Low-carbon ultra-high-performance concrete: raw materials, particle packing optimization, and data-driven performance assessment—a review Ahmed Fageeri, Ousmane Hisseine Abstract Ultra-high-performance concrete (UHPC) offers exceptional mechanical properties and durability; however, its cement-intensive formulations raise concerns regarding environmental sustainability. This study addresses this challenge by establishing quantitative relationships between UHPC mixture design, performance indicators, and embodied CO2 emissions by coupling a literature survey with data-driven analysis of over 200 mixtures. Unlike previous studies that examine the behavior of raw materials and the resulting performance in isolation, this study leverages data-driven approaches to link mixture design parameters to mechanical performance, durability indicators, and embodied CO2 emissions, thereby providing useful insights into low-carbon UHPC formulations. The influence of supplementary cementitious materials and mineral fillers is evaluated across rheological, mechanical, durability, and carbon footprint aspects. Machine learning (i.e., Random Forest) analysis demonstrates high accuracy in predicting compressive strength from mixture design parameters, while tensile properties and permeability remain influenced by microstructural factors not captured by mixture design parameters. Heat map correlation reveals strong interdependencies among mechanical and durability performance indicators, with compressive strength highly correlated (95%–98%) across curing ages and inversely related to permeability-related durability metrics. An exponential relationship between cement-to-binder and water-to-cement ratios is established (R2 of 95%), providing a strong predictive tool for mixture design. From a carbon footprint perspective, cement is the dominant contributor to CO2 emissions, as expected. Interestingly, cement content above 800 kg/m3 has negligible contributions to mechanical properties while significantly increasing the embodied CO2 emissions. By adopting a data-driven approach, this study provides novel insights into designing UHPC mixtures that foster higher mechanical and durability performance while reducing embodied carbon. Introduction Whereas concrete represents the most widely used construction material globally [], its eco-efficiency is currently challenged by the high CO2 emissions associated with cement production, which accounts for the majority of concrete’s embodied carbon (Fig. ). At the same time, the increasing demand for durable and resilient infrastructure further challenges conventional concrete, whose long-term performance under aggressive environmental conditions often leads to premature deterioration and significant maintenance burdens []. In this context, while ultra-high-performance concrete’s (UHPC) superior mechanical properties and durability [] are key for extending infrastructure service life and reducing maintenance, UHPC typically relies on cement-intensive formulations (in the range of 800–1 000 kg/m3) []. As a result, from a formulation standpoint, the environmental burden associated with UHPC can exceed that of conventional concrete, creating a challenging trade-off between performance and eco-efficiency [, ]. Achieving low-carbon UHPC (LC-UHPC) is inherently a multi-variable design problem that extends beyond simple cement replacement and is controlled by the types and replacement levels of supplementary cementitious materials (SCMs). The incorporation of SCMs—such as silica fume (SF), fly ash (FA), and metakaolin (MK)—provides an effective means to reduce clinker content through a combination of filler and pozzolanic effects, thus contributing to microstructural refinement and improved mechanical and durability performance []. In parallel, the use of inert mineral fillers alongside the implementation of particle packing design models (PPDMs) is key to improving packing density, thereby reducing voids and improving material efficiency []. PPDMs have been widely used to guide UHPC mixture design by optimizing the granular skeleton of the matrix []; although, the predictive accuracy of PPDMs may be influenced by their underlying assumptions, such as particle shape, interparticle interactions, and the influence of the liquid phase []. While the literature is replete with UHPC reviews on various individual aspects of UHPC, most of existing review studies predominantly focus on either raw materials, mechanical and durability properties, or mixture design approaches, and embodied carbon in isolation. For instance, recent review studies have focused on the influence of SCMs on the durability of UHPC [], proposed technical–economic-environmental roadmaps without quantitative analysis linking them to UHPC’s key compositional and performance indicators [], or applied a life-cycle assessment framework for UHPC decontextualized from UHPC’s controlling compositional and performance indicators []. Data-driven and machine learning approaches have been widely applied to predict the mechanical