The investigation utilized uniaxial step-wise compression creep tests on specimens subjected to temperatures ranging from 20 °C to 800 °C. The primary focus was to compare and analyze the effects of aggregate types on total strain, load-induced thermal strain (LITS), creep deformation, and rate patterns. This experimental design allowed for a direct comparison of how normal-weight and lightweight aggregates respond to thermal exposure and subsequent mechanical loading.

Below 200 °C, the LWAC exhibited a higher load-induced thermal strain than the NWAC. Specifically, the LITS of the lightweight concrete was approximately 83.3% higher than that of the normal-weight concrete. The study attributes this difference to the lower modulus of the lightweight aggregate, which results in a different initial strain response compared to the denser normal-weight material.

A significant divergence in behavior was observed above 400 °C. At this temperature threshold, the NWAC experienced a rapid increase in creep strain. The researchers identified the thermal expansion mismatch between the aggregate and the cement paste at the interface transition zone (ITZ) as the driving factor for this accelerated deformation. This mismatch creates internal stresses that alter the creep rate of the normal-weight concrete.

In contrast, the LWAC containing fly ash ceramsite aggregate displayed a different response mechanism. The porous structure of the fly ash ceramic aggregate was found to play a “thermal mitigating” effect. This porosity effectively inhibited the propagation of microcracks within the material. As a result, the LWAC was able to maintain a longer steady-state creep period compared to the NWAC, indicating a more stable deformation profile under sustained load after high-temperature exposure.

Scanning electron microscopy (SEM) analysis provided microstructural evidence supporting these mechanical observations. At high temperatures, the ITZ of the NWAC developed tensile cracks measuring several micrometers in width. These cracks are consistent with the rapid increase in creep strain observed in the normal-weight specimens. Conversely, the LWAC specimens maintained a relatively intact structure at the interface, with fewer signs of the microcracking seen in the NWAC.

The study also evaluated the applicability of the Burgers model to the creep data of the lightweight concrete. The fitting degree of the Burgers model for the LWAC creep data was high, with an R-squared value of 0.99. This high correlation verifies the model’s applicability in describing the deceleration and constant-speed creep behavior of LWAC after high-temperature treatment. The successful fit provides a mathematical framework for characterizing the specific creep phases of this material.

The findings reveal the deformation mechanism of fly ash ceramic aggregate in high-temperature environments. The research provides a theoretical basis for the design and post-disaster assessment of the fire resistance of LWAC structures. By understanding how the porous aggregate mitigates thermal stresses and inhibits crack propagation, the study offers insights into the long-term deformation performance of lightweight concrete after exposure to extreme temperatures.

Subject of Research: Creep behavior of lightweight concrete containing fly ash ceramsite aggregate after high-temperature exposure

Article Title: Creep behavior of lightweight concrete containing fly ash ceramsite aggregate after exposure to high temperatures

Article References: Xiao, D., Chen, S., Wen, L., Dou, Y., Liu, H., & Zhao, J. (2026). Creep behavior of lightweight concrete containing fly ash ceramsite aggregate after exposure to high temperatures. Scientific Reports. https://doi.org/10.1038/s41598-026-77256-4

Image Credits: AI Generated

DOI: 10.1038/s41598-026-77256-4

Keywords: lightweight concrete, fly ash ceramsite, creep behavior, high temperature, structural engineering, load-induced thermal strain, interface transition zone, Burgers model, Creep, behavior, lightweight, concrete

Both reciprocal effects were smaller than the corresponding estimated expanded engineering uncertainties. The experiment used reference masses, rigid targets and an instrumented frame; it included no human participants and did not assess stature, clinical utility or legal-metrology conformity. The custom wheeled platform integrated four nominal 40-kg load-cell elements, a load-cell balancing and amplification stage, an ESP32 controller, an optical distance sensor, a USB hub, a tablet interface and a pass-through power bank.

The mass measurand was the total applied working-reference load transferred through the seated frame. The distance measurand was the separation between the sensor reference plane and a rigid target along the optical axis, used here as a bench surrogate for a possible knee-height sensing path rather than as a direct measurement of human stature. Because the mass and distance subsystems shared the frame, sensing arm, wiring environment, embedded controller and power architecture, the study tested loading and geometry effects in both directions rather than treating the sensors as isolated components.

