Particle-size Analysis

Examples of fractionated samples in each sieve are shown in Fig. 3. The particle-size distribution for sample D (Fig. 4) revealed a D10 (effective diameter) value of 0.2 mm, indicating that 10% of the particles were of a diameter of 0.2 mm, considered to be a medium size. The values D50 and D90 were 6 mm and 37 mm, respectively. Once again, fine particles predominate in this sample. When the data for sample E were examined (Fig. 4), D10 was 0.4 mm, D50 was 18 mm, and D90 was 40 mm. This indicates that only 10% of the sample consists of particles with sizes > 40 mm, highlighting a predominance of smaller-sized particles. In contrast, sample F (Fig. 4) exhibited the following values: 0.3 mm, 7.5 mm, and 35 mm for D10, D50, and D90, respectively. In this case, undersized particles make up the majority (50%) of the sample with only 10% at quantitative values (10% cumulative passing of samples). One concludes, therefore, that sample D contains particles with larger sizes than samples E and F, but all samples are predominantly composed of fine particles.

Fig. 3

Bauxite sample during grain-size distribution tests

Fig. 4

Particle-size distribution of the bauxite samples D, E, and F

The Rosin-Rammler-Bennet studies (RRB) indicated that the grinding time affects particle characteristics, necessitating adjustments to the processing parameters (Reference Zhang, Bai, Dong, Wu and YeZhang et al., 2014). On the other hand, Wrebec’s mathematical model evaluates particles of <50 μm, resulting directly from crushing (Reference Mohan, Kloss, Khinast and RadlMohan et al., 2014). When the grinding time is increased for particles ranging from 0–3 to 3–10 μm, no variation in the mechanical fraction and particle-size values was observed. Increasing the grinding time for particles within this range therefore, only leads to unnecessary energy consumption, as it does not affect the mechanical fraction (Reference Liu, Li, Xie, Li and XiaoLiu et al., 2016; Reference Mohan, Kloss, Khinast and RadlMohan et al., 2014).

Several mathematical models can be used to determine the particle-size phases D10, D50, and D90 from grinding, such as the RRB (Eq. 2) and Gates-Gaudin-Schuhman (GGS) (Eq. 3) models. k, m, n, and D’ are the parameters to be adjusted to fit the experimental data (Reference Wu, Chau and YuWu et al., 2004). Of these two models, the RRB function performs better, particularly when a ball mill is used for grinding the samples. Combining multiple functions, however, is generally more effective than relying solely on a single mathematical model (Reference Chu, Nguyen, Vu, Dao, Dinh, Nguyen, Hoang, Le and PhamChu et al., 2019). So, consider: X = mass fraction of material finer than the sieve opening (kg/kg total); D’ = parameter representing the average particle size (μm); n = parameter representing dispersion (dimensionless); D = the opening of the sieve of order n (m); k = parameter representing the average particle size (μm); m = parameter representing dispersion (dimensionless), also called Schuhmann’s derivative.

(2) $X=1-\exp \left({-\left(\frac{D}{D\prime }\right)}^n\right)$

(3) $X={\left(\frac{D}{k}\right)}^m$

Mathematical models are essential to obtain the statistical parameters used in industrial processes. The plots using the mathematical models of granulometric distribution RRB and GGS for bauxites D, E, and F are available in the supplementary material (Figs S1, S2, and S3). The coefficient of determination (R²) indicated that the RRB model (Table 1) provided a better fit. The value of m was < 1 in all samples (Table 1), thus showing that the material does not have a uniform particle size.

Table 1

Equations for the mathematical models

The Fanrenwald model proposes a correlation between the reduction rate of residue on the sieve and the particle size, highlighting the relationship between grinding speed and the accumulated percentage on the sieve after a certain grinding period x (Reference Chen, Wang, Liao and HuangChen et al., 2020). These two factors are directly proportional during grinding (Reference Liu, Li, Xie, Li and XiaoLiu et al., 2016). Grinding efficiency is correlated with particle size, meaning that as particle size decreases, grinding efficiency also decreases (Reference Liu, Li, Xie, Li and XiaoLiu et al., 2016; Reference Wang, Lucas, Milward and CooperWang et al., 2021). When analyzing the D10, D50, and D90 percentages, the accumulated volumetric distribution and width were evaluated, facilitating the manipulation of the samples. The grinding time and the kinetic energy provided during this process are associated with the particle-size distribution values (Reference Fabre, Buche, Rouau and Mayer-laigleFabre et al., 2020).

