Multiscale damage evolution and chemo-mechanical constitutive modeling of acid-corroded sandstone Authors: Li Hao, Zeng Yifan, Wu Qiang, Cheng Yifan, Yang Donghui Published: 2026 ABSTRACT SummaryThis study investigated time-dependent chemo-mechanical deterioration of fine-grained sandstone exposed to dilute sulfuric acid representing the proton-sulfate component of acid mine drainage. Specimens were corroded for 0–20 days and examined by X-ray diffraction, scanning electron microscopy, endpoint solution chemistry, and triaxial compression at 20, 30, and 40 MPa. Results showed cumulative carbonate dissolution, cement degradation, microstructural deterioration, and mechanical weakening. After 20 days, Ca2+ and Mg2+ concentrations reached 156 and 25.3 mg/L, while peak strength decreased by 33.3%, 26.6%, and 29.9%, respectively. A nested Mohr-Coulomb comparison yielded F(4,5) = 1.284 and p = 0.388, supporting downward envelope translation but not statistically detectable slope variation. A time-dependent damage model reproduced pre-peak responses across all 15 conditions, with R2 = 0.767–0.961. The framework links mineral alteration and microstructural damage to mechanical degradation but requires further validation for multicomponent field acid mine drainage and long-term engineering service. Introduction High-sulfur coal is widely distributed in China (Figure 1). Oxidation of sulfur-rich components in coal seams releases sulfate (SO42−) and hydrogen ions (H+), which are the main drivers of acid mine drainage (AMD) formation.1 AMD is commonly characterized by a pH below 4.5 and high concentrations of sulfate and heavy-metal ions. Such corrosive water not only accelerates underground equipment corrosion and induces water-inrush hazards but also damages the structural integrity of coal-rock water-resisting strata through seepage, further aggravating pore expansion and mechanical degradation of the surrounding rock, ultimately threatening the long-term stability and ecological safety of deep underground engineering.2Figure 1Geographic distribution of high-sulfur coal resources in China Geographic distribution of high-sulfur coal resources in China After acidic fluids react chemically with rock mineral grains and cements, the original pore and fracture structures inside the rock mass can be substantially modified. This issue has been studied from multiple perspectives by researchers worldwide.3 Existing studies on fluid-rock interaction have confirmed that acidic aqueous solutions produce pronounced selective dissolution in sedimentary rocks containing carbonate components such as calcite and dolomite. This mineral dissolution at the mesoscale directly weakens the macroscopic load-bearing skeleton of the rock.4,5 Using nondestructive testing methods such as nuclear magnetic resonance and micro-computed tomography (CT), subsequent studies further found that acidic fluids preferentially dissolve soluble matrices and intergranular cementing materials, thereby promoting the initiation and propagation of subcritical cracks and ultimately causing obvious decreases in crack initiation stress and damage stress.6,7,8 Macroscopic mechanical tests have yielded consistent findings: as solution acidity increases, rock density, elastic modulus, compressive strength, and related parameters decrease to varying degrees.9,10 For example, Chen et al.11 used CT scanning to clarify the mesoscopic mechanism of sandstone damage evolution under chemical corrosion, and the long-term immersion tests conducted by Luo et al.12 showed that the uniaxial compressive strength of sandstone can decrease by approximately 43% in a strongly acidic environment with pH = 1.25. In summary, existing studies have clarified many effects of solution acidity, ionic composition, and immersion duration on rock mineralogy, pore structure, and mechanical properties, and several chemical-damage models have been proposed. Three issues nevertheless merit further study. First, fewer studies have followed damage accumulation through a controlled corrosion-duration series. Second, the cross-scale link among mineral dissolution, changes in particle contacts, characteristic stress thresholds, and macroscopic strength attenuation remains incompletely quantified. Third, whether acidic corrosion primarily translates the strength envelope or also changes confining-pressure sensitivity is difficult to determine without explicit nested statistical testing, wider pressure coverage, and replicate specimens. Recent reaction-chemo-mechanical studies suggest that mineral mass loss and microcrack propagation may interact, but these mechanisms and their transfer to sandstone under AMD-relevant acidic exposure require further verification. To address these research gaps, fine-grained sandstone was exposed to a fixed dilute sulfuric acid solution for 0–20 days and characterized by X-ray diffraction (XRD), SEM, endpoint ion analysis, and triaxial compression. The study retains three linked aims: to document the measured time-series mineralogical, microstructural, and mechanical changes; to compare time-specific, common-slope, and full interaction Mohr-Coulomb models using the 15 measured peak strengths; and to develop a time-dependent chemo-mechanical constitutive description with clearly stated calibration and extrapolation boundaries. Results Specimen consistency and experimental scope Pre-test screening data for the 15 triaxial specimens showed mean P-wave velocity, S-wave velocity, density, and Vp/Vs values of 4,895.02 ± 48.29 m/s, 2,747.21 ± 58.45 m/s, 2.658 ± 0.0237 g/cm3, and 1.783 ± 0.044, respectively. The corresponding coefficients of variation were 0.99%, 2.13%, 0.89%, and 2.45% (Table 1). These low descriptive dispersions support reasonable initial consistency of the selected specimens, but they do not replace within-condition replicates. Because one representative specimen was tested for each corrosion-duration-confining-pressure combination, the time-dependent differences described further are reported as observed trends rather than estimates of group means.Table 1Pre-test specimen-consistency statisticsStatisticP-wave velocity (m/s)S-wave velocity (m/s)Density (g/cm3)Vp/Vsn15151515Mean4,895.022,747.212.6581.783SD48.2958.450.02370.044CV (%)0.992.130.892.45Minimum4,814.532,657.832.621.72Maximum4,971.292,835.462.701.84 Table 1Pre-test specimen-consistency statisticsStatisticP-wave velocity (m/s)S-wave velocity (m/s)Density (g/cm3)Vp/Vsn15151515Mean4,895.022,747.212.6581.783SD48.2958.450.02370.044CV (%)0.992.130.892.45Minimum4,814.532,657.832.621.72Maximum4,971.292,835.462.701.84 Pre-test specimen-consistency statistics Evolution of mineral composition and ions in the immersion solution Main mineral reaction pathways The reactions described here are presented as possible stoichiometric pathways consistent with the identified minerals and the endpoint solution composition. Because solution chemistry was measured only initially and after 20 days, the data cannot resolve stage-specific reaction rates or establish a kinetic ranking among minerals. Possible ion-release pathways during the reaction of the constituent minerals with dilute sulfuric acid are summarized here. Carbonate dissolution is consistent with the large cumulative increases in Ca2+ and Mg2+, but the present endpoint measurements do not determine when or how rapidly each reaction occurred. The dissolution reaction of calcite (CaCO3) under dilute sulfuric acid corrosion is:CaCO3+H2SO4→Ca2++SO42−+CO2↑+H2O The dissolution reaction of dolomite [CaMg(CO3)2] under dilute sulfuric acid corrosion is:CaMg(CO3)2+2H2SO4→Ca2++Mg2++2SO42−+2CO2↑+2H2O Feldspar minerals (microcline) undergo slow hydrolysis under acidic conditions, releasing Na+ and K+ into pore water with groundwater seepage and generating secondary clay minerals such as kaolinite simultaneously.2KAlSi3O8+2H2SO4+9H2O→2K++Al2Si2O5(OH)4+2SO42−+4H4SiO4 Interlayer Na+ in montmorillonite is substituted by H+, raising the Na+ concentration in aqueous solution, while Al3+ and Mg2+ ions get leached out from the mineral octahedral sheets, and such cation exchange together with metal ion dissolution proceeds synchronously.Na0.3(Al,Mg)2Si4O10(OH)2+H2SO4→Na++Al3++Mg2++SO42−+silicagel Mineral contents and ion responses in the immersion solution As shown in Figure 2, the relative contents of dolomite and calcite decreased markedly, consistent with the ion-release mechanism controlled by preferential dissolution of carbonate minerals. Microcline and montmorillonite also showed decreasing trends, but the magnitudes were smaller. The change in microcline is consistent with progressive hydrolysis of silicate minerals, whereas the change in montmorillonite may be related to interlayer cation exchange and sheet leaching. Table 2 shows that after 20 days, the concentrations of Ca2+, Mg2+, and Na+ increased from 0.36, 0.038, and 0.45 mg/L to 156, 25.3, and 15.0 mg/L, respectively, providing independent evidence for carbonate dissolution and ion exchange in clay minerals. It should be noted that XRD phase proportions are semi-quantitative results; in this study, they are mainly used to identify relative change trends and are not interpreted as strict mass-conservation results. Overall, the apparent reaction sensitivity of the minerals follows dolomite and calcite > montmorillonite and microcline.Figure 2Evolution of mineral composition with acid-corrosion durationTable 2Inorganic ion analysis results for the initial solution and the solution after 20 days of immersionInorganic ion analysis resultsNo.S25035-001S25035-002Submitted sampleOriginal solutionSolution after immersionSample typeLiquidLiquidTest itemUnitDetection limitResultResultpH valuedimensionless–3.83.8Chloridemg/L0.0072.1711.8Sulfatemg/L0.01817.8519Bicarbonatemg/L5<5<5Carbonatemg/L5<5<5Calciummg/L0.020.36156Potassiummg/L0.050.353.12Magnesiummg/L0.0030.03825.3Sodiummg/L0.120.4515 Evolution of mineral composition with acid-corrosion duration Table 2Inorganic ion analysis results for the initial solution and the solution after 20 days of immersionInorganic