Applicability of carbothermic reduction for upgrading Sri Lankan ilmenite ores: towards converting ilmenite into synthetic rutile by mechanical activation

Ilmenite and rutile are naturally occurring titanium-bearing heavy minerals. Sri Lanka consists of ilmenite and rutile in placer deposits mainly along the northeast coast. The commercial value of rutile is higher than ilmenite. Therefore, the current study focuses to upgrade Sri Lankan ilmenite ores using commercially available activated carbon as a reducing agent. Ilmenite fraction was initially separated from raw beach sand using an industrial-scale magnetic separator (Wet high-intensity magnetic separator: sixteen pole model). The mixtures of ilmenite and activated carbon (4 to 1 ratio by weight) were milled using a planetary ball mill for one to four hours at one-hour intervals. The milled samples were subsequently heated for two hours continuously at temperatures of 800 °C, 900 °C, 1000 °C, 1100 °C, and 1200 °C, respectively. Initial and annealed samples were analysed using X-ray diffraction (XRD), Fourier-transform infrared (FTIR) spectroscopy, and X-ray fluorescence (XRF) analyses. XRD pattern of the initial sample was characterized by less percentage of rutile (TiO2) peaks and low crystallinity. However, the number of rutile peaks and crystallinity were increased with respective milling hours and annealed temperature. Besides, XRD analysis indicates a higher number of sharp and intense rutile and elemental iron peaks in the samples annealed above 1000 °C. FTIR analysis of raw ilmenite indicates mainly stretching vibrations of Fe–O. However, vibrational spectral bands around 700 cm−1 in samples annealed above 1000 °C are assigned to Ti–O–Ti stretching vibrations. High TiO2 and TiO2 + Fe2O3 (total) wt. (%) can also be observed in samples annealed above 1000 °C. The optimum conditions for carbothermic reduction were obtained at 4 h of ball milling of ilmenite with activated carbon and continuously annealing at 1000 °C for 2 h. Carbothermic reduction is applicable to upgrade Sri Lankan ilmenite ores, and thus this method adds value to heavy mineral resources in Sri Lanka.


Background
Ilmenite (FeTiO 3 ) (approximately, 40-60% of TiO 2 ), rutile (TiO 2 ) (~ 93-96% of TiO 2 ) and leucoxene (Fe 2 O 3 . nTiO 2 ) (~ over 65% of TiO 2 ) are the major ores of metal titanium and titanium dioxide (Wouterlood 1979;Ismail et al. 1983;Chen et al. 1997;Angusamy et al. 2005;Palliyaguru et al. 2017;Perks and Mudd 2019, 2020. Both ilmenite and rutile are usually mined from beach placer deposits. Titanium minerals characterise by properties such as exceptional scattering, chemical stability, and lack of toxicity (Adipuri et al. 2011). Titanium alloys are used in aerospace, biomedical, automotive and metallurgical industries, and titanium dioxide is used in paint, Wijewardhana and Ratnayake Bull Natl Res Cent (2021) 45:149 paper, pigments and plastic industries. Consequently, titanium minerals have become a valuable commodity (Adipuri et al. 2011;Perks and Mudd 2019). Because of that, a large number of high-quality titanium ores such as rutile are required for catering to the escalating global demands of both metal titanium and TiO 2 based products, especially in the Asia-Pacific region (Mackey 1994;Dooley 1975;Wang et al. 2008). In this case, the production of synthetic rutile (TiO 2 ) is gradually increasing due to a dearth of economically concentrated rutile deposits in the world. For example, ilmenite accounts for 94% of global titanium resources, while that of rutile is only about 4%, and the rest being supplied by minerals such as leucoxene Mudd 2020, 2021). Therefore, much attention has been taken to process low-grade titanium ores such as ilmenite (approximately US$ 130 per tonne in 2019) into synthetic rutile (approximately US$ 1150 per tonne in 2019) (Subasinghe et al. 2021). About 5% of the mined ilmenite is generally used to produce titanium metal while the rest is used in the production of titanium dioxide/synthetic rutile Mudd 2020, 2021). Consequently, readily available ilmenite gradually became an alternative raw material for synthetic rutile (Chen et al. 1997; Wang and Yuan 2006;Wang et al. 2008).
Several methods have been investigated in the literature to upgrade ilmenite into synthetic rutile such as smelting (Murty et al. 2007), acid leaching (Mahmoud et al. 2004), ion-exchange (Palliyaguru et al. 2017), sulphation (Lee and Sohn 1989;Nurdin et al. 2019), selective chlorination (Neurgaonkar et al. 1986;Rhee and Sohn 1990), reduction (Shiah 1978;Kahn 1984;Wijewardhana et al. 2021), and slagging (Wright et al. 1985). In addition, certain industrial-scale limitations and environmentally unfriendly waste generation (e.g., usage of corrosive and hazardous chemicals) make the existing processes expensive and problematic. For example, the sulphate process generates 3.5 tonnes of waste, while the chloride process generates only 0.2 tonnes of waste, per tonne of TiO 2 product (Mackey 1994). Although the chloride method limits waste generation, this method requires high titanium content in the feedstock to reduce the processing time (Tao et al. 2012). Despite these limitations and drawbacks, Sri Lankan ilmenite ores have not been the focus of value addition at an industrial scale using any of the methods stated in the literature (Subasinghe et al. 2021).
The current study thus focuses to apply a modified ball milling induced carbothermic reduction using commercially available activated carbon as an additive. In this study, the authors determine the optimum conditions of the reaction based on geochemical changes and crystallinity in milled and annealed samples using X-ray powder diffraction (XRD), Fourier transform infrared (FTIR), and X-ray fluorescence (XRF) analyses. The current study examines the value addition potential of Sri Lankan ilmenite at the laboratory scale. Consequently, this case study is a timely requirement to upgrade mineral sands in the country for gaining economic benefits. Moreover, determining the industrial viability of this methodology via a pilot plant would be a potential future work.