properties of conventional concrete [, ], and more recently to UHPC, where recent work has reviewed machine learning applications for predicting UHPC workability, mechanical, and thermal properties []. However, these efforts remain focused on algorithm development and prediction accuracy, with less attention paid to linking mixture design to UHPC’s key compositional and performance indicators. Therefore, limited attention has been paid to integrating raw material selection, particle packing optimization, mechanical and durability performance, and environmental impact within a single, data-driven framework. This study presents an integrated, data-driven review of LC-UHPC by linking input from mixture design, particle packing models, mechanical and durability properties, and carbon footprint assessment. First, the roles of cement, SCMs, and mineral fillers are reviewed with emphasis placed on how their physicochemical characteristics influence the end product as well as on their sustainability benefits. Second, particle packing design models are examined, and their applicability to UHPC mixture design is evaluated. Third, the influence of raw material characteristics and mixture proportions on key properties (including flowability, mechanical performance, and durability) is synthesized and analyzed. Finally, a database comprising around 200 UHPC mixtures was developed, and correlation heatmap analysis and Random Forest modeling were conducted to capture novel insights into the relationships between mixture composition, mechanical and durability performance, and carbon footprint. In an attempt to foster comprehensiveness, this review is based on peer-reviewed studies on UHPC materials, mixture design, and sustainability aspects. Emphasis was placed on publications reporting quantitative experimental data and embodied carbon assessments. Raw materials UHPC derives its unique mechanical performance and durability from a carefully engineered combination of cementitious materials and mineral fillers. Unlike conventional concrete, UHPC mixture design is governed by the synergistic interaction between material reactivity, particle size distribution, and packing density. This section examines the key constituents of UHPC with emphasis on their physicochemical characteristics and implications for low-carbon mixture design. Cement Representing the primary binding phase, cement governs early hydration kinetics and contributes significantly to strength development. As it is well established, cement consists mainly of tricalcium silicate (C3S, ~ 55%), dicalcium silicate (C2S, ~ 20%), tricalcium aluminate (C3A, ~ 10%), and tetra calcium alumino-ferrite (C4AF, ~ 8%) [], with CaO and SiO2 being the dominant oxides. Compared to the rest of the granular materials in UHPC, cement has a relatively high density (~ 3.15 g/cm3). Its mean particle size can vary depending on the cement type and ranges between 10 and 20 μm [–]. Various cement types have been reported for use in UHPC mixtures depending on the targeted application requirements and desired rheological and early-age mechanical performance. High early strength (HE) featuring superior fineness and faster hydration is commonly used to enable rapid strength gain []. Alternatively, general-use (GU) cement may be used when early strength is less critical [], while sulfate-resistant (HS) cement is selected for aggressive conditions due to its low C3A content, which also fosters controlled rheological properties []. Early UHPC formulations used to be cement-intensive, with mixture designs frequently containing 800–1 000 kg/m3 of cement []. While such high cement contents contribute to compressive strengths exceeding 150 MPa, they also result in significantly high levels of embodied CO2. This challenge has prompted the incorporation of SCMs to partially replace cement, while contributing to microstructural refinement and improved mechanical performance and durability. SCMs Through a combination of pozzolanic and hydraulic activity, SCMs contribute to the formation of additional calcium silicate hydrate (C–S–H), refine pore structure, and enhance the overall durability by impeding transport of aggressive ions such as chlorides and sulfates. At the same time, their partial replacement of cement reduces clinker content and lowers embodied CO2. As illustrated in Fig. , a wide range of SCMs has been explored in UHPC formulations. These materials can be broadly categorized into: (i) conventional SCMs, which are well-established in UHPC practice (e.g., SF, FA, ground granulated blast-furnace slag (GGBFS), and MK); and (ii) alternative emerging SCMs (e.g., ground glass pozzolans (GGPs), biochar (BC), rice husk ash (RHA), and recycled coral waste powder (RCWP)). SF SF is a by-product recovered during the production of alloys containing silicon or elemental silicon [] and is produced in the high-temperature reduction of quartz in electric arc furnaces, where the primary products are silicon or ferrosilicon alloys []. SF is composed predominantly of SiO2 (85%–98%) and also contains