ion analysis resultsNo.S25035-001S25035-002Submitted sampleOriginal solutionSolution after immersionSample typeLiquidLiquidTest itemUnitDetection limitResultResultpH valuedimensionless–3.83.8Chloridemg/L0.0072.1711.8Sulfatemg/L0.01817.8519Bicarbonatemg/L5<5<5Carbonatemg/L5<5<5Calciummg/L0.020.36156Potassiummg/L0.050.353.12Magnesiummg/L0.0030.03825.3Sodiummg/L0.120.4515 Inorganic ion analysis results for the initial solution and the solution after 20 days of immersion Microstructural evolution Figure 3 presents the available SEM fields obtained after 0, 5, 10, 15, and 20 days of acid exposure. The ×200 fields are shown in Figures 3A–3E, and an additional ×500 view of the 20-day specimen is shown in Figure 3F. The uncorroded sandstone specimen at 0 days (Figure 3A) showed a comparatively intact granular texture in the selected SEM field. Grain edges were distinct and surfaces were relatively smooth, with no obvious dissolution pits. Dark surface regions were generally small and visually isolated in this field. The visible cement-like material appeared comparatively continuous, and no obvious large cracks or grain detachment were observed. After 5 days of acid exposure (Figure 3B), grain edges appeared more rounded, and local surface depressions were more evident. Dark regions appeared larger and more irregular, and some cement-like material was locally discontinuous. These 2D features are visually consistent with local mineral dissolution and possible cement loss, but they do not establish bulk pore enlargement, seepage channels, or connectivity changes. At 10 days (Figure 3C), surface depressions appeared deeper or more numerous and locally formed honeycomb-like or sponge-like textures. Some grain boundaries appeared less clearly cemented. These observations may indicate continued surface alteration and local weakening of grain contacts; they do not demonstrate increased bulk porosity or a semi-suspended granular skeleton. At 15 and 20 days (Figures 3D and 3E), the selected fields showed more extensive irregular dark regions, local detachment-like features, and less continuous cement-like material at some grain contacts. The surfaces appeared less compact than the 0-day field. Figure 3F provides a higher magnification view of the 20-day surface morphology. Because each duration is represented by a single field acquired under nonidentical imaging conditions, these observations are qualitative and should not be interpreted as complete cement disappearance or bulk structural loosening.Figure 3SEM images of sandstone after different acid-corrosion durations(A) 0 days at ×200.(B) 5 days at ×200.(C) 10 days at ×200.(D) 15 days at ×200.(E) 20 days at ×200.(F) 20 days at ×500.Scale bars, 300 μm. SEM images of sandstone after different acid-corrosion durations (A) 0 days at ×200. (B) 5 days at ×200. (C) 10 days at ×200. (D) 15 days at ×200. (E) 20 days at ×200. (F) 20 days at ×500. Scale bars, 300 μm. Quantitative 2D SEM descriptors To supplement the qualitative SEM observations, the five available ×200 fields were processed using a uniform local-dark-feature workflow (STAR Methods). The resulting binary masks for 0, 5, 10, 15, and 20 days are shown in Figures 4A–4E, respectively. The 2D dark-area fraction increased from 38.62% at 0 days to 60.03% at 5 days and then varied between 53.99% and 58.61% from 10 to 20 days (Figure 5A and Table 3). Connected-region counts decreased from 142 at 0 days to 35 and 31 at 5 and 10 days, respectively, and then increased to 47 and 48 at 15 and 20 days (Figure 5B and Table 3). Because the later images differed in contrast and sharpness, the count increase may reflect both secondary small dark regions and segmentation sensitivity.Figure 4Binary dark region masks derived from the ×200 SEM fields(A) 0 days.(B) 5 days.(C) 10 days.(D) 15 days.(E) 20 days.Figure 5Time-dependent 2D SEM descriptors(A) Dark area fraction.(B) Connected region count.(C) Equivalent diameter percentiles D50, D90, and D95.(D) Area share contributed by regions ≥200 px2.Table 3Quantitative SEM descriptors and source-image quality indicatorsDuration (d)Dark area (%)Regions (n)D50 (px)D90 (px)D95 (px)Mean grayP95-P5LapVar038.621427.720.3426.02139.2762120.4560.03356.921.3422.72149.094777.61058.61317.826.6831.42135.655521.41554.53476.924.0639.6280.463306.92053.99487.834.2760.33141.131222.0 Binary dark region masks derived from the ×200 SEM fields (A) 0 days. (B) 5 days. (C) 10 days. (D) 15 days. (E) 20 days. Time-dependent 2D SEM descriptors (A) Dark area fraction. (B) Connected region count. (C) Equivalent diameter percentiles D50, D90, and D95. (D) Area share contributed by regions ≥200 px2. Table 3Quantitative SEM descriptors and source-image quality indicatorsDuration (d)Dark area (%)Regions (n)D50 (px)D90 (px)D95 (px)Mean grayP95-P5LapVar038.621427.720.3426.02139.2762120.4560.03356.921.3422.72149.094777.61058.61317.826.6831.42135.655521.41554.53476.924.0639.6280.463306.92053.99487.834.2760.33141.131222.0 Quantitative SEM descriptors and source-image quality indicators The median equivalent diameter remained between 6.9 and 7.8 px, whereas D90 changed from 20.34 to 34.27 px and D95 changed from 26.02 to 60.33 px across the series (Figure 5C and Table 3). The D95 sequence increased continuously from 5 to 20 days (22.72, 31.42, 39.62, and 60.33 px), and the complementary cumulative distributions show a generally extended large-diameter tail at later times (Figure 6). Regions ≥200 px2 accounted for 80.0% of the detected dark area at 0 days and more than 95% at 5–20 days (Figure 5D and Table 3). These quantities describe the selected 2D fracture surfaces; they are not measurements of three-dimensional porosity, connected-pore volume, or a connectivity threshold.Figure 6Complementary cumulative distributions of equivalent dark-region diameterThe distributions are interpreted descriptively because one field was analyzed per time. Complementary cumulative distributions of equivalent dark-region diameter The distributions are interpreted descriptively because one field was analyzed per time. Stress-strain behavior and evolution of characteristic stresses Stress-strain response stages The stress-strain curves at confining pressures of 20, 30, and 40 MPa are shown in Figures 7A–7C, respectively. All specimens displayed obvious brittle characteristics, and the curves can be roughly divided into four stages: pore and fissure compaction, elastic deformation, yield deformation (stable crack propagation), and post-peak deformation and failure. No clearly independent unstable crack propagation stage was observed. Stage 1: pore and fissure compaction: in this stage, the stress-strain curve is concave upward, and axial strain increases relatively rapidly with stress. This response is commonly associated with closure of initial microcracks and voids under loading. The selected SEM fields show surface depressions consistent with chemical alteration, but they do not directly identify the internal features being compacted. The weakly expressed confinement effect on the initial nonlinear segment may be related to exposure-induced changes in local cementation and grain-contact conditions, although this mechanism was not measured directly. Stage 2: elastic deformation: in this stage, the stress-strain relation is approximately linear, and the curve slope (elastic modulus) remains essentially stable, indicating that the specimen response is dominated by elastic deformation. The upper limit of this stage corresponds to the crack initiation stress. The lower crack-initiation stresses observed after exposure may be associated with chemically altered grain boundaries, cement degradation, and pre-existing defects, but the present measurements do not isolate their individual contributions. Stage 3: stable microcrack propagation: the curve gradually becomes concave downward and axial deformation accelerates, indicating a response dominated by stable microcrack propagation. Possible weakening of local cementation and grain-contact boundaries may facilitate crack growth along altered interfaces, increasing the deformation rate and flattening the curve. Some long-duration specimens exhibit double-peak behavior, especially at 20 days. The upper limit of this stage is the crack damage stress, which represents the characteristic state before unstable crack growth. Stage 4: post-peak deformation and failure: after reaching peak strength, the stress of each specimen dropped rapidly, showing typical brittle failure characteristics and indicating that a complete brittle-ductile transition did not occur within the range of this study. Brief plastic flow may occur after the peak, after which the stress remains at a certain residual level.Figure 7Stress-strain responses of acid-corroded sandstone at different confining pressures(A) 20 MPa.(B) 30 MPa.(C) 40 MPa. Stress-strain responses of acid-corroded sandstone at different confining pressures (A) 20 MPa. (B) 30 MPa. (C) 40 MPa. Evolution of characteristic stresses The crack initiation stress (σci) and crack damage stress (σcd) were determined using the volumetric strain method described in STAR Methods. These parameters mark turning points in the measured volumetric-strain response. Their evolution at 20, 30, and 40 MPa is shown in Figures 8A–8C, respectively. The observed σci and σcd values generally decreased with exposure duration at all three confining pressures, although the specimen-level paths were not monotonic in every case. At 20 MPa, the five observations were approximately linear with time (Figure 8A), whereas the higher pressure series showed larger condition-to-condition fluctuations (Figures 8B and 8C). The convergence of some 20 days values at 30 and 40 MPa is an observation within the present matrix and does not establish a change in confining-pressure sensitivity. Replicated tests and additional pressure levels would be required for that inference.Figure 8Evolution of crack initiation and crack damage stresses with acid-corrosion duration(A) 20 MPa.(B) 30 MPa.