Reference masses of 0, 20, 40, 60, 80 and 100 kg were applied, while rigid-target separations ranged from 300 to 600 mm. Across the 15 nonzero observations from three ascending linearity runs, the mass pathway had a mean signed error of 0.065 kg, a mean absolute error of 0.071 kg and a root-mean-square error of 0.081 kg. The maximum absolute error was 0.150 kg.

Linear regression returned an intercept of 0.010 kg, a slope of 1.00092 and an R-squared of 0.999998, indicating close agreement with the platform’s in-house working references over 20 to 100 kg. However, the archive did not preserve the reference instrument’s identity, external verification certificate or the derivation of its internal allowance, so these results characterize agreement with the working-reference system rather than traceable calibration accuracy. Path-dependent and geometry-dependent mass effects were larger than the pathway’s average bias.

Repeatability standard deviations were 0.032, 0.057 and 0.080 kg at reference loads of 20, 60 and 100 kg, respectively. The maximum paired hysteresis reached 0.240 kg during complete loading and unloading sequences, while the largest center-referenced shift under eccentric loading was 0.144 kg. A centered 60-kg load produced a maximum change of 0.190 kg during a 30-minute creep session.

The initial-zero indication remained small, reaching 0.010 kg. These observations show why a near-unity regression line alone was insufficient to characterize the system: loading history, load position and time under load introduced measurable effects. They are also consistent with the broader force-transducer literature cited by the study, in which creep, loading mode, calibration conditions and temperature can affect sensor performance.

Across 21 observations from three runs at seven target positions, the mean signed error and mean absolute error were both 0.952 mm, the root-mean-square error was 1.039 mm and the maximum absolute error was 1.8 mm. The fitted slope was 1.00119, the intercept was 0.417 mm and the R-squared was 0.999984. Repeatability standard deviations at target separations of 350, 450 and 550 mm were 0.408, 0.471 and 0.578 mm.

By comparison, changing the nominal angle setting produced a maximum contrast of 2.5 mm, mean readings for rigid matte white, gray and black targets differed by as much as 2.2 mm, and the recorded 30-minute temporal trend amounted to 1.3 mm. Those optical contrasts require careful interpretation. Pivot geometry and the complete test order were not retained, so the nominal angle result could not be isolated as a pure angular-sensor coefficient.

Material reflectance was not independently quantified, and the finish experiment was not fully documented as randomized or counterbalanced. The 2.2-mm difference therefore applies only to the three tested finishes and may include time or order effects; it is not a calibrated response across colors or surface properties. Ambient temperature, relative humidity, illuminance, supply voltage, warm-up duration, sampling frequency and filter settings were also not logged contemporaneously.

The 1.3-mm temporal change is consequently best described as system-plus-environment drift under the recorded bench conditions, not as drift attributable solely to the optical sensor. For the mass-to-distance experiment, the target was held at a 450-mm reference separation while the six mass levels were applied sequentially. Five readings at each level were averaged because they represented repeated observations of one mechanical state, avoiding the treatment of consecutive sensor readings as independent experiments.

The resulting run-level regression slopes were 0.0097, 0.0103 and 0.0094 mm per kilogram, giving a mean of 0.0098 mm per kilogram and a between-run standard deviation of 0.0005 mm per kilogram. Within-run R-squared values of 0.9995 to 0.9996 described the fit to six ordered design points rather than external reproducibility. The largest projected change over the full 0-to-100-kg range was 1.026 mm.

Adding a quadratic mass term did not materially improve the six-point fits. In the opposite direction, the sensing arm was placed in marked Home, Mid and Full states while a centered 60-kg load remained applied. Ten readings per state were averaged within each of three runs, and the primary result was the Full-minus-Home difference.

Those differences were 0.105, 0.105 and 0.120 kg, producing a mean of 0.110 kg, a sample standard deviation of 0.0087 kg and an observed range of 0.105 to 0.120 kg. The state analysis was restricted to three independent experimental blocks, and the archive contained no numerical arm angle or tip displacement. The result consequently applies to the physical configurations tested and cannot be converted into a universal geometric sensitivity coefficient.