minor amounts of Al2O3, Fe2O3, CaO, MgO, K2O, and SO3 [–]. It has a remarkably high specific surface area (202 000–239 560 cm2/g), a density of around 2.3 g/cm3, a mean particle size ranging from 0.22 to 0.35 μm, and a specific gravity of around 2.14 [, , ]. SF is the most widely used SCM in UHPC, with common dosages ranging from 5 to 25% [–, ], due to its ultrafine particle size (submicron scale) and high amorphous SiO2 content (around 85%–95%, depending on the source) [, ]. Literally, almost all UHPC formulations contain SF for its extremely high specific surface area enabling it to fill nanoscale voids, thereby significantly enhancing packing density. In addition, its high pozzolanic reactivity promotes rapid consumption of calcium hydroxide, leading to the formation of secondary C–S–H and a highly densified microstructure []. This combined filler and pozzolanic effect makes SF effective in enhancing compressive strength, densifying interfacial transition zone, and improving the impermeability []. However, its high surface area also increases water demand and can adversely affect workability at low water-to-binder ratios, necessitating the use of an increased amount of high-range water-reducing admixtures, which may interfere with the early-age strength development. SF content in UHPC also correlates with autogenous shrinkage triggered by SF’s pore refining effect, which leads to higher surface tension, self-desiccation, and internal volumetric instability. Furthermore, the high cost and limited availability of SF may restrict its use at high replacement levels []. FA FA is a byproduct of coal combustion and has two primary classifications, Class F and Class C, with Class F containing less calcium compared to Class C []. Depending on the source of coal and combustion conditions, the chemical composition can vary. Typically, Class C has a CaO content (mass fraction) of approximately 13.0%, along with notable amounts of Fe2O3 (~ 59.0%), Al2O3 (~ 18.0%), and SiO2 (~ 5.0%), while K2O, TiO2, Mn2O3, and SO3 are generally below 1.0% []. In contrast, Class F is predominantly composed of SiO2 (44.0%–67.0%), followed by Al2O3 (22.0%–30.0%), Fe2O3 (1.0%–14.0%), and CaO (0.4%–7.0%) [–]. FA has a density of approximately 2.4 g/cm3, a Blaine surface area of around 3 848 cm2/g, a specific gravity of 2.38, and a mean particle size (can vary depending on the source) of around 10 μm [, ]. The inclusion of FA in UHPC has been reported to impart beneficial effects on both fresh and hardened properties. FA is commonly used to replace cement at levels ranging from 10% to 60% [, ]. Its spherical particles improve workability and packing density, while its pozzolanic properties enhance long-term strength and durability through the formation of secondary C–S–H []. However, with the global shift toward cleaner energy sources and the gradual phaseout of coal-fired power plants, which is the primary source of FA, the availability of FA has sharply declined in many regions, particularly in North America []. GGBFS GGBFS is a byproduct of the steel production industry. It is derived from the blast furnace process, where iron ore, coke, and limestone are melted, forming molten slag. This slag is then separated, rapidly cooled by granulation, and finely ground to produce GGBFS []. The physicochemical properties of GGBFS vary depending on the source and processing techniques. It is highly reactive due to its fineness, as it can go up to a specific surface area of 23 100 cm2/g [], with a common range of 4 250–4 700 cm2/g []. It has a relatively high density of 2.89 g/cm3 and a mean particle size ranging from 11 to 32 μm [–]. GGBFS is primarily composed of CaO (38%–44%), SiO2 (34%–37%), Al2O3 (10%–13%), and MgO (8%–10%) [, ]. GGBFS enhances UHPC performance through its latent hydraulic reactivity, which promotes additional C–S–H formation and improves long-term strength, durability, and chloride resistance []. GGBFS can be used as a partial cement replacement at dosages ranging from 30 to 50% [, –], as it generally improves workability and reduces autogenous shrinkage. However, with high replacement levels or with ultrafine GGBFS, the improvement may diminish or even reduce flowability due to increased surface area and more irregular and rough-textured particles []. Its contribution to pore refinement and reduced heat of hydration further supports the production of dense and durable UHPC []. From a sustainability perspective, GGBFS's use as a byproduct of the steel industry provides a significant reduction in embodied CO2 []. MK and nano-metakaolin (NMK) MK is produced by calcining kaolinitic clay, one of the most abundant clays on Earth, at temperatures between 700 °C and 850 °C [–]. A more refined form, NMK, is developed through high-energy milling of raw kaolin [, ]. Both forms of MK offer significant pozzolanic activity, making them valuable SCMs in UHPC production. Typically, MK is incorporated as a partial cement replacement at levels ranging from 5% to 20% [, , , ], whereas NMK is primarily used as a nano-additive