The core difficulty is class imbalance. In ordinary concrete CT images, voids — the pores and microcracks that govern how water, chlorides and gases migrate through a structure — occupy only about 3.2 percent of the pixels. Cement paste fills roughly 64.4 percent and aggregate the remaining 32.4 percent. When a three-class network is trained on such skewed data, the optimizer can satisfy itself by simply predicting the dominant classes everywhere. The study’s own controlled experiment makes this vivid: a single-stage Vision Transformer–U-Net hybrid, trained on exactly the same data split, preprocessing and loss functions as the proposed method, collapsed onto the majority aggregate class. It achieved just 31.81 percent overall accuracy and a mean Intersection over Union of 10.61 percent, with F1-scores of essentially zero for both void and paste. In other words, the model learned to call nearly everything aggregate and nothing else.

The proposed alternative decomposes the problem along the physical hierarchy of the material itself. Stage 1 is a binary task: separate void from non-void, exploiting the fact that air and cracks attenuate X-rays weakly and appear dark. Stage 2 then classifies the remaining pixels as paste or aggregate, with void pixels masked out of the loss entirely through an ignore index so they cannot contaminate the second stage’s learning. At inference, the two predictions are merged: anything Stage 1 calls void becomes void, and everything else receives the Stage 2 verdict. Each stage therefore faces a milder imbalance than the original three-class problem, and void no longer competes with paste–aggregate boundaries for the network’s attention during optimization.

Both stages share the same backbone: a Vision Transformer encoder (ViT-B/16, pretrained on ImageNet) paired with a convolutional U-Net-style decoder. The choice is deliberate. Self-attention lets the encoder aggregate long-range context across a 224-by-224-pixel patch, useful when distinguishing fine pores from large aggregate grains, while the upsampling decoder reconstructs dense, full-resolution masks. Because the Vision Transformer produces a flat sequence of patch tokens rather than multi-scale feature maps, the decoder omits encoder skip connections and instead applies four bilinear upsampling blocks that rebuild the image from a 14-by-14 feature grid. Each stage carries roughly 86.6 million trainable parameters, all fine-tuned end-to-end on concrete patches with no concrete-specific pretraining.

Training was equally careful about imbalance. Each stage optimizes a weighted blend of cross-entropy, focal, Dice and IoU losses, with coefficients of 0.5, 0.25, 0.15 and 0.10 respectively. Stage 1 up-weights void pixels fifteenfold; Stage 2 mildly favors aggregate. The optimizer is AdamW with a learning rate of 5×10⁻⁵ under cosine annealing, early stopping on validation loss, and mixed-precision training in PyTorch. Crucially, the single-stage baseline received the same recipe, including void-upweighted losses — showing that reweighting alone, a common remedy in the literature, could not restore minority-class performance under this extreme skew. The gains come from task structure, not hyperparameter tuning.

The results on 151 held-out test patches are striking. The two-stage pipeline reached 94.58 percent overall accuracy, 87.50 percent mean IoU and 93.31 percent mean Dice, with class-wise F1-scores of 0.92 for void, 0.96 for paste and 0.92 for aggregate. Void recall hit 96.4 percent — the number that matters most, since missed pores translate directly into underestimated porosity and unreliable durability predictions. Bootstrap resampling over the test patches showed tiny standard deviations, with overall accuracy at 94.90 percent plus or minus 0.07 percent, indicating the performance is stable rather than a lucky draw. An ablation confirmed that removing either stage causes substantial degradation: Stage 2 alone cannot find voids at all, and Stage 1 alone cannot separate aggregate from paste.

The study also stress-tested the framework beyond its algorithmic training labels. On a second corpus of 82 images with human-prepared masks, held out twelve at a time for testing, the two-stage model outperformed a traditional enhanced-mask baseline on both 8-bit inputs and percentile-normalized 16-bit inputs, with the two encodings performing nearly identically on overall metrics. A full-slice demonstration on an unseen CT image predicted phase fractions of 3.64 percent void, 59.38 percent paste and 36.98 percent aggregate — closely tracking the training distribution and suggesting the pipeline can plausibly map whole specimens when preprocessing matches training assumptions. The authors are careful, however, to frame this as transfer plausibility rather than certified generalization.