Study area
The coastal tract of Sri Lanka consists of several heavy mineral placer deposits (Geological Survey of Ceylon 1970;Herath 1980;Wickremeratne 1986;Subasinghe et al. 2021). For example, Pulmoddai deposit spreads nearly 8 km from Arisimale to Kokkilai along the northeast coast of Sri Lanka with an average thickness of 6 m over Precambrian crystalline rocks (Fig. 1). Pulmoddai beach has been commercially exploited since 1957 (Herath 1980). The Pulmoddai deposit records high-grade values of heavy minerals such as 70-72% of ilmenite, 8-10% of zircon, 8% of rutile, 1% of sillimanite, and 0.3% of monazite (Sri Lanka Minerals Year Book 2014). This deposit contains nearly 6 million tonnes of heavy mineral sands, and the annual partial replenishment during the northeast monsoon makes this deposit more valuable (Sri Lanka Minerals Year Book 2014; Amalan et al. 2018). In 2015, Sri Lanka exported nearly 39,000 tonnes of ilmenite, 1,800 tonnes of rutile, and 600 tonnes of zircon (Lanka Mineral Sands Ltd 2018; Subasinghe et al. 2021). Two heavy mineral deposits are located towards the north (i.e., Nilaveli to Mulativu beach placer deposits) and south (i.e., Verugal deposit) of the Trincomalee Bay (Fig. 1). These deposits in combination can be considered as one of the largest heavy mineral deposits in the world. The current estimations suggest the availability of about 12.5 million tonnes of unexploited mineral sands in these deposits (Sri Lanka Minerals Year Book 2014).

Materials
In this study, ilmenite samples were collected from Lanka Mineral Sands Limited, Pulmoddai. Raw beach sands were subjected to screening for the removal of roots, stones, and shells. Screened raw beach sands were separated using spiral separation (Mark-6 Reichert and Mark-3A Reichert Spirals) to remove quartz. Ilmenite was separated using a high-intensity magnetic separator (Wet high-intensity magnetic separator: sixteen pole model) based on its paramagnetic property. Therefore, industrial-scale separated ilmenite ores were used for the geochemical analysis described here. In addition, commercially available activated carbon was used as an Page 3 of 10 Wijewardhana and Ratnayake Bull Natl Res Cent (2021) 45:149 additive (reductant) which is composed of over 95% of carbon.

Sample preparation
In this study, 10 g of ilmenite sand and 2.5 g of commercially available activated carbon were weighed and mixed thoroughly for the optimum weight ratio (Shahien et al. 2015;Wijewardhana et al. 2021). After that, twenty powdered samples were grouped considering different milling hours and annealed temperatures (Table 1). Samples  were initially milled for 1 h, 2 h, 3 h, and 4 h, respectively. Mechanical attrition was performed in a vertical laboratory planetary ball mill (semi-circle model) under an airtight state equipped with a ball mill tank with 1-30 mm zirconium balls and 100 ml zirconium jars. The cell was loaded with mixed samples and milled at room temperature (27 °C) at 500 rpm. The milled samples from 1 to 4 h with 1-h intervals were continuously annealed at 800 °C, 900 °C, 1000 °C, 1100 °C, and 1200 °C, separately for 2 h using a muffle furnace as shown in Table 1.

