There's a new paper out exploring some of the chemical mechanisms at work in Roman concrete. As per usual, it’s triggered a round of enthusiastic discussion of Roman concrete, and how its ability to last for millennia puts modern concrete (which often fails after a few decades) to shame.

These sorts of comparisons generally miss the important differences between modern and Roman concrete that are key for understanding why their lifespans are so different.

What the paper did

The paper investigated chemical structures that are frequently found within Roman concrete called lime clasts - essentially, small lumps of lime and other chemicals. Lime clasts have been studied previously in Roman maritime structures such as harbors, but this is (apparently) one of the first studies that looked at lime clasts in Roman concrete structures on land.

The chemical analysis of the clasts is beyond my ability to evaluate, but the researchers determined that the lime clasts might be the result of lime added directly to the concrete mix as quicklime (CaO), rather than lime that had been mixed with water (slaked lime, Ca(OH)2). (The MIT article suggests that the idea of Romans using quicklime rather than slaked lime in their concrete was novel, but the paper itself references several papers suggesting this mechanism, and I was able to find others as well.)

Previously it had been noticed that Roman concrete often had calcite filled (CaCO3) cracks in it. The researchers suspected that the lime clasts might act as a sort of self-healing mechanism for the concrete - water making its way through the concrete would pull along calcium ions from the lime clasts, which would then form calcite, sealing the cracks.

To test this, the researchers first made their own mixes of concrete which included various amounts of quicklime, resulting in a concrete with lime clasts. They then split apart test samples of the concrete, put them back together with a 0.5mm crack in them, and ran water through the samples. Over a period of 1-3 weeks, the cracks filled with calcite and healed. Cracks in conventional concrete samples, on the other hand, didn’t heal, and water continued to run through them.

Roman vs modern concrete

It's an interesting result, but it's important to put this in context. People have assumed that applying this mechanism to modern concrete might solve the problem of comparatively short lifetimes of modern concrete, and indeed the researchers appear to be moving forward with commercializing this idea. But at the moment its relevance to modern concrete is pretty unclear.

The overwhelming majority of modern concrete is reinforced concrete - concrete that has had some type of steel embedded in it. Usually this is in the form of bars (rebar), but it might also be mesh, or fibers, or steel cable. Steel is stronger than concrete, particularly in tension (reinforcing steel has perhaps 10-15x the compressive strength of concrete, but more than 100x the tensile strength of concrete), and a comparatively small amount of steel can greatly increase the strength of a concrete element. By adding steel, you can make shallow concrete elements (beams, slabs, etc.) that can still span long distances and that wouldn’t be possible if the concrete were unreinforced.

Concrete is also brittle, whereas steel is ductile - if a plain concrete element fails, it’s likely to fail suddenly without warning, whereas a steel element will (generally) stretch and sag significantly before it fails, absorbing a lot of energy in the process. This makes reinforced concrete fundamentally safer than unreinforced concrete - if you have a lot of warning before a structure fails, you have time to safely get out of the building. For this reason, structural concrete is often required by code to have some minimum amount of steel reinforcing in it, and concrete that might experience large sudden loads in unpredictable ways (such as from an earthquake) is required to have a LOT of additional reinforcing. Most buildings built in zones of very high seismicity aren’t actually designed to come through the earthquake undamaged - they’re merely designed to not catastrophically collapse so people can safely get out.

(Earthquake design might seem like something that you only need to worry about in a few places, but most of the US can theoretically see a surprisingly strong earthquake and the buildings must be designed accordingly.)

But while reinforcement provides a lot of benefits, it has drawbacks. The primary one is that, over time, the steel in concrete corrodes. This is the result of two mechanisms - chloride ions making their way through the concrete, and concrete absorbing CO2 over time (though the second one happens much more slowly). As the steel corrodes, it expands, putting internal pressure on the concrete, eventually resulting in cracking and spalls (chunks of concrete that have fallen off).

How quickly this happens depends on a lot of factors. Concrete exposed to weather or water will corrode faster than concrete that isn't. Concrete where the rebar is farther from the surface of the concrete will last longer than concrete where the steel is closer to the surface. Concrete exposed to harsh chemicals such as salts or sulfates will corrode faster than concrete that isn't.

The comparatively short lifespan of modern concrete is overwhelmingly the result of corrosion-induced failure. Unchecked, reinforced concrete exposed to the elements will often start to decay in a few decades or even less. Precast concrete parking garages, for instance, are exposed to a lot of weather, since they’re open-air structures and vehicles bring moisture and road salts inside them. And a precast garage will often have many exposed steel elements, since steel plates stitch the pieces of concrete together. A precast garage might have a design life of 50 years, and often need very substantial repairs much earlier. Roman concrete, however, is unreinforced, and doesn’t have this failure mechanism.

This type of failure is exacerbated by the fact that modern concrete is designed to come up to strength very quickly, which results in numerous small cracks caused by shrinkage strains in the hardened concrete. These cracks make it easier for water to reach the steel, accelerating the process of corrosion. They also make the concrete more susceptible to other types of decay like freeze-thaw damage. Roman concrete, on the other hand, cured much more slowly.

If we wanted to build more durable concrete structures, the most important thing would be to remove or minimize this failure mechanism, and structures designed for long lives often do. Buddhist or Hindu temples, for instance, will use unreinforced concrete, or concrete with stainless steel rebar, and often have 1000-year design lives (though whether they will actually survive 1000 years is another question). Stainless steel rebar advocates like to trot out a concrete pier in Mexico built in 1941 with stainless steel rebar, which has needed no major repair work despite being in a highly corrosive environment. The conventionally reinforced concrete pier next to it, built 20 years later, has completely decayed and collapsed:

If we have the possibility of building more durable concrete buildings, why don't we?

This doesn't seem that complicated to me. Using unreinforced concrete dramatically limits the sort of construction you can do - even if the code allows it, you’re basically limited to only using concrete in compression. Without reinforcing, modern concrete buildings and bridges would be largely impossible.

Other methods of reducing reinforcement corrosion also have drawbacks, especially cost. Stainless steel rebar is four to six times as expensive as normal rebar. Epoxy coated rebar (commonly used on bridge construction in the US) is also more expensive, and though it can slow down corrosion, it won’t stop it. Basalt rebar won’t corrode (as far as I know) but can apparently decay in other ways.

Adding cost to a building to potentially extend its lifespan is often tough to make the numbers work for a developer. Well-made reinforced concrete that’s protected from the weather can last over a century, so the net present value of any additional lifespan beyond that is pretty low. It's much more likely that the building will be torn down for other reasons long before the concrete fails.

(I think the case for corrosion-resistant construction for outdoor infrastructure is much stronger, and we do sometimes see DOTs requiring stainless steel rebar for bridge and overpass construction.)

One other key point is that self-healing isn’t a unique property of Roman concrete. Modern concrete will also self-heal small cracks, though it appears to do so with a different mechanism than lime clasts. The (apparently) notable thing about the Roman self-healing is the size of the cracks that it can heal - modern concrete can heal cracks up to 0.2-0.3mm, whereas the lime-clast mix could heal cracks up to 0.5mm. It’s possible that the ability to heal larger cracks will significantly extend the lifespan of concrete, but it just might not matter that much. After all, the self-healing ability of normal concrete doesn’t prevent it from failing.

And even if it does extend the lifespan of concrete, it’s not obvious how much a difference that would make in practice. As we’ve seen, we already have ways to make concrete last longer (including with coatings that can seal cracks up to 0.5mm), and it’s very unclear whether this would be a meaningful addition to that palette.

Construction Physics

Roman vs Modern Concrete