(C) 40 MPa. Evolution of crack initiation and crack damage stresses with acid-corrosion duration (A) 20 MPa. (B) 30 MPa. (C) 40 MPa. Peak strength and Mohr-Coulomb strength parameters Time evolution of peak strength As summarized in Figure 9 and Table 4, the peak strengths under confining pressures of 20, 30, and 40 MPa all generally decreased with acid-corrosion duration. The absolute losses from 0 to 20 days were 106.26, 88.44, and 111.39 MPa, corresponding to decreases of 33.3%, 26.6%, and 29.9%, respectively. The mean losses across the three confining pressures were 34.76, 9.56, 9.29, and 48.42 MPa over 0–5, 5–10, 10–15, and 15–20 days, respectively. Because each condition was represented by one specimen, these interval differences are descriptive specimen-level observations and do not establish statistically distinct degradation stages.Figure 9Evolution of measured peak strength with acid-corrosion duration under different confining pressuresTable 4Measured peak strengthsAcid-corrosion duration/d20 MPa30 MPa40 MPa0319.25332.94372.305270.84310.79338.5810265.72293.02332.8015253.34283.06327.2720212.99244.50260.91 Evolution of measured peak strength with acid-corrosion duration under different confining pressures Table 4Measured peak strengthsAcid-corrosion duration/d20 MPa30 MPa40 MPa0319.25332.94372.305270.84310.79338.5810265.72293.02332.8015253.34283.06327.2720212.99244.50260.91 Measured peak strengths Evolution of strength envelopes and parameters Under conventional triaxial compression, the peak strength was described by the linear Mohr-Coulomb relation in Equation 1, where k is the slope of the strength envelope and σcs is the intercept extrapolated to σ3 = 0. The internal friction angle and cohesion were calculated as φ = arcsin[(k−1)/(k + 1)] and c = σcs/(2k0.5), respectively. Because the lowest confining pressure was 20 MPa, σcs is an extrapolated intercept and should not be equated with the measured uniaxial compressive strength.(Equation 1)σ1=σcs+kσ3 Independent linear regressions were performed using the three confining-pressure points at each corrosion duration, yielding R2 values of 0.928–0.989 (Table 5). Figure 10 shows that the strength envelopes generally shifted toward the lower strength region with corrosion duration, but the slope did not decrease monotonically: the k values at 0, 5, 10, 15, and 20 days were 2.653, 3.387, 3.354, 3.697, and 2.396, respectively. Therefore, the time-specific parameters can be used for descriptive comparison, but they should not be interpreted as high-precision material constants solely because of high R2 values.Table 5Independent Mohr-Coulomb regression results for each corrosion durationTime/dkσcs/MPaR2φ/(°)c/MPa02.6525261.9220.927626.9080.4153.3870205.1270.989432.9655.73103.3540196.5600.988632.7353.66153.6965176.9950.987435.0446.03202.3960167.5870.968024.2754.13Figure 10Mohr-Coulomb strength envelopes based on measured data under different acid-corrosion durations Table 5Independent Mohr-Coulomb regression results for each corrosion durationTime/dkσcs/MPaR2φ/(°)c/MPa02.6525261.9220.927626.9080.4153.3870205.1270.989432.9655.73103.3540196.5600.988632.7353.66153.6965176.9950.987435.0446.03202.3960167.5870.968024.2754.13 Independent Mohr-Coulomb regression results for each corrosion duration Mohr-Coulomb strength envelopes based on measured data under different acid-corrosion durations As shown in Figures 11 and 12, the apparent cohesion and apparent internal friction angle obtained from independent regressions both showed nonmonotonic changes. To reflect the uncertainty associated with the limited sample size, the pooled residual variance from all time-group regressions was used, and 95% intervals were estimated through Monte Carlo propagation of the parameter covariance. The intervals for parameters at different times overlapped substantially, indicating that the current data are insufficient to support a monotonic time function for c or φ.Figure 11Time-specific estimates and the pooled common-slope estimate of apparent internal friction angleFigure 12Time-specific and common-slope-model estimates of apparent cohesion Time-specific estimates and the pooled common-slope estimate of apparent internal friction angle Time-specific and common-slope-model estimates of apparent cohesion To test whether the available data support a systematic change in confining-pressure sensitivity, corrosion duration was treated as a five-level categorical factor. For time group i and observation j, the reduced model M0 used time-specific intercepts and one common slope, whereas the full model M1 included time-specific intercepts and time-by-confining-pressure interaction terms (time-specific slopes), as given in (Equation 2), (Equation 3):(Equation 2)M0:σ1,ij=αᵢ+kσ3,ij+εᵢⱼ(Equation 3)M1:σ1,ij=αᵢ+kᵢσ3,ij+εᵢⱼ M0 contained six parameters (five intercepts and one common slope), with a residual sum of squares (RSS) of 473.129 and 9 residual degrees of freedom. M1 contained ten parameters (five intercepts and five slopes), RSS = 233.423, and 5 residual degrees of freedom. The nested comparison gave F(4,5) = 1.284 and p = 0.388. Thus, the additional interaction terms did not produce a statistically detectable improvement at α = 0.05, but this failure to reject M0 does not prove that the slopes are identical. The reduced-model results are summarized in Table 6, and Figure 13 compares the uncorroded and 20-day strength envelopes. The reduced model gave a pooled common slope k = 3.0972 (SE = 0.2293; 95% confidence interval [CI], 2.5785–3.6159), corresponding conditionally to a pooled apparent internal friction angle of 30.8°. The five intercept estimates were 248.581, 213.821, 204.264, 194.974, and 146.551 MPa; each had SE = 8.052 MPa and the respective 95% CIs were reported in Table 7. Under this simplification, apparent cohesion decreased from 70.6 to 41.6 MPa (41.0%) from 0 to 20 days. The more strongly supported conclusion is therefore a downward change in the overall load-bearing level. Possible envelope rotation and systematic evolution of apparent internal friction angle remain unresolved.Table 6Results of the common-slope Mohr-Coulomb modelTime/dIntercept/MPaCommon φ/(°)Apparent c/MPa0248.58130.78870.6245213.82130.78860.74810204.26430.78858.03315194.97430.78855.39420146.55130.78841.636Figure 13Comparison of Mohr-Coulomb envelopes for the uncorroded state and the 20-day acid-corroded stateTable 7Complete reduced and full Mohr-Coulomb model statistics and reduced-model parameter uncertaintyTerm/modelDefinitionParametersEstimate/RSSdf or SEStatisticp or intervalM05 intercepts + common slope6473.1299––M15 intercepts +5 slopes10233.42351.2840.388α0M0 intercept, 0 d–248.581SE 8.05295% CI230.366–266.796α5M0 intercept, 5 d–213.821SE 8.05295% CI195.606–232.036α10M0 intercept, 10 d–204.264SE 8.05295% CI186.049-222.479α15M0 intercept, 15 d–194.974SE 8.05295% CI176.759–213.189α20M0 intercept, 20 d–146.551SE 8.05295% CI128.336–164.766KM0 common slope–3.0972SE 0.229395% CI2.5785–3.6159 Table 6Results of the common-slope Mohr-Coulomb modelTime/dIntercept/MPaCommon φ/(°)Apparent c/MPa0248.58130.78870.6245213.82130.78860.74810204.26430.78858.03315194.97430.78855.39420146.55130.78841.636 Results of the common-slope Mohr-Coulomb model Comparison of Mohr-Coulomb envelopes for the uncorroded state and the 20-day acid-corroded state Table 7Complete reduced and full Mohr-Coulomb model statistics and reduced-model parameter uncertaintyTerm/modelDefinitionParametersEstimate/RSSdf or SEStatisticp or intervalM05 intercepts + common slope6473.1299––M15 intercepts +5 slopes10233.42351.2840.388α0M0 intercept, 0 d–248.581SE 8.05295% CI230.366–266.796α5M0 intercept, 5 d–213.821SE 8.05295% CI195.606–232.036α10M0 intercept, 10 d–204.264SE 8.05295% CI186.049-222.479α15M0 intercept, 15 d–194.974SE 8.05295% CI176.759–213.189α20M0 intercept, 20 d–146.551SE 8.05295% CI128.336–164.766KM0 common slope–3.0972SE 0.229395% CI2.5785–3.6159 Complete reduced and full Mohr-Coulomb model statistics and reduced-model parameter uncertainty Nested-model transparency: M0 (6 parameters) had RSS = 473.129 and 9 residual df; M1 (10 parameters) had RSS = 233.423 and 5 residual df. The common slope was k = 3.0972 ± 0.2293 (SE), with a 95% CI of 2.5785–3.6159. Full parameter standard errors and intervals are given in Table 7. Figure 13 shows that after 20 days of acid corrosion, the peak strength at each confining-pressure level decreased markedly, and the strength envelope shifted as a whole toward the lower stress region. Because the independent regression slopes at 0 and 20 days were 2.653 and 2.396, respectively, while the nested test did not detect a significant time-confining pressure interaction, the slope difference in the figure should be regarded as a descriptive result under a limited sample size. The core evidence at present is the decrease in the intercept and overall load-bearing level, rather than a proven systematic degradation of frictional parameters. Failure modes and peak strain Macroscopic failure morphology The macroscopic failure characteristics of sandstone specimens under different acid-corrosion durations and confining pressures are shown in Figure 14. At each acid-corrosion duration, increasing confining pressure generally made specimen failure less violent and more distributed, although no complete brittle-to-ductile transition was identified (Figures 14A–14O). Morphologically, the failure pattern changed from a single dominant macroscopic crack to a more distributed dendritic crack cluster. This suggests that higher confining pressure dispersed internal stress concentrations from a single local direction to multiple directions, and the failure mechanism gradually shifted from being dominated by axial stress toward joint control by axial stress and confining pressure. The amount of frictional rock powder and debris generated during failure also generally increased with confining pressure.Figure 14Failure characteristics of sandstone specimens under different acid-corrosion durations and confining pressures(A–C) 0 days; (D–F) 5 days; (G–I) 10 days; (J–L) 15 days; (M–O) 20 days. Within each group, the