A continuous arm-position metric would allow such a coefficient to be evaluated in a subsequent, independently replicated campaign with fully recorded mechanical conditions. The researchers supplemented the bench observations with a GUM-informed engineering uncertainty analysis. Type A repeatability terms came from repeated measurements, resolution terms used rectangular rounding models, and bounded zero, hysteresis, eccentricity, alignment, target-finish, creep and drift terms were represented by rectangular distributions.

The calculation used a root-sum-of-squares approximation with zero covariance. Under those assumptions, the combined standard uncertainty was 0.177 kg for the mass pathway and 1.269 mm for the distance pathway. These corresponded to expanded engineering values of 0.354 kg and 2.538 mm at a coverage factor of two.

The figures are illustrative engineering estimates, not certificate-derived uncertainties or exact 95-percent coverage intervals, because the internal reference allowances could not be reconstructed and covariance information was unavailable. Both observed reciprocal effects fell below their pathway’s expanded value, limiting claims about their significance for an individual reading. Taken together, the bench campaign supports continued engineering development of the integrated platform while defining what remains unresolved.

The mass system was close to the working references over its tested range, and the optical system tracked rigid targets with a maximum absolute linearity error of 1.8 mm, yet hysteresis, eccentric loading, creep, angle setting, target finish and temporal drift were all measurable. The shared frame also produced reciprocal trends relevant to future versions in which mounting stiffness, sensor alignment, zeroing or filtering might change. Exact component replication remains limited because load-cell and amplifier models, gain, sampling rate, firmware build, filter window, warm-up rules, optical field of view and display rounding were absent from the retained archive.

A confirmatory campaign would require traceable references, complete hardware and software documentation, logged environmental and power conditions, randomized or counterbalanced optical tests, more independently re-established runs and separate validation before any inference involving human anthropometry.

Subject of Research: Bench characterization and reciprocal cross-sensitivity of mass and optical-distance sensing in a shared seated measurement frame

Article Title: Engineering bench characterization and reciprocal cross-sensitivity of a sensor-integrated seated measurement platform

Article References: Widiyanto, W. W., Kuntoadi, G. B., & Ismawatie, E. (2026). Engineering bench characterization and reciprocal cross-sensitivity of a sensor-integrated seated measurement platform. Results in Engineering, 32, Article 113615. https://doi.org/10.1016/j.rineng.2026.113615

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113615

Keywords: sensor cross-sensitivity, seated measurement platform, load-cell testing, optical distance sensing, bench validation, measurement uncertainty, sensor integration, creep, hysteresis, Engineering, bench, characterization

The research team, led by Jie Cao, Jiwen Zhang and Ning Han of the China Jikan Research Institute of Engineering Investigation and Design, together with Qingyi Mu of Chang’an University, tackled a stubborn experimental gap. Laboratory element tests such as one-dimensional compression and triaxial tests can isolate the influence of moisture and compaction, but they cannot capture how an entire foundation evolves as a body. Ordinary 1-g physical models fail to reproduce the enormous self-weight stresses of a real fill, and field monitoring campaigns, while valuable, run for years, demand heavy resources, and often yield discontinuous records. Satellite techniques such as InSAR and UAV surveys mostly catch only post-construction surface movement. Centrifuge modeling solves the stress problem elegantly: by spinning a small model at up to 200 times Earth’s gravity, the machine makes a scaled-down fill feel its own full prototype-scale weight.

The experiments were carried out on the CKY-200 geotechnical centrifuge at the Yangtze River Scientific Research Institute, a machine with a nominal radius of 4 meters and a maximum acceleration of 200 g. The team used remolded Q2 loess sampled directly from an excavation face in Yan’an New District, the same kind of material actually being dumped into the gullies. The soil, with a specific gravity of 2.71, a maximum dry density of 1.72 grams per cubic centimeter and an optimum moisture content of 16.1 percent, was compacted layer by layer into a custom partitioned container, with each lift scarified to prevent weak interlayer planes. Laser displacement sensors tracked surface settlement continuously, miniature earth pressure cells recorded stresses inside the fill, and eleven layers of settlement markers embedded in one model allowed deformation to be mapped through depth against a glass sidewall.