The error analysis is refreshingly honest about where the method still struggles. The largest residual confusion lies at paste–aggregate interfaces, where similar X-ray attenuation and partial-volume mixing produced more than 365,000 swapped pixels in the test confusion matrix. Thin voids and microcracks are sometimes absorbed into paste, and beam-hardening streaks near dense aggregate can trigger false void calls. Because the merge rule forces every Stage 1 void prediction to remain void, errors cascade: the 3.6 percent of void pixels missed by Stage 1 can never be recovered, and false voids bypass Stage 2’s judgment entirely. For engineering use, the authors note, bulk porosity and aggregate grading from well-resolved regions are far more trustworthy than interface statistics or sub-resolution pore counts without local audit.

The limitations define a clear research agenda. All primary data come from a single specimen, one scan and one mix design, with patch-level rather than slice-level splitting, so neighboring windows from the same CT slice can straddle train and test folds. The reference labels themselves are algorithm-generated pseudo-labels, verified against manual annotation on only a 57-patch subset (88.64 percent pixel agreement), and the manual-label external check rests on just twelve test images. Inference is strictly two-dimensional, with no inter-slice consistency, and the two forward passes roughly double the computational cost of a one-pass model — 27.9 versus 14.6 milliseconds per patch on an RTX 2060, acceptable for offline analysis but not real-time deployment.

Even within those bounds, the study delivers a lesson that resonates far beyond concrete. When one class is vanishingly rare, the fix may not be a cleverer loss or a bigger model but a restructuring of the task itself — teaching the network to find the rare thing first, and only then sort the rest. As infrastructure agencies worldwide push toward digital twins of aging bridges, tunnels and nuclear containment structures, the microscopic pores inside concrete are where durability is decided. A two-stage trick that finally makes those pores visible to machines could quietly become one of the more consequential pieces of AI in civil engineering.

Subject of Research: Two-stage deep learning segmentation of void, paste and aggregate phases in concrete X-ray CT images under severe class imbalance

Article Title: Two-stage deep learning for segmenting in concrete X-ray CT images

Article References: Wang, Y. (2026). Two-stage deep learning for segmenting in concrete X-ray CT images. Results in Engineering, 32, Article 113430. https://doi.org/10.1016/j.rineng.2026.113430

Image Credits: AI Generated

DOI: Not provided

Keywords: concrete, X-ray computed tomography, deep learning, image segmentation, Vision Transformer, U-Net, class imbalance, porosity, cement paste, aggregate, civil engineering, digital twin

The research team, led by Tongning Cao and colleagues including Han Wang, Qi Dong, Fengjuan Wang, Zhiyong Liu, Chenyun Yu, Bin Peng, and Jinyang Jiang, set out to close a persistent gap in concrete durability science. Most laboratory studies of sulfate attack have been performed under simple immersion, wet-dry cycling, or atmospheric pressure conditions. Those regimes capture shallow-marine and terrestrial exposure reasonably well, but they ignore the defining feature of the deep-sea environment: sustained hydrostatic pressure ranging from several to tens of megapascals. At such pressures, the state of pore water, the morphology of the pore network, and the very mechanisms of ion transport inside concrete may be fundamentally altered, making conventional durability predictions unreliable for deep-water structures.

To probe these coupled effects, the researchers built a custom hydrostatic pressure-sulfate erosion apparatus capable of simultaneously controlling pressure, temperature, and sulfate exposure. The system, rated to a maximum of 25 megapascals with a pressure resolution of 0.1 megapascals, centers on a one-liter high-pressure reaction vessel pressurized by nitrogen gas. Cubic cement paste specimens with a water-to-binder ratio of 0.4 were cast, cured for 28 days, and then sealed with epoxy resin on five faces, leaving a single 20 by 20 millimeter face exposed. This clever geometry forced sulfate ions to travel along one well-defined direction, simplifying the interpretation of the depth-resolved measurements that followed.