confining pressures are 20, 30, and 40 MPa from left to right, respectively. Failure characteristics of sandstone specimens under different acid-corrosion durations and confining pressures (A–C) 0 days; (D–F) 5 days; (G–I) 10 days; (J–L) 15 days; (M–O) 20 days. Within each group, the confining pressures are 20, 30, and 40 MPa from left to right, respectively. With increasing acid-corrosion duration, the macroscopic failure morphology evolved from a single dominant crack toward more complex dendritic crack patterns. This evolution can be observed by comparing Figures 14A, 14D, 14G, 14J, and 14M at 20 MPa; Figures 14B, 14E, 14H, 14K, and 14N at 30 MPa; and Figures 14C, 14F, 14I, 14L, and 14O at 40 MPa. After 20 days of acid exposure, the failure characteristics under confining pressures of 20, 30, and 40 MPa are shown in Figures 14M–14O, respectively. Under the combined effects of high confining pressure and prolonged acid exposure, denser spiderweb-like crack networks and more pronounced accumulations of rock powder and debris could develop. Together with the SEM observations, these macroscopic changes are consistent with acid-induced mineral dissolution, possible cement loss, and weakening of grain contacts. Evolution of peak strain Figure 15 shows the peak strains of sandstone specimens under different confining pressures and acid-corrosion durations. Increasing confining pressure generally enhanced the deformation capacity of the specimens, whereas increasing acid-corrosion duration reduced the overall peak axial strain, indicating that microstructural deterioration weakened the pre-peak deformation capacity and energy-storage limit of sandstone. The confining effect and the weakening effect of acid corrosion coexist, but they are not simply equivalent and should be analyzed comprehensively in relation to the stress path, corrosion degree, and specimen heterogeneity.Figure 15Distribution of peak strain under different confining pressures and acid-corrosion durations Distribution of peak strain under different confining pressures and acid-corrosion durations Time-dependent chemo-mechanical coupled damage constitutive model Chemical damage variable and time evolution The reference modulus E0 = 13.1 GPa was obtained from the uncorroded zero-confinement test. To avoid mixing this value with confinement-dependent triaxial moduli, the zero-confinement quantities are denoted E0(0) and Et(0) in the chemical-damage definition used here. The 20, 30, and 40 MPa moduli are reported separately in Table 8 and are used as descriptive stiffness data rather than being divided directly by E0(0). The independent uncorroded reference tests are documented in Figure S2.Table 8Elastic moduli from the triaxial tests (GPa)Confining pressure (MPa)0 d5 d10 d15 d20 d2017.2415.316.2615.4914.563017.5416.2716.9816.7514.744016.8315.1315.515.114.55 Table 8Elastic moduli from the triaxial tests (GPa)Confining pressure (MPa)0 d5 d10 d15 d20 d2017.2415.316.2615.4914.563017.5416.2716.9816.7514.744016.8315.1315.515.114.55 Elastic moduli from the triaxial tests (GPa) Chemical reactions between acidic solutions and soluble minerals and cements in sandstone cause pore expansion, cementation weakening, and microcrack development. These microstructural changes are macroscopically manifested as a decrease in elastic modulus. Therefore, the chemical damage variable can be defined according to elastic modulus degradation. Here, the elastic modulus was obtained as the slope of the linear fit to the linear elastic segment of the stress-strain curve, typically within 30%–60% of the peak strength (Equation 4):(Equation 4)Dc=1−EtE0where Dc is the chemical damage variable induced by acid corrosion (0 ≤ Dc ≤ 1, with Dc = 0 representing the intact uncorroded state), Et is the elastic modulus of the rock after t days of acid corrosion, E0 is the initial elastic modulus of the intact rock specimen at 0 days, and t is the acid-corrosion duration (days). According to the time-accumulation characteristics of acid corrosion reactions, the chemical damage evolution is assumed to satisfy Equation 5:(Equation 5)dDcdt=kc(1−Dc)where kc is the chemical damage evolution coefficient. Separating variables in Equation 5 gives Equation 6:(Equation 6)dDc1−Dc=kcdt Integrating Equation 6 gives Equation 7:(Equation 7)−ln(1−Dc)=kct The corresponding definite-integral form is given in Equation 8:(Equation 8)∫0DcdDc1−Dc=∫0tkcdt According to fitting of the experimental data, the chemical damage evolution coefficient kc is 0.0093. The fitted chemical damage model is given in Equation 9:(Equation 9)Dc(t)=1−e−0.0093t For each curve, 30%–60% of the corresponding peak stress was used as an initial candidate interval for modulus identification and was checked to lie after pore compaction and below crack initiation. The exponential coefficient kc = 0.0093 d−1 summarizes the overall modulus-retention trend and does not require every single-specimen value to vary monotonically. Figure S3 displays the normalized retention values and residuals based on the rounded triaxial moduli. Mechanical damage variable and parameter calibration The strength of rock microelements is intrinsically random. During loading, microcracks gradually initiate, propagate, and coalesce. Assuming that the microelement strength follows a Weibull distribution,13 its cumulative distribution function, interpreted as the failure probability of microelements, is expressed as(Equation 10)P(ε)=1−exp[−(εε0)m] Defining the volume fraction of failed microelements as the mechanical damage variable gives Equation 11:(Equation 11)Dm=1−exp[−(εε0)m]where Dm is the mechanical damage variable, ε is the axial strain, ε0 is the scale parameter, and m is the shape parameter. The Weibull parameters ε0 and m were calibrated using the experimental stress-strain curve of uncorroded sandstone under 20 MPa confining pressure, and only data from the linear elastic stage and the stable crack propagation stage (ε ≤ 0.98εpeak) were used. The Weibull equation was linearized as ln[−ln(1−Dm)] = m lnε−m lnε0, and least-square regression gave m = 53.2 and ε0 = 0.0225, with a coefficient of determination R2 = 0.98. These parameters correspond to the present calibration condition and are not universal constants transferable to all confining pressures or lithologies. The calibrated mechanical-damage relation is presented in Equation 12 and in the subsequent in-sample fit assessment. The calibration can equivalently be written as Y = mX + b, where X = ln ε1 and Y = ln[−ln(1−Dm)], so that ε0 = exp(−b/m). The values m = 53.2 and ε0 = 0.0225 were obtained from the uncorroded 20 MPa pre-peak data (ε1 ≤ 0.98εpeak). They are condition-dependent curve-shape parameters rather than universal constants. A ±10% change in kc changes Dc(20 d) from 0.1541 to 0.1850, and the effects of ε0 and m are concentrated around the onset and steepness of pre-peak damage evolution.(Equation 12)Dm=1−exp[−(ε0.0225)53.2] Chemo-mechanical coupled constitutive relationship Acid-corrosion damage and mechanical-load damage jointly weaken the load-bearing capacity of the rock. Based on the Lemaitre strain equivalence hypothesis and using a multiplicative coupling principle, the total chemo-mechanical coupled damage variable D is defined in Equation 13:(Equation 13)1−D=(1−Dc)(1−Dm) Rearranging Equation 13 gives Equation 14:(Equation 14)D=1−(1−Dc)(1−Dm) According to the Lemaitre strain equivalence principle,14 the stress-strain relation of damaged rock is expressed in Equation 15:(Equation 15)σ1=(1−D)E0ε1+2νσ3 According to the strain equivalence hypothesis, the rock material first transforms from the initial undamaged state to a chemically corroded state and then develops into a fully damaged state under mechanical loading. Substituting (Equation 9), (Equation 12) into Equation 13 gives the integrated evolution equation in Equation 16:(Equation 16)1−D=e−0.0093texp[−(ε10.0225)53.2] The component terms in Equation 16 are given in Equation 17:(Equation 17)1−Dc=e−0.0093t1−Dm=exp[−(ε10.0225)53.2] Therefore, the time-dependent chemo-mechanical coupled damage constitutive model for sandstone under acid corrosion is given in Equation 18:(Equation 18)σ1=E0e−0.0093texp[−(ε10.0225)53.2]ε1+2νσ3 According to the test results, the initial elastic modulus is E0 = 13.1 GPa and Poisson’s ratio is ν = 0.21. Substituting these material constants into the derived model gives the final chemo-mechanical coupled damage constitutive equation (Equation 19):(Equation 19)σ1=13.1e−0.0093texp[−(ε10.0225)53.2]ε1+0.42σ3 Physical meaning of model parameters and calibration boundaries The model parameters are condition dependent. kc is influenced by solution chemistry, temperature, transport conditions, and soluble-mineral content; m and ε0 depend on confining pressure, lithology, and the initial-defect population; and ν affects the confinement contribution through 2νσ3. In this study, kc was fitted to the available 0- to 20-day modulus-retention trend, whereas m and ε0 were calibrated from the uncorroded 20 MPa pre-peak curve. Sensitivity expressions and the calibration formulas are provided in STAR Methods. These parameters must be recalibrated for other rocks, solutions, pressure ranges, or long-term field conditions. Model fit assessment at 20 MPa To assess the in-sample performance of the time-varying chemo-mechanical coupled damage model, the theoretical curves were compared with the measured responses at 0, 5, 10, 15, and 20 days under 20 MPa confining pressure, as shown in Figures 16A–16E, respectively. This comparison evaluates how the calibrated formulation represents corrosion-duration effects within the present dataset; it is not an independent predictive validation.Figure 16Comparison between theoretical calculations and experimental stress-strain responses at 20 MPa(A) 0 days.(B) 5 days.(C) 10 days.(D) 15 days.