Moisture Content, Loss on Ignition, and Total Organic Carbon Analyzer

The samples exhibited relatively low moisture content levels (Table 2), with values <1.0%. Comparing the samples revealed that sample D had a smaller moisture content (0.5%) than samples E and F (0.9 and 0.8, respectively). These values are generally considered to be smaller than other bauxite deposits. It is important to note that moisture content of 4–12% can negatively impact the transport of this material (Reference Manjare and DonolikarManjare & Donolikar, 2022).

Table 2

Moisture content and loss on ignition values for the three samples

The mass loss on ignition in this test is related to the loss of structural water (Table 2). In the studies by Reference Liu and PoonLiu and Poon (2016), the LoI in rejected red mud (generated after the clarification stage in the Bayer Process) originating from China was 8.14%. In the research of Reference Gräfe, Power and KlauberGräfe et al. (2011), after several analyses of red mud samples, the values of LoI varied from 4.4 to 14%. In rejected red mud originating from Brazil, the LoI varied with temperature; at 900 and 1100°C the LoI was 14.5 and 14.8%, respectively (Reference Botelho, Crocce and TenórioBotelho et al., 2021). In the studies by Reference Botelho, Espinosa and TenórioBotelho et al. (2020), the values for inorganic carbon and total carbon were 0.32% and 0.6%, respectively, so organic carbon remained at ~0.28%. Reference Power and LohPower and Loh (2010) evaluated the organic carbon values in bauxite from Guinea, Brazil, and Suriname, where the percentages were 0.25%, 0.18%, and 0.47%, respectively.

The total organic carbon was measured in the organic and inorganic matter in the sample and none was found (Reference Banerjee, Mankar and KumarBanerjee et al., 2023). The presence of organic matter would lead to an increase in the production costs of the alumina because the organic matter is considered a contaminant in the process. All lateritic bauxites contain a certain amount of organic carbon compounds, some of which are solubilized during the digestion process. As a result, organic compounds accumulate in the process liquor (Reference Castaldi, Silvetti, Santona, Enzo and MelisCastaldi et al., 2008; Reference Power and LohPower & Loh, 2010; Reference SantiniSantini, 2015). An assessment of the impact of these compounds on production costs revealed that the annual expenses for the Australian alumina industry alone exceeded AUD$ 500 million where organic products were concerned (Reference Power, Loh and VernonPower et al., 2012). The Brazilian bauxites do not consume more reagent during the digestion stage.

X-ray Diffraction (XRD)

The mineralogical composition of the three bauxite samples was determined using XRD (Fig. 5), and the phases were calculated using the Rietveld method (Reference Castaldi, Silvetti, Santona, Enzo and MelisCastaldi et al., 2008; Reference SantiniSantini, 2015). The Rietveld method identified Gbs as the predominant phase in all samples (Table 3), comprising ~50–80% of the composition. Smaller amounts of Kln (~20–30%), Ant (3–6%), and Hem (5%) were also present. This is in accordance with the literature (Reference César, Melo, Simões, Patrícia and PazCésar et al., 2020; Reference SantiniSantini, 2015).

Fig. 5

XRD patterns of the bauxite samples

Table 3

Rietveld profile analysis

Sample E contained most Gbs (75%), which is desirable due to its greater potential to yield aluminum (Table 3). However, a large (20%) Kln content in this sample poses a challenge during the Bayer Process. The reactive silica found in Kln consumes a significant amount of calcium hydroxide and subsequently affects the extraction of aluminum hydroxide. Reactive silica values (SiO₂) in the the lithium metaborate digestion data were 4.35, 3.16, and 3.23 wt.% in samples D, E, and F, respectively. Bauxite compositions containing <8% reactive silica are noteworthy and recommended in order to minimize excessive reagent consumption (Reference Tabereaux and PetersonTabereaux & Peterson, 2014).