Specimens were exposed to a 5 percent sodium sulfate solution at three gauge pressures: 0, 1.2, and 2.4 megapascals, for a total of 150 days. The solution was refreshed every 15 days to maintain aggressive conditions. At 60, 90, 120, and 150 days, replicate specimens were retrieved, sectioned into 2-millimeter-thick slices along the exposure direction, and treated with isopropanol to halt hydration. All sample handling took place in a nitrogen-controlled glove box to prevent carbonation from contaminating the chemistry. One portion of each powdered slice was analyzed for sulfate ion content by gravimetric barium sulfate titration, while the other was subjected to quantitative X-ray diffraction with Rietveld refinement, using aluminum oxide as an internal standard, to measure the amount of ettringite, also known as AFt, formed at each depth.

The results reveal a striking pressure dependence. At a depth of 9 millimeters, the sulfate concentration after 150 days rose from 1.6 percent under atmospheric pressure to 2.7 percent at 2.4 megapascals. Deeper in the material the contrast became even more dramatic: at 19 millimeters, the sulfate content after 150 days was 0.5 percent at 0 megapascals, but 0.9 percent at 1.2 megapascals and 1.2 percent at 2.4 megapascals. In other words, moderate pressures that are routine in coastal and shallow offshore engineering already boost sulfate penetration by roughly 50 to 80 percent, and the enhancement grows with both depth and time. Ettringite formation followed the same pattern, accumulating preferentially near the exposed surface and decaying inward, with concentrations at 5 millimeters reaching 1.69 percent at 0 megapascals, 2.6 percent at 1.2 megapascals, and 3.1 percent at 2.4 megapascals after 120 days.

Microscopy and porosimetry explained why. Field-emission scanning electron microscopy showed the corrosion products as interwoven needle-like and rod-like crystals, the classic habit of ettringite, while energy-dispersive spectroscopy confirmed the enrichment of calcium, aluminum, sulfur, and oxygen in ratios matching the mineral’s stoichiometry. Notably, ettringite crystals clustered around portlandite, suggesting that as sulfate ions consume calcium from the pore solution, portlandite dissolves to replenish it, sustaining the reaction and progressively degrading the matrix. Mercury intrusion porosimetry then quantified the structural toll: total porosity climbed from 14.93 percent at 0 megapascals to 19.02 percent at 1.2 megapascals and 25.81 percent at 2.4 megapascals, with the pore-size distribution shifting toward larger, better-connected channels.

This pore-structure evolution sets up a vicious feedback loop. Pressure-enhanced sulfate transport delivers more reactant deeper into the material, accelerating ettringite formation, which coarsens and connects the pore network. The modified network in turn raises the effective diffusion coefficient, allowing still faster sulfate ingress. According to Fick’s law, this increase in effective transport capacity strengthens concentration-gradient-driven migration, producing deeper penetration and higher interior sulfate contents. The authors’ simulations capture this nonlinear amplification: after 150 days, predicted sulfate concentrations at 20 millimeters were 112 percent higher at 1.2 megapascals and 290 percent higher at 2.4 megapascals than under atmospheric pressure. From a crystallization-pressure perspective, the elevated sulfate supply also drives the pore solution to supersaturation sooner, triggering earlier ettringite crystallization and shortening the material’s latent expansion period.

To turn these observations into a predictive tool, the team constructed a coupled sulfate transport-reaction model grounded in Fick’s second law and mass conservation. The model treats sulfate ions as either free in the pore solution or consumed by reaction with calcium and aluminate-bearing hydration products to form ettringite, following the stoichiometry that one mole of AFt requires two moles of external sulfate. The effective diffusion coefficient is computed from porosity, tortuosity, and the Stokes-Einstein estimate of sulfate diffusivity in water, with a phenomenological pressure term linking hydrostatic stress to porosity change through the circumferential tensile stress around idealized cylindrical pores. The equations were solved numerically with the finite volume method on hexahedral grids, accelerated by a multigrid V-cycle scheme implemented in MATLAB.

Validation against the 1.2 megapascal experiments showed good agreement between predicted and measured sulfate and ettringite profiles across the full 150-day exposure and the entire penetration depth, with minor discrepancies attributed to the finite thickness of the sliced samples. The framework provides, for the first time, an experimentally supported quantitative link between hydrostatic pressure, sulfate transport, reaction kinetics, and pore-structure evolution in cementitious materials. The authors caution that the study used cement paste under controlled laboratory conditions rather than full concrete in service, and that further validation across broader materials, environments, and timescales is needed before deep-ocean application. Even so, the message for engineers is clear and urgent: durability assessments for submerged and deep-buried structures that ignore hydrostatic pressure may substantially underestimate how fast sulfate attack proceeds, and how soon the ocean’s quiet chemistry begins to dismantle the concrete we trust to hold it back.

