD phase identification was performed with were performed. A continuous, uninterrupted flow of water
the HighScore Plus software using the ICDD PDF-4 database. through the test sample was used to verify the crack’s dimensional
stability.
Raman spectroscopy 3) Specimen preconditioning: To discriminate between intrinsic
Raman spectroscopy was performed with a WiTec Alpha 300R con- small crack healing mechanisms for sub–0.2-mm-wide cracks,
focal Raman microscope. The spectra were acquired with a thermo- which have been reported previously for most concrete formula-
electrically cooled charge-coupled device detector (DU401A-BV, tions (45) and other potential large-scale self-healing mechanisms,
Andor, UK) placed behind a grating (600 g/mm) spectrograph the samples were first subjected to a 2-month-long period of se-
(UHTS 300, WITec, Ulm, Germany) with a spectral resolution of quential wet/dry cycling, as described previously (45).
3 cm−1. A Research Electro-Optics 35-mW helium-neon 633-nm 4) Crack healing: After the initial preconditioning step, the
laser was used at 33% power to collect Raman spectra from the samples were subjected to a continuous source of water flow over
lime clasts. A Zeiss EC Epiplan-Neofluar 100× [numerical aperture the course of 30 days. In these studies, a measured decrease in
(NA), 0.9] objective lens was used for imaging, and the spectra were flow rate (compared to that of the lime clast–free controls) was pre-
collected for 1.0 s per point. The 400 μm–by–400 μm area presented dicted to correspond to potential self-healing (crack-filling) activity.

Downloaded from https://www.science.org on February 12, 2024

in Fig. 3E and fig. S4 was divided into a 250-by-250 grid, for a total
resolution of 1.6 μm per pixel. The 300 μm–by–300 μm area in fig. Petrographic analysis
S5 was divided into a 120-by-120 grid for a resolution of 2.5 μm per Petrographic analysis was performed using a Keyence VHX-950F
pixel. The Raman map shown in fig. S2 and spectra used for calcu- digital microscope on polished thin sections measuring ca. 25 μm
lating the peak widths in fig. S6 were collected using a Research in thickness to examine the mineralogy of the filled cracks, the lime
Electro-Optics 75-mW 532-nm laser at 25% power and a Zeiss clasts, and the surrounding matrix material.
LD EC Epiplan-Neofluar 50× (NA, 0.55) objective lens. The
Raman map in fig. S2 was generated from spectra collected for
0.6 s each, at a spatial resolution of 4 μm per pixel over the 200 Supplementary Materials
μm–by–200 μm scan area. Single-point spectra of various lime in- This PDF file includes:
Supplementary Text
clusions were collected for 1.0 s each for fig. S6. Using a 532-nm
Figs. S1 to S9
laser, only the υ1 carbonate peak was detectable above the fluores-
cence signal in the lime inclusions. All spectra used a grating at 600
g/mm for data collection.
REFERENCES AND NOTES
Processing of the Raman data was done using WiTec’s Project 5 1. J. Allwood, J. Cullen, Sustainable Materials With Both Eyes Open (UIT Cambridge, 2011).
software. Background subtracted spectra were identified using the 2. J. R. Mackechnie, M. G. Alexander, Using durability to enhance concrete sustainability.
built-in k-means clustering function. Known spectra (e.g., resin, J. Green Build. 4, 52–60 (2009).
calcite, and quartz) identified with the k-means clustering were sub- 3. L. Bertrand, C. Gervais, A. Masic, L. Robbiola, Paleo-inspired systems: Durability, sustain-
tracted from the rest until distinct spectra remained (demixing). ability, and remarkable properties. Angew. Chem. Int. Ed. Engl. 57, 7288–7295 (2018).
4. K. T. Faber, F. Casadio, A. Masic, L. Robbiola, M. Walton, Looking back, looking forward:
The spectra presented here are the demixed average spectra of the
Materials science in art, archaeology, and art conservation. Annu. Rev. Mat. Res. 51,
identified clusters. Raman maps were created in WiTec Project 5 435–460 (2021).
showing the intensity of the identified component spectra in the 5. A. Palomo, P. Monteiro, P. Martauz, V. Bilek, A. Fernandez-Jimenez, Hybrid binders: A
background-subtracted scan data. Color intensity is proportional journey from the past to a sustainable future (opus caementicium futurum). Cem. Concr.
to the intensity of the component spectrum’s intensity at that Res. 124, 105829 (2019).
point, and color mixing indicates that the spectrum at that point 6. V. Pollio, F. Granger, British Library. Harley 2767, On architecture (Harvard Univ. Press, 1931).
7. C. G. Malacrino, Constructing the Ancient World: Architectural Techniques of the Greeks and
is a combination of component phases.
Romans (J. Paul Getty Museum, 2010).
8. L. C. Lancaster, Mortars and plasters—How mortars were made. The literary sources. Ar-
Self-healing experiments chaeol. Anthropol. Sci. 13, 192 (2021).
After mechanical fracturing, the re-mated sides of the test samples 9. M. Jackson, D. Deocampo, F. Marra, B. Scheetz, Mid-Pleistocene pozzolanic volcanic ash in
(and lime clast–free controls) were introduced into a flow circuit ancient Roman concretes. Geoarchaeology 25, 36–74 (2010).
(Fig. 5B) and subjected to a water flow that was constantly measured 10. F. Massazza, M. Pezzuoli, Some teachings of a Roman concrete, in Mortars, Cemeilts and
Grouts Used in the Conservation of Historic Buildings (ICCROM, 1981), pp. 219–245.
to confirm the consistency of the crack. The self-healing experi-
11. M. D. Jackson, S. R. Mulcahy, H. Chen, Y. Li, Q. Li, P. Cappelletti, H.-R. Wenk, Phillipsite and
ments consisted of four sequential experimental phases: (i) speci- Al-tobermorite mineral cements produced through low-temperature water-rock reactions
men preparation, (ii) crack validation, (iii) specimen in Roman marine concrete. Am. Mineral. 102, 1435–1450 (2017).
preconditioning, and (iv) crack healing. 12. M. D. Jackson, G. Vola, E. Gotti, B. Zanga, Sea-water concretes and their material charac-
These experimental phases are reported as follows: teristics, in Building for Eternity: The History and Technology of Roman Concrete Engi-
1) Specimen preparation: After casting, the cylindrical test neering in the Sea, J. P. Oleson, Ed. (Oxbow Books, 2014), pp. 141–188.