(E) 20 days. Comparison between theoretical calculations and experimental stress-strain responses at 20 MPa (A) 0 days. (B) 5 days. (C) 10 days. (D) 15 days. (E) 20 days. The theoretical curves reproduced the principal pre-peak trends of the five measured responses, with R2 values of 0.767, 0.898, 0.944, 0.929, and 0.961. Agreement was weaker in the initial concave-up compaction segment because the strain-equivalence and Weibull formulation mainly describes matrix elasticity and progressive damage rather than explicit closure of irregular voids and cracks. Agreement improved in the linear-elastic and stable-propagation portions. The model represented the overall modulus-degradation trend and the earlier onset of nonlinear deformation, but these agreements should be interpreted as in-sample fit rather than proof of the proposed microscopic mechanism. At 20 MPa, the model reproduced the overall downward trend in peak strength, although agreement varied among durations and the uncorroded peak was underestimated. The comparison therefore supports a bounded description of the present dataset rather than an independently verified predictive capability. Cross-pressure prediction, post-peak residual behavior, and long-term extrapolation require evaluation with independent data. Cross-pressure in-sample assessment and error indicators The model was assessed in-sample across all 15 corrosion-duration-confining-pressure conditions (Figure S1 and Table 9). Across the individual pre-peak curves, R² ranged from 0.7670 to 0.9610, root mean square error (RMSE) ranged from 14.197 to 36.820 MPa, mean absolute error (MAE) ranged from 12.498 to 31.503 MPa, and normalized root mean square error (NRMSE), calculated by normalizing RMSE to peak stress, ranged from 6.67% to 12.30%. Assessment across all three confining-pressure levels provides a cross-pressure evaluation within the available data, but it does not constitute external validation.Table 9Pre-peak constitutive-model fit-assessment metrics for all 15 conditionsσ3 (MPa)t (d)RMSEMAER2NRMSE (%)Pred. peakMeas. peak20030.1826.3370.7679.45268.434319.25320522.13619.0120.8988.17256.238270.837201020.11816.790.9447.57244.596265.724201521.71118.3950.9298.57233.482253.339202014.19712.4980.9616.67222.874212.98730035.32331.5030.834910.61268.434332.94530524.95320.0480.94238.03335.958310.794301021.77517.950.94927.43320.693293.019301528.8324.140.900410.18306.123283.064302030.06227.1820.875512.3292.214244.49840036.8229.7940.90899.89363.886372.30440529.49624.3070.93268.71347.353338.584401026.46721.9460.94257.95331.571332.802401526.60422.1530.948.13316.506327.271402031.16629.2850.860311.95302.125260.91 Table 9Pre-peak constitutive-model fit-assessment metrics for all 15 conditionsσ3 (MPa)t (d)RMSEMAER2NRMSE (%)Pred. peakMeas. peak20030.1826.3370.7679.45268.434319.25320522.13619.0120.8988.17256.238270.837201020.11816.790.9447.57244.596265.724201521.71118.3950.9298.57233.482253.339202014.19712.4980.9616.67222.874212.98730035.32331.5030.834910.61268.434332.94530524.95320.0480.94238.03335.958310.794301021.77517.950.94927.43320.693293.019301528.8324.140.900410.18306.123283.064302030.06227.1820.875512.3292.214244.49840036.8229.7940.90899.89363.886372.30440529.49624.3070.93268.71347.353338.584401026.46721.9460.94257.95331.571332.802401526.60422.1530.948.13316.506327.271402031.16629.2850.860311.95302.125260.91 Pre-peak constitutive-model fit-assessment metrics for all 15 conditions The largest discrepancies remain in the initial pore-compaction segment, around some peak locations, and in the post-peak response. Quantitative assessment is therefore restricted to the aligned pre-peak portions and peak stress. The calculation of all data errors is based on the comparison between the original measured data and the results from numerical calculation models. The present formulation does not include residual fracture-surface friction, plastic sliding, strain localization, or residual load-bearing terms and should not be used as a whole-process post-peak constitutive model. Discussion Comparison with previous studies and contributions of this study Current research on acid-corroded rocks mainly follows three mainstream lines. The first compares rock strength degradation and energy evolution by controlling variables such as solution pH, environmental temperature, and corrosion-cycle number. The second focuses on fractured rock masses and constructs damage constitutive models coupling macroscopic mechanical responses and mesoscopic damage evolution. The third improves the theoretical framework of rock chemo-mechanical coupling from the perspectives of mineral reaction, solute transport, and fracture propagation mechanisms. Table 10 summarizes representative recent studies in this field. Existing conclusions indicate a basic consensus: mineral dissolution caused by acid solutions leads to different degrees of decrease in the elastic modulus and peak strength of rocks, lowers the crack initiation stress level, and advances crack initiation. However, different studies differ substantially in variable selection, tested lithology, and model construction scale, and a unified description of the evolution law has not yet been formed.Table 10Comparison between recent studies on acid-corroded rocks and the present workStudyObject and environmentMethods and main findingsRelationship to this studyChen et al.15sandstone; chemical corrosionrevealed, from mechanical and energy perspectives, that acid corrosion reduces strength and energy-storage capacityprovides a macroscopic degradation reference but does not emphasize the distinction between time path and confining-pressure effectChen et al.16sandstone; pH 1–7 coupled with high temperatureconducted triaxial tests and established a temperature-mechanical-chemical damage modelemphasizes multiphysics coupling; this study focuses on time effects under constant acidityLiu et al.17pre-fractured red sandstone; pH = 2established a coupled model of macroscopic fissures and Weibull microdamagefocuses on fissure inclination; this study focuses on corrosion duration and confining pressure for intact sandstoneTang and Hu18,19carbonate rocks; reactive fluids and stressed cracksrevealed the bidirectional reinforcing mechanism between mineral mass removal and microcrack propagationprovides theoretical support for the multiscale damage chain proposed in this studyLuo et al.12water-bearing sandstone; acidic environmentanalyzed deformation failure and energy evolution, confirming that a strongly acidic environment significantly weakens strengthprovides an experimental comparison for long-term acid corrosionThis studyfine-grained sandstone; nominal target pH 3.6 and measured initial pH 3.8, 0–20 dXRD/SEM/ion analysis/triaxial tests, MC-nested tests, and a time-dependent dual-damage modelemphasizes the time path, envelope translation, and statistical interpretation boundaries Table 10Comparison between recent studies on acid-corroded rocks and the present workStudyObject and environmentMethods and main findingsRelationship to this studyChen et al.15sandstone; chemical corrosionrevealed, from mechanical and energy perspectives, that acid corrosion reduces strength and energy-storage capacityprovides a macroscopic degradation reference but does not emphasize the distinction between time path and confining-pressure effectChen et al.16sandstone; pH 1–7 coupled with high temperatureconducted triaxial tests and established a temperature-mechanical-chemical damage modelemphasizes multiphysics coupling; this study focuses on time effects under constant acidityLiu et al.17pre-fractured red sandstone; pH = 2established a coupled model of macroscopic fissures and Weibull microdamagefocuses on fissure inclination; this study focuses on corrosion duration and confining pressure for intact sandstoneTang and Hu18,19carbonate rocks; reactive fluids and stressed cracksrevealed the bidirectional reinforcing mechanism between mineral mass removal and microcrack propagationprovides theoretical support for the multiscale damage chain proposed in this studyLuo et al.12water-bearing sandstone; acidic environmentanalyzed deformation failure and energy evolution, confirming that a strongly acidic environment significantly weakens strengthprovides an experimental comparison for long-term acid corrosionThis studyfine-grained sandstone; nominal target pH 3.6 and measured initial pH 3.8, 0–20 dXRD/SEM/ion analysis/triaxial tests, MC-nested tests, and a time-dependent dual-damage modelemphasizes the time path, envelope translation, and statistical interpretation boundaries Comparison between recent studies on acid-corroded rocks and the present work Compared with previous studies, the core academic contribution of this work is not merely to verify again the weakening effect of acid corrosion on the mechanical properties of sandstone. Instead, under test conditions with constant acid type and target pH, this study systematically reveals the time-evolution path of rock properties over a 0- to 20-day corrosion period and connects XRD phase analysis, solution ion concentration testing, SEM mesoscopic observation, characteristic stress evolution, measured peak strength, and constitutive model construction into an integrated analytical framework. In particular, a common-slope model and a nested F-test statistical method are introduced to quantitatively distinguish two evolution modes of the strength envelope: overall downward translation and systematic change in confining-pressure sensitivity. This avoids the common research pitfall of overinterpreting the time evolution of internal friction angle and cohesion solely on the basis of high coefficients of determination obtained from three-point linear regressions. Evidence-bounded interpretation of multiscale damage evolution Figure 17 summarizes the evidence-bounded multiscale interpretation. Together, the observations support a bounded multiscale interpretation rather than a directly demonstrated damage sequence. The endpoint increases in Ca2+ and Mg2+, and the XRD trends are consistent with carbonate dissolution. In the selected SEM fields, larger and more irregular dark regions and locally less continuous cement-like material are consistent with surface alteration and possible local cement loss. These features could reduce local