Physical and Mechanical Properties of High-Strength Concrete Incorporating Bottom Ash and Red Mud as Partial Cement Replacement

Concrete remains the most widely used construction material worldwide because of its high compressive strength, ease of forming [1], ready availability of constituent materials, durability, and resistance to biological decay [2]. The global scale of concrete consumption, however, places sustained pressure on the natural resources from which its constituents-particularly cement-are produced, motivating decades of research into supplementary cementitious materials (SCMs) derived from industrial by-products [3]. High-strength concrete (HSC), defined by Indonesian National Standard (SNI) 03-6468-2000 [4] as concrete with a compressive strength equal to or greater than 41.4 MPa, has become a structural material of choice for high-rise buildings, long-span bridges, dams, piers, chimneys, and tunnels [5] because its superior compressive capacity permits more slender structural elements and more efficient use of space and material [6]. As HSC adoption grows, so too does the need to identify SCMs that are both technically compatible with the demanding mix-design requirements of HSC and environmentally beneficial [7].

Indonesia possesses substantial bauxite reserves [8], with West Kalimantan Province identified as one of the country's most significant bauxite-bearing regions [9]. PT Indonesia Chemical Alumina (ICA), located in Tayan Hilir, Sanggau Regency, processes locally mined bauxite ore into alumina (Al₂O₃) via the Bayer process [10], in which raw bauxite is digested under high temperature and pressure in concentrated sodium hydroxide (NaOH) solution [11]. This process generates red mud (RM), a fine, reddish-brown, highly alkaline semi-solid residue, at an average rate of 0.3–0.5 tonnes per tonne of Al₂O₃ extracted, depending on the bauxite grade processed [12]. Globally, Al₂O₃ refineries generate more than 300 million tonnes of RM annually, and because the material carries high alkalinity and elevated heavy-metal ion content [13-15], Indonesian Government Regulation No. 18 of 1999 classifies RM as hazardous waste; uncontrolled accumulation in tailings ponds presents a recognized ecological liability [10].

The Al₂O₃ refining process is itself energy-intensive, and the energy demand at the ICA facility is met by a coal-fired steam power plant (Pembangkit Listrik Tenaga Uap, PLTU) [11]. Coal combustion in this plant produces fly ash and bottom ash (BA) as residues [16]. BA is the coarser, heavier combustion by-product that settles at the base of the furnace and accumulates in the ash hopper, distinguishing it from the finer, lighter fly ash that is carried with flue gases [16, 17]. BA typically exhibits a granular texture comparable to sand and is reported to carry silica (SiO₂)- and Al₂O₃-bearing phases that motivate its evaluation as a candidate pozzolanic material [17]; the present study does not itself verify pozzolanic reactivity, which requires mineralogical and reactivity testing beyond the scope of this work. Both RM and BA are therefore generated in tandem at the same industrial complex as functionally related by-products of the bauxite-to-Al₂O₃ value chain. Nevertheless, they originate from chemically distinct processes [18]-one hydrometallurgical, one combustion-based-and consequently may offer complementary, though not yet mechanistically confirmed, mineralogical contributions when considered as candidate SCMs [19].

Prior studies have evaluated RM and BA separately as partial cement replacements, generally in normal-strength concrete, with reported optimum dosages around 10% for each material individually [20-23]. Evidence on their combined use specifically within HSC remains limited, leaving open the question of whether the two industrial wastes interact synergistically, neutrally, or antagonistically when incorporated together-a question of direct practical relevance given that both materials are co-located at the same source facility and a combined utilization strategy would simplify waste-valorization logistics.

This study addresses that gap by experimentally evaluating HSC mixtures in which BA is held at a fixed 10% cement replacement-consistent with the optimum dosage reported in prior single-material studies [20, 22, 23]-while RM is varied from 0% to 15% of cement weight, in order to (1) analyze the influence of combined BA and RM addition on the physical and mechanical properties of HSC, and (2) identify, within statistically defensible limits, the best-performing RM dosage within this combined system. The physical properties examined are unit weight and fresh-state workability (mortar consistency, slump, and slump flow); the mechanical properties examined are compressive strength development through 28 days, splitting tensile strength, and modulus of elasticity, each analyzed using the individual replicate results (mean ± standard deviation, one-way analysis of variance (ANOVA), and Tukey's honestly significant difference (HSD) post-hoc test) rather than mean values alone. By isolating the marginal effect of RM against a fixed BA background and subjecting the results to inferential statistical testing, the experimental design provides an evidence base-appropriately qualified by its statistical power and single-laboratory scope-for a dual-waste utilization strategy for the Al₂O₃-refining and coal-power sectors of West Kalimantan.