Three test series isolated the key variables one at a time: moisture content at a fixed compaction degree of 0.80, compaction degree at optimum moisture, and fill height at 15, 25 and 35 meters in prototype scale. Loading mimicked real construction in cycles, with stepped 10-g increments representing filling periods and constant-load holds representing pauses between lifts. After construction, the centrifuge kept spinning to simulate the post-construction stage, and the models were then subjected to simulated rainfall and, in one case, full immersion to probe the notorious water sensitivity of loess. Data were recorded once per second throughout, giving an unusually complete settlement time series for a system of this kind.

The headline finding is stark: under these laterally confined conditions, 85 to 95 percent of total settlement is generated during construction itself, driven by consolidation of the unsaturated loess, while creep accounts for only the remaining 5 to 15 percent. That proportion of consolidation-driven settlement grows with wetter soil, poorer compaction and taller fills. Total settlement, unsurprisingly, increases with higher moisture content, lower compaction degree and greater fill height. But the study’s real contribution lies in the shape of the settlement curves. During early construction cycles, the curves undulate in a wave-like pattern of alternating fast and slow settlement. As the fill densifies, they transform into a step-like pattern: long, nearly stagnant platforms interrupted by abrupt settlement jumps, the signature of intermittent yielding events in which stored energy is suddenly released as the soil skeleton collapses into a more stable arrangement.

This step-like behavior persisted into the post-construction stage, and its character depended on the soil’s initial stiffness and thickness. Stiffer or thicker fills yielded in a high-frequency, low-amplitude mode, releasing energy in many small, frequent events and deforming uniformly. Softer or thinner fills showed long stable plateaus punctuated by concentrated, sudden settlement after prolonged quiet periods. The researchers also found pronounced hysteresis: settlement lagged behind loading because each construction cycle was too short for the soil to fully stabilize, meaning an apparently quiescent fill may simply be accumulating strain for its next yielding episode. For engineers, this is a warning that a surface that looks stable is not necessarily at equilibrium, and that premature loading or poor drainage during that apparent calm can trigger sudden differential settlement.

The internal anatomy of the settlement proved equally revealing. Deeper layers settled far more than shallow ones, with deep-layer settlement reaching roughly 80 to 95 percent of the shallow-layer values in relative terms, yet the settlement increment did not grow monotonically with depth. Instead it fluctuated, especially near the surface and the container sides. Horizontally, deep layers settled more in the middle than at the edges, while shallow layers did the opposite. The team interprets this differential movement as a spontaneous form of soil arching: faster-settling zones in the deep fill act like a trapdoor, deflecting overburden stress laterally toward stiffer supports and creating unloading zones above. Earth pressure measurements supported this picture, with deep pressures eventually exceeding theoretical overburden values while shallow pressures stayed below, and the earth pressure coefficient rising with depth as the fill densified.

Water emerged as the enduring villain. Rainfall infiltration caused settlement to increase sharply, with the wetting-induced increment largest where prior settlement had already loosened the soil structure, and rebound after unloading was correspondingly small because plastic deformation had exhausted the soil’s elastic recovery. Even more striking, soils with different initial compaction degrees converged to similar total post-construction settlement, yet the loosely compacted ones remained far more sensitive to rainfall. In other words, equal settlement does not mean equal risk: a poorly compacted fill that has finished settling can still lurch dramatically when wetted. After full immersion, the soil’s resistance to deformation weakened further, settlement became governed almost entirely by load increments, and even after construction, post-construction and rainfall phases, the fill retained measurable remaining settlement potential.

The authors frame the mechanism through the staged consolidation of unsaturated soil. Upon loading, the soil first deforms elastically through grain contacts and bound water films. Then pore air, connected to the atmosphere, is rapidly expelled or dissolved, driving the largest compression. Next, as saturation rises, pore water pressure dissipates while water attacks the soil skeleton itself, dissolving intergranular cements, thinning the contractile skin that generates matric suction, softening clay particles and lowering the yield stress, so the fill keeps compressing even under constant load. Finally, under sustained stress, grains creep and rearrange slowly toward an optimal packing. Compaction variability across the fill makes this process spatially uneven, producing the alternating concave and convex zones that drive the observed wave-like and step-like surface behavior.