at around 1% [, ]. The physicochemical properties of MK vary depending on the source and processing techniques, but it primarily consists of SiO2 (53.00%–74.30%) and Al2O3 (10.20–43.00%), with minor traces of Fe2O3 (0.38%–1.82%), CaO (0.40–2.40%), SO3 (~ 0.50%), MgO (0–0.33%), and K2O (0.10%–4.60%) [–, , , ]. The composition of NMK is similar but with a significantly higher SiO2 content (~ 89.6%), along with Al2O3 (~ 0.9%), Fe2O3 (~ 2.0%), CaO (~ 0.4%), MgO (~ 2.0%), and K2O (~ 4.6%) [, ]. MK has a density of approximately 3.58 g/cm3, a mean particle size of around 11 μm, a specific gravity of 2.50–2.57, and a specific surface area of roughly 200 000 cm2/g [, , , ]. GGPs GGPs are produced by grinding waste glass into fine powder, which can be processed to micro and nano levels to be used as a partial cement replacement, SF, or quartz sand [, ]. Improper disposal of waste glass remains an environmental concern, as glass does not naturally decompose. In this context, GGP offers a sustainable alternative for waste management while serving as a valuable SCM []. The physicochemical properties of GGP vary depending on the source and processing techniques, but it is generally composed of SiO2 (66.30%–89.40%), CaO (3.20%–12.20%), and Na2O (0.40%–14.90%), with traces of Al2O3 (0.20%–3.70%), Fe2O3 (0.04%–1.80%), SO3 (0.02%–0.30%), MgO (0.50%–2.90%), TiO2 (0–0.04%), and K2O (0.01%–0.70%) [–]. The specific gravity of GGP ranges from 2.52 to 2.63 [, , , , ]. Its specific surface area varies significantly based on grinding levels, reaching up to 130 500 cm2/g at the nanoscale and between 3 300 and 5 885 cm2/g at the micro-scale. The mean particle size of GGP ranges from 316 nm (0.316 μm), associated with nanoscale GGP used as a slurry, to 17 μm. This reflects the significant differences in processing intensity and application requirements [–, , ]. Due to its amorphous structure and high silica content [], GGP exhibits high pozzolanic activity, making it a viable partial cement replacement. Depending on the particle size, GGP can be used to replace cement, SF, or quartz sand at levels ranging from 10% to 50% [–, , ]. Its incorporation in UHPC improves packing density, reduces porosity, and contributes to long-term strength and durability []. However, the relatively high Na2O content in some glass sources has raised concerns about potential alkali-silica reaction (ASR) []. However, recent studies have shown that when waste glass is finely ground (typically < 300 μm), it does not promote ASR and can even mitigate expansion by reacting with calcium hydroxide to form additional C–S–H gel, thereby lowering pore solution alkalinity and enhancing durability []. BC BC is a highly carbonaceous by-product produced by the pyrolysis of biomass, which includes renewable plant-based materials such as wood and crop residues, under oxygen-limited conditions at 300–550 °C. The decomposition of cellulose, hemicellulose, and lignin during this process yields a stable, porous material with a high carbon content [–]. In addition to its carbon sequestration potential, which can offset up to 870 kg of CO2-equivalent per ton of biomass [], BC provides valuable benefits for cementitious composites, including improved water retention and internal curing [, ]. The composition (mass fraction) of BC varies based on feedstock and production conditions, typically containing 50.0%–93.0% carbon [], along with oxygen (16.0%–26.0%), hydrogen (~ 2.5%), nitrogen (~ 1.2%), and trace amounts (< 0.5%) of silica, calcium, aluminum, magnesium, and potassium [, , ]. Its particle size can range from 10 to 270 μm, with finer particles resulting from intensive grinding and milling [, ]. The porous nature of BC, which is formed during the decomposition of organic substances, contributes to its ability to enhance the hydration process. Its density varies significantly, ranging from 0.001 61 to 1.560 00 g/cm3, depending on its processing and feedstock characteristics [, ]. In UHPC, BC is typically used as a cement replacement at relatively low dosages (0.25%–2.00%), primarily due to its high absorption capacity and influence on rheological behavior [, , , ]. In addition to enhancing internal curing, BC contributes to improved UHPC performance by promoting hydration and refining the microstructure, which reduces porosity and mitigates autogenous shrinkage []. Its high carbon content and porous structure also improve UHPC's workability and reduce the mixture’s carbon footprint []. RHA Rice husk is the protective outer layer of the rice kernel and is an abundant agricultural byproduct, with approximately 200 kg of rice husk generated per ton of rice. Upon combustion, risk husk yields nearly 18%–20% RHA [, ]. With global rice production reaching 723 million tons in 2011, over 145 million tons of husks were generated, most of which were discarded, exacerbating solid waste stock []. Due to its unique composition, RHA has great potential to serve as an SCM and partially replace cement in UHPC. Depending on the production conditions, RHA can have a specific