3.1 Fresh-state properties and workability

Incorporation of BA and RM produced a visually detectable change in concrete color, progressing from light gray in the control and bottom-ash-only mixtures (V1, V2) to grayish-brown in the moderate red-mud mixtures (V3, V4) and to a reddish light-brown in the highest red-mud mixture (V5), consistent with the characteristic reddish-brown coloration of RM itself. This color shift is a convenient qualitative field indicator of red-mud dosage during batching, though it is not itself a performance measure and is not treated as such in the hypothesis testing below.

Table 6. Fresh-state properties of high-strength concrete (HSC) mixtures

Mortar flow consistency after superplasticizer addition, based on the mean of four dial readings per mixture (without an independent replicate structure suitable for computing a standard deviation), ranged from 100.50% (V5) to 110.50% (V3) (Table 6), comfortably within the 100-150% workability range reported by Neville and Brooks. Consistency increased from V1 (103.75%) to a peak at V3 (110.50%) before declining at V4 (105.00%) and V5 (100.50%). Slump after superplasticizer addition, reported as mean ± SD across n = 3 field readings, was highest for the control mixture V1 (23.50 ± 0.46 cm) and decreased across the substituted mixtures to 14.90-17.30 cm for V2-V5, consistent with the increased fines content associated with bottom-ash incorporation; BA is reported in the literature to have an angular, porous particle texture that increases internal friction in fresh mixtures [17], though this texture was not independently verified for the batches used in this study (Section 3.7), so the reduction is described here as consistent with, rather than demonstrated to result from, that mechanism. All slump values remained within the SNI 03-6468-2000 [4] allowance for HRWR-modified HSC mixtures. Slump flow, reported as mean ± SD across n = 2 field readings, followed a parallel trend, ranging from 30.00 to 38.00 cm, with the control mixture again showing the highest flowability. Given the small number of replicates available for these fresh-state indices (n = 2-3 per mixture), formal ANOVA is reported only for the properties measured on hardened specimens (Sections 3.2-3.5); the fresh-state trends described here should be read as descriptive rather than as statistically confirmed differences. Figure 1 presents workability indicators with their standard deviations.

tu_pian_1.png

Figure 1. Slump, slump flow, and mortar consistency across mix variations (mean ± standard deviation (SD))

3.2 Unit weight

Hardened unit weight at 28 days ranged from 2,321.54 ± 31.24 kg/m³ (V4) to 2,353.37 ± 12.05 kg/m³ (V1), as summarized in Table 7 and Figure 2. All five variations fall within the SNI 03-2847-2002 [36] normal-weight concrete classification of 2,200-2,500 kg/m³. A one-way ANOVA across the five mixtures did not detect a statistically significant difference in unit weight (F(4,10) = 2.09, p = 0.157); the apparent downward trend in unit weight from V1 through V4 is therefore descriptive and is consistent with, but not statistically proven by, the lower specific gravities of BA (2.546) and RM (3.019) relative to cement (3.062). Because the unit-weight variation across all mixtures spans only about 1.4% of the mean value and is not statistically distinguishable among mixtures, the practical structural-design implication of bottom-ash and red-mud substitution on self-weight is negligible within the tested dosage range.

tu_pian_2.png

Figure 2. Unit weight of high-strength concrete (HSC) at 28 days (mean ± standard deviation (SD)); shaded band shows the SNI 03-2847-2002 normal-weight range

Table 7. Unit weight of high-strength concrete (HSC) at 28 days (n = 3 per mixture)

3.3 Compressive strength development

Mean compressive strength results (± SD, n = 3) for each variation at curing ages of 3, 7, 14, 21, and 28 days are presented in Table 8 and Figure 3. All five mixtures show the expected pattern of strength gain with curing age. A one-way ANOVA (df = 4, 10) was performed separately at each age to test for differences among the five mixtures (Table 9); statistically significant differences were detected at 3, 14, 21, and 28 days (p < 0.05 in each case) but not at 7 days (p = 0.444), indicating that the mixture effect on compressive strength is not uniform across the curing period and should not be assumed present at every age from the 28-day result alone.