The Kln content of sample F is large at 26.29 wt.%, indicating the need to explore alternative methods to the Bayer process for recovering usable Al and removing impurities from this sample. In addition, sample D exhibited the greatest iron oxide content, ~5%. In comparison, sample E showed the removal of 3.9% titanium oxide during the Bayer process, resulting in the generated tailings known as red mud (Reference Ahmad, Hartge, Werther and WischnewskiAhmad et al., 2014). Another factor to consider is the presence of amorphous content, which may not be quantified accurately by the analytical techniques used in the present study. As a result, the calculated amount of crystalline content may be overestimated (Reference Alelweet and PaviaAlelweet & Pavia, 2022).

Rietveld quantitative analysis of bauxite from a mine in Russia revealed Gbs 55.2%, Kln 10%, titanium oxide 4.9%, and iron oxide 4.9% (Reference Abdulvaliyev, Dyussenova, Manapova, Akcil and BeisenbiyevaAbdulvaliyev et al., 2021). The XRD patterns of bauxite provided by Paragominas Mining (Pará, Brazil) revealed predominantly Gbs, with lesser amounts of Kln, Hem, Gt, and Ant (Reference Dodoo, Ellen, Seguwa, Yawson, Appiah, Suleiman and YayaDodoo et al., 2022). A mine in Ghana produced samples with 10 wt.% SiO₂, 50 wt.% Gbs, and 3.15 wt.% iron oxide, as well as ~2 wt.% oxides of Mg, Na, Ca, and K combined (Reference Melo, Angélica and PazMelo et al., 2020).

XRD patterns of samples from mines in central and northern Saudi Arabia revealed high crystallinity, with intense peaks from Gbs, Kln, and Bhm. Interestingly, those bauxites consisted of >50 wt.% boehmite and Gbs, with only 15 wt.% Kln along with other minor components of Ant, rutile (Rt), calcite (Cal), and Qz making up ~7 wt.% each (Reference Alelweet and PaviaAlelweet & Pavia, 2022). Bauxite from various provinces of China have a mineralogical composition of silicon oxide plus diaspore of ~58 wt.%, while iron oxide and titanium comprise 24.23 and 3.12 wt.%, respectively (Reference Wu, Lv, Xie, Li, Zhao and LiuWu et al., 2022b).

Thermogravimetric Analysis (TGA)

Thermogravimetric analysis (TGA/DTGA) (Fig. 6) revealed in the first stage, spanning 100 to 360°C, significant mass losses from samples D, E, and F, measuring 16.88, 11.27, and 8.50 wt.%, respectively. An intense endothermic peak at 300°C indicated the transformation of the Gbs into the first metastable alumina phase (χ-Al2O3). At 300°C, Gbs undergoes dehydroxylation and converts into boehmite, as described by the decomposition Eq. 4 (Reference Zhang, Cui, Hu, Wang, Prange, Wan, Jaegers, Zong, Zhang, Pearce, Li, Wang, Clark and RossoZhang et al., 2019).

Fig. 6

Thermogravimetric profiles of the bauxite: a TG and b DTG

(4) $AI{\left(O,H\right)}_3\to AIOOH+H_{2}O$

In the second stage, from 360 to 600°C, the mass losses were 6.79 wt.% for sample D, 6.74 wt.% for sample E, and 9.81 wt.% for sample F. Another endothermic event at ~515°C signified the formation of a second metastable phase of alumina (қ-Al₂O₃) (Reference Alelweet and PaviaAlelweet & Pavia, 2022; Reference Li, Tang, Jiang, Zhu, Zhou and XueLi et al., 2020a). With the loss of the hydroxyl groups from the samples containing Kln, the formation of amorphous metakaolinite occurred from 400 to 650°C (Reference Lagauche, Larmier, Jolimaitre and BartheletLagauche et al., 2017). In the third stage, ranging from 600 to 1200°C, the mass losses for D, E, and F were 1.17, 0.88, and 0.91 wt.%, respectively. During this stage, as the temperature increased from 900 to 1200°C, the amorphous alumina phases (ϒ-Al₂O₃) underwent regrouping to form the corundum phase α-Al₂O₃ (Reference Arogundade, Megat-yusoff, Ahmad, Bhat and AfolabiArogundade et al., 2021; Reference Kyriakogona, Giannopoulou and PaniasKyriakogona et al., 2017). This mass loss aligns with the LoI observed in the chemical analysis, attributed primarily to the release of hydroxyl groups present in the minerals. Relating the TGA test to the LoI test is possible.