samples were cured underwater at room temperature for 28 days

before being split longitudinally. After air drying, the two halves

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 11 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

13. G. Artioli, M. Secco, A. Addis, The Vitruvian legacy: Mortars and binders before and after the 42. U. Wehrmeister, D. E. Jacob, A. L. Soldati, N. Loges, T. Häger, W. Hofmeister, Amorphous,
Roman world, in The Contribution of Mineralogy to Cultural Heritage (Mineralogical Society nanocrystalline and crystalline calcium carbonates in biological materials. J. Raman Spec-
of Great Britain and Ireland, 2019), pp. 151–202. trosc. 42, 926–935 (2011).
14. J. P. Oleson, L. Bottalico, C. Brandon, R. Cucitore, E. Gotti, R. L. Hohlfelder, Reproducing a 43. M. D. Jackson, G. Vola, Appendix 4: Compositional Analyses of Concretes Drilled from
Roman maritime structure with Vitruvian pozzolanic concrete. J. Roman Archaeol. 19, Harbour Structures by ROMACONS, in Building for Eternity: The History and Technology of
29–52 (2006). Roman Concrete Engineering in the Sea, J. P. Oleson, Ed. (Oxbow Books, 2014), pp. 285–306.
15. R. L. Hohlfelder, C. Brandon, P. Oleson, Building a Roman Pila in the sea—Experimental 44. C. Pesce, M. C. Godina, A. Henry, G. Pesce, Towards a better understanding of hot-mixed
archaeology at Brindisi, Italy, September 2004. Int. J. Naut. Archaeol. 34, 123–127 (2005). mortars for the conservation of historic buildings: The role of water temperature and steam
16. H. Lechtman, L. Hobbs, W. D. Kingery, Roman concrete and the roman architectural revo- during lime slaking. Herit. Sci. 9, 72 (2021).
lution, in Ceramics and Civilization, Volume 3: High Technology Ceramics: Past, Present, Future 45. W. Zhang, Q. Zheng, A. Ashour, B. Han, Self-healing cement concrete composites for re-
(American Ceramics Society, 1986), vol. 3, pp. 81–128. silient infrastructures: A review. Compos. Part B Eng. 189, 107892 (2020).
17. Pliny, D. E. Eichholz, Natural History, Volume X: Books 36–37 ([electronic resource] translated 46. B. Xu, M. B. Toffolo, L. Regev, E. Boaretto, K. M. Poduska, G. Mor, L. Addadi, S. Weiner,
by D.E. Eichholz, Harvard Univ. Press, 1962). Y. Politi, M. Goder-Goldberger, Structural differences in archaeologically relevant calcite.
18. A. B. Leslie, J. J. Hughes, Binder microstructure in lime mortars: Implications for the in- Anal. Methods 7, 9304–9309 (2015).
terpretation of analysis results. Q. J. Eng. Geol. Hydrogeol. 35, 257–263 (2002). 47. K. M. Poduska, L. Regev, E. Boaretto, L. Addadi, S. Weiner, L. Kronik, S. Curtarolo, Decoupling
19. A. Forster, Hot-lime mortars: A current perspective. J. Archit. Conserv. 10, 7–27 (2004). local disorder and optical effects in infrared spectra: Differentiating between calcites with
20. J. Elsen, Microscopy of historic mortars—A review. Cem. Concr. Res. 36, 1416–1424 (2006). different origins. Adv. Mater. 23, 550–554 (2011).
21. J. Valek, A. Zeman, Characteristics of lime particles in hot mixed mortars, in Proceedings of 48. E. Dubina, J. Plank, L. Korat, J. Strupi-Šuput, L. Black, Influence of water vapour and carbon
the 12th Euroseminar on Microscopy Applied to Building Materials (Dortmund, 2009). dioxide on free lime during storage at 80°C, studied by Raman spectroscopy. Spectrochim.
22. L. M. Seymour, D. Keenan-Jones, G. L. Zanzi, J. C. Weaver, A. Masic, Reactive ceramic ag- Acta Part Mol. Biomol. Spectrosc. , 299–303 (2013).
gregates in mortars from ancient water infrastructure serving Rome and Pompeii. Cell Rep. 49. L. Randazzo, M. Ricca, S. Ruffolo, M. Aquino, B. Davidde Petriaggi, F. Enei, M. F. La Russa, An
Phys. Sci. 3, 101024 (2022). integrated analytical approach to define the compositional and textural features of mortars
23. D. Jackson Marie, R. Chae Sejung, R. Mulcahy Sean, C. Meral, R. Taylor, P. Li, A.-H. Emwas, used in the underwater archaeological site of Castrum Novum (Santa Marinella, Rome,