Table 8. Compressive strength by curing age and mix variation, mean ± SD (MPa), n = 3 per cell

Table 9. One-way analysis of variance (ANOVA) results for compressive strength at each curing age

tu_pian_3.png

Figure 3. Compressive strength development with curing age (mean ± standard deviation (SD), n = 3)

tu_pian_4.png

Figure 4. 28-day compressive strength by mix variation (mean ± standard deviation (SD), n = 3)

At 28 days, Tukey's HSD post-hoc test partitions the five mixtures into two statistically distinct groups (Figure 4): V1, V2, and V3 form one group (compact letter “a”), not significantly different from one another (all pairwise p > 0.19), while V4 and V5 form a second, significantly lower group (“b”, both significantly below V1, V2, and V3 at p < 0.05; V4 versus V5 itself is not significant p = 0.65). This is an important qualification of the numerically highest result: although V3 (10% BA + 5% RM) recorded the highest mean 28-day compressive strength (53.69 ± 2.24 MPa), exceeding V1 (52.20 ± 1.68 MPa) by 2.85% and V2 (52.84 ± 1.68 MPa) by 1.61%, this difference is not statistically significant given the observed within-group variability (n = 3 per mixture). V3 should therefore be described as the numerically best-performing composition among the tested mixtures at 28 days, and not as significantly superior to the control or to the bottom-ash-only mixture. What is statistically supported is that raising RM content to 10% and 15% (V4, V5, on a fixed 10% bottom-ash background) produces a significant reduction in 28-day compressive strength relative to V1, V2, and V3-both V4 (45.41 MPa) and V5 (42.87 MPa) fall below the 50 MPa design target, whereas V1, V2, and V3 meet or exceed it.

The quadratic and exponential regression models used in an earlier draft of this analysis to describe the relationship between RM content and 28-day compressive strength, and to model the strength-gain curve with age, have been removed from this revision. Refitting a second-order polynomial through the measured 28-day means for V2-V5 (BA fixed at 10%; RM at 0%, 5%, 10%, 15%) produces a curve that decreases monotonically over the entire tested range (0-15% RM) with its vertex located outside the tested domain; it therefore does not mathematically substantiate an optimum at 5% RM, even though 5% is the best-performing dosage among the four discrete levels actually tested. Likewise, an exponential strength-gain model of the form fc(t) = fc,28(1 − e−kt) fitted to the 3-28-day data produced “asymptotic” 28-day strength parameters that were, for several mixtures, lower than the actually measured 28-day strength-an internally inconsistent result that cannot be reported as a valid model. Both observations are described here as a methodological caution: with only four to five discrete dosage levels and no replication at intermediate red-mud contents, the present data can identify which of the tested compositions performed best, but cannot support a continuous mathematical optimum or growth-rate model without additional intermediate dosages and repeated testing.

Two hypotheses may explain the pattern in which V3 performs numerically, though not statistically significantly, better than V1 and V2, while V4 and V5 perform significantly worse: at low red-mud dosage, RM-rich in Al₂O₃ (61.24% Al₂O₃ by the literature-derived composition in Table 4) and iron oxide-may act as a fine filler that improves particle packing without a large dilution penalty, consistent with the corresponding increase in mortar consistency from V1 to V3 (Section 3.1); at higher combined replacement (10% BA plus 10% or 15% RM, i.e., 20-25% of cement mass), the cumulative reduction in reactive Portland cement may outweigh any packing or pozzolanic contribution. Because no scanning electron microscopy (SEM), XRD, thermogravimetric analysis (TGA), or actual XRF characterization of the RM used was performed in this study, these mechanisms remain hypotheses consistent with the observed mechanical trend rather than confirmed explanations, and are stated using hedged language (“may be attributed to,” “is consistent with”) throughout this discussion. A further, independent constraint applies specifically to any reading of this pattern as a BA-RM synergy: because BA was held fixed at 10% and only RM content was varied (Section 2.2), the experimental design cannot statistically test a BA×RM interaction term, and the numerically favorable V3 result should not be interpreted as statistical evidence of a synergistic effect between the two materials.

3.4 Splitting tensile strength

Splitting tensile strength results at 28 days, recalculated from the individual specimen loads, are presented in Table 10 and Figure 5. Unlike compressive strength, splitting tensile strength did not follow a simple monotonic trend with combined substitution: the control mixture V1 recorded the highest mean splitting tensile strength (3.87 ± 0.57 MPa), followed by V2 (3.49 ± 0.18 MPa), V4 (3.25 ± 0.14 MPa), V5 (3.04 ± 0.32 MPa), and V3 (2.88 ± 0.15 MPa) as the lowest. A one-way ANOVA indicated a statistically significant overall difference among mixtures (F(4,10) = 4.49, p = 0.025); however, Tukey's HSD post-hoc test identifies only the V1–V3 pair as individually significant (p = 0.022), while V2, V4, and V5 are not significantly different from either V1 or V3 (compact letters: V1 = a, V3 = b, V2/V4/V5 = ab). The practical reading of this result is that the reduction in tensile strength is clearly established between the control and the V3 optimum-compressive-strength mixture, but the intermediate mixtures cannot be confidently ranked against one another with the available replication (n = 3 per mixture). The ACI 318-1999 [41] reference value in Table 10 is evaluated using each mixture's own measured 28-day compressive strength (Table 8) rather than the 50 MPa design value, so the comparison reflects the concrete actually produced.