Multi-acid Microwave Digestion and Lithium Metaborate Fusion

The significant elements analyzed, including Fe, Al, Si, and Ti, were subjected to microwave and fusion digestion methods (Table 4). Oxides were calculated based on the amounts of the individual elements obtained in the analyses.

Table 4

Data obtained from the chemical analyses in the lithium metaborate and multi-acid digestions in ICP-OES

The results revealed that aluminum oxide was the most abundant compound in the samples, followed by silicon oxide, with smaller concentrations of titanium and iron oxide. A comparison between the two digestion methodologies showed variations in the element contents, indicating differences between the analytical techniques used to determine the elements present. Microwave digestion was the more sensitive method in terms of iron and titanium oxides, suggesting that metaborate fusion is not a consistent technique for elements present in smaller quantities (Reference Giels, Hertel, Gijbels, Schroeyers and PontikesGiels et al., 2022).

Based on the assays, atomic absorption spectrometry using a microwave-based opening methodology was deemed the most suitable technique for chemical analysis, specifically for samples D, E, and F. The multi-acid digestion in HNO₃, HCl, H₂SO₄, and HF solutions was an effective procedure for evaluating many elements in various mineral species. The addition of phosphoric and sulfuric acid facilitated the extraction of aluminum, while nitric acid reacted with other metals present, such as titanium, magnesium, calcium, sodium, and potassium. Hydrofluoric acid reacted primarily with the silica content in the samples (Reference Dillinger, Batchelor, Katrib, Dodds, Suchicital, Kingman and ClarkDillinger et al., 2020).

Silica, when in the form of Kln, poses a problem in the Bayer process as it reacts with NaOH during bauxite digestion, forming insoluble compounds that reduce the profitability of the process. In contrast, when silica is in the form of quartz, it exhibits inert behavior and does not significantly impact the consumption of NaOH, referred to as non-reactive silica. It is crucial to work with bauxites containing <8% reactive silica in their composition to mitigate these issues (Reference Tabereaux and PetersonTabereaux & Peterson, 2014).

Approximately 80–90% of calcium hydroxide is lost during the process, posing a significant issue. Consequently, utilizing bauxites with <8 wt.% reactive silica content is essential to mitigate this problem (Reference Ahmad, Hartge, Werther and WischnewskiAhmad et al., 2014; Reference Smith and PowerSmith & Power, 2021; Reference Wu, Zhang, Li, Bi and YangWu et al., 2010).

Iron oxides in bauxite are also problematic in alumina production. Although the samples used in the experiments had values of <5 wt.% Fe₂O₃, worldwide samples, especially from China and Iran, tend to have larger iron oxide contents. The iron oxide impurities are removed during the clarification process, resulting in the red mud.

Titanium oxides are generally non-reactive but can hinder precipitation and reduce reaction yields when present as boehmite. Although the samples showed titanium oxide values of <1 wt.%, global bauxites often have values of >5 wt.%.

Comparing the composition of bauxites from various locations worldwide (Table 5), Australia, China, and Saudi Arabia typically have aluminum oxide values of ~50 wt.%. In comparison, the bauxites from Porto de Trombetas exhibit 40 wt.% Al₂O₃. Regarding the iron oxide content, bauxites from mines in Iran, China, and Greece have values ranging from 20 to 30 wt.% of Fe₂O₃, while Russia and Saudi Arabia show values of 5 wt.%. The samples obtained in the experiments align with these findings, with iron oxide amounts of 4 wt.%. Titanium oxide values in global bauxites are usually >5 wt.%, whereas the samples analyzed in this study had <1 wt.%. Silicon oxide measurements in bauxites worldwide range from 20 to 26 wt.%, similar to the values obtained in the chemical analyses of the three samples. Dealing with the silicon oxide content is crucial as it represents one of the main challenges in the Bayer process.

Table 5

Oxide percentages in bauxites from around the world