gravity of around 2.16–2.28, a mean particle size of approximately 7.5 μm, and a specific surface area of around 15 991 cm2/g [, ]. Chemically, it is primarily composed of SiO2 (85%–95%), K2O (2%–6%) and traces of Al2O3, Fe2O3, and CaO [–]. Its high silica content enables it to promote the hydration process through high pozzolanic reactions, which improve the performance of the cementitious system. In UHPC, RHA enhances mechanical strength and durability by promoting additional C–S–H formation and refining the microstructure []. Its fine particles improve packing density and contribute to pore refinement, while partial cement replacement reduces the mixture’s carbon footprint []. RHA has been used as a partial cement replacement at levels of 10%–50%, while its complete replacement of SF has also been reported in some UHPC formulations [, ]. RCWP RCWP is a byproduct of marine construction activities and is abundant in tropical coastal regions. Composed primarily of aragonite, calcite, dolomite, and high-Mg calcite, coral has a naturally porous and absorbent structure [, ]. Large quantities of coral waste are often landfilled, posing great ecological risks, but its reuse as an SCM offers an environmentally responsible alternative, reducing cement demand and mitigating impacts from marine construction activities. Chemically, RCWP is close to cement, with Ca (47.70%–52.70%) as the dominant component, along with minor amounts of SiO2 (0.40%–1.70%), Al2O3 (0.10%–0.40%), Fe2O3 (0.04%–0.40%), SO3 (0.60%–0.70%), Na2O (0.60%–2.00%), MgO (0.20%–2.50%), and K2O (0–0.10%) [–]. Approximately, it has a mean particle size of around 9.0 μm and a specific gravity of around 2.48. In UHPC, RCWP has been used as a partial replacement of cement, SF, and quartz sand, with reported replacement levels ranging from 5% to 60% for cement, 15%–45% for SF, and 10%–30% for quartz sand []. Due to its composition and fine particle size, RCWP improves UHPC's performance by serving as a filler and also participating in the secondary hydration process. Role in UHPC design The aforementioned SCMs contribute differently to UHPC performance depending on their reactivity, particle morphology, and influence on particle packing and pore refinement. Ultrafine materials such as SF primarily densify the matrix, enhance the packing density, and improve early-age strength through both filler and pozzolanic effects, in addition to serving as nucleation sites. Moderately reactive SCMs such as FA and GGBFS contribute to long-term hydration, enhanced durability, and improved workability. Meanwhile, materials such as GGP, RHA, and BC have demonstrated great potential for producing LC-UHPC due to their favorable pozzolanic activity, microstructural refinement, and reduced carbon footprint. From a mixture design perspective, the effectiveness of SCMs extends beyond their chemical reactivity. Their particle size distribution and morphology directly influence packing density, which impacts the porosity and mechanical performance. Consequently, the selection and proportioning of SCMs must be considered within a particle packing framework, where multi-scale particle distributions are optimized to achieve maximum density and efficiency. From the carbon footprint side, SCMs incorporation is the main approach to reduce the clinker content and the elevated CO2 emissions in UHPC. However, achieving LC-UHPC requires balancing cement reduction with performance requirements. Table summarizes the key physical properties of the SCMs discussed, including specific gravity, particle size, and surface area, providing a comparative basis for their selection and optimization in UHPC mixture design. These properties are particularly critical for evaluating their contribution to particle packing density and overall material efficiency. However, reported optimal replacement levels for the same SCM vary across studies, reflecting differences in cement types, water-to-binder ratios, and curing regimes, which complicates direct comparison between sources. Furthermore, emerging SCMs such as BC, RHA, and RCWP are characterized far less consistently than conventional SCMs, with variable reporting of particle size distribution and chemical composition across studies. Most existing research also evaluates SCMs individually, and the synergistic effects of combining fast- and slow-reacting SCMs within a single mixture remain comparatively underexplored, representing a gap that limits the optimization of multi-SCM UHPC systems. Mineral fillers Unlike SCMs, which contribute chemically through pozzolanic reactions, mineral fillers are physically active components through filler and nucleation effects. Their incorporation enables the reduction of voids across multiple length scales, thereby improving mechanical performance and durability. As illustrated in Fig. , a wide range of mineral fillers has been investigated for UHPC applications, including limestone powder (LP), recycled concrete fines (RCFs), ceramic waste powder (CWP), granite powder (GRP), red mud (RM), and iron ore tailings (IOTs). These materials vary significantly in terms of particle size, morphology, and chemical composition, which in turn control their effectiveness in improving the packing density and hydration reactions. Mineral fillers in UHPC can be broadly classified into: (i) inert or weakly reactive fillers, which primarily improve packing density (e.g., LP, GRP); and (ii) partially reactive fillers, which combine filler effects with limited pozzolanic or hydraulic activity (e.g., RCFs, CWP, RM). LP LP is produced by crushing and grinding natural limestone, which mainly consists of skeletal fragments of organisms and mineral phases. Limestone forms in diverse environments, including marine, lacustrine, and evaporitic settings, or through chemical precipitation of calcite or aragonite. Its structure varies from crystalline and clastic to granular, depending on its formation process []. LP is one of the most widely used fillers in UHPC due to its availability, low cost, and favorable particle size. It is commonly used as a partial replacement of cement in the range of 10%, and it can reach an extreme level of 80% (accompanied by a drastic drop in compressive strength) [, , ]. Its relatively fine particles enable efficient filling of interstitial voids between cement and SCM particles, thereby enhancing packing density and reducing porosity. Although largely inert, limestone can participate in limited chemical interactions. The physicochemical properties of LP vary depending on the source and grinding levels. In general, its physical properties include a density of 2.68–2.72 g/cm3 [, ], and a mean particle size between 0.70 and 20.00 μm []. Chemically, LP is dominated by CaO, exceeding 80.0% of its composition, along with traces of SiO2 (2.0%–4.4%) and Fe2O3 (0.8%–1.6%). In cases with higher impurity levels, the mass fraction of CaO may decrease to about 50.0% [, ]. RCFs RCFs are obtained from the crushing and screening of concrete debris generated during construction, maintenance, or demolition of concrete structures [, ]. The construction sector is among the world’s largest waste producers; in Canada alone, over 4 million tons of construction materials were landfilled in 2020, accounting for approximately 1.8 million tons of embodied carbon []. Advances in crushing and sorting technologies have enhanced the recycling potential of this waste, enabling the production of fine recycled aggregates and promoting more sustainable construction materials []. RCFs primarily consist of SiO2 (~ 41.1%) and CaO (~ 33.4%), with smaller amounts of Al2O3 (~ 11.8%) and Fe2O3 (~ 4.0%) []. Particle sizes of RCFs typically range from 0.063 to 2.360 mm, depending on the process and degree of grinding [, , ]. RCFs exhibit considerable variation in density, with a bulk density around 1.37 g/cm3 and apparent density ranging from 2.28 to 2.33 g/cm3 [, ]. This variation is due to differences in sources and compositions of the demolished concrete. The high CaO and SiO2 contents govern RCF’s reactivity and make it suitable for incorporation in UHPC. Depending on the application and particle size of RCF, it can be used as a partial replacement for quartz sand, natural sand, mineral powders, and basalt aggregates. Reported replacement levels include approximately 11%–33% for mineral powder, 20%–100% for natural sand, 10%–30% for quartz sand, and 50%–100% for basalt aggregate [, –]. RCFs enhance particle packing and contributes to microstructural densification []. Its residual cementitious phases can participate in secondary hydration, improving strength and durability while reducing clinker demand. Thus, integrating RCFs into UHPC not only diverts construction waste away from landfills but also supports a lower-carbon and more circular construction industry. CWP CWP is obtained by crushing and grinding waste from the manufacturing and disposal of ceramic products. The high-temperature processing of ceramics (200–2 000 °C) imparts pozzolanic reactivity, contributing to long-term strength and durability [, ]. While primarily used as a filler in concrete, CWP’s pozzolanic activity increases with finer particle sizes and higher specific surface area, enhancing hydration at later stages []. CWP is mainly composed of SiO2 (61.70%–78.30%) and Al2O3 (15.90%–22.30%), with minor amounts of CaO (0.90%–6.67%), Fe2O3 (1.10%–4.32%), MgO (0.70%–3.50%), K2O (1.60%–2.10%), and Na2O (1.00%–3.12%) [, , –]. Its specific gravity ranges from 2.40 to 2.61, water absorption rate of 2% [], and its specific density ranges from 2.43 to 2.56 g/cm3 [, ]. Its specific surface area is between 4 580 and 5 790 cm2/g [, ], and the mean particle size is approximately 3.5 μm [], but it can vary depending on the grinding and screening level. In UHPC, CWP can partially replace cement or quartz sand, typically at levels of 10%–55% and 50%–100%, respectively [, , , ]. It can enhance particle packing and densify the microstructure []. Its semi-pozzolanic properties contribute to additional C–S–H formation, improving long-term strength and durability []. By utilizing