The practical implications are direct. Because settlement lags loading, construction rates on loess high fills should be strictly controlled with deliberate intermittent pauses, so that sudden yielding happens during scheduled breaks rather than beneath finished works. And because insufficient compaction can spontaneously generate soil arching that redistributes stresses even without any buried structures, designers of these laterally confined fills need to account for stress fields that depart substantially from simple overburden calculations. The team cautions that their rigid-walled models approximate one-dimensional compression in narrow gully terrain and that time scales were deliberately not converted to prototype values, so the results illuminate mechanisms and relative influences rather than delivering direct design numbers. Still, for projects with no engineering precedent, knowing that the danger is concentrated in construction-stage consolidation, that the soil yields in stealthy steps, and that water remains a long-term threat is exactly the mechanistic foundation the next generation of mountain-moving cities will be built on.

Subject of Research: One-dimensional settlement behavior of compacted loess high-fill foundations studied with geotechnical centrifuge model tests

Article Title: One-dimensional settlement characteristics of loess high-fill foundations: insights from centrifuge model tests

Article References: Cao, J., Zhang, J., Han, N., Dong, X., Chen, X., Duan, J., Zhou, Y., Zhou, L., Liu, Z., & Mu, Q. (2026). One-dimensional settlement characteristics of loess high-fill foundations: insights from centrifuge model tests. Environmental Earth Sciences, 85(16), Article 421. https://doi.org/10.1007/s12665-026-13167-x

Image Credits: AI Generated

DOI: 10.1007/s12665-026-13167-x

Keywords: loess, high-fill foundation, centrifuge model test, settlement, consolidation, compaction degree, moisture content, soil arching, Loess Plateau, geotechnical engineering, creep, wetting collapse

The root of the decades-long scatter lies in an experimental trap that is easy to describe but fiendishly hard to escape. To measure how creep rate changes with temperature, researchers need to compare deformation at different temperatures. The traditional approach was to prepare several snow samples that were supposedly identical and test each one at a different temperature. The problem is that no two snow samples are ever truly identical. Snow is a two-phase material of sintered ice crystals and air, and its densification depends intimately on the spatial arrangement of the ice skeleton. Recent work by the same group showed that even a 5 percent difference in density between two samples, which is common in practice, can shift the strain rate by about 50 percent, equivalent to a temperature difference of roughly 4 degrees. Add to that the temperature fluctuations of typical cold rooms, which reach plus or minus 0.6 degrees Celsius due to defrost cycles and heating from X-ray sources, and the uncertainty swamps the signal being measured.

Védrine’s team solved the problem with an elegant inversion of the protocol: instead of comparing different samples at different temperatures, they tested the same sample at a sequence of temperatures. The centerpiece is a newly developed thermo-mechanical cell, a slim aluminium cylinder 20 millimeters in diameter with walls just half a millimeter thick so that X-rays can pass through. Two Peltier thermoelectric modules at the top and bottom of the cell control the sample temperature independently of the surrounding cold room, keeping temperature differences within the cell on the order of a thousandth of a kelvin, with absolute accuracy better than half a degree. A hollow piston applies a controlled load, a miniature load cell beneath the sample verifies that the force stays constant and that the piston moves without friction, and a laser sensor tracks the vertical displacement of the sample with a resolution of one micrometer. The entire assembly rotates inside an X-ray tomograph, which images the evolving microstructure every eight hours.

The experimental protocol was deliberately gentle. Natural snow collected in the field was stored for months at minus 20 degrees Celsius, where it matured into decomposing and fragmented precipitation particles, then sieved and sintered for at least two weeks. Five samples with initial densities between 278 and 320 kilograms per cubic meter were loaded with a stress of 1.25 kilopascals, comparable to the overburden of a 45-centimeter layer of snow of that density. Each experiment lasted five days and comprised five one-day temperature steps from minus 6 to minus 18 degrees Celsius in 3-degree increments. Because the applied stress was low, the samples deformed only slightly, with a mean final vertical strain of just 4 percent, keeping microstructural evolution to a minimum. Tomography confirmed that density changed by less than 4 percent and specific surface area by less than 7 percent over the whole test, meaning the ice framework remained essentially unchanged while the temperature varied.