Subject of Research: Effect of hydrostatic pressure on sulfate transport and ettringite formation in cement paste

Article Title: Effect of hydrostatic pressure on transport-reaction processes of cement paste under sulfate attack

Article References: Cao, T., Wang, H., Dong, Q., Wang, F., Liu, Z., Yu, C., Peng, B., & Jiang, J. (2026). Effect of hydrostatic pressure on transport-reaction processes of cement paste under sulfate attack. Case Studies in Construction Materials, 25, Article e06588. https://doi.org/10.1016/j.cscm.2026.e06588

Image Credits: AI Generated

DOI: 10.1016/j.cscm.2026.e06588

Keywords: sulfate attack, hydrostatic pressure, cement paste, ettringite, concrete durability, deep-sea environment, ion transport, porosity, XRD-Rietveld, transport-reaction model, marine engineering, pore structure

The research, led by Rachida Idir of Cerema and University Gustave Eiffel together with colleagues from the University of Montpellier’s LMGC laboratory and IMT Mines Alès, including Katerina Ioannidou and Gwenn Le Saout, examined two commercially important types of carbon black in cement pastes. The first, furnace black, is the industrial standard, produced by partially burning heavy petroleum feedstocks in controlled furnaces. The second, acetylene black, is made by decomposing acetylene gas and is prized for its high purity and distinctive structure. The team varied the amount of carbon black, the water-to-cement ratio, and the dosage of superplasticizer, the chemical admixture used to keep workable concrete flowing without excess water, to map out how each parameter shaped the electrical behavior of the resulting paste.

The physics at the heart of the work is the percolation threshold, a concept borrowed from statistical physics that describes when disconnected particles suddenly link up into a continuous network. Ordinary cement paste is a poor conductor, relying mostly on ions dissolved in its pore water to carry current. Dispersed carbon black particles change that. Below a critical loading, the particles remain isolated islands, and the paste barely conducts better than before. Once that threshold is crossed, conductive pathways snake through the matrix and resistance plummets. In these experiments, the threshold generally appeared around 2 percent carbon black by weight of cement, and conductivity values above 1 siemens per meter, a level useful for practical sensing and heating applications, were typically reached at dosages of about 3 to 4 percent.

Under the conditions tested, acetylene black proved more effective than furnace black at building these conductive networks. That advantage matters because it means less additive is needed to reach a target conductivity, which in turn reduces both the material cost and the environmental footprint of the composite. The finding aligns with a growing body of literature on carbon-modified cementitious materials, which has explored everything from carbon nanotubes and nanofibers to carbon fibers and graphite as conductive fillers. Carbon black occupies an attractive middle ground in that landscape: it is far cheaper than nanotubes, easier to disperse than many high-aspect-ratio fillers, and available at industrial scale, making it a realistic candidate for construction applications where cost per cubic meter is decisive.

Conductive cement is not a laboratory curiosity. Researchers have already demonstrated electrically heated pavement systems that keep airport runways ice-free without salt, and self-sensing composites whose electrical resistance shifts as cracks form or loads change, allowing structures to report their own health. The piezoresistive effect in carbon-loaded cement means that squeezing the material alters its conductivity, so a bridge deck embedded with such sensors could, in principle, flag overloads or hidden damage continuously. Joule heating, where current passing through the resistive material generates warmth, underpins the deicing concept. What has been missing from many of these demonstrations is a rigorous accounting of the climate burden that the conductive additive adds to an already carbon-intensive material.

That accounting is the second pillar of the new study. The team performed a life cycle assessment focused on the global warming potential, or GWP, of the paste formulations, considering both the production of the carbon black itself and the full paste recipe. At the scale of the raw materials, the differences between acetylene black and furnace black were significant, reflecting the very different industrial processes behind them. But once the analysis moved to the scale of the complete paste, those differences shrank dramatically. The reason is sobering: cement clinker production, with its limestone calcination and kiln fuel emissions, so dominates the carbon footprint of the paste that the choice between two carbon blacks becomes a secondary consideration. Cement remained the overwhelming contributor to GWP across all formulations.

