In 2017 a team led by Marie Jackson (University of Utah) published in American Mineralogist an analysis of cores drilled from the Roman harbour piers of Caesarea Maritima and Portus: structures that had been submerged in seawater for nearly two thousand years, the most aggressive condition possible for a cementitious binder. The finding isn't that Roman concrete simply "held up" — it's that it grew stronger over time, not weaker. The key is a mineral that almost never forms in modern concrete: aluminous tobermorite.
The chemistry: pozzolana versus clinker
Modern Portland concrete relies on the hydration of clinker — calcium silicates fired at 1,450°C — which produces calcium silicate hydrate (C-S-H), a phase that's relatively stable but chemically "closed": once it forms, it no longer evolves significantly. Roman opus caementicium follows a different path. The binder isn't clinker but quicklime mixed with volcanic pozzolana from the Phlegraean Fields (ash rich in reactive silica and alumina). The pozzolanic reaction between lime and pozzolana is slower than Portland setting — it takes weeks, not hours — but it produces crystalline phases that keep transforming for centuries.
When seawater seeps into microcracks in Roman concrete, it doesn't degrade it: it triggers the growth of aluminous tobermorite and phillipsite crystals directly inside the fissures, reinforcing the matrix instead of corroding it. It's a form of passive chemical self-repair that requires no human intervention. In modern Portland concrete, seawater does the exact opposite: it carries chlorides that attack the steel reinforcement, the one weak point in an otherwise durable system.
Why we no longer build this way
If Roman cement is so durable, why did it disappear for over a thousand years, and why don't we use it today? The answers are less romantic than the chemistry. First: pozzolana from the Phlegraean Fields is a geographically localised resource, not available everywhere in the world at the same quality — Portland is standardisable and can be produced wherever there's limestone. Second: Roman opus caementicium has no tensile strength and can't be reinforced — it works for domes and piers in pure compression, not for the slabs, cantilevers or slender structures that reinforced concrete makes possible. Third: the slow setting times (weeks) are incompatible with industrial construction schedules. The Pantheon took years to be cast, layer upon layer, with aggregates chosen for decreasing density from the base to the top of the dome — a process about as far as you can get from casting a modern floor slab in a single day.
The point isn't "going backwards". It's that research into low-impact cements (see the in-depth piece on geopolymers and low-carbon cements) explicitly looks to Roman pozzolanic chemistry as a model: activated aluminosilicate binders that form durable phases without going through the high-temperature firing of clinker. The difference is that today the reactivity of pozzolana or its industrial substitutes (fly ash, blast-furnace slag) can be engineered, instead of depending on one specific volcanic deposit.
The real comparison: like-for-like structures
An honest comparison means not comparing apples with oranges. The Pantheon doesn't have to withstand traffic load cycles, has no reinforcement to protect from corrosion, and its mass (dome thickness up to 6 metres at the base) is oversized by any modern economic standard. The reinforced-concrete structures that show signs of decay today after 50-60 years — viaducts, multi-storey car parks, coastal buildings — almost always fail for the same reason: insufficient concrete cover, too high a water-cement ratio, an exposure class underestimated at the design stage. It isn't that "modern cement is inherently less durable" — it's that it's built to much tighter safety and cost margins, and preventive maintenance is often put off. Reinforced concrete designed to the EN 1992-1-1 exposure classes with correct cover (see the in-depth piece on concrete pathologies and diagnostics) has a design service life of 100 years even in a marine environment — a tenth of the Pantheon's, but still an order of magnitude beyond an ordinary building's service life.