Table 10. Splitting tensile strength at 28 days

tu_pian_5.png

Every mixture, including the control, fell below its own mixture-specific ACI 318-1999 reference value, with the shortfall most pronounced for V3 (70.1% of the reference) despite V3 having the highest compressive strength among all mixtures. This divergence between compressive and tensile performance is consistent with the general understanding that splitting tensile strength is governed strongly by the interfacial transition zone (ITZ) and the tensile capacity of the cement paste matrix, both of which the literature reports as more sensitive to angular, porous SCM particle texture than bulk compressive strength; this literature-based mechanism has not been verified for the present materials, since neither particle morphology nor microstructural imaging of the BA/RM batches or the hardened paste was performed in this study (Section 3.7), and it is offered here only as one plausible, unconfirmed explanation for the observed divergence between compressive and tensile results. The practical implication is that meeting or exceeding the compressive-strength target at V2 or V3 does not imply a proportional tensile-strength benefit, and designers should verify splitting tensile strength independently rather than inferring it from compressive-strength compliance when specifying these SCMs in HSC.

3.5 Modulus of elasticity

Modulus of elasticity was evaluated at 28 days using four calculation approaches under ASTM C469 [43] -secant modulus, chord modulus, initial tangent modulus, and tangent modulus-alongside the mixture-specific ACI 363R-92 [44] empirical prediction; the individual-method means are summarized in Table 11 and shown in full in Figure 6 (chord modulus) for statistical comparison. Because ASTM C469 [43] defines the standard reported elastic modulus as the chord modulus (measured between 0.05fc′ and 0.40fc′ on the stress–strain curve), the statistical analysis in this section is reported for chord modulus specifically. This choice avoids treating all four values as equivalent alternative outputs of the same standard method, consistent with reviewer guidance that presenting four unqualified moduli side by side makes interpretation ambiguous.

tu_pian_6.png

Of the 15 modulus specimens cast (three per mixture), 13 yielded valid chord-modulus readings: V2, V3, and V4 each contributed all three specimens, while V1 and V5 each contributed two, one specimen from each having failed to produce a valid reading in the source data (Table 11). A one-way ANOVA on these 13 chord-modulus readings indicated a highly significant overall difference among mixtures (F(4, 8) = 16.60, p < 0.001). Tukey's HSD post-hoc test shows that the control mixture V1 (32.47 ± 4.19 GPa) is significantly stiffer than all four bottom-ash/red-mud mixtures (V2-V5, all p < 0.01), while V2 (22.16 ± 1.36 GPa), V3 (17.93 ± 1.31 GPa), V4 (22.23 ± 2.54 GPa), and V5 (17.28 ± 0.38 GPa) do not differ significantly from one another (all pairwise p > 0.18). This indicates that incorporating BA and RM together-at any of the tested red-mud dosages-produces a statistically significant stiffness penalty relative to the unmodified concrete, but the present data cannot statistically distinguish among the different red-mud dosages themselves in terms of chord modulus. For the control mixture V1, the secant and chord moduli exceed the ACI 363R-92 [44] prediction (26.80 GPa) by roughly 20–24%, consistent with V1's denser, single-binder matrix. For the modified mixtures, the chord modulus falls below the ACI 363R-92 [44] prediction by roughly 8% (V2) to 34% (V3, V5). This indicates that the empirical ACI relationship, calibrated broadly across conventional HSC, tends to overestimate stiffness for bottom-ash/red-mud-modified mixtures at the higher substitution levels and should not be applied to this system without an appropriate reduction factor, which would need to be established from a larger dataset than the present study provides.

Table 11. Modulus of elasticity by calculation method and mix variation (MPa)

3.6 Synthesis: Best-performing composition and governing mechanisms

Considered together, the statistically tested results converge on a qualified narrative around the V3 mixture (10% BA + 5% RM) as the numerically best-performing composition within the tested range, rather than an unambiguous optimum across all properties. Table 12 consolidates the complete set of results, together with their statistical qualification, against the relevant reference standards.