To connect functionality with climate impact in a single metric, the researchers introduced a novel indicator they call the Performance Impact Indicator, or PII, defined as the ratio of global warming potential to electrical conductivity. This simple ratio captures a design tension: adding more carbon black raises the GWP of the paste, but it also raises conductivity, and the question is which effect wins. The experiments delivered a counterintuitive and encouraging answer. Higher carbon black contents led to lower PII values, meaning that the conductivity gains outpaced the added emissions. In other words, once past the percolation threshold, each additional increment of carbon black buys more electrical performance per unit of climate impact, at least within the range investigated. For designers of smart concrete, this suggests that under-dosing the filler may be the worst of both worlds, delivering neither strong conductivity nor an efficient carbon-to-performance trade.

The study also looked beyond conventional supply chains to two alternative routes for producing carbon black, examining them from an environmental perspective. The first is a process associated with hydrogen co-production, in which methane is pyrolyzed, splitting natural gas into solid carbon and hydrogen gas rather than burning it. This route, sometimes linked to so-called turquoise hydrogen, has attracted attention as a way to produce both a low-carbon fuel and carbon black with potentially far lower emissions than the furnace process. The second alternative is based on waste tire pyrolysis, in which end-of-life tires are heated in the absence of oxygen to recover oil, gas, steel, and a recycled carbon black. Both scenarios suggest that lower-impact carbon black supply routes may deserve serious consideration for conductive cement applications, potentially decoupling the growth of smart concrete from additional fossil feedstock consumption while giving waste tires a second life.

The broader significance of the work lies in its framing. Multifunctional materials are often evaluated on performance alone, with sustainability treated as an afterthought or a marketing claim. By pairing systematic electrical measurements with life cycle thinking, and by proposing a metric that binds the two together, the French team offers a template for how emerging construction technologies should be assessed. The PII approach echoes earlier eco-efficiency work in cement science, which sought to measure how much functional service a structure delivers per unit of environmental burden. Applying that logic to conductive cement reveals a genuinely useful insight: the environmental case for these materials improves as the conductive network matures, provided the dosage is chosen wisely.

Challenges remain before carbon-black cement becomes a routine building material. Dispersion of nanoparticles in the harsh, alkaline environment of fresh cement is notoriously difficult, and researchers continue to explore techniques such as sonication and optimized admixture chemistry to achieve uniform networks. Durability over decades of weathering, long-term stability of the conductive pathways, and the cost of scaling up alternative carbon black production routes all require further study. The authors also note that their environmental comparison at the paste scale was limited to the GWP indicator, leaving other impact categories for future work. Yet the direction of travel is clear. As infrastructure ages and cities demand smarter, more resilient materials, cement that can carry both loads and current may find its place, and this study shows that the greenest version of that future depends not just on what goes into the concrete, but on where its carbon comes from.

Subject of Research: Conductive cement pastes modified with carbon black and their electrical performance and global warming potential

Article Title: Carbon black from multiple production routes in conductive cement pastes: balancing multifunctional performance and environmental burden

Article References: Idir, R., Souane, S. F., Touati, F., Ioannidou, K., & Le Saout, G. (2026). Carbon black from multiple production routes in conductive cement pastes: balancing multifunctional performance and environmental burden. Journal of Industrial Ecology, 30(4), 2071-2088. https://doi.org/10.1007/s44498-026-00140-x

Image Credits: AI Generated

DOI: 10.1007/s44498-026-00140-x

Keywords: carbon black, conductive cement, cement paste, percolation threshold, electrical conductivity, life cycle assessment, global warming potential, furnace black, acetylene black, waste tire pyrolysis, turquoise hydrogen, smart materials

The mineral at the heart of the problem is pyrrhotite, a non-stoichiometric iron sulfide with the general formula Fe1−xS, where x can range from zero to 0.125. When even less than one weight percent of this mineral is present in concrete aggregates, its oxidation can trigger a chain of reactions known as internal sulfate attack, or ISA. As pyrrhotite reacts with oxygen and moisture, it releases sulfate ions and acidic species into the surrounding cement paste. These ions then react with hydration products in the cement to form gypsum and ettringite, minerals whose growth generates expansive pressures that crack the concrete matrix. The Canadian standard CSA A23.1:19, Appendix P formally recognizes this combined mechanism of iron-sulfide oxidation and subsequent sulfate reaction as the driver of ISA deterioration.

