Coal’s most stubborn footprint may not be the carbon in the atmosphere but the rubble on the ground. Coal gangue — the stony refuse discarded during mining and washing operations — now blankets China in stockpiles exceeding four billion tons, with another 150 to 200 million metric tons arriving every year. The heaps occupy farmland, leach pollutants into soil and groundwater, and in many regions smolder uncontrolled for decades. Now a team of civil engineers has charted a way to entomb that liability inside the world’s most-used manufactured material. In a study published in Case Studies in Construction Materials, researchers Zhang Tong, Wang Yujie, Gao Shan and Zhu Qingru present the most complete picture yet of how steel reinforcing bars grip concrete in which natural sand has been replaced by crushed coal gangue — the decisive physics that determines whether structures built from waste-derived material stand safely for a century or quietly come apart at the interface.

The timing is no accident. China pours roughly 1.4 billion cubic meters of concrete every year, and after cement the ingredient it devours fastest is natural sand. Decades of dredging rivers and stripping quarries for it have inflicted severe ecological degradation, pushing the industry toward manufactured alternatives. Coal gangue turns out to be chemically and physically suited for the role: dominated by silicon dioxide and aluminum oxide, its particles are rough, angular and somewhat reactive, allowing them to form high-quality interfaces with cement paste. Earlier studies established that concretes blended with gangue sand can achieve mechanical properties comparable to ordinary mixes, particularly at the medium-to-low strength grades common in construction. Yet compressive strength is only half the story. What actually holds a reinforced beam, column or bridge deck together is the bond between embedded steel and the surrounding concrete — and for gangue concrete that property had never been systematically measured, modeled or translated into design practice. Until now.

Steel and concrete form one of engineering’s great unlikely marriages, and their entire union rests on interfacial bond strength. Three mechanisms share the load: chemical adhesion between the steel and its skin of hydration products, friction at the contact surface, and above all the mechanical interlocking of the ribs rolled onto deformed reinforcing bars. In a standard pull-out test, the ultimate bond stress is the peak force divided by the embedded surface area — τ = F/(πd·la), where d is the bar diameter and la the anchorage length. Design codes such as the CEB-FIP Model Code 2010 distill decades of such experiments into bond-slip curves for engineers. But those curves were calibrated almost entirely on ordinary aggregates. Coal gangue sand is porous, absorbs up to 7.6 times more water than river sand, and its higher friability reshapes both the matrix’s pore structure and the thin interfacial transition zone hugging each bar — precisely the region where bond is born or destroyed. Applying the classical equations blind, the authors argue, was simply not defensible.

The team therefore began by taming the material itself. Mined gangue was crushed and sieved into a fine aggregate matching the gradation of natural sand, following the Chinese standard GB/T 14684–2022. Under 50- to 100-fold magnification the two sands tell visibly different stories: gangue particles are more angular, which aids mechanical keying, but carry micro-cracks from crushing, and their colors range from pale red to deep crimson — a fingerprint of spontaneous self-ignition during long years of open-air stockpiling. Laboratory characterization revealed water absorption of 11.4 percent against 1.5 percent for natural sand, a crushing index of 22.8 percent versus 15.2 percent, and porosity of 51.4 percent versus 39.0 percent. Chemically the sands are cousins, built mainly of silica and alumina, though gangue carries markedly less silicon dioxide and more aluminum oxide. To keep that thirst from sabotaging fresh concrete, the researchers pre-wetted the gangue sand for one hour, adding water equal to 60 percent of its absorption capacity — enough to stop the aggregate draining cement paste during hydration, without flooding the matrix with excess free water that would soften the interfacial transition zone.

The concretes were then mixed with ordinary Portland cement, granite coarse aggregate between 5 and 20 millimeters and a polycarboxylate superplasticizer, then cured for 28 days at 20 degrees Celsius and 95 percent relative humidity before mechanical testing. The gangue left a measurable mark on the matrix itself: cube compressive strength fell from 34.8 megapascals with zero substitution to 29.7 megapascals at full replacement in the C30 family, from 40.2 to 34.1 megapascals in the C40 family and from 49.5 to 44.1 megapascals in the C50 family, alongside reductions in elastic modulus and tensile strength. The steel side of the partnership consisted of HRB400 hot-rolled deformed bars in 14, 16 and 18 millimeter diameters, with yield strengths between 422 and 461 megapascals. Fifteen specimen groups — three identical 150-millimeter cubes each — then covered every combination of interest: three strength grades, five substitution ratios from 0 to 100 percent, three anchorage lengths of 3, 5 and 7 bar diameters, and the three bar sizes.