Particle size analysis
The particle size distribution of the initial ilmenite and four hours milled samples were analysed using a British standard mechanical sieve shaker. The weight of the initial sample and the weights retained in each sieve after 30 min of shaking were measured for calculating d 50 value in the cumulative distribution.

X-ray diffraction (XRD) analysis
The milled and annealed powder samples were placed in the ground glass depression in the sample holder. Afterwards, samples were flattened using a glass slide to arrange a smooth and well-packed sample. Mineral phases were identified for 20 samples using Rigaku Ultima (IV) X-ray diffractometer with Cu Kα radiation. The scanning speed of 2θ (the diffraction angle of Bragg's Law) was 10.00 degree min −1 with the range of 0° to 90°. The accelerating voltage and applied currents were 40 kV and 30 mA, respectively.

Fourier-transform infrared (FTIR) spectroscopy
Finely powdered samples were mixed with KBr in 1:10, and pressed into pellets for transmission. FTIR grade dried KBr (assay ≥ 99%) was used as alkali halide. FTIR analysis was performed for 6 samples using a Bruker Alpha spectrophotometer to identify functional groups over the range of 500-4000 cm −1 . The instrument has a resolution of 4 cm −1 over 64 scans. All individual FTIR spectra were corrected against the spectrum of KBr pellet and the automatic baseline. The background corrections were also carried out for the absorption of atmospheric water and CO 2 .

X-ray fluorescence (XRF) analysis
Loss on ignition (LOI) values (includes H 2 O, CO 2 , S, and other volatiles) were first determined from the weight loss after heating samples at 1000 °C for 2 h. The heavy absorber fusion technique was used to minimize the matrix effects of samples. The fusion disks were prepared by mixing a 0.75 g of heated sample with 9.75 g of a combination of lithium metaborate and lithium tetraborate with lithium bromide as a releasing agent. Samples were fused in platinum crucibles. Samples were analysed on a Panalytical Axios Advanced wavelength dispersive XRF at Activation Laboratories, Ontario, Canada. The intensities were then measured, and the concentrations were calculated against the standard G-16 of the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia. In general, the limit of detection is about 0.01 wt. (%) for most of the major elements (oxides).

Particle size distribution
According to the United States Department of Agriculture (USDA) textural classification in 2012, d 50 particle size distribution of initial ilmenite sample and 4 h milled samples are 0.14 mm and 0.08 mm, respectively (Fig. 2).