Downloaded from https://www.science.org on February 12, 2024

J. Moon, S. Yoon, G. Vola, H.-R. Wenk, J. M. Monteiro Paulo, Unlocking the secrets of Al- Italy). Minerals 9, 268 (2019).
tobermorite in Roman seawater concrete. Am. Mineral. 98, 1669–1687 (2013). 50. M. P. Cato, A. Dalby, On farming: De agricultura (Prospect Books, 1998).
24. M. D. Jackson, E. N. Landis, P. F. Brune, M. Vitti, H. Chen, Q. Li, M. Kunz, H.-R. Wenk, 51. H. Jedrzejewska, New methods in the investigation of ancient mortars, in Archaeological
P. J. M. Monteiro, A. R. Ingraffea, Mechanical resilience and cementitious processes in Chemistry: A Symposium, M. Levey, Ed. (University of Pennsylvania Press, 1967),
Imperial Roman architectural mortar. Proc. Natl. Acad. Sci. 111, 18484–18489 (2014). pp. 147–166.
25. T. Vanorio, W. Kanitpanyacharoen, Rock physics of fibrous rocks akin to Roman concrete 52. J.-M. Commandré, S. Salvador, A. Nzihou, Reactivity of laboratory and industrial limes.
explains uplifts at Campi Flegrei Caldera. Science 349, 617–621 (2015). Chem. Eng. Res. Des. 85, 473–480 (2007).
26. J. Elsen, O. Cizer, R. Snellings, Lessons from a lost technology: The secrets of Roman 53. M. J. Abdolhosseini Qomi, F. Ulm, R. J. Pellenq, Evidence on the dual nature of aluminum in
concrete. Am. Mineral. 98, 1917–1918 (2013). the calcium-silicate-hydrates based on atomistic simulations. J. Am. Ceram. Soc. 95,
27. M. D. Jackson, G. Vola, D. Všianský, J. P. Oleson, B. E. Scheetz, C. Brandon, R. L. Hohlfelder, 1128–1137 (2012).
Cement microstructures and durability in ancient Roman seawater concretes, in Historic 54. G. K. Smith, Calcite straw stalactites growing from concrete structures. Cave Karst Sci. 43,
Mortars (Springer Netherlands, 2012), pp. 49–76. 4–10 (2016).
28. M. Jackson, J. Oleson, J. Moon, Y. Zhang, H. Chen, M. Gudmundsson, Extreme durability in 55. P. L. Broughton, Morphogenesis and microstructure of concrete-derived calthemites.
ancient Roman concretes. Am. Ceram. Soc. Bull. 97, 22–28 (2018). Environ. Earth Sci. 79, 245 (2020).
29. L. M. Seymour, N. Tamura, M. D. Jackson, A. Masic, Reactive binder and aggregate inter- 56. H. Huang, G. Ye, C. Qian, E. Schlangen, Self-healing in cementitious materials: Materials,
facial zones in the mortar of Tomb of Caecilia Metella concrete, 1C BCE, Rome. J. Am. Ceram. methods and service conditions. Mater. Des. 92, 499–511 (2016).
Soc. 105, 1503–1518 (2022). 57. B. Lubelli, T. G. Nijland, R. P. J. van Hees, Self-healing of lime based mortars: Microscopy
30. S. Bruni, F. Cariati, P. Fermo, P. Cairati, G. Alessandrini, L. Toniolo, White lumps in fifth- to observations on case studies. Heron 56, 81–97 (2011).
seventeenth-century AD mortars from Northern Italy. Archaeometry 39, 1–7 (1997). 58. D. Miriello, A. Bloise, G. M. Crisci, R. De Luca, B. De Nigris, A. Martellone, M. Osanna, R. Pace,
31. J. Elsen, A. Brutsaert, M. Deckers, R. Brulet, Microscopical study of ancient mortars from A. Pecci, N. Ruggieri, New compositional data on ancient mortars and plasters from
Tournai (Belgium). Mater Charact 53, 289–294 (2004). Pompeii (Campania – Southern Italy): Archaeometric results and considerations about
32. G. Vola, E. Gotti, C. Brandon, J. P. Oleson, R. L. Hohlfelder, Chemical, mineralogical and their time evolution. Mater Charact 146, 189–203 (2018).
petrographic characterization of Roman ancient hydraulic concretes cores from Santa 59. R. Gagné, Expansive agents, in Science and Technology of Concrete Admixtures (Elsevier,