The damage is not hypothetical. Aggregate from the Maskimo quarry in St. Boniface, Quebec, has been linked to severe ISA affecting more than 1,700 housing foundations in Trois-Rivières, Quebec, many of which suffered serious damage after only five to ten years in service. Earlier work by Rodrigues and colleagues identified pyrrhotite as the principal reactive mineral phase in the damaged Trois-Rivières concrete. In the Maskimo rock, pyrrhotite and pyrite occur in an approximate ratio of one to two, with pentlandite and chalcopyrite frequently found as accessory inclusions within them. The quarry itself consists of a noritic microgabbro dominated by plagioclase feldspar, orthopyroxene, magnesiohornblende, and biotite, with total sulfur contents ranging from 0.85 to 1.8 weight percent, all bound in sulfides.

To capture the deterioration in action, a team led by Nikolas Å. Oberhardt of the Norwegian University of Science and Technology, together with colleagues including Klaartje De Weerdt, examined a sequence of samples spanning the entire life cycle of the reaction. They studied fresh Maskimo aggregate, mortar bars subjected to accelerated mortar bar testing for 12, 27, and 52 weeks, and a severely damaged concrete drill core taken from a Trois-Rivières foundation wall. The accelerated test, described in the Canadian standard, involves repeatedly immersing mortar prisms in sodium hypochlorite bleach, twice weekly for three hours, with interim storage first at 80 degrees Celsius and 75 percent relative humidity to accelerate sulfide oxidation, and later at 4 degrees Celsius and full humidity to promote thaumasite formation. The mortar bars expanded by 0.22 percent after 12 weeks, 0.27 percent after 27 weeks, and 0.47 percent after 52 weeks.

The analytical centerpiece of the study was scanning electron microscopy-based automated mineralogy, which uses quantitative energy-dispersive spectroscopy to classify every pixel of a polished thin section into a mineral phase. The researchers deployed several analytical modes: coarse field mapping at 10 and 20 micrometer step sizes for bulk mineralogy, line scanning at 1 micrometer for quantitative bulk composition, and high-resolution field mapping at 2 micrometers to resolve the fine-grained oxidation products and cement hydration phases. Complementing this, thermomagnetic analysis measured the magnetic susceptibility of powdered samples from room temperature to 650 degrees Celsius, allowing the team to detect and quantify two pyrrhotite superstructures, the hexagonal antiferromagnetic NC-type and the monoclinic ferrimagnetic 4C-type, using the characteristic lambda transition near 220 to 245 degrees Celsius and the Curie point at 315 degrees Celsius.

The results reveal a strikingly ordered oxidation sequence. In the pristine aggregate, pyrrhotite grains showed only incipient surface oxidation along mineral boundaries, with minor magnetite, goethite, and slightly oxidized pyrrhotite as the characteristic secondary phases. After 12 weeks of accelerated exposure, oxidation had penetrated the mineral interiors through etch pitting, and an oxidation rim of iron oxides, mainly hematite and magnetite, had grown around the grain peripheries. Ferrihydrite, an iron oxyhydroxide, appeared at around 1.4 weight percent, typically adjacent to the iron oxides rather than directly on the pyrrhotite. By 27 weeks, the grain interiors were severely altered and largely replaced by iron oxides and oxyhydroxides, leaving pyrrhotite only as relic-like structures. After 52 weeks, the crystals were almost entirely consumed, with ferrihydrite rising to 1.79 weight percent and ettringite increasing from 0.13 to 0.37 weight percent.

