The AI Journal

The breakwater at Pozzuoli, on the Bay of Naples, has been submerged in seawater for roughly two thousand years. Modern harbour concrete placed in similar conditions typically begins to deteriorate within decades: chloride ions penetrate the matrix, corrode the steel reinforcement, and the structure spalls from within. A 50-year service life is standard design expectation for modern marine concrete; some structures require major intervention within two decades. The Pozzuoli piers, built in the first century BCE from volcanic ash and hydrated lime, contain no steel, but that alone does not explain their condition. Something has been happening inside them that does not happen inside modern marine concrete, and it is not what most accounts of Roman concrete suggest.

When Marie Jackson and her colleagues examined drill cores from Roman harbour structures along the central Italian coast — cores collected by the ROMACONS project between 2002 and 2009 — they found crystals of a mineral called aluminous tobermorite growing in the relict lime particles of the mortar. Tobermorite is a layered calcium-silicate-hydrate that is exceptionally difficult to synthesise under ordinary conditions; it typically forms in hydrothermal environments at elevated temperatures and pressures. Yet here it was, in concrete that had cured at ambient seawater temperature, in structures that had been immersed for two millennia. Jackson's team showed that the tobermorite had not formed during the original setting of the concrete. Rather, it had crystallised progressively over centuries as seawater percolated through the porous mortar, dissolving components of the volcanic ash and precipitating new mineral phases from the highly alkaline pore fluids. A companion mineral, phillipsite — a zeolite — formed through the same low-temperature water-rock reactions, lining pores and crack surfaces with interlocking crystalline plates that increased the concrete's cohesion and resistance to fracture over time.

This is not pozzolanic reaction. The pozzolanic reaction — the chemical interaction between hydrated lime and reactive volcanic ash that produces calcium-aluminium-silicate-hydrate binder — occurs in the first months and years after mixing. It is the mechanism by which all Roman concrete, whether terrestrial or marine, gains its initial strength. What Jackson found in the marine structures was something else: a post-pozzolanic process, driven by the chemistry of seawater, that continued for centuries after the pozzolanic reaction had gone to completion. In the Pozzuoli cores, the pozzolanic reaction had consumed the hydrated lime within the first few years, as highly alkaline pore solutions reacted with the glassy volcanic ash. What followed — the dissolution of ash components by infiltrating seawater, the precipitation of phillipsite in pore spaces, and the eventual crystallisation of tobermorite at ambient temperature — is a process with no analogue in modern concrete chemistry. The concrete was not merely durable. It was, in a precise mineralogical sense, still curing.

The popular narrative about Roman concrete — that its volcanic-ash-and-lime formula produced a material inherently superior to modern Portland cement — treats this marine mineralisation as just one instance of a general phenomenon. It is not. The terrestrial Roman structures most often cited alongside the harbours — the Pantheon dome, Trajan's Markets, the Tomb of Caecilia Metella — do not contain aluminous tobermorite. Jackson's analysis of the Grande Aula of Trajan's Markets, built around 110 CE, revealed a different crystalline phase: strätlingite, a platy calcium-aluminosilicate mineral that precipitates in interfacial zones between volcanic scoria aggregate and the cementitious matrix, toughening the weakest links in the mortar fabric. Strätlingite forms through prolonged pozzolanic reaction at ambient temperature, not through seawater percolation. It is a genuine toughening agent — the Jackson group showed that mortar reproductions gained fracture toughness over 180 days through strätlingite crystallisation — but it is a mechanism that modern high-alumina pozzolanic cements can approximate. It does not require two thousand years of seawater immersion to produce.

The distinction matters because it changes what engineers should be trying to learn from Roman concrete. If the durability were a property of the recipe, the lesson would be broad: adopt the recipe. But the marine durability is a property of the interaction between the recipe and the marine environment. Jackson's group was explicit: the tobermorite and phillipsite crystallisation in the harbour mortars depends on open chemical exchange with seawater — a process that would be considered corrosive in a modern reinforced structure. The Romans' material was not inert, as modern concrete is designed to be. It was chemically reactive with its surroundings, and in the specific case of seawater immersion, that reactivity was constructive rather than destructive. The implication for engineering is not that we should copy the Roman recipe but that we should copy the Roman environment: design marine concrete to undergo controlled mineralogical exchange with seawater rather than to resist it.

The strongest objection is the Pantheon. Its unreinforced concrete dome, completed around 126 CE, has stood for nearly nineteen centuries without any seawater exposure. If Roman concrete's exceptional durability were marine-specific, the Pantheon would be a counterexample: a terrestrial structure that outlasts virtually everything modern engineering has produced, using the same pozzolanic ash-lime mortar. The objection is forceful because the Pantheon is genuinely remarkable — but it is remarkable for reasons that do not contradict the distinction drawn here. The dome's survival owes as much to structural geometry as to material chemistry: the progressively lighter aggregates graded from base to crown, the compression-only stress state that a hemispherical dome maintains, and the thick supporting walls that prevent tensile failure. These are engineering achievements, not mineralogical ones. The mortar itself, containing strätlingite and a low-calcium C-A-S-H binder, is good pozzolanic concrete, but it is not in a different category from what modern high-performance pozzolanic concrete can achieve. The Pantheon proves that Roman concrete was well made. It does not prove that the recipe produces a material that improves with age. The harbour piers do.

A more recent discovery complicates the picture without resolving it. In 2023, Admir Masic's group showed that Roman concrete was probably mixed with quicklime rather than slaked lime — a process they called "hot mixing" — leaving reactive lime clasts dispersed through the mortar. When cracks form, water infiltrates the clasts, dissolves calcium, and reprecipitates it to heal the fissure. This self-healing mechanism operates on a timescale of weeks, not millennia, and it works in terrestrial concrete without seawater. It is a genuine Roman innovation and it is transferable to modern practice. But it is a different mechanism from the tobermorite mineralisation, operating on a different timescale, solving a different problem. Hot mixing prevents crack propagation in the early life of a structure. Seawater mineralisation reconstructs the binding fabric over centuries. The two are complementary, not equivalent, and conflating them under the single banner of "Roman concrete durability" obscures which lesson applies where.

The lesson that actually transfers is narrow, and it is not the one most often drawn. Roman marine concrete was not superior because of its recipe. It was superior because its recipe, placed in seawater, initiated a mineralogical process that continues to this day — a process the Romans could not have fully understood and did not need to. The material specification matters, but so does the environment, and the environment is the variable that popular accounts routinely omit. The harbour at Pozzuoli is not a monument to lost knowledge. It is a two-thousand-year-old experiment that is still running, and the result is a material that is stronger now than when it was placed. A concrete that grows stronger over millennia is not a matter of formula alone. It is a matter of formula meeting seawater.