contact stiffness and facilitate crack initiation, but the present data do not measure particle-contact area, three-dimensional connectivity, or force-chain topology. Reaction-chemo-mechanical models reported in previous studies provide a plausible conceptual basis for coupling mineral removal and microcrack development; applying that mechanism to the present sandstone remains a hypothesis. Terms such as “pore-connectivity transition,” “degradation of load-transfer paths,” and “critical instability” are therefore used only as hypotheses, linking the observed surface morphology with the mechanical response. The available SEM images cannot establish bulk-connected porosity, three-dimensional contact area, or force-chain topology. Verification would require multiple independent fields, micro-CT or NMR, pore-network measurements, differential interference contrast (DIC), and/or grain-scale numerical analysis.Figure 17Conceptual illustration of the possible multiscale alteration pathways in acid-exposed sandstone Conceptual illustration of the possible multiscale alteration pathways in acid-exposed sandstone The quantitative 2D SEM descriptors suggest a possible two-part morphological path within the selected fields. From 0 to 5 days, the dark-area fraction increased by about 55%, the number of connected regions decreased by about 75%, and the area contribution of regions ≥200 px2 exceeded 96%, consistent with coalescence or enlargement of surface depressions. From 5 to 20 days, the area fraction varied within a narrower range while D90, D95, and the large-diameter tail generally increased. A plausible interpretation is the coexistence of continuing enlargement of large dissolution depressions, local wall detachment, surface rearrangement, and secondary small-pit formation. This interpretation remains conditional because the images are single fields acquired under nonidentical imaging conditions. The specimen-level peak-strength data do not establish statistically distinct rapid-slow-accelerated degradation stages. Differences between adjacent exposure durations may reflect cumulative chemical alteration together with specimen heterogeneity. Endpoint chemistry and single-field SEM cannot resolve time-dependent reaction kinetics or a connectivity threshold. Accordingly, the 0- to 20-day results are interpreted as an overall deterioration trend with condition-specific fluctuations; any staged mechanism remains a hypothesis requiring time-resolved chemistry, replicated mechanical tests, and three-dimensional pore characterization. Overall, the decreases in calcite and dolomite and the increases in dissolved Ca2+ and Mg2+ are consistent with cumulative carbonate dissolution during 0–20 days of acid exposure. The SEM observations and quantitative 2D descriptors indicate enlargement and redistribution of dark surface regions, but they do not establish three-dimensional pore connectivity or degradation of load-transfer paths. Possible competition between confinement and acid-exposure weakening Confinement and acid exposure have opposing observed effects on sandstone strength. Confining pressure tends to close cracks, increase normal contact forces, and restrict dilatancy, thereby increasing peak strength. Acid exposure is associated with the dissolution of susceptible minerals and weakening of cemented contacts. After 20 days, peak strength decreased by 33.3%, 26.6%, and 29.9% at confining pressures of 20, 30, and 40 MPa, respectively, while characteristic stresses and elastic moduli also showed overall deterioration. Intermediate differences remain specimen-level trends because only one specimen was tested per condition and should not be interpreted as statistically proven degradation stages. The present measurements do not directly quantify contact stiffness or internal load-transfer paths. The nested comparison gave F(4,5) = 1.284 and p = 0.388; under the reduced-model common-slope simplification, the slope was 3.0972 (95% CI, 2.5785–3.6159), and apparent cohesion decreased by 41.0%. These results support an overall downward shift of the strength envelope but do not demonstrate slope equality or a constant internal friction angle. Possible changes in confining-pressure sensitivity remain unresolved because of the limited pressure range, low residual degrees of freedom, and lack of replicates. The observed macroscopic failure patterns are consistent with a qualitative competition between confinement and exposure-related weakening. Confinement can suppress crack opening and promote more distributed failure, whereas chemically altered interfaces may reduce local crack-initiation resistance. Because internal crack paths and contact forces were not measured in situ, the relative roles of these processes cannot be quantified from the present tests and should not be converted directly into an equivalent confining pressure or burial depth. Model applicability and uncertainty Advantages and limitations of the constitutive model Most existing constitutive studies of chemically corroded rocks use solution pH, porosity, or mass-loss ratio as quantitative indicators of final-state damage. In this study, the elastic modulus degradation process is formulated as an explicit time function and coupled with a mechanical damage model based on the Weibull distribution, enabling separate characterization of chemical damage accumulated before loading and mechanical damage induced during loading. Chen et al.16 and Liu et al.17 verified the feasibility of combining the Lemaitre equivalent strain assumption with a statistical damage framework under thermo-chemical coupling conditions and in fractured sandstone specimens, respectively. Tang and Hu18,19 emphasized, from the mechanistic perspective of reactive transport, the bidirectional feedback between chemical dissolution and microcrack development. In comparison, the advantage of the present model is that it uses only a small number of physically meaningful parameters to describe the influence of corrosion duration on pre-peak stiffness and strength evolution. However, the model does not explicitly solve reactant transport, pore-connectivity evolution, or the feedback effect of the stress field on reaction rate. Across all 15 tested conditions, the time-dependent chemo-mechanical model reproduced the principal pre-peak responses, with R2 values of 0.767–0.961 and NRMSE values of 6.67%–12.30%. Its present scope is limited to pre-peak stiffness degradation, progressive damage, and peak-strength attenuation. Post-peak instability, residual strength, other lithologies and solutions, and long-term field prediction require additional experiments and recalibration. The applicability limitations of the model should also be recognized. First, the Weibull formulation and the assumption that failed microelements immediately lose all load-bearing capacity provide an idealized representation of post-peak behavior; actual specimens can retain strength through confinement and friction along rough fracture surfaces. Second, the chemical damage function was fitted over 0–20 days and does not include acid depletion, diffusion control, or secondary precipitation, so it cannot be extrapolated without constraint to multi-year corrosion. The parameters were calibrated mainly from the 20 MPa data and assessed in-sample across the available pressure range; parameter identification should be repeated and independently tested for cross-pressure applications. Long-term monitoring, independent validation, and back analysis remain necessary for reliable time-dependent parameters.20 To improve the model description of post-peak behavior, residual friction parameters can be introduced in future work, and a friction-degradation function related to plastic strain can be added to the effective-stress expression in Equation 20:(Equation 20)σ=(1−D)Eε+σrexp(−βεpε0)where, σr is the residual strength parameter, β is the friction degradation coefficient, and εp is the plastic strain. This form can realize a gradual post-peak stress decrease without disrupting the underlying damage-evolution framework. The influence of acid immersion duration on residual strength can also be further quantified, for example, by establishing an empirical relation between σr and time, so that the model can completely describe the whole mechanical behavior of acid-corroded sandstone from pre-peak damage to the post-peak residual stage. Limitations of the data and statistical interpretation The nested model has limited power: each time-specific regression uses three pressure points and one residual degree of freedom, and the full interaction model has five residual degrees of freedom. A nonsignificant interaction can therefore reflect either genuinely small slope differences or insufficient information to detect them. Future designs should include additional low and intermediate pressures (for example, 0, 5, 10, 20, 30, 40, and 50 MPa) and at least three replicate specimens per condition. In this study, only three confining-pressure levels, 20, 30, and 40 MPa, were set for each corrosion duration, and the residual degree of freedom of the time-specific Mohr-Coulomb regression was only 1. According to the current matrix, each corrosion duration-confining pressure combinations had only one representative specimen. Therefore, a high R2 only indicates approximate linearity of the data within the range of 20–40 MPa and does not imply high statistical precision for c and φ. The nested F-test and Monte Carlo intervals can restrict overinterpretation, but they cannot replace parallel tests. In addition, the Mohr-Coulomb intercept is extrapolated from the 20–40 MPa confining-pressure range to σ3 = 0 and should be referred to as a Mohr-Coulomb (MC) extrapolated intercept or apparent uniaxial strength. It cannot be equated with uniaxial compressive strength (UCS) obtained from direct uniaxial tests. Future studies should add low and intermediate confining-pressure points and set no fewer than three parallel specimens for each condition to identify the time evolution of c, φ, and confining-pressure