One subtlety remained: even a small amount of densification causes geometric hardening, an increase in the reference stress of the ice matrix that slows creep even at constant temperature. To disentangle this effect from genuine temperature sensitivity, the team used a state-of-the-art snow viscoplastic model that describes strain rate as a function of stress, solid fraction, and a temperature-dependent prefactor. By dividing the measured strain rate by the mechanical component, they isolated the pure temperature response. The approach proved robust: neglecting geometric hardening entirely would have produced an activation energy about 10 percent too high, but within the physically meaningful range of model parameters the estimated temperature sensitivity varied by only 3 percent. Digital volume correlation applied to the tomographic images independently confirmed that the laser-based displacement measurements faithfully captured the bulk deformation of the samples.

The results are striking. When the corrected strain-rate prefactor is plotted against inverse temperature on a logarithmic scale, the data from all five samples collapse onto a single master curve, revealing a clear Arrhenius-type relationship, but with a kink. A single Arrhenius fit gives an apparent activation energy of 105 plus or minus 6 kilojoules per mole, yet it fails to capture a change in slope around minus 12 degrees. A two-regime fit is statistically preferred according to the Akaike Information Criterion and yields an activation energy of 126 plus or minus 6 kilojoules per mole above a transition temperature of minus 13 plus or minus 1 degree Celsius, and 51 plus or minus 18 kilojoules per mole below it. In other words, snow creeps far more readily per degree of warming near the melting point than in colder conditions, and the transition between the two behaviors is sharp.

Remarkably, this two-regime pattern mirrors what has long been known for polycrystalline ice. Classic creep studies of glacier ice dating back to Glen’s work in the 1950s reported activation energies of roughly 60 kilojoules per mole below minus 10 degrees and around 130 kilojoules per mole above, values that bracket the new snow measurements almost perfectly. This is intriguing because earlier research by Védrine and colleagues showed that ice inside snow does not deform like dense polycrystalline ice, since the pore network must accommodate the deformation and creep is dominated by basal glide, the sliding of crystal planes along their base. The authors caution that the apparent similarity may be partly deceptive. When they allowed the stress exponent to vary with temperature, following single-crystal measurements of basal slip reported by Jones and Brunet in 1978, the entire dataset could instead be described by a single effective activation energy of 78 plus or minus 6 kilojoules per mole, consistent with basal-slip creep. Part of the apparent bilinear behavior may therefore reflect a temperature-dependent stress exponent rather than a true change in the underlying deformation mechanism.

Whatever the microphysical explanation, the practical implications for snowpack modeling are immediate. The team compared their new parametrization with the settlement laws used in two widely deployed detailed snowpack models, Crocus and Snowpack. The Brun et al. 1992 formulation implemented in Crocus corresponds to an activation energy of about 62 kilojoules per mole, close to the new low-temperature value but far below the high-temperature one, and it exhibits the weakest temperature sensitivity of all. The Kojima 1975 law for fresh snow in Snowpack behaves similarly at low densities. Only the new parametrization captures two distinct temperature regimes. In simulations of a one-meter isothermal snowpack with an initial density of 100 kilograms per cubic meter, the differences matter: after seven days, a snowpack held at minus 20 degrees Celsius is predicted to be 20 to 34 percent taller than one at minus 6 degrees, depending on which law is used, and the new measurements generally imply faster compaction at low densities than either existing formulation predicts.

Accurate snow compaction is not a niche concern. Settlement rates shape snow stratigraphy, which in turn governs avalanche forecasting, estimates of snow water equivalent for hydrology and reservoir management, and even the interpretation of polar ice cores, where firn densification laws link trapped air to paleoclimate records. The study’s authors note that their experiments did not extend above minus 6 degrees Celsius, leaving open whether the creep rate accelerates nonlinearly as the melting point approaches, as often reported for polycrystalline ice, or whether a basal-slip-dominated material like snow behaves differently. They also emphasize that dedicated experiments are still needed to constrain how the stress exponent of basal slip varies with temperature. But the methodological lesson is already secure: by holding the microstructure fixed and letting only the temperature change, the Grenoble team has turned a seventy-year-old scatter plot into a coherent physical law, and given snowpack models their most rigorous temperature calibration to date.