Investigation of ball milling and isothermal annealing by XRD
X-ray diffractogram of the initial sample shows peaks of ilmenite (FeTiO 3 ), rutile (TiO 2 ), and vanadium pentoxide (V 2 O 5 ) (Fig. 3). Vanadium pentoxide can be identified as a gangue component (Dewan et al. 2010). The diffraction peaks of ilmenite and rutile match with the standard XRD pattern of ilmenite (Joint Committee on Powder Diffraction Standards (JCPDS) card no 29-733), and rutile (JCPDS card no 21-1276). However, the initial sample contains less percentage of rutile, and their crystallinity is relatively low (Fig. 3)  The XRD patterns recorded from the milled samples (annealed at 800 °C, 900 °C, 1000 °C, 1100 °C, and 1200 °C) are shown in Fig. 4. Sharp (crystalline) ilmenite peaks can be identified at temperatures of 800 °C and 900 °C (Fig. 4a, b). The number of peaks corresponding to rutile is low in samples annealed at temperatures of 800 °C and 900 °C, and also the respective rutile peaks are not prominent in the XRD spectra (Fig. 4a, b).
In comparison, rutile peaks are more prominent in samples annealed at 1000 °C and above (Fig. 4c-e). Consequently, the occurrence of elemental iron (indicated as F at 2θ of 44.57° and 60.00° in Fig. 4) and rutile peaks indicate the smooth progression of carbothermic reduction at 1000 °C and above. However, the remaining one or more ilmenite peaks imply that the reaction of carbothermic reduction has not been fully completed at temperatures of 1000 °C and 1100 °C (Fig. 4c, d). Although rutile peaks are prominent in the spectra of both 1000 °C and 1100 °C, samples milled for 1 to 3 h imply lesser structural disorder in ilmenite followed by observed lower recrystallized rutile peaks after carbothermic reduction (Chen et al. 1997(Chen et al. , 2013Wijewardhana et al. 2021). Samples milled for 4 h have undergone better recrystallization of rutile with narrow and intense diffraction peaks. In addition, samples milled for 4 h show more rutile peaks after annealing above 1000 °C (Fig. 4d, e). Precisely, the peak intensity has increased with annealed temperature from 1000 °C to 1100 °C (Fig. 4c, d). Interestingly, all the ilmenite peaks have disappeared in the samples annealed at 1200 °C indicating that the carbothermic reduction has been fully completed above 1100 °C. However, elemental iron and rutile peaks are more prominent in samples milled for 3 h and 4 h after annealing at 1200 °C. Figure 5 shows FTIR transmittance spectrum of the initial ilmenite sample. For example, the characteristic transmittance band for CO 2 (Fig. 5) implies that atmospheric CO 2 has been attached to the surface of raw ilmenite. In addition, the Fe-O (from 520 cm −1 to 1450 cm −1 ) and Ti-O-Ti (from 550 cm −1 to 900 cm −1 ) characterises the FTIR spectrum of raw ilmenite (e.g., Chen et al. 2013;Wijewardhana et al. 2021). Furthermore, the transmittance bands around 1630 cm −1 and 3600 cm −1 can be assigned to Ti-OH stretching and bending vibrations of hydroxyl (OH), respectively (León et al. 2017). Figure 6 shows FTIR spectra of 4 h milled samples annealed at different temperatures of 800 °C, 900 °C, 1000 °C, 1100 °C, and 1200 °C, respectively. The broadband around 3600 cm −1 suggests the stretching vibration of the hydroxyl group (e.g., León et al. 2017). The transmittance band around 1630 cm −1 (Ti-OH vibration) appearing in all the treated samples indicates the presence of moisture during sample preparation and/ or analysis. In addition, stretching vibrations around 559 cm −1 and 586 cm −1 (attributed to Fe-O bonds) can be observed in samples annealed below 1000 °C (Fig. 6a,  b). Moreover, samples annealed above 1000 °C (Fig. 6c-e) show Ti-O-Ti stretching vibrations around 700 cm −1 (Siddick et al. 2018;Wijewardhana et al. 2021). Table 2 shows the chemical compositions of raw and treated ilmenite samples. Raw ilmenite concentrate is characterised by over 93 wt.% of FeOTiO 2 , ~ 4 wt.% of SiO 2 , ~ 1 wt.% of Al 2 O 3 , and the rest compensated by other minor oxides. TiO 2 and TiO 2 + Fe 2 O 3 (total) variations of annealed samples suggest that carbothermic reduction can be enhanced after annealing above 1000 °C (Table 2).