Two hedged hypotheses may explain the pattern in which mechanical performance diverges between compressive strength (numerically peaking at V3, though not significantly above V1/V2) and both splitting tensile strength and chord modulus (both significantly reduced by BA/RM incorporation relative to V1, largely independent of red-mud dosage beyond the initial addition of BA). First, RM at low dosage may act as a fine filler that improves particle packing in the paste, broadly consistent with the corresponding rise in mortar consistency from V1 to V3; this filler effect, if present, appears to benefit compressive strength more than tensile strength or stiffness, plausibly because compressive strength is governed by the bulk paste-aggregate system while tensile strength and modulus are more sensitive to the ITZ and micro-flaw distribution, which the literature associates with angular, porous SCM particle texture-a texture reported for BA and RM generally, but not independently verified for the batches used here, since particle morphology was not measured in this study (Section 3.7). Second, at higher combined replacement (V4, V5), compressive strength is significantly reduced (Section 3.3), while tensile strength and modulus-already reduced at the V2/V3 dosages-do not deteriorate much further; the cumulative dilution of reactive Portland cement is proposed as one plausible explanation for the compressive-strength reduction specifically, but this causal attribution is itself untested and should be distinguished from the statistical significance of the strength reduction it attempts to explain. Both explanations remain hypotheses consistent with the measured mechanical trends; confirming them would require SEM, XRD, TGA, and porosity measurements that were outside the scope of this study, and the language above (“may,” “plausibly,” “appears to”) is used throughout to reflect that these are proposed, not demonstrated, mechanisms. Likewise, because RM content was the only factor varied against a BA level held fixed at 10% (Section 2.2), neither explanation should be read as evidence of a statistically confirmed BA–RM interaction; the present one-factor-at-a-time design can describe the marginal effect of RM but cannot establish that the two materials act synergistically.

Table 12. Consolidated summary of physical and mechanical test results with statistical qualification

Taken together, and subject to the statistical qualifications established in Sections 3.3-3.5, the evidence supports 10% BA combined with 5% RM (V3) as the best-performing combined substitution level identified in this study for meeting the 50 MPa compressive-strength target while utilizing the maximum combined waste content (15% of cement mass) before mechanical performance deteriorates significantly; however, V3 does not show a statistically demonstrated compressive-strength advantage over the control, and it does show a significant tensile-strength penalty relative to the control. Any structural application of this mixture should therefore weigh the compressive-strength result against the tensile and stiffness trade-offs documented here, and should not describe the compressive result as a proven improvement over unmodified concrete.

3.7 Study limitations

Several limitations constrain the interpretive scope of this study and are consolidated here rather than left as isolated caveats. Characterization of the BA and RM materials used was limited to specific gravity and unit weight (Table 1); particle size distribution, particle morphology, and the pH and loss-on-ignition of these materials were not measured for the batches used, and the chemical compositions in Tables 3 and 4 are literature-derived rather than batch-specific XRF results (Section 2.2). This constrains the pozzolanic-filler mechanism proposed in Sections 3.3 and 3.6 to a hypothesis consistent with the observed mechanical trends rather than a microstructurally confirmed explanation, a constraint that also applies to the absence of SEM, XRD, and TGA of the hardened paste.

Because BA was held fixed at 10% while only RM content was varied, the experimental design isolates the marginal effect of RM against a constant bottom-ash background but cannot statistically confirm an interaction between the two materials; any synergy implied by the V3 result should be read as a hypothesis for future factorial testing rather than a demonstrated finding. The dataset further originates from a single undergraduate thesis conducted in one laboratory with n = 2–3 replicates per condition, limiting the statistical power to detect smaller effects, particularly for splitting tensile strength and the descriptively reported fresh-state indices (Section 3.1); two data-reconciliation discrepancies between the source thesis’s raw specimen data and its summary tables, affecting the 28-day compressive strength of V4 and the splitting tensile strength of V1, V2, and V5, should be resolved against the original laboratory records before submission (Sections 3.3-3.4).

Finally, the solid content and specific gravity of the superplasticizer and the capping compound and the thickness used before compressive and modulus testing are not recorded in the source data and should be obtained and reported by the author (Sections 2.2-2.3). None of these limitations undermines the mechanical performance results reported in Sections 3.2-3.5, which rest on directly measured, standardized test methods, but together they bound the strength of the mechanistic and generalization claims drawn from this study.