CWP as a supplementary material, UHPC mixtures benefit from reduced clinker content and a lower carbon footprint, without compromising mechanical performance. GRP GRP is an industrial byproduct from cutting granite blocks or grinding raw granite materials []. While primarily used as a filler in concrete, GRP exhibits pozzolanic activity, which increases with finer particle sizes due to the larger surface area available for reaction with calcium hydroxide. This enhances long-term concrete performance by promoting additional hydration products. GRP consists mainly of SiO2 (70.0%–83.0%), Al2O3 (2.0%–15.0%), and Fe2O3 (1.0%–14.8%). It has small amounts of CaO (0.4%–2.6%), MgO (0.2%–0.9%), SO3 (0.1%–0.3%), K2O (4.6%–5.8%), and Na2O (2.9%–3.9%). It has an approximate specific gravity of around 2.68 and a surface area of around 5 735 cm2/g (depending on the level of grinding) [, –]. In UHPC, GRP is used as a partial replacement for cement and quartz sand at replacement levels of approximately 20%–26% and 35%–100%, respectively. Due to its unique composition, GRP primarily contributes through filler and packing effects, improving matrix compactness and reducing void content, which supports enhanced mechanical performance and durability. It can also improve secondary hydration due to its rich composition of SiO2 [, , ]. RM RM is a highly alkaline byproduct generated during alumina production, with 1 to 1.5 tons of RM produced per ton of alumina [–]. Its large volume and chemical reactivity have drawn attention for sustainable construction applications, particularly as a partial cement replacement due to its pozzolanic activity []. The chemical composition of RM varies with source but is dominated by Fe2O3 (30.0%–50.0%), Al2O3 (18.0%–22.0%), and SiO2 (6.0%–20.0%). It also contains CaO (4.5%–13.0%), Na2O (6.0%–13.6%), TiO2 (3.9%–5.2%), and trace amounts of MgO, K2O, and SO3 (< 1%) [, , ]. In some cases, the mass fraction of Al2O3 may exceed that of Fe2O3, reaching up to 41.0% [, ]. RM has a specific density of approximately 2.97 g/cm3, a surface area of around 1 268 000 cm2/g (depending on the source) [], and a mean particle size of around 5.47 μm [], with its fine particles and high surface area enhancing its reactivity. In UHPC, RM is commonly incorporated as a partial cement replacement at levels ranging from 2.5% to 60% [, ]. Its fine particle size and high surface area enhance particle packing and microstructural densification []. Its pozzolanic activity contributes to additional C–S–H formation, improving long-term strength and durability []. Moreover, the high iron content of RM makes it an attractive option for special types of concrete, especially developing radiation-shielding UHPC. IOTs IOTs are mineral waste byproducts generated during the beneficiation process of iron ore. These tailings are typically stored in stacks, with limited utilization options []. In China alone, nearly 600 million tons are produced annually, representing over 50% of the global generation []. IOT properties might change depending on the source, but it is primarily composed of SiO2 (52.00%–75.00%), Fe2O3 (6.00%–9.00%), Al2O3 (3.50%–17.00%), CaO (4.00%–12.00%), SO3 (0.05%–10.60%), and MgO (3.00%–6.00%) [–]. It has a surface area ranging from 439 to 1 185 cm2/g, a mean particle size of 166.2 μm [, ], and a density between 2.62 and 2.70 g/cm3 [, , ]. In UHPC, IOT can partially replace cement and fine aggregates. Reported replacement levels range from 10% to 30% for cement and 20%–100% for quartz sand [, –]. While IOT exhibits low early-age pozzolanic reactivity, its fine particles act as nucleation sites and contribute a filling effect that densifies the microstructure, which supports long-term strength development and improved durability []. Role in UHPC design Mineral fillers are essential for achieving the dense particle packing that drives the great strength of UHPC. By filling voids across multiple size scales, they reduce porosity, enhance strength, and improve durability, while for some fillers, also enabling reductions in binder content. From a particle packing perspective, mineral fillers complement SCMs by occupying intermediate and fine particle size ranges, thereby facilitating continuous particle size distributions. This multi-scale packing approach is crucial for minimizing voids and optimizing material usage. From a sustainability standpoint, mineral fillers contribute to LC-UHPC by possibly reducing cement and valorizing industrial waste. However, their effectiveness depends on achieving an optimal balance between packing efficiency and reactivity. Table summarizes the key physical properties of the mineral fillers discussed, providing a comparative basis for their selection and integration into UHPC mixture design. Some fillers, such as GRP and CWP, exhibit semi-pozzolanic behavior, which obscures the conventional distinction between inert fillers and reactive SCMs and results in inconsistent classification across the literature. A further limitation is that many waste-derived fillers, including RCF, RM, and