Liberata, Italy, and Caesarea Palestinae, Israel. Period Mineral. 80, 317–338 (2011). 2016), pp. 441–456.
33. E. Uğurlu Sağın, H. E. Duran, H. Böke, Lime mortar technology in ancient eastern Roman 60. L. N. Plummer, E. Busenberg, The solubilities of calcite, aragonite and vaterite in CO2-H2O
provinces. J. Archaeol. Sci. Rep. 39, 103132 (2021). solutions between 0 and 90°C, and an evaluation of the aqueous model for the system
34. M. D. Jackson, J. Moon, E. Gotti, R. Taylor, S. R. Chae, M. Kunz, A. Emwas, C. Meral, CaCO3-CO2-H2O. Geochim. Cosmochim. Acta 46, 1011–1040 (1982).
P. Guttmann, P. Levitz, Material and elastic properties of Al-tobermorite in ancient Roman 61. M. Wu, B. Johannesson, M. Geiker, A review: Self-healing in cementitious materials and
seawater concrete. J. Am. Ceram. Soc. 96, 2598–2606 (2013). engineered cementitious composite as a self-healing material. Construct. Build Mater. 28,
35. K. Celik, M. D. Jackson, M. Mancio, C. Meral, A.-H. Emwas, P. K. Mehta, P. J. M. Monteiro, 571–583 (2012).
High-volume natural volcanic pozzolan and limestone powder as partial replacements for 62. G. Yildirim, M. Sahmaran, H. U. Ahmed, Influence of hydrated lime addition on the self-
portland cement in self-compacting and sustainable concrete. Cem. Concr. Compos. 45, healing capability of high-volume fly ash incorporated cementitious composites. J. Mater.
136–147 (2014). Civ. Eng. 27, 04014187 (2015).
36. M. D. Jackson, J. M. Logan, B. E. Scheetz, D. M. Deocampo, C. G. Cawood, F. Marra, M. Vitti, 63. C. Edvardsen, Water permeability and autogenous healing of cracks in concrete. ACI Mater.
L. Ungaro, Assessment of material characteristics of ancient concretes, Grande Aula, J. 96, 448–454 (1999).
Markets of Trajan, Rome. J. Archaeol. Sci. 36, 2481–2492 (2009). 64. M. Bechthold, J. C. Weaver, Materials science and architecture. Nat. Rev. Mater. 2,
37. L. C. Lancaster, Concrete Vaulted Construction in Imperial Rome. (Cambridge Univ. 17082 (2017).
Press, 2005).
38. J. MacFarlane, T. Vanorio, P. J. M. Monteiro, Multi-scale imaging, strength and permeability Acknowledgments: We would like to thank the archaeological area and the archaeological
measurements: Understanding the durability of Roman marine concrete. Construct. Build museum of Priverno, as well as the community of Priverno, particularly M. Cancellieri,
Mater. 272, 121812 (2021). N. Cassieri, and F. Di Mario for cooperation and assistance in providing access to the Privernum
39. F. M. Taylor, Cement Chemistry (Thomas Telford Publishing, ed. 2, 1997). archaeological site. The reconstruction data shown here were collected with the expertise of
Gianfranco Quaranta of Artech Lab and Roberto Scalesse of the Erresse Group, both of whom
40. C. Carde, R. François, Modelling the loss of strength and porosity increase due to the
are members of Associazione per la Ricerca e l’Educazione nell’Arte, Archeologia e Architettura
leaching of cement pastes. Cem. Concr. Compos. 21, 181–188 (1999).
(AREA3). The self-healing experiments were conducted with the help of P. Tudori of the Istituto
41. J. M. Maragh, J. C. Weaver, A. Masic, Large-scale micron-order 3D surface correlative
Meccanica dei Materiali (IMM). We dedicate this paper in memoriam of mastro Gilberto
chemical imaging of ancient Roman concrete. PLOS ONE 14, e0210710 (2019).