Discussion
In this study, a mixture of ilmenite and activated carbon was milled for 1 to 4 h to investigate the possible beneficial effects of milling on carbothermic reduction of ilmenite. Particle size changes, and the effect of particle size on annealing were compared with the initial sample to determine the effective number of milling hours. Accordingly, 4 h milled sample was observed to be laid at a very fine sand region of the USDA textural classification. Mechanical attrition during milling enhances the surface area, energy transferring and lattice distortions of materials (Tao et al. 2012). Moreover, carbon can be incorporated into the ilmenite structure through weak crystallographic planes (Wijewardhana et al. 2021).  Table 1 for sample labelling) Page 7 of 10 Wijewardhana and Ratnayake Bull Natl Res Cent (2021) 45:149 The recorded XRD pattern suggests that the energy required for completing the carbothermic reduction of ilmenite into synthetic rutile is not sufficient at temperatures of 800 °C and 900 °C. Carbothermic reduction reaction is almost completed at the temperature of 1000 °C and 1100 °C, and fully completed at the temperature of 1200 °C. Notably, the increment of intensity and the number of rutile peaks in the samples annealed at 1000 °C, 1100 °C and 1200 °C suggest that the crystallinity and TiO 2 phase have increased, with a well-crystallised structure. Consequently, defects and disorders during mechanical activation have been cured during annealing at high temperatures. However, 3 h of ball milling is sufficient for the complete carbothermic reduction of ilmenite at 1200 °C (Fig. 4e). Nevertheless, milling for 3 h and annealing at 1200 °C, and milling for 4 h and annealing at 1000 °C show similar results (Fig. 4c, e). Thus, the additional 1-h of milling would be efficient than rising and maintaining the temperature at 1200 °C for 2 h (Fig. 4c, e). The energy requirement is high to elevate the annealing temperature at 1200 °C than milling for an additional hour. Consequently, annealing at 1200 °C is less feasible due to high power prices in Sri Lanka (Subasinghe et al. 2021).
FTIR spectroscopy can be used for qualitative analysis of compounds, providing specific details on molecular structure, chemical bonding, and molecular environment Ramalla et al. 2015;León et al. 2017). Therefore, FTIR results suggest that ilmenite undergoes carbothermic reduction with activated carbon at temperatures above 1000 °C. In this regard, it is evident that Fe-O bond in ilmenite is broken, and new Ti-O bonds are formed subsequently during carbothermic reduction at 1000 °C and above.
In XRF analysis, all annealed samples were reheated at 1000 °C for 2 h to determine LOI values. Consequently, XRF results provide limited information for quantifying phase transition from ilmenite to synthetic rutile under low annealed (800 °C and 900 °C) temperatures.
Mechanical incorporation of reductants into ilmenite effectively enhances carbothermic reduction due to increment of effective surface area, reduction of crystallite size within individual particles, extension of crystal defects, and alteration of lattice distortions (Chen et al. 1997;Sasikumar et al. 2004;Tao et al. 2012;Low et al. 2017;Wijewardhana et al. 2021). Consequently, mechanical activation via ball milling increases the homogeneity of the mixture and thus lessens diffusion (Welham and Williams 1999).
The chemical reactions occurring in this reduction process are shown below.  The reactions denoted by Eq. 1 is a solid-state reduction reaction (Welham 1996;Chen et al. 1997) and that of Eq. 2 is a solid-gas reaction where carbon dioxide (CO 2 ) produced reacts with the remaining hot carbon to produce CO at high temperatures (Wouterlood 1979;Chen et al. 1997). In this regard, it is evident that the carbothermic reduction of ilmenite concentrates depends on temperature (Gupta et al. 1989;Wang and Yuan 2006). The optimum conditions of these reactions can also be controlled by physical and chemical characteristics and mineralogical composition of ilmenite ores (Ismail et al. 1983;Welham and William 1999;Wang and Yuan 2006). For example, the presence of manganese (Mn) above 1.24 wt% reduces the rate of carbothermic reduction of ilmenite (Wang and Yuan 2006;Wang et al. 2008). Literature (~ 0.95 wt.%, Herath 1980;Ismail et al. 1983) and XRF analysis of the present experiment (0.85 wt%, Table 2) prove the applicability of carbothermic reduction for ilmenite ores in Sri Lanka. This method can thus be identified as an industrially important process due to the solid-state separation of titanium dioxide (synthetic rutile) and iron present in ilmenite (Zhao and Shadman 1991). In this method, a magnetic separation followed by a simple leaching step is required to purify the synthetic rutile produced (Ismail et al. 1983;Sasikumar et al. 2004;Adipuri et al. 2011;Tao et al. 2012). Consequently, carbothermic reduction can be employed to upgrade the ilmenite resources in Sri Lanka and uplift the mineral processing industry.

Conclusions
The particle size distribution of the mixture of Sri Lankan ilmenite and commercially available activated carbon suggested a significant reduction of grain size after 4 h of ball milling (d 50 = 0.08 mm). Although milling is considered as an essential process for particle size reduction and to maintain homogeneity of the mixture, any direct evidence of carbothermic reduction was not recorded during the process of ball milling. However, XRD observations (i.e., more rutile peaks and increased crystallinity) suggested that carbothermic reduction has been completed by decomposition of ilmenite into elemental (1) FeTiO 3 (s) + C(s) → Fe(s) + TiO 2 (s) + CO g solid − solid reaction (2) C(s) + CO 2 g → 2CO g solid − gas reaction

(3)
FeTiO 3 (s) + CO g → Fe(s) + TiO 2 (s) + CO 2 g solid − gas reaction iron and TiO 2 after milling for 4 h and isothermal annealing at 1000 °C and above continuously for 2 h. Similarly, Ti-O-Ti stretching vibrations around 700 cm −1 and TiO 2 /TiO 2 + Fe 2 O 3 (total) variations suggest sufficient energy for carbothermic reduction at 1000 °C, despite a negligible amount of remaining ilmenite which can be easily decomposed by fine-tuning the timing of thermal treatment. Therefore, ball milling induced carbothermic reduction is applicable to produce a mixture of rutile and elemental iron from ilmenite at the optimum conditions of 4-h milling and annealing at 1000ºC for 2 h. However, magnetic separation followed by a simple leaching process would be encountered to purify the synthetic rutile produced by this mechanochemical method. Consequently, this method has the potential to be established for Sri Lankan ilmenite ores.