IOT, show considerable compositional variability depending on their source and processing history, which limits the generalizability of reported optimal replacement levels across studies and hinders their widespread use in standardized mixture design guidelines. PPDMs Traditional concrete mixture design approaches (Fig. ) are largely empirical and were developed for conventional concrete systems with relatively coarse particle distributions. In contrast, UHPC relies on densely packed, multi-scale granular systems, where performance is governed by particle size distribution (PSD), packing density, and interparticle interactions. As such, PPDMs have become essential tools for UHPC mixture optimization, enabling the reduction of voids, enhancement of mechanical performance, and improvement of durability []. Packing density (∅P), defined as the ratio of solid volume (Vsolids) to total volume (Vtotal), represents a fundamental parameter governing UHPC performance (Eq. ()) []. Optimized packing density reduces porosity, lowers water demand, and enables more efficient use of cementitious materials, an aspect that is directly relevant to the development of LC-UHPC. PPDMs are classified into discrete and continuous models. Discrete models describe particle interactions between individual size classes, whereas continuous models represent the whole spectrum of particle sizes using mathematical functions. While both approaches provide insight into packing behavior, their applicability to UHPC differs significantly. Discrete models Discrete packing models, including those proposed by Furnas, Toufar, and De Larrard, are based on the interaction between particles of different sizes. These models provide an understanding of the packing phenomena, particularly the role of fine particles in filling voids between coarser particles. Furnas model The Furnas model is one of the earliest binary packing models, describing systems composed of coarse and fine particles [, ]. It is based on the principle that fine particles can fill the voids between coarse particles, thereby increasing the overall packing density. Two main scenarios are considered in this model (Fig. ): (I) the volume fraction of the fine particles (VFF) is much lower than the volume fraction of the coarse particles (VFC), and (II) the dominant volume fraction is the fine particle fraction (VFF) [, ]. In scenario I, the partial volume of the coarse particles (VC) corresponds to the residual packing density of the coarse particles (∅PC) because the coarse particles rearrange to achieve their maximum packing state, creating voids that can accommodate the fine particles without displacing the coarse particles. Therefore, the packing density of the system (∅P1) is the sum of the maximum packing density of the coarse particles (∅PC) and the partial volume of the fine particles (VF), as shown in Eq. () [, ]. In scenario II, coarser particles are added to a system where fine particles are dominant. This means that the partial volume of the coarse particles (VC) will equal their impact on the system’s packing density (∅P2). The fine particles will fill the remaining volume with their maximum packing density (∅PF), as illustrated in Eq. () [, ]. While the model provides a useful conceptual framework, it requires a significant size difference between particle classes to achieve optimal packing. When this condition is not satisfied, several phenomena, namely the loosening effect, wall effect, and wedging effect (Fig. ) [], can reduce packing efficiency. These effects are particularly relevant in UHPC, where particle sizes often overlap across multiple scales. The loosening effect occurs when fine particles are larger than the voids between coarse particles, thus disturbing and displacing the coarse particles, increasing the system’s porosity. The wall effect prevails when fine particles are trapped between the coarse particles, leading to increased porosity. The wedging effect occurs when fine particles are wedged between coarse particles, increasing the matrix porosity. Furnas further extended the model to multi-modal cases (Eq. ()). In this context, monodisperse materials with a discrete distribution represent only a fraction of the multi-modal system, which forms a geometric progression similar to a continuous distribution encompassing all particle sizes []. where CPFT is the cumulative percentage finer than the particle diameter (DP); DS is the smallest particle diameter in the system; DL is the largest particle diameter in the system; and r is the ratio between the retained volume of particle diameter (DP) and the next smaller sieve (DP-1). Although the Furnas model can be extended to multi-component systems, its discrete nature limits its applicability to complex UHPC mixtures with continuous PSDs. Toufar model The Toufar model [] extends the Furnas approach by accounting for the loosening and wall effects when the fine (DF) and coarse (DC) particles are close in size. The model calculates the packing density for mixtures with diameter ratios of 0.22