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 12 of 13

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Quarneti (1941-2021) for his contribution to ancient materials studies. Funding: Partial “Building materials” (PCT/IB2022/000353), filed 7 April 2022, and claiming priority to U.S.
funding for this project was provided by the MIT Edward H Linde (1962) Presidential Fellowship provisional application no. 63/171,770, entitled “Building materials” and filed 7 April 2021. The
program to Linda Seymour. Author contributions: Conceptualization: A.M. Supervision: A.M. other authors declare that they have no competing interests. Data and materials availability:
Methodology: L.M.S., J.M., P.S., M.D.T., J.C.W., and A.M. Investigation: L.M.S., J.M., P.S., M.D.T., All data needed to evaluate the conclusions in the paper are present in the paper and/or the
J.C.W., and A.M. Visualization: L.M.S., J.M., and J.C.W. Writing—original draft: L.M.S. and A.M. Supplementary Materials.
Writing—review and editing: L.M.S., J.M., P.S., M.T., J.C.W., and A.M. Competing interests: P.S.
and A.M. are stakeholders in DMAT srl. A.M. and L.M.S. are inventors of patent applications Submitted 27 May 2022
related to the results presented herein, assigned to the Massachusetts Institute of Technology Accepted 28 November 2022
( published as WO 2019/204776). DMAT srl has certain rights to commercialize these patent Published 6 January 2023
applications. P.S. and M.D.T. are listed as inventors of the International (PCT) application entitled 10.1126/sciadv.add1602

Downloaded from https://www.science.org on February 12, 2024

Seymour et al., Sci. Adv. 9, eadd1602 (2023) 6 January 2023 13 of 13