Subject of Research: Temperature dependence of snow viscoplastic creep measured through controlled in-tomograph experiments

Article Title: Temperature sensitivity of snow viscoplasticity: Evidence from controlled creep experiments

Article References: Védrine, L., Brun, M., Bozon, M., Laurent, B., & Hagenmuller, P. (2026). Temperature sensitivity of snow viscoplasticity: Evidence from controlled creep experiments. The Cryosphere, 20(10), 5813-5828. https://doi.org/10.5194/tc-20-5813-2026

Image Credits: AI Generated

DOI: 10.5194/tc-20-5813-2026

Keywords: snow, viscoplasticity, creep, activation energy, Arrhenius law, snowpack modeling, X-ray tomography, ice mechanics, avalanche forecasting, The Cryosphere, Peltier temperature control, firn densification

The alloy in question, Mg-8wt.%Li, sits in a sweet spot of composition. Above roughly 8 percent lithium by weight, the material becomes dual-phase, containing both the hexagonal close-packed alpha phase typical of magnesium and the body-centered cubic beta phase that lithium stabilizes. That b.c.c. beta phase is the secret ingredient: it deforms far more readily than the hexagonal phase, endowing the alloy with ductility that no other magnesium alloy system can match. The Mg–Li family holds the record for elongation among all magnesium alloys, although its absolute strength remains modest, which is precisely why understanding and controlling its deformation mechanisms matters so much for lightweight structural applications.

To push the alloy into its most deformable state, the team turned to high-pressure torsion, a severe plastic deformation technique in which a disk of material is squeezed under a nominal pressure of 4 gigapascals while being twisted through five full turns. The process shears the microstructure into an ultrafine grain structure. Electron microscopy and transmission Kikuchi diffraction revealed that the extruded material, which started with grains of about 2.2 micrometers, was refined to an average spatial grain size of roughly 0.4 micrometers, with individual subgrains resolved in the range of 0.15 to 0.40 micrometers. The two phases became elongated and aligned with the torsion direction, a signature of the intense shear the disk experienced.

Miniature tensile specimens, cut from the disk mid-radius by electric discharge machining with a gauge cross-section of just 0.8 by 0.8 millimeters, were then pulled at room temperature across strain rates spanning from 10^-5 to 10^-3 per second. The results were striking. The peak stress varied dramatically with strain rate, a hallmark of high strain rate sensitivity, and the alloy exhibited large elongations at the slowest rates of 10^-4 and 10^-5 per second. This unusual increase in ductility at low strain rates, previously reported for similar alloys, is exactly the behavior that makes dual-phase Mg–Li alloys candidates for room-temperature superplasticity, a property normally reserved for ceramics and exotic superalloys at soaring temperatures.

Creep tests added another layer of intrigue. Under constant loads between 10 and 100 megapascals at room temperature, and at 373 kelvin under 10 and 25 megapascals, the creep rates steadily decreased as strain accumulated. Such decelerating creep is normally seen in coarse-grained metals at high temperature, where dislocation substructures build up during the initial stage of deformation, but that explanation should not apply to an ultrafine-grained material whose grains are already smaller than any stable subgrain size. The authors point to a recent study on ultrafine-grained aluminum that attributed similar behavior to strengthening by dynamic recovery, and they adopted the strain rate at 2 percent strain as representative of the as-processed state for further analysis.

The most consequential number to emerge from the creep experiments was the activation energy. Using the temperature dependence of the creep rate between room temperature and 373 kelvin, the team estimated activation energies of 64 kilojoules per mole at 25 megapascals and 77 kilojoules per mole at 10 megapascals. Both values fall well below the 135 kilojoules per mole for self-diffusion and the 92 kilojoules per mole for grain boundary diffusion in pure magnesium. The authors attribute this reduction to the large fraction of the b.c.c. beta phase, which has a lower melting temperature and correspondingly lower diffusion barriers. Reported grain boundary diffusion energies for related alloys, 65 kilojoules per mole for Mg-9%Li and 61 kilojoules per mole for Mg-8%Li-1%Zn, bracket the measurements, and the team settled on 75 kilojoules per mole for their modeling. The low activation energy is a clear signal that thermally activated mechanisms are at work even at room temperature.