The element heat maps told an equally important story about where the atoms went. Iron remained largely immobile, locked in place within the relic zones of the oxidizing grains as iron oxides and oxyhydroxides replaced the original sulfide. Sulfur, by contrast, leached away as sulfate anions into the cement paste, and the sulfur maps showed a progressive increase in sulfur concentration within the cement matrix as exposure time lengthened. In the field sample, the cement paste adjacent to reacted aggregates contained as much as 10 to 12 weight percent sulfur. The team also identified a cryptocrystalline composite phase, labeled FeOSCa, averaging roughly 34 percent oxygen, 10.5 percent sulfur, 24.5 percent calcium, and 31 percent iron, which formed along cracks and near oxidizing pyrrhotite and may warrant further study with transmission electron microscopy or electron probe microanalysis.

Comparing the laboratory samples with the field concrete proved especially revealing. The field sample contained between 1.06 and 2.30 weight percent pyrrhotite, far more than the 0.19 to 0.37 weight percent measured in the mortar bars and the 0.50 weight percent in the fresh aggregate, underscoring the link between the amount of reactive pyrrhotite available and the magnitude of expansion. The field sections showed gypsum and the FeOSCa composite phase together reaching approximately 1.87 weight percent, against less than 0.05 weight percent gypsum in the laboratory samples, along with distinct cracking in the interfacial transition zone between aggregates and paste. The mortar bars, by contrast, showed little to no expansion cracking, suggesting that their measured expansion stemmed primarily from the formation of secondary iron oxides and hydroxides rather than from ettringite-driven paste expansion. Thermomagnetic analysis indicated that roughly two-thirds of the pyrrhotite in all samples was the hexagonal NC-type, a variant previous work suggests is more prone to oxidation than the monoclinic 4C form.

The study also carries a practical warning for laboratories running accelerated tests. Sulfur measurements by high-temperature combustion showed that the aggregate batches used in the Norwegian pilot testing contained only 0.44 to 0.51 weight percent sulfur after cement paste removal, considerably lower than the 0.85 to 1.08 weight percent reported for the reference Maskimo aggregate, which plausibly explains why the Norwegian mortar bars expanded less than Canadian counterparts tested under the same conditions. The authors recommend routinely determining total sulfur content in every test batch before accelerated mortar bar testing to ensure results are comparable and not misinterpreted. They further note that pre-oxidation of sulfide minerals, whether in the geological deposit, the quarry stockpile, or after concrete production, could accelerate sulfate attack, making the degree of pre-oxidation an important quality assurance consideration.

Beyond its immediate findings, the work demonstrates that automated mineralogy can serve as a diagnostic tool for assessing aggregate reactivity and predicting ISA-related deterioration in pyrrhotite-bearing concrete. By correlating a continuous transition from incipient surface alteration to complete pyrrhotite replacement with distinct secondary mineral assemblages, magnetite and goethite in fresh rock, ferrihydrite, gypsum, ettringite, and the FeOSCa phase in mortar and concrete, the researchers have established a mineralogical fingerprint of internal sulfate attack at each stage. Although the study drew on a single, well-characterized aggregate source, the authors expect the observed oxidation progression to be relevant to other ISA cases, noting that reactive aggregates from regions such as Connecticut exhibit comparable mineralogical characteristics. For the homeowners of Trois-Rivières and engineers elsewhere facing similar mysteries of crumbling concrete, the invisible enemy has finally been caught in the act.

Subject of Research: Pyrrhotite oxidation and internal sulfate attack in concrete aggregates and cement paste

Article Title: Pyrrhotite oxidation in aggregate and cement paste of accelerated laboratory and concrete field samples

Article References: Oberhardt, N. Å., Danner, T., Lode, S., Lindgård, J., Aasly, K., & De Weerdt, K. (2026). Pyrrhotite oxidation in aggregate and cement paste of accelerated laboratory and concrete field samples. Case Studies in Construction Materials, 25, Article e06544. https://doi.org/10.1016/j.cscm.2026.e06544

Image Credits: AI Generated

DOI: 10.1016/j.cscm.2026.e06544

Keywords: pyrrhotite, internal sulfate attack, concrete deterioration, automated mineralogy, cement paste, Trois-Rivières, Maskimo quarry, thermomagnetic analysis, iron sulfide oxidation, ettringite, ferrihydrite, accelerated mortar bar test