The ingenuity lay in the instrumentation. Each 470-millimeter reinforcement bar was sliced lengthwise, milled with two grooves and fitted with strain gauges at 20-millimeter intervals, so that as the bar was pulled the researchers could convert measured strains into local bond stresses along the anchorage through a finite-difference method. Plastic sleeves of calibrated length defined the embedded zone, the remaining annular gap being packed with absorbent cotton and sealed with hot-melt adhesive, while displacement transducers capable of resolving 0.001 millimeters tracked slip at both the free and loaded ends. Loading proceeded in a 60-ton universal testing machine under displacement control at 0.5 millimeters per minute, following the national pull-out testing standard GB/T 50152–2012. The result was not a single number per specimen but a moving picture: the migration of bond stress along the bar, the emergence of secondary peaks, and the way different mixes redistributed load as damage accumulated.

The headline result concerned substitution ratio, and it was sobering. As gangue content climbed, bond strength fell steadily: at 100 percent replacement, peak bond stress dropped 33.6 percent for C30 mixes, 30.2 percent for C40 and 26.2 percent for C50 relative to natural sand references. Initial stiffness declined in step, and in the C30 series the slip at peak strength grew by roughly one to 1.2 millimeters at 50 and 100 percent substitution. The weaker, more crushable aggregates and their degraded interfacial transition zones could not resist the hoop stresses generated by a slipping rib. Failure modes shifted in parallel. High-substitution cubes overwhelmingly failed by pull-out, the bar simply grinding its way through crushed concrete, whereas stronger, more brittle C40 and C50 specimens tended to split, sending radial cracks racing to the surface. Conversely, raising the concrete grade lifted bond strength — C40 gained 16.1 to 22.2 percent and C50 gained 23.4 to 36.3 percent over C30 at equal substitution — although the stronger mixes then surrendered their grip more abruptly once past the peak.

Geometry proved just as consequential as chemistry. Compared with bars embedded three diameters deep, bond strength fell 21.7 percent at five diameters and 27.5 percent at seven diameters, because bond stress concentrates near the loaded end and longer embedments create steep gradients in which the rear segment contributes little to capacity. Bar diameter worked against intuition: 16-millimeter bars lost 28.6 percent and 18-millimeter bars lost 38.8 percent of bond strength relative to 14-millimeter bars. Larger bars dilute the cover-to-diameter ratio that provides lateral restraint, while amplifying the circumferential tension in a material whose splitting resistance is already compromised by porous aggregate. The strain-gauge maps added a subtle fourth dimension: with more gangue in the mix, the bond stress peak migrated toward the free end of the bar as the weakened matrix lost transmission efficiency, and with larger bars a secondary stress peak emerged ever closer to the unloaded end.

Assembled together, the measurements reconstruct a slow-motion catastrophe at the microscopic scale. On first loading, chemical adhesion carries the interface and bond stress rises almost linearly with slip. As load grows, adhesion yields to mechanical interlocking, and conical micro-cracks fan out ahead of each rib. In gangue concrete, the concrete wedged in front of the ribs fractures earlier and fragments faster — extracted bars bear shallower, blurrier rib impressions than in natural sand concrete — until the crushed zone collapses into a sliding plane and the bar pulls free. From this narrative the authors distilled a four-phase bond-slip constitutive model, anchored at three characteristic points: an elastic limit set at roughly 80 percent of peak stress, the peak itself, and the residual floor. The model spans an initial elastic stage, an elastoplastic transition, post-peak softening and a steady residual phase, expressed as regression-fitted equations ready for structural analysis. Implemented in ABAQUS finite-element simulations of the pull-out tests, the computed response tracked the experiments closely enough for the team to launch an expanded parametric sweep of combinations never cast in a laboratory.

The final flourish is a predictive tool for the machine-learning era: a back-propagation neural network trained on 1,376 experimentally derived bond stress–slip data points, reported by the team to achieve highly accurate predictions across the entire parameter space. The significance stretches well beyond one laboratory. With four billion tons of gangue waiting and sand reserves shrinking, any serious pathway to structural concrete from mining waste demands trustworthy anchorage rules — and this study now supplies the data, the constitutive model and the algorithm needed to draft them. The trade-offs are real: bond strength falls by up to a third at full substitution, and the authors position gangue concrete for medium-to-low strength applications where its profile balances out. But the direction of travel is unmistakable. The waste heaps of the coal age are being re-engineered into the load-bearing skeleton of the built environment, one interlocked rib at a time — with every reinforcing bar gripped just tightly enough.