To interpret these observations, the researchers deployed a two-pronged computational strategy. First, they applied a recently updated model for grain boundary sliding, the mechanism in which grains slide past one another along their boundaries, which is the canonical explanation for superplastic flow. The model treats the total flow stress as the sum of a threshold stress, capturing other strengthening contributions such as solid solutions, and a grain boundary sliding contribution that depends on shear modulus, temperature, grain size, Burgers vector and the grain boundary diffusion coefficient. A rectified linear function is used as a mathematical constraint to prevent the model from predicting negative stresses when thermal activation is strong. Second, they trained machine learning algorithms on 67 carefully curated data points from the literature, restricted to dual-phase Mg–Li alloys containing no alloying elements other than lithium, with lithium contents between 8 and 12 weight percent, to keep composition effects from contaminating the analysis.

The machine learning lineup included two black-box models, support vector regression and random forest, and two white-box symbolic regression models, one unconstrained and one constrained by dimensional consistency. The black-box models, evaluated with Shapley Additive Explanations, agreed that temperature is the dominant control on flow stress, that strain rate plays an intermediate role, and that grain size matters only marginally. On the training data the algorithms performed well, with mean absolute percentage errors ranging from 11.5 percent for the random forest to 35.3 percent for support vector regression. But when the models were asked to predict the new experimental data generated in this study, the black-box approaches collapsed, producing errors exceeding 250 percent, while the grain boundary sliding model and the dimension-constrained symbolic regression held their ground at roughly 35 percent error.

The deeper story emerged when the team plotted predicted flow stress against grain size, strain rate and temperature and compared the trends with experiment. The experiments revealed a fascinating crossover: for grains larger than about half a micrometer, refining the grains strengthens the alloy in classic Hall–Petch fashion, but below that threshold, further refinement actually softens it, the inverse Hall–Petch regime. Only the grain boundary sliding model captured this transition cleanly. The black-box models failed to reproduce the steep strain rate dependence of the flow stress, with support vector regression even predicting a physically implausible negative strain rate sensitivity at high rates. The dimension-constrained symbolic regression, guided by physically meaningful features such as the Zener–Holomon parameter and dimensionless groups derived from the grain boundary sliding equations, performed far better than its unconstrained counterpart, demonstrating that embedding physical knowledge into machine learning pays real dividends.

The take-home message resonates well beyond magnesium. Machine learning models are only as trustworthy as the data behind them, and with a sparse training set of 67 points, the black-box algorithms interpolated plausibly but extrapolated badly, especially outside the temperature and strain rate ranges they had seen. The century-old physics-based model, by contrast, reproduced every experimental trend once its activation energy was adapted to the dual-phase alloy. For engineers dreaming of lightweight magnesium–lithium components that can be superplastically formed at room temperature, the study offers both a validated design tool and a cautionary tale: in materials science, the white box and the physics still matter, and the smartest models are those that know the laws they are trying to learn.

Subject of Research: Deformation mechanisms and grain boundary sliding in ultrafine-grained dual-phase Mg–Li alloys processed by high-pressure torsion

Article Title: Analysis of deformation mechanism of dual-phase Mg–Li alloy, processed by high-pressure torsion, using experiments and machine learning

Article References: Carvalho, A. P., Edalati, K., & Figueiredo, R. B. (2026). Analysis of deformation mechanism of dual-phase Mg–Li alloy, processed by high-pressure torsion, using experiments and machine learning. Journal of Materials Science. https://doi.org/10.1007/s10853-026-13355-x

Image Credits: AI Generated

DOI: 10.1007/s10853-026-13355-x

Keywords: magnesium-lithium alloy, high-pressure torsion, grain boundary sliding, severe plastic deformation, machine learning, superplasticity, creep, activation energy, Hall-Petch, symbolic regression, ultrafine grains, flow stress