sensitivity more reliably. Recent studies on chemically corroded rocks also commonly emphasize that uncertainties in the inverse estimation of c and φ should be reduced through multiple confining-pressure points, repeated tests, and comparisons among multiple strength criteria.16,21 Therefore, in this study, the Mohr-Coulomb parameters are positioned as apparent parameters within a given confining-pressure range rather than as the only criterion for long-term rock-mass strength. Engineering implications The directly supported engineering conclusion is limited to the tested fine-grained sandstone, dilute sulfuric acid exposure, 0- to 20-day duration, and 20–40 MPa confining-pressure range. Closed mines affected by acidic mine water and acid-exposed sandstone formations are the closest application settings, but quantitative use requires site-specific calibration for mineralogy, water chemistry, seepage and recharge, temperature, in situ stress, and exposure history. CO2 storage and geothermal development are considered only as analogy-based chemo-mechanical settings because their fluids, temperatures, pressures, transport processes, and precipitation reactions differ substantially. The present 20-day immersion results should not be linearly scaled to multi-year service life. Based on the aforementioned conclusions, in engineering analysis, the confining pressure in laboratory triaxial tests should not be directly equated with formation burial depth or support resistance, and the degradation laws obtained from short-term tests should not be simply extrapolated linearly to cover the entire service life of a mine. In practical applications, corresponding scale conversion and parameter correction should be completed in combination with actual conditions such as in situ stress level, acid-solution recharge rate, seepage conditions, and formation temperature. High-sulfur coalfields are widely distributed in China. Long-term acidification of water accumulated in abandoned goafs can both increase the risk of mine water hazards and continuously corrode roof sandstone water-resisting strata. The findings of this study indicate that, for mining areas affected by AMD, long-term stability evaluations such as overburden movement after mine closure and safety assessment of water-resisting strata should not directly use rock mechanical parameters measured at the early exploration stage. When conducting engineering stability analysis, it is necessary to incorporate the time dimension and comprehensively characterize the long-term evolution of rock mechanical parameters in combination with acid-solution recharge conditions, seepage characteristics, environmental temperature, and in situ stress state. The constitutive model established in this study can provide theoretical support for stability analysis of water-resisting strata in closed mines, rock-mass assessment in areas affected by acidic groundwater, and time-dependent stability calculation of deep roadways. However, long-term parameter extrapolation still requires calibration using longer duration laboratory tests and field monitoring data. Such structural-integrity problems induced by chemo-mechanical coupling also exist in engineering fields, such as geothermal development and CO2 geological storage. The acidic environment formed when CO2 dissolves in formation water, together with salt precipitation, can significantly change the elastic parameters and local stress distribution of sandstone.22 This also indicates that engineering risks arising from the coupling among chemical environment, pore structure, and mechanical response have common features across different underground engineering scenarios. Overall, acid corrosion significantly changes the deformation laws and failure modes of underground engineering rock masses. For long-term stability evaluation of rock masses in acidic-water environments, the traditional purely stress-controlled analytical framework should be expanded to a time-dependent chemo-mechanical coupled analysis system. At the same time, engineering applications should pay attention to applicability boundaries and avoid unconstrained extrapolation beyond the corrosion duration and confining-pressure range of the present tests. Limitations of the study (1)Each corrosion-duration-confining-pressure combination contains one representative specimen, so within-condition variance, confidence intervals for mean mechanical responses, and formal stage comparisons cannot be estimated.(2)Each corrosion duration is represented by one ×200 SEM field. Imaging conditions differ, the images lack a reliable scale for μm conversion, and dark fracture shadows cannot be completely separated from true voids. The 2D metrics are not bulk porosity or connected-pore volume.(3)Solution chemistry was measured only initially and after 20 days. The endpoint data cannot determine stage-specific release rates, transport limitation, pH evolution, or intermediate precipitation.(4)The nested Mohr-Coulomb test has low residual degrees of freedom and a restricted 20–40 MPa pressure range. The regression-derived cohesion and friction angle are apparent parameters rather than robust, independently replicated material constants.(5)The constitutive model is calibrated to the present material and exposure conditions and is quantitatively evaluated only for pre-peak behavior. Residual friction, localization, post-peak softening, and long-term reactive transport are outside its present formulation.(6)Field AMD is chemically more complex than the exposure solution used here and commonly contains Fe, Al, Mn, and other coexisting ions in addition to H+ and sulfate. The present sulfuric-acid/distilled-water system was therefore designed to isolate sustained proton-sulfate attack under controlled conditions; it should be regarded as a simplified acidic sulfate environment relevant to one principal component of AMD, not as a complete simulation of field AMD. (1)Each corrosion-duration-confining-pressure combination contains one representative specimen, so within-condition variance, confidence intervals for mean mechanical responses, and formal stage comparisons cannot be estimated. Each corrosion-duration-confining-pressure combination contains one representative specimen, so within-condition variance, confidence intervals for mean mechanical responses, and formal stage comparisons cannot be estimated. (2)Each corrosion duration is represented by one ×200 SEM field. Imaging conditions differ, the images lack a reliable scale for μm conversion, and dark fracture shadows cannot be completely separated from true voids. The 2D metrics are not bulk porosity or connected-pore volume. Each corrosion duration is represented by one ×200 SEM field. Imaging conditions differ, the images lack a reliable scale for μm conversion, and dark fracture shadows cannot be completely separated from true voids. The 2D metrics are not bulk porosity or connected-pore volume. (3)Solution chemistry was measured only initially and after 20 days. The endpoint data cannot determine stage-specific release rates, transport limitation, pH evolution, or intermediate precipitation. Solution chemistry was measured only initially and after 20 days. The endpoint data cannot determine stage-specific release rates, transport limitation, pH evolution, or intermediate precipitation. (4)The nested Mohr-Coulomb test has low residual degrees of freedom and a restricted 20–40 MPa pressure range. The regression-derived cohesion and friction angle are apparent parameters rather than robust, independently replicated material constants. The nested Mohr-Coulomb test has low residual degrees of freedom and a restricted 20–40 MPa pressure range. The regression-derived cohesion and friction angle are apparent parameters rather than robust, independently replicated material constants. (5)The constitutive model is calibrated to the present material and exposure conditions and is quantitatively evaluated only for pre-peak behavior. Residual friction, localization, post-peak softening, and long-term reactive transport are outside its present formulation. The constitutive model is calibrated to the present material and exposure conditions and is quantitatively evaluated only for pre-peak behavior. Residual friction, localization, post-peak softening, and long-term reactive transport are outside its present formulation. (6)Field AMD is chemically more complex than the exposure solution used here and commonly contains Fe, Al, Mn, and other coexisting ions in addition to H+ and sulfate. The present sulfuric-acid/distilled-water system was therefore designed to isolate sustained proton-sulfate attack under controlled conditions; it should be regarded as a simplified acidic sulfate environment relevant to one principal component of AMD, not as a complete simulation of field AMD. Field AMD is chemically more complex than the exposure solution used here and commonly contains Fe, Al, Mn, and other coexisting ions in addition to H+ and sulfate. The present sulfuric-acid/distilled-water system was therefore designed to isolate sustained proton-sulfate attack under controlled conditions; it should be regarded as a simplified acidic sulfate environment relevant to one principal component of AMD, not as a complete simulation of field AMD. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Zeng Yifan (zengyf@cumtb.edu.cn). Materials availability This study did not generate new unique reagents or distributable biological materials. Rock specimens are finite geological materials and are not available as a standardized renewable resource. Data and code availability •The source data supporting the findings of this study are provided in Data S1.•The analysis code used to reproduce the reported statistical summaries and associated figures is provided in Data S1.•Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. •The source data supporting the findings of this study are provided in Data S1. The source data supporting the findings of this study are provided in Data S1. •The analysis code used to reproduce the reported statistical summaries and associated figures is provided in Data S1. The analysis code used to reproduce the reported statistical summaries and associated figures is provided in Data S1. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments This work was supported by the 10.13039/501100012166National Key Research and Development Program of China (2023YFC3012102), the 10.13039/501100001809National Natural Science Foundation of China (42372297), the Ordos New R&D Institution Construction Project (IMRI23001), the Ordos Landmark Innovation Team Project (TD20232309), and the Ordos Postdoctoral Research Fund (OPRF2404). Author contributions L.H., writing – review and editing, writing – original draft, methodology, and formal analysis; Z.Y., writing – review and editing, supervision, resources, and funding acquisition; W.Q., resources and funding acquisition; C.Y., investigation and data curation; Y.D., methodology, investigation, and data curation. Declaration of interests The authors declare no competing interests. Declaration of generative AI and AI-assisted technologies in the writing process Language-assistance tools were used to improve English clarity and document organization. The authors reviewed and take full responsibility for the scientific content, data interpretation, and final text. No generative AI tools were used to create or modify the graphical abstract or any other figure. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIEROtherFine-grained sandstoneRock blocks used in this studyNot applicableDistilled waterLaboratory supplyNot applicableSulfuric acid (analytical grade)Howei Pharm, Guangzhou, ChinaNot applicableJSR-DPR300 ultrasonic testerJSRJSR-DPR300YSC-1000 triaxial testing machineNational Engineering Research CenterYSC-1000Source dataThis studyData S1Analysis codeThis studyData S1 REAGENT or RESOURCESOURCEIDENTIFIEROtherFine-grained sandstoneRock blocks used in this studyNot applicableDistilled waterLaboratory supplyNot applicableSulfuric acid (analytical grade)Howei Pharm, Guangzhou, ChinaNot applicableJSR-DPR300 ultrasonic testerJSRJSR-DPR300YSC-1000 triaxial testing machineNational Engineering Research CenterYSC-1000Source dataThis studyData S1Analysis codeThis studyData S1 Experimental model and study participant details The experimental material was bluish-gray fine-grained sandstone obtained from intact rock blocks. No human participants, animals, cell lines, or biological model organisms were involved. Method details Test materials and specimen preparation Sandstone is one of the main lithologies in coal-measure overburden. Bluish-gray fine-grained sandstone was cored from intact blocks and processed into right circular cylinders with nominal diameter 50 mm and height 100 mm (height-to-diameter ratio 2.0; Figure S4). Diameter and height deviations were controlled within ±0.3 mm, maximum end-face unevenness within ±0.05 mm, and end-face perpendicularity deviation below 0.25°. Specimens with visible macroscopic fractures or anomalous ultrasonic velocities were excluded. Pre-test P-wave velocity, S-wave velocity, and density were recorded for consistency assessment (Table 1). Test equipment The tests were performed using a YSC-1000 computer-controlled servo rock triaxial testing machine. The equipment is located at the National Engineering Research Center of Coal Mine Water Hazard Prevention and Control and Water Resource Protection and is a comprehensive testing platform capable of simulating various environmental conditions and conducting multiphysics-coupled rock tests. Acid corrosion treatment and microstructural characterization The immersion solution was prepared from sulfuric acid (H2SO4) and distilled water, with a nominal target pH of 3.6; an independently measured initial value of 3.8 is reported in Table 2. The simplified solution was selected to isolate proton-sulfate attack and did not reproduce the Fe, Al, Mn, ionic-strength, redox, complexation, or precipitation processes of multicomponent field AMD. Specimens were exposed for 0, 5, 10, 15, or 20 days. The available historical record did not document solution volume, water/rock ratio, replacement schedule, temperature, sealing condition, or intermediate pH monitoring; these missing metadata are reported rather than inferred. Microstructural characterization included XRD and SEM tests, which were used to analyze changes in mineral composition and surface morphology under different corrosion durations. In addition, inorganic ion analysis was conducted on the initial acid solution and the 20 d immersion solution to assist in identifying the main mineral reaction pathways. SEM image processing Each of the five available ×200 SEM images was linearly stretched over its 1st-99th grayscale percentile range to 0–255 and denoized with a 3 × 3 Gaussian filter. A black-hat transform using a 31 × 31-pixel elliptical structuring element was used to extract locally dark features. Pixels were classified as dark regions when the black-hat response was ≥25 and the normalized grayscale was ≤90. Connected components smaller than 16 px2 were removed. The remaining regions were used to calculate 2D dark-area fraction, count, equivalent-diameter percentiles, CCDFs, and large-region area contribution. Parameter sensitivity was assessed over black-hat thresholds 18–30, absolute-darkness thresholds 80–100, and minimum component areas 8–64 px2. The principal sequence (lower at 0 day, higher at 5 days, followed by smaller fluctuations) remained stable within this range. Mean gray level, P95-P5 contrast, and Laplacian variance were also calculated (Table 3). Because one field was available per time and the 15 and 20 days images had lower quality, the analysis is descriptive and two-dimensional. Triaxial compression test scheme Five corrosion durations (0, 5, 10, 15, and 20 days) and three confining pressures (20, 30, and 40 MPa) formed 15 conditions (Figure S5). One representative specimen was assigned to each condition. Before loading, the platen was brought close to the specimen and a 1 kN preload was applied at 0.1 kN/s. After sealing the chamber, confining pressure was increased at 0.05 MPa/s to the target value. Following stabilization, axial displacement was applied at 0.02 mm/min until failure. Axial load/stress, axial and radial displacement/strain, circumferential strain, and confining pressure were recorded at 1 Hz. The one-specimen design supports descriptive trend analysis but not within-condition uncertainty estimation. Quantification and statistical analysis The crack initiation stress (σci) and crack damage stress (σcd) were determined using the volumetric strain method; that is, the corresponding stress thresholds were identified from characteristic turning points on the volumetric strain-axial strain curves. The peak strength was taken as the maximum axial stress of each stress-strain curve. The peak strength was described using the linear Mohr-Coulomb relation in Equation 1, where k is the slope of the envelope and σcs is the intercept extrapolated to σ3 = 0. The internal friction angle and cohesion were calculated as φ = arcsin[(k - 1)/(k + 1)] and c = σcs/(2k0.5), respectively. Because the lowest confining pressure was 20 MPa, σcs is only an MC extrapolated intercept and is not equivalent to the measured uniaxial compressive strength. To distinguish between the overall translation and slope change of the envelope caused by acid corrosion, time-specific independent regression models, a common-slope model for all times, and a time-confining pressure interaction model were established. A nested F test was used to determine the significance of the interaction term (significance level α = 0.05). Parameter intervals were estimated through Monte Carlo propagation based on the pooled residual variance and parameter covariance. Nested Mohr-Coulomb models Corrosion duration was represented by a five-level categorical factor. M0 used five time-specific intercepts and a common slope; M1 added four independent time-by-confining-pressure interaction degrees of freedom, equivalently allowing a separate slope at each time. Ordinary least squares was used for both models. The model forms are given in (Equation 2), (Equation 3), and the nested-test statistic is given in Equation 21.(Equation 21)F=[(RSS0−RSS1)/(p1−p0)]/[RSS1/(n−p1)] For n = 15, p0 = 6, and p1 = 10, the observed values were RSS0 = 473.129, RSS1 = 233.423, F(4,5) = 1.284, and p = 0.388. The reduced-model covariance matrix yielded k = 3.0972, SE = 0.2293, and 95% CI 2.5785–3.6159. The five reduced-model intercepts each had SE = 8.052 MPa (Table 7). Constitutive-model calibration and sensitivity (Equation 22)Eσ3,t=Σ(ε1,i−ε¯1)(σ1,i−σ¯1)/Σ(ε1,i−ε¯1)2 (Equation 23)RE(t)=E(t)/E(0);Dc(t)=1−RE(t)=1−exp(−kct) (Equation 24)Dm=1−exp[−(ε1/ε0)ᵐ] (Equation 25)ln[−ln(1−Dm)]=mlnε1−mlnε0 (Equation 26)σ1=E0exp(−kct)exp[−(ε1/ε0)ᵐ]ε1+2νσ3 Equations 22, 23, 24, 25, and 26 summarize the modulus, retention, mechanical-damage, calibration, and constitutive relations used for parameter estimation. The reported values were E0 = 13.1 GPa, ν = 0.21, kc = 0.0093 d−1, m = 53.2, and ε0 = 0.0225. E0 and ν came from the uncorroded zero-confinement reference test, kc summarized the modulus-retention trend, and m and ε0 were fitted to the uncorroded 20 MPa pre-peak curve. Parameter sensitivity and extrapolation limits are described in the Results and Discussion. Model-error metrics (Equation 27)RMSE=(1/n)Σ(yᵢ−ŷᵢ)2 (Equation 28)MAE=(1/n)Σ|yᵢ−ŷᵢ| (Equation 29)R2=1−Σ(yᵢ−ŷᵢ)2/Σ(yᵢ−ȳ)2 (Equation 30)NRMSE=RMSE/σpeak×100 The error metrics are defined in Equations 27, 28, 29, and 30. Metrics were calculated on aligned pre-peak data. Peak stress was used to normalize RMSE so that near-zero stresses in the initial segment did not dominate a percentage error. Post-peak data were excluded from quantitative assessment because residual-strength mechanisms are not included. Additional resources No study registration or protocol repository applies. Source data and analysis code are available as described under Resource Availability.